2.1.1.

Virus preparation

A vast variety of viruses for optogenetics are commercially available from vector core facilities (i.e., the UNC Vector Core¹⁴; or Penn Vector Core¹⁵). In general, there are three types of opsins: excitatory (i.e., ChR2, C1V1), inhibitory (i.e., NpHR, ArchT), and step-function opsins. Each opsin is sensitive to a different wavelength of light and with different response properties upon light illumination. For reviews that compare different opsins, please refer to Yizhar et al.,¹⁶ Fenno et al.,¹⁷ and Mattis et al.¹⁸ In most of our experiments with nonhuman primates, we have used the viral construct AAV5-$\mathrm{CaMKII}\alpha$-C1V1 (E122T/E162T)-TS-EYFP. C1V1 is a red-shifted excitation variant of ChR2 that peaks at a wavelength of approximately 540 nm.^19,20 This viral construct mainly targets excitatory neurons (determined by the $\mathrm{CaMKII}\alpha$ promoter and AAV5 virus), with which reliable neural and behavioral modulation has been previously reported.^10,12,21 However, the methods we detail next are applicable to all types of viruses used in optogenetics studies.

The handling of the viruses from their arrival from the vector core to the beginning of injections should be performed in the following way. The virus typically arrives from the vector core in a $100\mu \mathrm{L}$ vial, which we immediately aliquot into smaller amounts (10 or $20\mu \mathrm{L}$, depending on the amount we plan to use at one time plus some surplus). Aliquoting the original vial only once is critical, as multiple freeze–thaw cycles could potentially damage the virus and impair the efficacy of future injections.

2.1.2.

Virus performance verification

In our experience, there is a significant variation in the performance of each batch of virus obtained from a vector core facility. Some batches lead to no or weak efficacy in opsin expression. Therefore, it is important to verify the performance of a viral construct before injecting it into a primate. Our approach for verification is to inject the virus into at least two rats to confirm the efficacy of opsin expression with neuromodulation experiments and histology. We have found that if a virus shows good expression in rats, it has a high chance of success in primates. While this might not be true for every viral construct (note that we have only explored a few viral constructs in our studies), we believe that if a virus does not show good expression in rats, it will likely have problems in the monkeys. Therefore, we recommend checking the viruses in rats if possible. We note that the differences in performance between different batches of viruses are likely to be related to the titer of the virus, although we have not systematically tested this possibility. In our own experiments, we have successfully used titers of both $2.0\times {10}^{12}$ and $3.0\times {10}^{12}\text{molecules}/\mathrm{mL}$.

2.1.3.

Virus injection

Injections can be performed either in the operating room (OR) while the monkey is anesthetized^5,9 or in the primate chair while the monkey is seated under head restraint.^4,10 The latter approach has obvious advantages. First, chair injections are more flexible than injections in the OR, where additional veterinary personnel are needed. Second, it is also better for the animal, especially when the procedure requires multiple days to complete, as injections can be performed in the same conditions as recording sessions, without requiring stereotaxic placement, anesthesia, or sedation. These “in-chair” injections are, of course, only feasible when there is a pre-existing chamber or accessible burr hole in the skull, providing access to the brain. However, the “in-chair” injections require rigid animal restraint and calming animals; movement needs to be kept to a minimum to avoid potential tissue damage (also see Discussion).

Injection equipment used by different groups differs significantly. Typically, an injection needle and a micropump are used. Additionally, other equipment to position the injection needle at the precise target location is required. Both Han et al.⁴ and Diester et al.⁵ injected viruses through a needle which was connected to a syringe via an oil-filled polyimide-coated glass tube. We found that this method was not optimal for controlling the injection volume as the dead space between the syringe and the needle potentially results in a significant amount of virus wasted in the tube. Therefore, it is preferable to shorten the path and directly connect the needle with the syringe. Cavanaugh et al.⁸ performed the injection using a custom-built injectrode, which enabled them to inject immediately after identifying a target site. The main limitation of their injectrode is its size, which creates unavoidable tissue damage and makes it unsuitable for repeated use.

As an improvement over the existing paradigms, we developed our injection system as illustrated in Fig. 1, which could easily be applied in-chair with reduced tissue damage. Using this system, we control the injection needle as we normally control a recording electrode, and perform the injection while the animal is head-fixed as in a recording session. We modified our primate chair to hold a stereotaxic rail on which injection equipment could be rigidly held. The same setup could be adapted for OR injections by mounting it on the surgery stereotax. The virus was withdrawn into a $25\text{-}\mu \mathrm{L}$ syringe cemented to a 32-gauge needle (Model 702 SN, Hamilton, Reno, Nevada). The syringe was mounted on a microsyringe pump (UltraMicroPumps III, World Precision Instruments, Sarasota, Florida), which was used to control the injection volume and speed by directly controlling the movements of the plunger of the syringe. A micropositioner (In vivo manipulator, single axis, Scientifica, East Sussex, United Kingdom) holding the pump via a custom built piece (silver block attached to the micropositioner in Fig. 1) was mounted on the stereotaxic apparatus [Kopf, three-dimensional (3-D) manipulator] to accurately locate the target depth (with micrometer resolution), and the 3-D manipulator was used to adjust the rough position and angle. While targeting a particular location in the brain is straightforward with our system, before injection it is important to verify that the target brain site for injections is relevant for the particular study. To verify this, we recommend performing electrophysiological recordings beforehand to map the region functionally.

Fig. 1

Injection equipment and organization. A cemented syringe and needle are mounted on a micropump. A micropositioner and a three-dimensional (3-D) manipulator are used to control the positioning of the needle. A standard Kopf rail is attached to the chair and serves as a rigid base. The inset on the bottom right shows the Kopf X-Y stage, which is used to locate the position in the chamber.

In many cases, the target region covers a significant volume in the brain. Based on histology, we have found our injections routinely lead to opsin expression in a cylindrical brain volume of a 1 to 2-mm diameter and $\sim 1\mathrm{mm}$ height. Therefore, in cases where we would like to transduce neurons in a large volume, we inject the virus at different depths spaced 1 mm apart at each penetration site. The procedure for multidepth injections is as follows. Once the needle had been aligned with the guide tube (25 gauge), we lower the needle to the deepest target location (typically 8 to 9 mm in the lateral intraparietal area (LIP)). We start the injection from the deepest site, and then retract to a shallower site 1 mm above it. The reasons for starting with the deepest site are, first, not to interfere with sites that have already been injected. Passing the needle through these sites could indeed affect the diffusion of the viral solution and lead to adverse effects. Second, starting from the deepest site allows for deposition in the penetration track with subsequent diffusion and avoids potential reflux. Typically, one microliter (although larger volumes of $2\mu \mathrm{L}$ have also been used by our group) of virus per depth location is injected at a speed of 100 to $200\mathrm{nL}/\min$. A waiting time of 5 min after each injection seems to allow diffusion before retracting the needle. Usually, to obtain a larger transduced area, injections at multiple penetration sites (separated $\sim 1\mathrm{mm}$ from each other on the cortical surface) and at multiple depths are recommended.

2.1.4.

Optical stimulation and neural recording

A typical optical stimulation system includes a light source, delivery path and output, and control by an external TTL/analog signal. In all published optogenetic studies in primates, a laser coupled to an optical fiber has been utilized as the light source. This configuration can provide sufficient optical power [light-emitting diodes (LEDs) can be another option, but keep in mind that the LED wavelength spectrum is broader and coupling efficiency to an optical fiber is low). Lasers with fiber-optic couplers are commercially available (e.g., OptoEngine LLC, Coherent Inc., Omicron Laserage, and so on). An optical fiber connected to the coupler can deliver light into the monkey brain. To avoid excessive tissue damage, a small outer diameter is recommended for a fiber that needs to go into the brain. Previous studies mainly used fiber of 200 to $250\mu \mathrm{m}$ diameter.^4–6,8 The size of the fiber includes the core, the cladding, and the coating. The coating provides extra buffer layer to protect the fiber (i.e., the fiber mentioned previously has a coating diameter of about $500\mu \mathrm{m}$), and usually it will be removed before use. Note that different fibers having the same core size may have different cladding size. Caution should be used when indicating fiber size in publications. In our previous papers,^10,21 smaller fibers with $125\text{-}\mu \mathrm{m}$ cladding (10 or $50\mu \mathrm{m}$ core) were used. Another important parameter of the optic fiber is the numerical aperture (NA), which is a dimensionless number representing the range of angles at which the fiber can accept or emit light. In general terms, a larger NA means a broader dispersion of light emitted from the fiber.

One issue reported by previous studies is that fibers with blunt tips can cause tissue damage and subsequently make long-term recording difficult.^6,22 To address this, we taper the fiber to a tip angle of approximately 10 deg [$\sim 8\mathrm{deg}$ in Fig. 2(b)] using a laser puller (P-2000, Sutter Instruments). This modification, together with using a smaller-diameter fiber, significantly reduces tissue damage and renders it possible to repeatedly penetrate the same area for months (more than 6 months; generally, more than 100 penetrations were done for each monkey). Another benefit of such modification is that a significant amount of light can exit the fiber along the edges of the taper before reaching the tip (Fig. 2), which increases the volume of tissue illuminated compared to a blunt tip, and thereby reduces the peak light intensity in the tissue for a given power. In general, special attention should be given to the amplitude of peak light intensity in the tissue since high light intensity could cause tissue damage or nonspecific heat-induced effects.²¹

Fig. 2

Tapered fiber and optrode. (a) A pulled fiber after connecting to a ferrule and polishing of the ceramic end. The inset zooms in the tapered end, which also shows the fiber with and without coating. (b) The tip of the optrode under high-resolution microscope, which shows the relative position of the tapered fiber and the tungsten electrode. (c) Illumination of light scattering from the fiber tip.

Different stimulation parameters, including power, frequency, and duration, were applied by previous studies. In these studies, either “power” or “intensity” was used to report the amount of illumination, but these are different physical quantities which use completely different units. Power is the energy emitted per unit time, in units of watt (W), milliwatt (mW), and so on, which can be directly measured using a power meter. Intensity is the power per unit area, in units of $\mathrm{W}/{\mathrm{m}}^2$. One can easily measure the power of a laser by placing the output fiber into the sensor of a power meter, but further calculation is needed to determine intensity. For example, the light intensity at a blunt fiber end is the optical power divided by the area of the fiber core, but for a tapered fiber, the core is attenuated after pulling; therefore, the intensity will vary along the tip. In brain tissue, the light will scatter after exiting the fiber and the intensity will dramatically drop at increasing distances. To see a simulation of light intensity distribution in brain tissue, please refer to Ozden et al. for a blunt fiber²¹ and Dai et al. for a tapered fiber.¹⁰

In terms of optogenetic stimulation, it is of great interest which illumination parameters are optimal. Unfortunately, there is no clear-cut answer for this as these parameters depend on the fiber, opsin, brain area, and other factors. Before setting these parameters, one has to be careful about the heat accumulation around the fiber tip due to light absorption by the tissue, which potentially could damage the brain. To ensure safety, we have previously calculated that, given certain assumptions and with continuous light illumination, the power should not exceed 12 mW for a $10\text{-}\mu \mathrm{m}$ core fiber and 40 mW for a $200\text{-}\mu \mathrm{m}$ core fiber.²¹ Note that higher powers can be applied when pulsed (instead of continuous) stimulation is used, depending on the stimulation duty cycle. However, with our setup we have found that a few hundred microwatts is sufficient to activate neural activity when using the C1V1(T/T) opsin.

One advantage of optogenetics is that it allows simultaneous recording during stimulation. Several special readout devices, namely optrodes, have been designed for this purpose.^21,23,24 Lacking these more advanced and expensive tools, we have found an alternative solution is to build an optrode by gluing an electrode and a fiber together. Note that the optrode used in Diester et al.⁵ was a $250\text{-}\mu \mathrm{m}$ electrode glued with a $225\text{-}\mu \mathrm{m}$ fiber, so the overall size was at least $475\mu \mathrm{m}$, which is not ideal for an acute recording device. We refined this optrode by using a smaller fiber ($\text{cladding}=125\mu \mathrm{m}$) and a smaller electrode ($\text{shank diameter}=75\mu \mathrm{m}$), rendering the optrode diameter significantly smaller, around $200\mu \mathrm{m}$ in total.¹⁰ This smaller size is not dramatically larger than a standard recording electrode. Using this kind of device, we have been able to repeatedly obtain effective neural modulation from a transduced area approximately $2\times 2\times 4\mathrm{mm}$ for months without having the problem of serious cortical damage reported by Gerits et al.⁶ More importantly, virtually no artifacts were observed at the single/multiunit level using our filter setting (300 to 6000 Hz). Given that others have observed light-induced artifacts during recordings with metal electrodes and that same type of electrode was used in a similar optrode configuration,⁴ we think using a smaller size electrode is essential as decreasing the area of light exposure is effective in eliminating photoelectric reaction ($75\mu \mathrm{m}$ here versus $200\mu \mathrm{m}$ by Han et al.⁴).

2.1.5.

Expression verification

To verify that the injected virus does indeed transduce neurons in regions of interest, the most straightforward approach is to find optically induced neural modulation in those injected sites. At sites with sufficient opsin expression, neurons should be either activated (for excitatory opsins, Figs. 3 to 4) or suppressed (for inhibitory opsins) upon light stimulation. The question is how to find optical modulation after injection. For us, the general procedure is described as follows. We usually start by listening (using an audio monitor) to the change of background noise as we deliver light pulses through the optrode while being lowered in the brain. Typically, as the optrode gets closer to the transduced area, we begin to hear a weak hash, like the distant crash of waves, synchronized with light pulses. The sound gets stronger as the optrode gets deeper. There might be no visible spiking on the oscilloscope at this point, but by lowering the speed of progression, and listening patiently to the audio signal, eventually single/multiunit modulation can be detected in most cases after hearing this initial low-frequency noise. In our experience, the approach described previously is very efficient for identifying regions of opsin expression and finding spiking units responsive to light. As an alternative approach, one might first start looking for spiking units without checking the light responsiveness in the tissue, and then checking whether the isolated units are light-sensitive. In our hands, this approach was less efficient in finding light-responsive units, since the isolated units could be in a region of weak or no expression. Overall, the background noise during light pulses may reflect the network effect of multiunit response, thus it should be an effective source to identify optically modulated sites.

Fig. 3

Example of optogenetic modulation. Each column represents a different stimulation frequency. From top to bottom, we show: (a) a schematic of the pulse train delivered to the laser at each stimulation frequency, (b) the raw spike train from one randomly picked trial for each frequency, and (c) the raster and the spike density functions.

Fig. 4

Comparison of different optical stimulation powers. Neural responses to a 200 ms continuous stimulation with measured stimulation power of (a) 0.25, (b) 0.5, (c) 1.0, and (d) 1.5 mW.

In addition to multiunit modulation, previous studies have also reported the modulation of local field potentials (LFPs).^4,5,9 Typically, for sites expressing opsin C1V1(T/T), we have observed a negative deflection in LFPs upon green light illumination, followed by a rebound after stimulation [Fig. 5(a)]. However, as reported by Han et al., such a modulation pattern can also be observed in saline as artifacts, especially when a large-size fiber and electrode are used.⁴ Therefore, even in a region of opsin expression in the brain, it is not straightforward to distinguish opsin-induced LFPs from light-induced artifacts. Therefore, LFP-like signals in response to light pulses do not necessarily mean that the signal is neural or that the opsin is being properly expressed in nearby tissue. In fact, light-induced artifacts are more pronounced in the LFP bandwidth and depend on both the stimulation power and wavelength of light. Even in cases where the amplitude and shape of the artifact could be potentially different from optically induced LFPs [Fig. 5(b)], we still think LFPs cannot be reliably distinguished from artifacts and therefore are not a reliable way to identify regions of opsin expression. A more meaningful way may be to first characterize the artifacts at fresh sites that have not been injected and far away from injection sites. No artifacts observed at fresh sites could increase the reliability of LFPs observed in transduced sites [Fig. 5(c)]. However, given the variety of LFP patterns for different opsins,^5,9 the mechanism underlying optical-induced LFP is still unclear. The question of using LFP as a verification measurement remains open. We do recommend testing LFP artifacts in saline and in the brain beforehand.

Fig. 5

Typical local field potential (LFP) versus artifact. (a) Typical LFP recorded in a site transduced with C1V1(T/T). Black, purple, and green represent stimulation of continuous, 100 and 50 Hz, respectively. Red is the base line when no stimulation was applied. Stimulation period is indicated by the green shadow. Same conventions are used for b–c. (b) LFP recorded in saline. (c) LFP recorded in a nontransduced site.

As an alternative approach, in vivo fluorescence detection systems were proposed which allowed monitoring expression in vivo after injection.^5,21 Diester et al. reported that the fluorescence measurements correlated with neuron responses to light stimulation. Theoretically, this could be an ideal substitution for histology, especially for an ongoing experiment. However, our experience has shown that detecting an increase in fluorescence in a given location does not necessarily mean optically induced neuromodulation will be present.

The ultimate verification consists of histology, which is reliable and useful to characterize the efficacy of expression. However, we do not recommend sacrificing every individual subject for this purpose. Instead, we suggest testing the expression efficacy in a rat as we have observed very similar result in both recording and histology.

2.1.6.

Validation of the functionality of viral construct in rats

As previously mentioned, performing virus injections in rats is suitable for validation of virus efficacy and can be used in lieu of primate histology. Our protocol is detailed next. The viral construct to be tested was injected into two cortical areas (usually motor, somatosensory, or posterior parietal cortical areas) from the same hemisphere in two rats. As a control, a viral construct already tested and of known efficiency was injected into the same cortical areas on the other hemisphere. The injections were performed as follows: rats were mounted on a stereotaxic frame (Model 1730, Kopf Instruments) under isoflurane (2%) anesthesia, and under aseptic conditions, a small skin incision ($\sim 10\mathrm{mm}$) was made to expose the skull above the target areas for injections. Small burr holes ($\sim <1\mathrm{mm}$) were made at each site to expose the brain with the dura intact. Through each burr holes, the virus was injected at two depths (0.5 and 1.5 mm) with the same micropump-syringe system described previously mounted to a stereotaxic micromanipulator. The injection amount and speed were similar to that used in primates, i.e., $1\mu \mathrm{L}$ of virus per depth at a speed of $100\mathrm{nL}/\min$ and 5 min waiting time after injections. After injections were completed, the burr holes were covered with bone wax and the skin was sutured. After allowing at least 3 weeks for opsin expression, the rats were anesthetized with either isoflurane ($\sim 2\%$) or ketamine/xylazine ($10\mathrm{mg}/\mathrm{kg}$) and the sites injected with the investigated virus were exposed with two craniotomies ($\sim 2\times 2\mathrm{mm}$). The location of viral expression was first determined by investigating the yellow fluorescent protein (YFP) fluorescence (YFP is coexpressed with the opsin) at the brain surface. For this purpose, we delivered 473-nm laser light over the craniotomy and observed the YFP fluorescence with a stereomicroscope carrying a YFP filter. Usually, YFP fluorescence signal was visible at the surface throughout the craniotomy. After verifying the expression and finding the site of peak expression by observing the fluorescence, we validated neuromodulation by inserting an optrode into the region of peak fluorescence intensity and delivering occasional light pulses ($\sim <1\mathrm{mW}$, 561 nm) of 1 to 2 s duration and observing single and multiunit spiking activity. Once the functional expression was validated, the rats were perfused with 2% paraformaldehyde to prepare $60\mu \mathrm{m}$ thick histological slices. The slices were investigated for the extent of YFP expression under a fluorescence microscope. This time, the YFP fluorescence intensities and pattern of expressions due to injections of the tested and control viruses were compared. If they appeared to be similar, we concluded that the new viral construct was suitable for primate use.

2.2.1.

Reagents

Virus: AAV5-$\mathrm{CaMKII}\alpha$-C1V1(E122T/E162T)-TS-EYFP (UNC vector core);

Dilute bleach (10% dilute bleach);

95% ethanol (Fisher Science Education, cat. ID (catalog ID): A405F-1GAL);

Cidex solution (Advanced Sterilization Products CIDEXPLUS 28 day solution, cat. ID: 2683/2785);

Chlorhexidine solution (Vet Solution, cat. ID: 91010);

Sterile saline.

2.2.2.

Equipment

2.3.1.

Aliquoting virus

The purpose here is to aliquot the virus into individual $20\mu \mathrm{L}$ vials.

2.3.2.

Injecting virus

The injection plan here is to inject 2 to 3 locations covering 5 to 6 mm depth each.

2.3.3.

Making the optrode

The purpose here is to make a tapered fiber of $\sim 10\mathrm{cm}$ length and then glue with a tungsten electrode of approximately the same length. Readers are welcome to contact the authors to obtain further instruction.

2.3.4.

Stimulating and recording