Characterization of Channelrhodopsin and Archaerhodopsin in Cholinergic Neurons of Cre-Lox Transgenic Mice

The study of cholinergic signaling in the mammalian CNS has been greatly facilitated by the advent of mouse lines that permit the expression of reporter proteins, such as opsins, in cholinergic neurons. However, the expression of opsins could potentially perturb the physiology of opsin-expressing cholinergic neurons or mouse behavior. Indeed, the published literature includes examples of cellular and behavioral perturbations in preparations designed to drive expression of opsins in cholinergic neurons. Here we investigate expression of opsins, cellular physiology of cholinergic neurons and behavior in two mouse lines, in which channelrhodopsin-2 (ChR2) and archaerhodopsin (Arch) are expressed in cholinergic neurons using the Cre-lox system. The two mouse lines were generated by crossing ChAT-Cre mice with Cre-dependent reporter lines Ai32(ChR2-YFP) and Ai35(Arch-GFP). In most mice from these crosses, we observed expression of ChR2 and Arch in only cholinergic neurons in the basal forebrain and in other putative cholinergic neurons in the forebrain. In small numbers of mice, off-target expression occurred, in which fluorescence did not appear limited to cholinergic neurons. Whole-cell recordings from fluorescently-labeled basal forebrain neurons revealed that both proteins were functional, driving depolarization (ChR2) or hyperpolarization (Arch) upon illumination, with little effect on passive membrane properties, spiking pattern or spike waveform. Finally, performance on a behavioral discrimination task was comparable to that of wild-type mice. Our results indicate that ChAT-Cre x reporter line crosses provide a simple, effective resource for driving indicator and opsin expression in cholinergic neurons with few adverse consequences and are therefore an valuable resource for studying the cholinergic system.

All experiments and procedures were performed in accordance with protocols approved by the Northwestern University or Allen Institute Animal Care and Use Committee.
We also used anti-ChAT immunohistochemistry to identify and count cholinergic neurons in basal forebrain. We counted neurons in sections from 3 ChAT-Cre mice perfused at 196 ± 1 days of age, 3 ChAT-Cre/Ai32(ChR2-YFP) mice perfused at 179 ± 4 days of age, 3 ChAT-Cre/ Ai35(Arch-GFP) mice perfused at 186 ± 0 days of age, and 3 wild-type(C57BL/6J) mice at perfused at 183 ± 1 days of age. From each mouse, we counted ChAT-positive somata in a single 100 μm-thick coronal section 0.7 mm posterior to bregma. The dense patch of cholinergic neurons corresponding to basal forebrain was identified manually, ChAT-positive neurons within this region were counted in ImageJ using the cell counter plug-in, and the count was divided by area to yield cell density. Statistical significance was assessed using a 1-way ANOVA, with a significance criterion of 0.05.
For comparison of membrane properties between genotypes, we made measurements from ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice and compared to our previous measurements from wild-type (C57BL/6) mice [11]. Statistical significance was assessed using the unpaired two-tailed t-test, with a significance criterion of 0.05.
Reporter proteins were excited by wide field illumination through a microscope objective (Olympus, x20/0.95 NA) using a light-emitting diode (LED). Radius of illumination was 600 μm. ChR2 was excited with a blue LED (Thorlabs M470L2 or M470L3 and LEDD1B or DC2100 driver; maximum steady-state intensity 20 mW/mm 2 ) and Arch with a white LED (Thorlabs MCWHL1 or MCWHL5 and LEDD1B driver; maximum steady-state intensity 7 mW/mm 2 ).

Visual behavior
Behavioral data were collected from mice at postnatal day 70-140. Mice were maintained on a reverse light-cycle (12-hour on-off cycle, switching at 9 am and 9 pm). Surgery and behavior were performed at postnatal day 40-60. To head-restrain the mouse, a titanium plate was attached to the skull with C&B Metabond (Parkell) and a fiber optic guide cannula (22 gauge, part number C300GS-5/SPC, Plastics One) was implanted in the left hemisphere using stereotaxic coordinates (2 mm lateral, 0.5 mm posterior, 4 mm ventral to bregma). After 7-10 days of recovery, mice began water scheduling (1-1.5 ml water per day) and were habituated to head restraint (5-10 minutes per day) for 5-7 days before entering behavioral training.
Mice were trained to perform a visual discrimination task. Training occurred during the dark cycle in a dark box with sound attenuation. Mice were trained for 1 session per day, 5 days per week. The mouse was head-restrained and ran on a 16.5 cm diameter disk while visual objects were presented to the right eye on a monitor centered 15 cm from the eye. Luminance of the monitor was 0 to 90 cd/m 2 . The monitor was orientated parallel to the midline and contained a 50% grey background (45 cd/m 2 ).
Visual objects were circular stationary gratings, 10°in diameter that moved across the monitor, along the horizon (0°altitude), at a rate that was determined by the running speed of the mouse. Visual objects were presented in two orientations (vertical and horizontal) and only one of the two orientations was associated with a water reward. The mouse collected a 5-10 μl water reward by slowing its running to hold the rewarded object in a 'reward window' on the screen for 0.75-1.2 seconds. The reward window was invisible and subtended 20°in azimuth, centered at 90°in azimuth (relative to the midline). Mice observed 150-1000 objects per session. A session lasted until the mouse received its daily allotment of water, or until 60 minutes had passed, whichever was sooner. When a mouse failed to receive its daily allotment of water via rewards, supplementary water was provided in the home cage to bring the total daily volume up to 1 ml. 1.5 ml water per day was provided on weekends and mice were supplemented with additional water or high-calorie food, as necessary, to maintain the mouse at 80% of initial body weight.
Naive mice were initially rewarded for running. In this phase, mice were presented with only one orientation of visual object and were rewarded every time the object entered the reward window. Once a mouse was running routinely, the time for which the mouse was required to hold the object in the reward window, in order to receive a reward, was increased. After mice learned to select the rewarded target object (3-6 weeks of training), an unrewarded distracter (a novel orientation of the grating) was introduced randomly at a 1:1 ratio with the target. Mice learned to discriminate between the two objects within several sessions, as measured by the difference between the hit-rate for the target and the false-alarm rate for the distracter. Once a mouse learned to discriminate two high-contrast objects (criterion: difference between stop probability > 0.5), additional low (0%, 25%, 38%) contrast objects were added to the task. Mice were considered unable to learn the task if, after 6 weeks, they routinely failed to run on the disk or stop to select an object, of if they failed to display a difference in stop probability with high-contrast objects > 0.5.
To compare performance, running speed and time to criterion performance across genotypes, we used a one-way ANOVA with a significance criterion of 0.05.
There was also extensive fluorescence throughout neocortex, including a dense network of presumed cholinergic axons and terminals in all layers, and the sparse population of ChAT + local circuit interneurons (Fig 1C). This expression pattern suggests that local illumination of neocortex could be used to drive or suppress acetylcholine (ACh) release in ChAT-Cre/Ai32 (ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice, respectively, consistent with recent brain slice and in vivo studies [13][14][15][16][17].
In a small number (~5%) of ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice, we observed off-target expression of opsins. We previously reported off-target expression in ChAT-Cre/Ai32(ChR2-YFP) mice, in which ChR2 was expressed in glutamatergic neurons, resulting in light-evoked glutamatergic postsynaptic potentials in neocortical pyramidal neurons (Hedrick & Waters, 2015). In addition we have observed 'patchy' fluorescence scattered throughout the brain (or occasionally in only one hemisphere). Fluorescent patches were typically diffuse, unlike the more typical pattern of discreetly labeled neuronal somata and processes observed in most mice, suggesting that fluorescent patches were not associated with neuronal expression (Fig 1D). This patchy expression pattern occurred in ChAT-Cre/Ai32 (ChR2-YFP) and in ChAT-Cre/Ai35(Arch-GFP) mice. Mice with off-target expression were readily identified by their distinctive pattern of fluorescence, and were excluded from further analysis.

Modulating the activity of cholinergic basal forebrain neurons
To verify that the opsins are functional, we obtained whole-cell recordings from cholinergic neurons in acute slices of posterior basal forebrain.
In ChAT-Cre/Ai32(ChR2-YFP) mice, blue light evoked depolarization of cholinergic basal forebrain neurons. Spikes were followed by a large-amplitude and prolonged hyperpolarization, which limited the spike rate during repeated illumination: at 5 Hz, every 2 ms blue light stimulus evoked a spike, but at higher frequencies some stimuli failed to evoke a spike, such that spike rate failed to exceed 15-20 Hz (Fig 2B and 2C). A similar 15-20 Hz limit on spike rate was observed following viral expression of ChR2 in ChAT-Cre mice (Kalmbach et al., 2012) and for current-evoked spikes in wild-type mice [11]. Hence there is sufficient ChR2 expression in ChAT-Cre/Ai32(ChR2-YFP) mice to drive cholinergic basal forebrain neurons at their maximum spike rates.

Cholinergic cell health and mouse behavior
For reporter mice to be useful tools for studying the cholinergic system, it is important that they drive expression of reporter proteins without perturbing cholinergic neurons or mouse behavior. To check for an effect of Cre or opsin expression on the number of cholinergic cells in basal forebrain, we counted ChAT-positive neurons in fixed sections from ChAT-Cre, ChAT-Cre/Ai32(ChR2-YFP), ChAT-Cre/Ai35(Arch-GFP) and wild-type mice at 174-197 days of age ( Table 1). The densities of ChAT-positive neurons were similar across mouse lines   Previous experiments have demonstrated that the resting membrane properties, spike waveform and afterhyperpolarization of cholinergic basal forebrain neurons can all be altered by excessive expression of ChR2 in ChAT-Cre mice using virus [2]. To assay cholinergic cell health in ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice, we measured a suite of resting membrane properties, spike waveform characteristics and after-spike potentials in cholinergic basal forebrain neurons in whole-cell recordings from acute slices. Cholinergic neurons from ChAT-Cre/Ai32(ChR2-YFP) mice exhibited a depolarized resting membrane potential and low input resistance relative to cholinergic neurons from wild-type mice ( Table 2). These effects are less than those observed following strong expression of ChR2 in these neurons [2], suggesting that ChAT-Cre/Ai32(ChR2-YFP) mice have intermediate expression of ChR2 in cholinergic neurons. Importantly, spike waveforms and after-spike potentials were similar in wild-type and ChAT-Cre/Ai32(ChR2-YFP) mice (Table 2, Fig 5). The membrane properties and spike waveforms of cholinergic neurons of ChAT-Cre/Ai35(Arch-GFP) mice were indistinguishable from those of wild-type mice (Table 2, Fig 5).
Finally, we compared the performance of wild-type and transgenic mice on a visual task in which mice discriminate between two visual objects displayed on a monitor. Visual objects were displayed sequentially (one at a time), appearing at the left edge of the monitor and moving across the horizon (Fig 6A). The rate of movement was coupled to the speed at which the mouse ran on a disk. The mouse obtained a water reward by reducing its running speed to hold the visual object in the centre of the monitor for >0.75 s (Fig 6B), but a reward was available from only one of the two possible visual objects. The mouse learned to stop for rewarded objects and to ignore unrewarded objects and discriminated between the two objects even at low contrast (Fig 6C). We generated psychometric curves for each mouse by varying the contrast of the visual objects (Fig 6C and 6D). Once trained, the performance of wild-type, ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice was equivalent (Fig 6D; d' values were 2.48 ± 0.48 for wild-type, 3.04 ± 0.32 for ChAT-Cre/Ai32(ChR2-YFP) and 2.71 ± 0.23 for ChAT-Cre/Ai35(Arch-GFP) mice at a contrast of 1; P-values of 0.28, 0.38, 0.82 and 0.48 at contrast of 0, 0.25, 0.38 and 1, respectively).

Discussion
Our results indicate that crossing ChAT-Cre and Ai32 or Ai35 mouse lines results in expression of functional ChR2 and Arch, respectively, in cholinergic neurons. In basal forebrain, opsins were expressed in most ChAT-positive neurons and we were unable to identify any opsin-expressing neurons that were ChAT-negative. ChAT is a selective marker for cholinergic Spontaneously active, % 30 60 22 Resting membrane potential, mV -55.4 neurons, leading us to conclude that this transgenic mouse breeding strategy drives selective and widespread expression of opsins in cholinergic neurons. The cellular physiology of opsin-expressing ChAT-positive neurons in basal forebrain slices, assessed with an extensive collection of electrophysiological measurements, was comparable to the published cellular physiology of cholinergic neurons in slices from wild-type mice (identified by post-hoc immunocytochemistry [11]). There were just two parameters that were significantly perturbed in ChR2-containing cholinergic neurons: resting membrane potential and resting input resistance. The mechanistic bases of these two changes remain obscure. One possibility is that expression of ChR2 increases the permeability of the membrane to cations, resulting in tonic depolarization and reduced input resistance. However, comparable expression of ChR2 cellular physiology. Importantly, however, most cellular parameters were unaffected by ChR2 in ChAT-Cre/Ai32(ChR2-YFP) mice, including the distinctive spiking patterns of cholinergic neurons, which can be perturbed by strong overexpression of ChR2 [2]. Hence cellular physiology appears to be perturbed only marginally in ChAT-Cre/Ai32(ChR2-YFP) and not at all in ChAT-Cre/Ai35(Arch-GFP) mice.
We found no evidence for behavioral deficits in ChAT-Cre/Ai32(ChR2-YFP) or ChAT-Cre/Ai35(Arch-GFP) mice, probed using a visual discrimination task, which includes elements of sensation, perception, motor function, motivation and reward, and decision making. Our experiments are by no means an exhaustive analysis of possible behavioral deficits in these mice, but suggest that ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice do not display gross behavioral deficits, as observed in ChAT-ChR2-eYFP mice [7]. Together our characterization of opsin expression, cholinergic cell health and mouse behavior suggest that ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice are promising tools for studying the cholinergic system in mice.
However, there are also several limitations of ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/ Ai35(Arch-GFP) mice. Firstly, our results reveal that ChR2 and Arch are widely expressed in cholinergic neurons in multiple forebrain areas. For example, in neocortex opsins are expressed in basal forebrain axons ascending into cortex and also in local circuit ChAT-positive interneurons [18]. Hence widespread illumination of cortex will likely affect both long-range and local cholinergic connections. This lack of specificity can be overcome using viruses, which can be used to drive expression locally, such as in basal forebrain cholinergic neurons and their axons which extend into neocortex [2]. Hence the lack of areal specificity of opsin expression is a disadvantage of ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice relative to viral strategies.
A second, and perhaps even more significant limitation arises from off-target expression of opsins in a small subset of mice. We observed off-target expression in ChAT-Cre/Ai32 (ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice (and also in ChAT-Cre/Ai14(tdTomato) mice; not shown), which gain Cre-dependence from a loxP-stop-loxP sequence under control of the CAG promoter in the Gt(ROSA)26Sor locus. Hence it seems that the off-target expression results from either expression of Cre by ChAT-negative neurons or from expression of reporter proteins in the absence of Cre in this group of reporter lines. In an experiment in which the pattern of fluorescence cannot be assessed or where assessment must wait until after the experiment is complete, off-target expression may be problematic. For example, in behavioral experiments, with no cranial window through which to image fluorescence, assessment of the pattern of fluorescence might have to wait until the mouse is sacrificed at the end of the experiment. Under these conditions, off-target expression may limit the utility of ChAT-opsin mice derived from ChAT-Cre and Ai32 or Ai35 mouse lines. However, off-target expression is rare in our colonies and invariably results in a distribution of fluorescence which is readily distinguished from expression in only ChAT-positive neurons. Hence mice with off-target expression can be readily identified and can therefore be excluded from further experiments or analyses, thereby eliminating off-target expression as a significant limitation. We conclude that ChAT-Cre/Ai32(ChR2-YFP) and ChAT-Cre/Ai35(Arch-GFP) mice exhibit extremely selective expression of opsins in cholinergic neurons, little or no perturbation of cellular function and no obvious behavioral phenotype, making these lines among the most useful tools for studies that require optical modulation of forebrain cholinergic neurons.
G4 ES cell line, used in the development of the Ai32D and Ai35D mouse lines. Funding was