Genetically encoded optical probes of neuronal activity offer the prospect of simultaneous recordings of neuronal activity and behavior in intact animals. A central problem in simultaneous imaging is that the field of view of the high-power objective required for imaging the neuron is often too small to allow the experimenter to assess the overall behavioral state of the animal. Here we present a method that solves this problem using a microscope with two objectives focused on the preparation: a high-power lens dedicated to imaging the neuron and low-power lens dedicated to imaging the behavior. Images of activity and behavior are acquired simultaneously but separately using different wavelengths of light. The new approach was tested using the cameleon calcium sensor expressed in Caenorhabditis elegans sensory neurons. We show that simultaneous recordings of neuronal activity and behavior are practical in C. elegans and, moreover, that such recordings can reveal subtle, transient correlations between calcium levels and behavior that may be missed in nonsimultaneous recordings. The new method is likely to be useful whenever it would be desirable to record simultaneously at two different spatial resolutions from a single location, or from two different locations in space.
Genetically encoded optical probes of neuronal activity offer the prospect of simultaneous recordings of neuronal activity and behavior in intact animals. These include optically transparent, genetically tractable organisms such as nematodes (Kerr et al. 2000) and larval zebra fish (Higashijima et al. 2003), and also opaque organisms such as the fruit fly and mouse, in which the brain can be exposed by localized dissection (Fiala et al. 2002; Hasan et al. 2004; Ng et al. 2002). A wide range of probes have been developed that monitor key markers of neuronal activity such as transmembrane voltage, intracellular calcium, and synaptic vesicle cycling (Guerrero and Isacoff 2001; Miesenbock 2004; Miesenbock and Kevrekidis 2005; Miyawaki 2003). Nevertheless, the potential for using optical probes to record neuronal activity and behavior simultaneously remains largely untapped.
A central problem in imaging neuronal activity and behavior simultaneously is that the field of view of the high-power objective required for imaging the neuron is often too small to allow the experimenter to see more than a small portion of the animal. Thus it is generally not possible to ascertain, with confidence, the behavioral state of the animal throughout an imaging session. Here we present a method that solves this problem using a microscope with two objectives focused on the preparation: a high-power lens dedicated to imaging the neuron and a low-power lens dedicated to imaging the behavior. Images of activity and behavior are acquired simultaneously but separately using different wavelengths of light.
The new approach was developed in connection with the probe known as cameleon, a calcium-sensitive relative of green fluorescent protein (Miyawaki et al. 1997, 1999). Cameleon is a chimeric protein composed of two derivatives of green fluorescent protein—cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP)—bridged by calmodulin and the calmodulin binding domain of myosin light chain kinase (M13). In the presence of calcium, calmodulin binds to M13, which brings CFP and YFP closer to each other. When the molecule is excited at wavelengths that activate CFP, close apposition of CFP and YFP increases the efficiency of fluorescence resonance energy transfer between them, leading to an increase in the ratio of yellow to cyan fluorescence. A key strength of the cameleon protein is that it can be targeted to the neuron of interest by a cell-specific promoter for positive identification of the cell.
We used the new method to study the relationship between neuronal activity and behavior in Caenorhabditis elegans. In the past, the small size of the animal (1 mm in length) and its neurons (2 μm in diameter)—together with its tough, pressurized cuticle—was a barrier to neurophysiological analysis. This situation changed with the advent of four new techniques: 1) sharp-electrode recordings from pharyngeal muscle (Davis et al. 1995), 2) whole cell patch-clamp recordings from neurons and muscles in situ (Goodman et al. 1998; Richmond and Jorgensen 1999), 3) a method for culturing C. elegans neurons (Christensen et al. 2002), and 4) calcium imaging from neurons and muscles using genetically encoded probes (Kerr et al. 2000; Suzuki et al. 2003). However, so far it has not been possible to make simultaneous recordings of neuronal activity and whole-animal behavior in C. elegans, a significant obstacle to a comprehensive neuroethology of this widely studied genomic organism.
We present evidence that simultaneous recordings of neuronal activity and behavior are practical and illuminating in C. elegans. We found that calcium signals can be reliably detected in animals that are only partially immobilized and that such animals continue to respond behaviorally to chemosensory stimulation despite the presence of cameleon. We also found that simultaneous recording can reveal subtle correlations between neuronal activity and behavior that may be difficult to detect in nonsimultaneous recordings.
To make simultaneous recordings of neuronal activity and behavior, we used an inverted microscope (Zeiss Axiovert 135) with two objective lenses focused on the preparation: a high-power objective below the microscope stage and a low-power objective above the microscope stage (Fig. 1). We refer to this as the dual-objective approach. The high-power objective (63 × oil, 1.4 NA), situated on the nosepiece of the microscope, was used for calcium imaging. The low-power objective (10×, 0.25 NA, tube length 160 mm), supported by a hinged platform, formed an image of the whole animal for an analog video camera (Sony XC-ST70, Ft. Meyers, FL) that was used to record behavior to videotape. The purpose of the hinged platform was to enable the experimenter to interchange the low-power objective with the microscope's condenser for bright-field illumination when arranging the preparation on the microscope stage. A three-axis manipulator (not shown), incorporated into the hinged platform, was used to center and focus the low-power objective. The microscope was equipped with an epifluorescence illuminator, software-controlled shutter (ASI, Eugene, OR), beam splitter (Optical Insights, OI-DV-FC, Tucson, AZ), and digital video camera (Hamamatsu, ORCA-AG, Bridgewater, NJ) for ratiometric calcium imaging using cameleon as previously described (Kerr et al. 2000; Suzuki et al. 2003).
The optics contained two other key components. The first component was a 575-nm long-pass filter inserted between the low-power objective and the behavior camera. The function of this filter was to prevent saturation of the behavior camera by the intense blue light (426–446 nm) used to excite the cameleon protein. The second component was a red laser (635 nm, 4.9 mW) that served to illuminate the animal. We selected a red laser to avoid interference with the emission wavelengths of the cameleon protein (465–555 nm). Contrast in the behavior image was enhanced by aiming the laser beam almost parallel to the microscope stage. This arrangement minimized specular reflection from the polished surface of the high-power objective and achieved a pseudo-dark-field illumination effect; we were unable to obtain sufficient contrast using light-emitting diodes aimed at the preparation. The behavior of the animal, as assessed with bright-field illumination, did not visibly change when the laser was turned on or off. We manually synchronized the calcium imaging and behavior recordings. This synchronization was achieved by starting the videotape recording, then simultaneously activating the laser and the software routine that opened the shutter to begin frame acquisition for calcium imaging. We found that shutter opening was delayed relative to laser activation because of latencies in software. We measured this delay by removing the long-pass filter normally inserted between the low-power objective and the behavior camera so that this camera was now saturated whenever the shutter opened. Video data from the behavior camera were then replayed frame by frame to measure the time between laser activation (the start of the behavior record) and shutter opening (the start of the calcium imaging record). We observed a delay of 0.75 ± 0.01 s (means ± SE, n = 14); all calcium imaging and behavior recordings were therefore aligned in time by deleting the first 0.75 s from the behavior record. Additionally, the low variability in the delay indicated that experimenter reaction times were negligible.
We used adults of the strain AQ1444 (lin-15(n765) X; ljEx95 [lin-15(+); psra-6:YC2.12]), in which cameleon is expressed specifically in the ASH sensory neurons (Hilliard et al. 2005). The previously reported weak expression in the ASI sensory neurons (Troemel et al. 1995) was not observed; thus we were able to identify ASH neurons unambiguously. The emission intensity of the cameleon protein varied between individual animals, a likely consequence of the fact that the cameleon construct exists as an extrachromasomal array in this strain. In this study, we focused on animals in which expression was judged by eye to be bright enough to give good calcium signals.
Preparation and solutions
To stabilize worms for recording, we used the so-called glued worm preparation (Faumont et al. 2005), in which animals are glued to agarose-coated coverslips and submerged in a pool of saline. The glue is confined to the region of the head such that the head remains relatively still, whereas the rest of the body is free to move. The coverslip is sealed with wax under a hole in a glass plate, forming a recording chamber, which is filled with saline.
We previously showed that worms prepared in this way exhibit recognizable components of locomotory behavior and respond normally to chemosensory inputs (Faumont et al. 2005).
In this study, the vicinity of the worm was continuously perfused by a two-channel gravity-fed system. The worm was stimulated by switching between the channels containing normal saline and repellent saline. Normal saline was composed of (in mM): NaCl (50), CaCl2 (0.1), and sorbitol (100). To form the repellent saline, 10 mM CuCl2 was added to normal saline. Each animal was stimulated and recorded once, then discarded. In control experiments (Fig. 3A), both perfusion channels carried normal saline.
Approximately half the animals prepared in this way exhibited vigorous swimming behavior that had distinct forward and reverse states characterized, respectively, by a head-to-tail or tail-to-head progression of the body wave, as previously described (Faumont et al. 2005). The other animals appeared to move in an uncoordinated fashion, or not at all, so we excluded them. The proportion of vigorously swimming animals declined to zero in about 40 min. We attributed poor swimming behavior to the presence of glue on the animal because unglued individuals nearby in the recording chamber continued to swim normally; the AQ1444 strain appeared to be more sensitive to the glue than wild-type worms for unknown reasons. This sensitivity is unlikely to result from expression of the cameleon protein because other cameleon-expressing strains do not display it (Faumont and Lockery, unpublished data).
Images were acquired using MetaVue software (version 6.2r2, Molecular Devices, Sunnyvale, CA). Frames were taken at 1.25–33.0 Hz (corresponding to exposure times of 30–800 ms) with 8 × 8 spatial binning. This range of exposure times reflects the variability in the expression level of cameleon. Image stacks were processed using Jmalyze (written and kindly provided by R. Kerr, Salk Institute, San Diego, CA), as previously described (Kerr et al. 2000; Suzuki et al. 2003). The YFP/CFP emission ratio was computed as (YFP intensity)/(CFP intensity) − 0.65, where the latter term corrects for CFP bleed-through into the YFP channel. The emission ratio was compensated for photobleaching by fitting a single exponential function to the inactive portions of the emission ratio trace and dividing it by the fitted curve; thus all ratio changes were expressed in terms of ΔR/R. A spurious ratio change was sometimes detected in ASH neurons immediately after exposure of the preparation to the cameleon excitation beam, as previously reported (Hilliard et al. 2005). This ratio change was accompanied by a modest increase in reverse swimming probability that decayed in about 5 s; thus we waited 15 s after turning on the excitation beam before delivering the chemosensory stimulus. Spatial displacement of the neuron in the x–y plane was assessed by computing the radial distance between the present and initial positions of the cell body.
Video recordings were digitized and replayed frame by frame to note the times at which transitions between forward and reverse swimming occurred (Faumont et al. 2005). The time of transition from forward to reverse swimming was defined as the frame in which there appeared in the tail a sharp inflection that grew into a coordinated, tail-to-head body wave (see supplemental movie1 ). The time of transition from reverse to forward swimming was defined as the frame in which the flexed posture characteristic of reverse swimming began to unfold into a head-to-tail body wave. The probability of reverse swimming was computed in consecutive 5-s bins as Tr/T, where Tr is total time in the reverse state and T is bin length. Behavior scoring was not blind to the stimulus.
In the bottom panel of Fig. 3A1, differences between pairs of binwise means (stimulated vs. unstimulated; t ≥ 0 s) were assessed using multiple t-tests (m = 6). In accordance with the Bonferroni procedure (Winer et al. 1991) for a joint significance level of P = 0.05, the significance level required on individual tests was lowered from 0.05 to 1 − (1 − 0.05)1/m = 0.0085. The significance in the linear regressions of Fig. 3, A2 and B2 was assessed according to the statistic r, which is distributed like a Student's t-distribution with N − 2 degrees of freedom (Press et al. 1988).
We first asked whether calcium signals could be detected reliably in semirestrained worms. This question was significant because even when the cuticle of the worm's head is glued to the substrate, cells within the head often move as the rest of the body flexes or extends, and cell movements are larger in the semirestrained preparation used here than those in the fully restrained preparations typically used in C. elegans calcium imaging experiments.
We addressed the question of reliability by attempting to detect calcium signals in the ASH neurons, a left–right pair of sensory neurons that is sensitive to noxious chemical inputs and mechanosensory stimulation (Hart et al. 1995; Maricq et al. 1995; Sambongi et al. 1999). We selected ASH neurons for study because calcium signals were previously obtained from them in fully restrained worms (Hilliard et al. 2005) and because stimuli that activate ASH neurons also induce an avoidance response that is easy to observe and quantify in semirestrained worms (Faumont et al. 2005). In response to the chemical repellant CuCl2, each of the 27 ASH cells we imaged showed a clear ratio change that was characterized by a sharp rise and a prolonged decay, as exemplified by the recording shown in Fig. 2. The time course and amplitude of the ratio change were consistent with previously observed calcium transients in ASH neurons in fully restrained preparations (Hilliard et al. 2005).
It is well known, however, that calcium imaging is susceptible to movement artifacts. This is true even in the case of a ratiometric probe such as cameleon, where imperfections in background correction and chromatic aberration lead to spurious ratio changes, particularly for cell movements along the z-axis (Kerr 2002). To explore the question of whether the ratio changes we measured resulted from a change in calcium concentration or cell movement, we used the usual procedure of inspecting the YFP and CFP intensity traces that underlie the ratio trace. Theoretically, changes in calcium concentration usually cause reciprocal changes in YFP and CFP intensity, whereas cell movements usually cause changes in the same direction (Kerr 2002). In about half of the ASH cells we recorded (13/27, 48%), reciprocal changes in YFP and CFP intensity were coincident with the ratio change (Fig. 2); these ratio changes were thus likely to be the result of changes in calcium concentration. In the other cells, similar ratio changes were observed in the absence of reciprocal changes in YFP and CFP intensity. We believe that reciprocal changes in these cells were obscured by movement artifacts. In support of this view, a scatterplot of the data from all 27 ASH neurons (not shown) revealed that cells in which there was more movement in the x–y plane had higher joint variances in CFP and YFP intensities and were less likely to exhibit reciprocal changes. Nevertheless, the time course and amplitude of ratio changes in the nonreciprocal cases closely resembled the time course and amplitude in the reciprocal cases in this study, and in a previous one in which ASH neurons were imaged in fully restrained worms (Hilliard et al. 2005). This resemblance suggests that the nonreciprocal ratio changes we saw were caused mainly by changes in calcium concentration, not movement artifacts, and we thus included these data in our analysis. We conclude that calcium transients can be reliably detected in semirestrained worms.
We next asked whether behavioral responses persist in worms whose sensory neurons express cameleon, in the same set of 27 worms. This question was significant because cameleon, which contains the calcium binding protein calmodulin, could alter calcium concentration and its dynamics in the ASH neurons, which are responsible for triggering avoidance responses to CuCl2. Most worms (18/27, 67%) appeared to respond to the repellent by making an avoidance response, defined as a transition from forward to reverse swimming, with a latency of <10 s; one such response is illustrated in Fig. 2 and the supplemental movie. Of the other nine worms, six were engaged in a spontaneous bout of reverse swimming at the time the stimulus was delivered (so it was not possible to determine whether they responded to the stimulus), whereas three did not reverse. In addition, at the population level, we observed a significant increase in reverse swimming probability, relative to unstimulated control animals, in the 5 s after stimulus onset [Fig. 3A; t(33) = 2.815, P = 0.008]. We conclude that avoidance responses persist in worms that express the cameleon protein in the sensory neuron responsible for this behavior.
To assess the relative utility of simultaneous versus nonsimultaneous recordings, we analyzed the probability and ratio data in two different ways: first, using average probabilities and average ratios taken across animals (Fig. 3A) and, second, using probabilities and ratios taken from individual animals (Fig. 3B). The first method is statistically equivalent to seeking correlations in calcium and behavioral data recorded nonsimultaneously (i.e., from separate groups of animals) and serves as a point of comparison for the second method, which takes advantage of the dual-objective approach.
Using the first method, we plotted the average time course of the YFP/CFP ratio and the average time course of reverse swimming probability on the same time axis (Fig. 3A1). As noted earlier, this plot showed that a statistically significant increase in reverse swimming probability (relative to unstimulated controls) occurred in the first time bin after the stimulus, coinciding with the peak of the ASH calcium transient. However, a regression of average probability against average ratio demonstrated the absence of an overall correlation between calcium activity and behavior (Fig. 3A2; r = 0.011, P = 0.98). The absence of a correlation is somewhat surprising in light of previous evidence tightly linking ASH neurons to avoidance responses (Hart et al. 1999; Hilliard et al. 2002; Kaplan and Horvitz 1993; Maricq et al. 1995; Sambongi et al. 1999; Troemel et al. 1995). However, avoidance-response probability is also influenced by other neurons in C. elegans (Hilliard et al. 2002; Miller et al. 2005; Sambongi et al. 1999; Zariwala et al. 2003; Zheng et al. 1999).
Using the second method, we computed regressions of probabilities against ratios obtained simultaneously from individual worms in each of the 5-s time bins before and after the stimulus. We found a significant positive correlation between probability and ratio in the time bin immediately after stimulus onset (Fig. 3B1, r = 0.54, P = 0.008), but no significant correlations in the other time bins (Fig. 3B2, binwise values, P ≥ 0.07). Thus larger ASH calcium transients were associated with higher reverse swimming probability at precisely the time when reverse swimming probability was significantly elevated relative to unstimulated controls, suggesting that ASH contributes to the transient increase in reverse swimming probability. The fact that the second method was successful in revealing a correlation where the first method was not indicates that simultaneous recordings, made possible by the dual-objective approach, can detect subtle effects that may not be apparent in nonsimultaneous recordings.
This study demonstrates the practicality of making simultaneous recordings of neuronal calcium transients and whole animal behaviors using the dual-objective approach. This is the first time simultaneous recordings have been reported for animals and neurons on a spatial scale as small as that of C. elegans. An analogous technique has been used to record calcium transients and behavior in larval zebra fish (Higashijima et al. 2003; Ritter et al. 2001).
The dual-objective approach has two distinguishing features. The first feature is the simultaneous use of two objectives: 1) a low-power, wide-field objective for imaging behavior and 2) a high-power, narrow-field objective for imaging calcium transients. The second feature is the use of different wavelengths of light to prevent interference between the calcium imaging and behavioral optics. Wavelength separation was achieved by illuminating the preparation with red light, which was invisible to the calcium imaging optics. Conversely, a long-pass filter prevented the fluorescence excitation light from reaching the behavior camera.
The new approach has two main limitations. First, it is restricted to behaviors that survive semirestraint. In C. elegans, these include swimming, foraging, pharyngeal pumping, and egg laying. Second, it is restricted to recording neurons within, or quite near to, the glued region of the animal. In worms glued by the head, for example, tail neurons and many of the body-muscle motor neurons are likely to move too much to be reliably imaged.
The main advantage of the dual-objective approach is that it links with each frame in the optical recording of neural activity an image of the animal's behavior at the time the frame was captured. This linkage was achieved by synchronizing the starting point of the calcium and behavioral data streams and knowing the frame rates of the two data streams. By enabling us to plot probabilities against ratios obtained simultaneously from individual worms, the dual-objective approach revealed a correlation between reverse swimming probability and calcium transients in ASH chemosensory neurons that might otherwise have remained obscure. It should now be possible to seek correlations between the activity of other identified neurons and behavior in C. elegans.
The dual-objective approach is likely to be useful in many other types of experiments as well. These include situations in which it would be desirable to record simultaneously at two different spatial resolutions from a single location is space, such as when investigating the relationship between the activity of a single neuron and the network in which it is embedded, for example. Such experiments also include situations in which it would be desirable to record simultaneously from two different locations in space, such as when studying neurons separated by great distances, or a neuron and the appendage it controls. Finally, the dual-objective approach could also be adapted to photostimulation experiments in which the high-power objective is used to excite the phototrigger and the low-power objective is used to observe the resulting network activity or behavior (Boyden et al. 2005; Lima and Miesenbock 2005).
This work was supported by National Institute of Mental Health Grant MH-051383.
We thank C. Froekjaer-Jensen, R. Kerr, W. Schafer, and H. Suzuki for assistance with calcium imaging and R. Hodges (SD Instruments, Grants Pass, OR) for custom apparatus.
↵1 The Supplementary Material for this article (a movie) is available online at http://jn.physiology.org/cgi/content/full/01050.2005/DC1.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2006 by the American Physiological Society