This protocol presents a method for imaging neuronal population activity with single-cell resolution in non-transgenic invertebrate species using absorbance voltage-sensitive dyes and a photodiode array. This approach enables a rapid workflow, wherein imaging and analysis can be pursued over the course of a single day.
The development of transgenic invertebrate preparations in which the activity of specifiable sets of neurons can be recorded and manipulated with light represents a revolutionary advance for studies of the neural basis of behavior. However, a downside of this development is its tendency to focus investigators on a very small number of “designer” organisms (e.g., C. elegans and Drosophila), potentially negatively impacting the pursuit of comparative studies across many species, which is needed for identifying general principles of network function. The present article illustrates how optical recording with voltage-sensitive dyes in the brains of non-transgenic gastropod species can be used to rapidly (i.e., within the time course of single experiments) reveal features of the functional organization of their neural networks with single-cell resolution. We outline in detail the dissection, staining, and recording methods used by our laboratory to obtain action potential traces from dozens to ~150 neurons during behaviorally relevant motor programs in the CNS of multiple gastropod species, including one new to neuroscience – the nudibranch Berghia stephanieae. Imaging is performed with absorbance voltage-sensitive dyes and a 464-element photodiode array that samples at 1,600 frames/second, fast enough to capture all action potentials generated by the recorded neurons. Multiple several-minute recordings can be obtained per preparation with little to no signal bleaching or phototoxicity. The raw optical data collected through the methods described can subsequently be analyzed through a variety of illustrated methods. Our optical recording approach can be readily used to probe network activity in a variety of non-transgenic species, making it well suited for comparative studies of how brains generate behavior.
The development of transgenic lines of invertebrates such as Drosophila and C. elegans has provided powerful systems in which the neural bases of behavior may be optically interrogated and manipulated. However, these special preparations may have the downside of reducing enthusiasm for neural circuit studies of non-transgenic species, particularly with respect to the introduction of new species into neuroscience research. Focusing exclusively on one or two model systems is detrimental to the search for general principles of network function, because comparative studies represent an essential route by which such principles are discovered1,2,3,4. Our aim here is to demonstrate a large-scale imaging approach for obtaining rapid insight into the functional structure of gastropod neural networks, in an effort to facilitate comparative studies of neural network function.
Gastropod mollusks such as Aplysia, Lymnaea, Tritonia, Pleurobranchaea and others have long been used to investigate principles of neural network function, in large part because their behaviors are mediated by large, often individually identifiable neurons located on the surface of ganglia, making them readily accessible to recording techniques5. In the 1970s, voltage-sensitive dyes (VSDs) that can integrate into the plasma membrane were developed that soon enabled the first electrode-free recordings of the action potentials generated by multiple neurons6. Here, we demonstrate our use of VSDs to examine network activity in several species of gastropods, including one new to neuroscience, Berghia stephanieae. The imaging device is a commercially available 464-element photodiode array (PDA) that samples at 1,600 frames/second (Figure 1), which, when used with fast absorbance VSDs, reveals the action potentials of all recorded neurons7. The signals recorded by all diodes are displayed immediately after acquisition and superimposed on an image of the ganglion in the PDA acquisition software, making it possible to investigate neurons of interest with sharp electrodes in the same preparation8,9.
In the raw PDA data, many diodes redundantly record the larger neurons, and many also contain mixed signals from multiple neurons. A turning point was the development of an automated spike-sorting method using independent component analysis to rapidly process each raw 464-channel PDA data set into a new set of traces, in which every recorded neuron appears in a separate trace containing just its action potentials10,11.
In this article we outline the essential steps involved in obtaining large-scale action potential recordings from gastropod nervous systems with a photodiode array and fast absorbance VSDs. We moreover illustrate analytical methods that can be employed for clustering and mapping the optically recorded neurons with respect to their functional ensembles, and for characterizing population-level features that often are not apparent through simple inspection of the firing traces12,13.
NOTE: The workflow outlined below is summarized in Figure 2.
1. Minimize vibration
2. Run a pinhole test to enable proper alignment of ganglion photos with the PDA data
3. Dissections for three marine gastropod species
4. Stain the preparation with a voltage-sensitive dye
5. Flatten the preparation and set up for nerve stimulation
NOTE: The steps in this section should be performed under minimal illumination or with green light to minimize photobleaching.
6. Preparation and optimization for imaging
7. Optical recording
Tritonia
Skin contact with its seastar predator triggers Tritonia diomedea’s escape swim, consisting of a rhythmic series of whole-body flexions that propel it away to safety (Figure 3A). In isolated brain preparations, a brief stimulus to pedal nerve 3 (PdN3) elicits the rhythmic swim motor program (SMP) for this behavior, which is readily recognizable in optical recordings from the pedal ganglia. Figure 3B depicts the layout of a VSD imaging experiment designed to record the firing activity of neurons on the dorsal surface of the left Tritonia pedal ganglion, over which the PDA was positioned, as a stimulus to the contralateral (right) PdN3 elicits the SMP. Raw and filtered data (bandpass Butterworth filter, 5 and 100 Hz cutoffs) from 20 diodes recording activity before, during and after stimulation of PdN3 are shown in Figures 3C,D respectively. The nerve stimulus was delivered 20 s into the 2 min file. Immediately after acquisition, the signals measured by all 464 diodes of the recording array can be topographically displayed over an image of the preparation in the imaging software (Figure 3E). At this point, many traces contain spikes recorded redundantly from the same neurons, and some traces contain spikes from more than one neuron. Spike-sorting the filtered diode traces with ICA yielded 53 unique neuronal traces, 30 of which are shown in Figure 3F. The kinetics of individual spikes can be appreciated in Figure 3G, which expands an excerpt of four traces from Figure 3F (red box); the accuracy of the ICA spike-sorting algorithm was previously verified using simultaneous sharp electrode recordings, which showed that all spikes in the sorted traces correspond to intracellularly recorded spikes from individual neurons11,14.
Aplysia
A strongly aversive tail stimulus to Aplysia californica elicits a stereotyped two-part rhythmic escape response15. The first phase of the response is a gallop of several cycles of head lunges and tail pulls that move the animal quickly forward. This is typically followed by a period of crawling, involving repeated waves of head-to-tail muscular contractions that drive the animal forward at a slower speed for several minutes (Figure 4A). To capture these escape motor programs in optical recordings, the PDA was focused on the dorsal surface of the right pedal ganglion in an isolated brain preparation, and a suction electrode was placed on the contralateral (left) pedal nerve 9 (PdN9; Figure 4B). One minute into a continuous 20 min optical recording (Figure 4C), PdN9 was stimulated to elicit the gallop-crawl motor program sequence. The probabilistic Gaussian spatial distributions of the signals from all 81 recorded neurons were mapped onto the ganglion (Figure 4D). Dimensionality reduction applied to the full recording revealed that the gallop (cyan) and crawl (dark blue) phases of the escape program occupied distinct areas and formed different trajectories, spiral- and loop-like, respectively, in principal component space (Figure 4E).
Three videos based on the Aplysia recording depicted in Figure 4 demonstrate further types of analyses that can be performed on such data sets. Video 1 animates the firing of all recorded neurons over the full duration of the recording. The initial post-stimulatory period of the escape motor program was characterized by a gallop, in which activity in the ganglion was marked by alternating bursting of different functional clusters (Video 2). The gallop subsequently transitioned to a crawl, in which the activity across neuronal clusters remained broadly phasic but assumed a counterclockwise rotational trajectory in the ganglion (Video 3). The latter two videos also incorporate consensus clustering, which reveals the firing and locations of the different functional ensembles for the gallop and crawl phases of the escape response separately. Note that many neurons assigned to the same cluster in both the gallop and crawl phases exhibited physical proximity to one another in the ganglion, consistent with prior findings12.
Berghia
The aeolid nudibranch Berghia stephanieae (Figure 5A) represents a new model system for neuroscience. The imaging setup for a typical Berghia experiment is shown in Figure 5B. To elicit widescale neuronal activity, a suction electrode was placed on the most prominent left pedal nerve, and a nerve stimulus was delivered 30 s into a 2 min recording. ICA-processed traces revealed both spontaneous and stimulus-evoked activity in 55 neurons (Figure 5C). Community detection via consensus clustering identified ten distinct functional ensembles, which are depicted in Figure 5D in a network graph and in Figure 5E, which reorganizes the traces shown in Figure 5C based on their clustering assignments. Gaussian distributions of the signals from all recorded neurons are superimposed on an image of the preparation in Figure 5F to indicate the positions of all 55 recorded neurons.
Figure 1: Views of the optical imaging rig and photodiode array (PDA). (A) The optical imaging rig, featuring the PDA, digital camera, microscope, and stage. (B) The internal design of the PDA, in which fiber optics connect the imaging aperture to 464 photodiodes. A row of amplifiers is located above the photodiodes. (C) The hexagonal face of the imaging aperture, onto which the area being imaged is focused. Please click here to view a larger version of this figure.
Figure 2: A flowchart illustrating the essential workflow in obtaining optical recordings. The essential steps in the VSD imaging protocol, from dissection and staining through the details of imaging, are depicted in this flowchart. Please click here to view a larger version of this figure.
Figure 3: Results from Tritonia diomedea, illustrating raw, filtered, and spike-sorted data. (A) Tritonia escaping from the predatory sea star Pycnopodia helianthoides through its swim, which consists of alternating dorsal and ventral flexions of the body. (B) Schematic of the imaging setup. Ce=cerebral lobe of the cerebropleural ganglion; Pl = pleural lobe of the cerebropleural ganglion; Pd = pedal ganglion. (C) Raw data from 20 photodiodes, displaying activity in the left pedal ganglion to stimulation of the contralateral PdN3 (stimulus indicated by the arrow). (D) Filtered data from the same diodes as in C (5 and 100 Hz bandpass Butterworth filter). (E) Imaging software output in which compressed traces collected by all 464 diodes are superimposed topographically over an image of the preparation. The positions of the 20 diodes whose traces are shown in C and D are highlighted in red. (F) Thirty selected single-neuron traces generated by spike-sorting via ICA. (G) An expanded view of four single-neuron traces, corresponding to the red box in F, displays their action potentials at higher temporal resolution. Please click here to view a larger version of this figure.
Figure 4: Results from Aplysia californica, illustrating long-duration recording, signal mapping, and dimensionality reduction. (A) The two phases of Aplysia’s sequential escape motor program, the gallop and the crawl. (B) Schematic of the imaging setup. Ce = cerebral ganglion; Pl = pleural ganglion; Pd = pedal ganglion. (C) A 20 min recording of 81 neurons in the right pedal ganglion responding to a stimulus to the contralateral PdN9 (indicated by the arrow). Green, cyan, and dark blue bars below the traces indicate the pre-stimulus period, the gallop, and the crawl phases of the escape motor program, respectively. (D) An image of the preparation with mapped probabilistic Gaussian distributions of the locations of all 81 neuronal signal sources identified by ICA. The green outline represents the position of the hexagonal face of the PDA with respect to the ganglion. The numbers on each Gaussian correspond to the trace numbers in C. (E) Dimensionality reduction using principal component analysis plotting the first three principal components against each other over the course of the 20 min file. The pre-stimulatory baseline, gallop, and crawl epochs are shown in green, cyan, and dark blue, respectively. See Videos 1-3 for animations of neuronal firing corresponding to this recording. Please click here to view a larger version of this figure.
Figure 5: Results from Berghia stephanieae, a new species for neuroscience, illustrating network graphing, functional clustering, and bilateral signal mapping. (A) A Berghia specimen. (B) Schematic of the imaging setup. Ce = cerebral lobe of the cerebropleural ganglion; Pl = pleural lobe of the cerebropleural ganglion; Pd = pedal ganglion; Rh = rhinophore ganglion. (C) Traces displaying the spontaneous and stimulus-evoked activity of 55 bilateral neurons in the cerebropleural ganglia (delivery of the stimulus is indicated by the arrow). (D) A network graph displaying the ten functional ensembles, each assigned a unique color, identified through consensus clustering. Nodes in this plot represent neurons, where distance in network space represents the degree of firing correlation within and between ensembles. (E) The traces in C are rearranged and color-coded (following the color scheme of D) into functional ensembles. (F) An image of the preparation showing the mapped locations of the signals from every recorded neuron, and the trace numbers in C and E to which they correspond. Please click here to view a larger version of this figure.
Video 1: Animation of the full, 20-minute Aplysia escape locomotor program. The opacity of the white shapes overlying 81 individual neurons in the right pedal ganglion (left panel) was driven by the corresponding neuronal traces (right panel) and varied linearly as a function of average spike rate (binned per every 0.61 s of real time in the recording). For each neuron, full opacity was normalized to its maximum firing rate over the duration of the recording. One second of elapsed time in the video represents 12.2 s of real time. The scale bar corresponds to real time, with the green, cyan, and dark blue lines below the traces indicating the pre-stimulus baseline, gallop, and crawl phases of the escape locomotor program, respectively. The yellow boxes around the gallop phase and a portion of the crawl phase indicate the recording excerpts used to generate the animations in Videos 2 and 3. Please click here to view this video. (Right-click to download.)
Video 2: Animation of the gallop phase of the Aplysia escape locomotor program. Consensus clustering was performed on all 81 recorded neurons in just the gallop phase of the motor program to derive the functional ensembles, using the approach and software described and made available in ref.12. Neuronal ensembles exhibiting largely tonic or irregular firing patterns during this phase of the escape program were omitted from this video. The action potentials of the neurons belonging to the black and olive-green ensembles can be heard in the audio track of the video, with the corresponding neurons and traces highlighted. Average spike rates were normalized as in Video 1 and with equivalent time binning; 1 s of elapsed time in the video corresponds to 6.1 s of real time. Please click here to view this video. (Right-click to download.)
Video 3: Animation of the crawl phase of the Aplysia escape locomotor program. Consensus clustering was performed on all 81 recorded neurons in just the crawl phase of the motor program to derive the functional ensembles. Ensembles exhibiting largely tonic or irregular firing patterns during this phase of the motor program were omitted from this video. Average spike rates were normalized as in Videos 1 and 2 and with equivalent time binning; 1 s of elapsed time in the video corresponds to approximately 12.2 s of real time. Please click here to view this video. (Right-click to download.)
One of the most important details in implementing our large-scale VSD imaging approach is to minimize vibration, which produces movements of contrast edges across the diodes, resulting in large artifactual signals. Because absorbance VSDs produce very small-percentage changes in light intensity with action potentials, vibration artifacts, if not prevented, can obscure the neuronal signals of interest. We employ several methods to minimize vibration artifacts. First, our imaging room is situated on the ground floor, which isolates the preparation from vibrations related to building air-handling equipment and many other sources. Second, a spring-based isolation table was used, which other PDA users have confirmed provides better vibration dampening than the more common air table16. Third, water immersion objectives were used, which eliminate image fluctuations arising from surface ripples. Fourth, the preparation being imaged was lightly pressed between the chamber coverslip bottom and a coverslip fragment pressed down from above that is held in place by silicone plugs or petroleum jelly, further stabilizing the preparation. This also flattens the convex surface of the ganglion or ganglia being imaged, resulting in more neurons in the plane of focus of the objective, which increases the number of neurons recorded.
To maximize the signal-to-noise ratio for the very small changes in the degree of VSD light absorbance resulting from an action potential, it is essential to achieve near-saturating light through the preparation to the PDA, while at the same time minimizing photobleaching of the dye. To this end, we typically work at 3-4 V of resting light intensity, as measured with the PDA control panel gain switch in the 1x position (the PDA’s 464 amplifiers saturate at 10 V of light). During data acquisition this gain factor is changed to 100x. Getting sufficient light to reach 3-4 V as measured by the PDA can be accomplished in several ways. First, use an ultrabright LED light source that delivers a wavelength appropriate to the absorption properties of the absorbance dye in use. Accordingly, a 735-nm LED collimated lamp was used, which overlaps with the optimal absorption wavelengths of RH155 and RH482. Second, if necessary, use a flip-top substage condenser that concentrates the light from the LED light source to a smaller area. Third, adjust the condenser height to achieve Köhler illumination, which ensures high, even brightness and maximal image quality. Fourth, ensure there are no heat filters in the optical pathway, which can attenuate the LED lamp’s 735-nm wavelength. Fifth, remove diffusers, if more light is needed, from the optical pathway. Sixth, use high-NA objectives, which provide high spatial resolution, and allow sufficient levels of light to reach the PDA at lower lamp intensities. This has allowed us to minimize photobleaching to the extent that we can obtain several acquisition files of 10-20 min duration per preparation using the same light intensity across all files and without a significant loss of signal amplitude or the need for re-staining. Crucially, if the experimenter wishes to track neurons across these longer files, ensure that the focal plane does not change, and that the preparation does not move. Finally, an additional way to route sufficient light to the PDA is to use younger animals, which have thinner, and thus less opaque ganglia.
From time to time we find that the signal-to-noise ratio of the optical signals deteriorates and/or the motor program rhythms are suboptimal (e.g., slow or abnormal). When this begins to occur consistently, we mix fresh solutions of VSD. Aliquots of VSD typically remain viable for about 6 months in a -20 °C freezer. Relatedly, it is worth noting that for Berghia, the best results have so far been obtained with the absorbance VSD RH482. As RH482 is more lipophilic than RH155, it may better stain Berghia’s comparatively smaller neurons or remain in the neuronal membranes more effectively at the higher recording saline temperature used for this tropical species.
One limitation of PDA-based imaging of neural activity relates to the AC coupling of the voltage signals in hardware before the 100x preamplification step: although this represents a necessary feature to remove the large DC offset produced by the high resting light level required by this technique, the AC coupling intrinsic to the PDA precludes the measurement of slow changes in membrane potential, such as those associated with synaptic inputs. If recording slow or steady-state potential changes is desired, a DC-coupled CMOS camera imaging system can be utilized to capture subthreshold activity. Byrne and colleagues recently used such a setup with RH155 to image the activity of neurons in the buccal ganglion of Aplysia17,18. We have used both systems and found that the CMOS camera, due to its much higher density of detectors (128 x 128), generates 50x larger data files for the same imaging time7. The PDA’s smaller files facilitate faster processing and analysis. This also enables extended single-trial recordings (Figure 4) and studies of learning, in which data from multiple trials are concatenated into one large file before spike-sorting, allowing network organization to be tracked as learning develops19.
In other camera-based investigations, fluorescent VSDs have been used by Kristan and colleagues to examine network function in the segmental ganglia of the leech. In one influential study this led to the identification of a neuron involved in the animal’s decision to swim or to crawl20. In another study, Kristan et al. examined the extent to which the leech’s swimming and crawling behaviors are driven by multifunctional vs. dedicated circuits21. More recently, Wagenaar and colleagues used a two-sided microscope for voltage imaging that allows them to record from almost all neurons in a leech segmental ganglion22. In contrast to many camera-based imaging methods, an advantage of our PDA-based imaging method is rapid and unbiased spike-sorting by ICA, a form of blind source separation that involves no decisions about neuronal boundaries for results processing.
With respect to the choice of VSDs, one advantage of the absorbance dyes RH155 and RH482 is the little-to-no phototoxicity associated with them23,24, enabling longer recording times than is typical for fluorescent VSDs. Moreover, the fast absorbance VSDs we use are well suited for recording the overshooting somatic action potentials in gastropod preparations, which are typically 80 mV in amplitude. As shown in Figure 3G, our optical method can record action potential undershoots (none of our recordings are trace-averaged): this suggests that the VSDs we use should be able to discern action potentials in other model systems that attenuate to some degree and thus are not overshooting by the time they reach the soma. Nevertheless, our optical approach may not be ideal for species that are known to exhibit highly attenuated action potentials when recorded in the soma.
Much current research on neural networks is being focused on a small number of designer transgenic species. However, neuroscience benefits from the study of a wide variety of phylogenetically distinct species. Studying many different species provides insights into how circuits evolve25,26, and illuminates principles of network function that may be common across phyla1,2,3,4,27. We have so far applied our imaging method to a number of gastropod species, including Aplysia californica8,11,12,13,14,28, Tritonia diomedea8,9,11,14,19,28, Tritonia festiva28, Pleurobranchaea californica (unpublished data), and most recently Berghia stephanieae (Figure 5). An appeal of this approach is that it can be readily applied to many species, with no need for transgenic animals. We wish to acknowledge that our use of VSD imaging with fast absorbance dyes and a PDA follows in the footsteps of pioneering work that accomplished this in semi-intact, behaving Navanax29 and Aplysia30 preparations. Our emphasis on the rapidity of our approach is in part an answer to concerns that many investigators may be increasingly reluctant to initiate network studies in new species due to fears that years of study will be necessary to characterize basic network organization before being able to explore scientific questions of broad interest to neuroscience31. Accordingly, our goal here is to demonstrate a technique that greatly speeds the process – to the point that significant same-day insights into network organization may be obtained from single preparations.
The authors have nothing to disclose.
This work was supported by NSF 1257923 and NIH 1U01NS10837. The authors wish to acknowledge Jean Wang’s assistance in the laboratory.
Achromat 0.9 NA swing condenser | Nikon | N/A | |
Bipolar temperature controller | Warner Instruments | CL-100 with SC-20 | Controls perfusion saline temperature |
Chamber thermometer | Physitemp | BAT-12 with IT-18 microprobe | |
Digital camera | Optronics | S97808 | |
Dissecting forceps | Dumont | #5 | |
Dissecting scissors | American Diagnostic Corp. | ADC-3410Q | |
Imaging microscope | Olympus | BX51WIF | |
Imaging perfusion chamber | Siskiyou | PC-H | |
Instant Ocean | Instant Ocean | SS6-25 | Makes 25 gallons at a time |
Master-8 pulse stimulator | A.M.P.I. | Master-8 | |
Microdispenser | Drummond Scientific | 3-000-752 | Dye applicator for pressure staining |
Microdissection scissors | Moria | 15371-92 | |
Minutien pins (0.1 mm) | Fine Science Tools | NC9677548 | For positioning and stabilizing CNS |
Motorized microscope platform | Thorlabs | GHB-BX | Gibraltar platform |
NeuroPlex imaging software | RedShirtImaging | NeuroPlex | Compatible with the WuTech photodiode array |
Objective lenses | Olympus | XLPLN10XSVMP, XLUMPLFLN20XW, LUMPLFLN40XW, UAPON40XW340 | |
PE-100 polyethylene tubing | VWR | 63018-726 | Tubing to make suction electrodes |
Perfusion pump | Instech | P720 with DBS062SDBSU tube set | |
Petroleum jelly | Equate | NDC 49035-038-54 | |
Photodiode array with control panel | WuTech Instruments | 469-IV photodiode array | Contact jianwu2nd@gmail.com for ordering information |
RH155 | Santa Cruz Biotechnology | sc-499432 | Voltage-sensitive dye |
RH482 | Univ of Conn. Health Center | JPW-1132 | Voltage-sensitive dye; special order from Leslie Leow |
Silicone earplugs | Mack's | Model 7 | To be use for preparation compression |
Staining PE tubing | VWR | 63018-xxx | Different sizes depending on fit |
Sylgard 184 silicone elastomer kit | Dow Corning | Sylgard 184 silicone elastomer kit | |
Thorlabs LED and driver | Thorlabs | M735L2-C1, DC2100 | LED lamp and driver |
Tygon tubing | Fisher Scientific | 14-171-xxx | |
Vibration isolation table | Kinetic Systems | MK26 | Spring-based |