Multi-electrode patch-clamp recordings constitute a complex task. Here we show how, by automating of many of the experimental steps, it is possible to accelerate the process leading to qualitative improvement in performance and number of recordings.
The patch-clamp technique is today the most well-established method for recording electrical activity from individual neurons or their subcellular compartments. Nevertheless, achieving stable recordings, even from individual cells, remains a time-consuming procedure of considerable complexity. Automation of many steps in conjunction with efficient information display can greatly assist experimentalists in performing a larger number of recordings with greater reliability and in less time. In order to achieve large-scale recordings we concluded the most efficient approach is not to fully automatize the process but to simplify the experimental steps and reduce the chances of human error while efficiently incorporating the experimenter’s experience and visual feedback. With these goals in mind we developed a computer-assisted system which centralizes all the controls necessary for a multi-electrode patch-clamp experiment in a single interface, a commercially available wireless gamepad, while displaying experiment related information and guidance cues on the computer screen. Here we describe the different components of the system which allowed us to reduce the time required for achieving the recording configuration and substantially increase the chances of successfully recording large numbers of neurons simultaneously.
The capacity to record and stimulate multiple sites with micrometer precision is extremely useful for experimentally achieving a better understanding of neuronal systems. Many techniques have been developed to this end but none allow the submillivolt resolution achieved by the patch-clamp technique, essential for studying subthreshold activity and individual postsynaptic potentials. Here we cover the development of a twelve-electrode computer-assisted patch-clamp system aimed at simultaneously recording and stimulating a large number of individual cells with sufficient precision for the study of neuronal connectivity. While many other applications can be conceived for such a system, it lends itself particularly well to the study of synaptic connectivity given that the number of possible connections within a group of neurons grows proportionally to the square of the number of neurons in question. Therefore, while a system with three electrodes allows testing the occurrence of up to six connections and most often recording a single one, recording twelve neurons allows testing the occurrence of up to 132 connections and frequently observing over one dozen (Figure 1). The observation of dozens of connections simultaneously makes it possible to analyze the organization of small networks and infer statistical properties of the network structure that cannot be probed otherwise1. Moreover, precise stimulation of numerous cells also allows the quantification of recruitment of postsynaptic cells2.
1. Equipment Preparation
2. Patch-clamp Procedure
Following the methods described above we succeeded in performing whole-cell recording of up to twelve neurons simultaneously, nearly doubling the largest number of neurons simultaneously patch-clamped thus far. Examples of networks of direct synaptic connections between Pyramidal Neurons recorded in Layer V of the somatosensory cortex of rats are shown in Figure 6.
The determination of connection probability profiles as a function of inter-somatic distance for a given cell-type is a typical measurement of interest that can be performed efficiently with a multi-electrode patch-clamp system1. Recently photo-stimulation began to appear as an even more efficient means of obtaining such data3,4. Importantly, however, the data acquired with multi-electrode patch-clamp systems allows experimenters to obtain a more complete mapping of the network under study as well as high-resolution staining with intracellular diffused dyes. In multi-electrode patch-clamp experiments every cell can be recorded and stimulated, which is often not the case with other techniques such as photo-stimulation or calcium imaging. This feature allows experimenters to keep track of biases in connectivity, such as the high incidence of reciprocal connections, which cannot be measured otherwise. By stimulating and recording every studied neuron we showed that neurons are not only biased towards being reciprocally connected but also form clusters. The scale of our recordings also allowed us to observe that a higher connection probability existed among pairs of neurons that shared common neighbors, i.e. were both simultaneously connected to other individual neurons in the sampled network (Figure 7a). This observation was significant at every intersomatic distance bin of 50 mm (from 50 to 250 mm). We also observed pairs of neurons sharing many common neighbors occurred significantly more often than expected by chance (Figure 7b). Moreover, we detected an effect on connection probability according to the number of common neighbors shared by a given pair of neurons. The more common neighbors it shares the more likely a pair of neurons is to be interconnected (Figure 7c).
This tendency leads to the formation of groups of neurons that are more densely interconnected than average. Interestingly, highly interconnected groups of neurons not only exhibited more numerous connections but also, in average, stronger ones1. These findings led us to the conclusion that neocortical circuitry is not only organized in layers and columns but also into cell assemblies, i.e. groups of neurons sharing dense and strong synaptic interconnectivity as postulated by Donald Hebb several decades ago.
Other results of interest that can be achieved with multiple simultaneously patch-clamped neurons include the quantification of the recruitment of inhibition by supra-threshold activity in different numbers of excitatory cells2. A ubiquitous form of inhibition in the neocortex5 is mediated by Martinotti cells (Figures 8a and b, adapted from Berger et al.) which receive input from Pyramidal cells and integrate it in turn affecting, with some latency, other Pyramidal cells. We showed that, after brief bursts of spiking activity in as few as four Pyramidal Cells, every Pyramidal Cell in a local microcircuit receives this form of inhibition (Figures 8c, c, and f). We also showed that these inhibitory postsynaptic potentials tend to saturate when eight or more Pyramidal Cells are stimulated simultaneously (Figure 8E).
An overview of the system components can be seen in Figure 9. The software interface and hardware are shown in Figure 9a and b as well as the controller interface in Figure 9c guiding the electrodes towards cells for recording under microscope objective (Figure 9d).
Figure 1. Calculation of the number of connections observed in an experiment as a function of the number of neurons simultaneously recorded shows a quadratic growth. Numerical calculations (top). Illustrative diagram where further neurons of the same network are successively added (bottom).
Figure 2. Coordinate systems of each electrode manipulator (yellow) and microscope (red).
Figure 3. Screen capture of a pipette in approach towards assigned cell with overlaid approach vector, cell positions and scale bar.
Figure 4. Diagram illustrating pneumatic system for pipette pressure control.
Figure 5. Commands on the human interface device that centralizes controls used during patch-clamp experiments.
Figure 6. Example of three networks of direct synaptic connections mapped in individual experiments
Figure 7. Common neighbor effect. (a) Pairs of neurons that simultaneously connect to at least one other neuron in the sampled network (blue) exhibit a significantly increased probability of being interconnected. (b) Pairs of neurons sharing multiple common neighbors occur more often than expected by chance in the sampled networks. (c) Connection probability within pairs of neurons increases as a function of the number of common neighbors shared by the pair.
Figure 8. Quantification of the recruitment of Martinotti Cells. (a) Graphical representation of a Pyramidal Cell (red) that forms synapses onto a Martinotti Cell (blue) which in turn forms synapses onto a second Pyramidal Cell (black). (b) Supra-threshold stimulation of the Pyramidal cell (red) leads to recruitment of the Martinotti Cell (blue) through the integration of facilitating excitatory postsynaptic potentials. The Recruited Martinotti Cell then inhibits the second Pyramidal Cell (black). (c) Diagram representing the stimulation of increasing numbers of patch-clamped Pyramidal Cells and the effects on another Pyramidal Cell. (d) Average Inhibitory postsynaptic potentials recorded from a Pyramidal Cell as a function of the number of other nearby Pyramidal Cells that are stimulated. Adapted from Berger et al.2 (e) The amplitude of disynaptic inhibition in a local circuit tends to saturate when 9 or more Pyramidal Cells are stimulated indicating maximal recruitment of Martinotti Cells in probably reached at this point. (f) The fraction of cells receiving disynaptic inhibition quickly rises to 1 as increasing numbers of Pyramidal Cells are stimulated.
Figure 9. Illustration of ensemble of the system. (a) Graphical user interface with main window, live video display, log window and graphical representation. (b) Microscope and manipulators. (c) Human interface device with controls for experimental procedure. (d) Glass micropipettes in position for recording multiple neurons.
Figure 10. Binomial probability distribution functions that describe the fraction of experiments that successfully record a given number of cells given the use twelve electrodes. Comparison between the yield of experiments performed using visual feedback by experienced users are shown in blue and by less experienced users in non-ideal conditions in red.
An immediate question usually arises concerning the rate of success of the procedure we described. For high success rates preparation is essential. Pipettes must have tip openings that are adequate for the cells beings recorded. Filtering the intracellular solution to avoid clogged pipettes is also important. Extremely clean, freshly pulled pipettes are another requirement. A binomial distribution is the simplest model that can be used to understand how these issues affect the final yield. It is reasonable to expect an experimenter with experience and proper equipment to achieve a success rate of 80% or more in recording individual neurons with visual feedback. Beginners can be expected to achieve much lower success rates, especially if important preparation steps are overlooked. How these rates translate in numbers of cells recorded per experiment can be seen in Figure 10. Further increases in the number of simultaneously patch-clamped neurons will likely require miniaturization of manipulators and increased reliability of the procedure, which in turn requires substantial attention to detail, but are entirely feasible.
The versatility of the system presented here is still being explored and novel applications have frequently been found, in particular in the exploration of the relationship between extracellular signals and individual neuron activity7. Anatomical considerations become more important as the distance between recorded cells increases. Nevertheless, this system definitely allows investigation of long-range and inter-layer connectivity wherever connections remain intact following the slicing procedure.
Micromanipulator control
Enabling automatic positioning with the micromanipulators greatly accelerates the process of multi-electrode patch-clamp and increases its reliability reducing the occurrence of human errors. Different manufacturers provide different solutions in order to establish a connection from the PC to the manipulators. A common option is a serial or USB port. In order to ensure fast communication we dedicated a serial port to each manipulator controller. The most useful feature of automatic positioning is probably the elimination of most lateral and vertical movement of electrodes inside the tissue and limiting its distortion. As the movement of the electrodes inside the tissue takes place almost exclusively along the axial direction, mechanical interference is minimized. Lateral movements are only necessary to occasionally avoid blood-vessels and cells.
Visualize electrode positions
It is very useful to be able to track the position of each electrode during an experiment. In a traditional setup losing sight of an electrode can quickly become problematic. When using a large number of electrodes it is not possible to keep all electrodes within the field of view at all times. Relying on a graphic representation is the most intuitive alternative and also assists in keeping track of the progress of the experiment, instantly showing which electrodes have already been positioned in their final configuration and which ones have not.
Video acquisition and overlay
The ability to display live video from the microscope field of view on an application window is very useful. The display window was programmed to respond to mouse clicks enabling storage of positions of interest (cell somata) as well as quick update of the relative positions between microscope and electrodes removing accumulated errors whenever they appear. We also implemented the overlay of practical information on live video such as the approach trajectory for each electrode towards chosen cells or the marking of these cells (Figure 3). Registration of features and superposition of auxiliary images such as figures from anatomical atlases was also implemented to assist where regions of interest are not immediately clear.
Amplifiers
Computer controlled microscope amplifiers may allow control to take place from other applications as well. This drastically increases the speed and reliability with which multiple cells can be recorded simultaneously and eliminates the major source of human errors. Following computer-assisted positioning and pipette pressure control this is the step that produces the most noticeable gains in time but the gains in reliability are even more important.
Oscilloscopes
Ensuring that test signals can be visualized in real time at the appropriate scale and temporal resolution greatly enhances the efficiency with which an experimenter can execute experimental steps. By coupling the oscilloscope settings to amplifier modes (such as current or voltage clamp) we ensured that time-critical steps were executed and visualized with little effort such as holding cells at appropriate membrane potentials immediately after achieving the cell attached configuration. Proper visualization was ensured by sending the appropriate scaling and coupling commands through the serial port in the oscilloscope’s own protocol to fit the amplitude and offset of the test signals within the oscilloscope screen.
Pipette pressure control
As the number of electrodes employed in an experiment increases, ensuring that positive pressure is permanently applied to keep the tip of the glass electrodes clean becomes more demanding to the point of constituting an important hindrance. Sufficient positive and negative pressure (few hundreds of mbar) can be generated by simple membrane pumps. In order to stabilize the pressure, these pumps were coupled to reservoirs of approximately 100 ml whose opening and closing was controlled by pneumatic valves. The valves in turn were computer-controlled using a data acquisition card. The diagram of the pneumatic circuit can be seen in Figure 4. The pressure controller performs an important role not only ensuring pipette tips remain clean but also allowing the formation of Gigaohm seals quickly, upon the observation of dimples on the cell membrane as pipettes touch the cells, further accelerating the procedure.
The human interface device
Most experimental setups have the controls of experimental equipment widely dispersed over a large area. We concentrated the most commonly used controls on a single wireless gamepad (Figure 5) greatly reducing the time and effort required to record each cell but most importantly, eliminating sources of human error which can often bring large experiments to a premature end.
Programming language
We had very little choice in terms of the programming language for this application since the only language that supported integration of all the required equipment was C/C++. One great advantage of C/C++ is the possibility to implement multiple processing threads and fully profit from the performance improvement allowed by multiple-core processors.
Electrophysiological data acquisition
The system we described leaves the choice of electrophysiological recording software and data acquisition system up to the experimenter. Exchange of communications between data-acquisition software and our application can take place over serial ports or via socket communication over the network.
Future perspectives
More than 30 years since Sakmann and Neher’s seminal experiments, the patch-clamp technique is still alone in providing data with a particular combination of signal resolution and sampling frequency that are required for a wide variety of experiments, in particular those involving the detection of individual post-synaptic currents or potentials. By developing a computer-assisted system that enables recording of many neurons simultaneously we aimed at expanding the experimental possibilities of the patch-clamp technique. Combination of such new possibilities with recent developments in experimental neuroscience8-13 can open the way towards a more profound understanding of neuronal circuits with unprecedented speed and detail.
The authors have nothing to disclose.
We would like to thank Gilad Silberberg, Michele Pignatelli, Thomas K. Berger, Luca Gambazzi, and Sonia Garcia for valuable advice on improvements for the patch-clamp procedure automation. We thank Rajnish Ranjan for valuable advice and assistance with software implementation. This work was funded in part by the EU Synapse project and partly by the Human Frontiers Science Program.
Microscope | Olympus | BX51WI | 40X Immersion Objective |
Manipulators | Luigs & Neumann | SM-5 | Serial protocol used |
Amplifiers | Axon Instruments | MultiClamp 700B | SDK used |
Camera | Till Photonics | VS 55 | BNC analog output |
Framegrabber | Data Translation | DT3120 | SDK used |
Oscilloscopes | Tektronix | TDS 2014 | Serial communication |
Data acquisition | InstruTECH | ITC 1600 | |
Data acquisition | National Instruments | PCI-6221 | Library used (.dll) |
Pressure valve | SMC | SMC070C-6BG-32 | |
Pressure sensor | Honeywell | 24PCDFA6G | |
Membrane pump | Schego | Optimal |