The total possible number of connections probed using the traditional method of synaptic patch clamp recording can be expressed as a function of number of recorded cells (n), and number of pipettes in the multi-patch apparatus (p), as(1)possible connectionstraditional=np(p2−p).

Similarly, the total possible number of connections probed using the patch-walking method can be expressed as(2)possible connectionspatch−walk=(p2−p)+2(p−1)(n−p).

To visualize the advantage of patch-walking over the traditional method, these two equations can be represented as a matrix of potential probed connections. For example, the total number of possible connections using the traditional method and patch-walking for a two pipette apparatus is depicted in Figure 1A and B, respectively. Using these equations, patch-walking is always preferable in practice for n>p. Furthermore, one can expect the improvement in number of connections probed to approach double as n approaches infinity. For practical cases (apparatus with 2–8 pipettes), patch-walking yields 80–92% more probed connections, or efficiency, for experiments with 10–100 cells than the traditional synaptic connection searching method.

We built the apparatus (Figure 1C) and developed the software (Figure 1D) to perform patch-walking with two manipulators. We first conducted a dual-patch throughput experiment for two pipettes patching in a brain slice without testing for connectivity. In 33 patching attempts (18 attempts for pipette 1 and 15 attempts for pipette 2), we achieved whole cell success rates for pipette 1 of 44.4% (n=8/18 successful whole cells) and pipette 2 of 46.7% (n=7/15 successful whole cells). This is similar to success rates for manual patching as well as previously reported automated patch clamp robots (43–51% for Kolb et al., 2019). This result demonstrates the expected throughput and yield of these independent, uncoordinated pipettes.

Next we implemented the coordinated, dual-patching robot. In this series of experiments, we performed patch clamp attempts on 136 cells from 7 animals over a corresponding 7 days, with 2–3 slices per animal. Out of 136 patch clamp attempts for both pipettes, we achieved 71 successful whole cell recordings (52.2%). This is again comparable to previously reported automated patch clamp work such as Kolb et al., 2019 (51%), Wu et al., 2016 (43.2%), and Koos et al., 2021 (63.6% for rat, 37% for human). In addition, our success rates fall within the success range of manual users (30–80%) on differential interference contrast-based patch clamp systems (Vera Gonzalez et al., 2023). Thus, coordinating the motion of the pipette via the patch-walking algorithm does not deteriorate the success rate.

A representative brain slice with a box highlighting the experimental brain regions of interest is shown in Figure 2A. We patched in the somatosensory cortices as well as the primary visual cortex, primarily in L2/3, L4, and L5. In Figure 2B are histograms showing the distribution of the time it took to achieve a simultaneous recording (left, n=44), the intersomatic distance between neurons that were patch clamped simultaneously (center, n=44), and the time required to achieve gigaseal (time between increased resistance during neuron hunting step and achieving giga-ohm seal) for all cells (n=71). In Figure 2C are the distributions of whole cell properties (capacitance, tau, input resistance, resting membrane potential, and access resistance) of all cells. We were able to achieve paired patch clamp recordings between two pipettes in an average of 12.6±7.5 min as the pipettes walked across the slice. The average distance between two neurons for screened for connections was 91.6 ± 0.2 μm. The cells in paired recordings were held in whole cell configuration up to 45 min.

We demonstrate a connectivity matrix similar to those done by previous labs such as Peng et al. and the Allen Institute for Brain Science (Campagnola et al., 2022; Peng et al., 2019). Of the 71 whole cell recordings we recorded from the robot, we report a yield of 44 paired recordings using our patch-walking technique. In comparison, if we had screened the same 71 neurons for connections using the traditional method, we would have screened for 71/2=35 paired recordings. Therefore, our patch-walking algorithm screened for 9 more possible paired recordings for the 71 neurons we tested. Of those 44 recordings, 29 paired recordings (i.e. 58 probed connections) passed quality control checks and were used to validate the efficiency of the patch-walking algorithm, resulting in 3 found synaptic connections. From Equations 1; 2, we see the additional connections screened with the effect of patch-walking.

According to Perin et al., 2011, at an intersomatic distance of 91.6 ± 0.171 μm, the expected connection probability is 16.9% for each paired recording. Assuming they are independent, we would expect greater than 50% probability of getting at least one connection after just three paired recordings. According to binomial probability theory, we had a probability of 89% to find 3 connections with 29 paired recordings.

Figure 3 shows a connectivity matrix (as in Figure 1B), a spatial representation of the cells patched cells and connection probed, as well as a representative connection found between two cells. The matrix in Figure 3A shows the whole cell current clamp protocol described previously (black traces in leftmost column). During paired recordings, one cell would be stimulated in current clamp (traces along the diagonal). Recording color corresponds to a pair of cells tested for connectivity as in Figure 1, where each color has two traces because each pair of cells can be connected bidirectionally. The nomenclature for each row and column is n.p where n represents the cell number, in this case ranging from 1 to 7, since 7 total cells were patched between both pipettes, and p labeled either a or b represents each of the two pipettes. Cell 7 exhibited signs of decreased cell health, likely due to the duration of the experiment and increasing physical disruptions to the slice during patch-walking. The representative connection shown in more detail in Figure 3C was found between cells 1 and 2, with pre-synaptic cell 1 (black) stimulated and cell 2 (red) recording in voltage clamp the post synaptic currents elicited.

We designed and implemented an experimental apparatus to demonstrate the utility of patch-walking. The apparatus features a standard electrophysiology rig with two PatchStar micromanipulators. Samples (mouse brain slices) were imaged using a 40 X objective (LUMPLFL40XW/IR, NA 0.8, Olympus) on a motorized focus drive, illuminated under differential interference contrast microscopy (DIC) with an infrared light-emitting diode (780 nm), and captured with a Rolera Bolt camera (QImaging). We used a peristaltic pump (120 S/DV, Watson-Marlow) to perfuse the brain slices with buffer solution. We utilized the brain slice sample holder with integrated cleaning and rinse solution chambers as described previously (Kolb et al., 2016). We followed the cleaning protocol as suggested by Kolb et al., 2016, however we did not include rinsing in the cleaning protocol because recent literature found that there is no impediment to the whole cell yield or quality of recording (Landry et al., 2021; Peng et al., 2019).

Electrode pressure was controlled using a custom pipette pressure controller enabled up to four-channels, adapted from prior work (Kolb et al., 2019). Briefly, for each pipette, pressure was controlled by a±10 psi regulator (ProportionAir) using an analog (0–10 V) control signal. The control signal for each regulator was generated by a microcontroller (Arduino Due) via a digital-to-analog converter (MAX539, Maxim Integrated). In order to minimize valve switching to efficiently scale up the patcherBot to multiple manipulators and pipette pressure control, a custom printed circuit board was developed to control up to a maximum of four manipulators. Individual pressure regulators for each pipette were necessary to ensure that different pressures could be maintained on each pipette. The custom pressure controller regulates house-air line to deliver –500 to +700 mbar (relative to sea-level) using an inline venturi tube (SMC) and solenoid valve (Parker Hannifin) for rapid pressure switching (Kodandaramaiah et al., 2016; Kolb et al., 2019; Kodandaramaiah et al., 2012; Perszyk et al., 2021).

For real-time electrophsyiology feedback and collection, we used the Multiclamp 700B amplifier (Molecular Devices), cDAQ-9263/9201 and USB- 6221 OEM data acquisition boards (National Instruments), and Axon Digidata 1550B. The two data acquisition boards were each assigned to a manipulator in order to simultaneously acquire different signals from each pipette. This is particularly important for asynchronous and independent pipette control. Machine vision-based pipette calibration and position correction was performed according to Gonzalez et al., 2021 to correct for small micromanipulator position errors.

Following brain slice preparation and pipette fabrication, filling, and installation, pipette location was calibrated according to Kolb et al., 2019 for all pipettes, resulting in a ‘home’ position as described for each pipette.

All animal procedures were in accordance with the US National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at the Georgia Institute of Technology (A100359). For the brain slice experiments, male mice (C57BL/6, P19–P36, Charles River) were anesthetized with isofluorane, and the brain was quickly removed. Coronal sections (300 µm thick) were then sliced on a vibratome (Leica Biosystems VT1200S) while the brain was submerged in ice-cold sucrose solution containing (in mM) 40 NaCl, 4 KCl, 1.25 NaH₂PO₄·H₂O, 7 MgCl₂, 25 NaHCO₃, 10 D-Gluocse, 0.5 CaCl₂·2H₂O, 150 Sucrose (pH 7.3–7.4, 300–310 mOsm). The slices were incubated at 37 °C for 1 hr in neuronal artificial cerebro-spinal fluid (aCSF) consisting of (in mM) 124 NaCl, 2.5 KCl, 1.25 NaH₂PO₄·H₂O, 1.3 MgCl₂, 26 NaHCO₃, 10 D-Glucose, 2 CaCl₂·2H₂O, 1 L-Ascorbate·H₂O (pH 7.3–7.4, 290–300 mOsm). Prior to recording, the slices were maintained at room temperature for at least 15 min (22–25 °C). The sucrose solution and neuronal ACSF were bubbled with 95%O₂/5%CO₂. Recordings were performed in mouse primary visual area and somatosensory cortex.

Borosilicate pipettes were pulled on the day of the experiment using a horizontal puller (P-1000, Sutter Instruments) to a resistance of 4–6 MΩ. The intracellular solution was composed of (in mM) 135 K-Gluconate, 10 HEPES, 4 KCl, 1 EGTA, 0.3 Na-GTP, 4 Mg-ATP, 10 Na2-phosphocreatine (pH: 7.2–7.3, 290–300 mOsm). Recordings were performed at room temperature with constant perfusion of oxygenated neuronal aCSF. Pipette pressure during patch clamp steps was digitally controlled and pipettes were cleaned according to Kolb et al., 2016; Kolb et al., 2019, as previously described.

The patch-walking algorithm is depicted schematically in Figure 4. At the beginning of each patch-walking experiment, cells are first selected by the user. For each slice experiment, we selected 8–10 healthy cells located 20–100 µm below the surface of the tissue. These cells were spread across an area of approximately 200 µm x 200 µm (note the field of view under ×40 magnification is approximately 50 µm). These cell locations with three dimensional coordinates are stored in a cell queue for subsequent patch attempts. From these cell coordinates, the robot computes the distance between the pipettes’ respective home positions and each cell.

The assignment of cells to pipettes proceeds as follows. When a pipette is available for a patch clamp attempt, the cell with the shortest distance to the pipette home position is removed from the cell queue and assigned to the pipette. Thus, each pipette initially attempts to patch its closest cell. To probe connections between neurons, both pipettes must form a successful connection. Therefore, in the event that an attempted connection fails, the pipette is retracted and the next closest cell is assigned from the queue. This process is repeated until both pipettes are simultaneously connected to neurons. At this moment, the connections can be tested as described in ‘connectivity testing.’ A pipette that achieves whole cell configuration is held while patch attempts are made with the other pipette. Unsuccessful patch attempts result in the pipette being cleaned (Kolb et al., 2019), assigned a new cell and then a new attempt. For each patch clamp attempt, control of the microscope objective is assigned to the pipette in the ‘pipette finding’ or ‘neuron hunting’ phase, until the pipette has successfully completed ‘neuron hunting’ (when the pipette resistance increases by 0.2 MΩ over 5 descending 0.1 μm steps) (Kodandaramaiah et al., 2018). Once both pipettes have established whole cell patch clamp recordings, the connection test is performed. This algorithm then repeated the process until all viable neurons had been patched.

The ‘patch cell’ step in Figure 1 includes the following: neuron hunting, neuron detection, gigasealing, break in, whole cell protocol. As in Kolb et al., 2019, and briefly restated here, once the measured resistance reaches 1 GΩ, the algorithm waits (5 s) and proceeds to the break-in state. Break-in is accomplished by short pulses of suction (100–1000 ms, –345 mBar). A break-in is considered successful when the measured resistance drops to under 800 MΩ and the holding current remains low (<–200 pA at –70 mV in slices). The whole cell electrophysiology protocol consists of a voltage clamp protocol where cell parameters (access resistance, membrane resistance, holding current) are measured as well as a current clamp protocol (0 pA for 1 s, –300 to +300 pA step for 1 s, 0 pA for 1 s). Injected current pulses were 3 s pulses from –20 pA to +280 pA in 20 pA steps with a 2 s, –20 pA hyperpolarizing step 500ms prior.

We define a paired recording as a pair of cells which are simultaneously patch clamped using the patch-walking experimental method. We define probed connections as twice the number of paired recordings, because the connections can be bi-directional. We define possible connections as the theoretical upper limit of probed connections given a number of pipettes and number of cells recorded. Probed connections is, practically speaking, less than possible connections since patch clamp yield is 30–80%based on our experience with the PatcherBot, depending on sample preparation and tissue and cell type.

The access resistance for the neurons in paired recordings were below 40 MΩ, similar to the metric used by Kolb et al., 2019, and if the access changed above 50 MΩ, we stopped recording from that neuron. If the seal quality decreased during recording, the cell is excluded from analysis.

We tested for connectivity in a manner similar to that done previously by Perin et al., 2011 and the Allen Institute for Brain Science (Campagnola et al., 2022). To perform connectivity testing between two simultaneously patched neurons, we performed the following procedure. The BNC cables were manually moved from the NI DAQ to the Digitizer to enable Clampex control of the cells (rather than LabView). Two protocols were run in order to test for the two possible directions of connectivity. For each protocol, one pipette sent a stimulus in current clamp mode to elicit five action potentials at 20 Hz while the other pipette holding in voltage clamp recorded post-synaptic currents, held at –70 mV. Following this bi-directional measurement, the BNC cables were replaced manually to resume patch-walking.