Robotic Automation of In Vivo Two-Photon Targeted Whole-Cell Patch-Clamp Electrophysiology

Abstract

Whole-cell patch-clamp electrophysiological recording is a powerful technique for studying cellular function. While in vivo patch-clamp recording has recently benefited from automation, it is normally performed "blind," meaning that throughput for sampling some genetically or morphologically defined cell types is unacceptably low. One solution to this problem is to use two-photon microscopy to target fluorescently labeled neurons. Combining this with robotic automation is difficult, however, as micropipette penetration induces tissue deformation, moving target cells from their initial location. Here we describe a platform for automated two-photon targeted patch-clamp recording, which solves this problem by making use of a closed loop visual servo algorithm. Our system keeps the target cell in focus while iteratively adjusting the pipette approach trajectory to compensate for tissue motion. We demonstrate platform validation with patch-clamp recordings from a variety of cells in the mouse neocortex and cerebellum.

Keywords: automated neurophysiology; multiphoton microscopy; patch clamp; robotic automation.

Figure 1

Automated Two-Photon Guided Whole-Cell Recording In Vivo (A) Schematic of the apparatus, which consists of a conventional commercial two-photon microscope, a mode-locked Ti-Sapphire laser, a patch setup equipped with programmable three-axis micromanipulator, a signal amplifier, an analog to digital converter board, a computer, and a custom-made electro-pneumatic actuator for controlling micropipette internal pressure. (B) Block diagram of the two-photon guided robotic procedure. (C) Stages of the visually guided procedure: setup and pipette placement, tip and target coordinate acquisition, pipette positioning, automatic approach, position compensation, target cell engagement, seal formation, and break-in followed by whole-cell configuration. (D–F) Current-clamp traces (D) during current injection (400 ms-long pulses from −100 to +100 pA in 50 pA steps) and (E and F) at rest for a robotically patched gfp-positive neuron in the neocortex of a GAD67-gfp mouse (F is a zoomed-in detail of the underscored section of the trace in E). (G) Two-photon image of the patched neuron and the electrode.

Figure 2

Pipette Approach Trajectory and State during a Typical Robotic Two-Photon Targeted Patching Process (A) Automatic navigation of the pipette toward the target cell, with real-time feedback control of trajectory enabled. (B) Time course of pipette resistance, current, holding potential, internal pressure, and depth during the patching procedure (stages color coded; numeric labels correspond to points on the approach trajectory in A).

Figure 3

Further Examples of Robotic Two-Photon Targeted Whole-Cell Recording In Vivo (A and B) Current-clamp traces (A) during current injection (400 ms-long pulses from −100 to +100 pA in 50 pA steps) and (B) at rest for a patched neocortical interneuron in the V1 cortex of a GAD67-gfp (depth −113 μm from the brain surface; note compressed timescale relative to A). (C) Maximum-intensity z-projection of a two-photon stack image of the patched neuron and the electrode, acquired after the recording. (D and E) Current-clamp traces (D) during current injection (400 ms-long pulses from −100 to +25 pA in 25 pA steps) and (E) at rest for a patched neuron in the V1 cortex of a GAD67-gfp (depth −102 μm from the brain surface). (F) Maximum-intensity z-projection of a two-photon stack image of the patched neocortical interneuron and the electrode, acquired after the recording. (G and H) Current-clamp traces (G) during current injection (400 ms long pulses from −100 to +150 pA in 50 pA steps) and (H) at rest for a patched neuron in the V1 cortex of a GAD67-gfp (depth −142 μm from the brain surface). (I) Maximum-intensity z-projection of a two-photon stack image of the patched neocortical interneuron and the electrode, acquired after the recording.

Figure 4

Visually Guided Robotic Patching Achieves Similar Results to Manual Targeted Patching Comparison of robotic and manual two-photon targeted recordings obtained from neocortical interneurons in the V1 cortex (triangles for robotic, n = 15; x for manual, n = 14) and from Purkinje cells in the cerebellum (circles for robotic; n = 5). Comparisons are shown as a plot of input resistances obtained versus cell depth (A; left) and mean access resistance ± SEM for each condition (A; right), resting potential versus cell depth (B; left) and mean resting potential ± SEM for each condition (B; right), spike height versus cell depth (C; left) and mean spike height ± SEM for each condition (C; right), and holding time versus cell depth (D; left) and mean holding time ± SEM for each condition (D; right).