Performing reliable in vivo recordings with Neuropixels and other silicon probes

Why high-density probes?

Understanding how millions of neurons work together to allow our brain to function as it does – enabling learning, behavior, and ultimately consciousness – requires researchers to be able to measure the activity of these neurons. Despite all advances in imaging, electrophysiology remains the gold standard for such measurements.

Electrical recordings provide a temporal resolution and signal-to-noise ratio that optical methods are unable to compete with. Previously imaging methods unique strength was the ability to record from large numbers of individual neurons at the same time, compared with the low numbers of individual neurons accessible through single electrodes and tetrodes. Recent advances in the development of silicon probes have substantially increased the number and density of recording sites on single electrode shanks, enabling thousands to tens of thousands of individual isolated neurons to be recorded at the same time.

There is no doubt that imaging and electrophysiology bring different strengths to neuroscience research (indeed, the combination of the two is extremely powerful). The experiments where recordings with high-density silicon probes, such as Neuropixels, come into their own are those addressing questions where recordings with low signal-to-noise and high temporal resolution are required. Especially, if the neurons of interest are deep in the brain or spread over different brain regions.

Multiple probes to understand neural coding

The long shanks on silicon probe (up to several millimeters), make it possible to insert them in a manner that allows simultaneous recordings from different brain regions. Even with a single shank electrode. This can be taken one step further by using probes with multiple shanks, such as the new Neuropixels 2 probes (which have many other improvements such as smaller size, increased number of recording sites, tracking of neurons over days) or the custom probes by Neural Dynamics Technologies (which allow probes to be designed specifically for a given recording).

If an experiment calls for brain regions that are not on a single linear track to be recorded from (e.g. if they are in different hemispheres), or simply if more coverage is required, the next step is to move to multiple probes. This is increasingly more common, with detailed protocols being published (e.g. Durand et al. (2022), Nature Protocols).

Especially if the brain regions being targeted are not superficial, careful planning of insertion location and angles is required, to ensure that the regions of interested are all intersected. Mapping software such as Pinpoint (developed by the team behind Virtual Brain Lab) simplifies this task, by proving an overlay of a 3D brain map (multiple species are already available) and the geometry of the probes being used.

Making recordings reliable and reproducible

For the potential of such careful planning and mapping to transfer to successful experiments it is necessary that probe positions can be accurately and reliably adjusted – both for insertion locations and insertion angles. This is where the micromanipulators being used, as well as their mounting becomes important.

Stable design of both the mounting and manipulators minimizes vibrations, drift that would interfere with good recordings. Ensuring that manipulators are compact makes them less likely to pick up and amplify vibrations. Similarly, the mounting needs to be sturdy, allowing any adjustable components (e.g. for adjusting angles or coarse positioning) to be reliably and firmly clamped. Without changing the carefully adjusted setting.

The mounting of the manipulators should allow for accurate adjustment of insertion angles and positions. For this, it has proven useful to have the manipulators sliding on a spherical coordinate system, to adjust both the azimuth (phi) and elevation (theta) angles (see figure above). In addition, the orthogonal axes of the manipulator will allow parallel shifting of the probes. Using such a combination of coordinate systems centered on the target brain makes it straightforward to set the parameters determined in the initial planning and mapping phase. Clear readout of the set angles and manipulator positions is a big advantage here.

Especially when targeting tightly spaced locations, the ability to have several probes mounted at different elevation (“theta”) angles, but on the same azimuth (“phi”) angle is extremely convenient.

The final angle to consider is the rotation along the long axis of the electrode. This “spin” angle is particularly important when using multishank probes. Easy adjustment of this angle with a clear readout of the angle ensures that all the parameters determined in the mapping phase are adjusted correctly. This leaves on the insertion of the electrodes into the brain.

Slow insertion significantly improves the quality and amount of data – at insertion speeds of 2 um per second or less, more single units are detected with much better signal-to-noise ratio. Consequently, manipulators should be able to smoothly move with speeds down to 1 um per second (see Fiath et al. (2019) Scientific Reports). While this leads to a delay in the experiment of up to half an hour, the quality of data achieved fully justifies the care taken in the insertion.

If the dura mater is left intact, there will be notable dimpling of the brain surface. A number of strategies have evolved to reduce the dimpling (and associated potential damage), first and foremost sharpening the probe.

The final step towards ensuring a repeatable and controlled experiment is to rely on automation of the insertion phase.

Automating probe insertion

As described above, software such as Pinpoint helps with careful planning of the insertion coordinates. The more electrodes are used, the more important this planning gets, as risk of damaging expensive probes through collisions rapidly increases. Depending on the manipulators uses, Pinpoint and other software can receive position information from the different manipulators, and displays a live 3D image of the insertions. This visualization is approximate but helps keep an eye on the developing insertion.

Taking this one step further, the most recent version of Pinpoint and the associated Ephys Link software will actually drive supported manipulators (such as the full Sensapex uMp range). To achieve this, the software can work with a 3D model that includes the mounting angles of the micromanipulators with theta, phi and spin angles, as well as any orthogonal offsets.

The result of such automations are highly reliable and repeatable probe positions, making the resulting data substantially more reproducible.

Summary

The increasing use of high density silicon probes has been enabling research that just a few years ago would have appeared near impossible. But the use of increasing numbers of such probes in the same experiment also introduces new levels of complexity. A reliable recording rig, that provides stable and accurate manipulators, as well as a flexible mounting system that allows accurate positioning of the manipulators and thus the probes helps remove some of this complexity.

Today, a several in vivo recording rigs are commercially available that allow multiple manipulators and probes to be positioned around a brain. Considering the need for stability as well as the required adjustments, these solutions fulfil most or (in one case) all of these requirements. This allows researchers without easy access to mechanical workshops to undertake experiments that would previously have required substantial amounts of custom engineering work.

In the end, a good recording rig won’t automate all aspects of experiments, but it will help to ensure that the mechanics and basics are running smoothly, reliably, and most important repeatably. This frees up attention and time so that researchers can focus on acquiring good data.