This protocol describes a procedure for constructing carbon fiber microelectrode arrays for chronic and acute in vivo electrophysiological recordings in mouse (Mus musculus) and ferret (Mustela putorius furo) from multiple brain regions. Each step, following the purchase of raw carbon fibers to microelectrode array implantation, is described in detail, with emphasis on microelectrode array construction.
Multichannel electrode arrays offer insight into the working brain and serve to elucidate neural processes at the single-cell and circuit levels. Development of these tools is crucial for understanding complex behaviors and cognition and for advancing clinical applications. However, it remains a challenge to densely record from cell populations stably and continuously over long time periods. Many popular electrodes, such as tetrodes and silicon arrays, feature large cross-diameters that produce damage upon insertion and elicit chronic reactive tissue responses associated with neuronal death, hindering the recording of stable, continuous neural activity. In addition, most wire bundles exhibit broad spacing between channels, precluding simultaneous recording from a large number of cells clustered in a small area. The carbon fiber microelectrode arrays described in this protocol offer an accessible solution to these concerns. The study provides a detailed method for fabricating carbon fiber microelectrode arrays that can be used for both acute and chronic recordings in vivo. The physical properties of these electrodes make them ideal for stable and continuous long-term recordings at high cell densities, enabling the researcher to make robust, unambiguous recordings from single units across months.
Electrodes and electrode arrays are valuable tools for understanding how the brain processes information at the neuronal level. While electrophysiological recordings have been achievable for over two centuries1, it is still not possible to simultaneously measure the activity of entire neural circuits at the spatial and temporal resolution required to capture the spiking of individual neurons. Although non-invasive methods, such as electroencephalography2, positron emission topography3, and functional magnetic resonance imaging4 allow for whole-brain measurements, they cannot achieve the spatial and temporal resolution necessary for resolving the activity of neural circuits2,5. In contrast, imaging methods such as optical imaging using voltage-sensitive dyes or genetically encoded calcium indicators can achieve single-unit spatial resolution, but they pose issues such as low temporal resolution and poor selectivity3,4,5,6. Electrical recordings are a powerful alternative to these methods. Recording electrodes provide unparalleled temporal resolution and allow the user to make measurements with spike-time precision in any region of the brain7. Additionally, chronically implanted multielectrode arrays (MEAs) enable large-scale (tens to hundreds of cells), single-cell recordings in behaving animals over a period of days to months8,9. However, silicon probes that record at higher densities have a large footprint and are highly invasive, and chronically implanted arrays often generate an inflammation response, tissue encapsulation, and neuronal death10,11,12,13.
The limitations of existing electrodes have resulted in recent innovations that allow for stable, high-resolution, long-term recordings. Typical electrodes consist of a metallic conductor, such as tungsten or platinum-iridium, or are silicon- or polymer-based. While metal-based microwire arrays can maintain long-term, stable recordings, they have a much larger footprint, with a single wire's diameter ranging from 10-200 µm14. In contrast, silicon-based electrode arrays yield recordings with high spatial resolution, but due to their relatively rigid design, they are typically unable to maintain the signal and record from the same neurons over many months15. Recent developments in silicon-based arrays have resulted in electrodes that can reliably perform chronic recordings, but these arrays cannot be used to record from deep brain regions in larger animals and are intended for linear recordings9. Advances in polymer arrays have resulted in increased flexibility and recording stability of single units and offer the potential for high-density recordings in the near future but with limited availability at present8,16,17. Carbon fibers allow for high-density recordings with off-the-shelf materials that are described here.
Carbon fiber recording microelectrodes have been used for decades, with the first carbon fiber electrodes consisting of a single carbon-fiber inserted into a glass micropipette. These microelectrodes were used for single-unit extracellular recordings, and although the signal-to-noise ratio was comparable to the best tungsten-in-glass microelectrodes, they were advantageous due to their flexibility, lower impedance values, and simplicity to manufacture18,19. Efforts to develop carbon fiber electrode arrays have recently accelerated due to the biosensing capabilities of carbon fibers. In addition to their increased biocompatibility and exceptional electrical conductivity, they feature a unique set of properties, including high-temperature resistance, low relative density, high tensile strength, low bending stiffness, high detection sensitivity, and a small cross-sectional area10,12. All of these properties have motivated the development of carbon fiber microelectrode arrays (CFEAs) that facilitate chronic, stable, high-yield recordings of single neurons. Such CFEAs can now be crafted by hand20,21 (Figure 1), yielding microelectrode arrays that can hold single neurons over months. Described here is an accessible construction process for CFEAs that has been adapted in two ways for acute and chronic recordings of individual neurons in two species.
All experimental procedures were approved by the Brandeis University or Washington University Animal Care and Use Committee. Data shown were collected from one female ferret and one male mouse.
1. Preparation of carbon fibers and tools
2. Design and fabrication
3. Assembling the carbon fiber microelectrode array (CFEA)
NOTE: This step takes ~2 h for an experienced builder and ~6 h for a novice builder. Perform all CFEA assembly steps and fiber bundling steps under a 10x stereo microscope. Complete assembling the CFEA in an environment with minimal air movement, as this may disturb the building process.
4. Fiber bundle packaging
NOTE: It takes approximately 30 min to perform this step. Complete this step for the electrodes used in animal models with a thick layer of pia mater. Reinforce the fiber bundle to minimize bending. In mouse procedures, this step may not be necessary.
5. Electrode tip preparation
NOTE: It takes approximately 30 min per array to perform this step.
6. Insertion in the brain: Survival surgery, mouse (Mus musculus) and non-survival surgery, ferret (Mustela putorius furo)
NOTE: Surgical procedures should follow standard protocol in compliance with IACUC. For detailed information see Ma et al.22 for survival surgery protocol and Popovic et al.23 for non-survival surgery protocol. Follow the aseptic surgical procedures per the ASC guidelines for survival surgery in rodent species. These include autoclaving all surgical tools and materials at 135 °C for 15 min and treating the stereotaxic apparatus and surgical area with 70% ethanol. Use sterile surgical gloves, a disposable gown, and face mask during the procedure.
With the completion of this protocol, stable recordings of single-unit spiking activity will be possible. These microelectrode arrays are customizable in material, channel count, and headstage adapter based on the researcher's needs. Electroplating fibers in gold results in decreased impedances suitable for recording (Figure 4 and Figure 5). If the user intends to record chronically, measurements can be made after the animal has recovered from the surgical procedure. Chronic procedures have resulted in stable, single-unit recordings for at least 120 days. A representative recording is shown in Figure 6, illustrating stable 64-channel electrophysiological activity in the retrosplenial cortex of a freely behaving, adult male mouse. If an acute preparation is intended, recordings can begin shortly after implantation (~30 min). This will allow time for the electrode to settle in the brain. Figure 7 provides a representative example of an acute 16-channel CFEA recording acquired from the primary visual cortex of an adult female ferret. Spike sorting in mouse and ferret was performed with spike sorting software (see Table of Materials).
Figure 1: Anatomy of 16- and 32-channel carbon fiber microelectrode arrays (CFEAs). (A) Schematics of 32-channel (top) and 16-channel (bottom) CFEA from three different views. The 16-channel CFEA features an extended design for handling purposes. The 32-channel design features a flat face that allows for two jigs to be combined for a 64-channel CFEA. Both the diagrams have identifying structures labeled with dimensions. The connector end indicates the location of the connector insertion, and GND/REF channels indicate where the grounding wire is inserted. The funnel basin refers to the location that the fibers pass through to be overlaid with UV light-cured dental cement, and the funnel tip signifies the site from where the fibers exit the jig. The funnel tip is divided into quadrants to minimize fibers clinging together and creating damage. The fibers are later pulled into a single bundle with the use of the dental cement. Jigs are 3D printed using SLA resin printers. Diagrams are enlarged to show details. (B) Constructed CFEA. Diagram has identifying structures labeled. The blue bundle tip represents the segment of the carbon fibers that acquire recording measurements. The gray within the funnel basin and surrounding the connector is indicative of UV light-cured dental cement that holds carbon fibers in place in the funnel basin and secures the connector to the jig. The purple wire represents the grounding wire. Please click here to view a larger version of this figure.
Figure 2: Loading of raw carbon fibers into cassettes for parylene C coating. (A) Carbon fibers are loaded onto cartridges overlaid with two strips of double-sided tape (blue). Each cassette is loaded with ~25 fibers. (B) Cassettes are loaded into a laser-cut holder (gray) in preparation for parylene C coating. Each holds ten cassettes. Please click here to view a larger version of this figure.
Figure 3: Carbon fiber microelectrode array (CFEA) bundle construction schematic. (A) 16 individual, coated carbon fibers (black) are threaded through the 32-channel 3D-printed jig (gray). (B) Carbon fiber tips are cut with micro-scissors, leaving excess fiber equal to the height of the jig base, extending out of the jig base. (C) A standard plastic spark wheel lighter is quickly passed over the excess fiber to remove parylene C insulation. The top right schematic shows the removal of parylene from 9 of the 12 fibers. (D) Fibers are reinserted into the jig until the fiber end is flush with the base. The top-right schematic shows the reinsertion of 9 fibers with uninsulated (gray) fiber tips housed inside the jig base. The jig is then flipped over and steps A-D are repeated to thread the opposite 16 channels. (E) The jig is filled with dental cement to secure the fibers. Silver print is injected into each well of the jig base. (F) The male connector is inserted into the jig base. (G) CFEA and scalpel are frozen in a -20 °C freezer. The array tip is cut to the desired length, leaving 32 even fibers. Please click here to view a larger version of this figure.
Figure 4: Tip treatment and electroplating. (A) Electrode tips are first placed into 0.1 M PBS, where current is passed through each electrode. The tips are then rinsed and transferred to a gold plating solution, where they are electroplated with the current. (B) SEM images of prepared carbon fiber show gold plating solution concentrated at the tip. Scale bar represents 4 µm. (C) Impedance values from 168 channels after initial cutting (purple; 3.11 MΩ ± 0.42 MΩ, median ± SE, n = 168 fibers), positive current injection (pink; 1.23 MΩ ± 0.36 MΩ, median ± SE, n = 168 fibers), and electroplating (orange; 0.19 MΩ ± 0.15 MΩ, median ± SE, n = 168 fibers) show decreased impedance values after each processing step. Please click here to view a larger version of this figure.
Figure 5: Moderate gold electroplating durations produce small, rounded deposits on carbon fiber bundle tips. The carbon fiber tips pictured are all from different microelectrode arrays, reflecting different durations of injected current for impedance reduction or gold-plating. Images additionally depict the parylene C coating, which insulates the carbon fibers and prevents any acquisition of signal from a location other than the tips of the fibers. (A) Scanning electron microscopy image of carbon fiber tips after freezing and making a single cut with a razor blade. Scale bars represent 10 µm. (B) Same as A but then followed with injection of positive current for 10 s. (C) Same as B but then electroplated with gold for 5 s. (D) Same as B but then electroplated with gold for 15 s. (E) Same as B but then electroplated with gold for 30 s. (F) Same as B but then electroplated with gold for 120 s. We found that electroplating for 30 s at a current of -0.05 µA was optimal for electrophysiological recordings. Please click here to view a larger version of this figure.
Figure 6: Chronic extracellular recordings in freely behaving mouse retrosplenial cortex with carbon fiber microelectrode arrays show persistent, stable neural activity. (A) Eleven bandpassed voltage traces were recorded simultaneously. Subsequent traces recorded from the first channel (top row) are plotted in B to show durability across time. The remaining ten rows demonstrate the consistency of recording quality and show robust activity across the array. Scale bar to the left of each trace represents a 200 µV potential. (B) Bandpassed data from the same fiber as in the top trace in A, expanded to show robust activity across a 120-day continuous recording. (C) Clustering reveals robust single unit detection over months. Traces represent the average waveform of a continuously observable representative single unit across 120 days, extracted from the fiber plotted in B at each timepoint. (D) Mean, non-normalized spike waveforms from C stacked to demonstrate consistency over time. (E) Carbon fiber recordings demonstrate a stable noise floor over many months. Standard deviation of the noise floor (trace minus spiking activity) in B shows no progressive change in noise. Bars represent mean contamination. Error bars represent standard deviation. (F) Scale drawing of a mouse with a chronically implanted CFEA and headstage. (G) Raw voltage trace (top) 11 months after implantation shows robust LFP. Bandpassed voltage trace (bottom) shows steady neural activity. (H) Mean spike waveform of the neuron recorded on the fiber from C, underlaid by the first 1,000 incidences of spiking activity. Please click here to view a larger version of this figure.
Figure 7: Carbon fiber microelectrode array (CFEA) recordings from the ferret primary visual cortex. (A) Waveforms of spike sorted single units recorded from a 16-channel CFEA. Action potentials from single neurons were often evident on multiple channels at slightly differing amplitudes. (B) Direction tuning curves from selected neurons. Colors correspond to recorded units in A. Arrows indicate the direction of stimulus movement. Scale bars indicate the response rate. Error bars indicate the mean response with standard error. The horizontal dashed line represents the same cell's spontaneous firing rate during exposure to a blank screen. Please click here to view a larger version of this figure.
This protocol describes each step necessary for constructing a functional CFEA for both acute and chronic use. The process described is customizable to the researcher's needs, making it an accessible and inexpensive option for monitoring single neurons over months. The protocol demonstrates the feasibility of recording both robust single-unit activity within minutes of implantation in an anesthetized animal, and across four months in an awake, behaving animal, illustrating the potential of these CFEAs to study short-term and long-term changes in neural responses.
The steps of the protocol described have been thoroughly tested and improved upon over time to yield an efficient procedure that can be completed quickly, at a low marginal cost (<$100.00), with the capability of recording unambiguous single units, densely and stably over months. The construction steps can be completed in less than one day and will produce electrophysiological signals that are comparable to any leading commercial array. The CFEAs also have a much smaller footprint (16-channel bundle of fibers has a diameter of ~26 µm) than similar commercial arrays, and their biocompatibility makes them suitable for long-term use13. Importantly, there are several critical steps and instructions that must be followed in order to produce a functioning CFEA with comparable performance.
Due to the fragility of the carbon fibers, they must be handled with utmost care. Handling them with sharp forceps or other tools may result in breakage of the fibers. Additionally, it is important to construct the CFEAs in a space with limited air movement so that the fibers do not blow away. When flaming the back portion of the fibers, the lighter only needs to be moved in a back-and-forth motion very briefly, for approximately 1 s. The steps following this removal of insulation are crucial for constructing an electrode with working channels. The flamed tips should be fed into the jig without any additional contact. Then, when filling the basin with dental cement, it is important that the cement is carefully applied and completely fills the channels and funnel basin, closing off the openings without filling them. The dental cement should then be completely cured with UV light before proceeding. Once this is complete, silver paint should be injected into each channel until completely filled but not spilling out. This is the most variable step in the process. Any over-filling can produce crosstalk between channels, and insufficient filling can result in a connection failure. If unable to inject silver paint using a 25 G needle, it is likely that the solution is too viscous and, in this case, a small amount of paint thinner can be added to create a more fluid solution. Once all the channels are filled, and the headstage connector is inserted, it is important to allow the array to cure for 24 h prior to securing the connector with dental cement. We found that failure to do so lowered the number of connected channels. Applying a generous amount of dental cement is also important so that the connector does not disconnect when interfacing with the signal acquisition system. If they become detached, it is possible to attempt reconnection with the repeated filling of channels with silver paint, but the user should test the impedance values of the CFEA to assess the number of connected channels. Allowing the dental cement to cure overnight also serves to prevent potential detachment.
Measuring the impedance of the electrode will provide an accurate estimation of connected channels. This can be done after submerging the ground and reference wires and the carbon fiber tips in PBS. We have observed that a high impedance (>15 MΩ) is indicative of an open, unconnected channel. Prior to injecting current and electroplating, a connected channel can have a range of impedance values that should significantly decrease with this process. The average number of connected channels (impedance < 4 MΩ after current injection) per 16-channel electrode was 12.96 ± 2.74 (mean ± SD; N = 48 electrodes). A number of electroplating times were tested, and 30 s produced superior signal isolation among the recording sites (Figure 5). While it has been well established that PEDOT-pTS12,24,25,26 and PEDOT-TFB21 provide reliable options for preparing carbon fiber recording sites, we found that plating with gold, a proven and dependable method for electroplating electrodes for chronic implantation27,28, increased the ease of implantation and prevented the electrode tips from clumping together. In producing final impedance values of less than 0.2 MΩ on average, this method proves comparable to values achieved using PEDOT-TFB21 and PEDOT-pTS26.
When implanting the microelectrode array, it is important to visually follow the insertion of the carbon fiber tips under the microscope. Successful insertion should be apparent, with no bending of the fibers. If the fibers appear to be buckling, it is unlikely that they will successfully enter the brain. In this case, the angle of the probe should be adjusted for a second attempt. This process can continue until the insertion of the probe is successful. Once the electrode is at the desired depth, we have found that waiting at least 30 min will allow the probe to settle for optimal signal acquisition (acute recordings).
The CFEAs described, in addition to their small footprint and biocompatibility, offer a robust, customizable alternative to commercial arrays due to their ease of construction and low cost. The greatest limitation to the CFEAs detailed in this protocol is their scalability. Due to the manual nature of their construction, scaling up to designs with hundreds of recording sites may not be practical. Additionally, advances in microelectrode array fabrication using nanotechnology will enable larger-scale population recordings than the methods described here. However, this protocol delivers CFEA accessibility to labs interested in benchtop fabrication of carbon fiber electrodes. We observed no loss of stability or decreased robustness in spike amplitude over the duration of the 120-day chronic experiments, as indicated by a representative single channel typical of our observations on that time scale (Figure 6A–E). Additionally, the CFEAs show the capacity for persistent single-unit activity, as four single units remained discernible 11 months after implantation in mouse (Figure 6G,H). It is also possible to obtain stable, single-unit recordings acutely (Figure 7), which offers an advantage over many other commercial electrodes for the study of single neurons over short time periods. In the future, the development of such flexible, biocompatible probes with minimal diameters will enable the study of complex processes. These tools will provide substantial utility in the advancement of neural technology, including applications in brain-machine interfaces (BMIs), which require continuous, long-term stability29.
The authors have nothing to disclose.
We would like to thank Greg Guitchounts for guidance with electrode design and construction and Tim Gardner for opening up his lab and facilities to us. We would like to thank Christos Michas for his assistance with PDS use at the Bio-Interface and Technology core facility and Neil Ritter, Jon Spyreas, and David Landesman for their help in designing early versions of the 16-channel jig. We would like to thank Tim Cavanaugh for his assistance with SEM imaging at the Center for Harvard Nanoscale Systems at Harvard.
#10 scalpel blade | Fisher Scientific | 14-840-15 | Building tool |
16-channel CFEA Jig | Realize Inc. | CFMA component | |
16-channel Omnetics connector | Omnetics | A79014-001 | CFMA component |
25 G needle | Fisher Scientific | 14-840-84 | Building tool – sharp-tipped |
30 G needle | Fisher Scientific | 14-841-03 | Building tool |
31 G stainless steel 304 hypodermic round tubing | Small Parts Inc | B000FMYN38 | For guide tube |
32-channel CFEA jig | Realize Inc. | CFMA component | |
32-channel Omnetics connector | Omnetics | A79022-001 | CFMA component |
6 in cotton tip applicators | Fisher Scientific | 22-363-156 | Building tool |
Acetone | Fisher Scientific | A16P4 | Building tool |
AutoCad 3D printing software | Autodesk | Computer-aided design tool/ 3D modeling software | |
Autodesk Fusion 360 | Autodesk | Computer-aided design tool/ 3D modeling software | |
BD disposable syringes | Fisher Scientific | 14-823-30 | 1 mL |
Carbon fibers | Good Fellow USA | C 005725 | 7 μm epoxy sized |
Cassettes and cassette holder | For coating fibers | ||
Clear tape | Scotch | For coating raw fibers | |
Deionized water | Electroplating component | ||
Double-sided tape | Scotch | For coating raw fibers | |
Flowable Dental Composite | Pentron | Flow-It ALC | CFMA component/ UV cured dental cement |
Gold plating solution | Sifco ASC | 5355 | 10.0-20.0% glycerol, 1.0-5.0% ethylenediamine, 1.0-5.0% acetic acid (ethylenedinitrilo)tetra-, dipotassium salt, 5.0-10.0% butanoic acid, mercapto-monogold(1+) sodium salt, 1.0–5.0% potassium metabisulfite, 55.0-82.0% water |
Jewelry clamp | Amazon | B00GRABH9K | Building tool |
JRClust | Ferret spike sorting software | ||
Lighter | BIC | LCP62DC | Building tool |
Micromanipulator | Scientifica | PS-7000C | For guide tube |
Microscissors | Fisher Scientific | 08-953-1B | Building tool |
MountainSort | Mouse spike sorting software | ||
NanoZ 16-channel adapter | Multi-channel systems | ADPT-nanoZ-NN-16 | Electroplating component |
NanoZ 32-channel adapter | White Matter | NZA-OMN-32 rev A | Electroplating component |
NanoZ multi-electrode impedance tester | White Matter | Electroplating component | |
Parafilm | Fisher Stockroom | 13-374-10 | Semi-transparent, flexible film with adhesive properties |
Parylene 'C' Dimer | Specialty Coating Systems | 980130-C-01LBE | For coating raw fibers |
PEG 8000 | Fisher Scientific | 25322-68-3 | Electroplating component |
Phosphate-buffered saline | Electroplating component | ||
Polyimide tubing | MicroLumen | BRAUNI001 | For guide tube |
Rotary tool | Dremel | 300124 | For guide tube |
Scalpel handle | Fine Science Tools | 10003-12 | Building tool |
Silver conductive coating | MG Chemicals | 842AR Super Shield | CFMA component |
Stereo microscope with range 6.7:1 | Motic | SMZ-168 | Building tool |
Sticky notes | Post-it | Building tool | |
Tissue wipes | Kimtech Science | 34155 | Building tool |
Tungsten wire | A-M Systems | 797550 | CFMA component |
UV curing wand | Woodpecker | Building tool | |
Vacuum deposition chamber | Specialty Coating Systems | Labcoter 2 (PDS 2010) |