Hybrid fabrication of multimodal intracranial implants for electrophysiology and local drug delivery

New fabrication approaches for mechanically flexible implants hold the key to advancing the applications of neuroengineering in fundamental neuroscience and clinic. By combining the high precision of thin film microfabrication with the versatility of additive manufacturing, we demonstrate a straight-forward approach for the prototyping of intracranial implants with electrode arrays and microfluidic channels. We show that the implant can modulate neuronal activity in the hippocampus through localized drug delivery, while simultaneously recording brain activity by its electrodes. Moreover, good implant stability and minimal tissue response are seen one-week post-implantation. Our work shows the potential of hybrid fabrication combining different manufacturing techniques in neurotechnology and paves the way for a new approach to the development of multimodal implants.


Methods and Materials
Chemicals, Materials and Components Solvents were purchased from Fisher Scientific and used without further purification. Parylene-C dimer was purchased from Specialty Coating Systems, Inc. Photo resin Young Optics Inc BV-007A and Proc3dure GR 16 Xray for digital light processing 3D printing were purchased from IMakr and Formlabs Elastic ThermoFisher. High resolution film photomasks were purchased from JD Photodata. High purity oxygen (grade N6.0), sulfur hexafluoride (N3.0), tetrafluoromethane was purchased from BOC.

General Methods
Chemical vapor deposition of PAC was conducted with a Speciality Coating Systems PDS 2010 labcoter 2.
Digital light processing (DLP) 3D prints were conducted with an Asiga MAX X UV385 using a for Formlabs Elastic FLELCL01 and Young Optics Inc BV-007A. Filament-based 3D prints were manufactured with an Ultimaker 3. Spin coating was conducted with a Spin coater POLOS 200. Photolithography was conducted with the Karl Suss MA6/BA6 Mask Aligner and e-beam physical vapor deposition (PVD) with an K J Lesker PVD 75 E-beam Evaporator. Reactive ion etching was conducted with a Plasma Pro 80 RIE. For positioning and placing procedures a Fineplacer® Pico ma by FineTech GmbH was used. Optic microscopy was conducted with a Nikon SMZ1270 equipped with a Nikon DFK NME 24UJ003, while for scanning electron microscopy (SEM) a Hitachi TM4000Plus tabletop microscope was used. For washing procedure incl. ultra-sonication an Bandelin sonorex digiplus DL255H has been used. Dry and tempering procedures were conducted in a Fistreem Vaccum oven equipped with a KNF N840 labport or a Fisherbrand 65K Oven 230V. was used in a three-electrode configuration, equipped with the Ag/AgCl electrode ZA006 AS RE-1B by ALS Co., Ltd as reference electrode and an Metrohm Platinum-wire electrode as counter electrode. A sinusoidal voltage input of amplitude 10 mV at different frequencies ranging from 1 Hz to 100 kHz was used. All computer aided design were conducted with Autodesk Inventor Pro2020. Figure S1 Generic model of a rat brain and skull with implant. 1 Fabrication Insights DLP Printing of the electro-fluidic interface The parts of the interface were printed with Young Optics Inc BV-007A, using the parameter shown below.

Implant design
The prints were sonicated two times in IPA. Afterwards it was allowed to dry under vacuum for 30 min and post-cured (15 min 365 nm) Figure   Photolithography and e-beam PVD AZ nLof 2035 was spin-coated on a PAC coated 2in wafer (500 rpm, 1000 rpm/s for 5s; 3000 rpm, 8000 rpm/s for 30s ), followed by a soft baking step for 60s at 110°C on a hot plate. UV-Exposure were conducted with a mask aligner in one step of 160 mJ/cm 2 . Post exposure baking were conducted at 108°C for 80 s, followed by developing with AZ726 MIF for 12-16 s, followed by rinsing with DI water and air blow drying.  Heterogeneous integration of the implant for acute in vivo experiments Figure S6 Heterogeneous integration of the implant: a 3D printed socket; b and c assembly of M0.6 machine screws into socket; d-e assembly of the hybrid fabricated probe; f and g assembly of the double-sided adhesive tape; h 3D printed electro-fluidic interface; i and j insertion of the flex cable; k and l insertion of the pogo pins; m and n insertion of 1/32'' polyethylene tubing and encapsulation with low-viscosity UV curable resin(BV-007); o merging of interface with probe; p side view and q perspective view of the assembly with 1 pound sterling coin for scale.

Heterogeneous integration of the implant for long term in vivo experiments
For long term the socket has been exchanged. Here it is made from a tough, yet biocompatible resin (free form temp) to it withstand freely moving rates. The 3D printed socket allows the easy implantation, wire and tube management as well as a mounting of a standardized headcap with lid. The M0.6 fasteners were here replaced by a twisted wire, achieving a compressing force.  each; plot of the planned vs the measured total thickness using three z-resolution (20 µm, 30 µm and 50 µm) 5 ; plot of the planned vs the measured feature width using 50 µm z-resolution. 5 The individual errors are all smaller 3.5um and have been omitted in this figure to improve readability.
The measured feature width showing consistently a systematic error from the planed size by a mean (± standard error of mean) of 36 ± 2 µm for the channels and -36 ± 5 µm for the walls. Among different total thickness (300 µm, 200 µm, 150 µm, 100 µm, 50 µm using 50 µm z-resolution) the feature width for the 100 µm planed channels and walls deviate by 14 µm and 9 µm. Among four different wafers the feature width for the 100 µm planned channels and walls deviate by 6 µm and 10 µm Using three different z-resolutions (20 µm, 30 µm and 50 µm) the feature width for the 100 µm planned channels and walls deviate by 16 µm and 6 µm. (standard deviation). As shown in Fig S7 does

Accelerated Degradation Test of PAC-RESIN-PAC material composite 6
The degradation test was conducted using a comb-like specimen, where every comb tooth is 500 µm thick, similar to the width of the implant. The specimen was prepared as shown in fabrication schema in the main manuscript. The degradation was performed with PBS and 3 % aqueous hydrogen peroxide solution.
The test specimens were dried in vacuum at 80 °C to constant mass and weighted with an analytical balance (OHAUS Pioneer Plus analytical PA224C), giving minimum wight of 20 µm, a readability and repeatability of both 0.1 mg. The samples were immersed in 1.5 mL PBS or 3 % aqueous hydrogen peroxide solution, respectively, and shook for 24 h at 70 ± 2 °C. Afterwards the solution was decanted and washed twice with 1.5 DIW. The test specimens were dried in vacuum at 80 °C to constant mass. The test results are shown in Table S1. Qualitatively no changes have been observed.  Mean ± standard error of mean -3.6 ± 1.1

Mechanical properties
Flexural properties The flexural properties of the sandwich material were measured using a custom-made apparatus consisting of a 500 mN load cell (Aurora Scientific 402B-HD) mounted ona motorised stage which moves the load cell vertically. A horizontal stainless-steel rod of radius 0.915 mm was fixed to the load cell to act as a testing probe. Rectangular samples (9 x 25 mm 2 ) of the material with three different mid layer thicknesses of 120 µm, 164 µm, and 240 µm were fabricated, and placed horizontally on two parallelly oriented support rods with a radius of 0.915 mm and a spacing L of 15 mm. The testing probe was pushed vertically onto the sampled with a constant speed of 10 mmˑmin -1 . Figure S10 shows a photograph of the setup. Parylene-C (4.5 GPa). 10 With increasing thickness of the mid layer theproperties of the more stretchable 3D printer resin dominate over the less stretchable Parylene-C and hence the overall Young's modulus of the material decreases.

Pressure tests
In order to measure the maximum internal pressure, the microfluidic channels of the implant can sustain, blunt gauge 34 syringe needles were inserted into the inlets of four assembled implants and the connections were sealed with cyanoacrylate adhesive. The needles were then connected via an in-line pressure sensor (PendoTECH, PRESS-S-000) to a syringe pump (KD scientific, Legato 110). Deionised water was pumped into the implant in a step-wise fashion (repetitions of 15 s continuous flow of 100 µLˑmin -1 followed by 15 s without flow) and the pressure was recorded. Figure S11 shows theacquired data. The maximum measured pressure in each device was interpreted as an approximation of the implant burst pressure. A mean burst pressure of 1.1 ± 0.1 bar was determined. Similar pressures were observed for aged samples. The same experiments were conducted with a the pro3dure GR-16, giving far lower burst pressures below 0.5 bar.
Figure S11 Burst Pressure Test: left Pressure versus time data of four implants. The non-zero pressures as zero time arise from pressure created when assembling the testing system before data acquisition started: right repeating of the standard measurement and comparison with aged samples as well as different resin (pro3dure GR 16).

Insertion force measurements
For brain implants, reduction of damage to surrounding brain tissue caused by device insertion is paramount. Therefore, the insertion force which is necessary to insert the implants into brain tissue has been measured. To investigate whether the tip angle of the implant has an impact on the insertion force gradient, the measurement was repeated for devices with tip angles of 10°, 20°, 30°, and 50°. As insertion body, 0.6 % m/v agarose (Merk, Agarose A9539) gels were chosen as it has been reported thatgels of this density resemble the physical properties of brain tissue. 11 The implants were fixed onto a 500 mN load cell 10 https://formlabs-media.formlabs.com/datasheets/Elastic_Resin_Technical. pdf, 29.07.2021;1. W. Sim, B. Kim, B. Choi and J. O. Park, Microsystem Technologies, 2005, 11, 11-15. (Aurora Scientific 402B-HD) which was mounted on a motorised stage. The implants were inserted into the gels with a constant speed of 10 mmˑmin -1 while the insertion force was recorded. Figure 6a shows the corresponding data and figure 6b visualises the setup. It should be noted that the implants were not completely flat but slightly bent before inserted into the gel which might have had an impact on the outcome of the measurements. The samples with a tip angle of 10° and 20° were less stiff than samples with a blunter tip angle which resulted in the implants bending away fromthe insertion direction by almost 90° inside the gel.
(a) (b) Figure S12: a) Insertion force versus implant position of implants with a tip angle of 10°,20°, 30°, and 50°. b) The implants (centre) were mounted onto a load cell (thin metal needle on the right) and inserted into plastic cuvettes filled with an agarose gel.
Electrochemical Impedance Spectroscopy EIS

Animal surgery
All experimental procedures were performed in accordance with the UK Animals (Scientific Procedures) Act 1986. The work was approved by the University of Cambridge Animal Welfare and Ethical Review Body, and was conducted under the authority of a Project Licence (PFF2068BC). ~250 g Sprague Dawley rats (Charles River UK) were anaesthetised using isoflurane (2.5% in oxygen), and mounted on a stereotaxic frame. Their body temperature was monitored and maintained using a thermal blanket. A 1.5 mm 2 window was drilled into the skull, the dura carefully dissected, and the brain exposed. A hybrid fabricated probe was mounted onto the frame and gently lowered into the hippocampus CA1 (final coordinates for the tip of the probe: -4 mm anterio-posterior, 3 mm lateral, 3 mm depth).
For chronic implantations (tissue processing and immunohistochemistry), the space between the exposed brain and the body of the hybrid fabricated probe was covered in medical silicone (Kwik-cast silicone sealant, WPI). Two stainless steel screws (M0.8 x 2mm, US Micro Screw) were inserted into the skull, and the whole exposed skull and implant were sealed in place with surgical cement ( Figure S14).
For electrophysiology recordings (electrophysiology recordings and drug delivery), a stainless steel screw was drilled into the CSF above the cerebellum to act as a recording ground. Isoflurane anaesthetic was lowered to 1.25 % during and prior to recordings. Figure S14. Image of implanted hybrid probe after the application of silicone sealant (green, left), followed by the application of surgical cement (right). Tab used for implantation (top of both images) is removed from the implant before the animal is allowed to recover.

Electrophysiology recordings and drug delivery
The hybrid fabricated probe was connected to electrophysiology recording hardware (RHS Stim/Recording headstage and controller, Intan Technologies), with the cerebellar screw acting as a ground. Voltage signals were recorded and amplified (X192), bandpass filtered between 1 Hz and 7.5 kHz, and digitised at a 30 kHz sampling rate. To stimulate interictal-like activity, the RHS Stim/Record ground was connected to one of the electrodes in the hybrid fabricated probe. This formed a stimulating pair with one other electrode in the probe, locally stimulating the hippocampal tissue. The two most distant electrodes (bottom left and top right - Figure 1) were chosen to form this pair. Stimulation was carried out over 300 ms at 500 uA and 250 Hz (1 ms per pulse phase, followed 2 ms of interpulse period).
Local drug delivery was carried out by manually slowly injecting either 20 ul of 4AP (100 mM in phosphase buffered saline) or MPQX (54 uM in saline) over 30 s through one of the microfluidic channels of the hybrid fabricated probe.
For analysis, data was notch-filtered to remove mains noise and band-pass filtered between 1 and 400 Hz.
Spikes were identified and counted when they had an amplitude larger than three standard deviations above the baseline using a custom-written script. Electrophysiology data analysis was carried out using Spike2 software (Cambridge Electronic Design, UK, v9.04b). Statistical analysis and data plotting was carried out in MATLAB (Mathworks, R2016b).

Tissue processing and Immunohistochemistry
Rats were humanely killed at 7 days post-implantation by exposure to a rising concentration of CO 2 . Brains with probes still implanted were collected and transferred into 4% methanol-free paraformaldehyde in phosphate buffer saline with 0.1% sodium azide, and allowed to fix for 48 hours at 4ºC. Probes were then removed from fixed brains ( Figure S15), and the brains were soaked in 30% of sucrose in phosphate buffered saline for cryoprotection for 72 hours. Brains were frozen and sectioned in a cryostat into 10 um sections, which were allowed to defrost and dry prior to immunostaining.
For staining, brain samples were blocked in 5% donkey serum in phosphate buffered saline with 0.1% sodium azide for 1 hour at room temperature. This was followed by incubation in primary Anti-GFAP antibody (ab7260) used at 1/1000 dilution in 5% foetal bovine serum in phosphate buffer saline with 0.1% sodium azide overnight at 4 o C. The next day samples were incubated in secondary antibodies (Donkey anti-Mouse IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 used at 1/2000 in 5% foetal bovine serum in phosphate buffer saline with 0.1% sodium azide) for 3 hours at room temperature in the dark. Tissue was then incubated in Vector TrueVIEW Autofluorescence Quenching Kit (Vector Laboratories) for 3 minutes at room temperature, followed by a wash in 5µg/mL Hoechst for 5 minutes.
Between above incubation steps tissue was washed three times with 5% foetal bovine serum in phosphate buffered saline with 0.1%, 10 minutes per wash. Tissue was finally mounted and imaged in a Zeiss Axioscan Slide Scanner (Zeiss). Illumination intensity and exposure parameters were chosen such that GFAP+ astrocytes were visible in the naïve control brain samples.
For analysis, images of brain at 20X magnification were processed using a combination of custom-made Matlab (Mathworks, R2016b) and Fiji (v1.48, National Institutes of Health, USA) scripts. These produced a stain intensity profile at a range of depths from the edge of the implant, as indicated by the user in each image. This intensity was normalised against the stain intensity in a randomly-chosen 1 mm 2 area in the contralateral (non-implanted) brain hemisphere. Statistical analysis and data plotting was carried out in MATLAB (Mathworks, R2016b). Figure S15. Pictures of test specimen and hybrid fabricated probe before and after 7 weeks implantation: top left: high magnification of test specimen after removal from fixed brain profile from side of skull; top left: test specimen still adhered to the skull via cement; middle left: SEM picture of the microfluidic outlet before and middle right: after 7 days implantation; bottom left: SEM picture of the electrode before and bottom right: after 7 days implantation. Figure S16. Hippocampal recorded trace (top) and corresponding time/frequency plot (bottom) following delivery of 4AP. Generated using a 32768 FFT block size and a Hanning window.