Structurally Aligned Multifunctional Neural Probe (SAMP) Using Forest‐Drawn CNT Sheet onto Thermally Drawn Polymer Fiber for Long‐Term In Vivo Operation

Neural probe engineering is a dynamic field, driving innovation in neuroscience and addressing scientific and medical demands. Recent advancements involve integrating nanomaterials to improve performance, aiming for sustained in vivo functionality. However, challenges persist due to size, stiffness, complexity, and manufacturing intricacies. To address these issues, a neural interface utilizing freestanding CNT‐sheets drawn from CNT‐forests integrated onto thermally drawn functional polymer fibers is proposed. This approach yields a device with structural alignment, resulting in exceptional electrical, mechanical, and electrochemical properties while retaining biocompatibility for prolonged periods of implantation. This Structurally Aligned Multifunctional neural Probe (SAMP) employing forest‐drawn CNT sheets demonstrates in vivo capabilities in neural recording, neurotransmitter detection, and brain/spinal cord circuit manipulation via optogenetics, maintaining functionality for over a year post‐implantation. The straightforward fabrication method's versatility, coupled with the device's functional reliability, underscores the significance of this technique in the next‐generation carbon‐based implants. Moreover, the device's longevity and multifunctionality position it as a promising platform for long‐term neuroscience research.


Introduction
The rapid development of neural probes has emerged as a cornerstone in modern advancements in neuroscience.[16][17] Particularly, there is increasing efforts toward applying novel polymers and nanomaterials toward neural probe engineering in order to replace the conventionally used rigid substrates and electrodes. [14,18][33] In this study, we present a novel Structurally Aligned Multifunctional neural Probe (SAMP) using forest-drawn CNT sheets integrated onto thermally drawn polymer fibers, which features not only multifunctional integration and prolonged usability but also straightforward and reliable fabrication processes.The freestanding nature of the utilized forest-drawn CNT sheet simplifies the integration process into a direct wrapping step, while the continuity of the material ensures superior electrical and material properties.The resulting device enables year-long electrophysiological recordings, optogenetic neuromodulation, and chemical sensing functions, making it possible to study neural ensembles over durations comparable to a subject's lifespan, thereby paving the way for more accessible long-term neuroscientific research.

The Concept, Design, and Fabrication of SAMP
The proposed neural probe is designed to address key challenges in current neural interfaces: simple integration of multiple functions and longevity in in vivo operation.Advancements in these directions were realized by a straightforward fabrication approach, which combines thermally drawn polymer fiber with forest-drawn CNT sheets (Figures 1a and S1, Supporting Information).The probes incorporate multiple functionalities based on bidirectional communication into a single implant.This is achieved by utilizing optical waveguides for optogenetic interrogation, while wrapped CNT sheet electrodes are employed for neural signal recording and neurotransmitter sensing (Figure 1b).
[36] Subsequently, free-standing pure CNT sheets were obtained by teasing CNT bundles from the forest sidewall which is then continuously drawn out into a sheet (Figure 1c).[38] The SEM image of the CNT forest during air drawing process clearly shows the unidirectional alignment of CNT sheet, and the inter-fibrils between each set of parallel lines is also visible (Figure S2, Supporting Information).The electrical anisotropy of the CNT sheet is significantly influenced by its anisotropic structure, [34] which consists of main nanotubes aligned along the longitudinal direction and laterally branched inter-fibrils. [39]The width and length of the CNT sheet could be controlled with relative ease by adjusting the initial teasing conditions (in our case, a few nm thickness, 2 cm width, and 10 cm length, Figure S3a,b, Supporting Information). [37]To regulate the application thickness, the sheet could be stacked in multiple layers, with the thickness of the overall CNT layer increasing from 8.0 ± 0.5 to 24.5 ± 1.1 μm as the number of CNT sheets stacked increased from one to five layers (Figure S3c, Supporting Information).Concurrently, the optical absorption also visibly increased with the thickness (Figure S3d, Supporting Information).
For the polymer optical waveguide/backbone of the probe, flexible polymer fibers were produced using the thermal drawing process (TDP) (Figure S4, Supporting Information).Macroscale preforms (≈12 mm diameter) were initially constructed by wrapping PMMA sheets (Polymethyl methacrylate) around PC (Polycarbonate) rods in desired ratios (5:1 = Core radius: Cladding thickness).PC and PMMA were selected for their excellent optical transmission and their thermoplastic properties (with glass transition temperature T g,PC = 147 °C, T g,PMMA = 125 °C, respectively).The preform was consolidated and positioned in a custom-built drawing tower, where it was heated to above the glass transition temperature.By adjusting drawing parameters such as preform feed speed and fiber capstan speed, microscale waveguide fibers with a diameter of ≈160 μm were produced with high uniformity on a mass-production scale (≈10 2 m).
These drawn fibers were then wrapped with the previously drawn CNT-sheets.The step-by-step procedure for CNT sheet wrapping is demonstrated in Figure 1c and Video S1 (Supporting Information).To initiate the wrapping, thermally drawn fibers were secured at one end to minimize torsional stress on the polymer structure during the CNT wrapping process.CNT sheets (2 cm width) were laid onto the fiber waveguide, resulting in immediate adhesion based on Van der Waals forces.Here, the high specific surface area of the CNT sheets creates the phenomenon of substrate non-specific adhesion (Figure S5, Supporting Information). [40]Subsequently, the sheets were gradually wrapped around the fiber, as depicted in the optical and SEM images in Figure 1c.Unless otherwise noted, the CNTs were aligned parallel to the longitudinal direction of the fiber.The CNT layers were effectively densified by employing the ethanol evaporating densification method (Figure S6a,b, Supporting Information). [41,42]By applying ethanol to the fiber, the thickness of the CNT layer reduced from about 20 to 10 μm, which in turn decreased the fiber's overall diameter from 203.0 to 179.4 μm (Figure S7, Supporting Information).In addition, the height profiles from AFM analysis reveal that the average height of surface after densification (85.5 nm) markedly flatter and more uniform than that of the surface before treatment (33.0 nm) (Figure S6c,d, Supporting Information).This reduction can be attributed to the surface tension effect during ethanol evaporation.Furthermore, the decrease in the CNT layer's thickness enhanced the contact area between the nano-bundles and the fiber's surface, resulting in tighter adhesion between the concentric layers. [34]Finally, the densified CNT sheet-wrapped fiber probe was dip-coated with PDMS insulation (187.4 μm diameter).After insulation, only the end of the fiber was cut to expose the concentric cross section.The cross-sectional images of each step of the process show concentric structures without any physical obstruction from each other (Figure 1c).The actual exposed CNT area, ≈5.17 × 10 −5 cm 2 , accounts for about 18.7% of the cross-sectional area of SAMP.At the probe tip, this exposed CNT operates as the transducing electroactive surface, while the re-maining CNTs enveloping the probe body serve as a signal transmission line (Figure S8, Supporting Information).
Based on this manufacturing process, flexibility which is crucial to minimal micromotion-effect was ensured by the low modulus of the polymer fiber and thin CNT sheets layer.In addition to such, the self-supporting CNT sheets possesses excellent biocompatibility and cell-interfacing capabilities, which we interpret will facilitate chronic usage of the probes.Typically, an ordinary CNT substrate penetrates the cell and either directly disrupts its metabolism or usually contains residual contents from the synthesis that result in cellular damage. [43,44]Based on a cell viability test using 3T3 cells, we confirmed the high biocompatibility of the CNT sheet by comparing cell proliferation between Matrigel-coated glass substrates and CNT-sheet substrates (Figure 1d).Furthermore, thermogravimetric analysis (TGA) exhibited that our freestanding CNT sheets does not contain a notable amount of residual catalytic contaminants such as iron, nickel, and cobalt (Figure S9, Supporting Information).Excluding the decrease of ≈2.80%, indicative of amorphous carbon combustion at around 334.6 °C, the largest weight loss was observed near 684.7 °C which representing the combustion of CNTs. [45,46]The in vitro test also demonstrated enhanced interfacing and growth of branching that indicates the cell morphology were inclined to orient toward the CNT sheet's alignment (Figure 1d).We went on to verify this phenomenon in neural tissue by observing the neuronal branches and nanostructural CNT sheet electrode interface in an in vivo environment.A year after implantation, imaged brain slices showed no significant signs of tissue damage and rather displayed elevated neurofilament growth at the tip of the fiber (Figure S10, Supporting Information).Through these analyses, we found that the flexible and biocompatible nature of the CNT fiber probe had distinct benefits in allowing the probe to function over prolonged periods (Figure 1e).

Characteristic Properties of SAMP
The CNT sheet utilized in the device possesses distinct features that makes the device apt for neural implant applications (Figure 2a).First, the innately aligned orientation of the CNT sheet facilitates superior electrical transmission, consequently leading to enhanced neuroelectric interfacing (Figure 2b-j). [34,38]Moreover, the concentric structure of electrode placement allows the CNT sheet layer to simultaneously serve as a photo-jacket, effectively creating an anti-scattering cover that leads to targetspecific optical transmission (Figure 2h-j). [34,47]Lastly, the thin and highly flexible CNT sheet allows the devices to maintain stable performance even under severe mechanical deformation (Figure 2k-n). [48,49]o verify the effect of the orientation angle on electrical performances, we fabricated the CNT sheet-wrapped fibers with different orientation angles of 0°, 45°, and 90°(Figure S11, Supporting Information).The orientation angle () represents the angle between the longitudinal vector of the polymer fiber substrate and the alignment direction of the CNT sheet (Figure 2a).The density of the CNT sheet was controlled to maintain equal weight and thickness across sample groups.The CNT wrapped-fiber exhibited no significant differences in diameter and total weight (with diameters of 198.7 ± 6.3, 200.3 ± 5.2, and 205.1 ± 11.0 μm, and weights of 384.8 ± 4.7, 389.3 ± 5.2, and 399.1 ± 7.6 μg corresponding to the orientation angles of 0°, 45°, and 90°, respectively) (Figure S12a,b, Supporting Information), and the fiber density also remained largely unaffected with negligible changes (Figure S12c, Supporting Information).As illustrated in Figure 2b, the CNT sheet's alignment is well preserved after the wrapping process.Particularly, the enlarged SEM images reveal uniform CNT alignment and distinct orientation angles.The orientation angle of each CNT-wrapped fiber ( = i) 0°, ii) 45°, and iii) 90°, respectively) was visualized using surface SEM images and quantified through image processing.The analyzed results for contrast-enhanced SEM images and 2D FFT spectra of CNT wrapped fibers with varying orientation angles are compared in Figure S13 (Supporting Information).[52] The orientation angle of each wrapped fibers is dominant at −2°, 48°, and 94°, respectively.
To investigate the orientation-enhanced conductivity of the wrapped fibers, we analyzed the impedance and resistance under each condition.The impedance at 1 kHz, an evaluation criterion for assessing the single-unit recording capabilities of neural interfaces, was measured as 224.2 ± 62.0, 412.4 ± 118.0, and 484.4 ± 198.5 kΩ, respectively (Figure 2d and Figure S14, Supporting Information).The impedance of longitudinally wrapped fibers ( = 0°) were significantly lower compared with the ones with angled sheets, as well as conventional carbon compositebased electrode fibers (0.62/1.31MΩ with gCPE fiber of after/before soaking, about 8 MΩ with carbon nanofiber composite fiber). [19,20]The resistance of the longitudinally wrapped fiber ( = 0°), 63.0 ± 9.7 Ω cm −1 , also proved superior compared to other wrapped fibers, which was found to be 225.0 ± 29.9 Ω cm −1 ( = 45°) and 495.0 ± 56.2 Ω cm −1 ( = 90°), respectively (Figure 2e).With the electrical anisotropy of the CNT sheet, the longitudinally wrapped fiber ( = 0°) transported electrons directly and efficiently along the nanotubes (Figure S15a, Supporting Information).Meanwhile, the mismatch of the CNT sheet alignment orientation and the fiber direction ( = 45°or 90°) induces extended and more circuitous conductive pathways, leading to an increase in electrical resistance (Figure S15b, Supporting Information).These well-aligned pure CNT sheets were found to be suitable for neural recordings, even at the singleneuron level, upon completion with a PDMS insulation layer (Figure S16, Supporting Information).To demonstrate the uniformity of the wrapping process and the homogeneity of the CNT alignment, we examined multiple regions of the wrapped fiber using surface SEM images (Figure 2f).The unidirectional CNT bundles propagated parallel to the fiber length direction irrespective of the regions examined.The distribution function, fitted with a normal distribution at each region, was compared (Figure 2g).The orientation angles closely matched to 0°(parallel to the fiber length direction), with the error usually falling below 4°(the orientation angles of region 1, 2, 3, and 4 were 4°, −3°, 4°, and 1°, respectively).
In order to evaluate the device for optogenetic interfacing performances, we assessed the optical transmission and scattering properties.When tested at 20 cm length with a blue light laser (473 nm, Channel-rhodopsin 2 activating wavelength), the finalized fibers demonstrated excellent transmission, with tip-specific output (Figure 2h). [53]The transparency and appropriate refractive index of the core and cladding (PC; n = 1.58,PMMA n = 1.49) formed a waveguide, facilitating optical transmission.The concentric structure of CNT layer prevents lateral scattering, potentially averting undesired activation in the neighboring regions of implantation.Optical transmission is quantified by optical loss coefficients (dB cm −1 ) through measuring the output intensities of 1-10 cm long fibers with a fixed input of 473 nm light (Figure 2i).The CNT sheet fiber efficiently transmitted the stimulation light with 0.769 dB cm −1 .The absorbance of the CNT sheet was evaluated to elucidate the anti-scattering effect (Figure 2j, Figures S3e and S17, Supporting Information).A single layer of CNT sheet exhibited the absorbance of 0.4985, equating to a 473 nm light transmittance of 30-35%.Based on this data, the wrapped fiber, particularly following the densification process, (exceeding 10 layers CNT sheet) is expected to fully prevent lateral scattering.
The mechanical properties of the CNT sheet-wrapped fiber demonstrate that the resulting fibers are both flexible and durable.These fibers can withstand most mechanical deformations, maintaining structural integrity even under extreme conditions such as rounding, knotting, winding and bending (Figure 2k).To further validate the functional integrity is upheld during deformation, we evaluated the electrical and optical performances under bending conditions.The electrical impedance at 1 kHz remained stable under a 90°bending condition (Figure 2l).Regarding optical transmission properties, the flexible wadveguide withstood up to 2.5 mm bending radius without any significant relative loss at both 90°and 180°b ends (Figure 2m).The relative optical transmission under deformation was measured and calculated based on the value of 5 cm fiber with 12.5 mm radius of curvature bending.We also assessed the bending stiffness of these fibers, finding that they exhibit two orders of magnitude lower stiffness compared to stainless steel wire (Figure 2n).The presence of the CNT sheets exhibits negligible effect to the stiffness of fibers, owing to its extremely thin form factor.Despite such flexibility observed during mechanical testing, the CNT sheet-wrapped fiber proved to be firm enough for straight insertion into a phantom brain, eliminating the need for additional implantation guides or mechanically stiffening coats (Figure S18, Supporting Information). [54]Overall, the findings suggest that the proposed fiber represents an optimal balance point between soft mechanical properties and functional capabilities, implying that the choice of materials does not critically compromise performance while achieving a combination of soft yet robust characteristics.

In Vivo Assessment of SAMP as a Neural Probe
We integrated the fabricated fiber with the necessary backend components to form a functional device (Figure 3a).The fiber-type probe holds many benefits for actual applications as these head-mounted devices are compact, without requiring additional housing, and are lightweight, with a total weight of 0.45 g (which includes the 0.1 g of gold pin connector and the 0.3 g of optic ferrule).Herein, we utilized the probe in in vivo mouse models to validate its diverse functionalities, including electrophysiology, optogenetics stimulation, and electrochemical sensing.
[57] The SAMP was implanted into the hippocampus CA1 region (Cornu Ammonis-1, Coordinate AP: 1.7 mm, ML: 1.5 mm, DV: −1.5 mm) of ChR2expressing mice (transgenic Thy1-ChR2-YFP mice).We acquired stable recording of OEP and spontaneous activity simultaneously in both LFP (Local field potential, 3-300 Hz) and extracellular potential (300-5000 Hz) ranges with high Signal-to-Noise Ratio (SNR) (Figure 3c).Contrary to conventional metal electrodes, carbon-based electrodes are known to generate negligible light-induced artifacts. [27]To distinguish the recorded OEP signal from light induced artifacts, we recorded frequency-dependent potentials, considering the kinetics of ChR2 (Figure 3d,e). [8,58,59]The extracellular potentials, raster plot, and PSTH (Post Stimulus Time Histogram) across 40 trials and their successful spike rate indicate that the neuronal signals did not follow high frequency stimulation (100 Hz), due to the recovery period of ChR2 (Figure S19, Supporting Information).To further corroborate this, we also performed OEP recording with a control group of wild-type mice for comparison and analyzed latency in relation to pulse width (Figures S20 and S21, Supporting Information).The peak amplitude of the OEPs displayed a positive correlation with the light intensity at the tip, a phenomenon attributed to the characteristics of optogenetic electrophysiology (Figure 3f).In addition to the OEP, spontaneous activity of the neural system without external stimulation was also visualized through our device (Figure 3g).The neural streams were recorded with high stability, maintaining consistently low noise levels (below 25 μV), and relatively high amplitude spikes were detected.The extracted spikes illustrated the regular firing of two adjacent single neurons in the extracellular space, characterized by different polarity, waveform, and frequency (30.8 Hz and 32.5 ms of ISI, 4.3 Hz and 232.4 ms of ISI) (Figure 3g,h).Each neuron was recorded with high SNR, with values of 19.1 (−86.7 μV peak amplitude) and 17.0 (68.3 μV peak amplitude).Principal Component Analysis (PCA) clearly segregated these spikes into two distinct waveforms (Figure 3i).
The efficacy of the neuronal ensemble manipulation has been verified through experiments with freely moving animals, utilizing a head-mounted SAMP platform.We conducted the open field test (OFT) with Thy1-ChR2 transgenic mice that were device implanted in the supplementary motor cortex (M2 cortex layer 4/5, Coordinate AP: 2.5 mm, ML: 1.0 mm, DV: −1.0 mm), an area associated with general locomotor activities (Figure 3j). [60]he behavioral phenotype (i.e., rotation, velocity, distance moved, and trajectory) during OFF/ON photo-stimulation sessions (featuring a 10 ms pulse width at 20 Hz, 10 mW power) was compared to the behavioral phenotype of WT mice.The recordings demonstrate that the ChR2-expressing mice exhibited a significantly higher number of rotations during the ON session epoch compared to the WT mice (ChR2 mice: 3.5 ± 0.4 at OFF, 6.6 ± 1.3 at ON and WT mice: 2.8 ± 1.3 at OFF, 2.7 ± 0.5 at ON) (Figure 3k).Additionally, the average speed of transgenic mice was also visibly higher than that of WT mice during stimulation (ChR2 mice: 2.8 ± 1.3 cm s −1 at OFF, 10.5 ± 6.4 cm s −1 at ON and WT mice: 2.0 ± 2.2 cm s −1 at OFF, 2.3 ± 1.5 cm s −1 at ON) (Figure 3l).The trajectory of ChR2 mice at each epoch of OFT clearly demonstrated increased mobility and turning under stimulation (Figure 3m).Furthermore, we conducted experiments to validate the multifunctionality of our probes, specifically targeting the spinal cord circuits of mice (Figure 3n and Figure S22, Supporting Information).The probe was implanted in the lumbar spinal cord at segments L1/L2, and the sensory neuronevoked potentials were recorded in response to mechanical stimulation of the subject's paws (Figure 3o).Moreover, optical stimulation delivered to the spinal cords via our probes successfully evoked hindlimb responses, indicating successful manipulation of the spinal cord (Figure 3p and Figure S23, Supporting Information).

Electrochemical Sensing Capabilities of SAMP
To validate the neurotransmitter monitoring performance of our probes, we conducted dopamine detection experiments using Fast Scan Cyclic Voltammetry (FSCV) method, with a voltage sweep from −0.4 to 1.3 V with a scan rate of 400 V s −1 in 10 Hz (Figure 4a).Dopamine detection exhibited high sensitivity, with a measurement of 128.6 ± 2.2 nA μm −1 , and the oxidation peak current increased linearly with dopamine concentration below 1 μm-a range relevant to physiological environments (Figure 4b,c).We observed that the CNT sheets derived from forests demonstrated a shorter peak to peak distance of oxidation and reduction potentials compared to conventional carbon fibers (CF), owing to their high electron transfer rate.This makes them more ideal for electrochemical analysis (Figure S24, Supporting Information).Additionally, we tested other bio-active chemicals including uric acid (UA) and ascorbic acid (AA) to demonstrate the selectivity performance of SAMP (Figure S25, Supporting Information).
Subsequently, we demonstrated the in vivo real-time dopamine monitoring performance of the probes by using mice models targeting the dopaminergic pathway in ventral tegmental area (VTA) of mesocorticolimbic system. [61,62]For the selective activation of dopaminergic neurons, we prepared AAV2/9 mediated ChR2-virus injected DAT IRES Cre mice (Figure 4d).Real-time dopamine release was detected by the CNT sheet electrode under condition of 100 Hz and 100 pulses of optogenetic stimulation (Figure 4e).Additionally, we conducted an experiment to examine the dynamics of dopamine release with varying stimulation frequencies using the same probes.We confirmed changes in the frequency-dependent dopamine release in the VTA during optogenetic stimulation to the same region (Figure 4f-h).Voltammogram analysis revealed a dopamine-specific redox pattern during optogenetic stimulation (Figure 4g), and the maximum concentration of dopamine demonstrated a linear relationship with the stimulation frequency (Figure 4h).

Prolonged Operation of SAMP
The prolonged chronic operation of SAMP has been substantiated through a year-long implant tracking experiment.We monitored both optically evoked responses from neuronal population and endogenous action potential from individual neurons in mice implanted with a device targeting the hippocampus CA1 region.Throughout the year, the implanted devices appeared to cause no abnormal physical harm to the mice, and the device exhibited no significant signs of functional degradation or structural detachment (Figure 5a).As an initial validation, we calculated the SNR of optically evoked activities during optogenetics stimulation in Thy1-ChR2 mice from the neuronal response to 120 pulses of low-frequency stimuli (10 Hz), which allowed for sufficient recovery period.The data indicated that both SNR and peak amplitude remained consistent up to a year postimplantation, with values approximately ranging from 8 to 12 for the SNR (Figure 5b).Furthermore, the rate of successful OEP maintained uniformity during sweeps over a year-long period (Figure 5c and Figure S26, Supporting Information).Endogenous single-unit activities were also recorded at one-month interval (Figure 5d and Figure S27, Supporting Information).Over the span of the year-long recording period, the PCA results, overlap of the action potential waveform, and the waveform characteristics (SNR, amplitude, duration, peak-to-trough ratio) remained constant in the analysis (Figure 5e-h).This demonstrated that single neuron activity was adeptly tracked without visible degradation.These results suggests that the proposed device, as intended, exhibits more than sufficient capability to support prolonged operation for chronic neuroscience studies.

Conclusion
We have demonstrated the efficacy of a CNT sheet-wrapped fibertype neural interface, which shows optimal mechanical, electrical, optical properties stemming from its well-integrated longitudinally oriented features.The functionalities of these devices were affirmed through in vivo bidirectional communication with neural systems over a yearlong period, exhibiting no noticeable decrease in function.This was made possible through an innovative yet straightforward fabrication process concerning electrode integration with thermally drawn fibers.This process is substantially simpler than the traditional electrode manufacturing schemes while offering enhanced performance.Utilizing the CNT sheet electrode and the wrapping techniques, we managed to facilitate nano-structural communication, which produced high SNR neural recording with negligible light-induced artifacts.Furthermore, the designs and material properties resulted in enhanced biocompatibility, making it highly suitable for prolonged usage.These benefits potentially address the central issues present in conventional neural interfacing, paving the way for groundbreaking neuroscientific research targeting specific brain regions and complex circuitry over extended durations.

Experimental Section
SAMP Fabrication: A 10 mm of PC rod (Goodfellow) was wrapped with PMMA film (Goodfellow), following the aforementioned ratios.Consolidation was performed in a vacuum oven at 190 °C.The resulting preform was subjected to thermal drawing using custom-built towers at 175 °C.By controlling the feed and capstan speed, the scale was preformed down by ≈80 times scale, achieving ≈150 um optical waveguides.The fibers were then fixed to a motor tip in a custom twisting machine using adhesive carbon tape.A 2 cm wide CNT sheet was mechanically drawn from a CNT forest ≈300 μm in height (A-Tech System Co., Korea) and attached to one end of the fiber waveguide at a specific orientation angle, determining the dominant orientation direction of the CNT sheet layer.Subsequently, the sheets were wound around the fiber with gradually inserted twists.50 μL of ethanol was then applied onto a 10 cm fiber segment to densify the CNT sheets and induce tight adhesion via surface tension-based evaporating densification method.Finally, the fiber was dip-coated in PDMS (Dow, Sylgard 184) for electrical insulation and surface refinement.To ensure a uniform dip coating, the fiber was dipped in PDMS and attached to a DC motor to spin at 500 rpm for 10 s.
Neural Device Connection: The fabricated fiber was connected to a 2.5 mm ceramic ferrule (Thorlabs, CF270-10) using epoxy (Thorlabs, G14250).The opposite side of ferrule was polished to a smooth surface using a ferrule polishing puck (Thorlabs, D50-F) and a polishing sheet (Thorlabs).On the side of the fiber, the copper wire (Goodfellow, AWG-36) was connected to the CNT sheet sidewall with conductive silver paint (SPI Supplies, 04998AB), and other end of the wire was connected to a 4-pin connector (Digikey).The reference wire was wound around a screw and inserted into occipital side of the skull.
Device Property Measurement: Impedance Measurement: The impedance of the CNT sheet probe was measured using a LCR meter (HIOKI, IM3590).Impedance and phase angles were measured in sinusoidal AC wave of 10 mV, ranging from 10 Hz to 10 kHz in saline solution, utilizing Ag/AgCl electrodes as a reference.
Optical Property Measurement: The 473 nm blue light LASER (RWD, LIOG473-80-A1) was used for all optical inputs, and the output light intensity was measured using a power meter (Thorlabs, S121C and PM100D).The fiber device was connected to the patch cord from the laser through a ferrule-to-ferrule interconnect (Thorlabs, ADAF2).The optical loss coefficient was calculated in dB cm −1 units relative to a 1 cm output.Scattering intensity (representing the light emanating from the fiber wall) was analyzed by using imaging techniques to plot the spatial distribution of light intensity.
Alignment Measurement: The alignment of CNT sheet was quantified and analyzed by two-dimensional fast Fourier transforms (2D FFT) of magnified surface SEM images of CNT-wrapped fiber.The SEM images of the surface were first contrast-enhanced, then analyzed using MATLAB functions to generate a zero-frequency centered 2-D discrete Fourier transform plot.Here, the azimuthal distribution of intensity in spectrum relates to the dominant orientation angle and distribution function of CNT sheet.The power spectrum was represented in 8-bit grayscale image using logarithmic transformation.In order to obtain the orientation distribution function, the circular mask was applied, and the total intensity at each Electrochemical Characterization: The electrochemical monitoring of dopamine was carried out using a potentiostat (BioLogic, SP-300), equipped with ultra-low current and analog ramp generator options.Fast scan cyclic voltammetry (FSCV) was employed by applying a triangular waveform with a scan rate of 400 V s −1 within a potential window of −0.4 to 1.3 V.The cycle frequency was set at 10 Hz, holding the potential at −0.4 V between each sweep.Prior to FSCV recording, the fiber device underwent activation conditioning with a triangular waveform at 60 Hz for 10 min, followed by a frequency of 10 Hz for 15 min as presented in previous papers as presented in previous papers. [13,63,64]The CNT sheet electrodes in the fabricated multifunctional fibers were served as the working electrode.Ag/AgCl wires were used as reference electrodes, prepared by exposing 1 mm of Ag wire (A-M Systems, 787000) and chlorinating it in bleach overnight.For the electrochemical calibration for dopamine sensing tests, solutions of various dopamine concentrations were prepared by adding dopamine hydrochloride (Sigma-Aldrich, H8502) to phosphatebuffered saline (PBS) with a pH of 7.4.
In Vivo Animal Experiments: All animal experiments were conducted on the mouse model, with approval of the Institutional Animal Care and Use Committee (IACUC) of KAIST.The wild type C57BL/6NHsd mice (KOATECH, Gyunggi-do, South Korea) and the transgenic mice (Thy1::ChR2 mice) (JAX 007612) aged 8-12 weeks were utilized for in vivo validation.The CNT sheet fiber probes were both acutely and chronically implanted into the brain and spinal cord.The mice were anesthetized using isoflurane (4% induction and 1.5% maintenance).For in vivo electrochemical detection experiment, DAT-IRES-Cre mice (JAX 006660) aged 7-12 weeks were used.Virus injection for dopaminergic neuron selective activation was performed using AAV2/9-hEF1a-DIO-mCherry-WPRE-pA (KIST).After 4 weeks of recovery and virus expression period, in vivo FSCV for dopamine monitoring was performed.Craniectomy for brain and partially laminectomy for spinal cord were conducted.For chronic implantation, Super Bond (Sun Medical, C&B super bond) and dental cement (Lang Dental, Ortho-jet) were used for secure the fiber devices to the skull.After the surgery, the implanted mice were single-housed and were maintained at 22 °C and a 12-h light/dark cycle provided with food and water ad libitum.
In Vivo Electrophysiological Recording and Optical Stimulation: For opto-electrophysiology, laser pulses were applied through PC/PMMA waveguide to the central nervous system.All optical stimulations for optoelectrophysiological tests were performed with a power of 20 mW.Optically evoked potentials were recorded using a LABRAT instrument (Tucker-Davis Technologies).Faraday cages and high impedance back-end connections (Tucker-Davis Technologies, ZIF-Clip and ZCA-MIL16) were utilized to minimize electrical noises.Local field potential and single-unit spikes were obtained by processing the acquired signal through a 3-300 Hz bandpass filter and 300-5000 Hz band-pass filter, respectively.The sampling frequency was set to 24 414 Hz.Synapse Lite (Tucker-Davis Technologies) and MATLAB (MathWorks) software were used for analysis.The SNR of optically evoked potential and spontaneous action potential is calculated by formula as SNR = 20 × log Signal amplitude Noise amplitude In Vivo Fast Scan Cyclic Voltammetry (FSCV) Data: To address the presence of non-analyte signals and noise in raw fast scan cyclic voltammetry (FSCV) voltammogram data obtained in vivo, data processing was carried out using principal component regression (PCR) in MATLAB.Principal components (PCs) were extracted from the electrochemical calibration test data, obtained by injecting DA solutions with into a buffer.These PCs were classified into primary PCs, which captured analytically relevant variance, and secondary PCs, which were identified as other sources of noise using Malinowsk's F-test (at a significance level of 95%).Utilizing the selected primary PCs, PCR was then applied to the raw voltammogram data obtained in vivo, resulting in processed voltammograms that retained analytically relevant variance.
In Vitro Cell Experiments: Carbon Nanotube Sheet Preparation on Glass Slides: For the direct cultivation of 3T3 cells on CNT sheets, the glass slides (Marienfeld Laboratory Glassware) were prepared with CNT substrates.Three layers of CNT sheets were overlaid onto each glass slide and densified by direct application of ethanol.After preparation, the slides were sterilized by immersion in 70% ethanol and stored in a biosafety cabinet until use.
Cultivation of 3T3 Cells on CNT Sheet-Covered Glass Slides: 3T3 cells were cultured in Dulbecco's modified Eagle's medium (Gibco), supplemented with fetal bovine serum (Gibco) and 1% penicillin/streptomycin (Gibco) until they reached 70% confluency.Subsequently, the cells were trypsinized and seeded onto the CNT sheet-covered glass slides at a concentration of 1.0 × 105 cells per slide.Following a 3-h attachment period, the cells were transferred to a 100 mm culture dish filled with growth medium.The cultures were then incubated for 24 h before further analysis.
Assessment of Cell Viability and Morphology: Cell viability and morphology were assessed using a live-dead assay kit (Invitrogen).The staining solution was prepared by combining 5 μL of calcein AM and 20 μL of ethidium homodimer-1 in 10 mL of Dulbecco's phosphate-buffered saline (DPBS).This staining solution was applied directly to each slide glass, followed by a 30-min incubation at room temperature.Subsequently, the slides were imaged.Each image was subjected to analysis using ImageJ software to quantify live/dead cells and evaluate cell morphology and alignment.
Behavioral Test: The investigator conducting the behavioral tests had knowledge of the identity of experimental groups versus control groups.Mice used in the behavioral experiment were given a recovery time of 3 weeks after implantation surgery.Prior to the behavioral experiment, the ferrule to patch cord connection procedure was repeated to familiarize them with the process.The OFT box was routinely cleaned before and after all trials, and for all individual subjects, a habituation time of about Statistical Analysis: Pre-processing of data such as normalization was specified at manuscript and Experimental Section.The data were presented as mean ± SD and sample size (n) for each statistical analysis was described at figure caption.The name of the statistical test and value were clearly stipulated.

Figure 1 .
Figure 1.Concept, design, and fabrication of SAMP.a) Schematic illustration of SAMP and main considerations.b) Bidirectional communication between SAMP and neural system through integration of multifunctionality.c) Fabrication procedure for SAMP and cross-section image of each step.Scale bars: 5 mm, 100 μm, 100 μm, and 300 μm.d) The optical images of 3T3 cell on CNT sheet and glass substrate.Scale bars: 100 μm.Cell viability and morphology evaluation.The statistical analysis is two-tailed test.Cell viability: p = 0.4890, t = −0.7106,d.f.= 14 (N.S. p > 0.05, two sample t-test).e) Illustration describing prolonged operation of flexible SAMP.

Figure 3 .
Figure 3.In vivo assessment of SAMP.a) Photograph of backend-connected SAMP.0.5 g total weight.Scale bar: 5 mm.b) Schematic illustration of implanted CNT probe and possible output: optically evoked potential and spontaneous activity recording.c) The recorded neural stream of OEP and spontaneous activity in both local field potential and extracellular potential.d) Evoked potential and its raster plot in response to 10 and 100 Hz optical stimulation (40 trials).e) Successful evoked spike ratio with frequency of 5, 10, 20, 30, and 100 Hz (40 trials).f) Light intensity dependent peak amplitude of optically evoked potential.Endogenous single-unit action potential recording.g) Recorded neural stream.ISI histogram and firing rate of two individual units.h) Overlapped waveform of extracted spike units.i) PCA results and its clustering of the units.Optogenetic behavior test with SAMP in freely moving animal.j) Schematic illustration for behavioral test with ChR2 mice using implanted SAMP at M2 cortex.k) The number of rotations in each epochs with ChR2 mice and WT mice (n = 4).The statistical analysis is one-tailed test.ChR2 mice, rotation: p = 0.0223, t = −2.5285;d.f.= 6 (*p < 0.05, paired t test).l) The average speed in each epochs with ChR2 mice and WT mice (n = 4).The statistical analysis is one-tailed test.ChR2 mice, speed: p = 0.0006, t = −5.8015;d.f.= 6 (***p < 0.001, paired t test).m) Trajectory of ChR2 mice in open field arena during OFF/ON session (3 min time bin).The spinal cord application of CNT sheet fiber probe.n) The image of spinal cord implanted mouse.o) Neural signal evoked by mechanical stimulation on mice paws.p) Changes of joint position of hindlimb with optical stimulation (100 Hz) and the photographs of mouse hindlimb response at 100 Hz optical stimulation.

Figure 5 .
Figure 5.Long-term neural operation with SAMP.a) Photograph of year-long SAMP implanted mouse and explanted probe.Scale bar: 100 μm.b) SNR of optically evoked potential recorded from SAMP for 1 year (n = 3).c) Successful evoked spike ratio comparison between 1 month and 1 year postimplantation of SAMP with frequency of 5, 10, 20, 30, 100 Hz.Endogenous single-unit signal tracking for 1 year with SAMP.d) Averaged waveform of extracted spikes and its ISI histogram.e) Overlapped recorded waveform of each time point.f) PCA results of each time point.g) Defined parameters in the action potential waveform.h) SNR, peak amplitude, duration, and PT ratio of recorded waveform for 1 year.
30 min was given in a 40 cm × 40 cm × 40 cm open field box before each trial.A white noise generator set at 85 dB and illuminance at 200 ± 20 lx was maintained for consistent environmental conditions.Behavioral experiments were conducted only once for each subject to prevent bias and were reflected in statistics.All behavioral experiments were video recorded and further analyzed based on behavioral phenotypes.The following parameters were analyzed using EthoVision XT (Noldus, Wageningen, The Netherlands): rotation, speed change, time spent in mobility, distance traveled.