Adaptive and multifunctional hydrogel hybrid probes for long-term sensing and modulation of neural activity

To understand the underlying mechanisms of progressive neurophysiological phenomena, neural interfaces should interact bi-directionally with brain circuits over extended periods of time. However, such interfaces remain limited by the foreign body response that stems from the chemo-mechanical mismatch between the probes and the neural tissues. To address this challenge, we developed a multifunctional sensing and actuation platform consisting of multimaterial fibers intimately integrated within a soft hydrogel matrix mimicking the brain tissue. These hybrid devices possess adaptive bending stiffness determined by the hydration states of the hydrogel matrix. This enables their direct insertion into the deep brain regions, while minimizing tissue damage associated with the brain micromotion after implantation. The hydrogel hybrid devices permit electrophysiological, optogenetic, and behavioral studies of neural circuits with minimal foreign body responses and tracking of stable isolated single neuron potentials in freely moving mice over 6 months following implantation.

The authors have addressed my comments with additional data and analysis, and now demonstrate clear superiority of this hydrogel device over the earlier polymer version. The hydrogel implant now can achieve longer single neuron recording and experiences earlier wound healing as compared to the polymer device, two key metrics of in vivo viability. This is an exciting result, and one that I look forward to seeing in publication!! I do have some minor comments to the authors, summarized below. However, despite where the data is presented, I do think that the polymer and hydrogen Vpp amplitude and noise should be presented on the same axis. Having different axis for the two electrode types is confusing at best. Please revise panels a and b to compare the two device types on the same axis. The finding that the voltage amplitude is lower for hydrogel than for polymer is interesting, and likely due to neurons being pushed further away from the device. Notably, the amplitude normalizes to the polymer devices by roughly 8 weeks. This highlights that hydrogel devices are likely preferable for long-term recordings, but may be equivalent to polymer for <4 week duration recording experiments. Using a single y axis will also demonstrate that the noise level of the hydrogel devices is actually much lower, which accounts for the improved SNR starting at ~ 4 weeks compared to the polymer devices. Comment 1. It is still unclear to this referee how many channels per probe remain functional over implantation time and how many units can be simultaneously recorded with one probe (see also comment #5 of previous Referee#2). This needs to be explicitly stated in the results and commented in the discussion. It seems that the maximal number of 2 units was recorded from one probe per animal out of 21 recording sites.
Response 1. We appreciate the reviewer's recommendation to explicitly state about the specific information of recorded units. In our hydrogel hybrid probes, each microelectrode could measure 1 or 2 single units while the electrodes in each microelectrode array fiber (7 electrodes per each fiber, 3 array fibers per probe) could only measure the same pattern due to the nature of connectorization (all electrodes in each fiber shared the same electrical connection, i.e. were shorted together on purpose). Therefore, the total number of single units recorded per probe varied between 2 and 6 in the current study. The goal of our study was not to maximize the number of units in this particular design, a task much more readily accomplished by the sophisticated and costly CMOS platforms such as Neuropixels, but rather provide a framework for combining multiple functional features while maintaining high-quality recordings and low foreign-body response. Future collaborative studies will undoubtedly improve upon the channel count by outfitting our platforms with integrated backends, which are not the topic of research in our team. Nevertheless, we appreciate and agree with the reviewer's suggestion to explicitly state the maximum recordable single units of our probe in the manuscript. To clarify these points, we have added the following sentences (Result & Discussion) and the new Supplementary Table 3 in the revised manuscript: On page 10, the sentence "Each microelectrode fiber in the hydrogel hybrid probe measured 1 or 2 single unit-potentials (spikes). It should be noted that individual electrodes in each microelectrode fiber were not electrically distinguishable due to collective electrical connectorization procedure, and thus they measured the same pattern. This cross-talk was not a feature of the electrode array design or polymer insulation, which were previously shown sufficient to avoid cross-talk between neighboring channels, but instead was a fabrication choice for this proof-of-concept study." On page 11, the sentence "The number of functional channels and the maximum number of recordable single units per probe considering overlapped patterns (between 2 to 6) are provided in the Supplementary Table 3." On page 12, the sentence "Although the hydrogel hybrid probes offer functional advantages in tissue response, there are several points of further improvement left for the future development and applications. First, the complexity, size, and recording capacity of the microelectrode fibers are limited by the back-end electrical connectorization and resultant overlap in signal patterns from adjacent electrodes in each microelectrode fiber. An improved approach to establish highresolution connections between functional fibers and back-end interfaces would enhance the throughput of recorded units."  figure 16: the electrical signals triggered by optogenetic stimulation fall within the illumination window in panel b, but not in panel a. There seems to be a mistake there.

Supplementary
Response 2. Thank you for pointing out the issues in Supplementary Fig. 16. Because of the endogenous characteristics of optogenetically-evoked potential, there is the time-delay (~10 ms) between the onset of optical stimulation and the electrical potential changes. After a thorough inspection of the data, we confirmed that the panel a is correctly presented while the blue bars representing the optical stimulation in panel b appeared (erroneously) longer due to illustration. We have the corrected this error in Supplementary Fig. 16 in the revised manuscript.   Supplementary Fig. 19). We believe that the dataset was incorrectly copied from the Excel spreadsheet into Igor for plotting (that is the fourth column from the "drug" dataset was accidentally copied into the "saline" plot). We added the corrected version of the graph in the revised manuscript (in panel a, fourth bar from the left (Saline, light ON, eYFP). The velocity table used for this calculation is provided below for the reviewers' examination: Comment 4. In the statistics section, it is stated that error bars in all graphs represent s.e.m., but in figure legends (e.g., Fig. 1 h-j) it is reported that error bars represent SD.
Response 4. Thank you for pointing the inconsistency in our statistics section. We have revised the Statistical analysis description in Methods section as the follows and updated the size of error bars to the standard deviation (SD) throughout the revised manuscript: On page 33, the sentence "Normality of data distribution was tested via Shapiro-Wilk test. For the analysis of the behavioral data that showed normal distribution, two-way ANOVA followed by Bonferroni's multiple comparison test and Tukey's multiple comparison test were conducted with threshold of *p < 0.05, **p < 0.01, ***p < 0.001. For immunohistochemistry experiments, where data distributions were found to be non-normal, Kruskal-Wallis test followed by Bonferroni's correction and Dunn-Sidak multiple comparison test were conducted with threshold of *p < 0.05. For the fiber characterization, paired two-sided Student's t-tests were used, and significance threshold was placed at *p < 0.05, **p < 0.01. Error bars represent standard deviation." Comment 5. Lines 269-270: the decrease in SNR 4 weeks after implantation in polymer probes needs to be statistically quantified.
Response 5. We appreciate the reviewer's insightful comment. To clarify this point, we have added the statistical values and edited the following description in the revised manuscript as well as revised the Supplementary Fig. 21: On page 11, the sentence "Furthermore, SNR in the recordings acquired with the hydrogel hybrid probes remains stable at least up to 24 weeks following the implantation, whereas the recordings acquired with polymer probes exhibit substantial decrease in SNR 4 weeks following implantation ( Supplementary Fig. 21, p = 0.000578 and 0.000124 between 4 weeks and 8 or 12 weeks, respectively)."

Comment 6.
A normality test must be run before choosing the statistical test to be used (parametric vs non parametric), rather than assuming normality based on previous work.
Response 6. We thank the reviewer for this insight. Following the reviewer's suggestion, we have performed normality tests on our data (Shapiro-Wilk test was employed due to number of samples < 30). As a result, we found that the null hypothesis (normal distribution) is not rejected for an alpha level of 0.05 in the majority of our experiments. However, the immunohistochemistry data showed the non-parametric characteristics, which motivated us to switch from the two-way ANOVA followed by Bonferroni's multiple comparison test with Tukey's multiple comparison test to the Kruskal-Wallis test followed by Bonferroni's correction with Dun-Sidak multiple comparison test in the revised manuscript. This has been reflected in our revised manuscript: On page 33, the sentence "Normality of data distribution was tested via Shapiro-Wilk test. For the analysis of the behavioral data that showed normal distribution, two-way ANOVA followed by Bonferroni's multiple comparison test and Tukey's multiple comparison test were conducted with threshold of *p < 0.05, **p < 0.01, ***p < 0.001. For immunohistochemistry experiments, where data distributions were found to be non-normal, Kruskal-Wallis test followed by Bonferroni's correction and Dunn-Sidak multiple comparison test were conducted with threshold of *p < 0.05. For the fiber characterization, paired two-sided Student's t-tests were used, and significance threshold was placed at *p < 0.05, **p < 0.01. Error bars represent standard deviation."

Fig. 4 | Long-term investigation of single-unit potentials and foreign body responses. a,
Representative spontaneous electrophysiological activity recorded in the vHPC using a hydrogel hybrid probe 1 month following implantation. b, Principle component analysis of the single-unit potentials (spikes) recorded 3 days, 1 week, 2 weeks, 1-6 months following implantation. Closed circles indicate centers and ellipses represent two standard deviations (2σ) contours of the principle component distributions. c, Average spike waveforms recorded over time corresponding to clusters in (b). d, Average amplitude (red) and noise level (blue) of spikes in (c) recorded over time. e, Average inter-spike intervals for each unit. Significance is confirmed by one-way Student's t test (*** p < 0.001). f-i, Average immunofluorescence area quantifying the presence of GFAP (f), Iba1 (g), CD68 (h), and IgG (i) in the vicinity of the hydrogel hybrid probes, polymer probes, and stainless steel microwires for 3 days, 1 week, 1 month, 3 months, and 6 months following implantation, respectively. Values in f-i represent the mean and the standard deviation (n = 6 for each device and time point, Kruskal-Wallis test and Dun-Sidak multiple comparison test (*p < 0.05).

Comment 7. Line 265 and lines 234-235: please specify the statistical test used.
Response 7. Thank you for your comment regarding the statistical test. We have specified the statistical test we used for each experiment in the revised manuscript: On page 9, the sentence "Consistent with previous studies of the BLA-to-vHPC projection circuit, optical stimulation (20 Hz, 5 ms pulse width, 10 mW mm -2 ) of the BLA inputs in the vHPC delivered via the hybrid probes results in reduction of the time spent in the center of the open field by mice expressing ChR2-eYFP as compared to eYFP controls (two-way ANOVA and Bonferroni multiple comparison test, p = 0.0031) (Fig. 3h,j,k)." On page 10, the sentence "The different units further maintain their characteristics and statistically distinct inter-spike intervals (one-way Student's t-test, p < 0.001) (Fig. 4e)." Comment 8. Fig. 1I:

the text reports that the electrical properties of the hybrid device are not significantly affected (either increased or decreased, both directions) by bending. However, the test is one sided (as stated in the statistic analysis section of the Material and Methods), so what has been tested is if bending leads to an increase (or a decrease, but anyway only one direction) in the impedance of the electrode. Please amend. Moreover, shouldn't the test be paired rather than unpaired?
Response 8. Thank you for your insightful comment. We have used paired two-sided Student's t test for statistical analysis of fiber characterization in the revised manuscript: On page 17, caption Fig. 1i "Tip impedance of the electrodes within the fiber arrays in the hydrogel hybrid probes at 0° and 90° bending (paired two-sided Student's t-test: p = 0.5232)." On page 33, the sentence "For the fiber characterization, paired two-sided Student's t-tests were used, and significance threshold was placed at *p < 0.05, **p < 0.01. Error bars represent standard deviation." Comment 9. Please remove "similar trends are observed also for CD68 and IgG". At 3 and 6 months time there is no clear trend to decrease in CD68 and IgG expression.
Response 9. We appreciate the reviewer's suggestion about the clearer explanation for our immune-histochemistry experiments. In the revised manuscript, we removed the following sentence: "While the differences in GFAP and Iba1 expression are statistically significant, similar trends are also observed for CD68 and IgG." We greatly appreciate the reviewer's time and invaluable suggestions which have significantly improved the work. We are particularly grateful to the reviewer for identifying errors in our data presentation as well as suggestions for improving our statistical analyses. We have learned a lot and we sincerely hope that our revised manuscript fully addresses the reviewers' concerns.

Response to Reviewer #2:
General comment. As I summarized in my first review "Flexible array technologies, particularly those with multimodal function, is of keen interest to the research and translational neurotechnology communities. The incorporation of a hydrogel structure builds well on this lab's past work using polymer multimodal fibers, and offers a reduction in shear forces produced by mechanical mismatch between the brain and stiffer electrodes. This body of work shows that hydrogel probes are capable of being inserted into the brain, delivering light to stimulate photosensitive ion channels, delivering pharmacological agents, and recording evoked and spontaneous action potentials. In addition, these experiments demonstrate a reduction in lateral stress produced by simulated brain micromotion. Taken together, this is an exciting body of work demonstrating a technology with the possibility for great benefit for experimental protocols." The authors have addressed my comments with additional data and analysis, and now demonstrate clear superiority of this hydrogel device over the earlier polymer version. The hydrogel implant now can achieve longer single neuron recording and experiences earlier wound healing as compared to the polymer device, two key metrics of in vivo viability. This is an exciting result, and one that I look forward to seeing in publication!! I do have some minor comments to the authors, summarized below.

Response.
We are delighted that the reviewer has found our work important and exciting. To make our manuscript clearer for the readers, in the following paragraphs, we address each comment point-by-point. Sentences newly added to or revised in the manuscript and the supplementary information are marked in blue. Figure 3, rather than in Supp Fig 19. In my opinion, this is the single most important figure of the in vivo characterization.

Comment 1. The data presented comparing the polymer and hydrogel version of the probe is important and compelling. I would encourage the reviewers to include this comparison in the main
However, despite where the data is presented, I do think that the polymer and hydrogen Vpp amplitude and noise should be presented on the same axis. Having different axis for the two electrode types is confusing at best. Please revise panels a and b to compare the two device types on the same axis. The finding that the voltage amplitude is lower for hydrogel than for polymer is interesting, and likely due to neurons being pushed further away from the device. Notably, the amplitude normalizes to the polymer devices by roughly 8 weeks. This highlights that hydrogel devices are likely preferable for long-term recordings, but may be equivalent to polymer for <4 week duration recording experiments. Using a single y axis will also demonstrate that the noise level of the hydrogel devices is actually much lower, which accounts for the improved SNR starting at ~ 4 weeks compared to the polymer devices.
Response 1. We appreciate the reviewer's recommendations how to improve our data presentation. Following these suggestions, now we moved the previous Supplementary Fig. 19 to Main Manuscript Fig. 3. Furthermore, the data from both the hydrogel and polymer probes are now presented on the same axis. Also, the following sentences have also been revised in our manuscript: On page 9, the sentence "Potentials correlated with laser pulses are observed 1-2 weeks after the virus injection, and their amplitudes gradually increase up to 8 weeks, at which point they reach a stable level maintained for at least 4 more months (Fig. 3d). No light-evoked potentials are recorded in mice injected with the control virus (Supplementary Fig. 18). Consistent with prior studies 35 , the noise level decreases 2 weeks following the implantation and is maintained for at least 6 months (Fig. 3e). Signal-to-noise ratio (SNR) gradually increases over the first 4 weeks following the implantation and then plateaus for another 5 months, in contrast to the gradually decreasing SNR recorded from the polymer probes 5 (Fig. 3f)."

Fig. 3 | Optogenetic interrogation by hydrogel hybrid probes. a, A photograph of a mouse
implanted with the hydrogel hybrid probe. Scale bar: 2 cm. b, An illustration of optogenetic modulation and electrophysiological recording in the BLA-to-vHPC projection circuit. c, Representative electrophysiological recordings in the vHPC during optical stimulation (10 Hz, 10 mW mm -2 , 5 ms pulse width) using the hydrogel hybrid probes 3 days, 1-2 weeks, and 1-6 months following implantation and transfection with AAV5-CaMKII::ChR2-eYFP. d-f, Amplitude of optically-evoked potentials (d) and background noise (e), and signal-to-noise ratio (SNR) (f) in the opto-electrophysiological experiments recorded from the hydrogel hybrid (red) and polymer probes (  Response 2. Thanks for your insightful comment from the reviewer about the number of animals and recorded units per probe. We used 8 animals (1 probe per animal), and the presenting data is average value of one electrode from different probes. Multiple electrodes in a single device are not considered to exclude the device-dependency. We agree that specific information about recording performance of our devices is critical to their future applications. Therefore, we have added the Supplementary Table 3 and revised our manuscript accordingly.
Briefly, in our hydrogel hybrid probes, each microelectrode could measure 1 or 2 single units while the electrodes in each microelectrode array fiber (7 electrodes per each fiber, 3 array fibers per probe) could only measure the same pattern due to the nature of connectorization (all electrodes in each fiber shared the same electrical connection, i.e. were shorted together on purpose). Therefore, the total number of single units recorded per probe varied between 2 and 6 in the current study. The goal of our study was not to maximize the number of units in this particular design, a task much more readily accomplished by the sophisticated and costly CMOS platforms such as Neuropixels, but rather provide a framework for combining multiple functional features while maintaining high-quality recordings and low foreign-body response. Future collaborative studies will undoubtedly improve upon the channel count by outfitting our platforms with integrated backends, which are not the topic of research in our team.
On page 10, the sentence "Each microelectrode fiber in the hydrogel hybrid probe measured 1 or 2 single unit-potentials (spikes). It should be noted that individual electrodes in each microelectrode fiber were not electrically distinguishable due to collective electrical connectorization procedure, and thus they measured the same pattern. This cross-talk was not a feature of the electrode array design or polymer insulation, which were previously shown sufficient to avoid cross-talk between neighboring channels, but instead was a fabrication choice for this proof-of-concept study." On page 11, the sentence "The number of functional channels and the maximum number of recordable single units per probe considering overlapped patterns (between 2 to 6) are provided in the Supplementary Table 3." On page 12, the sentence "Although the hydrogel hybrid probes offer functional advantages in tissue response, there are several points of further improvement left for the future development and applications. First, the complexity, size, and recording capacity of the microelectrode fibers are limited by the back-end electrical connectorization and resultant overlap in signal patterns from adjacent electrodes in each microelectrode fiber. An improved approach to establish highresolution connections between functional fibers and back-end interfaces would enhance the throughput of recorded units." On page 31, the sentence "In the signal analysis of the device electrical performance (amplitude, noise, and SNR), 8 animals were used (1 probe implanted per animal) and the average value of one electrode from different probe was used. Multiple electrodes in a single device were not considered to exclude the device-dependency."