An Opto-Microperfusion Neural Probe for Multiplexed Detection of Optogenetically Evoked Neurochemical Release Dynamics

Abstract

Probing the spatiotemporal release dynamics of multiple neurochemicals simultaneously in vivo is essential for understanding the molecular basis of brain function and behavior. Microdialysis enables multiplexed neurochemical analysis when combined with advanced molecular tools such as mass spectrometry, but existing microdialysis probes typically exhibit low recovery rates and provide temporal resolution of several minutes, which is insufficient to capture rapid neurochemical release events. Moreover, it remains challenging to investigate the release of multiple neurochemicals upon cell type-specific neuromodulation with a single probe. Here, we present an opto-microperfusion platform that integrates time-sequential microfluidics with cell type-specific optogenetic neuromodulation and push–pull microsampling capabilities. We demonstrated sub-minute temporal resolution in vitro and near-minute temporal resolution in vivo, with in vivo sampling intervals limited by downstream analytical requirements. The platform achieved high recovery rates for both small molecules and neuropeptides, including 76% for serotonin, 69% for dopamine, and approximately 14% for neuropeptide Y, representing a significant improvement over conventional microdialysis probes. In vivo studies in rats demonstrated reliable microsampling of multiple neurochemicals with minimal sample-to-sample variation. Studies in mice demonstrated the simultaneous detection of the release dynamics of up to 23 different neurochemicals following optogenetic stimulation when coupled with mass spectrometry. Together, this multimodal neural probe paves the way for investigating the molecular mechanisms underlying behavior and neurological disorders.

Introduction

Tools and technologies for probing spatiotemporal dynamics of neurochemicals in vivo are now making it possible to understand where, when, and how the release of neurochemicals modulates diverse behavioral outputs of the brain at the cellular and circuit levels. Current measurement technologies, such as fast-scan cyclic voltammetry (FSCV) or genetically encoded fluorescent sensors, often have high spatiotemporal resolution but limited multiplexing capability (1, 2). FSCV, which typically employs a carbon-fiber electrode housed in a fused-silica capillary, enables real-time detection of redox-active neurochemicals with high spatiotemporal resolution (3, 4). However, many neurochemicals, such as glutamate, acetylcholine, γ-aminobutyric acid, and most neuropeptides, are electrochemically inert and therefore not detectable directly by FSCV, although enzyme-modified electrodes can expand the scope of electrochemical sensors (5). Recent advancements in genetically-encoded sensing technology have revolutionized the monitoring of neurochemicals to provide fast, cell type-specific detection. Many sensitive genetically-encoded fluorescent sensors have been developed for imaging the dynamics of small molecules and neuropeptides, including GRAB_DA (6), GRAB_NE (7), dLight1 (8), and NPRRs (9). Nevertheless, these sensors only enable the detection of one or two specific neurochemicals, and multiplex monitoring remains challenging (10). Furthermore, while the co-release of multiple neurochemicals has long been suggested to play an important role in circuit function (11, 12), the ability to detect coreleased neurochemicals simultaneously using these genetically encoded sensors is limited (10). In addition, genetically-encoded fluorescent sensors require complicated genetic modification of targeted cells, which limits their potential for translation into humans. Finally, quantifying basal neurochemical concentrations using this type of sensor remains challenging due to the lack of an absolute fluorescence baseline and difficulties in calibration within complex in vivo environments (10).

Microdialysis-based probes, comprising inlet and outlet capillaries housed within a hollow fiber sealed by semipermeable membranes, have been widely used in both clinical and fundamental neuroscience research. Unlike FSCV and genetically-encoded sensors, microdialysis allows multiplexed analysis when combined with liquid chromatography-mass spectrometry (LC-MS). To enhance the recovery efficiency, it is important to minimize the diffusion resistance imposed by the semipermeable membrane (13–15). To that end, membrane-free sampling technologies such as microperfusion have been developed (16, 17). The membrane-free design removes concerns about non-specific adsorption onto membrane surfaces and diffusion limitations, thereby enabling microsampling with high recovery rates, particularly for large-molecular-weight neuropeptides and neuroproteins (18, 19).

The temporal resolution of fluid sampling plays a critical role in studies focused on the dynamics of neurochemical release. For current microdialysis and microperfusion sampling probes, the temporal resolution of neurochemical release is limited by both Taylor dispersion (20, 21), in which the collected neurochemicals diffuse/spread in a tube with flowing liquid, and the amount of time required to collect sufficient samples for analysis. Typical microdialysis probes generally offer a temporal resolution of several minutes (22–24). Because neurochemical release could occur on the sub-minute or lower timescale (25), these traditional microdialysis methods lack the required temporal resolution to capture such release dynamics. Recent studies have coupled microdialysis or push-pull sampling probes with droplet microfluidics to solve this sample dispersion issue (15, 26, 27), but analysis of ultra-low concentrations of neurochemicals in a nanoliter-scale droplet of cerebrospinal fluid remains challenging.

Most importantly, existing microdialysis or microperfusion systems are limited in their ability to combine neuromodulation with multi-analyte neurochemical sensing within a single platform (28). This limitation has left many molecular mechanisms underlying behavior and neurological disorders unexplored. Optogenetic techniques are widely used for such studies, as they enable precise spatiotemporal control of specific cellular activities in heterogeneous tissues (29–31). This approach involves transfecting neurons with genes encoding light-sensitive ion channels or pumps, enabling selective activation or inhibition of neuronal activity in response to specific wavelengths of light (32). These manipulations can elicit action potentials and modulate the release of various neurotransmitters upon light stimulation (33–35). While platforms integrating optogenetics with electrophysiological recordings have been reported (17, 36), there has not been much progress in the integration of optogenetics with microperfusion probes to investigate the optically-evoked, simultaneous release of multiple neurochemicals within defined neuroanatomical structures. A previous study reported an optodialysis probe that integrates optical fibers with microdialysis probes (37), the platform suffers from limited temporal resolution (15 minutes) and low recovery rate due to the membrane barrier and non-specific adsorptions on the membrane surface.

In this study, we report an opto-microperfusion platform that incorporates time-sequential microfluidics for push-pull microsampling following optogenetically-evoked release of multiple neurochemicals with high spatial and temporal resolution. We have demonstrated in vitro recovery rates of 69% for dopamine, 76% for serotonin, and ~14% for neuropeptide Y (NPY), surpassing conventional microdialysis probes. The integrated time-sequential microfluidics enabled microsampling with sub-minute temporal resolution. We have also performed in vivo studies in rats that demonstrate reliable collection with small variations among different time-sequential microsampling reservoirs. Additional studies in mice demonstrated the successful capture of the release dynamics of up to 23 neurochemicals upon optogenetic stimulation. Overall, this study lays the foundation for the development of next-generation multimodal neural probes to investigate the molecular mechanisms underlying behavior and neurological disorders.

Results and Discussion

Overall design and operation principle

Our opto-microperfusion device consists of three key components: (1) a time-sequential collection chip with integrated reservoir fill-state indicators for microsampling with up to sub-minute temporal resolution; (2) a filament probe with an integrated microscale inorganic light-emitting diode (μ-ILED) for optical stimulation; and (3) a soft microfluidic probe incorporating push-pull microchannels (40 × 40 μm² cross-section SI, Fig. S1) for membrane-free microperfusion (Fig. 1A, B). This membrane-free push-pull microperfusion eliminates the diffusion resistance and non-specific adsorption associated with conventional microdialysis probes, thereby substantially improving recovery rates, particularly for large-molecular-weight neuropeptides in the brain.

Figure 1. Design and operating principle of the opto-microperfusion probe.

(A) Exploded schematic of the opto-microperfusion probe. (B) Pseudo-colored scanning electron microscopy (SEM) image of the probe, showing the push and pull microchannels (yellow), microscale inorganic light-emitting diode (μ-ILED; blue), and flexible printed circuit board (fPCB; orange). (C) Schematic of the time-sequential collection chip for discrete sample storage. (D) System architecture for wireless communication and control. Sampled fluid sequentially filled reservoirs R1–R6 on the collection chip and activated the surrounding fill-state indicators. The activated indicators generate electrical signals that are transmitted to the user interface through Bluetooth. The Bluetooth controller also controls the frequency and duty cycle of the μ-ILED for optical stimulation. (E) Optical images of the fully assembled opto-microperfusion device, showing top and bottom views. (F) In vitro demonstration of stereotactic positioning, probe implantation, microsampling, and optical stimulation in a brain phantom composed of 0.6% agarose gel.

During operation, an external syringe pump delivers perfusate through the push microchannel into the targeted brain region, where molecular exchange occurs at the probe tip. A second pump generates negative pressure through the pull channel, drawing the sampled fluid into a set of micro-reservoirs (6–21, each with a volume ranging from 0.1 to 1.0 μL) sequentially via passive capillary-burst valves (CBVs). At a flow rate of 1.0 μL/min, depending on the volume of the microreservoir, the chip fills each microreservoir within a few seconds to one minute, enabling microsampling with high temporal resolution. This time-sequential sampling facilitates temporal profiling of neurochemical concentrations (Fig. 1C). The time-sequential collection chip can be easily detached from the device and replaced, enabling several hours of microperfusion without additional surgical interventions. This modularity, combined with programmable μ-ILED control (detailed below), enhances versatility for neuroscience applications.

To accurately monitor the filling time of each micro-reservoir, we integrated a set of reservoir fill-state indicators based on impedance changes on a printed circuit board (PCB). This PCB also controls the operation frequency and duty cycle of the μ-ILED (Fig. 1D). Specifically, once the sampled fluid reaches the collection chip, the reservoir fill-state indicators detect impedance changes, which are processed through an impedance conversion chip and relayed to a microcontroller unit (MCU). Simultaneously, the MCU receives Bluetooth commands from the operator, enabling real-time, programmable optical stimulation via the integrated μ-ILED (Fig. 1D).

The fully assembled system features a compact, lightweight design (Fig. 1E), with a total device weight of 1.1 g (Fig. S2). This miniaturized construction enables use in small animal models such as mice and rats with minimal behavioral interference. The functionality of the device was first demonstrated for implantation, neurochemical sampling, and optical stimulation in an agarose gel-based brain phantom (Fig. 1F).

Fluidic characterization of the opto-microperfusion probe

The integrated opto-microperfusion probe comprises soft polydimethylsiloxane (PDMS)-based push-pull microchannels and a μ-ILED (205 μm × 305 μm; wavelength: 470 nm) mounted on a flexible polyimide-based filament (thickness: 100 μm) and a tungsten layer (thickness: 50 μm) that enhances mechanical rigidity and facilitates precise in vivo implantation (Fig. 2A, B). The integrated probe has a total cross-section of 400 μm× 350 μm, representing a 30% reduction in the area of potential tissue damage during in vivo studies compared to the guide cannula used for the commercial CMA 7 microdialysis probe with a guide cannula (diameter: 500 μm, Fig. 2C).

Figure 2. Fluidic characterization of the opto-microperfusion probe.

(A) Schematic of the probe structure. The μ-ILED and microperfusion channels were integrated on the top surface of the fPCB, and a tungsten steel stiffener was attached to the bottom surface to prevent bending during implantation. (B) Optical image of the opto-microperfusion probe, with the inset showing the illuminated μ-ILED and microperfusion channels. (C) Dimensional comparison between the opto-microperfusion probe and the commercial CMA 7 microdialysis probe. (D) Real-time in vitro recovery measurement using potassium ferricyanide solutions with opto-microperfusion probes (n = 3 biological replicates). (E) Average recovery rate of opto-microperfusion probes in potassium ferricyanide solutions at different concentrations (n = 3 biological replicates). As the external concentration changed (red), the opto-microperfusion probe showed high recovery and consistent performance (black). (F) Concentration response of opto-microperfusion probes to rising and falling potassium ferricyanide concentrations. (G) Recovery rate of opto-microperfusion probes and commercial CMA 7 probes for neurochemicals with different molecular weights: K⁺, 200 mM; DA, 500 nM; 5-HT, 50 nM; NPY, 5.0 ng/mL (n = 3 biological replicates).

To characterize the sampling dynamics of our opto-microperfusion probe, we developed a real-time concentration monitoring platform based on screen-printed electrodes (SI, Fig. S3). The probe performance was evaluated by immersing it in solutions of various concentrations of potassium ferricyanide. We chose the widely used redox reporter potassium ferricyanide as a demonstration example because it can be readily quantified using electrochemical methods (SI, Fig. S4). Fig. 2D shows real-time microsampling of potassium ferricyanide at a constant flow-rate of 1 μL/min. Following a step change in the target concentration, the sampled fluid signal showed an initial delay of approximately 1 min, which was mainly attributed to the dead volume between the microperfusion probe tip and the electrochemical detection reservoir. After this delay, the measured concentration increased roughly another 1 min before reaching a stable plateau. The probes achieved a steady-state recovery rate of approximately 80% across the tested input concentrations (Fig. 2E). Neurochemical concentrations can rapidly increase and/or decrease within the brain; thus, it is important for microsampling probes to accurately capture these rapid changes. To this end, using the same testing system, we introduced randomized fluctuations in the concentration of potassium ferricyanide and continuously monitored its levels in the sampled fluid. Our push-pull microperfusion system accurately captured local changes in chemical concentrations, regardless of the direction of fluctuation (Fig. 2F). This ensures that our system is capable of reliably tracking dynamic variations in neurochemical concentration.

Finally, we characterized the recovery rate of our push-pull microperfusion probe and benchmarked it against a conventional CMA 7 microdialysis probe (molecular weight cut-off: 6 kDa) by sampling representative neurochemicals, including ions (K⁺), small-molecule neurochemicals (dopamine and serotonin), and high-molecular-weight peptides (NPY) at a flowrate of 1.0 μL/min. Our push-pull microperfusion system demonstrated significantly higher recovery than the CMA 7 microdialysis probe (Fig. 2G), primarily due to its membrane-free operation, which eliminates diffusion resistance and non-specific adsorption to the microdialysis membrane. Overall, compared with the CMA 7 probe, which exhibits recovery rates of 30%, 16%, 10%, and 7.8% for K⁺, dopamine, serotonin, and NPY, respectively, our microperfusion probe achieved significantly higher recovery rates of approximately 81% for K⁺, 69% for dopamine, and 76% for serotonin, and 14% for NPY.

Design and characterization of time-sequential microfluidic collection chips

Our opto-microperfusion system integrates a time-sequential fluid-sampling device that immediately collects and stores the sampled fluid in individual microfluidic reservoirs. The chip contains multiple independent micro-reservoirs, each equipped with three precisely designed CBVs (Fig. 3A). These valves are positioned strategically to guide the sequential flow of sampled fluid through the chip, thereby enabling automated sample segmentation. The critical burst pressure (BP) of each valve is theoretically determined as follows (38, 39):

$$BP=-2\sigma \left[\left(\frac{cos\left({{\theta }_{\text{i}}}^{*}\right)}{b}\right)+\left(\frac{cos\left({\theta }_{\text{a}}\right)}{h}\right)\right]$$

where σ denotes the surface tension of water (72.8 mN/m), θ_a is the channel’s contact angle of PDMS (107°), and θ_i* represents the smaller value between (θ_a + β) and 180°. Here, β is the divergence angle of the channel, while b and h (150 μm) respectively correspond to the width and height of the diverging section. By adjusting these geometric parameters, we set the divergence angles of the three CBVs to 30°, 90°, and 135°. The channel widths of the three CBVs were set to 100 μm, 100 μm, and 20 μm, respectively. Consequently, the pressures applied to the three valves were finally estimated as 1,349 Pa, 1,740 Pa, and 7,564 Pa, respectively (Figure 3B). The sampled fluid that enters the device inlet will first open CBV #1 and flow into the collection reservoir (the burst pressure of CBV #2 is higher than that of #1). After filling this collection reservoir, the additional solution will burst CBV #2 and flow into the next collection reservoir (the burst pressure of CBV #3 is higher than that of #2). By repeating this process, multiple chambers can be filled in a sequential manner (Fig. 3A). Since the time-sequential operation relies on the difference among these three burst-pressure thresholds, sequential sampling can function properly even if the PDMS contact angle changes as a function of storage time after the surface treatment. In practice, as long as the PDMS surface remains hydrophobic (38), the time-sequential microsampling is preserved. Additionally, precise photolithography allows the three CBVs to be designed with distinct geometries, thereby widening the separation between their burst-pressure thresholds and improving operational tolerance to contact-angle variation. This time-sequential fluid sampling minimizes Taylor dispersion and avoids unwanted mixing between newly sampled fluid and the collected fluid. And in contrast to droplet microfluidics, the present design collects and stores discrete samples directly within on-chip reservoirs, without requiring an additional oil phase or transfer to external storage tubing. This feature facilitates compatibility with downstream LC-MS analysis and simplifies integration with a membrane-free push-pull microperfusion system. Finally, this design eliminates the need for external power or mechanical intervention beyond continuous fluid flow, allowing seamless integration with a membrane-free push-pull microperfusion system and greatly simplifying operation.

Figure 3. Design and operating principle of the time-sequential collection chip.

(A) Schematic of a microreservoir unit. Three capillary burst valves (CBVs; green, red, and blue) surround the microreservoir and regulate sequential filling by establishing distinct burst-pressure thresholds. (B) Optical micrographs of the three CBV designs used in this study (scale bar, 100 μm). Each valve features a different divergence angle (β), resulting in a unique burst pressure (BP). (C) Representative collection-chip designs with temporal resolutions at the minute, sub-minute, and second scales. Reservoir volumes and reservoir numbers are indicated for each design. (D) Filling times of individual microreservoirs at a flow rate of 1 μL/min for the three representative chip designs shown in C. The average temporal resolutions (σ_TR) were 63.2 s, 35.8 s, and 7.5 s, respectively. No significant differences in filling time were observed among reservoirs within each design (n ≥ 5 biological replicates; ns, not significant). (E) Time-lapse images of the sequential filling process. Different colors represent sampled fluid collected at different time points. After reservoir filling was completed, air was introduced to isolate adjacent samples and preserve temporal segmentation.

The design of time-sequential collection chips is versatile and highly adaptable, customized to meet the specific requirements of various experimental purposes. Here, we present three representative chip designs, each tailored to different temporal resolutions ranging from minutes to sub-minutes down to seconds with different reservoir volumes (Fig. 3C). The results indicate that the average temporal resolutions (σ_TR) for reservoir volumes of 1.0, 0.5, and 0.1 μL are 63.2 s, 35.8 s, and 7.5 s, respectively, when operating at 1 μL/min. The precise microfabrication process yielded highly consistent chip geometries (SI Fig. S5), which resulted in minimal variation in temporal resolutions across chips (Fig. 3D). Fig. 3E shows the representative operation of the time-sequential fluid-sampling device with different colored solutions. To eliminate potential cross-sample contamination across different micro-reservoirs, we manually injected air into the channels after fluid collection and detaching the chip from the system. This straightforward approach effectively separates each reservoir, thereby reducing the inter-reservoir diffusion (Fig. 3E). Overall, these in vitro characterizations demonstrate the feasibility of achieving sub-minute temporal resolution.

Electrical and thermal characterization of the miniaturized electronics and μ-ILED

To minimize experimental variability associated with manual handling during device operation, we developed miniaturized electronics and associated software to remotely monitor the filling status of the microfluidic chip. The system also enables wireless control of the operation frequency and duty cycle of the μ-ILED for the optogenetic stimulation. More specifically, the electronics consist of two layers. The top flow-indicator layer is precisely aligned with the time-sequential collection chip and features seven impedance sensors distributed around the six independent reservoirs to detect reservoir fill status. The bottom layer integrates a Bluetooth-enabled microcontroller unit (MCU) and an impedance converter for data processing and wireless transmission. These two layers are securely connected via a board-to-board connector, facilitating easy installation and replacement of the collection chip (Fig. 4A). Fig. 4B presents a circuit diagram of miniaturized electronics. Each of the seven impedance sensors is connected in parallel with a predefined calibration resistor. When sampled fluid contacts the sensor electrodes, a sudden decrease in impedance (Fig. 4C) is detected by the converter and transmitted wirelessly in real time to the user interface via the MCU.

Figure 4. Characterization of the miniaturized electronics and μ-ILED.

(A) Optical images of the electronic components. The flow-indicator layer contains seven reservoir fill-state indicators that interface with the time-sequential collection chip. The communication and control layer establishes mechanical and electrical connections to the flow-indicator layer through a board-to-board connector. (B) Schematic of the wireless control and sensing circuit. The reservoir fill-state indicators are connected in parallel with a calibration resistor. Activation of individual indicators decreases the total circuit impedance, enabling wireless monitoring of reservoir filling. (C) Total impedance measured during sequential activation of the seven reservoir fill-state indicators by fluid filling. Inset shows an optical image of a representative indicator (scale bar, 500 μm). (D) Temperature increase of the Bluetooth-enabled microcontroller unit (MCU) and impedance converter during continuous operation. (E) Thermal characterization of the time-sequential collection chip. Infrared imaging shows that the chip temperature remained within 1 °C of the ambient temperature after 30 min of operation. (F) Optical output power and power intensity of the μ-ILED as a function of driving voltage (n = 3, biological replicates). (G) Temperature increase induced by the μ-ILED under representative optogenetic stimulation conditions. (H) Graphical user interface (GUI) for wireless monitoring and control of the opto-microperfusion probe.

Electronics tend to generate heat, which can potentially cause thermal tissue damage and degrade neurochemicals in the sample housed within the collection chip. To address this potential concern, we monitored the temperature of the two primary components: the MCU and the impedance conversion chip. At steady state, the MCU and impedance conversion chip showed temperature increases of < 3 °C above ambient conditions (Fig. 4D), a range that is safe for mice and rats (40–42). We also measured the temperature of the fluid collection chip during device operation, which exhibited an increase of < 1 °C (Fig. 4E), confirming its suitability for preserving thermally sensitive neurochemicals (43).

We finally characterized the optical and thermal properties of the μ-ILED. By adjusting the input voltage, the μ-ILED achieves an optical output intensity of 0–400 mW/mm² (Fig. 4F), which is sufficient for most optogenetic stimulation paradigms (44, 45). By controlling the operation duty cycle and power intensity, the associated temperature increase can be maintained below 1 °C while still delivering adequate light intensity for effective optogenetic modulation (Fig. 4G). Fig. 4H shows a software graphical user interface (GUI) designed to streamline peripheral device operation and enable continuous, real-time impedance monitoring. Overall, the developed miniaturized electronics and associated GUI provide programmable control of μ-ILED operation and real-time monitoring of the filling status of the collection chip while minimizing thermal damage to surrounding tissues and collected samples.

In vivo microsampling in rat models

After completing the in vitro characterization, we assessed the in vivo performance of the devices using rat models. We targeted the nucleus accumbens (NAc) region for brain sampling, and the collected samples were analyzed using both high-performance liquid chromatography with electrochemical detection (HPLC-EC) and LC-MS (Fig. 5A). Histological analysis confirmed accurate targeting of the NAc with tissue damage comparable to that caused by the CMA 7 microdialysis probe, as shown in Nissl- and DAPI-stained coronal sections of the rat brain (Fig. 5B). HPLC-EC analysis of the collected samples revealed prominent peaks for dopamine and its metabolite 3,4-dihydroxyphenylacetic acid (DOPAC) (Fig. 5C). The DOPAC concentration in this sample was measured as 11.9 nM. The DA concentration in this sample was 1.18 nM, which is lower than DOPAC, as would be expected for extracellular dopamine levels in dopamine-rich regions such as NAc because of the high expression of the dopamine transporter (46). Although the DOPAC/DA ratio measured with the microperfusion probe was lower than values previously obtained in our laboratory using commercially available microdialysis probes (DOPAC/DA = 99.22 ± 18.7, n = 4, SI Table S1), DOPAC remained higher than DA concentrations, consistent with the expected extracellular neurochemical profile in the NAc. Together, these initial results demonstrate successful in vivo microsampling using our push-pull microperfusion probes.

Figure 5. In vivo microsampling using opto-microperfusion probes in rat models.

(A) Timeline of two in vivo sampling schemes coupled to distinct chemical analysis methods. HPLC-EC analysis used 30 μL samples collected in Eppendorf tubes (top), whereas liquid chromatography-mass spectrometry (LC-MS) analysis used 1 μL samples collected in time-sequential collection chips (bottom). (B) Comparison of implantation tracks for microperfusion and CMA 7 probes. Coronal brain sections stained with Nissl (left) and DAPI (right) show comparable tissue disruption after implantation of the probe. (C) HPLC-EC analysis of dopamine (DA) and 3,4-dihydroxyphenylacetic acid (DOPAC). Chromatograms from biological samples (red) matched DA and DOPAC standards (black), confirming neurochemical collection from the rat brain. (D) Filling times of individual microreservoirs during in vivo sampling. No significant difference in filling time was observed between chips used in the same implantation experiment. (E) Representative LC-MS extracted ion chromatogram from an in vivo sample. (F) Heat maps of neurochemicals across sequentially collected samples from three representative rats. Data was normalized across animals and analytes and are presented as z scores. HVA concentrations in Rat 2 remained below the limit of detection (LOD), precluding z-score calculation; these values are therefore shown as blank.

After validating the functionality of our microperfusion probes via HPLC-EC analysis, we performed time-sequential microsampling in rat models. Samples were collected from three rats, with two independent time-sequential collection chips used for each rat. Each microreservoir provided a temporal resolution of 50–80 s, enabling near-minute sampling with minimal chip-to-chip variation (Fig. 5D). We analyzed levels of as many as 23 different chemicals (SI Table S2) in each biological sample using LC-MS (Fig. 5E and SI Table S3). Z-score normalized heatmaps of the concentrations for the three animals (Fig. 5F) show that the concentrations for most samples collected fell within ±3 standard deviations of each other, demonstrating stable baseline microsampling throughout the experiment.

In vivo time-sequential microsampling following optogenetic stimulation in mice

Finally, we validated the in vivo performance of our opto-microperfusion platform to probe the release dynamics of multiple neurochemicals after optogenetic neuromodulation. To that end, we injected an adeno-associated virus (AAV) carrying a Cre-dependent construct, AAV₂-Ef1α-DIO-ChR2-EYFP, into the ventral tegmental area (VTA) of DAT-Cre mice. In this knock-in mutant mouse line, the enzyme Cre recombinase is exclusively expressed in dopamine neurons, which are defined by their expression of the dopamine transporter or DAT (gene: Slc6a3, protein: SLC6A3). The “DIO” (double-floxed inverted open reading frame) design of our viral construct acts as a genetic switch; only in Cre-expressing dopamine neurons does the construct flip to its active orientation to express its payload: channelrhodopsin-2 (ChR2), a light-sensitive ion channel that enables neuronal activation with blue light, and a fluorescent tag (EYFP) for histological verification. This strategy ensures precise, light-based control over only the VTA dopamine neuron population. Following a four-week incubation period, we confirmed successful and specific expression via fluorescence imaging of both the VTA and its major projection target, the NAc (Fig. S6 and Fig. 6A). Since dopamine rapidly oxidizes under non-buffered, antioxidant-free conditions (47), we added 300 μM ascorbic acid to the perfusate artificial cerebrospinal fluid (aCSF)during the operation. A syringe pump delivered perfusate at 1 μL/min through the device, and baseline sample chips were collected without light stimulation. Optical stimulation was then initiated using a μ-ILED pulsing at 20 Hz with a 20% duty cycle, while the same perfusion rate was maintained for sampling during the stimulation (Fig. 6B). Collected samples were extracted from the time-sequential collection chips and stored at −80 °C until LC-MS analysis. Although the system collected samples at 1 min per microreservoir and we have demonstrated the feasibility of analyzing 23 different chemicals using 1 μL samples in rat models, here, we combined two adjacent collected samples for LC-MS analysis to ensure analytical robustness. Therefore, the effective temporal resolution achieved in these in vivo studies was 2 min. Notably, in principle, we could achieve the same near-minute temporal resolution as in the rat studies. Across three mice, we determined a total of 23 neurochemicals (Fig. 6C and SI Table S4). Within each animal, comparisons of samples collected before and during optical stimulation revealed changes in the overall neurochemical profile in the NAc, with data from a representative mouse shown in Fig. 6D. Importantly, the opto-microperfusion platform enabled simultaneous characterization of the release dynamics of up to 23 neurochemicals. For example, in mouse #1, the dopamine levels increased upon optogenetic stimulation (Fig. 6E). Similarly, the concentrations of 3-Methoxytyramine (3MT) and DOPAC, both major metabolites of dopamine, also increased with the duration of optogenetic stimulation (Fig. 6F–G). Collectively, these results demonstrate the device’s capability for integrated optical stimulation and real-time neurochemical sampling. Combined with LC-MS analysis, this platform provides a powerful tool for investigating dynamic neuromodulation and in vivo neurotransmitter fluctuations.

Figure 6. In vivo time-sequential microsampling following optogenetic stimulation in mice.

(A) Confocal images confirming EYFP-ChR2 expression in the ventral tegmental area (VTA) and nucleus accumbens (NAc) of DAT-Cre mice. The probe track in the NAc is indicated. (B) Experimental workflow for in vivo optogenetic stimulation and time-sequential microsampling. Each mouse underwent baseline sampling followed by sampling during optical stimulation. Samples were extracted from the time-sequential collection chips, frozen at −80 °C, and analyzed by LC-MS. (C) Neurochemicals quantified by LC-MS in sampled fluid, with mean baseline concentrations ± SEM, from a representative DAT-Cre mouse (mouse #1; n = 9 samples, except serine, n = 8). (D) Heat maps of mean neurochemical concentrations before and during optical stimulation (n = 9 samples per condition, except baseline serine and dopamine during stimulation, n = 8) from DAT-Cre mouse #1. (E–G) Release dynamics of representative neurochemicals from DAT-Cre mouse #1: (E) dopamine (DA), (F) 3-methoxytyramine (3MT), and (G) 3,4-dihydroxyphenylacetic acid (DOPAC). Orange shading indicates baseline sampling, and green shading indicates sampling during optical stimulation. For E-G, concentrations were normalized to the mean of the first three baseline samples collected. One DA data point (13,600% of baseline) was identified as an outlier and the raw value (255.81 nM) is provided in SI Table S4.

Discussion

In this study, we report the development of a multimodal opto-microperfusion platform that integrates a time-sequential collection chip and push-pull microsampling with a μ-ILED, enabling cell type-specific optogenetic neuromodulation and neurochemical sampling with near-minute temporal resolution in vivo. These advanced features overcome the limitations of traditional microdialysis probes, such as low recovery rates, insufficient spatiotemporal resolution, and the lack of in vivo optogenetic stimulation capabilities. In contrast to previously reported microdialysis platforms (48, 49), our device minimizes Taylor dispersion through the use of integrated time-sequential collection chips. Moreover, compared to our earlier push-pull design (18), the current platform supports uninterrupted sampling over extended durations, allowing real-time tracking of neurochemical release dynamics associated with neural activity. This capability facilitates concurrent monitoring of changes in neurochemical signaling during optogenetic stimulation using a single integrated platform. To support automated operation, we developed a wireless μ-ILED control module and an impedance-based reservoir fill-state indicator array, which reduces the need for manual intervention and minimizes variability in data acquisition.

The opto-microperfusion platform probe is also compatible with a range of downstream analytical modalities, enabling the measurements of different neurochemicals. For example, when coupled with HPLC-EC, the platform enables the quantification of DA and DOPAC from in vivo samples, with a limit of detection (LOD) of 1 nM for DA in 30 μL of sample. However, because this analytical workflow requires a 30 μL sample for each measurement, the effective temporal resolution is limited to 30 min per analysis when operating at typical flow rate of 1 μL/min. In contrast, coupling the probe with LC-MS enables broader neurochemical coverage, with simultaneous analysis of up to 23 neurochemical compounds and favorable analytical sensitivity, including LODs of 0.5 nM for DA, 0.5 nM for 3MT, 30 nM for DOPAC, 30 nM for glutamate (Glu), and 3 nM for γ-aminobutyric acid (GABA) using 1 μL samples. Under this configuration, the system provides a near-minute temporal resolution. Together, these results indicate that both HPLC-EC and LC-MS provide viable analytical routes for in vivo samples collected by the opto-microperfusion platform, with LC-MS providing broader multiplexed neurochemical coverage with substantially improved temporal resolution.

Compared with recently developed neural microsampling probes (SI Table S5), our platform provides a balanced solution for in vivo studies by combining integrated optogenetic stimulation, high recovery rate, spatiotemporal precision, and compatibility with broadly targeted neurochemical analysis. For example, silicon-based nanodialysis probes or borosilicate glass-based microsampling platforms achieve microsampling with high recovery rate and small footprint (13, 19, 50, 51), but their relatively high modulus (~ 130–188 GPa for silicon and ~ 60 GPa for borosilicate glass) may be less favorable for long-term in vivo applications compared with PDMS-based probes (Young’s modulus: 1–3 MPa). On the other hand, PDMS- or polymer-based devices are typically larger in size, yet better matched to soft neural tissue which can reduce tissue reactions and enable long-term microsampling (18, 52). The choice of downstream analytical strategy also imposes important constraints. Droplet-segmentation methods can preserve rapid concentration changes and enable sub-second-scale temporal resolution (13, 27, 52), but the resulting pL–nL sample volumes can make downstream chemical analysis more challenging. Optical and colorimetric methods offer fast readout for specific analytes from ultra-small-volume samples, although their applicability is limited by the available scope of suitable chemistries (13, 52). Similarly, NanoESI-MS can analyze very small samples with multiplexing capability, but often with a narrower analyte scope than LC-MS (16, 27). In contrast, LC-MS provides broad neurochemical coverage and robust multiplexed quantification but requires larger sample volumes and therefore offers lower temporal resolution than droplet-segmentation approaches (50, 53).

The current system is optimized for high-efficiency neurochemical sampling with high spatiotemporal resolution in anesthetized animals. However, advanced push-pull microdialysis platforms have demonstrated the feasibility of sampling in freely moving animals (18). Future development will focus on incorporating miniature push-pull microactuators to enable fully wireless operation in freely behaving subjects. Additionally, replacing conventional photolithography-fabricated molds with molds produced by grayscale lithography or high-resolution 3D printing could enable the fabrication of rounded microchannels, which may reduce Taylor dispersion during sample transport and thereby better preserve temporal resolution. Finally, future iterations of the probe will support sampling and analysis of neuropeptides and proteins across larger domains and temporal windows. When coupled with advanced proteomic workflows, the system could provide deeper insight into the molecular underpinnings of a wide variety of neuropsychiatric and neurodegenerative diseases, across diverse freely moving animal models.

In summary, this work introduces a multimodal microsampling neural probe that captures release dynamics of multiple neurochemicals with a high recovery rate and near-minute temporal resolution in vivo. With further development, the platform may enable wireless neuromodulation and neurochemical tracking in freely moving animals, paving the way for next-generation tools to investigate the molecular processes underlying diverse behaviors and neurological disorders.

Materials and Methods

Detailed information on the fabrication and assembly of the opto-microperfusion probes, the design and characterization of the time-sequential collection chips, electrochemical monitoring, recovery-rate measurements, and HPLC-EC and LC-MS/MS analyses are provided in the SI Appendix. Detailed descriptions of the experimental animals, stereotaxic surgeries, viral injections, in vivo microsampling procedures, histological analyses, and fluorescence imaging are also provided in the SI Appendix. All experimental procedures were approved by the Institutional Animal Care and Use Committee of the University of Connecticut and complied with the guidelines of the National Institutes of Health.

Statistical analysis

All summary data are expressed as mean ± SEM. Statistical significance was taken as *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, as determined by two-tailed Student’s t-test (paired and unpaired). Statistical analyses were performed in GraphPad Prism 9.

Significance Statement.

Understanding the release dynamics of multiple neurochemicals simultaneously is essential for linking molecular signaling to brain function and behavior, yet existing methods lack the required temporal resolution or multiplexing capability. Here, we present an opto-microperfusion neural probe that integrates time-sequential microfluidics, cell-type–specific optogenetic neuromodulation, and push–pull microsampling into a single platform. This system enables the detection of diverse neurochemicals during optogenetic manipulation with a high temporal resolution. Studies in rodents demonstrated simultaneous monitoring of the release dynamics of up to 23 different neurochemicals following optogenetic stimulation when coupled with mass spectrometry. Together, our findings lay the foundation for the development of next-generation multimodal neural probes to study the molecular processes involved in behavior and neurological disorders.

Data availability

All study data are included in the article and/or SI Appendix.