Introduction

Parkinson’s disease is the fastest-growing neurological disorder globally, with over 10 million people affected1. Despite the focus on dopaminergic neurons, Parkinson’s disease is more accurately described as a multisystem disorder that features a profound albeit underappreciated loss of locus coeruleus noradrenergic (LC/NE) neurons in brainstem, as well as variable damage to other brain regions (the cholinergic neurons in nucleus basalis of Meynert, the serotonergic neurons in raphe nuclei, etc.)2,3,4,5. Postmortem studies have demonstrated that neuronal degeneration in LC/NE is comparable to that in substantia nigra pars compacta, and it may actually precede dopaminergic lesion and partially exacerbate dopaminergic loss in Parkinson’s disease6,7. Hence, it is of great importance to investigate the involvement of LC/NE neurons in Parkinson’s disease. Currently, the clinical management of Parkinson’s disease mainly relies on pharmacological interventions, complemented by other adjunctive therapeutic modalities8. However, the presence of blood-brain barrier (BBB) with high selectivity and low permeability restricts delivery of many pharmaceuticals to neuronal lesion area9,10,11,12,13. Regrettably, the BBB excludes nearly 100% of large-molecule neurotherapeutics and over 98% of all small-molecule drugs from entering the brain14. Consequently, during drug development, the majority of drug candidates fail to achieve sufficient concentrations with target cells, resulting in unacceptable therapeutic efficacy. For instance, topotecan for glioma had reached clinical trials, but failed due to poor penetration across the BBB15.

Traditional pharmacokinetics (PK) analyses, which are primarily based on plasma drug concentrations, often fall short in fully elucidating the pharmacological effect of drugs on specific tissues (such as brain, tumors, placenta), giving rise to frequent occurrence of pharmacokinetics-pharmacodynamics (PK-PD) non-correlation16, particularly for drugs targeting compartments protected by physiological barriers like the BBB. To address this disconnect, Wang et al.16, pioneeringly put forward the theory of micro pharmacokinetics (micro PK), advocating for a PK research shift from the “macro” level of plasma drug concentration to the “micro” level of target cells and subcellular compartments. Vauquelin et al.17, proposed that conventional PK profile and the target dissociation kinetics of a drug may fail to account for its long-lasting efficiency in vivo, and this lacuna can potentially be alleviated by incorporating micro pharmacodynamic and pharmacokinetic mechanisms. By comprehensively investigating the disposition of drugs within cells, this approach can effectively screen and evaluate novel targeted drugs, increase the success rate of new drug and formulation development, and elucidate the interactions and mechanisms of drugs within cells. Therefore, developing a high-fidelity model that precisely recapitulates the biology, physiology, immunology and anatomy of the human BBB is of paramount importance for micro PK-PD study toward neurotherapeutic drug screening and development.

Currently, experimental models, such as molecular docking, conventional two-dimensional (2D) cell, animal models and microfluidic organ-on-a-chip have significantly advanced drug screening and development18,19,20,21,22,23. Thereinto, microfluidic organ-on-a-chip has emerged as a revolutionary tool, distinguished by several distinct merits. Firstly, microfluidic platforms provide physiological relevance by closely recapitulating the complex microenvironments and tissue architectures24,25. For instance, a hypoxia-enhanced BBB organ-on-a-chip26 and BBB organ-on-a-chip integrating brain endothelial cells, pericytes and a three-dimensional (3D) astrocytic network27 have enabled more accurate investigations of drug distribution and efficacy in the brain. Secondly, the versatility and customization of microfluidic devices allow for recreating the pathologically relevant model, which is vital for translation applications28,29. Ingber et al.30, predicted human PK parameters to drugs via fluidically coupled organ chips that were highly similar to those obtained in human clinical studies. Finally, microfluidic organ-on-a-chip systems are amenable to integration with advanced analytical methods31,32,33, enabling real-time monitoring of cellular responses and drug interactions, facilitating dynamic assessments of micro PK-PD at the cellular level, and further advancing understanding the temporal profiles of drug effects.

In micro PK studies, drug concentration monitoring methods typically include spectrophotometry, colorimetry, liquid chromatography-mass spectrometry (LC-MS), immunoassays and electrochemical sensing technology34,35,36,37. Among these, LC-MS is widely employed due to its superior sensitivity, specificity and capability to analyze complex biological matrices38,39. In parallel, electrochemical sensing technology are gaining traction for micro PK applications owing to their portability, high sensitivity, rapid response, low cost and facile integration with microfluidic platforms, making it suitable for in situ, real-time and online monitoring drug concentrations in dynamic cellular microenvironments40,41. In micro pharmacodynamics (micro PD) studies, however, the challenge lies in the fact that most drug effects cannot be directly or continuously quantified in vivo. Therefore, it is necessary to utilize therapeutic targets and disease-relevant biomarkers for efficacy assessment. Of note, in the context of neurodegenerative diseases, the unique neurotransmitter secretion function of neuronal cells can serve as an indicator for monitoring cell viability, that is, drug efficacy indicator42,43. Such functional biomarkers provide a dynamic and quantifiable readout for assessing efficacy at the cellular level, enabling more precise micro PD monitoring in anti-Parkinson’s disease drug development.

In this work, we addressed two critical challenges in the drug screening process of Parkinson’s disease: the lack of PK-PD correlation and limited physiological relevance of existing models. To this end, we developed a highly biomimetic microfluidic BBB-brain organ-on-a-chip coupled with electrochemical sensing technology, combining with LC-MS, for micro PK-PD monitoring within LC/NE neuronal lesion microenvironment. As a proof of concept, pramipexole (PPX) was selected as the therapeutic drug owing to its potential regulation in LC/NE neurons44,45,46,47 and dysfunctional BBB48,49. As shown in Scheme 1, firstly, a detachable microfluidic chip combined with a microporous polyethylene terephthalate (PET) membrane was employed as the cell culture platform. The membrane was coated with collagen type I (ColⅠ) and fibronectin (FN) to form a basement membrane. Human cerebral microvascular endothelial (hCMEC/D3) cells were cultured on one side of the modified membrane to construct a BBB organ-on-a-chip, and its physical, metabolic and physiological barrier functions were evaluated. Human neuroblastoma SH-SY5Y cells were used as tool cells and seeded on the other side of the modified membrane to construct a BBB-brain organ-on-a-chip. LC/NE neuronal injury model of early-stage Parkinson’s disease was established using 1-methyl-4-phenylpyridinium iodide (MPP+). Subsequently, PPX concentrations and the secretion levels of neurotransmitter norepinephrine (NE) from SH-SY5Y cells as a pharmacodynamic indicator were respectively quantified using LC-MS for micro PK–PD study. In parallel, an electrochemical sensor based on Nafion-coated nitrogen-doped carbon quantum dots/overoxidized polypyrrole (N-CQDs/oPPy) modified screen-printed carbon electrode (SPCE), was developed (Nafion/N-CQDs/oPPy/SPCE). The sensor could be regenerated via electrochemical cleaning and integrated into the microfluidic chip for in situ, real-time and online monitoring of PPX. By coupling this electrochemical detection with LC-MS, the feasibility of micro PK-PD monitoring of PPX was validated within a biomimetic disease model (Scheme 2), providing valuable scientific guidance for drug screening and mechanistic studies of anti-Parkinsonian drugs.

Scheme 1
scheme 1

Schematic illustration of microfluidic BBB-brain organ-on-a-chip for micro PK-PD monitoring of anti-Parkinson’s disease drug PPX

Full size image
Scheme 2
scheme 2

Schematic illustration of microfluidic BBB-brain organ-on-a-chip combined with LC-MS and coupled electrochemical sensing technology

Full size image

Results and discussion

Reconstitution of microfluidic BBB organ-on-a-chip

Optimization of culture conditions

As shown in Fig. 1a, b and Supplementary Fig. S1, our microfluidic BBB organ-on-a-chip employed a detachable microfluidic chip, featuring an upper brain chamber and a parallel vascular chamber separated by a microporous PET membrane (pore size of 0.45 μm, pore density of 5 × 107/cm2 and thickness of 11 μm). In Supplementary Fig. S4, hCMEC/D3 cells cultured on FN/ColⅠ/PET membrane displayed a typical cobblestone-like morphology, uniform arrangement and tight connections, which are conducive to forming a structurally intact physical barrier. To improve the controllability, standardization and reproducibility of in vitro BBB model, we systematically investigated the culture time and cell density of hCMEC/D3. In Supplementary Fig. S5, a seeding density of 7.5 × 105 cells and a culture time of 3 h yielded a compact monolayer barrier with minimal apparent permeability (Papp) values, and thus chosen as the optimal conditions for BBB reconstitution in a microfluidic platform.

Fig. 1: Fabrication and functional verification of microfluidic BBB organ-on-a-chip.
figure 1

a Schematic illustration of the detachable microfluidic chip. b Photograph of microfluidic BBB-brain organ-on-a-chip. c Immunofluorescence micrographs of hCMEC/D3 cells labeled with ZO1 and occludin after dynamic culture on FN/ColI/PET membrane. d Immunofluorescence micrographs of SH-SY5Y cells labeled with TH after dynamic culture on FN/ColI/PET membrane. e Papp of 4 kDa and 40 kDa FITC-dextrans in microfluidic chips. The statistical analysis was Two-tailed Student’s t test. Compared with static chip group, ***P < 0.001, ns > 0.05, n = 3. f Papp of Rho123 and DiOC2 in static (upper) and dynamic (bottom) cultures. The statistical analysis was one-way ANOVA. Compared with Rho123 group, ##P < 0.01, #P < 0.05, ns > 0.05, n = 3. Compared with DiOC2 group, ***P < 0.001, ns > 0.05, n = 3. g Changes of Papp before and after mannitol hypertonic solution. The statistical analysis was one-way ANOVA. Compared with Papp at 3 h, ***P < 0.001, **P < 0.01, *P < 0.05, n = 8

Full size image

Functional verification of microfluidic BBB organ-on-a-chip

The functions of microfluidic BBB organ-on-a-chip were evaluated in terms of morphology, physical barrier, metabolic barrier and physiological function. Morphological characterization under static conditions revealed that hCMEC/D3 cells exhibited moderate expression levels of zonula occludens-1 (ZO1) and occludin, along with a clear cytoskeletal network. However, intercellular gaps were observed (Supplementary Fig. S6), indicating suboptimal barrier formation. In contrast, under dynamic cultivation (Fig. 1c), hCMEC/D3 cells were evenly adhered with tight cell-cell connections, and accompanied with enhanced expression levels of ZO1 and occludin, indicating the shear stress was responsible for inducing endothelial function with the barrier tightness50. Similarly, SH-SY5Y cells uniformly adhered and expressed tyrosine hydroxylase (TH) at a moderate level (Fig. 1d), demonstrating their neurosecretory characteristics51. Physical barrier function was evaluated using Papp of 40 kDa fluorescein isothiocyanate (FITC)-dextran. Papp value was significantly lower under dynamic culture condition compared to static condition (P < 0.001, Fig. 1e), suggesting the shear stress had a positive effect on the physical barrier function. In microfluidic BBB organ-on-a-chip, the trans-endothelial electrical resistance (TEER) value across the membrane between the upper and lower layers with an endothelial monolayer reached 134.0 ± 22.4 Ω·cm2, which was consistent with the report27. For metabolic barrier investigation, efflux transporter activity was assessed using rhodamine 123 (Rho123) and 3,3’-diethyloxacarbocyanine iodide (DiOC2) as substrates. While activity of P-glycoprotein (P-gp), multidrug resistance-associated proteins (MRPs) and breast cancer resistance protein (BCRP) transporters remained minimal in Transwell model (Fig. 1f, upper). However, microfluidic BBB organ-on-a-chip exhibited certain activities of P-gp and MRPs. This was evidenced by incremental Papp of Rho123 and DiOC2 following selective inhibition with verapamil and MK571, respectively (Fig. 1f, lower), indicating active drug transport and metabolism capacity. For physiological function assay (Fig. 1g), BBB reversible opening was realized using hypertonic mannitol solution (485 mOsmol/L) and isotonic solution (280 mOsmol/L)27. The above results manifested the microfluidic BBB organ-on-a-chip successfully recapitulated key characteristics of the native BBB, including morphology, monolayer barrier, efflux transporter expression and reversible osmotic opening, offering a physiologically relevant microenvironment for modeling neurodegenerative conditions.

Construction, intervention and treatment of Parkinson’s disease model

Construction of acute Parkinson’s disease model

MPP+, the active metabolite of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine in vivo, gives rise to progressively pathological changes resemble with those happened in Parkinson’s disease, that is, mitochondrial dysfunction, neurosecretory disorder, oxidative stress and degenerative death of neurons, and thus it is widely used for the construction of in vitro Parkinson’s disease models52,53. SH-SY5Y cells exist in both undifferentiated and differentiated forms54. To determine which form is more suitable for modeling, the cell counting kit-8 (CCK8) assay was used to screen the modeling time and concentration of MPP+ toward undifferentiated and all-trans retinoic acid (ATRA)-differentiated SH-SY5Y cells. ATRA-differentiated SH-SY5Y cells were obtained by culturing undifferentiated SH-SY5Y cells in Dulbecco’s modified Eagle medium/Nutrient mixture F12 (DMEM/F12) medium containing 10 μM ATRA and 2% fetal bovine serum (FBS) for 3 days55. As shown in Fig. 2a and Supplementary Fig. S7A, after 24 h of MPP+ exposure, cell viability was inhibited in a dose-dependent manner. The half maximal inhibitory concentration (IC50) of MPP+ for ATRA-differentiated SH-SY5Y cells (5.77 ± 0.32 mM) was lower than that of undifferentiated SH-SY5Y cells (6.72 ± 0.43 mM, P < 0.05), likely due to decreased cell tolerance associated with low serum culture conditions during differentiation. When the modeling time was extended to 48 h (Supplementary Fig. S7B, C), the IC50 values further decreased. Considering both cell tolerance and modeling time (Supplementary Fig. S7D), a modeling time of 24 h and undifferentiated SH-SY5Y cells (hereinafter referred to as SH-SY5Y cells) were selected.

Fig. 2: Construction, intervention and treatment of Parkinson’s disease model.
figure 2

Cell viability of undifferentiated SH-SY5Y cells damaged by different concentrations of MPP+ for 24 h via CCK8 (a) and LC-MS (b). Cell viability of SH-SY5Y cells after intervention with different concentrations of PPX via CCK8 (c) and LC-MS (d). The statistical analysis was one-way ANOVA. Compared with Model group, ***P < 0.001, **P < 0.01, *P < 0.05, ns > 0.05, n = 3. ROS expression (e) and relative gray values (f) of SH-SY5Y cells after PPX intervention at various modeling concentrations. The statistical analysis was Two-tailed Student’s t test. Compared with the intervention group, ***P < 0.001, ns > 0.05, n = 6. g Cell viability of SH-SY5Y cells after treatment with different concentrations of PPX via LC-MS. h Apoptosis rate assessed by Annexin V-FITC/PI staining. The statistical analysis was one-way ANOVA. Compared with Model group, *P < 0.05, n = 3. Compared with Ctrl group, ##P < 0.01, n = 3. Expression levels of inflammatory cytokines TNF-α (I), IL1β (J) and IL6 (K) detected by ELISA. The statistical analysis was one-way ANOVA. Compared with Model group, *P < 0.05, ns > 0.05, n = 3

Full size image

To further refine the modeling conditions, neurotransmitter secretion was quantified using LC-MS to screen the MPP+ concentration. SH-SY5Y cells were stimulated with 105 mM KHCO3 solution for 3 min (Supplementary Fig. S9), and neurotransmitter exocytosis level was examined. NE was the primarily secretory product (Supplementary Fig. S8G), whereas epinephrine and dopamine were minimally detected (Supplementary Fig. S8H, I), demonstrating SH-SY5Y cells could be used as tool cells to simulate the neurotransmitter exocytosis function of LC/NE neurons. Therefore, the level of NE exocytosis was investigated for micro pharmacodynamics monitoring. Taking the amount of NE secreted by the blank group as 100%, a cell viability curve was plotted based on LC-MS quantification (Fig. 2b). The IC50 value derived from NE exocytosis (3.71 ± 0.36 mM) was significantly different from the result obtained by the CCK8 assay (6.72 ± 0.43 mM, Fig. 2a). This discrepancy might be due to the fact that the CCK8 assay measures the activity of succinate dehydrogenase in cellular mitochondria to reflect metabolic activity, while LC-MS assesses the levels of neurotransmitter exocytosis to reflect changes in cellular biological functions. Notably, this divergence is consistent with our previous findings under 2D and 3D culture conditions43,56,57,58. In addition, the morphology of SH-SY5Y cells with MPP+-induced injury at concentration gradient was characterized. When treated with 6 mM MPP+, most of cell bodies shrank, suspended and even fragmented (Supplementary Fig. S7E), thus this condition was unsuitable for dynamic culture on microfluidic chips.

Intervention of Parkinson’s disease model

CCK8 assay results showed PPX at concentrations ranging from 50 to 400 μM exhibited no significant effect on cell viability (Supplementary Fig. S10). The intervention concentration of PPX was subsequently screened via mitochondrial enzyme activity, nitric oxide level (Supplementary Fig. S11D), neurotransmitter exocytosis capacity, reactive oxygen species (ROS) expression and cell morphology (Supplementary Fig. S11E). For CCK8 assay, in Fig. 2c, Supplementary Fig. S11A, B, the cell viability was significantly improved in 200 μM PPX intervention groups (P < 0.01) under MPP+-induced injury at concentrations of 3, 4 and 5 mM. However, at a modeling concentration of 6 mM MPP+, PPX intervention at any concentration failed to rescue cell viability (Supplementary Fig. S11C, P > 0.05), verifying its irreversible damage to SH-SY5Y cells at this higher level of neurotoxic insult. LC-MS analysis of NE exocytosis (Fig. 2d) revealed that PPX at 150–300 μM significantly increased the levels of NE exocytosis compared with the model group (P < 0.001), with the most pronounced effect observed at 200 μM PPX. PPX, as the dopamine receptor agonist, has been widely used for the symptomatic treatment of Parkinson’s disease. PPX binds to D2 subfamily with high selectivity, especially to D3 receptor, and possesses complete intrinsic activity, working by activating dopamine receptor and facilitating the secretion of dopamine46,56. Dopamine is the precursor for the biosynthesis of NE, and can be further converted into NE through the catalysis of dopamine β-hydroxylase47. Similarly, ROS assays (Fig. 2e, f) showed significantly reduced ROS expression following PPX intervention in the 3–5 mM MPP+ groups (P < 0.001). In the case of 6 mM MPP+ damage, no significant attenuation of ROS generation was observed, and the irreversible injury was consistent with mitochondrial membrane potential results (Supplementary Fig. S12). Based on the aforementioned results, 3 mM MPP+ was chosen as modeling concentration, and 200 μM PPX may be suitable as a therapeutic concentration for its relatively ideal intervention effect.

Treatment of Parkinson’s disease model

Through LC-MS measurements, the levels of NE exocytosis significantly increased after treated with 200 μM PPX compared to the 3 mM MPP+-induced injury group (P < 0.05, Fig. 2g). Cell apoptosis was assessed using the Annexin V-FITC/propidium iodide (PI) method. After damage by 3 mM MPP+, the apoptosis rate rose to 19.95 ± 1.06% (P < 0.05, Fig. 2h), whereas subsequent treatment with 200 μM PPX reduced the apoptosis rate to 13.95 ± 1.63% (P < 0.05). As neuroinflammation is a key pathological feature of Parkinson’s disease, the levels of inflammatory factors were examined using enzyme-linked immunosorbent assay (ELISA) to evaluate neuroinflammatory status after drug treatment. After PPX treatment, the expression of tumor necrosis factor-α (TNF-α) was downregulated (P < 0.05, Fig. 2i), while the levels of interleukin 1β (IL1β) and interleukin 6 (IL6) remained unchanged (P > 0.05, Fig. 2j, k). Given above, 200 μM PPX was selected as the therapeutic concentration for micro PK-PD monitoring.

Fabrication of renewable electrochemical sensor

Polypyrrole (PPy), a conductive polymer with its high conductivity, stable chemical properties and excellent biocompatibility, has been widely employed in biosensing applications59,60. Upon electrochemical over-oxidized, PPy is converted into overoxidized polypyrrole (oPPy), forming a porous membrane with large surface area, selective cation permeability and low background signal61,62. These properties are highly advantageous for the detection of PPX, a positively charged drug at physiological pH. N-CQDs exhibit high electron transfer rate, electrocatalytic activity and adsorption capacity63,64, which synergistically modulate electron density, 3D structure and stability of the PPy matrix. Hence, N-CQDs were used as anionic dopants during the PPy electropolymerization, and further over-oxidized on SPCE, followed by coating with an anti-fouling Nafion film (Fig. 3a). The resulting Nafion/N-CQDs/oPPy/SPCE was developed for the electrochemical drug detection.

Fig. 3: Preparation of renewable electrochemical sensor.
figure 3

a Schematic illustration of the preparation of electrochemical sensor. b TEM and high-resolution TEM (inset) images of N-CQDs. c Particle size distribution of N-CQDs. d XRD pattern of N-CQDs. e SEM images of different modified sensors. f DPV curves for different concentrations of PPX in CCM using microfluidic sensing chip coupled U-disk electrochemical workstation. g Linear relationship for different concentrations of PPX in CCM. Curves a–k: 5, 10, 25, 50, 100, 200, 300, 400, 500, 750 and 1000 μM PPX. Inset: Enlarged image of curves a–c. h Inter-group reproducibility and histogram of current values (inset). i Intra-group reproducibility and histogram of current values (inset). j Schematic illustration of electrochemical cleaning. k Intra-group reproducibility after electrochemical cleaning and histogram of current values (inset)

Full size image

Characterization of N-CQDs

Transmission electron microscopy (TEM) image showed N-CQDs exhibited a spherical shape with good dispersibility (Fig. 3b), and possessed an average size of ca. 3 nm (Fig. 3c) and a lattice spacing of 0.34 nm (inset of Fig. 3b), corresponding to the (002) plane of graphitic carbon65. Zeta potential of N-CQDs was −31.21 ± 0.83 mV (Supplementary Fig. S13A). Fourier transform infrared spectroscopy (FT-IR) and X-ray photoelectron spectroscopy (XPS) were shown in Supplementary Fig. S13. X-ray diffraction (XRD) pattern (Fig. 3d) displayed a distinct peak at around 26°, which might be related to the (002) crystal plane of graphitic carbon (PDF#99-0057). The aforementioned results showed that N-CQDs were successfully prepared and primarily composed of nitrogen-doped graphitic carbon, with the surface rich in nitrogen-containing groups and oxygen-containing functional groups (hydroxyl and carboxyl).

Characterization of electrochemical sensor

Scanning electron microscopy (SEM) was employed to characterize morphology of SPCE before and after modification. In Fig. 3e, bare SPCE exhibited relatively rough and mainly composed of carbon nanoparticles. PPy/SPCE and oPPy/SPCE showed no significant changes. N-CQDs/PPy composite film was uniformly deposited in the form of clustered particles on the electrode surface (N-CQDs/PPy/SPCE). Upon further electrochemical over-oxidation, the clustered particles became denser and more uniform (N-CQDs/oPPy/SPCE). After coating with Nafion, a film-like structure appeared, confirming the successful preparation of Nafion/N-CQDs/oPPy/SPCE. Additionally, the wettability, electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV) results were displayed in Supplementary Fig. S14. Optimization of conditions and mechanism of electrochemical detection of PPX were shown in Supplementary Figs. S15, 16.

Analytical performance

The analytical performance of Nafion/N-CQDs/oPPy/SPCE was shown in Supplementary Fig. S17 and Supplementary Table S1. The relative standard deviation (RSD) of five distinct sensors was as low as 3.53% (Fig. 3h). These results indicated Nafion/N-CQDs/oPPy/SPCE exhibited excellent specificity, accuracy and inter-group reproducibility for the detection of PPX. To achieve in situ, real-time and online electrochemical monitoring of PPX, a detachable microfluidic chip coupled with the electrochemical sensor and a U-disk electrochemical workstation was developed (referred to as electrochemical sensing chip, Supplementary Fig. S18A). To evaluate the analytical performance of this integrated sensing platform, differential pulse voltammetry (DPV) was employed to quantify PPX in the cell culture medium driven by a peristaltic pump (Fig. 3f). As shown in Fig. 3g, the peak current exhibited a linear relationship with PPX concentration in the range of 5–500 μM, with the linear equation of I (μA) = 5.52 × 10−3 CPPX (μM) + 0.14, R2 = 0.983. The aforementioned results revealed it is feasible to conduct in situ, real-time and online quantification of PPX on electrochemical sensing chip.

Renewability of the electrochemical sensor

To investigate the intra-group reproducibility of Nafion/N-CQDs/oPPy/SPCE, the current response of a single sensor was repeatedly measured five times in cell-contacting medium (CCM) containing 50 μM PPX. In Fig. 3i, the RSD was 29.8% and the current response in the fifth measurement was 47.6% of the initial value (IV), indicating signal attenuation due to surface fouling. To improve the intra-group reproducibility, the sensor was washed with blank PBS and regenerated via electrochemical cleaning using chronoamperometry (i-t, −0.4 V) after each detection (Fig. 3j). During the electrochemical detection process, PPX molecules can irreversibly adsorb or deposit on the surface of the overoxidized polypyrrole (oPPy) modified layer, blocking the pores and poisoning the active sites, resulting in signal attenuation. To overcome this issue, a reduction potential of −0.4 V is applied to induce de-doping of the oPPy layer. This process causes the contraction of the polymer chains and the expulsion of ions, which generates a synergistic “mechanical squeezing” and “electrostatic repulsion” effect. As a result, these adsorbed or deposited molecules are effectively scrubbed or expelled from the polymer membrane, thereby cleaning the electrode surface. This regeneration strategy effectively restored sensor performance and significantly improved reproducibility, reducing the RSD to 4.46% across five successive measurements (Fig. 3k). Meanwhile, the electron transfer resistance (Ret) values did not change significantly, and the RSD of five measurements was 6.04% (Supplementary Fig. S18A), indicating its excellent structural integrity of the electrode surface post-regeneration. Further, when integrated within a microfluidic chip, the RSD of sensor was 4.84% across five measurements (Supplementary Fig. S18B). The RSD of multiple time point measurements in cell culture media within 48 h was 5.63% (Supplementary Fig. S18C). These results demonstrated Nafion/N-CQDs/oPPy/SPCE possessed salient renewability after electrochemical cleaning. When integrated into the microfluidic device, the renewable electrochemical sensor enables stable and repeatable drug quantification, supporting longitudinal micro PK monitoring.

Micro PK-PD monitoring of the anti-Parkinson’s disease drug pramipexole

Micro PK-PD monitoring in static Transwell model

A Transwell model, combined with LC-MS and electrochemical sensing technology, was used to investigate the dynamic changes of drug concentration and pharmacological effects on the brain side (i.e., micro PK-PD) after administering 750 μM PPX on the vascular side. Four groups were set up as follows: (1) control (Ctrl)-1: after culture in serum-free mixed medium for 24 h, followed by incubation in complete mixed medium without PPX for 24 h; (2) Ctrl-2: after damage in serum-free mixed medium containing 3 mM MPP+ for 24 h, followed by incubation in complete mixed medium without PPX for 24 h; (3) Ctrl-3: after culture in serum-free mixed medium for 24 h, followed by incubation in complete mixed medium containing PPX for 24 h; (4) experimental (Expl): after damage in serum-free mixed medium containing 3 mM MPP+ for 24 h, followed by incubation in complete mixed medium containing PPX for 24 h.

As shown in Fig. 4b, within 0–24 h, the cell viability decreased in all groups, with significant declines observed in the Ctrl-2 and Expl groups. The moderate decline in the Ctrl-1 and Ctrl-3 groups might be due to basic medium, in which the deficiency of FBS led to the loss of essential nutrients such as growth factors, hormones, vitamins and lipids. In addition to the nutrient deficiencies, the Ctrl-2 and Expl groups used serum-free mixed medium containing 3 mM MPP+, which exacerbated the decline in neurosecretory function. At 24–48 h, cell viability in the Ctrl-1 group significantly improved, although there was not treated with PPX. This recovery might be due to the relatively mild damage in the absence of MPP+ and the nutrient-rich serum-contained medium, promoting the growth and metabolism of neuron-like cells. The cell viability in the Ctrl-2 group remained low, as the nutrient-rich complete mixed medium alone was insufficient to reverse the damaged neurosecretory function induced by MPP+ exposure.

Fig. 4: Micro PK-PD monitoring of the anti-Parkinson's disease drug PPX.
figure 4

a Transwell model combined with LC-MS and electrochemical sensing technology to study micro pharmacokinetics of PPX. b Transwell model combined with LC-MS to study micro pharmacodynamics of PPX. Ctrl-1 group: After incubation with basic medium for 24 h, incubation with complete medium without PPX for 24 h; Ctrl-2 group: After incubation with basic medium containing 3 mM MPP+ for 24 h, incubation with complete medium without PPX for 24 h; Ctrl-3 group: After incubation with basic medium for 24 h, incubation with complete medium containing PPX for 24 h; Expl group: After incubation with basic medium containing 3 mM MPP+ for 24 h, incubation with complete medium containing PPX for 24 h. EC: Detection results of electrochemical sensors. c Photographs of the microfluidic device and the electrochemical sensing chip (magnified view). d Microfluidic organ-on-a-chip combined with LC-MS and electrochemical sensing technology to study micro pharmacokinetics of PPX. e Microfluidic organ-on-a-chip combined with LC-MS to study micro pharmacodynamics of PPX. f DPV curves of micro pharmacokinetics monitoring. Ctrl group: After culture with basic medium containing 3 mM MPP+ for 24 h, incubation with blank complete medium for 24 h; Expl group: After culture with basic medium containing 3 mM MPP+ for 24 h, incubation with complete medium containing PPX for 24 h

Full size image

In the Ctrl-3 group, the brain-side concentration of PPX increased gradually (Fig. 4a), with consistent results obtained by electrochemical sensors (abbreviated as EC) and LC-MS, indicating the high accuracy of the developed sensor. After PPX treatment, cell viability maintained at a high level throughout the 24–48 h interval. In the Expl group, the brain-side concentration of PPX increased rapidly and plateaued at a level higher than in the Ctrl-3 group, because MPP+ exhibited a certain destructive effect on BBB, enhancing PPX permeability. After PPX treatment at around 32 h, although the optimal therapeutic concentration was reached, the cell viability continued to improve beyond this timepoint. This phenomenon may result from the dual effect of a gradually increasing drug concentration and duration of action. The above results manifested the Transwell model combined with LC-MS and the developed electrochemical sensor was feasible for micro PK-PD monitoring of PPX.

Micro PK-PD monitoring on microfluidic BBB-brain organ-on-a-chip

To enabled in situ, real-time and online monitoring of micro PK, a microfluidic BBB-brain organ-on-a-chip was connected with an electrochemical sensing chip (Fig. 4c and Supplementary Fig. S3). This biomimetic platform, in combination with LC-MS, was employed to investigate PPX concentrations and efficacy (micro PK-PD) on the brain side at different time points after administering 750 μM PPX on the vascular side. Two groups were set up as follows: (1) control (Ctrl): after damage in serum-free mixed medium containing 3 mM MPP+ for 24 h, followed by incubation in complete mixed medium without PPX for 24 h; (2) experimental (Expl): after damage in serum-free mixed medium containing 3 mM MPP+ for 24 h, followed by treatment in complete mixed medium containing PPX for 24 h. In Fig. 4e, within 0–24 h, due to the lack of nutrients and the destructive effect of MPP+, the cell viability of both groups significantly decreased, indicating the Parkinson’s disease model was successfully established on microfluidic organ-on-a-chip. Within 24–48 h, the cell viability in the Ctrl group slightly increased and then stabilized due to nutrient replenishment, but viability remained low in the absence of PPX treatment. The Expl group showed progressive recovery of cell viability, correlating with a gradual increase in the brain-side PPX concentration (Fig. 4d). Micro PK monitoring using DPV with the renewable electrochemical sensor coupled with U-disk electrochemical workstation were displayed in Fig. 4f. The electrochemical detection results were highly consistent with LC-MS measurements (Fig. 4d), validating the reliability of the electrochemical sensing chip for quantitative drug monitoring. The optimal therapeutic concentration of PPX was reached at around 37 h, slightly later than the ca. 32 h observed in the Transwell model (Fig. 4a). This delay is attributable to the enhanced barrier function of the BBB in the highly biomimetic microenvironment, leading to a decrease in permeability of PPX. The Papp values of PPX on microfluidic organ-on-a-chip and Transwell model were 7.28 × 10−7 ± 3.53 × 10−7 cm/s and 1.22 × 10−6 ± 3.53 × 10−7 cm/s, respectively. And the cell viability continued to recover beyond the point of optimal therapeutic concentration, a trend that was also observed in the Transwell model. The above results demonstrate that the microfluidic organ-on-a-chip coupled with a renewable electrochemical sensor and portable electrochemical workstation offers a powerful platform for in situ, real-time and online monitoring of micro PK. When coupled with LC-MS, it further enables micro pharmacodynamics monitoring, establishing a robust micro PK-PD monitoring strategy for clinical drugs against Parkinson’s disease. This approach hold promises for preclinical drug screening and action mechanism research.

Evaluation of the effects of Parkinson’s disease on BBB disruption

The BBB strictly regulates the selective transport of substances, which is crucial for maintaining the homeostasis of the intracerebral environment66. During the progression of Parkinson’s disease, the structural and functional integrity of the BBB may become compromised, allowing neurotoxins, pathogens or inflammatory factors to enter the brain, thereby exacerbating neurodegeneration67,68. In Fig. 5a, b, compared with the Ctrl group, the expression of ZO1 was markedly downregulated in the Model group (P < 0.001), while Papp of 4 kDa FITC-dextran significantly increased (P < 0.001, Fig. 5c). However, Papp of 40 kDa FITC-dextran remained unchanged (P > 0.05). The results collectively indicated the barrier function of the BBB was reduced and its permeability was slightly increased in Parkinson’s disease state. After treatment with PPX, the expression of ZO1 was upregulated (P < 0.001), Papp of 4 kDa FITC-dextran was correspondingly reduced (P < 0.001). Although PPX is not known to directly repair or enhance the barrier function of BBB, it may indirectly protect or maintain BBB function by reducing oxidative stress damage (Fig. 2e, f), enhancing mitochondrial function of cells (Fig. 2c and Supplementary Fig. S12) and decreasing the level of inflammatory factor (Fig. 2i).

Fig. 5: Evaluation of the effects of Parkinson's disease on BBB disruption.
figure 5

a Immunofluorescence micrographs and 3D surface plots of hCMEC/D3 cells labeled with ZO1. b Mean fluorescence intensity of immunofluorescence micrographs. c Papp of 4 kDa and 40 kDa FITC-dextrans. Compared with Ctrl group, ***P < 0.001, ns > 0.05. Compared with Model group, ###P < 0.001, ns > 0.05, n = 6 in Model group of 40 kDa FITC-dextran, n = 8 in other groups

Full size image

Conclusion

In this study, a highly biomimetic microfluidic BBB-brain organ-on-a-chip combined with LC-MS and electrochemical sensing technology, was proposed for micro PK-PD monitoring of PPX. Within this organ-on-a-chip system, physiological shear stress positively impacted the physical, metabolic and physiological barrier functions of the in vitro BBB, as evidenced by the high expression of tight junction proteins ZO1 and occludin, low permeability of fluorescent tracers, high expression of efflux transporters P-gp and MRPs, and reversible opening to hypertonic mannitol solutions. Building on this platform, a Parkinson’s disease model was established using MPP+-induced damage, replicating key pathological features of neurodegeneration and BBB impairment, thereby providing a pathophysiologically relevant microenvironment for evaluating drug behavior. Leveraging this disease-on-chip model, we achieved micro PK-PD monitoring of PPX via LC-MS. Furthermore, the exocytosis level of neurotransmitter norepinephrine as a pharmacodynamic readout was accurately quantified via LC-MS, supporting the establishment of a PK-PD correlation within this complex cellular context. The integration of a renewable Nafion/N-CQDs/oPPy/SPCE electrochemical sensor enabled in situ, real-time and online monitoring of PPX, with analytical performance matching that of LC–MS. Together, our work demonstrates a robust organ-on-a-chip platform that not only provides a technically versatile approach for micro PK-PD monitoring but also bridges the translational gap in neurodegenerative drug screening. This integrated strategy offers a universal approach for preclinical anti-Parkinson’s disease drug development and screening models.

Admittedly, there are still some limitations existed in the proposed micro PK-PD study. First, LC-MS for pharmacodynamic monitoring, while highly accurate, is limited by expensive instruments and inability to achieve miniaturization. Currently, Zn single-atom-modified TiO2 substrate69 and indium tin oxide-modified optical fibers bundles70 have been constructed for in situ and real time monitoring of NE dynamics. Meanwhile, aptamer and methylene blue-assisted nanosensor71 and nanozyme film with dicopper-coordinated amino-ligands72 have achieved in situ electrochemical detection of neurosecretory events. Future work may focus on the development of a highly sensitive dual-channel electrochemical sensor to realize in situ micro PK-PD monitoring. Second, the biological characteristics of human neuroblastoma SH-SY5Y cells significantly differ from those of normal neurons, and thus they cannot fully reproduce the pathological features of Parkinson’s disease. Human induced pluripotent stem cell (hiPSC)-derived noradrenergic neurons possess genetic fidelity and the ability to recapitulate human-relevant disease phenotypes73,74. Third, the simplified BBB model, ignoring the surrounding pericytes and astrocytes, may weaken the BBB integrity in vitro to some extent26,27. Our subsequent research aims to establish a higher-fidelity model comprising hCMEC/D3 endothelium interfaced with astrocytes and pericytes, thereby more accurately recapitulating barrier function of the in vivo human BBB. Finally, while this current study merely focused on a clinical drug, future directions will extend this micro PK-PD approach to novel nanodrug candidates with merits of targeting brain lesion areas, high biocompatibility and ease of functionalization. Hence, microfluidic organ-on-a-chip coupled multi-channel electrochemical sensor using hiPSC-derived neurons for micro PK-PD study of novel nanodrug candidates is the direction of our future efforts.

Materials and methods

Configuration of microfluidic chip

The configuration and size of the detachable microfluidic chip were shown in Fig. 1a and Supplementary Fig. S1A. Parts for apical and basal channels were cast in polydimethylsiloxane (PDMS). The hollow channels were 0.5 mm wide and 0.5 mm high. The circular cell culture chamber had a diameter of 1.1 cm and a height of 0.5 mm. A PET membrane was integrated between the PDMS layers, acting as a support for cell culture. In Supplementary Fig. S1B, the PET membrane featured a perpendicular pore size of 0.45 μm, a thickness of 11 μm and a pore density of 5 × 107 pores/cm2.

Reconstitution of microfluidic BBB-brain organ-on-a-chip

FN/ColⅠ/PET membrane was placed in a 6-well plate. A suspension of hCMEC/D3 cells was added to one side of the FN/ColⅠ/PET membrane and incubated in a CO2 incubator for 3 h. After cell adhesion, the FN/ColⅠ/PET membrane was flipped and transferred to another clean well in the same 6-well plate. A suspension of SH-SY5Y cells was added to the other side of the FN/ColⅠ/PET membrane and incubated in a CO2 incubator to allow full cell adhesion. Subsequently, the prepared FN/ColⅠ/PET membrane was integrated into a detachable microfluidic chip, with the SH-SY5Y cell layer oriented facing upwards, and secured with a fixture. The apical and basal channels of the chip were connected to a peristaltic pump via Luer tapers (outer diameter of 6 mm and inner diameter of 4 mm) and soft tubes (inner diameter of 0.8 mm). The mixed medium was flowed through the channels for 24 h with the flow rate of 120 μL/min.

Analytical conditions of LC-MS for neurotransmitters

The amount of neurotransmitter exocytosis in cell samples was analyzed via LC-MS. An Agilent Poroshell 120 SB-C18 column (3.0 × 100 mm, 2.7 µm) with a Poroshell SB C18 guard column (3 × 5 mm, 2.7 µm) was applied for the chromatographic separation. The flow rate was kept constantly at 0.2 mL/min. The column temperature was maintained at 30 °C. The mobile phase consisted of water containing 0.1% (v/v) formic acid (A) and methanol (B). The elution program was binary high-pressure isocratic elution. The system was flushed with 90% methanol, and the samples were injected after equilibrating with 15% B at an injection volume of 2 µL. Mass spectrometry was conducted using a Shimadzu LC-MS/MS 8040 triple quadrupole mass spectrometer equipped with an electrospray ionization (ESI) interface. Multi reaction monitoring (MRM) was utilized for the detection of analytes. The conditions of the ESI source were as follows: heat block temperature of 400 °C, desolvation line (DL) temperature of 250 °C, nebulizing gas (N2) flow of 3 L/min and drying gas flow (N2) of 15 L/min. According to the literature75, the qualitative ion pair of NE was selected as 170 > 152 and 170 > 107, and the quantitative ion pair was 170 > 152. For epinephrine, the qualitative ion pair was 184 > 166 and 184 > 107, and the quantitative ion pair was 184 > 166. For dopamine, the qualitative ion pair was 154 > 119 and 154 > 91.05, and the quantitative ion pair was 154 > 119.

Analytical conditions of LC-MS for PPX

An Agilent Poroshell 120 SB-C18 column (3.0 × 100 mm, 2.7 µm) with a Poroshell SB C18 guard column (3 × 5 mm, 2.7 µm) was applied for the chromatographic separation. The flow rate was kept constantly at 0.3 mL/min. The column temperature was maintained at 30 °C. The mobile phase consisted of water containing 0.1% (v/v) formic acid (A) and methanol (B). The elution program was binary high-pressure isocratic elution. The samples were injected after equilibrating with 90% B at an injection volume of 2 µL. Mass spectrometry was conducted using a Shimadzu LC-MS/MS 8040 triple quadrupole mass spectrometer equipped with an ESI interface. MRM was utilized for the detection of analytes. The conditions of the ESI source were as follows: heat block temperature of 400 °C, DL temperature of 250 °C, nebulizing gas (N2) flow of 3 L/min and drying gas flow (N2) of 15 L/min. According to the literature76, the quantitative ion pair of PPX was selected as 212.1 > 153.0.

Synthesis of N-CQDs

N-CQDs were synthesized via a one-pot microwave-assisted method. Briefly, 2.1 g citric acid was dissolved in 15 mL N,N-dimethylformamide. Subsequently, 1.25 mL hydrazine hydrate was added dropwise to form a gel-like precursor. The precursor was then placed in a microwave reactor (800 W) and reacted for 300 s, and a bulk porous solid was obtained. The solid was dissolved in 15 mL deionized water, and the large black precipitate was removed by centrifugation at 10,000 rpm for 10 min. The obtained supernatant was filtered through 0.45 and 0.22 μm filters, and dialyzed against deionized water using a dialysis bag (molecular weight cut-off 100–500 Da) for 24 h. The purified N-CQDs were then collected and dried under vacuum at 60 °C.

Preparation of electrochemical sensor

The bare SPCE was washed with PBS and dried with nitrogen. 100 μL ultrapure water containing 0.1 M pyrrole and 2 mg/mL N-CQDs was dropped onto the SPCE surface. CV was employed to form a polypyrrole (PPy) film doped with N-CQDs on the SPCE surface (N-CQDs/PPy/SPCE). The parameters for CV polymerization were as follows: a voltage range of 0–0.8 V, a scan rate of 0.05 V/s, 4 scan cycles, a sampling interval of 0.001 V, a quiet time of 2 s and a sensitivity of 10−3 A/V. After polymerization, 100 μL of 0.5 M NaOH was added to the surface of N-CQDs/PPy/SPCE, and CV was used for over-oxidation of PPy (N-CQDs/oPPy/SPCE). The parameters for CV over-oxidation were as follows: a voltage range of 0–1 V, a scan rate of 0.1 V/s, 8 scan cycles, a sampling interval of 0.001 V, a quiet time of 2 s and a sensitivity of 10−4 A/V. After washing with PBS, 2 μL of 0.1% Nafion solution was dropped on the N-CQDs/oPPy/SPCE surface to obtain Nafion/N-CQDs/oPPy/SPCE.

Electrochemical measurements

Electrochemical detection of PPX was performed using square wave voltammetry (SWV) on a CHI660b electrochemical workstation. The SWV measurements were conducted within a voltage range of 0–0.85 V, with an amplitude of 0.025 V, a frequency of 25 Hz, a quiet time of 2 s and a sensitivity of 10−5 A/V. Electrochemical detection of PPX was also carried out using DPV on a 3C66 portable U-disk electrochemical workstation. DPV scanning was performed over a voltage range of 0–1.1 V, with an amplitude of 0.075 V, a pulse width of 0.1 s, a pulse period of 0.2 s and a quiet time of 1 s.

Micro PK-PD monitoring in Transwell model

The static Transwell model combined with electrochemical sensing technology and LC-MS was used to monitor the micro PK-PD of PPX. In Transwell model, 7.5 × 105 hCMEC/D3 cells were seeded into the upper chamber, and 1 × 105 SH-SY5Y cells were seeded into the lower chamber. The cells were cultured with 5% CO2 at 37 °C for 3 h. After that, in the experimental (Expl) group, the medium in the upper chamber was replaced with 500 μL of complete mixed medium (DMEM/F12:ECM = 2:3, v/v), and the lower chamber was replaced with 1 mL of basic mixed medium containing 3 mM MPP+ for 24 h. Three control (Ctrl) groups were set up. Specifically, in the Ctrl-1 and Ctrl-3 groups, the upper chamber was replaced with 500 μL of complete mixed medium, and the lower chamber was replenished with 1 mL of basic mixed medium without MPP+ for 24 h. In the Ctrl-2 group, the upper chamber was replaced with 500 μL of complete mixed medium, and the lower chamber was replenished with 1 mL of basic mixed medium containing 3 mM MPP+ for 24 h. The time point at which the lower chamber was replenished with MPP+-containing or blank basic mixed medium was defined as 0 h. The levels of neurotransmitter exocytosis were examined using LC-MS at 0, 4, 8, 12 and 24 h of incubation. The specific operation was as follows. Firstly, the upper chamber was transferred to a blank well, and the medium from the lower chamber was transferred to a 1.5 mL EP tube. Subsequently, 100 μL of sterile normal saline containing 105 mM KHCO3 was added to the lower chamber and incubated with cells for 3 min. The cell incubation medium was collected into a new EP tube for LC-MS detection. Then, the MPP+-containing or blank basic mixed medium was re-transferred to the lower chamber, and the upper chamber was repositioned into the Transwell plate for continued incubation. After 24 h of modeling, drug treatment with PPX was immediately initiated to monitor its micro PK-PD. In the Expl group, 500 μL of complete mixed medium containing 750 μM PPX was added to the upper chamber, and 1.5 mL of complete mixed medium was added to the lower chamber. In the Ctrl-1 and Ctrl-2 groups, 500 μL of complete mixed medium without PPX was added to the upper chamber, and 1.5 mL of complete mixed medium was added to the lower chamber. In the Ctrl-3 group, 500 μL of complete mixed medium containing 750 μM PPX was added to the upper chamber, and 1.5 mL of complete mixed medium was added to the lower chamber. The levels of neurotransmitter exocytosis and drug concentrations in the lower chamber were examined at 1, 2, 4, 6, 8, 10, 12 and 24 h after PPX incubation, corresponding to time points recorded as 25, 26, 28, 30, 32, 34, 36 and 48 h, respectively. Thereinto, micro pharmacokinetics (drug concentrations) was monitored using electrochemical sensing technology and LC-MS, whereas micro pharmacodynamics (the levels of neurotransmitter exocytosis) was assessed using LC-MS. The specific operation was as follows. Firstly, the upper chamber was transferred to a blank well, and the medium from the lower chamber was transferred to a 1.5 mL EP tube. From this, 50 μL was aliquoted into a new EP tube for drug concentration analysis. Subsequently, 100 μL of sterile normal saline containing 105 mM KHCO3 was added to the lower chamber and incubated with cells for 3 min. The cell incubation medium was collected into another new EP tube for LC-MS detection. Then, the PPX-containing or blank complete mixed medium was re-transferred to the lower chamber, and the upper chamber was placed back into the Transwell plate for further incubation.

Micro PK-PD monitoring in microfluidic BBB-brain organ-on-a-chip

The microfluidic BBB-brain organ-on-a-chip coupled electrochemical sensing technology, and combined with LC-MS, were used to monitor the micro PK-PD of PPX. This system comprised a multi-channel peristaltic pump, two culture medium storage bottles, two bubble-trap chips, two single-channel peristaltic pumps, an electrochemical sensing chip, four Luer tapers and several tubes (inner diameter of 0.8 mm). All components were fully sterilized with UV light prior to assembly and the complete system was connected according to Scheme 2. Photograph of the microfluidic device, microfluidic BBB-brain organ-on-a-chip and electrochemical sensing chip were shown in Supplementary Fig. S3, Figs. 1b and 4c. The FN/ColⅠ/PET membrane (diameter of 25 mm) was placed in a 6-well plate. 7.5 × 105 hCMEC/D3 cells and 1 × 105 SH-SY5Y cells were respectively seeded on opposite sides of the membrane. After cell attachment, the membrane was loaded into the detachable microfluidic chip with the SH-SY5Y cell-seeded side facing upwards and secured with a fixture to form the microfluidic BBB-brain organ-on-a-chip (Fig. 1b). A mixed medium (DMEM/F12:ECM = 2:3, v/v) was flowed through the microfluidic channels at a flow rate of 120 μL/min for 24 h. After dynamic culture for 24 h, in the Expl and Ctrl groups, the inflow tube in the culture medium storage bottle was raised above the liquid surface. The medium in the upper chamber was drained using a multi-channel peristaltic pump, and the basic mixed medium containing 3 mM MPP+ was infused into the upper chamber, while the lower chamber continued to infuse complete mixed medium for 24 h. The time point at which MPP+-containing basic mixed medium was introduced into the upper chamber was defined as 0 h. The levels of neurotransmitter exocytosis were examined at 0, 4, 8, 12, 16, 20 and 24 h of incubation using LC-MS. The specific procedures were as follows. Firstly, the multi-channel peristaltic pump was paused, and the three-way valve between the bubble trap chip and the upper chamber of the chip was rotated to connect the single-channel peristaltic pump to the upper chamber of the chip. The liquid in the upper chamber of the chip was drained using the single-channel peristaltic pump. After the medium in the electrochemical sensing chip was completely drained, the three-way valve at the inlet of the electrochemical sensing chip was closed. Next, 100 μL of sterile normal saline containing 105 mM KHCO3 was pumped into the upper chamber of the chip. After incubation for 3 min, the cell incubation medium was pumped out into a new EP tube using the single-channel peristaltic pump, and stored at −20 °C for LC-MS measurement. Finally, the single-channel peristaltic pump was turned off, and the three-way valves were rotated to reconnect among the bubble trap chip, the upper chamber of the microfluidic BBB-brain chip and the electrochemical sensing chip. The multi-channel peristaltic pump was restarted for continued incubation with 37 °C at 5% CO2. After 24 h of modeling, drug treatment with PPX was immediately initiated to monitor its micro PK-PD. In the Expl group, the complete mixed medium containing 750 μM PPX was infused into the lower channel, and the complete mixed medium was infused into the upper channel for 24 h. In the Ctrl group, both upper and lower channels were perfused with the PPX-free complete mixed medium for 24 h. The time point at which the lower chamber was replenished with PPX-containing or blank complete mixed medium was defined as 24 h. For micro pharmacodynamics monitoring, the levels of neurotransmitter exocytosis were examined at 25, 26, 28, 30, 32, 34, 36, 37, 38, 40, 42, 44, 46 and 48 h of incubation using LC-MS. The specific procedures were as described above. For micro pharmacokinetics monitoring, the concentrations of PPX in the upper channel of the microfluidic BBB-brain chip were examined at 24, 25, 26, 28, 30, 32, 34, 36, 37, 38, 40, 42, 44, 46 and 48 h of incubation using LC-MS. The specific operation was as follows. The three-way valve at the outlet of the upper chamber of the microfluidic BBB-brain chip was rotated to connect the upper chamber to an EP tube. The culture medium was transferred into a new EP tube using the multi-channel peristaltic pump and stored at −20 °C for LC-MS analysis. After that, the three-way valve was rotated to reconnect the upper chamber to the electrochemical sensing chip for continued cultivation with 5% CO2 at 37 °C. Similarly, for micro pharmacokinetics monitoring, the concentrations of PPX in the upper chamber of the chip were determined in situ at 24, 28, 32, 36, 40, 44 and 48 h of incubation using electrochemical sensing technology. The specific operations for the detection and cleaning of renewable electrochemical sensors were as follows. Firstly, the prepared Nafion/N-CQDs/oPPy/SPCE was loaded into a detachable microfluidic chip to construct an electrochemical sensing chip (Fig. 4c). Electrochemical detection was performed using DPV. After the detection, the multi-channel peristaltic pump was paused, and the three-way valve at the inlet of the electrochemical sensing chip was rotated to connect the single-channel peristaltic pump to the electrochemical sensing chip. The medium inside the electrochemical sensing chip was then drained using the single-channel peristaltic pump. And the three-way valve at the outlet of the electrochemical sensing chip was closed. Subsequently, sterile PBS was pumped into the electrochemical sensing chip using the single-channel peristaltic pump, and Nafion/N-CQDs/oPPy/SPCE was cleaned and regenerated by scanning with chronoamperometry (i-t, −0.4 V). Then, the PBS in the electrochemical sensing chip was pumped out, and the single-channel peristaltic pump was turned off. The three-way valve was rotated to reconnect the upper chamber to the electrochemical sensing chip. Finally, the multi-channel peristaltic pump was restarted, and the microfluidic device was continuously cultivated with 5% CO2 at 37 °C.

Statistical analysis

Two-tailed Student’s t test or one-way ANOVA were performed to assess statistical significance between different groups using GraphPad Prism 8. All data were presented as the mean ± standard deviation (S.D.) of at least three independent experiments. The significance level is indicated as *** for P < 0.001, ** for P < 0.01, * for P < 0.05 or ns for P > 0.05.

Reagents, apparatus, methods, screening of PPX administration on vascular side, analytical performance of electrochemical sensor, and mechanism of electrochemical detection of PPX could be referred in Supplementary Information.