In-Situ Monitoring of Glucose Metabolism in Cancer Cell Microenvironments Based on Hollow Fiber Structure

Biosensors and Bioelectronics 162 (2020) 112261

Contents lists available at ScienceDirect

Biosensors and Bioelectronics

journal homepage: http://www.elsevier.com/locate/bios

In-situ monitoring of glucose metabolism in cancer cell microenvironments based on hollow fiber structure

Zhen Ma a, Ying Luo a,d, Qin Zhu a, Min Jiang a, Min Pan c, Tian Xie a,**, Xiaojun Huang b, Dajing Chen a,* a Medical School, Hangzhou Normal University, China b MOE Key Laboratory of Macromolecular Synthesis and Functionalization, Department of Polymer Science and Engineering, Zhejiang University, China c College of Biomedical Engineering & Instrument Science, Zhejiang University, China d School of Traditional Chinese Medicine, Guangdong Pharmaceutical University, China

ARTICLE INFO ABSTRACT

Keywords: Cancer cells alter their metabolism to promote rapid proliferation, resulting in significant amounts of glucose to Hollow fiber be used for aerobic glycolysis in the tumor microenvironment. Due to the spatial mismatch between electrodes Cancer cell and cells, existing methods for evaluating cancer cell metabolism lack kinetic and microenvironmental infor Metabolism microenvironment mation. In this paper, we present a hollow fiber structure loaded with sensing elements as a long-term cellular Glucose sensor metabolism monitoring platform. The unique gradient porous structure allowed the glucose sensor to be close to cultured cells but not directly in contact to prevent adverse effects. The liquid exchange channels in the porous fiber structure ensured continuous and real-time monitoring of glucose concentration changes in the culture media. Experimental results showed high electrochemical sensitivity and stability for continuously monitoring the glucose consumption rate. Furthermore, a continuous three-day long test quantified the change in glucose consumption in response to the anticancer drug Osimertinib. In addition to traditional endpoint cell counting and analysis, this hollow fiber sensing structure provided a real-time monitoring tool for cell metabolism. Such continuous monitoring of the cell metabolism microenvironment improves in-vitro toxicology models for personalized medicine and cancer therapy.

1. Introduction points at the end of cell culture (Butler et al., 2014; Opel et al., 2010). Indirect methods for assessing cell growth include the chemical analysis The fundamental abnormality of cancer is the uncontrolled prolif of cellular metabolic activity or cell culture media, which can be used to eration of cancer cells due to genetic mutations altering cell signaling continuously monitor throughout the culture progress (Bavli et al., pathways (DeBerardinis and Chandel, 2016). Rather than responding 2016; Kieninger et al., 2018; Tharmalingam et al., 2015). appropriately to the growth factors that control normal cell behavior, Glucose concentration change in the cell culture media is an indi cancer cells grow and divide in an uncontrolled manner. Therefore, the cator of cancer cell metabolic activity. Numerous sensors and method energy demand of the rapidly proliferating cancer cells results in higher ologies have been designed to determine glucose concentration in real- glucose consumption compared to normal cells (Fadaka et al., 2017; time including electrochemistry (Chen et al., 2015; Zaidi and Shin, Meadows et al., 2008). Monitoring cell metabolism and cell growth 2016), colorimetry (Xia et al., 2013), conductometry (Mahadeva and could provide insights into applications including drug screening, cell Kim, 2011), and fluorescent spectroscopy (Heo et al., 2011). Among therapy, and toxicity tests (Hanahan and Weinberg, 2011; Pavlova and them, the enzyme-based electrochemical glucose sensor has attracted Thompson, 2016). Direct methods to determine the cell proliferating much attention for the analysis of cell culture media due to its high status include manual counting, automated image analysis, electronic sensitivity and selectivity. In addition, the electrochemical sensor particle counting, in situ biomass monitoring and dielectrophoretic continuously measures the biochemical reaction current, providing cytometry, most of which are accurate but only provide single data real-time data for analysis. In the material aspect, conductive materials

* Corresponding author. D307, No. 1378, West Wenyi Road, Hangzhou, 311121 , China. ** Corresponding author. E-mail addresses: [email protected] (T. Xie), [email protected] (D. Chen). https://doi.org/10.1016/j.bios.2020.112261 Received 8 December 2019; Received in revised form 18 April 2020; Accepted 28 April 2020 Available online 6 May 2020 0956-5663/© 2020 Elsevier B.V. All rights reserved. Z. Ma et al. Biosensors and Bioelectronics 162 (2020) 112261

including metal nanoparticles (Fang et al., 2016), polymers (Li et al., SU-8000 field emission scanning electron microscopy (SEM). Fourier 2015), carbon materials (Mehmeti et al., 2017; Zhu et al., 2012), and transform infrared (FTIR) spectroscopic measurements were performed composite nanostructures (Zhai et al., 2013) have been explored as on Nicolet 380 Fourier transform spectrometer. The cyclic voltammetry enzyme immobilization interfaces exhibiting enhanced sensitivity, (CV) and amperometric response (I-t) were recorded using CHI660 selectivity and stability. In the system structure aspect, electrochemical electrochemical workstation with a three-electrode system. Ag/AgCl (in analysis has been implemented in microfluidic cell culture systems for saturated KCl solution) was used as a reference electrode, platinum wire on-chip analysis (Hiramoto et al., 2019; Ino et al., 2017). A microfluidic as counter electrode and hollow fiber sensor was used as working tumor-on-a-chip can mimic the key features of physiological environ electrode. Fluorescence images of cells on PHF were obtained by ments and be used to better predict drug response (Skardal et al., 2016). Olympus FV3000 confocal microscope. However, electrochemical glucose sensing in the cell culture system has two fundamental limitations: 2.3. Fabrication of PHF structure

1 In order to avoid glucose concentration gradient caused by cell The PHF was prepared by a non-solvent induced phase separation metabolism, sensors need to be placed close to the cells to obtain a method according to our group’s previous work (Zhu et al., 2016). precise measurement in the cell microenvironment (Rodrigues et al., Briefly, 18 wt% PSF, 6 wt% PVP-K17 and 16 wt% PEG-200 in DMAC 2008; Weltin et al., 2014). Sensors located at the downstream outlet solution was extruded from the spinning head to form hollow fiber, or far from cells can only detect general glucose changes in the bulk which were passed through coagulation baths and transferred to the culture media, not accurately reflecting the local concentration coiler for winding. After phase separation, the hollow fiberswere soaked changes in the cell microenvironment. in 15% aqueous glycerol solutions for 24 h. 2 The enzymatic reaction catalyzed by Glucose Oxidase (Gox) will oxidize glucose to hydrogen peroxide (H2O2) and D-glucono- 2.4. Fabrication of PHF-MWNT-PB-Gox biosensor δ-lactone. Therefore, in enzyme-based continuous electrochemical monitoring, H2O2 will accumulate and cause adverse effects on the Sensor electrode fabrication was carried out by layer-by-layer cells (Chang et al., 2003). One solution is increasing the distance deposition on the inner surface of the PHF through dead-end filtra between the sensors that produce H2O2 and the cultured cells, but tion. First, 30 μL of 1 wt% MWNT dispersion in water was injected into this is in contradiction with the above requirement. the hollow fiber by a syringe pump. The MWNT were filtered and trapped by the porous hollow fiber wall and formed a conductive layer Therefore, developing a sensing platform that overcomes the above on the inner surface. Next, 0.02 M of K3Fe(CN)6 and 0.2 M of KCl as a PB issues and enables continuous non-destructive monitoring of the cell precursor in 20 μL HCl solution (pH 1.7) were injected into the PHF and � metabolism microenvironment is still needed. In this paper, we devel dried at 40 C for 15 min. A 20 μL mixture of 0.02 M FeCl3 and 0.2 M KCl oped an enzyme-based Polysulfone hollow fiber(PHF) composite which was injected into the PHF to start the interfacial growth of PB on MWNT. served both as a sensing structure and a cell culture substrate. The cells The MWNT-PB-PHF electrode was further immobilized with the enzyme grow on the outer surface of the PHF, while sensing materials including by injecting 15 μL of Gox solution (1%), followed by 0.5% glutaralde Multi-walled carbon nanotubes (MWNT), Prussian Blue (PB), and hyde. Finally, a gold wire was inserted into the PHF to establish an Glucose Oxidase (Gox) were immobilized inside the hollow fiber. The electrical connection between the MWNT-PB-Gox layer and the mea � sensing layer and the cells, separated by porous structure, are not in surement circuit. The electrodes were stored at 4 C while not in use. direct contact. The porous structure also provides fluid exchange channels, so that the cell microenvironment changes can be detected by 2.5. Electrode performance test the sensor in real-time. The PB layer was used as the electron transport layer to reduce the working voltage and avoid interferences. Addition All in-vitro tests were carried out under ambient temperature using a ally, PB decomposed the hydrogen peroxide generated during the CHI660 electrochemical workstation. I-t curves were recorded in 10 mL enzymatic reaction to lower the adverse effects on the cells. Experi PBS (0.1M, pH ¼ 6.8) under -0.05 V working potential with 4 mM mental results showed a high electrochemical stability that enables long- glucose added every 600 s. Cyclic voltammograms of the PHF-MWNT term monitoring of the effect of anticancer drugs on cell metabolism and PHF-MWNT-PB electrodes in a 0.1M PBS solution were performed activities. This work developed a unique platform for cell metabolism in a potential range of -0.2 V–0.4 V with a 50 mV/s scan rate. Experi analysis and can be further applied in drug testing and compound mental details about the sensor stability test method are provided in the screening. supporting information. For the interference test, 0.1 mM of ascorbic acid (AA), acetaminophen (AP) and uric acid (UA) was added into the 2. Materials and methods PBS sequentially between two steps of 0.1 mM glucose concentration increases, and the current signal was collected for reference. The H2O2 2.1. Reagents and materials concentration in the solution was quantified by a Hydrogen Peroxide Assay Kit (ab102500). Glucose oxides (Gox), Polysulfone (PSF), poly-l-lysine solution (PLL), bovine serum albumin (BSA), glucose, Dimethylacetamide (DMAC), 2.6. Cell culture and real-time monitoring K3Fe(CN)6, KCl and FeCl3 were purchased from Sinopharm Chemical Reagent (Shanghai, China). Multi-walled carbon nanotubes (MWNT) In cell culture experiments, human lung cancer PC9 cells were � dispersion were provided by Xfnano Materials Tech Co., Ltd (Nanjing, cultured in a CO2 incubator at 37 C with high glucose DMEM containing China). Dulbecco’s modified eagle’s media (high glucose), Trypsin- 10% FBS. To investigate the cell adhesion on PHF, PC9 cells on the PHF EDTA (0.25%) and phosphate buffer solution (PBS) were purchased were stained with DAPI fluorescent stain after 72 h culture. PC9 cells from HyClone (Waltham, USA). Fetal bovine serum (FBS) was obtained were fixed in 4% formaldehyde for 20 min, followed by thorough from Gibco (USA). All aqueous solutions were prepared with Milli-Q washing. The hollow fibers were then incubated with DAPI staining water (18.2 MΩ cm, Millipore). solution (0.1% Tween-20 and 1% BSA in PBS) for 30 min and rinsed with PBS extensively. 2.2. Apparatus For continuous metabolism monitoring, PC9 cells were seeded in a PDMS chamber (fabrication procedure in the supporting information) Morphology of the hollow fiber sensor was investigated by Hitachi with a density of 6 � 104/ml and incubated for 24 h for cell adhesion on

2 Z. Ma et al. Biosensors and Bioelectronics 162 (2020) 112261

PHF. The system was kept in the incubator environment to maintain station. The porous structure also creates a selective-permeable barrier normal cell growth. Electrodes in the PDMS chamber were connected to between the cell culture media and sensing layer to further improve the the electrochemical workstation via wires through an incubator access sensing selectivity and stability. port for continuous detection as shown in Fig. S1. After 24 h, Osimertini, as a model anticancer-drug, was injected into the culture chamber. The 3.2. Characterization of PHF electrode current response reflectsglucose concentration change was recorded at an applied potential of -0.05 V through the whole culture procedure. The morphology and structure of the PHF electrodes were examined Continuous data were collected after a 1-h stabilization period. After 72 by SEM. Fig. 2(A) shows the cross-section image of a hollow fiber with h, cells were stained with Propidium iodide (PI) and detected by flow an outer diameter of 520 μm and an inner diameter of 400 μm. A cytometry (FCM) to obtain cell viability. At the beginning and end of the magnified image of hollow fiber wall is shown in Fig. 2(B). The pores experiment, the culture media was sampled and measured by YSI gradually became smaller from the inner surface (>1 μm) to the outer biochemical analyzer to obtain reference concentrations. surface (<100 nm). The outer layer acted as a cell adhesion layer and the porous structure allowed fluid exchange through the interconnected 3. Results and discussion pores. MWNT were used to form a 3D conducting framework at the inner surface by dead-end filtration.As shown in Fig. 2(C), unlike other bulky 3.1. Structure concept and design conductive materials, the MWNT retained a three-dimensional loose structure for fluidexchange. From Fig. 2(D), it can be observed that PB Hollow fiber-basedcell culture is one of the methods for large-scale was uniformly deposited in the MWNT mesh layer without pore block cell culture due to its large surface area to volume ratio (Unger et al., ing. The formation of PB on the MWNT was further demonstrated by FT- 2005). A precisely regulated gradient porous hollow fiberstructure has IR spectroscopy, as shown in Fig. 2(E). The original PHF showed char -1 been developed by our group with applications in enzyme immobiliza acteristic peaks at around 1151 cm , assigned to SO2 symmetrical -1 tion (Guo et al., 2018), microfiltration (Gao et al., 2016), and implant stretching, a strong peak at 1241 cm corresponded to O–C–O stretch -1 able biosensing (Zhou et al., 2019). In this paper, it has been utilized ing, and the peak at 1579 cm was associated with SO2 asymmetric with sensing materials to evaluate cell growth and metabolism proper stretching. In the FTIR spectra of PSF/MWNT/PB, besides similar peaks -1 ties during cell culture. As shown in Fig. 1, the gradient porous hollow to those of PHF, a sharp peak at 2100 cm was observed, which corre fiberprovides a unique asymmetrical structure suitable for carrying cells sponded to the stretching vibrations of Fe–CN in PB. and sensing material on the outer and inner walls of the hollow fiber, respectively. Cells grow on the outer surface of hollow fiber,constantly 3.3. Electrochemical studies consuming glucose in the culture media. The concentration of glucose inside and outside the hollow fiber is balanced due to the diffusion The electrochemical sensing performance was measured by succes through the porous structure. The Gox immobilized inside the hollow sively increasing the glucose concentration in PBS solution. Due to the fiber catalyzes the oxidation of glucose producing gluconic acid and peroxidase activity of PB, H2O2 generated during glucose oxidization H2O2. PB catalyzes the reduction of H2O2 at low voltage and generates can be catalytically decomposed. Fig. 3(A) shows the amperometric electron transfers through MWNT layer to electrochemical working curve of PHF responded to the successive addition of 4 mM of glucose

Fig. 1. Schematic illustration of the PHF electrode structure and sensing principle of the electrochemical reaction catalyzed by Gox and PB.

3 Z. Ma et al. Biosensors and Bioelectronics 162 (2020) 112261

Fig. 2. (A) SEM images of the PHF cross-section. (B)Magnifiedimage of the PHF wall showing a gradient porous structure. (C) MWNT layer deposited on the hollow fiber inner wall. (D) PB synthesized in the MWNT-PB layer. (E) FT-IR of original PHF and MWNT-PB-PHF. from 0 to 32 mM. The currents increased immediately after the addition glucose transmission. PHF achieved a linearity sensing range between 0 of the glucose, promptly reaching a steady state with an average mM and 32 mM (R2 ¼ 0.989) as shown in the insert image in Fig. 3(A). In response time of approximately 60 s. The sensing target of PHF was high order to verify the effectiveness of the PB deposition onto MWNT, cyclic glucose DMEM with initial concentration over 30 mM. Thus, a wide voltammetry was utilized by cycling the modifiedelectrodes from -0.2 V sensing range was required to perform accurate and linear cell meta to 0.4 V. Fig. 3(B) showed the CVs of PHF-MWNT-PB and PHF-MWNT bolism monitoring. According to our group’s previous study (Zhou et al., electrodes in absence and presence of glucose (1 mM) in PBS. 2019), the gradient porous structure could enhance the restriction of Compared with the bare MWNT, the PB-modified electrode showed a

Fig. 3. (A) Current-time response curve of PHF sensor as a function of glucose concentration. (B) Cyclic voltammograms of PHF-MWNT and PHF-MWNT-PB electrodes in absence and presence of glucose. (C) Working stability in different concentrations of glucose. (D) Amperometric responses of the PHF sensor to 0.1 mM of Glucose (Glu), UA, AP, and AA.

4 Z. Ma et al. Biosensors and Bioelectronics 162 (2020) 112261

pair of redox peaks indicating the oxidation and reduction of PB are peroxide content in the PHF sensor chamber was consistently kept below present. Fig. S5 showed the CVs of PHF sensor at different scan rates in 2 μM. This result proves that our proposed sensor structure is more the range of 20–200 mV/s. Square root of scan rate showed a linear suitable for long-term electrochemical sensing in the cell relation with peak current, indicating a diffusion-controlled behavior of microenvironment. the sensor. Current fluctuation interference will affect the accuracy of the analysis, so long-term measurement stability is required for continuous 3.4. Cell culture application test glucose monitoring. The structure of the PHF sensor immobilizes the electrode and enzyme inside the hollow fiber,which is more stable than The cell culture on the outer surface of the PHF was examined other direct contact configurations. The fluid diffuses through the through SEM and confocal microscopy. To verify the cell proliferation porous membrane also minimizing the interference in the complex cell on the surface of PHF, PC9 cells were cultured with PHF for three days culture environment. Fig. 3(C) shows the operational stability of the PHF and viewed with SEM. As shown in Fig. 5(A) and (B), the cells could sensor immersed in DMEM at different glucose concentrations (17.6, adhere to and spread on the PHF surface in a normal morphology, 20.3, 24.0, 27.0, 30.5 mM) for 5 h. After 1 h of the stabilization period, suggesting that the PHF nanoporous surface was a biocompatible sub the sensor current stayed at a constant value and did not show unex strate for cell adhesion and proliferation. Moreover, the cells cultured pected fluctuationsand drifts in the subsequent 4 h test. The influenceof directly on PHF for three days were stained by DAPI and observed by a Osimertinib on electrochemical response were verifiedin PBS buffer. As confocal microscope. Fig. 5(C and D) show the 3D volume rendering shown in Figs. S4 and S6, both experiments showed no interference with fluorescenceimages of the PC9 cells’ distribution along the entire length the electrochemical performance by adding 4 μM Osimertinib. of the PHF surface. The cell fluorescenceimage showed a typical curved For analyzing the glucose metabolism, sensor selectivity is an geometry structure indicating the close cell-PHF adhesion. In Fig. S3, important requirement considering the large number of electroactive fluorescence microscope images of FITC stained cells showed the cell species present in the cell culture environment. Ascorbic acid (AA), density increased throughout the three-day culture, indicating the cell acetaminophen (AP), and uric acid (UA) are by-products during cell viability on the PHF surface. Cell culture morphology characteristics culture, which are also the typical interferences for glucose electro demonstrated that PHF could be used as a proper platform for cell cul chemical sensing. As shown in Fig. 3(D), the addition of interfering ture in the subsequent studies. molecules induced only a small current, which was less than 1% of the current change magnitude caused by glucose addition. The results 3.4.1. Long-term monitoring in cell culture showed reliable selectivity attributed to the catalytic effect in the low To verify the continuous sensing and drug screening functionality, applied potential (0.05 V vs. Ag/AgCl) of PB covered MWNT structure. PC9 cancer cells were cultured on PHF and treated by anticancer drugs. Immobilized enzyme inactivation is a significantproblem that leads First, the response of the sensor in the culture media without cells was to frequent calibrations and shortened sensor lifetime. We compared measured for three days. The PHF sensor showed a stable current output different configurations of enzyme electrodes to optimize the structure during the three-day test. After 72 h of continuous work, the sensor of enzyme, electrode, and porous membrane. Gox was immobilized on current retained 99.7% of the initial value, shown as the red curve in the outer surface of PHF and the surface of a flat Polysulfone porous Fig. 6(A). The glucose concentrations measured by biochemistry membrane for comparison test (detailed preparation procedure in the analyzer at the beginning and the end of the continuous experiment supporting information and Fig. S2). In the storage stability test, all were 30.4 mM and 30.2 mM, respectively, in accordance with PHF � samples were immersed in 30 mM glucose solution at 37 C. The ac measured values. It also indicated that the amount of glucose consumed tivities of the enzyme electrodes were evaluated every day for one week. by the PHF electrode operation was much smaller than the amount Fig. 4(A) shows the enzyme activity in the PHF sensor retained above consumed by the cells, confirming that the continuous electrochemical 94% over one-week, while the other two enzyme immobilizing config monitoring does not act as a nutrient competitor with the cells. urations declined below 80%. The better stability of PHF sensor can be The practical applications of the PHF sensor were demonstrated in a attributed to the porous structure’s interspace confinement effect 72 h continuous anticancer drug screening experiment. Osimertinib, against enzyme inactivation. In an enzyme-based electrochemical re also known as AZD9291, was used as an anticancer drug in this exper action, the generated H2O2 will continuously accumulate in solution, iment. Osimertinib potently inhibits signaling pathways and cellular þ leading to cell damage. Due to the H2O2 decomposition by PB and growth in both EGFRm (Epidermal growth factor receptor) and þ þ encapsulation structure of the hollow fiber, the H2O2 generated in PHF EGFRm /T790M mutant cell lines in vitro, translating into profound can be effectively decomposed. As shown in Fig. 4(B), in the three-day and sustained tumor regression in EGFR-mutant tumors (Cross et al., � continuous working experiment, the sensor without the PB layer accu 2014). PC9 cells were seeded in the PDMS chamber with a density of 6 4 mulated 14.4 μM H2O2 in the cell culture chamber. The hydrogen 10 /mL and incubated for cell adhesion on the PHF. Cell metabolic rates were reflected by glucose concentration change, which was quantified

Fig. 4. (A) Enzyme stability study of Gox immobilized in different configurationswith polymer membranes. (B) H2O2 generated during three-day continuous work.

5 Z. Ma et al. Biosensors and Bioelectronics 162 (2020) 112261

Fig. 5. (A) SEM of cells on PHF; (B) magnified view of cell adhesion on the PHF outer layer; (C) Volume view of fluorescencestained cells on the fibersurface; (D) Magnified view of cells on a single fiber surface. by sensor current data. After the first 24 h, drugs were added to the 3.4.2. Dynamic monitoring in cell culture media to trigger cellular metabolism changes. In Fig. 6(A), as a control Besides long-term monitoring, dynamic response was evaluated by group, only PBS was injected into the chamber so that cells normally adding 1% of TritonX-100 (TX100) to lyse cells in a short time. TX100 is grew over 72 h of culture. In this control group, the PHF sensor current one of the most widely used nonionic surfactants for permeabilizing the dropped 12.5% from its initial value, which is equivalent to the con living cell membrane. A large amount of TX100 will result instant cell sumption of 5.8 mM glucose concentration in the media. The glucose death and release of protein and other cellular organelles. As shown in consumption measured by the biochemistry analysis is 5.65 mM, and the Fig. 6(H), before TX100 addition, cells consumed 0.95 mM glucose error from PHF sensor was within 3%. In the anticancer drug test groups, during first 24 h culture, which showed a similar metabolism rate as a high dose (4000 nM) and a low dose (60 nM) of Osimertinib were other groups. After the addition of 1% TX100, the cells were immedi added to the respective cultured media after the first 24 h. In the low ately lysed, so the glucose metabolism process was aborted and the dose group the sensor current dropped 3.1% in the first24 h, as shown in subsequent sensor currents no longer decreased. However, it can be Fig. 6(B). In the subsequent 48h after drug addition, the current dropped observed in Fig. 6(I) that there was a small current increase after the 2.4% (equivalent to 1.11 mM glucose concentration consumption), addition of TX100, which presumably caused by intracellular glucose which is only 25% of the control group consumption in the same period, release after the cells are lysed. This minor current increase also indi showing the drug inhibition effect on cancer cell proliferation. In Fig. 6 cated that the sensor structure could capture subtle changes in the (C), before high dose drug addition, the current dropped 3.0% in the first cellular metabolic status. As a control group, the cells were suspended in 24 h. In the subsequent 48 h, the current of the PHF sensor in the high the culture media instead of adhering on the PHF surface. This PHF dose group slightly dropped 0.9%, equivalent to 0.45 mM glucose con sensor did not record any glucose increase upon the addition of TX100, centration consumption after drug addition, indicating the strong inhi indicating that only the sensor located close to cells can detect micro bition effect on PC9 cell growth. Fig. 6(G) compared the daily glucose environment concentration changes. The measurement curves in Fig. 6 consumption in different groups. In the first 24 h before drug addition, showed that the system can continuously monitor glucose changes in the all three samples showed similar glucose concentration consumption cell metabolism environment and obtain specific time points of meta around 1.4 mM. After drug addition, both samples exposed to Osi bolic changes. Unlike endpoint analysis, the PHF sensor provided real- mertinib showed significant metabolite changes. The glucose concen time dynamic information about the glucose metabolism in the cell tration consumptions of the two samples on the second and third day culture microenvironment. were 0.79 mM, 0.37 mM and 0.36 mM, 0.09 mM, respectively. Cell proliferation and growth inhibition were further analyzed by flow 4. Conclusion cytometry. After 72 h culture, cells were washed off from the culture chamber and stained with PI. Dead cells displayed a high level of PI In this paper, a long-term cell metabolism monitoring system was staining compared to viable cells. The percentage of viable cells in the built to accurately assess the cell consumption rate of glucose. PHF three samples dropped according to the higher drug exposure sensor was fabricated based on the hollow fiber structure with sensing concentration. materials immobilized in the lumen. The combination of porous struc ture and in-situ deposited sensing layer showed remarkable

6 Z. Ma et al. Biosensors and Bioelectronics 162 (2020) 112261

Fig 6. (A–C) Current response to drugs upon the addition of PBS, low dose of Osimertinib, and high dose of Osimertinib. (D–F) Flow cytometry screening of cells after 72 h culture and drug treatment. (G) Daily glucose consumption chart of different drug usage. (H) Current response to the addition of TX100. (I) Changes in the current details of PHF sensors with or without cell adhesion after TX100 addition. electrochemical sensitivity and stability. The gradient porous structure administration, Writing - review & editing, Resources. of the hollow fiberallowed the glucose sensor to be close to the cultured cells but not in direct contact to prevent adverse effects to cells. Acknowledgements Comparative study of different electrode structures showed that the structure of PHF can provide a more stable enzymatic environment and This work was financially funded by Natural Science Foundation of greater efficiency in the decomposition of H2O2. This PHF sensor Zhejiang Province (Grant No.LY18H180009), National Natural Science addressed some of the challenges of implementing electrochemical Foundation of China (Grant No. 61801160) and Zhejiang Province Key sensor for long-term monitoring in high glucose concentration media. R&D Project (Grant No. 2017C03029). Compared with traditional endpoint analysis, continuous real-time monitoring offered insight into biological information on drug regu Appendix A. Supplementary data lated metabolic rates and cellular metabolism alterations. With further structural development and functionalization, PHF sensor can be used as Supplementary data to this article can be found online at https://doi. a platform technology for real-time monitoring of cellular metabolism org/10.1016/j.bios.2020.112261. and behaviors, achieving broader applications in drug screening and cancer research. References

Declaration of competing interest Bavli, D., Prill, S., Ezra, E., Levy, G., Cohen, M., Vinken, M., Vanfleteren, J., Jaeger, M., Nahmias, Y., 2016. Real-time monitoring of metabolic function in liver-on-chip microdevices tracks the dynamics of mitochondrial dysfunction. Proc. Natl. Acad. The authors declare that they have no known competing financial Sci. Unit. States Am. 113 (16), E2231–E2240. interests or personal relationships that could have appeared to influence Butler, M., Spearman, M., Braasch, K., 2014. Monitoring Cell Growth, Viability, and Apoptosis. Animal Cell Biotechnology. Springer, pp. 169–192. the work reported in this paper. Chang, D.K., Goel, A., Ricciardiello, L., Lee, D.H., Chang, C.L., Carethers, J.M., Boland, C. R., 2003. Effect of H2O2 on cell cycle and survival in DNA mismatch repair-deficient CRediT authorship contribution statement and-proficient cell lines. Canc. Lett. 195 (2), 243–251. Chen, D., Wang, C., Chen, W., Chen, Y., Zhang, J.X., 2015. PVDF-Nafionnanomembranes coated microneedles for in vivo transcutaneous implantable glucose sensing. Zhen Ma: Methodology, Investigation, Data curation, Writing - Biosens. Bioelectron. 74, 1047–1052. original draft. Ying Luo: Investigation. Qin Zhu: Investigation. Min Cross, D.A., Ashton, S.E., Ghiorghiu, S., Eberlein, C., Nebhan, C.A., Spitzler, P.J., Jiang: Investigation. Min Pan: Data curation. Tian Xie: Project Orme, J.P., Finlay, M.R.V., Ward, R.A., Mellor, M.J., 2014. AZD9291, an irreversible EGFR TKI, overcomes T790M-mediated resistance to EGFR inhibitors in lung cancer. administration, Resources. Xiaojun Huang: Validation, Writing - re Canc. Discov. 4 (9), 1046–1061. view & editing. Dajing Chen: Conceptualization, Project

7 Z. Ma et al. Biosensors and Bioelectronics 162 (2020) 112261

DeBerardinis, R.J., Chandel, N.S., 2016. Fundamentals of cancer metabolism. Sci. Adv. 2 Opel, C.F., Li, J., Amanullah, A., 2010. Quantitative modeling of viable cell density, cell (5), e1600200. size, intracellular conductivity, and membrane capacitance in batch and fed-batch Fadaka, A., Ajiboye, B., Ojo, O., Adewale, O., Olayide, I., Emuowhochere, R., 2017. CHO processes using dielectric spectroscopy. Biotechnol. Prog. 26 (4), 1187–1199. Biology of glucose metabolization in cancer cells. J. Oncol. Sci. 3 (2), 45–51. Pavlova, N.N., Thompson, C.B., 2016. The emerging hallmarks of cancer metabolism. Fang, L., Liu, B., Liu, L., Li, Y., Huang, K., Zhang, Q., 2016. Direct electrochemistry of Cell Metabol. 23 (1), 27–47. glucose oxidase immobilized on Au nanoparticles-functionalized 3D hierarchically Rodrigues, N.P., Sakai, Y., Fujii, T., 2008. Cell-based microfluidic biochip for the ZnO nanostructures and its application to bioelectrochemical glucose sensor. Sensor. electrochemical real-time monitoring of glucose and oxygen. Sensor. Actuator. B Actuator. B Chem. 222, 1096–1102. Chem. 132 (2), 608–613. Gao, Q.-L., Fang, F., Chen, C., Zhu, X.-Y., Li, J., Tang, H.-Y., Zhang, Z.-B., Huang, X.-J., Skardal, A., Shupe, T., Atala, A., 2016. Organoid-on-a-chip and body-on-a-chip systems 2016. A facile approach to silica-modifiedpolysulfone microfiltrationmembranes for for drug screening and disease modeling. Drug Discov. Today 21 (9), 1399–1411. oil-in-water emulsion separation. RSC Adv. 6 (47), 41323–41330. Tharmalingam, T., Wu, C.H., Callahan, S., Goudar, T.C., 2015. A framework for real-time Guo, Y., Zhu, X., Fang, F., Hong, X., Wu, H., Chen, D., Huang, X., 2018. Immobilization of glycosylation monitoring (RT-GM) in mammalian cell culture. Biotechnol. Bioeng. enzymes on a phospholipid bionically modifiedpolysulfone gradient-pore membrane 112 (6), 1146–1154. for the enhanced performance of enzymatic membrane bioreactors. Molecules 23 Unger, R., Huang, Q., Peters, K., Protzer, D., Paul, D., Kirkpatrick, C., 2005. Growth of (1), 144. human cells on polyethersulfone (PES) hollow fiber membranes. Biomaterials 26 Hanahan, D., Weinberg, R.A., 2011. Hallmarks of cancer: the next generation. Cell 144 (14), 1877–1884. (5), 646–674. Weltin, A., Slotwinski, K., Kieninger, J., Moser, I., Jobst, G., Wego, M., Ehret, R., Heo, Y.J., Shibata, H., Okitsu, T., Kawanishi, T., Takeuchi, S., 2011. Long-term in vivo Urban, G.A., 2014. Cell culture monitoring for drug screening and cancer research: a glucose monitoring using fluorescent hydrogel fibers. Proc. Natl. Acad. Sci. Unit. transparent, microfluidic, multi-sensor microsystem. Lab Chip 14 (1), 138–146. States Am. 108 (33), 13399–13403. Xia, Y., Ye, J., Tan, K., Wang, J., Yang, G., 2013. Colorimetric visualization of glucose at Hiramoto, K., Ino, K., Nashimoto, Y., Ito, K., Shiku, H., 2019. Electric and the submicromole level in serum by a homogenous silver nanoprism–glucose oxidase electrochemical microfluidic devices for cell analysis. Front. Chem. 7, 396. system. Anal. Chem. 85 (13), 6241–6247. Ino, K., Shiku, H., Matsue, T., 2017. Bioelectrochemical applications of microelectrode Zaidi, S.A., Shin, J.H., 2016. Recent developments in nanostructure based arrays in cell analysis and engineering. Curr. Opin. Electrochem. 5 (1), 146–151. electrochemical glucose sensors. Talanta 149, 30–42. Kieninger, J., Weltin, A., Flamm, H., Urban, G.A., 2018. Microsensor systems for cell Zhai, D., Liu, B., Shi, Y., Pan, L., Wang, Y., Li, W., Zhang, R., Yu, G., 2013. Highly metabolism–from 2D culture to organ-on-chip. Lab Chip 18 (9), 1274–1291. sensitive glucose sensor based on Pt nanoparticle/polyaniline hydrogel Li, L., Wang, Y., Pan, L., Shi, Y., Cheng, W., Shi, Y., Yu, G., 2015. A nanostructured heterostructures. ACS Nano 7 (4), 3540–3546. conductive hydrogels-based biosensor platform for human metabolite detection. Zhou, J., Ma, Z., Hong, X., Wu, H.-M., Ma, S.-Y., Li, Y., Chen, D.-J., Yu, H.-Y., Huang, X.- Nano Lett. 15 (2), 1146–1151. J., 2019. Top-down strategy of implantable biosensor using adaptable, porous Mahadeva, S.K., Kim, J., 2011. Conductometric glucose biosensor made with cellulose hollow fibrous membrane. ACS Sens. 4 (4), 931–937. and tin oxide hybrid nanocomposite. Sensor. Actuator. B Chem. 157 (1), 177–182. Zhu, X.-Y., Chen, C., Chen, P.-C., Gao, Q.-L., Fang, F., Li, J., Huang, X.-J., 2016. High- Meadows, A.L., Kong, B., Berdichevsky, M., Roy, S., Rosiva, R., Blanch, H.W., Clark, D.S., performance enzymatic membrane bioreactor based on a radial gradient of pores in a 2008. Metabolic and morphological differences between rapidly proliferating PSF membrane via facile enzyme immobilization. RSC Adv. 6 (37), 30804–30812. cancerous and normal breast epithelial cells. Biotechnol. Prog. 24 (2), 334–341. Zhu, Z., Garcia-Gancedo, L., Flewitt, A.J., Xie, H., Moussy, F., Milne, W.I., 2012. Mehmeti, E., Stankovi�c, D.M., Chaiyo, S., Zavasnik, J., Zagar,� K., Kalcher, K., 2017. A critical review of glucose biosensors based on carbon nanomaterials: carbon Wiring of glucose oxidase with graphene nanoribbons: an electrochemical third nanotubes and graphene. Sensors 12 (5), 5996–6022. generation glucose biosensor. Microchim. Acta 184 (4), 1127–1134.