Microneedle‐Integrated Device for Transdermal Sampling and Analyses of Targeted Biomarkers

Currently available point‐of‐care systems for body fluid collection exhibit poor integration with sensors. Herein, the design of a disposable device for interstitial fluid (ISF) extraction as well as glucose, lactate, and potassium ion (K+) monitoring is reported on. It is minimally invasive and appropriate for single use, minimizing the risk of infection to the user. This microscale device contains a 3D‐printed cap‐like structure with a four‐by‐four microneedle (MN) array, bioreceptor‐modified carbon fiber (CF)‐sensing surface, and negative pressure convection technology. These features are incorporated within a compact, self‐contained, and manually operated microscale device, which is capable of withdrawing ≈3.0 μL of ISF from the skin. MN arrays applied with an upward driving force may increase the ISF flow rate. Moreover, functionalized CF working electrodes (WE1, WE2, WE3) are shown to selectively detect lactate, glucose, and K+ with high sensitivities of 0.258, 0.549, and 0.657 μA μm−1 cm−2 and low detection limits of 0.01, 0.080, 0.05 μm, respectively. Ex vivo testing on porcine skin is used to detect the ISF levels of the biomarkers. The microscale device can be a replacement for current point‐of‐care diagnostic approaches.

Most MN-based direct ISF measurement systems, including the aforementioned systems, have been restricted to the evaluation of a single biomarker. These continuous glucose monitoring systems also did not exhibit good integration between ISF collection and sensing components. [5,6] Some wearable or on-body sensors developed so to this point have demonstrated the capability for detecting multiple analytes. [8] However, these arrangements usually involve sweat sensors, [10][11][12] complex electronic interfaces, and metabolically unrelated analytes (e.g., ethanol, cholesterol, dopamine, and glucose).
Analytes such as glucose, potassium ion, and lactate are important for diagnosing several chronic medical conditions. Therefore, detecting the levels of these analytes in body fluids can offer improved insight into a given individual's physiological state. [10] Here, we report on the development of a MN array-based point-of-care microscale device for ISF extraction and analyte monitoring. This device provides the analysis of three biomarkers, namely glucose, lactate, and potassium ion levels. The MN arrays utilized the pressure-driven convection technique to withdraw ISF. This facile method enabled the acquisition of a sufficient ISF volume (%3.0 μL) for downstream analyses. The multiplexed microscale device includes a 3D-printed four-by-four circular MN array cap at the one end, a thin cotton pad, asyringelike polydimethylsiloxane (PDMS) mold, a gel electrolyte slab with microchannels, as well as a three-electrode system containing a reference electrode, a counter electrode, and three-carbon fiber (CF)-based working electrodes (WEs). MNs were designed with similar features as the 2D MN arrays previously fabricated by Mukherjee et al.; in this study, a suction-induced negative pressure feature was incorporated in the microscale device. [13] The oblique structure of MNs enabled significant stretching of the epidermal layer at the MN tip to prevent skin from folding in proximity to the MN tip; this attribute enhanced the skin penetration process. The ISF that is released from the skin is absorbed into the thin flat cotton swab () and through fine microchannels in the agar gel with redox mediator; it is then made available to the surface of fiber WEs. CF, being conducting, flexible, exhibiting a high specific surface area, and exhibiting highly selective behavior, acted as an ideal biosensing material.
The CF-based electrodes were coated with siloxane to obtain a flat surface for maximum electrocatalytic efficiency. The integrated system, which contained three fiber electrodes with distinct components, was assembled after modification with appropriate receptors. Owing to the excellent adsorption properties of activated CFs, [14][15][16] they were functionalized with Pd@Au core shell nanoparticle dispersions along with specific catalytic bioreceptors glucose oxidase (GO x ), lactate oxidase (LO x ), and 4-aminobenzo-18-Crown-6-ether (AOCE-6). This integrated portable microscale device enabled the parallel electrochemical detection of three clinically relevant analytes. [15][16][17] The MN-CF electrochemical interface was used to obtain data from ISF and in vitro solutions. The microscale device is made from costeffective CF components; as such, the microscale device can serve as a single-use disposable sensor. In addition, a 2D symmetric model of the electroactive surface was developed using COMSOL Multiphysics 6.0 software to visualize the molecular transport. Appropriate external environment and boundary conditions were used to study the free diffusion of analytes according to analytical equations. [18] 2. Results and Discussion 2.1. Designing MN-Based Biosensor Device Figure 1 and 2 demonstrate the development of the cost-effective, user-friendly, and functional biosensor design. The setup resembles a conventional medical syringe and works under the influence of a pressure gradient for body fluid extraction. The ISF collection technique described in this study relies on the synergistic action of the negative back pressure and the positive pressure applied to the skin; an oblique MN array facilitates ISF acquisition. The captured ISF is absorbed by the device through the capillary action of the thin cotton pad placed beneath the MN cap (shown in Figure 2). From there, the ISF is directed to the electrode surfaces through microchannels created in gel electrolyte. The electrodes were functionalized with GO x , LO x , and AOCE-6 for sensing glucose, lactate, and potassium ions. The biorecognition components selectively detect the biomarkers in ISF in a concentration-dependent manner. Figure 2 illustrates the preparation of the MN array. Each MN is conical in shape with a shank height of 900 μm; the diameter of the hollow portion of the MN is 320 mm (Figure 3a,b). The cap consists of an array of 16 MNs in a 4 Â 4 arrangement; the array exhibits a center-to-center distance of 2.5 mm between two MNs and a bore hole with a base diameter of 800 μm (Figure 3c). Scanning electron microscopy (SEM) micrographs (Figure 3ce and S1, Supporting Information) show the characteristic topographic features of the MN array at various angles. Figure 3d,e shows SEM images of individual MN, depicting the sharpness of the MN tip and the oblique design at two magnifications. Figure 3f shows a 3D image of MN that was obtained using laser scanning optical microscopy. Figure 3g shows the detailed dimensional representation of the MNs obtained from keyence study. The needle is place upright on the stage and on exposure to laser the dimensions were recorded as such: the measurement of needle cone started at %500 mm including height, base diameter, tip diameter, and hollow MN diameter; these measurements are 902.5, 941.6, 9.7, and 355.2 μm, respectively. The distance between the hollow part and each side of the MN was also determined. The MN height was determined by rastering the laser in the XY plane and 0.5 nm in Z-direction.

Fabrication and Characterization of MN Array
The DLP 3D printing process provided polymeric MNs with sharp tips. [19] MNs for ISF acquisition 1) must be made of a material that exhibits a high Young's modulus value and 2) must be capable of skin penetration. [20] Therefore, a yellow transparent resin containing a mixture of two methacrylate oligomers, diphenyl (2,4,6-trimethylbenzoyl) phosphine oxide photo initiator, and lauryl methacrylate reactive diluent was used for fabricating the MN arrays. This material allows for the processing of structures with micrometer-scale features and the straightforward removal of unpolymerized material. Shen et al. investigated the role of a cardanol-based acrylate, UV oligomer methacrylate cardan phenolic polycondensate (MCPP), in 3D printing applications. [20] Methacrylate oligomeric resins are used for 3D printing of biomedical devices, as they exhibit high Young's modulus values, high glass transition temperature (T g ) values, low chemical reactivity, heat resistance, hydrophobicity, and straightforward processability. [21][22][23][24] Since these materials are hydrophobic, fewer adhesive interactions occur between the materials and biological molecules.  , GO x , and AOCE-6 to make WE 1 , WE 2, and WE 3 , B) Attachment of CFs with copper foil using silver epoxy resin, and C) preparation of electrolyte gel as well as assembly of three WEs into gel slab and device. Symbols were defined as: CFs, carbon fibers; HCl, hydrochloric acid; HAuCl 4 , tetrachloroauric acid; APTMS, 3-aminopropyl trimethyl siloxane; HCHO, formaldehyde; Pd(acac) 2 The hardness and Young's modulus values of the MN array material were determined using nanoindentation to be 3.29 AE 0.12 GPa (mean AE standard deviation) and 302.19 AE 10.44 MPa, respectively. These values are sufficient for piercing the porcine skin since the minimum Young's modulus value for puncturing human skin is %1 GPa. [24] We demonstrated the skin penetration properties of the MN array with fresh porcine skin; a water-soluble dye, trypan blue, was applied to the MN array-treated skin for visualization of the MN array-generated pores (as shown in Figure 3h-k). The MN arrays were inserted into the skin via manual application. The MN array penetrated the stratum corneum layer of the porcine skin without damage to the MN tips.

Preparation and Characterization of CF WEs and Electrolyte
The device was equipped with an electrochemical sensing system consisting of two parts, CF electrodes and gel electrolyte. To satisfy the requirements of a disposable, cost-effective, and leakproof device, solid agar gel electrolyte and CF-based WEs (WE 1 , WE 2 , WE 3 ) were used to prepare the device. The configuration of each CF electrode (WE 1 , WE 2 , WE 3 ) transducer facilitated the successful detection of biomarkers. The CF WEs were stepwise functionalized for the detection of the analytes. The WEs exhibited a cross-sectional dimension of 5 mm Â 5 mm and a electrochemical active surface area (ECSA) of 0.045 mm 2 . Figure 2 shows the slabs of gel electrolyte with three sets of microchannels arranged at %120°.
Channels were designed to direct the analyte flow to the electrodes. The conductivity and catalytic activity of CF WEs were improved by stepwise functionalization with specific recognition molecules for glucose, lactic acid, and K þ detection. Li et al. suggested that coating with these types of molecules increases the electrode surface area and enhances the mass transport channels, thereby increasing the electrocatalytic ability. [25] In this approach, all of the CF WEs were first smeared with a layer of siloxane ((3-aminopropyl) trimethoxysilane (APTMS), 1 mM) and dried under vacuum to generate a stable amino-terminated surface without compromising their electrocatalytic properties. In addition, bimetallic Pd@AuNPs [26][27][28][29] were in situ synthesized on CFs to improve their electroactivity. On an individual basis, LO x , GO x , and AOCE-6 were immobilized via residual amino groups of APTMS on CF WE 1 , CF WE 2 , and WE 3 , respectively; CF WE 1 , CF WE 2 , and WE 3 were used for successive multiplexed sensing of glucose, lactate, and K þ , respectively. The amino groups of APTMS could potentially combine with the residual carboxylic group at the gamma carbon (CG)/delta carbon (CD) of the acidic amino acids like aspartic (ASP) and glutamic acids (GLU) available on the structure of enzymes (GO x /LO x ) to form an amide bond. The possible reaction can proceed as shown above in Figure S2, Supporting Information, (molecules in yellow indicate ASP groups; the reaction is shown with one of them). Like the GO x and LO x , which display binding selectivity to specific substrates, crown ethers (AOCE-6) display binding selectivity for alkali metals (in this case K þ ).
The AOCE-6 cavity binds selectively with potassium ions, exhibiting ion-dipole interaction and 1:1 stoichiometry. [30] The available electrophilic center of the aldehydic group of glutaraldehyde enabled the attachment of AOCE-6 via the amino group. Figure 4 shows the SEM images of CF WEs at various magnifications. The topologies of the as-spun microfibrous CF/APTMS,  Figure 4(i-iii)) of CF WEs indicate nearly uniform modification, with some irregularities, including a series of grooves of various depths on the surface. [25] The uncoated CF exhibited a smooth texture ( Figure 4a); on the other hand, the modified CFs WEs had a rougher surface ( Figure 4b) along with an uneven appearance due to the presence of particle or layer deposition. Sekar et al. proposed that these surface aberrations could widen the molecule-to-surface interactions and decrease the polarization of the electrode. [31] Figure 4e,f shows the morphology, including the porous texture, of the agar-based semisolid gel slab surfaces. The 3D framework of agar gel was stable and did not swell; agar is known to possess biocompatibility, antifouling properties, antibacterial properties, and electrochemical catalytic properties. [32] The hydrated form of the gel encapsulated with redox species allows for electron transport through the 3D network; respective electrocatalytic conversions are triggered at lower potentials, leading to the effective performance of the device.
It does not solubilize the coatings on CF WEs, it can be stored at room temperature, and it is easily disposable; these attributes are appropriate for biomedical sensing. Figure 4g (1-6) and h shows the corresponding energy-dispersive X-ray spectroscopy (EDX) elemental mapping results. These results indicate the existence of elements C, Si, Cl, O, Au, and Pd, suggesting the dispersion of these elements on the CFs. A nearly uniform distribution of Si, O, Au, and Pd atoms could be observed throughout the examined area. Further, the modifications of CF WEs were assessed by X-ray photoelectron spectroscopy (XPS). As shown in Figure 5, modifications in the C1s spectra were reported after each change in the CFs. [33][34][35][36][37][38] The unmodified fibers show a well-resolved sharp C 1s singlet at %285 eV ( Figure 5a), which may be assigned to carbon-carbon bonded atoms in the hexagonal sheets or the graphitic structure of fibers. [33] For CF/APTMS/Pd@Au/glutaraldehyde/AOCE-6, the C1s spectra were fitted with peaks that were indexed at 284.2, 285.62, and 287.12 eV binding energies. These changes in standard C 1s spectra correspond to -C-N, -C-O, bridging interactions, which are attributed to the presence of amino-functionalized APTMS and AOCE-6 molecules ( Figure 5b).
For APTMS/Pd@Au/LO x or APTMS/Pd@Au/GO x -modified CFs, the C 1s spectra show five signals. Four sharp peaks and one less resolved peak were noted at energies 284.6, 285.9, 287.67, 291.07, and 288.47 eV, respectively ( Figure 5c). These peaks are characteristic of ─C═O and β-carbon bonding in addition to ─C─N and ─C─O─C bonded interactions with enzyme and siloxane molecules. [37] O 1s spectra [39] (Figure S3a  www.advancedsciencenews.com www.small-science-journal.com at a low binding energy (531.7 eV) belongs to Si─O bonding in the siloxane structure; peaks indexed at higher energies are related to ─C═O and enzymatic interactions. A calculated increase in the C/O ratio was detected with stepwise changes to surface topology. After treatment with APTMS, the distributions of C and O were observed as 80.0 atomic wt% and 16.3 atomic wt%, respectively. For CF WE 3 , the O content increased (C/O % = 55.9/24.51); for CF WE 1 /WE 2 , the O content was also found to increase (C/O % = 51.4/46.5). On modification with siloxane-based (APTMS) and noble metals (Pd@Au), the presence of silicon, palladium, and gold content ( Figure S4a-c, Supporting Information) confirmed the loading of the APTMS layer and in situ growth of nanoparticles. Figure S4a, Supporting Information, shows the Au 4f spectra for the growth of monometallic AuNPs with doublet peaks 4f 7/2 and 4f 5/2 as indicated at 87.10 and 90.75 eV, respectively. Peak shifts of %3 eV in lower and 2.25 eV higher binding energy from the standard values (84.1 and 88.5 eV) were attributed to interactions with the environment. While in the bimetallic Pd@Au arrangement, the introduction of Pd led to profound lower shifts in both signals (83.37 and 86.96 eV) ( Figure S3b, Supporting Information). [38] Additionally, the spectrum for palladium shows the doublet with 3d 5/2 and 3d 3/2 peaks at 335.1 and 340.45 eV, respectively ( Figure S4c, Supporting Information). Data from this study is summarized in Table S1, Supporting Information. The Fourier-transform infrared spectra of the CF WEs recorded are shown in Figure 5.  [31,32,[39][40][41][42] -C-H stretching of CH/CH 2 groups from the graphitic backbone of CFs, >C═O groups in the enzyme, -C-N amide bands (II), -C-H bending structures, and -C-N amide bands (III), respectively. Similarly, AOCE-6-modified CF shows peaks at 1220-1180, 2900-2860, 1500-1590 cm À1 , which are attributed to -C-O-C stretching of the crown ether, -C-H stretching of CH/CH 2 groups from the graphitic backbone of CFs, and -C═C stretching, respectively. Along with other CFs, APTMStreated blank CF shows prominent peaks for SiO 2 and -C-Si stretching at 1020 and 1160 cm À1 , respectively. Blank CF sonicated in the presence of mineral acid shows nonspecific broad bands in the functional region around 3020 cm À1 , which is attributed to carbon-carbon bonds in layered CF structures. [42] Shim et al. described the FTIR spectra of CFs on treatment with NaOH/organic acids, which showed the presence of carboxylic acid groups. [42] The spectra with detailed peak details are shown in Figure S5A,B, Supporting Information.

Device Assembly and Operation
The device was designed by integrating all of the components, including MN cap, thin cotton pad, micro channeled gel electrolyte, CF WEs with copper foil connectors, and hollow 1 mL syringe embedded with spring (with 1 mL outer diameter and inner diameter), through a fitted PDMS mold. Figure 2 illustrates the stepwise protocol for building and assembling the different parts of the device. All of these segments are installed within a compact 1 mL syringe-like holder (with 1.8 cm diameter and 3.5 cm height) to create a single-use device. Yu et al. previously www.advancedsciencenews.com www.small-science-journal.com prepared a miniature and disposable trinitrotoluene electrochemical sensor using polymer gel electrolyte. [43] Blicharz et al. reported on the fabrication of single-use point-of-care instrument for capillary blood collection, which employed a vacuum feature. [44] Li et al. demonstrated a PDMS-based hollow MN integrated with a self-powered prevacuum actuator for a blood extraction system. [45] Figure 2 demonstrates the basic arrangement and operation of the device for extraction of ISF through the pressure convection principle. One end of the spring is fixed with a solid circular PDMS mold, which is placed inside the main receptacle. The other end of the spring is attached to the second portion of the plunger-like holder, which can mimic the function of a vacuum suction tool to withdraw fluid. After placing the electrodes at their respective locations in order to form an airtight system, all of the orifices on the MN cap and holder are sealed with superglue. For fluid retraction, the device is pressed against the porcine skin (perpendicular to MN end); the plunger prototype is pulled in the opposite direction. With the expansion of the previously compressed coil, suction (negative pressure) is created inside the device, which is backed by compaction force that is created around the MN insertion site due to the oblique geometry of MNs. The upward drag built in the system draws the ISF into MN array; the skin becomes perforated as shown in Figure 3. The liquid gets absorbed by the cotton swab through capillary action. This dual approach acts as a fluid reservoir and controls the diffusion of the sample on the electrode, thereby preventing saturation. We observed that an optimum volume of ISF was collected and free from blood traces using MNs with a height of %900 μm. The capped MN device typically took %25 min for extraction (yield %3.0 μL) and analysis. The fluid collection was done by the method as described in the above response. It was repeated in three batches, with same sequence of steps (summarized in Table 1), and fluid collection profile over time is shown in Figure 6.  Figure 7 shows a representation of the CV responses on discrete electrodes. The voltammetric signals on inactivated blank CF electrodes showed poorly defined cathodic and anodic peaks, suggesting electrochemical irreversibility. In comparison, the reversible behavior appeared to improve after chemical modifications. The symbolic oxidation and reduction signals at %0.15 and À0.1 V were hierarchically resolved due to the increase in the electroactive surface area (0.0115-0.315 cm 2 ). [16] The corresponding ECSA values were calculated at every stage in terms of double-layer capacitance values using the following relation.

Electrochemical Characterization of CF WEs
The electroactivity of the CF WEs and the gradual improvement in redox electrochemistry of homogeneous Fe [(CN) 6 ] 3À/4À were demonstrated in the following order: CF-Pd@Au-GO x /LO x /AOCE-6 > CF-Pd@Au > blank CF. For CFs WE 1 , WE 2 , and WE 3 , the loading of metal catalyst led to an improvement in the signal symmetry of the redox pair, with an amplification in anodic/cathodic peak current (I pa /I pc ) of %58.9/67.8 μA and voltage separation of 0.248 V (Figure 2b and S6, Supporting Information). The resulting enhancement in current responses and shifting to lower potential validate the loading of self-assembled layers of the active materials.
Further, the presence of the enzyme showed a significant effect on the current responses for CFs (WE 1 , WE 2 ) due to rapid Table 1. Fluid collection using swabs.   electron transfer from the enzyme active centers. [46] An approximate twofold increase in current density was obtained for both GO x -and LO x -modified electrodes (%160.2/166.3 and 210.5/ 270.9 μA, respectively); peak separations of 0.103 and 0.201 V, respectively, were noted. AOCE-6-coated CF (WE 3 ) exhibited substantial enhancements in current responses such as 144.5/ 233.9 μA, respectively, with a potential gap of 0.031 V. Lower onset potentials and high Ipa/Ipc ratios resulted from superior charge transfer kinetics and rapid electrochemical reversibility. The faster kinetics on modified electrodes confirmed the ability to detect the analytes under smaller potential windows. The redox signal was improved with each increase in K þ concentration, as shown by the optimization curve in Figure S6, Supporting Information; gradual peak shifts toward lower potentials were noted. This result is attributed to the action of the positive electrode surface, which encouraged the oxidation/reduction of receptor ion [Fe(CN) 6 ] 3À/4À and facilitated the shuttling of electrons across the immobilized catalyst layer. This finding concurs with the finding previously reported by Kumbhat and Singh on electron transfer kinetics across the 11-MUA monolayer for potassium ion sensing. [47] The improvement in the reversibility of redox voltammetry may be attributed to regulation in proximal diffusion layers across CF WEs. Wightman proposed that if the area of the receptive region on WEs falls intermediate between that of a micro-and macroelectrode, the pattern of diffusion can be either linear or radial. [48] Janisch et al. explained the correlation of diffusion type, electrode size, and sweep speed of the signal. [49] At a constant scan rate (2 mV s À1 ), the peak reversibility improved with the successful stepwise coating of the electrodes. This result shows the transformation of initial hemispherical diffusion to the linear configuration. Thus, the thickness of diffusion layers decreased, which allowed better mass transport. Additionally, gradual switching to the rectilinear form of diffusion from radial (overlapping distribution) could potentially reduce the cross-talk between the electrode system adopted here. [9] Moreover, the precedent model of transport properties was further studied computationally via timedependent COMSOL studies. [50,51]

Amperometric Sensing
To optimize the detection of lactate, glucose, and K þ ion levels with LO x , GO x , and AOCE-6-modified CF WEs (WE 1 , WE 2 , and   Figure 8b. The peak value near the step increased since the concentration of the solution did not reach the equilibrium state instantaneously immediately after adding glucose. [24] For potassium ions, the amperometric signal appeared to strengthen slowly over time (Figure 8c Figure 8a(i),b(i)] were fit to the enzyme kinetics function using the Michaelis-Menten model. This plot determines the conversion rate of lactate or glucose to pyruvate or gluconate, respectively, and H 2 O 2 by enzyme catalysis. Based on this model, the current response to substrate concentration was calculated using the equation In this equation, y i is the reaction rate or current response in this study, [x] is the substrate (lactate/glucose) concentration, V max is the maximum rate obtained, and K m is Michaelis-Menten constant or substrate affinity of the enzyme. [9] The derivative curves obtained for both WE 1 and WE 2 show a steep increase in the rate of reaction with increasing substrate concentration. This result is attributed to the faster turnover rate and www.advancedsciencenews.com www.small-science-journal.com unsaturation of the catalytic site of the enzyme. The corresponding values of V max and constant K m were calculated as 1.2562 AE 0.03 and 1.223 Â 10 À5 for WE 1 and 0.6043 AE 0.07 and 1.708 Â 10 À7 for WE 2 , respectively. For K þ detection, the respective nonlinear plot (Figure 8f ) was fitted to an asymptomatic nonlinear regression model. The estimated lower values of K m for enzyme kinetics were attributed to faster conversion of substrate and the availability of exposed active sites. The steep response of reported biosensors (WE 1 , WE 2 , and WE 3 ) was characterized with high sensitivities of 1.258, 0.549 and 0.657 μA μM À1 cm À2 and low detection limits of 0.01, 0.080, 0.05 μM, respectively (with a probability value more than the F value from statistical analysis). Previous studies on glucose and lactate monitoring involved devices with lower sensitivity. Additionally, the interference studies were performed using 100 μM of uric acid, ascorbic acid, dopamine, and glucose ( Figure S7 Recently, Chen et al. reported a sensitivity of 18.0 μA cm À2 mM À1 over the working range of 0-100 μM for sensing glucose in sweat. [52] Wang et al. estimated the following essential parameters for the MN-based sensor: K m = 8.18 (AE2.30) mM and V max = 6.55 (AE1.17) μA. [9] Lee et al. compared the activity of enzymatic and nonenzymatic glucose sensing using an Au-Ni alloy electrode.
The sensitivity and LOD of the enzymatic and nonenzymatic sensors were observed as 1.30 μA mM À1 and 0.29 μM for the enzymatic sensor as well as 0.96 μA mM À1 and 5.84 μM for the nonenzymatic sensor, respectively. [53] Kim et al. described nonenzymatic sensing of lactate using a NiO and Ni(OH) 2 -based electroactive surface, which showed sensitivities of 9.08 and 35.76 μA (mM cm 2 ) À1 and detection limits of 0.53 and 0.59 mM, respectively [54] ; these values are severalfold lower than those described the present study. A detailed comparison of sensor performance is provided in Table 2. Also, for immobilized systems like these, the apparent K m is calculated in terms of diffusion of substrate. As such, the Michaelis-Menten kinetics is the one of the simplest and best-known models for the nonlinear reaction diffusion process (details are provided in the Experimental Section). The value of apparent K m was calculated as 0.25 and 1.5 mM for LO x and GO x , respectively, using the V max 1.256 mAcm À2 (LO x ) and 0.6043 mAcm À2 (GO x ). In the crosstalk test, in a one-by-one manner, all of the three WEs (WE 1 , WE 2 , WE 3 ) and the separate counter and reference electrodes were connected to potentiostat input. Only one WE was used at a given time. Since the WEs were connected to the virtual ground when they were not in use, no interelectrode crosstalk was observed. It was demonstrated that disconnecting each WE sequentially in this test did not cause any intermittent noise in the readings of the surrounding electrodes ( Figure S8, Supporting Information) since the WEs were connected to the virtual ground when they were not in use.

Ex vivo Performance of the Biosensor
The proof-of-concept CF-based three-in-one biosensor was tested with an ex vivo porcine skin sample. Fresh porcine skin was used www.advancedsciencenews.com www.small-science-journal.com for obtaining real-time biomarker measurements as the content of porcine ISF is analogous to that of ISF in humans. [5,11,13] The porcine skin was shaved and cleaned before the application of the MN-based device for ISF extraction. Figure 9a shows the insertion of the device into the stretched skin. The collected fluid was analyzed by connecting the respective electrodes (WE, CE, and RE). The CV responses from ex vivo studies were compared with those of in vitro samples to understand the concentrations of the biomarkers. Comparative cyclic voltammograms of the CF-Pd @AuNPs/GO x /LO x /AOCE-6 (WE 1 , WE 2 , and WE 3 ) electrodes in the absence and presence of glucose, lactate, and K þ are shown in Figure 9; a dotted curve was used to indicate the modified CF electrodes. The CV curves of both WE 1 and WE 2 showed  www.advancedsciencenews.com www.small-science-journal.com forward oxidative waves, corresponding to the irreversible oxidation of lactate and glucose. [52,53,[55][56][57][58][59] For WE 1 , an increase in the anodic peak current was observed at %0.31 V, indicating the onset of lactate oxidation to pyruvate and H 2 O 2 . (Figure 9c). For WE 2 , an amplification in oxidation current was observed at %0.18 V, marking the onset of glucose oxidation to gluconic acid and H 2 O 2 . These current signals continued to increase further up to an applied potential of 0.60 V. Additionally, there was corresponding increase in cathodic current in the reverse sweep of the CV curve for both WE 1 and WE 2 , which is attributed to the parallel reduction of generated H 2 O 2 in the system at the onset potentials of 0.08 and 0.01 V. Based on this finding, the device allows the oxidation of biomolecules at lower potentials than those reported in previous studies. [60][61][62][63] As shown in Figure 8, the peak current for WE 1 was lower (%370 μA) than the increase in current response (%610.29 μA) for WE 2 due to the low lactate level in ISF.
The low current density for the LO x -modified electrode is attributed to the low concentration of lactate, which correlates with the 0.01 mM lactate signal; the glucose content is similar to a fasting blood sugar level (3.15 mM). The performance of the sensor for detecting K þ levels was also evaluated. The relative responses of augmenting and decaying rates as a function of glucose uptake, glycolysis, and glycogenesis were observed (Figure 9). A high carbohydrate diet needs to be supplemented with potassium enrichment to facilitate glycogenesis and glucose homeostasis. A stimulated current signal of þ290.789/À318.71 μA was recorded on WE 3 , which corresponded to the K þ level. The volume of current obtained for K þ coincides with 6.5 mM (7.5 mEq L À1 ) in vitro, which is higher than the physiological concentration (3-5 mEq L À1 in ISF). This finding coincides with a glucose level of 7.1 mM of its PBS solution, which in turn is considerably higher than standard ISF concentration (3.9-5.6 mM). It is anticipated that this device can be used to better understand the performance of the sensor with food consumption, exercise, and fasting. [58] Also, the effect of height (300-900 μm), of these types of needles on the bleeding from the animal or human skin, has not been addressed in this study and will possibly be investigated from in future in vivo studies. Figure 10a,b shows the 3D and corresponding 2D symmetric work model or confined space representing the cross section of the working electrochemical cell in COMSOL Multiphysics (COMSOL Inc., Stockholm, Sweden). The proposed device operates in several phases; in particular, analyte diffusion on the electrode surfaces is crucial for the sensing process. [17] Therefore, surface diffusivity inside the gel slab was estimated in a time-dependent manner. In this simulation, ISF was diffused from the transdermal area to the inside of the electrolyte slab through capillary action of the cotton pad and then through a single curved channel to the electrode surfaces. [62] Research groups, including Tamba et al. and Karal et al., previously performed experimental and simulation studies to analyze the diffusion of drugs through channels. [63,64] It is important to estimate the transport behavior in a sensing device in a similar context. Figure 10c shows the typical simulation results of the transportation of fluid via flux in the defined space. Before onsetting the molecular transport computation, the contrast of the cell was the same throughout the allotted boundaries, as shown in Figure 10c at 0 s. The movements of diffusion currents were reflected in terms of the contours with undefined margins using a color pattern (0-7), as evident from Figure 10c. With an increase in time from 0.1 to 200 s, the currents from the red color region moved faster toward the demarcated electrode region and the color gradient moved in the opposite direction, showing complete coverage of the electroactive surface by analyte with time. Corresponding experimental evidence from amperometry curves in Figure 8 shows the increase in current with the available concentration of the analytes.  Further, the flux of a substance across a front, J (mol m À2 s), is given by Fick's law relation

Numerical Modeling for Analyte Transport
where P (m s À1 ) is the permeability coefficient of fluorescent probes in the nanopore, C i is the concentration of the analyte entering in a time 't' dependent manner, and D i (m 2 s À1 ) is the diffusion coefficient of the analyte in the cell. Here, we selected a concentration gradient across the cell from outside (MN arraycotton pad) to inside (electrode surface in gel slab). A similar approach was reported by Sharmin et al. for understanding the diffusion of oligoarginines along with a fluorescent probe through a permeable membrane to study the drug delivery rate. [65] The time of diffusion was studied in terms of Einstein's approximation, which is given as where x is the distance covered by the analyte in a nonlinear manner (in this study) following Nernst-Planck-Poisson equations after passing time t via microchannels and capillary action. Later, the analyte is anticipated to be distributed homogenously in the simulated 2D space or corresponding experimental 3D gel slab.

Conclusion
In summary, we report on the fabrication of an integrated device for minimally invasive monitoring of lactate, glucose, and K þ ion levels. Our device includes MN array-supported ISF withdrawal via transient convective force and selective biosensing on flexible CF electrodes that were modified with biorecognition factors. Each component of this first of its kind model device is coordinated to enhance the operation of the device, which provides an option for painless transdermal diagnosis. The MN array cap prevents a movement of air inward once the outlets along the device body are sealed. A low-pressure region developed inside the device, which facilitated skin puncture and ISF flow. The straightforward design and workflow of our device does not require skilled labor. Based on the data presented in this study, we intend to carry out experiments on human participants under different daily activities to assess the performance of the device.
Designing MN Array: All MN arrays (plate-and cap-like structures) were designed using Solidworks 2016 software (Dassault Systèmes, Vélizy-Villacoublay, France) and fabricated using commercial S130 3D printer (Boston Micro Fabrication, Maynard, MA, USA), which utilized digital light projection technology. A biocompatible resin named photoreactive resin BIO made by Boston Micro Fabrication was used for MN fabrication. This yellow transparent resin is a mixture of two methacrylate oligomers, diphenyl (2,4,6-trimethylbenzoyl)  The dimensions (e.g., height, base diameter, tip diameter, and hollow diameter) of the MN were assessed using a VK-X250 3D laser scanning confocal microscope (Keyence, Tokyo, Japan); laser confocal optics were used for field depth measurement. A 408 nm wavelength laser source was used for imaging. There were 16 hollow MN in each 4 Â 4 MN patch, which exhibited a MN interval distance of 2.5 mm. MN patches were disinfected with sterile 70% isopropyl alcohol (VWR, Radnor, PA, USA). To obtain hardness and Young's modulus values from the MN material, nanoindentation was performed. A Ubi-1 Nanoindenter (Hysitron, Minneapolis, MN, USA) with Berkovich-type tip was used for the nanoindentation measurement; a maximum force of 1000 μN, a loading time of 20 s, a dwell time of 10 s at maximum load, and an unloading time of 20 s were used in the study. These steps were repeated for 10 locations on the material. Through Oliver-Pharr analysis of the unloading curves, the hardness and Young's modulus values were obtained. [3,66] Fabrication of Carbon Fiber Working Electrodes: Prior to modification, the pristine CFs were activated as follows. The CFs (0.05 g) were immersed in HCl (1 N) solution and sonicated for 20 min. The CFs were rinsed with double-distilled Millipore (DDM) water, immersed in isopropyl alcohol (3 mL), and cleaned by sonication for 30 min. The fibers were again washed with DDM water and dried in a vacuum oven at 60°C for an hour. The modification of the CFs to give functionalized WEs (APTMS/CF; APTMS/CF/GO x ; APTMS/CF/Pd@Au/GO x ; APTMS/ CF/LO x ; APTMS/ CF/LO x /Pd@Au; APTMS/CF/AOCE -6; APTMS/ CF/AOCE -6/Pd@Au) was performed as follows. The CF bundle was treated with ethanolic solution of APTMS (1.5 M), by dipping for 20-30 min at room temperature. The bundle was flattened on a glass slide using tweezers and dried at 60°C. The CFs obtained after treatment at an elevated temperature appeared as a flat, hard, and malleable material.
For APTMS/CF/GO x or APTMS/CF/LO x , the fabricated APTMS/CF was further treated with 3 μL solution of GO x /LO x , which was prepared by adding lyophilized GO x /LO x to APTMS (1 M, 3 μL) in PBS. The CFs were dipped into this solution for 1 h and then dried at 4°C. The electrode APTMS/CF/ Pd@Au/GO x or APTMS/ CF/ Pd@Au/LO x was prepared by further functionalizing the APTMS/CF system via growing Pd@Au through a sequential pathway. APTMS-treated CFs were dipped in an ethanolic solution of HAuCl 4 (10 mM) for 20 min and dried in an oven. The obtained CFs were soaked in 30% HCHO solution for 5 min and left in air for 10 min. The fibers turned dark reddish in color and were dried under vacuum. These CFs were then immersed in a PVP (1% solution)-treated solution of Pd(acac) 2 (15 mM) in DMF for 1 h and allowed to dry. Modified CFs were treated using EETMS solution (2 M); the reaction was allowed to occur at 40-50°C. The obtained CFs were dried in vacuum. The protocol used for Pd@Au synthesis was reported in our previous studies. [26,29,[67][68][69][70][71] The Pd@Au-functionalized www.advancedsciencenews.com www.small-science-journal.com CFs were dipped in a solution of GO x /LO x [APTMS in PBS (3 μL)] for 60 min and allowed to dry at 4°C. For the APTMS/CF/AOCE-6 electrode, APTMS/ CFs were treated with methanolic solution of AOCE -6 (5 μL, 0.1 M) for 30 min and allowed to dry in an oven. The APTMS/CF/Pd@Au/AOCE-6 electrode, APTMS/CF was treated as described above under APTMS/CF/ Pd@Au/GO x or APTMS/CF/Pd@Au/LO x processing to give APTMS/CF/ Pd@Au and subsequently treated with glutaraldehyde to generate aldehydic group-terminated sites, followed by the steps for the APTMS/CF/AOCE-6 electrode.
Preparation of gel Electrolyte with Microchannels: An agar agar/redox couple/antifouling agent gel electrolyte was prepared using the following procedure. 0.625 g of agar powder in PBS (0.1 M, 19 mL, pH 7.4) at 70°C was prepared by constant stirring and kept for 15 min to completely dissolve the solute, resulting in a clear solution. Subsequently, potassium ferricyanide (0.5 M) was added dropwise into the agar solution and allowed to mix under stirring. Finally, the obtained slurry was poured into a container (with dimensions 50 mm Â 15 mm Â 15 mm in size) containing sections 5 Â 15 Â 15 mm in size, followed by the insertion of 3D-printed microchannel plates in each unit. The slurry was allowed to cool down and solidify at room temperature. The slabs of gel electrolyte with casted microchannels were removed; CF WEs were introduced at the available slots.
Device Assembly: The ISF extraction and biomarker sensing device was designed using the approach of a medical syringe with a hypodermic needle. A 6 cm-long syringe column was utilized and three rectangular orifices were made, which were nearly 10 mm below the top, each oriented 120°w ith respect to one another. The cap-like structure with a MN array at the top was adhered on the syringe column using a cyanoacrylate glue. After the assembly dried, a thin and flat piece of cotton swab (dimensions similar to the inner diameter of the holder with a capillary-like protrusion at the center) was inserted into the holder. The cotton swab ( Figure S9, Supporting Information) was used as reservoir for ISF collection in this study ( Figure 6). Similar sizes of thin cotton flat swabs were taken, and their dry weight (w 0 ) was recorded. Later, they were weighed again after absorption; the weight (w f ) reflects the sum of w 0 and the weight of the fluid (w l ) absorbed (w 0 þ w l ). Therefore, the volume absorbed (V l ) was calculated from the relation: where ρ is the density of fluid absorbed, which is taken as 1 g mL À1 . The experiment was performed with same size swabs and the deviations were calculated (Table 1 and 3). Subsequently, a slab of microchanneled gel electrolyte was placed to prevent reverse expulsion. Later, three replicas of three-electrode configurations (combination of working/auxiliary/reference microelectrode assemblies with specific analyte sensors) were inserted through the orifices and sealed to avoid any air exchange. The assembled syringe-type device was introduced with two additional features not seen in normal syringes: 1) a hook-shaped clip, which is meant to lock the retracted plunger at required position, and 2) a compression spring, in which one end was fixed inside the syringe column (close to the top) while the other end was attached with the head of plunger. This approach helped to enhance the negative pressure convection. Therefore, with the help of hook at the back, the plunger can be held consistently for long periods. The pressure in the device was measured with a digital manometer. The WEs were prepared as follows: 1) functionalization of CFs in the aforementioned manner, 2) dried and trimmed CFs were adhered to copper foil substrates using conductive silver epoxy glue, and 3) the assembly was treated to obtain the required dimensions. Later, 0.5 mm-thick, 1.5 cmlong silver and platinum wires, which served as reference and auxiliary electrodes, along with each of the Wes, were configured to prepare three sets of three-electrode systems. Each electrode set input was individually used to connect with the potentiostat circuit for electrochemical analysis.
Electrochemical Characterization and Performance of Biosensors In vitro: The electrochemical behavior of the modified and unmodified CF electrodes was evaluated using a DropSens SPELEC system (Metrohm AG, Herisau, Switzerland) via the cyclic voltammetric technique within a potential window of À0.3-0.6 V at a scan rate of 2 mV. Further, the sensitivity and selectivity of the WEs for glucose, lactate, and potassium ion sensing were analyzed using an amperometry scan. All of the WEs were incubated for 1000 s at certain fixed potentials to stabilize them before recording the analyte response. For glucose in vitro experiments, responses were recorded after 350 s with a physiological range of 0.075-350 μM using a potential step of À0.2 V over 1000 s. The electrochemical performance of the lactate biosensor toward lactic acid detection was evaluated over 1000 s at a stepping potential of À0.2 V with a concentration range of 0.01-75 μM. For potassium ions, responses were recorded with potassium chloride solutions ranging from 0.01 to 7.5 μM. Additionally, the apparent K m was calculated in terms of diffusion. The general form of nonlinear reaction-diffusion equations is given as This equation describes the density/concentration of substrate fluctuations in a material undergoing reaction-diffusion. A reaction-diffusion equation comprises a reaction term and a diffusion term. Δ 2 denotes the Laplace operator. [72][73][74] As such, the first term on the right-hand side describes the diffusion, including D as diffusion coefficient. The second term, f(u), is a smooth function and describes the processes with real change. The substrate and product concentrations S(t, x) and P(t, x) are functions of two variables, the distance to the biosensor electrode x and time t. The substrate and product concentrations S(t, x) and P(t, x) are functions of two variables, the distance to the biosensor [75,76] electrode x and time t. The values 0 < x < a e correspond to the enzymecontaining layer, which corresponds to the points inside the outer membrane. [77] For 0 < x < a e and t > 0, the dynamics of substrate and product concentrations are described by the following nonlinear reaction-diffusion equations.
where the functions D S (x), D P (x), and α(x) are defined as follows.
D Se and D Pe are the substrate and product diffusion coefficients in the enzyme-containing electrode and diffusion layer. It was assumed that the concentration of substrate in the solution (S 0 ) remains constant during the entire process time. At the beginning of process (t = 0), there is neither substrate nor product inside the enzyme-containing layer and outer surface. As a response of electrochemical biosensor, the steady-state current density (I) is used.
IðtÞ ¼ n e F D Pe ∂P ∂x (11) where ne is the number of electrons involved in the charge transfer at the electrode surface, and F is the Faraday constant. The diffusion module allows one to compare the rate of the enzyme reaction ( Figure 5), V max /K m , with mass transport through the enzyme-containing layer by the relation The value of diffusion module varied with respect to a e and diffusion coefficients, and apparent K m was calculated using a e : 2 mm, D Se : 3.5 μm 2 s À1 , and σ: 0.9782.
Ex vivo ISF Extraction and Testing of Biomarkers in Biofluid: The oblique design of an individual needle in an MN array (shown in Figure 3) combined with pressure-driven drag facilitated the successful withdrawal of ISF from porcine skin. The MN [78,79] -attached suction holder was pressed against the fresh porcine skin for %8-10 min. Similar to a disposable medical syringe (5 mL), a low-pressure area of À74 kPa (measured with digital manometer) was created through the suction setup (shown in Figure S10, Supporting Information). A hook-shaped clip was attached toward the back to lock the retracted plunger so that it could be held consistently for long and serve the purpose. After stopping the outward perpendicular force, clear ISF was drawn out from the pierced holes. Through the capillary action, ISF was absorbed into the cotton patch, transferred to the microchanneled gel electrolyte, and ultimately brought to the WE surfaces. The ISF from the punctured skin segment was quantified as %3.0 μL (per 4 Â 4 set of MNs). The ISF samples obtained from porcine skin were tested for ex vivo electrochemical measurements of glucose, lactate, and potassium ion levels. The CV responses from these fluids were matched with those of in vitro results to determine the approximate concentration of biomarkers present in the fluid.
Geometry and Numerical Assumptions: A microelectrochemical cell working system was considered to explore the diffusion or transport of analytes on the CF WEs. The 3D geometric model in the form of a cylinder with a radius 1 mm and a height of 0.5 mm was made to mimic the experimental cell design with one WE on its surface. For computational evaluations, a simple environment of 2D simulation domain (90°sector) was selected. The dimensions of this 2D frame included the radius (r c ) of the cylinder and a square electrode region with sides r e . As shown in Figure 10B, the electrode area domain was meshed using a free triangular type with an extremely fine element size. While the remaining portion was mapped with an extra fine mesh type to ensure that convergence was reached. In the cell surrounding, analyte transport was governed by capillary action, diffusion, and convection created in the system. The conversion of diffused analytes followed a fast irreversible transformation reaction. The time-dependent mass transport was computed using the equations where C i represents the analyte concentration in the sample (μmol m À3 ), Ji represents the flux of the analyte, R i represents the mass source of analyte, D i represents the compound diffusion coefficient in the sample (10 À9 m 2 s À1 ), and u represents the sample velocity field (m s À1 ).
r e ∶c lim ¼ 0 Equation (13) and (14) are Nernst-Planck-Poisson equations for the nonlinear distribution of the sample. Equation (15)-(17) assume a time-dependent fast irreversible reaction in the region r e 2 (electrode area).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.