SEEDING to Enable Sensitive Electrochemical Detection of Biomarkers in Undiluted Biological Samples

Electrochemical biosensors have shown great potential for simple, fast, and cost‐effective point‐of‐care diagnostic tools. However, direct analysis of complex biological fluids such as plasma has been limited by the loss of sensitivity caused by biofouling. By increasing the surface area, the nanostructured electrode can improve detection sensitivity. However, like a double‐edged sword, a large surface area increases the nonspecific adsorption of contaminating proteins. The use of nanoporous structures may prevent fouling proteins. However, there is no straightforward approach for creating nanostructured and nanoporous surfaces compatible with microfabricated thin‐film electrodes. Herein, the preferential etching of chloride and surfactant‐assisted anisotropic gold reduction to create homogeneous, nanostructured, and nanoporous gold electrodes is demonstrated, yielding a 190 ± 20 times larger surface area within a minute without using templates. This process, “surfactant‐based electrochemical etch‐deposit interplay for nanostructure/nanopore growth” (SEEDING), on electrodes enhances the sensitivity and antibiofouling capabilities of amperometric biosensors, enabling direct analysis of tumor‐derived extracellular vesicles (tEVs) in complex biofluids with a limit of detection of 300 tEVs µL−1 from undiluted plasma and good discrimination between patients with prostate cancer from healthy ones with an area under the curve of 0.91 in urine and 0.90 in plasma samples.


Introduction
Compact and affordable analytical tools that can rapidly perform sensitive, selective, and multiplexed detection of biological markers are essential for disease diagnostics and implantable sensors. Electrochemical readout platforms present an affordable and sensitive approach, along with miniaturized and simple instrumentation, promoting the fabrication and generalization of point-of-care medical devices. However, the application of biosensors in diagnostic platforms still faces limitations when measuring relevant clinical samples, typically comprised of complex biological fluids, such as plasma, because of surface inactivation and sensitivity loss from nonspecific adsorption and accumulation of abundant background proteins.
Sampling and molecular analysis of human biofluids hold great promise for cancer diagnosis, prognosis, and treatment response assessment. [1] Extracellular vesicles (EVs) secreted by all cells are present in large numbers in nearly all body fluids. They contain invaluable protein and genetic information and play an essential role in various physiological and pathological processes. [2] Therefore, EVs are potential biomarker candidates for liquid biopsy analyses; for instance, tumor-derived EVs (tEVs) are critically related to tumor progression, metastatic niche formation, and immune evasion. [3,4] However, EVs still represent a minor subpopulation of all circulating particles in the blood and are outnumbered by lipoproteins by up to six orders of magnitude. [5] Therefore, conventional analysis methods typically require EVs separation, concentration, and purification steps, which significantly affect the analysis time and sample volume required. [6] Studies on the detection of tEVs from plasma or serum have mainly been conducted using optical methods. [7,8] However, the low detection limits required, along with the large concentration of contaminants in clinical samples, impose a remarkable challenge in electrochemical biosensing. [9] An electrochemical biosensor with a low limit of detection and antibiofouling capabilities may address these requirements, providing a highthroughput tool for tEV diagnostics of small-volume liquid biopsies.
The sensitivity of amperometric biosensors can be enhanced by increasing their surface area, [10] thereby lowering their limit of detection [11] and yielding enhanced apparent electron transfer kinetics and improved bioreceptor functionalization yield. [12] However, a large available surface area also increases biofouling and the nonspecific adsorption of contaminating background proteins and molecules, which causes a subsequent decrease in the sensitivity. In this regard, the formation of nanoporous structures may be beneficial for limiting the diffusion of undesired fouling proteins [9,13] although the same principle limits the detection of desired analytes. [14,15] In this study, we present "surfactant-based electrochemical etch-deposit interplay for nanostructure/nanopore growth" (SEEDING), a method to fabricate gold electrodes that combine nanostructured surfaces and nanoporous substrates, achieving the benefits of both strategies, namely, large available electroactive areas as well as limited diffusion and adsorption of nonspecific proteins. As a proof-of-concept, we designed selective biosensors to detect a range of different targets: 1) electroactive molecules like ferrocyanide, 2) soluble proteins like the inflammatory marker interleukin 6 (IL6), 3) membrane proteins like tetraspanins CD9 and CD81 commonly found in EVs, as well as 4) epithelial cell adhesion molecules (EpCAM) used as a diagnostic marker for various cancers. All these markers were detected from undiluted biological samples, such as cell culture media, blood plasma, and urine. The process is a simple, costefficient, and rapid electrochemical method that can be applied to photolithographically patterned microfabricated thin-film electrode array chips for mass production of disposable assay chips that are typically used in portable or handheld devices.

SEEDING Mechanism: Formation of Nanostructured and Nanoporous Surfaces
The application of specific electric potentials to an electrode can transfer electrons between nearby molecules in the solution and the electrode, thereby changing the oxidation state of the species involved. For instance, the surface of the electrode can be electro-oxidized to etch materials and/or electrodeposit soluble materials atop. By repeating this process, we can gradually transform a flat electrode (Figure 1a) into a nanostructured Figure 1. SEEDING of microfabricated thin-film flat gold electrodes. a) Photograph of a microfabricated thin-film gold electrode chip with four circular working gold electrodes, a common reference, and counter electrode. b) Typical chronoamperograms (only showing anodic currents) during the SEEDING process conducted in microfabricated thin-film gold electrodes in different control solutions (one-component controls are greyed out). c) Cyclic voltammogram of a flat (black) and a nanostructured and nanoporous (NSG) electrode (blue) in acidic media. d) Photographs of the microfabricated thin-film gold electrodes before (top) and after (bottom) the SEEDING process. Scale bar: 250 µm. e) Top and f) Cross-sectional scanning electron micrographs of a gold electrode before (top) and after (bottom) the SEEDING process. Scale bar: 1 µm. g) Atomic force microscopy topography of the microfabricated thin-film gold electrodes before (top) and after (bottom) the SEEDING process. Scale bar: 1 µm. www.advmat.de www.advancedsciencenews.com and nanoporous gold (NSG) electrode. Electrochemistry provides a convenient tool for controlling the kinetics and thermodynamics of these reactions. However, it is essential for this etch-deposit process to be anisotropic in order to create NSG electrodes. In other words, the interplay between the etch and deposit processes must be balanced and exhibit spatial preferences. Otherwise, the processes would simply cancel each other, etching gold previously deposited, leading to simple surface roughening. [16] To achieve controlled and directed growth, we designed the SEEDING method. We used a sodium chloride solution for electro-oxidizing and etching gold from a flat electrode and chloroauric acid and cetyltrimethylammonium (CTA + ) surfactant for preferential growth orientation during electroreduction. By quickly repeating these steps, we see that the anodic and cathodic currents, representing the oxidation and reduction of gold, increase over time, reaching a stable current within 1 min (Figure 1b and Figure S1, Supporting Information). The current is also correlated with the available electroactive area, indicating the growth of nanostructures, a process not observed using control solutions containing only one or two of the three components. This process was quite reproducible, yielding a coefficient of variation of the current 8% across 14 different chips tested on different days ( Figure S2, Supporting Information).
Comparing the available electroactive area of the electrodes after the SEEDING, we achieved roughness factors on NSG of 200 ± 40 (n = 8) (Figure 1c), which is in reasonable agreement with the value calculated from the capacitive background current in the voltammogram (170 ± 40, n = 8) ( Figure S3, Supporting Information), and the value measured by electrochemical impedance spectroscopy (190 ± 20, n = 8). The reason for the higher variability in the measurements conducted by voltammetry may be associated with the irreversible formation and dissolution of gold oxides during the measurement, [17] a process that can be especially harmful to fine structures like the ones in our electrodes.
The fabricated NSG electrodes exhibited a distinctive dark homogeneous coloration on the surface (Figure 1d). At the microscopic level, the original gold surface comprised polycrystalline gold grains in a coral reef shape with ≈20 nm grain aggregates (Figure 1e) spanning homogeneously across the entire electrode surface ( Figure S4, Supporting Information). The SEEDING yielded large nanoporous structures on the electrode with ≈3 µm thickness ( Figure 1f) and presented sponge-like behavior when a water droplet was placed on top of the electrodes, quickly absorbing the liquid due to the large porosity of the substrate (Video S1, Supporting Information). Initially, the contact angle of NSG electrodes displayed a slight hydrophobic character (105°) ( Figure S5, Supporting Information), probably due to the roughness of the surface, and further functionalization of the electrode was difficult (Video S1, Supporting Information). However, modification with L-cysteine recovered the hydrophilic character of the surface (39°) and, in turn, helped for further functionalization of the electrode with bioreceptors ( Figure S5, Supporting Information).
The root mean square of the roughness R q , as measured by atomic force microscopy (AFM), reached values of 340 ± 90 nm for NSG versus flat gold (1.2 ± 0.3 nm) (Figure 1g).
To explore the SEEDING mechanism, we examined the three components used in detail and their interactions during the electrochemical process. Initially, CTA + displays weak adsorption on the gold electrode when Cl − is the counter ion. [18] Measuring the charge transfer resistance R CT , that is, the resistance to transfer an electric charge between the electrode and the diffusible redox molecule in the media, by electrochemical impedance spectroscopy, we can see a negligible change in R CT using [Ru(NH 3 ) 6 ] 3+ with gold electrodes between the presence or absence of 100 × 10 −3 m CTA + in solution (Figure 2a).
In solution, CTA + forms stable positively charged micelles that coordinate strongly with free chloroauric acid via electrostatic interactions. [19] The diameter of these micelles increased in the presence of chloride (Figure 2b), as expected due to counter-ion condensation, [20] but remained stable and monodisperse. The addition of chloroauric acid in small quantities (1:1, chloroauric: micelles) decreases the ζ-potential (Figure 2c) of the micelles, further confirming the coordination CTA-AuCl 4 , but at a ratio of ≈2:1, flocculation occurs due to neutralization of the positive charges of CTA + .
To analyze SEEDING in detail, we can conduct the electrochemical process slowly by scanning the voltage on a gold rod electrode within a potential window (Figure 2d). The overall mechanism of the SEEDING process is depicted in Figure 2e. When poising the electrode at the anodic potential, chloride can diffuse toward the gold surface, [21] generating compact chloride adatoms at preferential sites (Figure 2d-①,e-①). [22] At this point, the formation of chloride adlayers favors the co-adsorption of CTA + on the electrode. [23,24] When the onset potential for oxidation is reached, gold is etched with chloride ( Figure 2d-②,e-②), generating chloroauric acid in the process [22] following the reaction: [25] This process is anisotropic, preferentially etching away gold at specific step edges, kinks, and vacancies, where the formation of chloride adlayers is more stable, [22] generating, in turn, more nucleation points for growing nanostructures. The freshly dissolved chloroauric acid quickly coordinates with free micellar CTA + in solution, preventing chloroauric acid from diffusing away from the surface. Switching the polarity of the scan rate, we observed the micellar CTA-AuCl 4 being reduced and deposited on the surface of the electrode at ≈0.6 V (Figure 2d-③,e-③), following Equation (1) in reverse order.
Reduction of the CTA-AuCl 4 proceeds anisotropically, preferentially on the gold facets where access from gold-laden micelles is easier, such as at the edges of the crystalline faces, where the CTA + layer is less dense, the electrochemical potential is higher, and the surface energy is lower, promoting more efficient growth. [18,26] The iteration of this process yields the formation of increasingly large surfaces, and the cathodic current at this step is more significant after each cycle (Figure 2d). During this process, the morphology of the electrode changed; initially, during the first seconds, we observed that the grain boundaries were preferentially etched, creating large crevasses ( Figure 2f). The generated chloroauric acid is coordinated with CTA + micelles and quickly redeposited. At about 3 s, the nucleation points are visible at the grain edges. At these nucleation points, gold can be further deposited, yielding dense nanostructured and nanoporous electrodes within 1 min. www.advmat.de www.advancedsciencenews.com

Electrochemical Behavior of NSG Electrodes
The electrochemical behavior of the NSG electrodes was assessed by measuring the electron transfer kinetics using a diffusible redox mediator. The oxidation/reduction of an equimolar electroactive couple, potassium ferrocyanide and potassium ferricyanide, exhibits a reversible redox reaction in both electrodes, but with increased current densities for NSG compared to flat gold electrodes (Figure 3a). This is caused by the morphology of the electrode surface, which accounts for both the roughness [27,28] and the porosity. [29] The Cottrell equation defines a linear relationship between the current density j and the square root of the scan rate ν 1/2 for diffusion-controlled reversible systems, as observed in our experiments (Figure 3b).  On a flat electrode, this diffusion profile was planar at any scan rate (Figure 3c, left). Similarly, for NSG, at slow scan rates, the electrode displays large diffusion layers similar to a flat electrode ( Figure 3c, right). However, at faster scan rates, where the diffusion layers are smaller and become comparable to the size of the electrode's nanostructures and nanopores, the enhanced electron transfer kinetics arise from the morphology of the electrodes (Figure 3c, right). Therefore, for a flat electrode, the relationship α = jν −1/2 is constant ( Figure 3d). However, in the case of the NSG electrodes, we observed increased electron transfer kinetics and a deviation of α with ν ( Figure 3d).
The formation of oxide adlayers on polarized gold electrode surfaces exhibits crystalline face selectivity, that is, the voltammetric oxidation peak has different energies for each crystalline facet orientation Au{100} < Au{110} < Au{111}. [30][31][32] NSG electrodes prepared at 1.20 V oxidating potential show an increase in Au{110} compared to Au{100} and Au{111} (Figure 3e). The SEEDING process at potentials higher than 1.2 V led to the progressive co-formation of oxide layers along with the chloride adlayers up to 1.3 V, where the formation of the oxide film was preferential. This process has already been studied by Gaur and Schmid, [33] showing that this oxide layer passivates the electrode and prevents further dissolution of gold. The same passivation effect was observed after running the SEEDING at 1.3 V without a reduction step at the end, leading to a flat voltammogram ( Figure 3e). The formation of the oxide film was detrimental to the etching of gold and, therefore the formation of nanostructures, which in turn yielded smaller surface areas.
Grazing-incidence X-ray diffraction (XRD) on the NSG electrodes revealed the expected peaks for polycrystalline gold indexed to diffraction from the Au(111), Au(200), Au(220), and Au(311) planes of the face-centered cubic structure of metallic gold (JCPDS, card# 04-0784). The NSG surfaces exhibited an increased peak intensity compared to the flat gold electrodes, which could be explained by the increased thickness of the NSG electrode ( Figure 3f). The mean size of the crystalline domain for NSG was 21 ± 4 nm (Table S1, Supporting Information), comparable to the grain size observed by scanning electron microscopy (SEM) (Figure 1f and Figure S4, Supporting Information), but with different facet relative abundances (Table S1, Supporting Information). The crystalline orientation of Au{110} was the most abundant for both flat gold and NSG, which is in agreement with the electrochemical measurements (Figure 3f).  Randles-Sevçik plot of oxidation and reduction peak currents (i p ) (circles) versus the square root of the scan rate (extracted from the voltammograms in (b)) (n = 4 independent electrodes). The error bars represent the standard deviation of the mean. c) Schematic showing diffusion profiles on flat electrode (left) and NSG electrode (right) at different scan rates. d) Change in the scan rate behavior of j versus α for flat (black) and NSG electrodes (blue). e) Square-wave scan voltammograms of NSG electrodes prepared by chronoamperometry at different oxidation potentials versus a flat electrode as a control (*no reduction step). f) Grazing-incidence X-ray diffraction spectrogram of an NSG electrode versus a flat gold electrode.

Evaluation of Assay Strategy Depending on the Electrode Morphology
www.advmat.de www.advancedsciencenews.com with the diffusion of species toward the surface, that is, it is a diffusion-limited reaction. Therefore, at low concentrations, these species do not have sufficient time to diffuse within the entire porous substrate before the reaction occurs. In other words, there is almost no difference between the signal output using a flat electrode and an NSG electrode for a diffusionlimited reaction (Figure 4a), which is in agreement with the findings of previous studies. [14,15] The diffusion profile for this reaction on flat electrodes is planar (Figure 4b), whereas only a small contribution from the irregular surface of NSG is responsible for the increased sensitivity from 6 to 14 µA mm −2 mm −1 (Figure 4a).
The electrochemical biosensors we are developing are based on the use of antibodies as bioreceptors and therefore work similarly to traditional enzyme-linked immunosorbent assays (ELISAs). ELISA typically uses an enzyme substrate, for example, 3,3′,5,5′-tetramethylbenzidine (TMB), that is oxidized by horseradish peroxidase (HRP) at the end of the assay and changes the color. TMB is also electroactive and can be measured electrochemically (Figure 4b, left). However, considering the limitation of our NSG electrodes at diffusion-limited processes, we used a different formulation that generated a precipitated TMB instead of a soluble compound. This TMB precipitate was quickly adsorbed on the electrode surface during the detection process and could be measured electrochemically. Therefore the process was not diffusion-limited but rather reaction-limited (Figure 4b, right).

Evaluation of Antifouling Performance Based on Electrode Morphology
Surfaces in contact with biological samples, such as blood plasma, quickly adsorb proteins that generate a passivating multilayer. Furthermore, this process is typically enhanced in nanostructured surfaces because a larger surface area is exposed. Therefore, enhancing the sensitivity of the assay by only increasing the surface area with the nanostructures is not beneficial. Instead, it yielded an unexpectedly low sensitivity with real samples. However, if the electrode also has a large nanoporous structure beneath the surface, the diffusion of these contaminating proteins into the sensing area and passivation is limited. [13] To assess the antibiofouling performance of the NSG electrodes, we compared the electron charge transfer resistance R CT , that is, the resistance to transfer an electric charge between the electrode and a diffusible redox molecule in different media. Flat surfaces showed large adsorption of bovine serum albumin (BSA), increasing R CT to >4000% of its original value, whereas NSG remained almost undisturbed during 20 h of exposure (Figure 4c). The same trend was observed in plasma, with the NSG displaying one order of magnitude less R CT than the flat surface. The flat electrodes were easily blocked by the compact biofouling layers from the sample proteins (Figure 4d  www.advmat.de www.advancedsciencenews.com the nanopore substrate [13] and loose formation of the biofouling layers, preventing the blocking of available electroactive sites (Figure 4d, right).

Applications of NSG Electrodes for Biosensing
We tested NSG electrodes for biosensing and diagnostic applications using various analytes. We designed an immunoaffinity biosensor by functionalizing the surface of the NSG electrode with an appropriate bioreceptor for the detection of IL6, a soluble protein inflammatory marker commonly found in plasma. We also used a detection strategy based on the precipitation of an electroactive compound at the end of the assay. This allowed us to conduct more sensitive measurements, achieving a 17-fold higher sensitivity (0.062 ± 0.002 µA mL mm −2 pg −1 for NSG vs 0.0038 ± 0.0001 µA mL mm −2 pg −1 for flat gold) with a limit of detection (LOD) of 1 pg mL −1 for NSG versus 31 pg mL −1 for flat electrodes (Figure 5a). Moreover, one critical feature that we examined was the robustness of the biosensor when the same measurements were conducted in human plasma. Although the flat electrodes were completely passivated and detection of IL6 was not possible, the NSG electrodes showed only an increase in the LOD from 1 to 10 pg mL −1 .
By harnessing the antibiofouling and sensitive detection capabilities of this strategy, we employed the NSG-based Adv. Mater. 2022, 34, 2200981 Figure 5. Application of nanostructured porous gold electrodes for analysis. a) Calibration plot representing oxidation peak current (circles) recorded on NSG electrodes versus flat electrodes (magnified on the right side) at different concentrations of IL6 spiked in PBS or human plasma (n = 4 independent electrodes). Error bars represent the standard deviation of the mean. b) Calibration plot for CD9 + , CD81 + , or EpCAM + on CD9-captured EV, representing peak current densities (circles) versus particle concentration using NSG or flat electrodes (n = 4 independent electrodes). The error bars represent the standard deviation of the mean. c) Analysis of clinical urine samples using NSG biosensors showing current density values (circles) with different assay schemes, detection of CD9 + or EpCAM + on CD9-captured EVs (n = 20 for each box, five biological samples with four technical replicates for each). d) Electrochemical current density values (circles) for different clinical human plasma samples from healthy and patients with cancer regarding the detection of EpCAM + on CD9-captured EVs (n = 36 for each box, nine biological samples with four technical replicates for each). e) ROC curves showing the classification ability (healthy, cancer) for the three employed assays using either urine or plasma clinical samples. The boxes in (c) and (d) extend from the 25th to 75th percentiles, the middle line is the median, and the whiskers extend from the minimum to maximum values. www.advmat.de www.advancedsciencenews.com electrochemical biosensor to detect tEVs in undiluted blood plasma, which are promising biomarker candidates for minimally invasive cancer diagnostics. Here, we functionalized the NSG electrodes with anti-CD9, an EV-associated biomarker, to capture EVs on the surface and then used specific detection antibodies to investigate the presence of particular biomarkers on the EV. NSG biosensors could capture EVs from lymph node carcinoma of the prostate (LNCaP) cell culture supernatant spiked in human plasma and selectively detect biomarkers. We detected CD9 and CD81 for total EV quantification and EpCAM, a tumor-associated antigen, for tEV quantification (Figure 5b). NSG biosensors allowed us to detect total EVs and tEVs with high sensitivity using only 5 µL of plasma directly without EVs isolation, achieving a LOD for total EVs (CD9 + EVs) of 60 vesicles µL −1 and tEVs (EpCAM + EVs) of 300 vesicles µL −1 . For comparison, flat gold electrodes were electrically passivated and could not be tested for the assay in plasma and the ELISA achieving a LOD in phosphate-buffered saline (PBS) of 700 vesicles µL −1 for total EVs and 8000 vesicles µL −1 for tEVs ( Figure S6a, Supporting Information).
Finally, to investigate whether NSG biosensors could be used to detect tEVs in clinical samples, we analyzed a small cohort of clinical samples from patients with prostate cancer. [34] We first analyzed the concentration of CD9 + EVs and EpCAM + EVs in urine samples drawn from a small cohort (n = 10) that included patients with prostate cancer (stages II, III, and IV), aged 54-82 years. Urine CD9 + EV levels were not significantly higher in patients with prostate cancer than in healthy controls (Figure 5c), whereas EpCAM + EV levels were higher in patients with prostate cancer (p < 0.0001). Accordingly, we tested the blood plasma of a larger cohort (n = 18), which included patients with prostate cancer (stages II, III, and IV), aged 66-82 years. EpCAM + EV levels were significantly higher in patients with prostate cancer than in healthy controls (Figure 5d).
Receiver operating characteristic (ROC) curves indicated that EV detection using single associated marker CD9 + EVs showed extensive overlap across the groups, with no discriminatory power for classifying patients with cancer versus healthy controls (Figure 5e). Moreover, ELISA detection of either CD9 + EVs or EpCAM + EVs in the same clinical samples could not discriminate between healthy and cancer donors ( Figure S6b,c, Supporting Information). However, both urine and plasma EpCAM + EV levels constituted good classifiers (area under the curves of 0.91 and 0.90, respectively) for differentiating patients with prostate cancer from healthy cases (Figure 5e). Interestingly, EpCAM + EV levels measured by NSG biosensors in plasma were approximately ten times higher than those in urine. Finally, the serum prostate-specific antigen levels of all cancer cohorts (urine and plasma) differed widely, showing no correlation with prostate cancer stage (p = 0.7454) or Gleason sum (p = 0.6940). Therefore, these results indicate a correlation between circulating EpCAM + EV levels in both urine and plasma and the presence of prostate cancer. Notably, these clinical samples were analyzed directly with the NSG electrodes without dilution or preprocessing steps, suggesting the potential utility of this biosensing platform for sensitive detection of markers in complex matrices.

Conclusions
We developed SEEDING, a simple and fast (1 min) method to generate nanostructured and nanoporous surfaces from microfabricated thin-film gold electrodes without templates. The high electroactive area achieved on these surfaces increased the sensitivity of the bioassays because of the higher bioreceptor immobilization yield. Moreover, faster apparent electron transfer kinetics lead to amplification of the electrochemical signals, and porosity provides a size-exclusion mechanism that prevents biofouling. This allowed direct exposure to biological fluids containing large concentrations of contaminating proteins while maintaining high sensitivity.
Functionalizing these surfaces with specific bioreceptors allows for the specific detection of soluble proteins or EVs in low-volume complex matrices. Because the sensitivity of large surface electrodes is limited by the diffusion of species involved, we took advantage of a transduction mechanism based on the precipitation of an electrochemical mediator upon completion of the assay. Notably, the low detection limit for detection tEVs from small volumes of undiluted plasma and urine samples allowed us to discriminate between healthy controls and patients with prostate cancer. Even though other biosensing strategies report lower limits of detection (Table S3, Supporting Information), these are mostly based on the detection of generic EV biomarkers like tetraspanins, report their values in buffer, or are optical methods. Despite the high level of total EVs in the plasma, tumor EVs instead have a relatively low abundance but are important for clinical analysis. Plasma is more relevant for clinical settings than the buffer, but is a challenging medium that quickly passivates surfaces upon contact due to biofouling, reducing the sensitivity of the biosensors.
In this scenario, SEEDING offers key advantages over traditional high surface gold electrodes generation methods (Table S3, Supporting Information). Specifically, it is a relatively simple and fast process that works on microfabricated thinfilm electrode chips and generates nanostructured and nanoporous electrodes. The high sensitivity and low detection limit with antibiofouling performance is a key aspect for analyzing low-concentration biomarkers in real samples. Therefore, SEEDING may be used to develop quick, selective, sensitive, and miniaturized diagnostics in biological samples.

Experimental Section
Fabrication and Preparation of the Planar Electrodes: Planar electrodes were fabricated in a cleanroom using standard lithography processes on 4-inch soda-lime glass wafers (Product #1631, University Wafers), generating 44 chips (1.5 × 0.9 mm) containing four working electrodes. First, 20 nm of titanium and 400 nm of gold were deposited by e-beam evaporation (FC-2000, Temescal). Following spin coating (4000 rpm, 500 rpm s −1 , 1 min) of a 1.3 µm-thick photoresist (AZ 5214 E, MicroChemicals GmbH), and evaporation of the solvent on a hot plate (90 s at 105 °C), the wafer was exposed in a mask aligner (MA/BA6, SUSS MicroTec Korea Co. Ltd.) (h-line 405 nm, 90 mJ cm −2 ) using a Mylar mask (Advance Reproductions, USA) attached to a blank glass. After developing the photoresist (AZ MIF 326, MicroChemicals GmbH) to transfer the pattern, the exposed gold layer was etched with a gold etchant (cat# 651818-500ML, Sigma-Aldrich) for 1 min and the titanium layer was stripped with a buffered oxide etchant (BOE 7:1, www.advmat.de www.advancedsciencenews.com MicroChemicals GmbH) for 40 s. The remaining positive photoresist was removed in isopropyl alcohol, and a 2 µm-thick photoresist (SU-8 2002, K1 Solution) was spin-coated on the wafer (3000 rpm, 500 rpm s −1 , 1 min). Following a prebaking step of the negative photoresist (1.5 min at 65 °C and 3 min at 95 °C), the wafers were exposed in a mask aligner (i-line 365 nm, 120 mJ cm −2 ) and quickly postbaked (1.5 min at 65 °C and 1.5 min at 95 °C) before developing the photoresist (SU-8 developer, K1 Solution) to open up the electric contacts and limit the sensing area on the electrodes. Finally, the chip was rinsed with isopropyl alcohol, and the negative photoresist was further cured in a hard bake step (3 h at 180 °C). A protective layer of positive photoresist was spin-coated to protect the chips during dicing and keep them clean during storage. Before use, the chips were rinsed in acetone and cleaned in O 2 plasma (Cute, Femto Science) at 0.5 mbar and 50 mW for 2 min.
Preparation of Gold Rod Electrodes: Gold rod electrodes (Φ = 2 mm, cat# CHI101, Qrins) were prepared before use with a polishing kit (cat# CHI120, Qrins) by consecutively polishing their surfaces against alumina slurries of different sizes on a polishing pad. First, the authors used a 1 µm slurry on a CarbiMet disk, a 0.3 µm slurry on a nylon pad, and finally, a 0.05 µm slurry on a microcloth pad using an "8"-shaped motion for ≈30 s, and sonicating the electrode in a water bath (3510, Branson) for 1 min after each polishing step.
Preparation of SEEDING Solution: The solution for SEEDING electrodes was prepared by dissolving 36 mg of cetyltrimethylammonium chloride (TCI, cat# H0082, lot# 4LL50-MN) per mL of a 222 × 10 −3 m sodium chloride stock solution. Then, 1/9th of the volume was added as chloroauric acid 10 × 10 −3 m (stock stored in the dark at 4 °C) to obtain a final concentration of 100 × 10 −3 m cetyltrimethylammonium chloride, 200 × 10 −3 m NaCl, and 1 × 10 −3 m HAuCl 4 . This solution was stable at room temperature (25 °C) when stored in the dark. Preparation using cetyltrimethylammonium bromide (Sigma-Aldrich, cat# 52365-50G, lot# BCBT1510) generated a similar SEEDING performance, but the formulation was unstable over time.
Electrochemical Setup and Measurements: All the electrochemical processes and measurements were conducted in a potentiostatgalvanostat EC-Lab (VSP model with a low-current option, BioLogic, France) using a three-electrode configuration, with an external platinum wire as the counter electrode and a miniaturized leak-free Ag/AgCl electrode as the reference electrode (cat# ET072-1, Qrins). We used either microfabricated thin-film circular gold electrodes (Φ = 0.45 mm) or gold rod electrodes (Φ = 2 mm) for the working electrodes, with in-house-built connector boxes ( Figure S7, Supporting Information). For photolithographically patterned microfabricated thin-film electrodes, a laser-cut double-sided adhesive tape (DFM 200 clear 150 POLY H-9 V-95, FLEXcon) was attached with a rectangular shape and a main channel in the center to generate a 5-10 µL reservoir for incubation of samples on the electrodes. For gold rod electrodes, the bottom of the tube was used as an electrochemical cell (cat# CLS430828-500EA, Sigma-Aldrich).
Calculation of Diffusion of Species: The distance that electroactive diffusing particles moved away in one dimension from the surface of the electrode after a certain time (i.e., the thickness of the diffusion layer, δ), could be calculated using the root-mean-square displacement equation: [35] where D (cm 2 s −1 ) is the diffusion constant of the reacting species, and t is the time (s). Using the parameters D = 9.0 × 10 −6 cm 2 s −1 [36] and t = 0.001 s, the maximum diffusion layer thickness was estimated as 1.9 µm. The aggregation number of a micelle, N agg , could be estimated. [37] The volume of the hydrophobic tail was given by: with n being the number of carbon atoms in the tail. The radius of a micelle, with n c carbon atoms in the long apolar tail of its monomers and ′ n c carbon atoms inside the chains could be approximated by: , such that the radius was R = 23.5 Å and the N agg = 87. However, considering the concentration of the electrolyte [NaCl] = 0.2 m, one could estimate a N agg of ≈200, which was also in agreement with the findings of other studies (N agg = 75-170). [38] In this scenario, the diffusion constant D (m 2 s −1 ) of a CTA + micelle could be estimated using the Stokes-Einstein equation: where k B is the Boltzmann constant (1.38065 × 10 −23 m 2 kg s −2 K −1 ), T is the temperature (K), η is the viscosity of the medium (Pa s = kg m −1 s −1 ), and R 0 is the radius (m). Considering R 0 = 23.5 × 10 −9 m, T = 298 K, and η = 0.00089 kg m −1 s −1 , the authors obtained D = 1.0 × 10 −6 cm 2 s −1 . This diffusion constant was lower than that for free HAuCl 4 ; thus, the quick formation of CTA-AuCl 4 micellar complexes diffused ≈2% of the distance from its free counterpart, favoring the reutilization of etched gold in the SEEDING process.

SEEDING on Microfabricated Thin-Film Electrodes by
Step Chronoamperometry: SEEDING on photolithographically patterned microfabricated thin-film gold electrode chips was conducted in 10 µL of SEEDING solution by step chronoamperometry, consisting of 20 000 cycles of step voltages; first 1.2 V for 1 ms and then −1.2 V for 2 ms. Finally, a chronoamperometric step at −1.2 V was applied for 10 s to reduce gold oxides. This process was conducted using the external platinum wire as the counter electrode and a miniaturized leak-free Ag/AgCl electrode as the reference electrode, not the internal gold electrodes from the microfabricated thin-film chip.
SEEDING on Rod Electrodes by Cyclic Voltammetry: SEEDING on rod electrodes was conducted in 5 mL of SEEDING solution, by cyclic voltammetry, consisting of 20 cycles between 1.25 and 0 V, starting at 0 V and a scan rate of 0.1 V s −1 . Finally, a chronoamperometric step at −0.45 V was applied for 10 s to reduce the gold oxides.
Characterization of Electrochemical Surface Area: Electrodes were scanned by CV from 0 to 1.6-1.8 V in an acidic solution of H 2 SO 4 0.5 m at a scan rate of 0.1 V s −1 , integrating the area (charge) under the reduction curve at ≈0.9 V. The cathodic peak was proportional to the amount of gold being reduced; therefore, it was proportional to the surface area. Assuming a specific charge transfer of 400 µC cm −2 , [17,39] the electrode area could be calculated. By comparing it with the geometric area, the roughness factor was obtained. Additionally, the authors used the averaged values of the anodic scan where only capacitive current was observed (from 0 to 1 V) to measure the roughness factor of NSG using flat gold electrode as a reference.
As a comparison method, the surface roughness factor was also measured by electrochemical impedance spectroscopy at f = 100 Hz using background electrolyte (10 × 10 −3 m PBS + 0.1 m KCL) and extracted the capacitive reactance values. The area then could be calculated using the equation as follows: [40] π ( ) where A is the total electrolyte-accessible surface area, C is the capacitance of the NSG, C DL is the double-layer capacitance measured in flat gold electrodes, F is the frequency (100 Hz), and Im(z) is the imaginary part of the impedance (capacitive reactance). using SEM (S-4800, Hitachi High-Technologies) and AFM (DI-3100, Veeco). Before SEM characterization, the samples were sputter-coated with a 5 nm layer of gold (E-1045, Hitachi). Imaging was performed at a working distance of 3-4 mm at an accelerating voltage of 7 kV. The NSG electrode roughness and morphology were characterized by AFM in noncontact tapping mode using silicon AFM tips with a radius curvature <10 nm and an aluminum reflex coating (300AL-G-10, Woomyoung Inc.).
Contact Angle Measurements: Surface energy analysis was performed by water contact angle measurements: 1 µL drops of water were dropcast on different surfaces, and images were immediately recorded. The images were analyzed with ImageJ 1.53e, approximating the shape of the profile droplets to a sphere using a seven-point manual selection procedure. The contact angle was calculated using the following formula: where θ is the contact angle (rad), h is the height of the droplet, and l is the length of the base of the drop. Crystallite Size from X-ray Diffraction: NSG surfaces were prepared on diced chips (1 cm × 1.5 cm) with evaporated Ti/Au (20/400 nm) on a glass wafer. An adhesive tape with a squared 4 mm side opening was attached to the surface to generate a working electrode and cover the rest. The SEEDING process was conducted following the same method used for the electrode chips (see SEEDING on microfabricated thin-film electrodes by step chronoamperometry).
The crystallite size was calculated from the XRD spectra, fitting the peaks to a Gaussian model with Fityk 1.3.1, using the full width at half maximum from peak Au(220), and using the Scherrer equation: where τ is the mean size of the crystalline domain, K is the dimensionless shape factor (0.94), λ is the X-ray wavelength (λ = 1.5418 Å), β is the line broadening (rad), and θ is the Bragg angle (°). Electrochemical Characterization of Gold Crystalline Faces: The authors used SEEDING on microfabricated thin-film gold electrodes, but using different oxidation voltages, namely 1.20, 1.25, and 1.3 V. An additional set of electrodes was prepared at 1.3 V, but the reduction step for 10 s at −1.2 V was omitted. Formation of different crystalline gold faces generated on NSG electrodes was characterized electrochemically in an acidic media of H 2 SO 4 0.5 m by squared wave voltammetry, scanning from 0 to 1.3 V at a scan rate of 0.02 V s −1 (pulse height 30 mV, pulse width 100 ms, step height 4 mV).
Biofouling Behavior on NSG Electrodes: The biofouling behavior of NSG surfaces was characterized by Faradaic electrochemical impedance spectroscopy in two different media, 1% BSA or human plasma, containing an equimolar concentration of ferrocyanide K 4 Fe(CN) 6 and ferricyanide K 3 Fe(CN) 6 2.5 × 10 −3 m. Measurements were performed from 0.1 MHz to 0.1 Hz, at an amplitude of 5 mV versus an open circuit potential. The R CT of the NSG electrodes was determined by fitting the data from Nyquist plots to a Randles equivalent circuit, in which the R S modeled the resistance of the solution, and the constant phase element (nonideal capacitance) was used to model the double-layer capacitance (C DL ), and, in parallel, the r CT and a Warburg element (Z W ) modeled the diffusion of electroactive species in solution ( Figure S8, Supporting Information).
Electrode Functionalization with Capture Antibodies: The electrode chips were immersed in a fresh solution of l-cysteine 10 × 10 −3 m for 24 h under constant shaking, and rinsed in ultrapure water. A glutaraldehyde solution 2.5% prepared in PBS pH = 8 was drop-cast onto the surface of the chips and incubated for 30 min, after which the chips were rinsed and dried. A 20 µg mL −1 of anti-IL6 solution (Thermo Fisher Scientific, cat# CHC1263, lot# 172402) or anti-CD9 (BD Pharmingen, cat# 555370, lot# 9014503) was prepared in PBS, and deposited on individual electrodes with a hollowed ceramic needle (Φ i = 350 µm; LabNEXT, Inc., no. 007-350) and incubated overnight at 4 °C in a water-saturated atmosphere. After incubation, the electrodes were washed with 0.05% PBS-Tween 20 in a shaker for 30 min to remove unbound antibodies. Then, 10 µL of 1 m ethanolamine (Sigma-Aldrich, cat# E9508) in PBS, adjusted to pH 7.4, with HCl, was drop-cast on each electrode and incubated at room temperature for 30 min to neutralize the unreacted moieties. Finally, the chips were incubated in 1% BSA at room temperature for 1 h before use.
Cell Culture of LNCaP Cancer Cells, EV Isolation, and Concentration: The EVs for standards were extracted from LNCaP cell supernatant and enriched by centrifugal force filtration inside a disc containing a 20 nm anodized aluminum oxide membrane filter. [41] LNCaP cells were obtained from ATCC and cultured in Roswell Park Memorial Institute (RPMI) medium (Gibco, Thermo Fisher Scientific) supplemented with 5% Exo-Free fetal bovine serum (Systems Biosciences, Inc.) and 1% antibiotics/ antimycotics. The cells were incubated at 37 °C in 5% CO 2 for 48 h. The cell culture supernatant was collected, centrifuged at 300×g for 10 min, and then centrifuged at 2000×g for 15 min to completely remove dead cells and cellular debris. The supernatant was passed through a 450 nm filter.
Preparation of EV Standards: Enriched EVs from LNCaP cells were used to prepare standards. EVs were reconstituted in PBS to measure the particle count number by nanoparticle-tracking analysis (Nanosight NS500, Malvern Instruments) and spiked in PBS or human plasma samples at different dilution rates to prepare the calibration curves.
Clinical Samples: Plasma and urine samples were provided by the Biobank of Pusan National University Hospital, a member of the Korea Biobank Network, after approval by the Institutional Review Board (IRB) of Pusan National University Hospital (IRB 1802-004-063). Additional plasma samples from healthy donors were obtained from volunteers at the Yeungnam University Medical Center after approval by Yeungnam University Hospital IRB (YUMC2018-04-011). All samples were collected after receiving written consent from all participants.
Within 2 h of sample collection, urine was centrifuged at 500×g for 10 min at 4 °C to remove cellular debris and stored at −80 °C until analysis. Urine samples were thawed for analysis and centrifuged at 2500×g for 15 min at 4 °C, and the precipitated pellet was discarded.
Blood samples (3 mL) were collected in vacutainer ethylenediaminetetraacetic acid collection tubes and processed within 2 h of collection. Samples were centrifuged at 300×g for 10 min, then 2000×g for 10 min at 4 °C to remove cells and debris, and stored at −80 °C before use.
Electrochemical Enzymatic Sandwich Detection Using NSG Electrodes: The reagents and dilutions used for the IL6 detection assay and the washing steps were all conducted in assay buffer (1% BSA and 0.05% Tween 20 in PBS). IL6 (Thermo Fisher Scientific, cat# CHC1263) was spiked at different concentrations in PBS and human plasma and 20 µL was incubated on the electrode chips for 2 h. The biotinylated anti-IL6 detection antibody (Thermo Fisher Scientific, cat# CHC1263) was diluted to 1 µg mL −1 and incubated on the chips for 1 h. To complete the assay, HRP-streptavidin (Thermo Fisher Scientific, cat# CHC1263) was diluted 1:400 in assay buffer and incubated for 30 min. An enhanced one-component HRP membrane substrate (precipitating TMB, Sigma-Aldrich, cat# T9455) was incubated on the chips for 1 min. After each of the previously described steps, the electrode chips were washed in a shaking Petri dish with assay buffer for 15 min. The electrode chips were individually interrogated by cyclic voltammetry at a scan rate of 0.1 V s −1 between −0.5 and 0.5 V versus open circuit potential using the internal counter and pseudo-reference electrode of the chips. Other calibration curves were obtained under different conditions by spiking IL6 in PBS instead of plasma and using flat lithographic gold electrode chips for comparison.