Hollow Microneedle‐based Plasmonic Sensor for on Patch Detection of Molecules in Dermal Interstitial Fluid

Interstitial fluid (ISF) extraction and analysis are challenges that can be tackled by Hollow MicroNeedles (HMNs) technology, overcoming most of the difficulties associated with in situ detection. Herein, a plasmonic transducer, composed of gold nanoparticles embedded in poly(ethylene glycol) diacrylate (PEGDA) hydrogels, is integrated in the inner cavity of HMNs to detect biomarkers from the ISF‐based point‐of‐care. The wearable HMN‐based patch is used for minimally invasive pierce of the skin. The large swelling capability of the plasmonic transducer allows the uptake of ISF by capillarity. Biotin, as a small model molecule, is efficiently collected in the inner cavity of HMN and its high specificity with the streptavidin is exploited as a validation of the plasmonic nanocomposite functionality embedded within. The recognition of biotin is achieved in dual‐optical mode: the localized surface plasmon resonance (label‐free) and the metal‐enhanced fluorescence (label‐based). Overall, the proposed HMN‐based patch for target sensing in ISF can represent a novel point‐of‐use device for the detection of biomarkers as an alternative to conventional hospital or lab settings to help faster medical decision‐making.


Introduction
The diagnosis and monitoring of a disease are often conducted by measuring biomarkers found in blood, urine, saliva, and other body fluids. In particular, the interstitial fluid (ISF) is a rich source of biomarkers that surrounds cells and tissues of the body. [1] It is advantageous for biosensing applications since it does not contain any particulates and transports less proteins and increases the possibility of infection by microorganisms during the long sampling times required to collect enough ISF. [16,17] MNs made of water-swelling polymers extract the skin ISF during the swelling process [18][19][20] and additional steps such as incubation and centrifugation are required to extract and analyze analytes absorbed into MNs. [18] Hollow MNs (HMNs) have built-in cavities that allow the extraction of a small amount of the skin ISF (200 nL per MN) by capillary action over extended periods of time (15-20 min) or larger volumes (10 μL per MN) in a reducing time (2-10 min) using MNs in combination with vacuum suction. [21,22] HMNs can act as efficient biological fluid collectors creating a transdermal fluidic path at the interface between the dermis and the outer part of the skin. [23] Usually, HMN systems are integrated with an absorbent paper system capable of collecting the ISF for subsequent chemical analyzes. [24] The ISF is collected via the HMN device and almost always analyzed in a separate instrument, which requires an additional ISF transfer step, i.e., the extraction of the analyte from the MN paper patch by incubation in extraction medium, centrifugation, and measurement with an appropriate analytical method. [24] These devices require the transfer of collected ISF out of the MN lumen to the analyte detector. It extracts small volumes of accessible ISF with a slow time, leading to slow sensor response. To overcome the limitations of other MN technologies, this study presents an HMN patch, filled in its cavities by a 3D highly swellable network containing gold nanoparticles (AuNPs) as a plasmonic sensor. This device can directly detect the biomarker's extraction without any additional steps. The integration of a 3D optical transducer based on high molecular weight (Mw) poly(ethylene glycol) diacrylate (PEGDA) and spherical AuNPs brings together several advantages, such as good adaptability and flexibility within the HMNs cavity, higher surface area, and surface-to-volume ratio, and interrogation of the sensor with no needs for complex electrical circuits, which usually represent the bottleneck of wearable biosensors due to the interferences arising from the contact with the skin. [25][26][27] HMN patches, which are a miniaturized replica of hypodermic needles, are manufactured by a photolithographic method, exploiting photo-crosslinking properties of PEGDA at low Mw. [28][29][30][31][32] PEGDA is a biocompatible and non-toxic polymer. [33,34] AuNPs are included in high Mw PEGDA, which is then inserted in the HMNs cavities. The large swelling capability of high Mw PEGDA is used to extract the ISF. This technique avoids the separate analysis of the collected ISF and allows the target molecules of interest to be detected directly from the device. AuNPs are used as optical transducers based on a phenomenon known as localized surface plasmon resonance (LSPR), which is the oscillation of the electron density on the surface of the NPs occurring at a specific excitation wavelength. [35,36] Moreover, the electromagnetic field enhancement in the surroundings of AuNPs can lead to strong fluorescence enhancement of fluorophores, if suitable conditions are satisfied. [37,38] This phenomenon goes under the name of Metal-Enhanced Fluorescence (MEF), or plasmon-enhanced fluorescence and is generally exploited to boost the limits of detection (LODs) of plasmonic biosensors down to single-molecule levels. The fluorescent molecules near the metal surface show an efficient coupling between the electromagnetic field and spatially confined free electrons, which leads to higher emission intensity. The capability of PEGDA to retain the physical and chemical stability of AuNPs, which act as optical transducers in both label-free and label-based modes, has been already reported in previous works. [39,40] The combination of plasmonic transducers and MNs has been already reported in the literature, [5,41] as a promising optical alternative to electrochemical transducers. However, in the proposed platforms, the plasmonic transducers are arranged outside the MNs arrays. Here, instead, for the first time, the plasmonic nanocomposite transducer is integrated directly into the cavities of HMNs, resulting in an integrated refractometric biosensor, whose LSPR undergoes a redshift as a function of the medium effective refractive index surrounding the AuNPs (label-free sensing), due to the confined swelling of the high Mw hydrogel within the HMNs cavities. Moreover, plasmonic nanocomposites functioned as integrated optical antennas for fluorescence enhancement (label-based sensing). Finally, the bottom-up synthesis of AuNPs was optimized to achieve the desired concentration of AuNPs in one step, with no need for exceeding centrifugation steps, which could negatively affect the stability of the NPs.
The biotin-streptavidin complex was used as a proof-of-concept to validate the sensitivity of LSPR shift-based biosensors. Furthermore, AuNPs were efficiently used to enhance the optical properties of Cy3 fluorescent dye, which is linked to streptavidin. The possibility to detect molecules in the ISF of the dermis with minimally invasive devices constitutes an interesting opportunity since the proposed devices could be used in the context of Point-Of-Care-Testing (POCT), for diagnosis, disease monitoring, and drug dosage control. [42,43]

Result and Discussion
An innovative platform, composed of HMNs filled with a plasmonic bio-transducer for sensing biomarkers in ISF is herein proposed as an alternative method compared to the invasive and expensive technologies, which are currently available.

Fabrication and Characterization of HMNs Arrays
The HMNs arrays were fabricated using a photolithographic method by a one-shot fabrication process. [32,44] It is a versatile and low-cost method that allows the production of an array of HMNs with an asymmetric tip in a few seconds, with easily tuneable properties. 3D structures were defined by a chromium mask, consisting of a dark region with a transparent array of asymmetric rings (Figure 1a). A UV-curable solution composed of a pre-polymeric solution of PEGDA 250 Da and a photoinitiator Darocur 1173 was exposed for a few seconds to UV irradiation through the photo-mask pattern. HMNs were grown directly on a previously photocured flexible polymeric film, i.e., a patch, with a thickness of 250 μm ± 50 μm, to have HMNs arrays with closed cavities on the bottom side of the base plate (Figure 1a, b). The produced HMNs had an asymmetric (side-slanted) hole similar to a steel needle of a hypodermic syringe and a height of 1000 μm ± 50 μm (Figure 1b). This height allows HMNs to reach the dermis and encounter ISF without causing acute pain, inasmuch as, only two-thirds of HMN height penetrate into the skin. The height of HMNs and the inner cavity volume can be easily modified by changing the exposure time from 4 s to 11 s. By increasing Figure 1. a) On the left, a schematic representation of the fabrication process of an HMN based on UV-induced polymerization. Right, zoom-in of photomask Quartz/Chrome. b) Left, schematic of an HMN illustrating its typical geometry, and right, top-image of a patch of HMNs and lateral-view of a single HMN by optical microscopy. c) COMSOL simulation of HMNs indentation: initial condition, maximum penetration depth in the skin model, and von Mises Stresses distribution in the HMNs. d) SEM images of the different steps of mechanical testing on HMNs, and corresponding forcedisplacement graph. e) schematic representation of HMNs array insertion into the skin; digital photograph of the fabricated HMNs patch; histological analysis of ear pigskin before and after indentation with HMNs.
the exposure times even more (20 s), the needle lumen is completely closed. HMNs were characterized by optical microscopy images, as shown in Figure 1b. To predict the insertion depth in the skin, a 3D COMSOL simulation of HMN indentation into a multilayer skin model, consisting of two layers, which simulated the SC and dermis, was developed. [45] A single HMN with geometrical properties comparable to that of a fabricated MN was designed for the numerical simulations. In Figure 1c, different steps of HMN insertion into the skin during the COMSOL simulation are shown. HMN reached a depth of penetration of about 700 μm, encountering the stress at the tip of the HMN, as shown in Figure 1c. Generally, if the stress generated on the HMN, due to the applied force, is lower than the resistance of the material of which the needle is made, the HMN is able to penetrate the skin. [46] An estimation of the maximum HMN depth penetration was calculated to evaluate if the inner core of HMN, containing the plasmonic sensor, came in contact with the dermis. In Figure 1c, an overview of the obtained Von Mises stresses on the HMN walls is provided. The simulation results confirmed that the geometrical characteristics of HMNs were suitable to penetrate the skin. Further, after geometric optimization, the mechanical properties of HMN were evaluated to prove their ability to overcome the skin barrier. To test the mechanical properties of HMNs, a mechanical compression test was carried out up to a maximum load of 1 N per MN (Figure 1d). The analysis was performed by placing a series of HMNs between two metal plates and moving one plate toward the other with a controlled speed of 0.2 mm min −1 . The video recorded in real time showed that HMNs were subject to a deflection of tips along the z-axis as the compression force increased, without any break (see the Movie S1, Supporting Information and frames in Figure 1d). HMNs were perfectly intact after the test as shown by the optical images collected during the experiment ( Figure 1d). As HMNs were compressed, the two tips began to bend at 0.7 N, indicating that each needle tip can withstand at least 0.35 N without any mechanical failures. After that, tips were bent but not completely deformed since any drop in force was observed, as reported in Figure 1d. Once all tips were deformed, the detected force increased linearly, indicating that HMNs bodies still remained attached to the substrate without deformation. [47] Up to 1 N per single MN, which is a much greater force than that required to indent, no HMN failure was observed. Further, an indentation test on skin models was performed. To prove that HMNs were secure to penetrate the skin, they were analyzed after the indentation to prove that they were still morphologically intact without any leak of polymer fragments. Piercing tests were carried out on an agar skin model that is a versatile, easy-to-produce model, with a density similar to that of the skin; [48,49] fresh ear pigskin sample, which was one of the most accurate models for human skin from the perspective of anatomy, physiology, immunogenicity, cellular composition, morphology, and mechanical properties. [50][51][52] Both pigs and men have thick epidermis, [53] similarity in terms of the lipid composition of the stratum corneum, [53] and comparable localized mechanical properties during MN insertion. [54] Further, Ranamukhaarachchi et al., suggested that, for MN research, fresh porcine skin (at high humidity conditions) has more comparable mechanical properties to fresh than frozen human skin. [52] The piercing of HMNs on the skin model was analyzed by optical microscopy after the indentation test while the penetration depth of HMN was measured by histological analysis. In Figure 1e, the performed indentation test is schematically reported. Briefly, an HMN array was placed onto the skin model and pressed by the investigator's thumb for 1 min. After the indentation, the HMN array was raised to analyze the HMN integrity and piercings on the pigskin. A consistent and reliable piercing of a single MN on both agar and pigskin surfaces is shown by optical images in Figure S1, Supporting Information. To specifically highlight the sites of piercing, HMNs tips were stained with black ink. Further, histological analysis was conducted using hematoxylin and eosin stain (H&E stain) to evaluate the depth of penetration of the HMN array. As highlighted in Figure 1e, the SC disruption at poked sites was clearly evident and compared with intact skin. The depth of penetration was 20-30% of HMNs length. Further, micro conduits formed into the skin allowed the dispersion of ink that covered HMNs into the dermis, as observed after indentation in Figure 1e.
HMNs appeared to be robust systems with optimal mechanical characteristics to penetrate the skin. After the optimization of HMNs arrays, a plasmonic transducer made of high Mw PEGDA embedding AuNPs was included in the HMNs cavities. The hydrogel provided the device with the capability of collecting and evaluating the biomarkers of ISF. Therefore, the HMN array can pierce the SC and create micro conduits for the ISF flowing.

Fabrication of Plasmonic Nanocomposites Made of High MW PEGDA and AuNPs
The plasmonic nanocomposites embedded in the HMNs cavities can collect targets within the ISF, by means of the strong swelling capability of the high Mw PEGDA. Molecular targets can be fast detected from HMNs by means of AuNPs LSPR.
Thus, the inner cavities of HMNs were filled out with a pre-polymer solution of PEGDA 10 kDa and AuNPs. AuNPs were synthesized by the modification of a seed-mediated growth method proposed by Nikoobakht and El-Sayed. [55,56] Their finetuned synthesis allowed the growth of Nanorods (NRs) with controlled aspect ratios, reducing the number of spherical particles almost to zero. In this work, instead, the concentration of CTAB molecules in the seed and growth solutions, the amounts of silver nitrate, ascorbic acid, and chloroauric acid in the growth solution, and the amounts of sodium borohydride and chloroauric acid in the seed solution were adjusted to reduce nanorods formation. This approach allowed the obtaining of highly concentrated spherical AuNPs in a single growth step with good control of the size distribution and optical properties. Despite the use of several more reagents in the solution compared to other common protocols used in the synthesis of AuNPs, [57] the proposed approach was much faster and did not require several growth steps and long centrifugation times to achieve AuNPs with the desired concentration. As schematized in Figure 2a, the growth and seed solutions were prepared separately, then a defined amount of the first was added in the second one and an immediate color change of the growth solution was observed from transparent to intense violet, indicating the formation of highly concentrated AuNPs, as reported in the experimental section. In the inset of Figure 2a, the digital photograph of an AuNPs colloidal suspension diluted one hundred times is reported to show that even after a high dilution factor, it kept an intense violet coloration. The proposed bottomup synthesis led to a high concentration of spherical AuNPs with a principal component size distribution of 40 ± 10 nm, as highlighted by the particle analysis reported in Figure 2b and fitted by a gaussian curve. The histogram was obtained by using the particle analysis tool of the ImageJ freeware software on three different images. A representative SEM image is also reported ( Figure 2b). Moreover, the presence of CTAB moieties surrounding AuNPs provided the colloidal suspension with high stability, as confirmed by -potential measurements. In fact, the resulting net electrical charge measured by the zeta-sizer was 52 ± 8 mV, which ensured electrostatic repulsion between the synthesized AuNPs and their colloidal stability.
Thin film nanocomposite transducers (patches) were fabricated to study the stability and compatibility of the newly fabricated AuNPs within the hydrogel network. [39,40] By following the protocol reported in our previous work, [39,40] AuNPs aqueous solution was added to a prepolymer solution of high Mw PEGDA (10 kDa), containing the photoinitiator. The resulting mixture was cast between two coverslips and exposed to UV light for a few minutes until complete polymerization, to obtain a plasmonic patch (Figure 2c). The highly concentrated AuNPs, Figure 2. a) Schematic representation of the bottom-up synthesis of AuNPs obtained by modified seed-mediated growth method. b) Size distribution and Gaussian fitting of the synthesized AuNPs obtained from particle analysis of three SEM images (n = 3); and a representative SEM image of AuNPs (scale bar is 500 nm). c) Schematic representation of the methodology followed for the fabrication of PEGDA 10 kDa embedding AuNPs. d) UV-Vis absorption spectra of AuNPs in colloidal solution (black line) and embedded in PEGDA 10 kDa (red line). e) Measured swelling ratios of PEGDA 10 kDa with AuNPs evaluated at times 0, 12, and 24 h reported as mean values ± SD (n ≥ 3). fabricated herein, were physically retained in the polymeric networks and exhibited higher chemical and physical stability in the long term than their aqueous counterpart (data not shown here). High Mw PEGDA (10 kDa) was chosen for its excellent swelling capabilities, high chemical affinity with low Mw PEGDA (250 Da) of which HMNs were made, and excellent optical transparency. The preservation of the optical properties of AuNPs, when embedded in the polymeric network, was ensured by collecting their absorption spectra before and after their photocuring within the high Mw PEGDA (Figure 2d). While the absorption spectra of AuNP colloids were recorded by standard spectrophotometry, the thin hydrogel patches embedding AuNPs were characterized by a custom-made LSPR transmission mode set up at normal incidence. The optical spectra of the freshly synthesized AuNPs in solution and in polymerized high Mw PEGDA were compared in the interval between 400 and 900 nm, as reported in Figure 2d.
The AuNPs in solution had to be diluted 100 times to avoid saturation and their spectrum resulted in two LSPR peaks, one at 540 nm and one at 700 nm ( Figure 2d). The first peak is attributed to well-dispersed spherical AuNPs with an average diameter of 40 nm, while the second one is attributed to the formation of clusters of AuNPs within CTAB micelles, as confirmed by SEM images (Figure 2b). Although confirmation by SEM images was obtained, the cluster formation could be already hypothesized starting from the optical spectra of AuNPs once they were embedded in the hydrogel (Figure 2d). In fact, the first peak underwent a redshift of ≈10 nm, passing from ≈535 nm to ≈545 nm due to the change in the medium effective refractive index. The relative intensity of the second peak was reduced since during the stirring, the photocuring, and the washing steps of the hydrogel nanocomposites, the micelles forming AuNP clusters were partially desegregated. Interestingly, the LSPR of AuNPs perfectly matches the maximum excitation peak (≈550 nm) of the selected fluorescent dye (Cy3). This guarantees the verification of one of the two conditions required to achieve MEF by the Förster Resonance Energy Transfer (FRET) mechanism, which is the perfect matching between LSPR absorption spectra, and fluorescent dye excitation, [58] as denoted by the green dashed lines reported in Figure 2d. The second condition has been highlighted later on in the manuscript. The swelling capabilities of the plasmonic patch were investigated since it had to guarantee proper swelling in contact with ISF. The thin patch was used for measuring the swelling ratio over time until 24 h (Figure 2e). The swelling ratio (SR) was determined by measuring the weight of the swollen hydrogel (W s ) at 0, 12, and 24 h. The plasmonic nanocomposites (24 mm × 24 mm × 0.4 mm) were weighed immediately after their polymerization, time interval denoted as 0 h since they already contained a known percentage of water due to the prepolymer preparation phase. Then, they were soaked in ultrapure water at room temperature and weighed after 12 and 24 h. Finally, the weight of the dried hydrogel (W d ) to be used in the SR evaluation was measured after leaving the hydrogel in a vacuum oven at 40°C overnight. PEGDA 10 kDa achieved the full swelling equilibrium after 24 h, as observed by the SR reported in Figure 2e. The SR evaluation of the thin film nanocomposite was performed to provide the readers the information on the capability of PEGDA 10 kDa embedding AuNPs of absorbing up to three times the initial amount of water. Once in the HMNs, due to the smaller dimensions of the polymerized hydrogel nanocomposite comparable to the volume of the HMNs cavities, the swelling equilibria were assumed to have much faster kinetics and, therefore, were expected to be achieved at much lower times.

Integration of Plasmonic Nanocomposites within the Cavities of HMNs Arrays
After the evaluation of the proposed optical transducer optomechanical properties in terms of absorption spectra and swelling capability, the prepolymer solution was poured directly into HMNs cavities and polymerized by UV-light for a few minutes. This led to the integration of the plasmonic nanocomposite directly within HMNs cavities. At this stage, a high concentration of AuNPs within the prepolymer solution resulted in being able to fill the HMNs cavities with very tiny volumes and still appreciating plasmonic optical signals. In Figure 3a, a schematic representation of the HMN filling process was presented. After exposure to UV light, a crosslinked 3D purple network was obtained in HMNs cavities. It appeared purplish thus indicating a stable dispersion of AuNPs within the HMNs. Figure 3b,c shows an array of HMNs before and after the filling with and polymerization of PEGDA/AuNPs prepolymer solution. After the first polymerization, a retraction of the hydrogel nanocomposite, indicated by a reduction in its volume, was observed, and the complete HMN cavity filling was achieved after two pouring steps and polymerization. In this way, a homogeneous filling of the whole HMN cavity was obtained, as observed in Figure 3c. In more detail, in Figure 3b, the empty cavity of a single HMN is clearly visible, while a purplish filling into the HMN cavity is observed in Figure 3c. Furthermore, a filled single HMN was detached from the array and immersed in MilliQ for 24 h to verify the proper confinement and integrity of the hydrogel nanocomposite over time, as shown in Figure S3 (Supporting Information). From the wash solution, it does not detect any AuNP leakage during the process. To verify the swelling capability of the plasmonic nanocomposite confined into the inner bore, an extraction of a commercial dye, i.e., methylene blue (MB) from a skin model was simulated. The model was made of agar-agar containing 1 mg mL −1 of MB. To perform the test, an array of HMNs was only partially filled with high Mw PEGDA hydrogel (without AuNPs). Specifically, only two external perpendicular strings of the array were filled, while the remaining part of HMNs was left empty (Figure 3d). The HMN array was inserted into the skin model and pressed for 10 minutes. In Figure 3d, the digital photograph of the agar-agar surface after HMNs extraction is reported, clearly showing sites of the piercing of the whole array.
In the upper part of Figure 3d, only nine HMNs filled with the hydrogel nanocomposites absorbed MB from agar-agar assuming a blue color. The dye diffused homogeneously through the inner hydrogel during swelling. The confined hydrogel was able to uptake up to 3-4 times its weight of water during the test time, as shown in Figure 2e. On the contrary, empty HMNs appeared colorless ( Figure 3d). After proving the swelling capability of the high Mw PEGDA in a confined volume (i.e., the inner cavity of an HMN), the hybrid device made of HMNs filled with the plasmonic nanocomposite (Figure 3c) was obtained. All HMNs of the array were filled with the violet pre-polymer solution and cured by UV-light irradiation, as mentioned above. HMNs were characterized by a customized transmission mode setup. In this system, an optic fiber conveyed the light to the inner part of a single HMN or on a larger spot, which comprised three HMNs, by means of a collimator. The scattered light was collected by another optic fiber connected to an optical spectrum analyzer, as it is possible to observe in Figure 3e. The light spot was adjusted by using biconvex optical lenses. In Figure 3f, the absorption spectra of the plasmonic transducer of three HMNs and of a single HMN were compared with the spectrum of the plasmonic transducer patch used as a reference. As expected, the reduction of the investigated optically active material led to a decrease in the average LSPR signal relative intensity. However, the signals obtained from the measurement of the transmission spectra on single HMNs were sufficiently high to detect the first LSPR peak. This was selected as the optical signal to follow during label-free sensing experiments following in the manuscript. The collection of the LSPR signal by the single HMN was mainly possible thanks to the high concentration of the synthesized AuNPs, which turned into a high density of AuNPs embedded and stabilized within the HMNs cavities. The stability of AuNPs within the HMNs was also confirmed by optical measurements since the shape of the absorption spectrum remained like the hydrogel nanocomposite thin film with no other peaks detected at higher wavelengths.

Dual Optical Mode Sensing of Biotin-Streptavidin Interaction by Plasmonic HMNs Arrays
As a proof-of-concept, the specific biotin-streptavidin recognition was chosen as an analyte-receptor interaction system to validate the proposed diagnostic assay (Figure 4a). First, the HMNs filled with the plasmonic transducer were soaked overnight in MilliQ water to ensure the swelling equilibrium and the accurate removal of non-crosslinked moieties. At this stage, the LSPR peak, denoted as max , was located around 541 ± 1 nm, with a slight blue shift compared to the freshly synthesized plasmonic nanocomposite. The blue shift was due to the decrease in the medium effective refractive index since water possesses a lower refractive index than PEGDA. Then, AuNPs embedded within the high Mw PEGDA were chemically modified directly inside the HMNs cavities with cysteamine (cys) to obtain cysmodified AuNPs (cys-AuNPs). To do this, a filled HMNs patch was soaked into a solution of cys at a concentration able to fully cap the AuNPs surfaces with amino groups, by linking thiol groups AuNPs. The chemical modifications of the AuNPs were monitored, after accurate washing steps, by measuring the absorption spectra of the single HMNs (n ≥ 15) on a minimum of three different HMNs arrays (n ≥ 3). Differently from what was reported elsewhere, [40] [40] cysteamine-modified AuNPs exhibited a redshift in the LSPR absorption peak from about 541.3 ± 1.4 nm to 543.6 ± 1.0 nm. The redshift observed herein is justified by a novel concept: "confined" swelling. In fact, it has already been shown that hydrophilic molecules such as cys cause high Mw PEGDA/AuNPs nanocomposites absorption decrease due to an increase in the swelling ratio equilibrium, without significant redshift of the LSPR. [40] In high Mw PEGDA hydrogels embedding AuNPs, no significant redshift was observed after cysteamine functionalization. This does not mean that cys molecules had not capped the AuNPs. In the free high MW hydrogel, a cys concentration-dependent variation of the absorption intensity was observed. Indeed absorption intensity undergoes a significant decrease due to the intrinsic hydrophilic nature of cys molecules. Therefore, more water permeates the hydrogel and a decrease in the effective refractive index of the material is observed. This means that the redshift caused by the capping of cys molecules on the AuNPs is counterbalanced by the blue shift caused by the enhanced swelling of the nanocomposites. [40] Differently, when the nanocomposite is not free but confined in  10). c) Label-free calibration curve achieved by monitoring the mean LSPR wavelength as a function of the biotin concentration. A 4-parameters logistic curve was used to fit the data (red dashed line). d) Fluorescence and dark field (DF) images of HMNs bases after interaction with Cy3-STV at a fixed concentration. The images were collected for four different biotin concentrations (1 mm, 100 μm, 10 μm, and 1 μm, respectively) (n ≥ 3). Scale bars are 250 μm. e) Mean fluorescence intensities as a function of the biotin concentration. A 4-parameters logistic curve was used to fit the data (green dash-dotted line). In the inset, the fluorescence intensity obtained from a 1 mM biotin concentration is compared with the control signals obtained from HMN without cysteamine, and without biotin, respectively. SDs are reported as vertical bars. an MN, it is possible to observe the redshift caused by the cysteamine since the amount of water that effectively permeates the hydrogel is physically limited by the MNs. In this case, high Mw PEGDA was confined within a sub-millimetric cavity and exhibited a swelling up to the complete filling of the HMN cavity. Af-ter that point, cys diffused within the hydrogel and caused an increase in the medium effective refractive index with a subsequent LSPR redshift. Therefore, not only a decrease in the absorption intensity was observed, but also a redshift of the LSPR was achieved. A plasmonic redshift is observed also with the subsequent sensing step. After the capping of AuNPs with cys, the filled HMNs were soaked into sulfo-NHS-Biotin at different concentrations from 0.001 to 1 mm for 2 h (Figure 4c). Absorption spectroscopy was performed on the HMNs to monitor the LSPR wavelength shift as a function of biotin concentration to achieve a calibration curve for the label-free sensing of a small molecule model as biotin (Figure 4c). In the chosen range, a plasmonic redshift was observed as a function of the biotin concentration, following a four-parameters logistic fitting curve. Linearity of the plasmonic biosensor within the HMNs was observed in the range from 0.001 to 0.1 mm, thus covering two orders of magnitude. A final LOD of ≈20 μm was measured as reported in the experimental section. In Figure 4c, the gray box highlights the threshold between non-significant data comparable to the cys-AuNPs signal without biotin, and the significantly different signals. The LSPR wavelength shifts are reported as mean values ± SD on a minimum of ten measurements on three HMNs array replicates. The obtained result for label-free sensing of small molecules was already very competitive with most of the wearable biosensors reported in the literature and could be of great impact at the industrial level due to the ease of fabrication and use. [59,60] Moreover, glucose concentration, in ISF, ranges from 100 μm to 1 mm, [61] which is compatible with the achieved LOD. The linear range in label-free mode is sufficient enough for the proposed proof of concept. In fact, glucose concentration in ISF falls in the same concentration range. However, for some specific biomarkers found in the ISF, the critical and clinically relevant concentrations could be much lower than the micromolar range. In this context, it is of paramount relevance to amplify the signal obtained from the interaction of the few collected molecules on the selected transducer. To do this, the use of labels, e.g., fluorescent dyes, is highly recommended to increase the sensitivity of the transducing element and achieve lower LODs. To this aim, the strong electromagnetic field enhancement in the surroundings of the plasmonic AuNPs was coupled with Cy3 fluorescent dye labeling streptavidin (STV) molecules. STV was used at a fixed concentration and was selectively incubated within the HMNs array, in which biotin at the different reported concentrations had been incubated. Therefore, STV-biotin affinity was exploited to verify the effective recognition of biotin in solution by the plasmonic nanocomposite embedded within the HMNs cavities, but also to increase the plasmonic sensor sensitivity and performance. First, active sites that were not bound by biotin were passivated with a suitable concentration of BSA to avoid non-specific binding of Cy3-STV on the HMNs. Then, a very tiny concentration of 29 μg mL −1 of Cy3-STV was used to indirectly detect biotin on AuNPs. The coupling of the Cy3 and the AuNPs within HMNs cavities (MEF) occurred by the FRET mechanism since the average spatial distance between the AuNPs and STV was lower than 10 nm. In fact, biotin-STV complexes possess, on average, a size of ≈5 nm. [62] [62] MEF enabled the use of very low amounts of STV, thus reducing analytical costs and incubation volumes. Fluorescence signals were collected by fluorescence microscopy. First, the focus was adjusted on the MNs bases to collect the epi-fluorescence signal from the whole HMN, then the images were acquired and analyzed by ImageJ software. Exposure time, gain, and analysis modalities were kept constant for the duration of all the experiments. Some representative images collected both in the dark field mode and in fluo-rescence mode, at different biotin concentrations, are reported in Figure 4d. As evident, the fluorescence signal gradually decreased upon biotin concentration decrease. The HMNs walls also resulted in a really bright fluorescence signal, which can be explained by the intrinsic nature of fluorescence, which exhibits maximum signal intensity in a direction perpendicular to the laser line incidence. [63] [63] A minimum of ten fluorescence images was acquired per each HMNs array at the different biotin concentrations (in triplicates). The corresponding fluorescence intensities were evaluated as reported in the experimental section, and the results are reported in Figure 4e. Also in this case, a 4-parameter logistic regression was used to fit the mean fluorescence intensity values as a function of the biotin concentration. However, in this case, a LOD of ≈350 nm was achieved by gaining two orders of magnitude in both linear range and minimum detectable biotin concentration. The LOD in label-based mode was evaluated starting from the fitting curve reported in Figure 4e. As expected in this case, the gray box, dividing the significantly different signal intensities from the non-significant ones, resulted to be significantly smaller than the label-free calibration. Although the obtained results are already very competitive with wearable sensing devices, further optimization could be performed to reduce the obtained LODs both in label-free and in label-based sensing mechanisms. For example, in MEF-based sensing, it could be possible to increase the exposure time, since the non-specific interactions were really well hindered by BSA, as shown by the control experiments reported in the insets of Figure 4e, and whose corresponding fluorescence images are shown in Figure S4 (Supporting Information). Briefly, plasmonic HMN patches without cys modification of AuNPs and without biotin were used as independent controls. However, the optimization of the calibration sensing curves of the proposed platform in solution was out of the scope of this paper since the potentialities of combined LSPR and MEF in plasmonic nanocomposites made of AuNPs and PEGDA have been already reported in previous papers. [39,40]

Extraction and Detection of STV from a Skin Model: A Proof of Concept
The main aim of this paper was to demonstrate the capability of HMNs, integrated with an optical plasmonic transducer, to collect and efficiently detect a target molecule from the ISF starting from a skin model. As proof of concept, the plasmonic HMNs were used to extract the biotin from a skin model. The capture of HMNs after indentation in agar was proved both in label-free (LSPR) and in label-based (MEF) sensing mechanisms. To perform the analysis, a two-layers skin model, composed of agarose hydrogel, containing a 1 mm biotin, which was chosen from the previously reported calibration curves, and agar with a top layer of parafilm, exploited to simulate the SC, was prepared. The HMN patches were pressed on the skin model containing biotin for 3 h and then removed, as shown in Figure 5a. The reaction time was chosen, based on the previously reported results that were performed in solution with the same timing. Since HMNs arrays are painless, they should meet patient compliance. HMNs were able to penetrate the skin models, as already proved in Figure 1e and Figure S1 (Supporting Information), and, as reported in Figure 5. a) Schematic representation of the compared functionalization protocols: HMNs array in 1 mM biotin solution and HMNs array indentation in a skin model made of Agar containing biotin at the same concentration covered with parafilm. b) Mean normalized absorption spectra of single-filled HMNs after incubation with biotin (1 mM) in solution and after indentation in agar (black and red solid lines, respectively); the controls (black and red dashed lines, respectively) are performed with non-functionalized AuNPs (without cys) (n ≥ 3); on the right, histograms of mean LSPR wavelengths of plasmonic HMNs functionalized with biotin in solution and in agar with their relative controls. c) Fluorescence images of HMNs bases after interaction with Cy3-STV at a fixed concentration (n ≥ 3). The images were collected for a 1 mm biotin concentration incubated in solution and in agar with their relative controls. Scale bars are 250 μm. On the right, mean fluorescence intensities at fixed biotin concentration and STV concentration in solution and in agar, and their relative controls, are reported. SDs are reported as vertical bars. ****p < 0.0001 and ***p < 0.001 resulting from ANOVA tests were considered statistically significant. Figure 5, biotin was successfully collected and detected from the skin model. After HMNs extraction, and after accurate washing steps to remove unbound molecules and agar residues, absorption spectra of the filled HMNs were recorded. On average, an LSPR wavelength redshift comparable to the one reported from HMNs in solution was observed, confirming the extraction and interaction of biotin with cys-AuNPs. HMNs without cys were used as a control. Interestingly, the control was comparable with the control performed in the solution. Only a slight redshift, attributed to non-specific agar residues, was observed (Figure 5b). To validate this result, but also to prove the capability of the proposed platform to work in the MEF sensing mechanism, three arrays of HMNs after indentation in agar with biotin were soaked in a Cy3-STV solution (0.0029 μg mL −1 ). This step was performed after careful passivation of the sensor surface with BSA. From the results shown in Figure 5c, it was definitely evident that the biotin contained in the agar was efficiently extracted by the plasmonic HMNs. Also, in this case, the results after image analysis resulted to be consistent with the ones obtained in solution. As a control, HMNs filled with bare plasmonic nanocomposites (without cys) were indented in agar containing biotin and soaked in a Cy3-STV solution. With this sensing mechanism, the proposed platform indented in agar resulted to be as statistically significant as in solution since, even though some agar residues could be present after incubation. They did not show any auto-fluorescence when excited at 530 nm. Therefore, also after indentation in agar, a ≈30 times higher fluorescence signal than HMNs without cys-AuNPs was observed (p-value < 0.0001). Differently, in LSPR mode a bit of statistical significance (the p-value passed from < 0.0001 to < 0.001) was lost since agar residues were partially stacked in the surroundings of AuNPs. Despite this slight variation, overall, the results, in the model skin, resulted to be comparable to those obtained by the sample soaking. The high affinity of the biotinstreptavidin binding made this system an attractive model for validating the capability of HMNs to extract molecule targets from a complex matrix. However, this was only proof of concept paving the way to more clinically relevant target/receptor testing, such as glucose/glucose oxidase, antigen/antibody, and so on, which can be found also in ISF. The proposed device, relative to other methods that required ISF extraction, [22,64,65] exploited the HMNs capability of extracting and localizing tiny volumes of solution (e.g., ISF), and directly analyzing it by simple spectroscopic/optical techniques.
The integration of a plasmonic nanocomposite in an HMNs array, herein presented, was designed to measure biologically relevant analytes in the ISF based on the selectivity of biorecognition elements. Starting from the biotin-streptavidin model system, and based on the achieved results, it is possible to deduce that a biomarker from ISF could in principle flow through the HMNs conduits by capillary forces, interact with the embedded plasmonic nanocomposites, and be detected directly through the single HMN. After the system insertion into the skin model, the selective analyte-receptor interaction could be translated by the plasmonic transducer into a measurable optical signal based on LSPR and MEF mechanisms, whose intensities could allow the quantification of the target analyte.

Conclusion
An HMNs array integrating plasmonic nanocomposite transducers for on-patch LSPR and MEF-based optical detection of molecules from ISF was presented. The mechanical properties of the HMNs conferred the capability to break the SC barrier and penetrate the skin. The plasmonic nanocomposite transducer based on a high Mw PEGDA and highly concentrated spherical AuNPs was designed and fabricated to be operated in dual-optical modes. The integrated plasmonic HMNs device sensing performances were tested in solution by using biotin-streptavidin interaction, acting as a target/receptor coupled system, respectively. Finally, the capability of the proposed device to collect and capture the biotin target molecule from the skin was tested by using a skin model made of parafilm and agar (SC and dermis, respectively). Biotin, as a target, was successfully retrieved and optically detected in both label-free and fluorescence-based sensing mechanisms, thus demonstrating the functionality of the proposed platform. By basing on these results, plasmonic HMNs can represent a starting point to develop a simple, low-cost, large-scale, and versatile strategy for point-of-use (PoU) testing devices for patient monitoring of biomarkers from ISF, as opposed to conventional, expensive, and laborious hospital or lab settings. Further improvements in the label-free detection of small molecules with the proposed device could be achieved by exploiting differently shaped nanoparticles such as nanoprisms, nanotriangles, and/or nanostars, in which tip-shape phenomena are exploited or by increasing the transducer volume thus enhancing the collected ISF volume. These devices are receiving increasing attention due to population aging, lack of sufficient specialized personnel, and availability of places in hospitals. Therefore, the quest for PoU devices is becoming more and more urgent, and the proposed approach could pave the way to the satisfaction of this need.

Experimental Section
General Information: All chemicals were commercially available and used as received. Poly (Ethylene Glycol) Diacrylate (PEGDA), with average molecular weight, Mw = 250 Da and 10 000 Da, Methylene blue were purchased from Sigma-Aldrich, St. Louis, MO, USA. Bacto Agar was purchased from BD.
HMNs Fabrication: HMNs were fabricated by a one-shot fabrication process using a photolithography-based method. [32,44] HMNs were obtained using a photolithographic mask (JD Photo Data) containing an array of rings made of an outer transparent circle with a diameter of 750 μm and a dark inner circle of 500 μm. The centers of the two circles were not aligned and presented a center-to-center distance of 30 μm. PEGDA www.advancedsciencenews.com www.advmattechnol.de prepolymer solution (Mw 250 Da) was mixed with 2 v/v % of Darocur and stirred for 15 min at 300 rpm. The mixture (200 μL) was dropped on a glass slide and covered by another slide, then exposed to a high-power UV light source (365 nm) for 2.5 s using a Mask Aligner (MA6/BA6 by SÜSS MicroTec AG, Garching, Germany).
This resulted in a thin film acting as an HMN backing layer, which was referred to as a patch. Subsequently, the glass slides were removed, and the patch was placed to cover a glass vessel containing the PEGDA (Mw 250 Da) prepolymer solution. The vessel was aligned with a photomask and exposed to UV light for 9 s. Fabricated HMNs were treated with Milli-Q water and Isopropyl alcohol (IPA) to remove the residual prepolymer mixture. HMNs patches were post-reticulated by heat treatment for 3 h at 100°C.
HMNs Mechanical Characterization: Microcompression tests were performed using a DEBEN 200N stage for HITACHI TM3000 Scanning Electron Microscope. This test was performed to analyze the mechanical resistance of HMNs when subjected to compression. The test was carried out in real-time monitoring the applied force and the displacement. Each measurement was conducted by approaching the metal plate to the tips of HMNs strip of 2 MN and applying the compression load until 2 N.
Numerical Simulations: The numerical simulation of HMNs insertion into the skin was performed using an axial force on the base of HMN. [32,[66][67][68] COMSOL Multiphysics was used for the simulation and structural analysis of the HMNs. The simulations were performed on a single HMN to evaluate the penetration depth in a multilayered skin model. HMNs geometries having an a-symmetric cavity were reproduced by using AUTOCAD. The obtained geometry was subsequently imported into COMSOL. The abovementioned photolithographic mask, consisting of two circles (750 and 500 μm in diameter respectively) with a center-tocenter distance of 30 μm, was used to reproduce the HMN base. The single HMN structures are modeled using the mechanical properties of PEGDA (Young's modulus 500 MPa, Poisson's Ratio 0.3, density 1.11 g cm −3 ). The skin geometry is made by a composite cylindrical bilayer made of SC and dermis, whose Young's moduli are 26 MPa and 136 KPa, respectively, and their Poisson's Ratio is set at 0.49. [69,70] The thicknesses of the two layers are set at 100 and 1000 μm, respectively. The COMSOL Structural Mechanics Module was used to simulate the contact between the HMN and skin model to provide an estimation of the maximum MN depth penetration.
HMNs Penetration in Pig Ear Skin: Pigskin was used fresh, without any treatment. Micro-indentation tests were performed by pressing the patch of HMNs on the pigskin with the thumb. The skin was stretched and anchored to a support. Before the test, the HMNs arrays were immersed in a black ink solution to subsequently stain the skin punctures created by the HMNs penetration. The HMNs patch was slowly pressed on the skin for 1 min and then gently pulled out from the porcine ear skin. Finally, the skin was observed under a stereomicroscope (Leica DM6) to examine the sites of HMNs indentation.
Histological Analysis: Furthermore, histological analysis was conducted to evaluate the insertion depth of the HMNs. The skin was embedded in Optimum Cutting Temperature (OCT) compound and left to freeze in liquid nitrogen. After micro-CT analysis, 20-μm-thick sections were obtained by using a cryo-cut microtome (Slee Cut 6062, Slee Medical, Mainz, Germany). Sections were stained with hematoxylin-eosin (H&E). Whole slide images (WSI) were acquired using a stereomicroscope (Leica DM6).
AuNPs Synthesis and Characterization: Highly concentrated AuNPs with controlled size were chemically synthesized by modifying the seededgrowth method reported by Nikoobakht and El-Sayed. [55] The protocol was modified to achieve a high concentration of spherical AuNPs to be used as produced and to avoid the long centrifugation steps required to achieve the desired concentration. Growth and seed solutions were prepared separately as described below.
Growth Solution: CTAB solution (5 mL, 0.20 m) was stirred at 40°C at 550 rpm for 10 min until the complete dissolution of CTAB. Then, AgNO 3 solution (200 uL, 40 mm) was quickly added followed by HAuCl 4 (5 mL, 10 mm). Finally, AA (70 μL, 0.0788 m) was added. The initial intense orange solution became colorless quickly after AA addition, which acts as a mild reducing agent.
Seed Solution: CTAB solution (5 mL, 0.20 m) was mixed with 5.0 mL of 5 mm HAuCl 4 . To the stirred solution, 300 μL of ice-cold 100 mm NaBH 4 was added, resulting in an intense brownish solution.
At the end, 12 μL of the seed solution was added to the growth solution at 27-30°C. The color of the solution changed from nearly transparent to intense violet, indicating the formation of highly concentrated Gold Nanoparticles. Z-potential measurements and UV-vis spectroscopy were performed on the freshly synthesized AuNPs. The AuNPs Z-potential was measured by Zetasizer Nano ZS (Malvern Instruments, UK). The UV-vis absorption spectra were acquired using UV-vis spectrometer (Cary 100 spectrometer -VARIAN), from 400 to 900 nm.
Scanning Electron Microscopy: Morphological analysis of AuNPs was performed by means of scanning electron microscopy (SEM) (Quanta 200 FEG, 338 FEI, Eindhoven, The Netherlands). The AuNPs were observed in high vacuum mode using a secondary electron detector and an accelerating voltage of 30.0 kV. The imaging of AuNPs suspensions was performed by drop casting 5 μL of concentrated AuNPs on a glass. The suspension was left overnight to allow the complete evaporation of the solvent, prior to imaging. The particle analysis from SEM images was performed by Im-ageJ Software by applying a threshold to obtain a binary image, in which the mean area of the obtained AuNPs was measured. A histogram was obtained by using a 5 nm bin size and a gaussian curve was used to fit the data (Figure 2b). The statistical analysis was made on three different images (n = 3).
Plasmonic Nanocomposite Fabrication and Integration into HMNs: The plasmonic nanocomposite was obtained from a prepolymer solution made of PEGDA 10 kDa (200 μL, 400 mg mL −1 ), Darocur 1173 (8 μL, 4% w/v) as a photoinitiator, and freshly synthesized AuNPs (2 × 10 12 NPs mL −1 , 100 μL). PEGDA and the photoinitiator were stirred vigorously (1000 rpm) for 5 min before the addition of AuNPs, which were gently stirred (300 rpm) for 5 min more. The thin patch used for swelling ratio and AuNPs stability evaluation, and optical and MEF analysis was obtained by drop-casting 270 μL of the prepolymer solution on a glass coverslip (24 mm × 24 mm), closing with another glass coverslip, and UV-light irradiation (UV-exposure box -Belichtungsgerät 2) to polymerize the solution between the two coverslips. The polymerization time was set to 5 min. The integration of the plasmonic nanocomposite transducer was achieved by pouring the prepared prepolymer solution in the holes of the MNs under an optical microscope. The prepolymer solution was injected into the HMNs by means of a 1 mL syringe with sharp tips. After the removal of solution excess, the HMNs were exposed to UV light for 2 minutes to allow the complete polymerization of the plasmonic nanocomposite.
Absorption Spectroscopy: The HMNs arrays filled with plasmonic transducers made of high MW PEGDA and concentrated AuNPs were characterized by a customized transmission mode setup at normal incidence (Figure 3e). A halogen light source covering the visible and near-IR spectrum (400-900 nm) was connected to a Thorlabs optic fiber conveying light to the sample through a collimator. Prior to the sample, a biconvex lens was added to tune the spot size from 5 mm (for the integral thin film patch) to 500 μm (to focus on a single HMN), as reported in Figure 3f. The transmitted light was collected by another biconvex lens and sent to another Thorlabs optic fiber connected to a spectrometer (Filmetrics F20). The self-standing HMNs arrays were placed in between the two lenses, while the thin film patches for preliminary analysis were placed on a coverslip. The appropriate background was chosen, acquired, and subtracted, accordingly. The spectra were collected in the range of 400-900 nm with a resolution of ≈0.5 nm. The absorption of the plasmonic transducer patch was evaluated by using a bare polymerized PEGDA without AuNPs as a reference, while, for the HMNs, needles filled with PEGDA without AuNPs were used as a reference. The raw absorption spectra were smoothed to remove noise and the absorption peak position was measured. A minimum of three spectra was acquired per HMN and the measurements were performed on the whole array to achieve statistical significance. Peak analysis on all the HMNs arrays was performed and was reported as LSPR wavelength ± SD. The LOD was evaluated as three times the SD of the LSPR position in the control (blank) samples in absence of biotin.
Optical/Fluorescence Microscopy and Image Analysis: The HMNs arrays were observed using a stereomicroscope (Leica DM6) to record the HMNs geometric features. Fluorescence images were acquired by using a Leica AF6000LX-DM6M-Z microscope (Leica Microsystems, Mannheim, Germany), controlled by LAS-X (Leica Application Suite; rel. 3.0.13) software and equipped with a Leica Camera DFC7000T. Both dark field (DF) and Fluorescence images were acquired by focusing on the base of the HMNs with a 10× objective. Fluorescence images were obtained by using a Quadband LED filter cube and setting the excitation filter at 535(25) nm, the dichromatic mirror at 555 nm, and the suppression filter at 575(30) nm. Exposure time and gain were set at 300 ms and 1.0, respectively. Before imaging, the excess water from the HMNs arrays filled with the plasmonic transducer was removed. A minimum of three images was acquired on three independent samples. The obtained images were analyzed by the ImageJ freeware software by measuring the fluorescence intensity values from 9 fixed regions of interest (ROI = 300 pixels × 300 pixels) per HMN. Mean fluorescence intensity values ± SD were calculated starting from the fluorescence images of all the HMNs available in each array. The LOD was evaluated as three times the SD of fluorescence intensity in the control (blank) samples in absence of biotin.
Swelling Ratio Measurements: To measure the swelling ratio, i.e., the profile of the swelling capacity versus time of a hydrogel sample, absorbency capacity measurements at consecutive time intervals were performed. The mass swelling ratio (q) is measured as: where w d is the initial weight of the hydrogel (dry mass) and w s is the swollen hydrogel weight at different times. The as-synthesized hydrogel nanocomposites (n ≥ 3) embedding AuNPs were weighed to consider the initial amount of water contained in them. This first measurement was selected as t = 0 h, as reported in Figure 2e. Then, the hydrogels were dipped in an exceeding amount of MilliQ water for 12 and 24 h and weighed. Prior to each weighing step, the nanocomposites were placed on a dry cloth and gently wiped with another dry cloth to remove excess and weakly bound water. Finally, the hydrogel nanocomposites were put in a vacuum oven at 40°C overnight to achieve their complete drying and measure the w d value.
HMNs Penetration in Agarose: Agarose (agar) solution (1.4 wt.%) was prepared by dissolving agarose in sterile Milli-Q water by gentle stirring (30 min, 120°C) in a water bath. After dissolution, Methylene blue (10 mg) was added for the colorimetric (naked-eye detection) testing of the swelling capability of high Mw PEGDA when confined in the HMNs cavities ( Figure 3d). Moreover, the same agarose was mixed with biotin to achieve a final biotin concentration of 1 mm within the agar. Both agarose solutions (containing MB and biotin, respectively) were cooled at room temperature and solidified.
Functionalization of the Plasmonic HMNs Array and Biotin Sensing: Before their use, the HMNs arrays filled with the plasmonic nanocomposite transducer were soaked in MilliQ water for 24 h to remove the excess of Darocur and unreacted PEGDA, and to achieve the swelling equilibrium, as shown in Figure 2e. The functionalization of the AuNPs within PEGDA 10 kDa filling the HMNs array was performed by soaking the whole arrays into a water solution of cysteamine (1 mm, 1 mL) overnight to allow the thiol groups to interact with gold ( Figure 4a). The chosen concentration was already shown to be sufficient to fully cap the AuNPs within the hydrogel. [39,40] Then, three washing steps of 10 min at RT were performed to remove the unreacted moieties: one in a Tween20 aqueous solution (0.02% v/v) and two in water. The absorption spectra were measured and the LSPR wavelength was collected for each HMN. Then, the interaction of cysteamine-capped AuNPs with biotin was achieved in three different ways: in solution, in agar, and in agar with parafilm. In the first case, HMN arrays were soaked into biotin water solutions (1 mL) with concentrations ranging from 1 μm to 1 mm (2 h, at room temperature) to obtain a calibration curve. The same washing steps and optical analysis were performed to measure the absorption peak variations with increasing biotin concentrations. A four-parameters logistic type equation was used to fit the LSPR absorption peak redshift as a function of the biotin concentration. The LOD in label-free mode was obtained by considering three times the SD of the control point corresponding to the device functionalized with cysteamine (without biotin) divided by the slope of the linear fit obtained between 0.001 and 0.1 mm biotin concentrations. For label-based sensing, the plasmonic HMNs arrays were soaked in a BSA/PBS (1×, pH 7.4) solution (5 mg mL −1 , 1 mL) to reduce non-specific interactions (1 h at room temperature). Finally, fixed volumes (0.25 mL) of PBS (1×, pH 7.4) solutions of Cy3-STV (29 μg mL −1 , 50 nm) were incubated in the biotinylated HMNs arrays (2 h, at room temperature). The dose-response curve from the measurement of the fluorescence intensities of Cy3-STV in HMN, as a function of the biotin concentrations and after the washing steps, was obtained by the four-parameters logistic fit. The negative controls were performed by incubating Cy3-STV (50 nm), in absence of biotin, and in absence of cysteamine (2 h at room temperature, respectively).
Statistical Analysis: Mean absorption spectra were calculated as the average of 10 spectra (n ≥ 10) collected on different HMNs on three different samples treated in the same way. The plots of the absorption spectra were obtained by smoothing and, when specified, normalization by using OriginPro 8 free version. The LSPR peak positions max as a function of cysteamine, biotin, and streptavidin concentrations are reported by performing peak analysis on the smoothed absorption spectra. The mean ± SD is representative of at least three independent experiments (n ≥ 3). For fluorescence analysis, a minimum of three images on three independent samples was acquired in the same conditions (n ≥ 3). Fluorescence intensity was evaluated by using ImageJ software. Mean fluorescence intensities were analyzed and normalized by using OriginPro 8. The obtained data are reported as mean ± SD. The difference in data between groups for specificity tests was analyzed by ANOVA by using OriginPro 8 free version. p < 0.001 was considered statistically significant. Unless otherwise stated, the other data were expressed as mean ± SD. SDs are reported as vertical bars and are representative of at least three independent experiments (n ≥ 3).

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