Preventing Memory Effects in Surface-Enhanced Raman Scattering Substrates by Polymer Coating and Laser-Activated Deprotection.

The development of continuous monitoring systems requires in situ sensors that are capable of screening multiple chemical species and providing real-time information. Such in situ measurements, in which the sample is analyzed at the point of interest, are hindered by underlying problems derived from the recording of successive measurements within complex environments. In this context, surface-enhanced Raman scattering (SERS) spectroscopy appears as a noninvasive technology with the ability of identifying low concentrations of chemical species as well as resolving dynamic processes under different conditions. To this aim, the technique requires the use of a plasmonic substrate, typically made of nanostructured metals such as gold or silver, to enhance the Raman signal of adsorbed molecules (the analyte). However, a common source of uncertainty in real-time SERS measurements originates from the irreversible adsorption of (analyte) molecules onto the plasmonic substrate, which may interfere in subsequent measurements. This so-called "SERS memory effect" leads to measurements that do not accurately reflect varying conditions of the sample over time. We introduce herein the design of plasmonic substrates involving a nonpermeable poly(lactic-co-glycolic acid) (PLGA) thin layer on top of the plasmonic nanostructure, toward controlling the adsorption of molecules at different times. The polymeric layer can be locally degraded by irradiation with the same laser used for SERS measurements (albeit at a higher fluence), thereby creating a micrometer-sized window on the plasmonic substrate available to molecules present in solution at a selected measurement time. Using SERS substrates coated with such thermolabile polymer layers, we demonstrate the possibility of performing over 10,000 consecutive measurements per substrate as well as accurate continuous monitoring of analytes in microfluidic channels and biological systems.

Surface-enhanced Raman scattering (SERS) spectroscopy is a highly sensitive vibrational spectroscopy technique that facilitates the identification of trace analytes and allows the multiplexed detection of their characteristic vibrational fingerprints.[1] On this account, SERS has emerged as a promising chemical monitoring method, with applications in various fields including biosensing,[2] food control,[3] and detection of hazardous materials,[4] among others. SERS relies on the plasmonic properties of noble metal nanostructures to enhance the Raman signal of adsorbed molecules. The confinement of light at nanoscale volumes by plasmonic nanomaterials is responsible for a dramatic increase in sensitivity, which can go as far as single-molecule detection.[5]

Key features of SERS are its noninvasive character and label-free detection, which promise the potential of implementing in situ measurements, for example, in the clinic or in the field. However, to achieve in situ monitoring, not only further development in portable Raman equipment will be required, but also long-term SERS detection in real time also constitutes a complex challenge, toward acquiring chemical information on a probe solution at arbitrary time points.[6] An example of application would be the implementation of real-time SERS measurements in flow for water quality control.[7,8] On the other hand, SERS detection of biomolecules released by living organisms provides highly valuable information, for example, on their cellular state and function.[9] Thus, real-time measurements are likely to have significant implications in a wide variety of fields, from medical diagnosis to the biotechnology industry. However, a major drawback arises when molecules adsorb strongly on the surface of the plasmonic substrate,[10,11] so that their corresponding Raman signal would remain, even at later stages of incubation with other analytes, thereby interfering in subsequent measurements. Such a “memory effect” hampers real-time detection using standard SERS strategies, where analyte solutions are continuously in contact with the plasmonic substrate. Thus, a common approach to monitoring changes in the chemical composition of a solution comprises the use of a freshly made substrate every time a measurement is to be performed. Even if it works, this strategy does not allow continuous in-flow SERS measurements without operator intervention or automation. Obviously, using a different SERS substrate for every measurement imposes a heavy economic handicap, particularly when measurements for extended periods of time or detection of different analytes are required.

Different solutions have been proposed to render SERS substrates reusable and thus eligible for real-time sensing. The main strategy consists of cleaning the substrate to remove molecular adsorbates. Some cleaning techniques based on physicochemical treatments, such as UV-ozone, have allowed the reuse of plasmonic substrates for SERS measurements.[12] Other approaches involve the incubation of the plasmonic substrate with different solvents to alter electrostatic interactions so that the plasmonic system can be recycled.[13] Such techniques are interesting, but they do not allow for in situ measurements, as the substrates must be removed from the solution of interest for each cleaning process. Other innovative examples use photocatalytic materials, such as ZnO or TiO2, in combination with the plasmonic substrates to degrade sulfur bonds by photocatalysis.[14,15] On the other hand, Belder and co-workers have shown in situ real-time SERS measurements in microfluidic channels by electrical regeneration of silver wires.[16] Whereas these methods enable the monitoring of chemical species in real time, they can hardly be translated into efficient plasmonic systems, and long periods are required for cleaning, which slow down the detection process and may affect the quality of the substrate. A different strategy, followed by Gao et al., comprises the in situ synthesis of plasmonic nanoparticles (NPs) inside a microfluidic channel, from where the SERS signal is recorded. This means that a synthetic effort must be made for each measurement point. While resolving the SERS memory effect, such an approach requires a high consumption of material and leaves “useless” NPs in the microfluidic outlet, causing an additional drawback for real-time sensing.[17,18]

We propose herein an alternative concept based on covering the entire plasmonic surface with a thermolabile sheathing layer, which allows for not only long-term monitoring through the identification and/or quantification of analyte(s) but also efficient spatial and temporal control. On a selected plasmonic substrate, we spin-coated a layer of poly(lactic-co-glycolic acid) (PLGA), which has been broadly adopted for multiple applications due to its biocompatibility, low cost, and ease of handling.[19,20] Gupta et al. additionally demonstrated that site-selective PLGA degradation can be achieved by laser irradiation.[21] Photodegradation was achieved by first embedding gold nanorods (Au NRs) within the polymeric material so that PLGA was disrupted upon photothermal heating when Au NRs were irradiated with a resonant laser.[22] Other studies have demonstrated that the photothermal effect can be exploited to alter the permeability of different polymers, such as poly(N-isopropylacrylamide) (pNIPAM), upon NIR irradiation.[23]

We hypothesized that the deposited PLGA layer should protect the SERS substrate by preventing adsorption of analyte molecule(s) from solution. Upon laser irradiation at a sufficiently high fluence, the plasmonic photothermal effect would lead to formation of a micron-sized hole in the PLGA layer, while the plasmonic substrate itself remains intact. The so-created hole would act as an open window in the PLGA layer, rendering the plasmonic substrate underneath accessible to the target analyte(s) at that particular spot. By repeated irradiation at different locations (and different times), multiple windows can be opened at will so that freshly exposed areas of the plasmonic substrate are used each time for SERS measurements and chemical analysis of the solution. The versatility of this technique can be exploited for multiplexing experiments and real-time measurements of fluids within microfluidic channels, including the monitoring of metabolite fluctuations in biological systems or bioreactors.

Results and Discussion

Plasmonic Superlattices Coated with PLGA for SERS Sensing

The design and optimization of plasmonic substrates has attracted much interest toward the development of sensors with outstanding SERS performance.[24,25] Such substrates are expected to provide high and uniform near-field enhancements so that intense SERS signals can be registered from arbitrary spots, even for low concentrations of the target analytes.[2,26,27] However, sensitivity enhancement is often accompanied by compromised reproducibility, multiplexing ability, and reusability in practical applications. In this context, whereas the optimization of enhancement factors has been extensively studied, issues with real-time measurements and reusability have not been resolved so far. Hence, although current technology enables the ultrasensitive detection of multiple analytes,[28] its practical implementation for continuous monitoring remains limited. We introduce herein a simple methodology that allows the application of state-of-the-art SERS substrates as long-term detection platforms for in situ sensing applications (see Scheme ).

Scheme 1

Requisites for Real-Time SERS Substrates

An ideal SERS-based system would allow for continuous monitoring of chemical variations in solution, with time and space resolution.

We aimed at providing additional features to recently developed plasmonic substrates, comprising regular arrays of clusters made of hexagonally packed nanospheres, also known as plasmonic superlattices. Such plasmonic substrates have been optimized for efficiently enhancing the SERS signals of adsorbing analytes.[29] However, they also suffer from the above-described shortcomings, related to reliable measurements over time. This issue is illustrated in Figure . Upon incubation with a 100 μM solution of 4-mercaptobenzoic acid (4-MBA), the characteristic 4-MBA vibration at 1078 cm–1, corresponding to the ν12(C–C) ring stretching mode[30] (red highlighted region in Figure c), was unequivocally detected. However, the SERS memory effect did not allow us to reuse the same plasmonic superlattice, since after extensive rinsing with water and subsequent incubation with nicotinamide (NAm), the 4-MBA peak still dominated the SERS spectrum, whereas no peak associated with NAm (expected at the position marked by the orange bar in Figure c) could be identified. Similarly, when adding a mixture of 4-MBA + NAm after 4-MBA incubation, NAm could not be detected. We interpret these results in terms of 4-MBA molecules remaining anchored on the superlattice substrate, even after rinsing with water; such an irreversible adsorption of 4-MBA not only leads to an intense SERS signal but also impairs the adsorption of NAm molecules onto the NPs comprising the plasmonic substrate. A similarly deficient performance was repeatedly observed with different types of plasmonic substrates, analytes, and concentrations, including various biomolecules (see Figure S1), which indicates a widespread impact of the observed phenomenon.

Figure 1

(a) Scheme of an uncoated plasmonic substrate: The adsorbed (red) analyte prevents the attachment of a subsequently added (green) analyte. (b) Representative SEM image of a plasmonic superlattice. (c) SERS spectra from a superlattice, upon the sequential addition of 100 μM solutions of 4-MBA, nicotinamide, and a 50:50 mixed solution of 4-MBA and NAm. All measurements were performed with a 50× objective, 1 s acquisition time, and a maximum power of the 785 nm laser of 0.018 mW/μm2. (d) PLGA-SERS strategy: A plasmonic superlattice is covered with PLGA by spin coating; laser irradiation at a high fluence leads to local degradation of the PLGA layer, rendering the NPs exposed to analytes present in the solution; and finally, SERS is measured at a low laser fluence. (e) SERS spectra from a PLGA-coated superlattice incubated in 100 μM of 4-MBA solution, at each step described in (d), as labeled; the 4-MBA vibrational fingerprint is only registered after opening a measurement window by laser irradiation. (f, g) SEM images showing a hole created in the PLGA layer by laser irradiation (0.064 mW/μm2, 785 nm), at different magnifications. (h) SERS spectra from a PLGA-coated superlattice, upon the sequential addition of 4-MBA, NAm, and a 50:50 mixed solution of 4-MBA and NAm. Each SERS spectrum was recorded after addition of a new analyte solution, first creating a hole in the PLGA layer and then measuring SERS on the same spot.

As a solution to this commonly unwanted effect, and in general to the single-use limitation of SERS substrates, we propose the deposition of an impermeable polymer coating to protect the plasmonic substrate, which should act as a sheathing layer that can be readily disrupted by laser irradiation and photothermal degradation. Eventual removal of the polymer layer at the irradiation area would render the underlying NPs available for interaction with the probe solution, thereby avoiding potential interferences from previously adsorbed molecules. To this aim, we selected PLGA with a lactic/glycolic content ratio of 75:25 because of its well-described thermal degradation[31] and biocompatibility.[32] The deposition of a PLGA thin film on top of the superlattice was performed by spin-coating; a 12 wt% PLGA/ethyl acetate solution (200 μL) was spin-coated (1500 rpm; 30s) on top of the plasmonic superlattice. It should be noted that the nature of the solvent in the PLGA solution was found to strongly affect the degree of sample coverage: By using PLGA/acetone solutions, the superlattices were only partially coated by the obtained PLGA films, whereas PLGA/ethyl acetate solutions yielded homogeneous films with a uniform thickness of 1.5 μm over the whole substrate (as measured by cross-sectional SEM images, see Figure S2a,b). Homogenous PLGA coatings rendered the plasmonic substrates impermeable to analytes in solution, as evidenced by the absence of SERS signals from arbitrary spots on the sample, upon incubation with an adenosine solution (Figure S2d). On the contrary, incomplete coating of the plasmonic substrates by PLGA resulted in some impermeable regions, while others remained exposed to the probe solution, as exemplified by the corresponding SEM images and SERS spectra (Figure S2a,c). All samples were thus prepared using ethyl acetate as the solvent, thereby achieving complete and uniform coverage of the Au NP superlattices.

PLGA is a thermodegradable polymer, meaning that it can be degraded into glycolic acid and lactic acid upon heating.[33] In our setup configuration, heat is generated by photothermal conversion when the plasmonic superlattice is irradiated with an intense laser beam. Such a plasmonic heating effect leads to localized degradation of the PLGA layer at the illumination area, leading to local permeability and diffusion of dissolved molecules toward the underlying plasmonic substrate.[34,35] By using a laser excitation with an elliptical laser spot of 35 × 17 μm2 (Figure S3) and a high enough irradiance, an increase in local temperature can be generated that leads to degradation of the PLGA layer, precisely at the irradiated area. Although small variations in the number of particles per cluster might affect the exact local temperature around each cluster, considering the dimensions of the laser spot (much larger than single clusters), heating arises from an average contribution of many plasmonic clusters. Plasmonic heating was confirmed by means of an infrared camera, which showed a local temperature increase from 30 to 38 °C on the plasmonic superlattice, upon excitation with a 785 nm laser at an irradiance of 0.064 mW/μm2 for 1 min (Figure S4). No temperature increase was observed when the laser beam was focused outside of the plasmonic superlattice, thereby confirming that heating originated from photothermal effects at Au NPs and not from the laser alone. Please note however that, although the infrared camera can sense a temperature increase, such measurements do not reflect the actual temperature next to the NPs at the nanoscale, which has been reported to be considerably higher.[36−38] For example, Käll and collaborators used antistokes thermometry to show that bowtie antennas can increase the local temperature by more than 100 °C, resulting in the formation of nanobubbles when the bowties are immersed in water.[37]

The proposed SERS-PLGA strategy is schematically depicted in Figure d. As an experimental demonstration of this method, a plasmonic substrate covered with a PLGA (75:25) film was incubated with a 100 μM 4-MBA aqueous solution. The substrate was then irradiated with a 785 nm laser at 0.018 mW/μm2 for 1 s through a 50× objective (the same conditions for SERS measurements as in Figure c). This laser irradiance was not intense enough to induce a sufficient plasmonic heating that could degrade the polymeric chains and create a hole in the PLGA layer. As a consequence, no trace of 4-MBA signal was observed in SERS (spectrum i in Figure e). The sample was then irradiated with a higher laser power, 0.064 mW/μm2 for 1s, which created an open hole in the PLGA film due to polymer degradation by plasmonic heating. As soon as the hole was formed, the SERS signal from 4-MBA was recorded (spectrum ii in Figure e), due to molecular diffusion through the hole toward the exposed area of the Au NP superlattice. Although a consistent SERS signal of 4-MBA was recorded after 1 s of irradiation at 0.064 mW/μm2, laser illumination was prolonged until the SERS signal was stabilized, which was usually achieved after 5 s (see Video S1). The irradiance of the 785 nm laser was then lowered, back to 0.018 mW/μm2, and used to probe the detection of 4-MBA at a lower laser power (spectrum iii in Figure e). A SERS map of the substrate was generated by recording the intensities of the indicated 4-MBA peak at 1078 cm–1 in a region around the hole, confirming that the signal is recorded from the open window only (Figure S5). SEM imaging of the same area confirmed an opening in the PLGA film, with an elliptical shape of around 20 × 10 μm2, as illustrated in Figure f. This hole size is smaller than the laser spot size (35 × 17 μm2), likely due to the intensity decay of the laser power from the center of the spot outward. In contrast to holes created upon laser irradiation at 0.064 mW/μm2 for 5 s, a longer exposure time of 50 s was found to result in larger holes, which more closely match the laser spot size (Figure S3b). It should also be noted that slight changes of the laser focus may induce variations in the laser spot shape from an ellipse to a circle, which would in turn affect the shape of the hole (see Figure S3c,d). Higher magnification SEM images (Figure g) clearly reveal the lattice of plasmonic clusters underneath the PLGA film. No reshaping[39] or melting of the NPs was observed, which is essential to achieve an enhanced SERS signal.

Once the performance of the SERS-PLGA method was confirmed, we carried out the same sequence of incubations with different analytes, as described above for the uncovered plasmonic substrate (no PLGA sheathing layer). During the sequential incubation in 4-MBA, NAm, and a 50:50 4-MBA/NAm mixture, a specific measurement window was generated for each analyte at a different position of the PLGA-coated substrate. Using the PLGA-SERS strategy, we could monitor the sequential presence of the analytes, as shown in the corresponding SERS spectra (Figure h). The characteristic peak of 4-MBA around 1078 cm–1 was detected after incubation with a 4-MBA solution. Upon subsequent incubation with NAm and laser irradiation at a different spot, only the NAm vibration at 1032 cm–1, assigned to an aromatic ring bending,[40] was observed. Further irradiation and measurement in the presence of both 4-MBA and NAm (50:50) revealed the SERS signature of both analytes. We additionally confirmed that a similar performance of the PLGA-protected plasmonic substrate could be obtained when working with different pairs of analytes (an example is shown in Figure S6), in agreement with our initial hypothesis and validating the efficacy of the method in preventing memory effects in SERS substrates. It should be noted that the SERS analytical enhancement factor (AEF) of plasmonic superlattices was not affected when using the PLGA-SERS method (see Figure S7). Specifically, the AEF using PLGA-SERS with plasmonic superlattices was determined to be 3.38 × 106, which lies within the upper range of values reported in the literature.[41]

As described above, a specific irradiance threshold was required to reach a sufficient temperature for PLGA degradation and formation of the measurement window. Such a threshold strongly depends on the light-to-heat conversion efficiency of each specific substrate and thus on its optical absorption properties. For this reason, the selected plasmonic material, made of hexagonally packed Au NP clusters, gives rise to higher temperatures than single Au NPs under the same laser irradiation conditions.[36] Still, the process can be equally applied to other plasmonic substrates, such as random NP clusters created by simple drop casting of a colloidal dispersion on top of a glass slide (Figure S8). These substrates display a lower SERS AEF of 5.8 × 105, almost an order of magnitude lower than those determined for plasmonic superlattices, as shown in Figure S7. However, the simplicity of fabrication, with no lithography process involved, may justify their use when high sensitivity is not required. This result shows the broad potential for application of the PLGA-SERS method with arbitrary plasmonic substrates, thus providing it with a general character.

We additionally note that other polymers, such as poly(methyl methacrylate) (PMMA), can be deposited on top of plasmonic substrates, in line with previous studies.[42,43] PMMA coatings showed a similar behavior as those made of PLGA, forming micron-sized holes upon laser irradiation and thereby allowing SERS detection (Figure S9). As a result, we propose that multiple designs can be implemented, by a combination of different plasmonic substrates and different polymers. However, additional drawbacks may arise, either from substrates with low photothermal efficiency or from highly thermostable polymers, and therefore these characteristics must be taken into consideration for the selected materials. For example, polymers with high thermal resistance would require a higher photothermal heating for the window to be opened, which may cause damage on the plasmonic component.[39,44] Such damages, for example, reshaping or degradation, would negatively affect the performance of the plasmonic substrate as a SERS platform in subsequent measurements.

Real-Time Detection Mediated by Localized PLGA Degradation

The small size of the measurement windows opened via laser irradiation along with the precision of the microscope stage provide a sufficient spatial resolution to carry out a large number of sequential measurements at different times. A simple example is provided in Figure S10, showing a SERS map of a PLGA-coated substrate in which two holes were created in close proximity to each other (20 μm). In the same example, the SERS signal of NAm was only identified within the two windows, in agreement with the above results for a single hole.

We thus propose that such a micron-scale control over PLGA degradation can be used to open windows at defined positions, every time SERS measurements are needed. This idea is illustrated in Figure a, for the multiplex SERS detection of different analytes. In this example, the analytes, 4-MBA, crystal violet (CV), thiabendazole (TBZ), and NAm, were sequentially injected at 100 μM concentration inside a silicone chamber covered with the plasmonic substrate (see Experimental Section). When one analyte was added, a laser irradiance of 0.064 mW/μm2 was applied for 1 s so that a measurement window was created. The analyte was then removed from the silicone chamber, and, after a cleaning step by flowing water through the chamber, the following analyte was injected. During the process of sequential incubations, a hole was formed at a selected spot of the PLGA layer for each analyte. Finally, we mapped the whole area by using a laser power of 0.018 mW/μm2. The map was generated by using the SERS intensities at the corresponding wavenumbers of the respective characteristic peaks for each analyte (Figure S11). The SERS map in Figure a displays four regions (corresponding to the laser-irradiated regions) with meaningful and different signals, which we coded by different colors. Note that, due to the strong SERS memory effect, only the characteristic Raman signals of the analyte present at the time of laser irradiation were enhanced in the corresponding new hole (no additional analytes were detected at previously opened areas). This result also indicates that both the plasmonic substrate and the PLGA coating were sufficiently stable to play their corresponding roles during consecutive hole opening and SERS mapping. It should be stressed that four different SERS measurements could be recorded from an area of 120 × 40 μm2, which could potentially translate into thousands of measurements on a substrate with an area of 1 cm2. We thus propose that our approach can largely expand the reusability and lifetime of SERS substrates. By using larger plasmonic substrates and smaller laser spots, even more measurements could be made on a single substrate.

Figure 2

Multiplex SERS detection with high spatial resolution. (a) SERS map of a NP superlattice on which four holes were created by laser irradiation and degradation of the PLGA film (see text for details). The map was generated by integration of the characteristic vibrational modes of 4-MBA at 1078 cm–1 (1), CV at 1183 cm–1 (2), TBZ at 1015 cm–1 (3), and NAm at 1032 cm–1 (4). We used a laser power of 0.018 mW/μm2 with 1 s integration time and a 50× objective. (b) SERS map of a NP superlattice with 2 unshielded areas of ca. 200 × 200 μm2, created by consecutive laser irradiation through a 10× objective, with a power of 0.026 mW/μm2. (c) SERS spectra from each hole in (b). Area 1 was created during incubation with 4-MBA and area 2 during incubation with both MB and NAm.

On a different example, we demonstrated that PLGA-coated substrates can be used for multiplex sensing. We show in Figure b the detection of 4-MBA in area 1, while subsequent incubation with a mixture of NAm and methylene blue (MB) is readily registered in area 2. In this example, we show not only the high spatial control that can be achieved with our PLGA-coated substrates but also the simultaneous detection of multiple analytes from each hole (Figure c). In the same figure, we demonstrate that the holes in the PLGA film can be easily tailored to different sizes and shapes. Therefore, we created larger holes by reducing the objective magnification through which the laser beam was focused and then scanning multiple points at the desired area of the substrate. In this particular example, holes were devised to have a square shape, 200 × 200 μm2; irradiance and illumination time were selected on the basis of the results obtained in Figure S12. In addition, this strategy greatly lowers the detection limit, down to the nM range (see Figure S13).[45]

Long-Term SERS Monitoring: From Microfluidic Chips to Bioreactors

The spatial and temporal control on SERS measurements of different analytes, enabled by the PLGA-SERS method, can be applied to continuous monitoring of solutions, even in complex environments. As a proof of concept, a microfluidic device was attached on top of a PLGA-coated NP superlattice. The resulting microfluidic plasmonic device was mounted along with a syringe pump system to modulate the flow of an analyte solution in the chip (see Experimental Section for details). A schematic representation of the experiment is shown in Figure a. For each SERS measurement, the position of the sample was varied by using a piezoelectric stage, to find a pristine region of the sample. A new measurement window was then created at this spot, and SERS was measured to detect the analytes present in the solution at that precise time. In this manner, changes in the injected solutions with different analytes—here 4-MBA and TBZ—or water, were monitored by SERS on the basis of the most intense SERS peaks from 4-MBA (1080 cm–1) and TBZ (1014 cm–1). As can be observed in Figure b, the sequential presence of TBZ or 4-MBA in the microfluidic channel was readily identified by the PLGA-SERS method over various cycles of injection. Accordingly, no SERS signal was detected when water was flowing in the microfluidic channel. As a control, we used an uncoated plasmonic supercrystal (no PLGA layer), in which case we found that the signal of the first molecule attaching to the substrate (4-MBA) was persistently recorded, even after flowing water or TBZ solution (Figure S14). We propose that an example of application for this setup would be the in situ detection of pollutants or hazardous molecules in wastewater treatment plants.

Figure 3

(a) Scheme of the setup for sensing in flow. The plasmonic substrate was mounted on a microfluidic chip, and the fluid flow was provided by a syringe pump at 20 mL/h. Introduction of both water and analyte solutions in the microfluidic channel was performed by syringe pumps. (b) SERS intensity of the characteristic mode of thiabendazole at 1014 cm–1 (blue line) and 4-MBA at 1080 cm–1 (red line), as a function of the introduction cycles (n) by the syringe pump, for a plasmonic superlattice coated with a PLGA sheathing layer. The black arrows on top of each graph indicate the introduction of a different analyte solution (water (W), TBZ, or MBA) in the microfluidic channel at the indicated cycle. (c) Schematic view of the methodology used to combine a silicone chamber with a plasmonic superlattice to perform SERS measurements, with the laser radiation passing through the support layer. The Au NP-PLGA side of the plasmonic substrate must be oriented toward the inner compartment. We represent in the inset the incorporation of HeLa cells inside the silicone chamber, thereby generating a bioreactor. The bioactive environment causes the interconversion of Ado and HX. (d, e) SERS spectra recorded in situ from plasmonic superlattices with (d) and without (e) the PLGA sheathing layer, at different times (0 and 24 h) after Ado (200 μM) supplementation into the bioreactor containing HeLa cells. (f) SERS spectra of the extracellular supernatant extracted from the bioreactors and measured on fresh substrates without PLGA layer.

In the above-described experiments, the versatility of the detection system was probed for short periods of time, that is, time frames of minutes or a few hours. However, other interesting applications of this technology, such as monitoring the extracellular milieu in cell cultures, would likely involve analyses extending for longer time periods. For such long-term experiments, we propose employing a different PLGA (95:5) solution, due to its higher stability in water.[46] As shown in Figure S15, PLGA 95:5 films maintain their impermeability over longer times, at least for several days. A threshold laser irradiance of 0.064 mW/μm2 was also found to be appropriate for this polymer formulation, allowing molecules in solution to pass through the PLGA 95:5 layer after irradiation (Figure S15, spectrum ii). For experiments with cell cultures, a silicone chamber was manufactured to obtain a biologically active cellular environment, that is, a bioreactor. HeLa cells (1 × 106 cells/mL) were then laden inside the silicone chamber, and the whole system was assembled with the plasmonic substrate, placing the side with the Au NP clusters directly in contact with the extracellular milieu (Figure c). Such devices were previously shown to accurately monitor changes in a similar bioreactor, as long as the plasmonic component was renewed prior to each measurement,[9] which was likely to alter the sample under investigation, even if moderately. In this regard, we envisioned that the PLGA-SERS strategy could be a smart solution to such an invasive procedure.

As a proof of concept, we investigated purine derivative fluctuations in the cell milieu of the bioreactor. Purine metabolites are important modulators of physiological and pathological processes by regulating diverse cell functions.[47,48] Hence, continuous monitoring of extracellular purine levels can provide valuable information about the cellular state within bioreactors. To this end, we initially challenged the culture media with adenosine (Ado), reaching a high extracellular concentration of 200 μM. Under such conditions, the presence of active enzymes in the bioreactor caused a quick decrease of the extracellular Ado, converting it into hypoxanthine (HX),[49] as validated by high-performance liquid chromatography coupled to a mass spectrometry detector (HPLC-MS) (Figure S16). For SERS studies, we measured in situ the spectra of the cultured medium at different incubation times (0 and 24 h) after Ado supplementation. As detailed below, we followed three different strategies for recording these measurements, the results of which are presented in Figure d–f. For completeness, we show the SERS spectra of the pure metabolites Ado and HX in Figure S17a.

For the experiments performed with the bioreactor and 95:5 PLGA coating of the plasmonic substrate (Figure d), irradiation with a 785 nm laser at 0.064 mW/μm2 for 5 s, through a 50× objective, ensured complete degradation of the sheathing layer at the selected spots. Once this procedure was completed, we could readily detect by SERS the high concentration of Ado at time zero, responsible for an intense peak at 735 cm–1 in the SERS spectrum. After 24 h and upon laser irradiation to generate a new measurement window, the metabolic conversion of Ado was recorded, as indicated by a significant peak shift to 725 cm–1. On the other hand, SERS spectra of bioreactors without the protective PLGA layer were still dominated by the characteristic Ado peak after 24 h and could not sense any metabolic activity in the culture medium of the bioreactor (without PLGA in Figure e). To verify these contradictory results, cell supernatants were collected and re-evaluated on fresh plasmonic sensors, which were not found to present interferences from adsorbed molecules. As observed in Figure f, SERS spectra of collected supernatants were in agreement with those obtained by 95:5 PLGA-coated plasmonic substrates, while confirming the lack of accuracy of the results obtained without the PLGA layer, due to a SERS memory effect. These results validate SERS-PLGA as a strategy to consistently measure varying conditions inside bioreactors, even at lower concentrations than those employed in this experiment (Figure S17b). The obtained metabolic parameters may offer highly valuable information to remotely control manufacturing requirements in biotechnology processes.

Conclusions

The present study provides compelling evidence supporting the use of PLGA-coated plasmonic substrates as suitable SERS platforms. Thanks to the impermeable PLGA sheathing layer, such structures were capable of overcoming the common problem of “SERS memory effect”. As a result, this system allows for efficient real-time detection. Using a single SERS substrate in combination with next-generation portable Raman devices could expand the applications of SERS to fields where in situ measurements can be applied. Additionally, the great simplicity and versatility of the developed method will allow others to easily adopt and exploit this technology, even with nanostructured plasmonic substrates or polymers, different to those used in this proof of concept. This study will thus contribute to accelerating the development of real-time SERS monitoring systems. Finally, we envision that the accuracy in the spatial distribution of holes in the sheathing layer might also serve for selective functionalization of plasmonic substrates with targeting molecules such as antibodies or aptamers. This approach will facilitate the creation of rapid multiplex assays, such as immune assays, with multiple possibilities.

Experimental Section

Chemicals

HAuCl4·3 H2O (≥99.9%, trace metal basis) was purchased from Alfa Aesar. Sodium borohydride (ReagentPlus, ≥99%, NaBH4), l-ascorbic acid (ACS reagents, ≥99%, AA), poly(ethylene glycol) methyl ether thiol average Mn 6000 (PEG-6K), sodium hyprochlorite (6–14% active chlorine, Emplura), and cetyltrimethylammonium chloride (≥98%, CTAC) were purchased from Sigma-Aldrich. All solutions, except HAuCl4 and CTAB, were prepared immediately before use. Purified Milli-Q water was used in all experiments (Millipore, 18.2 MΩ cm). Glassware (Menzel-Gläser 24 × 24 #1) was cleaned with aqua regia and rinsed extensively with Milli-Q water before use. Polymethyl methacrylate (PMMA A2 950) was purchased from EM Resist and used as supplied. Polylactide-co-glycolide acid end-cap (75:25, Mn 25,000) and polylactide-co-glycolide acid end-cap (95:5, Mn 25,000) were purchased from Polysciences. Adenosine (suitable for cell culture, Ado), 4- mercaptobenzoic acid (90%, 4-MBA), 4-nitrothiophenol (80%, 4-NTP), thiabendazole (>99% powder, TBZ), crystal violet (dye content >90%, CV) methylene blue (>95%, MB), and nicotinamide (>98% powder, NAm) were purchased from Sigma-Aldrich.

Fabrication Methods

Synthesis and Functionalization of Au NPs

Gold nanospheres were synthesized by seeded growth, as previously reported.[50] First, 2 nm seeds were prepared by adding 50 μL of a 0.05 M HAuCl4 solution to 5 mL of a 100 mM CTAC solution. Subsequently, 200 μL of a 0.02 M NaBH4 (0.75 mg/mL) was added under vigorous stirring. After 3 min, the mixture was diluted 10 times with a 100 mM solution of CTAC. The seeds were then overgrown to 10 nm nanospheres. For this purpose, 900 μL of seeds was added to a mixture of 40 μL of a 0.1 M AA solution and 10 mL of a 25 mM CTAC solution. Next, 50 μL of a 0.05 M HAuCl4 solution was added under vigorous stirring. The 10 nm seeds showed a localized surface plasmon resonance at 520 nm. The dispersion was left undisturbed for at least 1 h. Thereafter, the nanospheres were centrifuged at 12,000 rpm and washed at least 3 times with CTAC 25 mM to obtain a narrower size distribution prior to overgrowth into larger nanospheres. Therefore, 125 μL of 10 nm nanospheres was added to a solution containing 40 μL of a 0.1 M AA solution and 10 mL of a 25 mM CTAC solution. Subsequently, 50 μL of 0.05 M of HAuCl4 was added. The resulting NPs had rough edges, which were removed by oxidative etching. To this end, 10 μL of a dilute solution of sodium hypochlorite (1 to 1.5% of available chlorine) and, 10 min later, 5 μL of a 0.05 M solution of HAuCl4 were added under stirring. After 30 min, the nanospheres were centrifuged at 4800 rpm for 15 min and redispersed in 500 μM CTAC. The particles were then concentrated to ca. 5 mM Au0 in a 500 μM CTAC solution. PEG functionalization was carried out by the addition of 1 mg/mL of PEG-6K and stirring overnight at room temperature. The excess of unbound PEG was removed by 4-fold centrifugation at 4800 rpm for 15 min and redispersion of the sedimented NPs (30 nm) in CTAC 500 μM.

Assembly of Au NP Superlattices

NP superlattices were prepared as previously reported.[29] A 2 μL droplet of NP dispersion (50 mM gold nanospheres of 30 nm, calculated from the absorbance at 400 nm, 66% EtOH, 200 μM CTAC) was casted on a nanostructured PDMS stamp. The PDMS stamp featured a square lattice of 270 nm holes, with a center to center spacing of 500 nm. After 40 s, a glass slide (24 × 24 mm2 and a thickness of 0.13–0.16 mm) was placed on top of the droplet, creating a NP dispersion film between the glass substrate and the PDMS stamp. After 2 h and complete evaporation of the solvent, the glass slide was carefully lifted off the PDMS template, resulting in transfer of the plasmonic superlattice onto the glass substrate, displaying the inverse shape of the nanostructured mold. Immobilization and cleaning of the NP clusters onto a glass slide were achieved by an oxygen plasma process for 20 s, followed by UV-ozone cleaning (ProCleanerTM chamber) for 5 min. The oxygen plasma process was operated using a Diener Electronic nanoplasma cleaner at 100 W and 0.4 mbar oxygen pressure.

Fabrication of Polymer Films

Commercially available, solid PLGA (75:25 or 95:5) was dissolved in ethyl acetate by mechanical stirring of PLGA granules in the selected solvent for 1 h. Stock solutions were prepared at 12 wt% in ethyl acetate and kept in the fridge at 8 °C, and each vial was wrapped with parafilm to avoid solvent evaporation.

PLGA films were created by spin coating (Laurell WS-400B-6NPP LITE) the 12 wt% PLGA/ethyl acetate solution on top of the NP superlattice. To this end, a 300 μL droplet was placed on top of the superlattice, such that it wetted the whole surface of the sample. The spin coating process was then started at a speed of 1500 rpm for 30 s. The thickness of the PLGA film was measured to be 1.5 μm by SEM cross section analysis using a 5 kV acceleration voltage. PMMA coatings were created by spin coating 300 μL of the commercial solution at 1500 rpm for 30 s.

Microfluidic Chip

For the microfluidic chip, the channels were etched into PDMS Sylgard 184, purchased from Sigma-Aldrich. Microdevices were fabricated according to a previously reported protocol.[51] To this end, soft lithography was used to develop positive SU8 240 μm relief patterns with the desired geometry on a silicon wafer. Subsequently, PDMS was mixed at a 10:1 weight ratio of base to curing agent. The mixed solution was poured into the SU8 master and cured in the oven. Finally, the replica-molded layer was trimmed and perforated. The PDMS devices were then exposed to a plasma cleaning treatment (2 min) and subsequently bound to the PLGA-plasmonic substrate by applying a soft pressure on the device with a sterile pair of tweezers, finally presenting a capacity of 10 μL. The flow in the microfluidic channel was generated using a Cetoni Nemesys syringe pump with a low-pressure module. The flow was set at 10 mL/h, and the outlet was connected to another syringe pumping at the same flow rate of 10 mL/h.

Silicone Chamber

To perform the controlled incubation of selected analytes, a silicone chamber was 3D-printed to hold the liquid analyte solution. This silicone chamber was prepared using an elastomer base silicone (Advanced Proser, AS 5702) loaded into a 10 mL clear syringe (PSY-E; Musashi Engineering, Ltd.) and printed with a diameter of 2 cm by a multiheaded 3D Discovery bioprinter (RegenHU, Switzerland) on a glass microslide (26 × 76 mm). To ensure that the total area occupied by the silicone chamber lies within the area of the plasmonic substrate employed in subsequent steps, the area of the internal space within the silicone chamber was smaller than the area of the plasmonic substrate.

Equipment

Scanning electron microscopy was performed using an environmental SEM (FEI Quanta 250). 5–20 kV acceleration voltages were used to take the images.

SERS was measured with a confocal Raman microscope (Renishaw inVia) equipped with a 1024 × 512 CCD detector, using a 785 nm laser excitation source. This system contains an optical microscope equipped with a motor-controlled stage. The optical setup provides motorized XYZ locations of the samples, allowing a high accuracy in position. Control on the Z-axis enables focusing the laser on the surface of the plasmonic substrate, whereas XY displacement governed the location of the laser beam over the substrate. Once focus is achieved on a specific area of the substrate selected by an optical image, a hole can be created by laser irradiation at high irradiance. On the other hand, XY scanning allows acquisition of multiple SERS spectra at varying positions. This procedure can be used to generate a SERS map of the scanned area. SERS maps can be processed to detect areas of the substrate which were exposed to the analyte solution and where the characteristic SERS fingerprint of the absorbed molecules can be recorded. The presented spectra were baseline corrected by an automatic fitting procedure.

SERS Measurements

PLGA-SERS Method

For the irradiation of plasmonic substrates covered with a PLGA sheathing layer, the laser power was regulated as follows: Once the laser spot was focused on the surface of the plasmonic substrate, the 785 nm laser at 0.064 mW/μm2 was irradiated for 1 s, through an 50× objective, so as to create a measurement window in the PLGA layer. Although already at the first exposure, a SERS signal was detected, and exposure time was extended until the intensity of the SERS signal was stabilized (see Video S1). The total irradiation time was ca. 5 s, thereby ensuring complete degradation of the PLGA layer over the selected spot. Upon generation of the measurement window, the SERS signal of the analytes was recorded using an irradiance of 0.018 mW/μm2 with the 785 nm laser for 1 s, in static mode and through a 50× objective (numerical apertures of NA = 0.5, Leica Microsystems, Wetzlar, Germany). In parallel, SERS spectra from control plasmonic substrates (with no PLGA coating) were registered with the same settings: 50× objective with 0.018 mW/μm2 irradiance for 1 s.

Measurement of the PLGA Layer Permeability

To measure the ability of the covering layer to prevent the analytes in solution to reach the plasmonic component, a 50 μL drop of the analyte solution (either 100 μM Ado or 100 μM 4-MBA solution) was deposited on a PLGA-coated plasmonic substrate (95:5 lactic/glycolic acid ratio). Fifty SERS spectra were then randomly acquired from different regions at different times (24 h and 48 h), with a low laser intensity (0.018 mW/μm2) at 785 nm through a 50× objective, thereby preventing degradation of the PLGA layer. The absence of characteristic SERS signals from the analytes indicated impermeability of the sample to the solution. This strategy allowed for a quick evaluation of the coverage degree of sheathing layers on plasmonic substrates. Thus, it was followed to compare the coverages obtained with different solvents of the PLGA solution: ethyl acetate and acetone.

Multiplex SERS Mapping

To confirm the spatial control at different times and analytes, a PLGA-coated plasmonic substrate was sequentially incubated with different molecules at 100 μM concentration. Such analytes were chosen such that their characteristic Raman peaks do not overlap with each other (4-MBA, CV, TBZ, and NAm). Once the silicone chamber was stuck onto the plasmonic substrate, 500 μL of the analyte solution was introduced by microppipeting into the inner chamber of the device. Upon addition of the analyte, a laser irradiance of 0.064 mW/μm2 was applied until SERS intensity stabilization, thus creating a measurement window. The analyte was then removed from the silicone chamber by aspiration with a micropipette, and, after a cleaning step consisting of flowing Mili-Q water through the silicone chamber, the following analyte was injected. During this process of sequential incubations, a hole was created at a different spot of the PLGA layer for each analyte. Control of the hole position was achieved using the piezoelectric stage in the Raman microscope. The distribution of holes along the x-axis was chosen with a step of 20 μm without changing their position in the y-axis (with this step size, no overlap or interference occurred between holes). Once all the different analytes were sequentially incubated and the corresponding PLGA holes created, the whole area was mapped with a laser irradiance of 0.018 mW/μm2, thereby detecting the analytes retained in the plasmonic layer. Scanning measurements were performed using a 785 nm laser excitation source (maximum 119.50 mW), recorded in static mode using a 50× objective (numerical apertures of NA = 0.5) with 1 s integration time, at 0.018 W/cm2 or 0.00390 mW/μm2 laser power. The map of a selected area was acquired with a resolution of 10 μm in x and y. Larger areas can be irradiated by reducing to 10× (NA = 0.25) the objective through which the laser beam is focused. Despite considerably decreasing the power of irradiation on the plasmonic surface when illuminating through this lower magnification objective (0.026 mW/μm2), continuous exposure of the selected area to the laser beam rendered larger measurement windows. For example, a square shape of 200 × 200 μm2 was achieved through 40 successive irradiations of 5 s over the defined area by 785 nm laser line scanning. The resolution between two irradiated points was 10 μm in x and y. Finally, when the scanning was completed, a SERS map was acquired in static mode using a 10× with 1 s integration time, at 0.0039 mW/μm2 laser power.

HPLC-MS

Ado and HX identification and quantification were performed on an Acquity UHPLC chromatograph equipped with a photodiode-array system and coupled to a LCT Premier XE ESI-TOF mass spectrometer (Waters, Milford, MA, USA). Chromatographic separation was carried out using an Acquity BEH C18 (100 × 2.1 mm, 1.7 μm) reverse phased column (Waters, Mildford, USA). The elution buffers were 0.1% formic acid in water (A) and acetonitrile (B), and the linear gradient method consisted of 99% A over 1.5 min, 99–1% over 1.5–6 min, 1% for 2 min, and 99% for 2 min before next injection. Total run time was 10 min, injection volume was 5 μL, and the flow rate was set at 300 μL/min. Both metabolites were detected and quantified after monitoring the UV signal at 254 nm of wavelength.

Bioreactor and Measurements of Cell Metabolic Activity

In cellular experiments, a silicone chamber such as that obtained following the previously described protocol, using PLGA 95:5, was manufactured to obtain a biologically active cellular environment, and its operability as a bioreactor was tested. To that end, HeLa cells (1 × 106 cel/mL) were laden inside the silicone chamber, and the whole system was then assembled with the plasmonic substrate (obtained as described in previous sections), placing the Au NP covered with a PLGA layer 95:5 directly in contact with the extracellular milieu.

Cells were cultured in 500 μL of Dulbecco’s modified Eagle medium with 10% fetal bovine serum, supplemented with 200 μM of Ado. SERS measurements were recorded at 0 and 24 h after Ado suplementation to monitor changes in extracellular concentration over time. For this experiment, irradiation with the 785 nm laser at 0.064 mW/μm2, for 5 s and through a 50× objective, ensured complete degradation of the sheathing layer at the irradiated spots. Subsequently, SERS spectra were recorded at these positions by irradiating with the same 785 nm laser at 0.018 mW/μm2 for 10 s.