# Fileset

[Iwanaga_Front_Bioeng_Biotech12(2024)1484638-2.pdf](https://mdr.nims.go.jp/filesets/6e6bac51-4b08-4bed-a554-d8deb01ff9fe/download)

## Creator

[Masanobu Iwanaga](https://orcid.org/0000-0002-8930-6940)

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Productive biosensing techniques empowered by all-dielectric metasurfaces](https://mdr.nims.go.jp/datasets/94698f2e-e383-4685-86c2-e8d3d431c1a4)

## Fulltext

Productive biosensing techniquesempowered by all-dielectricmetasurfacesMasanobu Iwanaga*Research Center for Electronic and Optical Materials, National Institute for Materials Science (NIMS),Tsukuba, JapanArtificially designed, functional nanostructured surfaces, called metasurfaces, arean emerging platform for biosensing. Twomajor types of metasurface biosensorshave been reported: one is based on resonant-wavelength shift and the other isspecialized for fluorescence (FL) detection. The all-dielectric metasurfaces thatcomposed of periodic arrays of silicon nanocolumns have a series of opticalmagnetic-mode resonances, some of which were found to significantly enhancecapability for FL detection of diverse target biomolecules, ranging from nucleicacid to antigens and antibodies. Here, we mainly address the recent advances inproductive metasurface FL biosensors, provide an overview of the pivotal results,and discuss the future prospects, including artificial-intelligence-driven big dataanalysis for the next-generation healthcare services.KEYWORDSall-dielectric metasurface, fluorescence, biosensor, DNA, antibody, antigen, big data, AI1 IntroductionBiosensing data are growing at a high rate and forming big data. To exploit these,rapidly developing artificial intelligence (AI) will play a key role. To date, the establishedfunction of AI has been to extract reliable averaged information from large amounts of data.As is widely known, human languages are outstanding big data, being successfully made useof in large language models of AI. Thus, for optimal use of AI in the field of biotechnology, itis crucial to collect and accumulate scientific data using reliable biosensing and/orbioimaging techniques. In this context, the biosensing techniques are expected to enablethe efficient acquisition of numerous precise data on biomolecules related to living bodies.Biosensing has been studied extensively for several decades. Several biosensingtechniques have been used to detect biomolecules: gel-based mass analysis such assodium-dodecyl-sulfate poly-acrylamide gel electrophoresis (SDS-PAGE), immunoassayrepresented by enzyme-linked immunosorbent assay (ELISA), optical resonance shift,Raman scattering, fluorescence (FL) detection, and electrochemical techniques. Accordingto the weight of the target biomolecules and the sensing precision, the biosensing techniquesare arranged as shown in Figure 1A, which schematically illustrates the positions of eachmethod in the plane of molecular weight and sensing precision, where the precision isdefined as the amount that is inversely proportional to limit of detection (LOD) in eachmethod. Large molecules, such as DNA and immunoglobulin G (IgG)-type antibody, arefunctional biomolecules in living cells and bodies. Relatively small molecules, such assucrose and amino acids, play a role in maintaining the living systems. Qualitatively, thebiosensing techniques based on the optical resonance shift, ELISA, FL detection, andelectrochemical measurement are mainly used for the huge molecules, whereas Ramanscattering including surface enhanced Raman scattering (SERS) is applied to smallOPEN ACCESSEDITED BYYu-Feng Yu,Guangzhou University, ChinaREVIEWED BYGuojin Zhong,Hunan University, China*CORRESPONDENCEMasanobu Iwanaga,iwanaga.masanobu@nims.go.jpRECEIVED 22 August 2024ACCEPTED 20 December 2024PUBLISHED 09 January 2025CITATIONIwanaga M (2025) Productive biosensingtechniques empowered by all-dielectric metasurfaces.Front. Bioeng. Biotechnol. 12:1484638.doi: 10.3389/fbioe.2024.1484638COPYRIGHT© 2025 Iwanaga. This is an open-access articledistributed under the terms of the CreativeCommons Attribution License (CC BY). The use,distribution or reproduction in other forums ispermitted, provided the original author(s) andthe copyright owner(s) are credited and that theoriginal publication in this journal is cited, inaccordance with accepted academic practice.No use, distribution or reproduction ispermitted which does not comply with theseterms.Frontiers in Bioengineering and Biotechnology frontiersin.org01TYPE PerspectivePUBLISHED 09 January 2025DOI 10.3389/fbioe.2024.1484638https://www.frontiersin.org/articles/10.3389/fbioe.2024.1484638/fullhttps://www.frontiersin.org/articles/10.3389/fbioe.2024.1484638/fullhttps://www.frontiersin.org/articles/10.3389/fbioe.2024.1484638/fullhttps://crossmark.crossref.org/dialog/?doi=10.3389/fbioe.2024.1484638&domain=pdf&date_stamp=2025-01-09mailto:iwanaga.masanobu@nims.go.jpmailto:iwanaga.masanobu@nims.go.jphttps://doi.org/10.3389/fbioe.2024.1484638https://creativecommons.org/licenses/by/4.0/https://creativecommons.org/licenses/by/4.0/https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.org/journals/bioengineering-and-biotechnology#editorial-boardhttps://www.frontiersin.org/journals/bioengineering-and-biotechnology#editorial-boardhttps://doi.org/10.3389/fbioe.2024.1484638FIGURE 1(A) Practical biosensing techniques classified by target molecular weight (horizontal axis) and sensing precision (vertical axis). Each biosensingtechnique is positioned, according to its target molecular weight and sensing precision, which is inversely proportional to LOD. The highest sensingprecision is sub aM, reaching single-molecule concentration. Arrows indicate technical improvements in sensing precision with help of newtechnologies. Large molecules, such DNA and IgG-type antibody, and a small molecule are illustrated. (B) Target biomolecules detected by the all-dielectric metasurfaces to date. The targets include diverse biomolecules, ranging from proteins, such as antibody and antigen, to nucleic acids, such asDNA and RNA. Antibody is indicated using the abbreviation Ab. BSA and Cys-SA denote bovine serum albumin and cys-strepravidin, respectively. Top-right and bottom-left antibody images are cited from Iwanaga (2023b), Iwanaga (2020), respectively. In the antigen region, Spike-protein image is fromIwanaga and Tangkawsakul (2022) and PSA-detection image is from Yavas et al. (2017). Copyright (2017) American Chemical Society. DNA image is fromIwanaga et al. (2023a). RNA image is from Iwanaga (2022).Frontiers in Bioengineering and Biotechnology frontiersin.org02Iwanaga 10.3389/fbioe.2024.1484638https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://doi.org/10.3389/fbioe.2024.1484638molecules to detect their vibrational signals, which are often referredto as molecular fingerprints.To seek for higher precision in the detection techniques, the FLdetection and electrochemical measurement are mostly employed inrecent biosensing studies. Regarding the electrochemical sensing, itwas often claimed that high sensitivity was obtained over a widerange of target biomolecule concentrations (Miao et al., 2018; Khanet al., 2018). However, the electrochemical sensors generally exhibitvery low responsivity, that is, the detected signal changes aretypically 20% or less for one order of target concentrations.Consequently, it is difficult to provide a precise multi-order-digitanalysis for target concentrations. In contrast, FL detection yieldsmore precise and specific results for the target concentrations, owingto the significant development of FL-protein-labeling techniquesthat started from the finding of green fluorescent proteins(Shimomura, 1979). This is one of the reasons that methods todetect FL are considered to be most precise and are capable ofachieving the lowest LOD among the various biosensing techniquesto date (Giljohann and Mirkin, 2009). Single-molecule sensing wasclaimed to use SERS before 2000 (Nie and Emory, 1997; Kneippet al., 1997); however, the SERS sensing was conducted at high targetconcentrations, indicating that the SERS signals were observed asrare events involving many target molecules at a large LOD.In 2010s and later, further technical improvements in FLdetection were ma made, substantially advancing the sensingprecision to detect target biomolecules. One of the approacheswas the fraction section method (Rissin et al., 2010), whichmeasured the FL signals in sectioned areas up to 1,000,000,conducted statistical analysis, and enabled a wide dynamic rangeof the FL signals. Conducting the elaborate measurement andanalysis procedures, the signal-to-noise ratio was substantiallyimproved. This sectioning method is often referred to as thedigital method; for instance, digital ELISA and digital polymerasechain reaction (PCR) were commercialized. Another approach wasrealized by employing new FL-enhancing platforms, particularlymetasurfaces (Choi et al., 2015; Iwanaga et al., 2016; Iwanaga, 2018).Incorporating microfluidic chips, the metasurfaces formed compactFL biosensors and enabled highly precise biosensing (Iwanaga,2020). In particular, single cell-free DNA (cfDNA) wassuccessfully detected by discriminating one cfDNA from zerocfDNA (Iwanaga et al., 2023); to the best of our knowledge, suchultimate single-molecule detection has not been attained in anyother biosensing platform. These recent developments are shown inFigure 1A and listed in Supplementary Table S1.In this paper, wemainly address the recent advances in all-dielectricmetasurface biosensors. They were applied to resonance shift and FLbiosensing. In Section 2, the working principles are described. In Section3, recent advances in the metasurface FL biosensors are overviewed. InSection 4, good practices to use AI are discussed.2 Biosensing techniques usingmetasurfacesIn this section, we briefly address sensing targets and theworking principles of biosensing techniques using the all-dielectric metasurfaces. In addition, the limit of detection (LOD)in each technique is referred to.2.1 Resonance-shift detection2.1.1 Detection of spectral changesOptical resonances in metasurfaces are observed in theirreflection or transmission spectra. The prominent resonancesrespond to changes of the outermost surface of nanostructuresconstituting the metasurfaces, resulting in spectral shifts of theresonance. By calibrating the shift with amount of moleculesabsorbed onto the outermost surface, the concentrations of targetmolecules can be determined. This method was initiated usingsurface plasmon resonance (SPR) on flat gold film in 1980s(Raether, 1988). The SPR method is now commercialized, beingused as an analyzing instrument to evaluate biomolecule bindingreaction (Myszka et al., 1998).Although the SPR method is an established analytical method,the LOD is typically in the nanomolar (nM, M�mol/L) to sub-nMrange. Furthermore, the optical configuration requires deep obliqueincidence, which is complicated and demanding in terms of space.Therefore, further improvements in LOD and optical configurationswere pursued. In 2010s, dielectric nanostructures were recognized asoptical resonators (Gomez-Medina et al., 2011), which areassociated with prominently resonant spectra. All-dielectricmetasurfaces of silicon nanodisk arrays were used to detect acancer marker protein, prostate-specific antigen (PSA) (Yavaset al., 2017); analysis of the transmission spectrum shift indicatedthat the LOD was approximately 100 picomolar (pM), comparableto the LOD of commercially available ELISA kits. The benefit ofusing the all-dielectric metasurfaces is the simple optical setup thatallows normal incidence.A new twist in resonance-shift sensing is to use the bound statesin the continuum (BIC) in all-dielectric metasurfaces. The BIC playsa role in realizing modes with high-quality (Q) factors (or narrowlinewidths), in accordance with slightly symmetry-breakingstructure(s). The high-Q modes facilitate tracing resonance shift,compared to low-Q modes, in accordance with the change in theenvironmental refractive index or the absorption of molecules on theoutermost surface of the metasurfaces. Realizing this concept, BIC-based resonance-shift sensing was reported (Hsiao et al., 2022;Watanabe and Iwanaga, 2023; Watanabe and Iwanaga, 2024).The physical limit of resonance shift sensing is given by arelation of Δλ � a × Δn where Δλ, a, and Δn denote thewavelength shift, periodic length of the metasurface, and changein environmental refractive index, respectively (Iwanaga, 2023a).The BIC-based metasurfaces showed 50%–70% realization of thephysical limit. Thus, there is a limit, even if the sensitivity isincreased; therefore, narrow linewidth associated with the high-Qmode is used to obtain better figure of merit (or signal-to-noiseratio) in the BIC metasurfaces. Even when these improvements areincorporated, the LODwill remain at a pM range. Thus, it is unlikelythat extremely high sensitivity at an attomolar (aM) range is realizedusing the BIC metasurfaces.2.1.2 Detection of molecular absorption linesMolecules have their light absorption lines in the infrared range.The absorption spectroscopy is valid for rather small moleculesbecause large molecules with molecular weight more than50,000 have complicated absorption spectrum and are difficult tobe identified. Detection of a series of absorption lines of a particularFrontiers in Bioengineering and Biotechnology frontiersin.org03Iwanaga 10.3389/fbioe.2024.1484638https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://doi.org/10.3389/fbioe.2024.1484638molecule may be useful to prepare a series of metasurfaces with abroad absorption band corresponding to the molecular absorptionlines. This concept was demonstrated using a set of all-dielectricmetasurfaces, each of which had a narrow infrared absorption bandoriginating from BIC in accordance with the asymmetric units (Tittlet al., 2018). Although the LOD was not stated, the shown datasuggested that the LOD was in a nM range. This means that themethod is valid only for high-concentration targets.Thus, the biosensing techniques based on resonance shifts aresuitable for analyses at rather higher target concentrations. In thissense, they are not highly precise sensing techniques but arecharacterized as analyzing techniques regarding molecularbinding reaction and molecular identification at moderate andhigher concentrations.2.2 FL detectionMethods for efficient FL detection have been pursued for tens ofyears. In 1970s, it was found that FL quenching becomes dominantwhen FL molecules were placed on a flat metal or dielectric material(Chance et al., 1978). The effect was understood as ultrafast transferof photoexcited states.In the era of nanotechnology since 2000, many nanostructuredsurfaces on substrates were tried to improve the FL-detectionefficiency; some of them showed prominent FL-enhancing effects(Kinkhabwala et al., 2009; Fu et al., 2010; Zhang et al., 2012; Zhouet al., 2012; Punj et al., 2013; Choi et al., 2014; Choi et al., 2015;Iwanaga et al., 2016; Iwanaga, 2018; Dong et al., 2021). In thesestudies, more than 1,000-fold FL-intensity enhancing effects werereported, where the enhancement was defined as the FL-intensityratio of FL intensity on the nanostructures to the FL intensity onrelevant flat references. Reproducibility, which is defined here aspercentage to measure the best FL enhancement, was sufficientlyhigh for only two cases: plasmo–photonic metasurfaces (Choi et al.,2015; Iwanaga et al., 2016) and all-dielectric metasurfaces thatfeature higher magnetic-mode resonances (Iwanaga, 2018). Thesetwo cases adopt a strategy to optimize the entire photoexcitedprocess from excitation to FL emission. In contrast, the othercases focused on the so-called hot spots, which were electric-fieldenhanced local spots, emerging in nanogaps or nanocorners.3 Metasurface fluorescence biosensorsFigure 1B summarizes the representative results on the detectionof biomolecules using the all-dielectric metasurfaces. The targetswere diverse, ranging from antigen/antibody (Section 3.1) to DNA/RNA (Section 3.2). Labeling FL probes on the antibodies or DNAprobes specific to the target ensures that the FL sensing is specific tothe target. The practical potential of the metasurface biosensors isalso referred to (Section 3.3).3.1 Antigen/antibody detectionBiosensing primarily targets proteins of diverse sizes andvarieties. When focusing on biosensors for healthcareapplications, antigen and antibody among the numerous proteinsare suitable sensing targets. On the left side of Figure 1B, antibodyand antigen detection using the all-dielectric metasurfaces of siliconcircular nanocolumn array is illustrated. The typical structuralparameters were as follows: periodic length of the metasurfaceswas 300 nm, and the diameter and height of the nanocolumns were220 and 200 nm, respectively. The parameters were selectedconsidering the excitation wavelength of 532 nm and FL-emission wavelengths of 570–630 nm. A higher magnetic-moderesonance was tuned to the FL-emission wavelength (Iwanaga,2018). Nanofabrication procedure of the metasurfaces wassubjected to a top-down approach using electron-beamlithography (Iwanaga, 2020).The antigen/antibody detection on the metasurfaces wasconducted in sandwich configurations where an antigen wasbound to the two specific antibodies, as illustrated in Figure 2. Inthe direct sandwich configuration, an antibody labeled with biotinand another antibody labeled with FL molecules; the biotin-labeledantibody was bound to the streptavidin (SA, represented by yellowmarks in Figure 2) that were first immobilized on the metasurfaces;the FL-labeled antibodies emitted FL signals specific to the target onthe metasurfaces in an FL-enhanced manner. In the indirectconfiguration, the FL-labeled antibodies were bound to the non-conjugated primary antibody that formed the sandwich body. Theconcrete microfluidic protocols were described in previous reports(Iwanaga, 2020; Iwanaga, 2023b).The antigen/antibody detection results shown in Figure 1B arebriefly described below.(i) Cancer marker antigens, such as carcinoembryonic antigen(CEA) and PSA, were detected in a dynamic range of fourorders of target concentrations, and the LOD were10 femtomolars (fM) and 4 pM, respectively. Importantly,both CEA and PSA were successfully detected withoutsubstantial reduction of the FL signals in human serum,which is a practical medium for medical examinations(Iwanaga, 2023b). We note that the medical diagnostic criteriaregarding CEA and PSA are 1,000- and 25-fold higher than theLOD, respectively. Therefore, the biosensing on the metasurfacesis sufficiently precise for the medical cancer diagnoses.Additionally, in the experiment for direct comparison, theCEA was detected using a commercial ELISA kit, and theLOD was approximate 5 pM, being 500-times higher thanthat by the metasurface biosensors (Iwanaga, 2023b).(ii) The spike proteins of COVID-19 were detected and the LODwas identified to be 0.8 pg/mL, which is estimated to beapproximately 292 fM. In a similar configuration, theantibodies of the spike proteins were also detected in therange of 10.7–686 pM (Iwanaga and Tangkawsakul, 2022),which can serve as a test of the antibody level. This is one ofthe advantages of the metasurface FL biosensors thatmultiple targets are detected in the same platform byadjusting the microfluidic protocols.(iii) IgG antibody was detected in a wide range from pg/mL totens of ng/mL, and the LOD was determined to be 34 fM.This detection capability of IgG was better than that of thecommercial ELISA kits, which exhibited the LOD of 1.56 pMin the experiment for direct comparison (Iwanaga, 2020).Frontiers in Bioengineering and Biotechnology frontiersin.org04Iwanaga 10.3389/fbioe.2024.1484638https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://doi.org/10.3389/fbioe.2024.1484638The all-dielectric metasurfaces have recently been applied tohuman serum albumin (HSA) detection in urine (Fu et al., 2024). Inthis experiment, the molecular configuration used for detection wasdifferent from that used for previously described sandwich assays.Instead of the ordinary FL probes labeled on antibodies, we used amaterial to show aggregation-induced emission (AIE) on HSA,which was termed as TPE-4TA and emitted green FL peaked at530 nm (Tu et al., 2019). The AIE material binds to the HSA in aspecific manner, induces the deformation of HSA, further aggregateson the HSA, and increases FL emission. In conventionalmeasurements using microwells, the AIE material enabled HSAdetection at 0.25–1,000 μg/mL. Using the metasurfaces, a lowerconcentration range of 0.0188–160 μg/mL was successfullymeasured and the LOD was identified as 18.75 ng/mL, whichwas more than 10-fold lower than that of the conventionalmeasurements. Thus, the metasurface platform exhibited theprecise sensing without loss of robustness for a type of realisticbiopsy samples, urine.3.2 Nucleic acid detectionNucleic acid detection using all-dielectric metasurfaces hasfrequently been reported (Iwanaga, 2021; Iwanaga, 2022;Iwanaga et al., 2023), which are based on the same design asthat of the antigen/antibody detection described in Section 3.1. Inthe basic detection scheme, the binding molecule SA was firstimmobilized on the outermost surface of silicon nanocolumns,biotin-conjugated DNA probes were efficiently immobilized onthe metasurface via the biotin-streptavidin coupling, the targetDNAs with a complementary sequence to those of the biotin- andFL-probes were immobilized on the metasurface, and FLmeasurement was conducted. An illustration of the FLdetection is provided in Figure 1B. Overall, the scheme issimilar to that for the sandwich protein detection, being basedon target-specific detections.There is an option that the target DNA sequence can beamplified through thermal cycling in a manner similar to PCR ifthe original target concentrations are too low, such as in the aMrange. With help of FL enhancement of the metasurfaces, theamplification cycles are reduced, which is important forsuppressing false positive reaction that could occur in ordinary PCR.As typical results, we here refer to three results.(i) Direct DNA detection (without amplification) resulted inDNA detection in a pM range. LOD for a target DNA was0.1 pM (Iwanaga, 2021). This procedure was the simplest forDNA detection.(ii) A low-concentration target of COVID-19 complementaryDNA was detected in the range of aM to fM withamplification cycles shorter than those of conventionalquantitative PCR (qPCR). The LOD of 5.86 aM (or14 copies/test) is better than qPCR and equivalent tothat of digital PCR (Iwanaga, 2022); typical results thatreported COVID-19 RNA detections were 40±10 copies/test using the qPCR (Bruce et al., 2020) and approximate10 copies/test using the digital PCR (Yin et al., 2021). It isto be noted that thermal cycling for amplifications weredifferent in the three cases; that is, the metasurfacebiosensors were used after 35 cycles, the qPCRconducted 40 cycles, and the digital PCR did 45 cycles;therefore, the ratio of amplified products is 1: 32: 1, 024,respectively, meaning that the metasurface biosensorswere far less relied on the thermal cycle amplification;however, the metasurface biosensors exhibited the highsensitivity, thanks to the exceptional FL-enhancementcapability. The short cycle operation significantlyreduces falsely positive results without loss of therobustness and is one of the advantages to use themetasurface biosensors. As is widely known, thecomplementary DNA is directly produced via reversetranscription of RNA. The acquired data were FLimages of the metasurfaces captured by a charge-coupled device (CCD) camera, as shown in Figure 1B,and analyzed in an automated manner. The dataacquisition and straightforward analysis enabled us toFIGURE 2(A) Illustration of a commonly considered scheme to use AI in biosensing and bioimaging. (B) Scheme involving productive biosensors, big data, AI,and healthcare information services for people. The FL and metasurface images are adapted from Iwanaga (2022), Iwanaga (2020), respectively.Frontiers in Bioengineering and Biotechnology frontiersin.org05Iwanaga 10.3389/fbioe.2024.1484638https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://doi.org/10.3389/fbioe.2024.1484638obtain massive biosensing data. Thus, the metasurfacebiosensor system is a productive sensing platform.(iii) cfDNA is a short DNA fragment of 150–200 base pairs thatcirculates in human blood. The cfDNAs are released fromvarious organs and have signatures specific to diseases orreactions. They are considered as next-generationbiomarkers (Cristiano et al., 2019). However, theconcentration is extremely low and requires ultrahigh-precision detection techniques. Under this requirement,we applied the metasurface platform for cfDNA detectionand achieved single cfDNA detection (Iwanaga et al., 2023).In particular, one cfDNA was definitely discriminated fromzero cfDNA with high statistical confidence. Such anultimately precise detection has not been shown in anyother biosensing platform; for example, the digital PCRusually exhibits the LOD at approximate 10 copies/test(Hu et al., 2020; Wei et al., 2022).3.3 Practical potentialWe here address the metasurface FL biosensors from apractical viewpoint. The metasurface substrates and themicrofluidic chips are mass productive in foundriesconducting nano/microlithography. The measurement setup ofmetasurface biosensors has been built in a compact, cost-effective, and automated form Iwanaga et al. (2023b), enablinghigh-throughput operation in biosensing. Thus, the total cost isable to be suppressed, being sufficiently practical in comparisonwith the existing commercial biosensing techniques.4 Discussion and future scopeHere, we discuss best practices for the use of AI in biosensing.Figure 2 shows two plausible uses of AI in situations related tobiosensing. Figure 2A illustrates the AI involved in biosensing orbioimaging. Figure 2B shows a scheme for AI usage in a cycle ofhealthcare services, starting with a productive biosensor.Owing to the rapid development of AI technologies, trials toinvolve AI in biosensing and/or bioimaging will increase over thenext few years. Indeed, massive stacked medical images such asX-ray and magnetic resonance images are being learned andanalyzed by AI in machine-learning schemes. There are alreadynumerous images regarding ordinary medical examinations, so AI isexpected to assist medical doctors in make diagnoses. This is astereotype expectation for AI.When considering the AI applications shown in Figure 2A tonew biosensors or bioimaging devices, it is necessary to prepare alarge amount of teaching data, along with accurate humanknowledge. This would be a demanding task in developing a newbiosensing/bioimaging application associated with AI. To avoid thisdemanding task, it may be helpful to link newly acquired data withthe existing, known data in a reasonable manner.When building thislink, AI could be a handy and cost-effective tool because the currentAI has already learned from the existing data including images.As concluding remarks, we address outlook for themetasurface FL biosensors in the near future. In our view, oneof the best practices is a scheme as shown in Figure 2B, where AIcan contribute to the analysis of big data collected using theproductive biosensors. In principle, AI can identify correlationsin big data without any preconception and prejudice. To realizethis best practice, scientifically reliable, precise big data should beacquired. The metasurface FL biosensors meet this requirementand can function as productive sensing platforms, which areconstructed by simple, cost-effective elements, such as CCDcameras, and associated with an application for automateddata analysis and collection. This usage of AI has not beenemphasized, probably because the productive biosensingsystems have not yet become widespread. In the near future,this usage will be one of the standards.Very recently, this type of AI usage, such as Figure 2B, hasbeen recognized worldwide. The Nobel prize in chemistry2024 was presented in part to the AI development forcomputational protein design (Nobel, 2024). Computationalmolecular dynamics produced big data, which are based on afirm scientific base, were analyzed using AI to explorecandidates for new medicines. Just as the computationalmethod served as a productive data generator, themetasurface FL biosensors can function as productive data-taking instruments for opening the next-generationhealthcare services.Data availability statementThe original contributions presented in the study are included inthe article/Supplementary Material, further inquiries can be directedto the corresponding author.Author contributionsMI: Conceptualization, Data curation, Formal Analysis,Funding acquisition, Investigation, Methodology, Projectadministration, Resources, Software, Supervision, Validation,Visualization, Writing–original draft, Writing–reviewand editing.FundingThe author(s) declare that financial support was received for theresearch, authorship, and/or publication of this article. This studywas partially supported by NIMS Priority Research Project“Biomaterials.” Nanofabrication of the metasurfaces wassupported by ARIM (number JPMXP1223NM5310).AcknowledgmentsThe author thanks Takashi Hironaka and WanidaTangkawsakul for conducting the experiment employing themetasurface FL biosensors. Numerical designs and analysis formetasurface were implemented on the supercomputers inCyberscience Center, Tohoku University.Frontiers in Bioengineering and Biotechnology frontiersin.org06Iwanaga 10.3389/fbioe.2024.1484638https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://doi.org/10.3389/fbioe.2024.1484638Conflict of interestThe author declares that the research was conducted in theabsence of any commercial or financial relationships that could beconstrued as a potential conflict of interest.Publisher’s noteAll claims expressed in this article are solely those of the authorsand do not necessarily represent those of their affiliatedorganizations, or those of the publisher, the editors and thereviewers. Any product that may be evaluated in this article, orclaim that may be made by its manufacturer, is not guaranteed orendorsed by the publisher.Supplementary materialThe Supplementary Material for this article can be found onlineat: https://www.frontiersin.org/articles/10.3389/fbioe.2024.1484638/full#supplementary-materialReferencesBruce, E. A., Huang, M.-L., Perchetti, G. A., Tighe, S., Laaguiby, P., Hoffman, J. J., et al.(2020). Direct rt-qpcr detection of SARS-CoV-2 RNA from patient nasopharyngealswabs without an RNA extraction step. PLOS Biol. 18, e3000896. doi:10.1371/journal.pbio.3000896Chance, R. R., Prock, A., and Silbey, R. (1978). Molecular fluorescence andenergy transfer near interfaces. Adv. Chem. Phys. 37, 1–65. doi:10.1002/9780470142561.ch1Choi, B., Iwanaga, M., Miyazaki, H. T., Sakoda, K., and Sugimoto, Y. (2014).Photoluminescence-enhanced plasmonic substrates fabricated by nanoimprintlithography. J. Micro/Nanolithogr. MEMS MOEMS 13, 023007. doi:10.1117/1.JMM.13.2.023007Choi, B., Iwanaga, M., Miyazaki, H. T., Sugimoto, Y., Ohtake, A., and Sakoda, K.(2015). Overcoming metal-induced fluorescence quenching on plasmo-photonicmetasurfaces coated by a self-assembled monolayer. Chem. Commun. 51,11470–11473. doi:10.1039/c5cc04426jCristiano, S., Leal, A., Phallen, J., Fiksel, J., Adleff, V., Bruhm, D. C., et al. (2019).Genome-wide cell-free DNA fragmentation in patients with cancer. Nature 570,385–389. doi:10.1038/s41586-019-1272-6Dong, Z., Gorelik, S., Paniagua-Dominguez, R., Yik, J., Ho, J., Tjiptoharsono, F., et al.(2021). Silicon nanoantenna mix arrays for a trifecta of quantum emitter enhancements.Nano Lett. 21, 4853–4860. doi:10.1021/acs.nanolett.1c01570Fu, C.-C., Ossato, G., Long, M., Digman, M. A., Gopinathan, A., Lee, L. P., et al.(2010). Bimetallic nanopetals for thousand-fold fluorescence enhancements. Appl. Phys.Lett. 97, 203101. doi:10.1063/1.3495773Fu, Q., Iwanaga, M., and Tang, Y. (2024). Metasurface platform incorporatingaggregation induced emission based biosensor for enhanced human serum albumindetection. Adv. Opt. Mat. 12, 2400868. doi:10.1002/adom.202400868Giljohann, D. A., and Mirkin, C. A. (2009). Drivers of biodiagnostic development.Nature 462, 461–464. doi:10.1038/nature08605Gomez-Medina, R., Garcia-Camara, B., Suarez-Lacalle, I., González, F., Moreno, F.,Nieto-Vesperinas, M., et al. (2011). Electric and magnetic dipolar response ofgermanium nanospheres: interference effects, scattering anisotropy, and opticalforces. J. Nanophot. 5, 053512. doi:10.1117/1.3603941Hsiao, H.-H., Hsu, Y.-C., Liu, A.-Y., Hsieh, J.-C., and Lin, Y.-H. (2022).Ultrasensitive refractive index sensing based on the quasi-bound states in thecontinuum of all-dielectric metasurfaces. Adv. Opt. Mat. 10, 2200812. doi:10.1002/adom.202200812Hu, F., Li, J., Zhang, Z., Li, M., Zhao, S., Li, Z., et al. (2020). Smartphone-baseddroplet digital LAMP device with rapid nucleic acid isolation for highly sensitivepoint-of-care detection. Anal. Chem. 92, 2258–2265. doi:10.1021/acs.analchem.9b04967Iwanaga, M. (2018). All-dielectric metasurfaces with high-fluorescence-enhancingcapability. Appl. Sci. 8, 1328. doi:10.3390/app8081328Iwanaga, M. (2020). All-dielectric metasurface fluorescence biosensors for high-sensitivity antibody/antigen detection. ACS Nano 14, 17458–17467. doi:10.1021/acsnano.0c07722Iwanaga, M. (2021). High-sensitivity high-throughput detection of nucleic-acidtargets on metasurface fluorescence biosensors. Biosensors 11, 33. doi:10.3390/bios11020033Iwanaga, M. (2022). Rapid detection of attomolar SARS-CoV-2 nucleic acids inall-dielectric metasurface biosensors. Biosensors 12, 987. doi:10.3390/bios12110987Iwanaga, M. (2023a). A design strategy for surface nanostructures to realizesensitive refractive-index optical sensors. Nanomaterials 13, 3081. doi:10.3390/nano13243081Iwanaga, M. (2023b). Robust detection of cancer markers in human serums usingall-dielectric metasurface biosensors. Biosensors 13, 377. doi:10.3390/bios13030377Iwanaga, M., Choi, B., Miyazaki, H. T., and Sugimoto, Y. (2016). The artificial controlof enhanced optical processes in fluorescent molecules on high-emittance metasurfaces.Nanoscale 8, 11099–11107. doi:10.1039/C6NR01318JIwanaga, M., Hironaka, T., Ikeda, N., Sugasawa, T., and Takekoshi, K. (2023).Metasurface biosensors enabling single-molecule sensing of cell-free DNA. NanoLett. 23, 5755–5761. doi:10.1021/acs.nanolett.3c01527Iwanaga, M., and Tangkawsakul, W. (2022). Two-way detection of COVID-19 spikeprotein and antibody using all-dielectric metasurface fluorescence sensors. Biosensors12, 981. doi:10.3390/bios12110981Khan, M. S., Dighe, K., Wang, Z., Srivastava, I., Daza, E., Schwartz-Dual, A. S., et al.(2018). Detection of prostate specific antigen (PSA) in human saliva using an ultra-sensitive nanocomposite of graphene nanoplatelets with diblock-co-polymers and auelectrodes. Analyst 143, 1094–1103. doi:10.1039/C7AN01932GKinkhabwala, A., Yu, Z., Fan, S., Avlasevich, Y., Müllen, K., and Moerner, W. E.(2009). Large single-molecule fluorescence enhancements produced by a bowtienanoantenna. Nat. Photonics 3, 654–657. doi:10.1038/nphoton.2009.187Kneipp, K., Wang, Y., Kneipp, H., Perelman, L. T., Itzkan, I., Dasari, R. R., et al.(1997). Single molecule detection using surface-enhanced Raman scattering (SERS).Phys. Rev. Lett. 78, 1667–1670. doi:10.1103/PhysRevLett.78.1667Miao, P., Jiang, Y., Wang, Y., Yin, J., and Tang, Y. (2018). An electrochemicalapproach capable of prostate specific antigen assay in human serum based onexonuclease-aided target recycling amplification. Sens. Actuat. B Chem. 257,1021–1026. doi:10.1016/j.snb.2017.11.064Myszka, D. G., He, X., Dembo, M., Morton, T. A., and Goldstein, B. (1998).Extending the range of rate constants available from BIACORE: interpreting masstransport-influenced binding data. Biophys. J. 75, 583–594. doi:10.1016/S0006-3495(98)77549-6Nie, S., and Emory, S. R. (1997). Probing single molecules and single nanoparticles bysurface-enhanced Raman scattering. Science 275, 1102–1106. doi:10.1126/science.275.5303.1102Nobel (2024). Nobel prize in chemistry. Available at: https://www.nobelprize.org/prizes/chemistry/2024/summary/(Accessed December 10, 2024).Punj, D.,Mivelle,M.,Moparthi, S. B., van Zanten, T. S., Rigneault, H., vanHulst, N. F., et al.(2013). A plasmonic ‘antenna-in-box’ platform for enhanced single-molecule analysis atmicromolar concentrations. Nat. Nanotechnol. 8, 512–516. doi:10.1038/nnano.2013.98Raether, H. (1988). Surface plasmons on smooth and rough surfaces and on gratings.Berlin: Springer-Verlag.Rissin, D.M., Kan, C.W., Campbell, T. G., Howes, S. C., Fournier, D. R., Song, L., et al.(2010). Single-molecule enzyme-linked immunosorbent assay detects serum proteins atsubfemtomolar concentrations. Nat. Biotechnol. 28, 595–599. doi:10.1038/nbt.1641Shimomura, O. (1979). Structure of the chromophore of aequorea green fluorescentprotein. FEBS Lett. 104, 220–222. doi:10.1016/0014-5793(79)80818-2Tittl, A., Leitis, A., Liu, M., Yesilkoy, F., Choi, D.-Y., Neshev, D. N., et al. (2018).Imaging-based molecular barcoding with pixelated dielectric metasurfaces. Science 360,1105–1109. doi:10.1126/science.aas9768Frontiers in Bioengineering and Biotechnology frontiersin.org07Iwanaga 10.3389/fbioe.2024.1484638https://www.frontiersin.org/articles/10.3389/fbioe.2024.1484638/full#supplementary-materialhttps://www.frontiersin.org/articles/10.3389/fbioe.2024.1484638/full#supplementary-materialhttps://doi.org/10.1371/journal.pbio.3000896https://doi.org/10.1371/journal.pbio.3000896https://doi.org/10.1002/9780470142561.ch1https://doi.org/10.1002/9780470142561.ch1https://doi.org/10.1117/1.JMM.13.2.023007https://doi.org/10.1117/1.JMM.13.2.023007https://doi.org/10.1039/c5cc04426jhttps://doi.org/10.1038/s41586-019-1272-6https://doi.org/10.1021/acs.nanolett.1c01570https://doi.org/10.1063/1.3495773https://doi.org/10.1002/adom.202400868https://doi.org/10.1038/nature08605https://doi.org/10.1117/1.3603941https://doi.org/10.1002/adom.202200812https://doi.org/10.1002/adom.202200812https://doi.org/10.1021/acs.analchem.9b04967https://doi.org/10.1021/acs.analchem.9b04967https://doi.org/10.3390/app8081328https://doi.org/10.1021/acsnano.0c07722https://doi.org/10.1021/acsnano.0c07722https://doi.org/10.3390/bios11020033https://doi.org/10.3390/bios11020033https://doi.org/10.3390/bios12110987https://doi.org/10.3390/bios12110987https://doi.org/10.3390/nano13243081https://doi.org/10.3390/nano13243081https://doi.org/10.3390/bios13030377https://doi.org/10.3390/bios13030377https://doi.org/10.1039/C6NR01318Jhttps://doi.org/10.1021/acs.nanolett.3c01527https://doi.org/10.3390/bios12110981https://doi.org/10.1039/C7AN01932Ghttps://doi.org/10.1038/nphoton.2009.187https://doi.org/10.1103/PhysRevLett.78.1667https://doi.org/10.1016/j.snb.2017.11.064https://doi.org/10.1016/S0006-3495(98)77549-6https://doi.org/10.1016/S0006-3495(98)77549-6https://doi.org/10.1126/science.275.5303.1102https://doi.org/10.1126/science.275.5303.1102https://www.nobelprize.org/prizes/chemistry/2024/summary/https://www.nobelprize.org/prizes/chemistry/2024/summary/https://doi.org/10.1038/nnano.2013.98https://doi.org/10.1038/nbt.1641https://doi.org/10.1016/0014-5793(79)80818-2https://doi.org/10.1126/science.aas9768https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://doi.org/10.3389/fbioe.2024.1484638Tu, Y., Yu, Y., Zhou, Z., Xie, S., Yao, B., Guan, S., et al. (2019). Specific andquantitative detection of albumin in biological fluids by tetrazolate-functionalizedwater-soluble AIEgens. ACS Appl. Mat. Interf. 11, 29619–29629. doi:10.1021/acsami.9b10359Watanabe, K., and Iwanaga, M. (2023). Nanogap enhancement of the refractometricsensitivity at quasi-bound states in the continuum in all-dielectric metasurfaces.Nanophotonics 12, 99–109. doi:10.1515/nanoph-2022-0565Watanabe, K., and Iwanaga, M. (2024). Optimum asymmetry for nanofabricatedrefractometric sensors at quasi-bound states in the continuum. Appl. Phys. Lett. 124,111705. doi:10.1063/5.0158793Wei, C., Yu, C., Li, S., Meng, J., Li, T., Cheng, J., et al. (2022). A droplet-basedmultivolume microfluidic device for digital polymerase chain reaction. Sens. Actuat. BChem. 371, 132473. doi:10.1016/j.snb.2022.132473Yavas, O., Svedendahl, M., Dobosz, P., Sanz, V., and Quidant, R. (2017). On-a-chipbiosensing based on all-dielectric nanoresonators. Nano Lett. 17, 4421–4426. doi:10.1021/acs.nanolett.7b01518Yin, H., Wu, Z., Shi, N., Qi, Y., Jian, X., Zhou, L., et al. (2021). Ultrafast multiplexeddetection of SARS-CoV-2 RNA using a rapid droplet digital PCR system. Biosens.Bioelectron. 188, 113282. doi:10.1016/j.bios.2021.113282Zhang, W., Ding, F., Li, W.-D., Wang, Y., Hu, J., and Chou, S. Y. (2012). Giant anduniform fluorescence enhancement over large areas using plasmonic nanodots in 3Dresonant cavity nanoantenna by nanoimprinting. Nanotechnol 23, 225301. doi:10.1088/0957-4484/23/22/225301Zhou, L., Ding, F., Chen, H., Ding,W., Zhang,W., and Chou, S. Y. (2012). Enhancement ofimmunoassay’s fluorescence and detection sensitivity using three-dimensional plasmonicnano-antenna-dots array. Anal. Chem. 84, 4489–4495. doi:10.1021/ac3003215Frontiers in Bioengineering and Biotechnology frontiersin.org08Iwanaga 10.3389/fbioe.2024.1484638https://doi.org/10.1021/acsami.9b10359https://doi.org/10.1021/acsami.9b10359https://doi.org/10.1515/nanoph-2022-0565https://doi.org/10.1063/5.0158793https://doi.org/10.1016/j.snb.2022.132473https://doi.org/10.1021/acs.nanolett.7b01518https://doi.org/10.1021/acs.nanolett.7b01518https://doi.org/10.1016/j.bios.2021.113282https://doi.org/10.1088/0957-4484/23/22/225301https://doi.org/10.1088/0957-4484/23/22/225301https://doi.org/10.1021/ac3003215https://www.frontiersin.org/journals/bioengineering-and-biotechnologyhttps://www.frontiersin.orghttps://doi.org/10.3389/fbioe.2024.1484638 Productive biosensing techniques empowered by all-dielectric metasurfaces 1 Introduction 2 Biosensing techniques using metasurfaces 2.1 Resonance-shift detection 2.1.1 Detection of spectral changes 2.1.2 Detection of molecular absorption lines 2.2 FL detection 3 Metasurface fluorescence biosensors 3.1 Antigen/antibody detection 3.2 Nucleic acid detection 3.3 Practical potential 4 Discussion and future scope Data availability statement Author contributions Funding Acknowledgments Conflict of interest Publisher’s note Supplementary material References