Combining Electrochemiluminescence Detection with Aptamer‐Gated Indicator Releasing Mesoporous Nanoparticles Enables ppt Sensitivity for Strip‐Based Rapid Tests

Abstract The combination of electrogenerated chemiluminescence (ECL) and aptamer‐gated indicator delivering (gAID) magnetic mesoporous silica nanoparticles embedded into glass fibre paper functionalised with poly(ethyleneglycol) and N‐(3‐triethoxysilylpropyl)diethanolamine allowed the development of a rapid test that detects penicillin directly in diluted milk down to 50±9 ppt in <5 min. Covalent attachment of the aptamer “cap” to the silica scaffold enabled pore closure through non‐covalent electrostatic interactions with surface amino groups, while binding of penicillin led to a folding‐up of the aptamer thus releasing the ECL reporter Ru(bpy)3 2+ previously loaded into the material and letting it be detected after lateral flow by a smartphone camera upon electrochemical excitation with a screen printed electrode inserted into a 3D‐printed holder. The approach is simple, generic and presents advantages with respect to sensitivity, measurement uncertainty and robustness compared with conventional fluorescence or electrochemical detection, especially for point‐of‐need analyses of challenging matrices and analytes at ultra‐trace levels.


Introduction
Fuelled by their key role in helping to contain the SARS-CoV-2 pandemic especially through the fast identification of contagious yet potentially asymptomatic individuals,r apid tests have perhaps received unprecedented attention in societies around the globe during the last year. [1] However, besides their current prominence in medical diagnostics,rapid tests and assays have also become increasingly important in other areas such as food, [2] agriculture and forestry, [3] security and forensics [4] or the environment [5] in the last decade.T heir advantage is that they can be used outside of alaboratory by untrained personnel directly at ap oint of need, minimizing time between first suspicion and first decision taking.I nt his regard, paper-based sensors are an attractive and emerging class of devices [6] because they fulfil the prerequisites of the World Health OrganizationsA SSURED principle:t hey are affordable,s ensitive,s pecific,u ser-friendly,rapid and robust, equipment free and deliverable to end-users. [7] Thep hysical, chemical and mechanical properties of cellulose or glass fibre paper in combination with the simplicity of preparation render these materials very interesting also in terms of resource-effective alternatives for device production technologies. [8] Furthermore,t he ubiquity of mobile communication devices with powerful computing capabilities and onboard cameras have led to as ituation in which al arge majority of the global population in principle has ap owerful detector in hand that is especially suited for taking photographs of flat substrates such as paper strips and analysing their content. While the use of smartphones as detectors has started ad ecade ago, [9] only the advent of affordable 3D printing technologies and OTG( on-the-go) electronics have boosted the field, [10] making the fabrication of cases to fit on aphone simple and affordable and the adaption to new phone models with different size or camera position straightforward. [11] OTGelectronics allow for facile and autonomous integration of microelectrodes and light-emitting diodes (LEDs) for electrochemical and fluorescence measurements into such holders, [12] leaving much more room for assay development than the photographing of coloured areas.
Because of these advantages,m any efforts have been undertaken to design sensory nanomaterials for paper-based point-of-care diagnostics. [13] Better robustness,sensitivity and multiplexing capabilities have been achieved by tuning the properties of the membranes in combination with the use of mobile phones for data analysis, [14] improving the performance of such devices. [15] Most of the current paper-based rapid tests rely on colorimetric, [16] fluorescence [17] or electrochemical detection. [18] However,m any of these assays show weaknesses in terms of specificity,s ensitivity,a ccuracy and precision or the capability for multiplexed detection. [19] Whereas specificity is connected to the recognition (bio)chemistry,c olorimetric detection, which is still the prevalent method in rapid tests,i sa lmost exclusively relying on gold nanoparticles (AuNPs) which are decorated with biomacromolecular binders,m ainly antibodies.A lthough AuNPs possess distinctly higher molar absorption coefficients than organic dyes or coloured inorganic ions,s uch assays are limited with respect to sensitivity,e specially when used with visual (naked eye) inspection. In addition, such colorimetric tests are also primarily employed for biomolecular analytes as the immunoassay formats commonly employed, sandwich and competitive,often show inferior sensitivity in small-molecule analysis. [20] Furthermore,multiplexing of AuNP-based colorimetric assays requires spatial separation and cannot rely on an identification via different colours.Fluorescence detection in contrast can be measured straightforwardly with as martphone camera and allows for fluorescence colour multiplexing but has drawbacks in signal-to-noise ratio especially when used with paper supports,s cattering light significantly. Electrochemical detection on the other hand would require additional accessory and multianalyte detection is achallenge. Thus,apromising approach is the combination of both techniques in electrogenerated chemiluminescence or electrochemiluminescence (ECL) detection. [21] ECL dispenses with al ight source for excitation, thus reducing noise significantly,a nd, when compared with the widely used colorimetric assays,p ossesses all the advantages of fluorescence. [21,22] Although ECL detection for paper-based rapid tests has been realized ad ecade ago, [22] up to now only considerably few examples have been reported, most of them for the detection of heavy metal ions,D NA or protein biomarkers as well as whole cells. [23] However,many of these reports only show the applicability of ECL sensing on paper and many of the examples are characterized by rather high detection limits.
In ECL, aluminescence signal is generated by achemical reaction that is initiated and controlled by the application of an electrical potential. Since signal generation is only taking place at the electrode and only for the duration of an applied potential, ECL is ah ighly localized and controlled detection method. Thep aramount advantage of ECL compared with fluorescence is that it does not require al ight source for excitation, allowing to reach exceptionally high signal-tonoise ratios.T his is important in two aspects,i .e., for nontransparent and scattering supports such as paper and for turbid sample media such as milk, wastewater or body fluids, when these need to be analysed without clean-up.A so nly comparatively few compounds show ECL emission under ambient conditions in aqueous media, tris(2,2'bipyridine)ruthenium(II) chloride (Ru(bpy) 3 2+ )i st he most frequently used ECL reporter. [24] Furthermore,powerful ECL sensing is only possible when Ru(bpy) 3 2+ is combined with aco-reactant, such as asecondary or tertiary amine, [25] leading to adesired enhancement of the ECL signal and allowing for the detection in aqueous solution under aconstant potential. Another practical advantage of Ru(bpy) 3 2+ -based ECL is the wavelength range of emission (600-750 nm) that can be easily detected by smartphone cameras. [26] Based on our experience in developing powerful yet simple test strip-based sensing systems for point-of-need scenarios, [27] ac hallenge remains the ultra-trace detection of analytes in complex media, e.g.,o fp ollutants such as pesticides or antibiotics in foodstuff such as milk. Consider am ilk truck driver who can use at est during every stop at afarm when collecting fresh milk on the daily tour, to screen for the presence of antibiotics such as penicillin before accepting the milk to be filled into the tank, thus avoiding potential contamination of the entire load. Theeconomic and health benefits are immediately obvious.Such atest should be simple to use for an untrained person, provide areliable result in short time,record the result automatically for documentation purposes and should be very sensitive as,for instance,the maximum residue level (MRL) for penicillin in milk is as low as 4 mgkg À1 . [28] To develop such at est, we combined the advantages of ECL detection with our highly sensitive, selective and modular nanoparticle-based signalling approach that utilizes gated indicator release [27b] and implemented them on alateral flow-type test strip with smartphone readout. As we have recently demonstrated, such an approach can be expanded facilely into atest that remains simple yet allows to analyse asmall number of analytes at the same time. [29] Such low-number multiplexing would be highly desirable also for our milk truck driver as it would allow the testing for various lead contaminants at every farm in at ime-saving and straightforward way.W er eport here for the first time how the synergistic use of ECL detection and gated indicator release in rapid paper-based assays allows to determine antibiotics such as penicillin in achallenging matrix like milk down to the ppt level in less than 5min of overall assay time.

Results and Discussion
An aptamer-gated indicator delivery (gAID) system was chosen as chemical recognition element because aptamers, which are DNAs equences with high selectivity and affinity for target proteins, [30] small molecules [31] or metal ions, [32] are an attractive alternative to antibody-gated systems while showing superior versatility and modularity. [33] Thep rinciple of gated indicator delivery signalling is as follows (Scheme S1). [34] Ap orous scaffold material, commonly mesoporous silica nanoparticles,i sl oaded with indicator molecules and coated with (bio)chemical entities as so-called gatekeepers,w hich are grafted covalently to the scaffolds outer surface.B ulky entities such as biomacromolecules are then bound to these gatekeepers,u sually in an on-covalent fashion, capping the pores and blocking release of the indicator cargo (Scheme S1a). Thesystem is designed in such away that atarget analyte binds stronger to either gatekeeper or cap than the two gating partners bind to each other, leading to ad issociation of this pore closing ensemble and hence allowing for ar elease of the cargo.T he result is ac hemical signal amplification as few analyte molecules react with the gating ensemble and lead to the delivery of many more reporter molecules,d iffusing from the pores into the surrounding solution. [35] Especially in combination with biochemical gating,s uch systems have shown already good sensitivity and selectivity in simple assay formats.W ith respect to gAID systems,afew have been already reported in the literature. [36] However,i nt hose cases,t he aptamer as such has been non-covalently coated onto the surface of the porous host material and is fully displaced from the inorganic scaffold after binding of the corresponding analyte,t he binding event entailing ar efolding of the aptamer that facilitates desorption from the surface (Scheme S1a). This approach is inconvenient when other components of ar eal sample can lead to unintended desorption and thus falsepositive release of indicator. When aiming at the detection of antibiotics in milk, such ascenario is rather likely because of the electrolyte content and the presence of various surfaceactive compounds in milk. Instead of antibody-based gAID systems with which we have worked so far, [27b,29] aptamers seemed more attractive to us here because they offer better possibilities of defined covalent chemical attachment to the outer surface of the porous host, including adjustment of linker length between anchor point and binding region. Our principle is thus different (Scheme S1b,S cheme 1): in the closed state,t he aptamer is in al oose,o pen, unfolded form and is capping the pores by electrostatic interaction with an excess of functional groups on the surface,l ike for singlestrand DNA-gated systems. [37] In essence,e ach aptamer is covalently attached to the particle surface via its 5' terminus through one anchor point (green fragment, Scheme 1a)while the negatively charged phosphate groups on the oligonucleotide backbone (black dots) can non-covalently interact (red arrows) with positively charged ammonium groups of the excess amino silane moieties of the primary functional surface coating (blue fragments). Use of apropyl-amido-decyl linker (magenta fragment) allows the aptamer strand to bend over and orient horizontally to the surface so that electrostatic interactions can occur (red arrows). While the covalent bond fixes the aptamer,t he multiple non-covalent interactions are dynamic,t he many aptamers on one particle thus forming aconstantly changing monolayer-type network of nucleotide strands on the particle surface.Ifananalyte molecule (yellow shape) comes close to the binding region and can dock to the motif,aconformational rearrangement of the aptamer strand takes place that leads to af olding up of the aptamer in the Scheme 1. (a) Gating mechanism of the gAID system as described in the text. Forbetter understanding, the relevant dimensionsa re as follows: length of aptamer with 39 nucleotides = ca. 13 nm, pore diameter = 2.4 nm (see text), wall thickness = ca. 2.2 nm (see text), length of APTES group = ca. 0.45 nm, diameter of Ru(bpy) 3 2+ = ca. 1.2 nm, ratio of surface coverage of aptamer to APTES = ca. 1:1000 times (Table S1). (b) Preparation sequence (1-3) and mode of operation (3,4) of the gAID system. 1) Loading of Ru(bpy) 3 2+ into the mesopores.2 ) Functionalisation of the outer surface with APTES moieties. 3) Covalentg rafting of an aptamer moiety through its 5'-terminus appended 1-carboxy decyl linker onto the surface via EDC/NHSc oupling chemistry.4 )Upon advent of penicillin,t he aptamer folds up and the analyte is bound in the designated binding pocket, entailingr elease of the dyes. As each aptamer is bound to the particle surface through only one covalent bond but multiple non-covalent interactions, whenever an analyte molecule approaches and binds to its designated binding motif, the aptamer folds up, the binding pocket closes, and the aptamer is locked in aconformationw here minimal electrostatic interactions with surface APTES groups can occur.The pore is open. complex, away from the particle surface,i nhibiting the noncovalent interactions and opening the pores for the dye (orange dots) to diffuse out. Thec ovalent attachment of the aptamer thus ensures that, instead of lifting the entire gate from it hinges,a nalyte binding simply induces that the gate swings open.
Specifically,t he newly developed material comprises am agnetic mesoporous silica nanoparticle (MSN) scaffold that is loaded with Ru(bpy) 3 2+ and contains ap enicillinselective aptamer [38] covalently attached to the outer surface, ensuring residence of the cargo in the pores in the absence of an analyte (Scheme 1b). Upon binding of penicillin, the aptamer changes its conformation, folds up at the distal end of the surface thus opening the pores and leading to the release of al arge number of Ru(bpy) 3 2+ reporters.T he detailed synthesis,f unctionalisation and characterisation of the materials are given in Sections 2-4, Figures S1-S5, Tables S1, S2 of the Supporting Information.
Before testing the performance of the material, the ECL signalling of Ru(bpy) 3 2+ in solution was optimised for sensitivity by screening several co-reactants and SPE electrodes (see Sections 5.1-5.3, Figures S6-S8, Table S3 in Supporting Information). As aresult, we employed an SPE electrode containing gold as aw orking electrode,p latin as ac ounterelectrode and silver as ar eference electrode (AT250), in conjunction with N-butyldiethanolamine (NBEA) as ac oreactant at ac oncentration of 5mM.
Thesensing material MMRAA was prepared by suspending magnetic mesoporous nanoparticles MM of MCM-41type in ah ighly concentrated Ru(bpy) 3 2+ solution in acetonitrile to load the maximum amount of dye into the calcinated scaffold, yielding MMR (see Section 3i nS upporting Information). Mesoporous particles with am agnetic core were chosen because they endow the system with more flexibility in particle handling.N ext, amino groups were covalently attached to the outer surface by condensation of 3-aminopropyltriethoxysilane (APTES), resulting in MMRA,b efore the penicillin-selective aptamer,C OOH-C10-5'-TTT TCT GAA TTG GATC TC TCT TCT TGA GCG ATCT CC ACA-3', [38] was grafted covalently through its terminal carboxylic acid group to the amino groups of MMRA via an active-ester-method (see Section 3i nSupporting Information). Thep resence of the mesoporous structure of MM was confirmed with nitrogen adsorption-desorption isotherms, small and wide-angle X-ray scattering (SAXS/WAXS), scanning electron microscopy (SEM) and transmission electron microscopy (TEM) analysis.SEM and TEM images revealed that the as-prepared nanoparticles MM were spherical with radii between 70-110 nm, i.e., with an average size of 205 AE 34 nm, encapsulating iron oxide nanoparticles (IO-NPs)with diameters of 6.5 AE 1.1 nm. Furthermore,t he specific surface area (1009 m 2 g À1 ), pore diameter (2.4 nm) and pore volume (0.67 cm 3 g À1 )w ere determined by porosimetry studies using the Barrett-Joyner-Halenda (BJH) and Brunauer-Emmett-Te ller (BET) models on the adsorption branch of the isotherm for analysis.S AXS measurements provided the lattice cell parameter (4.59 AE 0.06 nm) which, together with the pore size,a llowed the wall thickness to be estimated to 2.18 nm (see Section 4i nS upporting Information). The respective contents of APTES and Ru(bpy) 3 2+ on and in the material were determined to 1.03 and 0.26 mmol gsolid À1 for MMRA from elemental analysis,t hermogravimetry and spectrophotometric measurements,r espectively.S uccessful condensation of the aptamer was revealed by zeta potential and STEM-EDX measurements.T he latter showed an increase in the phosphorus and sulphur content for the material. Thea ptamer content was estimated from spectrophotometry through as tandard addition method by measuring the increase in absorbance at 260 nm, yielding contents of 0.7 AE 0.3 mmol gsolid À1 .This amount corresponds to acoupling efficiency of ca. 70 %o ft he aptamer added during the synthesis (see Section 4i nS upporting Information). Zeta potential measurements were performed in water and buffers of different pH (water at pH 7; MES 100 mM, pH 5; PB 10 mM, pH 8), all of them showing anegative displacement of the zeta potential after covalent anchoring of the aptamer moieties to the surface of MMRAA (displacement of ca. 40 mV in both buffered media;8 0mVi nw ater), due to reduction of the net positive charge of the aminated particles. Thelatter includes covalent amide bond formation as well as electrostatic interaction between phosphate groups of the aptamer backbone with the excess amino groups in their protonated ammonium state on the surface.After addition of 2ppm of penicillin, asmall displacement to less negative zeta potential values was observed at neutral pH (displacement of ca. + 4mVinPB10mM; + 15 mV in water), ascribed to the binding with the aptamer producing ac onformational rearrangement that leads to ab reaking of the non-covalent interactions and af olding up of the aptamer in the complex, thus de-shielding amino surface groups during the opening of the pores.
To assess whether the optimized amount of co-reactant might have to be adjusted for best performance of the gAID system under realistic conditions, MMRAA was tested by suspending the material 5min in the presence and the absence of 1ppm of penicillin in am ixture of buffer containing different amounts of co-reactant (0.8-25 mM NBEA) and milk (25 %), keeping in mind that pH control is important because many aptamers retain their integrity and binding behaviour until ca. pH 9 [39] and the ECL emission yield of Ru(bpy) 3 2+ is highest at pH 9( Figure S9). Figure 1 shows the ECL emitted in absence and in presence of penicillin as afunction of the NBEA concentration. Whereas c NBEA < 1.6 mM was not able to produce as ignificant response,t he optimum was found again in the range of 3mM < c NBEA < 6mM, which let us use an NBEA concentration of 5mMinthe subsequent studies.Itshould be noted that the interplay of electrolyte content (buffer concentration) and co-reactant concentration (the co-reactant being abase) is critically important for system operation. Tooh igh electrolyte content could interfere with the non-covalent interaction of aptamer and surface amino groups as well as complex formation [40] and atoo high or too low pH would also strengthen or weaken these interactions (pK a of APTES = 9.6), [41] which would influence blank release or binding kinetics.
Having established the chemical parameters of the assay in suspension, the response kinetics were studied. The response time is an essential feature of every rapid test as it majorly decides about acceptance by the end user.T he experiments were carried out as explained before,i .e., in buffered solution (10 mM PB,p H8;5mM NBEA) containing 25 %ofcow milk, respectively. Figure S10a shows that the presence of penicillin induced the opening of the pores with the subsequent release of Ru(bpy) 3 2+ ,w hereas release was negligibly low in the absence of the analyte,d ye release happening on the order of several minutes.T his is not extremely fast, but it should be noted that the response kinetics in suspension usually differ from those on strip,t he latter often being faster because of active transport in the lateral flow.S ince as atisfactory level of signal was reached after 5min, we proceeded with this timing.T he kinetics were also assessed for different amounts of penicillin, at final concentrations between 1p pb and 10 ppm, showing that the response accelerates with analyte concentration which hints at diffusion control of such assays in suspension (Figure S10b).
Following as imilar procedure,s ystem sensitivity was studied by recoding dye release from MMRAA as afunction of penicillin concentration after 5min of reaction by both, ECL ( Figure 2) and fluorescence ( Figure S11). Acorrelation between dye release and analyte concentration was observed in all cases,inagreement with conformational changes of the surface-bound aptamer upon binding of penicillin, opening the pores.H owever,w hen the dose-response curves were fitted to af our-parametric logistic function, [42] ah igher sensitivity was found for ECL compared with fluorescence, yielding limits of detection (LODs) of 0.18 AE 0.07 mgl À1 as well as 0.42 AE 0.10 mgl À1 for ECL using the smartphone as well as the spectraECL cell and 3.1 AE 0.4 mgl À1 when employing the fluorometer ( Figure S11). This difference of ca. one order of magnitude is tentatively ascribed to the scattering of milk contained in the medium, which affects fluorescence measurements.Whenpenicillin was present in the lower ppb range (7 ppb), an amplification factor of 120 molecules of Ru(bpy) 3 2+ delivered per molecule of penicillin in the sample was estimated. With amounts of 1ppb of penicillin, an amplification factor of 350 was obtained.
With the aim to reinforce our theory of the sensing mechanism, in which in the closed state of the gAID system the aptamer is in an open form and is capping the pores by interaction with an excess of functional amino groups on the surface,two control materials MMRAAc and MMRCA were prepared. For MMRAAc,amixture of two different oligonucleotides was grafted to the surface of MMRA,the aptamer used for MMRAA and as hort c-DNA, COOH-C10-5'-TTT TGT GGA GATC-3',that is partially complementary to the sequence of the penicillin aptamer (Schemes S1c,S 2). Using a1:1 mixture of aptamer and c-DNAitwas expected that the aptamer would (partially) hybridize with the short c-DNA,  inhibiting dye escape through base strand pairing instead of electrostatic interaction with surface amino groups.K inetic control experiments showed ar elease of the dye that is virtually identical for MMRAA and MMRAAc in the absence of analyte,y et that the presence of analyte led to as lower and reduced release for MMRAAc in comparison with MMRAA ( Figure S12a). When MMRAAc was treated with different concentrations of penicillin in asimilar way as MMRAA,ar elease of dye as af unction of c penicillin was also observed. However,the sensitivity of MMRAAc was ca. one order of magnitude lower than that of MMRAA (LOD = 24 AE 6ppb, Figure S13). This reduced sensitivity is tentatively ascribed to an aggravated competition of penicillin for the aptamer once the latter has hybridized to the short c-DNA strand.
Fort he second control material MMRCA,t he aptamer was grafted with an amino group at the 5' terminus (instead of COOH as in MMRAA and MMRAAc)t ot he surface of MMRC,w hich was synthesised by treating MMRA with succinic anhydride,converting the surface amino largely into COOH groups (Scheme S2). Thea im was to support our consideration of non-covalent capping by electrostatic attraction between surface ammonium groups and phosphate groups on the aptamer backbone.E xpressing ac arboxylaterich surface it was expected that pore closure would be much less efficient for the aptamer grafted to MMRC,i .e., in MMRCA.Already during its preparation, distinctly more dye was released when washing MMRCA than for MMRAA or MMRAAc,i ndicating that the gate was not properly closing the pores.F urthermore,i nb inding studies ah igher dye release compared with MMRAA or MMRAAc was observed for MMRCA in the absence of the analyte and amuch lower release in presence of the analyte,indicating that besides pore closure also the gating mechanism was not efficient (Figure S12b). Zeta potential measurements of the control materials supported these observations ( Figure S4).Whereas for MMRAA and MMRAAc,n egative displacements of the zeta potential by ca. 75 mV and 85 mV (in H 2 O, pH 7) were observed after covalent anchoring of the aptamer or the aptamer/c-DNAmoieties to the surface of MMRA, MMRCA showed ad isplacement of only ca. À40 mV with respect to MMRC,p resumably because conjugation of one aptamer introduced 39 phosphate groups per one carboxylate anchor group and not because of shielding by non-covalent interaction. Thee ffect of the analyte was also different, the smallest changes in zeta potential occurring for MMRCA, demonstrating that the binding of penicillin by the aptamer produces ac onformational rearrangement and pore opening for MMRAA and MMRAAc,yet not for MMRCA.
Having established the sensing behaviour in solution, we moved towards applicability and incorporated the hybrid particles with modified glass fibre strips.B ased on previous work in which we have shown that the chemical functionalization of at est strip matrix can significantly improve assay performance, [27a,43] we also went through several matrix tailoring cycles by sterically adsorbing or covalently anchoring PEG and N-alkyldiethanolamine moieties on the fibres to arrive at the best material used here.W hereas the PEG moieties preserve the stability of the sensing material MMRAA in the paper matrix and facilitate the transport of Ru(bpy) 3 2+ due to areduction of the electrostatic interaction of the positively charged dye with an et negatively charged silanol-and silanolate-expressing surface of the glass fibre paper,t he N-alkyldiethanolamine groups enhance the ECL signal of the dye.For that purpose,glass fibre papers (Fusion 5 grade, GF)w ere first modified by adsorption of NBEA and PEG groups on the membranes,y ielding (NP)GF membranes.A lternatively, N-(3-triethoxysilylpropyl)diethanolamine (NPEAS) and 3-[methoxy(polyethyleneoxy)propyl)] trimethoxysilane (PEGS) were grafted covalently to GF membranes in toluene,y ielding the membrane NPGF (see Section 6, Figures S14-S16, Table S4 in Supporting Information).
Thei mprovement of the ECL efficiency of the modified papers was assessed by suspending 2.5 mlRu(bpy) 3 2+ solution (1.2 mM) at ca. 7mmfrom the bottom and dipping the strips for 2min into 300 mlofasolution of PB 10 mM (pH 8) or PB 10 mM containing NBEA (25 mM) before recording the ECL as well as the fluorescence emission of released Ru(bpy) 3 2+ in zone B( Figure 3). As can be seen in Figure 3a,b,Ru(bpy) 3 2+ was strongly adsorbed on the membranes containing no modification (GF)o rN BEA and PEG moieties sterically adsorbed ((NP)GF). In contrast, when NPEAS and PEGS were covalently grafted to the membranes (NPGF), Ru-(bpy) 3 2+ was much less retained. Interestingly,l ess fluorescence was observed in both cases for the covalently functionalised strips NPGF.F igure 3c shows the corresponding ECL emission measured for three different strips after development with PB (10 mM) or PB (10 mM) containing NBEA (25 mM). Thesignal increase from GF via (NP)GF to NPGF is apparent, and afavourably strong reddish orange ECL was especially seen for NPGF,which is advantageous with respect to handling and on-site measurements.M oreover,b oth, the measured for the strips placed on an AT250 SPE electrode at the end of their development using either PB (10 mM, "PB") or PB (10 mM) and NBEA (25 mM;" PB + NBEA") as mobile phase with the spec-traECL cell;m embranenumbering as in a,b). d) ECL images registered with the smartphone setup for the membranes (numberingasina ,b) measured at the end of the developed strips using either PB or PB + NBEA (see a,b) as mobile phase. Note that ISO was reduced ca. 16 when NBEA was present, to avoid ECL signal saturation on the images.
presence of NBEA and PEG adsorbed in (NP)GF as well as the presence of NPEAS and PEGS grafted to NPGF were able to enhance the ECL efficiency. In addition, af urther strong enhancement was observed in both cases when NBEA was also used in solution (Figure 3c,e xamples 2, 3). Thus, membranes NPGF containing covalently grafted NPEAS and PEGS moieties were further used.
To improve performance through better control of the flow,h ydrophobic wax patterns were printed onto NPGF membranes by lamination of the printed patterns on aluminium foil. Curing the strips at 110 8 8Cfor 1hled to amelting of the wax into the membrane,imprinting the features across the entire thickness of the paper. Sensing material MMRAA was then incorporated into strips NPGF by depositing 5 mlo f as olution of MMRAA in PB (2 mg ml À1 )a tt he interaction zone (zone A) of the strip,located ca. 5mmfrom one end of the strip (Scheme 2; see Section 6inSupporting Information). Them embranes were analysed with an optical microscope and by SEM, revealing that the incorporation of PEGS and NPEAS groups and the subsequent impregnation with wax led to an expansion of the fibres ( Figure S15a). Energydispersive X-ray spectroscopy (EDX), thermogravimetric (TGA) and elemental analysis (EA) were conducted to qualitatively estimate the amounts of PEGS and NPEAS groups.Anincrease in mass loss of 6.7 %for NPGF compared with neat GF membranes was found by TGA (see Section 6in Supporting Information). Furthermore,t his result was in good agreement with EA, which provided amounts of PEGS and NPEAS groups as 0.14 AE 0.02 and 0.09 AE 0.01 mmol g À1 glass fibre membrane,c orresponding to at otal mass loss of 7.1 %.
Thel ayout of the sensing membranes was based on designs recently reported by us for gAID-based lateral flow assays (LFAs), [27b, 29] containing two different zones,azone A, in which the sensing material is deposited, and az one B, in which the released indicators are confined at the distal end of the strip after traveling with the solvent front. Chemical recognition happens in zone Aa nd the ECL signal is measured in zone B ( Figure 4a). When the strip is dipped into as olution that does not contain an analyte,n od ye is released, and no signal is detected in zone Bb ecause MMRAA remains capped. In the presence of the analyte in the sample,asignal proportional to the analyte concentration is detected in zone Bb ecause aptamer moieties are rearranged on the surface and dye molecules are released. The much larger MMRAA particles remain at the spot of deposition. As ar esult, the ECL (or the fluorescence) of the reporter molecules released can be quantified with the onboard camera of asmartphone,when the latter is equipped with the respective miniaturized accessories.I na ddition, the strips were also evaluated with the spectraECL cell, verifying the ECL spectra.
Like for MMRAA in suspension, the optimum amount of co-reactant had to be found for adipping time of 2min which yields optimum contrast ( Figure S17). Figure 4s hows the corresponding reddish orange ECL light emitted in the absence and the presence of penicillin as af unction of the concentration of NBEA. In contrast to the results observed in suspension, aq uenching effect was found in the paper experiments at much higher c NBEA > 25 mM, providing also better signal-to-noise ratios when using higher amounts of coreactant as in solution (12.5 mM < c NBEA < 37.5 mM). On strip,l ower amounts of NBEA (3-6 mM) induced only am oderate ECL signal;s till higher amounts (50 mM) produced as trong release also in absence of the analyte. Following Figure 4, 25 mM of NBEA was chosen for further studies.T hat the quenching only occurs at higher NBEA concentrations on strip compared to solution is tentatively attributed to the preconcentration of the released dye in zone Ba nd to ap artial retention of the co-reactant in the fibrous matrix itself.T ok eep the assay simple,N BEA has to be Scheme 2. Design and principle of operation of the LFA. i) Composition of the strips containing MMRAA in zone Aand working principle of dye release from MMRAA in ii)the presence and iii)the absence of the analyte. As chematic representationo fhow zone Bo fthe membrane is placed on the electrode is also shown. In the presence of an analyte, ECL emission is generated at the electrode area upon applying avoltage. provided together with the buffer in as ingle sample preparation step,i .e., 1+ 3d ilution of the sample so that the use of higher amounts is necessary.H andling of another solution and including ap ipetting step to apply NBEA directly to zone Bw as no option for us.
Because of the considerably higher concentration optimum of NBEA for the strip experiments,the influence of pH on the performance of MMRAA@NPGF was assessed. 25 mM NBEA in solution were equivalent to pH 9, see above.Onthe strip with its coated fibres,the microscopic pH can be different, or the strip material can have ac ertain buffering effect. However, as the pH in the strips is difficult to measure,w er epeated the strip assay for dipping solutions with an adjusted pH of 7.0-9.6 in the presence and absence of 250 ppb of penicillin in diluted milk as described before ( Figure S18). It was found that the strips showed best and stable performance between pH 8.5-9.0. While the behaviour between pH 7.0-9.0 seems to be primarily dictated by the ECL efficiencyo fR u(bpy) 3 2+ (see also Figure S9), the high blank release in absence of analyte at pH 9.6 suggests that capping becomes inefficient, because too few surface amino groups are still in their protonated form, the pK a of APTES being 9.6. [41] Up to pH 9.0, the binding efficiency of the aptamer thus seems to be largely unaffected by pH.
Thes ensitivity of MMRAA@NPGF was evaluated next, dipping them into 300 mlo fb uffered solutions (PB,1 0mM, pH 8; NBEA, 25 mM) containing 25 %ofmilk and different amounts of penicillin as described before.A sc an be seen in Figure 5, in both cases,E CL and fluorescence detection, an increase of dye release was observed as af unction of the penicillin concentration. However,w hen integrated density values were plotted vs. c penicillin ,ahigher sensitivity was found for ECL in comparison with fluorescence,arriving at quantitation ranges of 0.2-3.7 ppb and 6-119 ppb as well as LODs of 0.05 AE 0.01 mgl À1 and 3.1 AE 0.7 mgl À1 for ECL and fluorescence measurements,r espectively.M oreover,E CL did not only outperform fluorescence again, but even more advanta-geously,the sensitivity on paper was improved with respect to that of the gAID system in suspension. Ac omparison of the images in Figure 5a,b further reveals that the use of an electrode for excitation leads to am ore homogeneous spottype signal.
Aiming to compare the signal amplification in suspension (see above,inconjunction with Figure 2) and on the strips,the ECL signal of MMRAA@NPGF at the two different penicillin concentrations used above,1and 7ppb,w as converted into aR u(bpy) 3 2+ concentration by comparison with ac alibration curve constructed from applying different known amounts of Ru(bpy) 3 2+ to NPGF and measuring ECL under identical conditions.T he amplification factors found were 400 molecules of Ru(bpy) 3 2+ delivered on average per molecule of analyte for c penicillin = 7ppb and 1600 for 1ppb of penicillin. Theg AID system thus shows consistently a4 -fold higher amplification on strip which is ascribed to ab etter concentration of the dye in the detection zone.
Finally,c ross-reactivities against other antibiotics were investigated by analysing several samples containing 250 ppb of ampicillin, amoxicillin, enoxacin, oxacillin, cefazolin, cefapirin, sulfamethazine and sulfathiazole with strips MMRAA@NPGF (for chemical structures,s ee Figure S19). Figure 6r eveals that only penicillin was able to significantly release the indicator from the pores with enoxacin and sulfathiazole showing am inor cross-reactivity.T he other antibiotics showed negligible dye release,similar to the blank release of the gAID material, in accordance with the reported selectivity of the aptamer. [38a] Forp ractical utility,t he reproducibility of the materials production process and its long-term storage are decisive. MMRAA showed unaltered performance in suspension when stored at 8 8 8Ci narefrigerator under normal air atmosphere for 1m onth, before blank release of dye became more pronounced. Concerning reproducibility,a ne rror of 6-8 % found between replicates among as ingle batch of material increased to 7-15 %between assays using different batches of material, which is acceptable for such simple tests.W hen MMRAA was deposited on the strips,t hey showed unchanged performance over as torage period of ca. 3months (same conditions as above), before the flow became slower, with errors of ca. 5% between replicates and 6-12 %between strips with different batches.U pon storage for ! 11 months, the system still showed avery good performance with an only slightly reduced efficiency (LODs in lower ppb range), the aging of the coated fibres and/or wax barriers being most likely responsible for this.H owever, as the general performance remains unaltered, we tentatively assume that the stability of the material as well as the final test strip can be improved when packaged under CA (controlled atmosphere) conditions.

Conclusion
In summary,t he present work reported for the first time the favourable synergisms that can be obtained by combining electrogenerated chemiluminescence (ECL) detection on paper-based test strips with gated indicator releasing materials.T oward ageneric approach, covalently attached aptamers were used for the gating of the analyte-induced release of an ECL reporter,e ndowing the system with ap ronounced robustness that allowed for the direct determination of penicillin in ac hallenging matrix such as milk. Thea ssay shows fast response times,g ood selectivity,a nd exceptional sensitivity,r eaching an LOD of 50 AE 9ppt in an LFAw ith < 5min overall assay time.The recognition mechanism relies on the folding up of the aptamer upon penicillin binding, which leads to ad isruption of the non-covalent interactions that closed the pores and locked the ECL dye in them. Besides the intrinsic features of chemical signal amplification of gAID systems,o ptimization of ECL co-reactant, SPE electrode and component concentration brought about as trong gain in sensitivity.F or the strip assays,i np articular the covalent modification of the paper fibres with hydrophilic yet uncharged PEG moieties and the additional immobilization of co-reactant moieties allowed to outperform the solution assay.I nc omparison to other gAID systems for small-molecule detection, the present approach performs significantly better than approaches that require alaboratory environment while reaching LODs such as 10.5 ppm for adenosine. [44] Even the most sensitive laboratory-based gAID assay for thrombin with an LOD of 0.13 ppb requires an analysis time of > 120 min. [45] Our system also outperforms other paper-based ECL assays with readout via mobile communication and other handheld devices,b eing up to many orders of magnitude more sensitive. [46] In addition, commercially available lateral flow tests for penicillin or other antibiotics relying on gold nanoparticle aggregation are also limited to ppb sensitivity. [47] Keeping in mind that alarge number of aptamers are reported in the literature for other types of analytes,t his concept is easily generalizable,t hus rendering it very attractive for point-of-care diagnostics, environmental or illicit drug analysis.L ike demonstrated by us recently, [29] the approach harbours atremendous potential for low-number multiplexing, which would require advancements in electrode and reporter design that are currently addressed in our laboratory.T he successful detection of penicillin in milk suggests that such assays might indeed have abroad applicability especially for complex realistic matrices.