Highly Sensitive Naked Eye Detectable Colorimetric Biosensors Made from Macroporous Framework Melamine Foams for Onsite and Simultaneous Detection of Multiple Environmental Hazards in Flowing Through Sensing Systems

Personal‐use, naked eye‐readable, low‐cost, highly sensitive, and selective biosensors for rapid detection of environmental toxicants are relevant for many application scenarios. Here, the recent developments of highly sensitive and naked eye distinguishable colorimetric sensors by using commercially available melamine foam (MF) as basic sensing materials for instant and volume‐responsive simultaneous detection of multiple targets in fluid systems are reported. The MF possesses a unique reticulated three‐dimensional (3D) macroporous framework structure enabling rapid mass transfer of large biomolecules through the structures in all directions, ensuring easy access of numerous active binding sites of the chemically modified framework to the proteins and target molecules, and subsequently providing significantly increased sensitive and volume‐responsive detection of target molecules in flowing through sensor systems. Promising results in direct, sandwich, and competitive ELISA tests demonstrated the great application potential of the materials. Besides, an additive and simultaneous detection of two targets in one system is achieved by using different layers of the sensor materials in a flowing‐through filtering device. The novel biosensors are expected to significantly improve the sensitivity and broaden the applications of ELISA in the rapid detections of trace amounts of toxicants in liquid and aerosol systems.


Introduction
Enzyme-linked Immunosorbent Assay (ELISA) is a common tool that can selectively and sensitively detect a variety of target analytes, such as antibodies, pesticides, antibiotics, proteins, and pathogens. [1,2,3,4]Conventional ELISA sensors can detect hazardous chemicals in foods environment and are used in biomedical diagnosis and chemical quality controls. [5,6,7]However, certain limitations of the current ELISA sensing materials exist, including relatively high cost, lack of scalability and flexibility for personal use, and simultaneous examinations of multitarget chemicals in very low concentrations, as well as dependence on specialized instrumentation. [8]s an alternative, paper-based ELISA (p-ELISA) uses fibrous and microporous platforms with advantages of increased surface areas of fibers, reduced cost, ease of use, and improved detection sensitivity. [9,10,11,12,13]However, the heterogeneous structures of papers and fibrous membranes, especially along the vertical direction, inhibit penetration of large biomolecules through the membranes, resulting in lower than the expected amount of biomolecules incorporated onto surfaces of fibers inside the media. [14,15,16,17]uch a structural feature could consequently lower the sensitivity of ELISA sensors made by nitrocellulose membranes, filter papers, and even nanofibrous membranes, resulting in inhomogeneous colorimetric signals of p-ELISA sensors. [18,19,20]hus, an ideal media with a three-dimensional homogenous and open macroporous structure that can allow large biomolecules migrate freely inside in all directions was envisioned as a better material for such ELISA biosensors.Macroporous aerogels have some 3D structural features.However, most aerogels are highly hygroscopic, causing increased non-specific adsorption of molecules and potentially high false-positive rates, together with reduced sensitivity and accuracy in applications. [21]In addition, most aerogels are not structurally macroporous in all directions, blocking large molecules from moving freely. [22,23,24]Reticulated melamine foam (MF) is an ideal fit to the envisioned materials for biological sensors.MF has 3D uniformed pore sizes of 60∼150 μm, large enough for rapid mass transfer of biomolecules within the media; proper hydrophilicity, high porosity, and high nitrogen content for various chemical modifications, enabling covalent immobilization of biomolecules for immunoassay interactions (Figure S1, Supporting Information); high elasticity and excellent mechanical properties for fabricating into different filtering materials; as well as being chemically stable and nonflammable. [25]MFs have been chemically modified for a wide range of applications in water treatments, such as oil/water separation, water disinfection, adsorption, strain/stress sensing, catalysis, and so on. [26,27]Herein, a commercially available MF was chemically modified for competitive, sandwich, and direct ELISA sensing applications.A SARS-CoV-2 spike protein, a transmembrane protein of SARS-CoV-2 virus, attached with a histidine tag, and an antibiotic, chloramphenicol (CAP), were employed as representatives of various detecting targets.The results revealed that the sensors made of the MF materials could detect SARS-CoV-2 spike protein receptor binding domain (SP-RBD-His) at 0.1 mg L −1 level by naked eyes, with a limit of detection (LOD) at 0.047 mg L −1 when supplemented by a smartphone.In the case of chloramphenicol (CAP), the sensors could detect concentrations as low as 1 ng mL −1 by naked eyes and 0.096 ng mL −1 with smartphone assistance.The materials can be fabricated in an additive testing system to simultaneously detect two or more targets.

Molecule Diffusion in Melamine Foam Membranes
Diffusion of large biomolecules in electro-spun microporous and nanofibrous membranes was proven heterogeneous and very slow in vertical directions due to the fact of layered nanofibrous mats and significantly reduced effective pore sizes. [18]The framework MF materials possess a unique 3D macroporous fibrous structure and could allow large biomolecules to penetrate through them without much resistance.A side-by-side diffusion chamber was employed to measure the transport behavior of biomolecules through different membranes (Figure 1a).Fluorescein isothiocyanate linked dextran (FITC-Dextran, 40KDa) and Human IgG (HIgG, 150KDa) were employed as sample biomolecules to study their diffusion patterns through the MF membranes because the FITC-dextran and HIgG have similar molecular sizes as horseradish peroxidase (HRP) (≈40KDa) and immunoglobulin (≈150KDa), respectively, which are widely used in immunoassays.The concentration changes of HIgG and FITC dextran in the receiver chamber versus diffusion times of the biomolecule through MF membranes in varied thicknesses of 1 mm to 3 mm, nanofibrous membrane (NF), and nitrocellulose paper (NP) are plotted and shown in Figure 1b and Figure 1c, respectively.With the thickness increase of the membranes from 1 mm to 3 mm, the diffusion times of HIgG to reach the steady-state diffusion slightly rose from 8 min to 10 min, while the FITC-dextran needed shorter times (6 min, 7 min, and 8 min) to reach the steady state diffusion pattern, respectively (Figure 1c).The large HIgG molecules showed a slower diffusion rate through the membranes.However, compared to the diffusion performances of the large biomolecules in nanofibrous membranes (PVA-co-PE) and nitrocellulose papers, [18,30] which required hours to reach the steady state of diffusion, the thicknesses of membranes and sizes of biomolecules did not show any significant impact and can be ignored if the time of interaction between the MF and substrate is longer than 10 min.Thus, the open framework structure, high porosity, and large pore size of the MF allow large biomolecules to penetrate through the membranes without significant mass transfer resistance.

Modification and Protein Immobilization on MF
The melamine foam (MF) is a framework-structured material consisting of active secondary amine groups as shown in Figure S1, Supporting Information.To covalently immobilize proteins on the MF, chemical modification of the secondary amino groups on the material is necessary (Figure 2a), which is achieved by using N, N′-disuccinimidyl carbonate (DSC) to introduce the N-hydroxysuccinimide (NHS) functional groups on the material (NHS@MF) that can readily react with amino groups in proteins. [31]The reactions of chemical modification and protein immobilization on the MF are shown in Figure 2b.Fouriertransform infrared spectroscopy (FTIR) proved successful incorporations of the reactive groups (NHS) and immobilization of the protein onto the MF, based on the carbonate peak of NHS@MF at 1730 cm −1 and amide I peak at 1625 cm −1 (Figure 2c). [32,33]SC reagent could increase hydrophobicity of surfaces of materials such as hydrophilic PVA-co-PE NF membranes. [30]As shown in Figure 2d and the Video S1, Supporting Information, the NHS@MF indeed showed slightly reduced hydrophilicity of the MF with water contact angle increased to 81.3°from 0°.Despite the decrease of hydrophilicity, the liquid drop could still completely spread out on the NHS@MF membranes after ≈40 s (Video S2, Supporting Information), and the membrane still retains the desired hydrophilicity, which can reduce nonspecific protein adsorption, promote protein diffusion through the membrane, and ease the complete removal of any unbonded substances in each step, subsequently improving sensitivity in detection of target molecules.After the immobilization of antibodies onto the NHS@MF, there was an increase in hydrophilicity, which is attributed to the inherent hydrophilic domains of the antibodies and the interactions between the protein's hydrophilic regions and water.The SEM images shown in Figure 2a indicate that before and after the modification and protein immobilization, the morphologies of the MF framework structures stay intact, with a pore size of ≈150 μm and a fiber diameter of ≈5 μm in the framework.
The diffusion and penetration of proteins through the framework membrane and the binding between the chemically modified surfaces of the MF and the substrate during the diffusion process could be visualized by confocal imaging.As shown in Figure 3a, Ab-His-647 could completely penetrate inside the NHS@MF membrane and homogeneously distribute and become immobilized on the skeleton of the entire NHS@MF membrane.With the Video S3, Supporting Information (scanning from the bottom to the top of the membrane), the immobilization of the protein is uniformly homogeneous in the MF system vertically.
The amount of NHS immobilized on the NHS@MF was measured and compared to a nanofibrous membrane with the same area and treatment conditions (5% DSC in 100 mL 1,4-dioxane at 70 °C), [34] as nanofibers have smaller diameters and higher surfaces areas than nitrocellulose paper.Per a unit square meter, the NHS@MF showed a higher content of immobilized NHS groups than the nanofibrous membrane (NHS@NF) (Figure 3b), which was exactly as we have speculated based on structural analysis.Structurally speaking, the average diameter of the fibrous frames in MF is significantly coarse in micrometers, and the surface areas per mass should be several scales smaller than that of NF and even NP materials.However, even though the nanofibrous membranes have a higher specific surface area than that of the MF, the overlaid nanofiber webs could be further reduced observed and measured pore sizes when the NF membranes are formed in varied thicknesses.The effective pore sizes, which are corresponding to the diffusion of molecules through the NF membranes, could be reduced to one hundredth or thousandth of the measured pore sizes in the vertical direction. [18]Such a structure is perfect for serving as filters, however, not ideal for use as biological sensor materials, since diffusion of large biomolecules through NF, such as HIgG could be significantly reduced, leading to reduced amounts of proteins loaded onto the inside surfaces of NF membranes. [19]ifferent from the NF membrane materials, the MF framework structure can allow diffusion and penetration of large molecules through the membranes at high speed and freedom.As demonstrated in Figure 1, there is no size exclusion effect in the MF.When the NHS@MF membranes were employed in the immobilization of Ab-His-647 in varied concentrations (5 mg L −1 , 1 mg L −1 , 0.5 mg L −1 ), the amounts of the antibody used and immobilized on MF were correlating well (Figure 3c), indicating that the large biomolecules are homogeneously distributed in the MF membrane.Such a structural feature is quite unique for development of the biological sensors involving the use of large biomolecules and even cells.Furthermore, as shown in Figure S3, Supporting Information, the loading capacity of Ab-His-647 on NHS@MF was higher than both NHS@NF and nitrocellulose paper per mass, generating more reactive sites for target molecules than the regular materials used in the p-ELISA sensors.
A direct visualization comparison test was conducted for both MF and NHS@MF in the immobilization of antibodies. [35]100 μL of 0.5 mg L −1 Ab-His-HRP solution was added onto these two membranes in the same size and thickness, respectively.Subsequent additions of TMB substrate resulted in blue color in varied intensities.Figure 3d shows that after thoroughly washing, the NHS@MF membranes revealed color signals with much higher intensity than that of the pristine (MF) membranes (insert in Figure 3d), proving the importance of the DSC modification on the MF structures for immobilization of proteins.

Direct ELISA on the Melamine Foam Membranes
To demonstrate the applicability of the MF as biological sensor material, direct ELISA was employed on the NHS@MF membranes.A SARS-CoV-2 spike protein receptor-binding domain with C-His tag (SP-RBD-His) in different concentrations from 0 to 100 mg L −1 was employed in the immobilization reaction on the NHS@MF membranes, and Ab-His-HRP was introduced to specifically bind with the immobilized protein and generate colorimetric signals from the reaction between the HRP and TMB substrate (Scheme 1a).To find a proper concentration of HRP and enzyme-substate reaction time, optimization experiments were conducted as shown in Table S1, Supporting Information.The concentration of the Ab-His-HRP at 1 mg L −1 was identified as the optimal concentration and a reaction time of 5 min between the TMB substrate and HRP was chosen accordingly. [24,26,35,36]Besides, to demonstrate the specificity of the assay, different control assays were conducted, and the results were collected and are shown in Figure S4, Supporting Information.In addition to the negative control experiments without the use of the HRP, there was no color or very low response in terms of color change in the absence of SP-RBD-His (Figure S4, Supporting Information).
To explore the sensitivity of the material in detecting target agents, a detection assay with the use of varied SP-RBD-His concentrations (0 to 100 mg L −1 ) was conducted, and the naked-eye readable blue color signals corresponding to different concentrations of the SP-RBD-His are shown in Figure 4a.By examining the color intensities via the photoshop software following Equation 1, [19] where RGB background is the R-value of the white background (no HRP), and RGB membranes is R-value of the NHS@MF membranes, the linear equation for the colorimetric assay was fitted as y = 13.67x+ 2.48 (R 2 = 0.97) between the protein concentrations of 0.1 mg L −1 to 1.5 mg L −1 .Naked eye recognizable SP-RBD-His reached at 1 mg L −1 level with a limit of detection (LOD) at 0.52 mg L −1 with the help of a smartphone and further analysis from software for the direct ELISA sensor (Figure 4a).

Sandwich and Competitive ELISA on the Melamine Foam Membranes
With these promising results in direct ELISA tests, the NHS@MF membranes were also employed in sandwich and competitive ELISAs.Again, SP-RBD-His was also used as the detecting target with a testing protocol shown in Scheme 1b.In the presence of the Ab-SP immobilized on the NHS@MF, the SP-RBD-His could be recognized by the antibody. [37]Then the introduction of the Ab-His-HRP would generate colorimetric signals from the reaction between the HRP and TMB substrate as shown in Scheme 1b.To minimize any background signal of control groups (without primary antibodies), the type and concentration of the blocking buffers were optimized.As shown in Figure S5, Supporting Information, the colorimetric signals of the sample groups blocked with skim milk were lower than that of the groups blocked with BSA under the same treatment, while 3% of skim milk presented the lowest colorimetric signal among other concentrations.Therefore, we chose skim milk (3%) as the blocking and dilution buffer in the following experiments.
To explore the sensitivity of the MF as a sandwich ELISA material, different SP-RBD-His concentrations (0 to 100 mg L −1 ) were employed, and the naked-eye readable color signals corresponding to the concentrations of the target are shown in Figure 4b.By examining the intensity of colorimetric signals following the same procedure as in the direct ELISA, the linear equation for the colorimetric assay was fitted as y = 101.65x+ 18.03 (R 2 = 0.99) between 0.01 mg L −1 and 0.20 mg L −1 of the protein.Naked eye recognizable SP-RBD-His concentrations reached at 0.1 mg L −1 level with a limit of detection (LOD) at 0.047 mg L −1 with the help of a smartphone and further analysis from software for a sandwich ELISA sensor, which is consistent with the fact that the sandwich ELISA tests normally produce more sensitive results than that of the direct ELISA. [38]he MF was employed in competitive ELISA for quantitative detection of chloramphenicol (CAP) as well, an antibiotic used in aquacultural farming but banned in the US currently.An antibody (Ab-CAP) was immobilized on NHS@MF, and the detection procedure is schematically described in Scheme 1c.Different from the other two assays, an unlabeled antigen (CAP) in samples and a labeled antigen (CAP-HRP) competes for binding to the immobilized antibody on the MF.A decrease in color signal from the MF membranes indicates the presence of the antigen in samples when compared to control groups with the labeled antigen.To explore the sensitivity of the MF in the competitive ELISA, a detection assay using different CAP concentrations (0 to 10000 ng mL −1 ) and measuring corresponding naked-eye readable color intensities are shown in Figure 4c.By examining the intensity of colorimetric signals following the same procedure employed in both direct and sandwich ELISA assays, a linear equation for the colorimetric assay was fitted as y = −122.88*x+ 226.13 (R 2 = 0.96) between 0.01 ng mL −1 and 0.20 ng mL −1 of CAP.Naked eye recognizable CAP concentrations reached at 1 ng mL −1 level with a limit of detection (LOD) at 0.096 ng mL −1 with the help of a smartphone and further analysis from the software for a competitive ELISA sensor.
Compared to literature results of other developed ELISA biosensing materials that can perform onsite detection of both Spike protein and CAP without using any specialized instruments, the NHS@MF-based sensor has the advantage of high sensitivity in a short testing time (Table 1).With the use of commercially available material, the commercial scalability and low cost make the MF more competitive and advantageous to be a new material for the development of biological sensors.

Impact of Sample Volumes in Immunoassays
Traditional ELISA assays use a very narrow range of sample volumes.Recently, p-ELISA managed to further scale down the volume through the miniaturization of sample sizes. [44]However, different from point-of-care clinical analysis, pollutants could be in very low concentrations in various scenarios, such as ground and surface water, treated industrial wastes, and food samples, [45,46] but in abundant volumes of samples.The framework structure of the NHS@MF retains high content of active sites (Figure 3b), and the MF membranes in varied thicknesses did not result in a significant increase in resistance to fluids (Figure 1).While increasing test sample volume could increase amounts of targets bound on the MF structure and magnify the intensity of signals, detection of trace amounts of targets could be achieved by flowing varied volumes of testing solutions through a filtering setup of the MF sensing materials.The volume-responsive performances of the MF materials in three immunoassays were further investigated with target sample volumes varied in 100 μL, 500 μL, 1 mL, and 2 mL, respectively.
Here, except for the added volumes of the analytes, the testing steps were the same as the protocols used in the earlier discussions.As shown in Figure 5a, in a direct ELISA sensing test, by changing the sample volume from 100 μL to 2 mL, the color intensities of the MF sensing material changed dramatically under varied target concentrations of 0.1 mg L −1 , 0.5 mg L −1 , and 1 mg L −1 of SP-RBD-His, respectively.The colorimetric sensing signal intensities of a direct ELISA sensor made of the NHS@MF showed consistent increases as the volumes of the samples in three different concentrations were raised.Higher target concentration resulted in much stronger signal intensity, while for the very low concentration (0.1 mg L −1 ) of SP-RBD-His, the intensity was increased coordinately with the increase of the sample volume (Figure 5a).Similar results were observed on the sandwich assay tests of the SP-RBD-His on the NHS@MF (Figure 5b) as well.In competitive ELISA tests, the signal intensity inversely changed corresponding to increased concentrations of CAP from 50 ng mL −1 to 200 ng mL −1 , respectively (Figure 5c).In all three types of ELISA sensing tests, increasing sample volumes led to profoundly stronger colorimetric signal differences for target concentrations respectively, demonstrating the potential in improving the sensitivity and broadening the ranges of detection limits.Compared to other sensing materials, such a unique feature of the NHS@MF allows sensors to handle varied sample volumes (Table 1) and potentially as flowing-through filtering sensor system for large-volume target solutions.

Simultaneous Detection of Multiple Targets
The structural features of the MF membranes also provide potential in additive sensing of multiple targets simultaneously in a flowing-through filtering sensor system.As illustrated in Figure 6a, antibodies (Ab-CAP and Ab-CPS) of CAP and chlorpyrifos (CPS) were immobilized on two different NHS@MF membranes, respectively.After blocking with the BSA buffer, these two membranes (5 mm in diameter) were mounted into a syringe needle pocket as a filtering sensing device, and 2 mL  of a mixture of CAP and CPS in varied concentrations, together with CAP-HRP and CPS-HRP in a concentration of 1 mg L −1 each, was filled into a 20 mL syringe and flow through the filtering needle with a flow rate of 6 mL h −1 , controlled by a Sy-ringeONE programmable syringe pump (NewEra Instruments, USA) (Figure 6b).As shown in Figure 6c, nine groups of the sam-ple mixtures were tested and collected through the sensing device following the varied concentrations of CAP and CPS in Figure 6c.The intensity of the colorimetric signals of the first-layer membranes showed an increasing trend, and that of the second-layer membranes showed a decreasing trend, indicating that simultaneous detection of CAP and CPS could be achieved without any interference of the two targets in the same system, an advantage of potential additive detection of multiple targets in one system.The total testing time was less than 40 min from running sample liquid through the sensor to the completion of tests.

Conclusion
Unique rapid, sensitive, additive, and volume-responsive colorimetric sensor materials were fabricated from using chemically modified framework melamine foam (MF), which can be applied in competitive, direct, and sandwich ELISA biosensors.The MF sensor materials demonstrated promising detection sensitivity to a SARS-CoV-2 spike protein with histidine tag (SP-RBD-His), a transmembrane protein of SARS-CoV-2 virus, and chloramphenicol (CAP), often used as an antibiotic.Naked eye recognizable SP-RBD-His reached at 1 mg L −1 level with a limit of detection (LOD) at 0.52 mg L −1 when supplemented by a smartphone for the direct ELISA sensor.In the case of the sandwich ELISA sensor, it's capable of detecting SP-RBD-His at a concentration of 0.1 mg L −1 by the naked eye and can reduce the LOD to as low as 0.047 mg L −1 with the help of a smartphone.In addition, using a competitive ELISA, chloramphenicol (CAP) can be detected at 1 ng/mL level with the naked eye and at 0.096 ng mL −1 with the help of a smartphone.Moreover, due to the excellent mechanical properties and framework structure of the MF, diffusion of the analyte through the different membrane layers is fast and homogeneous in all directions, making the MF suitable for the simultaneous detection of trace amounts of two or more targets in samples with large volumes in one integrated system.The successful fabrication of such sensor materials is expected to improve the sensitivity and broaden the applications of ELISA sensors for onsite and personal uses.

Experimental Section
Information on materials and chemicals, protocol for measuring diffusion of biomolecules in the chamber through different membrane materials, and characterization of the materials are shown in the supporting information file (S1-S3.).
Modification of Melamine Foam Membranes: A 0.5 g of MF in 1 mm thick slices and 5 mm diameter circular membranes were immersed into a N, N′-disuccinimidyl carbonate (DSC) modification solution (prepared by dissolving 5 g DSC and 0.4 g triethylamine (TEA) in 100 mL of 1,4 dioxane).The mixture was stirred for two hours at 70 °C.The modified membranes (NHS@MF) were thoroughly washed with 1,4-dioxane for 15 min twice and with acetone for 10 min and vacuum dried.
Immobilization of Proteins: The chemically modified MF membranes (NHS@MF) were immersed into His Tag monoclonal antibody with Alexa Fluor 647 (Ab-His-647) solution (200 μL) for 30 min and were washed several times using a phosphate-buffered saline (PBS) buffer before the following measurements.A confocal microscope (FV 1000 system, Olympus America) was used to observe the distribution of immobilized proteins on the membranes.Using a 60X bright field objective and 633 nm (Ar laser) excitation, 665-755 nm emission was collected for the Ab-His-647 conjugate used in this experiment. [28]The images were acquired at 640 × 640 pixels with 12.5 μs per pixel scanning speed.FTIR was employed to characterize the membrane before and after the modification and immobilization of proteins following the protocols.Fluorescent signals from Ab-His-647 conjugate were used to determine the concentration of proteins that are covalently immobilized on MF membranes according to the calibration curves (Figure S2, Supporting Information).NHS@MF in Direct, Sandwich, and Competitive ELISA Applications: Direct and Sandwich ELISA assays were tested on the NHS@MF membranes to detect a SARS-CoV-2 spike protein receptor-binding domain with His tag (SP-RBD-His). [29]For direct ELISA, 100 μL varied concentrations (ranging between 0 and 100 mg L −1 ) of the SP-RBD-His were added to the NHS@MF membranes (5 mm in diameter), and an incubation lasted for 30 min under gentle agitation.Then the membrane was exposed to 3% BSA (200 μL) to block the remaining active sites on the NHS@MF.Subsequently, 100 μL of 1 mg L −1 His Tag monoclonal antibody HRP (Ab-His-HRP) was added to each membrane.After 20 min, the membranes were first washed with a tween-20 (0.05%) solution and then washed with the PBS buffer and dried in air.25 μL of TMB substrate (ThermoFisher) was then applied onto the membranes, and membranes were placed in an LED lightbox (E mart).The colorimetric signal from the interaction between HRP and TMB substrate was captured by a smartphone (iPhone 8) and analyzed using Photoshop (Adobe) software. [47,48]To take pictures of each result, the smartphone was placed over membranes at a fixed distance of 50 cm.For Sandwich ELISA, 100 μL of the 5 mg L −1 SARS-CoV-2 spike protein RBD recombinant human monoclonal antibody (Ab-SP) was added to the membrane platform and incubated for 30 min.Then the membrane was exposed to 200 μL of 3% skim milk to block the remaining active sites. [49,50]After the blocking step, 100 μL varied concentrations (ranging between 0 and 100 mg L −1 ) of SP-RBD-His were added to the NHS@MF membranes, and the incubation lasted for 30 min under gentle agitation.Subsequently, 100 μL of 1 mg L −1 Ab-His-HRP was added to each membrane.After 20 min, the membranes were first washed with tween-20 (0.05%) and then washed with a PBS buffer and dried in air. [51,52]o obtain the outcome of colorimetric signals, the following steps were the same as in the direct ELISA.
A competitive ELISA assay was used to detect chloramphenicol (CAP) in aqueous systems, an antibiotic banned in use in the USA but is still used in other countries.First, 100 μL of 25 mg L −1 Anti-CAP antibody (Ab-CAP) was added to the membranes and incubated for 30 min.Then the membrane was exposed to 200 μL of 3% BSA to block the remaining active sites.After 30 min, 50 μL varied concentrations (ranging between 0 and 100 mg L −1 ) of CAP were mixed with 50 μL of 2 mg L −1 CAP-labelled horseradish peroxidase (CAP-HRP) conjugate, and 100 μL of the mixed solution was then added to each membrane.After 20 min, the membranes were first washed with a tween-20 (0.05%) solution and then washed with the PBS buffer, and lastly dried in air.The subsequent experimental steps were the same as the first two experiments.The red channel values (Rvalue) were read by using the Photoshop color histogram.The R-values were correlated to the concentrations of analytes. [30]To further investigate the dependence of the sample volumes of the materials in different types of immunoassays, the varied volumes of samples (100 μL, 500 μL, 1 mL, 2 mL) were applied to each experiment.In this study, except for the addition of varied volumes of analytes, the rest steps followed the same protocols as we mentioned above.The sample size of all experiments was 3.
Simultaneous Detection of Multiple Targets: A competitive ELISA assay was used to achieve simultaneous and on-site detection of multiple targets in samples.First, 100 μL 25 mg L −1 Ab-CAP and anti-chlorpyrifos monoclonal antibody (Ab-CPS) were added into two different groups of NHS@MF membranes separately.Then both groups of the membranes were exposed to 3% BSA to block the remaining active sites.Afterward, one membrane from the Ab-CAP immobilized group and one membrane from the Ab-CPS immobilized group were selected and placed together into a 20 mL syringe needle as a filtration column, as shown in Figure 6b.The order of different layers should be remembered.Then 2 mL of a mixture of CAP and chlorpyrifos (CPS) in specific concentrations, and 40 μL of a mixture solution of CAP-HRP and chlorpyrifos HRP (CPS-HRP) in a concentration of 100 mg L −1 each were filled into the syringe.The filtration flow rate was controlled by a SyringeONE programmable syringe pump (NewEra Instruments, USA) with a flow rate of 6 mL h −1 .Then the column was successively washed with 20 mL tween-20 (0.05%) and the PBS buffer.The membranes mounted in the syringe needles were collected separately, and 25 μL of TMB substrate (ThermoFisher) was then applied to the membranes.By analyzing the colorimetric signals obtained from the picture of a smartphone (iPhone 8), simultaneous detection of multiple targets can be achieved on-site.
Colorimetric Data Processing: When TMB was added to the membranes, the membranes were placed in an LED lightbox (E mart), and images were captured through the smartphone camera.The R channel value of the area of interest was obtained by using Photoshop software.
The red channel (R) values from RGB values represent the color intensity. [19,53]Here, the red channel intensity change could be represented by ΔRGB value, which was obtained by the RGB value difference between the white background and each membrane, as the equation of ΔRGB = RGB background − RGB membranes (1)   Statistical Analysis: All experiments were repeated three times.Data are expressed as mean ± standard deviations (SD).Intergroup comparison was analyzed by Student's t-test (two-tailed).The level of significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001.
The limit of detection (LOD) was calculated based on the standard deviation of the response and the slope using 3.3 /S, where  is the standard deviation of the response and S is the slope of the calibration curve.
The correlation coefficient (R) was used to measure the linear correlation between observed and predicted values.A value of P < 0.05 was considered statistically significant.All statistical analyses were performed using GraphPad Prism 8.0.2.

Figure 2 .
Figure2.a) Schematic illustration of the protein immobilization on NHS@MF and SEM images of MF, NHS@MF, and protein immobilized NHS@MF.b) Reaction of MF with DSC and proteins.c) FTIR results of MF at different steps: pristine MF, NHS@MF, and protein immobilized NHS@MF.d) Water contact angles of MF, NHS@MF, and protein immobilized NHS@MF.

Figure 3 .
Figure 3. a) Protein Immobilization distribution visualized by a laser scanning confocal microscope (lower ones are in higher magnification ratio).b) Loaded NHS amount on NHS@MF and NHS@NF after the modification of DSC (5%).Data are presented as mean ± SD, with n = 3 independent experiments.*P < 0.05 (two-tailed Student's t-test).c) Immobilized antibody amounts on NHS@MF from 5 mg L −1 , 1 mg L −1 , and 0.5 mg L −1 of 100 μL of antibody solutions.Data are presented as mean ± SD, with n = 3 independent experiments.d) Optical images and colorimetric signals generated from the interaction between immobilized HRP and TMB substrate on NHS@MF and pristine MF.Data are presented as mean ± SD, with n = 3 independent experiments.***P < 0.001 (two-tailed Student's t-test).

Scheme 1 .
Scheme 1. Mechanism of NHS@MF based a) direct ELISA, b) sandwich ELISA, and c) competitive ELISA.

Figure 4 .
Figure 4. Optical images and the calibration curve of membranes in the detection of SARS-CoV-2 spike protein RBD using a) direct ELISA approach and b) Sandwich ELISA.c) optical image and the calibration curve of membranes treated by varied concentrations of CAP using a competitive ELISA approach.All data are presented as mean ± SD, with n = 3 independent experiments.

Figure 5 .
Figure 5.Effect of the sample volume in color signal intensities of a) Direct ELISA, b) Sandwich ELISA, and c) Competitive ELISA.All data are presented as mean ± SD, with n = 3 independent experiments.

Figure 6 .
Figure 6.a) Schematic illustration of the mechanism of simultaneous multiple on-site targets detection.b) Photograph demonstrated the fast-flow device driven by a syringe pump.c) Optical image and ΔRGB values of membranes treated by the mixture of varied concentrations of CAP and CPS using a competitive ELISA approach.All data are presented as mean ± SD, with n = 3 independent experiments.

Table 1 .
Comparison of colorimetric sensors for detection of SARS-CoV-2 spike protein (SP RBD) and chloramphenicol (CAP) without employing any instrument.