Pluronic F127, potassium tetrachloroplatinate(II) (K₂PtCl₄), sodium tetrachloropalladate(II) (Na₂PdCl₄), hydrochloric acid (HCl), ascorbic acid, potassium carbonate (K₂CO₃), acetone, sucrose, HRP, hydrogen peroxide (H₂O₂), 3,3′,5,5′-tetramethylbenzidine (TMB), Tween-20, sucrose, phosphate-buffered saline (1× PBS, pH 7.4, 0.01 M), and bovine serum albumin (BSA) were purchased from Sigma-Aldrich (St. Louis, MO). Mouse anti-S. Enteritidis monoclonal antibody, mouse anti-E. coli O157:H7 monoclonal antibody, and goat antimouse IgG antibody were obtained from Abcam Inc. (Cambridge, MA). Goat anti-S. Enteritidis polyclonal antibody and goat anti-E. coli O157:H7 polyclonal antibody were purchased from KPL Inc. (Gaithersburg, MD). Nitrocellulose membrane, glass fiber, backing cards, and an absorbent pad were purchased from Millipore (Billerica, MA).

S. Enteritidis PT30 was from American Tissue Culture Collection (ATCC, Manassas, VA). E. coli O157:H7 EDL933 was obtained from the STEC center at Michigan State University. Listeria innocua (NRRL B-33197) was obtained from USDA ARS culture collection. Staphylococcus aureus (91.48) was a mastitic isolate from the cow mastitis case.³² These strains were maintained at −80 °C in trypticase soy broth (Becton, Dickinson and Company, Sparks, MD) supplied with 0.6% yeast extract (Fisher Scientific, Fair Lawn, NJ) (TSBYE) and 15% (v/v) glycerol. Bacteria were first activated in TSBYE at 37 °C for 8 h statically; then, 1:1000 were transferred to TSBYE for the second activation at 37 °C statically for additional 14 h. The twice activated culture was washed once with phosphate-buffered saline (PBS, pH 7.4) and resuspended in sterile PBS. The resulting bacterial suspension was used as a live cell in the recovery experiment of spiked food samples. For inactivated bacterial cell preparation, the above prepared bacterial suspension was heat-inactivated in boiling water for 30 min, followed by addition of formalin (J.T. Baker, Phillipsburg, NJ) to a final concentration of 0.5% (v/v). Inactivated bacterial cells were used to optimize the conditions and evaluate the performance of the developed dual LFIA.

Mesoporous Pd@Pt nanoparticles were synthesized according to our previous published method with modification.³³ Briefly, 20 mg of Pluronic F127 was ultrasonically dissolved in aqueous solution containing 1.8 mL of K₂PtCl₄ (20 mM), 0.2 mL of Na₂PdCl₄ (20 mM) solution, and 44 μL of HCl (6 M). The mixture was processed by an ultrasonic method for 4 h after adding 2.0 mL of ascorbic acid (100 mM). The final product was centrifuged at 10 000 rpm for 5 min, the resulting pellet was washed with 4 mL of acetone three times. After freeze-drying and resuspending in 5 mL of water, the concentration of the as-synthesized Pd@Pt nanoparticles was 2 mg/mL.

Antibody-modified Pd@Pt nanoparticle conjugations were prepared by a modified method according to the literature.³⁴ First, the pH of the Pd@Pt nanoparticles solution (2 mg/mL) was adjusted to 8.2–8.5 by adding 0.02 M K₂CO₃. Then, 5 μL of mouse anti-S. Enteritidis monoclonal antibody (1 mg/mL) or mouse anti-E. coli O157:H7 monoclonal antibody (1 mg/mL) was added into the 1 mL of adjusted Pd@Pt nanoparticles solution. After incubation for 60 min at room temperature, 10.0 wt % BSA was added to the mixture, followed by 30 min of incubation. After that, the mixture was centrifuged at 10 000 rpm for 20 min and washed with 1× PBS containing 1% BSA twice. Finally, the prepared antibody-modified Pd@Pt nanoparticle conjugations were collected and suspended in 100 μL of PBS buffer containing 2% BSA and 3% sucrose.

The dual LFIA consisted of a triangle-shaped sample pad (9 mm × 20 mm), two conjugate pads (4 mm × 5 mm), two nitrocellulose membranes (4 mm × 30 mm), two absorbent pads (4 mm × 15 mm), and an interoperable backing (Scheme 1A). The sample pad was made from glass fiber and saturated with 1× PBS containing 1% BSA and 0.25% Tween-20 and then dried overnight at room temperature. Two test lines (S and E) and two control lines were prepared by dispensing 30 μL of goat anti-Salmonella polyclonal antibody (or goat anti-E. coli O157:H7 polyclonal antibody) solution (1.0 mg/mL) and 30 μL of goat antimouse IgG antibody solution (0.5 mg/mL) at different locations on the nitrocellulose membrane using a BioDot BioJet BJQ 3000 dispenser (Irvine, CA). The distance between the lines was approximately 1 cm. The nitrocellulose membranes were then dried overnight at 37 °C and stored at 4 °C. Then, 2.5 μL of each antibody-modified Pd@Pt nanoparticle conjugation (20 mg/mL) was precoated on the corresponding conjugate pads after performing a short-timed ultrasound. These pads were assembled on the backing with an overlap of approximately 2–3 mm. Single LFIAs were cut at a width of 4 mm using a BioDot Paper Cutter module CM4000 (Irvine, CA). Finally, two LFIAs with different test lines (S or E) were assembled together on the triangle-shaped sample pad. The assembled dual LFIA was either used immediately or stored under dry conditions at room temperature for further tests.

The smartphone accessory was designed with computer-aided design software SolidWorks (Dassault System, MA) based on Apple iPhone 5s (Apple Inc., CA). The iPhone 5s was equipped with a rear iSight 8 megapixel sensor, 1.5 μm pixels, and an f/2.2 aperture. A commercial fused filament deposition 3D printer (Einstart S, Shining 3D, Hangzhou, Zhejiang, China) was used for the rapid prototyping of the smartphone accessory. A black poly(lactic acid) (PLA) filament was used for the smartphone holder to minimize light leakage, and a white PLA filament was used to utilize the ambient light for sample illumination (Scheme 1C). A plano convex lens with a diameter of 10 mm is placed in front of a camera to shorten the imaging distance and thus reduce the overall size of the device. The lens is fixed in a 3D-printed tube with three plastic threaded screws positioned in a triangular form to prevent lens movement from the operation vibration. The dual LFIA minicartridge was fabricated by laser cutting and engraving (trotec Speedy 300, Trotect., MI) a poly(methyl methacrylate) (PMMA) board for consistent and tight fit with the smartphone accessory. The smartphone accessory has a dimension of 73 mm (L), 35 mm (W), and 20 mm (H) and a weight of a total of 45.9 g, excluding the smartphone and the dual LFIA cartridge. An app was used to control the image capture parameters (ISO 200, exposure time of 1660.035 s, redGain: 2.77828, greenGain: 3.36689, and blueGain: 2.68713).

In a typical test, a dual LFIA was preset in the minicartridge. Afterward, 90 μL of running buffer (1× PBS containing 0.25% Tween-20) was mixed with 10 μL of sample, then loaded to the triangle-like sample pad of dual LFIA and migrated up due to capillary force. After 1 min, the black lines appeared visually. After adding 1 μL of TMB solution on test lines, the final blue signals were observed on the lines of dual LFIA within 10 min. When S. Enteritidis and E. coli O157:H7 are present in the sample, they will bind to their own antibody-modified Pd@Pt nanoparticle conjugations and then combine with capture antibodies on the test line-S and test line-E, thereby forming two visible blue lines (Figure S1A). If only S. Enteritidis or E. coli O157:H7 is present in the sample, only one blue line can be formed (Figured S1B,C). When a negative control sample was applied onto the sample pad, no blue line was observed in the test lines (Figure S1D). The control line indicated that the assay worked properly and the signal at the test line is reliable. Meanwhile, the minicartridge was inserted into the 3D-printed smartphone reader to obtain the results. An iPhone 5S (Apple) with an 8 megapixel (8 MP) camera was used. The image analysis software, ImageJ (NIH, MD), can provide a graphical readout of the peak area.

To demonstrate the proof-of-concept and its potential application in food samples, milk and ice cream samples were spiked with a known number of live bacteria and then detected using the proposed dual LFIA. Two percent low fat milk and double strawberry light ice cream samples were purchased from Walmart supermarket (Pullman, WA). Only negative samples were selected for being spiked by live bacteria. Milk (10 mL/g) and ice cream were added to 90 mL of sterile PBS from which we prepared samples in the final concentrations of bacteria at 1 × 10⁴ and 1 × 10⁵ cfu/mL. Then, the spiked samples were evaluated using the same above-mentioned condition and the recovery (%) of bacteria from milk and ice cream samples was obtained according to the above calibration curve results.

The morphologies of the nanomaterial were characterized by transmission electron microscopy (TEM) using a Philips CM200 UT (Field Emission Instruments) at an accelerating voltage of 40 kV. The ζ potential was measured using a zeta potential analyzer (Brookhaven Instruments Co.). Fourier transform infrared (FTIR) spectra were recorded in the range of 400–4000 cm⁻¹ with 0.5–4 cm⁻¹ resolution using a Tensor 27 FTIR spectrophotometer. All of the measurements were performed at room temperature.

To investigate the morphology of Pd@Pt nanoparticles, transmission electron microscope (TEM) images were collected. As shown in Figure 1A–C, the crystal growth was happened on the surface of Pd nanocrystals and on tips of the active branches³⁵ and the as-synthesized Pd@Pt nanoparticle size was approximately 35 nm. To verify the successful conjugation of antibody, we measured ζ potential and FTIR spectra of Pd@Pt nanoparticles before and after modification. As shown in Table S1, the ζ potential decreased from −40.6 to −32.5 mV for S. Enteritidis antibody conjugations and from −40.6 to −33.5 mV for E. coli O157:H7 antibody conjugations. As shown in Figure S2, a small peak around 1550 cm⁻¹ corresponding to amino groups was observed after antibody modification in the FTIR spectra,³⁶ which further confirmed antibody functionalization successfully.

The function of antibody-modified Pd@Pt nanoparticle conjugation is a critical factor for dual LFIA. Antibody-modified Pd@Pt nanoparticle conjugation is composed of two parts: anti-S. Enteritidis antibodies (or anti-E. coli O157:H7 antibodies) with recognition function and Pd@Pt nanoparticles with catalytic activity. First, we test whether the proposed antibodies could recognize S. Enteritidis and E. coli O157:H7. As shown in Figure 1D–G, TEM images confirmed that S. Enteritidis and E. coli O157:H7 could be captured by antibody-conjugated Pd@Pt nanoparticles. Then, we use antibody-conjugated Pd@Pt nanoparticles to catalyze the oxidation of TMB. The results demonstrated that the antibody-modified Pd@Pt nanoparticle conjugation had similar catalytic activity with the solution of HRP and unmodified Pd@Pt nanoparticles (Figure S3). Together, these results indicated that antibody-conjugated Pd@Pt nanoparticles possessed the dual functions (capturing bacteria and catalytic activity).

To achieve double capacity of lateral flow immunoassays, we banded single plexes together in parallel instead of adding two test lines in a single LFIA (Scheme 1A). The design of such dual-detection systems was based on careful consideration of cross-reactivity and the limitation of Washburn’s theory.^24,37 The proposed dual LFIAs could provide the same sensitivity of multiplexed detection as that of individual single LFIA detection. To further improve the sensitivity of the dual LFIA, we employed Pd@Pt nanoparticles instead of AuNPs as signal amplification (Scheme 1B). The Pd@Pt nanoparticles modified by antibodies owned dual functions: recognition and signal amplification. Owing to the intrinsic peroxidase-like catalytic activity, the antibody-modified Pd@Pt nanoparticle conjugations can generate stronger blue color in the presence of TMB and H₂O₂.^38,39 The enhancement of signals on the two test lines are critical for the detection of low level of pathogens. To provide a field-portable, cost-effective platform with high sensitivity, the smartphone-based device was developed to image and record results. The device consists of three parts, a 3D-printed cover, holder, and smartphone accessory, resulting in an integrated tool (Scheme 1C). Together, the proposed assay is portable and sensitive without any complex sample pre-enrichment and specialized instruments, making it a valuable detection tool for foodborne pathogens.

To acquire sensitive detection with enhanced signal, we compared the catalytic capability of different-sized Pd@Pt nanoparticles first, which were synthesized by adding various amounts of surfactants, Pluronic F127 (10, 20, and 30 mg) in solution. As shown in Figure S4, the as-synthesized Pd@Pt nanoparticle size was approximately 40 nm for 10 mg surfactants, approximately 35 nm for 20 mg surfactants, and approximately 15 nm for 30 mg surfactants. Insufficient gametes (10 mg) lead to larger particles and less surface area, which results in weaker catalytic capacity than that of 20 mg of gametes. Excessive gametes (30 mg) led to smaller particles and fewer branches, which were also not suitable as signal elements in our system due to weak catalytic capacity. After optimization, the amount of Pluronic F127 used to prepare Pd@Pt nanoparticles was selected to be 20 mg in the following experiments. To achieve the best performance of the dual LFIA, the reaction system were further optimized. The amount of antibody-modified Pd@Pt nanoparticle conjugations on the conjugate pad could affect the hybridization efficiency between bacteria and corresponding antibody and thus the performance of dual LFIA. As seen from Figure 2A, the peak area increased up to 2.5 μL and then decreased at 3.0 μL. The peak area loss occurred at a high concentration probably due to stereo hindrance between bacteria and antibody-modified Pd@Pt nanoparticle conjugations. Therefore, 2.5 μL of antibody-modified Pd@Pt nanoparticle conjugations was the optimal amount and was used in the further experiments. Figure 2B showed the results for the time after adding TMB solution. The peak area increased as the reaction time increased until 12 min and decreased as the reaction time increased after 12 min. No significant difference was observed in the peak area for assays between the reaction times of 8 and 12 min. A hybridization time of 10 min was selected for all subsequent tests. Running buffer also plays an important role, as it can minimize nonspecific adsorption and increase the sensitivity of the dual LFIA. As shown in Figure 2C, with increasing BSA, the corresponding response decreased significantly, especially when adding TMB solution. BSA-free 1× PBS was found to be the optimal running buffer, whereas 1% of BSA was added in the solution to prepare the sample pad. Figure 2D shows that the responses remained almost the same with varying amounts of Tween-20 in the running buffer with TMB solution. However, in the condition of the signal without TMB enhancement, the best result was obtained with 0.25% Tween-20, which was used as the optimal running buffer in subsequent detections.

The sensitivities of optimized dual LFIA were further estimated using sample solutions containing various concentrations of targets under the optimized experimental conditions. Figure 3A displays the typical photo image of the dual LFIAs with different bacterial concentrations ranging from 0 to 10⁸ cfu/mL. The detection limit of S. Enteritidis and E. coli O157:H7 without TMB and H₂O₂ were both approximately 10⁶ cfu/mL (Figure 3B,C), which was dramatically improved after adding TMB and H₂O₂ (Figure 3D). The blue band in the dual LFIA test zone was observed with as low as 10 cfu/mL of S. Enteritidis and E. coli O157:H7 without any pre-enrichment (Figure 3D), which was 100-fold more sensitive than previously published AuNP-based LFIAs and similar to reported HRP-enhanced LFIAs (Table 1). Quantitative detection was performed by recording the peak areas of the two test lines (S and E) (Figure 3E). The resulting calibration curves were exponential for both pathogens over the 10–10⁷ cfu/mL range (Figure 3F). The LODs based on 3.3 × standard deviation (SD)/slope of the calibration curve were calculated to be ~20 cfu/mL for S. Enteritidis and ~34 cfu/mL for E. coli O157:H7, indicating that approximately 2–3 cfus present in 100 μL of the sample can be quantified using this dual LFIA. These results well demonstrate the sensitive performance of our assay for simultaneous detection of two pathogens, which is contributed by the peroxidase-like catalytic activity of the Pd@Pt nanoparticles for signal enhancement and the parallel design of dual detection for noninterference.

The specificity of dual LFIA was further evaluated with other bacteria and PBS. We detected a mixture of S. Enteritidis and E. coli O157:H7 as well as S. Enteritidis, E. coli O157:H7, a mixture of S. Enteritidis and Listeria, a mixture of E. coli O157:H7 and S. aureus, and Listeria and S. aureus by dual LFIAs. No distinct blue band was observed on the test lines of dual LFIAs in the samples in absence of corresponding bacteria, as expected (Figure S5). No cross-reactivity with other bacteria was confirmed by Listeria and S. aureus (Figure S5). In addition, no signal was observed in the PBS control test, indicating negligible nonspecific adsorption of the optimized dual LFIA.

The stability test was performed by storing the dual LFIA at room temperature during the recent 30 day period using the same batch of strips. Dual LFIAs were stored in Ziploc bags under dark and dry conditions and intermittently measured (every 10 days). During the stability examination, their responses did not change significantly and the intensities of the test lines on the dual LFIA for 10⁴ cfu/mL S. Enteritidis and 10⁴ cfu/mL E. coli O157:H7 showed the coefficient of variation as less than 7% compared to that obtained with the newly prepared dual LFIA (data not shown). Thus, the dual LFIA was still usable after 30 days of storage at room temperature, indicating that the dual LFIA approach has a good stability.

To demonstrate the potential application of the dual LFIA in food samples, the device was used to detect live S. Enteritidis and E. coli O157:H7 spiked in milk or ice cream samples from 10² to 10⁵ cfu/mL. The estimated recoveries of the dual LFIA range from 91.44 to 117.00% (Table 2), which clearly indicated that the developed method is capable of detecting live S. Enteritidis and E. coli O157:H7 in real food samples.