Advancing In Situ Food Monitoring through a Smart Lab‐in‐a‐Package System Demonstrated by the Detection of Salmonella in Whole Chicken

With food production shifting away from traditional farm‐to‐table approaches to efficient multistep supply chains, the incidence of food contamination has increased. Consequently, pathogen testing via inefficient culture‐based methods has increased, despite its lack of real‐time capabilities and need for centralized facilities. While in situ pathogen detection would address these limitations and enable individual product monitoring, accurate detection within unprocessed, packaged food products without user manipulation has proven elusive. Herein, “Lab‐in‐a‐Package” is presented, a platform capable of sampling, concentrating, and detecting target pathogens within closed food packaging, without intervention. This system consists of a newly designed packaging tray and reagent‐infused membrane that can be paired universally with diverse pathogen sensors. The inclined food packaging tray maximizes fluid localization onto the sensing interface, while the membrane acts as a reagent‐immobilizing matrix and an antifouling barrier for the sensor. The platform is substantiated using a newly discovered Salmonella‐responsive nucleic acid probe, which enables hands‐free detection of 103 colony forming units (CFU) g−1 target pathogen in a packaged whole chicken. The platform remains effective when contamination is introduced with toolsand surfaces, ensuring widespread efficacy. Its real‐world use for in situ detection is simulated using a handheld fluorescence scanner with smartphone connectivity.


Introduction
Foodborne illness represents a growing crisis, with an estimated annual caseload surpassing 600 million globally-largely attributed to the consumption of pathogencontaminated food products. [1]In the United States alone, there are an average of 37 million annual incidences of foodborne illness, with associated treatment costs exceeding $14 billion. [2,3][6] While various procedural and regulatory interventions have been introduced to reduce the incidence of food contamination, testing all foods as they transition through the food production pipeline remains instrumental in illness prevention.That being said, ready-to-eat (RTE) food products are of particular concern, given the lack of a cooking step prior to consumption. [7]Unfortunately, culturing protocols remain the industry standard for such testing, despite their costly, laborious, and timeconsuming nature.
10][11][12][13] However, such systems yield significant food waste and operate under the assumption that the food products selected and opened for testing are representative of all products within a given batch.[16][17][18][19][20] Specifically, functional bacteriophages, oligonucleotides, and antibodies have all been used in situ as highly specific biorecognition probes for the detection of pathogens within complex food matrices.23] Namely, these probes all rely on complementary reagents such as buffers and metal ions for functionality, meaning that food samples must be extensively treated with these agents prior to testing-a measure that severely alters their organoleptic properties.Further, reported tests largely involve the contamination and testing of small food samples, wherein target bacteria are localized to the test site.In reality, contamination can occur at any given site on bulk food products with large surface areas.Given that placing sensors across an entire food surface is commercially unfeasible, testing in a manner that provides results representative of the entire food sample is difficult.Last, these sensors largely detect target pathogens within liquid test media, which is not easily isolated within closed packaging systems.Consequently, there exists a critical gap between the myriad of sensing platforms discovered in recent works and their real-world implementation into food packaging for real-time food monitoring, necessitating the innovation of traditional packaging.
[26][27][28][29] Under this premise and based on the aforementioned limitations, we propose that an ideal food monitoring packaging system should: 1) enable sensor visualization without disrupting the closed package, 2) localize all sample solution released by the food matrix onto the sensor, 3) retain necessary reagents within the food packaging in a manner that minimizes organoleptic alternations to the adjacent food, and 4) facilitate analyte diffusion from the food matrix onto the sensor surface.Achieving these four objectives would then enable 5) the in situ integration of novel biorecognition elements, to advance complete hands-free, real-time monitoring of packaged foodsthe primary objective of this work.Importantly, the proposed sys-tem must employ food safe materials to ensure regulatory and commercial viability.
Consequently, we have developed a three-pronged food-safe packaging system consisting of a newly designed food packaging tray to meet objectives (1) and ( 2) and a membrane interface to meet objectives (3) and ( 4).The developed "Lab-ina-Package" platform is completed through the incorporation of pathogen sensors, which act to facilitate sensing within individual packages.Importantly, the generalized nature of the developed packaging tray and membrane interface enable the universal incorporation of various pathogen sensing platforms.For this study, a new synthetic fluorescent nucleic acid probe (FNAP) highly specific to S. Typhimurium was developed to meet objective (5).The probe functions as a highly specific substrate for cleavage by the RNase H2 protein of S. Typhimuriuma reaction that can be monitored through the integration of a fluorophore-quencher pairing into the FNAP construct.Incorporation of the newly discovered probe with the tray and membrane, yielded the complete Lab-in-a-Package platform.Artificially contaminated bulk, RTE chicken products were subsequently stored within this sensor-embedded packaging system, wherein high detection performance was observed.Next, recognizing that different avenues of contamination localize pathogens on food products in different ways, bulk, RTE chicken products were exposed to S. Typhimurium-contaminated culinary instruments, gloves, and surfaces and subsequently monitored using Lab-in-a-Package.The platform's high detection performance was maintained.Finally, the platform's real-world use was fully simulated using a handheld fluorescence scanner with smartphone connectivity, wherein successful target detection was observed without any disruption to the closed package.

Lab-in-a-Package Platform Design
Overarchingly, we sought to develop Lab-in-a-Package to facilitate test sample localization, in situ reagent incorporation, target diffusion, and fouling prevention on the sensing interface.Collectively, these efforts contribute to pathogen detection-all while eliminating the need to open food packaging or manipulate food products.The integration of the aforementioned universally applicable novel food packaging tray design, membrane interface, and pathogen sensor is shown in Figure 1.Importantly, all materials considered in this study were selected from the Indirect Food Substances provisions, ensuring regulatory viability in accordance with the Generally Recognized As Safe designation from the US Food and Drug Administration. [30]

Test Sample Localization Optimization
A dramatic shift in packaging design compared to traditional packaging was performed to facilitate sensor visualization and localization of sample fluids.A sensing window was first introduced to enable the integration of fluorescent sensing interfaces that can be monitored without opening packaged foods.Optimization of fluid localization from target food matrices was then explored through three packaging trays with varying levels of incline-45°, 60°, and the traditional 90°, which were fabricated using 3D printing (Figure 2a; Figures S1 and S2, Supporting Information).Here, the angle refers to the incline at the fluidpackage interface.The 60°model was developed as an intermediary model to substantiate any trends observed with changes in incline angle.
Fluid transfer efficiency of the models was first assessed, where the 90°model significantly outperformed the other models (P < 0.01).This was attributed to its steeper angle, inducing the strongest forces of downward acceleration (Figure 2b).Accordingly, the 60°model also outperformed the 45°model.Yet, while fluid transfer efficiency offers an important preliminary understanding of the models' fluid transport capabilities on a droplet scale, macroscale fluid accumulation at a central collection site better defines suitability for the desired application.As such, subsequent tests focused on characterizing the localization efficiency of each model.This was first accomplished by measuring fluid localization over time (Figure 2c).Here, the 45°model significantly outperformed the 60°(P < 0.001) and 90°models (P < 0.0001).This improved performance was attributed to the 45°model offering a consistent decline directly into the sensor window for the milliliter-scale test volumes.Fluid localization was then evaluated based on the volume localized onto the sensing window within a fixed period of time with both phosphate-buffered saline (PBS) (Figure 2d) and chicken purge (Figure 2e).Specifically, this was represented as the percentage of volume localized and was calculated according to Equation (1) below.
Here, V f represents the final volume collected from the base of the packaging tray after a constant timepoint, and V i is the initial volume of fluid applied to the trays at the start of the study.In both studies, the 45°model exhibited significantly higher fluid localization compared to both the 60°(P < 0.001, 0.01) and 90°models (P < 0.0001, 0.001).However, overall localization with chicken purge was lower due to a combination of a viscosity-induced reduction in fluid transfer rate and a longer incubation time.Together, these factors led to purge drying along the fluid-package interface prior to reaching the sensor window, decreasing the fluid volume available for localization.A timepoint analysis of this study was also briefly conducted (Figure S3, Supporting Information).Ultimately, the 45°model was selected for use within the final Lab-in-a-Package platform given its superior localization when compared to the collective properties exhibited by the other models as quantitatively summarized in Table S1, Supporting Information.

Absorption and Diffusion-Focused Materials Characterization
With regards to the membrane interface, the buffer absorption, macromolecular filtration, and target diffusion properties of five candidate materials were evaluated (Figure 1c).Namely, cotton, cotton-cellulose, cellulose, cellulose-polyester, and polyester materials were considered, owing to their inherent biocompatibility, filtering potential, and permeability.The candidate membranes were imaged via optical microscopy (Figure S4, Supporting Information) and scanning electron microscopy (SEM) (Figure 2f) to visualize their fibrous structures.Cotton exhibited the most unique structure due to its convoluted fibril arrangement, with comparatively larger pores. [31]All materials exhibited low fluorescence across the visible light spectrum, substantiating further characterization given their potential applicability to fluorescent systems (Figure 2g).
After optical characterization, the absorption capacity of candidate membranes was assessed by volume (Figure 2h) and over time (Figure S5, Supporting Information).Here, cottoncellulose significantly outperformed all other candidate membranes (P < 0.0001).This was attributed to the abundance of cellulose hydroxyl groups present within the matrix, giving the material high hydrophilicity and affinity for moisture. [32]While pure cellulose offers a larger number of hydroxyl groups, cottoncellulose has comparatively higher crystallinity, resulting in the formation of complex hydrogen bonds that further elevate hydrophilicity and in turn, buffer absorption. [33]As polyester fibers are densely packed and lack polar groups, they are intrinsically hydrophobic and thus had the lowest absorption. [34]Buffer retention was subsequently evaluated within a closed package environment, wherein all of the candidate materials exhibited limited changes in buffer volume after 120 h (Figure S6, Supporting Information).
Next, diffusion testing was performed to ensure that agents of interest can permeate through the membrane matrix onto the sensing interface.Polyester, cellulose-polyester, and cotton membranes performed best in preliminary buffer diffusion studies (Figure 2i).Bacterial diffusion was subsequently evaluated using E. coli.A preliminary control study confirmed that the candidate materials did not influence bacterial proliferation or survival (Figure 2j).Two bacterial diffusion studies were then performed with both unsaturated and buffer-saturated membranes to evaluate the effects of hydration on bacterial diffusion (Figure 2k,l).In both studies, cotton significantly outperformed all other candidate materials, facilitating the diffusion of E. coli at a concentration of 10 9 CFU mL −1 in both unsaturated (P < 0.001) and saturated states (P < 0.01), given its unique fibrous arrangement.Membrane performance was also tested with S. Typhimurium (Figure 2m-o).Again, no effect on bacterial proliferation or survival was observed.Similar trends were observed with the five membrane candidates with regard to bacterial diffusion, except that the magnitude of diffusion was much lower in the unsaturated state with S. Typhimurium.This is attributed to its larger size, requiring saturation-mediated pore expansion for substantial diffusion opportunities. [35,36]inally, porosity was also directly quantified as the percentage of the total top surface covered by pores within previously collected SEM images (Figure S7, Supporting Information).Here, the cotton membrane had the best combination of surface poros-ity and low fiber density, making its pores better suited for the facilitation of buffer and bacterial diffusion.Consequently, cotton membranes were selected given their high buffer retention, diffusion, and porosity when compared to the other materials, as summarized in Table S2, Supporting Information.While the material offered comparatively lower absorption capacity, we hypothesized that this drawback could be overcome through the use of a more concentrated reagent mix.
Last, the benefits of a cotton membrane as a physical barrier between a food sample and the sensing substrate were explored.Specifically, recognizing that food represents a very complex matrix, fouling of the sensing interface is a major concern.The optical density of crude chicken purge and chicken purge processed through cotton membranes was measured for each sample (Figure S8a, Supporting Information).The collected values showed that the cotton membrane significantly decreased the presence of macroscale entities present within the chicken purge test samples (P < 0.0001).This effect can be expected to be amplified with the saturation of the cotton membrane, as wet cotton fibers offer a larger physical footprint and higher mechanical strength. [31]To visualize anti-diffusive effects toward macroscale entities, buffer-saturated cotton membranes were visualized via SEM after chicken purge processing (Figure S8b, Supporting Information).These images indicated lipid deposition onto the cotton fibers at a macroscale, while retaining sufficient porosity to permit analyte and buffer diffusion at the microscale, further substantiating the use of this membrane within the proposed platform.

S. Typhimurium Sensor Development
With packaging and membrane optimization complete, we next sought to incorporate a compatible sensor into the developed platform.While we have previously reported E. coli-responsive, real-time fluorescence sensors for potential in situ incorporation, growing concerns surrounding S. Typhimurium contamination led us to develop a new sensor altogether. [15,23]Using systematic evolution of ligands by exponential enrichment (SE-LEX), a highly sensitive nucleic acid probe that cleaves in the presence of RNase H2 from S. Typhimurium was identified.This probe was hybridized with a substrate strand embedded with a fluorophore-quencher pairing.S. Typhimurium-induced cleavage enabled quencher separation, yielding an increase in fluorescence (Figure 3a).The complete sequence for this construct is provided in Table S3, Supporting Information.
The surface-immobilized sensor was developed through the covalent attachment of an aminated, FITC-labeled version of the ) Time required for a water droplet to fall down packaging edge.c) Time required for 5 mL of buffer to reach sensing window when dispensed at a rate of 0.5 mL s −1 .d) Percentage of original PBS volume localized on sensing window after 1 min when dispensed at a rate of 0.2 mL s −1 .e) Percentage of original chicken purge volume localized after 24 h at 37 °C.f) SEM images of candidate membranes at 100× with overlays at 500×.g) Mean background fluorescence of candidate membranes.h) Absorption capacity of candidate membranes.i) Volume of buffer diffused through candidate membranes after 2 min.j) Membrane effects on bacterial growth following a 6 h incubation with E. coli.k,l) Bacterial diffusion through unsaturated (k) and buffer-saturated (l) membranes onto underlying substrates following a 6 h incubation at 37 °C with E. coli.m) Membrane effects on bacterial growth following a 6 h incubation with S. Typhimurium.n,o) Bacterial diffusion through unsaturated (n) and buffer-saturated (o) membranes onto underlying substrates following a 6 h incubation at 37 °C with S. Typhimurium.All reported values represent the mean of all samples with error bars representing sample standard deviation.All asterisks represent significant differences at corresponding significance levels.S. Typhimurium-responsive probe to polyethylene food packaging substrates.To ensure an adequate sensor signal, probe surface density was optimized and subsequently quantified to be 1.3 × 10 −5 nmol per array spot (Figure S9, Supporting Information).The sensitivity, stability, and specificity of this surface sensor were then evaluated.Given that chicken food matrices are most commonly contaminated by S. Typhimurium, and RTE foods offer the highest risk of illness, RTE chicken products were selected as the target matrix for our studies.Sensitivity testing was thus performed using contaminated chicken purge samples.The effects of chicken purge on bacterial proliferation and survival were first assessed, to ensure consistency between reported and experimental bacterial concentrations.Chicken purge spiked with bacteria, bacteria resuspended in buffer, and chicken purge alone were all selectively plated (Figure S10, Supporting Information).No significant changes in bacterial count were observed within any of the tested conditions.
Subsequent sensitivity testing was performed using spiked chicken purge, wherein sensors were incubated with the bacterial test solutions for 8 h at 37 °C-environmental conditions in line with RTE chicken storage within grocery stores (Figure 3b).The limit of detection of the developed sensor was confirmed to be 10 3 CFU mL −1 of S. Typhimurium, where a distinct 2.34-mean fold-change was observed following incubation-significantly higher than the non-specific 1.50-mean fold-change of the control condition (P < 0.05).Chicken purge samples contaminated with S. Typhimurium concentrations ranging from 10 8 to 10 4 CFU mL −1 were also tested, showing a linear relationship between bacterial concentration and mean fluorescence fold-change, with the 10 8 CFU mL −1 showing the highest significant mean fold-change at 4.36 (P < 0.0001).The developed sensor exhibited a significant linear operating range based on linear regression analysis (R 2 = 0.98, P < 0.001), further validating its detection efficacy (Figure S11, Supporting Information).The detection range of this sensor is summarized by Equation ( 2) below, where y represents the fluorescence signal fold-change and x is the logarithmic S. Typhimurium concentration in CFU mL −1 .
Importantly, higher variations in fluorescence signals were observed at high bacterial concentrations.This is of limited real-world concern, however, as industry guidelines for S. Typhimurium monitoring seek positive versus negative detection results.Variations at these high concentrations do not affect the accuracy of such Yes/No test results.The functionality of the sensors was then further evaluated at 4, 25, and 45 °C (Figure 3c).Sensing performance was maintained at 45 °C, whereas reduced activity was observed at 25 °C.Unfortunately, sensing activity was not observed at 4 °C.These results are in line with the temperature-dependent properties of the nucleic acid probe, wherein its activity is limited at low temperatures and peaks around 37 °C.
Next, the stability of these sensors was further established to ensure long-term storage viability.First, covalent attachment of the nucleic acid probe onto the polyethylene substrate was confirmed to ensure the sensors could withstand induced shear stresses (Figure 3d).Here, probes that were covalently attached with crosslinker retained 85% of their original fluorescence signal, significantly higher (P < 0.0001) than the 14% retention of the probes that were not covalently attached.To evaluate sensing performance after prolonged storage, sensors were tested 3 months after fabrication (Figure 3e).Sensors tested with 10 6 CFU mL −1 S. Typhimurium exhibited a slight increase in signal, which can be attributed to the variation that is observed at higher target concentrations.With all other concentrations, sensing performance deteriorated to a degree, but a significant limit of detection of 10 4 CFU mL −1 was obtained (P < 0.0001), confirming sensor resilience.
Specificity testing then involved incubating the sensors with S. Typhimurium and other common foodborne pathogensnamely, Klebsiella pneumoniae, E. coli O157:H7, Pseudomonas aeruginosa, Listeria monocytogenes, and Bacillus subtilis (Figure 3f).The developed sensors exhibited a significantly higher 2.59-mean fold (P < 0.0001) change in fluorescence when incubated with S. Typhimurium, compared to the less than 1.10-mean fold-change observed with other bacterial species.Accordingly, after sensitivity, stability, and specificity were confirmed, the sensors were then applied for proof-of-concept testing of the developed packaging platform using commercial-scale, solid food products.

Proof-of-Concept Testing of Sensor-Embedded Packaging Platform
The complete Lab-in-a-Package platform was then tested by integrating the newly developed FNAP sensor into the aforementioned packaging and membrane system (Figure 4a).The final in situ detection platform consisted of a concave packing tray with a 45°incline and a sensing window, an S. Typhimurium-responsive real-time fluorescence sensor embedded within this window, and an adjacent buffer-infused cotton membrane (Figure 4b).Recognizing that chicken exhibits significant fluorescence noise at the FITC emission/excitation wavelengths of 490 nm/525 nm, we transitioned to a TAMRA fluorophore-labeled version of the probe with excitation and emission wavelengths of 557 nm/576 nm (Figure 4c). [37]Importantly, the activity of enzymatic oligonucleotide probes largely relies on the presence of divalent metal ions, substantiating the need for a buffer immobilizing membrane within this in situ monitoring platform.The developed S. Typhimurium-responsive probe in particular, offers excellent performance with Mg 2+ ions-a metal safe for consumption, which made the leaching of buffer from the membrane onto adjacent foods of limited concern.The concentration of the MgCl 2 ions used within the Labin-a-Package platform was optimized for surface-based sensing, to a value of 30 mm (Figure 4d).
RTE rotisserie chicken products were weighed and placed within inclined packaging trays, which already contained FNAP sensors and MgCl 2 saturated membranes (Figures S13,S14 and Video S1, Supporting Information).The collected chicken purge was spiked with S. Typhimurium, and this contaminated purge was then applied to the test products to perform a sensitivity analysis of the complete system, with concentrations ranging from 10 6 to 10 2 CFU g −1 .Concurrently, uncontaminated chicken purge was applied to control chicken samples.The samples were then incubated at 37 °C for 8 h to simulate grocery store RTE chicken storage environments.After 8 h, sufficient localization was visually confirmed through the accumulation of significant chicken purge on the cotton membrane and adjacent sensing interface.Specifically, the top surface of the membrane was coated with macroscale fouling agents such as lipids, confirming the membrane's anti-fouling capabilities within an in situ environment.With regards to sensitivity, sensors incubated with contaminated chicken samples exhibited a significantly higher (P < 0.0001) mean fold-change of up to 2.83 at 10 6 CFU g −1 compared to the 1.09 mean fold-change of the control samples (Figure 4e).The limit of detection of Lab-in-a-Package was determined to be 10 3 CFU g −1 , which exhibited a significant mean fold-change of 1.54 (P < 0.0001) compared to the control.The operating range of the complete system was significant (R 2 = 0.98, P < 0.05) based on linear regression analysis (Figure S12, Supporting Information).The linear operating range of Lab-in-a-Package is summarized by Equation ( 3).In this model, y represents the fluorescence signal fold-change and x is the logarithmic S. Typhimurium concentration in CFU g −1 .
Further, a bacterial growth study was conducted to assess opportunities for improved sensitivity using this platform.It was determined that 10 2 CFU mL −1 of S. Typhimurium suspended in chicken purge grows to 10 4 CFU mL −1 within 4 h.This suggests that extended incubation periods may offer improved detection limits (Figure S15, Supporting Information).
The specificity of Lab-in-a-Package was also studied through sensor incubation with a mixture of common food contaminants.Specifically, FNAP sensors were placed at the base of the packaging and then incubated with control chicken samples contaminated with a 10 6 CFU g −1 mixture of E. coli O157:H7 and L. monocytogenes.Test samples also contained S. Typhimurium (Figure S16, Supporting Information).Here, the test samples which contained S. Typhimurium exhibited a significantly higher mean fold-change of 2.70 compared to the 1.25 mean fold-change of the control samples (P < 0.0001).To further confirm the specificity of the system, a target verification study was also performed (Figure S17, Supporting Information).Here, spiked purge and purge localized on the sensing window after 8 h of incubation were selectively plated.Insignificant differences were observed between the two samples.These results confirmed that the developed packaging platform successfully mediates the localization of representative test solution onto the sensing interface for detection, alongside buffers required for biomolecular functionality with a degree of sensitivity and specificity.
Next, to further simulate real-world food contamination, the platform was used to detect contamination within samples that were spiked via means of handling and processing.Specifically, the test chicken samples were contaminated through contact with a contaminated knife, glove, and surface (Figure 4f) spiked with a solution corresponding to 10 7 CFU g −1 of the corresponding chicken sample.The contaminated chicken samples exhibited high mean fold-changes of 4.03 (P < 0.0001), 3.73 (P < 0.0001), and 3.00 (P < 0.001) following contamination induced by the knife, glove, and surface, respectively, further validating the presented platform as a means of in situ contamination detection in real-world settings (Figure 4g).
To further evaluate the generalizability of the presented platform, contamination detection was performed in spiked lettuce samples at 25 °C (Figure S18, Supporting Information).Test samples that were spiked with 10 6 contaminated wash water exhibited significant mean fold-changes of 1.51 (P < 0.001) compared to the 1.15 fold-change of the control samples.These results suggest the potential use of this system for other food products including produce and for consumer goods.
Finally, we sought to simulate real-world use of the developed platform to evaluate its full in situ sensing capabilities using a portable handheld fluorescence scanner that visualizes images onto a smartphone (Figure 4h).The capabilities of the handheld scanner were first evaluated with the sensor alone, wherein a significant mean fold-change of 2.76 (P < 0.0001) was observed following incubation with a 10 8 CFU mL −1 solution of S. Typhimurium (Figure 4i).When used to monitor a sealed Lab-in-a-Package set-up containing 10 6 CFU g −1 contaminated RTE rotisserie chicken, a significant fold-change of 3.27 (P < 0.01) was observed without any disruption to the closed food package (Figure 4j).Using such a portable system over a laboratory-scale microscope makes sensor monitoring possible across the entire food production pipeline on an individual product level, emulating real-time, hands-free, in situ detection.

Conclusion
We have developed Lab-in-a-Package, a revolutionary solution to advance in situ, real-time food contamination detectionbridging the gap between the myriad of developed food sensors and their adoption into food products at the retail and consumer levels.The platform combines a newly designed food packaging tray and a buffer-infused membrane to address the complete lack of an in situ monitoring-compatible packaging platform.This combinatory approach meets the key objectives required to facilitate real-time, hands-free detection as it: 1) enables sensor imaging within a closed package format, 2) localizes sample solution onto the sensing interface, 3) retains all necessary buffers inside the food packaging, 4) facilitates sample diffusion from the food matrix to the sensor, and 5) enables the in situ incorporation of novel biorecognition elements.The redesigned 45°i nclined packaging model displayed the highest levels of fluid localization and fluid transfer when compared to a traditional food packaging tray and an intermediately inclined one.Moreover, the selected cotton membrane was rigorously tested through a variety of experiments to insure adequate diffusivity and buffer retention.Complete proof-of-concept testing with a newly developed S. Typhimurium sensor demonstrated the successful in situ detection of this target pathogen within food products, with high sensitivity and specificity.The efficacy of the developed solution was also confirmed via application-specific testing that involved contaminating food samples with an S. Typhimurium-contaminated glove, surface, and knife, to better simulate real-world conditions.Finally, real-world use was simulated using a handheld fluorescence scanner attached to a smartphone for sensor visualization.This solution has immense application for a variety of food samples beyond RTE chicken, including the packaging trays used for other meats and seafood products, as well as for the plastic containers used for produce.Its generalized design, costeffectiveness, and food-safe nature positions it well to forward commercial implementation of in situ food sensors.Overarchingly, Lab-in-a-Package represents a paradigm shift, as the first packaging technology designed for in situ monitoring.Whereas existing traditional food testing protocols only test select products at central laboratories via time-consuming and laborious procedures, this system offers: 1) continuous monitoring of packaged foods, 2) with results available on an hour-scale, 3) without the need for any product processing, 4) in a manner that does not destroy tested foods.Collectively, these traits define the system's delivery of 5) individual product monitoring in situ.
Packaging Tray Fabrication: All three packaging trays and their associated 2D drawings were developed using 3D computer-assisted design (CAD) software (Autodesk Fusion) and then 3D printed using PLA filament (Ender 3 V2, Shenzhen Creality 3D Technology Co., Ltd., China).These packages were then smoothened using acetone to lower the coefficient of friction on the fluid-interface.All packages were printed at a 50% scaleddown rendering to improve characterization efficiency.
Fluid Transfer and Localization Efficacy: The fluid transfer efficiency involved recording the time it took for a droplet to transport down the edge of each packaging model, when dispensed at a rate of 16.54 μL s −1 using an automated syringe (DSA30, Krüss Scientific, Hamburg, Germany).This study was video tapped using the Krüss Advance software and then viewed in slow motion to accurately quantify time measurements.The time required to localize 5 mL of deionized water into the sensing window was quantified through the time required for 5 g of water to collect within a weigh boat positioned directly below the sensing window, as PBS was dispensed onto the trays' edges from above.Here, PBS was dispensed at a rate of 0.5 mL s −1 .All studies had at least triplicate measurements.Fluid localization based on volume was characterized as the percent of solution that reached a weigh boat collection basin that was attached to the bottom of each packaging model, relative to the total applied volume.PBS studies involved dispensing the buffer over 1 min at a rate of 0.2 mL s −1 .Chicken purge studies involved applying 4 mL of chicken purge onto 62 g samples of RTE chicken and assessing the percent volume collected within an attached collection basin after 24 h storage at 37 °C.Timepoint readings were also done to develop trends for volume localization across a 60 s time span.Here, 5 mL of solution were applied for 10 and 20 s on each packaging model and 10 mL of solution were applied for timepoints from 30 to 60 s and the total volume of fluid collected in collection basins after each time period was recorded.All values were considered relative to the initial volume applied.
Preliminary Membrane Characterization: All membrane candidates were first cut to the same size of 4 cm × 2.5 cm.Their thicknesses were then matched to the same general thickness of ≈1.1 mm.The samples were then microscopically imaged using an inverted microscope (Nikon Eclipse Ti2, Nikon Instruments Inc.) at 4× and 10× magnifications.Fluorescence analysis was done on the same samples using the same imaging system, with at least three samples imaged for each membrane.Samples were imaged across DAPI, FITC, TRITC, and Cy5 fluorescence wavelengths.Membrane samples were also cut to ≈1 × 1 cm sizes and then mounted using carbon tape and nickel paste.A sputter coater (Polaron model E1500, Polaron Equipment Ltd., Watford, Hertfordshire) was then used to coat the samples with 10 nm of gold, which were then imaged using the TESCAN VEGA-II LSU SEM.
Membrane Absorption Quantification: The absorption capacity of each candidate membrane was assessed using 3.5 cm × 2.0 cm × 1.1 mm samples, wherein samples were weighed when dry, submerged in PBS for 1 min, and then reweighed.The density of PBS was then used to convert the weight readings into volumetric measurements, which yielded the total absorption capacity of each membrane.To prevent measurement error, the hydrated samples were briefly shaken to remove residual, unabsorbed PBS.Additionally, a timepoint study was also created to confirm that the membranes were saturated in 1 min.This involved repeating the above protocol for different submersion times spanning 5 s to 30 min.
Membrane Retention Analysis: Membranes were submerged in PBS for 1 min, shaken to remove unabsorbed solution, and then weighed as the initial starting weight.Membrane weight was measured and then converted to volumetric values using the density of PBS.These samples were then stored within packaging.Membranes were reweighed at 24 and 120 h to quantify the volume of buffer retained within the membranes over time.
Membrane Diffusion Analysis: Buffer diffusion was assessed by quantifying the volume of PBS that diffused through 5 cm x 2.5 cm pre-saturated membrane samples over 2 min of continuous flow at a rate of 0.1 mL s −1 .Samples were supported by a plastic scaffold and placed on top of a collection basin.PBS was then pipetted onto the top surface of the membrane and the amount of buffer which diffused through each membrane into the basin below was collected and quantified.Triplicate measurements were obtained to reduce experimental error.
Membrane Bacterial Studies: To assess membrane effects on bacterial proliferation and survival, 1 cm × 1 cm membrane samples were incubated with 10 8 CFU mL −1 of bacteria for 6 h.The contaminated membranes were then vortexed for 1 min to extract bacteria from the membrane into solution.This solution was then serially diluted and plated onto Gram-negative selective MacConkey agar (MilliporeSigma) plates.A control (no membrane) condition consisting of 10 8 CFU mL −1 of bacteria was maintained for the same incubation period and concurrently plated.The bacterial plates were incubated overnight at 37 °C.Following incubation, the total number of colony-forming units was counted for each membrane and compared to that of the control.Bacterial diffusion through the membranes was assessed across both unsaturated and saturated membranes, where the latter had 1 mL of PBS buffer added to the surface.Both groups of membranes were then placed on top of glass substrates and 10 6 CFU mL −1 of bacteria was distributed on top of each membrane candidate.After 6 h, any solution which diffused through the membrane onto the glass substrate below was collected, serially diluted, and plated onto the same selective plates.Once again, the total number of colony-forming units formed after the overnight incubation was used to determine the overall bacterial diffusion through each membrane.
Bacteria Preparation: S. Typhimurium, E. coli K12 and O157:H7, K. pneumoniae, P. aeruginosa, L. monocytogenes 1/2a, and B. subtilis were cultured in appropriate media for 18 h at 37 °C under constant agitation at 180 RPM from glycerol stock solutions.The bacteria from these overnight incubations were then centrifuged at 7000 RPM for 15 min to form a bacterial pellet.This pellet was then resuspended in PBS buffer solution for use in all bacterial studies.
Membrane Antifouling Assessment: Chicken purge was extracted from chicken samples and heated in a water bath at 60 °C to melt any solidified lipid molecules.The filtered chicken purge was then pipetted through a cotton membrane and then collected.Both the filtered and unfiltered membrane samples were pipetted into a well plate and had their optical density measured using a Synergy Neo2 plate reader (Aligent Technologies).Samples were measured across an absorbance spectrum ranging from 400 to 700 nm in increments of 10 nm.Deionized (DI) water was also assessed for baseline readings.Quadruplicate readings were obtained.SEM images of cotton membranes saturated in chicken purge were obtained to visualize antifouling properties. 1 × 1 cm 2 cotton membrane samples were saturated with chicken purge and then dried in ambient conditions for 24 h.These samples were then mounted, coated, and imaged in a method identical to the SEM procedures outlined in "Preliminary Membrane Characterization".
FNAP Synthesis: All relevant sequences are listed in Table S3, Supporting Information.3′ amino-modified probe fragments were phosphorylated using ATP, T4 polynucleotide kinase buffer A, and T4 polynucleotide kinase in-solution, over 30 min at 37 °C.Substrate fragments (FQ30, TB30) and ligation template fragments were then added, heated for 1 min at 90 °C, and cooled at ambient temperature, to mediate the annealing of the three fragments.T4 DNA ligase buffer, T4 DNA ligase, and water were then added and incubated at ambient temperature for 1 h to mediate ligation of the probe and substrate fragments.The sample was then ethanol precipitated and centrifuged for 20 min at 4 °C, 20 000g.A polyacrylamide gel was used to purify the ligated product.The final nucleic acid probe product was resuspended in water.
Sensor Development and Preliminary Characterization: Nucleic acid probe was first mixed with EDC-NHS crosslinker in MES buffer to facilitate covalent attachment to polyethylene substrates.A GeSiM Nano-Plotter piezoelectric printer was used to deposit nucleic acid probe onto the sensor surface.The sensors were then incubated in a 75% humidity environment for 2 h and then washed in a water bath at 220 RPM for 30 min on a platform shaker (VWR International) to remove any unbound probe molecules.They were then dried and imaged using an inverted fluorescent microscope.Covalent attachment was confirmed by comparing the fluorescence of nucleic acid probes both with and without the EDC-NHS covalent crosslinker before and after the aforementioned water washing step.Next, a calibration curve was developed to confirm the density of nucleic acid probe added to the sensor surface.This curve was created using intensity measurements of arrays composed of known concentrations of fluorescent, single-stranded TRITC DNA molecules.Maximal fluorescence intensity of the nucleic acid probe was obtained using 1 m NaOH, at which point probe density was quantified using the curve.
Sensor Sensitivity and Specificity Testing: The effects of chicken purge on bacterial proliferation and survival were assessed through the resuspension of 10 6 CFU mL −1 E. coli in chicken purge and in PBS.These solutions were then plated on selective MacConkey agar plates, alongside chicken purge alone.These plates were incubated overnight at 37 °C and then the colony-forming units were quantified.Sensor sensitivity was tested by incubating printed, pre-imaged sensors with S. Typhimurium concentrations ranging from 10 7 to 10 3 CFU mL −1 , where all dilutions were performed using chicken purge.100 mm MgCl 2 was also added to the incubation solution.Control samples were composed of chicken purge and MgCl 2 alone.After an 8 h incubation at 37 °C, all test solution was removed from the sensor surface.The sensors were then briefly washed in DI water and re-imaged to assess their fluorescence fold-change.Sensor selectivity was tested in a similar manner except this time a constant bacterial concentration of 10 6 CFU mL −1 was tested using S. Typhimurium, K. pneumoniae, E. coli O157:H7, P. aeruginosa, L. monocytogenes 1/2a, and B. subtilis, while the control solution remained the same.
Sensor Stability and Temperature Profile Development: Printed sensors were washed, imaged, and then stored at 4 °C.After 3 months of storage, the sensors were incubated with S. Typhimurium in concentrations ranging from 10 6 to 10 3 CFU mL −1 and were then re-imaged to confirm sensor viability.A temperature profile of sensor performance was also developed to confirm their functionality in a variety of food storage conditions.To this end, sensors were incubated with 10 7 and 10 5 CFU mL −1 contaminated chicken purge and incubated for 24 h at 4, 25, 37, and 45 °C and then reimaged.
Proof-of-Concept Testing: Full-scale models of the 45°packaging tray were 3D printed for the proof of concept testing.A printed sensor was placed in the window at the base of the packaging and an MgCl 2 buffersaturated membrane of the same size as the sensor was placed directly on top.Cooked RTE rotisserie chicken purchased from local grocery stores was cut into 250 g samples and placed on the packaging trays.These chicken samples were contaminated with 10 mL volumes of S. Typhimurium chicken purge suspensions to yield the desired CFU g −1 concentrations.Control samples were treated with 10 mL of uncontaminated chicken purge.The samples were incubated at 37 °C for 8 h.After 8 h, the packaging was opened, the membrane was removed, and the extracted sensor was fluorescently imaged.Application-based testing followed the exact same protocol, except introduction of the bacteria onto food samples was done via a contaminated surface, glove, and knife, rather than through contaminated chicken purge.
4.0.0.1.Typhimurium Growth Study: An original concentration of 10 2 CFU mL −1 of S. Typhimurium suspended in chicken purge was selectively plated at a timepoint of 0 h.It was then incubated at 37 °C, with selective plating repeated at 2 and 4 h timepoints.The total number of colonyforming units formed after an overnight incubation at 37 °C was used to quantify the growth of the original 10 2 CFU mL −1 of S. Typhimurium.
Full System Specificity Testing: Full system specificity was tested in a similar manner to sensor selectivity testing.Equal amounts of 10 6 CFU mL −1 of E. coli O157:H7, L. monocytogenes 1/2a, and S. Typhimurium were resuspended in chicken purge and applied onto chicken samples within sensor and membrane-containing packaging trays for an 8 h incubation at 37 °C.After this incubation, sensors were fluorescently images as previously described.
Target Verification Study: Chicken samples were contaminated with 10 6 CFU g −1 S. Typhimurium.The final target that reached the sensor interface was collected after the 8 h incubation period and selectively plated along with some of the originally contaminated chicken purge that was applied.After the plates were stored for a standard overnight incubation at 37 °C, the total number of colony-forming units formed for both the initial and post-incubation samples were compared to both assess that the collected target contained S. Typhimurium and that there was no significant change in the overall bacterial concentration.
4.0.0.2.Typhimurium Detection in Lettuce Samples: Bagged lettuce was obtained from a local grocery store and washed thoroughly.Water used to wash the lettuce was collected as the wash fluid, which was then spiked with 10 6 CFU mL −1 of S. Typhimurium to simulate contaminated lettuce.Lettuce leaves were placed within the Lab-in-a-Package system and 10 mL of the contaminated wash fluid was readministered onto the samples.The sensors were reimaged after a 24 h incubation at 25 °C to assess contamination detection.
Handheld Fluorescence Detection: A handheld fluorescence scanner (Dino-Lite Edge, Dino-Lite US, Dunwell Tech., Inc.) was used to image S. Typhimurium contamination in sensor samples and Lab-in-a-Package.Initial characterization was performed with 10 8 CFU mL −1 contaminated FNAP sensor samples.In situ detection was performed with chicken samples that were contaminated with 10 6 CFU g −1 chicken purge.In this case, the handheld microscope was used to image the sensor window on the base of the packaging, without opening the package or extracting the sensor.The scanner can be connected to either an associated computer software or smartphone application for sensor visualization and final signal readout, through which all images were obtained.

Figure 1 .
Figure 1.Schematic illustration of the Lab-in-a-Package platform.a) Complete Lab-in-a-Package in situ detection platform with inclined packaging tray, reagent-saturated membrane, and sensor incorporation shown for RTE chicken products.Imaging procedure involving fluorescence scanning is also shown.b) Inclined food packaging trays with angles ranging from 45°to 90°to optimize test sample localization.c) Depiction of membrane saturation with reagent components, diffusion of buffer components and target analyte to sensor surface, and fouling prevention.d) FNAP sensor development with corresponding material surface and biochemical modifications.

Figure 2 .
Figure2.Characterization of packaging models and membrane candidates based on application-relevant properties.a) CAD models for all packaging models with top, bottom, and orthogonal views shown.b) Time required for a water droplet to fall down packaging edge.c) Time required for 5 mL of buffer to reach sensing window when dispensed at a rate of 0.5 mL s −1 .d) Percentage of original PBS volume localized on sensing window after 1 min when dispensed at a rate of 0.2 mL s −1 .e) Percentage of original chicken purge volume localized after 24 h at 37 °C.f) SEM images of candidate membranes at 100× with overlays at 500×.g) Mean background fluorescence of candidate membranes.h) Absorption capacity of candidate membranes.i) Volume of buffer diffused through candidate membranes after 2 min.j) Membrane effects on bacterial growth following a 6 h incubation with E. coli.k,l) Bacterial diffusion through unsaturated (k) and buffer-saturated (l) membranes onto underlying substrates following a 6 h incubation at 37 °C with E. coli.m) Membrane effects on bacterial growth following a 6 h incubation with S. Typhimurium.n,o) Bacterial diffusion through unsaturated (n) and buffer-saturated (o) membranes onto underlying substrates following a 6 h incubation at 37 °C with S. Typhimurium.All reported values represent the mean of all samples with error bars representing sample standard deviation.All asterisks represent significant differences at corresponding significance levels.

Figure 3 .
Figure 3. S. Typhimurium sensor development and testing.a) Schematic illustration of S. Typhimurium-responsive nucleic acid probe cleavage activity within food matrices, with associated pre-cleavage, cleavage, and quencher separation states.b) Sensitivity testing of nucleic acid probe using bacterial dilutions in chicken purge, with associated images with 100 μm scale bars.c) Temperature profile of nucleic acid probe with bacterial species of 10 7 and 10 5 CFU mL −1 at 4, 25, 37, and 45 °C.d) Covalent attachment confirmation of nucleic acid probe on substrate surface.e) Stability testing of developed sensor tested with 10 6 to 10 3 CFU mL −1 of bacteria after storage for 3 months at 4 °C.f) Specificity testing of nucleic acid probe using various bacterial species at 10 6 CFU mL −1 , with associated images with 100 μm scale bars.All reported values represent the mean of all samples with error bars representing standard error of the mean.All asterisks represent significant differences at corresponding significance levels.

Figure 4 .
Figure 4. Lab-in-a-Package platform development and testing.a) Schematic illustration of in situ sensing interface with FNAP-based S. Typhimurium detection.b) Images of packaging platform assembly, involving: i) sensor implantation within sensing window, ii) membrane incorporation, and iii) food addition into the package.Scale bars represent 3 cm on printed packaging tray.c) Inherent fluorescence of chicken purge at four fluorescence wavelengths.Mean fluorescent values of overlayed cotton membranes shown with gray boxes.d) MgCl 2 concentration optimization for membrane absorption and diffusion.e) Sensitivity testing following in situ full platform testing of contaminated whole chicken sample, with associated images with 100 μm scale bars.f) Contamination of food products from i) various avenues of contamination, introduced ii) during stages of the production process.g) Induced real-world contamination detection in situ with Lab-in-a-Package platform.h) Optical image of experimental set-up for handheld fluorescence scanner with associated smartphone readout.i) S. Typhimurium detection using FNAP sensor as visualized using a handheld scanner, and associated images with 3.33 mm scale bars.j) Handheld fluorescence detection of S. Typhimurium in Lab-in-a-Package, with associated sensor images with 3.33 mm scale bars.All reported values represent the mean of all samples with error bars representing standard error of the mean.All asterisks represent significant differences at corresponding significant levels.(a) and (f) were created using BioRender.com.