Recurrence monitoring for ovarian cancer using a cell phone-integrated paper device to measure the ovarian cancer biomarker HE4/CRE ratio in urine

Ethics statement

No human or animal materials were used in this study.

Reagents

Monoclonal mouse anti-human epididymis protein 4 (HE4) was purchased from HyTest (Turku, Finland). The reporter reagent used in the LFA was BioReady150 nm gold nanoshells by nanoComposix Inc (San Diego, California). Ethyl-N;-(3-dimethylaminopropyl) carbodiimine hydrochloride (EDC) and Sulfo-NHS chemistry was used (sigma Cat# E1769-1G, Thermo, Prod# 24510) to covalently bind the antibody to the nanoshell. Hydroxylamine (Sigma, cat# 467804) was used to quench remaining amine groups during the covalent conjugation protocol.

HE4 antigen standards were provided by Fujirebio Diagnostics Inc. (Malvern, PA). For the enzymatic creatinine reaction, we purchased a Creatinine LiquiColor kit from Stanbio (Boerne, TX). The creatinine test paper used was Whatman no. 3, purchased from Cytiva (Marlborough, MA). The dipping soy wax was purchased from Hearts & Crafts (Brooklyn, NY). Parafilm tape was purchased from Hach (Loveland, CO). Anhydrous creatinine was purchased from Thermo (Prod # C4225, lot# SLCF5841). All solutions were prepared with deionized water unless otherwise noted.

Lateral flow materials and design

To create the lateral flow assay standards for HE4, we used lyophilized standards in the Fujirebio ELISA kit, which contained HE4 antigen in a phosphate buffered salt solution with bovine serum albumin, an inert yellow dye, and a non-azide antimicrobial preservative. To reconstitute, 1.0 mL of deionized water was added to each standard vial. The vial was vortexed and then allowed to stand at room temperature for 15 min. Prior to use, the vial contents were gently mixed by pipette. Standard dilutions were created from the standards for the simulated patient samples with creatinine solutions or deionized water. Vials were stored at 4° C before and after use as recommended by the manufacturer.

Both HE4 and creatinine were measured using paper-based methods (Fig. 4). For HE4, we used standard LFA methods to spot reagents and assemble the test strips. An automatic antibody dispenser (IsoFlow from Imagene Technology Inc) was used to deposit HE4 sheep-derived polyclonal antibody (R&D Systems, Cat# AF6274) across the membrane as the capture antibody at the test line. To make the control line, the antibody dispenser deposited a goat anti-mouse antibody (Lampire, cat# 7455507, Lot# 17H40070) across the membrane. The mouse antibody was chosen for the control line so that the excess mouse antibody would bind to it. The striped membrane was dried for 1 h in an oven at 37° C. Next, the striped membrane was placed onto an adhesive backing card. A sample pad was added to the right edge of the membrane and the wicking pad was added to the left edge, overlapping the membrane by 1 mm to allow for capillary action along the test strip. The assembled card was cut into individual test strips 3.9-mm wide by 70-mm long with an automated guillotine (Matrix 1201 Membrane cutter, Kinematic, Twain hart, CA). The strips were stored at room temperature in sealed pouches with desiccants until use.

Next, we optimized the selection of the reporter antibody and nitrocellulose membranes. To identify the best reporter, we screened three monoclonal HE4 reporter antibodies (R&D systems, LsBio, and Hytest). To test the binding and non-specific binding of each antibody we conducted a dot blot test. We spotted 1 μl spot of polyclonal antibody on the membranes and dried the membrane for 1 h in an oven at 37° C. We then mixed 1 μl of antigen with each of the reporters and spotted the solution onto the sample pad (Supplementary Fig. S1) for the positive control test. For the negative control test, we spotted the reporter without antigen. Non-specific binding occurred when a colored spot was generated without antigen present. Unbound reporters were washed away by the addition of a running buffer to the sample pad. Since the Hytest antibody showed strong signal intensity in antigen binding in the positive control tests and the least non-specific binding in the negative control tests, it was chosen for further testing.

We screened three different nitrocellulose membranes (supplementary Fig. S1) and subsequently chose MDI CNPC (10 μm) membranes (MDI membrane technologies) since they exhibited the least non-specific binding in the dot blot test, and the antigen signal intensity was higher in the positive control tests.

To run the LFA HE4 test, we used 20 μl of nanoshell + HE4 antibody incubated with 25 μl of sample for 10 min on a rotator prior to pipetting it onto the sample pad of the lateral flow test strip. After deposition onto the sample pad, the lateral flow test was left at room temperature for 20 min before measurement on a flatbed scanner or cell phone.

Creatinine test design

For the design and fabrication of the creatinine test, we used an enzymatic creatinine kit that comprises a two-reagent system that is typically found in automated analysis equipment³¹. Since chemicals can interfere with creatinine determination using the Jaffe reaction³², we chose to use enzymatic reagents to measure creatinine. Enzymatic reaction has been chosen over the picrate reaction for the determination of creatinine in clinical labs as the results of enzymatic methods have been reported to match the gold standard method more closely, isotope-dilution-mass spectrometry (IDMS)^31,33,34. In the first reaction mixture, reagent 1 (R1), the creatinine amidohydrolase, converts creatine to sarcosine, and the oxidation of sarcosine by sarcosine oxidase produces hydrogen peroxide. The hydrogen peroxide generated in R1 is then reacted with reagent 2 (R2) in the presence of peroxidase to react with Quinoneimine dye as shown in Fig. 4. This reaction results in a light to dark purple dye in which the dark purple corresponds to high creatinine concentration in the sample.

Whatman no. 3 paper was used for its high absorbency, medium porosity and medium flow rate. To create the creatinine paper strips, Whatman no. 3 paper was cut with scissors into individual 2.5-mm long by 10-mm wide pieces. The test strip end was then briefly dipped in heated liquid soy wax. The wax end created a barrier for the chemical products generated during the chemical reaction³¹ and allowed for tweezer handling of the test strips.

As powdered creatinine dissolves readily in deionized water, we created a stock solution of creatinine at 80 mg/dL. Standard dilutions were made from the stock solution with deionized water into individual Eppendorf tubes and refrigerated at 4° C until use. The creatinine enzymatic kit uses a 300 μl cuvette to measure absorbance and the manufacturer instructions suggest using 270 μl of R1 and 90 μl of R2. For our small paper-based tests, we reduced the volumes for the two creatinine enzymatic reagents (R1 and R2) but preserved the ratio as suggested by the manufacturer. This reduction in volume was necessary to prevent leaching out of the paper onto the cell phone housing or scanner bed during measurement. We pipetted 18 μl of R1 onto the test strip followed by 6 μl of sample. The test strip was then placed in a 37° C oven for 4 min. Then the test strip was removed from the oven, and 6 μl of R2 was pipetted on the test strip. The test strip was then placed back into the 37° C oven for 4 min. The incubation in the oven was used to decrease the time to measurement: if left at room temperature, the result can be read in 30 minutes³⁵. The test strip was finally removed from the oven and allowed to sit at room temperature before measurement on the flatbed scanner and cell phone. To prevent leaching out, a small 3 mm × 11 mm section of Parafilm tape (Hach, Loveland, CO) was added to the phone cassette well where the creatinine test is placed. The hydrophobicity of Parafilm tape ensures that the liquid remains inside the creatinine test strip during measurement.

Quantification of paper-based tests

The quantitative LFA method involves measuring the amount of an analyte in a sample against a standard curve of known amounts of the analyte³⁶. To measure the intensity of the reaction in both the LFA and creatinine tests, we used a CanoScan LIDE 300 scanner from Canon (Ōta, Tokyo, Japan) and a cell phone (Samsung galaxy s20).

In this report, we compare quantification using our cell phone app to a flatbed scanner as the “gold standard”. Flatbed scanners have been used widely to quantitate LFAs and other colorimetric tests³⁷. For the scanner measurement of HE4, the individual paper strips were placed face down. The auto scan button was pressed to create a PDF scan. The document was then opened in Adobe Photoshop to convert the PDF to JPEG. The JPEG was then opened in ImageJ. The test line in the image was measured with the “Measure tool”. The output of the measured result was then copied and pasted into Excel for analysis.

In this study, determination of HE4 concentration was based on a sandwich assay format in a lateral flow method. Therefore, the timing of when the reaction is considered “complete” and ready to read on a scanner or cell phone app influences the intensity values. Per the manufacturer’s instructions for the membrane of choice, MDI-10, the strip can be read at 15 min after sample application. Therefore, we analyzed the intensity associated with a concentration of 280 pM (the medium standard) over 5-min intervals from the earliest it could be read (Fig. 5). We chose the 20-min time mark because subsequent drying of the test reduced the intensity of the test value by about 12% as shown in Fig. 5.

To determine the optimal time for reading the creatinine test strips, three concentrations of a standard curve were analyzed every 2 min until 14 min. The time point at 8–10 min was selected as the change from 8 to 16 min is less than 2% change (Fig. 5).

To make a standard curve for HE4, we serially diluted the 573 pM HE4 standard from the ELISA kit. We then compared the output of the value calculated from the standard curve to the known value (Fig. 6A). For the scanner measurement of Creatinine, an 80 mg/dL stock solution of creatinine was made and serially diluted to create a standard curve for creatinine. After incubation, the test strips were allowed to sit at room temperature for 8–10 min before being measured in the scanner. Creatinine concentrations up to 40 mg/dL were analyzed, and triplicate assays were performed for each concentration. The results demonstrated a second order polynomial fit range from 0 mg/dL to 40 mg/dL of creatinine (r² = 0.9893), as shown in Fig. 6B. We determined the limit of detection (LOD) of each test strip by analyzing a blank 5 times and calculated a 3 SD limit³⁸. The LOD for HE4 is 15 pM and for creatinine test the LOD is 1.2 mg/dL. Average intensity of the blank for is 226.4856 ± 2.9178 for HE4 and 243.31 ± 2.99 for creatinine (Fig. 2).

In addition to quantification with a flatbed scanner, we also developed a cell phone-based system for quantification of both HE4 and creatinine. Figure 7A shows a photograph of the developed device, which includes a 3D-printed black enclosure and a test strip holder. The black enclosure is used to mitigate the problem with varying external lighting conditions. The current system was implemented on an Android platform-based cell phone (Redmi Note 7 from Xiaomi, camera resolution 4000 pixels × 3000 pixels), but the same system can be used with other cell phones with a small adjustment in the design of the phone attachment. As shown in Fig. 7B, an inexpensive plano-convex lens of focal length 75 mm was used to magnify the area of the test strips. The lens was attached to the phone attachment A, which is mechanically connected to the cell phone. Phone attachment B is connected to attachment A. The attachments are positioned perpendicular to the phone to eliminate the need to tilt the phone and to standardize the focal distance to reduce imaging inconsistencies. Both attachments are 3D-printed with black polylactic acid (PLA) material. The material of choice is not significant. Black was chosen to decrease the reflectance of light inside the test strip compartment. Once the test-strip holder is positioned inside of Phone attachment B, ambient light is blocked from entering the inside of the phone attachment B.

It is possible to use a cell phone flashlight as a light source for colorimetric applications³⁹. However, in the current application, it is not possible to use the flashlight to illuminate the test strips. This is attributed due to the use of the external lens to acquire an image with a large field of view to cover both the test strips and the calibration standards. Another problem in using the flashlight is that its position varies from phone to phone, therefore the illumination over the detection area will be different for different brands of cell phones⁴⁰. Instead, here we achieve uniform illumination by using a novel optical fiber-based illumination scheme as shown in Fig. S2. Light from a low-cost external battery-powered white LED (Finware LED, Amazon) is coupled to two plastic multimode optical fibers (980-µm core diameter, Edmund Optics), and the other ends of the optical fibers are connected to the phone attachment. The distances between the test strips and the optical fibers were optimized to achieve a uniform illumination area. The battery and the LED compartment are located on the outside of the attachment so that they are easily operable by the user and the battery. While the cell phone battery can be used to power an external LED⁴¹ but we have observed that long term use of cell phone battery for powering external components heats up the phone and the output power of the LED is significantly reduced when the cell phone battery power is low (< 30%). Therefore, in the current application, we have used an external coin cell battery to power the LED, which can easily be replaced when needed.

Figure 7C shows the schematic of the test-strip holder. The test strips are inserted from the side and manually slid into the field of view. The test-strip holder contains the HE4 test, CRE test and calibration stickers to allow for simultaneously measurement of both tests and calibration stickers. We prepared calibration stickers indicating the color of both the HE4 and CRE standard solutions and attached them to the test-strip holder. The phone application captures both the test strips and the calibration stickers together in a single image frame to evaluates the intensity values. The intensity values from the calibration stickers are used to generate the respective calibration equations and the unknown HE4 and creatinine concentrations are evaluated by interpolating corresponding intensities values in respective calibration equations. Our device evaluates the intensity of both tests and determines the corresponding concentration value by using the calibration sticker as a reference standard. The calibration standards are printed as a vinyl sticker to increase stability over time compared to typical inkjet printer paper. To facilitate on-board data processing, an Android application (app) was developed and installed on the cell phone. The detailed procedures for using this app are given in the supplementary information. To simulate an at-home user, both HE4 and creatinine were measured simultaneously on both the scanner and cell phone imaging modality (Fig. 3).

Figure 6 shows the calibration plots generated from the HE4 (pM) and CRE (mg/dL) standard solutions using the cell phone app. The analysis algorithm in the phone is simple, as the concentration can be determined by a linear relationship. In the case of HE4 we have selected the concentration range of 33.75 pM-543 pM, which indicates a co-efficient of determination (R²) of 0.973 as shown by Fig. 8A. This range was within the range of the ELISA kit. Similarly, for CRE, we evaluated a concentration range of 2.5–40 mg/dL, which shows a co-efficient of determination (R²) of 0.97 as shown in Fig. 8B. For the phone, a linear fit was chosen over a polynomial fit since the linear fit provided a high R² value and was computationally easier since there is only one answer given from the intensity calculation.

Surrogate sample preparation

In this work, we tested the performance of our test using clinically relevant ratios of ovarian cancer recurrence. Surrogate patient samples were created based on previously published work by Liao et al., who measured patient HE4 urine values in pM and creatinine values in mg/dL as a ratio¹¹. The range of 2–47 ratios was selected to ensure that the paper-based test was operating in the range of ratio values associated with recurrence testing in urine¹¹. Because both creatinine and HE4 can be present in high amounts, Liao et al. diluted clinical samples by 1:40 prior to testing¹¹. Therefore, we assume the same level of dilution for our test method and CRE values from 0 to 40 mg/day³¹ were selected for testing as they are in the ranges of diluted urine samples as described by others^10,11. To create simulated patient samples, we made a 3 × 3 matrix of HE4/CRE solutions. For the three HE4 solutions, we included a high, medium, and low CRE version of each of the HE4 concentrations.