A paper-based HPV E7 oncoprotein assay for cervical precancer detection at the point of care

Abstract

Cervical cancer, while preventable through screening and treatment of cervical precancer, remains a global challenge with a disproportionately high burden of disease in resource-limited settings, especially in low- and middle-income countries (LMICs). Lack of affordable, easy-to-use screening and diagnostic tests contributes to this disparity. Most commercially available tests are not appropriate for use in LMICs due to resource constraints. Specifically, HPV mRNA and oncoprotein tests that have high specificity for cervical precancer and cancer require complex sample preparation protocols and expensive instrumentation. To address these limitations, an HPV E7 oncoprotein assay for HPV16, 18, and 45 was developed that is appropriate for use at the point of care. The assay is paper-based, involves only five simple steps, and does not require instrumentation. A clinically relevant limit of detection was demonstrated with cellular samples. Additionally, clinical performance was demonstrated with a small pilot study (n = 19), in which the HPV E7 paper-based assay was found to have 95% accuracy when compared to histopathologic diagnosis of cervical intraepithelial neoplasia grade 2 or more severe (CIN2+). With further clinical validation, this assay could enable highly specific point-of-care testing for cervical precancer and cancer that is instrumentation-free, affordable, and ideal for use in resource-limited settings.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-79472-2.

Introduction

Cervical cancer remains the fourth most common cancer among women globally with approximately 662,000 new cases and 349,000 deaths annually^1,2. Low- and middle-income countries (LMICs) and low-resource settings in high-income countries (HICs) have a disproportionate burden of disease³. Approximately 90% of cervical cancer deaths occur in resource-limited settings, often due to a lack of accessible tools for early screening and diagnosis^1,4. Common resource constraints include a shortage of trained personnel, limited laboratory infrastructure, and high per-test costs, preventing women in these settings from receiving potentially life-saving early detection measures^5,6.

While the traditional method of cervical cancer screening has been cytology via Pap smear, newer screening methods, such as molecular tests for high-risk HPV DNA, HPV mRNA, and HPV oncoproteins, have been introduced in recent decades^6,7. High-risk HPV DNA tests are highly sensitive and have excellent negative predictive value for cervical precancer and cancer^8–10. Women with a negative HPV DNA test have an extremely low risk of developing cervical cancer at 3- and 5-year intervals¹⁰. However, HPV DNA tests are not specific to cervical precancer and cancer because most women clear HPV infections within 1–2 years¹¹. Screen-and-treat programs based on HPV DNA testing alone can, therefore, cause overtreatment and waste resources^6,12.

In contrast, HPV mRNA and HPV oncoprotein tests are more specific for cervical precancer and cancer. The integration of HPV into cellular hosts prompts the overexpression of HPV mRNA, production of oncoproteins E6 and E7, inhibition of tumor suppressors, and subsequent malignant transformation of infected cells^13,14. HPV mRNA and E6/E7 oncoproteins are therefore potentially key biomarkers for identifying patients at high risk of cervical precancer and progression to cancer¹³. While HPV mRNA and E6/E7 oncoprotein tests have great potential for use in screen-and-treat programs, the feasibility of using them in low-resource settings has historically remained limited by high per-test costs and the need for complex laboratory instruments¹⁵; only recently have low-cost tests such as Arbor Vita’s OncoE6 and OncoE6/E7 been evaluated in resource-limited settings, and these tests still require the use of a centrifuge for sample preparation¹⁶.

To address these resource limitations, a low-cost, sample-to-answer, paper-based HPV E7 oncoprotein assay was developed. Expanding upon previous work, the assay is a paper-based enzyme-linked immunoassay (ELISA) with high sensitivity due to signal amplification¹⁷. The assay has five simple steps including sample preparation and lysis. No instrumentation or infrastructure is needed, making the assay appropriate for use in resource-limited settings. Here, the workflow is first described and the point-of-care sample preparation and lysis protocols are characterized. Next, the performance of the assay with HPV16, 18, and 45 cell lines is assessed in both a traditional 96-well ELISA and the paper-based ELISA format. Finally, the assay is validated with clinical samples from patients with biopsy-proven high-grade cervical intraepithelial neoplasia grade 2 or more severe (CIN2+) in a pilot clinical study.

Methods

Cell lines

Five cell lines were used to evaluate the oncoprotein assay: HeLa (HPV18, HTB-35), SiHa (HPV16, CCL-2), Ca Ski (HPV16, CRL-1550), MS751 (HPV45, HTB-34), and C-33 A (HPV negative, HTB-31). All cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA). Cells were cultured using DMEM (Corning, Tewksbury, MA) with 10% fetal bovine serum (FBS, Bio-Techne, Minneapolis, MN) and Penicillin-Streptomycin (Thermo Fisher Scientific, Waltham, MA), and passaged no more than ten times. After passaging, cells were counted and pelleted, media was removed, and the dry pellets were stored at -80ºC until use. Pellets were thawed at room temperature before use.

Lysis evaluation

Four conditions were tested for point-of-care lysis: (1) Tissue Protein Extraction Reagent (T-PER, Thermo Fisher Scientific, Waltham, MA); (2) Mammalian Protein Extraction Reagent (M-PER, Thermo Fisher Scientific, Waltham, MA); (3) NP-40 (Thermo Fisher Scientific, Waltham, MA); and (4) xTractor Buffer (Takara Bio, Mountain View, CA). Each buffer was compared to a no lysis control (NLC) and to a freeze-thaw positive lysis control. Five different cell types were tested, including HeLa (HPV18), SiHa (HPV16), Ca Ski (HPV16), MS751 (HPV45), and C-33 A (HPV negative).

For each point-of-care lysis condition, buffer was added to a cell pellet at 10 million cells/mL, briefly mixed, and incubated for 10 min at room temperature. No-lysis controls were reconstituted in Phosphate Buffered Saline (PBS); the freeze-thaw samples were reconstituted into ice-cold PBS with 0.05% Tween 20 (PBST) with 1 mg/mL EDTA-free protease inhibitor (Roche, Basel, Switzerland). For the freeze-thaw method, samples were frozen with liquid nitrogen and thawed in a 37 °C water bath four successive times. After sample preparation, all samples were centrifuged at 13,000 RCF for 10 min, and the resultant supernatant was diluted 1:2 in PBS before assessment using a bicinchoninic acid (BCA) protein test kit (Thermo Fisher Scientific, Waltham, MA). The assay was performed according to the BCA protocol with 25 µL per sample, so approximately 125k cells per well. Total protein concentration in the supernatant was used to characterize the lysis ability of each buffer. The fold change in lysis compared to freeze-thaw was calculated for each buffer by taking the ratio of its supernatant protein concentration to the freeze-thaw supernatant concentration of the corresponding cell type. Of note, all sample preparation started with a cell pellet which was originally stored at -80ºC; however, no lysis and freeze-thaw controls were included to ensure lysis could be attributed to the conditions tested.

Traditional 96-well ELISA for HPV E7 oncoprotein

Traditional 96-well ELISAs were performed using the protocol detailed in Appendix S1. The capture antibody was an anti-HPV18 E7 monoclonal capture antibody (MBS310529, MyBioSource, San Diego, CA). Samples were tested in triplicate. The detection antibody was an unconjugated IgG detection antibody (anti-HPV E7 detection antibody, Ab100953, Abcam, Cambridge, MA), biotinylated with 20 mM biotin using the EZ-Link™ Sulfo-NHS-Biotin biotinylation kit (Thermo Fisher Scientific, Waltham, MA). Two-tailed t-tests were performed between each concentration to determine whether differences in absorbance were significant.

Lysis buffer comparison

To evaluate the effect of the point-of-care lysis buffers on E7 oncoprotein assay sensitivity, a traditional 96-well ELISA was performed on the cell lysate for cells lysed in all four point-of-care lysis buffers. A small range of HeLa cells were spiked into C-33 A cells, so that the total cell number remained constant at 50,000 cells. Cellular samples were lysed using the point-of-care buffers with a 10-minute incubation step at room temperature and added directly to ELISA plate for sample incubation. As a control, the same cellular range was prepared using standard freeze-thaw lysis.

Paper-based assay for HPV E7 oncoprotein

Paper devices were designed to perform ELISA reactions to detect HPV E7 oncoprotein using a two-dimensional paper network described previously¹⁷. Briefly, devices consist of a nitrocellulose membrane (backed CN140, Sartorius, Goettingen, Germany), glass fiber pads (grade 8951, Ahlstrom, Helsinki, Finland), adhesive-backed plastic backing (5 mm Dura-Lar, Blick Art Supplies, Galesburg, IL), and a cellulose wicking pad (C083, Millipore, Billerica, MA), all cut using a CO2 laser cutter (Universal Laser Systems, Scottsdale, AZ). A QR code can be used to provide directions for use. An example of the paper device is shown in Fig. 1.

Fig. 1.

Components of the HPV E7 paper test. The test consists of a cellulose wicking pad, six glass fiber pads with stored reagents, and a nitrocellulose membrane printed with anti-HPV-16/18/45 E7 antibodies at the test line and anti-streptavidin antibodies at the control line. A QR code can be used to provide directions for use.

Capture lines were printed onto the nitrocellulose membrane using a sciFLEXARRAYER S3 (scienion, Berlin, Germany) printer. The control line consisted of 80 nL of 250 µg/mL streptavidin monoclonal antibody (S10D4, Thermo Fisher Scientific, Waltham, MA), and the test line consisted of 400 nL of 1 mg/mL anti-HPV18 E7 monoclonal antibody (MBS310529, MyBioSource, San Diego, CA). After printing, strips were dried for 1 h in a 37° C incubator. Next, nitrocellulose strips were incubated in a solution of 0.5% BSA, 4% trehalose, and 1% sucrose in PBST for 30 min with gentle shaking on an orbital shaker. Finally, strips were dried for 1.5 h in a 37 °C incubator before being stored, in a foil pouch with desiccant, at 4 °C until use.

To run the assay, nitrocellulose strips and glass fiber pads were added onto the adhesive-backed Dura-Lar backing. The following reagents were then added to glass fiber pads as follows: 15 µL of 10 µg/mL biotinylated detection antibody (Ab100953, Abcam, Cambridge, MA), 20 µL of 20 µg/mL streptavidin poly-HRP80, 25 µL of wash buffer (1% BSA, 1% trehalose, 1% sucrose in PBST), 30 µL of the colorimetric solution, and 35 µL wash buffer (1% BSA, 1% trehalose, 1% sucrose in PBST). The colorimetric solution, consisting of 2 mg/mL solution of diaminobenzidine (DAB, Sigma-Aldrich, St. Louis, MO) with 0.5% sodium percarbonate (Sigma-Aldrich, St. Louis, Missouri), was added immediately before running the assay. Alternatively, lyophilized antibody, enzyme, colorimetric reagent, and wash pads were placed upon the acetate backing and rehydrated with PBST to run the assay.

After adding 50 µL of sample to the first glass fiber pad, the paper covering for the adhesive Dura-Lar was removed, and the paper test was folded in half. Each component of the ELISA then flowed sequentially down the nitrocellulose to the test zone, where a reaction occurred if any oncoprotein was captured on the test line. The colorimetric solution reacts with the streptavidin HRP captured at the control or test lines to form a brown precipitate; the results can be read visually. If HPV E7 oncoprotein is present in the sample, two lines appear: a control and test line. If the sample does not contain oncoprotein, only one line appears: the control line. Absence of any lines indicates issues with the stored reagents, and results should be considered invalid. Paper-based ELISAs were imaged using a flatbed color scanner at 600 dots-per-inch (DPI). A complete workflow is shown in Fig. 2.

Fig. 2.

HPV E7 paper test workflow. (A) Complete set of reagents needed to run the HPV E7 paper test includes a cervical swab, a tube containing lysis buffer, a paper device, disposable pipettes, and rehydration buffer (PBS with Tween20 (PBST)). To perform the test, the user performs five steps: (B) Place the cervical swab into the lysis buffer, mix, and incubate for 10 min at room temperature. (C) Add PBST to rehydrate stored reagents in Pads 2–6 using a disposable pipette. (D) Add sample to the Pad 1 using a second disposable pipette. (E) Peel off the paper backing and (F) fold the paper test in half. (G) After an hour, observe the test and control lines. (H) Test line signal can be visually inspected or (I) quantified. Colorimetric signal appears at the test line if HPV E7 oncoprotein is present in the sample and at the control line if the test result is valid.

Lyophilization

Biotinylated detection antibody, streptavidin poly-HRP80, DAB, sodium percarbonate, and wash pads were lyophilized as following. Detection antibody and streptavidin poly-HRP80 were diluted into a lyophilization solution (1% BSA, 5% trehalose, and 5% sucrose in PBS) at 10 µg/mL and 40 µg/mL, respectively. DAB and sodium percarbonate were prepared in water with 5% trehalose at 2 mg/mL and 2.5 mg/mL (0.25%), respectively. Wash pads consisted of 1% BSA in PBST. Reagents were added to glass fiber pads with the following volumes: 15 µL for biotinylated detection antibody, 20 µL for streptavidin poly-HRP80, 30 µL for DAB, 15 µL for sodium percarbonate, and 25 µL and 35 µL for the wash pads. DAB and sodium percarbonate were lyophilized onto separate glass fiber pads to prevent interaction before rehydration. Reagents were flash frozen in liquid nitrogen for at least 20 s and lyophilized for a minimum of 24 h (LabConco FreeZone 12, Kansas City, MO). Reagents were stored, in a foil pouch with desiccant at -20 °C, until use. During assembly, the lyophilized sodium percarbonate pad was placed onto the adhesive backing and covered with the lyophilized DAB pad. When rehydrated, the two reagents mixed before moving down the nitrocellulose to the capture zone.

Assay performance with lyophilized reagents was compared to that with freshly prepared reagents on a paper ELISA platform using positive (HeLa) and negative (C-33 A) samples. For each sample type, cell pellets were reconstituted at 1 million cells/mL using xTractor buffer, incubated for 10 min at room temperature, and added directly to the sample pad. Lyophilized reagents were reconstituted with PBST.

Reagent optimization for paper-based assay

To reduce any false positive results on the paper ELISA, various concentrations (1–3% w/v) of the blocking agent BSA were added to the reagent and wash pads and tested with 50,000 total HeLa and C-33 A cells in duplicate (Figure S1). HeLa and C-33 A cells were lysed with xTractor buffer as described previously. The optimal condition was defined as one that minimizes the signal-to-background ratio (SBR) of HPV-negative (i.e., C-33 A) samples, while maximizing SBR for HPV-positive (i.e., HeLa) samples.

The concentrations of paper ELISA components were also optimized to maximize the signal-to-background ratio of HPV-positive cell lines while retaining a negative signal for C-33 A samples (Figure S2). HeLa and C-33 A samples were lysed with xTractor buffer and run in duplicate on the paper ELISA platform with the following conditions: baseline, 2X detection antibody concentration, 2X streptavidin poly-HRP80 concentration, 2X DAB concentration, and 0.1X sodium percarbonate concentration. As described previously, the baseline condition included 10 ug/mL detection antibody, 20 ug/mL streptavidin HRP, 1 mg/mL DAB, and 0.5% sodium percarbonate. Similarly, the optimal condition was defined as one that minimizes the SBR of HPV-negative (i.e., C-33 A) samples, while maximizing the SBR for HPV-positive (i.e., HeLa) samples.

Assay performance with a range of cellular and recombinant protein concentrations

Samples with a range of HPV-positive cell concentrations were created by diluting HeLa (HPV18), SiHa (HPV16), Ca Ski (HPV16), or MS751 (HPV45) cells into C-33 A (HPV negative) cells, so that the total cell number remained constant at 50,000 total cells. Each HPV-positive cell type was tested over the following range: 50,000 cells, 25,000 cells, 10,000 cells, 5,000 cells, 2,500 cells, 1,000 cells, 500 cells, and 0 cells, plus a no-cell control. Cells were lysed using xTractor buffer for 10 min at room temperature, then added directly to the 96-well ELISA plate or to the sample pad of the HPV E7 paper test. Additionally, a range of HPV18 E7 recombinant protein (Biomatik, Wilmington, DE) was created by linear dilution into xTractor buffer. Each HeLa cell has approximately 1 fg of HPV18 E7 protein²¹, so the following amounts of total recombinant protein were tested to correspond to the cellular HeLa range: 50 pg, 25 pg, 10 pg, 5 pg, 2.5 pg, 1 pg, 0.5 pg, and 0 pg. Cellular and recombinant protein ranges were tested in both traditional 96-well ELISA and the HPV E7 paper test, using the respective protocols described above. Ideally, HPV16 E7 recombinant protein would also be tested to compare to the SiHa and Ca Ski cellular ranges; however, a source for HPV16 E7 recombinant protein was not found at the time of the study.

Clinical testing and validation

Provider-collected exfoliated cervical samples were acquired from a screening population at Basic Health International and the Instituto del Cáncer de El Salvador (El Salvador Cancer Institute, ICES) in El Salvador. Nonpregnant women, 30–49 years of age, with no history of prior cryoablation, excisional procedure, or invasive cervical cancer were eligible for participation. Informed consent was obtained. Use of the specimens was approved by Internal Review Boards at Rice University and The University of Texas MD Anderson Cancer Center. All methods were performed in accordance with the relevant guidelines and regulations.

Samples were collected into PreservCyt buffer. Cervical samples were tested for high-risk HPV DNA with Qiagen careHPV™. In addition, all patients that were either VIA + and/or HPV + and 10% of the double negatives underwent colposcopy with cervical biopsy of any abnormal lesions or of one colposcopically normal region if there were no visible lesions. If the colposcopic exam was normal, a four-quadrant examination was performed with a high resolution microendoscope and the highest scoring area was biopsied.

All pathology specimens underwent standard processing with hematoxylin and eosin staining. All cervical biopsies and ECCs were reviewed by two expert pathologists: the local institutional pathologist and a second pathologist from the United States. Each specimen was classified as being normal/benign (diagnoses including normal, negative and inflammation), CIN1, CIN2, CIN3, adenocarcinoma in situ (AIS), or cancer. Discrepant results were resolved by a third pathologist from the United States, with the final result being based on two-thirds agreement. If all three pathologists arrived at different diagnoses, all three met in person to review the specimen and reach a consensus diagnosis. Patients were treated or scheduled for follow-up based on the final histopathology results per standard of care.

Of the nineteen clinical tested samples, eight were hrHPV-negative with a corresponding biopsy with < CIN 2, three were hrHPV-positive with a corresponding biopsy with < CIN 2, and eight were hrHPV-positive with a corresponding biopsy with CIN 2+ (two hrHPV-positive with CIN2 pathology, six hrHPV-positive with CIN3 pathology). Partial genotyping was conducted on all hrHPV-positive samples and two hrHPV-negative samples using the AmpFire HPV High Risk Genotyping kit (Atila BioSystems, Mountain View, CA). One sample was recorded as hrHPV-positive but tested negative with AmpFire, shown in Fig. 6 (sample 737). This sample was considered hrHPV-negative as the gold standard result for sensitivity and specificity analyses.

Fig. 6.

Clinical Sample Testing. (Top) Signal-to-background ratio for C-33 A control and 19 clinical samples measured with the HPV E7 paper test, stratified by HPV positivity (tested by either Qiagen careHPV™ or AmpFire kit) and pathologic diagnosis. (Bottom) Scanned images of clinical samples tested with HPV E7 paper test.

For oncoprotein testing using samples collected into PreservCyt buffer, a brief buffer conversion protocol was required. This conversion process with instrumentation would not be necessary for a sample collected via dry swab or a swab placed directly into xTractor lysis buffer; however, the conversion was required with the samples described in this study to prevent interference from the high methanol content in PreservCyt. Two mL of each sample were aliquoted and centrifuged for 10 min at 4,000 RCF to pellet the cells. The supernatant was removed and replaced with 60 µL of xTractor buffer. The samples were flicked and incubated at room temperature for a minimum of 10 min for lysis. After incubation, the samples were centrifuged at 16,000 RCF for 3 min. 50 µL of the supernatant was applied to the paper assay for testing. Sensitivity and specificity were determined using histopathology as the gold standard.

Signal-to-background and statistical analyses

Signal-to-background analysis of the HPV E7 paper strips were determined as previously described in Grant et al.¹⁷. Briefly, a custom MATLAB code was used to assess the pixel intensities from a region-of-interest (ROI) at the test line and from a corresponding background ROI. A ratio of the two ROIs then determined the signal-to-background value. To evaluate whether differences in means were significant between conditions, a two-sided t-test was performed; p-values < 0.05 were determined to be significant. For limit-of-detection analyses, a positivity threshold was first created using the average negative signal plus three standard deviations. Using that threshold, values were binarized, and probit analysis was performed to determine limit of detection using a probability value of 0.95 (XLSTAT, Addinsoft, Paris, France).

Results

Point-of-care sample preparation

Of the four conditions tested for point-of-care lysis, all four achieved lysis equivalent to or greater than the freeze-thaw positive control (Fig. 3A,B, n = 2). xTractor buffer showed the best performance across all five cell types, with a 1.35-1.45-fold change in lysis compared to freeze-thaw. These results indicate that the 10-minute point-of-care protocol at room temperature is able to effectively lyse cellular samples.

Fig. 3.

Point-of-care sample preparation for cell lysis. Four buffers were tested for point-of-care lysis: T-PER, M-PER, NP-40, and xTractor. All differences are significant unless otherwise labeled. (A) The total protein concentration in the resulting supernatant was compared to a standard no lysis control (NLC) and a positive lysis control (Freeze-Thaw) using a BCA test. (B) Lysis efficiencies for each buffer were assessed by comparison to the freeze-thaw condition for each cell type. All four buffers resulted in equivalent or greater lysis than the freeze-thaw positive control. (C) HPV E7 detection was evaluated using a traditional 96-well ELISA. A range of HeLa cells spiked into C-33 A cells (50 K total cells) were tested using four lysis buffers and Freeze-Thaw. Freeze-Thaw and xTractor sample preparation methods resulted in a statistically significant difference in absorbance between 2% HeLa cells and 0% HeLa cells. Therefore, xTractor was selected as the lysis buffer for future experiments. HeLa = HPV18+; SiHa = HPV16+; Ca Ski = HPV16+; MS751 = HPV45+; C-33 A = HPV negative; ns = no significant difference. NLC = no lysis control.

To determine the effect of the point-of-care lysis buffers on assay sensitivity, a traditional 96-well ELISA was performed over a range of cellular samples using all four lysis buffers as well as freeze-thaw lysis (Fig. 3C, n = 2). All lysis methods produced a quantitative response in absorbance to HPV E7 oncoprotein levels in the HeLa samples. However, freeze-thaw and xTractor were the only lysis methods that had a significant difference (p < 0.05) in absorbance between 1,000 HeLa cells (2%) and 50,000–33 A cells (0%). Additionally, xTractor had a strong positive signal at higher HeLa concentrations of 50,000 HeLa cells and 10,000 HeLa cells compared to other lysis options. Therefore, xTractor was selected as the lysis buffer for future experiments.

Limit of detection with 96-Well ELISA

Next, a range of HPV18 E7 recombinant protein was tested in the traditional 96-well ELISA format (Fig. 4A, n = 2). In addition, a range of HeLa (HPV18), SiHa (HPV16), Ca Ski (HPV16), and MS751 (HPV45) cells were spiked into C-33 A cells (HPV negative) to keep the total cell count constant and tested in a 96-well format; results are shown in Fig. 4B-E, respectively (n = 2). The positivity threshold was determined to be the average plus three standard deviations of the C-33 A signal, and probit analysis was performed using this threshold for positivity. Limits of detection for HPV-positive cells were determined as: 135 total HeLa cells, 2,533 total SiHa cells, 6,210 total Ca Ski cells, and 1,823 total MS751 cells. The limit of detection for HPV18 E7 recombinant protein (135 fg) correlated well to that of HeLa cells (135 total cells).

Fig. 4.

Limit of detection for HPV E7 oncoprotein in a traditional 96-well HPV E7 ELISA with xTractor sample preparation.(A) A range of HPV18 E7 recombinant protein spiked into xTractor buffer (n = 2). The limit-of-detection for HPV18 E7 recombinant protein is 0.135 pg. (B-E) A range of HeLa (HPV18), SiHa (HPV16), Ca Ski (HPV16), and MS751 (HPV45) cells were spiked into C-33 A (HPV negative) cells so that total cell number remained constant. After point-of-care lysis using xTractor buffer, the samples were immediately tested in a traditional 96-well E7 ELISA (n = 2). Using probit analysis, the total number of HPV-positive cells required to detect E7 oncoprotein (limit of detection) were determined as: 135 HeLa cells, 2,533 SiHa cells, 6,210 Ca Ski cells, and 1,823 MS751 cells (indicated in box). The positivity threshold (dashed line) represents the average negative signal ± three standard deviations. HeLa = HPV18+; SiHa = HPV16+; Ca Ski = HPV16+; MS751 = HPV45+; C-33 A = HPV negative; Dashed Line = positivity threshold determined as average negative signal + three standard deviations.

Limit of detection with HPV E7 paper test

All samples from the 96-well ELISA in Fig. 4 were also tested in the HPV E7 paper test (Fig. 5A-E, n = 3). First, the optimal amount of BSA in the glass fiber pads was determined to be 1% w/v BSA in both wash and reagent pads, as this concentration reduced false positive signal (Figure S1). In addition, all paper components were optimized to achieve maximum signal-to-background for HPV-positive cellular samples while minimizing signal for HPV negative (i.e., C-33 A) cellular samples (Figure S2). Again, the positivity threshold was the average plus three standard deviations of the C-33 A, signal and probit analysis was performed on results using the positivity threshold. Limits of detection for HPV-positive cells were determined as: 328 total HeLa cells, 15,968 total SiHa cells, 12,287 total Ca Ski cells, and 3,513 total MS751 cells.

Fig. 5.

Limit of detection for HPV E7 oncoprotein in paper test with xTractor sample preparation.(A) The corresponding linear range of HPV18 E7 recombinant protein spiked into xTractor buffer correlated well to the HeLa range when tested in the HPV E7 paper assay (n = 3). The limit-of-detection for HPV18 E7 recombinant protein is 0.331 pg which correlates to 331 total HeLa cells. (B-E) A range of HeLa (HPV18), SiHa (HPV16), Ca Ski (HPV16), and MS751 (HPV45) cells were spiked into C-33 A (HPV negative) cells to maintain a constant number of cells, lysed using xTractor buffer, and tested on the HPV E7 paper assay (n = 3). Using probit analysis, the limits-of-detection for the HPV E7 paper test were: 328 total HeLa cells, 15,968 total SiHa cells, 12,287 total Ca Ski cells, and 3,513 total MS751 cells (indicated in box). The positivity threshold (dashed line) was determined to be the average negative signal ± three standard deviations. HeLa = HPV18+; SiHa = HPV16+; Ca Ski = HPV16+; MS751 = HPV45+; C-33 A = HPV negative; Dashed Line = positivity threshold determined as average negative signal ± three standard deviations.

Finally, the performance of the paper ELISA device using fresh reagents was compared to that of fully lyophilized reagents (Figure S3). There were no significant differences in either positive (HeLa) signal or negative (C-33 A) signal between freshly prepared reagents and lyophilized reagents.

Clinical assessment

Paper assay performance was validated with clinical samples using histopathology as the gold standard (Fig. 6). Nineteen samples were tested, including eight hrHPV-negative with corresponding biopsies that showed < CIN 2 pathology, three hrHPV-positive with corresponding biopsies that showed < CIN 2 pathology, and eight hrHPV-positive with corresponding biopsies that showed CIN 2 + pathology (two hrHPV-positive with CIN2 pathology, six hrHPV-positive with CIN3 pathology).

A summary of results with clinical samples is presented in Table S1. Using histopathology as the gold standard, the paper-based HPV E7 assay demonstrated a 100% sensitivity and 90.9% specificity (95% accuracy) for identification of patients with CIN 2+. Positive and negative predictive values were 88.8% and 90.9%, respectively. Out of eleven < CIN 2 samples, one HPV negative sample with a biopsy with < CIN 2 tested positive. All CIN 2 + samples tested positive. The accuracy of discriminating samples with CIN3 + from < CIN2 was 94% (16/17 samples).

Discussion

The advent of HPV testing in cervical cancer screening has improved the sensitivity of cervical cancer screening over cytology alone^18–20. However, HPV DNA testing presents low specificity for cervical precancer and cancer²¹. In otherwise healthy women, many HPV infections typically clear within 1–2 years, and using only the presence of HPV DNA to guide management decisions may result in overtreatment, use of limited resources, and unnecessary stress for patients^11,12. More targeted markers such as oncoprotein E6/E7 expression may help improve the triage of patients testing positive for HPV who are at higher risk of cervical precancer and early cancer²². Particularly in low-resource settings where patients have limited follow-up opportunities, efficient use of screening and diagnostic tests should be prioritized to prevent progression to cervical cancer.

This study details the development of a point-of-care assay for HPV E7 oncoprotein that can detect cervical precancerous lesions with minimal user input, instrumentation, or infrastructure. For samples stored as dry swabs or placed directly into lysis buffer, sample preparation with xTractor buffer effectively lyses the cellular samples without the need for centrifugation, a key component for its use at the point-of-care. In addition, the successful lyophilization of reagent pads ensures a simple, 15-minute workflow with five user steps, and one hour to results. The HPV E7 paper test costs less than $1 per test with small-scale manufacturing, and $1.47 when including costs for the cervical collection brush, lysis tube, and disposable pipettes (Table S2). The lack of instrumentation, simple workflow, and low cost make this test uniquely appropriate for use in resource-limited settings.

This assay shows comparable performance to the commercially available oncoprotein tests. The Arbor Vita OncoE6 detects oncoprotein E6 from HPV 16/18 with a limit of detection of 2,000 cervical exfoliated cells per test or 30 pg of E6 protein²³. Clinical testing of OncoE6 from a screening and referral population showed high specificity, 98.9–99.4%, though much lower sensitivity, 31.3–53.5%, when compared to CIN2 + on histopathology^24,25. Additionally, a systematic review of several oncoprotein studies, which included OncoE6, showed high specificity in screening and referral populations (99.1% and 82.8% respectively), though lower sensitivity (52.2% and 61.7% respectively) compared to CIN3 + histopathology¹⁶. Another novel test in development, the Arbor Vita OncoE6/E7 Eight HPV Type Test, detects oncoprotein associated with additional HPV types (31/33/35/52/58) at 2,000–10,000 total cells per test²⁶. With a pilot study (n = 259, 31 CIN 2+), the sensitivity for the assay was 67.7%, and specificity was 89.3% when compared to CIN 2 + pathology; notably the sensitivity increased to 100% when compared with CIN 3 + pathology (n = 259, 10 CIN3+)²⁵. In the pilot clinical study, the E7 oncoprotein assay had a sensitivity of 100% and specificity of 90.9% for detection of patients with CIN 2+.

The Arbor Vita tests involve a 45-minute sample preparation process requiring extensive user interaction and centrifugation when processing samples²⁷. These requirements for instrumentation and trained personnel limit use in settings where diagnostic testing is most desired⁶. The goal of this work was to match the performance of these tests without the need for complex sample preparation or instrumentation. With point-of-care lysis and lyophilized reagents, the HPV E7 paper test requires minimal user input while retaining performance. After probit analysis, the limits-of-detection of the HPV E7 paper oncoprotein assay were determined to be: 328 HeLa cells, 15,968 SiHa cells, 12,287 Ca Ski cells, 3,513 MS751 cells. Therefore the desired limit-of-detection was achieved for HPV18 E7 (HeLa cells) and close to the limit of detection with HPV45 E7 (MS751 cells). SiHa and Ca Ski (HPV16) limits of detection were slightly higher at less than 16,000 cells, although this value is still reasonable for a point-of-care assay that requires no sample manipulation. With these data, the HPV E7 paper test was able to sufficiently quantify HPV 16/18/45 E7 oncoprotein.

One sample out of 19 was misidentified. The < CIN 2 sample that falsely tested positive for HPV E7 by paper-based assay was HPV-negative by clinical standard evaluation. This may indicate the presence of HPV18 E7 in the sample, as oncoproteins can be present before progression to CIN 2+^11,13. Hui et al. found hrHPV oncoproteins E6/E7 present in 11.1% of their CIN 1 samples and 97% of CIN 2 + samples²⁸. The false positive result could also result from sample contamination, and processing samples individually may reduce contamination²⁹.

Future work will focus on preparing this assay for use in remote environments and testing in larger clinical studies, including the implementation of lyophilized reagents and evaluation of stability after storage. To be most effective in low-resource areas, the assay should not be reliant on cold storage and work well in high heat and humidity, in addition to working without electricity. Future stability studies will inform assay storage conditions and performance in high heat and humidity. HPV16 detection can be improved in the future with the addition of a secondary HPV16 E7 detection antibody to further reduce SiHa and Ca Ski limits of detection, as the antibodies used in the text were developed using HPV18 E7 antigen. In addition, performance would likely improve if the paper tests were produced under commercial manufacturing conditions with standardized production processes. To expand its relevance in resource-limited settings, a self-sampling collection option could be incorporated, and the assay could be evaluated for use with mobile quantitative readers to provide objective results at the point-of-care^30–32.

Larger clinical studies will be necessary to further evaluate the sensitivity and specificity for CIN 2 + detection. The implementation of cellular controls and testing against gold standards for HPV 16/18/45 E7 will be critical to assessing test performance. Testing for HPV 16/18/45 E7 mRNA, as well as other hrHPV E7 mRNA, in future clinical studies could provide valuable information about test performance. Higher levels of mucus, blood, and other components in patient samples may interfere with flow or target detection. Increasing sample concentration could also be explored to improve sensitivity. This could include sample collection directly into a small volume of xTractor buffer or additional concentration of samples stored in PreservCyt buffer. Increasing sample concentration would increase the number of cells and amount of E7 applied to the test and could improve sensitivity of CIN 2 + detection. Threshold and reagent optimization following sample concentration in larger clinical studies could improve specificity and reduce false positives.

While larger-scale validation is still needed, the HPV E7 oncoprotein test performs well with clinical samples, detecting CIN 2 + pathology with high sensitivity and specificity. The assay could serve as a follow-up test for women positive for high-risk HPV DNA to stratify those at greater risk of preinvasive disease, or potentially as a standalone test in a same-day screen and treat program after further clinical validation. A paper-based, low-cost test to identify women likely to have CIN 2 + lesions would allow patients to be screened, diagnosed, and treated within the same visit, reducing loss to follow-up while preventing overtreatment in already resource-limited settings.

Conclusion

This study demonstrated the successful creation of a sample-to-answer HPV oncoprotein assay. The assay consists of five simple user steps with a 15-minute workflow, has no infrastructure requirements, and uses a low-cost platform. The assay was validated with HPV16, 18, and 45 cellular samples and with a pilot clinical study, producing sensitivity of 100% and specificity of 90%. Further clinical validation with larger sample size and field testing is necessary, as the current feasibility study is limited with only 19 samples. However, with promising performance and a truly point-of-care format, the HPV E7 paper oncoprotein assay could prove a helpful tool for diagnosing cervical dysplasia in resource-limited settings, expediting referral to treatment pathways for women at highest risk of preinvasive disease progression.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

CAS, SP, PEC, KMS, and RRK conceived of the project and its scope. CAS and SP designed and performed experiments and analyzed data leading to the development and optimization of the assay. CAS and SP developed the sample preparation protocol. SGP and LL coordinated the clinical sample collection in El Salvador with supervision from MM, and CAS and SP analyzed the pilot study data. JF, PR, and PE provided clinical expertise on pathology review of clinical samples. CAS, SP, and KEH prepared the manuscript with input and supervision from KMS and RRK and editing from PEC, MB, and SGP. Funding for this research was acquired by KMS and RRK.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and its Supplementary Information files. Should any raw data files be needed in another format, they are available from the corresponding author upon reasonable request. Source data are provided with this paper.

Declarations

Competing interests

Dr. Castle has received HPV tests and assays at a reduced or no cost for research from Roche, Becton Dickinson, Cepheid and Arbor Vita Corporation. All the remaining authors declare no conflict of interest.