An integrated isothermal nucleic acid amplification test to detect HPV16 and HPV18 DNA in resource-limited settings

Abstract

High-risk human papillomavirus (HPV) DNA testing is widely acknowledged as the most sensitive cervical cancer screening method but has limited availability in resource-limited settings, where the burden of cervical cancer is highest. Recently, HPV DNA tests have been developed for use in resource-limited settings, but they remain too costly for widespread use and require instruments that are often limited to centralized laboratories. To help meet the global need for low-cost cervical cancer screening, we developed a prototype, sample-to-answer, point-of-care test for HPV16 and HPV18 DNA. Our test relies on isothermal DNA amplification and lateral flow detection, two technologies that reduce the need for complex instrumentation. We integrated all test components into a low-cost, manufacturable platform, and performance of the integrated test was evaluated with synthetic samples, provider-collected clinical samples in a high-resource setting in the United States, and self-collected clinical samples in a low-resource setting in Mozambique. We demonstrated a clinically relevant limit of detection of 1,000 HPV16 or HPV18 DNA copies per test. The test requires six user steps, yields results in 45 minutes, and can be performed using a benchtop instrument and minicentrifuge by minimally trained personnel. The projected per-test cost is <$5, and the projected instrumentation cost is <$1,000. These results show the feasibility of a sample-to-answer, point-of-care HPV DNA test. With the inclusion of other HPV types, this test has the potential to fill a critical gap for decentralized and globally accessible cervical cancer screening.

One Sentence Summary

We developed an HPV16 and HPV18 test and demonstrated its potential for cervical cancer screening in low-resource settings.

INTRODUCTION

Despite the efficacy of primary and secondary prevention methods, cervical cancer remains a major public health challenge, with ~604,000 women diagnosed and ~342,000 women dying from cervical cancer in 2020 (1). In high-income countries, screening with cytology and high-risk human papillomavirus (HPV) DNA testing has led to decreases in cervical cancer incidence and mortality (2, 3). Low- and middle-income countries (LMICs), particularly in sub-Saharan Africa, bear the greatest burden of cervical cancer because effective HPV vaccination and cervical screening programs have not yet been established (2, 4).

Evidence shows that testing for HPV DNA is more effective than cytology (5–8) and visual inspection with acetic acid (VIA) (8, 9) for detection of pre-cancerous lesions and early-stage cervical cancer. Accordingly, HPV DNA testing is increasingly recommended and adopted as the standard-of-care for cervical cancer screening in LMICs (4, 10, 11). HPV DNA testing can be conducted on self-collected cervicovaginal samples with comparable accuracy to provider-collected swabs when using HPV nucleic acid amplification tests (12, 13). Numerous studies show self-collection circumvents many of the barriers to provider-performed screening, including access to a healthcare provider, availability of clean speculums and other supplies, modesty concerns, and discomfort (14). Self-collection could thus decentralize sample collection, making HPV testing more feasible in LMICs.

Several barriers hinder the use of HPV DNA testing for screening strategy in LMICs, including the cost of current platforms and the need for laboratory equipment, technical staff, and protocols to run industry standard tests, like digene Hybrid Capture 2 (HC2)® and Roche cobas®. Thus, they remain too costly and complex for widescale and decentralized use in LMICs. Lower-cost assays, such as careHPV™, require sample batching, which can delay test results by days and up to weeks, depending on testing volume. Rapid, decentralized testing (i.e., at the point-of-care) is critical for effective cervical cancer screening, as it allows immediate treatment of women with abnormal results at the same visit (screen & treat). The screen & treat model decreases the loss-to-follow-up associated with longer sample-to-result time (15). This is especially important in LMICs, where patients may need to travel long distances to reach care providers and is now recommended by the World Health Organization (16). Rapid HPV testing methods, such as GeneXpert® HPV (Cepheid, Sunnyvale, CA), can produce results in one hour with similar clinical performance to industry-standard HPV tests (17–20). However, cost remains prohibitive. There is thus an urgent need for an effective, affordable point-of-care cervical cancer screening test.

To be effective at the point-of-care, nucleic acid tests must accept samples collected by patients and rapidly deliver clinically actionable results with high sensitivity and specificity. Instrumentation, including disposable test cartridges, should be affordable, rugged, field-serviceable, and designed to prevent user error. Disposable test cartridges should fully contain the sample and resulting amplicons to prevent biohazards and workspace contamination that can lead to future false positive test results. While great progress has been made to design tests that meet many of these criteria (21–23), there remains a need for an integrated test that meets all these criteria for cervical cancer screening. Additionally, to meet the critical needs of cervical cancer screening programs globally, a point-of-care test should achieve comparable analytical sensitivity and specificity to commercially available HPV DNA tests for the detected genotypes, which we benchmarked as 1,000 copies for each 10 μL reaction and no cross-reactivity with off-target genotypes based on the digene HC2 DNA test (Qiagen).

To address the gaps for point-of-care cervical cancer screening and meet the stated test requirements, we developed a prototype test for HPV16 and HPV18 DNA—which cause approximately 70% of cervical cancer worldwide (24). To develop sample-to-answer detection with a limited number of user steps, we designed an extraction-free assay to detect HPV16 and HPV18 DNA. The assay relies on recombinase polymerase amplification (RPA), an isothermal amplification technology that tolerates traditional PCR inhibitors, reducing the need to extract and purify nucleic acids. Additionally, instead of thermocycling, RPA relies on recombinase enzymes to separate DNA strands, allowing primers to bind, and polymerase enzymes to extend complementary DNA strands from the primers, allowing amplification at a single temperature. The RPA “nfo” assay allows for detection of amplified DNA on lateral flow strips by adding the nfo enzyme, which cleaves a modified, sequence-specific probe bound to target DNA. The result is an amplicon with dual antigenic labels, one from the cleaved probe and the other from the reverse primer, which can be captured in lateral flow.

The integrated test is performed using a platform called NATflow (Axxin Pty Ltd., Victoria, Australia). The system is designed to prevent workspace contamination by amplified DNA and subsequent false positive results, which remains a major challenge for point-of-care molecular testing, including previously reported isothermal amplification tests for HPV16 and/or HPV18 (22, 25, 26). The test is comprised of a disposable NATflow cartridge containing reagents for sample preparation, amplification, and lateral flow detection and the NATflow two-temperature heater. The projected cost at scale is <US$5 per-test and US$1,000 for instrumentation including US$500 for the NATflow heater and US$500 for a mini centrifuge.

We demonstrated successful integration of all steps required for sample-to-answer testing in a reliable, manufacturable, and contamination-free format. Working in both high- and low-resource settings, we demonstrated the validity of the test for HPV16 and HPV18 detection in patient samples from provider- and patient-collected cervicovaginal swabs with a sample-to-answer time of 45 minutes. The HPV16 and HPV18 DNA test integrates sample-to-answer processes reliably and in a format that could be manufacturable at scale and extended to include other HPV DNA genotypes.

RESULTS

Here, we describe results that characterize: 1) LoD and cross-reactivity of the isothermal amplification assay; 2) a sample preparation method that can be performed in low-resource settings and is compatible with downstream amplification and detection; 3) assay integration within a disposable, self-contained cartridge; and 4) assay performance in high- and low-resource settings with patient samples from provider- and patient-collected cervicovaginal swabs. Together, these results demonstrate an HPV16 and HPV18 DNA test that meets the technical parameters, can be used in low-resource settings, and could be extended to include other HPV DNA genotypes.

HPV16 and HPV18 RPA assays detect as few as 50–500 copies per reaction

We designed candidate RPA nfo primer and probe sets for the E7 gene of HPV16 and HPV18 using the RPA assay design software, PrimedRPA (28). We experimentally screened primer and probe combinations to maximize analytical sensitivity and specificity, first in singleplex format, then in multiplexed format. Performance of the optimized multiplexed primer sets is described in Fig. 1, and the sequences are included in Table S1. The RPA assay for HPV16 reliably detects as few as 50 copies of a 299-bp synthetic gBlock (Integrated DNA Technologies, Coralville, IA) of HPV16 DNA or 500 copies of an HPV16 transcript extracted from SiHa cells per reaction. The HPV18 assay detects as few as 50 copies per reaction of a 328-bp synthetic gBlock of HPV18 DNA or 50 copies of an HPV18 transcript extracted from HeLa cells per reaction (Fig. 1A–B). The LoD of both assays meets the digene HC2 benchmark of 1,000 copies per reaction. In this RPA nfo assay format, the probe and the reverse primer each contain one of the antigenic labels present on a dual-labeled amplicon that is captured in a sandwich lateral flow assay. The higher sensitivity of the HPV18 assay is likely due to the more favorable amplification efficiency between the probe and reverse primer (also known as the “cut” product) compared to the reverse primer and forward primer (the “uncut” product) (Fig. 1C). The uncut product only contains one of the antigenic labels necessary for capture in lateral flow; therefore, the resulting uncut amplicon does not bind to antibodies at the lateral flow test line (Fig. S1).

Fig. 1: Limit of detection and cross-reactivity of the HPV16 and HPV18 multiplex RPA nfo reaction.

(A) Example lateral flow strips for each tested input copy number of gBlock Gene Fragment (gBlock) DNA (HPV16 and 18) and extracted cellular DNA from SiHa (HPV16) and HeLa (HPV18) cells, amplified and detected using commercially available lateral flow strips (Milenia HybriDetect 2). Image contrast for all strips increased +20%. NTC: no-template control. (B). Signal-to-background ratio (SBR) for each condition shown in (A). Mean ± standard deviation, n=3 for each condition. Positivity threshold, indicated by the dashed line, was calculated to be 1.17. The limit of detection was calculated to be 50 copies of gBlock DNA or 500 copies of extracted SiHa cellular DNA for the HPV16 assay and 50 copies of gBlock DNA or extracted DNA from HeLa cells for the HPV18 assay. (C) Gel electrophoresis of the same targets shown in (A) and (B). The “uncut” primer-to-primer product should be 200 base pairs for both assays and is not detected by lateral flow. The “cut” probe-to-primer product is detected by lateral flow and should be 125 base pairs for both assays. The limit of detection on lateral flow is lower for the HPV18 assay in part due favorable formation of the cut product compared to the uncut product. bp: base pairs; L: Low molecular weight ladder. (D) Specificity of the HPV16 and HPV18 assays. Vertical labels refer to HPV type and horizontal labels refer to test (HPV16 or HPV18) or control (C) lines. For each target and off-target HPV type listed next to the lateral flow strips, 10⁶ input copies of gBlock DNA were amplified and detected using commercially available lateral flow strips. Positive signal is only seen at the corresponding test line when the HPV type of the input target corresponded to that targeted in the assay (dashed box), demonstrating high analytical specificity (n=1 for each target).

HPV16 and HPV18 assays are specific for the targeted genotypes

We found low cross-reactivity of the multiplexed HPV16 and HPV18 assay with other high- and low-risk HPV genotypes. As indicated by the dashed box in Fig. 1D, HPV16 and HPV18 reactions only form products at the appropriate test lines when genotype-specific target is present; this fulfills the specificity benchmark of no cross-reactivity with off-target HPV genotypes at a starting copy number of 10⁶ DNA copies per reaction.

Sample preparation: Achromopeptidase (ACP) is effective for cellular lysis and is compatible with direct amplification by RPA

The input to our HPV test is cells collected with a cervicovaginal swab, and the cells must first be lysed to release DNA for detection. To identify a lysis strategy compatible with downstream isothermal amplification, we compared the efficacy of several chemical and enzymatic lysis methods using cultured cells and preserved cells collected with a cervicovaginal swab (Fig. S2). In most cervical cancer screening programs, cervicovaginal swabs are collected directly into preservation media. For preserved cells in SurePath or PreservCyt media, we found enzymatic lysis approaches to be most effective (Fig. S2). Of the enzymatic approaches tested, achromopeptidase (ACP) required the shortest incubation times and was, therefore, judged most appropriate for point-of-care testing. Previous studies have shown ACP is an effective lysis agent and is compatible with use at the point-of-care (21, 29, 30). We demonstrated effective ACP lysis with a lower deactivation temperature of 75 °C, which simplifies point-of-care instrumentation development (Fig. S2). With this modified heat profile, we found that ACP provided comparable lysis to the standard of probe sonication (Fig. 2A). Moreover, we demonstrated that crude cellular lysate resulting from ACP treatment can be directly amplified in the RPA assay for HPV16 (Fig. 2B).

Fig 2: Sample preparation with ACP.

(A) Starting quantity of HPV16 DNA in SiHa lysate, as determined by qPCR, following sample preparation by either sonication or achromopeptidase (ACP). ACP lyses SiHa cells comparably to probe sonication (p=0.77; no target control (NTC) n=3, remaining conditions n=9 with 3 biological x 3 technical replicates). (B) Images of lateral flow strips containing RPA reactions run with ACP-produced SiHa cell lysate (NTC, +: n=1, SiHa lysate: n=3), and signal-to-background ratios (SBR) of the HPV16 test lines. Signal-to-background ratio (SBR) above the threshold indicates successful amplification of HPV16 DNA from the unpurified lysate. NLC: no lysis control; N.S.: not significant; +: positive control.

Custom NATflow-compatible lateral flow assay performs comparably to commercially available lateral flow strips

We developed custom lateral flow strips to be used with the NATflow detection cartridge (31). The HPV16 test line contains immobilized anti-fluorescein (FAM) antibodies, which capture FAM-labeled HPV16 amplicon when present. The HPV18 test line contains immobilized anti-digoxigenin (DIG) antibodies, which capture the DIG-labeled HPV18 amplicon when present. The control line contains immobilized biotinylated anti-mouse antibodies, which along with the biotin labels on both amplicons, capture streptavidin-conjugated gold nanoshells. Therefore, the control line produces a colorimetric signal with or without target DNA detection, and the HPV16 and HPV18 test lines only produce colorimetric signal in the presence of amplified, dual-labeled target DNA from RPA nfo reactions. The capture chemistry is modified from the commercially available Hybridetect 2 lateral flow strips (Milenia Biotec, Gießen, Germany), which immobilize streptavidin, anti-DIG, and anti-Rabbit antibodies at the first test line, second test line, and control line, respectively, and gold nanoparticle reporters contain anti-FAM antibodies.

We compared the analytical sensitivity of the custom NATflow-compatible lateral flow strips to that of Hybridetect 2 lateral flow strips, and we found that the LoD using the custom NATflow-compatible lateral flow strips was comparable to the LoD using Milenia lateral flow strips (Fig. S3).

Test integration: amplification and lateral flow detection are combined onto integrated NATflow platform with a LoD of 500–1,000 input copies per reaction

We integrated the HPV amplification and detection assays onto the NATflow platform (Axxin Pty Ltd). The NATflow system is composed of a consumable cartridge and a two-temperature heat block with integrated timers (Fig. 3A). The cartridge has three components: a lysis tube, an amplification chamber, and a lateral flow cartridge. An optional lateral flow reader (Axxin Pty Ltd. AX-2X-S) can be used to improve analytical sensitivity and reduce inter-user variability. In our workflow, we also used a mini centrifuge for sample preparation (Fig. 3B). All test reagents— including those for lysis, amplification, and amplicon detection—are lyophilized. The HPV test workflow on the NATflow platform is shown in Fig. 3C and Movie S1.

Fig. 3: Sample-to-answer HPV16 and HPV18 test process on NATflow.

(A) NATflow instruments and consumables. (B) Mini-centrifuge used in this application. (C) Images of the six user steps of the NATflow workflow: 1) add sample to ACP, 2) incubate sample with ACP at room temperature and then at 75 °C, and centrifuge the lysed sample, 3) transfer lysate to amplification reagents, 4) add amplification chamber cap and place amplification chamber in 39 °C heater, 5) twist chamber into lateral flow cartridge, and 6) read result. The total time-to-result is approximately 45 minutes.

First, a cellular sample collected with a cervicovaginal swab is added to lyophilized ACP within the lysis tube. The tube is incubated at room temperature (20–23 °C) for 5 minutes and then placed into the first heater port (75 °C) on the NATflow heat block for 5 minutes. An LED ring visually indicates the five-minute countdown. Samples are spun in a mini centrifuge for 5 minutes to pellet cellular debris to minimize risk of inhibition of target amplification. Ten microliters of lysate are manually transferred to the lyophilized HPV16 and HPV18 RPA amplification reagents in the amplification chamber. The amplification chamber cap is added, and the amplification chamber is placed into the second NATflow heater port (39 °C). A second countdown timer provides a visual cue of the remaining time, and visual and audio cues indicate when the amplification time of 20 minutes has expired. Finally, the amplification chamber is inverted and twisted into the lateral flow cartridge. The threads between the amplification chamber and lateral flow cartridge ensure that the system remains closed during elution, preventing workspace contamination by amplified DNA. The total sample-to-result time is approximately 45 minutes, and the cartridge provides a visual readout of HPV16 and/or HPV18 positivity.

Integrated NATflow cartridge reliably detects 500–1,000 input copies of target DNA

Integrating an RPA-based assay into a fully enclosed sample-to-answer test involves two challenges: first, RPA reactions need to be diluted for detection by lateral flow strips, and second, RPA efficiency is improved with agitation throughout incubation. Previous work has shown that reducing the RPA reaction volume is an effective alternative to agitation (33, 34). Using this approach, we optimized elution buffer volume and reaction volume, as well as gold nanoshell volume, to maximize signal strength and minimize test cost (Fig. S4).

To evaluate performance of the optimized, integrated NATflow cartridge, we lyophilized amplification reagents and gold nanoshells and stored all lyophilized test components in desiccated foil pouches. We tested a range of 100 through 10,000 input copies of HPV16 and HPV18 DNA extracted from SiHa and HeLa cells, respectively (Fig. 4A). In triplicate, we found reliable detection with at least 1,000 input copies of target per 10 μL reaction (Fig. 4B).

Fig. 4: Integrated cartridge performance to amplify and detect DNA extracted from SiHa and HeLa cells.

(A) Replicates of three cartridges with increasing copy numbers from 0 (no-target control, NTC) to 10,000 HPV16 and HPV18 copies per reaction using lyophilized amplification reagents. Cartridges were scanned after 15 minutes, and images were cropped around the test windows. (B) Signal-to-background (SBR) analysis of triplicate cartridges in (A). Average ± standard deviation of each test (16, 18) and control (C) line is plotted. Consistent amplification, using lyophilized reagents, and detection are visible at 500–1,000 copies and higher.

Finally, we evaluated reproducibility of 10 positive control and nine negative control samples with liquid RPA reactions in the integrated NATflow cartridge. Repeated measurements demonstrated complete, consistent separation between the positive and negative signals at the HPV16 and HPV18 test lines (Fig. S5).

Lateral flow strips remain stable at room temperature for a minimum of one month

We evaluated stability of custom NATflow-compatible lateral flow strips with an optimized conjugate buffer containing 5% dextran and 5% bovine serum albumin. Assembled strips were stored in sealed foil bags with desiccant under two conditions: room temperature (20–23 °C) with ambient humidity and high heat and humidity (37 °C). We found similar signals that remained positive throughout a month of storage in both conditions (Fig. S6).

Sample-to-answer NATflow test detects >500–1,000 copies of HPV16 and HPV18 from provider-collected cervicovaginal swabs in a high-resource setting

We demonstrated sample-to-answer HPV16 and HPV18 testing on the NATflow platform with preserved and non-preserved cellular samples. Briefly, ten μL of lysate or supernatant were added to lyophilized RPA reagents within the NATflow amplification chamber and placed into the amplification chamber of the NATflow heat block at 39 °C for 20 minutes. The amplification chamber was then inverted and twisted into the lateral flow cartridge to elute amplicons onto the lateral flow strip. Fifteen minutes later, strips were read visually and the lateral flow cartridge was scanned on a flatbed scanner and on the AX-2X lateral flow reader (Axxin Pty Ltd).

Results from testing provider-collected cervicovaginal swabs collected in preservative are summarized in Fig. 5. The HPV DNA copy number per reaction was quantified by genotype-specific HPV qPCR assays (Fig. S7). Only samples that previously tested positive by the standard-of-care test, Roche cobas HPV, were quantified with qPCR. With preserved samples, the integrated RPA test detected HPV16 in 12 out of 12 samples that had >1,000 copies of HPV16 DNA per reaction and one out of two samples that had 500–1,000 copies of HPV16 DNA per reaction. The integrated test was negative for six out of six samples that had fewer than 500 copies of HPV16 DNA per reaction, as well as all nine HPV16 negative clinical samples (Fig. 5A). The integrated test detected HPV18 DNA in the single sample with 500–1,000 copies of HPV18 DNA per reaction. The integrated test was negative for two out of two samples with fewer than 500 copies per reaction and for all 27 HPV18 negative clinical samples (Fig. 5B). Stratifying by qPCR-quantified starting copy number, the HPV16 sensitivity was 100% for samples with at least 1,000 copies per reaction and 93% for samples with at least 500 copies per reaction. Test specificity for HPV16 was 100%. The positive and negative predictive values for HPV16 were 86% and 56%, respectively. The low number of HPV18-positive samples precludes robust sensitivity and predictive value characterization, but HPV18 specificity was 100%. The overall percent agreement between NATflow and cobas for HPV16 and HPV18 combined in preserved samples was 85%. All discordant samples had fewer than 500 copies per reaction, and false negatives had lower starting HPV DNA amounts than true positives (Fig. 5C). Cartridge images are included in Figs. S8 and S9.

Fig. 5: Preserved cervicovaginal sample results processed by the Axxin AX-2X-S reader.

(A) Number of samples determined to be positive and negative by the HPV16 assay on the NATflow, stratified by number of starting copies as determined by qPCR. (B) Number of samples determined to be positive and negative by the HPV18 assay on the NATflow, stratified by number of starting copies as determined by qPCR. (C) Log copies of HPV DNA, measured by qPCR, for samples that were true positives (TP) and false negatives (FN) by the NATFlow HPV16 and HPV18 assay. There was a statistically significant difference (*) between mean Ct values for TP and FN samples by the NATflow HPV16 assay (p<0.01); statistical analysis was not conducted between the TP and FN samples by the NATflow HPV18 assay due to small number of positive samples. Arrow indicates increasing amounts of HPV DNA.

Similar results were observed with non-preserved provider-collected samples (Fig. S10). For HPV16, three of three samples with >1,000 copies per reaction, one out of two samples with 500–1,000 copies per reaction, and one of two samples with fewer than 500 copies per reaction were detected by the integrated RPA assay. Only one HPV18-positive sample was collected, which had fewer than 500 copies per reaction, and HPV18 was not detected in the integrated RPA test. The overall percent agreement between the NATflow and cobas for non-preserved samples was 87.5%. All discordant samples had fewer than 1,000 copies per reaction.

With both preserved and non-preserved provider-collected cervicovaginal swabs, the LoD was estimated to be 1,000 copies of HPV16 or HPV18 DNA per reaction, as 15/15 HPV16 samples and 1/1 HPV18 sample with at least 1,000 copies per reaction were detected.

Field evaluation with self-collected samples in Maputo, Mozambique detected 80% of samples below a Ct value of 35

The developed HPV16/18 NATflow test was optimized for use in a low-resource setting with patient-collected samples, and we evaluated the optimized test at Hospital Geral de Mavalane in Maputo, Mozambique. Results of NATFlow were compared to a composite gold standard based on GeneXpert testing, with reflex Ampfire genotyping only for samples that tested HPV 18/45 positive by GeneXpert. Samples were self-collected by women undergoing primary cervical cancer screening by HPV testing into PreservCyt preservative media. In this setting, changes to the methods included a longer centrifugation (10 minutes), higher temperature following ACP incubation (95 °C), and processed supernatant was not diluted prior to testing.

Results from the field evaluation are summarized in Fig. 6. There was no difference in the HPV16 detection between GeneXpert and RPA detection of HPV16 (p=0.80, exact McNemar’s chi-square test). The integrated RPA test detected HPV16 DNA in: four of five samples with GeneXpert Ct values < 30, six of seven samples with Ct values between 30–35, and one of five samples with Ct values > 35. Additionally, the HPV16 assay tested negative for 34 of 38 samples that were negative for HPV16, i.e., negative for all HPV genotypes, positive for HPV18, or positive for other hrHPV genotypes (Fig. 6A).

Fig. 6: Field testing in Mozambique.

Cervicovaginal samples collected into PreservCyt buffer (n=55) were tested using NATflow and results compared with GeneXpert HPV test results. Agreement between GeneXpert HPV and the NATflow (A) HPV16 assay and (B) HPV18 assay are stratified by GeneXpert Ct value. (C) GeneXpert Ct values for samples that were true positives (TP) and false negatives (FN) by the NATflow HPV16 and HPV18 assay. There was a statistically significant difference (*) between mean Ct values for TP and FN samples by the NATflow HPV16 assay (p=0.05) but not by the NATflow HPV18 assay (p=0.14). Arrow indicates increasing amounts of HPV DNA.

HPV18 detection was marginally greater by GeneXpert and AmpFire than RPA (p=0.07, exact McNemar’s chi-square test). The integrated test detected HPV18 DNA in three of four samples with Ct values < 30, one of two samples with Ct values between 30–35, and three of eight samples with Ct values > 35. The HPV18 assay tested negative for 40 of 41 samples that were negative for HPV18 (Fig. 6B). All four samples that were positive for both HPV16 and HPV18 were accurately detected (Fig. S11).

Truly positive samples that tested positive for either HPV16 and HPV18 with the NATflow test had mean Ct values that were lower than samples with false negative results for each genotype (Fig. 6C). Stratifying by HPV Ct value, the sensitivity of the NATflow HPV16 assay was 80% for samples with Ct < 30 and 83% for samples with Ct ≤ 35. Test specificity for HPV16 was 89%. Positive and negative predictive values for HPV16 were 73% and 85%, respectively. Low numbers of HPV18 positive samples precluded sensitivity, positive predictive value, and percent agreement calculations. Test specificity for HPV18 was 98%, and the negative predictive value for HPV18 was 85%.

Additionally, samples that tested falsely negative with the NATflow test for either HPV16 and HPV18 had higher cellular content, as measured by the GeneXpert sample adequacy control, than samples that returned true positive results. Samples with lowest cellular content (high Ct values) and highest HPV DNA content (low Ct values) had the highest percent of true positives at 92% (Fig. S12). Samples with highest cellular content and lowest HPV DNA content had the lowest sensitivity, indicating samples with higher cellular content may have been inhibitory.

DISCUSSION

We developed and evaluated an HPV16 and HPV18 DNA point-of-care test that integrates sample preparation, isothermal amplification, and lateral flow detection in a format that is manufacturable at scale, self-contained, and affordable. The process of operating the test is simple and requires few user steps. We demonstrated a high level of agreement with a reference HPV test in a high-resource setting using provider-collected cervicovaginal samples, and we demonstrated moderate agreement with reference HPV tests in a low-resource setting in Maputo, Mozambique using patient-collected samples. The study in Mozambique was the first field effort to evaluate this developed test, and the resulting data will be used to improve detection of HPV16 and HPV18.

In developing this test, we designed a multiplexed isothermal amplification assay for HPV16 and HPV18 with high analytical sensitivity and specificity. In addition, we designed and validated a sample preparation method that is appropriate for direct addition to an isothermal amplification reaction in a low-resource lab setting. We then integrated sample preparation, isothermal amplification, and lateral flow detection onto the NATflow platform, optimizing several volume and cartridge parameters to allow for the necessary dilution of RPA products for lateral flow detection. We demonstrated a LoD between 500–1,000 copies of HPV16 and HPV18 DNA per reaction for genomic DNA from cultured cells. A similar LoD was confirmed with provider-collected cervicovaginal swabs collected into cellular preservation media (n=30) and into a non-preservative buffer (n=11) when evaluated in a high-resource setting.

The demonstrated LoD is comparable to digene HC2, but lower than some other industry standard HPV tests, including GeneXpert HPV (Cepheid, Sunnyvale, CA), which is used in many resource-limited settings. GeneXpert HPV relies on nucleic acid extraction and PCR-based amplification to achieve a LoD of .01–.03 DNA copies/μL or 10–30 plasmid DNA copies per 1 mL reaction (35). Cepheid has made strides to improve the accessibility of the GeneXpert family of tests; however, challenges remain in expanding use outside centralized laboratories, including the need for climate control (36) and the high per-test cost ($15/test) (37). The reduction in analytical sensitivity of the NATflow test relative to GeneXpert HPV is primarily due to the extraction-free approach—which allows for lower per-test cost (<$5/test, Table S2) and less reliance on the infrastructure of a centralized laboratory. However, HPV DNA tests do not need to be ultra-sensitive, as low viral-load infections have a low likelihood of causing clinical disease. Detecting low viral loads has been shown to increase positivity rates by two- to threefold without increasing detection of clinical disease (i.e., high-grade abnormality detection) (38). Given the comparable sensitivity between digene HC2 and the developed test, we anticipate that the LoD of the test is clinically relevant, as the digene HC2 analytical sensitivity has been previously determined to optimize clinical sensitivity and specificity (39). For self-collected samples or lower limits of detection, a DNA extraction or concentration step will likely be necessary. Recent work toward point-of-care nucleic acid concentration could be readily incorporated to improve the LoD (23, 30, 40–42). For provider-collected samples in this study, we prioritized an extraction-free workflow for simplicity and cost, while maintaining a clinically relevant LoD.

When evaluating the test in high-resource settings, most of the clinical samples used were originally collected for routine cytology evaluation into SurePath media, which contains formalin to crosslink nucleic acids and proteins for long-term storage (43). Therefore, buffer exchange, centrifugation, and heat treatment at 120 °C were required to isolate the HPV DNA prior to test evaluation. However, preservative media is not necessary in a point-of-care testing scenario, in which samples are tested immediately following collection. We therefore included a small number of samples with direct collection into a non-preservative buffer, Tris-EDTA. We included a dilution and centrifugation step to reduce inhibition by cellular components and found a comparable LoD between preserved and non-preserved cellular samples. Further evaluation of cellular inhibitors in a larger patient population is needed to develop instrumentation-free sample preparation methods.

When evaluating the developed test in a low-resource setting, samples were originally collected into PreservCyt buffer, and a similar buffer exchange process was needed. Additional work is needed to evaluate compatibility with dry swab collection or collection into a molecular diagnostics-friendly buffer. Overall, we saw a lower level of agreement with the reference test in a low-resource setting, and we identified that false negative samples had higher cellular content than true positive samples. This is a particular challenge with self-collected samples; other studies have found the median cellular concentration of self-collected samples to be more than five-fold higher than in provider-collected samples from the same patients (p<0.001) (44). Additional optimization is needed to determine if dilution alone can circumvent cellular inhibition, or if an extraction step is needed to achieve clinically relevant sensitivity. If an extraction step is needed, this may add cost and complexity to the test.

HPV16 and HPV18 account for ~70% of invasive cervical cancers globally; as such, identification of these genotypes is an important part of triage and treatment guidelines in the US (45) and in LMICs (16). An effective cervical cancer screening test will need to include additional HPV genotypes. Recent research has demonstrated that to maximize clinical sensitivity and specificity for detecting cervical cancer and pre-cancer, eight HPV genotypes (16, 18, 31, 33, 35, 45, 52, and 58) that cause approximately 90% of cervical cancer should be targeted (46). We envision several strategies to achieve detection of eight types, including higher levels of multiplexing in 1–3 total cartridges per patient, as well as designing consensus primers. These approaches would rely on the same assay optimization parameters employed in this report and would be readily achievable with additional primer design. Moreover, incorporating a cellular control into the test, especially in the extraction-free format, would be helpful to differentiate between true negatives and inadequate samples, as well as to differentiate between false negatives and inhibitory samples. We note that cellular controls are included in the GeneXpert HPV test but not the digene HC2 test.

By demonstrating an effective solution to integrate sample preparation, isothermal amplification, and lateral flow detection that is appropriate for use in resource-limited settings, this work represents a major step toward a cervical cancer screening test that could be translated to clinical use in resource-limited settings and implemented at scale. With incorporation of additional genotypes, optimization of sample preparation, and additional field testing, the format and affordability of this test could allow for greater decentralization and reach of cervical cancer screening in resource-limited settings, a critical step in pursuit of global cervical cancer elimination.

MATERIALS AND METHODS

Study design

The hypothesis of this research was that an integrated point-of-care HPV16 and HPV18 test could match the analytic sensitivity and specificity of the digene HC2 assay (i.e., 1,000 copies of HPV DNA per 10 μL reaction, no cross-reactivity with other HPV genotypes). Additional hypotheses included that all molecular testing components could be integrated into a contamination-free platform and that pilot clinical evaluations would show reasonably high (>80%) agreement with standard-of-care HPV tests. These hypotheses were tested experimentally, first by establishing analytic sensitivity and specificity, then by conducting pilot field evaluation in high- and low-resource settings to qualitatively assess environmental cross-contamination and agreement with reference standard tests. Both field assessments were carried out under protocols approved by Institutional Review Boards (IRBs), including The University of Texas MD Anderson Cancer Center IRB (2014–0021) for the field evaluation in a high-resource setting and the Rice University IRB (2020–143) for the field evaluation in a low-resource setting. All participants provided informed consent.

Clean reaction setup

A dedicated room was used for pre-amplification reaction setup. Separate biosafety cabinets were used for sample preparation and for amplification reaction setup. Post-amplification activities were physically separated from pre-amplification activities to prevent workspace contamination by amplified DNA. When possible, all reagents were prepared aseptically in single-use aliquots. Lab spaces were routinely disinfected with 10% bleach and/or RNaseAway (ThermoFisher Scientific, Waltham, MA).

Cell passaging and preparation

SiHa, HeLa, and CaSki cells were acquired from the American Type Culture Collection (ATCC, HTB-35, CCL-2, and CRL-1550, respectively, Manassas, VA). Cells were passaged up to ten times, pelleted in quantities of 0.5–10 million cells, and stored at −80°C until use.

Target DNA preparation

Quantitative synthetic DNA standards were acquired from ATCC for HPV16 (VR-3240SD) and HPV18 (VR-3241SD). gBlocks Gene Fragment (gBlock) DNA for the full E7 genes of HPV6, 11, 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 72, and 82 based on the sequences published in the PapillomaVirus Episteme (PaVE, available: pave.niaid.nih.gov) (47, 48), were purchased from Integrated DNA Technologies, Inc. (IDT, Coralville, IA). DNA was extracted from SiHa and HeLa cells (ATCC HTB-35 and CCL-2) using the DNeasy Blood and Tissue kit (Qiagen) per manufacturer’s instructions, including a final elution step into 200 μL of nuclease-free water. HPV16 and HPV18 gBlock and cellular extract DNA were quantified against ATCC quantitative synthetic standards via qPCR, as described in qPCR below. gBlock DNA for all other HPV genotypes was quantified by a NanoDrop ND-1000. Single-use aliquots of ATCC, gBlock, and cell extract DNA were prepared and stored at −20 °C or −80 °C for up to six months prior to use. Dilutions of target DNA were prepared in nuclease-free water on each day of experiments.

qPCR

qPCR primers for HPV16 and HPV18 were designed by Primer3 software (Supplementary Table 1). All qPCR reactions were set up using the PowerUp SYBR Green Master Mix (Applied Biosystems). Each reaction contained 10 μL 2X Master Mix, 1 μL forward and 1 μL reverse primer (each at a 10 μM working concentration in 1X TE), 3–6 μL nuclease-free water, and 2–5 μL sample. When amplifying unpurified lysate, 2 μL of sample and 6 μL of nuclease-free water were added to each reaction. When amplifying purified target, 5 μL of sample and 3 μL of nuclease-free water were added to each reaction. The thermocycling protocol included the following steps: 50 °C for 2 minutes, 95 °C for 2 minutes, 40 cycles of 95 °C for 15 seconds and 60 °C for 1 minute, and a melt curve.

RPA

RPA nfo primers and probes were designed using PrimedRPA software (28) and are listed in Table S1. All primers were diluted to 10 μM working concentrations in 1X TE. 50 μL RPA nfo (TwistDx, Ltd, Maidenhead, UK) reactions were assembled with 1.95 μL 16rF, 1.95 μL 16rR, 1.3 μL 16rP, 1.05 μL 18rF, 1.05 μL 18rR, 0.7 μL 18rP and 29.5 μL rehydration buffer. For detection on Milenia HybriDetect lateral flow strips, 18rRM and 18rPM were used. For detection on custom NATflow lateral flow strips, 18rRN and 18rPN were used. To prepare 10 μL reactions, a master mix comprised of a discrete number of 50 μL reactions was prepared as described without magnesium oxaloacetate, then an equal number of enzyme pellets as the number of 50 μL reactions was added to the master mix. The mixture was vortexed and placed on ice until the addition of target. Target (10 μL per discrete 50 μL reaction) and magnesium oxaloacetate (2.5 μL per discrete 50 μL reaction) were added, the mixture was vortexed and spun in a minicentrifuge, and 10 μL of the combined RPA reagents and target were aliquoted into individual PCR tubes.

For experiments outside of NATflow cartridges, amplification was carried out for 20 minutes at 39 °C either on a benchtop heat-block or on a Bio-Rad CFX Touch 96 (Hercules, CA) with the heated lid set at 105 °C. Reactions were removed from the heat-block or Bio-Rad CFX four minutes into the incubation, vortexed, quickly spun, and replaced in the thermocycler for the remaining 16 minutes of incubation. For all other experiments, RPA reactions were incubated within the NATflow heat block for 20 minutes at 39 °C without agitation.

Lateral flow assays

To produce NATflow-compatible lateral flow strips, paper and plastic device components were laser cut (Universal Laser Systems VLS 3.60, Scottsdale, AZ) and assembled by hand. Devices were constructed from glass fiber (Alstrohm 8951, Mt Holly Springs, PA), nitrocellulose (Sartorius UniStart CN95, Goettingen, Germany), cellulose (Millipore CFSP223000), and .005” clear adhesive-backed film (Blick Art Supplies 55525–1021, Galesburg, IL). Glass fiber sample pads were cut to a dimension of 41 × 5 mm with a 4 mm-diameter circular cutout offset 2 mm from the end of the pad. Nitrocellulose membranes were cut to a dimension of 45 × 5 mm. Cellulose wicking pads were cut to a dimension of 18 × 5 mm. Sticky acetate backings were cut to a dimension of 87 × 5 mm with a 4 mm circular cutout offset 2 mm from the end of the backing and notches etched 16 mm from the edge opposite the circular cutout for nitrocellulose alignment.

Biotinylated anti-mouse anti-FAM, and anti-digoxigenin were prepared as previously described (31) and approximately 8, 8, and 20 μg, respectively, were deposited in successive 400 pL drops by a sciFLEXARRAYER S3 (scienion, Berlin, Germany) across 50 lateral flow strips. Conjugate dilution buffer was prepared by dissolving 1.25 g bovine serum albumin (BSA, Sigma-Aldrich A3912, St. Louis, MO) in 22.5 mL of nuclease-free water and 2.5 mL of 10X PBST (Genesee Scientific, 18–173, San Diego, CA). Streptavidin-functionalized 150 nm gold nanoshells (NanoComposix GSIR150, San Diego, CA) were diluted 1:1 in the conjugate dilution buffer. Forty μL of diluted gold nanoshells were pipetted onto each glass fiber sample pad and lyophilized for 24 hours as described in Lyophilization section.

To assemble the lateral flow strips, nitrocellulose was aligned 16 mm offset from the edge of the sticky acetate opposite the circular cutout. Second, the glass fiber sample pad was adhered to the sticky acetate backing such that the circular cutouts on each piece were aligned. Finally, the cellulose wicking pad was aligned with the edge of the sticky acetate opposite the glass fiber sample pad (Fig. S3). NATflow lateral flow strips were compared against commercially available lateral flow strips (HybriDetect 2, Milenia, Giessen, Germany). 10 μL RPA reactions with either 0 or 10⁴ input copies of HPV16 or HPV18 gBlock DNA were assembled as previously described. Products were diluted 1:50, 1:500, or 1:5,000 in either 1X PBST (NATflow lateral flow strips) or Milenia buffer (Milenia lateral flow strips). Ten μL of each diluted product was added to each type of lateral flow strip in triplicate. NATflow strips were placed into 200 μL of 1X PBST, and Milenia strips were placed into 80 μL of Milenia buffer within a 96-well plate. NATflow strips were imaged 15 minutes and Milenia strips were imaged 5 minutes after placing into running buffer. Strips were imaged on a flat-bed scanner (Epson Perfection V550 Photo, Suwa, Nagano, Japan).

Lateral flow assay stability evaluation

Lateral flow strips were prepared in a batch during study initiation. Negative control (nuclease-free water) and positive control (10,000 starting copies of HPV16 and HPV18 DNA) amplicons were diluted and stored at 4 °C for the duration of the study. Each week, stored amplicon was added to three stored lateral flow strips, and SBR was measured after 45 minutes.

Following the initial evaluation, the gold nanoshell conjugate buffer was optimized to improve conjugate release following storage. Sugars or surfactants that were anticipated to improve stability, including dextran, sucrose, trehalose, and tween, were evaluated at various densities in combination with BSA for impact on control line signal formation following up to two weeks of storage. Subsequently, the best performing combination (5% dextran and 5% BSA) was tested for impact on positive and control line formation across four weeks of storage in sealed foil pouches with desiccant at room temperature with ambient humidity, as well as high heat (37 °C) and humidity (>80%).

Lysis buffer preparation

ACP lysis buffer preparation was adapted from Buser et al. (30). First, 0.48 mg lyophilized ACP (MilliPore Sigma A3547, Burlington, MA) was dissolved into 100 μL of 1X TE buffer (ThermoFisher Scientific BP2473) with 5% trehalose (Millipore Sigma 90210) to create a 24 U/μL working solution. The 24 U/μL solution was then diluted 1:48 into 1X TE buffer with 5% trehalose, yielding the 0.5 U/μL solution used for lysis. Thermolabile Proteinase K (P8111S, New England Biolabs, Inc., Ipswich, MA) was diluted into 1X TE buffer (1 μL enzyme into a total of 50 μL).

A TCEP and sodium hydroxide (NaOH) buffer was adapted from Rabe and Cepko for comparison with enzymatic lysis buffers (49). 100X TCEP was prepared by first dissolving 358 mg TCEP in 3.43 mL nuclease-free water. Next, 1 mL of 0.5 M EDTA and 525 μL of 10 N NaOH (ThermoFisher Scientific SS267) were added. To prepare 1X TCEP+NaOH, the 100X buffer was diluted in nuclease-free water.

Lysis evaluation

To test lysis efficacy of several chemical and enzymatic buffers on preserved samples, a pooled negative sample was created from the stored clinical samples. 1 mL of cell suspension from ten patients with negative HPV results were combined, and 1.75 mL of the pooled sample was pelleted by centrifuging at 4,000 RPM for 5 minutes. SurePath supernatant was removed, and the pellet was resuspended in 500 μL PBS and split into equal volume aliquots. Each aliquot was again pelleted by centrifuging at 4,000 RPM for 5 minutes and was reconstituted in 50 μL of each lysis buffer. If a room temperature incubation was included in the heating profile for a particular condition, tubes were placed into heat blocks set at 23 °C for the designated time. After lysis, cell lysates were placed on ice for up to one hour prior to use.

To test lysis of stored clinical samples, lysate was amplified using beta actin primers in qPCR as previously described. To test lysis ability of cultured cells, a single CaSki cell pellet containing 1–2 million cells was reconstituted and split into equal volume aliquots. Each aliquot was re-pelleted and suspended into one of the following: lysis buffer, 1X TE buffer or 1X PBS. For probe sonication, cell aliquots were resuspended in 1X TE and placed in beakers of ice. Cells were sonicated for 30 seconds total over three 10-second pulses (Qsonica Q125, Newtown, CT). For all ACP lysis conditions, cells were incubated in 0.5 U/μL ACP for 5 minutes at 23 °C followed by a five-minute inactivation at 75 °C. For thermolabile Proteinase K lysis conditions, cells were incubated at 37 °C for 15 minutes followed by inactivation for 10 minutes at 56 °C. Chemical lysis incubations are noted in each experiment. CaSki cell lysate was quantified in the previously described HPV16 qPCR assay.

Lyophilization

An RPA nfo master mix for 24 reactions was prepared without adding enzymes, i.e. including rehydration buffer, primers and probes, and magnesium acetate. The master mix was vortexed thoroughly, then 8 μL were aliquoted into PCR tubes. Separately, 24 RPA nfo enzyme pellets were reconstituted in 1.2 mL of nuclease-free water and vortexed thoroughly. 10 μL of the reconstituted enzymes were aliquoted into 96 PCR tubes. Six milliliters of 0.5 U/μL ACP were prepared as previously described and aliquoted in 25 μL increments into individual PCR tubes. Master mix, enzyme, and ACP aliquots were frozen at −20 °C for a minimum of 2 hours, −80 °C for a minimum of 2 hours, and liquid nitrogen for 10 seconds prior to lyophilizing for a minimum of 24 hours (LabConco FreeZone 12, Kansas City, MO). As previously described, 40 μL of diluted gold nanoshells were deposited onto glass fiber pads and lyophilized for 24 hours without freezing prior to lyophilization. Lyophilized reagents were stored at −20 °C (amplification reagents, enzymes) or room temperature (glass fiber pads) in foil pouches with desiccant for up to one month.

NATflow cartridge assembly and use

NATflow cartridges were provided by Axxin Pty Ltd. (Victoria, Australia) and assembled according to manufacturer instructions. Amplification reaction volume, elution buffer volume, and gold nanoshell volume were optimized within the NATflow cartridge. Amplification reaction volumes between 5 and 50 μL were loaded into amplification cartridges and with 450 μL elution buffer in the cap. Next, elution buffer volumes between 250 to 450 μL were loaded into the amplification chamber cap and tested with 10 μL amplification reactions. Finally, gold nanoshells were diluted to an optical density of 5 OD, and 20 to 80 μL were loaded onto sample pads of lateral flow strips.

With optimized conditions, 250 μL of 1X PBST was loaded into the elution buffer ring and used with 10 μL RPA reactions and 40 μL of gold nanoshells on the custom NATflow-compatible lateral flow strips, described in Lateral flow assays. Lateral flow strips were placed within the NATflow cartridge during assembly on the same day as use. Sample preparation, amplification, and lateral flow detection were carried out using lyophilized reagents on NATflow instrumentation and with NATflow cartridges. Cartridges were scanned on a flat-bed scanner (Epson Perfection) and/or an AX-2X-S lateral flow reader (Axxin) after 15 minutes.

Reproducibility evaluation

Liquid RPA reactions were assembled as described in the RPA section. Briefly, 7.5 μL of master mix was aliquoted into PCR tubes, and 2 μL target—either nuclease-free water for no-target controls or 1,000 copies each of purified SiHa and HeLa DNA for positive controls—were added. Finally, 0.5 μL of magnesium oxaloacetate was added to each reaction, and reactions were mixed by vortexing and briefly spun in a microcentrifugte. The reactions were incubated at 39 °C for 20 minutes on the NATflow instrument, and the products were eluted into NATflow cartridges, which were scanned on the AX-2X-S lateral flow reader (Axxin) after 15 minutes.

Sample-to-answer testing with provider-collected clinical samples in high-resource settings

Cervicovaginal swabs were collected from a referral population under a protocol approved by the Institutional Review Board (IRB) at The University of Texas MD Anderson Cancer Center (2014–0021). All participants provided informed consent. De-identified cervical samples were transferred to Rice University under an exempt protocol reviewed by the IRB at Rice University.

Preserved cervicovaginal swabs were collected into SurePath preservation media (Becton Dickinson, Franklin Lakes, NJ) and stored at −80 °C until use. Standard of care results were obtained by Roche cobas testing. Prior to use, 500 μL of clinical samples were aliquoted. Sample aliquots were centrifuged for 10 minutes at 4,000 RPM, supernatant was removed, and cells were resuspended in 1X PBS; this process was repeated a second time to remove SurePath buffer from cells. Samples were then centrifuged for 10 minutes at 4,000 RPM, supernatant was removed, and cells were resuspended in 1 mL of nuclease-free water. To reverse chemical linkages between proteins and nucleic acids that form due to the formalin present in SurePath buffer, samples were heated at 120 °C for 20 minutes, incubated at room temperature for 10 minutes, then vortexed for 5 seconds, a protocol adapted from Gilbert et al (43). This preparation method yields a cellular sample that is ready for lysis and DNA amplification.

Non-preserved cervicovaginal swabs were collected with flocked nylon swabs (Puritan, Guilford, ME 25–3316-H). Swabs were placed into 300 μL of 1X TE buffer in 1.5 mL microcentrifuge tubes (Sarstedt, Newton, NC, 72.692.415), cut with sterile scissors to allow the microcentrifuge tube to close. Samples were transported to our lab on the same day as collection and stored at −80 °C, with a maximum of three freeze-thaw cycles, until use. Samples were diluted 1:4 in nuclease-free water prior to testing. Reference results were obtained by in-house qPCR.

Fifty μL of preserved samples or 25 μL of non-preserved samples were added to lyophilized ACP, pipetted up and down three times to mix, and incubated at room temperature (approximately 23 °C) for 5 minutes followed by a five-minute inactivation at 75 °C. Non-preserved samples were centrifuged for 5 minutes at 5,000 RCF to pellet cellular material. Ten microliters of the resulting preserved sample lysate or non-preserved sample supernatant was then transferred to lyophilized RPA nfo reagents within the NATflow amplification chamber. The lysate and lyophilized RPA reagents were mixed by pipetting prior to incubating the chamber on the NATflow heat block at 39 °C for 20 minutes to amplify DNA. At the end of the incubation, the NATflow chamber was inverted and twisted into the lateral flow cartridge. After 15 minutes, the lateral flow result was read visually, scanned, and read by a flat-bed scanner (Epson Perfection) and the Axxin AX-2X-S lateral flow reader.

To quantify HPV copies per reaction for preserved samples, 25 μL of the ACP-produced lysate was purified in a Monarch DNA Cleanup Kit (New England Biolabs T1030) per manufacturer’s instructions, with elution into 25 μL nuclease-free water. Purified samples were quantified by HPV16, HPV18, and beta actin qPCR assays as previously described.

To quantify HPV copies per reaction for non-preserved samples, DNA was extracted from 50 μL of the unprocessed sample in a DNeasy Blood and Tissue Kit (Qiagen) per manufacturer’s instructions, with elution into 200 μL nuclease-free water. Purified samples were quantified by HPV16, HPV18, and beta actin qPCR assays as previously described.

Sample-to-answer testing with self-collected clinical samples in a low-resource setting

Test evaluation was carried out using self-collected cervicovaginal samples from a screening population that were banked and stored for future research at Hospital Geral de Mavalane. Testing of the banked samples was conducted under a protocol approved by the IRB at Rice University (2020–143). Samples were collected into PreservCyt media, initially tested on GeneXpert HPV to inform clinical decision-making, concentrated into 1 mL aliquots, and stored at −20 °C for up to two years.

Two reference tests were used in this study: GeneXpert HPV and Multiplex High Risk HPV Real Time Fluorescent Detection with HPV16/18 Genotyping (Atila Biosystems, Mountain View, CA). For GeneXpert HPV testing, concentrated stored samples were re-diluted by adding 67.4 μL of concentrated sample to 1.13 mL of PreservCyt and tested, following manufacturer instructions. Samples that tested positive in GeneXpert channel P2 (HPV18 and 45) were then tested with the Atila genotyping kit per manufacturer’s instructions, with the exceptions that 100 μL of the kit’s lysis buffer was used with 50 μL of concentrated sample, and testing was carried out on an Axxin T8-ISO instrument.

The gold standard HPV result was determined by the following algorithm. Samples that tested positive in GeneXpert channel P1 (HPV16) of HPV were considered positive for HPV16 and samples that tested positive in GeneXpert channels P3 (HPV 31, 33, 35, 52, and 58), P4 (HPV 51 and 59), or P5 (HPV 39, 56, 66, and 68) were considered positive for other HPV genotypes. Samples that tested positive on GeneXpert channel P2 (HPV18/45) were with the Atila genotyping kit, and the Atila result was used to determine whether the sample was positive for HPV18 and/or HPV45. Samples positive by GeneXpert channel P2 (HPV18/45) but not tested with the Atila genotyping kit were excluded from analysis. GeneXpert Sample Adequacy Control (SAC) values were also recorded to analyze cellular content among true positive and false negative samples.

NATflow test materials were prepared in Houston, Texas, USA and were transported to Maputo, Mozambique. Briefly, solid ACP was added to 1.7 mL Eppendorf tubes, and the mass was recorded for reconstitution to 0.75 U/μL ACP in 1X TE buffer with 5% trehalose in-country. Liquid amplification reaction reagents were prepared as described in RPA. During air travel, reaction and lysis reagents were stored in medicinal travel thermoses with ice packs for approximately 36 hours. Lateral flow strip components were prepared as described in Lateral Flow Assays and assembled in Mozambique immediately prior to use.

To prepare samples for testing on NATflow, 250 μL of concentrated, stored sample were centrifuged at 4,000 RPM for 10 min. PreservCyt buffer was removed, and the cell pellet was resuspended in 1 mL 1X TE buffer. The sample was centrifuged again at the same speed and time, supernatant was removed, and the cell pellet was resuspended in 25 μL 1X TE buffer. Fifty μL of 0.75 U/μL ACP lysis buffer was added to the sample, which was incubated at room temperature for 5 minutes, then at 95 °C on a heat block for 5 minutes. Finally, the sample was centrifuged at 10,000 RPM for 10 minutes to pellet cellular debris, and the supernatant was re-aliquoted and vortexed for testing. To run the reaction, the same methods described in Reproducibility evaluation were used.

Image analysis

Signal-to-background (SBR) measurements on scanned cartridge images were calculated with a custom image analysis program in MATLAB. Raw RGB images of the Axxin cartridges were converted to grayscale, and the inner region of interest (ROI) containing the lateral flow strip inside of each cartridge was automatically identified through template matching. The complement of the inner ROI was taken, morphological top-hat filtering was performed to remove background noise, and the mean image pixel intensities across the image were calculated. SBR values for the HPV16, HPV18, and the control lines of the lateral flow strips were calculated by finding the three highest peaks of the mean intensity line plot, corresponding to the test and control lines. The background value was measured by finding the peak of the inverse of the mean pixel intensities across the image.

Cartridges images obtained by the AX-2X-S lateral flow reader were processed using an algorithm developed by Axxin Pty Ltd to find localized peaks within the read window. The peak value obtained for each test line was used for image analysis.

Statistical methods

To evaluate statistical significance between lysis conditions at the 0.05 significance level, a one-way analysis of variance (ANOVA) was used (Microsoft Excel version 16.38). To set positivity thresholds on signal-to-background (SBR) measurements, SBR was calculated from the HPV16 and HPV18 test lines of a minimum of three lateral flow strips with negative samples—either no-target controls or clinical samples that tested negative for HPV16 and HPV18 by the standard-of-care test—from the same experiment. Positivity threshold was set three or four standard deviations above the mean SBR value. For images scanned under controlled lighting, i.e., on the Axxin AX2X-S, positivity threshold was set three standard deviations above the mean peak SBR value. For cartridge images obtained on a flatbed scanner, the positivity threshold was set four standard deviations above the mean SBR value to circumvent false positives resulting from shadows. To evaluate statistical significance of the true positive and false negative means during the field evaluation in Mozambique at the 0.05 significance level, a two-tailed student’s t-test assuming equal variances was run (Microsoft Excel version 16.70).

Data and materials availability

All data are available in the main text or in the supplementary materials.