Detection of 14 High-risk Human Papillomaviruses Using Digital LAMP Assays on a Self-digitization Chip

Abstract

Cervical cancer is the fourth-leading cause of cancer deaths among women worldwide and most cases occur in developing countries. Detection of high-risk (HR) HPV, the etiologic agent of cervical cancer, is a primary screening method for cervical cancer. However, the current gold standard for HPV detection, real-time PCR, is expensive, time-consuming, and instrumentation-intensive. A rapid, low-cost HPV detection method is needed for cervical cancer screening in low-resource settings. We previously developed a digital loop-mediated isothermal amplification (dLAMP) assay for rapid, quantitative detection of nucleic acids without the need for thermocycling. This assay employs a microfluidic self-digitization chip to automatically digitize a sample into an array of nanoliter wells in a simple assay format. Here we evaluate the dLAMP assay and self-digitization chip for detection of the commonly tested 14 high-risk HPVs in clinical samples. The dLAMP platform provided reliable genotyping and quantitative detection of the 14 high-risk HPVs with high sensitivity, demonstrating its potential for simple, rapid, and low-cost diagnosis of HPV infection.

Cervical cancer is the fourth most common cancer among women worldwide, with over 500,000 new cases annually, most of which occur in developing countries.¹ Almost all cervical cancer is caused by persistent high-risk (HR) human papillomavirus (HPV) infection, and HR-HPVs have been detected in 99.7% of cervical squamous cell carcinomas.^2–3 Over 200 HPV types have been identified, and some 18 HPV genotypes are classified as high-risk (HR-HPVs). HR-HPVs include 12 HPV genotypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58 and 59) as causal agents of cervical cancer, and another six (26, 53, 66, 68, 73 and 82) as probable or possible causes of cervical cancer. Of these, 14 HR-HPVs (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68) are commonly tested for primary cervical cancer screening, based on the frequency of their association with cervical cancer.⁴ HPVs 16 and 18 together account for nearly 70% of cervical squamous cell carcinomas globally and are thus genotyped individually in many recent tests.⁵

Early detection of pre-cancerous cervical lesions is important for preventing cancer progression. Additionally, many studies have shown that HPV testing is more sensitive than cytology for detecting cervical pre-cancer^6–8.Thus, primary screening strategies that include HPV testing are now recommended per “International Agency for Research on Cancer” guidelines, and have been implemented in the U.S. and other high-income countries.^9–16 In 2014, the U.S. Food and Drug Administration approved the Roche cobas HPV Test for primary cervical cancer screening.^17–18 This test measures HPV based on multiplex real-time PCR, to detect the 14 HR-HPV types (genotyping for HPVs 16 and 18, and one pool to detect the other 12 HR-HPV types). Recently, an increasing number of HPV testing kits have been developed and some are now validated for primary HPV screening. The INNO-LiPA HPV Genotyping Extra II kit, for example, detects and genotypes 32 types of HPVs based on sensitive PCR followed by reverse hybridization using type-specific probes. The Digene HC2 HPV DNA Test is widely used globally and does not rely on PCR but uses chemiluminescence for the qualitative detection of 18 HPV types. However, none of these tests provide a quantitative result and they require expensive equipment, expensive infrastructure, and complex procedures which are unsuitable for many low-resource settings. There remains a need for a simple, fast, and low-cost method for HPV detection. Additionally, viral-load information from a quantitative assay can provide further diagnostic and prognostic information, such as HPV persistence and clearance.

Isothermal amplification methods such as loop-mediated isothermal amplification (LAMP) allow fast, low-cost nucleic acid detection.^19–22 Standard, real-time LAMP, however, does not permit accurate quantitation because quantitation depends on the reaction time which in turn depends on amplification efficiency. Our group previously developed a microfluidic self-digitization (SD) chip that automatically partitions a sample to allow digital nucleic acid quantification. Here we combine this SD chip with LAMP to create a digital LAMP (dLAMP) assay that enables quantitative detection of HPVs.^23–25 Digital LAMP is an absolute quantitation method, and is therefore more robust than relative quantitation methods that require a reference. We used 14 sets of LAMP primers for the 14 commonly tested HR-HPVs, and evaluated the performance of the dLAMP assay for genotyping and quantitative detection of the HPVs.

EXPERIMENTAL SECTION

Chemicals and reagents.

WarmStart LAMP Kits (DNA & RNA) and bovine serum albumin (BSA) were purchased from New England BioLabs (Ipswich, MA). Calcein (high purity) was purchased from Thermo Fisher (Waltham, MA). Betaine, Tween-20, light mineral oil, and MnCl2 were purchased from Sigma-Aldrich (St. Louis, MO). Primers and probes were synthesized by Sigma-Aldrich or by Integrated DNA Technologies (Coralville, IA). Abil WE 09 and Tegosoft DEC were purchased from UPI Chem (Somerset NJ). Poly(dimethylsiloxane) (PDMS; Sylgard 184) was purchased from Dow Corning (Midland, MI). SU-8-2050 photoresist was purchased from Microchem (Newton, MA).

LAMP primer design and screening.

LAMP primer sets including loop primers were designed using LAMP Designer v. 1.15 (Premier Biosoft). Input sequences were obtained from GenBank and consisted of either the complete reference genome for each HR-HPV type, or only the HPV L1 gene. LAMP design parameters were set using BLAST such that detrimental binding of the six primers in each set to each other as well as to the human genome was avoided. Six LAMP primer sets were designed against the whole HPV genome, and six sets were designed against only the HPV L1 gene. The highest LAMP Designer scores were tested for fast detection with no false-positives during a 90 min assay run time with negative-control samples (DNA from a cervical cancer cell line devoid of HPV (C33A; ATCC), or water). For each HPV type, 2–3 primer sets were optimized for LAMP assay temperature and for MnCl₂, calcein, and primer concentrations. The assays were then optimized for best results on the SD chip. Table S1 lists the final LAMP primer sequences.

HPV reference plasmids and quantification.

Purified plasmid DNA containing full-length genomes of 12 HR-HPV types, the half-genome of HPV 35, and the L1 gene of HPV 68 were kindly provided by Dr. Denise Galloway (Fred Hutchinson Cancer Research Center, Seattle, WA). The DNA was transformed into 10-beta competent E. coli (C3019H, New England BioLabs), and the cells were streaked onto agarose plates containing 50 μg/mL ampicillin and cultured overnight. Single colonies were picked and grown overnight in 2 mL of LB broth containing 50 μg/mL ampicillin. Plasmid DNA was isolated using a Qiagen Plasmid miniprep kit. The concentration of purified plasmid DNA (60 μL) was measured using a Nanodrop 2000 (Thermo Scientific) and converted to copy number based on the average base pair weight and plasmid length. All plasmid copy numbers were confirmed via quantification by ddPCR (Bio-Rad, QX200 droplet generator, QXDx droplet reader) and adjusted in TE buffer pH 8 to a starting concentration of 10⁸ copies per μL. DNA from human C33A cells containing no HPV was used as a no-template control and was added to each plasmid at 50–100 ng/μL to mimic the host cell DNA background normally present in a clinical sample.

Clinical samples and nucleic acid extraction.

Fifty clinical samples previously tested by PCR (Roche Linear Array) for a wide range of HPV types were obtained from the University of Washington HPV Research Group Specimen Repository for use in assay validation (Table S2). Six samples were negative for HPV, three samples contained low-risk HPV types only, and 41 samples contained at least one HR-HPV type. HPV 35 was present only in samples positive for multiple HPV types, whereas HPV 68 was present only in samples without any other detectable HPVs. Samples positive for multiple types contained 2–7 HPV types. The samples were a mix of cervical swabs, self-collected vaginal swabs, and pelleted urine cells pre-digested with proteinase K in 200 μL buffer. DNA was extracted from 200 μL of each clinical sample using a QiaAmp Blood Mini Kit (Qiagen), eluted in 100 μL AE buffer and stored at −20 °C.

Digital droplet PCR quantification.

Digital droplet PCR (ddPCR), an established commercial method, was used to quantify reference plasmid copies and HPV copies in clinical samples. We designed new primer pairs and TaqMan probes targeting the L1 gene of HPV 45 and HPV 68, and used published primers and probes targeting the remaining 12 HR-HPVs (Table S3).^26–27 For each ddPCR reaction, 25 μL were prepared containing1X ddPCR Supermix (No dUTP, BioRad), 0.9 μM HR-HPV type-specific primers, 0.25 μM TaqMan probe (Sigma), and 1 μL DNA template. 20 μL of each reaction was then used to generate droplets on the Bio-Rad QX200. Emulsified droplets were subjected to PCR on a Bio-Rad thermocycler using 1 cycle at 95 °C for 2 min; 45 cycles of 95 °C for 15 sec, 55 °C for 30 sec, and 72 °C for 30 sec; and 1 cycle at 37 °C for 30 sec. Subsequently, droplets were read on the Bio-Rad QX200 droplet reader and analyzed with BioRad Quantsoft Analysis Pro 1.0.596. The QX200 reads and analyzes 15000–20000 droplets per sample.

Real-time LAMP assays.

Optimized LAMP reactions were performed in duplicate in a total volume of 25 μL. Each reaction contained 0.15 μM F3/B3, 1 μM FIP/BIP, and 0.4 μM LF/LB primers (Sigma), 1X NEB WarmStart LAMP mix (NEB), 600 μm MnCl₂ (Sigma), 30 μM calcein (Molecular Probes), and 1 μL DNA template. The FIP/BIP primers were pre-heated at 95 °C for 8 min in 1x LAMP buffer together with the DNA template, then cooled to 30 °C for 3 min before adding the reaction mix. Non-template controls included H₂O, C33A cell DNA (50–100 ng/μL), and clinical samples with no HPV or low-risk HPV types. Reactions were run on a LC96 LightCycler (Roche) or a Bio-Rad CFX96 Real-Time PCR instrument for 90– 180 min at 65 °C and fluorescent signal at 560–580 nm was recorded at 30 sec (LC96) or 60 sec (CFX96) intervals.

Chip fabrication.

The SD chip microstructure was printed onto a Mylar photomask (Fine Line Imaging, Colorado Springs, CO) with a soft lithography technique described previously.²⁸ To make the reservoir for filling, we used a 3D printer (Lulzbot) to add extended 3D reservoir structures to the patterned reservoir area of the fabricated master mold. The patterned wafer with reservoir structure was treated with silane [(tridecafluoro-1,1,2,2-tetrahydrooctyl) trichlorosilane; Gelest Inc., Morrisville, PA] under vacuum overnight and placed in a custom aluminum-based mold that created a thin cavity in which the chips were cast from poly(dimethylsiloxane) (PDMS) (Sylgard 184 silicone elastomer kit, Dow Corning Corp., Midland, MI). The PDMS chips were allowed to cure at 70 °C for 3 h. Holes were punched into the reservoir of the molded PDMS replica to create an inlet and outlet. The molded PDMS replica was then exposed to oxygen plasma and bonded to PDMS coated glass slides. A polychlorotrifluoroethylene (PCTFE) plastic film (Ted Pella, Inc.) was bonded to the top of the PDMS chip to prevent evaporation during the LAMP reaction.

dLAMP assay on SD chip.

Calcein was mixed with MnCl2 solution at a 1:20 molar ratio and stored at 4 °C until use. The LAMP assay was performed in a 25 μL reaction containing 1× WarmStart LAMP Master Mix, 25 μM calcein, 0.5 mM MnCl₂, 0.2 M betaine, 1.0 mg/mL BSA, 0.5% Tween-20, HPV DNA, and indicated LAMP primers (0.02 μM F3, 0.02 μM B3, 1.6 μM FIP, 1.6 μM BIP, 0.8 μM LF, 0.8 μM LB). Prior to dLAMP experiments, the SD chip was primed by injecting oil (91% Tegosoft DEC, 9% light mineral oil, and 0.04% Abil WE 09) into the inlet reservoir. Subsequently, the air was displaced by oil, which eliminates air bubbles that might interfere with the reaction, from the chip. The chip was then filled with 8 μL of LAMP reaction mix including sample DNA, then digitized by the SD chip (1520 (8×190) 3.04-nL chambers). Similar to the real-time assay, FIP and BIP primers were allowed to pre-anneal to the template by a hot-start step. The SD chip was then heated at 65 °C for 90 min. Images were acquired every 10 min with a digital camera at a resolution of 3872×2764 pixels.

Well volume calculations.

A custom software program was used to calculate the filled sample volume in the SD chip (LabVIEW, National Instruments Corp.). The rolling ball background subtraction method was applied to the image acquired before amplification to create a background subtracted image of the chip. An adjustable threshold was applied to the image that maximized the number of fluorescent spots that matched the size and shape of the wells in the device. The locations of the spots were offset by an adjustable common displacement in orientation of the chip in the image and in the X and Y directions to produce the best match between spot locations and expected well locations. Spots which did not align with a well were discarded, and the remaining spots were identified as droplets in wells.

Sample in some wells may evaporate, thus affect the fluorescent intensity of the well. Therefore, only wells with fluorescent spots covering more than 50% of the well area were analyzed. From our prior research,²³ such wells contain droplets large enough to ensure full vertical filling in the brightest portion of the fluorescent spot. The program measures pixel intensity and selects the 20% of the spot area with the brightest pixels. The “Top Intensity” of the droplet was calculated as the average fluorescent intensity of that 20% area. If the entire well was filled completely, the total fluorescent intensity (“Filled Intensity”) of the droplet would equal the Top Intensity multiplied by the area of the well in pixels. However, droplets invariably do not fill the corners of wells; therefore, the actual total intensity of a spot will be less than its Filled Intensity. The filling fraction for a given droplet is defined as its actual total intensity divided by its Filled Intensity. The filling fractions for the droplets in each array on the device are averaged to produce an average filling fraction for those droplets. The average filling fraction is multiplied by the well volume for that specific chip to obtain an average volume for the droplets in the array. The average filling fraction of droplets in a single array is typically 0.8–0.9.

Data analysis.

To distinguish between positive and negative wells after amplification, each well was analyzed using two thresholds: a low threshold was applied to collect all of the wells analyzed (n), and a high threshold to collect only the positive wells (k). The number of HPV DNA molecules present in each well in the chip follows a Poisson distribution. Therefore the HPV DNA copy number can be calculated based on Poisson statistics following the equation λ = − ln (1 − k/n), where λ is the mean number of target DNA molecules per well, n is the total number of wells analyzed, and k is the number of positive wells. The confidence interval can be estimated as

$$\left(p + \frac{z^2}{2n} \pm z\sqrt{\left(\frac{p\left(1-p\right)}{n}+\frac{z^2}{4n^2}\right)}\right)/\left(1+z^2/n\right)$$

where p is the probability that a partition is empty, which can be approximated by p = e^−λ, and z is equal to 1.96 for a 95% confidence interval.²⁹

Limit of detection calculation for dLAMP.

Based on the false-positive rates for negative controls, we set 5 positive wells as the cutoff, so that arrays with 4 or fewer positive wells were deemed negative. The LOD at 95% confidence was calculated using the equation: LOD = − ln (1 − 5/1500)/(v﹒FR﹒R1﹒R2), where 5 is the positive signal cutoff for LOD, 1500 is the average total number of wells analyzed, v is the well volume (3.04 nL), FR is the filling ratio (0.85), R1 is the ratio of dLAMP to ddPCR with a single LAMP primer set, and R2 is the ratio of dLAMP using multiplexed LAMP sets to dLAMP with a single primer set.

Importantly, the cutoff used to calculate the LOD at 95% confidence interval was set based on the highest false-positive rate observed (HPV35). Accordingly, we expect the confidence level to be higher for HPV types with higher sensitivity (lower false-positive rate).

RESULTS AND DISCUSSION

To evaluate the utility of the digital LAMP SD chip platform for quantitative detection of the 14 high-risk HPVs, we designed and optimized HPV-specific LAMP primer sets.

LAMP primer design and benchtop screening.

Several groups have previously published HPV type-specific LAMP primer sets; however, these primers were not optimized for digital LAMP on a chip, and most primer sets did not target all 14 HR-HPV types.^{20–21, 30–33} To generate a large pool of candidate LAMP primers for optimization on the SD chip, we designed 12–18 new LAMP primer sets for each of the 14 HR-HPV types, targeting different genomic regions. Using reference plasmids containing whole genomes of each HPV type, we employed a standard LAMP protocol to screen all primer sets for time-to-detection (TTD). At 10⁶ plasmid copies per 25 μL reaction, all HPV types except HPV 33 were detected in 28 min or less. The shortest TTD was 14 min (for HPVs 56 and 66), and the longest TTD was 47 min (for HPV 33). In this first round, primer sets were selected based on fast TTD and no false-positive signals in a 90 min run time. The assays were run for 3 h to gauge each assay’s propensity for false-positives at later time points. For each HPV target, the 2–3 best-performing LAMP primer sets were selected for optimization of assay parameters including reaction temperature and concentrations of calcein, MnCl₂, dNTP, betaine, and primers. Subsequently, one LAMP primer set for each HPV type with the shortest TTD and no false-positives within 90 min for 10⁶ copies per reaction was selected for further evaluation.

Benchtop specificity and limit of detection.

We evaluated the specificity and limit of detection (LOD) of all 14 LAMP primer sets for HPV reference plasmids using the optimized LAMP protocol (see EXPERIMENTAL SECTION) at up to 90 min of assay run time and a reaction temperature of 65 °C. Since most swab samples contain 10²–10⁶ HPV copies/μL DNA, we used 10⁶ HPV copies per reaction for specificity testing. Each of the 14 LAMP primer sets detected the targeted HPV type and none of the other 13 types, at 10⁶ HPV plasmid copies/reaction. The LOD varied between the 14 HPV types, ranging from 10 to 10³ copies/reaction (Table 1).

Table 1.

Limits of detection of dLAMP and real-time LAMP

	FPR (%)	R1 (%)	R2 (%)	LOD of dLAMP (copies/μL)	LOD of real-time LAMP (copies/μL)
HPV 16	0.036	63.9	N/A	10.2	100
HPV 18	0.071	97.7	N/A	6.7	10
Multiplex 1-HPV 33	0	32.1	35.0	58.0	1000
Multiplex 1-HPV 39	0.31	67.8	30.8	31.1	100
Multiplex 1-HPV 45	0.072	85.5	49.5	15.4	1000
Multiplex 1-HPV 51	0.14	129.0	43.6	11.6	10
Multiplex 2-HPV 35	0.34	61.5	63.5	16.7	10
Multiplex 2-HPV 56	0.27	79.2	89.8	9.1	100
Multiplex 2-HPV 58	0.077	126.7	102.2	5.0	100
Multiplex 2-HPV 68	0.27	36.0	62.7	28.8	100
Multiplex 3-HPV 31	0.09	56.4	23.3	49.4	1000
Multiplex 3-HPV 52	0.20	64.7	79.6	12.6	100
Multiplex 3-HPV 59	0.070	63.5	53.3	19.2	1000
Multiplex 3-HPV 66	0	107.3	25.9	23.4	100

False-positive rates (FPR), ratio1 (R1), ratio2 (R2), and limit of detection (LOD) of dLAMP assays targeting 14 high-risk HPVs.

Digital LAMP workflow and output. Figure 1 shows the workflow of HPV DNA detection using dLAMP on the SD chip. The SD chip has two reservoirs, with inlet and outlet on one side and an enclosed reservoir on the other side that serves as overflow space for excess sample volume. The sample mixture was pipetted into the inlet and filled into wells. Fluorescence signals during dLAMP reactions on the SD chip are shown in Figure 2A. The reaction volume was calculated before amplification (see EXPERIMENTAL SECTION). At the start of the reaction, the temperature reached 65 °C in 80 sec and was held at that temperature until the end of the reaction. An increase in fluorescence was observed after 20–25 minutes in most partitions. The reactions were stopped at 90 min because nonspecific amplification in negative controls was observed after 90 min for some LAMP primer sets. Therefore, images acquired at 90 minutes were used to calculate HPV DNA concentration. Positive and negative wells were easily distinguished (Figure 2B), allowing automated counting.

Figure 1.

Schematic showing the workflow of HPV DNA detection with dLAMP on a self-digitization chip.

Figure 2.

(A) Fluorescence images of negative and positive samples during a dLAMP reaction on an SD chip. (B) Fluorescence image of an array loaded with a positive sample before and after the dLAMP reaction. (C) Enlargement showing individual wells, and a line scan showing the fluorescence intensity difference between positive and negative wells.

Optimization of dLAMP efficiency on the SD chip.

One limitation of the dLAMP method is that not all HPV DNA molecules present in the chip are amplified, and the HPV copy number obtained with dLAMP is lower than expected. Therefore, we next focused on optimizing the “digital efficiency” (amplified templates/total templates) of dLAMP in our system. We found that preheating the HPV plasmid DNA together with the LAMP primers significantly enhanced the digital efficiency. This enhancement indicated that HPV DNA strand separation (necessary for LAMP primer binding) was insufficient at 65 °C (Figure S1). To investigate which primers were responsible for the increased “digital efficiency”, we preheated DNA template with different combinations of LAMP primers (FIP/BIP, FIP/BIP+F3/B3, FIP/BIP+F3/B3+LF/BF). We found that preheating the template DNA together with FIP/BIP generated the best results (Figure S2B). We then optimized the loop primer (LF/BF) and F3/B3 primer concentrations (Figure S2A, C). These combined changes increased the average digital efficiency 10-fold, and improved the amplification speed, measured using real-time LAMP (Figure S2D).

Evaluation of dLAMP for specific and quantitative HR-HPV detection in clinical samples.

Next, we evaluated the performance of dLAMP on an SD chip for quantitative detection of the 14 HR-HPV DNAs in clinical specimens. The final set of 14 HR-HPV LAMP primer sets satisfied the following criteria: (1) efficient detection with an early positive signal and late false-positive signal in real-time LAMP; (2) good digital efficiency and low false-positive rate in dLAMP; (3) specificity of primer sets for only the intended target; and (4) no cross-reaction of primer sets. HR-HPV DNA copies in DNA extracted from 41 archived HPV-positive samples were quantified by dLAMP and by droplet digital PCR (ddPCR), a commercial method that provides accurate quantitation but is equipment-intensive. Correlation analysis showed strong association between the two assays (Pearson’s r = 0.92–0.99), indicating the reliability of the dLAMP SD chip method for viral quantification (Figure 3A–C). In most samples, copy numbers obtained by dLAMP were slightly lower than those obtained by ddPCR (Figure 3D). This difference could be attributed to lower template-to-volume ratio in dLAMP than in ddPCR, and/or to the stochastic nature of LAMP amplification initiation, which could cause a late fluorescent signal in some wells. The only sample with a dLAMP:ddPCR copy number ratio over 2 (HPV 16) had a low HR-HPV DNA concentration (12 copies/μL for ddPCR, 37 copies/μL for dLAMP). The HPV 16 LAMP primer set provided accurate quantification at 100 copies/μL. Overall, the dLAMP assay on an SD chip provided robust quantitation of HR-HPV DNA.

Figure 3.

Comparison of quantification of HPV in clinical samples by dLAMP in an SD chip and by ddPCR. Correlation between dLAMP and ddPCR (A) HPV 16 viral copy number (Pearson’s r = 0.99), (B) HPV 18 viral copy number (Pearson’s r = 0.99), and (C) viral copy numbers of the other 12 HR-HPVs (Pearson’s r = 0.92). The grey dotted line is the 45-degree line. Error bars represents 95% confidence intervals. (D) Ratio distribution (dLAMP/ddPCR copy numbers) in 41 clinical HR-HPV-positive samples.

Benchtop LAMP assay multiplexing.

To align with the Roche cobas HPV Test that identifies HPV 16 and HPV 18 individually and the remaining 12 HPV types as a pool, we combined the remaining 12 HPVs into three multiplexed assays: Multiplex 1 (HPVs 33, 39, 45, and 51), Multiplex 2 (HPVs 35, 56, 58, and 68), Multiplex 3 (HPVs 31, 52, 59, and 66). The multiplexed assays were tested for specificity using reference plasmids. None of the multiplexed assays showed cross-reaction with HPV 16, HPV 18, or any HPV type targeted by the other multiplexed assays. We also confirmed that combinations of three of the four assays in each multiplex did not produce a signal from the fourth group within 90 min (Figure S2), indicating primer set specificity. The assays were deemed ready for optimization on the SD chip.

dLAMP multiplexing.

Five assays using five groups of primers were run on the SD chip: HPV 16 (group 1), 18 (group 2), Multiplex 1 (group 3), Multiplex 2 (group 4), and Multiplex 3 (group 5). We designed and fabricated an SD chip with eight arrays. Five arrays were used for these five HPV assays, and two arrays were used for controls: a positive control (mix of HPV 16 and 18 plasmids) and a negative control (HPV-negative cervical swab DNA) (Figure 4A, B). To prove the feasibility of this design, clinical sample #44 containing HPVs 16, 18, 31, and 66 was tested on the SD chip. The dLAMP assay in this SD chip detected all four HPV types present (Figure 4C). HPV 16 and 18 were identified and quantified. For HPVs 31 and 66, the assay reads positive for Multiplex 3, indicating that one or more of the four HPV types are also present in the sample. The assay thus provides a specific quantitative result for HPVs 16 and 18, and a qualitative result for the other 12 HR-HPV types.

Figure 4:

(A) Schematic of dLAMP assay on an SD chip for detection of HR-HPVs in clinical samples. Red wells indicate fluorescence signal. (B) Photograph of SD chip containing eight arrays of 1520 (8×190) 3.04-nL nanowells connected by channels. (C) Image of dLAMP assays on an SD chip.

Limit of detection of multiplexed dLAMP.

To obtain the LOD of dLAMP for each HPV target, we determined the analytical cutoff, which depends on the false-positive rate (FPR). The FPR for each LAMP primer set was calculated based on negative control experiments (Table 1). The highest FPR was 0.34% (for the HPV 35 primer set). Based on this value, we set the cutoff at 5 positive wells; i.e., the array must have at least five positive wells to be deemed positive. Copy numbers obtained by dLAMP were typically lower than those obtained by ddPCR in most samples (Figure 3), indicating that not all DNA template molecules loaded into the chip were amplified, lowering the sensitivity of dLAMP. Therefore, we introduced a “Ratio1” correction factor (R1), defined as the ratio of templates amplified by individual HR-HPV-specific dLAMP to total templates (measured by ddPCR). Unsurprisingly, we also found that the digital efficiency decreased somewhat with multiplexing, indicating interference of multiple primer sets with LAMP amplification. Therefore, we introduced a second correction factor for multiplexed arrays, Ratio 2 (R2), defined as the ratio of positive wells obtained with multiplexed LAMP sets to positive wells obtained with individual LAMP sets. The LOD for each HR-HPV type (Table 1) was calculated according to the equation in the Materials and Methods. The LODs ranged from 5–50 copies/μL sample DNA input, which exceeds the LOD needed for clinical diagnosis of HPV (100 viral copies/μL). It should be noted that the SD chip can be scaled up in terms of well number and total volume to increase the sensitivity as needed for various applications.²³ The LOD of real-time LAMP were also measured with a series of plasmid samples with 10 times dilution (Table 1). Overall, dLAMP on an SD chip was more sensitive than real-time LAMP performed on a benchtop PCR instrument.

Conclusion

In summary, we developed a digital LAMP assay on a self-digitization chip for the specific and quantitative detection of HPVs 16 and 18 - the two HR-HPVs responsible for nearly 70% of cervical squamous cell carcinomas⁵ - as well as semi-quantitative detection of the other 12 HR-HPVs in three pools. We optimized LAMP primer sets for each of the 14 HR-HPV types for use in the dLAMP assay on an SD chip. We further optimized the digital efficiency of dLAMP to ensure accurate quantification of viral copies run of HPV16 and 18. The final SD chip-based dLAMP method had 100% assay sensitivity and 100% assay specificity for the detection of 41 clinical cervical swab samples containing HR-HPVs compared to the Roche Cobas results for the same samples. As the Roche Cobas results also showed excellent positive and negative predictive value for the detection of lesions of grade CIN2 or higher as well as cervical cancer in this clinical sample set, we can assume that the HPV dLAMP assay would perform likewise. Our results demonstrate that dLAMP is a reliable method for the quantitative detection of HPV DNA. The current layout of the SD chip is based on the layout of the most commonly used FDA-approved assays for cervical cancer screening used in the U.S. and globally, with HPV16 and HPV18 typing and pooled results for another 12 HR-HPVs. However, with injection molding, it is easy to reconfigure the SD chip to allow for extended HPV typing (for example for HPV31, 33, 45, or 52) as guidelines change. The specificity and sensitivity of the SD chip-based dLAMP assay is comparable to that of current commercial assays. This dLAMP SD chip platform can also be used for assays that target other pathogen families. Importantly, the LAMP method is particularly amenable to POC use not only for its high specificity, speed and efficiency similar to PCR, but also its robustness that includes a high tolerance for impurities in biological samples which simplifies nucleic acid extraction. With the multiplexed dLAMP reactions developed here, we expect this platform to be further developed as a fully automated system that includes on-chip nucleotide isolation for use in diagnostic applications in low-resource settings.