Introduction

Cervical cancer is one of the leading causes of cancer death in women, with 342,000 deaths in 2020 alone1. It is reported that 95% of cervical cancers are caused by Human Papillomavirus (HPV), with HPV16 being more prevalent. The infection can be sexually transmitted and could cause pre-cancerous lesions, which may ultimately progress into cancer if left untreated due to the uncontrolled infected cervical cell replication in some cases. The World Health Organization (WHO) recommended HPV screening methods include the Papanicolaou smear test, colposcopy, and nucleic acid amplification test (NAAT)2. There are some shortcomings to some of these tests, as the sensitivity of methods relying on visual tests depends on the practitioners, and they cannot discriminate between HPV subtypes. NAAT can distinguish subtypes, however, it requires amplification steps like polymerase chain reaction (PCR). This typically requires an expensive thermocycler, and there are possible risks of false positives due to non-specific amplification3.

Recently, several HPV DNA sensors that utilized probes for improving specificity in combination with different detection techniques, including electrochemical4,5,6, colorimetric7,8, and fluorescent9 sensors were reported. While the use of probes enhances selectivity, these aforementioned works suffered from one common drawback, i.e., the target DNA must be present in the form of single-stranded DNA (ssDNA) to allow probe binding. Since HPV is a circular double-stranded DNA (dsDNA) virus10 ssDNA must be obtained from the denaturation of dsDNA using chemicals or heat.

Peptide nucleic acid (PNA) is a DNA mimic with a completely redesigned backbone that makes it different from natural DNA. PNA/DNA hybrids showed higher thermal stability and selectivity compared to natural DNA/DNA hybrids11. Therefore, PNA has been extensively used as a probe for the detection of specific DNA sequences12,13,14. However, the DNA target must be present in single-stranded form (ssDNA) to allow probe binding. It is known from the literature that PNA probe(s) can bind directly with target dsDNA via triplex formation, triplex invasion, duplex invasion, and double duplex invasion11. The triplex formation and triplex invasion require a homopurine/homopyrimidine stretch, and duplex invasion requires an AT-rich region in the target dsDNA sequences15. With these restrictions, finding a short gene sequence for the specific detection of certain DNA targets may not always be possible16,17. On the other hand, in the double duplex invasion mode, two PNA probes that could be paired with each strand of dsDNA of the same region by Watson–Crick base pairing are required. Such a pair of PNA probes is inevitably complementary to each other, causing self-pairing. The issue can be overcome with the use of the new pyrrolidinyl PNA with an α/β-dipeptide backbone (acpcPNA) that offers better selectivity towards sequence mismatch and highly prefers antiparallel pairing with DNA formation of a stable self-pairing hybrid without additional modification required18,19. However, DNA and PNA probes on their own possess no intrinsic signals. Recently, a series of dicationic styryl dyes have been reported for high-performance fluorescent detection of dsDNA20. As opposed to conventional monocationic styryl dyes, the additional positive charge improves the dyes’ binding and fluorescence response towards DNA due to the favorable electrostatic interactions with the negatively-charged phosphate backbone of DNA. Since the dyes are not selective towards particular sequences of DNA, a PNA is needed to get a selective response. The redox-active property of similar styryl dyes also suggests the possibility of electrochemical detection21. This paves the way for dual detection, which allows users to choose the detection method according to their specific need or use both as a self-validation mechanism22. Paper-based analytical devices (PADs) offer several advantages, including cost-effectiveness, simple fabrication, low reagent usage, ease of use, portability, and disposability, making them suitable in resource-limited areas23,24,25. Either fluorescence- or electrochemical-based assays have been widely adapted for the paper-based platform due to their aforementioned advantages26,27,28,29.

Our main objective is to develop a dual-mode paper-based sensor for a simple, rapid, and accurate detection of HPV-DNA without requiring DNA denaturation. To the best of our knowledge, the DNA detection by double duplex invasion has not been adapted to a paper-based platform yet, while topics on the quantitative detection of dsDNA using double duplex invasion are also lacking30,31,32. Herein, for the first time, we report a dual-mode fluorescent/electrochemical paper-based detection of HPV16 dsDNA utilizing a pair of acpcPNA probes that can bind to dsDNA via double duplex invasion without the need for denaturation into single-stranded DNA, thereby simplifying the sample preparation step. Using this platform, the detection of multiple samples can also be performed simultaneously with multiple devices, offering high sample throughput. The dual-mode detection also offers increased flexibility by allowing users to choose between fluorescence detection for a fast, visual screening or electrochemical detection for extra sensitivity, depending on the available equipment and needs. Both detection modes can also be done on a single device for self-validation purposes. The detection of dsDNA products from LAMP-amplified dsDNA from clinical cervical swabs was performed to demonstrate the application of the developed sensor in real-world settings.

Results and discussion

Fluorescent and electrochemical behavior of dicationic PY2+(C4) dsDNA-binding fluorescent dye

DNA-binding dye was chosen as an indicator to provide label-free detection without the need for labeling of the PNA probes or the DNA target. Supabowornsathit et al.20 have synthesized many dicationic styryl dyes that showed remarkable fluorescence enhancement in the presence of DNA. The reasoning behind the choice of dye for this work, PY2+(C4), is discussed in the SI section 1.

In a proof-of-concept experiment, the ability of the fabricated sensors to detect HPV dsDNA was evaluated with HPV DNA samples consisting of 4 µM synthetic 53 bps target dsDNA and negative samples without the dsDNA. Figure 1a shows the difference in the intensities of the red signal obtained from positive and negative samples (referred to as Δred). The camera images in Fig. 1a show the remaining fluorescence after washing, which could be readily observed by the naked eye. In addition, the red intensity could be measured by ImageJ to differentiate between positive and negative samples with more clarity. The grayscale image showing only a red channel makes differentiation by eye easier as the dye emits red fluorescence. Subsequent images in this work will show only the red channel for clarity. The negative sample also contains some red intensity as a background signal, possibly due to the autofluorescence of the paper substrate under black light. Thus, the Δred intensity obtained from subtracting the negative control signal from the sample signal was used for subsequent experiments to minimize interference. A positive dsDNA sample, when mixed with PNA1, undergoes double duplex invasion with PNA2 on the sensor. The double duplex invasion by PNAs was possible because dsDNA was known to “breathe” in some regions33. The phenomenon leads to local denaturation of the dsDNA, allowing the PNA probes to partially invade, which further destabilizes the neighboring bases, leading to more opening of the dsDNA and finally completing the invasion process34. As PNA2 has already been immobilized on the sensor, the invaded dsDNA would then be fixed on the sensor surface together with PNA1. The dye added to the sensor then binds to the immobilized dsDNA through the minor groove binding mechanism20,35 and would also be fixed on the sensor by the dsDNA. The washing step was then performed to eliminate unbound dye from the sensor surface, leaving behind only the dyes that are bound to the dsDNA on the paper surface. More immobilized dsDNA can trap more dye molecules, leading to higher fluorescence intensity. In the absence of the dsDNA (negative samples), no dye molecules will be present on the sensor surface after the washing step, thereby generating no signals. The use of paper-based format is particularly suitable for this detection scheme as it allows facile surface immobilization and washing from its cellulose component, which would be difficult to perform in tube or plate formats. The wax-printed zone on the paper serves to limit the area of the sample to a small area for detection. Figure 1b shows the fluorescence spectra of various solutions with different mixtures of components. The photo of the solutions can be seen in Fig. S1, and it can be seen that solutions containing the dye and PNA without DNA do not exhibit fluorescence since this dye could not bind to acpcPNA. Solutions containing dye and DNA with or without PNA exhibit high fluorescence. The solution, including the dye, DNA, and both PNAs, exhibits less fluorescence signal than solutions lacking one of the PNAs since the dye could not bind to the PNA-DNA hybrid region. Therefore, the fluorescence response is primarily due to the binding of the dye to the dsDNA region. The synthetic dsDNA used in this experiment consisted of 53 bps, while the PNA probes had only 13 bps, hence, the dye could still bind with the extra base pairs outside the PNA invasion site. This leads to slightly less fluorescence than the uninvaded DNAs. In real samples, the number of base pairs in the viral genomic DNAs, as well as their amplified products, is much larger. Longer base pairs translate into more binding sites for the dye, leading to better sensitivity.

Fig. 1
figure 1

(a) Photographs in RGB and red channel only (top) and red intensity bar graph (bottom) of a sample containing 4 µM dsDNA (positive) and a sample without dsDNA (negative). The blue light is from the blacklight (b) Fluorescence spectrum at 480 nm excitation of dye solution in the absence and presence of dsDNA and PNAs. The insets show zoom-ins of each spectrum’s peak for clarity. (c) Square wave voltammogram between positive and negative samples.

Full size image

Electrochemical detection has also been evaluated on the same paper-based sensor device, which contains the screen-printed electrodes with both positive and negative samples. Square wave voltammetry was used due to its inherent ability to reduce non-faradaic current, which leads to a decrease in background signal, enhancing sensitivity36. Figure 1c shows that the negative sample gave a very low current signal while the positive sample exhibited a very strong signal. This is due to the different amounts of dyes being trapped on the sensor surface, as previously explained in the fluorescence detection. The PY2+(C4) dye is a redox-active molecule, and the negative current, which peaked around 0.4 V, indicates that the electrochemical reduction of the immobilized PY2+(C4) had occurred. Acidic conditions were required for electrochemical detection to denature dsDNA, which leads to the release of the dye into the solution to accept the electrons at the electrode’s surface more efficiently than when the dye is bound to the dsDNA37.

Characterization of the acpcPNA-immobilized cellulose paper

The fabricated sensor was characterized by X-ray photoelectron spectroscopy (XPS) at different stages. Figure 2a shows C1s and O1s spectra of the unmodified paper. There are 4 peaks for C1s at the following binding energies: 284.804, 285.792, 287.029, 289.523 eV, which can be attributed to C–C/C–H, C–O/C-O–H, β-glycosidic bond (C–O–C), and O–C–O/C = O bonds, respectively. The O1s contains 3 peaks at 531.906, 533.082, and 534.384 eV for C=O, C–O–C, and C–O–H, respectively38,39,40,41. All bonds can be found in cellulose molecules.

Fig. 2
figure 2

XPS spectra of the paper surface at various stages of preparation, including (a) unmodified paper (b) aldehyde-functionalized paper (c) PNA-immobilized paper, and (d) C = O : C–O–C XPS peak area ratio graph to illustrate the change in C = O of the sensor at each stage of preparation.

Full size image

The NaIO4 oxidized paper (Fig. 2b) contains the same number of peaks at similar binding energies. The C1s contain 284.831, 286.513, 288.017, 290.147 eV for C–C/C–H, C–O/C–O–H, β-glycosidic bond (C–O–C), O–C–O/C=O bonds while the peaks for O1s are 532.070, 533.178, 534.482 eV for C=O, C–O–C, C–O–H, respectively. Though the peak corresponding to C=O for O1s (532.070 eV) is more prominent than that of unmodified paper, confirming the successful installation of the aldehyde groups by the oxidation process42.

The PNA immobilized paper (Fig. 2c) showed similar peaks as the oxidized paper, but with the addition of nitrogen bonds from the PNA probe. The C1s peaks are 284.600, 285.912, 286.852, 288.488, 290.278 eV for C–C/C–H, C–O/C–O–H, β-glycosidic bond (C–O–C)/C-N, C=N, O–C–O/C=O bonds, respectively. O1s are 531.908, 533.219, and 534.580 eV for C=O, C–O–C, and C–O–H, respectively43. Similarly, the peaks’ intensity that corresponds with C=O is reduced due to the reductive amination between NH2 at the end of PNA2’s tail and the C=O consumes the C=O. This reaction bonded the PNA2 with the paper substrate covalently44. The ratio between C=O and C–O–C peak area for each paper was calculated and shown in Fig. 2d to illustrate the change in C=O peak as discussed. There is an N1s peak at 399.299 eV, which corresponds to the presence of C-N bonds45.

FTIR and SEM had also been used to characterize the fabricated sensor as well. The results and accompanying discussions can be found in the SI section 3 and Fig. S2 and Fig. S3.

Based on these characterizations, it can be certain that the PNA probe was successfully immobilized on the paper surface. This accentuates the advantage of paper devices since there is no known simple method to covalently immobilize the PNA probes on microtubes or microcells.

Parameters optimization

The full discussion of parameter optimization can be found in the SI section 4. In brief, the following parameters were optimized (in that order): washing solution, PNA probe concentration, and hybridization time for fluorescence-based detection. Tween 20 (0.5%) in 0.01 M PBS solution (PBST) and methanol were evaluated as washing reagents. Figure 3a shows that methanol gives the highest Δred intensity for sensors with synthetic dsDNA, and thus it was chosen as the washing reagent. Next, 40% to 100% methanol was then prepared by dilution with milli-Q water to be evaluated as the washing reagent as well, and Fig. 3b shows that undiluted methanol gave the highest Δred intensity and thus it was chosen as the washing reagent in this work. However, this means that extra care must be exercised during the washing step to remove the dye from the device, as methanol can penetrate the wax layer if left in contact with the sensor for too long. Then, the PNA concentrations of the probe to be mixed with the DNA in solution (PNA1) and the probe to be immobilized on paper (PNA2) from 2 to 8 µM were evaluated. Figure 3c shows that the Δred intensity increased steadily from 2 to 4 µM and only slightly changed at 8 µM PNA, and thus, 4 µM PNA concentration was chosen for this work. Finally, the hybridization time between PNA1, dsDNA, and the immobilized PNA2 was evaluated from 10 to 120 min. Figure 3d shows that 60 min of incubation gives the highest Δred intensity and thus was selected for subsequent experiments. Although double duplex invasion takes more time than normal hybridization between probe and ssDNA, the reduction of steps and minimizing the use of heating sources are advantageous, and multiple tests can be performed in multiple devices that are easily mass-produced for high-throughput screening.

Fig. 3
figure 3

Δred intensity influenced by (a) washing reagent, (b) methanol to ultrapure water percent ratio, (c) PNA1 and PNA2 concentration, (d) PNA1/dsDNA/PNA2 incubation time, (e) electrochemical detection step potential, (f) amplitude, and (g) frequency. The dsDNA concentration was 10 µM for experiments (c) and 4 µM for other experiments. The green bars indicate conditions chosen for subsequent experiments.

Full size image

Electrochemical parameters, including step potential, amplitude, and frequency, were also optimized in that order, and it was revealed that 0.015 V, 0.04 V, and 30 Hz, respectively, were found to be optimal conditions as shown in Fig. 3e-g.

Performance evaluation with synthetic dsDNA

The performance of this proposed detection platform in the quantitative detection of synthetic linear 53 bps dsDNA was evaluated over the concentration range from 0.01 µM to 4 µM. This experiment was conducted before using LAMP-amplified plasmids to validate the sensor without complications from any potential interference, such as buffers and enzymes. As shown in Fig. 4, a linear relationship between the Δred intensity and the logarithmic dsDNA concentration was observed with the linear regression equation Δred intensity = 7.1424 log10(DNA concentration) − 4.9621 and R2 = 0.9992. The limit of detection (LOD, 3 × SD/slope) and limit of quantification (LOQ, 10 × SD/slope) were calculated as 1.1 nM and 1.7 nM, respectively. The LOD was in the same order as the solution phase detection of DNA by PY2+(C4)20. In this case, the washing step to remove the unbound PY2+(C4) could reduce the background signal further when compared to the solution phase, in which the unbound PY2+(C4) cannot be removed.

Fig. 4
figure 4

Representative image of fluorescence detection steps and red channel images (top) and analytical curve (bottom) of various synthetic linear 53 bps dsDNA concentrations converted to log base 10.

Full size image

Selectivity, reproducibility, and stability test

The selectivity was evaluated using three additional synthetic linear 53 bps dsDNA oligonucleotides corresponding to the gene sequences of HPV types 1846, 3147, and 3348 in a similar region (between E7 and E1 genes) to the target HPV16 dsDNA49. All HPV types used in this test are also classified as oncogenic subtypes. This region was chosen following the previous research by Kumvongpin et al.50 which was reported to show the best performance and specificity for LAMP-based detection of HPV16. The PNA probe sequence is then selected from the region that would not overlap with the primers’ binding sites. All primer and probe sequences were subjected to NCBI BLAST to ensure specificity towards HPV16 detection. Figure 5a shows that this sensor is selective towards HPV16 dsDNA even at high concentration (4 µM) of other DNA sequences because the probes were designed to target HPV16 thus demonstrating the high specificity of the acpcPNA probe which is generally capable of discriminating similar DNA targets to a single base mismatch level51. It should be noted that real sample DNA would need to be amplified by LAMP in the final application, which is an additional layer of sequence discrimination of DNA as well.

Fig. 5
figure 5

(a) Selectivity of the sensor towards 4 µM linear dsDNA of various HPV subtypes. (b) Reproducibility of sensors by repeating the same procedure on 10 sensors. (c) Storage stability of the sensors from day 1 to day 21.

Full size image

Reproducibility tests were also conducted by measuring the red intensity of 10 sensors tested with samples containing the same 4 µM target dsDNA. In Fig. 5b, the raw red intensity values were reported instead of the Δred intensity to demonstrate the consistency of the signal obtained from different pieces of sensor under controlled conditions. The results show little variation of red intensity values among different sensors at %RSD = 1.82, which suggests good reproducibility of the device fabrication. The %RSD of electrochemical detection was found to be 4.43, which is in the acceptable range according to the AOAC method.

Stability tests were conducted by testing the sensors immobilized with 4 µM PNA2 after storage at 4 °C for up to 21 days. Figure 5c shows little variation over the course of three weeks, indicating decent sensor shelf life owing to the high chemical and biological stability of acpcPNA18. The changes in Δred intensity obtained from the aged sensors were within ± 5% compared to the first day. A long shelf life would be greatly beneficial to logistics since the sensors can be mass-produced, stored, and distributed to remote locations without timing constraints.

Analytical performance of amplified sequence-inserted plasmid dsDNA by fluorescence

Loop-mediated isothermal amplification (LAMP) is one of the DNA amplification methods that does not require thermal cycling52. Only simple water baths or heating blocks can be used for the sample preparation, which would significantly simplify and lower the cost of the sample preparation compared to PCR53. Before testing the sensor with the real clinical samples, the analytical curve was generated using pUC57 plasmid with 378 bps HPV16 inserted at the concentration from 102 to 108 copies/µL. The plasmids were amplified by LAMP, and the detection was performed. The LAMP experiment was shown to be successful by comparing the fluorescence of the two solutions containing the same 108 copies of plasmid DNA, LAMP primers, and enzyme, and PY2+(C4) dye, where one was subjected to heating at 65 °C for 30 min while the other was left at room temperature. As shown in Fig. S4, high fluorescence intensity was observed in the heated mixture due to the increase in DNA concentration from successful LAMP amplification, while the other did not fluoresce because of the low concentration of the plasmid DNA.

As in the linear dsDNA detection, a similar linear relationship between Δred intensity and logarithmic concentration of the plasmid DNA with the linear regression equation Δred intensity = 3.1679 log10(DNA concentration) + 0.0148 and R2 = 0.9981 was observed as seen in Fig. 6a. The LOD (3 × SD/slope) and LOQ (10 × SD/slope) were calculated as 4.7 copies/µL and 17.8 copies/µL, respectively. This shows that PNA, PY2+(C4), and double duplex invasion by acpcPNA could tolerate biomolecules, salts, and surfactants presented in the LAMP amplified DNA samples.

Fig. 6
figure 6

Representative image of fluorescence and electrochemical detection steps and (a) Red channel images (top) and analytical curve (bottom) of various plasmid dsDNA concentrations subjected to LAMP converted to log base 10. (b) Square wave voltammogram (top) and analytical curve (bottom) from electrochemical detection of various plasmid dsDNA concentrations subjected to LAMP converted to log base 10.

Full size image

Analytical performance of amplified sequence-inserted plasmid dsDNA by square wave voltammetry

Electrochemical detection has also been applied for the detection of LAMP-amplified plasmid dsDNA as well. Similar to the fluorescence detection, a linear relationship between the current readout and logarithmic plasmid dsDNA concentration was also obtained with a linear regression equation Δcurrent = 4.6234 log10(DNA concentration) + 7.6229 and R2 = 0.9985, as shown in Fig. 6b. The LOD (LOD, 3 × SD/slope) and LOQ (LOQ, 10 × SD/slope) were calculated as 1.0 copies/µL and 4.9 copies/µL, respectively. The LOD and LOQ are lower than those obtained from the fluorescence detection, as well as giving a broader linearity range from 100 to 108 copies/µL. This is due to the high sensitivity inherent to electrochemical detection. The response curves before taking the logarithm base 10 on the concentration for both fluorescence and electrochemical detection have been included in Fig. S5 and Fig. S6, respectively. The comparison of sensing performance between the proposed paper-based sensor and previous work based on the detection of HPV16 dsDNA was also discussed in the SI, section 5 (Table S1).

Real HPV16 sample detection

Cervical swab samples from Phramongkutklao Hospital were used for the performance evaluation of the developed dual-mode sensor for real sample detection. The calibration curves from Fig. 6 were used for the quantification of the DNA by fluorescence and electrochemical detection, respectively. The presence of HPV16 DNA in each sample was confirmed using PCR followed by gel electrophoresis. The detection of dsDNA extracted from clinical samples in this work was carried out following the same procedure for LAMP and detection without denaturation or additional sample preparation steps. The results were compared with the aforementioned gel electrophoresis method.

Table 1 shows the comparative results between the gel electrophoresis method and this work. HPV16 positive (+) samples as defined by those with higher Δred intensity or Δcurrent than their respective LOD from the analytical curve of LAMP-amplified plasmids. The results from both fluorescent and electrochemical methods were in complete agreement with the results from gel electrophoresis; thus, the first fluorescence/electrochemical dual-sensor utilizing acpcPNA probes for the detection of dsDNA by double-duplex invasion has been successfully validated.

Table 1 Detection of HPV16 dsDNA from clinical cervical swabs.
Full size table

Conclusion

The first paper-based fluorescence/electrochemical dual sensor for the direct detection of synthetic and LAMP-amplified dsDNA is reported. The developed sensor features the use of a pair of inherently pseudocomplementary acpcPNA probes to capture the DNA target by double duplex invasion, followed by staining with a fluorescence and redox-active PY2+(C4) styryl dye. The proposed sensor exhibited good sensitivity, especially with electrochemical detection. It also showed good selectivity towards HPV16 dsDNA detection as well as high stability and reproducibility. The new sensor is competitive with previous works in terms of sensitivity while being simple to use, highly selective, and less prone to false positives from non-specific amplifications, and no extra enzymes or additional components are required (e.g., CRISPR/Cas, RNA guide). The elimination of the dsDNA denaturation step greatly simplified the operation. The unique inability to form self-pairing hybrids of acpcPNA means that no further modification of the PNA probe or the base is necessary. In addition, the label-free system requires no complicated labeling of the probes or targets. The use of isothermal amplification, a smartphone, and an inexpensive blacklight source, or a simple potentiostat, means the whole system is easy to construct, portable, and does not require complex instrumentation. The dual detection modes (fluorescence and electrochemical detection) offer flexibility to the user to choose, depending on the available resources and purpose. Fluorescence is sensitive enough and cheap. On the other hand, electrochemical detection requires more sophisticated instrument setups but also offers better sensitivity. Both methods could be performed on the same sample in the same device for self-validation. The detection of HPV16 dsDNA from clinical cervical swabs showed results that are in good agreement with the gel electrophoresis results, suggesting the sensor’s suitability for real-world applications that could be readily adapted for other problems. The sensor operation, however, still requires multiple operations, including dye addition and a manual washing step. For future development, the sensor device would benefit from new techniques that allow simpler operation and/or automation.

Methods

Chemicals and materials

The detailed information about chemicals and materials used in this work was provided in the supplementary information (SI), section 6. Oligonucleotide and primer sequences are provided in Tables S2 and S3, respectively. pUC57 plasmid vector for analytical performance experiments was inserted with a region of the HPV16 sequence shown in the SI section 7. Clinical cervical swabs for real sample tests were kindly provided by Dr. Nattapong Sreamsukcharoenchai (Obstetrics and Gynecology Division, Phramongkutklao Hospital, Thailand), with anonymization before accession by the researchers to ensure patient confidentiality and was approved by The Research Ethics Review Committee for Research Involving Human Research Participants, Group I, Chulalongkorn University (COA No. 125/66). All methods were carried out in accordance with guidelines and regulations established by Chulalongkorn University and the Institute for the Development of Human Research Protections (IHRP, Department of Medical Science, Thailand). Written informed consent was obtained from all subjects.

Apparatus

The list of apparatus employed in fabrication of paper-based device, characterization, and detection of HPV16 dsDNA was described in SI section 8.

Preparation of paper-based sensor

The paper-based sensor was custom-designed to perform both fluorescence and electrochemical detections in a single device using the Adobe Illustrator program. The device was designed to be foldable, with each side having a dimension of 20 × 15 mm. The design was then printed on Whatman grade 4 filter paper using the wax printer. The paper was made to be folded in half. One half was screen-printed with a graphene counter electrode and silver/silver chloride reference electrode, and punctured in the middle, while the other half was screen-printed with a graphene working electrode with a diameter of 3 mm. The detection zone was located behind the working electrode. Figure 7a–b shows diagrams for both sides of the paper-based sensor.

Fig. 7
figure 7

(a) Diagram of the assembled sensor (b) Diagram of the flipped sensor (c) Procedure of direct HPV16 dsDNA fluorescence detection by acpcPNA probes and PY2+(C4) dye on the paper platform.

Full size image

Immobilization of one of the acpcPNA strands (designated as PNA2) was performed according to the method of Teengam et al., with some modifications45. Briefly, 10 µL of 2.1 M LiCl in 0.04 M NaIO4 aqueous solution was pipetted onto the paper’s detection zone and left in a dark and humid chamber for 15 min. The detection zone was washed twice with ultrapure water and dried at room temperature. Next, 3 µL of 4 µM PNA2 in DMF containing NaBH3CN (1 mg/mL) was pipetted onto the aldehyde-functionalized detection zone and left in a dark and humid chamber overnight. The PNA-immobilized paper was then washed with 1:1 (v/v) acetonitrile:water twice and dried at room temperature. Finally, the sensor was stored in a refrigerator at 4 °C until use.

Detection of HPV16 dsDNA by fluorescence and electrochemical readout

A mixture including various concentrations of target HPV16 dsDNA and 4 µM of another acpcPNA strand complementary to PNA2 (designated as PNA1) was pipetted on the PNA2-immobilized paper sensor, followed by incubation at room temperature for one hour to allow the double duplex invasion of dsDNA by PNA1 and PNA2. Next, 3 µL of 40 µM PY2+(C4) dye in 0.01 M PBS solution was pipetted onto the sensor, followed by 5 min of incubation at room temperature. The sensor was washed with 10 µL of methanol, followed by 10 µL of ultrapure water, to eliminate unbound dsDNA and PY2+(C4), and then left to dry for 30 min. For fluorescence detection, the sensor was placed under a blacklight lamp inside a box then the image was taken using a smartphone. Blank samples were prepared without the dsDNA target to measure the background signals. ImageJ software was used to evaluate the intensity from the red channel of the images. The red intensity of the sample was subtracted from the intensity of the blank sample to obtain the Δred intensity. In the case of electrochemical detection, the paper-based sensor was folded in a way that all electrodes faced outward. Then, 1 M HCl was pipetted into the puncture hole to ensure appropriate pH for the electrochemical reaction. The potentiostat was then connected to the electrodes, and square wave voltammetry was performed to observe the current readout. The current obtained from blank samples was subtracted from the measured current from samples to give the Δcurrent. All results are based on triplicate experiments. A simplified diagram showing the detection procedure is shown in Fig. 7c.

Preparation of LAMP for real sample plasmid dsDNA analysis

LAMP primer sequences were referenced from Kumvongpin et al. and shown in Table S250. The LAMP condition was summarized in the SI section 9.

DNA extraction of clinical cervical swabs and gel electrophoresis test.

The real sample DNA extraction and detection using gel electrophoresis were described in the SI section 10.

Statistical analysis

Graphs, regression equations, R-squares, means, and standard deviations were calculated by Microsoft Excel with error bars representing standard deviation. Data points represent a sample size of 3 (n = 3).