Ultrasensitive and label-free biosensor for the detection of Plasmodium falciparum histidine-rich protein II in saliva

Malaria elimination is a global public health priority. To fulfil the demands of elimination diagnostics, we have developed an interdigitated electrode sensor platform targeting the Plasmodium falciparum Histidine Rich Protein 2 (PfHRP2) protein in saliva samples. A protocol for frequency-specific PfHRP2 detection in phosphate buffered saline was developed, yielding a sensitivity of 2.5 pg/mL based on change in impedance magnitude of the sensor. This protocol was adapted and optimized for use in saliva with a sensitivity of 25 pg/mL based on change in resistance. Further validation demonstrated detection in saliva spiked with PfHRP2 from clinical isolates in 8 of 11 samples. With a turnaround time of ~2 hours, the label-free platform based on impedance sensors has the potential for miniaturization into a point-of-care diagnostic device for malaria elimination.

While PfHRP2 ELISAs [16][17][18] , PCR 19,20 and loop-mediated isothermal amplification (LAMP) 21 assays have shown superior sensitivity to RDTs for detection of low-density infections, the platforms are slow and technically complex assays typically performed by highly-skilled technicians in centralized laboratories. Implementation of these methods is therefore impractical in the low-resource settings of many countries striving for malaria elimination.
Additionally, detection of PfHRP2 in saliva is gaining substantial interest due to ease of collection, lower biohazard risk, and higher likelihood of testing compliance particularly in communities where blood collection poses cultural objections 22,23 . Although salivary PfHRP2 has been evaluated for the detection of low to high-density P. falciparum parasitemias 10 no RDTs have been developed for this purpose, due to the lack of sensitivity. An in-house ELISA assay developed for detection of salivary PfHRP2 had a sensitivity of 0.17 pg/mL 7 , but the laboratory-based nature of the platform hinders effective field implementation.
Impedimetric biosensors are promising options to help close current diagnostic gaps, due to their high sensitivity, low cost, and amenability to miniaturization. Table 1 summarises previous reports on the development of ultrasensitive PfHRP2 biosensors. The biosensors were used in direct or sandwich immunoassay formats to target PfHRP2 protein in various sample matrices. To achieve high-level sensitivity, most of these sensors required additional labeling and signal amplification, which in the long run may incur additional costs during miniaturization.
In this study, we aimed to develop an interdigitated electrode (IDE) sensor for impedimetric detection of PfHRP2 at low concentrations suitable for malaria elimination, with a focus on the utilization of saliva as sample medium. Compared with other technologies, the IDE sensor geometry has demonstrated high level of sensitivity and specificity for label-free detection of various targets including nucleic acids [24][25][26][27] , cells [28][29][30] , and proteins [31][32][33] . Here, the detection platform utilized anti-PfHRP2 monoclonal antibodies (MAbs) immobilized on sensor surface as capture probes towards circulating PfHRP2 protein. The application of periodic small AC voltage allowed measurement of the sensor impedance, defined as a complex of the circuitry's resistance and capacitance. Specific detection of surface-bound PfHRP2 on the sensor was achievable through frequency-dependent characterization of the impedimetric changes. The platform demonstrates promising ultrasensitive detection of PfHRP2 protein in saliva. Fabricated using low-cost techniques, the platform is amenable to future automatization and miniaturization for point-of-care applications.

Results
Sensor preparation and platform development. IDE sensors were fabricated using UV-lithography on borosilicate glass wafers as previously described 27 , with modification. An additional SiO 2 layer was deposited at a thickness of 25 nm. Each sensor consisted of paired electrode arrays with finger length (L) of 980 µm, finger width (W) of 8 µm and gap width (G) of 8 µm (Fig. 1a). The sensors were functionalized according to our established protocol 30,34 by covalent immobilization of anti-PfHRP2 antibodies on the surface, followed by blocking to reduce non-specific binding to the antibodies. Optimization results of capture antibody concentration are provided in Supplementary Figure 1.
Impedance of the sensors was measured using our previously established circuit setup 30,34 shown in Fig. 1b, in which the sensor is represented as a resistor (R) in series with a capacitor (C) connected in series with a 1 kΩ reference resistor (R ref ). A function generator provided the input sinusoidal AC excitatory signal (V in ) at 20 mV www.nature.com/scientificreports www.nature.com/scientificreports/ peak-to-peak voltage (V pp ) at multiple frequencies (1,10, 20, and 40 kHz). Sensors were measured in wet-state (in PBS 1×) at time points T1 (baseline measurement obtained before sample incubation) and T2 (after sample incubation and washing) to detect binding associated change occurring between the two points. A lock-in amplifier recorded the amplitude of the output voltage (V out ) and phase across the R ref , which was used for the acquisition of frequency (ω) dependent impedimetric parameters including impedance magnitude (Z), capacitance (C), and resistance (R) using the following equations 34 Development of the sensor platform was divided into two phases (Fig. 1c). In Phase I, phosphate buffered saline (PBS) samples spiked with PfHRP2 were used. The optimal detection frequency was determined, and parameters for specific PfHRP2 detection were established and used to obtain the analytical sensitivity. The protocol was then applied in Phase II in which saliva samples spiked with PfHRP2 derived from cultured cell lines were used. Optimal parameters for saliva-based detection were determined to obtain the sensitivity. Specificity was then assessed using a panel of saliva samples spiked with isolate-derived PfHRP2.
Phase I -Detection of PfHRP2 protein in PBS buffer. Among the range of excitation frequencies applied (1 kHz to 40 kHz), 20 kHz was found to elicit the most significant difference in the change in impedance magnitude between PBS samples with and without spiked PfHRP2 (Fig. 2a). This frequency was used for all subsequent experiments. The impedimetric parameters (impedance magnitude, impedance phase, resistance and capacitance) were also compared in their ability to differentiate PBS samples with and without spiked PfHRP2, and the change in impedance magnitude was found to be the optimal parameter for this purpose (Fig. 2b).
The finding that frequency-dependent characterization of PfHRP2 binding is optimal within the frequency range of 10-100 kHz is consistent with similar previous studies 30, 35 , and supports the notion that the signal www.nature.com/scientificreports www.nature.com/scientificreports/ is primarily due to the changes in bulk medium conductivity as a result of dipole accumulation 36 from the surface-bound PfHRP2.
The optimal excitation frequency and impedimetric parameter were then used to determine the limit of PfHRP2 detection in PBS medium. The limit of detection, defined as the concentration at which the test sample exhibited a signal change significantly distinct from the blank, was found to be 2.5 pg/mL (Fig. 2c). Starting at that PfHRP2 concentration, a significant increase in the impedance signal was observed, indicating the sensor response towards the solid-state capture. No signs of saturation or prozone effect were observed at the higher end of the tested concentrations. This detection limit was concordant with our previous finding that the IDE sensors were able to detect another protein at the minimum concentration of 2.9 pg/mL 33 .
The Phase I results provided proof-of-concept data for ultrasensitive PfHRP2 protein detection using the label-free electrical biosensor platform. The detection limit was 3 logs lower than that of the average RDTs (~ 4 ng/mL) 13 , and is also lower than previously developed label-free electrical biosensor for PfHRP2, which reported a sensitivity of 12 ng/mL 37 .
Phase II -Detection of PfHRP2 protein in saliva. Next, the ability of the platform to detect culture-derived PfHRP2 in human saliva samples was evaluated to support future real-world applications. Saliva can be collected non-invasively, making it an ideal biospecimen for population screening. The protocol in Phase I was modified for use in saliva samples owing to differences in the ionic composition and protein content between PBS and saliva, and the need to reduce degradation of PfHRP2 by protease enzymes present in saliva 7 . The modification included more stringent blocking protocol (Supplementary Figure 2), longer incubation time of 2 hours (Supplementary Figure 3), sample pretreatment with protease inhibitor and re-characterization of the impedimetric detection parameters. The optimized parameters were then used to identify the limit of detecting PfHRP2 in saliva and to validate the protocol using saliva samples spiked with PfHRP2 expressed by different P. falciparum isolates.
Unlike findings in Phase I that used PBS, change in sensor resistance was found to be a more specific parameter in differentiating saliva samples with and without spiked PfHRP2 (Fig. 3b). The finding that optimal differentiation is parameter-specific is one of the advantages of impedance-based measurements, as it permits optimal detection despite differences in ionic composition across sample types.
The adapted parameters were used to determine the lowest limit of PfHRP2 detection, which was found to be 25 pg/mL (Fig. 3c). Starting at this concentration of PfHRP2, the change in sensor resistance was significantly different between the positive sample and the negative (unspiked) sample. In contrast to the PBS-based dynamic range assessment, signal saturation was observed earlier in the saliva concentration curve, at 2.5 ng/mL PfHRP2. This saturation, which resulted in a less optimal concentration-dependent response, is most likely due to the presence of interfering proteins in the saliva matrix compared to the PBS, which may have been the cause of reduction in binding magnitude. The presence of multitude of substances in saliva can result in failure of antibody and ligand complex formation on the sensor surface, which may contribute to the reduced dose-dependent effect.
To further validate the protocol for real-world implementation, 11 saliva samples spiked with PfHRP2 (samples P 1-9 are PfHRP2 present in supernatants of clinical parasite isolates, P 10-11 are PfHRP2 isolated from culture supernatants of 3D7 and CS2 laboratory adapted parasites lines) at a concentration of 25 pg/mL were incubated with the sensor platform. Among the 11 samples tested, 8 resulted in significantly higher change in resistance (%ΔR) compared to the un-spiked saliva sample and the sample spiked with Plasmodium lactate dehydrogenase (PLDH) (Fig. 4). The variations among the sensor signals from the different PfHRP2 samples might be attributed to the differences in PfHRP2 protein size since previous characterizations have shown the protein to be of multiple bands ranging from 50-80 kDa 7 , although this assumption will need to be assessed further.

Discussion
The sensor platform developed in this study represents a highly sensitive impedimetric sensor capable of label-free detection of surface-bound PfHRP2. Promising sensitivity was demonstrated with a PfHRP2 detection limit of 2.5 pg/mL in PBS, an order of 3 logs lower than current PfHRP2 RDTs. Furthermore, the study also demonstrated the feasibility of PfHRP2 detection in saliva with a detection limit of 25 pg/mL, in 8 out of 11 tested samples. Specific detection was achieved using an optimized, constant excitation frequency of 20 kHz across both PBS and saliva samples.
Differences in assay sensitivity are commonly observed in the development of PfHRP2 immunosensors as previous studies have also reported the variations in assay performance among various matrices 38,39 , with few studies examining saliva samples. Adaptation of the platform for other types of biospecimens, such as whole blood or plasma, would require re-optimization of impedimetric parameters, incubation time, and blocking protocol due to the differences in the ionic properties and protein content of different specimen types.
Saliva-based PfHRP2 detection is an attractive and potentially cost-effective testing approach. To date, quantitative data on salivary PfHRP2 in cases of low-density parasitemia is lacking, as previous salivary PfHRP2 quantification using commercial 10 or in-house developed 7 ELISAs has been limited to testing of symptomatic individuals. Although there is currently no consensus regarding the correlation between saliva PfHRP2 and parasitemia, the sensitivity shown by the platform is adaptable towards the range of salivary PfHRP2 detected using the laboratory-based ELISA (17-1167 pg/mL) 7 . To futher guide the development of salivary PfHRP2 diagnostics for elimination and screening purposes, more research is required to study the concentration of salivary PfHRP2 protein in asymptomatic versus symptomatic individuals, and in low versus high density parasitemias. Additionally, more work is required on assessing the effect of patient conditions such as dehydration or hormonal fluctuations on the sensor readings, as minor physiological fluctuations can affect the saliva sample matrix.
Future work on the platform will focus on the miniaturization of sensors and integration of the platform with microfluidics. The lock-in amplifier readout technique has good miniaturization potential, with setup possible (2019) 9:17495 | https://doi.org/10.1038/s41598-019-53852-5 www.nature.com/scientificreports www.nature.com/scientificreports/ within small-scale portable devices 40,41 . Miniaturization and integration of the sensor platform has the potential to (1) reduce variation due to automatization of the electrical reading and washing steps, (2) reduce the required incubation time by avoiding the diffusion limited process of static incubation, and (3) improve efficiency due to reduction in the reagent and sample volume. Additionally, future work can also be geared towards adaptation of the IDE sensor platform to detect other proteins relevant for malaria diagnostics, including PLDH, and aldolase.
Ultrasensitive diagnostics are considered to play an important role in the global effort towards malaria eradication. Efforts continue to strive for a balance between assay performance, cost-effectiveness, and practicality. We have developed the PfHRP2 IDE sensor platform with these factors in mind. Our study has provided proof of concept that the platform may be a potential technology to help achieve this goal.

Methods
Sample preparation. In Phase I, recombinant PfHRP2 protein (CTK Biotech, California, USA) was suspended in PBS 1× buffer. Antigen quantification was performed using commercial PfHRP2 ELISA (Cellabs Pty. Ltd., Brookvale, New South Wales, Australia). An additional negative control was prepared by suspending 2.5 pg of PLDH protein (CTK Biotech, California, USA) in 1× buffer. Change in sensor resistance is the optimal impedimetric parameter for detection of PfHRP2 (25 pg/mL) protein in saliva. Red bar = sensors incubated in unspiked saliva, blue bar = sensors incubated in 2.5 pg PfHRP2 per mL of saliva. Bars represent mean ± SEM, n = 7 sensors per group, p-values calculated with Welch's two-tailed t-test. (c) Sensor response towards various concentrations of PfHRP2 protein in saliva. A serial dilution of 2.5 pg/mL to 2.5 ng/mL PfHRP2 protein in saliva was measured. Bars represent mean ± SEM, n = 7-8 sensors per concentration. Statistical analysis by one-way ANOVA followed by Dunnett's multiple comparisons test with blank saliva used as the comparator. *p ≤ 0.05, **p ≤ 0.01, ns = not significant. All incubations were performed for 2 hours within a parafilm-wrapped petri dish to avoid sample evaporation, and all measurements were performed at 20 kHz frequency. All specimens were cultured for 36 hours, to obtain samples at 6% parasitemia at mature trophozoite stage. Spent culture medium supernatants were collected. Control medium was prepared similarly by incubating medium with uninfected erythrocytes. Supernatants were stored at −80 °C and used to quantify PfHRP2 by ELISA, and to detect HRP2 using the sensors. Control medium was used to spike negative saliva controls.
To prepare spiked saliva samples, unstimulated fresh saliva was collected and subjected to mechanical filtration to remove residues and mucus (Corning ® 0.2 µm filter). Protease inhibitor 100× (P8340 Protease Inhibitor Cocktail, Sigma Aldrich, Missouri, USA) was added immediately followed by the culture-harvested PfHRP2 antigens. The PfHRP2 samples (n = 11) were diluted using saliva to the final concentration of 25 pg/mL. Negative control blank samples (n = 11) were prepared by diluting the control medium in saliva to match the dilution of each PfHRP2 sample. An additional negative sample was prepared using PLDH protein spiked saliva at 25 pg/mL. Spiked samples were immediately added to sensor after baseline electrical readings.

Sensor fabrication.
Fabrication was performed at the Melbourne Centre for Nanofabrication (MCN).
Wafers of BOROFLOAT ® borosilicate glass (UniversityWafer, Massachusetts, USA) were cleaned with isopropyl alcohol, dried, and then coated with hexamethyldisilazane (MicroChemicals GmbH, Ulm, Germany) and AZ1512HS (MicroChemicals GmbH, Ulm, Germany) photoresist. A chrome mask of the sensor design was applied on the substrate followed by UV exposure (75 mJ/cm 2 ). After development, a thin film of chrome (5 nm), gold (100 nm), and titanium (5 nm) was deposited on the substrate, followed by a lift-off process to reveal the IDE pattern. This was followed by an addition of a thin layer (25 nm) of SiO 2 using e-beam evaporation (Intlvac Nanochrome TM II, Colorado, USA).   www.nature.com/scientificreports www.nature.com/scientificreports/ impurity. Following functionalization, a 2 mm diameter incubation area was established using medical grade pressure sensitive adhesive tape (Adhesive Research, Pennsylvania, USA).

Sensor functionalization and antibody immobilization.
Anti-PfHRP2 IgG MAb receptor (AB-0445, Vista Diagnostics International, Washington, USA) at a volume of 15 µl (50 µg/mL) was applied on the modified sensing area and incubated in a humid chamber at 4 °C overnight. Sensors were then washed with PBS 1× for 5 min and blocked for 30 min using either 5% ethanolamine (Sigma Aldrich, Missouri, USA) for Phase I or 5% ethanolamine and 2.5% normal goat serum mix (Sigma Aldrich, Missouri USA) for Phase II.
Data processing and statistics for PfHRP2 detection. Acquisition of impedance properties from the lock-in amplifier V out and phase output was calculated based on Eqs. 1 and 2 using MATLAB. The baseline measurement obtained before sample incubation (T1) were first evaluated for assessment of sensor quality, then T1 and T2 (after sample incubation and washing) values for each parameter were processed in Microsoft Excel to obtain the percentage changes in impedance magnitude (%∆Z = ABS Z T2 /Z T1 ), impedance phase (%∆Z θ ), resistance (%∆R), and capacitance (%∆C). The absolute percentage change in impedance (|%∆Z|), resistance (|%∆R|), and capacitance (|%∆C|) were then used to assess the sensor performance statistically.
GraphPad PRISM 7 was used to generate all plots and perform statistical analysis, and p-values ≤ 0.05 were considered statistically significant. To implement quality control on the sensors, a baseline outlier test was performed using the regression and outlier test (ROUT) 42 on all derived T1 values, with the maximum false discovery rate (Q) set to 1%. An example of the PRISM ROUT test performance can be seen in Supplementary Table 1 and Supplementary Figure 4. Sensors with baselines excluded using the ROUT test were not included in the rest of the analysis.
In Phase I, Welch's two-tailed t-test was used to compare the percentage change in impedance magnitude (%∆Z) between blank and test sensors. A serial dilution of PfHRP2 in PBS 1× buffer was incubated on the sensors, and the %∆Z obtained was assessed using one-way ANOVA with Dunnett's multiple comparisons test. Detection limit is defined as the lowest tested concentration showing statistically significant difference from blank sensor reading.
In Phase II, Welch's two tailed t-test was used to determine the optimal sample pretreatment and detection parameter in saliva. The optimized parameters were used to determine the detection limit in the same manner as in Phase I, using the optimized parameter for saliva (%∆R). Platform performance was then assessed in a panel of PfHRP2-spiked saliva using the Dunnett's multiple comparisons test to determine the degree of differentiation against the un-spiked saliva control.

Data availability
Relevant data are available from the authors on request.

code availability
The codes used for impedimetric calculations are available on request from the authors.