Electrochemical affinity biosensors for fast detection of gene-specific methylations with no need for bisulfite and amplification treatments

This paper describes two different electrochemical affinity biosensing approaches for the simple, fast and bisulfite and PCR-free quantification of 5-methylated cytosines (5-mC) in DNA using the anti-5-mC antibody as biorecognition element. One of the biosensing approaches used the anti-5-mC as capture bioreceptor and a sandwich type immunoassay, while the other one involved the use of a specific DNA probe and the anti-5-mC as a detector bioreceptor of the captured methylated DNA. Both strategies, named for simplicity in the text as immunosensor and DNA sensor, respectively, were implemented on the surface of magnetic microparticles and the transduction was accomplished by amperometry at screen-printed carbon electrodes by means of the hydrogen peroxide/hydroquinone system. The resulting amperometric biosensors demonstrated reproducibility throughout the entire protocol, sensitive determination with no need for using amplification strategies, and competitiveness with the conventional enzyme-linked immunosorbent assay methodology and the few electrochemical biosensors reported so far in terms of simplicity, sensitivity and assay time. The DNA sensor exhibited higher sensitivity and allowed the detection of the gene-specific methylations conversely to the immunosensor, which detected global DNA methylation. In addition, the DNA sensor demonstrated successful applicability for 1 h-analysis of specific methylation in two relevant tumor suppressor genes in spiked biological fluids and in genomic DNA extracted from human glioblastoma cells.


Results
Two different electrochemical biosensing strategies for evaluating DNA methylation in a direct and independent PCR mode are described in this work. Specific promoter sequences of RASSF1A and MGMT genes were selected to verify the suitability of the designed strategies. The RAS association (RalGDS/AF-6) domain family member 1A RASSF1A gene is a tumor suppressor gene frequently detected to undergo epigenetic silencing by aberrant hypermethylation of its promoter region in many human solid tumors. It has been reported to have clinical sensitivity of 67-75% for stage IV breast cancer 2,41,42 . On the other hand, methylation of the O6-methylguanine-DNA methyltransferase (MGMT) of the MGMT promoter of malignant glioma appears to be a useful predictor of the responsiveness to alkylating agents that reverse epigenetic alterations. Numerous studies have demonstrated that patients with silencing MGMT respond better to therapy 13,43 .
The first strategy (Fig. 1a) is an immunosensor involving the use of two different antibodies. The one immobilized on the surface of carboxylic acid-modified magnetic beads (HOOC-MBs), is specific to 5-methylcytosine (anti-5-mC) and, therefore, capable of capturing any ss-DNA sequence bearing this type of methylation. A second antibody conjugated with peroxidase (HRP-anti-ssDNA), able to recognize any ss-DNA, was used as detector antibody.
The second method (Fig. 1b) is a DNA sensor involving immobilization of a biotinylated DNA capture probe, specific to the methylated sequence to be detected, on the surface of Streptavidin-modified MBs (Strep-MBs). Methylation in the captured target DNA was recognized by means of the specific anti-5-mC tagged with a secondary HRP-conjugated antibody (HRP-anti-mouse IgG). In both strategies, amperometric determination was carried out using the hydrogen peroxide/hydroquinone (H 2 O 2 /HQ) system after magnetic capture of the modified MBs on the surface of a screen-printed carbon electrode (SPCE), by measuring the cathodic current generated by the enzymatic reduction of H 2 O 2 mediated by HQ, this current being proportional to the concentration of methylated DNA in the sample.
Optimization of the experimental conditions. Because methylation frequently is present in a little subset of cells in a clinical specimen, the sensitivity is critical to design an analytical method for monitoring the methylation status of the target gene 11,44 . Therefore, seeking for a high sensitivity and for simple and short assay protocols, the relevant experimental variables involved in the biosensors preparation were optimized. The taken selection criterion was the largest ratio between the current values measured at a potential value of −0.20 V (vs. the Ag pseudo-reference electrode) using the H 2 O 2 /HQ system, in the absence (B) and in the presence of 5.0 nM of synthetic target RASSF1A (S). The evaluated variables, the tested ranges and the values selected for further work are summarized in Table 1. Other experimental variables not included in the Table, such as the MBs suspension volume 45,46 and the potential applied for the amperometric response 47 were optimized in previous works.
Illustrative examples of these optimization studies, such as the capture and detector antibodies loadings in the immunosensor and the biotinylated capture probe (bCp) concentration and number of steps for the DNA sensor, are displayed in Fig. 2a-d. Figure 2a and b show as the S/B ratio increased with the concentration of the capture and detector antibodies up to a certain concentration and then decreased significantly due to the sterically hindered antigen binding when high loadings of antibodies are immobilized 48 . The somewhat higher signal obtained in the absence than in the presence of target methylated DNA without anti-5-mC (Fig. 2a) may be attributed to a slightly higher non-specific adsorption of HRP-anti-ssDNA on HOOC-MBs in this case.
Regarding the optimization of the bCp concentration in the DNA sensor (Fig. 2c), while no significant differences were apparent between the B signals, the S signals increased significantly with the bCp concentration up to 0.1 μM, then decreasing for larger concentrations, which is most likely due to the restricted hybridization efficiency when large amounts of bCp are immobilized 49,50 . Results achieved in the optimization of all the other experimental variables are shown in Figure S1 (in the Supplementary Information).
Looking for a protocol simplification with reduced assay time, the influence of the number of steps used in the preparation of the DNA sensor was also investigated. All the evaluated protocols started after preparation of bCp-MBs and involved 30 min-incubation steps. The tested protocols consisted of: (a) incubation of bCp-MBs with a mixture solution containing the target DNA, the anti-5-mC and the HRP-anti-mouse IgG (1 step); (b) a former incubation with the target DNA solution and a second one with an anti-5-mC and the HRP-anti-mouse IgG mixture solution (2 steps); (c) independent and successive incubations with target DNA, anti-5-mC and HRP-anti-mouse IgG solutions (3 steps). Figure 2d shows as the working protocol involving two incubation steps provided the largest S/B current ratio. This protocol, additionally, allows reducing largely the total assay time. The observed results could be attributed to a less efficient hybridization of the target DNA labelled with both antibodies onto the bCp-MBs (1 step), and to a better recognition of the anti-5-mC by the HRP-anti-mouse IgG when both antibodies are free in solution. Consequently, a 2-step protocol involving the former hybridization of the target DNA onto the bCp-MBs and further labeling of the captured methylated DNA by incubation in an anti-5-mC and HRP-anti-mouse IgG mixture solution was selected for further work.
Analytical characteristics. The calibration plots and the analytical characteristics obtained with the two developed biosensing strategies for the methylated synthetic sequences of the promoter regions of the selected genes are displayed in Fig. 3 and summarized in   Table 2, the relative standard deviation (RSD) values demonstrated great reproducibility in both biosensors preparation protocols as well as in the amperometric transduction.
When we compared the sensitivity achieved with the developed immunosensor with a commercial ELISA methodology for the same methylated DNA standard (E. coli genomic DNA with 346,670 methylated cytosines), we observed that the immunosensor provided a 2,500 times lower LOD (0.004 vs. 10 ng) in a 4-times shorter assay time (45 min vs. 3 h). Also, the LOD achieved with the immunosensor is 4.7-times lower (6.8 vs. 32 pM) than that reported for an electrochemical immunosensor for 5-hmC methylation (determination of 5-hydroxymethyl-2′ -deoxycytidine-5′-triphosphate) 36 .
It is important to remark that both developed methodologies allowed methylated DNA to be determined at picomolar level in about 1 h, with no need for bisulfite and amplification pretreatments and excellent reproducibility throughout the entire protocols. However, the DNA sensor exhibits better sensitivity (see comparative amperograms in (Fig. 3b). Moreover, the existence of 5 methylated cytosines in the part of the sequence that remains unhybridized (the only ones recognized by the anti-5-mC) 14 in the synthetic target RASSF1A compared with the 4 in the MGMT, apart from the longest length of the hybrid fragment, explains the slightly higher sensitivity achieved for the determination of the target RASSF1A.  Interestingly, a comparison between the electrochemical methods reported so far for the determination of different types of targets (synthetic oligonucleotides, free DNA bases, PCR products and genomic DNA), demonstrates that the LODs achieved with the DNA sensor for the synthetic short target methylated DNAs are much lower than those reported with PCR amplification (25 pg) 11 , as well as for PCR-free methods using direct oxidation of DNA bases of short methylated oligonucleotides (0.11 μM) 35 and free un-methylated C (0.6 μM) 38 , digestion by a restriction enzyme of a synthetic target DNA (10 nM) 27 , bisulfite conversion of synthetic target DNAs (18 pM) 34 and PCR products (0.5 nM) 37 , and a paired-end tagging and amplification electrochemical strategy for methylated genomic DNA (40 pg) 23 . It is important to note also that the most sensitive strategies among these required assay times between 1.5 and 24.5 h 11,23,34 compared to the 1 h of the DNA sensors developed in this work.
It is worth to mention also that although the LODs achieved with the developed biosensors are higher than those claimed for other two reported affinity electrochemical biosensors for synthetic target methylated DNAs determination, 2 fM 4 and 35 fM 24 , the preparation of these biosensors required multiple reagents and complex and time-consuming working protocols that included amplification strategies involving nanomaterials. Moreover, in both cases the assay time is 3-5 times longer than that needed with the developed biosensors (2 h 40 min 4 and 4 h 50 min) 24 . It is important to highlight also that, apart from the shorter assay time and the inherent simplicity of the biosensors construction, these allow an accurate and straightforward determination of the synthetic target methylated DNA in spiked biological samples without previous extraction or amplification of the genetic material, as it will be shown in further on.
In addition, the storage stability of the anti-5-mC-MBs and the bCp-MBs employed for the preparation of the immuno-and DNA sensors, respectively, was evaluated by storing the modified MBs at 4 °C in microcentrifuge tubes containing 50 μL of filtered phosphate-buffered saline (PBS) or Binding and Washing buffer (B&W), respectively. Subsequently, they were used to prepare biosensors on different working days and to measure the amperometric responses for 0.0 and 0.5 nM of target RASSF1A or 1.0 nM of target MGMT solutions, respectively. No significant decrease in the measured S/B ratio was observed during at least 35 days in both cases (no longer times were assayed), suggesting the possibility of preparing the conjugated MBs in advance and storing them under the above-described conditions, until the biosensor preparation is required.
Selectivity. The selectivity of both biosensors towards 5-mC was evaluated by comparing the amperometric responses they provided for 1.0 ng of denatured synthetic DNA standards containing unmodified cytosines (C), 5-mC or 5-hmC. Results shown in Fig. 4 demonstrate the specificity of the developed methodologies to detect only 5-mC sequences. In addition, as predicted, the immunosensor detected the presence of any oligonucleotide with 5-mC methylation without any sequence selectivity. Accordingly, this sensor was able to quantify the total amount of 5-mCs in the analyzed DNA. Conversely, the DNA sensor only recognized the methylated sequence complementary to the bCp immobilized on the functionalized MBs and thus it can be used to simultaneously detect the presence of 5-mCs and their position in the DNA sequence.

Detection of 5-mC DNA methylation in biofluids and cells.
The evaluation of the usefulness of the developed methodologies was restricted to the DNA sensor because of its higher sensitivity and suitability for quantification of gene-specific methylation as compared with the immunosensor. Such evaluation was accomplished in biological fluids supplemented with synthetic methylated target RASSF1A.
The comparison of the amperometric responses obtained with the DNA sensor in the absence and in the presence of 5.0 nM (125 fmol) of the synthetic target RASSF1A prepared in buffered solution and in the different biological fluids tested 25%-diluted with 5-mC-ELISA buffer is shown in (Fig. 5a). The slope values corresponding to calibration plots prepared between 139 and 5,000 pM of the synthetic target RASSF1A in each media are summarized in Table 3. The comparison of the obtained slope values revealed the existence of matrix effects in serum and urine samples and, therefore, in these biological samples the determination should be carried out by interpolation in the representative calibration plots constructed in the 25%-diluted biological fluids instead of the calibration constructed for synthetic target in buffer (case of saliva samples). The results obtained in the recovery studies performed in these biological samples spiked with 2.5 nM of the synthetic target RASSF1A for a confidence level of 0.95 (summarized in Table 3) outline the reliability of the approach to determine a low concentration of the synthetic target in biological fluids just after a simple dilution and without prior DNA extraction and preconcentration. These results are considered particularly relevant taking into account that the commercial ELISA methods, as well as most of the biosensors described so far for methylation determination, have proved to be suitable only for the analysis of previously extracted DNA, but not directly in biological samples, and after applying bisulfite and/or amplification treatments.
It is important to note that most of the work reported in the literature only provides semi-quantitative data or comparative percentages of methylation in healthy and cancer patients. Therefore, it is extremely difficult to find any reference indicating the absolute value of the methylated target DNA concentration in liquid biopsies samples. Some indication about the concentration of circulating hypermethylated RASSF1A in serum from breast cancer patients, between 1 and 200 ng mL −1 , is given by Kristiansen et al. 42 , while the best cut-off for circulating serum RASSF1A to differentiate the hepatocellular carcinoma is 13 pM according to Mansour et al. 51 .
Despite the interesting results achieved in spiked serum, saliva and urine, it is important to mention that in most clinical applications DNA methylation detection at specific gene positions is much more relevant than quantification of DNA methylation. Therefore, in order to check the clinical applicability of the developed DNA sensor, the implemented methodology was applied to analyze the endogenous MGMT status in the reference cell line U87. Figure 5b shows that a larger amperometric response was measured when 100 ng genomic DNA extracted from these cells were analyzed, in comparison with the currents measured for both in the absence of target DNA and genomic DNA extracted from HeLa cells (non-methylated MGMT gene promoter cells used as control) 52 , which is consistent with the reported specific hypermethylation of the MGMT gene of U87 cell line 53,54 . All these results confirmed that the developed electrochemical DNA sensor exhibits suitable sensitivity and specificity for the determination in just 1 h of gene-specific methylations directly in biological fluids without previous DNA extraction and pretreatments (bisulfite and/or amplification), and in genomic DNA extracted from cells.

Discussion
This paper describes the development of two electrochemical biosensing strategies free of bisulfite and/or amplification pretreatments for the simple, sensitive and quick detection of DNA methylation using functionalized MBs, the anti-5-mC as affinity bioreceptor and amperometric detection at SPCEs using the H 2 O 2 /HQ system.  While the immunosensor uses the anti-5-mC immobilized onto HOOC-MBs as capture bioreceptor, the DNA sensor employs the anti-5-mC as detector bioreceptor to label the methylated DNA previously captured by a complementary DNA capture probe immobilized on the surface of Strep-MBs. Both sensors types imply simple and reproducible working protocols free of the complex and time consuming procedures required by the conventional methodologies. However, the DNA sensor exhibits a higher sensitivity and a clear discrimination only for the methylated sequence complementary to the immobilized capture probe. The obtained results demonstrated the suitability of the DNA sensor for detection of the gene-specific methylation without PCR amplification, bisulfite or labeling processes, directly in genomic DNA extracted from cells and in liquid biopsies without previous DNA extraction or amplification. In comparison with conventional ELISA available approaches, the developed sensors can greatly benefit research on DNA methylation in terms of convenience, low-cost, ease of operation and suitability for universal analysis of any DNA sequence. These sensors offer also very interesting features over other electrochemical biosensors described to date in terms of sensitivity and assay time. Moreover, the fact that they do not require complicated nanomaterials preparation, conventional methylated DNA pretreatments or sophisticated instruments for signal readout makes them economical and compatible with portable and low-cost devices to perform studies of DNA methylation-related functional genomics and epigenomics in different settings or at the point-of-need. These attractive capabilities can make these biosensing strategies to serve as a basis for developing easy and affordable tests in an easy-to-use format, not too technically demanding and requiring equipment readily available at most academic institutions for cancer suspects as well as for management and better outcome of cancer patients, complementing current and future pathological and molecular assessments.

Methods
Apparatus and electrodes. Amperometric measurements were performed with a CH Instruments (Austin, TX) model 812B potentiostat controlled by software CHI812B. SPCEs (DRP-110), consisting of a 4-mm diameter carbon working electrode, a carbon counter electrode and an Ag pseudo-reference electrode were used as electrochemical transducers, and the specific cable connector (DRP-CAC), which acted as interface between the SPCEs and the potentiostat, were purchased from DropSens (Spain). All the electrochemical measurements were performed at room temperature. A neodymium magnet (AIMAN GZ) embedded in a homemade Teflon casing was used for the reproducible magnetic capture of the modified-MBs on the surface of SPCEs.
A Bunsen AGT-9 Vortex for homogenization of the solutions, a Raypa steam sterilizer, a biological safety cabinet Telstar Biostar, a thermocycler (SensoQuest LabCycler, Progen Scientific Ltd.), an incubator shaker Optic Ivymen ® System (Comecta S. A, Sharlab) and a magnetic particle concentrator DynaMag ™ -2 (123.21D, Invitrogen Dynal AS) were also employed.   All DNA synthetic oligonucleotides used were purchased from Integrated DNA Technologies and their sequences are summarized in Table 4. All of them were reconstituted upon reception in TE buffer to a final concentration of 100 μM, divided into small aliquots and stored at −80 °C.

Reagents and
It is worth to mention at this point that the probe and target methylated DNA strands were designed according to the specific promoter sequence of RASSF1A 4 and MGMT 55 genes, and that methylated and unmethylated target DNAs have similar hybridization capacity to probe oligonucleotides 27 .

MBs modification.
MBs modification was different depending on the type of biosensor prepared (Fig. 1).
Genomic DNA was isolated from these cells with QIAamp DNA FFPE Tissue Kit (Valencia, CA, USA) according to the manufacturer's instructions with minor modifications. DNA concentrations and quality were measured using a Nanodrop 1000A spectrophotometer (Wilmington, DE, USA), obtaining ratio values confirming pure DNA in all cases. Due to DNA from tissue biopsies is fragmented when isolated from FFPE, genomic DNA from U87 was sonicated 3 cycles of 10 s at 30% amplitude for fragmentation.
The analysis of genomic DNA extracted from cells and the 897 bp linear dsDNA standards involved their previous denaturation by heating at 97 °C for 5 min in a thermocycler and transferring immediately to ice for 10 min just before making the determination with the biosensor using a similar protocol to that followed with the synthetic target DNAs.
This study and all the experimental protocols used were performed according to the guidelines and regulations and approved by the University Complutense of Madrid. It is worth to mention that since the samples analyzed were commercial serum samples and urine and saliva collected from one of the authors of this paper (V. Ruiz-Valdepeñas Montiel) no other written informed consents were required.