A Urinary Bcl-2 Surface Acoustic Wave Biosensor for Early Ovarian Cancer Detection

In this study, the design, fabrication, surface functionalization and experimental characterization of an ultrasonic MEMS biosensor for urinary anti-apoptotic protein B-cell lymphoma 2 (Bcl-2) detection with sub ng/mL sensitivity is presented. It was previously shown that urinary Bcl-2 levels are reliably elevated during early and late stages of ovarian cancer. Our biosensor uses shear horizontal (SH) surface acoustic waves (SAWs) on surface functionalized ST-cut Quartz to quantify the mass loading change by protein adhesion to the delay path. SH-SAWs were generated and received by a pair of micro-fabricated interdigital transducers (IDTs) separated by a judiciously designed delay path. The delay path was surface-functionalized with monoclonal antibodies, ODMS, Protein A/G and Pluronic F127 for optimal Bcl-2 capture with minimal non-specific adsorption. Bcl-2 concentrations were quantified by the resulting resonance frequency shift detected by a custom designed resonator circuit. The target sensitivity for diagnosis and identifying the stage of ovarian cancer was successfully achieved with demonstrated Bcl-2 detection capability of 500 pg/mL. It was also shown that resonance frequency shift increases linearly with increasing Bcl-2 concentration.

with benign gynecological disorders, early-and late-stage ovarian cancer [18]. The average urinary level of Bcl-2 was found to be 0.59 ng/mL in healthy controls, 1.12 ng/mL in benign disorders, 2.60 ng/mL in early-stage ovarian cancer and 3.58 ng/mL in late-stage ovarian cancer. The highest Bcl-2 concentration observed in the study was around 10 ng/mL. The number of samples, average concentration, and standard deviation of Bcl-2 for these four patient groups are listed in Table 1. Signs of poor early stage diagnosis can be observed from the table of samples included in this study which represents actual availability of samples from tissue banks containing each stage of samples. Thus, analyzing the values in Table 1, the minimum detectable target concentration of Bcl-2 was chosen to be 0.5 ng/mL for design and for experimental characterization studies reported herein. Table 1. Elevated urinary Bcl-2 in cohorts for healthy controls, benign diseases, and early-and late-stage ovarian cancer (N:388) [18]. The efficacy of Bcl-2 as a biomarker for ovarian cancer was further validated by comparison to CA125 serum levels using ELISA tests on 35 samples from the same cohort [18]. The comparison of Bcl-2 and CA125 levels for the same samples shows efficacy of Bcl-2 as a urinary ovarian cancer biomarker for reliable dual screening with CA125.

Number of Samples
The studied biosensor employs shear horizontal (SH) surface acoustic waves (SAWs) to identify mass loading changes caused by Bcl-2 binding specifically to antibodies on the sensor surface. It is composed of a pair of interdigital transducers (IDTs) microfabricated on ST-cut Quartz wafers in the direction 90° off x-axis and delay path specifically functionalized to capture Bcl-2 proteins while minimizing non-specific adsorption ( Figure 1). An experimentally-verified optimized surface functionalization scheme was employed for effective capture of Bcl-2 protein while maximizing sensitivity and selectivity. The developed surface functionalization technique also minimizes the non-specific binding to the sensor surface. The sensor's electrical connections were made by low-conductivity silver-reinforced epoxy. The experimental characterizations of the sensor's response to varying Bcl-2 concentrations were performed in a custom-designed oscillatory circuit. The oscillatory circuit was composed of two RF amplifiers connected in series, a frequency counter, an oscilloscope (to monitor the signal), and the sensor, which was used as the feedback element determining the oscillation frequency. The characterization was performed by using multiple sensors with up to 10 tests conducted on a sensor by cleaning the delay path with 1.5 M NaCl solution in de-ionized (DI) water. The tests were done by placing 80 µL droplets of Bcl-2 solutions (in Dulbecco's phosphate-buffered saline-DPBS) with various concentrations on the delay path. Quantification of the Bcl-2 concentration was achieved by monitoring the frequency shift for each solution. The frequency shift was caused by the change in surface density of the delay path (mass loading). As surface density increased by protein adhesion, SAW velocity decreased, resulting in a reduction in the oscillation frequency that was measured by the frequency counter. The frequency shift for each tested concentration was measured, and the sensor was successful in detecting Bcl-2 concentrations as low as the target concentration, 0.5 ng/mL. It was observed that the frequency shift had a linear trend corresponding to increasing Bcl-2 concentration. Additionally, minimal frequency shift was observed for the control DPBS solution with no Bcl-2 present. In the following section, important design parameters, fabrication of the sensor, and surface functionalization are discussed in detail. In Section 3, the electrical characterization of the sensor and results are presented. The final section covers the discussions and conclusion along with the future work.

Sensor Design
The sensor uses shear horizontal surface acoustic waves, which are frequently used for liquid-loaded biosensing applications. In SH-SAWs, the particle displacement is in the plane of the surface (unlike normal-to-surface displacement of Rayleigh waves). SH-SAWs are not affected or damped by liquid loading, as compared to Rayleigh waves, in which the particle displacement is directly coupled with the liquid on top and highly damped by mass loading of the liquid itself. Thus, Rayleigh waves are virtually insensitive to mass loading changes in liquid sensing applications. On the other hand, almost all SH wave propagation on various substrates results in leaky waves (not pure waves like Rayleigh waves), which also leak into longitudinal and shear vertical wave components when excited. For this reason, special cuts of typical wafer types of wafers are typically used for SH waves, in which the energy is highly concentrated on the SH mode. Typical wafer types used in SH-SAWs are ST cut Quartz, 41° and 36° Y-cut lithium niobate (LiNbO 3 ), and 36° Y-cut lithium tantalate (LiTaO 3 ). The sensing mechanism of SH-SAW sensors relies on the change of SAW speed either by change in mass loading (most biological and chemical sensors) or by changing physical parameters. In general, the majority of SAW sensors include surface treatments and extra layers to effectively and specifically sense the target analyte. Several SH-SAW sensors have been reported using 36° Y-cut LiTaO 3 [20], ST-cut Quartz [21,22], 36° Y-Cut Quartz [23], 41° Y-cut LiNbO 3 [24,25], 64° Y-cut LiNbO 3 [26,27], potassium niobate (KNbO 3 ) [28], and langasite [29]. In the design stage, different alternative SH-SAW generating wafers (such as ST-cut Quartz, 36° Y-Cut LiNbO 3 , and LiTaO 3 ) were tested with identical delay path designs and surface functionalization steps. It was observed that ST-cut Quartz is the most stable and the easiest to operate among those tested. ST-cut Quartz is also favorable for narrower bandwidth operation, and it does not need additional layers or gratings to concentrate the energy in the surface. SH waves are present in the direction of 90° off the x-axis (parallel to primary flat) in ST-cut Quartz, so the features were designed to obtain wave propagation in this particular direction.
The substrates used in this study were 3-inch, single-side-polished, 500 µm-thick ST-cut Quartz wafers. The SH-SAWs were generated and sensed by a pair of interdigital transducers separated with a delay path on these wafers. The pitch (corresponding to the wavelength of the SAW) was chosen as 300 µm, ensuring fabrication yield and tolerable wave attenuation through the delay path. Each finger of the IDT was 75 µm wide (corresponding to the one quarter of wavelength) for the most efficient SAW generation, as reported by others [30]. The design parameters of the sensor are illustrated in Table 2.

Fabrication
The IDTs were microfabricated using conventional MEMS fabrication techniques with a single-mask photolithography process. The fabrication steps are illustrated in Figure 2. The fabrication process started with cleaning the wafer in an acetone bath placed in an ultrasound cleaner for 10 min, followed by rinsing with methanol and DI water and dried by a stream of nitrogen ( Figure 2(a)). The metal film (Cr) was then sputtered by DC sputtering (CRC Sputter, Torr International, New Windsor, NY, USA) for 5 minutes at 200 mA constant current to obtain a film thickness of 1,000 Å ( Figure 2(b)). After sputtering, the wafers were cleaned once again with acetone, methanol, and DI water and dried with nitrogen. A positive photoresist S1813 (Shipley, Marlborough, MA, USA) was then spun on the wafer. Photoresist was spun initially for 10 seconds at 700 rpm to be spread around the wafer, and then for 40 seconds at 3,000 rpm to reach the desired thickness ( Figure 2(c)). A photoresist thickness of 1.6 µm was obtained using this recipe. The wafers were then soft-baked at 90 °C for 60 seconds before exposure. After soft bake, the wafers were exposed to UV light for 5 seconds with an average intensity of 25 W/m 2 to transfer the features from the mask to the wafers ( Figure 2(d)). The exposed wafers were then developed in developer MF-319 (Shipley) for 70 seconds to remove the exposed portions of the photoresist. The wafers were then hard-baked for 5 minutes at 100 °C. The features were formed by etching for 40 seconds (Figure 2(e)). Then, the remaining photoresist was removed in an acetone bath (Figure 2(f)). After completing the fabrication process to realize the sensors, the wafers were coated with photoresist to protect the features during the dicing process. The dicing of the ST Quartz wafer was performed using resin-bonded diamond blades at a spindle speed of 20 K rpm and a feed rate of 1 mm/s in the dicing saw. One of the fabricated ST-cut Quartz sensors used in this study is illustrated in Figure 3. It should be noted that the yield was found to be very close to 100% with these optimized fabrication parameters.  The electrical connections to the sensors were obtained with a low-conductivity silver-reinforced conductive epoxy (Duralco 120, Cotoronics Corp., Brooklyn, NY, USA). This connection method was preferred over the wire bonding method because stronger connections are desired during tests and it is not possible to solder chromium. Unlike the wire bonding method, it was observed that the conducting silver epoxy successfully withstood the solvents used in the surface functionalization of the delay path.

Surface Functionalization
The sensor will quantify Bcl-2 levels in aqueous solutions, so it is necessary to apply surface treatments to the delay path section of the sensor to specifically and selectively sense the target protein.
Molecular self-assembly techniques involving bioconjugation were employed to develop an effective Bcl-2 capture method. Bioconjugation can be defined as molecular linking of two or more components to construct a compound-namely, molecular self-assembly of several reagents for a desired purpose [17,31]. A scheme was developed to isolate the Bcl-2 protein from a solution via specific binding to a monoclonal capture antibody that is immobilized on the sensor surface. More specifically, self-assembled monolayers (SAMs) and blocking agents were used to optimize Bcl-2 capture on the sensor surface. The surface functionalization developed in this study is novel as it is the first of its kind to isolate and capture Bcl-2 protein using functionalization by surface assemblies. The method was optimized for most effective Bcl-2 capture, with trials of several different SAMs and recipes, which were presented in detail in [17]. The optimization was handled by modified sandwich ELISA tests. The surface functionalization recipe employed in this study is summarized in the following steps ( Figure 4): • First, oxygen plasma cleaning of the surface was performed to remove organic residues and form hydroxyl groups on the surface. The sensors were oxygen plasma-cleaned for 5 minutes. The beaker in which the silanization was performed also was plasma-cleaned for 5 minutes. The hydroxyl groups serve as a foundation for organofunctional silanization. The sensors were kept in the solution for 30 minutes, while IDTs were protected by Kapton tape. If the IDTs were not protected, it was observed that the chromium was attacked by HCl formed during the silanization process, and the sensor did not work properly. After the silanization step, the sensors were rinsed with acetone and ethanol and dried with nitrogen. • Immunoglobin G (IgG) antibodies (polyclonal rabbit anti-human Bcl-2, Sigma Aldrich, St. Louis, MO, USA) were used for Bcl-2 capture, which are Y-shaped with two antigen binding (F ab ) regions and one F c . The F c regions were immobilized with the help of Protein A/G onto the surface, resulting in properly-oriented free F ab regions for the most effective Bcl-2 capture and therefore maximum sensor surface affinity. A working concentration of 5 µg/mL of anti-Bcl-2 in DPBS was placed on the sensor surface as a droplet covering only the delay path for 1 hour. The sensor surface was then rinsed with DPBS before the final step.
• Pluronic F127 (Sigma Aldrich) was adsorbed to obtain a non-fouling surface for highly selective Bcl-2 capture, which is essential for a diagnostically applicable sensor. The Pluronic, a tri-block copolymer whose non-fouling nature is mediated by its two polyethylene glycol (PEG) chains, prevents other molecules from non-specifically attaching to the sensor surface. The sensor was submerged in 10 µg/mL Pluronic F127 in DI water for one hour and then rinsed with DI water.

Oscillat
The SH-S Bcl-2 conce compared to oscillator c configuratio oscillation fr counter, wh Bcl-2 charac Olympus 50 53220A, Ag Textronix I Figure 5. Tw As the S order to ide response is d he function oscillation fr

Results
The tests DPBS rangi biosensor w functionaliza with other a droplets of B and surface DI water, rin was able to he antibodie he shape an Bcl-2 sol control (DPB frequency w be observed ncreasing B also be obse end of the s which is a oscillation fr calculated b frequency in

Discussion and Conclusions
In this study, an ultrasonic MEMS biosensor for detection of urinary anti-apoptotic protein Bcl-2 was successfully designed, fabricated, and experimentally characterized. SH-SAWs were employed with microfabricated IDTs on ST-cut Quartz to quantify the Bcl-2 concentration. SH-SAWs were generated and sensed by a pair of micro-fabricated IDTs separated by a surface functionalized delay path. An optimized recipe using SAMs of ODMS, Protein A/G, monoclonal antibodies, and Pluronic F127 was employed for the most effective Bcl-2 capture. The method was optimized for specificity and selectivity, with trials of several different similar SAMs.
The sensor was experimentally characterized in a resonator circuit by placing buffer solutions of Bcl-2 of known concentration (in DPBS) on the delay path. Bcl-2 concentrations were characterized by the resulting resonance frequency shift caused by the mass loading increase of biomarker binding, which reduces the speed of the SH-SAWs. The target sensitivity for diagnosis and quantifying the stage of ovarian cancer was achieved with successful detection of Bcl-2 in the concentration range of 0.5 to 12 ng/mL. It was also observed that there is a linear relationship between the shift in resonance frequency and Bcl-2 concentration. Each sensor was used up to 10 times after applying 1.5 M NaCl solutions in DI water to remove the proteins attached to the antibodies. It was also observed that due to

Bcl-2 Concentration (ng/mL)
optimized fabrication process, the inter-and intra-array variation was minimal and had no measurable impact on the experimental results. The sensor developed was successful in detecting and quantifying Bcl-2 in the target concentration range.
The sensor can potentially be employed in a point-of-care test device for monitoring and diagnosis at the patient's bedside. The electrical components of the sensing system-RF amplifiers, frequency counter, and analog filter can potentially be miniaturized, assembled, and packaged in a standalone device with the sensor itself. A new, low-cost, accurate, safe, simple, and reliable testing platform to diagnose ovarian cancer by urinary Bcl-2 levels would benefit all women not only in the USA, but worldwide, including medically underserved geographical areas and women at high risk for developing ovarian cancer. This is especially important for detection of early-stage ovarian cancer, which is associated with high survival (>95%) and reduced lifelong medical costs, but currently accounts for less than 10% of diagnosed ovarian cancer cases. In addition to our sensor's ability to accurately detect initial ovarian cancer cases, ovarian cancer monitoring during the course of the disease may indicate recurrent disease and, possibly, therapeutic efficacy. In 2009, 21,550 women were diagnosed with ovarian cancer in the USA [1]; this biosensor could potentially detect thousands of previously-undiagnosed cases. However, it should be noted that the efficacy of the method is limited by biomarker capability/availability. Future work includes plans to collect fresh urine samples from ovarian cancer patients for further validation of the sensor as a detection assay for Bcl-2. The urine samples with concentrations of Bcl-2 characterized by ELISA tests will be tested with this SAW-based sensor.