A CMOS-Based Multiomics Detection Platform for Simultaneous Quantification of Proteolytically Active Prostate-Specific Antigen and Glutamate in Urine

By combining information such as genomics, proteomics, and metabolomics data, multiomics methodologies have the potential to reveal the otherwise invisible processes of carcinogenesis. Using these data, a new-generation point-of-care (POC) device may support patient care through rapid diagnosis involving accurate and simultaneous multiomics measurements. In this work, we present a proof-of-concept portable system capable of measuring a metabolite (glutamate) and a protein [proteolytically active prostate-specific antigen (aPSA)] simultaneously on a single device in a 2-min interval using on-chip optical methods. The device has the potential to improve the current diagnostic pathways of prostate cancer (PCa) by reducing the high false-positive rate with the widely used total prostate-specific antigen (tPSA) test. We demonstrate a simultaneous assay of glutamate and aPSA in buffer and synthetic urine. For the glutamate assay, the limit of detection (LOD) and resolution were found to be 0.9 and $2.7 {\boldsymbol {\mu }}\text{M}$ , respectively. The LOD and resolution for aPSA were 0.5 and 2.0 ng $/$ ml, respectively. The LOD and resolution achieved using this device are, therefore, in line with human physiological ranges. The platform is suitable for accommodating two independent optical assays, and, therefore, chemistry can be modified to measure different analytes regardless of their nature. This technological breakthrough has the potential to deliver an accurate multiomics portable device for PCa and may be extended to other diseases.

and metabolites. The chemistry required to quantify each 31 of these molecular types is different, hence most diagnostic 32 devices investigate only a single class of biomarker. How-33 ever, multiomics approaches to measuring different classes 34 of biomarkers simultaneously have the potential to provide 35 combined diagnostic information to improve the sensitivity 36 and specificity of diagnostic testing. 37 Multiomics strategies show promise to untangle the 38 complicated pathways of carcinogenesis by combining mul-39 tilevel omics approaches, including genomics, proteomics, 40 and metabolomics [1], [2]. Multiomics diagnostics can be 41 particularly beneficial for prostate cancer (PCa) where the 42 current clinical standard blood test, total prostate-specific 43 antigen (tPSA), is notoriously unreliable [3]. PCa is the most 44 common cancer in developed countries in men, claiming more 45 than 300 000 lives each year. Typically, an elevated tPSA result 46 triggers a sequence of invasive testing including digital rectal 47 examination, multiparametric magnetic resonance imaging, 48 and biopsy, leading to a diagnosis [4]. The high false-positive 49 rate of tPSA (∼70%) results in many unnecessary invasive  [11], [12]. Glutamate has drawn particular atten-71 tion as it directly correlates with the energy cycle of cancer 72 cells [13]. A correlation between urinary glutamate and PCa 73 aggressiveness (Gleason score) has also been presented [14]. 74 As cancer research enters the multiomics era [15], tech-75 nology also needs to evolve. Accurate and cost-effective 76 point-of-care (POC) technology has been developed to deliver 77 rapid measurement of multiple biomarkers in readily available 78 body fluids for different omics types, including genomics, 79 proteomics, and metabolomics [10], [16], [17], [18]  (a) Architecture and working principle of the device. (b) Normalized responsivity of the photodiode array (red) and absorbance spectra for glutamate (dotted blue) and aPSA (solid blue) assays measured with a spectrophotometer (ffTA-1 from Foster and Freeman). A 1616-array of photodiode was designed and fabri-110 cated using a commercially available CMOS 350-nm high-111 voltage four-metal process provided by austriamicrosystems 112 AG (AMS). Each pixel of the array integrates a photodiode 113 that is 100 μm × 100 μm in size, leading to a total sensitive 114 area of 1.6 mm × 1.6 mm. The size of the entire CMOS chip 115 is 3.4 mm × 3.6 mm. The sensitive area of the sensor array 116 was divided into two independent regions accommodating a 117 glutamate colorimetric assay and an aPSA chromogenic assay, 118 as shown in Fig. 1(a). The CMOS chip, including the noise 119 level of the platform, has been fully characterized in previous 120 work [20], [21]. Fig. 1(b) shows the average spectral response 121 of the photodiode array. The responsivity was normalized to 122 its maximum at 575 nm. A second responsivity peak was 123 observed at 620 nm.

124
To fabricate the two independent reaction zones on the 125 sensing area, the CMOS chip was wire-bonded onto a ceramic 126 120-pin grid array (PGA) chip package. Two sacrificial 127 polydimethylsiloxane (PDMS) microblocks of 600 μm × 128 600 μm and height of 1500 μm (SYLGARD 184 Silicone 129  turer's guidelines [22] and used to prepare test solutions. The sensor array was connected to a custom hand-held 181 reader using a zero-insertion force socket. The reader was 182 composed of a custom printed circuit board (PCB) and a 183 microcontroller (ST Nucleo F334R8 board) programed before 184 use with custom firmware. The reader was used to interface 185 the sensor array for addressing, data digitization (12 bits, 3.3 V 186 range), and transmission to a personal computer (HP EliteBook 187 i7-8650u 16 GB) via a universal serial bus (USB) link. The 188 reader was powered by the USB link (5 V) that in turn powered 189 the sensor array (3.3 V). Data were collected with an average 190 frame per second of 8.3 frames/s.

191
To verify that both glutamate and aPSA assays were work-192 ing as expected, the absorbance spectrum for both the assays 193 was initially assessed using a micro-spectrometer (ffTA-1 194 from Foster and Freeman). The absorbance spectrum of the 195 glutamate assay was measured by mixing 50 μL of 0.4 mM 196 L-glutamic acid with 50 μL of the reagent solution in a 197 384-well plate. The absorbance spectrum of the aPSA assay 198 was measured by mixing 50 μL of 800 ng/ml aPSA with 199 50 μL of the reaction solution in a 384-well plate. Spectra 200 were measured after a 24-h incubation at room temperature. 201 Both the spectra for glutamate and aPSA are shown in 202 Fig. 1(b). For glutamate, an absorbance peak was observed 203 at 499 nm and the full-width at half-maximum (FWHM) of 204 the curve was 160 nm. For a PSA, an absorbance peak was 205 observed at 419 nm and the FWHM of the curve was 34 nm. 206 For both glutamate and aPSA assays, experiments on the 207 chip were performed in buffer and synthetic urine. Test solu-208 tions were obtained by spiking assay buffer (50 mM Tris, 209 1 M NaCl, pH 8.0) or synthetic urine (artificial stabilized 210 urine BZ104 from Biochemozone) with a known concentration 211 of glutamate and aPSA stock solutions. The synthetic urine 212 did not contain any glutamate or aPSA before being spiked. 213 Additional details regarding the preparation of synthetic urine 214 samples are reported in the supplementary material.

215
For each experiment, 50 μL of the reaction solution and 216 50 μL of the test solution were introduced into the respec-217 tive microwell by manual pipetting. No further wash step 218 or incubation was needed. Optical absorbance at 430 nm 219 was measured using the photodiode array for 2 min after 220 the reaction was started. Measurements were performed at 221 room temperature. An 8-mW light emitting diode (LED) 222 (λ = 430 nm, FWHM = 20 nm) from Thorlabs was used to 223 illuminate the sensing area of the sensor array. The distance 224 between the LED and the sensor array was 10 cm. No lens 225 was used in this setup. All the measurements were performed 226 in dark conditions. The LED power was adjusted to have 227 the array working in the linear region of the characteristic in 228 Fig. 1(c) (irradiance value 30 μW/cm 2 ). The irradiance value 229 was kept constant in all the measurements. The wavelength 230 of 430 nm was selected as a tradeoff between the absorbance 231 peak of the two reactions and the responsivity peak of the 232 photodiode array [see Fig. 1(b)]. The output characteristic of 233 the photodiode array at 430 nm was measured and is shown in 234 Fig. 1(c). The measurement setup is shown in Fig. 1(c) (inset). 235 The CMOS chips employed in this work were all from the 236 same fabrication batch.

237
The device was cleaned and reused. A cleaning proce-238 dure after every measurement was used to avoid cross-239 contamination. The cleaning process involved a sequential 240 rinse in deionized (DI) water, isopropyl alcohol (IPA), and then 241 ethanol. The washing sequence was repeated multiple times.

242
Nitrogen was used to blow dry the device after the cleaning 243 procedure.

E. Data Analysis 245
The measure of the initial reaction rate was performed using 246 a custom MATLAB program. Output data from the pixel in 247 the same microwell were low-pass-filtered and averaged. Unre-248 sponsive sensors or sensors affected by strong artifacts were 249 excluded from the averaging process after visual inspection.

250
The use of a sensor array is particularly beneficial to improve 251 the signal-to-noise ratio as shown in previous works [23]. Data  Fig. 2(b).

284
For the aPSA assay, the platform was tested and character-  Table I.

293
For the glutamate assay, LOD and resolution were found  that urine glutamate might be more than doubled in patients 300 with a Gleason score of 8 or higher with respect to healthy 301 men [14]. Therefore, the LOD and resolution data obtained in 302 this work are suitable for measuring glutamate in urine. For 303 the aPSA assay, LOD and resolution were 0.5 and 2.0 ng/ml, 304 respectively. A urinary tPSA concentrations <150 ng/mL 305 in conjunction with elevated serum tPSA concentrations 306 >4 ng/mL have been linked to PCa diagnosis [27]. In urine, 307 PSA mainly exists in its free form and is enzymatically 308  Fig. 3. The first test solution did not contain any 323 glutamate or aPSA, and, therefore, only a background signal 324 was observed [see Fig. 3(a)]. The microwells used for the 325 glutamate and aPSA assays each provided a clearly detectable 326 signal when a solution containing only glutamate or aPSA was 327 introduced to the device [see Fig. 3(b) and (c)]. To complete 328 the study, a solution containing both glutamate and aPSA was 329 introduced to the device. A strong signal was observed from 330 both the microwells [see Fig. 3(d)].

331
Data were processed and initial reaction rates were mea-332 sured for each experiment for the first 2 min of the experiment. 333 Reactions rates for the four testing conditions are shown in 334 Fig. 3(e)-(h). The measured initial reaction rate from these 335 experiments was similar to those observed in the calibration 336 curves.

C. Simultaneous Assay in Urine 338
Further validation of the platform for multiomics testing 339 in synthetic urine was then conducted. Urine was selected 340 as the medium of interest since there is growing interest in 341 improved urine testing for PCa [29]. Urine testing benefits 342 from easy sample collection and elevated concentrations of 343 numerous biomarkers of interest compared to blood. aPSA is 344 predominantly present in urine [7], and urinary glutamate has 345 also been indicated as a promising PCa biomarker [14].

346
The platform was tested on seven synthetic urine samples. 347 The synthetic urine did not contain any glutamate or aPSA.

348
Therefore, test samples were obtained by spiking synthetic 349 urine with known concentrations of glutamate and aPSA. 350 Glutamate and aPSA solutions were prepared as described in 351 the materials and methods section and experiments were con-352 ducted using the same procedure outlined for the simultaneous 353 experiments in buffer (Section III-B).

354
The concentration of glutamate in the first six test samples 355 was 0, 10, 20, 60, 100, and 200 μM, respectively. The 356 concentration of aPSA used in the same six test samples was 0, 357 20, 40, 150, 100, and 400 ng/ml, respectively. The test samples 358 were tested in the same order as reported above. To confirm 359 that the sensor was not subject to any form of hysteresis when 360 exposed to high glutamate or aPSA concentrations, the seventh 361 sample with lower concentrations at 75 μM of glutamate 362 and 10 ng/ml aPSA was then tested. The first test sample 363 (0 μM glutamate and 0 ng/ml aPSA) was used as background 364 measurement [see Fig. 4(a)]. The other test samples produced 365 clearly detectable signals dependent on the concentrations of 366 glutamate and aPSA [see Fig. 4

367
For each experiment, the initial rate of reaction was mea-368 sured using the first 2 min of the recordings. The initial 369 reaction rates observed in the simultaneous assays in urine are 370 shown in Fig. 5(a) and (b). In Fig. 5(a) and (b), samples were 371 sorted with the respective concentration of the biomarker in 372 ascending order. The initial reaction rates measured simultane-373 ously in synthetic urine were comparable to those observed in 374 the buffer calibration curves. This provides a proof of principle 375 that the platform is suitable for quantifying glutamate and 376 aPSA simultaneously in urine.

377
D. Discussion 378 We have presented a proof-of-concept that simultaneous 379 measurement of metabolites and proteins is feasible using a 380  to other GLOX-based glutamate sensors, which are typically 383 below 1 μM [25]. The buffer aPSA LOD is also similar 384 to the one achieved in similar works [7]. The test time 385 (2 min) and the absence of washing steps or incubation 386 match commercial POC platform requirements [19]. Unlike 387 other state-of-the-art devices, this platform can measure both 388 a metabolite (glutamate) and a protein (aPSA) simultane-389 ously. This feature is currently missing in state-of-the-art 390 devices.

391
The use of a CMOS-based photodiode array is particu-392 larly advantageous for affordability. If mass-produced, the 393 device could reach cost levels allowing for single-use disposal 394 devices. The use of an optical sensor array with physical 395 separation ensuring low crosstalk is beneficial for scalabil-396 ity. Potentially more microwells could be integrated on the 397 platform, for instance, by extending the CMOS-sensitive area 398 or further miniaturizing sample application capability. Since 399 the techniques adopted in this work, including CMOS tech-400 nology, epoxy molding, and bio-specific assays, are suitable 401 for large-scale production, we believe that the platform we 402 have demonstrated is not limited to only two analytes; it can 403 be potentially scaled up to accommodate many optical assays, 404 regardless of their nature.

405
In this work, we have demonstrated the multiomics capa-406 bility of a CMOS sensing platform. The platform was tested 407 on urine samples. Since concentrations of aPSA in blood and 408 urine of healthy subjects and men affected by PCa are still 409 under research, we have assumed the aPSA concentration to 410 be comparable with the tPSA concentration. We believe that 411 the platform can also be used for multiomics testing of other 412 body fluids, including blood serum and plasma. For example, 413 plasma aPSA and plasma glutamate have also been linked to 414 PCa [30]. Processing techniques are also available to activate 415 other inactive forms of blood PSA using, for instance, human 416 kallikrein 2 (hK2) [31].

417
While urinary tests are undoubtedly more convenient than 418 blood tests, they do present challenges associated with wide-419 ranging physiological variation in water content that will 420 result in a wide reference range of "normal" for any diag-421 nostic urinary analyte. To control for this, the multichip 422 could include a physiological control, such as urinary crea-423 tinine using microfluidics multiplexing [24]. The biomarker 424 panel we propose here is a valuable multiomics proof-of-425 concept that might be improved by introducing additional 426 analytes.

427
Further studies leading to a full clinical trial employing 428 the diagnostic method suggested in this research are required 429 to demonstrate that the platform can deliver an effective 430 POC diagnostic tool for PCa and to determine specificity 431 and robustness against other potential interference including 432 sample-to-sample variations, batch-to-batch discrepancies, and 433 stability over time.

434
The use of multiomics testing may also have the advantage 435 of characterizing the nature of underlying cancer beyond the 436 simple binary outcomes of current POC tests. Ultimately, this 437 could result in a test that not only identifies whether cancer 438 is present or not, but one that could distinguish between 439 potentially life-threatening and more benign forms of the 440 disease. Indeed, it is possible that such a test could go further 441 and become a predictive marker, identifying a subgroup of men