Bladder cancer (BC) is the seventh most common cancer among men worldwide and results in a yearly incidence of approximately 430,000 cases [1
]. The high recurrence rate of BC means that patients often require lifelong surveillance. Currently, there are several options for BC diagnosis and follow-up and the gold standard is transurethral resection of the bladder tumor (TURBT) [2
] (Table S1
). TURBT is an invasive procedure where a cystoscope is inserted through the urethra into the bladder and a sample of the bladder tumor is removed by resectoscope. BC types are widely divided into two pathologic entities: Non-muscle-invasive bladder cancer (NMIBC) and muscle-invasive bladder cancer (MIBC). Upon detection, 75–85% of cases are classified as NMIBC, which is associated with a five-year survival rate of 90% [3
]. However, NMIBC has a recurrence rate as high as 50–70% and 10–25% progress to muscle-invasive MIBC [2
]. Therefore, BC patients still require routine monitoring for recurrence and progression after TURBT procedures.
Liquid biopsy using blood samples has been used to isolate circulating tumor cells (CTCs) for the detection of other cancer types, namely breast, lung and colorectal cancer [4
]. Malignant cells of BC tumors are suggested to be exfoliated spontaneously into the urine at low concentrations [6
], although there is a lack of conclusive studies on the exact frequency of these exfoliated bladder cancer cells (EBCCs). Hence, EBCCs detection from urine has not been fully evaluated for use in the clinics.
The enumeration and cytological analysis of EBCCs may serve as a complementary means for diagnosis, along with cystoscopy for the detection and surveillance of bladder cancer. However, urine cytology has low sensitivity of disease detection (0–50%) [7
], and often lead to false negatives in cases of low-grade bladder tumors [8
]. Moreover, none of the currently available markers used in cytology can conclusively determine the presence of BC tumors, as the markers are not unique to this disease [9
]. Common markers for the detection of BC include the Nuclear Matrix Protein 22 (NMP22) and bladder tumour antigen (BTA), both of which are proteins secreted directly into the urine. The sensitivity of the NMP22 ELISA is typically 60–70% [10
]. Although assays with BTA demonstrates higher sensitivity than the urine cytology test, the reported specificity was much lower [11
]. Other markers such as survivin, BLCA1 and BLCA2 demonstrate higher sensitivity and specificity but were unable to provide further characterization of the cell phenotypes. Hence, the invasive and costly procedure of cystoscopy remains necessary in clinical settings. There is an urgent need to develop a procedure that can guide the characterization of EBCCs for a novel biomarker discovery, as well as the non-invasive method to better evaluate the presence and status of the disease.
One of the key caveats of using urine samples for disease management is cellular heterogeneity. A urine sample is a heterogeneous collection of different cell constituents [12
] and may include white blood cells (WBCs) and red blood cells (RBCs) due to hematuria and inflammation, as well as various casts, mucus threads and squamous epithelial cells (SECs) (diameters ranging from 30 µm to 60 µm) exfoliated from the distal urethra [13
]. These contaminants can severely influence the sensitivity of the cytological analysis. EBCCs also comprise various epithelial-mesenchymal transition (EMT) subpopulations, the latter a phenomenon characterized by the loss of cell-to-cell adhesion, increasing cell motility and invasiveness of cancer [14
]. EBCCs may express a range of different biomarkers depending on their EMT status, of which cytokeratin (CK) and vimentin (VIM) are two protein markers that have a pivotal role in the development and progression of BC. The expression level of VIM is also associated with grade, recurrence, and progression [10
Recent efforts to isolate EBCCs from urine include methods such as filtration [15
] and immune-capture [17
] (Table S2
), but the techniques still face technical limitations. For example, filtration-based methods that involve smaller pore sizes induce shear stress that will reduce cell viability and functionality, as well as alter the morphology of cells. Other procedures based on antibody recognition may be highly specific but lack sensitivity [19
]. Approaches have been developed for the sensitive and specific detection of bladder cancer [20
], but the procedures are based on genomic assays which require lysis of the target cell components. Hence, a more efficient technique is highly warranted for routine and rapid clinical usage.
Here, we designed a new strategy using an EBCCs sorting (ES) device to isolate malignant EBCCs for real-time detection within hours to facilitate disease management. The procedure integrated a microfluidic assay based on the principles of inertial focusing for the high-throughput separation of EBCC subtypes based on the cell size. We aimed to develop a label-free and non-invasive procedure with the potential for EBCCs detection from urine with high robustness and sensitivity. Isolated cells were characterized with the downstream phenotypic analysis using antibodies or probes targeting established BC associated markers such as epidermal growth factor receptor (EGFR) and survivin to confirm the presence of EBCCs. Further screening with EMT markers also revealed selective enrichment of more aggressive EBCCs subtypes in one of the outlets, which could be utilized to provide insights for the clinician and assist disease management. The functional assays carried out supported the observation of different phenotypes between sorted populations.
The uniqueness of a rapid and non-invasive method permitting the separation of different EMT phenotypes bears high clinical implication. To our knowledge, no prior study has reported on the specific enrichment of mesenchymal BC cells in a label-free manner. Using the ES device, we aimed to process and characterize enriched cells on the same day, for real-time detection application and to enhance personalized treatment.
Bladder cancer has a high recurrence rate, and errors in clinical staging as well as pathological grading of bladder cancer are frequent [37
]. The main problem of an invasive therapeutic procedure for BC diagnosis and prognosis is burning of the tissue during cauterization, making it difficult to determine whether the superficial layer or stroma has been invaded, moreover invasive procedures are very painful for the patient. The development of a simple and robust platform capable of detecting BC with non-invasiveness and high sensitivity remains a challenge. Here we demonstrate the application of a non-invasive, label-free procedure using inertial focusing microfluidics for the isolation of BC cells from urine. Due to the scope of this project for demonstrating the efficacy of bladder wash procedures in the reduction of EBCC counts over time, bladder wash urine samples were utilized throughout this study. Four samples were obtained from each patient at different time points of a bladder wash procedure, where the first unwashed sample correlates to the voided urine for cytology. This is one of the approved methods of collection for urine cytology. The significance of this microfluidic sorting device is the ability to isolate intact bladder cancer cells in a label-free manner, under minimal shear stress and high sensitivity (93.3 ± 4.8%), which will allow cells to remain viable for downstream analysis, including culture. This provides unprecedented opportunities for the efficient enrichment of these rare cells, allowing easier downstream characterization and prognosis of BC. Our two-step method captures cells of interest while eliminating background cells (larger SECs and smaller leucocytes) for detection and downstream phenotype characterization.
Compared with other gold standards of diagnosis, the method we developed has the following technical merits: 1. Efficient processing of clinically relevant urine sample volumes within a short period of time (20 mL urine in 20 min); 2. high recovery, consistency, and sensitivity; and 3. viable enriched cell fraction which can be used for downstream analysis. The procedure can also be modified to remove squamous epithelial cells instead via negative selection, which will prevent the loss of bladder cancer cell clusters. Clustered cancer cells have been shown to correlate with clinical outcomes in various cancer types [38
], and may be important parameters for diagnosis. Overall, the fast processing time will help to translate this enrichment platform to in-clinic and ‘point-of-care’ applications. The simplicity in the manufacturing of the spiral microchannel device and the ease in handling of the equipment and device also make this enrichment method attractive for clinical applications requiring a one-time-use operation.
The rapid enrichment procedure allows selective enrichment of mesenchymal and intermediate EMT phenotypes. EMT occurs during the progression of tumors, endowing cancer cells with increased motility and invasiveness [39
]. Since mesenchymal and epithelial cells have different shapes and deformability properties [40
], we were able to isolate different fractions of epithelial and mesenchymal cells into respective outlets of the device. EMT phenotypes have been widely shown to correlate with tumor-initiating capabilities [41
]. Using EMT markers, we classified subpopulations of the sorted cells (Figure 5
B) and further evaluated their metastatic properties via wound healing assays, which confirmed the heightened metastatic properties of sorted cells from the target outlet (Figure 7
). The specific enrichment of these subpopulations may be valuable for the expansion of cell lines to be used in the design of new therapies targeting EMT [43
]. The EMT is a reversible process and new strategies inducing EMT reversal are promising to suppress cancer cell migration and metastasis [45
]. Moreover, accumulating evidence indicates that conventional therapies often fail to eradicate cancer cells that have entered the stem state, which is activated via EMT [46
]. These cells, termed as cancer stem cells (CSCs), are a major hurdle to traditional therapies [47
]. Therapies include the induction of CSCs apoptosis and induction of CSCs differentiation by interrupting signals from the microenvironment that regulate important properties like self-renewal, differentiation and apoptosis resistance. They hold great promise for cancer therapy whereas traditional therapies against cancer (chemotherapy and radiotherapy) have multiple limitations that lead to treatment failure and recurrence [48
]. Moreover, EMT-targeting therapy may also be useful as a personalized medicine approach that complements conventional BC treatments.
The evaluation of clinical samples revealed a vast amount of tumor heterogeneity. Such heterogeneity has crucial consequences in terms of adaptive drug resistance and tumor dormancy [45
]. With these clinical samples, we also showed that BC cells with an intermediate EMT phenotype were enriched in the innermost outlet, a population affiliated with the CSC subtype. Although the current study is focused on BC samples, it may also be relevant to the detection of prostate cancer cells as some reports have suggested that prostate cancer cells can also be shed into the urine [49
]. A combination of markers, including those specifically for bladder cells, could be implemented when realizing actual clinical utility to distinguish between various cancer types. CKs and genomic profiles associated with patient survival could also be used in combination to evaluate the association with clinicopathologic parameters and prognosis [50
Future efforts may also be directed towards the design of primary trials to test the diagnostic capacity of our microfluidic device. Developing a method for early detection of new tumors and for effective surveillance for recurrences could reduce the morbidity of bladder cancer. Further improvements also include increasing the throughput of the spiral microfluidic device so that it is capable of processing a larger sample volume (e.g., > 100 mL) within a relatively short time.
4. Materials and Methods
4.1. Device Fabrication and Characterization
The silicon master was fabricated using standard microfabrication techniques, as described previously [52
]: Six-inch-diameter silicon wafers were patterned using standard UV lithography and etched using deep reactive-ion etching (DRIE) to define the channels on the wafer (170 µm etch depth). After etching, the patterned silicon wafers were cleaned using acetone and isopropanol and treated with trichloro (1H, 1H, 2H, 2H perfluorooctyl) silane for 2 h to facilitate the polydimethylsiloxane (PDMS) mold release. After silanization, the mold was used to fabricate PDMS devices using standard soft lithography techniques. For standard soft lithography, the PDMS base and curing agent (Sylgard 184, Dow Corning Inc., Singapore, Singapore) were mixed in a 10:1 ratio and degassed for casting on the spiral microfluidic chip mold. The PDMS mixture was baked in an oven for 2 h at 70 °C. After curing, the PDMS was peeled from the mold, inlet and outlet holes were punched with the 1.5 mm Uni-CoreTM
Puncher (Sigma-Aldrich Co. LLC., Singapore, Singapore) and the PDMS device was irreversibly bonded to Petri dishes covered with PDMS using an oxygen plasma machine (Diener Electronic, Ebhausen, Germany) to complete the channels. The settings for the plasma treatment were 21% pure oxygen and 21% air for 1 min 30 s. Then 15 cm and 12 cm long sections of TYGON flexible plastic tubing with an inner diameter of 0.06 IN were coupled to the fluidic inlets and outlets, respectively. Devices were first primed with 70% ethanol, followed by phosphate buffered saline (PBS), both at a flow rate of 1.7 mL/min.
The sorting devices were mounted on an inverted phase contrast microscope (Olympus IX71) coupled with a high-speed CCD camera (Phantom v9, Vision Research Inc., Wayne, NJ, USA). UMUC3 cells were spiked in PBS and pumped through the device using flexible Tygon®
tubing at different flow rates ranging from 0.5 mL/min to 2 mL/min. High-speed videos were captured using the Phantom Camera Control software and analyzed subsequently with the ImageJ software (Figure S3
4.2. Cell Culture
UMUC3 bladder cancer cells were cultured in Minimum Essential Medium (MEM) (Thermofisher Scientific, Waltham, MA, USA) supplemented with 10% FBS (Invitrogen, Carlsbad, CA, USA) and 1% penicillin-streptomycin (Invitrogen, Carlsbad, CA, USA). Cells were cultured in a 5% CO2-humidified air atmosphere at 37 °C. They were passaged every 2–3 days at an 80% confluence. The medium was aspirated and the cultures were washed briefly with PBS. After removal of the PBS solution, 1 mL of trypsin was added and the cultures were placed in the incubator at a temperature of 37 °C during 5 min. Adherent cells were then gently dissociated from the flask bottom by inserting 1 mL of MEM media and diluted in a tube with 5 mL of media. After centrifugation at 1200 rpm for 3 min, the supernatant was aspirated and the remaining pellet was suspended in new culture flasks filled with 5 mL of MEM culture medium.
4.3. Characterisation of Sorted UMUC3 Cells
For the evaluation of cell loss under various conditions, UMUC3 cells were harvested and stained with Hoechst. Cells were then collected from each outlet respectively, centrifuged down at 1200 rpm for 3 min and the supernatant was removed to a volume of 50 µL for imaging. The exact number of UMUC3 cells spiked was obtained by loading 50 µL of unprocessed cells to a 96-well plate, scanning the well under a fluorescent microscope and counting the cell number using ImageJ. The recovered UMUC3 cell number from each outlet was then compared with the initial spiked number to identify the cell loss. The error bars were determined with triplicate experiments. To examine the effects of viability after sorting, sorted cells were stained with 0.4% Trypan Blue (Thermofisher Scientific, USA) and viewed under a microscope. The percentage of viable cells was calculated both at the start and after 3 h.
4.4. Processing of Clinical Samples
This study was approved by respective institutional review boards (IRB) and the local ethics committee (National Healthcare Group (NHG)) (2016/00380). Informed and written consent was obtained from all patients. Seven patients with BC were recruited for this study (Table 1
). Four samples (100 mL, 500 mL, 300 mL and 300 mL respectively) were collected from each patient during the process of trans-urethral removal of a bladder tumour (TURBT), each sample corresponding to one of the four steps of bladder washing processed (undiluted, first wash, second wash, third wash). Samples were collected in the morning and the entire processing was completed within 12 h. Samples were first concentrated to 25 mL and diluted with PBS in a 1:1 ratio. Diluted samples were further concentrated to 5 mL for the estimation of cell concentration with an automated cell counter. Samples were eventually resuspended into volumes to reach a cell concentration of 2.5 million per mL or below. BSA was added to achieve a concentration of 1.5% prior to filtration with the 40 µm cell strainer. Filtered cells were processed in the microfluidic spiral device. Outlets one, two and three were collected as a sample and outlets four and five are discarded as waste. The collected sample was split evenly in a 1:1 ratio for histopathology cytology and imaging, respectively.
4.5. Immunostaining of Sorted Cells
Cells collected from each of the outlets were stained with Hoechst 33,342 for enumeration. To determine the EMT phenotype, cells were also stained with anti-CK-FITC (130-098-802 Miltenyi Biotec 1:250) and anti-VIM-PE (30-106-369 Miltenyi Biotec 1:250) antibodies. 2.5% of BSA was added to avoid non-specific binding of the antibodies. To identify white blood cells, clinical samples were also stained with CD45-APC-Vio770 (130-113-677 Miltenyi Biotec 1:100). CK+ or VIM+ positive cells were counted as cancer cells only if they were also negative for CD45. Clinical samples were stained with the bladder cancer marker anti-survivin antibody (SAB5500179 SIGMA 1:100) and anti-EGFR antibody. Cells trapped by the filter were stained in situ with Hoechst 33,342 for 30 min (Figure S2
), and subsequently imaged with a fluorescence microscope (Olympus IX81).
4.6. Imaging of Samples and Cell Counting
Imaging of the samples was done with 96-wells plates using a confocal microscope under 10× objective. An automated cell-counting algorithm was generated with an image processing software (ImageJ; Figure S8
) to measure the cells counts, cell size distribution, cell intensity gradient and percentage of each phenotype. Fluorescence intensity was normalized relative to the background noise and expressed in arbitrary units (AU).
4.7. Wound Scratch Test
Sorted cells from the first two outlets were split into two separate wells of a 48-well plate. After 2 to 4 h, cells were attached and a cross-directional scratch was created. Images were taken every 30 min for 7.5 h on four points of each scratch with a confocal microscope under optimal cell culture conditions.
4.8. Statistical Summary
Data were summarized as mean ± STD (Standard Deviation). Groups were compared using a t-test with a p-value of < 0.05 taken to reflect the presence of significant heterogeneity.