Design and performance evaluation of microfluidic chip
The microfluidic-assay system was constructed according to Figure 1 (a-c), of which the self-designed microfluidic chip acted as the core component. The basic capture unit of the microfluidic chip is roughly a semicircular arc of three independent micro-pillars (Figure 1(d)). The upward and the bottom openings are respectively the inlet and the outlet of the fluid flow. The outlet size was specially set at 9 um which allows the smaller background cells to pass through while keeping the larger CTCs stuck. To minimize the detrimental clogging commonly encountered in the blood sample, we proposed two solutions. The first was to set dozens of rows of larger capture unit (with an inlet of 60 um and two outlets of 21 um) at the very beginning of the flow pathway. Once the clogging occurs, the debris will be captured and cleaned prior to entering the CTC capturing region. The second was to enlarge the flow passage width between two adjacent capture units to 30 um. Although CTCs might be missed due to wider flow passage, the capture probability can still be compensated and enhanced by having more rows of the capture units. In the optimized design of the microfluidic chip, over 200 rows of the capture units were set in a parallel and staggered manner.
To assess the impact of hydrodynamics on the captured cells, we performed a numerical simulation to reveal the fluidic characteristics. As illustrated int Figure 1 (e-h), the velocity profile and the probable cell trajectory were depicted to reflect a probable capture of the cells. Further analyses showed that in the microchannel, the shear stress caused by the flow ranged from 0 to 4.6 Pa and the shear rate was consistently smaller than 95.3 1/s. These values were within the safe range of human’s normal physiological state of <7.0 Pa stress [30] and <2000 1/s shear rate [31], respectively. These observations convinced us that the microfluidic chip was capable of isolating CTCs in a harmless and intact way.
The performance of the microfluidic chip was characterized from two aspects: the capture efficiency and the inter-assay variability. In spiked cell line experiments, the capture efficiency gradually declined with the increase of the flow rate, but in all the groups, the chip’s capture efficiencies were consistently higher than 75% (Figure 2(a)). To balance the capture efficiency and the time consumption, we chose the flow rate of 2ml/h for the subsequent processing of patients’ blood samples. Further, the experiments conducted on the five groups with different cell concentrations ranging from 5 to 100 cells/3ml showed our method achieved an RSD smaller than 10% (Figure 2(b)), indicating a desirable consistency between assay repeats and a reliable experimental result.
Identification of CTCs by immunofluorescent staining
Due to the epithelial origin of bladder cancer [32], an epithelial marker is capable of distinguishing the CTCs of the bladder tumor from the non-epithelial background blood cells. Pan-CK is a subgroup of intermediate filament proteins, characterized by the diversity and abundance of polypeptides presented in human epithelial tissues [33]. Using anti-Pan-CK antibody as a biomarker would be amply adequate to realize a wide coverage recognition of bladder CTCs. Besides, CD45 has been well confirmed as a reliable marker of the white blood cells and DAPI is widely used to stain the nucleus [34]. Therefore, a combined marker-panel of “DAPI+/Pan-CK+/CD45-/” enabled us to identify CTCs in the peripheral blood and to eliminate the interference caused by background blood cells (Figure 2(c)).
To validate the performance of the combined marker-panel, three different human bladder cancer cell lines (UMUC-3, 5637 and T24) were tested. The expected staining and identification of all the cell lines verified the efficacy of our immunofluorescent protocols (Figure 3 (a)).
Correlation between CTC enumeration and the clinical outcomes of bladder cancer
With the successful isolation of CTCs from the peripheral blood (Figure 3 (b)), correlations between CTC enumeration and clinical prognostic outcomes were assessed based on a cohort of 48 bladder cancer patients with varied degrees of disease progression. The baseline demographics and clinicopathological characteristics of eligible patients are summarized in Table 1.
There is a significant elevation in the CTC count for MIBC versus NMIBC patients [4.67 (95% CI, 1.41-7.93) vs. 1.88 (95%CI, 0.76-3.00) CTCs/3 mL; P=0.019] (Figure 4(a)). Similarly, the CTC count increased significantly in the high-grade bladder cancer patients verses the low-grade and PUNLMP (Papillary urothelial neoplasm of low malignant potential) bladder cancer patients [3.69 (95% CI, 1.89-5.49) vs. 1.18 (95% CI, 0.19-2.17) vs. 0.20 (95% CI, -0.36-0.76) CTCs/3mL; P=0.024;] (Figure 4(b)). By contrast, there were no significant correlations between the CTC enumeration results and other clinical prognostic outcomes such as BC history, tumor multifocality, risk level of NMIBC and tumor size (Figure 4(c-f)).
Table 1 Baseline clinicopathological characteristics of the cohort
|
Characteristics
|
NMIBC1 (n=33)
|
MIBC2 (n=15)
|
P-value
|
Age, mean (SD)
|
65.7 (10.2)
|
65.6 (10.2)
|
0.493
|
Gender, n (%)
|
|
|
0.143
|
Female
|
2 (6.1)
|
3 (20.0)
|
|
Male
|
31 (93.9)
|
12 (80.0)
|
|
Body mass index, mean (SD)
|
24.1 (3.5)
|
24.4 (3.7)
|
0.654
|
Smoking history, n (%)
|
|
|
0.834
|
Yes
|
23 (69.7)
|
10 (66.7)
|
|
No
|
10 (30.3)
|
5 (33.3)
|
|
Drinking history, n (%)
|
|
|
0.875
|
Yes
|
14 (42.4)
|
6 (40.0)
|
|
No
|
19 (57.6)
|
9 (60.0)
|
|
Hematuria, n (%)
|
|
|
0.688
|
Yes
|
20 (60.6)
|
10 (66.7)
|
|
No
|
13 (39.4)
|
5 (33.3)
|
|
Urine, median (IQR)
|
|
|
|
Leucocyte (/uL)
|
20.8 (4.5–52.5)
|
34.6 (6.5–171.7)
|
0.317
|
Bacterium (/uL)
|
42.2 (13.6–317.8)
|
100.2 (28.1–513.5)
|
0.247
|
Blood, median (IQR)
|
|
|
|
Serum creatinine (umol/L)
|
81.0 (73.0–93.5)
|
85.0 (77.0–92.0)
|
0.456
|
Serum urea (mmol/L)
|
5.8 (5.2–6.8)
|
5.6 (4.8–7.3)
|
0.841
|
Serum uric acid (umol/L)
|
366.0 (287.0–435.5)
|
344.0 (281.0–380.0)
|
0.312
|
Pathological grade, n (%)
|
|
|
0.004
|
PUNLMP
|
5 (15.2)
|
NA
|
|
Low grade
|
11 (33.3)
|
NA
|
|
High grade
|
17 (51.5)
|
15 (100.0)
|
|
Initial BC, n (%)
|
|
|
0.259
|
Yes
|
14 (57.6)
|
6 (40.0)
|
|
No
|
19 (42.4)
|
9 (60.0)
|
|
Tumor focus, n (%)
|
|
|
0.724
|
Nonmultifocality
|
18 (54.5)
|
9 (60.0)
|
|
Multifocality
|
15 (45.5)
|
6 (40.0)
|
|
Tumor size, n (%)
|
|
|
0.040
|
< 20mm
|
17 (51.5)
|
3 (20.0)
|
|
≥ 20mm
|
16 (48.5)
|
12 (80.0)
|
|
Surgical options, n (%)
|
|
|
0.151
|
Radical Cystectomy
|
4 (12.1)
|
5 (33.3)
|
|
TURBT3
|
29 (87.9)
|
10 (66.7)
|
|
CTCs/3 mL, median (IQR)
|
1.88 (0.76-3.00)
|
4.67 (1.41-7.93)
|
0.019
|
1MIBC, muscle-invasive bladder cancer; 2NMIBC, non-muscle-invasive bladder cancer; 3TURBT, Transurethral resection of bladder tumor
|
CTC count as a prognostic marker of bladder cancer
To assess whether the CTCs count could be used as a supplementary biomarker for the stratification of bladder cancer, we performed ROC analysis of CTC enumeration and patients with bladder cancer in different clinical stages and grades (Figure 5(a-b)). The AUC [95% confidence interval (CI)] were calculated by comparing NMIBC with MIBC group, and high-grade patients with combined PUNLMP and low-grade groups. The AUC in comparing the NMIBC and MIBC cohort was 0.707 (95% CI, 0.545-0.869; P=0.023) with a sensitivity and specificity of 80.0% and 66.7%, respectively (Figure 5(a)). Similarly, the AUC comparing the PUNLMP/low-grade and high-grade cohorts was 0.717 (95% CI, 0.576-0.858; P=0.015) with a sensitivity and specificity of 62.5% and 81.2%, respectively (Figure 5(b)). The optimal cutoffs for distinguishing NMIBC vs. MIBC and high-grade vs. low-grade bladder cancer were both at 1.5 CTCs/3 mL blood.