Nanowire Substrate-based Laser Scanning Cytometry for Quantitation of Circulating Tumor Cells

Fabrication and functionalization of transparent quartz nanowire arrays

The fabrication procedure for the transparent quartz nanowire (QNW) arrays is illustrated in Figure 1. It went through a series of processes including nanoparticle coating, metal deposition, pattern transfer, and deep reactive ion etching to generate vertical nanowires. Polystyrene nanoparticles (PS NPs) were applied onto a quartz wafer using either spin-casting or dip-coating. The resulting pattern exhibits short-range ordering in a close-packed manner. The size of the PS NPs can be further shrunk using oxygen plasma etching. Then these particles served as a template to deposit chromium metal forming a nanohole pattern that was inverted to yield a nanodot pattern using nickel deposition and selective chrome etching. Finally, the nanodot pattern was transferred down to the quartz substrate using oxide reactive ion etching, producing an array of quartz nanowires. The typical diameter and length are ranging from 80 to 100 nm and 250 to 350 nm, respectively (Figure 1f). The QNW array substrate is completely transparent and suitable for optical examination (both phase-contrast and fluorescence) of captured tumor cells.

The nanowire substrate was then functionalized with monoclonal anti-human EpCAM antibody using a streptavidin-immobilization method³⁴ (Figure 1g). During this procedure, as-prepared quartz nanowires were treated with oxygen plasma to render a surface with high density of silanol group. Then we applied (3-aminopropyl)-triethoxysilane (APTES) to aminate the nanowire surface, which can be further functionalized with streptavidin(STR) via a two-step aldehyde/amine reaction and using glutaraldehyde (GA) as the linker. Finally, biotinylated anti-human EpCAM was introduced to the STR-functionalized nanowires, which, through the high affinity biotin-streptavidin binding, yielded a immunofunctionalized nanowire surface coated with anti-human EpCAM antibodies. EpCAM is a pan-epithelial cell surface marker and presumably EpCAM+ cells captured from blood are carcinoma cells shed into the blood stream.

Automated imaging and quantitation of tumor cells captured on nanowire substrates

In order to accurately count and characterize very rare tumor cells such as CTCs with good statistic power, large quantities of blood samples need to be analyzed per individual patient, suggesting the need of large-area nanowire substrates (>1″×1″) to handle the large volume of sample and ensure high-yield capture of CTCs. However it is not reliable for clinicians to perform counting or characterization of captured cells over such a large area using a conventional hemocytometer method. Here we exploited a rapid laser scanning cytometry approach to automatically quantitate all single cells captured on the substrate using a simple microarray scanner. Two laser beams (635nm and 532nm) were employed to measure fluorescence-stained tumor cells (see Supporting Method).³⁵ This approach can be fully automated and standardized for the use in clinical laboratories.

To evaluate cell capture and imaging capability, we first conducted a set of experiments using pure tumor cell samples. Cell suspensions were prepared at densities (A549, human lung carcinoma cells) ranging from ~200 to 4,300 cells per test sample (100μL) onto a set of PDMS microwells assembled on EpCAM functionalized QNW arrays. Figure 3a shows scanned images of A549 tumor cells captured on an anti-human EpCAM-functionalized QNW substrate as well as quantitative analyses of cellular parameters. The first and second columns show original fluorescence images and the third column shows the result of automated identification of cell shape (cell boundary shown in green) for morphometric analysis. In our experiments, we were able to simultaneously extract a dozen physical parameters and in Figure 3a (the fourth and fifth columns) we show two parameters for all single tumor cells we captured – one is the cell size and the other is cell circularity. It was found that the average size of immobilized A549 cells was ~ 176.3 ± 133.3 μm², which is in good agreement with the estimation using epifluorescence imaging (Figure S1). Cell circularity analysis indicates that most as-captured cells are round with a typical roundness factor ~0.88. The observed homogeneity of morphometric parameters in all single tumor cells are what we expected because the same cell line was used in all these experiments. However, CTCs from patients are expected to be highly heterogeneous and the capability for multiparameter functional (including morphometric) analysis is useful for distinguishing CTC subsets and differential diagnosis.

We also performed A549 tumor cell capture using anti-EpCAM functionalized flat glass substrates and the results are shown in Figure S2. Then we conducted a quantitative comparison of tumor cell capture efficiency between these two platforms and at varying cell densities (Figures 3b–c). The result from nanowire substrates shows a nearly linear correlation between the number of captured cells vs loaded cells for up to 4,000 cells (n=3, R²=0.94). The flat glass platform yields a lower regression, of approximately 0.75, than that of nanowire arrays, which indicates a higher cell-substrate interaction on nanowire arrays as compared to the flat glass substrate. This result is consistent with prior reports.^20,21 Figure 3c shows the cell capture yield, which is defined as the percentage of captured cells to all cells initially loaded. It was found that the capture yield remains relatively constant over the wide range of cell density. The nanowire substrate gives rise to substantially higher yield than the flat glass substrates.

In order to image small numbers of tumor cells captured on nanowire arrays, we added human lung cancer cells (A549) pre-labeled with Dil (invitrogen) to culture medium with nominal cell numbers ranging from 10 to 64 per test well (100uL). Figure 3d shows the representative scanned images of two microwells. The number of loaded cells in each test well is shown in red at the lower right corner while the number of captured cells shown in yellow at the upper right corner. The average capture yield is ~65.1.0 ± 25.2% (Figures 3e and f). These results demonstrated that we can accurately count captured cells using the imaging cytometry approach when the cell number is very low.

Quantitation of rare tumor cells from mixed cell populations

Here we take one step further to assess the utility of nanowire substrate-based laser imaging cytometry for CTC analysis in the settings close to clinical diagnosis – capture from mixed cell populations that can mimic patient specimen to some extent. Lung cancer cells (A549) were spiked at varying cell densities into the suspensions of human peripheral blood mononuclear cells (PBMCs) to perform tumor cell capture experiments. We also studied the mixed cell populations by spiking A549 cells to human monocytic cells (U937) (Figures S3 and S4). As shown in Figure 3a, the first and the second columns show single channel scanned images of all cells captured in a representative microwell loaded with the mixture of A549 cells and PBMCs. Tumor cells were pre-stained with green membrane dye Dil and all PBMCs pre-stained with red membrane dye DiD. The third column is the overlay of both channels. The histogram showing the quantitative image analysis of cell size distribution indicates tumor cells (~20–40mm) are generally larger than PBMCs (~5–15mm), manifesting the potential utility of single cell morphometric analysis for distinguishing heterogeneous cell populations. Next we systematically studied the effect of background cells on the capture yield and purity by varying the ratio of tumor and background cells. U937, a human monocytic cell line, was used as the background cells. The result indicates the capture yield of target cells (tumor cells) is relatively constant whereas the background cells do show elevated non-specific capture with increasing their density (Figure 3b and Supporting Figures S8 and S9). Finally we demonstrated the capture of rare tumor cells spiked in PBMCs at a density equivalent to clinical cancer patient samples. In this experiment, the entire chip containing 18 PDMS microwells was used to measure one sample. Automated imaging cytometry can visualize the whole chip and rapidly identify the total number of tumor cells captured from all the microwells. Figure 3c shows the raw scanned fluorescence images of each microwell (labeled 1–18) after cell capture experiment and the tumor cell capture yield is 9 out of 15 cells spiked in the PBMC sample, demonstrating the applicability of our technology for quantitation of rare tumor cells from heterogeneous cell samples.

Human whole blood samples to assess the utility for clinical CTC analysis

We used fresh whole blood samples from two volunteers and two brain tumor patients to assess the utility of our technology for quantitative analysis of rare tumor cells by spiking ultra-low abundance A549 cells in whole blood (several A549 cells per mL of blood). These spiked samples are almost identical to the blood samples from metastatic cancer patients. The experiment procedure is the following. First, these whole blood samples were spiked with A540 cells at a level of ~10 cells per mL. Second, RBC-lysis solution was added to the spiked whole blood samples and incubated for 10 min at room temperature to remove red blood cells that may complicate fluorescence imaging. Third, the remaining cells were spun down to remove debris and re-suspended in the same volume of medium for cell capture experiment. Fourth, these samples containing rare tumor cells (~ 7 or 15 cells per mL) was introduced onto the entire nanowire substrate chip, on which there are 18 PDMS wells and each contains ~ 60 μL of sample. The total volume of blood analyzed using a nanowire substrate is 18×60=1,080 μL or ~1mL. Cells were incubated on the nanowire array for 30–60min for cell capture followed by a rinsing step. For each sample, we also measured un-spiked blood for negative control (Figure S7). Finally, all the captured cells were fixed by 4% PFA in PBS buffer, stained with Phycoerythrin(PE)-conjugated anti-cytokeratin and YO-PRO-1 for 30 min at room temperature. Afterwards, captured cells were imaged using laser scanning cytometry to obtain cell count, cell size, shape, fluorescence intensity of each cell over the entire substrate (Figures S5 and S6).

Figure 4a shows the scanned images of all the microwells used to analyze a human blood sample (volunteer V1) using PE-conjugated anti-cytokeratin, whereas Figure 4b is the raw images for negative control (no tumor cells spiked). Figure 4c summarized the quantitative counts of captured tumor cells from all four human blood samples (P1, P2, V1 and V2) as well as from a negative control sample. Then we can determine the capture yield for these whole blood samples spiked with low-abundance tumor cells (Figure 5a) and the average capture yield is ~ 67. 5 ± 15.0 %. This result indicates a greater performance as compared to previous stationary nanowire-based CTC capture presumably due to the large-area of nanowire substrate and large-scale imaging capability. The imaging cytometry approach also allows for rapid quantitation of cellular parameters such as cell size (Figure 5b). Apparently the size distribution is similar for these samples, which is as expected because the same type of tumor cells (A549) were used in all the human blood sample experiments. Finally, we almost could not detect tumor cells from negative control samples according to the fluorescence signal. Only in one of all the negative control samples we observed a false signal (Figure 4b). However quantitation indicates the size of that “tumor cell” is much larger than others (Figure 5b) suggesting that imaging cytometry-based morphometric characterization may help remove false positive signal and more accurately quantitate tumor cells.

In summary, we report on a nano-enable technology platform for high-yield capture and rapid analysis of rare cells by integrating biofunctionalized quartz nanowire arrays and laser scanning cytometry. We demonstrated it can successfully capture and characterize rare tumor cells (several cells per mL) from mixed cell populations or even human whole blood, which fully justified the utility of this technology in clinical settings. This approach can also be applied to a variety of cytopathological examinations of rare cells such as tumor stem cells from tissue biopsies and immune cell subtypes for informative immune monitoring. Using a high-content imaging cytometry approach, we show not only rapid counting of captured tumor cells but also quantitative analysis of functional parameters such as cell shape and circularity at the single cell level. It has been reported that nanostructured substrates can alter cellular behavior such as adhesion, spreading and migration^36–38. We performed capture of tumor cells on nanowire substrate and found size and morphology of A549 cells on quartz nanowire substrates were generally smaller and more rounded (less spreading) than the cells captured on smooth substrate (i.e. glass). (see Figure S10, Supporting Information), indicating that nanoscale topographic features do influence the cell capture and spreading. This is an interesting biological phenomenon, but also means that we need to be cautious when we interpret the morphometric signatures of captured tumor cells in hope to identify CTC subtypes. In addition, it could be exploited to analyze other functional parameters such as surface receptor expression and intracellular signaling proteins using immunocytochemistry, further enhancing the power of this integrated approach for distinguishing phenotypic diversity and functional heterogeneity of rare cells in hope of detecting the stem-like, metastasis-initiating circulating tumor cells from patients.

Supplementary Material