Cancer is a class of diseases characterized by the uncontrolled growth of cells that ultimately invade surrounding tissues and metastasize to distant sites within the body [1] [2] . Early cancer detection is crucial for improved prognosis and cancer management due to the small tumor size and localization of the tumor at the primary site [3] [4] . Conventional cancer cell sorting techniques, which have been reviewed elsewhere [5] [6] including centrifugation, chromatography, and fluorescence and magnetic-activated cell sorting, are limited in yield and purity and further rely on the expertise and subjective judgments of highly skilled personnel. The small sample volumes, fast processing times, multiplexing capabilities, and large surface to volume ratios inherent in micro- fluidic systems [7] [8] offer new opportunities for cytology and cyto-pathology [9] - [18] particularly for in vitro cell sorting and detection [17] [19] - [25] . Leveraging these advantages, various microfluidic platforms have been developed for capturing rare cells including circulating tumor cells (CTCs), circulating fetal cells, and stem cells. Microfluidic sorting of rare cells has been reviewed elsewhere [26] - [29] .

In this review, we focus on the application of microfluidic systems for cancer cell detection and sorting. We first present the development and working principle of several key microfluidic platforms including those based on magnetic activated cell sorting [30] - [39] and differences in cellular biophysics (e.g., cell size [40] - [53] and dielectrophoresis (DEP) [54] - [89] . We discuss the performance and capabilities of each system in terms of throughput, yield, purity, cell viability, and the capability for on-chip post-processing after cancer cell capture.

Magnetics based flow detection can enable several improvements in cytometry-based analyses. Magnetic detectors are extremely small (tens of micrometers) and rugged, representing an intriguing opportunity to reduce the size and complexity of cytometers for field-deployment. The small footprint also points to the potential to create highly multiplexed systems with hundreds of parallel channels and dramatically increase sample throughput. Magnetic activated cell sorting relies on the interaction between cell surface antigens and antibodies conjugated to suspended magnetic particles Compared to cell-affinity micro-chromatography, where the retrieval of captured cancer cells can be difficult, magnetic bead-based techniques readily permit the manipulation of captured cancer cells using local magnetic fields Table 1. Liu and Pang et al. demonstrated the first microfluidic device for isolating low abundance cancer cells from a red blood cell (RBC) suspension using magnetic cell separation Figure 1(A). In this system, a hexagonal array of nickel micro-pillars was integrated onto the bottom of a micro-fluidic channel and used to generate magnetic field gradients to efficiently trap super paramagnetic beads. The trapped magnetic beads functioned as a capture zone, followed by in situ chemical and biological modifications to functionalize the surface of beads with specific antibodies. Based on the interaction between the specific antibodies and N-acetylglucosamine on the cell membrane, A549 cancer cells spiked in RBCs were effectively captured and sorted on the microfluidic device with a capture rate between 62% and 74%. Antibody-coated magnetic beads were also used in a micro-fluidic device for the serial selection of cell subpopulations. As illustrated in Figure 1(B), this separation system consists of two separate compartments, each containing magnetic beads functionalized with different surface membrane protein receptors specific to prostate cancer cells (PSMA and CD10). As a cell suspension is introduced to the first array, the cells expressing CD10 are immobilized onto the magnetic beads while CD10 cells pass through this chamber and into the second compartment. PSMA+ cells bind to the magnetic beads located in the second compartment after which the remaining cells are flushed from the system. Thus, PSMA+/CD10 and CD10+ prostate cancer cell subpopulations can be isolated.

In order to further increase the surface-to-volume ratio of magnetic beads for cell sorting, Saliba and Viovy et al. developed a method using columns of bio-functionalized super-para-magnetic beads self-assembled in a microfluidic channel. In this system, a hexagonal array of magnetic ink was first patterned at the bottom of microfluidic channels. Beads coated with anti-bodies were then injected into the channel and allowed to settle down. Upon application of an external vertical magnetic field, the magentic beads assembled on top of the ink dots to form a regular array of columns.

Tests using cell line mixtures demonstrated a capture recovery rate greater than 94% and the capability to cultivate the captured cells on chip. Furthermore, clinical samples (blood, pleural effusion, and fine needle aspirates) from healthy donors and patients with B-cell hematological malignant tumors were analyzed in the microfluidic chamber. Lien and Lee reported Multi-functional, integrated microfluidic devices capable of cancer cell separation, cell lysis and genetic identification. This platform consisted of an incubation module where target cancer cells are selectively captured onto functionalized magnetic beads, a control module for sample transportation, and a nucleic acid amplification module for cell lysis and genetic identification Figure 1(C). Cancer cells (e.g., lung and ovarian carcinoma) were spiked into whole blood samples and loaded into the incubation chamber with pre-loaded magnetic beads coated with monoclonal antibodies. The cancer cells were specifically immobilized onto the surface of the magnetic beads with a recovery rate higher than 90%.

Differences in cell size can be exploited for microfluidic cancer cell selection without the knowledge of target cells’ biochemical characteristics. Size-based cell separation is attractive, for instance, for capturing CTCs since these cells are much larger than other cells found in whole blood Table 2. Mohamed et al. reported the first size-based microfluidic cancer cell separation device which featured on-chip micro-filters. The device consisted of four regions with decreasing channel widths (20 mm, 15 mm, 10 mm, and 5 mm) and a constant channel depth (20 mm). Cultured neuroblastoma cells mixed with whole blood were injected into the device where the 10 mm wide channels trapped the cancer cells. Zheng and Tai et al. developed a parylene membrane micro-filter device with circular holes (10 mm diameter) with a center to center distance between adjacent pores of 20 mm.

The size difference between CTCs and human blood cells was exploited to test 57 blood samples from patients with metastatic prostate, breast, colon, or bladder cancer. The results demonstrated CTC capture and identification in 51 of 57 patients compared with only 26 patients in 57 patients using the cell membrane during the trapping process, and the device enabled via CTC conventional Cell Search method. However, this process resulted in low capture cell viability due to the large stresses that developed in the cell membrane during the cell capture process. Zheng and Tai et al. further developed a double-membrane device to decrease stresses experienced by capture Figure 2(A). In this device, a second porous membrane was incorporated below the first membrane. The por positions between the two membranes were intentionally misaligned. This bottom membrane provided support for the trapped cells to effectively reduce flow-induced stress on the cell membrane. Tan and Lim et al. developed a microfluidic device with multiple arrays of crescent-shaped wells Figure 2(B) to isolate cancer cells from spiked blood [68] and patient whole-blood samples. Gaps (5 mm) were made within each of the crescent-shaped traps to ensure the complete removal of other blood constituents due to their ability to traverse narrow constrictions. After cancer cell capture, a reverse flow was used to retrieve the captured cancer cells from the device. Isolation efficiencies higher than 80% were achieved for breast and colon cancer cell lines. In addition, this device was able to successfully detect and retrieve CTCs from the peripheral blood of patients with

metastatic lung cancer. Di Carlo et al. utilized microscale laminar vortices combined with inertial focusing to selectively isolate and trap larger cancer cells spiked into whole blood while smaller blood cells were flushed out of the device Figure 2(C). Multiple micro scale laminar vortices were created on chip with processing rates as high as 7.5 × 10⁶ cells per second. The reported cell recovery rates for these devices were 23% for MCF-7 cells and10% for HeLa cells.

Dielectrophoresis (DEP) uses the polarization of cells in non-uniform electrical fields to exert forces on cells. DEP forces depend on factors such as cell membrane and cytoplasm electrical properties as well as cell size.

DEP devices have been developed for separating cancer cells Table 3, based on differences in cells’ response to electric fields [90] . Becker and Gascoyne et al. reported the first dielectric affinity column Figure 3(A) for cancer cell separation in which human leukaemia cells suspended within normal blood cells were retained on microelectrode arrays while normal blood cells were eluted [91] . The cancer cells were subsequently released for collection by the removal of the DEP field. Becker and Gascoyne et al. further demonstrated the applicability of this method for the separation of epithelial cancer cells (MDA-231 cells) from diluted blood and reported a recovery rate of 95% [92] [93] .

Gascoyne et al. proposed DEP flow-field fractionation (DEP-FFF) wherein DEP forces are generated to levitate suspended cells to different equilibrium heights within amicrofluidic chamber, based on variations of cells’ electrical properties [94] . The levitated cells are transported at different flow velocities upon the application of fluid flow Figure 3(B). Using this approach, human leukemic (HL-60) cells, 99, 106 MDA-435 cells, 101, 102 MDA-468 cells and MDA-231 cells 113 were successfully separated from background cell populations. To enhance sorting sensitivities, a 3D-asymmetric microelectrode setup was developed for cancer cell separation Figure 3(C). An alternative method for separating cancer cells has been demonstrated by combining multi-orifice flow fractionation (MOFF) with DEP.

Figure 3(D), when cell samples were introduced through the inlet, most of the blood cells were separated via

MOFF and extracted through outlet I while MCF-7 cells with residual blood cells (not fully separated) proceeded to the DEP separator. At the DEP separator, cancer cells exited through outlet II while the residual blood cells passed through outlet III. Since the DEP technique leverages differences in both cellular size and dielectric properties, it could potentially lead to a higher cancer cell separation yield and purity compared to micro-filtra- tion methods that are based on cell size differences only. However, in practice, due to the limited dielectric differences between target cells and carrier cells, this technique’s yield and purity are not as high as expected in Table 3. Among the detection techniques discussed in this review, on-chip DEP is the only technique that has not yet undergone verifications with clinical samples. Thus, an approach that utilizes a combination of multiple cell-capture methods may prove viable for improving the performance of cancer cell capture devices. For example, to improve device selectivity and cell-capture efficiency, one may envision a multi-module microfluidic system for cancer cell capture in which the first module performs high-throughput concentration and purification of target cells while a second module enables the selective capture of cancer cells. Such a device can be realized by integrating DEP with cell affinity micro-chromatography, such as for CTC detection. The DEP module would function as a pre-concentrator to increase the concentration of CTCs by flushing samples through channels patterned with electrodes. The concentrated samples would then enter the cell affinity micro-chromato- graphy module for high-purity CTC capture.

This review summarized the working principles and experimental results of key microfluidic technologies for cancer cell separation and detection. These microfluidic devices are based on magnetic activated micro-cell sorting, size-based microfluidic separation, and dielectrophoresis. Despite the recent technological advances, the development of a single device capable of simultaneously achieving high throughput, high target cancer cell recovery, high purity, and high cell viability remains challenging. Magnetic activated cell sorting readily permits the manipulation of captured cancer cells by controlling local magnetic fields for post-capture processing. Lien and Lee et al. proposed a multi functional, integrated magnetic bead-based microfluidic device capable of cancer

cell separation, cell lysis, and genetic identification. Microfiltration methods also permit easy retrieval of captured cancer cells, as demonstrated by Tan and Lim et al., using a reverse flow to release captured cancer cells in multiple arrays of crescent-shaped wells. However, on-chip post-capture processing capabilities have yet to be developed.

However, existing systems are only capable of processing small numbers of cells within a reasonable time frame. For example, the total number of cells tested by the optical stretcher was 36 for MCF-10, 26 for MCF-7, and 21 for Mod-MCF-7.81. Reported electrical impedance spectroscopy differences of head and neck cancer cell lines with different metastatic potentials (686LNvs.686LN-M4e) were also based on the testing of low sample numbers (n = 72 for the 686LN-M4e cell and n = 57 for the 686LN cell). Furthermore, most microfluidic devices to date have been only capable of characterizing a single biophysical parameter.

The broad spectrum of cell separation technologies described in this review illustrates the high level of interest and activity in this area. The described size based approaches offer a great potential for separation of cell subpopulations for which specific markers are not known or cannot be used (e.g., to prevent cell activation). Affinity-based approaches (magnetic and electrophoretic) can be employed for fast (~minutes) and continuous separation with high specificity (~99%). For all of the approaches, the design of the devices is such that they can be operated in a massively parallel fashion to increase scale and throughput without compromising purity and efficacy; although each technique has some limitations e.g. in sized based, the probability of cells damages is high and it is considered as nonspecific technique. The magnetic activated micro-cell sorters are easy to develop but the time required for screening the tumor is relatively long. Regarding DEP, the need to control precisely laminar flow conditions and the electric field frequencies is a more critical point. A challenge for microfluidic cancer cell’s biophysical characterization is existing devices that have low sample throughput. To obtain clinically relevant information, these devices must be able to measure biophysical properties of a large number of cells with true high throughputs.

This work was done by the Nanotechnology Research Group in School of Life Science, Beijing Institute of Te- chnology.