In-flow measurement of cell-cell adhesion using oscillatory inertial microfluidics

Abstract

Multicellular clusters in circulation can exhibit a substantially different function and biomarker significance compared to individual cells. Notably, clusters of circulating tumor cells (CTCs) are much more effective initiators of metastasis than single CTCs, and correlate with worse patient prognoses. Measuring the cell-cell adhesion strength of CTC clusters is a critical step towards understanding their subsistence in the circulation and mechanism of elevated tumorigenicity. However, measuring cell-cell adhesion forces in flow is elusive using existing methods. Here, we report an oscillatory inertial microfluidics system which exerts a repeating fluidic force profile on suspended cell doublets to determine their cell-cell adhesion strength (F_s), without any biophysical modifications to the cell surface and physiological morphology. Using our system, we analyzed a large number (N > 500) of doublets from a patient-derived breast cancer CTC line. We discovered that the cell-cell adhesion strength of CTC doublets varied almost 20-fold between the weakly adhered (F_s < 28 nN) and strongly bound subpopulations (F_s > 542 nN). Our system can be used with other cancer or noncancer cells without restrictions, and may be used for rapid screening of drugs aiming to disrupt the highly-metastatic CTC clusters in circulation.

Graphical Abstract

Cell-cell adhesion strength of freely suspended cell clusters can be measured using an oscillatory inertial microfluidic system.

Introduction

Blood cells (e.g. erythrocytes, leukocytes) and other cells (e.g. endothelial, progenitor, tumor-originated) found in the circulation take part in a range of physiological and pathological processes. While these circulating cells commonly exist as individually suspended cells in blood, they can also be found as doublets, triplets or larger clusters. Recent studies show that the mere existence and the frequency of these clusters can be an important disease biomarker. Furthermore, being in a cluster can also significantly affect the function, potency or survival of the cells in circulation. For example, circulating tumor cells (CTCs) and CTC clusters drive metastasis by getting into the bloodstream from a primary tumor, and reaching distant organs via circulation where they extravasate and colonize ¹. Recent studies have shown that CTC clusters are substantially more (~50 to 100-fold) potent initiators of metastasis compared to single CTCs ^2-4. On the other hand, neutrophils are also shown to pair with CTCs in circulation, and enhance their tumorigenicity ⁵. Since metastasis is responsible for the majority of cancer related deaths, inhibiting metastatic occurrence is a primary goal for cancer therapies ⁶. Thus, the adhesion interactions between CTCs or CTCs and neutrophils can be viable therapeutic targets for inhibiting metastasis. Tumor originated endothelial cell clusters were also found in the blood of colorectal cancer patients, which could provide information about the mechanobiology of the tumor endothelium ⁷. Circulating cell clusters are not limited to tumor associated cells, for instance T cell-monocyte complexes are also found to be associated with immunological perturbations ⁸. Consequently, characterization of cell-cell adhesion interactions, especially when they are in suspension, is critical for understanding the survival of cell clusters in circulation. For CTC clusters, these measurements can reveal more subsistent (and potentially more metastatic) subpopulations, and establish a baseline for antimetastatic drugs which can disrupt and dissociate CTC clusters in circulation ⁹. Collectively, studying cell-cell adhesion interactions in flow can lead to significant scientific and clinical outcomes.

Characterization of the cell-cell adhesion forces in flow is not currently possible using existing methods. In the case of CTC clusters, after intravasation into a blood vessel, they travel in suspension along with other blood components, while accelerating and decelerating as they pass through various sections of the circulatory system. During this travel, they are not attached to a surface and exposed primarily to fluidic stresses, except for the capillaries which they are able to transit through ¹⁰. Considering this physiological environment, force characterization methods which require cells to be attached to a functionalized surface (e.g. shearing flow over surface, atomic force microscopy probes), or deform their morphology during testing (e.g. micropipette aspiration) are not physiological because the cells are forced to alter their cytoskeletal organization during measurements. Alternatively, force spectroscopy methods which can manipulate cells with minimal modifications to cell structure (e.g. optical or magnetic trapping) are conducted in stagnant fluids, limited to pico-Newton (pN) levels of forces, and low throughput ¹¹. Throughput is especially important for the characterization of clusters from patient derived CTC lines as CTCs are shown to exhibit significant phenotypic variations, thus need to be analyzed in large numbers to obtain the full biophysical landscape of the metastatic process ^12,13. As a result, a method which can characterize the cell-cell adhesion forces in flow, while preserving cell morphology, and at high throughputs is highly desirable.

In this study, we developed a microfluidic system which exerts a predetermined fluidic force on freely suspended two-cell clusters (i.e. doublets) in a cyclic fashion, in the range of hundreds of nano-Newtons (nN). The applied force is the result of the differential drag force acting on the cells of a doublet in a rapidly accelerating fluid, as it travels through a flow constriction. The microfluidic chip is designed to achieve this rapid flow acceleration, while also limiting the spatial distribution of the doublets in the channel via a combination of oscillatory inertial focusing and geometric constraining ¹⁴. A pressure-driven flow control system is developed to create an oscillating flow field with a net forward flow bias, which enables: (i) upstream cell focusing, and (ii) doublets experiencing the fluidic stress for a preset number of times at the constriction. The system is coupled with a high-speed camera, which allows determination of the dynamic parameters such as the position and velocity of doublets, as well as tracking of the stress cycles and eventual cell dissociation. This microfluidic platform is a unique tool to assess cell-cell adhesion strength in flow, with minimal effect on the cell surface and morphology during testing.

We used the microfluidic platform to characterize doublets from a previously established breast cancer CTC line (BRx-142) ¹². Biophysical characterization of cultured CTC clusters is an important application as it can enable the development of personalized antimetastatic therapies by targeting the cell-cell adhesion in a patient’s clusters. Our analysis with a large number of CTC doublets (N > 500) showed that when the applied separation force ranged from 28 nN to 271 nN, from ~5% to ~70% of the CTC doublets were dissociated. Due to the observed heterogeneity in the adhesion strength of the doublets within the population, we determined the separation force required to dissociate 50% of the CTC doublets as: F_s,50 = 173 nN. Surprisingly, ~30% of the CTC doublets remained united, even after separation force (F) was increased to 542 nN, indicating a robust subpopulation of doublets with stronger adhesion. Finally, by analyzing the CTC doublets under different levels of exerted stress we determined that their dissociation mode can be both gradual (over multiple cycles) or abrupt (over a single cycle), but transitions to the latter mode as F increases.

Experimental

Oscillatory inertial microfluidic system

The pressure/flow control system is designed to generate an oscillatory flow field with a forward bias, such that the flow has a net forward motion after each oscillation cycle. This is achieved by using two controlled pressure sources (Flow-EZ, Fluigent, France) and a high-speed, three-way valve (LHDA0533315H, The Lee Company, CT, USA). First pressure source (P₁) is connected to the doublet suspension, which is directly connected to the microfluidic chip. Therefore, the CTC doublets do not travel through the valve (or any other high fluidic stress region) before the measurements are taken, in order to eliminate any potential adverse effects on doublet integrity. The second pressure source (P₂) is set to: P₂ ≈ 2P₁, and is connected to a cytocompatible buffer solution (phosphate buffered saline) which is connected to the three-way valve. Thus, as the three-way valve alternates between two connectivity modes, the flow in the microchannel is oscillated under a net pressure difference ∆P₁. The valve is driven by a square signal with a frequency between 5 Hz and 10 Hz. The flow frequencies are selected to ensure that the doublets remain in the microchannel during measurements, and pulsatile effects can be neglected. The forward bias in the flow is achieved by either adjusting the duty cycle of the signal (such that the valve is ON in the forward flow direction slightly longer), or by slightly reducing P₂ (which results in a slightly higher pressure in the forward direction). For the high gradient chip (Chip-HG), the width (W_C) of the constriction and the height (H) of the microchannel were selected as H = W_C = 30 μm; and for the low gradient chip (Chip-LG), W_C = 40 μm was selected. For both chips, the width of the large section of the microchannel was: W_L = 300 μm. All microfluidic PDMS devices were fabricated using standard soft lithography techniques ¹⁵.

Experimental procedure and data collection

BRx142 CTC line was maintained as previously described ¹². Cultures were routinely checked for mycoplasma with the MycoAlert kit (Lonza, Switzerland). After harvesting the cells, the concentration of clusters in the suspension were characterized using a standard hemocytometer chip. Typically, cell doublets constituted approximately 10% of the suspension. After determining the concentration, cluster suspension and phosphate buffered saline were connected to the pressure control system which was used to drive the flow and the counterflow required for the oscillating flow field. The components of the system were connected to each other via Tygon Tubing (Cole Parmer, USA). Videos of doublets were obtained by recording with a high-speed camera (Phantom 4.2, Vision Research Inc.) at a high frame capture rate (10,000 fps) for flow velocimetry experiments and at a lower rate (500 – 1,000 fps) for cell-cell adhesion characterization experiments to enable longer observation and data acquisition times. After obtaining the videos, they were analyzed using an image processing algorithm to determine which frames contained doublets, and evaluate their geometric and dynamic properties. Subsequent visual analysis was also performed to track individual doublets and verify their fate under stress cycles.

Results

Oscillatory inertial microfluidics system and chip design

The microfluidic system exerts a predetermined cyclical separation force on the suspended CTC doublets, while tracking their morphology and eventual dissociation as they go under the stress cycles. The passive separation force (F) is exerted on the cells by the surrounding fluid, which is the result of the differential drag force arising in a rapidly accelerating flow. Magnitude of this force is adjusted by the combination of the input pressure/flowrate and the geometry of the microfluidic chip. Repeated application of the stress cycles is achieved by applying an oscillatory flow field with a forward bias, using a pressure control system and a high-speed microfluidic valve. Oscillatory inertial focusing of the doublets and dimensional constraints on the chip geometry ensures uniform exertion of the force profile. Overall, the microfluidic platform consists of a pressure/flow control system, a microfabricated chip, and a microscope coupled with a high-speed camera for recording cell morphology during stress cycles (Figure 1A and Figure S1).

Figure 1 – Overview of the microfluidic system, operating principle and chip design:

A) System schematic showing the experimental setup, B) Illustration of the path of a CTC doublet during the forward (left to right) motion of a stress cycle and its subsequent dissociation, C) Microchip design and flow simulations demonstrating the drastic increase in the flow velocity which is the cause of the differential drag force acting to separate the doublet. (Top) Velocity gradient along the centerline of the chip. Maximum x gradient is obtained at x_max,du. (Bottom) Velocity field at the mid z-plane of the channel. D) Image sequence of a real CTC doublet (boxed) travelling in the microchannel. In its forward pass, the doublet rotates and accelerates as it passes through the constriction but retain its integrity. During its backward pass, it dissociates around the point where the flow acceleration and the resulting separation force is maximum.

The microfluidic chip consists of a constriction section where the flow is rapidly accelerated, and a wide upstream section where the doublets are inertially focused (Figure 1B). At the constriction, the width of the microchannel is drastically reduced to attain rapid acceleration of the flow in a very small length-scale comparable to the size of a single cell. The dimensions of the constriction are selected such that the cells are not physically deformed by the walls of the channel while being constrained in their geometrical position and orientation. The constriction is also symmetrical so that the doublets experience the identical stress in both flow directions. The wide upstream section is designed to enable oscillatory inertial focusing of the doublets before they reach the constriction, while exerting minimal fluidic stress on the cells. Upstream cell focusing is necessary to minimize the spatial variability of the doublets in the microchannel, specifically during their travel through the constriction.

Rapid and drastic acceleration of the flow allows the doublets to be exposed to a significant separation force at the constriction section, while ensuring that the stress levels upstream are negligible. At the constriction section, x gradient of the x-component of velocity (u), du/dx, increases several orders of magnitude (Figure 1C). This rapid acceleration is the fundamental principle that causes the separation force (F) on the doublets. As a cell doublet passes through the constriction, the drag force on the leading cell is higher than the drag force on the trailing cell, because of the difference in the flow velocity. Consequently, the maximum magnitude of the differential drag force is attained at the x-position corresponding to the maximum acceleration of the flow (x_max,du). This position is located slightly before the narrowest section of the constriction, as expected.

In the experiments, CTC doublets are loaded to the microfluidic system as shown in Figure 1A, and the forward bias of the flow is adjusted such that each doublet passed through the constriction a fixed number of stress cycles. We observed that the doublets: (i) inertially focus upstream, (ii) rotate to a horizontal orientation and accelerate as they approach the constriction, and (iii) retain their integrity or dissociate into single cells as they are passing through (Figure 1D). We were able to characterize, track the stress-cycles and the eventual fate (i.e. dissociated or retained integrity) of every unique CTC doublet using the integrated high-speed camera and analysis of the post-processed videos. Using this analysis, we determined the adhesion strength (F_s) of several hundred CTC doublets under flow conditions.

Flow velocimetry analysis of inertially focused CTC doublets

We first characterized the geometric and dynamic properties of the CTC doublets, as they travelled through the constriction section of the channel. Limiting the analysis to CTC doublets (two-cell clusters) ensured that the cells were not geometrically constrained by the channel walls through any section of the microchip, thus the force acting on the cells was only the result of the fluidic stresses. We recorded high-speed (~10,000 fps) videos of the CTC doublets while going through the stress cycles, and determined their (i) position in the microchannel, (ii) velocity, and (iii) cell diameter and horizontal orientation (i.e. the angle between the CTC doublet and the x-axis). During these experiments, the applied differential pressure to the system was P = 2.9 psi, which resulted in an upstream channel Reynolds number of Re = 8.9, and particle Reynolds number of Re_p = 0.72.

Dynamic characterization results verified that the CTC doublets are laterally centered, horizontally oriented and have uniform dynamic properties as they pass through the constriction. The lateral (y-axis) position of the CTC doublets was consistently in the middle of the channel along the x-axis, due to the upstream inertial focusing (Figure 2A). The velocity profile of the doublets was also consistent along the x-position of the constriction, as a result of their positioning on proximate streamlines. The position and velocity of the doublets where the maximum separation force was applied (at x_max,du) was determined as: y = 149.7 ± 6.4 μm and u = 0.39 ± 0.038 m/s (all indicated errors are standard deviation). Similarly, we also determined the diameter and the horizontal orientation of the doublets while they are passing through the constriction. The diameter of the CTC doublets was evaluated as: D = 16.4 ± 2.01 μm, which is consistent with the design criteria used for designing the channel dimensions. Finally, we observed that the orientation of the doublets was close to horizontal alignment, and their angle with the x-axis was determined as: θ = 1.2 ± 10.9°.

Figure 2 – Flow velocimetry analysis of CTC doublets (N > 20):

A) Lateral (y-axis) position graph showing uniform positioning of the doublets via inertial focusing (constriction region of the microchip is shown in the background for easier visualization of the doublet positions), B) Velocity of the CTC doublets as the pass through the constriction. (Box plots show inter-quartile ranges.)

Oscillatory inertial focusing of the doublets enables them to have uniform dynamic properties during the stress cycles. In the upstream section of the channel where the width is an order of magnitude greater than the doublet size, focusing is achieved by extending the travel length via the oscillatory flow field. Note that this is not practically achievable by steady-flow inertial focusing in a straight channel as the ratio of the particle size (a) to focusing dimension (H) is: a/H < 0.1 and the particle Reynolds number is small (Re_p < 1). Therefore, oscillatory flow field is not only critical for repeated exertion of the stress, but also for focusing the CTC doublets. It should be noted that the attained focusing width of the doublets is slightly broader than the typical focusing width obtained with single cells. This is due to the fact that single cells have spherical symmetry, whereas the CTC doublets do not. Consequently, inertial lift forces on doublets vary based on their rotational position, as previously reported ¹⁶. Nevertheless, our results show that the variation caused by this effect is sufficiently small, and has minimum impact on the dynamic properties of the doublets passing through the constriction.

Separation force evaluation using computational fluid dynamics simulations

The separation force (F) applied to the CTC doublets is evaluated using experimental data and computational fluid dynamics (CFD) simulations. The fluidic drag force exerted on a suspended particle in flow can be determined by inserting the geometric and dynamic properties of the system into a CFD model. In our setup, these properties are determined by analyzing high-speed videos via standard image processing methods for flow characterization. Then, these properties are implemented in a CFD model using a commercially available software (COMSOL Multiphysics), and the solution for the system is obtained. An important decision in this evaluation is the compromise between model accuracy and analysis time. While it is theoretically possible to run a CFD simulation for each geometric and dynamic parameter combination, the required computational processing time makes it impractical for analyzing a large number of CTC doublets in high throughput. Thus, we initially verified that the variations in the geometric and the dynamic properties of the system on F is minimal. Then, we proceeded with developing a doublet model based on the ideal and/or median values of the dynamic properties of the system, and compared the model approximation with the exact solutions to individual cases.

Initially, we determined the sensitivity of the separation force (F) profile to the variations in the geometric and dynamic properties observed in our system. We evaluated the F on the CTC doublets using the exact parameters extracted from each recorded high-speed frame during the cycle, which yields the highest accuracy estimation of the separation force. Specifically, in the CFD model, the surface of the CTC doublet was set as a moving wall boundary condition based on the experimentally recorded velocity of the doublet, and the net force acting by the surrounding fluid was evaluated by integrating the pressure and viscous stress on the surface of the individual cells. Initially, the CTC doublet was placed at its true position and orientation on the xy-plane, and at the middle of the channel on the z-plane (Figure 3A, light blue circles), and its velocity was set based on experimental measurements (see the actual images of a CTC doublet in the microchannel and the corresponding CFD model below, Figure 3A-right panel). Then, the same evaluation was repeated where the doublet was moved to the bottom (5 μm shift in z-direction) of the z-plane (Figure 3a, dark blue circles), to assess the effect of uncertainty on the z-axis position (i.e. channel height). Later, the same unique CTC doublet was evaluated during its second and third stress cycles, in which its position and orientation was slightly different (Figure 3a, blue triangles and squares). Lastly, we repeated the same evaluation with two different CTC doublets (Figure 3a, green and orange circles).

Figure 3 – Separation force evaluation:

A) Separation force exerted on the CTC doublets calculated for individual cases with high-accuracy models (individual dots), and an ideal model approximation (solid line). Right panel shows the actual images from an experiment on top, and its high accuracy CFD model at the bottom, B) Attained force profiles with the changing input pressure to the microfluidic chip (Chip-HG), C) Peak applied separation force resulting from different applied pressures to the high and low gradient microfluidic chips. Chip-HG exerts a higher separation force to the doublets under the same pressure compared to the Chip-LG because of its higher acceleration gradient.

After obtaining the individual solutions for different stress cycles, we developed an approximation model by implementing a combination of the ideal and/or the median values of the geometric and dynamic properties of the doublets. Specifically, we modeled a CTC doublet with the median cell size (D = 16.4 μm), and set it to follow the experimentally measured median velocity profile (Figure 2B) along the channel. For the position and orientation, we used the ideal values because the experimentally determined median values (y = 149.7 μm and θ = 1.2°) were sufficiently close. Namely, we placed the model doublet at the center of the channel (y = 150 μm) with a horizontal orientation (θ = 0°). Based on these geometric and dynamic properties, we resolved the net separation force acting on the model CTC doublet at several locations along the x-position of the constriction (Figure 3A, solid line).

Our results show that the exerted force profile is uniform among observations, and it can be approximated by an ideal model. During the first stress cycle of the first doublet, at its closest point to maximum, F was evaluated as: 262 nN and 214 nN for the middle z-plane and bottom z-plane cases respectively (Figure 3A). In comparison, at the same x-position, F estimated by the ideal model was 264 nN. These evaluations suggest that the error of the ideal model is less than 1% when the doublet is at the middle of the z-plane, and approximately 18.9% when the doublet is at the bottom of the z-plane. Thus, even though there is an uncertainty in the z-position of the doublet in the microchannel, its effect on F is small by the virtue of the limited channel height (H = 30 μm). Thus, we proceeded with middle z-plane assumption for the remainder of the calculations. During the next two cycles, F was evaluated as 217 nN and 251 nN at points near peak separation force. At the same x-positions, the ideal model estimation for the separation forces were: 198 nN and 246 nN respectively. Note that the ideal model diverges from the observations primarily near the left and right edges of the constriction. This is due to the mismatch between the horizontal orientation of the actual doublets and the model, because the model is always oriented horizontally whereas the real doublets rotate. Since the model accuracy is mostly important close to the constriction region where F is high, we determined that the deviation close to the edges are not consequential.

Using the ideal model, we were able to evaluate the magnitude of the separation force profile with changing applied pressure, and the corresponding velocity and acceleration profile. The applied pressure in the experiments ranged from 1.45 psi (100 mbar) to 11.6 psi (800 mbar), which corresponded to a flowrate (in forward or backward direction of the oscillation) of Q = 8 μL/min to 65 μL/min. The force profiles obtained under different flow fields were calculated based on the ideal model (Figure 3B). In order to increase the dynamic range of the exerted separation force under similar pressures, we also used a secondary chip (Chip-LG). In this secondary low gradient chip, the acceleration of the flow is designed to be more gradual at the constriction, thus the magnitude of the acceleration profile (Figure S2) is lower than the high gradient chip (Chip-HG). Specifically, at its maximum, x-gradient of the velocity is equal to 1.3 x 10⁴ s⁻¹, which is lower than the corresponding value (2.0 x 10⁴ s⁻¹) in the Chip-HG. Therefore, under the same applied differential pressures, the peak applied force of the Chip-LG is also lower than Chip-HG (Figure 3C).

Characterization of the adhesion strength and dissociation modes of CTC doublets

The adhesion strength of the CTC doublets was characterized by applying a predetermined separation force profile for a set number of stress cycles, and evaluating the consequent integrity of each doublet. We tested two microfluidic chips with high and low velocity gradients (Chip-HG and Chip-LG) under the same differential pressures, and determined the required force to break the cell-cell adhesion forces. We opted to evaluate the CTC doublet strength (F_s) in terms of the fraction of the dissociated doublets under the peak applied force of a stress cycle (Figure 4A). We selected lower and upper limits of the applied force as: 28 nN – 227 nN and 67 – 542 nN for Chip-LG and Chip-HG respectively. At the lowest applied force (28 nN) using Chip-LG, only 5.9% of the doublets were dissociated. When we doubled the applied force (56 nN) on Chip-LG, dissociated doublets increased to 20%. On the other hand, when the applied force on Chip-HG was comparable to Chip-LG (67 nN), the fraction of dissociated doublets was also similar (18%). The fraction of dissociated doublets increased rapidly with increasing force for both chips, until it reached to 64% to 70% of the population (at F = 227 nN and 271 nN for Chip-LG and Chip-HG respectively). These observations confirmed that the peak applied force is a suitable metric for characterizing cell-cell adhesion strength, as we obtained comparable results from both low and high velocity gradient microchips. Increasing the applied force beyond F = 271 nN resulted in only a marginal change in the number of dissociated doublets, reaching to 72% dissociation of the population at F = 542 nN.

Figure 4 – Cell-cell adhesion strength analysis of CTC doublets (N > 500):

A) Fraction of the dissociated doublets with respect to the peak applied force obtained from two different (low velocity gradient: Chip-LG and high velocity gradient: Chip-HG) microfluidic chips. B) Number of stress cycles resulting in doublet dissociation at low (F ≤ F_s,50) and high (F > F_s,50) levels of exerted separation force.

Our results show that the cell-cell adhesion strength among the CTC doublets used in this study (BRx-142) is highly heterogeneous. At the lower limit of the tested force range where F was 28 nN, we observed only 5.9% doublet dissociation, indicating a fraction of the population with a relatively weak adhesion strength. In contrast, at the higher limit of the force range where F equaled 542 nN (~20x lowest applied force where we observed dissociation), 30% of the doublets still remained undisrupted. This indicates a strongly adhered subpopulation of doublets which has an adhesion strength close to the milli-Newtons range. Due to this heterogeneous distribution, it was not possible to determine an adhesion strength (F_s) applicable to the majority of the population. Thus, we determined F_s,50 instead, defined as the force required to dissociate half of the population. In order to determine F_s,50, we fitted a linear curve with the equation: y = ax+b to our data (where y is the percent dissociated doublets and x is peak separation force) using a standard least-squares regression method. For linear curve fitting, we only included the data where increasing the separation force correlated with dissociated doublets (i.e. F ≤ 271 nN). Based on the fitted curve (a = 0.264, b = 4.316, R² = 0.87), F_s,50 was determined as 173 nN.

We observed both abrupt and gradual dissociation modes in the lower and higher ranges of the exerted separation force. Our system enables tracking of individual CTC doubles going through the stress cycles; thus we were able to quantify the exact number of stress cycles that the doublets experienced before dissociation. During the experiments, the forward bias of the flow was adjusted such that each CTC doublet goes through at least five full stress cycles before leaving the constriction region. When the applied separation force was low (F < F_s,50), the dissociation happened over several stress cycles more frequently (Figure 4B). This suggests that at each passage, only a portion of the molecular adhesion interactions were broken until the cells were fully dissociated. In contrast, when the applied force was high (F > F_s,50), most of the doublet dissociations happened within one full stress cycle (i.e. one forward and backward passage through the constriction), indicating an abrupt separation. The evidence of the gradual separation mode is important because the CTC clusters experience fluidic stresses repetitively while they are in circulation. Thus, it is possible that even though a single exertion of a fluidic stress is not sufficient to dissociate a CTC cluster, repeated exertion may lead to the generation of single CTCs from the cluster.

Discussion

Our microfluidic platform enables rapid characterization of cell-cell adhesion in flow, and is distinctly different than other force spectroscopy technologies. Our system’s capabilities include: (i) fluidic force exertion with no disruption to physiological cell morphology, (ii) in flow measurement of cell-cell adhesion, (iii) applied force in the range of tens to hundreds of nN, and (iv) high throughput observations. Currently, optical tweezers are the benchmark technique for force spectroscopy applications ¹⁷. It is well acknowledged that high intensity radiation in optical traps can cause photodamage in trapped cells, including opticution (i.e. cell death) ^18-20. These studies have revealed that the photodamage can be attributed to the photochemical generation of reactive oxygen species. Importantly, there is no damage threshold of laser power in optical tweezers and cells start to experience the photodamage from the moment they are trapped ¹⁹. On the other hand, dual pipette aspiration techniques significantly disrupt the cell membrane and can cause separation of the membrane and the cortex ^21,22. It is currently unknown to what extent these damages affect the force measurement ²³. Our system introduced in this work does not suffer from such artifacts, preserves cell morphology during measurement, and provides sufficiently high throughput to capture the observed heterogeneity of the CTC doublets. Currently, our system can record observations from up to ~100s of doublets/min, but our throughput is significantly lower because of the labor-intensive, manual downstream analysis for tracking unique doublets travelling in an out of the field of view. We anticipate that this limitation could be overcome by developing and implementing a machine-learning based, fully automated image processing algorithm for video analysis, which would bring the overall system throughput closer to flow cytometry applications. Overall, we expect that this system will be a uniquely useful tool for studying various physiological and pathological processes involving cell-cell adhesion of circulating cell complexes, including cancer metastasis.

Our results show that the cell-cell adhesion strength of CTC doublets is extremely heterogeneous, and CTC doublet dissociation can occur either abruptly or gradually over several stress cycles. For the majority (~70%) of the doublet population, the adhesion strength (F_s) of CTC doublets varied between 28 nN to 271 nN. These results are comparable (within one order of magnitude) to previously reported results from different cell lines. Daoudi et al. used a micropipette aspiration method to measure the cell-cell adhesion strength of various human embryonic kidney (HEK) cell clones, and determined them to be varying between 2 nN to 12 nN ²⁴. Chu et al. used a similar micropipette aspiration technique and evaluated the cell-cell adhesion strength of murine sarcoma S180 doublets and their stronger (transfected to express E-cadherin) clones as 50 nN and 350 nN respectively ²⁵. Importantly, we also identified a robust subpopulation (~30%) of CTC doublets with a significantly higher adhesion strength (F_s > 540 nN), which we were not able to dissociate using our system. CTC clusters are held together in circulation by adherens junctions, and the variable expression of adherens junction components, including plakoglobin, could contribute to the observed differences in the intercellular adhesion strength ². Previous findings using the same (BRx-142) and other established breast cancer CTC lines showed that different subpopulations existed within the CTC populations ^12,13, which were shown to have variances in their drug resistance and proliferative pathways. Our results reported here expand on these findings and demonstrate that they also have significant heterogeneity in their cell-cell adhesion strength.

Our results raise the intriguing hypothesis that some CTC clusters may have a better survival chance in the circulation based on their cell-cell adhesion strength, and consequently a higher tumorigenicity. Metastatic disease is clinically observed to be highly heterogeneous, where some lesions appear to be clonal (comprised of cells with identical mutations) and others appear to be oligoclonal (comprised of mixed cells with different mutations). CTC clusters with different genetic compositions could contribute to this variability by having a biophysical advantage for survival based on their intercellular adhesion strength. Specifically, the CTC clusters with weaker intercellular interactions may dissociate after a short time period in the circulation, while the strongly adhered clusters may subsist longer. Theoretically, for stronger CTC clusters, this would lead to more interactions with various tissues as a coherent cell cluster, leading to higher metastatic potential. While it is not known whether longer subsistence in circulation correlates with higher tumorigenicity of CTC clusters, previous reports suggest that cell-cell adhesion is an important target for inhibiting metastasis. In a previous study, it was shown that by knocking down an adherens junction (plakoglobin) in CTC clusters, significant suppression of distant metastases (80% reduction in lung nodules) was achieved while the primary tumor remained unaffected ². These observations suggest that understanding the intercellular adhesion forces within CTC clusters in different cancers is critical for reducing the risk of metastatic disease, and can potentially lead towards to better identification and targeting of more metastatic subpopulations of CTC clusters. It should also be noted that cell-cell adhesion strength is not the only biophysical factor affecting CTC cluster mechanobiology. For example, rearrangement of the cell-cell junctions in CTC clusters was shown to play an important role to enable clusters to pass through capillary-sized channels as a single file ¹⁰. Therefore, the interplay between the cell-cell adhesion strength and flexibility is also critical, as it may lead to more immediate arrest of clusters in the capillaries. We propose that further studies are warranted where designed clusters with phenotypic and genetic variations are cultured, characterized using a system such as ours, and subsequently sorted and evaluated for their tumorigenicity using animal models. Furthermore, the underlying biology and adhesion mechanisms which cause the heterogeneity in the intercellular adhesion strength need to be investigated, along with potential links between the CTC clusters’ intercellular adhesion strength and their ability to rearrange their cell-cell junctions.