Photoconversion and chromatographic microfluidic system reveals differential cellular phenotypes of adhesion velocity versus persistence in shear flow

Abstract

An integrated photoconversion and cell sorting parallel-plate chromatography channel enabling the measurement of instantaneous and average velocities of cells mediating adhesion in flow fields was engineered to study the mechanisms underlying adhesion to selectins by metastatic cancer cells. Through the facile enrichment of cells into subfractions of differing adhesive behaviors and a fluorescent velocity probe amenable to off-chip analysis, underlying, causal molecular profiles implicated in differing adhesive phenotypes of metastatic cancer cells could be interrogated. This analytical method revealed selectin-mediated rolling adhesion to be strongly associated with expression of selectin ligands, correlations that vary with ligand type and rolling velocity magnitude. Discrete selectin ligand expression profiles were also found to underlie persistent versus non-persistent adhesion on selectins, suggestive of divergent regulatory mechanisms. This integrated cell sorting and photoconversion microfluidic platform thus enables in vitro analysis and comparisons of adhesive phenotypes as they relate to mechanisms of cancer cell metastasis in the context of selectin mediated adhesion, revealing new insights into potential cancer dissemination pathways.

Graphical Abstract

Introduction

Systemic metastasis—a multistep progression in which cancerous cells disassociate from the primary tumor, infiltrate the circulatory system, leave the vasculature, and establish metastatic tumors in secondary tissues—is the cause of over 90% of all cancer-related deaths.^1–3 In order to travel to distant sites in the body during the process of metastatic cancer extravasation, circulating tumor cells utilize a highly orchestrated adhesion cascade that, like leukocytes, can involve slow rolling interactions with endothelial cells under hemodynamic forces.^4–8 This process is mediated by interactions between endothelial-presented selectins and selectin ligands [such as sialofucosylated CD44 variants and carcinoembryonic antigen (CEA)] present on the tumor cell’s surface.^9–11 The interaction of selectins with their ligands slows the cells to allow cell-cell signaling or firm adhesion to occur within in the dynamic fluid flow environment of the circulatory system, and thus represents a critical regulatory step leading to subsequent cell transmigration and eventual formation of distant metastatic tumors.^7,9,12–15 Identification of cellular attributes and/or pathways contributing to metastatic dissemination in the context of selectin-mediated adhesion are therefore of high interest for its potential to identify cancer cell-selective drug targets inhibiting cancer’s spread.^16–18

Although they are highly simplified compared to in vivo mouse models, engineered in vitro systems that mimic the in vivo microenvironment offer numerous advantages for the investigation of mechanisms underlying malignant progression in a highly controlled manner.^14,19 The application of microfluidics to such problems allows the interrogation of biomechanical effects on metastatic cell transport under defined cellular, molecular, and biophysical conditions.^12,20–22 For example, in vitro systems have been used to sense morphological cell changes in a hypoxic tumor microenvironment, to probe force interactions between cells and the extracellular matrix, and to elucidate the impact of hemodynamic flow induced shear stress on regulating mechanotransduction mechanisms of circulating cancer cells.^23,24 Further, and relevant to this work, they can be designed to recapitulate selectin-mediated cell adhesion in flow, permitting the examination of adhesion mechanisms, and when coupled with high speed video microscopy, can be used to visualize the adhesive behavior of cells in flow fields that simulate the microenvironment of the vasculature.^14,15,25,26 Despite their numerous advantages, approaches to date have been limited to indirect analysis of perturbation of associated pathways or drugs on measured adhesion quantity and quality. Few approaches have been described that allow cells to be recovered and analyzed, fewer still that enable high content analysis of metastatic cancer cell adhesion and expression phenotypes to enable mechanistic disease and drug response modeling.

To fill this technical gap, a cell sorting microfluidic was previously engineered by our group to fractionate cells based on their residence time in a selectin functionalized channel as a proxy for average velocity.¹⁵ This platform permits the separation and enrichment of subsets of cells exhibiting little to no adhesion versus those that are highly adhesive based on their elution time through the functionalized microfluidic for interrogation of their metastatic potential in vivo and profiling of cellular molecular profiles.¹⁵ This approach solely separated cells into fractions of cells exhibiting any adhesion versus that did not. A natural extension of this proof-of-principle would be to perform such analyses with finer resolution in elution times (e.g. fast versus slow rolling adhesion). The persistence of adhesion by metastatic colon cancer cells to selectins in shear flow is highly variable, however, complicating interpretation of elution time alone as a readout of cell adhesion phenotype^25,26 and the application of the cell adhesion chromatography system to analyze complex or even subtle adhesive mechanisms.

To establish a methodology to analyze cellular phenotypes associated with different quantities and qualities of adhesion in flow, we sought to reconcile and address the potential confounding effects of sustained versus diminished adhesion persistence on fractionating and profiling cell subsets with differing adhesive phenotypes among a heterogeneous population of metastatic cancer cells. To do so, a fluorescence-based velocity probe recently reported by our group as enabling the off-chip analysis of velocities with which cells move through spatially defined shedding areas of photoconverting light was incorporated into the cell adhesion chromatography microfluidic system.¹⁴ Individually, the cell sorting microfluidic quantifies adhesion in flow over long time- and length-scales on a large, subpopulation basis. Conversely, the photoconversion technique quantifies velocities of rolling adhesion in flow over short time- and length-scales on a single cell basis. The integration of the cell sorting and photoconversion systems thus permits the simultaneous evaluation of cellular properties associated with average versus instantaneous adhesion behaviors of metastatic tumor cells. Using this methodology, analysis of persistent adhesion via photoconversion and adhesion chromatography converged to yield the same analyzed cellular velocities and phenotypes associated with rolling adhesion by human LS174T metastatic colon carcinoma cells that are frequently used to model selectin-mediated tumor dissemination.^13,27–29 This analytical method also revealed rolling adhesion on E-selectin to be strongly associated with expression of E-selectin ligands, correlations that vary with ligand type and rolling velocity magnitude. Further, discrete selectin ligand expression profiles were found to underlie persistent versus non-persistent adhesion by LS174T cells to E-, P-, versus L-selectin in flow, suggestive of divergent regulatory mechanisms. This integrated cell sorting and photoconversion microfluidic platform thus enables in vitro analysis and comparisons of adhesive phenotypes as they relate to mechanisms of cancer cell metastasis in the context of selectin mediated adhesion to reveal new insights into potential cancer dissemination pathways.

Materials and Methods

Cell Culture

Human colorectal adenocarcinoma LS174T cells (both parental and those transfected to stably express Phamret14) were cultured in Dulbecco’s Modified Eagle Medium supplemented with 10% heat inactivated fetal bovine serum (FBS) and 1% penicillin-streptomycin-amphotericin (Life Technologies). Cells were harvested via mild trypsinization with 0.25% trypsin EDTA, centrifuged at 400 X G for 5 min, and resuspended in culture medium. Prior to use in perfusion experiments, the cells were incubated in suspension for 2 h at 37°C to allow regeneration of adhesive cell surface ligands. All cells were obtained from American Type Culture Collection.

Channel Fabrication

Microfluidic channels were fabricated as detailed previously.¹⁵ Briefly, the microfluidic channels were made using 100 µm thick double-sided adhesive tape (3M) backed with a release liner into which a U-shaped channel was cut, using a craft cutter (Silhouette America), consisting of two parallel 14 cm long by 2 cm wide sections connected by a 1.5 cm long by 2 cm wide straight portion of adhesive tape. The channel was designed as a U-shape to optimize the size of the platform for ease of use with the spatial constraints of the high speed videomicroscopy, while permitting settling of the cell pulse before reaching the functionalized portion of the channel, as experimentally and quantitatively validated previously.¹⁵ The adhesive channels were then affixed to PDMS (Ellsworth Adhesives), which was previously cured by mixing PDMS base with curing agent at a ratio of 9:1 and curing at 90˚C for 3 h in a Pyrex dish. To complete fabrication, an inlet hole was punched into the end of one of the long, straight portions of the channel using a 3 mm biopsy punch. The platform was attached to a non-tissue culture treated polystyrene plate into which an outlet hole was drilled prior to assembly to permit the collection of the cell fractions into easily interchangeable reservoirs without disturbing flow, as described previously.¹⁵

Channel Functionalization

The long, straight portion of the channel nearest the outlet was functionalized through the following steps: incubating at 4°C overnight with Fc specific anti-IgG (R&D Systems) diluted in Dulbecco’s Phosphate Buffered Saline (D-PBS) without calcium and magnesium to a concentration of 2.5 µg mL⁻¹ for E-selectin experiments or 25 µg mL⁻¹ for P- and L-selectin experiments, which is non-specifically adsorbed onto the surface of the polystyrene plate; blocking for 1 h at room temperature with 1% BSA in D-PBS; incubating for 2 h at room temperature with either 2.5 µg mL⁻¹ of E-selectin (R&D Systems), 25 µg mL⁻¹ of P-selectin (R&D Systems), or 25 µg mL⁻¹ of E-selectin (R&D Systems), all of which were diluted in D-PBS with calcium and magnesium; and finally, blocking the entire device for 1 h at room temperature with 1% BSA in D-PBS at room temperature. In between each step, the channel was washed three times with 1 mL of D-PBS with calcium and magnesium. Only the straight portion nearest the outlet was functionalized by pipetting 200 µL of anti-IgG, 1% BSA and selectin solutions into the drilled outlet hole which, by virtue of this volume covering only the terminal straight channel portion, prevented functionalization of the rest of the channel. In so doing, cells that had all settled to the bottom of the channel prior to reaching the functionalized portion of the channel were ensured to experience a uniform channel length with which to mediate adhesive interactions.¹⁵

Perfusion Experiment Workflow

Perfusion experiments were performed in a similar manner to that formerly described.¹⁵ An inlet syringe connected to tubing filled with perfusion media (0.1 % BSA in D-PBS), was connected to a syringe pump (PhD Ultra Harvard Apparatus) and used to withdraw a cell pulse of 250,000 cells diluted in 200 µL of perfusion media into the inlet tubing at a rate of 0.5 mL min⁻¹. The tubing connected to the syringe was inserted into the inlet hole of the channel. An outlet reservoir was made by connecting a 5 mL test tube to the bottom of the drilled outlet hole on the polystyrene plate. The assembled platform was placed on an Eclipse TI optical microscope (Nikon) with an objective magnification of 10X and linked to NIS-Elements software (Nikon) to acquire videos at a frame rate of 25 frames per second at an exposure time of 0.281 µs and 2x2 binning of a 500 by 376 pixels. To complete the set-up, the system was integrated with a 405 nm light source was placed under the channel, directly upstream of the microscope objective, at the beginning of the functionalized portion of the channel. To ensure an instantaneous velocity was ascribed to the cells, a mask that is placed atop the laser and has a rectangular cutout 1.2 cm wide by 1 cm long to permit light exposure across the entire width of the channel but only 1 cm down the length of the channel in the direction of flow to occlude light exposure outside the defined window.

To begin perfusion, the 405 nm was turned on at maximum power and the syringe pump was set to a flow rate of 1 dyn cm⁻² for E-selectin experiments or 0.5 dyn cm⁻² for P-selectin and L-selectin experiments to initiate inflow from the syringe-tubing assembly. The flow rate to achieve desired wall shear stresses was determined from Equation 1, with the assumption that the apparent viscosity of the perfusion media was 0.009 P.

$$\text{Volumetric flow rate }=\text{ (wall shear stress)}*{(\text{channel}\hspace{0.17em}\text{height})}^2*\hspace{0.17em}\text{(channel width) }/(6*\text{apparent viscosity of perfusion}\hspace{0.17em}\text{media})$$ Equation 1

Perfusion was continued for 10 min for experiments conducted at 1 dyn cm⁻² or 20 min for experiments conducted at 0.5 dyn cm⁻², as these were the predetermined elution times for the free flow cells to elute through the channel (as measured previously)¹⁵, after which perfusion was stopped and the test tube containing the cell solution collected during perfusion was removed and a new test tube was attached. Finally, the syringe with inlet tubing was replaced with a new syringe-inlet tubing apparatus filled with only perfusion media and used to eject the cell solution out of the channel and into the second collection tube for experiments in which only one adherent (AD) fraction was collected. For experiments in which two AD fractions were collected, after the predetermined free flow cell elution time, the cell collection reservoir was changed, perfusion was continued for an additional 10 min (the time required for 50% of the adherent cells to elute). For experiments in which three adherent fractions were collected, after the predetermined free flow cell elution time, another fraction was collected after 28 min, the time for 66% of adherent cells to elute, to collect Adherent 2/3.

In experiments where cells were statically pre-photoconverted and photoconverted with perfusion, the cell pulse of 250,000 Phamret LS174T cells diluted in 200 µL of 0.1% BSA perfusion media were first statically photoconverted for 4 or 8 min in 96 well plate prior to perfusion through an E-selectin functionalized channel as described above.

Confirmation of Retention of Anti-IgG Surface Functionalization After Perfusion

Retention of anti-IgG adsorption was verified by functionalizing dishes with either 2.5 µg mL⁻¹ or 25 µg mL⁻¹ of FITC labelled anti-IgG. The channels were then blocked with 1% BSA for 1 h, and subsequently, perfusion experiments were conducted at either 1 dyn cm⁻² for 28 min (the longest time experiments are performed at 1 dyn cm⁻²) or 0.5 dyn cm⁻² for 20 min (the longest time experiments are performed at 0.5 dyn cm⁻²). During the experiments, the perfusion media was collected, and the mean fluorescent intensity (MFI) of the perfusion media was measured using a plate reader. The MFI of the collected perfusion media was correlated to an anti-IgG concentration using a standard curve. The perfusion media was found to contain a negligible amount of anti-IgG (Figure S8).

Elution Time Quantification

Similar to previous work¹⁵, 2 hour videos of 250,000 cells perfused as a 200 µL cell pulse at a flow rate corresponding to a wall shear stress of 1 dyn cm⁻² were taken 0.5 cm from the channel outlet. Our custom written cell tracker code was used to analyze collected videos, setting to threshold of 4 and a blurring factor of 7 to identify objects that had a diameter between 12–18 µm that had traveled more than 400 μm during the course of the video and record the time it took the object to leave the field of view after perfusion had begun. Cummulative frequency plots generated in Prism (GraphPad Software Inc) based on the elution time of each cell quantified using cell tracking software. For experiments where 2 AD fractions were collected, the elution time of the adherent 1/2 (AD1/2) fraction was 10 min after the free flow cells had eluted. For experiments where 3 adherent fractions were collected, the elution time of the adherent 1/3 (AD1/3) fraction was 8 min after the free flow cells had eluted and the elution time of the adherent 2/3 (AD2/3) fraction was 10 min after the AD1/3 cells had eluted.

Flow Cytometry Analysis

Collected cell solution fractions were centrifuged at 400 X G for 5 min and resuspended in diluted antibody solutions on ice for 45 min, washed with D-PBS with calcium and magnesium two times, and resuspended in 0.1% BSA perfusion media for analysis on the flow cytometer. The antibodies (BD Biosciences) were diluted in D-PBS with calcium and magnesium in the following ratios – 1:20 for PE CD44 and PE CEA, 1:40 for APC HECA452, and 1:10 for APC CD24. Obtained flow cytometry data was analyzed using FlowJo software (Treestar Inc.).

Quantification of Adhesive Cell Behavior

Cell adhesive phenotypes were characterized based on both their average and instantaneous velocities of rolling adhesion in flow when perfused over a selectin functionalized substrate. The instantaneous velocity of perfused cells was quantified by their green fluorescent protein to cyan fluorescent protein (GFP/CFP) ratio compared to an unphotoconverted control and converted to velocity using a photoconversion standard curve. Average cell velocities were quantified based on their population average residence times in the functionalized channels.¹⁵

Development of Standard Curve

Channels were blocked with 1% BSA (diluted with D-PBS with calcium and magnesium) to inhibit any non-specific cell interaction with the polystyrene dish during perfusion. The light source was set to maximum power and masked, as described above, to allow light only on 1 cm of the channel in the direction of flow. By varying the flow rate within the channel and, therefore, the velocity of cells, the residence time of cells exposed to the 10 mm window of light from the light source can be controlled. Using this setup, cells were perfused in the channels at differing flow rates which corresponded to desired velocities of cells, as derived previously²⁶, and desired residence times of cells exposed to light source according to Equations 2 and 3, where the cell radius is assumed to be 8 µm and medium viscosity 0.009 P.

$$\text{Instantaneous velocity }=\text{ (mask length) }/(\text{residence}\hspace{0.17em}\text{time in light})$$ Equation 2

$$\text{Instantaneous velocity }=\text{ (wall shear stress) }/[(\text{channel height})*(\text{medium}\hspace{0.17em}\text{viscosity})]*[(\text{channel}\hspace{0.17em}\text{height}/4)-(\text{cell}\hspace{0.17em}\text{radius})\hspace{0.17em}\text{^}2]$$ Equation 3

Each perfusion was allowed to run for the duration of the time experimentally determined necessary for the cell pulse to traverse the light source window. After which, the cells and perfusion media was expelled from the channel at high flow rates into the reservoir. The cells and perfusion media in the reservoir were collected, centrifuged at 400 X G, and resuspended in FACS buffer (0.1% BSA diluted in D-PBS with calcium and magnesium) for flow cytometry analysis of extent of photoconversion. Data from flow cytometry was collected and analyzed using FlowJo (Treestar Inc). Graphing the relationship between the extent of photoconversion—as given by ratio of GFP to CFP, normalized to a control sample perfused with the light source off—and velocity of cells at different flow rates using Prism (GraphPad Software Inc), a nonlinear regression was fit to the data. This was done for Phamret LS174T cells of different passage age and different cell thaws to ensure the standard curve was broadly applicable (Figure S7).

Ligand Expression and Extent of Photoconversion Correlations

FlowJo (Treestar Inc) and Prism (GraphPad Software Inc) were used for gating and subsequent analysis of flow cytometry data. For all experiments, collected events from the BD LSR II Flow Cytometer were gated for live cells. The live cell population was then gated into photoconverted and unphotoconverted populations, and coefficients of correlation in linear models were quantified as previously described.¹⁴ Briefly, after an unphotoconverted sample of LS174T Phamret cells were analyzed on the flow cytometer and gated to include only live cells, logarithmic plots of GFP versus CFP signal were made using FlowJo, and the resultant population was used gate the photoconverted negative population, likewise, any cells with a greater GFP/CFP ratio as compared to the unphotoconverted cells were gated as photoconverted positive. In order to determine correlation coefficients of the linear, the photoconverted negative or photoconverted positive cells, which were labeled with a PE- or APC-conjugated antibody for the ligand of interest, from each collected fraction were subsequently analyzed on the flow cytometer at the single cell level for fluorescent intensity of PE or APC, GFP, and CFP. The resulting data points were then pooled and divided into four equal subgates based on their extent of photoconversion, as quantified by the GFP/CFP ratio such that the center of each subgate was the mode of the cell population. The range for each quartile was determined by a perfused, photoconverted, but unsorted control that was conducted on each day for each experimental condition. This was done each time an experiment was performed to account for inherent variability between cells. For each subgate, the mean PE or APC fluorescent intensity and GFP/CFP ratio was calculated. Finally, Prism was used to fit a linear regression model to the data, as well as perform post-hoc tests for non-zero slopes and differences between slopes.

Outlier Detection and Elimination

ROUT outlier tests with criteria Q = 1 were conducted on all data sets used for linear regression models. Recognized outliers were removed from the data sets prior to model fitting.

Enzymatic Treatments

Prior to perfusion in functional adhesion experiments with enzymatic treatment, Phamret LS174T cell suspensions at 1X10⁷ cells mL⁻¹ were incubated with or without 0.1 U mL⁻¹ Vibrio Cholerae Neuraminidase (Roche Applied Science) at 37°C for 1 hr, as described previously.^12,30 After treatment, cells were washed to remove the enzyme and resuspended in 0.1% BSA diluted in D-PBS with calcium and magnesium for perfusion. To confirm complete digestion, cells were stained with a fluorescent mAb HECA452 (which recognizes LS174T expressed sialic-acid–bearing epitope sialyl Lewis x) and analyzed on a flow cytometer.

Results and Discussion

Adherent relative to free flow cell fractions exhibit greater photoconversion proportion and extent

A previously engineered cell sorting microfluidic platform based on a parallel plate flow chamber configuration was modified by adding a 405 nm laser to the beginning of the selectin functionalized portion of the channel (Figure 1A,B).¹⁵ The start of the functionalized channel portion was intentionally chosen as the light exposure location in order to standardize when cells would be exposed to light during experimentation, thus minimizing potential variance in when cells were light exposed (e.g. after different times of adhesion) that may confound interpretation of experimental results. Note that by virtue of the upstream settling feature, all cells perfused through the device were within ~10% of a cell radius from the inferior substrate, ensuring all cells had the opportunity to mediate adhesive contacts.^15,26 Metastatic human colon carcinoma LS174T cells of epithelial origin and approximately 14 cell µm diameter were used as the model cell line, as they have been extensively studied in adhesion assays and for interrogation of mechanisms underlying colon cancer metastasis.^{7,12,14,15,25,26,31,32} LS174T cells that stably expressed Phamret, a photoconvertible fluorescent protein formed from fusion of photoactivatable green fluorescent protein (GFP) with cyan fluorescent protein (CFP), exhibited no difference in adhesion behavior compared to untransfected parental cells (Figure S1A,B). Use of this engineered cell line thus enabled a readout of instantaneous velocity to be ascribed individually to each cell in proportion to their transit time through the illuminated region during perfusion, as previously described and quantified by an increase in GFP signal relative to an unchanged CFP signal resulting in an increased GFP/CFP ratio (Figure 1A, 2A).^14,33,34

Figure 1. Integrated photoconversion, cell sorting adhesion chromatography microfluidic for in vitro interrogation of metastatic cancer cell adhesive phenotypes.

(A) The instantaneous and average velocity of an infused cell pulse can be simultaneously quantified by utilizing a 405 nm laser to photoconvert Phamret-expressing cells to ascribe an instantaneous velocity and interchangeable cell fraction tubes to ascribe an average velocity in a (B) hemodynamic microenvironment-mimicking, parallel-plate microfluidic device, into which metastatic cancer cells are infused into the selectin functionalized channel and visualized via integrated high speed videomicroscopy. (C) LS174T cells perfused over a functionalized channel were sorted into a free flow (FF) and adherent (AD) fraction based on their elution time from the cell sorting system.

Figure 2. Adherent relative to free flow cell fractions exhibit increased extents of photoconversion and proportions of photoconverted cells.

(A) Photoconversion standard curve relating cell velocity (calculated based on perfused cell exposure time within light exposure window) and GFP/CFP ratio normalized to an unphotoconverted control. (B) Flow cytometry scatter plot gating of photoconverted (PC+) and unphotoconverted (PC-) Phamret-expressing LS174T cells based on unphotoconverted control population. (C) Representative flow cytometry plots of the extent of photoconversion from cells collected in the FF (left) or AD fractions (right). (D) GFP/CFP ratio measured by flow cytometry and (E) percent of PC+ cells from collected fractions after perfusion, photoconversion, and separation over E-, P-, or L-selectin. Intercept of d dashed line: E, 1; F, 2.5%. (F) Schematic depicting persistent versus non-persistent adhesion along adhesive ligand-functionalized substrate. Cell A represents a free flow cell, while cell B represents a lowly persistent, photoconverted cell that would elute into the FF fraction. Cell C and cell E are also lowly persistent cells. Cell D is a fast rolling highly persistent cell, whereas cell F is a slow rolling highly persistent cell. (G) PC+ cells from a perfused and photoconverted, but unsorted, cell pulse used to create four extents of photoconversion quantile bins. (H-J) Percent of PC+ cells from collected fractions contained within each bin after perfusion, photoconversion, and separation over (H) E-, (I) P-, or (J) L-selectin. (D,E,H-J) Perfusion, photoconversion, and separation of a Phamret-expressing LS174T cell pulse of 250,000 cells over either 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻² or 25 μg mL⁻¹ L- or P-selectin at 0.5 dyn cm⁻². Each dot or bar represents the mean ± s.e.m. of n≥3 individually run experiments. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001 by (D,E) one sample t-test (*) with (D) h₀ = 1 or (E) h₀ = 2.25% or paired t-test (#). (I-K) One-way ANOVA with post hoc t-test with Bonferroni corrections for multiple comparisons.

As selectins mediate cell adhesion over a wide range of rolling velocities (order of 1–100 µm/sec) and efficiencies,²⁵ parameters under which this device was implemented to interrogate mechanisms underlying of long- versus short- scale adhesive interactions by LS174T cells to E-, P-, and L-selectin were adjusted in order for the collected cell population sizes to be roughly similar amongst selectins and sufficiently large to enable subfractionation and fluorescent labeling analyses described below. Specifically, all experiments were performed at concentrations of 2.5 versus 25 μg mL⁻¹ and 1.0 dyn cm⁻² versus 0.5 dyn cm⁻² for E- versus P- and L-selectin, respectively, and cells sorted over 20 min into Free flow (FF) versus Adherent (AD) fractions (Figure 1A) for analysis (Figure S1C–E). The chosen wall shear stresses of 0.5 and 1 dyn cm⁻² recapitulate the low venular shear stresses under which metastatic circulating tumor cell recruitment occurs.^25,35,36

Within each collected subfraction, the extent and proportion of total cells that were photoconverted was measured based on unphotoconverted control cells (Figure 2B,C). Expectedly, the extent of photoconversion (GFP/CFP ratio) was significantly greater for the AD relative to FF subpopulation, though not to a statistically significant extent on P-selectin (Figure 2D), potentially due to the wider range of velocities typically observed for cells mediating adhesion to P-selectin^25,26 and diminished adhesion persistence²⁵ resulting in cells not mediating adhesion within the photoconversion window eluting into the AD fraction or vice versa. Additionally, measured ratios in the FF fraction were not statistically different from unphotoconverted cells when perfused over E- and P-selectin (Figure 2E). However, a small increase was measured over L-selectin (Figure 2D), in line with the tethering type interaction mediated by L-selectin and very low adhesion persistence mediated by LS174T cells on L-selectin.²⁵ The frequency of photoconversion positive (PC+) cells (Figure 2E) was also increased in AD relative to FF fractions after perfusion, though again not to a statistically significant extent on P-selectin (Figure 2E), changes not associated with differences in cell viability (Figure S1F). Notably, the frequencies of PC+ cells were significantly non-zero in the FF fraction (Figure 2E). Thus, despite high (>220 µm/sec) average velocities indicated by their elution into the FF fraction, a low but substantial fraction of cells within the FF fraction did exhibit some level of photoconversion above background. Consistent with previous reports,²⁵ a subfraction of cells thus exhibit non-persistent adhesion which in this case is captured by a subfraction of cells that elute into the FF fraction but have been photoconverted as a result of adhesion within the light exposure window that slows them down to some level below minimum detectable instantaneous velocity (Figure 2F). In its capacity to identify and measure different adhesion qualities (persistent versus non-persistent adhesion), this system also substantiates previous results that a higher proportion of LS174T cells exhibit diminished adhesion persistence on P- and L- relative to E-selectin (Figure 2E).

In an effort to further quantify differences instantaneous velocity magnitudes exhibited by FF versus AD cells, the extent of photoconversion was assessed by subdividing the PC+ cells into GFP/CFP ratio quantile bins based on perfused and photoconverted but unsorted cells (Figure 2G). The proportion of cells exhibiting high extents of photoconversion (bin 4) corresponding to low instantaneous cell velocities was substantially higher in AD but not FF cells collected on all three selectin types (Figure 2H–J). This integrated system can thus assess both the quantity (frequency of AD versus FF and PC+ cells of total) as well as quality [persistent (PC+ in AD) versus non-persistent (PC+ in FF)] of rolling adhesion by cells in shear flow.

Velocity of rolling adhesion magnitude on E-selectin correlates differentially with ligand type

The expression of selectin ligands within subfractions/subsets of perfused cells was assessed in the context of E-selectin adhesion, which is highly persistent (Figure 2D–F).²⁶ Thus, enrichment of specific selectin ligands over differing rolling velocity magnitudes was measured and results between chromatography versus photoconversion methodologies compared.

First, the fractionation resolution was increased to further elucidate phenotypic differences amongst late and early eluting cells within the AD fraction. Specifically, collection of either two fractions of interacting cells (Figure 3A, S2A,B) – the first and later 50% of interacting cells to elute (AD 1/2 and 2/2, corresponding to average velocities of 83 – 220 and <83 µm/s, respectively) – or three fractions of interacting cells (Figure 3B, S2A,C) – the first, middle, and last third of interacting cells to elute (AD 1/3, 2/3, and 3/3, corresponding to average velocities of 90 – 220, 60 – 90, and <60 µm/s, respectively) was performed (Table 1). In so doing, the extent of photoconversion for cells in AD 2/2 but not AD 1/2 was found to be greater than that of FF (Figure 3C) with a greater percentage of AD 2/2 cells being photoconverted, specifically to high extents (Figure 3D,E fourth quartile) as would be expected given the upper limit of photoconversion measurable velocities (Figure 2A). Similarly, when the perfused cells were fractionated into three adherent subsets, again, only the latest eluting fraction (AD 3/3) had significantly greater extents of photoconversion (Figure 3F) and greater percent of cell photoconverted (Figure 3G,H fourth quartile) compared to the FF cells and earlier eluting adherent cells (AD 1/3 and AD 2/3). In converting these photoconversion extents to velocities, analysis of all collected cells in PC+ fractions demonstrated rolling behavior over E-selectin that was in alignment at both long and short time and length scales (e.g. average and instantaneous rolling velocity, respectively), in contrast to all cells (PC+ and PC-) that exhibited instantaneous velocities higher than those predicted by their elution time (Figure 3I–K).

Figure 3. Instantaneous and average cell velocities converge when measured using photoconversion and chromatographic methodologies in combination.

Phamret-expressing LS174T cells perfused over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻² were sorted into (A) two or (B) three AD fractions based on their elution time from the selectin functionalized integrated photoconversion and cell sorting system. (C-H) Extent of photoconversion measured by flow cytometry (C,F) and percent of PC+ cells (D,G) within each fraction normalized to the unphotoconverted control population when either (C,D) two or (F,G) three AD fractions were collected. FF denotes free flow, AD denotes adherent. (E,H) Binning and analysis of PC+ cells from each fraction based on extent of photoconversion measured by flow cytometry normalized to the unphotoconverted control after collection of (E) two or (I) three AD fractions. (I-K) Instantaneous velocity ± s.e.m and average velocity ± s.e.m among cells in each fraction when (I) one, (J) two, or (K) three AD fractions were collected. (C- K) Each dot or bar represents the mean ± s.e.m. of n≥3 individually run experiments. Perfusion, photoconversion, and separation of a Phamret-expressing LS174T cell pulse of 250,000 cells over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻². * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001 by one-way ANOVA with post hoc t-test with Bonferroni corrections for multiple comparisons.

Table 1.

Average rolling velocity ranges for eluting cell fractions.

Cell fraction	Average rolling velocity (µm/s)
Free flow (AD)	> 220
Adherent 1/2 (AD 1/2)	83–220
Adherent 2/2 (AD 2/2)	< 83
Adherent 1/3 (AD 1/3)	90 – 220
Adherent 2/3 (AD 2/3)	60 – 90
Adherent 3/3 (AD 3/3)	< 60

Expression levels of selectin ligands among the AD fractions were next assessed by flow cytometric analysis post sorting (Figure 4A). PC+ cell populations eluting in the latest fractions demonstrated the most pronounced enhancements in sLe^x and CD44 expression compared to the non-interacting (PC- and FF) and early eluting AD cells (Figure 4B,C,E,F). CEA ligand expression was also enriched in the latest fraction of eluting AD cells compared to levels measured for FF cells (Figure 4D,G), a trend not seen when only one AD fraction was collected.¹⁵ Notably, no differences were observed between adhesive ligand expression on E-selectin among PC- cells (Figure 4B–G).

Figure 4. Selectin ligand expression increased in cells mediating persistently slow rolling on E-selectin.

(A) Phamret-expressing LS174T colon carcinoma cells perfused over selectin-functionalized channels, photoconverted in a manner proportional to instantaneous velocity, sorted into fractions based on average velocity, and gated using flow cytometry into photoconverted (PC+) and unphotoconverted (PC-) populations. Cells collected in each fraction are stained with fluorescently tagged antibodies, allowing correlations between ligand expression, elution time, and photoconversion to be elucidated. (B-G) Perfused, photoconverted, and sorted Phamret-expressing LS174T cells were collected into (B-D) two or (E-G) three AD fractions fluorescently labeled for (B,E) sLe^x, (C,F) CD44, or (D,G) CEA to measure ligand expression levels among PC+ or PC- cells between the collected fractions. MFI denotes mean fluorescent intensity. (B-G) Each bar represents the mean ± s.e.m. of n≥3 individually run experiments. Perfusion, photoconversion, and separation of a Phamret-expressing LS174T cell pulse of 250,000 cells over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻². * p < 0.05, ** p < 0.01, *** p < 0.001 by one-way ANOVA with post hoc t-test with Bonferroni corrections for multiple comparisons.

Relationships between adhesive ligand expression with instantaneous velocity magnitude, as measured by a cell’s mean GFP/CFP ratio, in various fractions were next assessed. After further dividing the PC+ and PC- subsets of cells separately based on GFP versus CFP expression into four evenly distributed bins (quantiles with respect to the frequency of an unphotoconverted parent population of Phamret LS174T cells, Figure 5A) to quantify the GFP/CFP ratio spread among the cells, mean ligand expression levels were quantified for each bin in the FF and AD fraction of PC+ or PC- cell populations. In so doing, the relative correlations (when statistically non-zero) between adhesion molecule expression and cellular velocity was quantified, with greater slope differences between subfractions indicating different relative roles for each ligand in mediating selectin adhesion within each fraction, e.g. slow (PC+ in AD 2/2 and AD 3/3) versus fast (PC+ in AD 1/2, AD 1/3, and AD 2/3) rolling adhesion.

Figure 5. Magnitude of persistent, slow cell rolling mediated by E-selectin adhesive interactions inversely correlates with selectin ligand expression.

(A) PC+ and PC- cells from the photoconverted but unsorted cell fraction (A, middle) binned into four equal gates (left/right) based on their extent of photoconversion after perfusion over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻² or 25 μg mL⁻¹ L- or P-selectin at 0.5 dyn cm⁻². The extent of photoconversion (GFP/CFP ratio) of FF PC- cells and (B-D) two or (E-G) three fractions of AD PC+ cells related to the expression of (B,E) sLe^x reactive FIECA452, (C,F) CD44, and (D,G) CEA. (B-G) Binned, flow cytometry data of PC+ cells was pooled from independent experiments and plotted with corresponding linear fits; C_ADand C_F represent the regression slope of the PC+ cells in the AD and FF fractions, respectively, * denotes non-zero slopes of the linear fit, † denotes significance of comparison between the C_AD values and C_F. * p < 0.05, ** p < 0.01. Outlier tests with criteria Q = 1 were conducted on all data sets, and outliers were removed prior to regression and statistical analysis. Perfusion, photoconversion, and separation of a Phamret-expressing LS174T cell pulse of 250,000 cell over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻².

Regression analyses comparing ligand expression levels amongst PC+ cells within various AD fractions revealed sLe^x and CD44 expression being proportional to the GFP/CFP ratio of PC+ cells perfused, photoconverted, and sorted into either two (Figure 5B,C) or three (Figure 5E,F) adherent fractions on E-selectin in flow. These relationships were more pronounced with increasing the number of collected AD fractions. Notably, significant correlations with CEA expression were also found, a relationship not observed when only one AD fraction was analyzed¹⁵ but consistent with results using photoconversion methodology alone (Figure 5D,G).¹⁴ No correlations were seen between ligand expression and velocity among PC- cells (Figure S3). Overall, these data demonstrate that increased expression of adhesive ligands sLe^x and CD44 to be inversely correlated with slower rolling on E-selectin over a broad range of velocity magnitudes, whereas CEA expression is associated with E-selectin adhesion in the context of slow rolling only.

Slow and fast rolling adhesion on E-selectin is associated with sLex expression

Within the experimental configuration described so far, only fractions that exhibit velocities within the dynamic range of the photoconversion system are measurable via photoconversion (Figure 6A). Under all fractionation conditions, the latest eluting cells, and therefore the cells with the slowest average rolling velocities, exhibited the most significant correlations between increasing extent of photoconversion and adhesive molecule presentation (Figure 5B–G). However, with the photoconversion methodology, velocities above ~200 µm/s are poorly discernible (Figure 2A). As such, although the photoconversion methodology reduces noise within subfractionated populations to reveal subtle trends, it is limited to the direct analysis of cells interacting with slow instantaneous velocities only (Figure 6A).

Figure 6. Cellular sLe^x expression inversely correlated with magnitude of fast rolling adhesion on E-selectin revealed by pre-photoconversion.

(A) Measured GFP/CFP ratios and average velocities (calculated by elution time) of collected cell fractions. Circles, measured average velocities; Squares, average velocities adjusted based on pre-photoconversion time and measured average velocities. (B) Experiment workflow depicting static photoconversion of cells pre-perfusion to increases the baseline photoconversion of the cell pulse, followed by perfusion and photoconversion over an E-selectin functionalized channel in order to maximally photoconvert late eluting adherent cells. (C) Percent of Phamret-expressing LS174T cells photoconverted after exposure to the 405 nm laser for 4 min of pre-photoconversion. FF denotes free flow, AD denotes adherent. PC denotes photoconversion. (D) Quantification of mean fluorescent intensity of sLe^x reactive FIECA452 in each collected fraction of either PC+ or PC- subsets. (E) The extent of photoconversion (GFP/CFP ratio) of PC+ cells within the FF and 2 AD fractions related to the expression of sLe^x (FIECA452). Binned, flow cytometry data of PC+ cells was pooled from independent experiments and plotted with corresponding linear fits; C_AD1/2, C_AD2/2, and C_F represent the regression slope of the PC+ cells in the AD 1/2, AD 2/2, and FF fractions, respectively, * denotes non-zero slopes of the linear fit, † denotes significance of comparison between C_AD1/2, C_AD2/2, and C_F. Outlier tests with criteria Q = 1 were conducted on all data sets, and outliers were removed prior to regression and statistical analysis. (F) Quantification of mean fluorescent intensity of sLe^x expression normalized to a perfused, but not pre-photoconverted control after 4 or 8 min of pre-photoconversion. (C-F) 4 min of pre-photoconversion followed by perfusion, photoconversion, and (C-E) fractionation of a Phamret-expressing LS174T cell pulse of 250,000 cell over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻². (A,C-F) Each dot or bar represents the mean ± s.e.m. of n≥3 individually run experiments. * p < 0.05, ** p < 0.01, *** p< 0.001 **** p < 0.0001 by (C-E) one-way ANOVA with post hoc t-test with Bonferroni corrections for multiple comparisons or (F) one sample t-test with h₀ = 1. Dashed line at 1.

To expand the working parameters and versatility of the system to elucidate expression relationships for cells exhibiting higher velocities of rolling adhesion, photoconversion of cells prior to perfusion was performed (Figure 6B). In so doing, GFP/CFP ratios were increased for all cells depending on the time of light pre-exposure (Figure S4A). As a result, different fractions of cells are adjusted to GFP/CFP levels within the working dynamic range of the fluorometric velocity measurement technique based on pre-exposure time length (Figure 6A), broadening the range of adhesive phenotypes (in this case to include on average fast rolling cells) over which expression correlations can be quantified using this photoconversion method. For example, with pre-exposure for 4 min, nearly all cells were PC+ when fractionated on E-selectin into two AD fractions (Figure 6C), with minimal differences in viability amongst collected fractions (Figure S4B). However, photoconversion extent varied widely with GFP/CFP ratio of AD 1/2 within the photoconversion dynamic range (Figure 6A). Regression analysis revealed sLe^x expression positively correlated with photoconversion extent for not only AD 1/2 but also the AD 2/2 fraction with 4 min pre-photoconversion (Figure 6D,E), in contrast to analyses performed on fractions that were not pre-photoconverted (Figure 5B).

Of note, overexposure to photoconverting light results in loss of cell viability (Figure S4A). When using a light pre-exposure time that results maximal photoconversion but 50% cell loss (8 min, Figure 6A, S4A), perfusion over E-selectin resulted in loss of cell viability. In the context of collection into multiple fractions, nearly half of the free flow cells (FF) and 40% of early eluting, adhesive cells (AD 1/2) remained viable, whereas only about 20% of the late eluting adhesive cells (AD 2/2) were viable (Figure S4C). Notably, pre-photoconversion does not affect the adhesive behavior of cells in the channel, as the percent of pre-photoconverted cells that elute into the FF, AD1/2, and AD2/2 fractions after perfusion through the channel without photoconversion during perfusion is not statistically different than the percent of cells that elute into the aforementioned fractions without pre-photoconversion (Figure S4D,E). Without subfractionation, expression of sLe^x by viable cells post perfusion was lower in cells pre-exposed to the photoconverting light for 8 min versus those without pre-photoconversion (Figure 6F). This supports sLe^x enabling adhesive interactions to E-selectin in flow as the viability of cells exhibiting the most extensive adhesive interactions would be compromised. Findings using this pre-photoconversion technique thus confirm inverse correlations between sLe^x expression and velocity of rolling adhesion across a broad range of velocity magnitudes.

Selectin ligands enriched in cells mediating persistent but not non-persistent selectin adhesion

The capacity of this system to differentiate subsets of LS174T cells in persistent versus non-persistent adhesion with selectins was next leveraged in order to interrogate mechanisms underpinning these different qualities of selectin adhesion (Table 2). Ligand expression levels amongst subfractions were compared and relationships between measured instantaneous velocities evaluated. Specifically, a subset of non-persistent cells (e.g. those within the PC+ subpopulation of the FF fraction) were compared to those of the PC+ cells within the AD fraction, which exhibit sustained rolling adhesion, the caveat being for reasons described above that the latter is limited to only slow rolling cells (<200 µm/s) for fractionation experiments performed without pre-photoconversion.

Table 2.

Correspondence of photoconversion integrating cell sorting microfluidic readout to adhesion persistence.

Fraction	Photoconversion	Adhesion persistence
Free flow	PC−	No adhesion
Free flow	PC+	Non-persistent
Adherent	PC−	Non-persistent
Adherent	PC+	Persistent

Within fractions collected on selectin functionalized substrates of AD cells only, expression of sLe^x and CD44 was increased in PC+ subpopulations (Figure 7A,B). Expression of sLe^x was also increased in PC- cells collected on P-, but not E- or L-, selectin (Figure 7A). No other correlations were found between velocity and selectin ligand expression among cells that remain unphotoconverted (PC-, Figure 7A–C). Additionally, CEA expression was found to be increased in PC+ AD but not FF subpopulations on P- and L-, but not E-, selectin (Figure 7C), a surprising trend given CEA’s function as an E- but not P-selectin ligand in these cells. These data demonstrate sLe^x but not CD44 to be enriched in LS174T cells mediating not only persistent but also non-persistent slow rolling adhesive interactions with P-selectin in shear flow and CEA is associated with persistent adhesive phenotype on P- and L-selectin.

Figure 7. Selectin ligands sLe^x, CD44, and CEA are enriched in cells mediating persistent slow rolling selectin adhesion in flow.

(A-C) Expression levels of selectin ligands (A) sLe^x, (B) CD44, and (C) CEA among PC+ and PC- cell subsets collected after exposure to a 405 nm light during perfusion of a cell pulse of 250,000 Phamret-expressing LS174T cells over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻² or 25 μg mL⁻¹ L- or P-selectin at 0.5 dyn cm⁻². Each bar represents the mean ± s.e.m. of n≥3 individually run experiments. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001 by one-way ANOVA with post hoc t- test with Bonferroni corrections for multiple comparisons.

Regression analyses revealed sLe^x and CD44 expression to vary directly versus inversely with cell velocity magnitude (measured by GFP/CFP ratio) on E-selectin in the context of persistent (PC+ cells in AD fraction) versus non-persistent (PC+ cells in FF fraction) adhesion, respectively (left panels, Figure 8A,B). No correlations direct or otherwise were seen for CEA expression (Figure 8C, left panel). In the context of P- and L-selectin adhesion, sLe^x and CEA but not CD44 expression correlated directly with GFP/CFP ratio for persistent adhesion (middle and right panels, Figure 8A–C). Contrastingly, inverse correlations in expression of all three ligands with GFP/CFP ratio were found in the context of non-persistent adhesion (PC-, Figure 8D). Consistent with results from the collection of multiple adherent fractions, no correlations are observed in PC- cells subsets (Figure S5). Together, these results demonstrate the differential overall expression levels of selectin ligands and associations with velocity magnitudes between cells mediating persistent versus non-persistent adhesion. For example, although CD44 expression overall is enriched in cells in cells exhibiting persistent adhesion to P-selectin, its expression does not correlate with slower rolling velocities. Furthermore, although sLe^x expression is increased in cells mediating slow rolling P-selectin non-persistently, its expression is associated with increased, not decreased, velocity magnitudes. This integrated analysis platform can thus reveal complex interdependent relationships in cellular phenotypes associated with both rolling adhesion quantity (e.g. velocity) and quality (e.g. persistence).

Figure 8. Opposing ligand expression correlations with rolling velocity magnitudes between persistently versus non-persistently adhesive cells.

Sorted cell fractions labeled with fluorescently tagged antibodies for (A) sLe^x, (B) CD44, and (C) CEA, then applying the photoconversion gates and binning allowed the extent of photoconversion and elution time to be correlated to the relative ligand expression. Binned, flow cytometry data of cells was pooled from independent experiments and plotted with corresponding linear fits; C_AD and C_F represent the regression slope of the PC+ cells in the Adherent and or PC- cells in the Free flow fractions, respectively, * denotes non-zero slopes of the linear fit, † denotes significance of comparison between C_AD and C_F. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Outlier tests with criteria Q = 1 were conducted on all data sets, and outliers were removed prior to regression and statistical analysis. Perfusion, photoconversion, and separation of a Phamret-expressing LS174T cell pulse of 250,000 cells over either 2.5 μg mL⁻¹ E- selectin at 1 dyn cm⁻² or 25 μg mL⁻¹ L- or P-selectin at 0.5 dyn cm⁻². (D) Summary of P-values of comparisons between the slopes of the linear relationships for interrogated ligands cells in the Adherent (C_AD) versus Free flow (C_F) fractions; dashed line represents p=0.05.

Screening ligand degrading enzyme effects on adhesion mechanisms

To demonstrate the versatility of this microfluidic approach to assess drug treatment effects on cell adhesion extent and mechanism, cells were treated with neuraminidase (NA), a sialidase that cleaves the terminal sialic acid residues on sLe^x epitopes,^{12,37,38,39–41} and cell adhesion to selectin-functionalized surfaces analyzed via the integrated microfluidic and photoconversion based approach (Figure 9A). With NA treatment, sLe^x expression was completely lost (Figure 9B), while surface expression levels of CD44 and CEA ligands remained unchanged compared to untreated cells (Figure 9C,D) in all collected cell fractions. As previously reported,^12,13,42 NA treated LS174T cells retained the ability to mediate adhesion to E-, but not P- or L-, selectin (Figure 9E–F, S6). When fractionated on E-selectin, approximately 25% of perfused cells eluted into the AD fraction (Figure 9E). With respect to the extent of photoconversion, GFP/CFP ratios were diminished for AD but slightly increased for FF fractions as a result of NA (Figure 9F). Although the proportion of PC+ cells in AD fraction was also strikingly decreased as a result of NA, the proportion was strongly increased for FF (Figure 9G), suggestive of NA treatment diminishing LS174T adhesion persistence and velocity of rolling adhesion on E-selectin. The overall relative frequency of PC+ cells in the AD compared to FF fraction remained increased, however, though the relative difference much reduced (Figure 9H).

Figure 9. Sialyl Lewis x and CD44 correlated in highly but not lowly persistent slow rolling selectin-mediated adhesion.

(A) Experimental workflow for neuraminidase (NA) treatment perfusion experiments. (B-D) quantification of mean fluorescent intensity of (B) sLe^x, (C) CD44, or (D) CEA expression after treatment with/without NA and perfusion over E-selectin. FF denotes the Free flow, AD denotes Adherent, US denotes an unsorted group. (E) Percent of cells collected in the AD and FF fractions, (F) Mean GFP/CFP ratio after perfusion over E-selectin, (G) percent of cells photoconverted, and (FI) percent of cells photoconverted in the AD fraction normalized to the FF fraction with/without NA treatment. (I,J) The extent of photoconversion (GFP/CFP ratio) of PC- cells within the Free flow fraction or PC+ cells within the AD fraction related to the expression of (I) CD44 or (J) CEA sorted over E-selectin. Binned flow cytometry data of cells was pooled from independent experiments and plotted with corresponding linear fits after perfusion over E-selectin with/without NA treatment. Outlier tests with criteria Q = 1 were conducted on all data sets, and outliers were removed prior to regression and statistical analysis. (B-J) Perfusion, photoconversion, and separation of aPhamret-expressing LS174T cell pulse of 250,000 cell over 2.5 μg mL⁻¹ E-selectin at 1 dyn cm⁻². (E-J) Each dot or bar represents the mean ± s.e.m. of n≥3 individually run experiments. (B-J) * p < 0.05, ** p < 0.01, **** p < 0.0001 by (B-K) paired t-test or (J) one sample t-test with h₀ = 1; dashed line at y=l. (G) # denotes difference between -NA AD group and +NA AD group by paired t-test.

As CD44 expression has a strong inverse correlation with E-selectin rolling velocities by LS174T cells that holds across velocity magnitudes (Figure 4,5,6), effects of NA treatment on regression trends with CD44 were assessed. As expected given the role of sialylfucosylated CD44 variants in mediating E-selectin adhesion by these cells^14,15, NA treatment eliminated correlations between CD44 and extent of photoconversion within PC+ subsets of cells collected into AD cell fractions on E-selectin (Figure 9I). The lack of correlation of CEA expression was also sustained (Figure 9J). No changes were seen in correlations were seen in treated cells that eluted into the FF fraction, as they continued to show a lack of correlation between selectin-ligand expression and rolling velocity (Figure 9I,J). Together, these data suggest that sialylated CD44 mediates sustained adhesion to E-selectin and non-persistent adhesion that results from NA treatment is mediated by another herein unexplored selectin ligand.

Conclusions

The ability to evaluate within a heterogenous population the cellular features associated with different qualities of rolling adhesion using a novel, integrated cell sorting and photoconversion adhesive chromatography microfluidic platform was demonstrated here through the juxtaposition of measured long and short time- and length-scale adhesive rolling behaviors. Building off of a previously designed cell sorting system that fractionates subsets of cells based on their time-averaged cell adhesive behavior,¹⁵ a 405 nm laser was added to photoconvert cells in a manner directly proportional to their velocity under the small window of light exposure,¹⁴ thereby “labeling” the cells with their instantaneous velocity. Cells with slow rolling velocities within the light exposure window could thus be compared across elution times, i.e. adhesive versus non-adhesive fractions. Furthermore, cellular features could be analyzed using this preparative technique to distinguish expression pattern differences between cells with different adhesion phenotypes, as was explored here using flow cytometry.

First, the photoconversion readout allowed ligand expression signal in measured population subfractions to be filtered to exclude those that were exhibited non-persistent adhesion (and were thus PC-). In so doing, expression of CEA was found to be enriched in cell subfractions mediating adhesion to E-selectin, increases not seen without photoconversion.¹⁵ Even low levels of diminished persistence thus diluted signal and photoconversion could distinguish CEA expression by cells mediating slow, persistent rolling adhesion on E-selectin (Figure 4). Additionally, pre-photoconversion as well as increasing the number of collected adhesive subfractions allowed trends to be explored across rolling adhesion magnitudes (Figures 3–6). These techniques revealed adhesive ligands sLe^x and CD44 to be inversely correlated with slower rolling on E-selectin over a broad range of velocity magnitudes, whereas CEA expression is associated with E-selectin adhesion in the context of slow rolling only.

Concurrent measurement of instantaneous and average velocities also permitted the elucidation of cellular molecular phenotypes related to different adhesion persistence phenotypes. We previously reported colon carcinoma cells to exhibit in their interactions with P- and L- but not E-selectin diminished adhesion persistence,^25,26 defined as the capacity for cells to sustain adhesion and this case rolling adhesion over time and distance. In the instance of highly persistent adhesion, cells would mediate the same adhesive behavior along the entire channel length. Thus, so long as rolling velocities are below the threshold detectable by photoconversion, AD fractions would contain PC+ cells. Cells exhibiting low adhesion persistence on the other hand would interact for only portions of their transit time through the channel. Cells eluting into the FF fraction that are PC+ would thus be considered very lowly persistent as they were mediating adhesion within the light exposure window but deadhered. It should be noted that PC- cells cannot be assigned as either persistent or non-persistent, as they may have not photoconverted as a result of lack of adhesion or their rolling adhesion velocity being higher than the maximum velocity detectable by photoconversion.

This photoconversion and chromatographic method confirmed previous results using videomicroscopy only^25,26 that LS174T cells exhibit a persistent phenotype over E-, but not P- or L-selectin (Figure 2). Concurrent cellular phenotyping also revealed that although CD44 expression overall is enriched in cells in persistent P-selectin adhesion, its expression does not correlate with slower rolling velocities. Furthermore, although sLe^x expression is increased in cells mediating slow rolling P-selectin non-persistently, its expression is associated with increased, not decreased, velocity magnitudes. CEA expression was also associated with persistent adhesive phenotype on both P- and L-selectin (Figure 7), despite not serving as a ligand for the former.¹³ Correlations in selectin ligand expression with rolling velocity magnitude were also found to be inverted between persistently versus non-persistently adhesive cells on all selectins (Figure 8). Furthermore, enzymatic digestion of sialofucosylated epitopes also reduced persistent but not non-persistent adhesion by LS174T cells to E-selectin in flow (Figure 9), the latter presumably mediated by another selectin ligand unexplored herein. Molecular expression characteristics thus vary by adhesion quality amongst selectin types, which has ramifications for the potential contribution of adhesion types and molecular pathways in dissemination mechanisms. This integrated analysis platform can thus reveal complex interdependent relationships in cellular phenotypes associated with both rolling adhesion quantity (e.g. velocity) and quality (e.g. persistence) as well as drug effects on the molecular mechanisms underpinning such processes.

In closing, this integrated system is capable of isolating and identifying subsets of cells with different adhesive behaviors to enable associated cellular phenotypes to be discerned. Separating cells based on biophysical characteristics is a broadly used technique for probing mechanisms underlying multiple disease types, and this platform improves upon former methods by enabling both quantification of rolling adhesion velocities and assessment of their qualities, e.g. persistent versus non-persistent. Our results demonstrating divergent molecular profiles associated with various adhesive phenotypes suggests the potential for this system to be applied to identify druggable targets on distinct subsets of circulating cancer cells (tumor initiating, for example) or provide nuance to the utility of various biomarkers in distinguishing circulating cells with distinct malignant phenotypes.