Our scRAFA exploits a microfluidic platform integrated with versatile optothermal manipulation and optical imaging to trap and rotate single cells while monitoring the sequential cell rotation and cell-substrate adhesion [19, 20] (Fig. 2a). As a demonstration, Figs. 2b, c show the successive images of light-driven trapping and rotation of a single S. cerevisiae above a substrate (Additional file 2: Movie S1). We achieve high-efficient trapping and rotation of the targeted cell using two working laser beams with wavelengths of 785 nm and 532 nm. The substrate is designed to strongly absorb the 532 nm laser beam while being mostly transparent to the 785 nm laser beam. Specifically, a focused 785 nm laser beam is applied to trap the cell with optical force [21,22,23]. To achieve the stable rotation of the cell while being trapped, we further apply a focused 532 nm laser beam to heat the light-absorbing substrate near the trapped cell to generate a temperature gradient field (Additional file 1: Fig. S1), where thermophoretic force repels the cell from the laser-heated hot region while thermo-osmotic force attracts the cell to the hot region [24,25,26]. Thermo-osmosis is a surface-driven effect where a temperature gradient along the substrate induces a thermo-osmotic flow that is parallel to the substrate. With the increased laser power, the temperature gradient along the substrate increases and hence results in a stronger thermo-osmotic flow velocity. A balance among the optical force, thermophoretic force, and thermo-osmotic force leads to a stable optothermal trapping of the cell at the side of the heating laser beam (Additional file 1: Fig. S2). Without the optical force from the 785 nm laser beam, the 532 nm-laser-induced thermophoretic force and thermo-osmotic force cannot achieve stable trapping of the cells (Additional file 1: Fig. S2). Moreover, we manage an unbalanced thermo-osmotic flow along the cell surface to exert a torque on the trapped cell and to optothermally drive the out-of-plane rotation of the cell (Fig. 2d) [27]. The rotational speed is proportional to the heating laser power (Fig. 2e). To quantify the force on the cell, we extract the rotational speed of the cell from the video and calculate the force through Stokes’ law. Specifically, when the optical power of 532 nm and 785 nm lasers is 0.2 mW/μm2 and 1 mW/μm2, the corresponding rotational frequency is ~ 1 Hz and the calculated shear force is ~ 4 pN for 5 μm cell [28]. Similar to our previously theoretical analysis [27], the rotation is anticlockwise for a cell trapped at the left sides of the laser beam. To further verify the mechanism for optothermal manipulation, we experimentally track the central position of a cell being trapped and rotated relative to the laser beam center. The temporal trajectory distribution of the cell center shows that the stable cell-trapping position is away from the laser center (Additional file 1: Fig. S3), which matches well with our force analysis (Additional file 1: Fig. S2). In contrast to conventional optical tweezers where the cell is trapped at the focused laser beam center, the stable trapping position of the cell in our assay is away from the laser beam center. The distance between the heating center and the trapped cell center is over 2 μm due to the repelling thermophoretic force. The actual temperature increase on the cell membrane is below 10 degrees when the laser power is 0.2 mW/μm2, which causes much less thermal damage to the cell (Additional file 1: Fig. S4).
Precise control of the interacting distance between cell and substrate is pivotal for measuring the adhesion strength in our scRAFA due to the strong dependence of the adhesion strength on the cell-substrate distance, which can be precisely controlled by moving the trapped cell up and down through optical tweezer. We conduct in situ optical transmission measurements to extract the cell-substrate distance during the cell rotation (Additional file 1: Note 1) [29]. As an example, we collected the time-dependent transmission spectral shift for cell 1 and cell 2 during cell rotation (Fig. 2f). The closer distance between the cells and substrate, the larger the red shift observed in the transmission dip. The distance between the cell and the substrate can be extracted once there are Δλ matches between the simulated and experimental results (Fig. 2g). By sweeping the distance between cells and substrates d in the simulation, we show that the spectral shift has different values. Specifically, when d = ~ 7.4 nm, the simulated spectral red shift matches with the experimental data. Therefore, the distance between the cell and substrate is ~ 7.4 nm ± 0.1 nm, indicating that our technique can precisely control the interacting distance between cell and substrate. Different from cells with rough surfaces like neutrophils, the cell in the current study is topologically spherical and homogeneous. This is consistent with the previous AFM measurement, which revealed the surface roughness of S. cerevisiae as a root mean square of ~ 0.3 nm [30]. Therefore, our rotational adhesion measurements on the spherical cells can preclude the inaccuracy caused by the anisotropic cell shape.
In biology, cell-adhesion molecules vary and show significant difference in their effects on the duration of the cell adhesion and cell–cell interactions. To demonstrate how scRAFA quantifies the cell adhesion, we functionalize the substrate with ligands and investigate the cell-receptor-ligand interactions through tracking the light-driven cell rotation and the associated cell-substrate adhesion events. We first study the interaction between mannosides on the yeast cells and concanavalin A (ConA) immobilized on the substrate, where the mannosides are uniformly distributed on the cell surface (Additional file 1: Fig. S5). Such homogeneous yeast cells are chosen to preclude the anisotropic cell shape effect in our assay [31, 32]. Once a targeted cell was trapped and driven into rotation mode with the working laser beams, we observed three behaviors: direct adhesion (Additional file 3: Movie S2), transient adhesion (Additional file 4: Movie S3), and continuous rotation with no adhesion (Additional file 5: Movie S4). To quantify the behaviors, we retrieved the time-dependent light intensity signals from the cell images using the in situ recorded video once the cell was trapped at the laser center (Additional file 6: Movie S5). For a cell with continuous rotation, there is no adhesion between the substrate surface and cell receptors. The recorded optical oscillatory signals correspond to a continuous rotation of the cell without adhesion from 0 to 40 s (Fig. 3a). For the cell with transient adhesion, our recorded multiple transient adhesion events indicate the adhesion is mediated by a series of weak and low-affinity adhesion. The alternating oscillatory and constant signals correspond to rotation and transient adhesion events, respectively (Fig. 3b). Specifically, once the cell is trapped close to the substrate, the cell will experience continuous rotation in the initial time series (0–8 s) followed by multiple transient adhesion events (8–33 s). The cell will finally stop rotation and adhere to the substrate (33–40 s). This time-dependent behavior is like in vivo cell rolling adhesion process where the cell sliding promotes the formation of new interactions, thereby slowing dissociation and prolonging bond lifetime to reach the permanent adhesion. For the cell that directly adheres to the substrate without any rotation, there is a constant signal once the cell is trapped at the laser position (Fig. 3c). At 3.2 s, we remove the laser beam and the cell remains at the original position confirming the stable cell-substrate adhesion. For the scRAFA applications, we are mostly interested in the transient adhesion behavior, which dominates in vivo cell adhesion and cell–cell interactions. Therefore, we optimized the concentration of the ConA on the substrate to maximize the transient adhesion events (Additional file 1: Fig. S6). To prove that our observed rotation-adhesion events arise from the specific interaction between mannoside on the cell and ConA molecules on the substrate rather than other effects (e.g., non-specific interaction), we performed a control experiment in which excess D-mannose was added into the cell solution to block the mannoside on the cell membrane and thus the mannoiside-ConA interaction. As a result, most of the yeast cells in the control experiment (ConA + D-mannose), once optically trapped, underwent continuous rotation without adhesion onto the substrate (Fig. 4a).
To quantify the cell adhesion using our scRAFA, we collected the time-dependent intensity signals (Fig. 3b) for the duration that feature cell rotation with transient adhesion for 40 individual cells and analyzed the dissociation constants (koff) as a parameter of the adhesion kinetics for both non-pathogenic (S. Cerevisiae) and pathogenic (C. Albicans) yeast cells (Fig. 4b). Both strains show single exponential distribution with similar dissociate constants (koff), indicating both have the similar levels of manosides [31]. With the larger shear force arising from the increased laser power, the measured dissociate constant shows a higher value, revealing that the mannoiside-ConA interaction is through slip-bonds whose lifetime decreases with the increased shear force (Additional file 1: Fig. S7). We further compare our result with previously measured value by AFM [33]. Interestingly, our measured dissociation constants of mannosides for both strains are one order of magnitude larger than those in the literature [33]. We believe that this variation arises because the dissociation constants measured using scRAFA reflect the cell interaction in the lateral direction (Fig. 1d) while AFM-based measurement was carried out in the normal direction (Fig. 1b). In the presence of the flow-induced shear stress, which mimics in vivo biological process, the interacting molecules (i.e., receptors on cells and ligands on substrates) can reorientate and slip apart, becoming shorter-lived bonds and corresponding to the larger dissociation constants.
We further apply our scRAFA to successfully quantify the heterogeneous adhesion between chitin on single yeast cells and wheat germ agglutinin (WGA) immobilized on the substrate. In contrast to the uniformly distributed mannoside on the cell surface, chitins are non-uniformly distributed on the cell surface and high-concentrated chitins are localized in the bud scars (Additional file 1: Fig. S8). We first conducted a control experiment to confirm the role of specific interaction between chitin on the cell and WGA molecules on the substrate in the cell adhesion. As shown in the control experiment (WGA + GlcNAc), where excess GlcNAc was added into the cell solution to block the chitin on the cell membrane and thus the chitin-WGA interaction, most of the yeast cells underwent continuous rotation without adhesion (Fig. 4c). Then, we collected the adhesion time during the transient adhesion using the scRAFA and analyzed the survival curves of chitin for both S. Cerevisiae and C. Albicans yeast cells (Fig. 4d). The double exponential survival curves indicate that there are two types of chitin distributions along a single cell surface, which is consistent with the previous report on the heterogeneous distribution of chitins on the cell membrane [34]. koff-1, corresponding to the lower adhesion of cell walls with the substrate, is approximately 5–6 times larger than koff-2, corresponding to the higher adhesion of chitinous bud scars with the substrate, which is also consistent with previously reported data [34]. We further analyze the fitting parameters A and B of \({Q}_{a} \left({\tau }_{a}\ge \hspace{0.17em}t\right)=A{e}^{{k}_{off-1}t}+B{e}^{{k}_{off-2}t}\) (see Materials and Methods) to obtain the percentage of normal cell walls and chitinous bud scars over the whole cell area, respectively. The calculated A and B are 0.78 and 0.22, respectively, indicating the normal cell walls account for 78% of the total cell area while chitinous bud scars account for 22%. The higher adhesion associated with chitinous bud scars is also responsible for the fact that the longer transient adhesion of over 0.5 s occurs at the specific orientations of the rotating cells where the scars are closest to the substrate (Additional file 7: Movie S6).
Furthermore, we extend the applicability of scRAFA to measure in-vivo-like shear force occurring in blood vessels and urinary bladders. For this purpose, we collected a series of clinical biofluids (i.e., blood and urine) with a wide range of organisms from bacteria to neutrophils of different sizes and aspect ratios. We succeeded in trapping and rotating these organisms as required for the scRAFA-based adhesion measurement (Fig. 5a and b, Additional file 8: Movie S7). When applying scRAFA to measure the transient adhesion time of organisms in clinical samples, we noticed that multiple factors can contribute to their adhesion and thus cause inaccurate measurement of the adhesion kinetics. Specifically, many bacteria and neutrophils in the clinical samples have irregular shapes. As a result, the organism-substrate distance may vary at different organism orientations during their rotation [35, 36]. In addition, some of the bacteria and neutrophils show complex time-dependent dynamics. For example, the flagella of the bacteria can constantly change the bacterial orientation [37]. The receptors on neutrophils can also dynamically redistribute their spatial distribution on the cell membranes [38, 39]. Therefore, to acquire the more accurate force analysis and adhesion measurement, a more complex modeling of the organisms and their interactions with the substrate will be required to preclude these factors, which is beyond the scope of this report.
Despite current challenge in quantifying the adhesion forces on bacteria and neutrophils, we have successfully measured the adhesion strength of yeast cells in as-collected human urine without any pre-treatment, which is highly relevant to study of the Candida urinary tract infections (UTI) [40]. Our scRAFA has revealed that the urinary yeast cells exhibit diverse durations of transient adhesion (Additional file 1: Fig. S9). Some urinary yeast cells show continuous rotation (Fig. 5c), while others exhibit multiple transient adhesion events during the rotation (Fig. 5d). Optical images of these cells also indicate their different properties (Additional file 1: Fig. S10): the cell in Fig. 5c looks brighter and smoother than that in Fig. 5d. Our measured dissociate constants of mannosides and chitins receptors on urinary yeast cells (Fig. 5e, Additional file 1: Fig. S11) are larger than those of mannosides and chitins receptors on the previously measured two strains (Fig. 4), indicating that in vivo environmental conditions can significantly change the cell adhesion strength compared with the cultured environment. A possible reason for the yeast cells in human urine to have the lower mannosides and chitins adhesion strengths is the inhibition of mannosides and chitins due to their binding with urinary proteins (Additional file 1: Note S2). A comprehensive analysis of the clinical urine solutions at the different Candida UTI stages will be required to further prove the protein-binding effect on the cell adhesion.