Spatial confinement and temporal dynamics of selectin ligands enable stable hematopoietic stem cell rolling

Hematopoietic stem/progenitor cell (HSPC) homing is initiated by tethering and rolling of the cells on endothelium through selectin-ligand interactions. Although multiple factors that affect the rolling behaviour of the cells have been identified, molecular mechanisms that enable slow and stable cell rolling remain elusive. Here, using a microfluidics-based single-molecule live cell fluorescence imaging, we reveal that unique spatiotemporal dynamics of selectin ligands on the membrane tethers and slings, which are distinct from that on the cell body, play an essential role in the rolling of the cell. Our results suggest that the spatial confinement of the selectin ligands to the tethers and slings together with the rapid scanning of a large area by the selectin ligands increases the efficiency of selectin-ligand interactions during cell rolling, resulting in slow and stable rolling of the cell on the selectins. Our findings provide novel insights and contribute significantly to the molecular-level understanding of the initial and essential step of the homing.


Introduction
Delivering circulatory cells to specific sites in the body is central to many physiological functions, from immunity to cancer metastasis, which is achieved by their interactions with the surface of endothelium under the presence of external shear forces. (Ley et al., 2018;Quail & Joyce, 2013) These cellular interactions are controlled by a number of adhesion molecules that include selectins and integrins and their corresponding ligands. (Kolaczkowska & Kubes, 2013) So far, molecular mechanisms of cell adhesion have been investigated by characterizing migration behaviour at the cellular level in the presence of shear force exerted to the migrating cells and/or characterizing binding behaviour of the adhesion molecules with their ligands at the molecular level by applying external force to the bonds using, for example, single-molecule force spectroscopy technique. (Alon, Hammer, & Springer, 1995;Marshall et al., 2003;McEver & Zhu, 2010) However, under physiological flow conditions, interactions between adhesion molecules and their ligands occur under a spatiotemporal rather heterogeneous cellular environment. Thus, without capturing real-time nanoscopic spatiotemporal interactions of adhesion molecules and their ligands at the molecular level in the cellular environment, one cannot develop a complete picture of complicated cellular interactions that exist.
The initial step of homing is mediated by the binding of homing receptors expressed on endothelial cells, E-and P-selectin (Lehr & Pienta, 1998;Schweitzer et al., 1996), to their ligands expressed on the HSPCs that include CD44 (Dimitroff, Lee, Rafii, Fuhlbrigge, & Sackstein, 2001;Ponta, Sherman, & Herrlich, 2003) and PSGL-1 (Goetz et al., 1997;Yago et al., 2010). HSPCs are tethered to the endothelium by the selectin-ligand interactions resulting in their rolling along the endothelium at shear stress of several dynes cm -2 generated by the blood flow (Frenette, Subbarao, Mazo, von Andrian, & Wagner, 1998;Hidalgo, Weiss, & Frenette, 2002;Mazo et al., 1998). Although multiple factors that affect the rolling behaviour of the cells have been identified, including spatial clustering of the ligands (Abbal et al., 2006), formation of membrane tethers and slings (Sundd et al., 2012), and shear force-dependent selectin-ligand interactions (Marshall et al., 2003), spatiotemporal dynamics of selectin ligands during this initial step of homing and its contribution to slow and stable cell rolling are not well understood. This is due the lack of an experimental method that enables capturing real-time nanoscopic spatiotemporal interactions of adhesion molecule and their ligands at the molecular level in the cellular environment.
In this study, using a microfluidics-based single-molecule live cell fluorescence imaging technique, we showed that the unique spatiotemporal dynamics of selectin ligands on the membrane tethers and slings play an essential role in the rolling of the cell. We demonstrated that the membrane tethers are formed from single microvilli on the cells and this provides a mechanism to spatially localise selectin ligands, in particular PSGL-1 and CD44, on these tethers and slings. Furthermore, this work also established that due to the detachment of the cell from the actin cytoskeleton during the formation of the tethers, fast and random diffusional motion of selectin ligands is exhibited along these structures (i.e., tethers and slings) in contrast to the slow and confined motion of the ligands on the cell body. Our results suggest that the spatial confinement of the selectin ligands to the tethers and slings together with the rapid scanning of a large area by the selectin ligands increases the efficiency of selectin-ligand interactions during cell rolling, resulting in slow and stable rolling of the cell on the selectins.

Results and Discussion
We developed a microfluidics-based single-molecule fluorescence imaging platform to capture and characterize molecular level spatiotemporal dynamics of selectin ligands that occur during cell rolling on selectins (AbuZineh et al., 2018). To this end, we deposited recombinant human   (Sundd et al., 2012). Membrane tethers that appear during cell rolling on selectins are believed to play a critical role in the stable rolling of the cells. Previous studies demonstrated that live cells roll on a selectin surface more stably compared to fixed cells or to microspheres coated with selectin ligands (Sundd et al., 2013). These observations suggest that tethers help decrease the tension exerted on selectin-ligand bonds and thereby reduce the probability of breaking these interactions. This stronger binding is consistent with previous studies from our lab illustrating that the off rate of binding between selectin ligands and E-selectin is low (AbuSamra et al., 2015;AbuSamra et al., 2017).
Two-color epi-fluorescence imaging of the cell membrane and CD44 of the KG1a cells showed perfect colocalisation of the two images, demonstrating that the tethers and slings are fully covered by CD44 molecules (Fig. 1b, c, Supplementary Note 2). The tethers and slings were formed at all the shear stresses used in this study (1 -8 dyne cm -2 , Supplementary Fig. 7, Supplementary Movie 1). The formation of tethers and slings was also observed for primary human CD34 pos -HSPCs at a shear stress of 2 dyne cm -2 ( Supplementary Fig. 8). These are in contrast to the previous study on neutrophils rolling over P-selectin, which showed the formation of the tethers and slings only at shear stresses higher than 6 dyne cm -2 (Supplementary Note 3) (Sundd et al., 2012). The length of the tethers and slings range in size between several micrometres to tens of micrometres ( Supplementary Fig. 9, 10). Interestingly, on occasion, tethers and slings longer than 100 micrometres were observed ( Supplementary Fig. 9). Timelapse fluorescence images of CD44 clearly demonstrated that a tether detached from the Eselectin surface is converted into a sling within several hundreds of milliseconds in both KG1a cells and primary human CD34 pos -HSPCs ( Supplementary Fig. 11, 12 Supplementary Movie 2, Supplementary Note 4) (Marki, Gutierrez, Mikulski, Groisman, & Ley, 2016). Given that the primary human CD34 pos -HSPCs showed very similar structures to our model CD34 pos cell line and that access to these primary cells is limited, we chose to focus on KG1a cells for all subsequent experiments.
While the slings are persistent structures, we also observed a retraction of the slings during cell rolling (Supplementary Fig. 13,Supplementary Movie 3). In contrast to the contiguous distribution of CD44 on the tethers and slings (Fig. 1d), we found that PSGL-1 shows a discrete spatial distribution on the tethers and slings (Fig. 1e, Supplementary Movie 4, 5). The PSGL-1 molecules on the tethers and slings displayed perfect spatial overlap with the cell membrane and CD44 (Fig. 1f, Supplementary Fig. 14, 15).
By reconstructing 3D fluorescence images of CD44 (Fig. 1g, h, Supplementary Fig. 16) and PSGL-1 ( Supplementary Fig. 17), a clear and obvious formation of the tethers and slings during cell rolling along with the spatial distributions of the selectin ligands on the tethers and slings was possible. The side views of the 3D images visibly show the formation of multiple tethers and slings during cell rolling over E-selectin (Fig. 1g,h,Supplementary Fig. 16,Supplementary Note 5). These images also unambiguously demonstrated that the tethers and slings are entirely covered with the CD44 and PSGL-1 molecules in a distinct spatial distribution (i.e. contiguous and discrete distributions for CD44 and PSGL-1, respectively). In addition, the time-lapse 3D images captured the elongation of tethers (Supplementary Fig. 18) and change in the Z-axis position of the slings (Supplementary Fig. 19) during cell rolling. The entire tethers were consistently covered by the selectin ligands during their elongation. We also often observed the formation of anchoring points on the tethers (Fig. 1d, Supplementary Fig. 20).
The fluorescence images of CD44 and PSGL-1 on the tethers and slings sometimes show bright spots at their tips ( Fig. 1h and 2a, Supplementary Fig. 21), indicating clustering of the selectin ligands at the tethering points. The clustering of selectin ligands at the tethering points was further investigated by calculating the number of PSGL-1 molecules at each spot on the tethers and slings. Our calculation demonstrated that each PSGL-1 spot distributed on the tethers and slings mainly corresponds to a single PSGL-1 molecule (Fig. 2b). The distributions were fitted to Poisson distribution, suggesting that the number of PSGL-1 molecules at each spot is determined stochastically (i.e. absence of a mechanism that colocalislocalies the PSGL-1 molecules together, Supplementary Note 6). On the other hand, we found that multiple PSGL-1 molecules are present at the tethering point (4.1 molecules, Fig. 2c) and anchoring point (2.3 molecules, Fig.   2d), suggesting the spatial clustering of the selectin ligands at the tethering and anchoring points. The significant deviation from Poisson distribution observed for the tethering and anchoring points (Fig. 2c,d) indicates the presence of a specific mechanism that supports the spatial clustering of PSGL-1 at these points, which facilitates the formation of the tethering and anchoring points (Supplementary Note 6). Data from our lab has shown that the binding of Eselectin to its ligands is limited by a slow-on rate (AbuSamra et al., 2015;AbuSamra et al., 2017) and this observation that clustering of the ligands (i.e. CD44 and PSGL-1) occurs at anchor points helps to explain how the cell in flow can overcome this "slow on" rate and enhance the binding of E-selectin to its ligands ultimately through the increase in the local concentration of the ligands.
To investigate the effect of the clustering of the selectin ligands on the formation of tethers/slings and on the rolling behaviour of the cell, we disrupted the clusters of CD44 by treating the cells with methyl-β-cyclodextrin (MβCD) (Abbal et al., 2006;AbuZineh et al., 2018;Setiadi & McEver, 2008). Cholesterol existing in the cell membrane is extracted by this treatment, leading to the disruption of lipid rafts domains. Super-resolution fluorescence images of CD44 on the  (Fig. 2h, Supplementary Fig. 22b). These results are reminiscent of previous data from our lab showing that CD34 plays a key role in the formation of microvilli structures since once they are knocked down, microvilli structures disappear (AbuSamra et al., 2017). This result strongly suggests that the CD44 molecules localise on the microvilli in the control cells.
We found that the formation of the tethers and slings is extremely inefficient under the MβCDtreated condition (Fig. 2i, Supplementary Note 7) as well as when CD34 is knocked down (Fig.   2j, Supplementary Fig. 23, 24). While the tethers and slings are formed in more than 90 % of the control cells rolling over E-selectin (Supplementary Movie 6), they are observed in less than 10 % of the MβCD-treated cells and CD34 knockdown cells. Further, the length of the tethers and slings formed during the rolling of the MβCD-treated cells were much shorter than those formed in the control cells (Fig. 2k). The fluorescence images of CD44 on the MβCD-treated cells and CD34 knockdown cells never showed bright spots (i.e. clusters of CD44) at the tethering points ( Fig. 2i and 2j). These results demonstrate that the clustering of the selectin ligands contributes to both initial binding to the surface E-selectin (i.e. tether formation) as well as the elongation of the tethers against the tension exerted to the tethering point during cell rolling. Since CD34 knockdown KG1a cells (AbuSamra et al., 2017) and MβCD-treated KG1a cells (AbuZineh et al., 2018) exhibit approximately 3-5 times faster rolling velocity compared with control cells and showed more unstable rolling (i.e. faster detachment) on E-selectin, our findings suggest that the efficient formation of the tethers and slings and their strong resistance to shear stress due to the spatial clustering of the selectin ligands on the microvilli enables the slow and stable rolling of the cells (Supplementary Note 8). Our finding also suggests that the microvilli play a key role in the spatial clustering of the selectin ligands and the formation of the tethers and slings during cell rolling over E-selectin.
Based on these findings, we next sought to investigate the mechanism of the formation of the tethers. To that end, we calculated the total number of the PSGL-1 molecules on a single tether during its elongation (Fig. 3a). The time lapse images of PSGL-1 showed that the total number of the PSGL-1 molecules on the tether estimated by the integrated fluorescence intensity of PSGL-1 over the entire tether is almost constant during the tether elongated from 75 μm to 110 μm (Fig. 3b,c,d). We also calculated the total number of the PSGL-1 molecules on single tethers and slings at different shear stresses ( Fig. 3e) using the lengths of the tethers/slings (Fig. 3f,Supplementary Fig. 9) and the distances between the adjacent PSGL-1 molecules on the tethers/slings at different shear stresses (Fig. 3g, Supplementary Fig. 25). The calculation showed that the total number of the PSGL-1 molecules on single tethers and slings is almost constant over the 1 -8 dyne cm -2 although the tethers and slings were two times longer at 8 dyne cm -2 compared with those at 1 dyne cm -2 (Fig. 3e). Given the localisation of PSGL-1 and CD44 on the microvilli (Abbal et al., 2006;AbuZineh et al., 2018;Miner et al., 2008;Moore et al., 1995), our results strongly suggest that single tethers and therefore slings are formed from single microvilli upon their binding to the surface E-selectin (Fig. 3h). This is supported by our observation that the tethers and slings are always covered entirely by CD44 and PSGL-1 (Fig. 1b, f). If the tethers are formed from multiple microvilli, this would result in discrete patches of CD44 and PSGL-1 clusters on the tethers and slings.
Previous studies suggested that the localisation of selectin ligands to microvilli enables them to interact with selectins during the initial step of homing (Miner et al., 2008;Moore et al., 1995).
Our findings here extend this to the tethers and slings. During cell rolling, CD44 and PSGL-1 are spatially redistributed to the entire tethers and slings by the conversion of single microvilli to the tethers. This redistribution significantly increases the exposure of these selectin ligands to the surface E-selectin, promoting clustering and efficient selectin-ligand binding that leads to mechanisms that help overcome the slow-on rate (AbuSamra et al., 2015;AbuSamra et al., 2017) leading to slow and stable rolling.
Single-molecule imaging revealed that the PSGL-1 molecules show diffusional motion on the tethers and slings ( Due to the distinct rates and modes of the diffusion, the PSGL-1 molecules on the tethers and slings cover much larger displacement areas compared with those restricted to the microvilli (0.14 μm 2 and 0.26 μm 2 for tethers and slings, respectively compared with 0.038 μm 2 for microvilli) during the several hundreds of milliseconds of the lifetime of the tethers (Chen & Springer, 1999;Schmidtke & Diamond, 2000). The difference becomes even larger when more stable structures (i.e. tethers with multiple anchoring points and non-retracting slings that exist for more than several seconds) are formed during cell rolling.
The larger motional freedom (i.e. faster and random diffusion) of the PSGL-1 molecule on the tethers and slings compared with those localised to the microvilli and the merger of multiple tethers into a single tether (i.e. occurring during cell rolling; Fig. 4g, Supplementary Movie 7) imply the involvement of the actin cytoskeleton. We thus conducted a two-color fluorescence imaging experiment of CD44 (labelled by either Alxa-Fluor-488-or Alxa-Fluor-647-conjugated antibodies) and actin (labelled by either silicon-rhodamine-conjugated jasplakinolide (Lukinavicius et al., 2014) or Alexa-Fluor-488-conjugated phalloidin). The fluorescence image of CD44 and actin on the control KG1a cells showed perfect colocalisation (Fig. 4h). On the other hand, we did not observe any fluorescence signal of actin from the tethers and slings in most cells rolling over E-selectin. In some rare cases, we found that actin exists along the tethers and slings of the rolling cells, but in a form of small fragmented patches ( Fig. 4i). This result unambiguously demonstrates that the cell membrane is detached from the cortical actin cytoskeleton during the formation of the tethers, consistent with previous force spectroscopy studies that indicated the detachment (Evans, Heinrich, Leung, & Kinoshita, 2005;Shao, Ting-Beall, & Hochmuth, 1998). Since PSGL-1 and CD44 on the control cells are directly or indirectly anchored to the actin cytoskeleton in the microvilli by actin binding ERM proteins (Moore et al., 1995;Snapp, Heitzig, & Kansas, 2002;Wang et al., 2014), the formation of the tethers and therefore the detachment of the membrane from the actin cytoskeleton would cause a significant enhancement in the motional freedom of PSGL-1 and CD44, consistent with the results obtained from the single-molecule tracking analysis.
To form a selectin-ligand bond, the ligand molecules on the rolling cell have to be located within tens of nanometres distance from the surface E-selectin molecules (Cummings & McEver, 2009).
If the locations of the PSGL-1 molecules are fixed, the probability to find the surface E-selectin molecules within this distance is relatively low. Thus, the fast and random motion of the selectin ligands on the tethers and slings, which enables them to scan large surface area during a limited lifetime of tethers (with anchoring points) and slings, should facilitate efficient binding of the selectin ligands to the surface E-selectin and thus enable slow and stable rolling of the cell (Fig.   4j).
In conclusion, our results suggest that during the initial step of HSPC migration, the selectin ligands are spatially confined to the tethers and slings through the development of the tethers from single microvilli and their motional freedom on the tethers and slings is enhanced by the detachment of the membrane from actin cytoskeleton during the formation of the tethers. These mechanisms enable the efficient utilization of the limited amount of the selectin ligands, which contributes to overcome the slow on rate of binding to the underlying E-selectin (AbuSamra et al., 2015;AbuSamra et al., 2017) to achieve effective selectin-ligand binding, resulting in this slow and stable cell rolling.

Materials and Methods
Cell culturing and treatments. Human acute myelogenous leukemia cell lines, KG1a cells (ATCC CCL-246.1) were maintained in the lab using cell culturing facility in RPMI 1640 (1×) media (Gibco) that contains 10 % fetal bovine serum (Corning) and antimicrobial agents (100 U ml -1 penicillin, 100 μg ml -1 streptomycin, Hyclone) incubated at 37 o C in presence of controlled 5% CO2. For the disruption of lipid raft microdomains, the KG1a cells were treated with methylβ-cyclodextrin (MβCD). Briefly, 10 6 ml -1 KG1a cells were washed three times by HBSS buffer Transfection was done using the SE Cell Line 4D-Nucleofector TM X Kit (Lonza). The company protocol was followed to achieve the knockdown. Cells were then collected after 48 h and prepared for Western blot analysis to determine the extent of the CD34 inhibition.
Preparation of human CD34 pos HSPCs. Primary human CD34 pos HSPCs isolated from umbilical cord blood (CD34 pos HSPCs) and the mononuclear cells (MNC) from whole cord blood were purchased from ALL Cells (USA). For CD34 pos HSPCs isolation from whole cord blood, the MNCs were washed and filtered through a 100 m cell strainer (BD Falcon) followed by lineage depleton by negative selection using a lineage cell depletion cocktail (Milteny Biotec). Cell rolling assay using bright-filed microscope. The rh E-selectin-deposited microfluidic chambers were connected to a syringe pump (Harvard Apparatus, PHD Ultra) using a silicone tubing (inner diameter of 0.8 mm, ibidi GmbH) and mounted on an inverted optical microscope (Olympus, CKX41). KG1a cells were suspended in HBSS buffer containing 0.7 mM of Ca 2+ (anhydrous CaCl2, Sigma Aldrich) and 1 % (w/v) BSA. The rolling behaviour of the cells in this perfusion buffer was observed using the inverted light microscope that is equipped with a 20× objective lens (Olympus, LCAch N 20X PHP) and XC10 CCD camera (Olympus). The transmitted images were recorded by the CCD camera using CellSens imaging software provided by Olympus. All images were recorded at the frame rate of 15 fps with 30 ms exposure time. The pixel size of the CCD camera is 6.45 µm. The KG1a cells were perfused into the microfluidic chamber mounted on the microscope stage at the flow rate of 100 dyne cm -2 , then the flow rate was decreased gradually to reach a constant shear stress of 4 dyne cm -2 , 2 dyne cm -2 , 1 dyne cm -2 , 0.5 dyne cm -2 , or 0.25 dyne cm -2 where cell rolling behaviour was monitored for 78 seconds in mg ml -1 . 30.7 μl of the antibody solution and 2 µl Alexa-Fluor dye solution (i.e. mixing ratio of ten to one, which corresponds to the dye to antibody molar ratio of approximately 120 to 1) were mixed and incubated for 1 hour at room temperature. After the labelling reaction, 467 μl of HBSS was added to the reaction mixture. Free dyes in the mixture were removed by using a spin filter (Pall Corporation, Nanosep centrifugal devices with Omega membrane, OD010C34) at a centrifuge speed of 10,000 xg for 5 min. The solution was resuspended in HBSS and the centrifugation was repeated several times. The final concentration of the antibodies and the labelling degree were determined by UV-vis absorption spectra of the Alexa-Fluor dyeconjugated antibodies. The degree of labelling was in the range of 4 -9 dyes per antibody. The kit uses immobilized ficin protease to efficiently digest mouse IgG1 into Fab or F(ab')2 fragments, depending on the concentration of cysteine and solution pH. Briefly, the digestion buffer was prepared by dissolving 43.9 mg of cysteine•HCl in 10 ml of the supplied Mouse IgG1 digestion buffer (pH = 5.6) to produce Fab fragments. Then, the immobilized enzyme was dispensed into the spin column, centrifuged and washed using the prepared digestion buffer. The antibody (1 mg ml -1 , 125 μl) was desalted using the accompanied Zeba column. The flow through of at least 100 μl that contained the antibody was incubated with the immobilized enzyme for 3-5 hours at 37 o C on an end-over-end mixer, which digested the antibody into Fab fragments. After the incubation, the column was put into a 2 ml collection tube and centrifuged at 5000 × g for 1 min to collect the digested antibodies. The collected flow through was incubated with protein A column for 10 min at room temperature on an end-over-end mixer, followed by centrifugation at 1000 × g for 1 min to separate the Fab fragments from Fc and undigested antibodies. The flow through that contained Fab fragments was collected. After the fragmentation and purification, concentrations of the anti-PSGL-1 antibody (KPL-1 clone, IgG1) and its fragments were determined by measuring the absorbance at 280 nm using UV-vis spectrophotometer (Thermo Scientific, NanoDrop 2000).
Fluorescence labelling of the Fab fragment of the anti-PSGL-1 antibody by the Alexa-Fluor dyes conjugated to N-hydroxysuccinimide (NHS) was conducted in a manner similar to the labelling of the whole antibody. The Fab fragment and the dye were mixed at the mixing ratio of thirty to one, which corresponds to the dye to antibody molar ratio of approximately 120 to 1. The degree of labelling with this mixing ratio was in the range of 1 -3 dyes per Fab fragment. Fluorescence imaging. Fluorescence imaging experiments were conducted using a home-built wide-field fluorescence microscopy setup (Serag, Abadi, & Habuchi, 2014). Continuous wave (CW) solid-state laser operating at either 488nm (60 mW, Cobolt, MLD), 532 nm (100 mW, Cobolt, Samba), or 638 nm (60 mW, Cobolt, MLD) that passed an excitation filter (Semrock, LD01-640/8, FF01-532-12.5 or FF01-488/6 for the 638 nm, 532 nm or 488 nm excitation, respectively) and a beam expander (Thorlabs) was introduced into an inverted microscope (Olympus, IX71) through a focusing lane (f = 300 mm). The laser was reflected by a dichroic mirror (Semrock, FF660-Di02-25x36, Di01-R532-25x36, or FF506-Di03-25x36 for the 638 nm, 532 nm or 488 nm excitation, respectively) and the sample was illuminated through an objective lens (Olympus, 100× NA = 1.49, UAPON 100XOTIRF, 60× NA = 1.3, UPLSAPO60XS2 or 40× NA = 1.25, UPLSAPO40XS). The excitation power was adjusted to 1 -2 mW cm -2 at the sample plane using an acousto-optical tuneable filter (AOTF; AA Optoelectronics) inserted in the excitation beam pass. The fluorescence from the sample was captured by the same objective, separated from the illumination light by the same dichroic mirror, passed an emission bandpass filter (Semrock, FF01-697/58-25, FF01-580/60, or FF01-550/88 for the 638 nm, 532 nm or 488 nm excitation, respectively), and detected by an EMCCD camera (Andor Technology, iXon3 897). The fluorescence images were recorded with either 133-nm or 333-nm pixel size at 30 ms exposure time. The exposure of the EMCCD camera and the illumination of the sample by the excitation laser were synchronized by the AOTF using a laser control system (Andor Technology, PCUB-110). The image acquisition was done using the Andor iQ3 software. 3D fluorescence images were obtained by recording epi-fluorescence images of the cells at different Z-axis positions with 0.5 -1.0 μm step size using a piezo objective scanner (PI PIFPC® P721) and reconstructing 3D images using ImageJ plugin.

Fluorescence labelling of the primary cells and KG1a cells. Prior to immunostaining of the
Two-color fluorescence imaging experiments were conducted by introducing 638-nm and 488nm lines of the lasers coaxially into the inverted microscope in the same way as the single-color excitation. The samples were excited through the objective lens (Olympus, 100× NA = 1.49, UAPON 100XOTIRF or 40× NA = 1.25, UPLSAPO40XS). The fluorescence from the sample was captured by the same objective, separated from the illumination light by a multiband dichroic mirror (Semrock, Di03-405/488/561/635-t1-25x36), and passed through a TuCam dualcamera adaptor (Andor Technology) equipped with a filter cassette containing a dichroic mirror (Semrock, FF580-FDi01-25x36) to separate the fluorescence into two channels. The separated fluorescence from the samples was detected by two EMCCD cameras (Andor Technology, iXon3 897) through emission bandpass filters (Semrock,.
Determination of the number of PSGL-1 molecules on tethers and slings. The number of PSGL-1 molecules in each spot on the tethers and slings formed during KG1a cells rolling over E-selectin was calculated by comparing the integrated fluorescence intensity obtained from each PSGL-1 spot with the intensity obtained from single Alexa-Fluor-dye-conjugated anti-PSGL-1 antibody. To determine the fluorescence intensity obtained from single Alexa-Fluor-dyeconjugated anti-PSGL-1 antibody, the antibody was deposited on the surface of the microfluidic chamber at a concentration of 0.02 μg ml -1 . Fluorescence signals from the single deposited antibodies were captured at the experimental conditions identical to those for the imaging experiment on the immunostained KG1a cells rolling over E-selectin. We used only well-focused fluorescence images of the PSGL-1 spots on the tethers and slings for this analysis to ensure an accurate estimation of the number of the PSGL-1 molecules.

Single-molecule tracking analysis.
Single-molecule tracking analysis of the PSGL-1 molecules moving along the tethers and slings of KG1a cells rolling over E-selectin and localised on the microvilli of control KG1a cells were conducted by generating single-molecule diffusion trajectories using a versatile tracking algorithm (Jaqaman et al., 2008;Serag & Habuchi, 2017). The algorithm uses a mixture-model fitting algorithm to localise and track multiple particles in the same field of view. It can detect the merging and splitting of particles during motion. To achieve these tracking targets, the algorithm localises and tracks all the local maxima in the single-molecule image including maxima that are partially overlapping along the tethers and slings. The diffusion rate (i.e. diffusion coefficient) and mode were analyzed by mean square displacement (MSD) analysis (Kusumi, Sako, & Yamamoto, 1993). The MSD was calculated from the generated diffusion trajectories by using equation (1) (∆ ) = 〈( + − ) 2 + ( + − ) 2 〉, (1) where xi+n and yi+n demote the spatial positions after time interval Δt, given by the frame number, n, after starting at positions xi and yi. . The MSD-Δt profiles were fitted to random (equation (2)), and confined (equation (3)) diffusion models.
(∆ ) = 4 ∆ , (2) where D is the diffusion coefficient, where L is the side length of the confined area.
Super-resolution fluorescence imaging and analysis. Super-resolution (SR) fluorescence localisation microscopy imaging of fixed immunolabelled KG1a cells was performed in an imaging buffer composed of TN buffer (50 mM tris (pH 8.0) and 10 mM NaCl), oxygen scavenging system (glucose oxidase (0.5 mg ml -1 , Sigma), catalase (40 μg ml -1 , Sigma), 10% (w/v) glucose), and 100 mM β-mercaptoethanol (Sigma) as a reducing reagent. The imaging solution was prepared immediately before the imaging experiments. The imaging experiments were conducted on a custom-built wide-field illumination fluorescence microscope on an inverted optical microscope platform described above (Abadi, Serag, & Habuchi, 2018;AbuZineh et al., 2018). The fluorescence images were recorded using a 150 × 150 pixel region of the EMCCD camera with 83-nm pixel size at 10-ms exposure time. The fluorescence image sequences with 10,000 frames were recorded for the reconstruction of SR localisation microscopy images.
The SR images were reconstructed by using either a custom-written MATLAB (MathWorks) code or Localizer software (Dedecker, Duwe, Neely, & Zhang, 2012). The positions of the AF-647 molecules were determined by 2D Gaussian fitting of the images. We removed fluorescence spots whose width was significantly larger (>200 nm) than the point spread function (PSF) of the optical system (PSF, ~130 nm) from the analysis. The effect of the stage drift in the xy plane was corrected by reconstructing the subimages using 5000 localisations. In the two-colour SR imaging, TetraSpeck microspheres (diameter, 100 nm) deposited on a cleaned coverslip was used to calibrate the shift between the two channels. Using fluorescence images of the TetraSpeck microspheres recorded simultaneously on the two cameras, we generated a registration map that corrects the shift between the two images and applied the registration map to the images obtained from the cell samples. Schematic illustration describing the custom-built microfluidics-based single-molecule imaging setup. Continuous-wave lasers were introduced into an inverted optical microscope through a focusing lens, resulting in a wide-field illumination of the samples through the objectives (40x, NA 1.25, silicone immersion, or 100x, NA 1.49, oil immersion). Fluorescently-labelled or stained KG1a cells were injected to a microfluidic chamber whose surface was coated by rh Eselectin molecules (15 molecules μm -2 ). The shear stress was set to 1 -8 dyne cm -2 . The fluorescence images were recorded by an EM-CCD camera at a frame rate of 33 Hz. The axial position of the microscope stage was controlled by an objective piezo scanner. (b) Two-color fluorescence images of the cell membrane (stained by Vybrant DiO dye) and CD44 (immunostained by Alexa-Fluor-647-conjugated anti-CD44 antibody, clone 515) captured during the KG1a cell rolling over the surface-deposited E-selectin molecules. The stained and labelled cells were injected into the microfluidic chambers whose surface was coated by the rh E-selectin molecules with the density of 15 molecules μm -2 . The cells were injected into the chambers at a shear stress of 2 dyne cm -2 . (c) Time-lapse two-colour fluorescence images of the cell membrane (stained by Vybrant DiO dye) and CD44 (immunostained by Alexa-Fluor-647-conjugated anti-CD44 antibody, clone 515) captured during the cells rolling over the surface-deposited E-selectin molecules. The merged images are displayed in the right panels. The stained and labelled cells were injected into the microfluidic chambers whose surface was coated by the rh E-selectin molecules with the density of 15 molecules μm -2 . The cells were injected into the chambers at a shear stress of 2 dyne cm -2 . (d) Immunofluorescence image of the CD44 molecules on the tethers (top left) and slings (