Effector T-cells rely on integrins to drive adhesion and migration to facilitate their immune function. The heterodimeric transmembrane integrin LFA-1 (αLβ2 integrin) regulates adhesion and migration of effector T-cells through linkage of the extracellular matrix with the intracellular actin treadmill machinery. Here, we quantified the velocity and direction of F-actin flow in migrating T-cells alongside single-molecule localisation of transmembrane and intracellular LFA-1. Results showed that actin retrograde flow positively correlated and immobile actin negatively correlated with T-cell velocity. Plasma membrane-localised LFA-1 forms unique nano-clustering patterns in the leading edge, compared to the mid-focal zone, of migrating T-cells. Deleting the cytosolic phosphatase PTPN22, loss-of-function mutations of which have been linked to autoimmune disease, increased T-cell velocity, and leading-edge co-clustering of pY397 FAK, pY416 Src family kinases and LFA-1. These data suggest that differential nanoclustering patterns of LFA-1 in migrating T-cells may instruct intracellular signalling. Our data presents a paradigm where T-cells modulate the nanoscale organisation of adhesion and signalling molecules to fine tune their migration speed, with implications for the regulation of immune and inflammatory responses.
The heterodimeric integrin lymphocyte functional antigen-1 (LFA-1; αLβ2 integrin) is a transmembrane heterodimer highly expressed on T-cells. It is essential for T-cell function and is required for naïve lymphocyte antigenic activation, recruitment of immune cells to sites of infection and cytotoxic killing. LFA-1 expressed on effector T-cells binds to intracellular adhesion molecule (ICAM)-1 to induce rapid migration in vivo and in vitro (Hons et al., 2018; Teijeira et al., 2017). The force required for migration is generated by actin polymerisation and myosin-driven actin flow, and integrin adhesion molecules on the cell surface, which couple the actin flow to the underlying substrate (Shannon et al., 2015). Actin and integrins are indirectly linked via a group of proteins collectively known as the molecular clutch, which regulates this mechanical coupling (Ishibashi et al., 2015; Chen et al., 2012; Case and Waterman, 2015). The control of this linkage is integral to achieving high-speed migration (Hons et al., 2018) and is especially important during inflammation (Teijeira et al., 2017). It is also known that integrins can exist in low, intermediate and high affinity states that describe a specific pattern of localisation within migrating T-cells. Low and intermediate affinity integrin molecules are located in the leading edge (LE), whereas high affinity LFA-1 is located in the mid cell body (focal zone; FZ) (Smith et al., 2005).
Control of adhesion is achieved through actin linkers (talin and vinculin) and regulatory kinases [FAK (also known as PTK2) and Src family kinases] (Nordenfelt et al., 2016; Raab et al., 2017), a mechanism originally described in other, much more slowly migrating cell types that characteristically form large micron sized focal adhesions (Kanchanawong et al., 2010). These components were originally described in slow migrating adherent cells that characteristically form large micron-sized focal adhesions (Kanchanawong et al., 2010). In T-cells, these complexes are small (nanoscale) (Burn et al., 2016), similar in size and composition to nascent adhesions (Changede et al., 2015), and their spatial organisation may represent a novel paradigm for the nanoscale regulation of adhesion and migration.
Conventional fluorescence microscopy is limited in resolution to ∼200 nm. Recently, a number of super-resolution approaches have been developed to overcome this limitation. Single-molecule localisation microscopy (SMLM), here implemented as stochastic optical reconstruction microscopy (STORM) (Rust et al., 2006), utilises fluorophore dark-states and sequential imaging to generate lists of molecular coordinates with a mean of 10 nm precision in fixed cells. By using Bayesian statistical cluster analysis (Griffié et al., 2016; Rubin-Delanchy et al., 2015), it is possible to accurately quantify protein clustering at the nanoscale. This includes clustering of intracellular molecular populations via 3D interferometric SMLM imaging featuring isotropic resolution (Shtengel et al., 2009; Griffié et al., 2017). Such populations of intracellular, membrane proximal LFA-1 and FAK may represent distinct nanoscale arrangements recruited to form new adhesions in the membrane.
Here, we investigated the dynamics of retrograde actin flow alongside the nanoscale organisation of actively signalling LFA-1 nanoclusters in primary murine T blasts migrating on ICAM-1-coated coverslips. A range of conditions were used to modulate cell speed including addition of Mn2+ (which push integrins into the high affinity state; Dransfield et al., 1992), high density ligand (to increase adhesion), the addition of the chemokine CXCL12 (to induce signalling through CXCR4) and cytochalasin D (CytoD, which inhibits actin polymerisation; Goddette and Frieden, 1986). We then included in our analysis cells lacking the phosphatase PTPN22, a negative regulator of migration (Burn et al., 2016) where a loss-of-function mutation (Svensson et al., 2011) causes a rapid migration phenotype in T-cells and predisposes humans and mice to a range of autoimmune diseases (Dai et al., 2013; Burn et al., 2011; Brand et al., 2005; Bottini et al., 2006; Begovich et al., 2004). By characterising the nanoclustering of LFA-1 and its regulators as a previously un-investigated aspect of the molecular clutch model, we are able to link changes in their membrane and submembrane arrangement to the functional output of cell speed and actin flow/engagement changes during rapid polarised migration.
The speed of actin retrograde flow is positively correlated with T-cell velocity and negatively correlated with immobile actin
Inside-out and outside-in activation of LFA-1 regulates its affinity maturation and translates into functional effects on adhesion and migration behaviour. Filamentous actin (F-actin) forms networks of actin that concurrently ‘slip and grip’ (Ponti et al., 2004; Jurado et al., 2005) to facilitate movement by linking highly regulated intracellular flowing actin to the extracellular substrate via transmembrane integrin adhesions (Wiseman et al., 2004; Lawson et al., 2012; Alexandrova et al., 2008; Goult et al., 2013; Changede et al., 2015; Sun et al., 2014). Transient engagement of the actin through adhesions is part of a mechanism that transduces force from flowing actin and is termed the molecular clutch (Chan and Odde, 2008; Hu et al., 2007; Brown et al., 2006; Ishibashi et al., 2015; Chen et al., 2012; Case and Waterman, 2015; Havrylenko et al., 2014). Flowing actin therefore represents a population that is not engaged with adhesions, whereas immobile actin (in the external reference frame) represents a population in contact with adhesions.
With these features in mind, we first investigated cell speed on a population level (thousands of cells measured) in polarised migrating murine effector T-cells exposed to a range of stimuli, all in the presence of an ICAM-1 coating on a glass coverslip. Fig. S1 shows that cells slow down when exposed to MnCl2 or high surface concentrations of ICAM-1, and speed up in the presence of chemokine CXCL12 as well as in cells in which PTPN22 has been genetically deleted. To characterise the dynamics of actin flow in more detail, primary mouse T-cells, expressing a Lifeact–GFP transgene were allowed to migrate on glass surfaces coated with ICAM-1 before being imaged by TIRF microscopy (Fig. 1A). As a proxy for actin linked to the molecular clutch, Fourier analysis was used to determine the fraction of immobile actin in cells, since stationary portions of the actin cytoskeleton must necessarily be engaged with extracellular substrate while the cell migrates over that position. Subsequently, the centre of the cell was continually reset, shifting the frame of reference from the laboratory frame, to the cell frame (Fig. 1B), while spatiotemporal image correlation spectroscopy (STICS) was used to measure the velocity and directionality of flowing (unengaged) actin relative to the cell, that is, from the frame of reference of the cell (Fig. 1C).
For cells migrating on ICAM-1 (2 µg/ml), 75±5% (median±interquartile range) of pixels contained more than 50% immobile F-actin over 5 s time windows (Fig. 1D). Slower cells, treated with CytoD, Mn2+ or a high ICAM concentration (100 µg/ml), displayed a significantly increased percentage of pixels displaying immobile actin (P<0.0001; Fig. 1D). In contrast, faster migrating cells (treated with CXCL12 or PTPN22−/− cells) had significantly lower percentages of pixels containing immobile actin (P<0.0001).
STICS was then used to analyse the flow speeds of unengaged actin, relative to the cell (Fig. 1E). T-cells migrating on ICAM-1 (2 μg/ml) alone had an average cell speed of 13.19 μm/min (blue), with retrograde flowing actin moving faster at 18.95 μm/min (orange). Conditions that slowed cell migration also slowed actin retrograde flow speed. However, in all cases, actin flow speed remained much higher than cell speed. The percentage values above the bars show that, in slower cells, treated with CytoD, Mn2+ or a high concentration of ICAM-1, actin speed was 192%, 239%, and 91% higher than the cell speed, respectively. Adding CXCL12 to cells or measuring PTPN22−/− cells led to faster cell migration which was also coupled to faster actin retrograde flow. Furthermore, Fig. 1F shows that in all conditions, the fraction of pixels in which actin was immobile was inversely correlated to cell speed (r values for correlation shown in table, P<0.0001). Therefore, we concluded that decreased actin engagement correlates with increased cell speed. Taken together, these data indicate that retrograde actin flow is always greater than the overall velocity of the T-cell and that the proportion of immobile actin is inversely related to cell speed.
The size and number of LFA-1 membrane nanoclusters increase with T-cell speed
When effector T-cells migrate on ICAM-1-coated glass, they adopt a polarised morphology (Hogg et al., 2004; Comrie et al., 2015; Valignat et al., 2013; Ridley et al., 2003). Previously, studies of integrin affinity have shown a specific pattern where LFA-1 within the LE exists predominantly in the intermediate affinity form, whereas high affinity LFA-1 is detected in the mid cell body/FZ (Smith et al., 2007). It has been reported, in slow moving cells, that integrin-based adhesions change their size as they mature (Kanchanawong et al., 2010). We therefore tested whether T-cells use differential membrane LFA-1 nano-clustering to regulate cell migration.
We investigated integrin LFA-1 nanoscale clustering in conditions that speed up or slow T-cell migration by performing dSTORM and Bayesian cluster analysis (Rubin-Delanchy et al., 2015; Griffié et al., 2016). Effector T-cells were allowed to migrate on ICAM-1-coated glass for 15 min in the presence of conditioned or nonconditioned medium, and these were compared with PTPN22−/− cells. Cells were fixed, surface LFA-1 detected with anti-LFA-1 antibody conjugated to Alexa Fluor 647 (AF647) (clone 2D7), STORM acquisition performed and coordinate pointillist maps generated for Bayesian cluster analysis. Stationary effector T-cells were used as non-polarised T-cell controls (no ICAM-1 on the coverglass). Representative STORM images and cluster maps of a 2×2 µm region of interest (ROI) within a stationary and migrating T-cell are shown (Fig. 2A,B). ROIs from round stationary cells were taken at random, as they do not have a LE and FZ. T-cell velocities were measured for each condition by automatic tracking before fixation for SMLM (as previously shown in Fig. S1).
In polarised T-cells, the FZ displayed different LFA-1 clustering properties to the LE. The number of LFA-1 clusters per ROI was consistently increased in the FZ (Fig. 2C). Compared to the LE, the size of clusters was significantly decreased in the FZ (Fig. 2D), whereas the number of molecules per cluster was significantly increased (Fig. 2E). Together, this represents a median increase in cluster density in FZ clusters (blue bars), compared to LE clusters (orange bars). This is achieved both through a size contraction and an increase in molecular content of the clusters. Compared to stationary cells, clusters in the cell membrane of polarised migrating T-cells increase in number, in size and in molecular content. Together, this indicates that T-cells adopt a regionally discriminated LFA-1 nanoclustering programme to achieve polarised T-cell migration.
The number of clusters significantly decreased in cells that migrated faster owing to treatment with CXCL12 or lack of PTPN22 (Fig. 2C) as compared to slower cells (Mn2+ treated, high ICAM-1 concentration and untreated controls). Slower moving cells had a similar number of LFA-1 nanoclusters to untreated control cells (Fig. 2C). The size of clusters was significantly increased in the LE and FZ of T-cells treated with CXCL12 [32% LE and 22% FZ increase with CXCL12 compared to ICAM-1 alone (P<0.0001)] and PTPN22−/− [33% LE and 23% FZ increase in PTPN22−/− compared to ICAM-1 alone (P<0.0001)] (Fig. 2D). The number of localisations per cluster was increased in faster moving T-cells to a greater extent in the LE than in the FZ [29% LE and 15% FZ increase with CXCL12 compared to ICAM-1 alone (P<0.0001); (24% LE and 5% FZ increase in PTPN22−/− compared to ICAM-1 alone (P<0.0001)] (Fig. 2E).
In slower moving T-cells treated with CytoD or MnCl2, we observed a minor increase in cluster size (<9% LE and FZ), and an intermediate increase of 25% and 17% in the LE and FZ of cells migrating on higher concentrations of ICAM-1 (100 μg/ml), respectively. In CytoD-treated cells, the molecules per cluster increased by 35% and 25% in the LE and FZ, respectively. In MnCl2-treated cells, the molecules per cluster increased by <6%, whereas in cells migrating on an increased concentration of ICAM-1, molecules per cluster was increased by 35% and 15% in the LE and FZ, respectively. Together, this indicates a more substantial increase in cluster size, and a concomitant increase in molecular content (representing maintained cluster density) in fast moving cells. Slow moving cells showed negligible changes in cluster size (Mn2+ or CytoD treated). Mn2+-treated clusters had negligible changes to molecular content, whereas CytoD-treated cells experienced an increase in cluster density. All changes were more substantial in the LE compared to the FZ. All changes are summarised in Table S1.
These results suggest that integrin LFA-1 membrane clusters are modulated on the nanoscale and this influences the migration speed of the cell. Clusters adopt a small dense profile in the high affinity integrin-rich FZ, and a larger less dense profile in the reduced affinity LE. Clusters are larger and less dense in faster migrating cells. The data suggest that these zone-specific cluster patterns may be hardwired into polarised T-cells and indicate that the biophysical arrangement of LFA-1 is regulated in conjunction with integrin affinity to control cell speed. As there are fewer clusters in faster moving cells, we next investigated the 3D profile of nanoclusters of LFA-1 to determine whether this phenomenon relates to the recruitment of LFA-1 to the membrane from submembrane stores.
The size of 3D intracellular LFA-1 clusters increases above the FZ as cell speed decreases
LFA-1 is constantly recycled to and from the cell membrane to facilitate cell migration. Accordingly, we investigated whether the intracellular pool of membrane-proximal LFA-1 became nanoclustered, and if so, whether it was similarly zonally distributed to surface LFA-1 and whether that distribution was modulated with cell migration dynamics. Interferometric photoactivatable light microscopy (iPALM) (Shtengel et al., 2009) was used for the acquisition of 3D single-molecule data with ∼20 nm z-precision to a depth of ∼500 nm from the height of the coverslip. The same battery of conditions were used to slow down and speed up migrating T-cells, which were tracked to confirm their speed prior to fixation. Fig. 3A illustrates a representative iPALM z-projection of cytosolic LFA-1 nanoclusters in a migrating T-cell (left) and the result of Bayesian 3D cluster analysis (Griffié et al., 2017) in a representative 2000×2000×400 nm ROI (right). Cluster descriptors for the size and number of molecules per cluster were then extracted from ROIs in the FZ or the LE and plotted as a function of the z-height from the coverslip.
Clustering in migrating cells was regionally distinct between the LE and FZ. In the LE, clusters in migrating cells compared to stationary cells exhibited almost no increase in size (Fig. 3B, top panel), but this was accompanied by an increase in the number of molecules per cluster (Fig. 3C, top panel; P<0.0001). Clusters in migrating cells also exhibited an increase in size, focused at a region 200 nm above the coverslip (Fig. 3B, lower panel; P<0.0001). The number of molecules per cluster was also increased in the FZ of migrating cells (Fig. 3C; P<0.0001). Together, this indicates that, in the switch from stationary to migrating phenotype, 3D LFA-1 clustering becomes region specific, where clusters above the LE increase in density and clusters above the FZ increase in size and slightly increase their molecular content, concentrated close to the cell membrane.
In cells with a slower migration (treated with CytoD or Mn2+), effects on nanoclusters occur above the FZ, whereas LE clusters remain constant (orange box, top panel Fig. 3B). FZ-specific nanoclusters increase in size, especially in the range 210 to 410 nm above the cell membrane (Fig. 3B, orange box, lower panel; P<0.0001), with a similar number of molecules at all z-heights in the range 0 to 420 nm. Thus, intracellular LFA-1 nanoclusters become larger and less dense in cells that migrate slower. In faster migrating PTPN22−/− cells, the size of intracellular LFA-1 nanoclusters remains similar to those in PTPN22+/+ cells in both the LE and FZ, with a decrease in the number of molecules per cluster (Fig. 3B,C, orange compared to light blue bars). Activating T-cells with CXCL12 (Fig. 3B, orange compared to green) slightly increased the size of LFA-1 molecules per cluster high above the membrane (z=210 to 420 nm) with an LE focused reduction the in molecules per cluster (Fig. 3D, orange to green). Intracellular LFA-1 clusters in PTPN22−/− fast cells stay the same size and remain the same density, and in CXCL12-treated cells become slightly larger higher up.
Deleting PTPN22 increases T-cell velocity, and the nanoscale colocalisation of pY397 FAK, pY416Src and LFA-1
Integrin-based adhesions are regulated by an array of kinases and phosphatases. Phosphorylation of FAK is a feature of active ICAM-1–LFA-1 signalling (Zhang and Wang, 2012; Sánchez-Mártin et al., 2004), and is implicated in the first stages of nascent adhesion formation in non-leukocytes (Swaminathan et al., 2016). FAK is basally activated by the Src family kinase LCK in effector T-cells at Y576 or Y577 (depending on species), and Y925. This primes FAK for full activation though phosphorylation at the auto-phosphorylation site Y397 (herein denoted pY397 FAK) (Chapman and Houtman, 2014), which in turn regulates LFA-1 adhesion and de-adhesion (Raab et al., 2017). Expression of dominant-negative mutants of FAK in T-cells reduces the speed of migration on ICAM-1-coated glass (Rose et al., 2003), and overexpression of a negative regulator of FAK phosphorylation impairs LFA-1 clustering and reduces ICAM-1–LFA-1-driven T-cell homo-aggregation (Giannoni et al., 2003). PTPN22 is a cytosolic phosphatase that acts with C-terminal Src kinase (CSK) to negatively regulate LCK and ZAP-70 downstream of LFA-1 engagement (Burn et al., 2016). CSK is also regulated by FAK (Chapman and Houtman, 2014). Auto-phosphorylated pY397 FAK has also been described as being associated with active integrin in vesicles above the membrane (Nader et al., 2016; Kleinschmidt and Schlaepfer, 2017).
We therefore employed iPALM and Bayesian cluster analysis to study 3D pY397 FAK localisation in stationary and migrating PTPN22+/+ or PTPN22−/− cells. An example pseudocoloured 3D localisation map is shown in Fig. 4A. ROIs were chosen from the cells, and Bayesian cluster analysis carried out to produce cluster maps (Fig. 4B). pY397 FAK staining was verified using FAK inhibitor 14, which blocks Y397 phosphorylation (Fig. S2). ROIs from the LE or FZ regions were quantified in terms of their radius and molecules per cluster as a function of the z-height (Fig. 4C,D).
When comparing stationary cells to cells migrating on ICAM-1, pY397 FAK cluster size (Fig. 4C) was largely unchanged above the LE and FZ, in both PTPN22+/+ and PTPN22−/− cells. The number of molecules per cluster was higher close to the cell membrane (z=0 to 210 nm) in the LE and FZ, meaning that such clusters retained their size but increased in density (Fig. 4D). The size of and number of molecules per cluster was similar between stationary cells and that seen in the LE of migrating cells. Clusters became less dense (of similar size but with fewer molecules per cluster) close to the membrane in the FZ of migrating cells (Fig. 4D, blue shaded box, z=0 to 210 nm; P<0.0001).
PTPN22−/− stationary cells (no ICAM-1) adopted pY397 FAK nanoclusters high above the membrane (orange box in Fig. 4C, comparing blue versus orange bars at z=270 to 390 nm; P<0.0001), which were not present in PTPN22+/+ cells. This population of high clusters present in PTPN22−/− stationary cells has the same number of molecules per cluster as those lower down in z-height, representing a maintenance of density (Fig. 4D). During migration, lack of PTPN22 did not change the size or density of intracellular pY397 FAK clusters.
As nanoclusters resemble nascent adhesions in non-leukocytes, but occur throughout the cell, we next investigated how they are regulated in T-cells. We tested whether the lack of PTPN22 causes changes in the colocalisation of pFAK with LFA-1 and a second set of signalling intermediates, the Src family kinases, using multi-colour SMLM microscopy (Yi et al., 2016, 2017). Wild-type or PTPN22−/− primary murine T-cell blasts were plated on ICAM-1, allowed to migrate and tracked by live-cell imaging. They were then fixed, stained and imaged using madSTORM (Yi et al., 2016), permitting the sequential imaging of pY397 FAK, Src family kinases phosphorylated at pY416 (denoted pY416 Src) and LFA-1. Fig. S3 shows the overall cell speeds, where PTPN22−/− cells were faster than PTPN22+/+ cells directly prior to fixation. Fig. 5A shows a representative cell post fixation after three rounds of sequential staining and imaging.
In wild-type migrating T-cells, LFA-1, pY397 FAK and pY416 Src clusters co-existed on a similar size scale and with similar numbers of molecules per cluster (Fig. 5B,C). LFA-1, pY397 FAK and pY416 Src clusters were consistently larger with more molecules in the LE compared to the FZ. PTPN22−/− cells displayed significantly larger clusters of all three species in both the LE and FZ, combined with significantly more molecules per cluster. PTPN22 deficiency therefore increases the clustering of all three proteins, but especially pY397 FAK and pY416 Src family kinases. Extraction of Pearson's correlation coefficients (PCCs) for the colocalisation of clusters of each species showed that LFA-1 clusters were more colocalised with pY397 FAK clusters when PTPN22 was not present (Fig. 5D). pY397 FAK was also more colocalised with pY416 Src family kinases in PTPN22−/− cells (Fig. 5E). The colocalisation of LFA-1 clusters with pY416 Src family kinase clusters was unchanged (Fig. 5F). Taken together, faster migrating PTPN22−/− cells display larger LFA-1, pFAK and pSrc clusters, which are more often colocalised spatially at the nanoscale. The increase in the number of molecules per cluster in the case of pY416 Src family kinases and pY397 FAK supports existing literature describing the action of PTPN22 on LCK, a well characterised participant in early integrin adhesion signalling that itself upregulates FAK phosphorylation (Chapman and Houtman, 2014). Additionally, we do not believe that the absence of PTPN22 is compensated for by upregulation of any other phosphatase activity, as the data show a clear increase in total phosphorylation of direct (Src family kinases) and indirect (FAK) PTPN22 substrates in PTPN22−/− cells (Fig. S4).
We employed advanced and super-resolved imaging methods to study the nanoscale migration machinery in T-cells. Specifically, we quantified the nanoscale distribution of membrane-localised and intracellular LFA-1 alongside actin flow dynamics in migrating T-cells to understand the relationship between integrin affinity-clustering and engagement with flowing actin and the speed of T-cell propulsion. Our data indicate that: (1) actin flow and engagement drive migration, where engagement is inversely correlated to cell speed, and the flow rate of unengaged actin is positively correlated; (2) T-cells exploit differential clustering of hundreds of tiny nano-adhesions to translate force from the actin cytoskeleton to the substrate; (3) LFA-1 and pY397 FAK nano-adhesions adopt region-specific clustering patterns both in 2D in the membrane and in 3D above the membrane; (4) changes to cell speed induced by a battery of signals modulate adhesion nano-cluster size and density; and (5) the deletion of PTPN22 phosphatase causes faster T-cell migration, and modulates LFA-1 and pY397 FAK clustering in 2D and 3D. Importantly, lack of this phosphatase induced an increase in colocalization of LFA-1, pY397-Src family kinase and pY416-FAK in nanoclusters, coupled to an actin flow speed increase, actin engagement decrease and cell migration speed increase (summarised in Fig. 6).
The front (LE)/middle (FZ) trends in LFA-1 (2D and 3D), pY397 FAK (2D and 3D) and pY416 Src (2D) indicated here are characterised for the first time. For LFA-1, reported trends for affinity occur in the same zones: high affinity integrin occurs in the FZ, whereas low/intermediate affinity integrin predominates in the LE (Persson et al., 2018; Evans et al., 2011; Smith et al., 2005). In addition, nano-adhesions here differ from adhesions in slower moving cells, which are founded as ‘nascent adhesions’ and grow as they mature into larger multi-molecular clutch assemblies towards the middle of the cell. The data here indicate a different system, where high affinity LFA-1 adopts smaller, denser and more numerous cluster arrays in the FZ. Intracellular LFA-1 clusters are larger above the FZ, and intracellular pY397 FAK clusters are denser above the LE. Coupled with the actin work, this indicates that these clusters provide an area of increased adhesion in the FZ; it may be that this adhesion imbalance between the LE and FZ enables fast forward cell migration.
In conditions that increase cell speed, nano-adhesions become larger and are joined by larger clusters of phosphorylated kinases (pY397 FAK and pY416 Src family kinases). The integration of multiple molecules into the same nanocluster may necessitate its increase in size in this way. That this effect is also seen in cells knocked out for PTPN22, which directly targets LCK, a Src family kinase that itself phosphorylates FAK, indicates that the size, density and number of LFA-1 clusters is modulated by such kinases and phosphatases on the nanoscale. Pools of intracellular LFA-1 decrease in density in rapidly migrating cells, and intracellular pY397 FAK clusters stay the same size, indicating a further level of control of adhesion based on the nature of LFA-1 undergoing recycling.
It is also important to note that T-cells are capable of several modes of migration, depending on their environmental niche. Here, we study ICAM-1–LFA-1-dependent migration, coupled to actin treadmilling and engagement. High speed polarised T-cell crawling is a phenomenon that occurs in the lymphatic capillaries, during lymph node entry, tissue entry and fast dendritic cell (DC) scanning by T-cells (Teijeira et al., 2017). Such integrin-dependent migration is especially important during inflammation, including in autoimmune disease (Teijeira et al., 2017), and is a physiological requirement for high speed migration (Hons et al., 2018). Integrin-free slower migration relies mainly on RhoGTPase-mediated actin remodelling and predominates in the dense, fibrous interstitium (ICAM-1 blockade does not affect cell migration in this location). In the lymph nodes, both integrin-independent and integrin-dependent fast migration occur, where ICAM-1 blockade reduces T-cell migration speed considerably. In vitro, the presence of ICAM-1 potentiates migration speed by 50 to 30% in confined 3D (under agarose; Hons et al., 2018), although in such a system the cells do not require integrin to migrate slowly. Previously, the simple unconfined ICAM-1-coated glass system used here has been used to characterise integrin affinity states (Persson et al., 2018; Evans et al., 2011). Here, we used it to characterise the contribution of nanoclustering and its effect on actin engagement and cell speed.
Overall, we demonstrate that T-cells use individually regulated integrin-based, kinase/phosphatase-coupled ‘nano-adhesions’ to achieve fast migration. Such nano-adhesions are similar to previously described nascent adhesions (Sun et al., 2014) in size and molecular composition, but differ in that they are regulated on a sub-diffraction length scale throughout the cell membrane, and not only in the LE. The differential inclusion of proxy markers for integrin activity (phosphorylated FAK and Src family kinases) indicates the highly spatial nature of such adhesion regulation (summarised in Fig. 6). We posit that this nano-adhesion system, as opposed to a classical adhesion system, with maturation from nascent adhesions to focal complexes and then focal adhesions, represents a novel way for T-cells to quickly turn over individual adhesions to achieve dynamic feats of migration depending on external or internal cues. Our results agree with the conventional understanding of how integrins associate with the actin treadmill as a ‘molecular clutch’ to facilitate adhesive traction, with the important difference being that T-cells utilise nanoscale ‘clutches’ of LFA-1 adhesive units that are at least 100-fold reduced in size and polarised in their cellular location. We suggest that these features potentiate the speed and agility of T-cell migration.
Determining what are the live-cell dynamics and rare events that occur in the regulation of nanoclusters will be key to understanding how cells use spatiotemporal molecular organisation to change their behaviour. New high-throughput techniques (Holden et al., 2014; Gunkel et al., 2014; Yasui et al., 2018) to low-power imaging (Chen et al., 2014) and new fluorescent proteins (Zhang et al., 2016; Chang et al., 2012; Tiwari et al., 2015) may soon enable these kind of investigations, and will be especially important in connecting nanoscale regulation to whole-cell function in diverse settings (Shannon and Owen, 2019), including in autoimmune disease.
MATERIALS AND METHODS
Mice were bred and maintained in the Biological Services Unit of King's College London in compliance with UK Home Office regulations and local ethically approved guidelines. Ptpn22−/− mice maintenance and breeding have been described elsewhere (Clarke et al., 2017; Sanchez-Blanco et al., 2018). Male β-actinCre/+; Lifeact-mEGFP/Y C57BL6 males were a gift from the laboratory of Dr Karen Liu (Centre for Craniofacial and Regenerative Biology, King's College London, UK), which were crossed with Ptpn22−/− C57BL6 females; offspring were backcrossed to generate Ptpn22+/+ β-actinCre+;Lifeact-mEGFP+ and Ptpn22−/− β-actinCre+; Lifeact-mEGFP+ males for experiments. Tail blood was genotyped for EGFP expression by flow cytometry. Single-cell suspensions of mouse lymph nodes were stimulated with 1 μg/ml of concanavalin A. After 48 h, T-blasts were re-suspended in medium supplemented with 20 ng/ml human recombinant IL-2 (Proleukin) for an additional 3 to 5 days before being used for migration assays.
Primary murine T-cell culture
Single-cell suspensions were made from lymph nodes harvested from Ptpn22+/+, Ptpn22−/−, and Ptpn22+/+ Lifeact-mEGFP EGFP+; Ptpn22−/− EGFP+ mice and stimulated with 1 μg/ml of concanavalin A (Sigma-Aldrich, L7647) at 3×106 cells/ml for 48 h in complete medium (RPMI 1640-Glutamax, 10% FBS; penicillin, streptomycin, 100 μM sodium pyruvate, 20 mM Hepes and 20 μM 2-mercaptoethanol). After 48 h, T-blasts were washed and re-suspended at 1×106 to 2×106 cells/ml in complete medium constantly supplemented with 20 ng/ml human recombinant IL-2 for an additional 3 to 5 days before being used in migration assays.
T-cell migration assay
Migration medium was RPMI 1640-Glutamax, 100 μM sodium pyruvate, 25 mM Hepes, 20 μM 2-mercaptoethanol; 2 μg/ml human recombinant IL-2; equilibrated overnight at 37°C 5% CO2. Ibidi μ-Slide eight-well glass bottom chamber slides of #1.5H thickness were coated with 2 μg/ml murine recombinant ICAM-1-Fc (R&D systems 796-IC-050) for 1.5 h at 37C. Conditioned migration medium was made by adding supplements [150 ng/ml CXCL12 (Peprotech 250-20A-10), 5 μM MnCl2 (Sigma M1787) and 0.5 μM CytoD (Sigma-Aldrich, C2618)] to cells in migration medium for 10 min prior to addition of 2×105 T-cells to ICAM-1-coated slide chambers.
Wide-field live-cell phase-contrast time-lapse microscopy and automatic tracking
Imaging was performed using a Nikon Eclipse TI-e microscope. 200,000 T-cells were plated onto slide chambers, transferred to a pre-warmed microscope chamber and settled for 5 to 10 min before acquisition of time-lapse videos at 37°C in 5% CO2 under a 10× objective for 15 min (50 ms exposure; 25 s between frames). Automatic tracking to quantify T-cell migration was performed using a custom ICY modular program. Image files were imported in the ND2 format with attached metadata to the ICY interface and cell tracking data was then exported to Graphpad Prism or MATLAB for statistical analysis.
Analysis of actin engagement and flow
The analysis windows or ‘time of interest’ (TOI) for actin flow and engagement was five frames. Immobile actin was calculated in the external reference frame of the cell by taking the zero component of the temporal Fourier transform. A threshold of 50% was applied to each pixel for it to be counted as immobile. For STICS, we adapted a stationary cell programme (Wilson and Theriot, 2006), which keeps the cell centroid in the centre of the ROI. Actin flow analysis was then carried out using STICS (Hebert et al., 2005).
Fixation and labelling for STORM
Migrating T-cells were fixed using 3% PFA plus 0.1% glutaraldehyde in 80 mM kPIPES, 2 mM MgCl2, EGTA 0.5 mM and 2 mM sucrose at pH 6.8 for 10 min followed by 3%-PFA+0.1% glutaraldehyde in Borax 100 mM for 10 min. Slides were washed once with PBS between each treatment.
For immunofluorescence, T-cells were quenched for auto-fluorescence with 1 mM NaBH4 for 15 min, 0.1 M L-glycine for 15 min, blocked with 10% normal goat serum (Thermo Fisher Scientific, 50062Z) for 30 min at room temperature then stained with primary directly conjugated antibody at a 1:100 dilution in 0.1% BSA for 1.5 h at room temperature.
Antibodies used were anti-mouse CD11a (LFA-1) 2D7 clone (1:1000, Biolegend, UK, 101002), ABfinity rabbit monoclonal anti-pTyr397-FAK clone 31H5L17 (1:1000, Thermo Fisher Scientific, 700255) and rabbit monoclonal anti-pY416-Src clone D49G4 (1:1000, Cell Signaling Technology, 6943BF) antibody. Immunofluorescence validation profiles for these antibodies are available at the company websites upon entering the product codes.
STORM and iPALM
Fixed and stained cells were placed in an oxygen scavenging buffer comprised of three parts, which were stored separately. Part A was comprised of 4 mM Tris (2-carboxyethyl) phosphine hydrochloride (Sigma 51805-45-9), 50% glycerol (Sigma 56-81-5), 25 mM KCl (Sigma 7447-40-7), 20 mM Tris-HCl (Sigma 1185-53-1), 20 μg/ml catalase (Sigma 9001-5-2) and 1 mg/ml glucose oxidase (Sigma G2133) all at pH 7.5, made up to a total volume of 5 ml and stored in aliquots of 50 μl at –20°C. Part B contained 100 mg/ml glucose and 10% glycerol and in practice was made to a total volume of 40 ml and stored in aliquots of 400 μl volume at –20°C. Part C contained 1 M Cysteamine-HCl, and in practice was made up directly prior to imaging to 1 ml volume. For imaging, an aliquot of each of Part A and Part B was thawed and added to 100 μl of freshly made Part C, supplemented with 450 μl PBS to a total volume of 1 ml. Imaging was performed on a Nikon N-STORM microscope using a Nikon Plan Apo 100× 1.49 NA oil immersion TIRF objective. Cells were imaged under TIRF illumination with a 15 mW 647 nm laser with the 405 nm laser set to maintain photo-blinking and transition to the triplet state by gradually increasing its power from 0.1 to 1 mW. Emitted fluorescence was collected at wavelengths above 660 nm, with EM gain 300 on an Andor iXon EM-CCD camera at room temperature. Acquisition time was for ∼2.5 min; 10,000 frames were collected at 10 ms exposure time per frame and pixel size was 160 nm. Molecular coordinates were calculated using ThunderSTORM localisation software (Ovesný et al., 2014) with multi-emitter fitting analysis (Huang et al., 2011) enabled. Cluster analysis was performed using Bayesian analysis in 2D (Rubin-Delanchy et al., 2015; Griffié et al., 2016) or 3D (Griffié et al., 2017). Cluster analysis software is available for download as described in the articles listed above. For iPALM, custom gold nanorod fiducial embedded coverslips (Moore et al., 2018) were coated with ICAM-1. iPALM imaging was performed at the Advanced Imaging Centre at Janelia Research Campus including a Nikon 60× 1.49 Apo TIRF Objective, 647 nm excitation and 405 nm activation and 50 ms exposure time. x, y and z coordinates were derived using custom Peakselector localisation software, which is available upon request. It is important to note that clustering patterns in the first 60 nm of iPALM data closely matched those in the TIRF data. While detection efficiencies were different between the two techniques, reliable TIRF data closely matched iPALM data at the coverslip.
Single-cell correlative live tracking and multicolour dSTORM by multiplexed madSTORM
T-cells were first tracked during migration. Cells were fixed, and their positions were recorded, and then multiplexed antibody madSTORM, using fiducial nanodiamonds for image alignment (Yi et al., 2016, 2017), was used to achieve multicolour imaging. Three-colour pointillist maps were analysed channel per channel with Bayesian cluster analysis (Griffié et al., 2016), and colocalisation was analysed (Rossy et al., 2014).
D'Agostino and Pearson tests were used to check for normality in pooled data (from three or four mice in each condition, where LE and FZ measurements were treated as separate conditions). Kruskal–Wallis testing and Dunn's post hoc testing was used to compare pooled, non-normal data between each condition. All statistical analyses were performed with Graphpad Prism software. It is important to note that due to high numbers of clusters being tested in the case of the radius and molecules per cluster, statistical testing gave very low P-values between conditions, therefore these results are also discussed in terms of percentage changes. For multiple comparisons between conditions, one-way analysis of variance (ANOVA) and Kruskal–Wallis testing was used. All statistical analyses were performed with Graphpad Prism software.
PTPN22-knockout mice were a kind gift of Rose Zamoyska. Nanodiamonds were a kind gift of Keir Neumann, NIH. Male β-actinCreER/+; Lifeact-mEGFP/Y mice were a gift from Dr Karen Liu. We acknowledge use of the Nikon Imaging Centre (NIC), King's College London. For the STICS code and support, we acknowledge Paul W. Wiseman and Elvis Pandzic (McGill University), with cell stationarity code written by Laurence Yolland and Brian Stramer (King's College London). iPALM imaging was done in collaboration with the Advanced Imaging Center at Janelia Research Campus, a facility jointly supported by the Gordon and Betty Moore Foundation, and the Howard Hughes Medical Institute.
Conceptualization: A.P.C.; Methodology: M.J.S.; Software: J.G., J.A., D.J.W., D.M.O.; Validation: M.J.S.; Formal analysis: M.J.S., J.P., J.G.; Investigation: M.J.S.; Resources: J.A., T.P., D.J.W., R.Z.; Writing - original draft: M.J.S., A.P.C., G.H.C., D.M.O.; Writing - review & editing: A.P.C., G.H.C., D.M.O.; Supervision: A.P.C., G.H.C., D.M.O.; Project administration: D.M.O.; Funding acquisition: D.M.O.
This work was supported by a European Research Council Starter Grant #337187, and Arthritis Research UK Programme grant 20525. M.J.S. was supported by the King's College London Bioscience Institute and the Guy's and St Thomas’ Charity Prize PhD Programme in Biomedical and Translational Science.
Supplementary information available online at http://jcs.biologists.org/lookup/doi/10.1242/jcs.232991.supplemental
The authors declare no competing or financial interests.