Lymphocytes perform reverse adhesive haptotaxis mediated by LFA-1 integrins

An efficient immune response requires the rapid recruitment of circulating leukocytes from the blood system to lymph nodes and inflamed tissues. Leukocytes arrest on the endothelium of blood vessels (the cellular monolayer covering the lumen) then migrate spontaneously with a rapid ‘crawling’ mode and eventually cross the endothelium – an event called diapedesis. On the walls of blood vessels, leukocytes crawl distances of several tens or hundreds of micrometers, proposedly in search for optimal diapedesis sites (Massena and Phillipson, 2012; Sumagin and Sarelius, 2010; Sumagin et al., 2010) called ‘transmigratory cups’ (Carman and Springer, 2004) or ‘portals’ (Sumagin and Sarelius, 2010). These sites are composed of endothelial cells with vertical microvilli-like projections enriched in adhesion molecules, notably the integrin ligands ICAM-1 (intercellular adhesion molecule-1) and VCAM-1 (vascular cell adhesion molecule-1). ICAM-1 is particularly important for transmigration in nearby portals (Park et al., 2010; Sumagin and Sarelius, 2010), suggesting that integrins might guide leukocytes towards and through extravasation sites. After diapedesis, leukocytes migrate in three-dimensional environments of tissues and organs, where integrin-mediated adhesion was shown to be dispensable for motility (Lämmermann et al., 2008; Park et al., 2010). Nevertheless, integrins remain crucial in vivo for efficient homing, and the microenvironment of inflamed tissues is known to enhance motility in an integrin-dependent manner (Hons et al., 2018; Lämmermann et al., 2013; Overstreet et al., 2013). An important role of integrins has been further established in the context of tumors, like the chronic lymphocytic leukemia (CLL), where VLA-4 (also known as integrin α4β1) expression on CLL cells had prognostic impact by favoring B cell adhesion and proliferation on stromal cells (Burger et al., 2009; Gattei et al., 2008). Furthermore, the traffic of lymphocytes in lymph nodes is orchestrated by specialized stromal cells expressing integrin ligands and chemokines, and it has been hypothesized that integrins participate in guiding lymphocytes on the stromal network (Bajénoff et al., 2006). Altogether, guided migration by integrins is potentially relevant in several steps of leukocyte survey functions; however, there is no direct evidence yet that integrins can guide leukocytes in environments that present modulations in adhesion ligand density. The first goal of this work was therefore to investigate the existence and conditions of emergence of integrin-mediated guidance in amoeboid cells such as leukocytes.

Directed motion is a hallmark of the immune response, and several guiding cues are well characterised. Leukocytes are sensitive to mechanical cues such as blood flow (mechanotaxis) (Dominguez et al., 2015; Gorina et al., 2014; Valignat et al., 2013, 2014), to soluble biochemical cues including bacterial fragments and chemokines (chemotaxis) (Liu et al., 2012; Malawista et al., 2000; Malet-Engra et al., 2015; Massena and Phillipson, 2012; Poznansky et al., 2000; Tharp et al., 2006), and to anchored signaling molecules such as the chemokines CCL21 and IL8 (haptotaxis) (Canton, 2008; Roy et al., 2017; Schwarz et al., 2017; van Gils et al., 2013; Weber et al., 2013). Directed motion guided by adhesion molecules is documented for mesenchymal cells, such as cancer cells, fibroblasts, muscle cells and endothelial cells (King et al., 2016; MacNearney et al., 2016; Oudin et al., 2016; Wen et al., 2015; Wu et al., 2012), as well as for Schwann cells (Motta et al., 2019) and neurons (Aznavoorian et al., 1990; Brandley and Schnaar, 1989; Carter, 1965, 1967; Klominek et al., 1993; Mccarthy and Furcht, 1984; O'Connor et al., 1990; Smith et al., 2004; Thibault et al., 2007), but it has not been demonstrated yet for amoeboid cells in general or for leukocytes in particular. For clarity, hereafter we refer to any guidance phenotype triggered by anchored molecules as ‘haptotaxis’, and to the guidance phenotypes specifically triggered by adhesion molecules as ‘adhesive haptotaxis’. For mesenchymal migration, adhesive haptotaxis was initially explained by a tug-of-war in the cell adherence zone; the areas with lower grip destabilizing spontaneously under traction in favor of areas with a higher grip (Caballero et al., 2015; Carter, 1965, 1967; O'Connor et al., 1990). Mesenchymal cells are indeed characterized by strong cell–substrate adhesion mediated by mature focal adhesions and by strong traction forces transmitted to the substrate by contractile actin stress fibers. In contrast, amoeboid cells migrate at high speed with weak adhesion and low traction forces transferred to the substrate (Lämmermann et al., 2008; Paluch et al., 2016; Ricoult et al., 2015; Smith et al., 2007). Leukocytes can even swim without interaction with a substrate (Aoun et al., 2019 preprint). A tug-of-war mechanism, with an interplay between strong adhesion and high traction, is therefore irrelevant for amoeboid cells, and the existence of adhesive haptotaxis is all but evident for leukocytes.

Adhesive haptotaxis might, however, rely on mechanisms more sophisticated than a tug-of-war, especially when integrins are involved (King et al., 2016). Integrins are indeed more than adhesion molecules, as they can transduce signal when they engage to their ligands and are submitted to a force – a phenomenon called mechanotransduction (Ross et al., 2013; Zaidel-Bar et al., 2005). Adhesive haptotaxis mediated by integrins can therefore involve signal transduction by integrins relayed by internal signaling pathways to transform external cues into cytoskeletal reorganization, as in chemotaxis (Artemenko et al., 2016; Aznavoorian et al., 1990; Klominek et al., 1993; Malawista et al., 2000; Massena and Phillipson, 2012; Thibault et al., 2007). Fundamentally, different paradigms can therefore sustain integrin-mediated adhesive haptotaxis, either active mechanisms relying on mechanotransduction through integrins acting as mechanoreceptors, or passive mechanisms based on physical competition between cell edges attached with integrins acting as adhesion linkers only.

To scrutinize guided migration mediated by integrins, we performed quantitative in vitro experiments and took advantage of technological advances in surface-patterning methods (Strale et al., 2016). Our aim was not to answer whether leukocytes were guided by adherent versus non-adherent substrates, but more specifically whether integrins could sense modulations of adhesion on a globally adherent substrate. This latter case of globally adherent substrates is indeed directly relevant for leukocytes crawling on blood vessels, where adhesion is required at all times to resist the shear stress of blood flow (Granger and Kubes, 1994), as well as in tissues around scars (Berry et al., 2006; Sullivan et al., 2014; Wen et al., 2015) and in lymphoid organs, where cells are generally exposed to integrin ligands (Bajénoff et al., 2006). In all these cases, haptotaxis requires mechanosensitivity to variable adhesion and cannot be attributed to a trivial migration bias by lack of adhesion. For this task, usual microcontact-printing techniques with stamps were not adapted, because they produce substrates with binary patterns of adherent and non-adherent areas. In order to produce patterns with multiple levels of integrin ligands on a glass substrate, we developed an original ‘subtractive printing protocol’ based on the light-induced multiple adsorption setup (Strale et al., 2016). We focused our study on coatings of ICAM-1 or VCAM-1 proteins, ligands of integrins LFA-1 (α_Lβ₂) and VLA-4 (α₄β₁), respectively, and studied human primary effector T lymphocytes. Our results demonstrate the existence of sensitive and versatile adhesive haptotaxis mediated by integrins towards increasing adhesion with VLA-4 and, quite counterintuitively, towards decreasing adhesion with LFA-1. Mechanisms proposed for mesenchymal haptotaxis cannot explain this latter ‘reverse haptotaxis’ phenotype, and a novel mechanism remains to be deciphered.

In order to pattern substrates with controlled amounts of adhesion molecules, we used a set-up of light-induced molecular adsorption, or LIMA (Strale et al., 2016), which optically coupled a 375 nm UV light source with a digital micro-mirror device (DMD) on an inverted microscope (Fig. 1A). A specific patterning protocol of ‘subtractive printing’ was developed to engineer substrates with modulated and integrin-specific adherence for lymphocytes. Glass coverslips were first treated with aminopropyltriethoxysilane (APTES), then sequentially incubated with solutions of protein A, bovine serum albumin (BSA) and chimeric Fc–ICAM-1 or Fc–VCAM-1 (Fig. 1B). UV illumination was then used to degrade the functionality of ICAM-1 or VCAM-1 in presence of the photoactivator. Degradation was UV-dose dependent, and gray-leveled illumination allowed us to obtain substrates with gradual adhesion properties. Fig. 1C shows an immunofluorescence microscopy image of an ICAM-1 printed substrate, illuminated with stripes of different UV doses. The maximum dose was of 800 mJ/mm², and the pattern of illumination corresponded to nine stripes of doses ranging between 10, 20, 30, 40, 50, 60, 70, 80, 90 and 100% of the maximum dose intercalated with stripes without illumination (Fig. 1A). The different levels of fluorescence brightness in Fig. 1C,D correspond to different surface densities of functional protein A, that then lead to different adsorbed amounts of Fc–CAM adhesion molecules. The amount of grafted CAM molecules was further assessed by imaging substrates after incubation with specific antibodies against ICAM-1 and VCAM-1. A calibration procedure, based on microfluidic channels filled with known amounts of fluorescent molecules, allowed us to convert immunofluorescence data into densities of CAM molecules on the substrates (see details in Materials and Methods). The accessible densities of ICAM-1 and VCAM-1 molecules ranged between 0 and ∼1000 molecules/μm² (Fig. 1D), which corresponds to physiological levels found on the walls of high endothelial venules (Dustin and Springer, 1988).

For ICAM-1 and VCAM-1, the adhesion rate was null below densities of 150–200 molecules µm⁻², and increased until reaching a plateau at 0.6 for ligand densities around 1000 molecules µm⁻² (Fig. 1F,G). These data attest that ligand density variations allowed us to tune cell adhesion.

Haptotaxis corresponds to cell guidance by gradients in surface concentration of a component, and a complete characterization of a haptotaxis phenotype requires screening of different parameters of the stimulus sensed by individual cells, namely the local concentration of the component, and the steepness and direction of its gradient. Determination of the existence of haptotaxis can, however, be screened using fewer variables by selecting the optimal conditions for its emergence. Optimal conditions require a maximal cue difference between cell rear and front, as well as cue levels that are neither too low (i.e. below detection level) nor too high (above saturation level). We therefore prepared crenel profiles of adhesion, with alternating stripes of two different adhesion rates that were between the minimal and maximal adhesion rates that sustained cell adhesion. These patterns yielded adhesion gradients of maximal steepness at the boundaries of the stripes. Furthermore, in order to optimize the occurrence of events of cells having to choose direction at boundaries between stripes, the width of the stripes was set at 20 µm, slightly larger than the average diameter of adherent cells (15±2 µm).

For all assays, the adhesion rate of stripes with high adhesion was kept at the maximum value of 0.6, whereas the adhesion rate of the stripes with low adhesion was varied from one assay to the other between 0 and 0.6. The range of assays varied, therefore, from an adhesion contrast of 0%, which corresponded to a homogeneous substrate (with the maximal adhesion rate of 0.6), and an adhesion contrast of 100%, which corresponded to alternating stripes of maximal and null adhesion rates. Adhesion rates and CAM molecule densities were systematically measured in each experiment and for each stripe type. Hereafter, guidance properties are plotted versus substrate adhesion properties, and the correspondence between adhesion properties and ligand densities are shown in Fig. 1G.

Control experiments corresponded to substrates with homogeneous adhesion (i.e. an adhesion contrast of 0%). Rose plots of direction were isotropic for these controls, and the anisotropy index was equal to 0 (Fig. 2A; Movies 1 and 2, left panels), which demonstrated that cell orientation was random. The ‘adhesion/non-adhesion assays’ corresponded to substrates with alternating stripes of null and maximum adhesion rate (i.e. an adhesion contrast of 100%) (Fig. 2B; Movies 1 and 2, middle panels). Cells crossed both stripe types and explored large areas; however, their migration was not random. The corresponding rose plots were highly anisotropic with an anisotropy index equal to 0.5 for ICAM-1 and VCAM-1, which revealed that lymphocytes were sensitive to the presence of non-adherent zones. Migration of mesenchymal cells are known to be guided by adhesion stripes because they are excluded from non-adherent stripes. Here, lymphocytes were not excluded from non-adherent zones, due their ability to swim over non-adhesive substrates (Aoun et al., 2019 preprint). Orientation in adhesion/non-adhesion assays might then result from a difference in the speed of leukocytes on adherent and non-adherent zones, which are significant (Aoun et al., 2019 preprint), and from other migration biases induced by the contrast of adhesion. However, further investigations on the mechanisms involved would distract us from our main goal. These assays of adhesion/non-adhesion are indeed not informative on adhesive haptotaxis, which refers to guidance on globally adherent substrates. The ‘homogeneous control’ assays and ‘adhesion/non-adhesion’ assays are only of interest here as the two extreme controls of actual haptotaxis assays.

We then focused on globally adherent substrates, for which the alternating stripes were both adherent for lymphocytes, albeit with different adhesion rates. This configuration is reminiscent of blood vessel endothelia, where a minimal adhesion is always required for crawling leukocytes to resist blood flow. The ‘haptotaxis assay’ of Fig. 2C corresponds to alternating stripes with adhesion rates of 0.6 and 0.3 (i.e. adhesion contrast of 50%). The trajectories of cells do not display an obvious directional trend (see also Movies 1 and 2). Some fractions of trajectories followed stripes, whereas others appeared rather perpendicular to stripes. Phenotypes along and across stripes were different from one cell to another, and individual cells displayed different phenotypes over time. Nevertheless, the rose plots, corresponding to the average of hundreds of cell trajectories, were clearly anisotropic for both ICAM-1 and VCAM-1, with an anisotropy index of ∼0.4. To ensure that cells adhered on each type of stripe, the adhesion footprint was then probed in situ by reflection interference contrast microscopy (RICM). In Fig. 2D (see also Movies 3 and 4), the cell body was located using brightfield images (gray), the distribution of adhesion by fluorescence images of protein A (red) and the footprint of cell adhesion by RICM images (green). Merged images show clearly that cells adhered on both types of stripes. These data confirm unambiguously that the crawling lymphocytes are sensitive to modulations of substrate adhesion on a globally adherent substrate.

To identify the range of molecular coatings that stimulate integrin-mediated haptotaxis, we then varied the adhesion contrast of patterned substrates (Fig. 3). For haptotaxis assays in which the substrates were globally adhesive, the anisotropy index increased steadily with the adhesion contrast, from 0 for an adhesion contrast of 0% to around 0.5 for an adhesion contrast of ∼55%. For larger adhesion contrasts, data corresponded to ‘adhesion/non-adhesion assays’, because the stripes with low molecular density coatings were not adhesive. We previously argued that guidance in the ‘adhesion/non-adhesion’ assay might result from a crawling bias introduced by non-adherent zones, whereas a real haptotaxis sensing phenotype is revealed in the ‘haptotaxis assay’. However, the data from ‘adhesion/non-adhesion’ assays seem to extrapolate from the ‘haptotaxis assay’ data in terms of anisotropy index. At this point, this observation suggests instead that haptotaxis on globally adherent substrates and orientation bias by stripes of null adhesion might result from a similar mechanism and differ only in terms of magnitude.

I_PA extreme values of 1 and −1 corresponded to cells positioned exclusively on high and low adhesion stripes, respectively, whereas a value of 0 corresponded to an equal distribution on both types of stripes. In Fig. 4B, each I_PA data point corresponds to the average result from 100 images taken every 10 s, each image containing more than 200 cells. I_PA increased with increasing adhesion contrast from the minimal value 0 (when there was no difference between stripes) to a maximum value of 0.5. Hence, lymphocytes on VCAM-1 always preferred higher adhesion zones. This tendency resembles the one usually reported in the literature for mesenchymal cells. However, the I_PA of mesenchymal cells would be equal to 1 in ‘adhesion/non-adhesion’ conditions because they are excluded from non-adherent zones. For lymphocytes, the maximum value of I_PA reached only 0.6 in ‘adhesion/non-adhesion’ conditions because lymphocytes can swim over non-adhesive substrates (Aoun et al., 2019 preprint), and are not excluded from non-adherent zones.

In a representative haptotaxis assay on ICAM-1 at adhesion contrast of 40% (Fig. 4A left and middle), a quasi-static state was reached with 70% of cells on less adherent stripes and 30% on more adherent stripes. Here again, cell velocity was independent of the adhesion contrast of the stripes (Fig. 4A right), so that the imbalance of cell distribution between stripes could not be due to a speed difference. This result indicates, therefore, a counterintuitive preference of cells for low adhesion zones. Systematic assays confirmed this stunning behavior of ‘reverse haptotaxis’ over a large range of adhesion contrasts (Fig. 4B). I_PA started at 0 (when there was no difference between stripes), then decreased with increasing adhesion contrast to a minimal negative value of −0.35 for an adhesion contrast of ∼20%, then eventually increased to a plateau value around 0.1 (Fig. 4B). To the best of our knowledge, such a phenotype of adhesive ‘reverse haptotaxis’ has never been observed in the literature. Furthermore, reverse haptotaxis is remarkably sensitive, because it appears for adhesion contrasts as low as 10%. In order to rule out possible artifacts due to the patterning method of subtractive printing, we performed similar experiments using substrates obtained by LIMAP additive printing on anti-fouling PEG-coated substrates (Strale et al., 2016). This method has the advantage of avoiding any exposure of ICAM-1 proteins to UV/PLPP photo-scission effects. Similar phenotypes of reverse haptotaxis were obtained in the full range of adhesion contrast between 10 and 50% (Fig. S1). These results strongly support that ‘reverse haptotaxis’ is not dependent on the patterning method but is specifically controlled by LFA-1.

The fact that two different integrins mediate haptotaxis phenotypes with opposite orientations suggests that the mechanism underlying integrin-mediated haptotaxis might involve properties of integrins beyond their adhesion function. In light of the ICAM-1 results, the hypothesis that guidance on VCAM-1 substrates might simply result from an adhesive bias also becomes more arguable. To shed light on a potential role of integrin-mediated mechanotransduction, we monitored intracellular calcium activity of lymphocytes crawling on stripe patterns. Fig. 5 shows that fluctuations in calcium activity for cells crossing stripes of various adhesions remained unaffected on both ICAM-1 and VCAM-1 patterned substrates, whereas control assays with ionomycin showed an instant and intense peak in calcium flux. These data show that intracellular calcium flux is not involved in integrin-mediated haptotaxis. Because calcium signaling is shared by many intracellular signaling pathways (Munaron, 2011), these results challenge the existence of a role for mechanotransduction in integrin-mediated haptotaxis of lymphocytes.

All previous experiments were performed using crenel profiles of adhesion to maximize the sensitivity of the detection of haptotaxis. In order to check that lymphocytes also haptotax on adhesion gradients with finite steepness, we then patterned substrates with 20 µm wide stripes of low and high adhesion, as in previous experiments, separated by 40 µm wide gradients of increasing and decreasing adhesion (Fig. 6A,B).

Because cells had an average spread diameter of 10–15 µm, they experienced in these assays a gradual change of adhesion in the gradient zones extending over 40 µm. Haptotaxis towards decreasing or increasing gradients eventually led to accumulation of cells in the areas of lower or higher adhesion rate, respectively. With a crawling speed of 15 µm min⁻¹, cells scanned the gradient zone within a few minutes, so the transient period between an initially homogeneous distribution of cells (just after sedimentation/adhesion) and a quasi-static distribution of cells after sorting by haptotaxis was short and not easily recorded in experiments. The direction of cell haptotaxis was instead assessed by comparing the quasi-static distribution of cells (MacNearney et al., 2016) in the lower and upper halves of the adhesion gradients (Fig. 6B). While the analysis of the transient phase can only give access to a limited number of events because it is limited in time, the analysis of the quasi-static regime has no time limit, which allows one to increase statistical power just by increasing recording time. In a single representative experiment (Fig. 6C), with 20 gradient stripes and ∼200 cells, the distribution of cells was markedly unbalanced in each frame, with marked preferences for high adhesion zones on VCAM-1 and low adhesion zones on ICAM-1. These data confirm direct haptotaxis on VCAM-1 and reverse haptotaxis on ICAM-1 in response to an adhesion gradient instead of crenels. In order to further probe the robustness of reverse haptotaxis on ICAM-1 for different steepness of gradients, systematic assays were then performed on substrates patterned with saw-tooth profiles and an adhesion contrast of 25% between extremes. Reverse haptotaxis was consistently observed for adhesion gradients extending over 0, 40 and 100 µm (Fig. 7).

Observation of the directed paths of single cells required tracking cells during the transient phase between the initial adhesion of cells and the arrival in their respective accumulation zone (either up or down the gradient depending on their orientation versus a gradient). After this transient phase, cells remained, on average, in the same preferential accumulation zone due to the on-going guiding effect, so that their average direction was random, and directed path analysis was ill adapted to reveal haptotaxis. The time window to observe directed migration of single haptotaxing cells was therefore limited. This time window could be increased by increasing the spatial width of gradients, but then the gradient slopes and cell directivity would decrease. Linear gradients of width 100 µm were a good compromise because they offered several minutes to observe transient directed paths (given a cell speed of 15 µm min⁻¹), and haptotaxis is significant (Fig. 7). To grasp the transient regime in its entirety (i.e. the path of each cell from initial adhesion to arrival at the edge of the gradient), we injected cells into the devices at room temperature to keep cells in an inactive and non-adherent state, then the temperature was increased to 37°C to activate the cells. As cells switched one-by-one to an active state, they acquired a polarized morphology, adhered to the substrate and migrated. Cells were individually tracked for 5 min, and the few cells reaching the edge of the gradients before the end of the 5 min were rejected from analysis to avoid bias by limit conditions. We prepared ICAM-1 coated substrates with a homogeneous adhesion rate of 0.6 as control, and 100 µm wide linear gradients with adhesion rate extremes of 0.3 and 0.6. Fig. 8 shows that the directions of cells during the initial 5 min of motion were random along the x and y axis on a homogeneous substrate, as well as along the y axis on a gradient substrate. Only directivity along the x axis of the gradient substrate gave a non-null forward migration index (FMI) value (x_FMI=0.14), which was significant as compared to the x_FMI in the control and to y_FMI on the gradient. Furthermore, the preferential direction of the directed paths on the gradient oriented cell migration towards lower adhesion. These analyses of cell direction during the transient regime are less sensitive than the analysis of cell distribution at equilibrium (Fig. S2) but they confirm the existence of reverse haptotaxis mediated by LFA-1.

S.B. Carter (Carter, 1965) defined adhesive haptotaxis in 1965 using cancer cells migrating on cellulose acetate substrates and gradients of palladium. Cells moved in the direction of increasing adhesion, and Carter proposed that directed movement was controlled by the relative strength of their peripheral adhesion, hence the name ‘haptotaxis’ (from the Greek: haptein, to fasten; taxis, arrangement). Carter foresaw that movement towards surfaces offering greater adhesion could be a general phenomenon applicable to all metazoan cells whose mobility depends on contact with a surface. Indeed, adhesive haptotaxis was later acknowledged for fibroblasts, neurons and stem cells (Amarachintha et al., 2015; Doyle et al., 2009; King et al., 2016; O'Connor et al., 1990; Smith et al., 2004; Thibault et al., 2007), and is often attributed to a tug-of-war between rival parts of the leading edge, those parts that have the strongest grip on the substrate being expected to win (Caballero et al., 2015). Mesenchymal cells are strictly excluded from non-adhesive zones and follow adhesive stripes by default because their motility requires adhesion (Doyle et al., 2009). Similarly, growth cones of neurons also avoid non-adherent substrates (O'Connor et al., 1990). In contrast to these systems, it was recently shown that leukocytes can travel across non-adhesive zones and propel themselves without interaction with a substrate by a motility mode described as swimming (Aoun et al., 2019 preprint). Directed motion controlled by adhesion was therefore not conceptually straightforward for leukocytes and its existence remained unproven.

We developed here in vitro assays with adhesion patterns characterized by a high spatial resolution (1.5 μm). Repetition of sharp adhesion gradients allowed us to maximize the probability of haptotaxis events, which was instrumental in revealing the subtle adhesive haptotaxis phenotype of leukocytes. In these assays, lymphocytes crawled adhesively on patterns with ligand concentrations between 200 and 1200 molecules/µm² without change of speed, and that they migrated back and forth between zones of different ligand concentrations. This ubiquitous capacity to explore various adhesive areas in vitro might be important for immune cells in vivo to achieve an exhaustive spatial survey independently of the adhesion properties of the environment. Nevertheless, our assays also showed that lymphocytes were sensitive to modulations of substrate adhesion, their orientation being significantly biased by adhesive patterns. These results attest that amoeboid cells, such as the lymphocytes used here, are capable of adhesive haptotaxis, even if their exploration capacity is barely restricted by adhesion properties.

Efficient guidance by adhesion modulations on globally adhesive substrates is most relevant for leukocytes crawling on blood vessel endothelia because adhesion is required at all times to resist detachment by blood flow. Interestingly, the densities of integrin ligands of 1000 molecules/µm² tested in our in vitro assays correspond to ICAM-1 and VCAM-1 expression levels found on activated endothelial cells in vivo (Dustin and Springer, 1988). These results suggest, therefore, that integrin-mediated haptotaxis can participate in physiological processes involving spatial modulation of integrin ligands, such as transmigration at transmigration cups (Sumagin and Sarelius, 2010), guidance by stromal networks in secondary lymphoid organs (Bajénoff et al., 2006; Hons et al., 2018; Malhotra et al., 2013), or homing to inflamed, wounded or tumoral tissues (Majumdar et al., 2014). In contrast, and according to our results, the idea that lymphocytes could be strictly guided by paths paved with integrin ligands, like trains on a railway, seems invalidated because lymphocytes can easily migrate ‘off-track’ on adhesive patterns in vitro. The guidance in vitro was nevertheless found to be sensitive to minute modulations of adhesion with adhesion contrasts as low as 10%. Taken together, our results show that lymphocytes are highly sensitive to subtle modulations of adhesion in a globally adherent landscape, and that the conditions where integrin-mediated haptotaxis is functional exist in vivo. There is, however, no proof yet that adhesive haptotaxis does contribute significantly to physiological processes, and this hypothesis deserves further exploration.

In complete contrast to Carter’s foresight that movement towards surfaces offering greater adhesion would be a general phenomenon, lymphocytes with LFA-1 mediated adhesion displayed a marked preference towards surfaces offering lesser adhesion. This unique phenotype of ‘reverse haptotaxis’ is clearly incompatible with control of directed migration by the relative strength of peripheral adhesions – the tug-of-war mechanism. An alternative mechanism is therefore at work, in which integrins cannot only play a passive role of adhesion linkers. King et al. (2016) proposed an original mechanism of adhesive haptotaxis for fibroblasts, in which the kinetics of lamellipodial growth was actively regulated by integrins. In this model, engagement of integrins with their ligands promoted activation of the Arp2/3 complex and lamellipodial growth (via intermediate activation of FAK, SFK and Rac). This signaling pathway yielded cell reorientation towards higher density of integrin ligands by differential kinetics of lamellipodial growth. This active lamellipodial sensing mechanism proposes an original alternative to the passive tug-of-war mechanism proposed by Carter. However, it explains the orientation of cells towards increasing adhesion, and it is therefore not applicable to the reverse haptotaxis of lymphocytes.

Chemotaxis and chemokine-driven haptotaxis are known to rely on signaling pathways that can orient cells towards or against gradients of a chemical cue (Amarachintha et al., 2015; Basara et al., 1985; Malet-Engra et al., 2015; Poznansky et al., 2000; Schwarz et al., 2016, 2017; Tharp et al., 2006; Weber et al., 2013; Woolf et al., 2007). A tempting hypothesis would be that integrin mechanotransduction might similarly be relayed by specific downstream signaling pathways, orienting cells towards or against gradients of adhesion. In support of this hypothesis, the work of Dixit et al. (2011) describes the involvement of integrin mechanotransduction in the directed migration of neutrophils, and the work of Artemenko et al. (2016) supports that mechanical and chemical cues trigger common signaling elements. The hypothesis that mechano- and chemo-taxis share common signaling machineries to orient cells versus an external cue is therefore plausible, and it proposes an elegant explanation for the versatility of cells to orient towards or against a cue gradient. However, this scenario remains largely hypothetical, because a full link between molecular mechanotransduction and cellular orientation has not yet been deciphered for leukocytes. Moreover, we addressed here the question of mechanosignaling by monitoring intracellular calcium signaling, and found that adhesive haptotaxis of T lymphocytes was not correlated to intracellular calcium signaling. This result is uncommon for cell chemo- and mechano-sensitivity and is rather surprising given that ion channels usually act as ‘hubs’ in pathways of cell responses to mechanical and chemical agonists (Dixit et al., 2011; Lima et al., 2014; Munaron, 2011; Ranade et al., 2015). One option is that haptotaxis involves integrin mechanotransduction with downstream signaling pathways independent of intracellular calcium signaling (Roy et al., 2018). Another option is that mechanotransduction is not involved at all. The question remains open.

Alternatively, a mechanism of versatile adhesive haptotaxis without mechanotransduction could rely on spatiotemporal regulation of integrin affinity states. Nordenfelt et al. (2016) showed that actin-dependent force appears to regulate integrin activity, and they argue that a coupling between integrin activity and cytoskeletal dynamics could regulate eukaryotic cell orientation, which relaxes the requirement for centralized control of complex pathways. This idea is conceptually rich but it lacks direct evidence at this point. In a related vein, we recently described a mechanism that explains versatile decisions of lymphocytes to orient with or against a flow in response to the integrin ligand (ICAM-1 and VCAM-1) composition of the substrate (Hornung et al., 2020). Our mechanism relied on the spatial polarization of integrin affinity states along the cell polarization axis, and crosstalk signaling between integrins LFA-1 and VLA-4. Opposite activation of LFA-1 and VLA-4 at cell edges controlled opposite orientations with flow on ICAM-1 and VCAM-1, and an inhibiting crosstalk between integrins triggered a bistable decision of cells between opposite orientations on mixed ICAM-1/VCAM-1 substrates. Complex spatiotemporal regulation of integrin affinity at the cellular scale explained lymphocyte guidance under flow without mechanotransduction. Mechanistic elements of this rheotaxis phenotype might also play a role in the adhesive haptotaxis phenotype of lymphocytes. A detailed investigation of integrin activation dynamics in the leading and trailing edges of cells performing haptotaxis would be necessary to shed more light on the interplay between ligand density, integrin affinity distribution, and migration orientation.

In conclusion, our data revealed that integrins can act as efficient and multidirectional guidance mediators that may be involved in the guidance processes of leukocytes. These findings could be inspirational for studies of other cell types guided by integrins, such as muscle cells (Chen et al., 2012) or neurons (Huang et al., 2007). Important roles of integrins as adhesive linkers are well identified for leukocytes in the recruitment from blood and lymphatic vessels, where VLA-4 transient adhesion contributes to initial rolling, and LFA-1-mediated firm adhesion is required for crawling and diapedesis. However, these functions concern leukocyte capture, whereas our work sheds light on leukocyte guidance. Integrin-mediated adhesion is also involved in leukocyte trafficking in lymphoid organs and homing to inflamed tissue; however, their exact function remains unclear. We show here that integrins cannot promote rigid migration along pathways paved with their ligands because lymphocytes have the ubiquitous ability to explore their environment irrespective of its adhesive properties. At the same time, integrins are efficient multidirectional guidance mediators and, as such, their role might have been underestimated in various processes of leukocyte migration and guidance.

Blood from healthy volunteers was obtained through a formalized agreement with French Blood Agency (Établissement Français du Sang, agreement 2017-7222). Blood was obtained by the agency after informed consent of the donors, in accordance with the Declaration of Helsinki. All experiments were approved by the INSERM Institutional Review Board and Ethics Committee. Peripheral Blood Mononuclear Cells (PBMCs) were recovered from the interface of a Ficoll gradient (Eurobio, Evry, France). T cells were isolated using a Pan T cell isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany), then activated with antiCD3/antiCD28 Dynabeads (Gibco by Thermo Fisher Scientific, Waltham, MA), according to the manufacturer's instructions. Cells were subsequently cultivated in Roswell Park Memorial Institute Medium (RPMI; Gibco by Thermo Fisher Scientific, Waltham, MA) 1640 supplemented with 25 mM GlutaMax (Gibco by Thermo Fisher Scientific, Waltham, MA) and 10% fetal calf serum (FCS; Gibco by Thermo Fisher Scientific, Waltham, MA) at 37°C, 5% CO2 in the presence of IL-2 (50 ng/ml; Miltenyi Biotec, Bergisch Gladbach, Germany) and used 7 days after activation. At the time of use, the cells were >99% positive for pan-T lymphocyte marker CD3 and assessed for activation and proliferation using CD25, CD45RO, CD45RA and CD69 makers, as judged by flow cytometry.

Glass coverslips (NEXTERION coverslip, #1.5H glass D263, SCHOTT Technical Glass Solutions, Jena, Germany) were first activated by air plasma (Harrick Plasma, Ithaca, NY) for 5 min. Activated glass coverslips were treated in gas phase with 3-aminopropyltriethoxysilane (APTS; Sigma-Aldrich, St Louis, MI) for 1 h and then heated for 15 min at 95°C on a heating plate. Sticky-Slides VI 0.4 (Ibidi GmbH, Martinsried, Germany) were then mounted on treated glass coverslips. The prepared flow chambers were incubated sequentially at room temperature with an Alexa Fluor 647-conjugated protein A (Thermo Fisher Scientific, Waltham, MA) solution at 50 µg/ml for 1 h, a bovine serum albumin (BSA; Sigma-Aldrich, St Louis, MI) solution at 4% (w/v) for 15 min, and Fc-ICAM-1 or Fc-VCAM-1 (intercellular adhesion molecule 1; vascular cell adhesion molecule 1; R&D system, Minneapolis, MN) solution at 10 µg/ml, overnight at 4°C. The flow chambers were rinsed extensively with phosphate buffered saline (PBS; Gibco by Thermo Fisher Scientific, Waltham, MA) after each incubation.

We used an inverted microscope (TI Eclipse, Nikon, France) coupled to a UV laser source and a Digital Micromirror Device (Primo™, ALVEOLE, Paris, France) (Strale et al., 2016). For subtractive printing, gray level patterns – either alternating stripes with infinite slope or alternating gradients with finite slope – were projected on ICAM-1- or VCAM-1-treated substrates in the presence of a soluble photo-activator (PLPP™; ALVEOLE, Paris, France) to gradually degrade the proteins (Pasturel et al., 2020). Samples were then rinsed with PBS solution and passivated with 4% (w/v) BSA (Sigma-Aldrich, St Louis, MI) for 15 min at room temperature. In the additive printed experiments, glass coverslips (NEXTERION coverslip, #1.5H glass D263, SCHOTT Technical Glass Solutions, Jena, Germany) were first activated by air plasma for 5 min, then treated for 2 h in liquid phase with APTS diluted to 1% in water with 0.03% acetic acid and heated for 15 min at 95°C on a heating plate. Sticky-Slides VI 0.4 were then mounted on treated glass cover-slips. The prepared flow chambers were incubated at 4°C overnight with of 0.23 g ml⁻¹ PEG-SVA (mPEG-SVA, molecular weight 5000 Da; Interchim) solution with NaHCO₃ at 10 mM, then rinsed with milliQ water, incubated with 4% (w/v) BSA for 15 min at room temperature and rinsed with PBS. Gray level patterns were then projected on substrates in the presence of photo-activator (PLPP™; ALVEOLE, Paris, France) to locally activate the substrate (Strale et al., 2016). Samples were then incubated with an Alexa Fluor 647-conjugated protein A solution at 50 µg/mL for 1 h, a BSA solution at 4% (w/v) for 15 min, and Fc-ICAM-1 (R&D system, Minneapolis, MN) solution at 10 µg/mL overnight at 4°C. The flow chambers were finally rinsed extensively with PBS after each incubation.

Cell adhesion and migration assays were systematically performed on the same microslide printed for each set of experiments with homogeneous areas for adhesion assays and patterned areas for migration assays. Cells were seeded into the flow chamber at the concentration of approximately 1.5 million cells/ml and incubated for 10 min at 37°C. Flow of prewarmed and CO2 equilibrated culture medium through the flow chamber was controlled using an Ibidi pump system (Ibidi GmbH, Martinsreid, Germany). Cell migration was recorded at 37°C with a Zeiss Z1 automated microscope (Carl Zeiss, Oberkachen, Germany) equipped with a Snap HQ CCD camera (Photometrics, Tucson, AZ), pE-300 white LED microscope illuminator (CoolLED, Andover, UK) and piloted by µManager (Edelstein et al., 2010).

For the cell adhesion assay, during each experiment, upon cell seeding and incubation, the flow chamber was then rinsed with culture medium with a gentle flow at 1 dyn cm⁻² (0.1 Pa) to remove non-adherent cells. After rinsing, 100 brightfield images (Plan-Neofluar 10×/0.3 objective; Carl Zeiss, Oberkachen, Germany) were collected every 10 s without flow then under a shear stress at 8 dyn cm⁻² (0.8 Pa). Fluorescence images for each pattern were collected at the end of the experiment using the same objective at the recorded position. The cell adhesion was quantified as the ratio between the number of adherent cells under 8 dyn cm^-2 and the number without flow after rinsing. This ratio is defined as the adhesion rate. For patterns with modulated adhesion, the adhesion contrast is defined as the normalized difference of adhesion rate between areas with maximal adhesion and modulated adhesion.

For cell migration assays, during each experiment, 20 brightfield images (Plan-Neofluar 10×/0.3 objective, Carl Zeiss, Oberkachen, Germany) were first collected every 10 s upon incubation to verify the state of the cells. Then, the flow chamber was rinsed with culture medium with a gentle flow of 1 dyn cm⁻² to remove non-adherent cells. After rinsing, 100 brightfield images (Plan-Neofluar 10×/0.3 objective, Carl Zeiss, Oberkachen, Germany) were collected every 10 s. Fluorescence images for each pattern were collected at the end of the experiment with the same objective and at the same position. Additionally, brightfield and RICM images (Neofluar 63×/1.25 antiflex, Carl Zeiss, Oberkachen, Germany) were collected every 5 s for each pattern to reveal the cell adhesion fingerprint, and fluorescence images were collected at the end of the experiment to localize the protein patterns.

PE-labeled anti-human CD54 (ICAM-1) and anti-human CD106 (VCAM-1) antibodies (eBioScience by Thermo Fisher Scientific, Waltham, MA) were used for adhesion molecule quantification. First, we set up a bulk calibration curve by measuring the fluorescence intensity of antibody solutions inside thin channels of 48 µm height at concentrations of 0, 1.5, 3, 5 and 7 µg/ml. Channels were pre-treated with 1% Pluronic F127 (Sigma-Aldrich, St Louis, MO) to limit adsorption of antibodies on the channel surface. Subsequently, channels were rinsed extensively with PBS. Residual fluorescence intensity due to adsorbed antibodies was measured and then subtracted from the previous measurements. A previous study (Hornung et al., 2020) showed a linear relation between the fluorescence intensity and the bulk concentration. We assume that the signal is given by the total number of molecules in the thin channel, then the volume concentration can be converted to a surface concentration for a channel 48 µm in height. Then, for each sample used for cell adhesion and migration assays, the patterned surfaces coated with ICAM-1 or VCAM-1 were first rinsed extensively with cold PBS solution. Then, the sample was stained with the corresponding antibody at 10 µg/ml and incubated overnight at 4°C. Images were taken the next day using the Zeiss Z1 microscope setup. The fluorescence intensity was analyzed using ImageJ software (National Institutes of Health, Bethesda, MD) at five different positions. The average intensity was converted into surface density of the adhesion molecules according the calibration data.

A home-made program developed using MATLAB software (The MathWorks, Natick, MA) was used to track migrating cells, as previously described by Valignat et al. (2013). Briefly, the program (1) performs image quality enhancement using background division and intensity normalization; (2) binarizes the images using a given threshold to distinguish cells from the background; (3) detects and numbers cells in the first frame and tracks them in the following frames; and (4) saves the coordinates and time points for each identified cell to calculate different migration parameters. Only cells migrating for at least 30 µm were considered. In this study, to quantify cell adhesion, the program counted the number of adherent cells at 0 dyn cm⁻² (0 Pa) and 8 dyn cm⁻² (0.8 Pa) to determine the adhesion rate of patterned substrates. To determine the directional bias of migrating cells on a patterned surface, the migration angle was defined as the one between the patterned stripes and the total cell trajectory between the first and final image. The program calculated the fraction of cells with a migration angle <45° and >45°. The anisotropy index was defined as the difference between these two fractions. To further characterize the preference of cells on patterns with modulated adhesion, the program used a binarized image of the fluorescence pattern as a mask to identify the position of each cell on the pattern. It then displayed color-coded cell trajectories according to their position on the patterns. It also calculated the number of cells on patterns with higher and lower adhesion. The Index of Preferential Adhesion was defined as the normalized difference between these two numbers.

For calcium flux imaging experiments, cells were first seeded in channels with RPMI medium and were incubated for 10 min at 37°C to allow adhesion, then they were rinsed with HBSS (1×HBSS; Gibco, ThermoFisher, Waltham, MA) containing 1% BSA then incubated with Oregon Green^® 488 BAPTA-1, AM (ThermoFisher, Waltham, MA) diluted in HBSS containing 1% BSA to a concentration of 5 µM for 15 min at 37°C in the dark. After rinsing with HBSS containing 1% BSA, the medium was replaced with HBSS containing 10% FCS. The control experiment was achieved by injection of ionomycin (ThermoFisher, Waltham, MA) at a concentration of 1 µg ml⁻¹. The calcium intensity fluctuation for each migrating cell was calculated for the whole duration of the experiment as follows: first, the minimal fluorescence intensity for the first 20 images was calculated; then, the fluorescence fluctuation was calculated as the normalized difference between the intensity of the migrating cell and the minimal intensity; finally, the average fluctuation for all migrating cells was generated for the whole duration of the experiment.

We are grateful to the Cell Culture Platform facility (Luminy TPR2-INSERM).

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.242883.reviewer-comments.pdf