WASp triggers mechanosensitive actin patches to facilitate immune cell migration in dense tissues.

When crawling through the body, leukocytes often traverse tissues that are densely packed with extracellular matrix and other cells, and this raises the question: How do leukocytes overcome compressive mechanical loads? Here, we show that the actin cortex of leukocytes is mechanoresponsive and that this responsiveness requires neither force sensing via the nucleus nor adhesive interactions with a substrate. Upon global compression of the cell body as well as local indentation of the plasma membrane, Wiskott-Aldrich syndrome protein (WASp) assembles into dot-like structures, providing activation platforms for Arp2/3 nucleated actin patches. These patches locally push against the external load, which can be obstructing collagen fibers or other cells, and thereby create space to facilitate forward locomotion. We show in vitro and in vivo that this WASp function is rate limiting for ameboid leukocyte migration in dense but not in loose environments and is required for trafficking through diverse tissues such as skin and lymph nodes.

Introduction

Cells embedded in tissues are tightly confined by complex three-dimensional (3D) microenvironments. These can be dense fibrillar networks as in mesenchymal interstitia such as the dermis or cell-packed environments such as organ parenchyma or lymphatic tissues (Weigelin et al., 2016). Due to the limited availability of open space in such environments, cells frequently experience compressive loads from their surroundings, which is particularly relevant for cells that actively migrate.

Single-cell migration is an almost ubiquitous phenomenon in eukaryote biology and follows a continuum of biophysical strategies, with the mesenchymal mode on one pole and the ameboid mode on the other (Friedl and Wolf, 2010). Ameboid cells, such as metastatic cancer cells or leukocytes, migrate much faster and are more autonomous from their extracellular environment than mesenchymal (or epithelial) cells and rather than remodeling their environment they adapt to it (Paluch et al., 2016). Survival and locomotion of ameboid cells do not depend on adhesive ligands, and they migrate efficiently even in ectopic environments or artificial materials (Lämmermann et al., 2008; Reversat et al., 2020). The quick shape changes that ameboid cells owe their name to (Amoeba, from Greek amoibe, meaning “change”) are entirely autonomous from the environment and the cellular envelope behaves as an active surface that rapidly deforms even when the cell floats in suspension (Barry and Bretscher, 2010).

These active adaptations of cell shape seem necessary to negotiate tissues, as ameboid cells do not digest their environment in order to create a path. Instead, in obstructive environments with heterogeneous geometry they effectively seek the path of least resistance choosing larger pores over small ones (Renkawitz et al., 2019). Whenever there is no other choice and the cell faces a small pore, it actively deforms its cell body and/or transiently dilates the pore in order to pass through (Pflicke and Sixt, 2009; Thiam et al., 2016; van den Berg et al., 2019). However, how ameboid cells generate forces underlying these processes is incompletely understood.

In animal cells, most intracellular forces are generated by the actin cytoskeleton, which produces pulling forces via actomyosin contraction and pushing forces via actin polymerization. Both pulling and pushing forces are tightly regulated by mechanical feedback. In order to test the geometrical and mechanical features of its surrounding, mesenchymal cells probe their substrate by pulling on it (Plotnikov and Waterman, 2013). To do so, they use their integrin-mediated adhesion sites as mechanosensitive organelles, which emit signals in response to traction force. Substrate probing by pulling forces seems less relevant for leukocyte migration. Whenever they move in an adhesion-free mode there is clearly no opportunity to pull. Also, whenever leukocytes do use adhesion receptors to generate friction with the substrate, they minimize traction: they tune their integrins so that they just transmit the extremely small traction forces required to move the cell body forward (Hons et al., 2018).

Pushing forces are generated whenever actin filaments polymerize against the plasma membrane (Mogilner, 2006). The prototypic pushing structure is the lamellipodium, the flat actin protrusion that forms the leading front of most ameboid and mesenchymal cells (Krause and Gautreau, 2014). Lamellipodial actin can adapt its protrusive force to the counterforce it experiences, making the tip of a cell mechanosensitive (Heinemann et al., 2011). However, lamellipodia are strictly two-dimensional (2D) leaflets, and the molecular machinery that drives actin polymerization at the lamellipodial tip does not support growth in the third dimension (Fritz-Laylin et al., 2017b, 2018). Hence, lamellipodia can protrude into open spaces, but they are insufficient to create the active surface of an ameboid cell that adapts and deforms in 3D. In order to do that, the cell requires both mechanical sensitivity and activity across its whole cortex and not only at its lamellipodial tips. Here, we use dendritic cells (DCs) and T cells as model systems to address how ameboid cells respond to geometrical constraints of their environment. We investigate how these constraints are met by the molecular machinery that drives actin polymerization across the cellular cortex.

Results

Dendritic cells form actin patches in response to restrictive environments

To study how ameboid cells respond to compressive loads, we observed the migratory behavior of dendritic cells (DCs) under a patch of non-degradable agarose cast on a serum-coated coverslip (Hons et al., 2018) (Figure 1A). In this reductionist approach, no pre-formed space is available and the highly flattened cells need to actively lift and deform the agarose to create sufficient space between coverslip and agarose to move forward (Figure 1A) (Laevsky and Knecht, 2003). To control the mechanical load that the cells have to counter, we adjusted the stiffness of the agarose around the range of cellular stiffness as measured by atomic force microscopy (AFM) (Blumenthal et al., 2020; Guimarães et al., 2020) (Figures 1A and S1A). With increasing stiffness, the mean migratory speed declined (Figures 1B, 1C, and S1B) and the cells shifted from continuous locomotion to a stop and go pattern where cells collapsed and stalled after short periods of movement (Figure 1D; Video S1). Notably, time spent in arrest rose only slightly from low (2.5 kPa; below cellular stiffness) to intermediate (10 kPa; slightly exceeding cellular stiffness) but drastically at high agarose stiffness (17.5 kPa) (Figure 1D). Morphometric analysis of average cell shape revealed that at low and intermediate load, DCs adopted an elongated shape with a wide leading edge and a slender cell body (Figure 1E; Video S1). Only under high load, average cell shape became substantially widened (Figures 1E, 1F, and S1C). These data indicate that DCs can successfully counter restrictive environments of stiffnesses that are in the range of cellular stiffness but frequently collapse when external load becomes exceedingly high.

Figure 1

Dendritic cells form actin patches in response to restrictive environments

(A) Confinement of dendritic cells (DCs) under agarose. Agarose stiffness is adjusted around cellular stiffness (also see Figure S1A). Top view: SEM; scale bar, 50 μm. Side view: SEM image of a sagittal section; scale bar, 5 μm.

(B) Plot of single-cell tracks from representative experiments (Δt = 40 min); scale bar, 200 μm.

(C) Mean track speed. Each data point represents one track. Low agarose stiffness: n = 407 pooled from 6 experiments; intermediate agarose stiffness: n = 182 pooled from 4 experiments; high agarose stiffness: n = 205 pooled from 4 experiments; Kruskal-Wallis/Dunn’s multiple comparisons test; ∗∗∗∗p < 0.0001; mean ± SD. Also see Figure S1B.

(D) Fraction of time in arrest was extracted from tracks in (C); mean ± SEM; Kruskal-Wallis/Dunn’s multiple comparisons test; ∗∗∗∗p < 0.0001.

(E and F) Mean shape of cell contours (E). Contours were further analyzed for mean cell width along the centerline (F). Kruskal-Wallis/Dunn’s multiple comparisons test; ∗p = 0.0189; ∗∗p = 0.0016. ns, not significant. Mean ± SD.

(G) Actin patches (LifeAct-eGFP) form in response to confinement. Representative images from spinning disk confocal microscopy movies of live cells on PMOXA-coated (non-adhesive) coverslips (confined cells). Unconfined cells adhere to PLL-coating. Kymograph shows retrograde actin flow. Notably, actin patches move with the bulk actin flow.

(H) Dimensions of actin patches were quantified from confocal z stacks (also see Figure S1J); n = 246 pooled from 7 cells (mean ± SD).

(I) Correlative light (epifluorescence) and (scanning) electron microscopy (CLEM). Note, LifeAct-eGFP signal correlates with knob- and ridge-like protrusions. Right panels: SEM of unconfined cells (spreading and in suspension). Scale bars, 10 μm.

(J) Ultrathin section analyzed by SEM. Upper panel: section across the leading edge. Lower panel: section across the cell body with nucleus. Notably, knobs (black arrow heads) form orthogonally to lamellipodia and indent the agarose. Mirrored indentations are visible in the nuclear lamina, indicating pushing forces generated by knobs (see magnified SEM images and scheme of a single actin knob). Scale bar, 1 μm. Right: TEM of ultrathin section of mDC in suspension. Unconfined cells form multiple lamellipodial veils (arrow heads) but lack knob-like protrusions of confined cells.

(K) Nucleus-free cytoplasts migrating under agarose form actin patches (spinning disk microscopy) correlating with knob-like protrusions (SEM).

(L) Of note, cytoplasts resemble the morphology of intact cells both under confinement and during unconfined spreading or in suspension; scale bars, 5 μm (L). Also see Figure S1 and Video S1.

Video S1. Dendritic cells form actin patches in response to restrictive environments, related to Figures 1 and S1

(A) Epifluoresence movies (LifeAct-eGFP and Hoechst) of mature dendritic cells (DCs) migrating under agarose of low, intermediate and high stiffness. Lower panels: overview; upper panels: higher magnification.

(B) Dendritic cells form actin spikes in response to restrictive microenvironments (related to Figures 1 and S1). Spinning disk confocal microscopy of actin dynamics of migrating dendritic cells without confinement (2D), with partial confinement (3 μm, 3D), with complete confinement (intermediate-stiff agarose, 3D).

(C) Cytoplasts form actin patches similar to intact cells.

When migrating under intermediate-stiff agarose LifeAct-eGFP expressing DCs displayed small actin-rich patches embedded in a homogeneous actin cortex. These patches were scattered across the whole cell, with peak intensities in the cell body (Figure 1G; Video S1). Notably, actin patches were restricted to regions of cellular confinement (Figure 1G; Video S1), formed independent of myosin II activity (Lomakin et al., 2020; Venturini et al., 2020) (Figure S1D) and were not sites of clathrin- or caveolin-mediated endocytosis (Ferguson et al., 2017) (Figures S1E–S1G). Actin patches were independent of adhesive interactions with the substrate as they were equally pronounced when cells migrated between (inert) agarose and passivated coverslips (Figure 1G; Video S1). DCs migrating under non-adhesive conditions display a substantial retrograde flow of actin as both cortical and lamellipodial actin slide backward in relation to the substrate (Renkawitz et al., 2009). Actin patches moved together with this bulk actin flow, indicating linkage of the patches to the rest of the actin cortex (Figures 1G and S1H). Over time, patches occasionally evolved into elongated stripes (Figures 1H and S1I; Video S1). Confocal z scans as well as correlated light and scanning electron microscopy (CLEM) indicated that patches and stripes corresponded to short knob- or ridge-like surface structures that protrude normal to the imaging plane into the agarose (Figures 1H, 1I, and S1J). Ultrastructural analysis of serial sections suggested that these knobs generated vertical pushing forces: when actin knobs were located on top of the nucleus, the protrusions into the agarose overlay were mirrored by an indentation in the nuclear lamina (Revach et al., 2015) (Figure 1J). Supporting this notion, visualization of the cell-substrate interface by interference reflection microscopy (IRM) revealed that actin patches were, despite the lack of cell-substrate adhesions, in closer proximity to the underlying substrate than the surrounding ventral membrane (Yu et al., 2013) (Figure S1K). Collectively these experiments suggest that cells confined under a restrictive overlay respond by forming cortical actin patches that push against the compressive load.

We next tested if the nucleus is required for the formation of actin patches and generated nucleus-free cytoplasts by gradient ultracentrifugation (see STAR Methods) (Figures 1K, 1L, and S1L). Cytoplasts were viable for hours and resembled intact cells in morphology and function (Figures 1K, 1L, and S1L). When migrating under a patch of intermediate-stiff agarose cytoplasts formed actin patches that were scattered over the entire projected cell area and showed an actin flow pattern similar to intact cells (Figure 1K; Video S1). Importantly, cytoplasts accumulated prominent actin patches in the area corresponding to the peri-nuclear region of intact cells. Similar to intact cells CLEM revealed that actin patches correspond to short knobs on the surface of cytoplasts that grow normal to the imaging plane (Figures 1L and S1L). Together these data confirm that actin patches triggered by confinement can evolve independent of the presence of a nucleus.

WASp-driven actin patches grow orthogonal to WAVE-driven lamellipodia

Formation and dynamics of cortical actin heavily depends on the Arp2/3 nucleation complex (Bovellan et al., 2014; Laplaud et al., 2021). Upon pharmacological inhibition of Arp2/3 with CK666 actin patches in DCs under intermediate-stiff agarose were abolished, and treated cells spent most of their time in arrest (Figure S2A). We therefore compared the contribution of the two main nucleation-promoting factors (NPFs) upstream of Arp2/3: Wiskott-Aldrich syndrome protein (WASp) and WASp-family verprolin-homologous protein (WAVE) (Rotty et al., 2013). To this end, we analyzed DCs deficient in WASp or the WAVE complex subunit Hem1, which in hematopoietic cells is essential for the stability of the pentameric WAVE complex (Leithner et al., 2016). Both mutants showed reduced migration speeds and increased time in arrest under agarose of intermediate stiffness (Figures 2A and S2B; Video S2), while chemotaxis toward CCL19 was not affected (Leithner et al., 2016; Weiner et al., 2007) (Figure S2C). Hem1−/− DCs were devoid of lamellipodia and formed multiple, spiky protrusions at the leading edge (Figure S2D), while WASp−/− DCs formed normal lamellipodia (Figure S2D). Morphometric analysis showed that despite their spiky protrusions the mean overall cell body shape of Hem1−/− cells was comparable to WT cells, while WASp−/− cells were significantly wider and shorter (Figure 2B). Hence, regarding the shape and migratory pattern, WASp−/− cells under intermediate-load conditions phenocopied WT cells under high-load conditions (Figures 1B–1F).

Figure 2

WASp-driven actin patches polymerize orthogonal to WAVE-driven lamellipodia

(A) WASp−/− and Hem1−/− DCs were derived from bone marrow of transgenic mice. Left panel: WT: n = 118 pooled from 3 experiments; WASp−/−: n = 175 pooled from 4 experiments; Hem1−/−: n = 347 pooled from 6 experiments; mean ± SD; Kruskal-Wallis/Dunn’s multiple comparisons test; ∗∗∗∗p < 0.0001. Right panel: Fraction of time in arrest was extracted from tracks; mean + -SEM; Mann-Whitney test and Kruskal-Wallis/Dunn’s multiple comparisons test; ∗∗∗∗p < 0.0001.

(B) Mean shape of contours; WT: n = 75; Hem1−/−: n = 100; WASp−/−: n = 84. Mean cell width along the centerline; mean ± SD; Kruskal-Wallis/Dunn’s multiple comparisons test; (g) ∗p = 0.0491; ns, not significant.

(C and D) TIRF microscopy of agarose-confined DCs expressing (C) Abi-eGFP (to label the WAVE complex) or (D) WASp-eGFP. Note, while Abi1-eGFP is restricted to the lamellipodial tip, WASp-eGFP forms dot-like structures scattered across the cell body and moving with the retrograde flow; resembling LifeAct-eGFP signal (see Figure 1G). Scale bar, 10 μm.

(E) Dual labeling of LifeAct-mCherry and WASp-eGFP (TIRF microscopy); scale bar, 10 μm; also see Figure S2F.

(F) WT, WASp−/− and Hem1−/− DCs under intermediate-stiff agarose on PMOXA-coated (non-adhesive) coverslips. Representative images from spinning disk confocal microscopy movies (LifeAct-eGFP) of live cells are shown (also see Video S2). Notably, Hem1−/− DCs are devoid of lamellipodia but form prominent actin patches at the leading edge and the cell body. Scale bar, 10 μm.

(G and H) Quantification of area (G) and dynamics (H) of actin patches in WT and WASp−/− cells. Jaccard index measures frame-to-frame overlap of segmented actin patches (1 = 100% similarity).

(G) Area of actin patches. WT: n = 25 cells; WASp−/− : n = 23 cells. Mean ± SD; Mann-Whitney; ∗∗∗∗p < 0.0001.

(H) WT: n = 2,144 frames pooled from 25 movies; WASp−/−: 2,055 frames pooled from 23 movies. Mean ± SD; Mann-Whitney; ∗∗∗∗p < 0.0001.

(I) Ultrastructural analysis of WASp−/− DCs using CLEM (upper panel; scale bars, 10 μm) and SEM of an ultrathin section (lower panel; scale bar, 1 μm). While unconfined WASp−/− cells morphologically resemble WT cells (Figure 1I), they do not form knob- and ridge-like structures upon confinement.

(J) Graphical summary. Also see Figure S2 and Video S2.

Video S2. WASp-driven actin patches polymerize orthogonal to WAVE-driven lamellipodia, related to Figures 2 and S2

(1) WT, Hem1−/− and WASp−/− dendritic cells migrating under agarose (intermediate stiffness); epifluorescence microscopy. Lower panels: overview; upper panels: higher magnification.

(2) TIRF microscopy of agarose confined DCs expressing Abi1-GFP (to visualize the WAVE complex) and WASp-eGFP (to visualize WASp). While Abi1-eGFP was strictly localized to the tip of lamellipodia, WASp-eGFP formed punctae scattered across the cell surface.

(3) TIRF microscopy of agarose confined DCs co-expressing LifeAct-mCherry and WASp-eGFP. Cells migrate on inert, PMOXA-coated substrate. LifeAct and WASp co-localize in puncta at the leading edge, cell body and the peri-nuclear region of the cell. WASp dots precede actin patch formation (arrow head).

(4) WASp drives actin spike formation scattered across the cell body while WAVE drives lamellipodia formation. Cells are confined under the load of intermediate-stiff agarose, on inert substrate (PMOXA) (Spinning disk microscopy).

(5) Epifluorescence movie of WT and WASp−/− DCs migrating under agarose. Actin patches were segmented in Ilastik. Cell contours were stabilized and motion of actin patches is shown in the cell-frame of reference.

Next, we were interested in the spatiotemporal organization of WAVE and WASp-dependent actin networks under compressive loads. While the WAVE complex (detected with Abi1-eGFP) strictly localized to the tip of lamellipodia (Leithner et al., 2016), WASp (detected with WASp-eGFP) formed dots scattered across the cell surface, matching the distribution and flow pattern of actin patches (Figures 2C–2E, and S2E; Video S2). Simultaneous monitoring of WASp-eGFP and LifeAct-mCherry revealed co-localization of both signals, with WASp dots preceding polymerization of actin patches by about 5–10 s (Figure 2E; Video S2). Accordingly, deletion of WASp, but not the WAVE complex abrogated formation of actin patches as revealed by LifeAct-eGFP signal (Figures 2F and 2G; Video S2), and the few actin patches remaining in WASp−/− DCs were transient and short-lived (Figures 2H and S2F; Video S2). Ultrastructural analysis of serial sections and CLEM confirmed the morphological absence of short knobs on the dorsal surface of WASp−/− DCs under confinement (Figure 2I), while there were no apparent morphological differences to WT cells in the absence of confinement (Figures 1I and 2I).

Together, these data reveal a characteristic organization of branched actin networks in 3D: while WAVE-driven actin nucleation powers the horizontal forward protrusive component of cellular locomotion, WASp-driven actin nucleation counters the mechanical load of the overlay (Figure 2J).

WASp-driven actin patches are triggered by mechanical load

We next addressed if WASp-driven actin patches form in response to a mechanical trigger or if they develop spontaneously. We seeded LifeAct-eGFP-expressing DCs on poly-L-Lysine coated coverslips and locally pinched the cell body with the blunted tip of a microneedle (Figure 3A). Within seconds after indenting the cell, the actin reporter increased around the pipette tip of WT DCs, leaving behind a localized actin cloud (Gérard et al., 2014) (Figure 3B; Video S3). This mechanically induced burst of actin polymerization was substantially decreased in WASp−/− DCs (Figure 3C). This finding supported a role of WASp in polymerizing actin in response to mechanical loading (Figure 1G).

Figure 3

WASp-driven actin patches are triggered by mechanical loading

(A) DCs attached to poly-l-lysine (PLL)-coated coverslips were pinched from top with a microneedle (schematic). Micropipette was held in place during time-lapse recordings.

(B) Time traces (spinning disk confocal microscopy) of representative LifeAct-eGFP expressing DCs. The last frame shows an overlay of brightfield images of the micropipette (highlighted in orange). Right panel: time traces of LifeAct-eGFP signal at the site of micropipette indentation (normalized by mean LifeAct-eGFP signal of non-indented area). Intensity of first frame was set to 1. WT: n = 24; WASp: n = 20; mean ± SEM. Scale bar, 10 μm.

(C) Dot plot showing fold increase of LifeAct-eGFP intensity following indentation (last frame/first frame). Each data point is one indentation experiment (see B). Mean ± SD; Mann-Whitney test; ∗∗∗∗p < 0.0001.

(D) Schematic shows DCs confined on inert nano-ridges with 2 μm pitch (PMOXA coating) (see STAR Methods).

(E) Representative WT and WASp−/− DC migrating on nano-ridges. Notably, actin patches formed with high precision on top of ridges (also see Figure S3C). Arrow head indicates actin patches. Kymograph shows retrograde flow of actin patches. Fraction of ridges covered by actin patches (segmented in Ilastik) (also see Figure S3B). WT: n = 19 cells; WASp−/−: n = 19 cells; Mean ± SD; Mann-Whitney test; ∗∗∗∗p < 0.0001. Scale bar, 20 μm.

(F) Representative micrograph of WASp-eGFP-expressing DCs migrating on nano-ridges (width: 1 μm; height: 250 nm; pitch: 2 μm) (TIRF). WASp-eGFP shows same dynamic pattern as actin patches (kymograph) with sharply delineated dots matching the ridge structure. Scale bar, 20 μm.

(G) Dual labeling of LifeAct-mCherry and WASp-eGFP reveals co-localization on nano-ridges. Kymograph shows that WASp dots precede the formation of actin patches. Scale bar, 10 μm.

(H) Dual labeling of LifeAct-eGFP (false colored in magenta) and Cip4-mCherry (false colored in green) reveals co-localization on nano-ridges. Kymograph shows that Cip4 clusters precede the formation of actin patches. Scale bar, 10 μm.

(I) Cip4- and WASp dots precede actin patches. Time traces of fluorescence intensities of kymographs shown in (G) and (H).

(J) Schematic summary showing the spatiotemporal sequence of actin patches formation on nano-ridges.

(K) Schematic of an AFM cantilever indenting a DC adherent to a PLL-coated coverslip. Left: using a Hertz contact mechanics model, the elastic modulus was estimated by fitting the force indentation curves up to 1 μm. Each data point represents one measurement. WT: n = 160 from 37 cells (also see Figure S1A); WASp−/−: n = 106 from 26 cells. For cellwise analysis see Figure S4F. Violin plot with median ± quartiles; Mann-Whitney test; ∗∗∗∗p < 0.0001. Right: time series cellular response to repetitive indentations of the AFM cantilever (elastic modulus). WT = 35 cells; WASp−/− = 23 cells. Mean ± SEM; multiple t tests; ∗∗∗p < 0.01. Also see Figure S3 and Video S3.

Video S3. WASp-driven actin patches are triggered by mechanical loading, related to Figures 3 and S3

(1) LifeAct-eGFP expressing WT and WASp−/− DCs were indented by the blunted end of a micropipette (Spinning disk microscopy). Last frame shows overlay with the outline of the micropipette. Lower panel shows color-coded LifeAct-eGFP signal.

(2) LifeAct-eGFP expressing WT and WASp−/− DCs migrating on inert surfaces (PMOXA) with nano-ridges (Spinning disk microscopy).

(3) WASp-eGFP expressing WT DCs migrating on inert surfaces (PMOXA) with nano-ridges (TIRF microscopy).

(4) Co-expression of WASp-eGFP and LifeAct-mCherry reveals co-localization of WASp and F-actin in clusters accumulating on top of ridges. WASp dots precede the formation of actin patches on ridges.

(5 and 6) Curvature-sensing protein Cip4 accumulates in dot-like structures preceding actin patches. Membrane deformation was induced by the topography of the agarose overlay (V, Spinning disk) and nano-ridges on coverslips (VI, TIRF).

Migrating leukocytes constantly encounter submicron-sized obstacles such as collagen fibers that restrict their passage and therefore locally indent the cell body. We mimicked this scenario in vitro and confined cells onto inert substrates with a ridged topography matching the size of collagen fibrils (Figures 3D, S3A, and S3B). This setup allowed the precise localization of the fluorescent signals. We found that ridges of 600-nm height (and 250-nm width) induced actin patches that matched the pattern of ridges (Figure 3E; Video S3). Notably, these patches exclusively formed on top of ridges and were virtually absent in the interjacent grooves (Figures 3E and S3C). This response was independent of adhesive interactions as actin patches formed efficiently on passivated substrates, where they moved with the retrograde actin flow (Figure 3E). WASp-eGFP followed the same dynamic pattern, with sharply delineated dots precisely matching the ridge structure and preceding actin patch formation by 5–10 s (Figures 3F, 3G, and 3I; Video S3). In contrast, WAVE (detected with Abi1-eGFP) was not recruited to ridges but remained restricted to the very tip of the leading edge and occasionally to small lamellipodial protrusions forming within interjacent grooves (Figure S3D). While actin patches remained targeted to ridges after deletion of WAVE (Figure S3E), they were virtually absent after deletion of WASp (Figures 3E and S3C; Video S3). We rarely observed sparse formation of actin patches at the cell body of WASp-deficient DCs (Figure S3C), but these remnants were negative for Abi1-eGFP, arguing against a compensatory function (Veltman et al., 2012; Zhu et al., 2016) (Figure S3F). In summary, we demonstrate two non-interchangeable functions of WAVE and WASp in DCs: while WAVE drives explorative lamellipodial protrusions (Leithner et al., 2016), WASp is recruited to sites of cellular indentation where it remains on target to locally trigger actin polymerization.

Membrane curvature is a key upstream regulator of NPFs (Suetsugu and Gautreau, 2012). To analyze if curvature sensing acts upstream of WASp activation we transfected DCs with fluorescently tagged Cip4, a F-BAR domain protein expressed in leukocytes, including DCs (Koduru et al., 2010). Cip4 enriched at the leading edge of confined DCs and accumulated in dot-like structures preceding the formation of actin patches (Figure S3G). On topographic slides Cip4 formed clusters aligning with nano-ridges and co-localizing with actin patches (Figure 3H). Similar to WASp, Cip4 preceded actin polymerization by approximately 10 s, but quickly dissociated from actin patches once actin polymerization reached its peak (Figures 3I and 3J).

Previous studies have shown that mechanical load can induce global stiffening of the cellular cortex, which is dependent on actin reorganization (Hu et al., 2019). To test if load-induced actin polymerization increases cortical stiffness in DCs we repetitively indented DCs immobilized on poly-L-Lysine with the tip of an AFM cantilever and measured cellular stiffness. WASp-deficient DCs showed a significantly reduced cortical stiffness, confirming previous reports (Blumenthal et al., 2020). While cortical stiffness gradually increased following repetitive indentation, this response was significantly blunted in WASp−/− cells, indicating a crucial role of WASp in adapting cell mechanics to mechanical load (Figures 3K and S3H). In summary, we demonstrate that cells respond to mechanical load by locally polymerizing actin, a process requiring WASp.

WASp-driven actin patches are negligible for the generation of cell-substrate friction under agarose

WASp deficiency was shown to alter integrin-dependent leukocyte functions (Zhang et al., 2006). Dependent on the availability of adhesive ligands, leukocytes can flexibly shift between non-adhesive locomotion modes and modes where actin flow couples to the substrate via integrins (Hons et al., 2018). To test whether WASp and integrins lie in the same pathway we placed WASp−/− cells under intermediate-stiff agarose confinement on inert substrates (PMOXA coating). Similar to migration on adhesive (serum-coated) substrates (Figures 2A–2C), WASp−/− DCs were slower than WT cells and arrested frequently (Figure 4A), while adopting a rounded morphology (Figures 4B and S4A). This additive phenotype demonstrates that the contribution of WASp is neither upstream nor downstream of integrins but represents an independent mechanism. We saw that WASp-dependent actin knobs at the dorsal surface of the cell occasionally reached into holes in the overlaying agarose (Figures 1I and S4B), raising the possibility that WASp function can compensate for lack of integrin-mediated adhesion. We therefore addressed if, in analogy to the molecular clutch model of integrins, actin patches can act as footholds that couple into the retrograde actin flow to generate traction forces (Figure S4C). The most sensitive readout for force-coupling in leukocytes is the rate of actin polymerization. Whenever cells slip due to a disengaged adhesive clutch, actin polymerization massively increases to compensate for the retrograde slippage (Hons et al., 2018; Maiuri et al., 2015; Renkawitz et al., 2009; Reversat et al., 2020). When we measured actin flow in WT versus WASp-deficient DCs on slippery, PMOXA-coated substrates, there was no sign of accelerated actin polymerization. In contrast, polymerization rates showed a slight drop (Figure S4C). These data indicate that the main function of WASp-dependent actin patches is not to generate traction tangential to the membrane but rather to facilitate protrusion by pushing orthogonal to the membrane.

Figure 4

Vertical pushing facilitates locomotion by deforming restrictive environments

(A) Under agarose assay on inert PMOXA-coated substrates. WT: n = 642 pooled from 3 experiments; WASp−/−: n = 597 pooled from 4 experiments; mean ± SD; Mann-Whitney test; ∗∗∗∗p < 0.0001. Fraction of time in arrest was extracted from tracks; mean + SEM; Mann-Whitney test; ∗∗∗∗p < 0.0001.

(B) Mean shape of contours; WT: n = 107; WASp−/−: n = 195; mean cell width along the centerline and aspect ratio; mean ± SD; Mann-Whitney test; ∗∗∗∗p < 0.0001.

(C) Representative color-coded WT DC outlines (time). Scale bar, 25 μm.

(D) Smoothed time traces of speed and aspect ratio extracted from WT DC contours (epifluorescence movies (LifeAct-eGFP); frame rate = 30 s).

(E) Cross-correlation between the (1) cell speed and aspect ratio, (2) cell speed and mean fluorescence of actin patches, and (3) mean fluorescence of actin patches and area of nucleus is shown (WT: n = 25 cells; mean ± SEM). The positive lag time (gray curve) means an increase in mean fluorescence of actin patches precedes cell speed (also see Figures S4D–S4F).

(F) Smoothed time traces of area of nucleus (normalized by the mean of all time points) and mean fluorescence of actin patches (normalized by the mean of all time points) of the same cell (see A and B).

(G) Representative color-coded WASp−/− DC outlines (time). Scale bar, 25 μm.

(H) Smoothed time traces of speed and aspect ratio extracted from WASp−/− DC contours (epifluorescence movies [LifeAct-eGFP]; frame rate = 30 s).

(I) Smoothed time traces of area of nucleus (normalized by the mean of all time points) and mean fluorescence of actin patches (normalized by the mean of all time points) of the same cell.

(J) Cross-correlation analysis (also see E) (WASp−/−: n = 22–23 cells; mean ± SEM).

(K) PDMS-based (non-deformable) microfluidic device with a constant height and constrictions.

(L) Representative images of WASp-eGFP- and LifeAct-eGFP-expressing cells squeezing through constrictions (TIRF). Arrow heads indicate increased fluorescence at constrictions. Scale bar, 10 μm. (M) Mean speed of cells migrating in straight channels (mean ± SD; Mann Whitney test; p = 0.82) and passage time of constriction (mean ± SD; Mann Whitney test; ∗∗p < 0.01). Schematic shows dimension of microfluidic device in μm. Also see Figure S4 and Video S4.

Vertical pushing facilitates locomotion by deforming restrictive environments

The decrease in cell speed under mechanical load was accompanied by a change in cell shape (Figures 1E, 1F, and S1C). While WT cells were still able to maintain an elongated shape with a slender cell body in intermediate-stiff agarose, WASp−/− DCs were significantly shorter and had a widened cell body (Figures 2C, 4B, and S2D). To address if and how vertical pushing is embedded into the morphodynamic cycle of cellular locomotion, we next used a cross-correlation approach of cellular morphology. DCs migrating under agarose undergo repetitive phases of acceleration and deceleration with a mean phase duration of 4–5 min (Figures S4D and S4E). Acceleration of WT cells under agarose is initiated by leading-edge protrusion and elongation of the cell body (Figure 4C and 4D; Video S4). Accordingly, cross-correlating speed with aspect ratio revealed a clear positive and instantaneous correlation (Figures 4E and S4F). Incorporating the number of cortical actin patches into the correlation showed that patches preceded acceleration by 1–2 min (Figures 4E and 4F), indicating that patches “facilitate” rather than drive forward protrusion. We next measured nuclear-projected area as a proxy for how much a cell is able to lift the agarose: the nucleus of suspended DCs is approximately spherical, but naturally flattens with increasing confinement, making it a good indicator of cell height (Lomakin et al., 2020) (Figures 1A, 1J, and S4G). WT DCs showed steady fluctuations of nuclear-projected area when migrating under agarose (Figure S4H; Video S4). Nuclear area was negatively cross-correlated with the appearance of actin patches on the surface area of the cell (Figures 4E and 4F), supporting the concept that vertical WASp-driven protrusions lift the agarose so that the nucleus can expand vertically. Accordingly, cross-correlations between (residual) actin structures (Figure S4E) speed and projected nuclear area were blunted in WASp−/− DCs, while the speed to aspect ratio cross-correlation was retained (Figures 4G–4J; Video S4). This morphodynamic sequence suggested that actin patches are a “prerequisite” to lift and deform the agarose, which then “facilitates” forward protrusion.

Video S4. Vertical pushing facilitates locomotion by deforming restrictive environments, related to Figures 4 and S4

(1) WT and WASp−/− dendritic cells migrating under the load of intermediate-stiff agarose; (epifluorescence Microscopy). Right panels show color-coded LifeAct-eGFP signal.

(2) Nuclear shape fluctuations of dendritic cells migrating under agarose. Dendritic cells show steady fluctuations of nuclear-projected area when migrating under agarose (epifluorescence microscopy).

(3) Dendritic cells migrating in non-deformable microfabricated PDMS devices. Migration of WASp−/− mutants in pillar forests is indistinguishable from WT cells (brightfield microscopy).

(4) WASp (WASp-eGFP) and F-actin (LifeAct-eGFP) are recruited to narrow constrictions (1–3 μm) (TIRF and brightfield microscopy).

We next tested if the WASp contribution to locomotion is indeed specific for migration in deformable restrictive environments. We thus seeded DCs into PDMS-based (non-deformable) microfluidic devices large pore sizes (100-μm2 cross-section). These can be passaged without substrate deformation. Here, migration parameters of WASp−/− mutants were indistinguishable from WT cells (Figure S4I; Video S4). Next, we reduced the pore sizes to cross-sections as small as 5–15 μm2 (Figure 4K). DCs squeezing through these tight constrictions recruited WASp and actin to sites of compression (Figure 4L; Video S4), matching the observations under agarose. To functionally test if WASp-dependent pushing forces are required for squeezing through non-deformable constrictions (Thiam et al., 2016), we measured migration in channels with tight constrictions (6-μm2 cross-section). Mean time of passage was not prolonged for WASp−/− compared with control cells but rather slightly reduced (Figure 4M). Together, we show that WASp-driven vertical pushing forces counter mechanical load and displace deformable microenvironments to generate space for the cell to enter into the protrusive phase of locomotion. Whenever the substrate is non-restrictive, WASp function is not required for locomotion. Whenever the substrate is restrictive but non-deformable, WASp is recruited but is not effective in facilitating locomotion.

Deformation of collagen fibers is required for migration in fibrous environments

We next challenged our findings in physiological environments. To initiate adaptive immunity, DCs need to migrate from the site of antigen retrieval to the draining lymph node (LN) (Worbs et al., 2017), which requires fast transit through diverse microenvironments. DCs can reach the LN without proteolytically generating a path (Pflicke and Sixt, 2009), suggesting that they predominantly use a mechanical strategy. We therefore adoptively transferred differentially labeled WT and WASp−/− DCs into footpads of WT recipients and measured homing to draining LNs (Figure S5A). After 24 h, recruitment of WASp−/− DCs was significantly reduced compared with WT controls, confirming previous reports by others and demonstrating physiological relevance (Bouma et al., 2007; de Noronha et al., 2005; Pulecio et al., 2008; Snapper et al., 2005).

We next employed fibrillar collagen, the most abundant matrix scaffold of mammalian tissues, as a proxy for the interstitial matrix. Collagen gels were polymerized at densities resulting in pore sizes substantially smaller than the minimal cross-section of deformed DCs (Lang et al., 2015; Renkawitz et al., 2019; Thiam et al., 2016; Wolf et al., 2013). Chemotactic locomotion under these conditions should depend on the deformation of obstructive fibers (Figure 5A). Indeed, fast confocal imaging revealed that DCs locally displaced collagen fibers, thereby generating space for the passage of the cell body. Locally, the deformation of collagen fibers was accompanied by a burst in actin polymerization (Figures 5B and 5C; Video S5). We then performed bulk measurements of large numbers of DCs migrating along CCL19 gradients (Figure 5D). In low-density collagen gels, where pore sizes between fibers are sufficiently large to allow unconstrained locomotion, WASp−/− cells were not significantly slower than WT controls (Figure 5E). With increasing collagen gel densities (decreased pore size and increased stiffness), DCs were more reliant on displacing fibers to generate space (Figure S5B), became substantially slower (Figure 5E), and often arrested (Figure 5F), indicating that matrix deformation became a rate-limiting parameter. Accordingly, under these conditions, WASp-deficient DCs were less efficient in recruiting actin to restrictive fibers (Figures 5C and S5C). Of note, similar to migration under agarose, directionality toward CCL19 was not affected by WASp deletion, indicating a negligible role of WASp-dependent actin patches in chemotaxis (Figures 5D and 5G). Together, our data show that cells require WASp-dependent pushing to mechanically create space in dense 3D matrices and provide a mechanistic explanation for impaired migration of WASp−/− DCs in vivo (Figure S5A).

Figure 5

Deformation of collagen fibers is required for migration in fibrous environments

(A) Schematic of experimental setup.

(B) Time-lapse spinning disk confocal microscopy reveals local deformation of obstructive collagen fibers (Alexa-594 labeled; 3.0 mg/mL) and increase of pore sizes for passage of the cell body (lower panel). Arrow heads indicate pore. Kymograph (left) shows dilation of the pore. Notably, fiber deformation was accompanied by local actin polymerization (middle panel). Upper panel shows merged images of LifeAct-eGFP and Collagen-Alexa-594. Scale bar, 10 μm.

(C) Left: time traces of LifeAct-eGFP signal at sites of obstructive fibers (normalized to intensity of the first frame; fold change of intensity is shown). Right: each data point shows the local maximum of LifeAct-eGFP signal from one trace (fold change of intensity; first frame as reference). WT: n = 12 cells; WASp−/−: n = 10 cells; mean ± SD; Mann-Whitney test; ∗p = 0.0358.

(D–G) Bulk analysis of DCs migrating in fibrous 3D collagen matrices with increasing densities (gray = WT; turquoise = WASp−/−). (D) Trajectory plot and directionality plot (inset) of DCs migrating in 1.5 mg/mL collagen gels.

(E) Shows mean speed of tracks pooled from 3 experiments. WT (0.75: n = 694; 1.5: n = 1,181; 3.0: n = 1,551; 5.0: n = 981); WASp−/− (0.75: n = 599; 1.5: n = 1,130; 3.0: n = 1,086; 5.0: n = 853). Mean ± SD; Kruskal-Wallis-test/Dunn’s multiple comparisons test; ∗p = 0.0221; ∗∗∗∗p < 0.0001.

(F) Fraction of time in arrest was extracted from tracks in (E). Mean ± SD; Kruskal-Wallis-test/Dunn’s multiple comparisons test; mean ± SEM; ∗p = 0.047; ∗∗∗∗p < 0.0001.

(G) Mean directionality of DCs along CCL19 gradients. n = 3 experiments; mean + SEM; Watson’s Large-sample non-parametric test; ns, not significant. Also see Figure S5 and Video S5.

Video S5. Deformation of collagen fibers is required for migration in fibrous environments, related to Figures 5 and S5

(1) Migrating dendritic cells (LifeAct-eGFP) deform collagen fibers (Alexa-594-labeled) (Spinning disk confocal microscopy).

(2) WT DC (LifeAct-eGFP) migrating in a fibrous collagen gel (Alexa 594-labeled) (Spinning disk confocal microscopy). Local actin polymerization at sites of a narrow pore deforms restrictive collagen fibers thereby providing space for the cell body to move forward as indicated by shape changes of the nucleus (Hoechst).

Orthogonal actin patches drive T cell migration in crowded LN

WASp deficiency affects multiple hematopoietic lineages however, the dominant phenotype of WAS patients is a congenital immunodeficiency due to defective T cell functions (Ochs, 2001). Cell migration is inevitably linked to activation of naive T cells, as T cells need to constantly scan cell-packed secondary lymphoid tissues to search for cognate antigen on antigen presenting cells (Krummel et al., 2016). While WASp plays an established role in T cell development and immune synapse formation its role in interstitial T cell migration remained unclear (Thrasher and Burns, 2010). To test if the mechanism we established for DCs is relevant for lymphocytes, we purified naive T cells from WASp−/− mice and confined them under intermediate-stiff agarose. In the presence of homogeneous CCL19, naive T cells become polarized and migrate randomly without the formation of substrate adhesions (Hons et al., 2018). Locomotion of WASp−/− T cells was significantly impaired compared with WT controls, and cells were frequently arrested under the load of the compressive overlay (Figure 6A; Video S6). Similar to confined DCs, WT T cells formed highly dynamic actin patches traveling backward with the retrograde actin flow (Figure 6B) and pushing vertically against the substrate (Figure S6A). These actin patches were virtually absent in WASp−/− cells (Figure 6B; Video S6). Subcellular mechanical loading with both micropipette tips (Figure S6B; Video S6) and nanometer-sized ridges (Figures 6C and S6C; Video S6) triggered the formation of WASp-dependent actin patches. Finally, cortical stiffness measured by submicron indentation with the tip of an AFM cantilever was significantly reduced in WASp−/− compared with WT cells (Figures 6D and S6D). Together, these data support a crucial role of WASp-dependent actin patches in the cortical mechanics of T cells. Next, we adoptively transferred labeled WT and WASp−/− naive T cells into WT recipient mice (Figure 6E). Since WASp−/− T cells show unimpaired homing to popliteal LN (Snapper et al., 2005), these experimental conditions allowed us to compare intranodal migration of both genotypes side by side (Figure 6E; Video S6). WASp−/− T cells were able to reach maximum speeds comparable to WT controls, indicating that WASp was not strictly required for locomotion of T cells in vivo (Figure S6D). However, a significantly larger fraction of cells showed periods of minimal displacement (<2.5 μm in 1 min) (Figure 6F), resulting in an overall reduced mean speed (Figures 6G and S6E). These data confirm that WASp-dependent actin patches are of general relevance for 3D migration under compressive loads across both myeloid and lymphoid hematopoietic lineages and provide a mechanistic framework for defective cell migration observed in X-linked WAS (Bouma et al., 2009).

Figure 6

Orthogonal actin patches drive T cell migration in crowded lymph nodes

(A) Schematic of experimental setup. Epifluorescence images (CMTMR) from representative time series: single-cell tracks are overlaid in light-blue; scale bar, 100 μm. Mean track speed was quantified. Each data point represents one track. WT: n = 290, WASp−/−: n = 335 pooled from 4 experiments; mean ± SD; Mann-Whitney test; ∗∗∗∗p < 0.0001. Fraction of time in arrest was extracted from tracks; mean + SEM; Mann-Whitney test; ∗∗∗∗p < 0.0001.

(B) Actin patches were segmented from LifeAct-eGFP movies (spinning disk confocal microscopy) in Ilastik and area was quantified. Actin patches traveled backward with the retrograde actin flow (kymographs); WT: n = 21 cells, WASp−/−: n = 22 cells; mean ± SD; Mann-Whitney test; ∗∗p = 0.0030; scale bar, 5 μm.

(C) Fraction of ridges covered by actin patches (segmented in Ilastik) (also see Figures S4B and S6B). WT: n = 11; WASp−/−: n = 13; Mean ± SD; Mann-Whitney test; ∗∗p = 0.0015.

(D) Schematic of an AFM experiment. Using a Hertz contact mechanics model, the elastic modulus was estimated by fitting the force indentation curves up to 500 nm. Each data point represents one measurement. WT: n = 182 from 16 cells; WASp−/−: n = 165 from 15 cells. For cellwise analysis see Figure S6C. Violin plot with median ± quartiles; Mann-Whitney test; ∗∗∗∗p < 0.0001.

(E) Adoptive transfer of fluorescently labeled T cells (WT and WASp−/−) into wild-type recipient mice. After 24 h, T cell migration in LN parenchyma was analyzed using intravital two-photon microscopy. Representative 3D reconstruction of adoptively transferred T cells migrating in LN parenchyma (2-photon intravital microscopy). Scale bar, 50 μm.

(F) Fraction of time in arrest was extracted from tracks (see H); mean + SEM; Mann-Whitney test; ∗∗∗∗p < 0.0001.

(H) Mean track speed. Each data point represents one track. WT: n = 262, WASp−/−: n = 237 pooled from 5 experiments; mean ± SD; Mann-Whitney test; ∗∗∗∗p < 0.0001. Also see Figure S6 and Video S6.

Video S6. Orthogonal actin patches drive T cell migration in crowded lymph nodes

WASp-dependent actin patches drive T cell migration under mechanical load, related to Figures 6 and S6

(1) Naive T cells (CMTMR-labeled) migrating under agarose (intermediate stiffness) (epifluorescence microscopy).

(2) Naive T cells (LifeAct-eGFP) migrating under agarose (intermediate stiffness) on inert (PLL-PEG) substrates. T cells show retrograde movement of actin patches (Spinning disk confocal microscopy). Actin patches were segmented in Ilastik (show in yellow).

(3) LifeAct-eGFP expressing WT and WASp−/− T cells were indented by the blunted end of a micropipette (Spinning disk microscopy). Lower panel shows overlay with the outline of the micropipette (LifeAct-eGFP is color coded).

(4) LifeAct-eGFP expressing WT and WASp−/− T cells migrating on inert surfaces (PLL-PEG) with nano-ridges (Spinning disk microscopy).

(5) WASp drives T cell migration in crowded lymph nodes. Adoptive transfer of fluorescently labeled T cells (WT = CMTMR [red] and WASp−/− = CFSE [green]) into wild-type recipient mice. After 24 h, T cell migration in LN parenchyma was analyzed using intravital 2-photon microscopy.

Discussion

First described in 1937, WAS is one of the most thoroughly investigated severe congenital immunodeficiencies (Ochs, 2001). WAS patients suffer from thrombocytopenia, eczema, and recurrent bacterial infections. Although underlying defects in cell-cell interactions and motility of the hematopoietic compartment have been attributed to the well-established function of WASp as one of the upstream activators of Arp2/3 nucleated actin polymerization, the precise cell biology behind WAS remained obscure (Fritz-Laylin et al., 2017a; Graziano and Weiner, 2014; Machesky and Insall, 1998; Thrasher and Burns, 2010). We here show that WASp drives cortical actin polymerization in the third dimension and that it does so in a mechanosensitive manner. WASp-dependent forces act orthogonal to the direction of cellular locomotion, and they are dispensable when the cell migrates in environments where the pore size is sufficiently large for the cell body to passage. WASp becomes rate limiting for migration when the cell needs to pass through restrictive environments that require the cell to deform its viscoelastic surrounding. WASp assembles into dot-like structures that are embedded into the actin cortex and act as nucleation sites for actin patches that then protrude against the external obstacle.

This function of WASp contrasts with that of the Arp2/3-activating WAVE complex. The WAVE complex does not assemble in discrete spots but in traveling waves: as WAVE activity is self-amplifying and at the same time negatively regulated by polymerized actin, it creates an excitable system that travels as a linear front (Graziano and Weiner, 2014). The resulting “in-plane of the membrane” polymerization pattern only allows protrusion at the tip of a strictly flat lamellipodium, where the sheet of plasma membrane curves to form an envelope. Accordingly, the orientation of Arp2/3 branches in lamellipodia is exclusively horizontal (Svitkina, 2018). WASp lacks the self-organizing feature of the WAVE complex. It stays confined in dot-like assemblies, and its lateral movement seems restricted to passive co-migration within the surrounding cortex. This lack of self-organization might allow WASp activity to “remain on target” when the cell pushes against a structure as delicate and small as a collagen fiber. When cells migrated on the nanotopographies, the precision with which the WASp dots located to the ridges was remarkable and suggested that WASp assembles in response to a highly localized signal. Importantly, this putative signal is independent of adhesive interactions, which is well in line with the ameboid principle that can work in the absence of any cognate adhesive interaction with a substrate. Topographical changes of the plasma membrane as sensed e.g., by BAR domain-containing proteins are the most attractive candidates for such an alternative sensory function (Lou et al., 2019; Suetsugu and Gautreau, 2012; Brunetti et al., 2021). Supporting this hypothesis, we show that the F-BAR domain protein Cip4 accumulates at sites of membrane indentation where it precedes the formation of actin patches.

How cells adapt to compressive load in 3D tissue environments is still incompletely understood. Recent studies proposed an evasion reflex in response to cellular compression: upon deformation of the nucleus cells increase their cortical contractility to move away and squeeze out of tight spaces or crowded tissue regions (Lomakin et al., 2020; Venturini et al., 2020). However, to reach their target sites, leukocytes often need to traverse tissues of high density. Mechanosensitive actin patches pushing against obstructing barriers facilitate this and might as such represent a very primordial invasive program. Indeed, until to date, one of the best-established functions of WASp and its non-hematopoietic isoform N-WASp are invasive podosomes and invadosomes (Murphy and Courtneidge, 2011). Podosomes are adhesive organelles, composed of a protrusive actin core, which is surrounded by an obligate adhesive ring-structure, presumably to counter the pushing force of the core (van den Dries et al., 2019). While matrix-degrading activity, due to targeted delivery of proteases is a further defining feature of podosomes (Linder, 2007), some non-proteolytic podosome-like organelles have been described, but these structures were strictly coupled to cell-cell adhesions (Carman et al., 2007; Kumari et al., 2015; Poulter et al., 2015; Sens et al., 2010). Consequently, the WASp patches we describe may represent the most rudimentary podosome-like structure, reduced only to the protrusive actin core. It seems plausible that in very-fast-migrating and low-adhesive leukocytes WASp-driven actin patches are not locked in place by an adhesive ring and thus lack the maturation signal that would turn them into degradative podosomes. Notably, work in invasive C. elegans anchor cells showed that in WASp-dependent sites of basement membrane invasion, protrusive forces and proteolytic degradation synergize to drive barrier-penetration (Cáceres et al., 2018). Upon inhibition of proteolysis, invasion can partially proceed in an entirely mechanically driven fashion (Kelley et al., 2019).

The protrusive actin patches we describe also share some features with early stages of clathrin-mediated endocytosis, where the positive curvature of the plasma membrane is associated with WASp-driven actin polymerization (Almeida-Souza et al., 2018). Actin polymerization on clathrin coated pits (CCP) is oriented normal to the cell surface, allowing it to push the CCP inward (Collins et al., 2011; Picco et al., 2015; Akamatsu et al., 2020). This orientation is consistent with our findings that WASp plays a unique role in driving cortical actin polymerization in the third dimension.

Taken together, our data support the general notion that, in contrast to inside-out activated lamellipodia that protrude in the direction of migration, outside-in activated actin patches push orthogonally to the plasma membrane, thereby rendering cells mechanoactive in 3D.

Limitations of the study

While our study shows that positive membrane curvature recruits WASp-dependent actin patches to the plasma membrane, the underlying mechanisms acting upstream of WASp remain unclear. Curvature-sensing proteins, such as F-BAR domain proteins, are promising candidates, but future work still needs to decipher their functional role in this process.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Michael Sixt (michael.sixt@ist.ac.at).

Materials availability

This study did not generate new unique reagents.

Experimental model and subject details

Mouse strains

C57BL/6 (Janvier); WASp-/- (B6.129S6-Was/J; No. 019458; The Jackson Laboratory); LifeAct-eGFP (Riedl et al., 2010); LifeAct-eGFPxWASp-/-; LifeAct-eGFPxHEM1-/- (Leithner et al., 2016). All strains were backcrossed to and maintained on C57BL/6-background. For in vivo studies mice were sex-matched and experiments were performed at the age of 8-12 weeks. Mice were bred and maintained at the local animal facility in accordance with the IST Austria ethics commission and the Austrian law. Permission was granted by the Austrian Federal Ministry of Science, Research and Economy (identification code: BMWF-66.018/0005-II/3b/2012).

Primary cells

All primary dendritic cells and T cells originated from 6-12-week-old, male or female mice (see above) and were isolated and cultured as described in detail below. No cell lines used in this study were found in the database of commonly misidentified cell lines that is maintained by ICLAC and NCBI biosample. The cell lines were not authenticated. Cell lines tested negative for mycoplasma. Cell culture medium: R10 medium, consisting of RPMI 1640, supplemented with 10% fetal calf serum (FCS), 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, and 50 μM 2-mercaptoethanol (all Invitrogen), was used as basic medium for all cells unless stated otherwise. All cells were grown and maintained at 37 °C / 5% CO2 unless noted otherwise.

Method details

Differentiation and culture of mature dendritic cells

Dendritic cells (DCs) were differentiated from bone marrow (2.0 x 106 BM-cells in 10 ml / dish) or transiently immortalized hematopoietic precursors (0.5 x 106 precursor cells in 10 ml / dish) (Leithner et al., 2018; Redecke et al., 2013), both originating from 6-12-week-old, male or female C57BL/6J, LifeAct-eGFP, WASp-/-, LifeAct-eGFPxWASp-/- or LifeAct-eGFPxHem1-/- mice. Immortalization of hematopoietic precursors: the bone marrow of 6-12-wk-old mice was isolated and hematopoietic progenitor cell lines were generated by retroviral delivery of an estrogen-regulated form of HoxB8 (Leithner et al., 2018; Redecke et al., 2013).

Cells were differentiated in 9 ml R10, supplied with 1 ml in-house-generate granulocyte macrophage colony-stimulating factor (GM-CSF) hybridoma supernatant. On day 3, 8 ml R10, supplied with 2 ml GM-CSF, was added. Half of the medium was replaced on day 6. On day 8, lipopolysaccharide (LPS) from Escherichia coli 0127:B8 (Sigma) was added to an end concentration of 200 ng/ml to mature DCs overnight.

Purification and culture of naïve T-cells

Peripheral LNs and spleens were harvested and homogenized with a 70 μm cell strainer. Untouched primary naïve T cells were isolated with an EasySep Mouse T cell Isolation Kit according to the manufacturer’s protocol (STEMCELL Technologies, 19851 A). T-cells were harvested 1 d before imaging and incubated overnight at 37 °C under 5% CO2 in R10 medium consisting of RPMI 1640 supplemented with 10% FCS, 2 mM l-glutamine, 100 U/mL penicillin 100 μg/mL streptomycin, 1 mM sodium pyruvate, 100x nonessential amino acids and 50 μM 2-mercaptoethanol (all from Invitrogen). T cells were fluorescently labeled with 10 μM CMTMR (CellTracker Orange or 2 μM CFSE for 15 min at 37 °C.

Plasmids and lentivirus production

The following fusion constructs were used: eGFP-Abi1 (Lai et al., 2008), eGFP-WASp (Benesch et al., 2002), Cip4-mCherry (Taylor et al., 2011), CLC-mCherry (Taylor et al., 2011), LifeAct-mCherry (Riedl et al., 2008). The MLC-mKate2 fusion construct was generated by exchanging GFP against mKate2 from a previously generated construct (Croft et al., 2005). Fusion-protein-coding lentiviruses were produced Lenti-X™ 293T Cell Line (Takara Bio) by co-transfection of the expression plasmids (pLenti6.3, Invitrogen) with pCMV-dR8.91 packaging (Creative Biogene) - and pCMV-VSV-G envelope plasmids (a gift from B.Weinberg, MIT, USA, Addgene plasmid no. 8454) (Stewart et al., 2003).

Transfection

For transgene delivery, bone marrow-derived DCs were transfected with 4 μg DNA, according to manufacturer guidelines using the nucleofector kit for primary T cells (Amaxa; Lonza Group). Briefly, 5 × 106 cells were resuspended in 100 μl reconstituted nucleofector solution and transferred to an electroporation cuvette, and a total amount of 4 μg plasmid DNA was added. Cells were transfected by using a protocol specifically designed for electroporating immature mouse DCs (program X-001). Transfected DCs were used one day after transfection and enriched by FACS for GFP-expressing cells. Immortalized hematopoietic precursors were spin-infected (1,500 g, 1 h) with lentiviruses in the presence of 8 μg/ml Polybrene. Following transduction, cells were selected for stable virus insertion using 10 μg/ml Blasticidin (pLenti6.3) for at least 1 week.

Enucleation of matured dendric cells

Enucleation was performed as described previously (Graham et al., 2018). From a 50% (wt/vol) solution Ficoll-400 with PBS (Fisher scientific BP525) a 30% (vol/vol) stock solution was made with D10 (DMEM containing 10% FBS and 100ul/ml Pen-strep). The stock solution was filtered with a 0.4 PES filter. From the stock solution 30%, 20%, 18%, 15% Ficoll working solutions were prepared freshly on the day of experimentation by adding D10 containing cytochalasin B (10μg/ml) (#5474, Tocris) and DMSO (0.2%). 2 ml of each concentration were layered into an ultracentrifuge tube (13.2 ml thin wall, Thermofisher scientific), starting with 30% followed by 20%, 18 % and 1ml of 15%. The tube was covered and the gradient was incubated overnight at 37°C. The next day, 1-2x107 matured DCs were pelleted and the pelleted was resuspend with 1 ml pre-warmed 15% Ficoll and layered on top of the gradient. The tube was then filled with D10 containing cytochalasin (10μg/ml). The samples were loaded into a pre-warmed SW641 rotor of a Sorval wx100 (Thermos Scientific). The cells were centrifuged for 1 h at 31°C with 27000rmp (started with acceleration of 9 and stopped with deceleration of 1). After centrifugation the visible cell fraction was pipetted into a 50 ml tube. And the cells were washed 3x with 40 ml PBS at 330g. The cell pellet was resuspended in 1ml R10 medium and transferred into a 24 well plate, 1 drop of nucBlue (#R37605, Life tech.) and vitamin C (50μM) was added. After 30 min cells were spun down and used for the under-agarose assays.

Pharmacological inhibitors

The following small molecule inhibitors were used to perturb actin dynamics. Inhibitors were mixed with the cell suspension and the agarose (before polymerization) using the indicated final concentration. The Arp2/3 Complex Inhibitor I, CK666 (100 μM; #SML0006; Sigma); the Arp2/3 Complex Inhibitor I, Inactive Control; CK689 (100 μM; #US1182517; Merck Millipore); the Formin FH2 Domain Inhibitor, para-nitro-Blebbistatin (#ax494693; Axol)

Antibodies

Monoclonal rabbit anti-Cav1 (clone D46G3) (#3267) and monoclonal rabbit anti-CHC (clone D3C6) (#4796) antibodies were both purchased from Cell Signaling. Goat anti-rabbit IgG coupled to Alexa488 (#A11008; Invitrogen) was used as secondary antibody. F-actin was labelled using AF-594-coupled Phalloidin (#A12381; Thermo Fisher).

Under-agarose migration assay of naïve T-cells

Glass coverslips were overlaid with 4 μg/mL rmICAM-1/Fc (R&D, 796-IC). An 1% agarose block was formed by mixing (i) one part 2x HBSS buffer (Sigma), (ii) two parts RPMI (Invitrogen) supplemented with 20% BSA (instead of FCS) (Sigma) with 2x concentrations of all other supplements used in R10 medium (see above) and (iii) one part 4% high-molecular weight agarose (Biozym Gold Agarose, 850152) in water at 52 °C. CCL19 (20 ng/ml) (Peprotech, 250-27B) was added to soluble agarose before casting. Liquid agarose was subsequently poured into a dish, covering the coated coverslip. The agarose blocks were allowed to solidify at room temperature and were equilibrated first at 4 °C for 1 h and subsequently at 37 °C and 5% CO2 for 30 min. T cells were injected under the agarose block with a micropipette and allowed to polarize for at least 30 min at 37 °C under 5% CO2 before imaging. Epifluorescence movies were recorded using the same settings as described above. Images were taken every 30 s at 6 multi-positions with NIS Elements software (Nikon Instruments). Spinning disc microscopy was performed on an inverted spinning-disc confocal microscope (Andor) using a 100x/1.4 NA objective and a 488 nm laser line in a custom-built climate chamber (37°C under 5% CO2). Time-lapse movies were recorded every two seconds.

Under-agarose migration assay of mature dendritic cells on adhesive substrate (invasion under agarose)

Glass coverslips were washed with isopropanol, ethanol, and dH2O; subsequently, plasma-cleaned (pdc-002 plasma cleaner, Harrick) and glued to a petri-dish with a 17mm hole. To obtain humid migration chambers, a 17 mm plastic ring was attached to a glass-bottom dish using paraffin (Paraplast X-tra; Sigma). Agarose blocks were formed by mixing (i) one part 2x HBSS buffer (Sigma), (ii) two parts RPMI (Invitrogen) supplemented with 20% FCS (Invitrogen) with 2x concentrations of all other supplements used in R10 medium (see above) and (iii) one part 4% UltraPure Agarose (Invitrogen) in water at 52 °C. Increasing agarose stiffnesses were achieved by mixing 2% agarose for low stiffness (2.5 kPa), 4% for intermediate stiffness (10 kPa), and 6% for high stiffness (17.5 kPa) (Biozym Gold Agarose, 850152) (Hons et al., 2018). 500 μl liquid agarose was subsequently poured into a dish, covering the coverslip. The agarose blocks were allowed to solidify at room temperature for 5 min. After polymerization, two 2-mm holes (5 mm apart) were punched into the agarose pad followed by 30 min equilibration at 37°C, 5% CO2. 2.5 μg/ml CCL19 (PeproTech) was placed into one hole to generate a soluble chemokine gradient. 0.5 x 106 mDCs were placed in the second hole opposite to the chemokine hole. Before the acquisition, dishes were incubated at least 2 h at 37°C, 5% CO2 to allow invasion under the agarose (cells are now confined between the coverslip and the agarose). During the acquisition, dishes were held under physiological conditions at 37°C and 5% CO2. Epifluorescence movies were recorded with an inverted wide-field Nikon Eclipse Ti-2 microscope in a humidified and heated chamber at 37 °C and 5% CO2 (Ibidi Gas Mixer), equipped with a 20x/0.5 NA PH1 air objective, a Hamamatsu EMCCD C9100 camera and a Lumencor Spectra X light source (390 nm, 475 nm, 542/575 nm; Lumencor). Images were taken every 30 s or 3 min at 6 multi-positions with NIS Elements software (Nikon Instruments).

Under-agarose migration assay of mature dendritic cells on slippery substrate (injection under agarose)

Plasma-treated (pdc-002 plasma cleaner, Harrick) glass coverslips were incubated with Poly-2-methyl-2-oxazoline (1 mg/ml for 1h at RT; PAcrAm™-g-(PMOXA); SuSoS Surface Technology) to generate an inert, non-adhesive coating. Agarose blocks were generated as described above. After polymerization (5 min at RT), a 2-mm hole was punched into the agarose pad followed by 30 min equilibration at 37°C, 5% CO2. 2.5 μg/ml CCL19 (PeproTech) was placed into the hole to generate a soluble chemokine gradient. The cell suspension was injected under the agarose opposite the chemokine hole to confine migrating DCs between the coverslip and the agarose. Before the acquisition, dishes were incubated at least 2 h at 37°C, 5% CO2 to allow recovery and persistent migration of cells towards the chemokine source. TIRF microscopy was performed with a 60/1.46 NA oil objective, optovar 1x or 1.6x in a humidified and heated chamber at 37°C and 5% CO2 using an inverted Axio Observer (Zeiss) microscope, a 488 nm laser and an Evolve EMCCD camera (Photometrics) controlled by VisiView software (Visitron Systems). Images were recorded every two seconds. Spinning disc microscopy was performed using the same settings as described for under agarose assays. To record actin spike formation z-stacks of 3 images (0.5 μm step size) were recorded every two seconds. Interference Reflection Microscopy (IRM) was performed on a Leica SP5 inverted confocal microscope. Image acquisition was performed as previously described (Barr and Bunnell, 2009).

Migration in microfabricated pillar forests and straight and constricted channels

Microfluidic devices with pillars were micro-fabricated with polydimethylsiloxane (PDMS) (Leithner et al., 2016). Photomasks were designed using Coreldraw X18, printed on a chrome photomask (1 μm resolution; JD Photo data), followed by a spin coating step using SU-8 2005 (3,000 rpm, 30 s; Microchem) and a prebake of 3 min at 95°C. The wafer was then exposed to 100 mJ/cm2 ultraviolet light on an EVG 610 mask aligner. After a postexposure bake of 3 min at 95°C, the wafer was developed in propylene glycol methyl ether acetate (PGMEA). A 1h silanization with Trichloro(1H,1H,2H,2H-perfluorooctyl)silane was applied to the wafer. The devices were made with a 1:10 mixture of Sylgard 184 (Dow Corning), and air bubbles were removed with a desiccator. The PDMS was cured overnight at 85°C. Microdevices were attached to isopropanol/ethanol-cleaned coverslips and incubated for 1 h at 85°C after plasma cleaning (pdc-002 plasma cleaner, Harrick). Before the introduction of cells, devices were flushed and incubated with complete medium for at least 1 h. Figure S4I: Dimensions of the pillars were 5 x 30 μm (height x width). The spacing between pillars was 20 μm. Figure 4L: 5 x 5 μm (height x width) and 1 μm, 2 μm or 3 μm spacing. Figure 4M: The dimensions of the straight channels are 5 μm width, 4 μm height and the dimensions of constrictions are 1.5 μm width, 4 μm height, 15 μm length. Brightfield movies of DCs migrating pillar mazes were acquired by time-lapse acquisition (time interval of 60 s) using inverted cell culture microscopes (DM IL Led, Leica Microsystems) equipped with cameras (ECO415MVGE, SVS-Vistek) and custom-built climate chambers (5% CO2, 37°C, humidified).

Collagen migration assay

Custom-made migration chambers were assembled by using a plastic dish containing a 17-mm hole in the middle, which was covered by coverslips on each side of the hole (Sixt and Lämmermann, 2011). 3D scaffolds consisting of 0.75 / 1.5 / 3 / 5 mg/ml bovine collagen I (PureCol, Nutragen, Fibricol; all AdvancedBioMatrix) were generated by mixing 1.5 × 105 cells in suspension (R10) with collagen I suspension buffered to physiological pH with Minimum Essential Medium and sodium bicarbonate in a 1:2 ratio. To allow polymerization of collagen fibers, gels were incubated 1 h at 37°C, 5% CO2. Directional cell migration was induced by overlaying the polymerized gels with 0.63 μg/ml CCL19 (R&D Systems) diluted in R10. To prevent drying out of the gels, migration chambers were sealed with Paraplast X-tra (Sigma). Bright-field movies were acquired by time-lapse acquisition (time interval of 60 s) using inverted cell culture microscopes (DM IL Led, Leica Microsystems) equipped with cameras (ECO415MVGE, SVS-Vistek) and custom-built climate chambers (5% CO2, 37°C, humidified).

To visualize collagen fibers using spinning disc microscopy, collagen was directly conjugated to Alexa Fluor 594 NHS Ester (Succinimidyl Ester, ThermoFisher). Collagen was added to SnakeSkin Dialysis Tubes, 10K MWCO, 16mm (ThermoFisher), and immersed in 100mM NaHCO3 overnight at 4°C to allow polymerization. Alexa Fluor 594 NHS Ester (1.5mg/mL) was added to the polymerized collagen and incubated for 3h. To remove the unconjugated dye, the collagen mixture was placed in 0.2% acetic acid in deionized water for further dialysis overnight at 4°C.

Imaging of LifeAct-eGFP expressing DCs in Alexa-594-labeled collagen matrices was performed on an inverted spinning-disc confocal microscope (Andor Dragonfly 505) using a 60x/1.4 NA objective and a 488 nm / 561 laser line in a custom-built climate chamber (37°C under 5% CO2). Z-stacks (1.5μm step size) of migrating cells in labeled collagen matrices were acquired using an Andor Zyla camera (4.2 Megapixel sCMOS) every 60 seconds for 20-25min.

Migration on nano-ridges

Fabrication of coverslips with nano-ridges

Clean glass coverslips were coated with a 50 nm reflective layer of chromium. An electron beam resist (AR-P 6200.13) was spin-coated on the chromium covered coverslip. The inverse pattern of ridges was written using the e-beam lithography tool and subsequently developed using AR600-546. Chromium was dry-etched in Cl2-O2 plasma inside an ICP (Inductively coupled plasma) chamber. The chamber was pumped to a pressure of 10 mTorr with a gas flow of 26 sccm for Cl2 and 4 sccm for O2. The forward power and ICP power were 20 W and 400 W, respectively. Glass was subsequently etched using a SF6-Ar plasma inside an ICP chamber, pumped to a pressure of 10mTorr with a gas flow of 100 sccm for SF6 and 67 sccm for Ar. The forward power and ICP power were 100 W and 1500 W, respectively. After dry-etching, the remaining chromium layer was wet etched at room temperature using Chromium etchant. The height of the ridges is determined by the etching time of the glass surface, on average the etching rate of the glass is around 118.5 nm/min. After the wet etching step, the height of the ridges is verified using an atomic force microscope (NX10 from Park Systems) in non-contact mode.

Under agarose assay on nano-ridges

Glass coverslips with ridges were rinsed with isopropanol, ethanol and dH2O, air-dried and plasma cleaned (pdc-002 plasma cleaner, Harrick). They were then incubated with Poly-2-methyl-2-oxazoline (1 mg/ml for 1h at RT; PAcrAm™-g-(PMOXA); SuSoS Surface Technology) or PLL-PEG (1 mg/ml for 1h at RT; PLL-PEG; SuSoS Surface Technology) to generate an inert, non-adhesive coating. Agarose blocks (1%) were generated as described above and matured DCs or purified naïve t-cells were injected under the agarose using a micropipette. Spinning-disc confocal microscopy and TIRF microscopy were performed as described above.

Micropipette indentation assay

Glass bottom dishes (50 mm dish diameter, 14 mm glass diameter, glass coverslips No. 1, Mattek) were plasma cleaned (pdc-002 plasma cleaner, Harrick) and coated with 1x poly-L-lysine (P8920, Merck) in dH2O for 10 min. Dishes were washed twice with dH2O and then dried for at least 4h at room temperature. Cells in R10 (mDCs or t-cells expressing LifeAct-eGFP) were incubated for 15 min at 37 °C and dishes were carefully washed once with R10 containing HEPES (10mM; Sigma) to remove floating cells. Dishes were immediately mounted on an inverted spinning-disc confocal microscope (Andor) equipped with a micromanipulator (Eppendorf) and maintained at 37°C in a custom-built climate chamber. Micropipettes (blunt; inner diameter 4 μm; bent angle 30°) (BioMedical Instruments) were centrally positioned over the cell and carefully lowered to indent the cell body. Movies were recorded using a 100x/1.4 NA objective and a 488 nm laser line. Z-stacks of 3 image (0.5 μm step size) were recorded every two seconds.

Electron microscopy

Sample preparation and light microscopy examination (CLEM)

Under-agarose assays were performed as described above with minor modifications. Removable culture inserts (Cat.No: 81176; Ibidi) were attached to plasma-cleaned (pdc-002 plasma cleaner, Harrick) glass coverslips or tailored Aclar foil (Science Services) (used for serial sectioning). Culture inserts were then filled with 200 μl of agarose-mix and experiments were performed as described above. Dendritic cells were allowed to migrate under the agarose for > 2 H. Samples were subsequently fixed in cytoskeleton buffer (10 mM MES buffer, 150 mM NaCl, 5 mM EGTA, 5 mM glucose and 5 mM MgCl2; pH=6.1) containing 2% PFA; 2.5% glutaraldehyde (EMS); 0.01% Triton-X 100 (Sigma); phalloidin-Alexa488 (1:40; Invitrogen) for 30 min at 37°C. After fixation, agarose pads and removable culture inserts were carefully removed using a coverslip tweezer. Z-Stacks of palloidin-Alexa488 stained cells were performed on an inverted spinning-disc confocal microscope (Andor) using a 100x/1.4 NA objective and a 488 nm laser line. For CLEM (correlative light and electron microscopy) epifluorescence images of phalloidin-Alexa488 stained cells were recorded with an inverted wide-field Nikon Eclipse Ti-2 microscope equipped with a 20x/0.5 NA PH1 air objective, a Hamamatsu EMCCD C9100 camera and a Lumencor Spectra X light source (475 nm Lumencor). Images were taken at multi-positions with NIS Elements software (Nikon Instruments). Fluorescent and SEM images (see below) were manually aligned using FIJI.

Scanning electron microscopy

Samples were dehydrated in a graded ethanol series at RT. For further chemical drying, the samples were first incubated in a 1:1 mixture of 100% ethanol and hexamethyldisilazane (HMDS) for 30min at RT and then transferred to 100% HMDS for 1h at RT. Access HMDS was removed with a pipette followed by overnight evaporation at RT. The cover slips with the dried cells were coated with platinum to a thickness of 5nm using an EM ACE600 coating device (Leica Microsystems). The samples were observed with a FE-SEM Merlin compact VP scanning electron microscope (Zeiss) at 5kV using a secondary electron detector.

Contrast enhancement and resin embedding for serial sectioning

Samples on Aclar foil were post-fixed with 1% glutaraldehyde (GA) (EMS) in phosphate buffer (PB; 0.1M, pH 7.4) for 10 min at RT. After a brief wash with PB, samples were contrast-enhanced with 0.5% tannic acid (TA; Sigma) in PB (w/v) for 45 min at 4 °C in the dark. The solution was replaced with freshly prepared 0.5% TA in PB and samples incubated for another 45 min at 4 °C in the dark. After a wash with PB, samples were incubated in 1% aqueous osmium tetroxide (w/v; EMS) for 30 min at 4 °C in the dark. Then, they were washed with Milli-Q water and incubated in 1% aqueous uranyl-acetate (w/v; AL-Labortechnik) overnight at 4 °C in the dark. After wash with Milli-Q water, samples were contrast-enhanced further according to en bloc Walton’s lead aspartate staining (Walton, 1979). Samples were incubated in lead aspartate solution for 30 min at 60 °C. After a wash with Milli-Q water, samples were dehydrated in graded ethanol (10%, 20%, 50%, 70%, 90%, 96% and 100%) for 5 min each at 4 °C. They were then placed in propylene oxide puriss. p.a. (Sigma) twice for 10 min at 4 °C and embedded in Durcupan™ ACM epoxy resin (Sigma). To that, hard grade Durcupan™ was formulated by weight as follows: component A 11.4 g, component B 10 g, component C 0.3 g, component D 0.1 g. Samples were consecutively infiltrated in mixtures of 3:1 propylene oxide/Durcupan™, 1:1 propylene oxide/Durcupan™ and 1:3 propylene oxide/Durcupan™ for 1.5 h each at 4 °C. Then they were infiltrated in mere Durcupan™ overnight at RT. The Aclar foil was removed from the resin block with a razor blade under a stereo microscope (Nikon SMZ 800). Trimmed samples were placed on PELCO® Cavity Embedding Molds (Ted Pella Inc.), cavities filled with freshly prepared Durcupan™ and resin cured at 60 °C in an oven over two days.

Serial sectioning and scanning

Embedded samples were trimmed with an Ultratrim diamond knife (Diatome) to a rectangle (70μm x ∼1mm) with a slanted side for orientation using an EM UC7 ultramicrotome (Leica Microsystems). Prior to serial sectioning, a 25 x 25mm silicon wafer was plasma treated using an ELMO glow discharge cleaning system (Agar Scientific) for increasing the hydrophilicity of the wafer. Serial sections were cut at a thickness of 80 nm using a 4mm Leica AT-4 35° diamond knife (Diatome). Section ribbons were collected onto a plasma treated wafer using the water drain device of the knife. Then the wafer with the serial sections was coated with carbon to a thickness of 5nm using an EM ACE600 coating device (Leica Microsystems) to ensure conductivity. The serial sections were observed under a FE-SEM Merlin compact VP scanning electron microscope (Zeiss) equipped with the Atlas 5.3.2.9 Array Tomography software (Zeiss). The images were acquired using a backscattered (5nm pixel size) and a secondary electron detector (7nm pixel size) at 5kV.

Transmission electron microscopy

For transmission electron microscopy, samples were cut with 70 nm using an EM UC7 ultramicrotome (Leica Microsystems GmbH, Austria) and analyzed with a Tecnai 12 (FEI/Thermo Fisher Scientific, The Netherlands). Large area montaged images were collected using SerialEM (Mastronarde, 2003) (Webpage: https://bio3d.colorado.edu/SerialEM/), and then stitched in IMOD software (Kremer et al., 1996) (http://bio3d.colorado.edu/imod). The acquisition was carried out using a pixel size of 2.094 px/nm and a 20% overlap of the tiles.

Atomic force microscopy

Glass bottom dishes (30 mm dish diameter, 14mm glass diameter, glass coverslip No. 1, Mattek) were plasma cleaned (pdc-002 plasma cleaner, Harrick) and coated with 1x poly-L-lysine (P8920, Merck) in dH2O for 10min. Dishes were washed twice with dH2O and then dried for at least 4h at RT. Cells in R10 (mDCs or t-cells expressing LifeAct-eGFP) were incubated for 15 min at 37°C and dishes were carefully washed once with R10 containing HEPES (10mM; Sigma) to remove floating cells. Dishes were immediately mounted on the atomic force microscope equipped with a climate chamber (37°C). AFM nanoindentation was performed on a Nanowizard4 AFM microscope from JPK Instruments (Bruker) interfaced to an inverted optical microscope (IX81, Olympus). We used cantilevers with spring constants of 0.1 N m−1 and a tip radius of 10 nm (qp-BioAC, Nanosensors). Cantilever actual spring constants were determined using the thermal noise method implemented in the JPK software. Using brightfield microscopy, we positioned the tip of the cantilever over the central region above the cell body and performed 5-10 indentation measurements. Force-distance curves were acquired with an approach speed of 2 μm s −1 until reaching the maximum set force of 10 nN. Measurements were restricted to indentation depths of 500 nm for T-cells and 1000 nm for DCs (<10% of the height of the cell) to minimize the contribution of the substrate, and to maximize the contribution of the cortex stiffness to cantilever deflection. Elastic moduli (young’s modulus) were determined from force-distance curves with the Hertz model as implemented in the JPK analysis software (parabolic model).

Dendritic cell lymph node homing assay

Bone marrow-derived dendritic cells were differentiated and matured as described above. Mature DCs were harvested, counted, and adjusted to 1x107 cells/ml at RT 1xPBS. For labeling, Tetra-Methylrhodamine (TAMRA; TAMRA, SE; 5-(and-6)-Carboxytetramethylrhodamine, Succinimidyl Ester (5(6)-TAMRA, SE; Invitrogen) and Oregon Green® (CellTrace™ Oregon Green® 488 carboxylic acid diacetate, succinimidyl ester carboxy-DFFDA, SE; Invitrogen) was added to a final concentration of 10 μM TAMRA or 3 μM Oregon Green®, respectively. After a 10 min incubation fresh R10 medium was added to the cell suspension to stop the reaction, and the cells were pelleted. Subsequently, cells were resuspended in pre-warmed (37°C) R10 medium and incubated for another 30 min at 37°C for esterification. Finally, cells were washed twice with 1xPBS and subsequently used for experiments.

WT and WASp-/- were differently labeled, mixed at 1:1 ratio, and adjusted to a final concentration of 4x107 cells/ml in 1xPBS. 25 μl (= 1x106 cells) were injected into the mouse hind footpads. Draining popliteal lymph nodes (LNs) were harvested after 24 h and transferred to a polystyrene FACS tube (Falcon BD) containing 0.5 ml complete DMEM (2.5% FCS (Gibco), 10 mM HEPES (Sigma-Aldrich), 5% penicillin/ streptomycin (Gibco) and 5 % glutamine (Gibco)) for subsequent isolation of the DCs for flow cytometry analysis (one LN per tube). LNs were then opened and cut into pieces in the tube using sterile scissors. Afterwards, LN fragments were incubated in digestion buffer (complete DMEM, 3mM CaCl2 (Sigma-Aldrich), 0.5 mg/ml collagenase D (Roche), 40 μg/ml DNase I (Roche), EDTA 0.5 M, pH 7.2: Ethylenediaminetetraacetic acid (EDTA, Sigma-Aldrich)) for 30 min at 37 °C in a water bath and solution was thoroughly pipetted every 10 min using a 1 ml pipette to further disrupt the fragments. The enzymatic reaction was stopped after 30 min by the addition of 10 mM EDTA. The cell solution was again thoroughly pipetted to disrupt remaining LN fragments and subsequently flushed through a cell strainer into a fresh FACS tube (tube with cell strainer cap, BD Falcon). After washing with FACS-buffer, cells were directly stained with labeled primary antibodies (CD11c (Antigen: CD11c; conjugated: APC; eBioscience; Cat-No 17-0114-82; Clone N418) and mouse MHC II (I-A/I-E) (Antigen: MHC II; conjugated: eFluor450; eBioscience; Cat-No 48-5321-82; Clone M5/114.15.2)), resuspended in an appropriate volume of FACS buffer and used for FACS analysis. Flow-cytometric analysis was performed on FACS Canto II (Becton Dickinson) or FACS Aria III (Becton Dickinson) using FACS DIVA software (Becton Dickinson) for acquisition and FlowJo (Treestar) for analysis.

Intravital two-photon microscopy of popliteal lymph nodes

Freshly purified T cells were fluorescently labeled with 20 μM CMTMR (CellTracker Orange or 5 μM CFSE for 15 min at 37 °C. After being washed, labeled T cells were i.v. injected retro-orbitally into sex-matched 5- to 10-week-old WT C57BL/6 recipient mice and were allowed to home to lymph nodes for at least 18 h. Recipient mice were anesthetized by intraperitoneal injections of ketamine (50 mg kg-1), xylazine (10 mg kg-1) and acepromazine (4 mg kg-1). The right popliteal lymph node was prepared micro-surgically for intravital microscopy and positioned on a custom-built microscope stage. Care was taken to spare blood vessels and afferent lymph vessels. The prepared LN was submerged in normal saline and covered with a glass coverslip. A thermocouple was placed next to the LN to monitor local temperature, which was maintained at 37°C. To label blood vessels, 50 μl of Evans blue (1mg/ml) (Merck) were i.v. injected retro-orbitally before imaging. Two-photon microscopy of right popliteal LNs was performed with a Trimscope II multi-photon imaging platform (LaVision Biotech) on an upright Olympus stand. Images were acquired using a Plan-Apochromat 20x/1.0 NA objective (Carl Zeiss Microscopy) with H2O as an immersion medium. A Chameleon Ti:Sapphire laser (MaiTai) was tuned to 840 nm, and an optical parametric oscillator (OPO) tuned to 1100 nm for simultaneous excitation of CFSE, CMTMR, and Evans blue. Fluorescent signals were collected using four external/non-descanned photomultipliers (PMTs) (3 Hamamatsu H7422-40 GaAsP High Sensitivity PMTs and 1 Hamamatsu H7422-50 GaAsP High Sensitivity red-extended PMT). For four-dimensional analysis of cell migration, z-stacks with 11–25 slices (spacing 4μm) of 250–300 x 250–300 μm x–y sections were acquired every 20 s for 30–45 min.

Image analysis

FIJI imaging processing software (https://fiji.sc/) was used for basic image and video microscopy analysis (Schindelin et al., 2012).

3D analysis of T-cell migration in LNs

Imaging sequences of image stacks were transformed into volume-rendered four-dimensional movies using Imaris software (v9; Bitplane), which was also used for semiautomated tracking of cell motility in three dimensions. The average track velocity and instantaneous velocities were calculated from the x, y, and z coordinates of cell centroids.

2D analysis of migration assays (under agarose; pillar forests; collagen assays)

The average track velocity and instantaneous velocities were calculated from the x, y-coordinates of the nucleus (under agarose assays of dendritic cells) or the cell centroid (under agarose assay of T-cells / dendritic cells migrating in pillar forests or in collagen gels) tracked in TrackMate (Tinevez et al., 2017) (https://imagej.net/TrackMate).

Analysis of directionality

TrackMate files (XML) were imported to TraXpert (a track analysis software based on R v4.03 and R Shiny v1.6.0). Directionality plots were generated using ggplot2 package v3.3.3 with polar coordinates. Each track was grouped in 12 cardinal directions ranging in ±15° (e.g. -15° to +15° is a cardinal direction at 0°) for each replicate and number of tracks were calculated. Percentage of tracks in each cardinal direction compared to all tracks were calculated. Replicates in each group (e.g. genotype) were aggregated by mean for each cardinal group. Watson’s Large-sample non-parametric test was used to test for common mean direction (Pewsey et al., 2013). Trajectory plots were generated by plotting each trajectory starting at the origin of Cartesian coordinate system.

Fraction of time in arrest

T-cells were classified as being in arrest when the cell centroid remains confined within a radius of 2.5 μm during an interval of 1 min. Dendritic cells were classified as being in arrest when the cell nucleus remains confined within a radius of 2.5 μm during an interval of 3 min. “Fraction of time in arrest“ was calculated by ratio of the time spent in arrest to the total time.

Segmentation of actin patches

To quantify the area and dynamics of actin patches we first segmented actin patches from LifeAct-eGFP expressing cells with interactive machine-learning using Ilastik (Berg et al., 2019). Training was performed on WT cells and the same trained workflows were then applied to WASp-/- cells. To measure the dynamic changes of actin patches from frame to frame, we calculated the Jaccard similarity coefficient, defined by the overlap of segmented actin patches of one frame with the segmentation of the previous frame divided by the area of the union of both. Hence, highly persistent actin patches would lead to a Jaccard coefficient of 1.

Shape analysis

Dendritic cell bodies were segmented from fluorescent images (LifeAct-eGFP) by thresholding and conversion into binary images using FIJI (Schindelin et al., 2012). Polygonal outlines were extracted from segmentation masks and sampled at evenly spaced 200 points. To ensure that all polygons were orientated equally, an algorithm based on Procrustes analysis was used to rotate and translate the polygons until corresponding points were optimally aligned. We used the nucleus as a characteristic landmark defining the cell body of a migrating dendritic cell. Nuclei were extracted from fluorescent images (Hoechst) in FIJI and converted to another set of binary images. This additional landmark was used in the Procrustes procedure and improved the alignment. Finally, the alignment was manually verified. The following cellular characteristics were measured (1) aspect ratio = long axis/short axis, (2) the average diameter of the contour along the central axis, (3) length of the central axis and (4) normalized polygon curvature at the cell front was measured as an approximation of leading-edge roughness (The average of the absolute values of the point wise curvature of the contour is computed over a specified range, and multiplied by the contour length over the same range. Absolute values must be used because otherwise positive and negative curvatures would cancel out; the sum is multiplied by the arc length to make the measurement scale-invariant.) All algorithms are implemented in the ’’Celltool’’ software package (Pincus and Theriot, 2007).

Cross-correlation analysis and bootstrapping statistics

To compare the temporal dynamics of two parameters (speed, aspect ratio, fluorescence of actin patches, nuclear area) and test for the statistical significance of the temporal offset we used cross-correlation analysis with a custom-written MATLAB script (MATLAB; R2020a) (Mueller et al., 2017). Bootstrapping statistics was performed as described in (Tsai et al., 2019). Bootstrapping was used to obtain the 95% confidence interval of each cross-correlation value. Time traces of x and y were randomly permuted, and the same cross-correlation analysis to obtain the maximal correlation with all possible temporal offsets was performed. This process was then repeated 2000 times to obtain a distribution of maximal correlation value. The 95% confidence interval indicated a correlation value better than 95% of the maximal correlation values one can obtain with a pair of randomly permuted x and y.

Particle Image Velocimetry of collagen matrices

The displacement vectors of collagen fibers deformed by migrating DCs were calculated with the software Davis 8 (LaVision) applying Particle Image Velocimetry (PIV) on spinning-disc confocal images. Further post-processing was carried out using a custom-written Python script for extracting the maximum of deformation from frame-to-frame. Considered vectors were limited to the vicinity of the cell boundary corresponding to twice the cell area, which initially was determined by the LifeAct-eGFP expression.

Quantification and statistical analysis

All of the statistical details of experiments can be found in the figure legends, including the statistical tests used, exact value of n, what n represents, definition of center, and dispersion and precision measures. Appropriate control experiments were performed for each biological replicate. All replicates were validated independently and pooled only when showing the same trend. Statistical analysis was conducted using Prism7 (GraphPad). D’Agostino Pearson omnibus K2 test was used to test for Gaussian or non-Gaussian data distribution, respectively. When data were normal distributed t test was used. When data were non-normal distributed Kruskal-Wallis with Dunn’s test or two-tailed Mann-Whitney test was used. Statistical tests used for individual experiments are indicated in the figure legends.

REAGENT OR RESOURCE	SOURCE	IDENTIFIER
Antibodies
Monoclonal rabbit anti-Cav1 (clone D46G3)	Cell Signaling	#3267; RRID: AB_2275453
Monoclonal rabbit anti-CHC (clone D3C6)	Cell Signaling	#4796; RRID: AB_10828486
Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488	Invitrogen	#A11008; RRID: AB_143165
CD11c-APC (clone N418)	eBioscience	#17-0114-82; RRID: AB_469346
MHC II (I-A/I-E)-eFlour450 (clone M5/114.15.2)	eBioscience	#48-5321-82; RRID: AB_1272204
Chemicals, peptides, and recombinant proteins
Lipopolysaccharide (LPS) from Escherichia coli 0127:B8	Sigma	#L3129
CellTracker™ Orange CMTMR Dye	Invitrogen	#C2927
CellTrace™ CFSE Cell Proliferation Kit	Invitrogen	#C34554
5-(and-6)-Carboxytetramethylrhodamine, Succinimidyl Ester (5(6)-TAMRA	Invitrogen	#C6123
CellTrace™ Oregon Green® 488 carboxylic acid diacetate	Invitrogen	#C34555
Evans blue	Merck	#E2129-10G
Blasticidin	Invivogen	#ant-bl
Polybrene	Merck	#TR-1003-G
Ficoll-400	Fisher scientific	#BP525
Cytochalasin B	Tocris	#5474
NucBlue™ Live ReadyProbes™ Reagent (Hoechst 33342)	ThermoFisher	#R37605
CK666	Sigma	#SML0006
CK689	Merck Millipore	#182517
para-nitro-Blebbistatin	Axol	#ax494693
rmICAM-1/Fc	R&D Systems	#796-IC
Biozym Gold Agarose	Biozym	#850152
UltraPure Agarose	Invitrogen	#16500500
CCL19	Peprotech	#250-27B
Alexa Fluor™ 594 Phalloidin	Thermo Fisher	#A12381
PAcrAm™-g-(PMOXA)	SuSoS Surface Technology	N/A
SYLGARD™ 184 Silicone Elastomer Kit	Dow Corning	#1317318
PureCol®Type I Collagen	AdvancedBioMatrix	#5005
Nutragen®Type I Collagen	AdvancedBioMatrix	#5010
FibriCol®Type I Collagen	AdvancedBioMatrix	#5133
Alexa Fluor™ 594 NHS Ester (Succinimidyl Ester)	ThermoFisher	#A37572
Collagenase D	Roche	#11088858001
DNAse I	Roche	#10104159001
Critical commercial assays
EasySep Mouse T cell Isolation Kit	STEMCELL Technologies	#19851
Mouse T Cell Nucleofector™ Kit	Lonza	#VPA-1006
Experimental models: Cell lines
Lenti-X™ 293T Cell Line	Takara-Bio	# 632180
Experimental models: Organisms/strains
Mouse: C57BL/6JRj	Janvier	N/A
Mouse: B6.129S6-Wastm1Sbs/J	The Jackson Laboratory	#019458
Mouse: LifeAct-eGFP	(Riedl et al., 2010)	N/A
Mouse: HEM1-/-	(Leithner et al., 2016)	N/A
Recombinant DNA
Plasmid: eGFP-Abi1	(Lai et al., 2008)	N/A
Plasmid: eGFP-WASp	(Benesch et al., 2002)	N/A
Plasmid: Cip4-mCherry	(Taylor et al., 2011)	RRID: Addgene_27685
Plasmid: CLC-mCherry	(Taylor et al., 2011)	RRID: Addgene_27680
Plasmid: MLC-eGFP	(Croft et al., 2005)	N/A
Plasmid: LifeAct-mCherry	(Riedl et al., 2008)	N/A
Plasmid: pLenti6.3	Invitrogen	#V53306
Plasmid: pCMV-dR8.91	Creative Biogene	#OVT2971
Plasmid: pCMV-VSV-G	(Stewart et al., 2003)	RRID: Addgene_8454
Software and algorithms
SerialEM 3.x	(Mastronarde, 2003)	https://bio3d.colorado.edu/SerialEM/
IMOD 4.7	(Kremer et al., 1996)	http://bio3d.colorado.edu/imod
FIJI imaging processing software 1.53f	(Schindelin et al., 2012)	https://fiji.sc/
Imaris software v9	Oxford Instruments	https://imaris.oxinst.com/packages
TrackMate v6	(Tinevez et al., 2017)	https://imagej.net/TrackMate
Ilastik 1.3.2rc2	(Berg et al., 2019)	https://www.ilastik.org
Celltool	Pincus and Theriot, 2007	https://zplab.wustl.edu/celltool/
MATLAB R2020a	The MathWorks	https://ch.mathworks.com/products/matlab
Prism7	GraphPad	https://www.graphpad.com/scientific-software/prism/
Python 3.9.7	Python Software Foundation	https://www.python.org
TraXpert	Original code	https://github.com/SarenT/TraXpert/10.5281/zenodo.5721237
DaVis 8	LaVision	https://www.lavision.de/en/products/davis-software/imaris
Cross-correlation analysis (MATLAB script)	(Mueller et al., 2017)	N/A
Other
Culture-Inserts 2 Well	Ibidi	#80209
ACLAR® Fluoropolymer-Film	Science Services	#A50426
Cantilevers (spring constants: 0.1 N m−1; tip radius: 10 nm)	Nanosensors	#qp-BioAC