Article

Cellular locomotion using environmental

topography

https://doi.org/10.1038/s41586-020-2283-z Anne Reversat1,7 ✉, Florian Gaertner1, Jack Merrin1, Julian Stopp1, Saren Tasciyan1,
Juan Aguilera1, Ingrid de Vries1, Robert Hauschild1, Miroslav Hons1,2,8, Matthieu Piel3,4,
Received: 13 January 2018
Andrew Callan-Jones5, Raphael Voituriez6 & Michael Sixt1 ✉
Accepted: 9 March 2020

Published online: xx xx xxxx

Eukaryotic cells migrate by coupling the intracellular force of the actin cytoskeleton
Check for updates to the environment. While force coupling is usually mediated by transmembrane
adhesion receptors, especially those of the integrin family, amoeboid cells such as
leukocytes can migrate extremely fast despite very low adhesive forces1. Here we show
that leukocytes cannot only migrate under low adhesion but can also transmit forces
in the complete absence of transmembrane force coupling. When confined within
three-dimensional environments, they use the topographical features of the substrate
to propel themselves. Here the retrograde flow of the actin cytoskeleton follows the
texture of the substrate, creating retrograde shear forces that are sufficient to drive
the cell body forwards. Notably, adhesion-dependent and adhesion-independent
migration are not mutually exclusive, but rather are variants of the same principle of
coupling retrograde actin flow to the environment and thus can potentially operate
interchangeably and simultaneously. As adhesion-free migration is independent of
the chemical composition of the environment, it renders cells completely
autonomous in their locomotive behaviour.

Mechanistic understanding of cell motility started with Abercrom- collagen gels (Supplementary Video 1). To eliminate integrin-based
bie’s demonstration that particles placed on the dorsal surface of a cell force transmission, we deleted Tln1, which encodes talin 1, a cytosolic
undergo retrograde transport2. The authors concluded that cellular adaptor protein essential for integrin functionality10 (Extended Data
material is added to the front of the cell, feeding a rearwards flow, which Fig. 1a–c). Talin-deficient (referred to here as ‘talin KO’) cells were
couples to the substrate and thereby generates friction to drag the cell completely unable to adhere to and migrate on 2D surfaces (Fig. 1a,
body forwards. Later studies showed that the added material is the b, Extended Data Fig. 1d, e). However, once incorporated into 3D col-
growing polymer of F-actin and that F-actin retrograde flow couples lagen gels, their migratory characteristics were indistinguishable from
via transmembrane adhesion receptors, mainly of the integrin fam- wild-type cells (Fig. 1c, Extended Data Fig. 1f, Supplementary Video 1).
ily. Although this adhesion-dependent principle of motility turned In the following, we engineered artificial environments to dissect the
out to be universal on two-dimensional (2D) surfaces, some cells can geometrical transition between 2D surfaces and fibrillar environments,
migrate in the absence of integrin-mediated adhesion when confined while keeping the chemical composition of the substrate unchanged.
within a three-dimensional (3D) context3,4. This applies especially to We first designed a microfluidic setup to confine leukocytes between
the amoeboid (shape-changing) class of cells such as leukocytes5. two parallel surfaces separated by an adjustable space11 (‘2.5D’; Fig. 1d).
Force transmission under these conditions is not understood, and While this setup supported efficient migration of wild-type cells,
alternative membrane-spanning molecules that mediate specific or talin-KO cells were unable to translocate, although they showed mor-
even unspecific friction have been suggested6,7. Notably, in his seminal phological polarization (Fig. 1d, Extended Data Fig. 1g, Supplementary
study, Abercrombie contemplated a second, qualitatively different, Video 2). When T cells expressing the actin reporter Lifeact-GFP were
scenario: backward-moving waves of deformation might passively imaged with total internal reflection fluorescence (TIRF) microscopy,
entangle with the applied particles (or a substrate) and propel the cell actin within the rapidly moving cell body remained static in relation to
in a manner analogous to paddling and swimming2. the adhesive substrate and flowed backwards in the cell frame of refer-
To explore adhesion-independent locomotion, we initially inves- ence (Fig. 1e–g, Supplementary Video 3). Talin-KO cells on adhesive
tigated T lymphocytes. When these cells migrate within lymphatic substrates or wild-type cells on passivated surfaces were immobile but
organs, integrin depletion causes them to slow down by only 15%, mean- showed similar actin dynamics in the cell frame of reference8, mean-
ing that they do employ integrins but retain a substantial capacity to ing that actin slid backwards in relation to the substrate (Fig. 1e–g,
transmit forces via other means8,9. We used a mouse T cell line, which Extended Data Figs. 1g–l, 2a–c, Supplementary Video 3). These findings
migrates vigorously in 3D environments, such as in vitro-assembled demonstrate that, when placed on 2D substrates or when confined
1
Institute of Science and Technology Austria (IST Austria), Klosterneuburg, Austria. 2Institute of Scientific Instruments of the Czech Academy of Sciences, Brno, Czech Republic. 3Institut Curie,
PSL Research University, CNRS, UMR 144, Paris, France. 4Institut Pierre-Gilles de Gennes, PSL Research University, Paris, France. 5Laboratoire Matière et Systèmes Complexes, UMR 7057 CNRS,
Université Paris Diderot, Paris, France. 6Laboratoire de Physique Theorique de la Matière Condensée et Laboratoire Jean Perrin, CNRS/Université Pierre-et-Marie Curie, Paris, France. 7Present
address: Institute of Translational Medicine, Department of Cellular and Molecular Physiology, University of Liverpool, Liverpool, UK. 8Present address: BIOCEV, First Faculty of Medicine,
Charles University, Vestec, Czech Republic. ✉e-mail: [email protected]; [email protected]

Nature | www.nature.com | 1
Article
a b c d a b 3 μm 4 μm 5 μm 6 μm
2D 2D 3D 2.5D

–100 0 100 –100 0 100

Talin KO Control
5 μm
2.5D + pillars

5 μm
Spreading area per cell (μm2)

5 μm
5 μm

y (μm)
Fc-ICAM1 3D view Side view
Side view
Adhesive cells (N mm–2)

*** 250 **** 20,000 Control 50,000 Control

50 3 μm
200

MSD (μm2)

MSD (μm2)
15,000 Talin KO 40,000 Talin KO Pillar spacing 4 μm
40 3D view 5 μm
150 30,000 6 μm
30 10,000
100 20,000 2.5D (no pillar) ∞ –100 0 100 –100 0 100 –100 0 100 –100 0 100
20
5,000 10,000
10 50
c x (μm)
d

Cell speed (μm min–1)

0 0

Displacement (μm)
0 0 0 20 40 60 0 20 40 60
20 20 ** 500 500 **
Ta ol
KO

l
KO ****
tro

Time (min) Time (min)

tr

400 400
on

on

15 15
lin

lin
C

Ta

300 300
e Control Talin KO
f Lifeact-GFP
g 10 10
200 200
Lifeact-GFP Lifeact-GFP 20
**** 5 5 100 100
****

Velocity (μm min–1)

Control
00:00 00:00 0 0 0 0
02:00 02:00 15 ∞ ∞ ∞ ∞

μm
μm
μm
μm

μm
μm
μm
μm

μm
μm
μm
μm

μm
μm
μm
μm
3
4
5
6

3
4
5
6

3
4
5
6

3
4
5
6
10
Control Talin KO Control Talin KO
5 e f
Talin KO

5 μm
0 15 ****

Cell velocity
Cell Actin Cell Actin

(μm min–1)
Control
Control Talin KO 10
44:00
5
Fig. 1 | Adhesive and migratory properties of talin-KO T cells. a, The number Talin KO
0
of cells adhering to the 2D surface. Data are mean ± s.e.m. of four independent 60:00

lin l
KO
Ta tro
experiments; ***P = 0.00076, paired t-test. b, Surface area (μm2) of cells

on
C
spreading on Fc-ICAM1-coated glass. Control (n = 139) and talin-KO (n = 81) cells g Serrated Smooth h NS
25 ***

5 μm

Cell velocity
of three independent experiments are shown. Data are mean ± s.d.;

(μm min–1)
20
15
****P < 0.0001, Mann–Whitney U-test. c, Mean square displacement (MSD) of Control
6 μm
10
control (n = 51, in black) and talin-KO (n = 52, in red) cells migrating in 3D 60:00 5
0
collagen gel from three independent experiments. Data are mean ± s.e.m.; Talin KO

o d
Se oth

oo d
th
Smate

Sm te
60:00

rra
P = 0.2243, Mann–Whitney U-test. d, MSD of control (n = 290, in black) and

rr
Se
talin-KO (n = 284, in red) cells migrating in a 5-μm confiner from four Serrated Smooth Control Talin KO
i 2.5D channel channel j
independent experiments. Data are mean ± s.e.m.; P < 0.0001, Mann–Whitney
NS

EDTA (%)
U-test. e, Snapshots of TIRF microscopy of a control cell (left) and a talin-KO cell 3D view
200 **** *** 2.5D
(right) expressing Lifeact-GFP under 5-μm confinement. Representative of Side view 150 Serrated channel
three independent experiments. The green arrowheads mark the uropod and 20 **** 20 * 10 ****

before
after
Smooth channel
Cell speed
(μm min–1)

15 15 100
the orange arrowheads indicate the cell front. TIRF images in cyan are at 10 10 5 50

Cell speed
t = 0 min and in red at t = 2 min. The yellow dashed line is used for the kymograph 5 5
0
0 0 0
in f. Scale bars, 5 μm. Time is shown in min:s. f, Kymographs of a control cell and EDTA – + – + – +
talin-KO cell under 5-μm confinement as shown in e. Cell velocity is indicated by
the green dashed line, and actin retrograde flow is shown by the cyan dashed Fig. 2 | Migration of T cells in textured confinement. a, Experimental setup
line. Horizontal scale bar, 5 μm; vertical scale bar, 1 min. g, Retrograde flow for b–d. b, Cell trajectories analysed in c and d during 1 h. The black crosses
velocities for control (n = 42) and talin-KO (n = 36) cells from three independent indicate the start of the individual tracks. c, d, Cell speed (c), mean ± s.d.;
experiments. ****P < 0.0001, Mann–Whitney U-test. **P = 0.0021, otherwise not significant (P > 0.9999); and displacement (d),
mean ± s.d.; **P = 0.0089 and ****P < 0.0001, otherwise not significant, Kruskal–
Wallis test followed by post hoc Dunn’s test. n = 25, 20, 75 and 61 for control cells
in respective 3-μm, 4-μm, 5-μm and 6-μm zones. Data are from three
between two surfaces, T cell locomotion depends on integrin-mediated independent experiments. n = 64, 115, 124 and 94 for talin-KO cells in respective
force transmission6,8,9,12 (see also Supplementary Discussion). This find- 3-μm, 4-μm, 5-μm and 6-μm zones. Data are from six independent experiments.
ing was in contrast to our collagen gel data and to the in vivo finding e, f, Migration in smooth channels. Data are from four independent
that T cells retain substantial migratory capacity following integrin experiments. In e, 5 × 5-μm smooth channels are shown. Top, scheme of the
depletion8,9. We reasoned that the major difference between confine- channel. Representative snapshots at t = 44 min of Lifeact-GFP-expressing
control cells (middle) and at t = 60 min talin-KO cells (bottom) are displayed.
ment within a lymph node or within a fibrillar matrix and confinement
Tracks are shown in yellow. Scale bars, 20 μm. In f, the cell velocity is shown.
between two planar surfaces is geometrical complexity. Hence, we
n = 99 for control cells, n = 24 for talin-KO cells. Data are mean ± s.d.;
introduced arrays of variably spaced pillars that intersected the two
****P < 0.0001, Mann–Whitney U-test. g, h, Migration in serrated channels. Data
surfaces (Fig. 2a), thereby mimicking the geometry of a fibrillar gel are from three (control) and four (KO) independent experiments. In g, the
while keeping the surface chemistry unchanged13. Between the pil- serrated topography with a 6-μm period in a 5 × 5-μm channel is shown. Top,
lars, the migratory capacity of talin-KO cells was restored (Fig. 2b–d, scheme of the channel. A representative snapshot at t = 60 min of the
Extended Data Fig. 2d, e, Supplementary Video 4). While translocation Lifeact-GFP control cell (middle) and the talin-KO cell (bottom) is displayed.
velocities of wild-type cells were not affected by increased pillar spac- Tracks are shown in yellow. Scale bars, 20 µm. In h, the cell velocity in channels
ings, the performance of talin-KO cells dropped in wider-spaced pillars, is shown. n = 59 (control serrated), 25 (control smooth), 8 (talin-KO serrated)
suggesting that tight contact with curved surfaces enables propulsion and 5 (talin-KO smooth). Data are mean ± s.d.; ***P = 0.0009, Kruskal–Wallis test
in the absence of integrins (Supplementary Video 4). We next turned to followed by post hoc Dunn’s test. i, j, Cell speed before and after the addition of
channels in which cells are fully enclosed by surfaces (Fig. 2e). Within 10 mM EDTA. Data are from three independent experiments. 2.5D: n = 73 cells;
smooth-walled 5 × 5-μm channels, control cells migrated persistently, serrated channel: n = 23; and smooth channel: n = 27. In i, the cell speed before
(−) and after (+) the addition of EDTA in respective devices is shown.
while talin-KO cells were immobile (Fig. 2e, f, Supplementary Video 5).
****P < 0.0001 and *P = 0.0167, Wilcoxon matched-pairs signed-rank test.
We next introduced surface texture as a serrated pattern with a 6-μm
In j, the change in cell speed after versus before the addition of EDTA is shown.
period and found that the migration of talin-KO cells was restored
Data are mean ± s.d.; not significant (NS; P > 0.9999), ***P = 0.0004 and
(Fig. 2g, h, Extended Data Fig. 2f, Supplementary Video 5). In channels
****P < 0.0001, Kruskal–Wallis test followed by post hoc Dunn’s tests.
that contained both smooth and serrated sectors, wild-type cells trans-
ited freely between the sectors. Following cation depletion with EDTA,
which inactivates integrins (and several other cell-surface adhesion a smooth area (Fig. 2i, j, Supplementary Video 6). The same principle
molecules), cells within smooth areas slowed down or stalled, while was conserved when we tested dendritic cells and neutrophil granulo-
cells within serrated sectors kept migrating until they encountered cytes, two other types of amoeboid leukocytes. Notably, EDTA-treated

2 | Nature | www.nature.com
 a Confining b
wall O ∞

w(x)
U=0
v0

y
F-actin flow x
L

c
O<L
Force h+(x)
Cell cortex h–(x) U>0
v0
y
O = 2π/k x
Cytoplasm

Fig. 3 | Physical model of adhesion-free cell migration in a complex arrows) but cannot move forwards. c, Scheme of the modelled non-adhesive
environment. a, Scheme of the model: the channel has a varying wavy cell in a serrated channel. Cell propulsion is created by the shear stress in the
cross-section w(x) and period λ = 2π/k. The cell contains cortical F-actin that actin cortex, which is caused by the bending of flow lines (black arrows) by the
presents a retrograde flow of length L, an average speed v0 and is modelled as a environmental topography. This induces a pressure gradient in the cortical
viscous fluid layer of thickness h(x). b, Scheme of a modelled cell in a smooth F-actin mesh and thus non-vanishing normal forces onto the substrate. U, cell
channel with an adhesion-free coupling of the actin cytoskeleton with the speed.
environment. The cell is polarized and has a high retrograde actin flow (black

dendritic cells retained a weak locomotive capacity even in smooth of talin-KO and EDTA-treated cells in textured channels and found an
channels (Extended Data Fig. 3a, b, Supplementary Video 7). This is in increase in the actin retrograde flow relative to the substrate when-
line with previous findings that, compared with T cells, dendritic cells ever the cells slowed between the topographical features (Fig. 4e–g,
have a higher capacity to transmit forces via unspecific transmem- Extended Data Fig. 5g, Supplementary Video 9). This suggests that, in
brane coupling8,12,14. These data demonstrate that force transmission analogy with the molecular clutch formed by transmembrane recep-
of leukocytes follows a continuum of strategies. They can flexibly use tors, topographical features allow the coupling of retrograde actin flow
integrin-mediated adhesion, unspecific transmembrane force coupling to a substrate. Finally, we devised an entirely orthogonal approach,
and, if both strategies fail due to the absence of suitable ligands, can in which we placed primary mouse T lymphocytes on surfaces that
efficiently utilize surface topography to transmit forces. were either passivated or coated with the integrin ligand intracellular
As a framework to understand topography-based locomotion, we adhesion molecule 1 (ICAM1) and overlaid them with a pad of agarose.
propose a minimal model (Fig. 3a–c, Supplementary Text) in which As shown previously8, cells migrated on ICAM1, but slipped and did
actin is described as a viscous gel that travels from the leading to the not migrate on passivated substrate (Extended Data Fig. 6a–g). We
trailing edge of the cell in the absence of any tangential friction force then used passivated surfaces with engraved linear ridges to introduce
with the substrate. In serrated but not in smooth channels, this flow anisotropic topography. We found that on these surfaces, migratory
generates local shear stresses and thus a pressure gradient in the actin capacity was rescued and that cells predominantly migrated perpen-
gel, which drives locomotion by imposing non-vanishing normal forces dicular but not parallel to the ridges (Extended Data Fig. 6a–g, Sup-
onto the substrate. Here the actin flow generated in the cell can couple plementary Video 10). Within the same cell, actin showed retrograde
to the environment when its topographical features are smaller than flow in relation to the substrate when the flow was aligned with the
the flow scale of the actin cortex. To test this prediction, we varied the ridges, but not in orientation perpendicular to the ridges (Extended
period of serrated topography from 6 to 12 to 24 μm. Serrated sectors Data Fig. 7, Supplementary Video 10). Together, these data demonstrate
were followed by smooth areas as an internal control (Fig. 4a). Wild-type that the cells slipped in an orientation along the ridges and coupled in
T cells effectively traversed all (including smooth) channel designs, an orientation perpendicular to the ridges.
while talin-KO cells migrated in 6-μm and 12-μm period patterns but In our experiments, we demonstrated topography-based force trans-
rarely in the 24-μm patterns (Fig. 4b–d, Extended Data Fig. 4a), in which mission by controlling the topography of the environment, meaning
the spacing exceeded the average length scale of cortical actin flow in a that we extrinsically imposed shape changes of the cell. To directly link
cell (Extended Data Fig. 4b). In channels with a successively increased this to the amoeboid (shape-changing) principle, we tested whether
serration period, continuous single-cell observation showed that, in cells can autonomously generate appropriate deformations. Morpho-
qualitative agreement with model predictions, cells slowed from 6 to metric analysis of T cells showed that they have the intrinsic capacity
12 μm, presented even slower and saltatory movement in the 24-μm to produce rearwards-propagating deformation waves (Extended Data
stretches and were unable to enter the smooth channel (Fig. 4c, d, Fig. 8a–e and previously published studies18–23) and that the travelling
Supplementary Video 8). The model also predicts a linear depend- speed of these waves scaled with the speed of actin flow. Such deforma-
ence of the propulsion force on actin flow speed. Actin flow has two tion waves could intercalate with any textured environment and propel
mechanical components: polymerization pushes filaments from the a cell as conjectured by Abercrombie2.
front to the back, while actomyosin contraction pulls at the back12,15–17. Altogether, our findings demonstrate that cells can transmit forces
Accordingly, inhibition of myosin II slowed down but did not stall cells by coupling the retrograde flow of actin to a geometrically irregular
under adhesive conditions (Extended Data Fig. 5a–f). Following dual environment, and that this can happen in the complete absence of any
inhibition of adhesion and contractility, cells still performed better in transmembrane receptors that link the cytoskeleton to the substrate.
the serrated than in the smooth channels (Extended Data Fig. 5d–f), Notably, force transmission by specific adhesion receptors, by unspecific
suggesting a direct relation between actin flow speed and locomo- friction and by shape change are not mutually exclusive mechanisms.
tive force, as predicted. Next, we quantitatively imaged the actin flow Rather, they are variants of the same fundamental principle of coupling

Nature | www.nature.com | 3
Article
a b Migration c d NS NS
* **
Serrated Smooth No migration Control * **
30 30 ****

Talin KO cells (%)

6 μm 12 μm 24 μm Channel
100

Cell velocity
zone

(μm min–1)
80 20 20
60
Talin KO 10
40 10
20
0 Cell velocity (a.u.) 0 0
5 μm 6–12 24 μm to 0 1

μm

μm

μm

th

μm

μm

th
μm
oo

oo
μm smooth

12

24

12

24
6
Sm

Sm
Control Talin KO
e 6 μm 12 μm 24 μm Smooth f g
Channel zone 6 μm Channel zone 24 μm NS
Time Bright field Bright field *
25
(min:s) 00:00 07:30 15:00 22:30 30:00 **
20

(μm min–1)
Lifeact-GFP Lifeact-GFP

Velocity
00:00 15
10 Cell
Channel zone 12 μm Channel zone smooth
Actin flow
Bright field Bright field 5
0
Lifeact-GFP Lifeact-GFP Channel 6 μm 12 μm 24 μm Smooth
Bright field zone
31:20 Lifeact-GFP TIRF Talin KO

Fig. 4 | Force transmission of T cells in varying geometries. a, The channel the 6-μm zone to the smooth zone. From top to bottom: scheme of the channel,
design that is used in b–g. b, The proportion of talin-KO cells migrating in 6-μm colour-coded snapshots every 7.5 min and a kymograph of Lifeact-GFP
and 12-μm zones (n = 41) and in the 24-μm zone and the smooth zone (n = 22). (TIRF is shown in cyan and bright field is shown in grey). Scale bars, 20 μm.
****P < 0.0001, Fisher’s exact test. c, Snapshots of control cells (top) and Representative of three independent experiments. f, Bright-field imaging
talin-KO cells in channels with a varying period. Lifeact-GFP is in red, and of channel sectors and the corresponding kymograph of the Lifeact-GFP-
Hoechst is in cyan. Scale bars, 20 μm. Bottom, tracks are colour-coded for cell expressing talin-KO cell shown in e during 50 s. The red dashed lines indicate
velocity. Data are representative of four and eight independent experiments, cell displacement, and the cyan dashed lines indicate actin retrograde flow.
respectively. d, The velocity of control (n = 88) and talin-KO (n = 79) cells in Scale bar, 20 μm. g, Actin retrograde flow and cell velocities of talin-KO cells
zones with different periodicity. Data are representative of four and eight in different channel zones. Data are representative of four independent
independent experiments, respectively. Control: *P = 0.0234; KO: *P = 0.0188, experiments. n = 7. *P = 0.0423 and **P = 0.0036, Kruskal–Wallis test followed
**P = 0.0013 (6 µm versus 24 µm) and **P = 0.001 (24 µm versus smooth), by post hoc Dunn’s test. Boxes in d and g extend from the 25th to the 75th
**P < 0.0001, otherwise NS, Kruskal–Wallis test followed by post hoc Dunn’s percentile, with the middle line showing the median and the whiskers
test. e, TIRF microscopy of a Lifeact-GFP-expressing talin-KO cell moving from indicating the minimum to maximum values.

retrograde actin flow to the surrounding and can therefore operate 10. Calderwood, D. A., Campbell, I. D. & Critchley, D. R. Talins and kindlins: partners in
integrin-mediated adhesion. Nat. Rev. Mol. Cell Biol. 14, 503–517 (2013).
alternately or even simultaneously. This endows amoeboid cells with 11. Le Berre, M., Aubertin, J. & Piel, M. Fine control of nuclear confinement identifies a
enormous adaptability when migrating in heterogeneous environments. threshold deformation leading to lamina rupture and induction of specific genes. Integr.
Biol. (Camb.) 4, 1406–1414 (2012).
12. Renkawitz, J. et al. Adaptive force transmission in amoeboid cell migration. Nat. Cell Biol.
11, 1438–1443 (2009).
Online content 13. Renkawitz, J., Reversat, A., Leithner, A., Merrin, J. & Sixt, M. Micro-engineered “pillar
Any methods, additional references, Nature Research reporting sum- forests” to study cell migration in complex but controlled 3D environments. Methods Cell
Biol. 147, 79–91 (2018).
maries, source data, extended data, supplementary information, 14. Vargas, P., Barbier, L., Sáez, P. J. & Piel, M. Mechanisms for fast cell migration in complex
acknowledgements, peer review information; details of author con- environments. Curr. Opin. Cell Biol. 48, 72–78 (2017).
tributions and competing interests; and statements of data and code 15. Medeiros, N. A., Burnette, D. T. & Forscher, P. Myosin II functions in actin-bundle turnover
in neuronal growth cones. Nat. Cell Biol. 8, 215–226 (2006).
availability are available at https://doi.org/10.1038/s41586-020-2283-z. 16. Henson, J. H. et al. Two components of actin-based retrograde flow in sea urchin
coelomocytes. Mol. Biol. Cell 10, 4075–4090 (1999).
17. Cramer, L. P. Molecular mechanism of actin-dependent retrograde flow in lamellipodia of
1. Lämmermann, T. & Sixt, M. Mechanical modes of ‘amoeboid’ cell migration. Curr. Opin. motile cells. Front. Biosci. 2, d260–d270 (1997).
Cell Biol. 21, 636–644 (2009). 18. Driscoll, M. K. et al. Cell shape dynamics: from waves to migration. PLoS Comput. Biol. 8,
2. Abercrombie, M., Heaysman, J. E. & Pegrum, S. M. The locomotion of fibroblasts in e1002392 (2012).
culture. 3. Movements of particles on the dorsal surface of the leading lamella. Exp. Cell 19. Howe, J. D., Barry, N. P. & Bretscher, M. S. How do amoebae swim and crawl? PLoS One 8,
Res. 62, 389–398 (1970). e74382 (2013).
3. Liu, Y.-J. J. et al. Confinement and low adhesion induce fast amoeboid migration of slow 20. Bae, A. J. & Bodenschatz, E. On the swimming of Dictyostelium amoebae. Proc. Natl Acad.
mesenchymal cells. Cell 160, 659–672 (2015). Sci. USA 107, E165–E166 (2010).
4. Friedl, P. & Wolf, K. Plasticity of cell migration: a multiscale tuning model. J. Cell Biol. 188, 21. Mandeville, J. T. H., Lawson, M. A. & Maxfield, F. R. Dynamic imaging of neutrophil
11–19 (2010). migration in three dimensions: mechanical interactions between cells and matrix.
5. Lämmermann, T. et al. Rapid leukocyte migration by integrin-independent flowing and J. Leukoc. Biol. 61, 188–200 (1997).
squeezing. Nature 453, 51–55 (2008). 22. Tozluoğlu, M. et al. Matrix geometry determines optimal cancer cell migration strategy
6. Bergert, M. et al. Force transmission during adhesion-independent migration. Nat. Cell and modulates response to interventions. Nat. Cell Biol. 15, 751–762 (2013).
Biol. 17, 524–529 (2015). 23. Jankowiak, G., Peurichard, D., Reversat, A., Schmeiser, C. & Sixt, M. Modelling
7. Schmidt, S. & Friedl, P. Interstitial cell migration: integrin-dependent and alternative adhesion-independent cell migration. Math. Model. Methods Appl. Sci. 30, 513–537
adhesion mechanisms. Cell Tissue Res. 339, 83–92 (2010). (2020).
8. Hons, M. et al. Chemokines and integrins independently tune actin flow and substrate
friction during intranodal migration of T cells. Nat. Immunol. 19, 606–616 (2018). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
9. Woolf, E. et al. Lymph node chemokines promote sustained T lymphocyte motility published maps and institutional affiliations.
without triggering stable integrin adhesiveness in the absence of shear forces.
Nat. Immunol. 8, 1076–1085 (2007). © The Author(s), under exclusive licence to Springer Nature Limited 2020

4 | Nature | www.nature.com
Methods insertion using 5 μg/ml puromycin (Gibco, Thermo Fisher Scientific) for
7 days and were subjected to western blotting. Cells were then cloned
Cell culture using fluorescence-activated cell sorting (FACS Aria III, BD Biosciences)
Cells were grown and maintained in a humidified incubator at 37 °C and and expanded in R10 medium containing puromycin (5 μg/ml).
5% CO2. LMR7.5 T cell hybridomas, a gift from A. M. Lennon-Duménil
(Institut Curie, Paris, France), were cultivated in R10 medium (RPMI T cell infection and selection. For infection of LMR7.5 T cells with
1640 supplemented with 10% FBS, 2 mM l-glutamine, 100 units/ml the Lifeact-mCherry reporter construct, the same infection protocol
penicillin, 100 μg/ml streptomycin and 50 μM 2-mercaptoethanol; was applied except cells were selected with 5 μg/ml blasticidin (Gibco,
Gibco, Thermo Fisher Scientific). The Lifeact-GFP24 parental T cell line Thermo Fisher Scientific) and sorted for fluorescence.
was generated by nucleofection with an eGFP reporter construct (Kit V,
Lonza). Alternatively, LMR T cells were infected with a plasmid encod- Western blot
ing Lifeact-mCherry (pLenti6.3, Invitrogen, Thermo Fisher Scientific). For immunodetection of talin and GAPDH proteins, 1 × 106 control and
LX-293 HEK cells (Clontech) were used for lentivirus production and talin-KO cells were centrifuged and lysed in RIPA buffer (New England
maintained in D10 medium (DMEM with 2 mM l-glutamine, 10% FBS, Biolabs) and the supernatant was denatured in 2× Laemmli buffer
and 100 units/ml penicillin and 100 μg/ml streptomycin; Gibco). All cell (161-0737, Bio-Rad,) at 95 °C for 5 min. Protein lysates were loaded on
lines were mycoplasma-free and frequently tested. Neutrophil-like PLB- a 10% Tris-glycine gel (Invitrogen, Thermo Fisher Scientific), sepa-
985 cells were obtained from the DSMZ (ACC 139). In brief, cells were rated by SDS–PAGE and further transferred by electrophoresis to a
maintained in RPMI 1640, supplemented with 10% FBS, 20 mM HEPES polyvinylidene difluoride membrane (iBlot, Invitrogen, Thermo Fisher
and 1% glutamine (all Gibco, Thermo Fisher Scientific). For differentia- Scientific) according to the manufacturer’s instructions, with a 9-min
tion, PLB cells were kept for 6 days in medium containing 1.25% DMSO transfer time. Membranes were blocked in PBST-BSA 5% for 1 h and incu-
(cell culture grade; Sigma-Aldrich). Differentiation status was validated bated with primary antibodies overnight at 4 °C. Antibodies to talin
using flow cytometry (CD11b, Miltenyi Biotec). Before experimenta- (mouse anti-pan-talin, clone 8d4; T3287, Sigma-Aldrich) and GAPDH
tion, differentiated cells were washed three times to remove DMSO. (mouse, clone GA1R; ab125247, Abcam) were diluted 1:200 and 1:3,000 in
Dendritic cells were differentiated from bone marrow that originated PBST-BSA 1%, respectively. After washing with TBST, secondary antibody
from 6–12-week-old male or female C57BL/6J mice25. Primary mouse was applied for 1 h in PBST-BSA 5% at room temperature (goat anti-mouse
T cells were obtained from peripheral lymph nodes and spleens8 and IgG HRP conjugate, diluted 1:5,000; 170-6516, Bio-Rad). Protein detec-
homogenized with a 70-μm cell strainer. Untouched primary naive tion was performed by enhanced chemiluminescence (ECL; Thermo
T cells were isolated from 6–12-week-old male C57BL/6J Lifeact-GFP Fisher Scientific) detection using a VersaDoc imaging system (Bio-Rad).
mice with an EasySep Mouse T cell Isolation Kit according to the manu-
facturer’s protocol (19851 A, STEMCELL Technologies) and incubated Adhesion assay
overnight at 37 °C in 5% CO2 in R10 medium. Control and talin-KO T cells (5 × 105) were plated in pre-warmed R10
medium in a 25-cm2 culture flask (TPP, Sigma-Aldrich) and placed in
Animals a humidified incubator at 37 °C, 5% CO2 for 24 h before observation.
Mice were bred and maintained at the local animal facility in accordance Cells were imaged with an inverted microscope at low magnification
with the IST Austria ethics commission and the Austrian law. Permis- (DM IL Led and DFC450 digital camera, Leica Microsystems) and cells
sion was granted by the Austrian Federal Ministry of Science, Research adhering to the bottom of the dish were counted manually.
and Economy (identification code: BMWF-66.018/0005-II/3b/2012).
Spreading assay
Reporter and CRISPR–Cas9 cell line generation The 2D spreading assay was performed on glass coverslips
CRISPR–Cas9 design and production. Single-guide RNAs (sgRNAs) (Menzel-Glaser no. 1, VWR) that were previously coated with 2 μg/ml
were designed as described elsewhere26. In brief, sgRNAs were designed of recombinant mouse ICAM-1/human Fc chimaera (796-IC, R&D
to induce the double-strand break in the first exons of the mouse gene Systems) overnight at 4 °C, and washed three times with PBS. Con-
encoding talin 1 (Tln1). Then sgRNAs were scored for high on-target trol (Lifeact-mCherry or Lifeact-GFP) and talin-KO (Lifeact-GFP or
effects (https://portals.broadinstitute.org/gpp/public/analysis-tools/ Lifeact-mCherry) T cells (0.2 × 106) were gently mixed in 200 μl R10
sgrna-design) and low off-target effects (http://crispr.mit.edu). sgRNAs medium and allowed to settle on the coated dish for 60 min at 37 °C
were cloned into the LentiCRISPRv1 vector (a gift from F. Zhang, Ad- and 4.5% CO2. Cell spreading was then imaged by simultaneous TIRF
dgene plasmid no. 49535). Guide sequences: sgRNA-Scramble s(5′–3′) and epifluorescence microscopy.
GCCGTGGCGCATGGGTAGCA; sgRNA-Talin1g1 s(5′–3′) ATAATGCCCTAC
GAGCCGT (off-target score of 91); sgRNA-Talin1g2 s(5′–3′) CTCACT Photolithography
GTTTCCCCGGGTA (off-target score of 82); sgRNA-Talin1g3 s(5′–3′) Photolithography was performed as previously described13,25,27,28 with
GTCGAGGCTGGGCGACTGG (off-target score of 98). the following adaptations.

Lentivirus production. Lentivirus production was performed as Photomasks. Patterns were designed with Coreldraw X8 (Corel Cor-
described previously25. In brief, LX-293 HEK cells (Clontech) were poration) exported to DXF, then converted to GERBER format with
co-transfected with LentiCRISPRv1 or LentiCRISPRv2, packaging- ps- LinkCad. Chrome photomasks (100 mm or 125 mm; PhotoData/JD
PAX2 (Addgene no. 12260) and pCMV-VSV-G envelope plasmids using Photo-Tools) were used.
Lipofectamine 2000 (Thermo Fisher Scientific) as recommended by the
manufacturer, and cells were resuspended in R10 medium 1 day before Confiners. The 3-μm confiner masters were produced by spin-coating
transfection. Supernatant was collected after 72 h and stored at −80 °C. SU8-GM1050 (Gersteltec) for 40 s at 3,000 rpm, pre-baking for 1 min at
120 °C, exposing to 35 mJ/cm2 of UV, post-exposure baking for 5 min at
CRISPR-based talin KO. Lifeact-eGFP LMR7.5 T cells were spin-infected 95 °C, developing in SU8 developer, and then hard-baking at 150 °C for
at 1,500g for 1 h in the presence of the lentivirus-containing supernatant 5 min. The 5-μm confiner master was produced by spin-coating SU8-
and 6 μg/ml Polybrene (Sigma-Aldrich). Spin infection was carried out in 2005 (Microchem) for 30 s at 3,000 rpm, pre-baking for 2 min at 95 °C,
12-well plates with 3 × 105 cells/ml per reaction plus 500 μl of undiluted exposing to 105 mJ/cm2 of UV, post-exposure baking for 3 min at 95 °C,
virus. Three days after infection, cells were selected for stable virus developing in SU8 developer, and then hard-baking at 150 °C for 5 min.
Article
layer in a tight aluminium mould and further degassed in a vacuum
EDTA device. A 4-μm high master was produced by spin-coating desiccator to pump out the remaining bubbles. Both layers were baked
SU8-2005 for 30 s at 5,000 rpm, pre-baking for 2 min at 95 °C and at 80 °C for 45 min, and the flow layer was removed from the mould,
UV exposure of 550 mJ/cm2 through a PL-360-LP (Omega Optical) trimmed with a razor blade, and entry and exit holes were punched
optical filter on an EVG mask aligner 610 (EVG Group). The wafer was with a homemade arbor press. Debris from both the flow device and the
post-exposure baked for 3 min at 95 °C, developed in SU8 developer control layer on its wafer were carefully removed with Scotch tape, the
and then hard-baked at 150 °C for 5 min. The optical filter was necessary two layers were plasma-cleaned (2 min at medium intensity), aligned
for high-resolution features in SU8. manually and bonded together. After overnight baking in 80 °C, the
chip was removed from the wafer, the control layer entry holes were
Multilayer devices. First, the 25-μm high control master was made by punched, and the chip was bonded to the glass coverslip by oxygen
spin-coating SU8-3025 (Microchem) for 30 s at 3,000 rpm. The wafer plasma and kept at 80 °C until further use. Before the experiments, all
was soft-baked for 15 min at 95 °C, then exposed to 200 mJ/cm2 UV, entry ports were connected with metal pins (NE-13-1003, New England
then post-exposure baked at 95 °C for 5 min, then developed in SU8 Small Tube Corporation).
developer. Second, the fluid-layer master was first fully coated with
a submicron layer of GM1040-SU8 by spin-coating at 5,000 rpm for Migration assays
2 min. The wafer was then baked at 95 °C for 5 min, exposed to 100 mJ/cm2 Before experiments, 2 × 105 T cells in 2 ml R10 medium were stained
UV, post-exposure baked at 95 °C for 30 min, and then developed with Hoechst 33342 for 30 min (1 drop, NucBlue, R37605, Invitrogen,
in SU8 developer. The 5-μm fluid-layer features were made on top Thermo Fisher Scientific) when nucleus visualization was needed for
of the coated wafer by spin-coating SU-2005 for 30 s at 3,000 rpm, cell tracking.
soft-baking for 2 min, UV exposure of 550 mJ/cm2 through the optical
filter, post-exposure baking for 3 min at 95 °C, and developing in SU8 Collagen assay. A 3D collagen scaffold (final concentration of
developer. Third, the fluid-layer master was then prepared for the next 1.7 mg/ml) was obtained by mixing bovine collagen (PureCol, Advanced
layer by spin-coating HMDS at 3,000 rpm for 30 s and baking for 1 min at BioMatrix) in 1× minimum essential medium Eagle and 0.4% sodium
126 °C. Scotch Magic tape (3M) was applied over the alignment marks. bicarbonate (both Sigma-Aldrich), with 3 × 105 cells in R10 medium at
AZ-40XT-11D (MicroResist Technologies) was then coated at 3,000 rpm a 2:1 ratio5,29. After casting the mix in homemade migration chambers,
for 30 s. The Scotch tape was then peeled off to reveal the alignment gels were allowed to polymerize for 45 min at 37 °C, 5% CO2. Migration
marks. The wafer was then baked at 126 °C for 7 min, aligned and was observed by time-lapse video microscopy.
exposed to UV at 500 mJ/cm2. After a post-exposure bake for 10 min at
120 °C, it was developed in 726 MIF for about 5–10 min. The features Cell confiner assay. PDMS micropillars were placed on the soft pillar
were then rounded to make parabolic channels at 130 °C for 2 min on a homemade magnetic glass lid. When indicated, dish and confiner
(centre height of 26.6 μm). were coated with 2 μg/ml of recombinant mouse ICAM-1/human Fc chi-
maera (796-IC, R&D Systems). The setup was placed in a dish containing
Silanization. Before applying PDMS, wafers were placed in a desic- R10 medium in a humidified incubator at 37 °C, 5% CO2, 1 h before the
cator with 10 μl of Trichloro(1H,1H,2H,2H-perfluorooctyl)-silane experiments. Cells (5 × 104 in 5 μl R10 medium) were then pipetted onto
(Sigma-Aldrich), a vacuum of 100 mbar was applied, and then the the medium-freed micropillars and confined on a dish above a magnetic
desiccator was sealed for 1 h. One permanent coating was sufficient ring, and R10 medium was added back in the dish ready for imaging.
for later uses. For the myosin inhibition experiment, 50 μM para-nitroblebbistatin30
was added to the medium (Optopharma).
PDMS device microfabrication
Microfabricated devices were generated as previously described11,13,28, Primary T cell confinement assay. To measure actin flow speed and
with 1:10 PDMS and degassed for 2 min at 2,000 rpm (mix) and for cell curvatures, we used an agarose-confinement setup8. In brief,
2 min at 2,200 rpm (defoam) in a mixer/defoamer (ARE-250, Thinky) plasma-activated glass was overlaid with a 0.2 mg/ml solution of
before use. poly(l-lysine)-graft-PEG copolymer (PLL(20)-g[3.5]-PEG(2); SuSoS) at
4 °C overnight. To form a 0.5% agarose block with 10% FBS (and standard
Confiner. PDMS (Sylgard 184, Ellsworth Adhesives) was gently poured concentrations of other supplements), one part 2× HBSS buffer (Sigma)
onto the wafers, then 10-mm coverslips were activated by plasma clean- and two parts RPMI supplemented with twice the concentrations of
ing for 2 min at medium intensity (Harrick Plasma Cleaner, pdc-002, all other supplements used in R10 medium were mixed together with
Harrick Plasma) and pressed upside-down onto the PDMS-covered one part 1% high-molecular-weight agarose (850152, Biozym Scientific
wafers. The wafer was baked on a hot plate at 95 °C for 15 min, and the GmbH) in water at 50 °C and subsequently cast onto the dish then al-
300-μm width micropillar-coated coverslip was gently removed from lowed to solidify at room temperature. CCL19 (250-27B, Peprotech) was
the wafer. A soft PDMS 10-mm diameter pillar (1:30) was poured into a added to soluble agarose before casting. T cells were injected under the
homemade metal mould, degassed under vacuum and baked at 80 °C agarose block with a micropipette and incubated for 30 min at 37 °C
for 1 week. under 5% CO2 before imaging.

EDTA device. Of 1:10 PDMS (Sylgard 184, Ellsworth Adhesives), 20–25 g EDTA migration assay. T cells or PLB cells (5 × 104) were intro-
was poured onto the wafer contained in an aluminium mould in a Petri duced into the 2-mm entry port of the microchamber, and the de-
dish, degassed in a vacuum desiccator and baked overnight at 80 °C. vice was soaked in R10 medium. Cells were left to enter freely into
The devices were then diced with a razor blade and 2-mm entry and exit the confinement and channel zone for 1 h and then imaged with a
holes were punched (Harris Unicore biopsy puncher, Sigma-Aldrich). video microscope. After 1 h of imaging, chemical disruption of the
Following punching, they were cleaned with tape, sonicated in ethanol, cell adhesions was performed by removing R10 medium from the
blown dry and plasma-bonded to a coverslip. dish and adding R10 medium containing 10 mM EDTA (prepared
from 0.5M stock; EDS-100G, Sigma-Aldrich) with or without 50 μM
Multilayer devices. PDMS (80 g; RTV615, Techsil) was used; 10 g of para-nitroblebbistatin30 (Optopharma). Imaging was then resumed
PDMS was put on the control layer and spin-coated (500 rpm for 15 s for 1 h. For dendritic cells, experiments were performed 1 day after
followed by 2,300 rpm for 60 s), and 70 g was applied on top of the flow overnight LPS activation (200 ng/ml; Sigma-Aldrich) with or without
2.5 μg/ml CCL19 (250-27B, Peprotech) added on the opposite hole of
cell entry and with or without 30 mM EDTA13. Time-lapse video microscopy
Bright-field video of T cells for collagen assays (×10 objective), 3-μm
Multilayered microfluidics: maze and microchannels migration high confiner assays and primary T cells migrating under agarose
assay. As talin-KO cells failed to freely enter the confinement zone, a (×20 objective) were acquired by time-lapse acquisition (time interval
homemade microfluidic setup was used to push the cells by applying of 20 s) using inverted cell culture microscopes (DM IL Led, Leica
a pressure differential to the inlet and outlet ports of the device. The Microsystems) equipped with cameras (ECO415MVGE, SVS-Vistek)
microfluidics designs in Extended Data Fig. 2d were used, as well as and custom-built climate chambers (5% CO2, 37 °C, humidified). A
separated chambers containing the four pillar mazes (identical to 5-μm high confiner, EDTA and microfluidics assays were recorded
ref. 28 with four chambers). In brief, the core component of those every 30 s or 60 s at 2–6 multi-positions with NIS Elements software
devices is the flow layer with a 5-μm height pillar maze (Fig. 2a–d, (Nikon Instruments). Experiments were performed with an inverted
Extended Data Fig. 2e) or channels (Figs. 2e–g, 4, Extended Data wide-field Nikon Eclipse Ti microscope in a humidified and heated
Figs. 2f, 4) with an exit, cell and medium entry ports and on one side chamber at 37 °C and 5% CO 2 (Ibidi Gas Mixer), equipped with a
and on the other side a sink channel connected to two or three exits. ×20/0.5 NA PH1 air objective, a Hamamatsu EMCCD C9100 camera
Flow and sink channels are rounded, 26 μm high. For a controlled and a Lumencor Spectra X light source (390 nm, 475 nm, 542/575 nm;
flow of cells and fluids, all ports are equipped with independently Lumencor). TIRF microscopy was performed with a ×60/1.46 NA oil
controllable push-up PDMS membrane valves. In brief, the control objective, optovar ×1 or ×1.6 in a humidified and heated chamber at
channels are connected to solenoid valves (MH1, miniature, Festo) 37 °C and 5% CO2 using an inverted Axio Observer (Zeiss) microscope,
controlled with a MATLAB graphical user interface (MathWorks). a TIRF 488/561-nm laser (Visitron Systems) and an Evolve EMCCD
First, the chip was mounted into the 37 °C, 5% CO2 chamber of the camera (Photometrics) controlled by VisiView software (Visitron
microscope, and optimal closing pressures of 0.15–0.2 MPa of the Systems). To visualize actin retrograde flow, assays were recorded
water-filler PDMS membrane valves were determined for each chip; every second for TIRF acquisition only (Fig. 1e–g, Supplementary
each valve was checked individually by microscopy. Second, the whole Video 3 example 3) or every 2 s or 3 s when acquisition of both TIRF
chip was saturated with a pre-warmed medium by applying a pressure or wide-field fluorescence and bright field was necessary (Fig. 4e–g,
of 0.04 MPa and incubated in R10 medium for 30 min. Cells were first Extended Data Figs. 4b, 5g, Supplementary Videos 3 (examples 1 and
loaded via the ports on one side, resulting in their distribution along 2), 10). When cells migrated outside the field of view (Supplementary
the chamber, then slowly pushed with a pressure of <0.01 MPa from Video 10), they were manually followed via a joystick controller.
the medium port towards the confinement zone, and further pushed For the primary T cell confinement assay on nanoridges (Extended
in the microchannels or micropillars maze. Finally, all valves were Data Fig. 7), a video of actin flow (Lifeact-GFP) were recorded (2-s
closed to ensure that no external pressure or flow was applied to the frame rate) on an inverted spinning-disc confocal microscope
cells during imaging. (Andor) using a ×100/1.4 NA objective and a 488-nm laser line in a
custom-built climate chamber (37 °C under 5% CO2). For the primary
TIRF microfluidics: microchannels migration assay. For Fig. 4e–g, T cell flow analysis (Extended Data Fig. 8), inverted spinning-disc
TIRF microscopy was not possible with the multilayered microfluidic confocal microscopy was performed at 37 °C under 5% CO2 with an
devices, so devices with the flow layer only were generated. First, iMIC (TILL Photonics, FEI) instrument with a ×60/1.35 NA objective
R10 medium was forced into the device with a vacuum desiccator, and a 488-nm laser line.
and then the chips were incubated in humidified 5% CO2 at 37 °C for
2 h before the experiments. Experiments were performed on the Image analysis
microscope stage (humidified environment 5% CO2 at 37 °C) and a FIJI imaging processing software (https://fiji.sc/) was used for image
low cell flow was delivered and stopped after talin-KO cell loading and video microscopy analysis.
in the microchannel using a LA120 syringe pump (5 μl/h; Landgraf
Laborsysteme). Spreading area. TIRF images of single cells were binarized after
de-speckling and background subtraction, and the spreading area
Primary T cell confinement assay on nanoridges was measured using the Analyse Particles tool.
T cells were confined under agarose as previously described with minor
modifications8. In brief, plasma-activated coverslips with nanoridges Cell size measurement. To measure cell size, fluorescent images of
(800 nm/800 nm/600 nm (groove/ridge/depth)) (ANFS-CS25-50, 3-μm-confined control (Lifeact-mCherry) and talin-KO (Lifeact-GFP)
NanoSurface Biomedical) were passivated with a 1 mg/ml solution of cells were binarized, and the single-cell spreading area and perimeter
poly(l-lysine)-graft-PEG copolymer (PLL(20)-g[3.5]-PEG(2), SuSoS) were measured using the Analyse Particles tool.
at 4 °C overnight. Flat control coverslips (FLAT-CS25-50, NanoSur-
face Biomedical) were either passivated as described above or coated Manual tracking. A 1-h-long bright-field video of T cells in collagen
with rhICAM1-Fc (2 µg/ml) (720-IC-050, R&D) to generate adhesive and 3-μm high confiners were reduced to a 1-min and 2-min interval,
substrates. Next, 0.5% agarose was prepared by mixing A, 5 ml of 2× respectively, and cells were tracked manually by using the ‘Manual
HBSS buffer (Sigma), and mixing B, 10 ml of serum-free RPMI supple- tracking’ plugin provided by Fiji.
mented with 20% BSA and twice the concentrations of all other sup-
plements used in R10 medium (see above), and mixing C, 5 ml of 1% Automated tracking. For the 5-μm high confiner, pillar maze and
high-molecular-weight agarose (850152, Biozym Scientific GmbH) in EDTA assays, cell migration was analysed by nucleus tracking using
PBS. Agarose solution (0.5%) was kept at 50 °C in a water bath before TrackMate31 (https://imagej.net/TrackMate). For the EDTA assay, tracks
it was layered on top of the coated coverslips and allowed to solidify before and after EDTA treatment were analysed and the speed was
for 1 h at 4 °C. CCL19 (10 nM; 250-27B, Peprotech) was added to solu- extracted at the single-cell level in the different zones (2.5D confine-
ble agarose before casting. T cells were isolated from the spleen of ment, smooth and serrated microchannels containing only one cell).
wild-type and/or Lifeact-GFP mice (T cell isolation kit mouse; MACS, For Fig. 4b, cell migration in the channel zones was regarded as efficient
Miltenyi Biotec) and kept in R10 medium at 37 °C and 5% CO2. Before for velocities above 0.5 μm/min. The mean square displacement was
imaging (30 min), 0.5 µl of highly concentrated T cells was injected calculated from the tracking files with a homemade script (MATLAB,
under the agarose block. MathWorks).
Article
Actin flow versus cell velocity analysis. For Figs. 1e–g, 4e–g, Extended Data availability
Data Figs. 1l, 4b, 5g, 8a, kymographs were generated along the cell Data that support the findings of this study are available within the
trajectories, and both cell and actin velocities were extracted from Article and its Supplementary Information. Source data for Figs. 1, 2,
those kymographs. To follow cells migrating in a channel with a high 4 and Extended Data Figs. 1–8 are provided with the paper. Original
magnification (Fig. 4e–g, Supplementary Video 10), the channel was videos will be made available on reasonable request due to the large
realigned with a homemade script (MATLAB, MathWorks). file size.

Microchannel assay analysis. Cells going in both directions of the

24. Riedl, J. et al. Lifeact: a versatile marker to visualize F-actin. Nat. Methods 5, 605–607
microchannels were analysed, and channels containing more than one (2008).
cell were excluded. Kymographs were generated along the channel, and 25. Leithner, A. et al. Diversified actin protrusions promote environmental exploration but are
cell velocity was calculated for each cell in the different zones. dispensable for locomotion of leukocytes. Nat. Cell Biol. 18, 1253–1259 (2016).
26. Shalem, O. et al. Genome-scale CRISPR–Cas9 knockout screening in human cells.
Science 343, 84–87 (2014).
Actin flow in primary T cells confined on nanoridges. Actin flow 27. Leithner, A., Merrin, J., Reversat, A. & Sixt, M. Geometrically complex microfluidic devices
for the study of cell migration. Protoc. Exch. https://doi.org/10.1038/protex.2016.063
vectors were measured by two methods: A, which was automatic par-
(2016).
ticle tracking with TrackMate plugin (3.8.0) of FIJI31 (ImageJ 1.52p, java 28. Schwarz, J. et al. A microfluidic device for measuring cell migration towards
1.8.0), and B, which was optical flow analysis (using a custom MATLAB substrate-bound and soluble chemokine gradients. Sci. Rep. 6, 36440 (2016).
29. Sixt, M. & Lämmermann, T. in Cell Migration: Developmental Methods and Protocols (eds.
code). Only frames where optical flow analysis was confirming auto-
Wells, C. M. & Parsons, M.) 149–165 (Humana, 2011).
matic particle tracking (deviation of <30°) were subjected to further 30. Képiró, M. et al. para-Nitroblebbistatin, the non-cytotoxic and photostable myosin II
analysis. To estimate vectors of forwards cell movement, cell contours inhibitor. Angew. Chem. Int. Ed. Engl. 53, 8211–8215 (2014).
31. Tinevez, J. Y. et al. TrackMate: an open and extensible platform for single-particle
were segmented with Ilastik32, and cell displacement was measured by
tracking. Methods 115, 80–90 (2017).
tracking the centroids. Angular distributions of actin flow vectors and 32. Sommer, C., Straehle, C., Koethe, U. & Hamprecht, F. A. Ilastik: interactive learning and
angular distributions of cell forwards movement vectors were then segmentation toolkit. In 2011 IEEE International Symposium on Biomedical Imaging: From
Nano to Macro 230–233 (IEEE, 2011).
analysed and plotted with respect to the orientation of nanoridges
using a custom MATLAB code.
Acknowledgements We thank A. Leithner and J. Renkawitz for discussion and critical reading
of the manuscript; J. Schwarz and M. Mehling for establishing the microfluidic setups; the
Curvature analysis. For Extended Data Fig. 8, we set up a home- Bioimaging Facility of IST Austria for excellent support, as well as the Life Science Facility and
made script that performs binarization of fluorescent images of the Miba Machine Shop of IST Austria; and F. N. Arslan, L. E. Burnett and L. Li for their work
during their rotation in the IST PhD programme. This work was supported by the European
Lifeact-GFP-expressing cells. The local curvature was calculated from
Research Council (ERC StG 281556 and CoG 724373) to M.S. and grants from the Austrian
a spline fit to the cell outline, and the cortical actin flow was quantified Science Fund (FWF P29911) and the WWTF to M.S. M.H. was supported by the European
with optical flow analysis (MATLAB, MathWorks). Regional Development Fund Project (CZ.02.1.01/0.0/0.0/15_003/0000476). F.G. received
funding from the European Union’s Horizon 2020 research and innovation programme under
the Marie Skłodowska-Curie grant agreement no. 747687.
Statistical analysis
All statistical analyses were performed with GraphPad Prism (Graph- Author contributions A.R. and M.S. conceived the experiments and wrote the manuscript, with
critical feedback from all authors. A.R. designed, performed and analysed the experiments,
Pad Software). The normality test was performed, and the statistical
with the help of I.d.V., J.S., M.H., R.H. and S.T. A.R. and J.M. designed the microfluidic devices.
strategy used accordingly (for example, non-parametric Mann–Whit- J.M. performed the photolithography. F.G. designed and performed the nanoridge
ney U-test and one-way ANOVA with Kruskal–Wallis test with post hoc experiments. J.A. fabricated the nanoridge substrates. R.H. wrote the image analysis scripts.
A.C.-J., R.V., J.M., R.H., M.S. and A.R. discussed the physical model. A.C.-J. and R.V. wrote the
Dunn’s test) as indicated in Supplementary Table 1. For the box and
physical model.
whiskers graphs in Fig. 4d, g and Extended Data Figs. 2f, 4a, 4b, 5g, 6f, the
box extends from the 25th to the 75th percentiles, with the middle line Competing interests The authors declare no competing interests.
showing the median, and the whiskers indicating the minimum to maxi-
Additional information
mum values. See Supplementary Table 1 for additional information. Supplementary information is available for this paper at https://doi.org/10.1038/s41586-020-
2283-z.
Reporting summary Correspondence and requests for materials should be addressed to A.R. or M.S.
Peer review information Nature thanks Kenneth Yamada and the other, anonymous,
Further information on research design is available in the Nature reviewer(s) for their contribution to the peer review of this work.
Research Reporting Summary linked to this paper. Reprints and permissions information is available at http://www.nature.com/reprints.
 a b c e
d
bl g1 g2 3

RNA expression (a.u.)

125
am lin1 lin1 1g Control Talin KO
CD4+ primary T cell r lin
CRISPR Wash, Puromycin 100 CD4+ T cell line Sc Ta Ta Ta
virus infection Selection stock re-culture
75 BMDCs Talin1 and 2
d1 d7 Experiments
d0 d3 d10 to d30 50
188kDa
Media Storage (-175ºC) 25
Renewal 0
39kDa GAPDH

2
lin

lin
Ta

Ta
e Control Lifeact-mCherry Talin KO-Lifeact-GFP
f
Control Talin KO

Brightfield
3D
EPI

Epi mCherry Epi GFP

Merged
TIRF

120:00
TIRF mCherry TIRF GFP

g h i
2.5D Control Talin KO
ns ns

Cell speed (µm/min)

Cell speed (µm/min)
25 25

20 20
3D view 15 15
10 10
5
5µm

5
Nucleus
side view Lifeact-GFP 0 0

m
rry
60:00

FP
l
tro

5µ

8µ
he
t-G
on

C
C

ac

t-m
fe
ac
Li
fe
Li
j k l
Control Talin KO
2.5D
300 ****
ns 1500 25
Cell perimeter (µm)

Velocity (µm/min)
ns ****
Cell area (µm2)

200 20
TIRF contact area (µm2)

1000 ns
500
3D view 15
100 500 400
10
300
Fc-ICAM1
5µm

0 0 200 5
l
KO

l
KO
tro

tro

100 Fc-ICAM1
0
on

on
in

in

0
side view Cell Actin Cell Actin
C

l
Ta

Ta

l
KO

Control Talin KO
tro
on

in
C

l
Ta

Extended Data Fig. 1 | Effects of Tln1 deletion on T cell adhesion and and imaged by video microscopy. Representative of four experiments.
migration. a, Workflow scheme of virus production, cell infection and Snapshot is at t = 60 min. Nucleus tracks are displayed in yellow. Scale bars,
analysis. d, day. b, RNA-seq analysis of talin 1 and 2 expression in mouse CD4+ 50 μm. h, Cell speed of control T cells (n = 62) and T cells expressing Lifeact-GFP
primary T cells, the T cell line and bone marrow-derived dendritic cells (n = 100) or Lifeact-mCherry (n = 138). Data are mean ± s.d.; ns P = 0.0775,
(BMDCs). c, Western blot analysis of talin 1 and talin 2 expression in T cells Kruskal–Wallis one-way ANOVA followed by two-sided post hoc Dunn’s test;
infected with lentivirus containing scrambled and talin 1-targeting CRISPR control versus Lifeact-GFP or Lifeact-mCherry: P > 0.999; Lifeact-GFP versus
guides (guides 1–3 (Talin1g1–Talin1g3)). Data are representative of three Lifeact-mCherry: P = 0.0717. i, Speed of T cells placed under 5-μm (n = 300) or
independent experiments. d, Related to Fig. 1a. Phase-contrast images of 8-μm (n = 274) confinement. Data are mean ± s.d.; ns P = 0.8460, two-sided
control and talin-KO cells. Scale bars, 50 μm. Representative of four Mann–Whitney U-test. j, Perimeter (left) and area (right) of control and talin-KO
experiments. e, Related to Fig. 1b. Representative snapshots of cells placed under 3-μm confinement (control cells: n = 145, talin-KO cells:
epifluorescence (EPI) and TIRF microscopy of control cells (left) and talin-KO n = 127). Representative of three experiments. Data are mean ± s.d.; ns
cells (middle) expressing mCherry and eGFP Lifeact fusions, respectively, P = 0.4261 for perimeter and P = 0.2558 for area, two-sided Mann–Whitney
plated on a Fc-ICAM1-coated surface. Representative of three experiments. U-test. k, Spreading area observed by TIRF microscopy of control (n = 27) and
The right panel shows bright field and merged fluorescence. Scale bar, 5 μm. talin-KO (n = 20) cells under 5-μm confinement. Representative of two
f, Related to Fig. 1c. Representative snapshots of control and talin-KO cells experiments. Data are mean ± s.d.; ns P = 0.2025, two-sided Mann–Whitney
embedded in 3D collagen gel at t = 120 min; individual tracks are displayed in U-test. l, Scheme of the 5-μm high Fc-ICAM1-coated confinement (left) used to
different colours. Representative of three experiments. Scale bars, 100 μm. measure velocities and F-actin retrograde flow (right) in control cells (n = 49)
g, Related to Fig. 1d. Left, scheme of 5-μm high confinement used in f, g and j. and talin-KO T cells (n = 59). Representative of five experiments. ****P < 0.0001,
Right, control cells (left) and talin-KO T cells (right) expressing Lifeact-GFP two-sided Mann–Whitney U-test.
(red), stained with Hoechst (cyan), were placed under 5-μm height confinement
Article
a b c
2.5D Control Talin KO
3µm
**** ****
10 25.103

Cell speed (µm/min)

600 Control

Displacement (µm)
8 20.103 TalinKO

MSD (µm2)
3D view
6 400 15.103
4 10.103
200
3µm

2 5.103
side view
0 0 0
60:00
0 20 40 60

KO

KO
l

l
tro

tro
on

on
Time (min)

lin

lin
C

C
Ta

Ta
d
Entry Exit
Channel Channel
Flow channels Microchannels
Control valves
Medium entry ports

Cell loading entry port or

top view
Exit ports Micropillars

25µm PDMS
flow layer

5µm
side view

Cell loading outlet port PDMS

control layer

e
pillar spacing 3µm 4µm 5µm 6µm pillar spacing 3µm
Talin KO 4µm 5µm 6µm
Control

Nucleus Nucleus
Lifeact-GFP Lifeact-GFP
60:00 60:00

f
25 25 *
Cell velocity

Serrated 20 20
(µm/min)

6µm Smoooth 15 15
10 10
5 5
Channel
width 5µm 6µm 7µm 5µm 0 0
Channel
width
5µm 6µm 7µm Smooth 5µm 6µm 7µm Smooth

Control Talin KO

Extended Data Fig. 2 | Topography rescues T cell locomotion in the absence microchannels used in e and f. e, Snapshots of video microscopy at t = 60 min of
of adhesion. a–c, Migration of control (n = 71) and talin-KO (n = 74) cells under control and talin-KO cells in the pillar device. The nucleus is shown in cyan
3-μm confinement. Representative of four independent experiments. In a, the (Hoechst), Lifeact-GFP reporter in red and the bright-field image in grey.
left panel shows the scheme of the 3-μm confinement used in a–c. The right Individual cell tracks are displayed in yellow. Scale bars, 50 μm. Time is in min:s.
panel shows representative snapshots of control and talin-KO cells under 3-μm Representative of three experiments. f, Related to Fig. 2g, h. Velocities
confinement at t = 60 min. Individual tracks are displayed in different colours. (μm/min) of control and talin-KO cells in serrated channels of different
Scale bars, 50 μm. In b, the speed and displacement of control and talin-KO cells diameters (with a conserved serration period of 6 μm). Representative of three
under 3-μm confinement are shown. Data are mean ± s.d.; ****P < 0.0001, experiments. n = 57 for control cells and n = 9 for talin-KO cells. *P = 0.0356;
two-sided Mann–Whitney U-test. In c, the MSD of control cells (black) and otherwise not significant, Kruskal–Wallis one-way ANOVA followed by post hoc
talin-KO cells (red) is shown. Data are mean ± s.e.m.; P < 0.0001, two-sided Dunn’s test. Boxes extend from the 25th to the 75th percentiles, with the middle
Mann–Whitney U-test. d, Related to Fig. 2a–d. Scheme of multilayered line showing the median and the whiskers representing the minimum to
microfluidic devices used to push cells into the pillar maze or into maximum values.
 a b
No chemokine + CCL19 + CCL19

** **** **** **** **** ****

ns ***
20 20

Cell speed (µm/min)

Cell speed (µm/min)

15 15

10 10

5 5

0 0
EDTA - + - + - + EDTA - + - + - +
Channel Channel
6-12µm 24µm Smooth 6-12µm 24µm Smooth
zone zone
Cell frequency (%)

Cell frequency (%)

50 50
40 Control 40 Control
6-12µm 6-12µm
30 +EDTA 30 +EDTA
20 20
10 10
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Cell frequency (%)

Cell frequency (%)

50 60
40
24µm 24µm 40
30
20 20
10
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20

60
Cell frequency (%)

Cell frequency (%)

50
40
Smooth 40 Smooth 30
20 20
10
0 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20

Cell speed (µm/min) Cell speed (µm/min)

Extended Data Fig. 3 | Topography rescues dendritic cell locomotion in the scheme of a mature dendritic cell migrating in a channel designed for the EDTA
absence of adhesion. a, Top left, scheme of a mature dendritic cell migrating experiment, along a CCL19 gradient (indicated in red). Mature dendritic cells
in a channel designed for the EDTA experiment. Mature dendritic cells migrate migrate towards the chemokine CCL19 into a device that contains
in a device containing channels with different serration periods, with or microchannels with different serration periods, with or without 30 mM EDTA.
without 30 mM EDTA. Top right, single-cell migration speed with (+, blue) and Top right, migration speed with (+, blue) and without (-, grey) EDTA treatment.
without (-, grey) EDTA treatment. Bottom, histograms of cell speed in different Bottom, histograms of cell speed in the different serrated channels with (blue)
serrations with (blue) or without (grey) EDTA. Representative of three or without (grey) EDTA. Representative of three independent experiments.
independent experiments. n = 35 cells without EDTA, n = 17 cells with EDTA. Without EDTA: n = 44 cells; with EDTA: n = 52 (6–12 μm), n = 44 (24 μm), n = 33
Data are mean ± s.d.; **P = 0.0022, ****P < 0.0001 and ns P = 0.6363, one-way (smooth channels). Data are mean ± s.d.; ***P = 0.0001 and ****P < 0.0001,
ANOVA with Kruskal–Wallis test followed by post hoc Dunn’s test. b, Top left, one-way ANOVA with Kruskal–Wallis test followed by post hoc Dunn’s test.
Article
a Control Talin KO
b
Actin cortex
30 30 ***

Cell velocity (µm/min)

n=28 ** n=28
20 20 TIRF Lifeact-GFP
10 10
TIRF Actin
0 0 retrograde flow

th
m

th

oo
oo

6-

-
Cell length

6-

12

24
12

24

Sm
Sm
30 * *** Smooth
30 2.5D

Cell velocity (µm/min)

n=13 ** n=15 channel
20 20 5µm

5µm
5µm
10 10

0 0 ns

th
m

th
39µm 49µm

oo
oo
80 80

6-

-
6-

Cell length (µm)

12

24
12

24

Sm
Sm
60 60
30 30 * 27µm
Cell velocity (µm/min)

n=15 n=22 40 40
20 20 20 20
10 10 0 0
0 0 ns
m

th
m

th

19µm

oo
oo

30 30
6-

-
6-

12

24
12

24

Sm
17µm
Sm

flow length (µm)

Actin retrograde
20 14µm 20
30 30 **
Cell velocity (µm/min)

n=31 ** n=15
20 20 10 10

10 10 0 0

TA

TA

KO
0 0

ED

ED

in
l
m

th
m

th

Ta
l+

l+
oo
oo

tro

tro
6-

-
6-

12

24
12

24

Sm
Sm

on

on
C

C
Extended Data Fig. 4 | Effect of topographical features on T cell migration. Kruskal–Wallis test followed by post hoc Dunn’s test. Boxes extend from the
a, Related to Fig. 4d. Scheme of channel designs with decreasing complexity. In 25th to the 75th percentiles, with the middle line showing the median and the
brief, serrations with a period of 6, 12 and 24 μm were arranged in consecutive whiskers representing the minimum to maximum values. b, Top, scheme of cell
zones of a 5-μm wide channel ending with a smooth zone. The left panel shows and actin retrograde flow length. Bottom, control cells expressing the
the different geometries. The right and middle panels show the velocity of cells Lifeact-GFP reporter with 10 mM EDTA (grey) or talin-KO cells (red), under 5-μm
migrating in zones with a different period. The total number of cells in each confinement (left, n = 12) or in a 5 × 5-μm channel (right, n = 9 for control cells
design is shown. Representative of four (control cells) and eight (talin-KO cells) and n = 8 for talin-KO cells), were observed by TIRF microscopy, and the mean
independent experiments. Control: *P = 0.0415. KO arrow ‘right’: ***P = 0.0008 cell length (ns P = 0.0555, two-sided Mann–Whitney U-test) and the actin
and **P = 0.0013; KO arrow ‘left’: ***P = 0.0003 and **P = 0.0018; KO bulb: retrograde flow length (ns P = 0.7247, two-sided Mann–Whitney U-test) were
*P = 0.0344; KO half-bulb: **P = 0.0016 (6 μm versus smooth) and **P = 0.0013 measured using kymograph analysis. Representative of two independent
(12 μm versus smooth), otherwise not significant, one-way ANOVA with experiments. Data are mean ± s.d.
 Micro-
a Confinement Confinement
Cell loading port Outlet channels

Loading port Outlet

b Serrated Smooth c
2.5D 2.5D
channel channel
Serrated channel
3D view
Smooth channel

side view ns
ns

pNB (%)
ns
**** **** **** 300
Cell speed (µm/min)

20 20 20

15 15 15

before
200

after
10 10 10
100

Cell speed
5 5 5

0 0 0 0
pNB - + - + - +
d Serrated Smooth e 2.5D
2.5D
channel channel
Serrated channel
3D view
Smooth channel
pNB+EDTA (%)

ns
side view
**** ****
**** 20 **** 20 **** 300
Cell speed (µm/min)

20

15 15 15
200
before
after

10 10 10
100
Cell speed

5 5 5

0 0 0 0
pNB+EDTA - + - + - +
f +paraNitroBlebbistatin (50µM) +EDTA (10mM)
+EDTA (10mM)
50 +paraNitroBlebbistatin (50µM)
Cell frequency (%)

50 50
40 40 40
2.5D
30 30 30 Serrated channel
20 20 20
Smooth channel
10 10 10
0 0 0
20 40 60 80 100 120 20 40 60 80 100 120 20 40 60 80 100 120
before before before
Cell speed treatment (%) Cell speed treatment (%) Cell speed treatment (%)
after after after

g
Brigthfield t=0min t=8min t=20min t=70min
ns
25 ***
*
Velocity (µm/min)

20
Lifeact-mCherry
15
Cell
10
Actin
0s 15s 30s 45s 60s
5 flow
0
Channel
m

µm

µm

th
6µ

oo

zone
12

24

Sm

0s
60s EDTA

Extended Data Fig. 5 | See next page for caption.

Article
Extended Data Fig. 5 | Actin polymerization versus contractility during e, Migration speed change after versus before pNB + EDTA treatment. Data are
topography-based locomotion. a, Scheme of the device used for the mean ± s.d. ns P = 0.1170 and ****P < 0.0001, one-way ANOVA with Kruskal–
experiments in Fig. 2i, j and Supplementary Figs. 3, 5. b, Control T cells Wallis test followed by post hoc Dunn’s test. f, Histogram of migration speed
migrating under confinement (2.5D) or in smooth versus serrated channels as variation after versus before pNB treatment alone (left, see b, c), EDTA
shown in the top panels. After 60 min, 50 μM para-nitroblebbistatin (pNB) was treatment alone (middle, related to Fig. 2i, j) or pNB + EDTA treatment (right,
added. The migration speed before (-) and after (+) treatment is shown. see d, e). g, Related to Fig. 4e–g. Left, snapshots of a control T cell expressing
Representative of four independent experiments. n = 68 for 2.5D, n = 79 for Lifeact-mCherry incubated with 10 mM EDTA in different channel zones. The
serrated channel and n = 44 for smooth channel. ****P < 0.0001, two-tailed, left panel from top to bottom shows bright field, Lifeact-mCherry in black,
Wilcoxon matched-pairs signed-rank test. c, Migration speed change after colour-coded snapshots of F-actin in the different channel zones (15-s
versus before pNB treatment. Data are mean ± s.d. ns P > 0.9999 (2.5D versus intervals), and the corresponding kymograph of a cell migrating in the channel.
serrated channel), ns P = 0.1414 (2.5D versus smooth channel) and ns P = 0.1044 Scale bars, 5 μm. Right, cell and actin retrograde flow velocities observed by
(serrated vs smooth channel), one-way ANOVA with Kruskal–Wallis test wide-field microscopy and measured with kymograph analysis. n = 8 cells;
followed by post hoc Dunn’s test. d, Control T cells migrating under 3 cells in all zones, 4 cells in the 6–12–24-μm zone and 1 cell in the 12–24-µm
confinement or in smooth versus serrated channels as shown in the top panels. zone. *P = 0.0138 and ***P = 0.0002, one-way ANOVA with Kruskal–Wallis test
After 60 min, 50 μM pNB and 10 mM EDTA were added. The migration speed followed by post hoc Dunn’s test. Boxes extend from the 25th to the 75th
before (-) and after (+) treatment is shown. Representative of four independent percentiles, with the middle line showing the median and the whiskers
experiments. n = 166 for 2.5D, n = 45 for serrated channel and n = 22 for smooth representing the minimum to maximum values.
channel. ****P < 0.0001, two-tailed, Wilcoxon matched-pairs signed-rank test.
 a b
parallel to ridges

Agarose

ICAM-1 perpendicular
WT and Lifeact-GFP WT and Lifeact-GFP to ridges
primary T cell primary T cell
or
top view top view

Agarose 800nm Agarose

Agarose Agarose
Pll-PEG Pll-PEG 600 nm
side view side view
800nm

c
Smooth Smooth Ridges
ICAM-1 PEG PEG

50µm

d
-600 -600 -600

-300 -300 -300

Y (µm)

0 0 0

300 300 300

600 600 600

-600 -300 0 300 600 -600 -300 0 300 600 -600 -300 0 300 600
X (µm)

90° 90° 90°

120° 60° 120° 60° 120° 60°

150° 30° 150° 30° 150° 30°

180° 0° 180° 0° 180° 0°

-150° -30° -150° -30° -150° -30°

-120° -60° -120° -60° -120° -60°

-90° -90° -90°

f g ****
30 **** 30 ****
ns
****
****
Cell velocity (µm/min)

cell velocity (µm/min)

Mean instantaneous

20 20

10 10

0 0
Smooth Smooth Ridges Smooth Smooth Ridges
ICAM-1 PEG PEG ICAM-1 PEG PEG

Extended Data Fig. 6 | See next page for caption.

Article
Extended Data Fig. 6 | Topography-based migration of primary T cells. the top right corner. e, Rose diagram (with circular lines from the centre
a, Scheme (top and side view) of control and Lifeact-GFP cells confined under showing 6%, 12% and 18% of total cells) of cells migrating on Fc-ICAM1-coated
agarose on a smooth surface coated with Fc-ICAM1 or passivated with PLL-PEG. smooth surfaces (left, n = 419), PLL-PEG-coated smooth surfaces (middle,
WT, wild type. b, Scheme (top and side view) of control and Lifeact-GFP cells n = 650) and PLL-PEG-coated ridge-baring surfaces (right, n = 687). The
confined under agarose on a coverslip with linear ridges passivated with orientation of ridges is indicated in the top right corner. n = 6 independent
PLL-PEG. c, Minimum-intensity projection showing the trace of control cells experiments. f, g, Velocities of cells migrating under agarose on
migrating for 15 min under agarose on Fc-ICAM1-coated smooth surfaces (left), Fc-ICAM1-coated smooth surfaces (n = 419), PLL-PEG-coated smooth surfaces
PLL-PEG-coated smooth surfaces (middle) and PLL-PEG-coated ridge-baring (n = 650) and PLL-PEG-coated ridge-baring surfaces (n = 687). n = 6 independent
surfaces (right). Representative of six independent experiments. Scale bars, experiments. Cell velocities (cell displacement divided by track duration) (f)
50 μm. d, Randomly selected tracks of mixed control and Lifeact-GFP cells and mean instantaneous velocities (g) are shown. ****P < 0.0001 and ns
migrating under agarose on Fc-ICAM1-coated smooth surfaces (left, n = 183), P = 0.0757, both one-way ANOVA with Kruskal–Wallis test followed by post hoc
PLL-PEG-coated smooth surfaces (middle, n = 183) and PLL-PEG-coated Dunn’s test.
ridge-baring surfaces (right, n = 183). The orientation of ridges is indicated in
 a b c
15 90°

Normalized actin speed (a.u.)

parallel to ridges 2.0
90° 10 0°
1.5
5
0°
Lifeact-GFP

Y (μm)
perpendicular 0 1.0
primary T cell to ridges
-5 r=-0.65
top view 0.5
-10
800nm Agarose
Agarose -15 0.0
Pll-PEG 600 nm -20 -15 -10 -5 0 5 10 15 20 0 5 10 15
side view X (μm) Cell speed (µm/min)
800nm
d
5µm
Cell core
Detected spots
Spots inside cell core
Spots direction
Cell direction
Optical flow direction
Lifeact-GFP Actin: spots tracking Actin: optical flow

e f
120 120
90° 90° 2.0

Normalized speed (a.u.)

Cell Speed
100 100
Number of detections

Actin Speed
Number of detections

0° 0°
1.5
80 80

60 60 1.0

40 40 0.5

20 20
0.0
0
15

0
0

-3

-4

-6

-7

-9
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90 0-

15

30

45

60

75
Angle of actin flow (°) Angle of cell movement (°) Angle of cell movement (°)

Extended Data Fig. 7 | Actin-based force transmission in primary T cells. shows predominant retrograde actin flow when aligned along the ridges (90°).
a, Scheme (top and side view) of cells migrating under agarose on a Retrograde actin flow declines towards 0° (perpendicular to the ridges), while
non-adhesive, topography-baring substrate. b, Tracks of migrating cells forwards locomotion (right histogram) steadily increases, indicating that
analysed in c–f. n = 18; mean cell speed = 6.58 μm/min; mean actin retrograde retrograde actin flow couples to topographical barriers and drives
flow speed = 6.53 μm/min. Starting points of tracks are shifted to the origin. locomotion. n = 422 events of actin flow and cell movements obtained in n = 18
c, Mean speed of cells in relation to their mean actin retrograde flow speed (in cells from three independent experiments. f, Frame-to-frame speed of cells
relation to the substrate). Pearson's rank correlation coefficient r = −0.6484. migrating on nanoridges increases when forwards locomotion aligns
A drop in retrograde flow with increasing locomotion speed demonstrates perpendicular to topographical barriers (90° → 0° ≃ 15% increase). The
that, as in the transmembrane clutch paradigm of force transmission, actin orientation-dependent increase of cell speed is paralleled by a steady decrease
slippage is inversely related to locomotion. d, Snapshots of actin flow analysis of retrograde actin flow (90° → 0° ≃ 55%), indicating that retrograde actin flow
on ridged surfaces of cells expressing Lifeact-GFP. Left, cell core segmented couples to topographical barriers and drives locomotion. Frame-to-frame cell
with Ilastik (pink); the yellow arrow indicates cell direction, the blue arrow speed and actin retrograde flow speed were recorded in 18 cells from 3
indicates the mean optical flow direction and the green arrow indicates actin independent experiments, and pooled to n = 412 (cell speed) and n = 422 (actin
retrograde flow. Middle, optical flow analysis (obtained with a customized retrograde flow speed) events. The centre shows the mean, and both s.e.m.
MATLAB code). Right, actin single-spot tracking (TrackMate). Representative (solid lines) and s.d. (thick transparent lines) are shown. Before pooling, cell
of three independent experiments. e, Angular distribution of actin retrograde speed was normalized to the mean speed of the tracked cell, and actin
flow and forwards locomotion of cells migrating on nanoridges. Only frames retrograde flow speed was normalized to the mean actin retrograde flow
where optical flow analysis was confirming automatic particle tracking speed.
(deviation of <30°; see d) were subjected to analysis. The histogram on the left
Article
a b
Lifeact-GFP Rear Front Rear + Curvature -
Front Rear
Distance (µm)
0 10 20 30 Left semiperimeter (µm) Right semiperimeter (µm)
0
20 s 30 20 10 0 10 20 30 40
0 20 s
Time (s)

Time (s)
30
25 s 30
25 s

60 60
30 s
30 s
Front Rear
Front Rear

c d e
front back

Actin retrograde flow velocity (µm/min)

40
i
30

ii Binary mask
20

iii
+ Curvature - Cortical F-actin
10

iv 0
-0.1 -0.05 0 +0.05 +0.1
Curvature (a.u.)
Curvature speed (µm/min)

100
80 r=0.55 Analysis
60
1 nM vs
40
CCL19 10 nM
20 100 nM

0
0 20 40 60 80 100
Actin speed (µm/min)

Extended Data Fig. 8 | Retrograde actin flow and shape changes in confined on an inert substrate and exposed to the indicated chemokine
adhesion-free confinement. Related to Fig. 4h. a–c, Primary T cells (CCL19) concentrations. n = 51 cells from three independent experiments:
expressing Lifeact-GFP confined on PLL-PEG. n = 51 cells from three n = 22 cells for 100 nM CCL19, n = 16 cells for 10 nM CCL19 and n = 13 cells for 1 nM
independent experiments: n = 22 cells for 100 nM CCL19, n = 16 cells for 10 nM CCL19. Pearson's rank correlation coefficient r = 0.5542. d, e, Cell curvature
CCL19 and n = 13 cells for 1 nM CCL19. a, Representative snapshots (right, t = 20, versus cortical actin retrograde flow analysis in channels (obtained from
25 and 30 s) and kymographs (left), of primary T cells expressing Lifeact-GFP three cells in three experiments). A scheme of the segmentation used to
confined on PLL-PEG, showing actin retrograde flow (red arrows). Scale bar, measure channel-dependent cell curvature and cortical actin retrograde flow
5 μm. b, Representative snapshots (right, t = 20, 25 and 30 s) of the cell outline velocity used in e is shown (d). Fluorescent images were binarized and the local
and the colour-coded curvature. Kymograph analysis (left) of the migrating curvature was calculated from a spline fit to the cell outline. Measurement of
T cell shows retrograde movement of cell body deformations (right, red channel-dependent cell curvature versus cortical actin retrograde flow shows
arrowhead). c, Top, scheme of the travelling cell body deformations. Bottom, no correlation (Pearson's correlation coefficient c ≈ −0.006425) (e). The green
retrograde actin flow versus curvature flow velocities in primary T cells line shows mean ± s.d.
 12345656762589 56 5317425
C$4)&71$B(*2$4)&7:$B(n$(:&4:&(+
!"#$"%&'()*+, >""06&(*&"":6&(n%2&$7:42+
-&('!#&(#./&'()*+,opopDpoDqr
03&('0!&(4)5$$")%1 ' 22&  /
 ($2!6()!#'4$.$7$(/8()59()&(5!'.7$):;)$82!6$# ('4('84"$("4/&"#(&"!&"4/
$"!($"%:<8'()$"82&($""3&('0&4)!7$4$=>'()?08 &"#()@#$($&7A7$4/)497$(:

1(&($($4
<&77(&($($4&7&"&7/=4"8$2()&(()8775$"%$(2 &!"($"()8$%'7%"#=(&.77%"#=2&$"(B(=C()# 4($":
"D& "8$2#
s ;)B&4(&2!7  $E*F+8&4)B!$2"(&7%'!D4"#$($"=%$6"& &#$4("'2.&"#'"$(82&'2"(
s > (&(2"("5)()2&'2"(5(&9"82#$($ "4(&2!7 5)()() &2 &2!75& 2&'#!&(#7/
s; ) (&($($4&7((*+'#>3G5)()()/&"H(5H$##
IFJKLMNOONFLPQRPRLRSNTJULVQLUQRMWXVQULRNJQJKLVKLFYOQZLUQRMWXVQLONWQLMNO[JQ\LPQMSFX]TQRLXFLPSQL^QPSNURLRQMPXNF_
s >#4$!($"8&7746&$&( ((#
s >#4$!($"8&"/& '2!($" 44($"='4)& ((8"2&7$(/&"#&#`'(2"(82'7($!742!&$"
s >8 '77#4$!($"8() (&($($4&7!&&2($"47'#$"%4"(&7("#"4/*:%:2&"+().&$4($2&( *:%:% $"488$4$"(+
>3G6&$&($"*:%:(&"#&##6$&($"+& 4$&(#($2&( 8'"4(&$"(/*:%:4"8$#"4$"(6&7+
s <"'77)/!()$(($"%=()(((&($($4*:%:a=P=W+5$()4"8$#"4$"(6&7=884($E=#% 88#2&"#b6&7'"(#
cXdQLbLdYJTQRLYRLQ\YMPLdYJTQRLeSQFQdQWLRTXPYVJQ_
s <f&/$&"&"&7/$=$"82&($""()4)$48!$&"#C&964)&$"C"(&7 (($"%
s <)$&4)$4&7&"#42!7B#$%"=$#"($8$4&($"8()&!!!$&(7678((&"#8'77!($"%8'(42
s @($2&( 8884($E *:%:)"gU=A&"gW+=$"#$4&($"%)5()/54&74'7&(#
ITWLeQVLMNJJQMPXNFLNFLRPYPXRPXMRLhNWLVXNJNiXRPRLMNFPYXFRLYWPXMJQRLNFLOYFKLNhLPSQL[NXFPRLYVNdQ_
18(5&&"#4#
A7$4/$"82&($"&.'(&6&$7&.$7$(/842!'(4#
G&(&4774($" G&!$(8&4 774($"8$2&%$"%5& !82#5$()3t1H@72"(u:po:pp*3$9"@47$!2$44!+=v$$v$5o:q:q*v$$(";t0<D
7'4"42$44!+=>"#$wo:x*>"#!$""$"%H#$42$44!+=&"#$Ct-$6>4m'$$($"18(5&*;t--A)("$4=<@t=
l$77.=y0+z&7()2()# 4($"gt2&%$"%g:77($"%5& !82#'$"%<>1G$6&x:q:{:G$%"82$487'$#$4#6$4
5& !82#5$()7#&5|}*7!&($"=~1>+:>4'(2 4$!(5& '#(4"(7()6&76 *C&(7&.=C&()j9=~1>+:
G&(&&"&7/$ >"&7/$5& !82#5$()<$̀$*t2&%+q:pzC&(7&.0opq}&*C&()j9=~1>+z$7&($96q:q:u:'(2H2&# 4$!(8&"&7/$
*C&(7&.=C&()j9=~1>+&&6&$7&.7'!"m'(:
>77(&($($4&7&"&7/ 5!82#5$()k&!)A&#A$2€*k&!)A&#18(5&=~1>+:
<2&"'4$!('($7$E$"%4'(2&7%$()2 8(5&()&(&4"(&7(()&4).'("(/(#4$.#$"!'.7$)#7$(&('=8(5&2'(.2&#&6&$7&.7(#$(D6$5:
j ("%7/"4'&%4##!$($"$"&422'"$(/!$(/*:%:k$(l'.+:1()3&('0&4)%'$#7$" 8'.2$(($"%4#? 8(5&88'()$"82&($":
G&(&
A7$4/$"82&($"&.'(&6&$7&.$7$(/8#&(&
>772&"'4$!(2'($"47'#&#&(&&6&$7&.$7$(/(&(2"(:;)$ (&(2"()'7#!6$#()8775$"%$"82&($"=5)&!!7$4&.7,
H>44 $"4#='"$m'$#"($8$=5.7$"98!'.7$47/&6&$7&.7#&(&(


H>7$(88$%' ()&()&6& 4$&(#&5#&(&

H>#4$!($"8&"/($4($" "#&(&&6&$7&.$7$(/
G&(&()&('!!(()8$"#$"% 8()$ ('#/&&6&$7&.75$()$"()&($47&"#$(1'!!72"(&/t"82&($":1'4G&(&8<$%:q=o=&"#@B("##G&(&<$%
q‚}=&!6$##5$()()!&!:>"/&##$($"&7$"82&($"&"#7&(##&(&&&6&$7&.782()4!"#$"%&'()'!"&"&.7m'(:

0
<A7$&7#7H4((!)"4.$87$5(4  !  ($ " %

12345656762589 56 5317425
)&($().(8$(8/'&4):t8/'&"('=&#()&!!!$&( 4($" .82&9$"%/'74($":
s -$8 4$"4 f)&6$'&7? 4$&74$"4 @47%$4&7=67'($"&/?"6$"2"(&74$"4
<&8"44!/8()#4'2"(5$()&774($"="&(':42D#4'2"(D"H!($"%H'22&/H87&(:!#8

->7$78('#$2'4$(#"$447"()(!'#$"/#  $% "
(6"5)"()#$47'$"%&($6:
1&2!7 $E j$"#$4&($" '!!72"(&/(&.71(&($($4&"#0!#'4$.$7$(/&"B)&'($6 '22&/8&77&2!7 $E&"# (&($($48&4)B!$2"(:
3!#(2$"&($" 8&2!7 $E 5#"z6/B!$2"(&7!($($"$"47'##!&&777$2&%$"%82'7($!7477='7($"%$"
&2!7 $E 82(7/$"()&"%8(" ()'"##:
G&(&B47'$" 5&
G&(&5"7/B47'##$"()4&82$487'$#$4B!$2"(.&#""!H(&.7$)#4$($'2,5)"2$487'$#$4H.&#477!')$"%
'"'44 8'7='4)& 6&767&9&%)$%)477#&2&%=()B!$2"(5& 26#82&"&7/$
0!7$4&($" 04!&7!$4$&('($7&"($"8&787#B&!(&5& '44 8'7:>77#&(&5!7$4&( &(7&({($2= '!!72"(&/(&.71(&($($4&"#0!#'4$.$7$(/8
$2"(:
0&"#2$E&($" 3&"#2$E&($"5& !82#,&77477& $%"#(4"(7&"#B!$2"(&7%'! 5&"&7/E#:
f7$"#$"% t"6($%&(5"(.7$"##(%'!&774&($"82&`$(/8() ('#/& #&(&4774($"&"#&"&7/$5& !82#./() &2
$"#$6$#'&7& $%"$"%()%'!z)56=9/8$"#$"% *:%:477674$($ 5$()$"#$88"(E"+5!#'4#$"&"'".$&#H*.7$"##+
2&""./()'8<$̀$&"#4'(2H2&#&"&7/$ 4$!(:

0j!m'$$"8(2&$"($%8   !  4$8$42& (   $& 7 = / ( 2 & " #2 ( )  #

"82&'()&.'(2(/! 82&($&7=B!$2"(&7/(2 &"#2()# '#$"2&"/('#$:l=$"#$4&(5)()&4)2&($&7=
/(22()#7$(#$76&"((/'('#/:t8/'&"('$8&7$($(2&!!7$ (/'&4)=&#()&!!!$&( 4($".8 74($"%&!":
C&($&7?B!$2"(&7/(2 C()#
"D& t"676#$"() ('#/ "D& t"676#$"() ('#/
s >"($.#$ s )tAHm
s @'9&/($44777$" s <754/(2(/
s A&7&"(7%/ s C0tH.&#"'$2&%$"%
s >"$2&7&"#()%&"$2
s l'2&"&4)!&($4$!&"(
s 7$"$4&7#&(&

>"($.#$
>"($.#$ '# C';&7$"*&"($H!&"H(&7$"=47"}#=;{o}€=1$%2&H>7#$4)=~1>z#$7'($"qDopp+:
C'k>AGl*47"k>q0=&.qouo€=>.4&2=~„z#$7'($"qD{ppp+:
k&(&"($H2't%kl0A4"`'%&(*q€pHxuqxzf$&#=~1>z#$7'($"q,uppp+
v&7$#&($" >77&"($.#$ !"(4224$&7(&"#&#6&7$#&(#&"($.#/2&9=6&7$#&($"#&(&&&6&$7&.7"6"#5.$(:
)((!,DD555:$%2&&7#$4):42D4"("(D#&2D$%2&H&7#$4)D#4D1$%2&DG&(&)(DqD({o}€#&(:!#8
)((!,DD555:&.4&2:42D%&!#)H&"($.#/H%&qH7&#$"%H4"(7H&.qouo€:!#8
)((!,DD555:.$H&#:42D5.(D5.D!#8D7D7$(&('D-t;{€o:!#8

@'9&/($44777$"


A7$4/$"82&($"&.'(4777$"
777$" '4*+ ;)-C0€:u;477)/.$#2&5& &%$8(82>:C:-"""HG'2 "$7*t"($('('$=A&$=<&"4+&"#4&(#./3:
k7&$4)")&' *~"$6$(†(#g>E'=3$4=<&"4+:;)-"($H|‡or{;77-$"5& .(&$"#82;f~1>=827/9"5"
& 7"(4)-&.&($=t"4:*x{oq}p+:A-fHr}u4775.(&$"#82()G1Cˆ*>
q{r+

ƒ
 >'()"($4&($" 777$"&'()"($4&($"5& !82#
"(
" )-"($H|‡or{;.
5& !82#:A-fHr}u*>Hq{r+5& 6&7$#&(#../(
./(
/) 77:<()-C0€:u;477)/.$#2&=03> m'"4$"%
/)G1Cˆ*477'8&42&9=G3>8$"%!$"(+:

12345656762589 56 5317
C/4!7&2&4"(&2$"&($" >774777$" 5((#&"#"%&($682/4!7&2&:
22"7/2$$#"($8$#7$" 3
3422"7/2$H$#"($8#4777$"5'#:
*1t->%$(+
>"$2&7&"#()%&"$2
A7$4/$"82&($"&.'(('#$ $"676$"%&"$2&7z>00tv@%'$#7$" 422"##8!($"%&"$2&7&4)
-&.&(/&"$2&7 x‚qoH59H7#=2&7
82&7j;
j;
-$8&4(Hk<AHB! $"%u€f-Dx2$45'#8*q+!$2&/;47774($"82
7/2!)"# &"# !7"=&"#*o+#"#$($4477#$88"($&($"82."2&5!%"$(:

7425
j$7#&"$2&7 ;) ('#/#$#"($"676#5$7#&"$2&7:
<$7#H4774(# &2!7 ;) ('#/#$#"($"676 &2!74774($"82()8$7#:
@()$46$%)( >77&"$2&7B!$2"(&(t1;>'($&5$
()>'($&"<#&7C$"$(/
$"&
" 44#&"45$()>'($&"7&58&"$2&7B!$2"(:A2$$"5& %&"(#../
814$"4=0&4)&"#@4"2/*$#"($8$4&($"4#,fCj<Hxx:pq}DpppuHttD{.Dopqo+:
3(()&(8'77$"82&($"
"(
" )&!!6&78() ('#/!(472'(&7..! 6$##$$"(
" )2&"'4$!(:
<75/(2(/
A7(
"8$2()&(,
;)&B$7&.7 (&(()2&9&"#87'4)2'#*:%:GH<t;+:
;)&B$ 4&7 &47&7/6$$.7:t"47'#"'2.&7"%&B "7/8.((278(!7(8%'!*
*&g
&%'!g$&"&
" "&7/$8$#"($4&72&9+:
>77!7(&4"('!7(5$()'(7$!'#47!7(:
>"'2$4&76&7'8"'2.8477!4"(&%*5$() (&($($4+$!6$##:
C()#7%/
1&2!7!!&&($" >77477'!"$" 58$7(#()'%)&4(("2)( (
 26477&%%%&( &"#47'2!=" (&$"$"%5& !82#
t"('2"( <>1>$&ttt5& '#8477($"%*5$()"EE7}uŠ2&"#7&7$" }}&"#uxq"2+ +:
18(5& G&(&54774(#5$()<>1G$6&x:q:{:
77!!'7&($"&.'"#&"4 -$8&4(Hk<A-$8&4(H2)/!$($647754774(#:
k&($"% (&(%/ G'.7(&"##.$5B47'##.&# "<
" 1&"#114)&&4($($4:
;$49()$.B( (4
 "8$2()&(&8$%'B2!7$8/$"%()%&($"% (&(%/$ $!6$##$ $"(
" )1'!!72"(&/t"82&($":