nature chemical biology

Article https://doi.org/10.1038/s41589-023-01353-y

Polarized focal adhesion kinase activity

within a focal adhesion during cell migration

Received: 26 October 2022 Xiaoquan Li1,2, Joseph Dale Combs III3, Khalid Salaita 3
& Xiaokun Shu 1,2

Accepted: 3 May 2023

Published online: 22 June 2023 Focal adhesion kinase (FAK) relays integrin signaling from outside to
inside cells and contributes to cell adhesion and motility. However, the
Check for updates
spatiotemporal dynamics of FAK activity in single FAs is unclear due to
the lack of a robust FAK reporter, which limits our understanding of these
essential biological processes. Here we have engineered a genetically
encoded FAK activity sensor, dubbed FAK–separation of phases-based
activity reporter of kinase (SPARK), which visualizes endogenous FAK
activity in living cells and vertebrates. Our work reveals temporal dynamics
of FAK activity during FA turnover. Most importantly, our study unveils
polarized FAK activity at the distal tip of newly formed single FAs in the
leading edge of a migrating cell. By combining FAK–SPARK with DNA tension
probes, we show that tensions applied to FAs precede FAK activation and
that FAK activity is proportional to the strength of tension. These results
suggest tension-induced polarized FAK activity in single FAs, advancing the
mechanistic understanding of cell migration.

Signal transduction from the environment to inside cells has essential so that FAK activity can be reported during FA turnover. And it should
roles in modulating cell adhesion and motility1. Within this signaling have a good temporal resolution that is within the temporal range of FA
pathway, focal adhesion kinase (FAK) is a critical component in sensing dynamics. Although Förster resonance energy transfer (FRET)-based
and integrating extracellular cues to control cell motility2,3. FAK activity FAK activity reporters using autofluorescent green fluorescent protein
regulates dynamic formation of cell adhesions and membrane protru- (GFP) and its variants have been developed6, these FRET-based report-
sions, which controls cell movement2. In particular, FAK influences focal ers suffer from weak signal, with small fluorescence changes upon FAK
adhesion (FA) turnover, including FA assembly and disassembly4. FA is activation, and they lack spatial resolution in imaging FAK activity in
a critical structure in connecting extracellular matrix to intracellular single FAs. Here we designed a versatile and robust FAK activity reporter
structures, including actin filament5. During cell movement, includ- by using a new principle-based approach—GFP phase separation via
ing spreading and migration, integrin–matrix interaction triggers FA multivalent interactions that are induced by FAK activity-dependent
formation, which anchors cells to the substrate and provides traction phosphorylation. We named this FAK reporter FAK–separation of
force for actin polymerization-induced membrane protrusion. There- phases-based activity reporter of kinase (SPARK). FAK–SPARK forms
fore, spatiotemporal control of FAK activity is essential in regulating FA intensely bright droplets upon FAK activation in living cells and verte-
dynamics. To further understand the role of FAK during cell spreading brate animals. FAK–SPARK achieves excellent spatiotemporal resolu-
and migration, it is ideal to visualize real-time FAK activation in single tion and visualizes endogenous FAK activity within single FAs during
FAs with spatial and temporal resolution. their assembly and disassembly.
To achieve spatiotemporal resolution, it is preferred to use a genet-
ically encoded fluorescent reporter in imaging FAK activity. Such a Results
reporter should have large dynamic range, that is large fluorescence Designing a FAK reporter by GFP phase separation
change upon FAK activation, which will enable robust reporting of FAK To develop FAK–SPARK so that we can induce GFP phase separation
activation. It should also provide high spatial resolution in single FAs by active FAK, we designed a multivalent interaction system. First, we

Department of Pharmaceutical Chemistry, University of California San Francisco, San Francisco, CA, USA. 2Cardiovascular Research Institute, University
1

of California San Francisco, San Francisco, CA, USA. 3Department of Chemistry, Emory University, Atlanta, GA, USA. e-mail: xiaokun.shu@ucsf.edu

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1458

Article https://doi.org/10.1038/s41589-023-01353-y

fused a FAK-derived substrate peptide containing tyrosine 397 (Y397) FAK activation follows the assembly of FAs during cell
to enhanced GFP (EGFP). This substrate peptide has been widely used spreading
in the FRET-based biosensors6 and is well known to be specifically phos- We first demonstrated that FAK–SPARK achieved spatiotemporal reso-
phorylated by FAK (and thus we named it FAKsub). Second, we fused lution in detecting FAK activity using the synthetic system by targeting
the phosphopeptide-binding domain SH2 to a nonfluorescent variant the constitutively active FAK (CA–FAK, amino acid 355–690 of FAK) to
of GFP (denoted as EGFP* that contains Y66F mutation). Third, we specific locations in cells10. First, we targeted CA–FAK to the centro-
fused a multivalent tag named homo-oligomeric tag (HOTag) to each some by fusing it to Aurora A kinase and a red fluorescent protein mKO3
of the above constructs, which introduces multivalency (Fig. 1a)7,8. (refs. 11,12). Co-expression of this CA–FAK fusion and FAK–SPARK
Here the hexameric tag homo-oligomeric tag 3 (HOTag3) is fused to revealed GFP droplets in the centrosome (Extended Data Fig. 3a)13. In
FAKsub–EGFP. And a tetrameric tag homo-oligomeric tag 6 (HOTag6) contrast, no FAK–SPARK droplets were observed in the centrosome
is fused to SH2–EGFP*. The two HOTags are de novo-designed coiled of cells without co-expression of the CA–FAK fusion protein. Second,
coils8. Finally, to express the two parts in one construct, we linked the we targeted CA–FAK to nuclear condensates by fusing it to nuclear
two parts with the ‘self-cleaving’ 2A sequence (Supplementary Table 1). homo-oligomeric tag 1 (HOTag1) condensates, tagged with mKO3.
Thus, when FAK is activated, the FAKsub is phosphorylated, resulting in Previously, we showed that HOTag1 formed condensates8. For nuclear
its interaction with SH2 (Fig. 1b). Together with the multivalent HOTags, localization, we fused the nuclear localization signal (NLS) to HOTag1.
this drives phase separation of EGFP through multivalent interaction. Co-expression of the CA–FAK fusion and FAK–SPARK showed green
Specifically, each hexameric FAKsub–EGFP–HOTag3 recruits six SH2– droplets in the nucleus, which are colocalized with the red condensates,
EGFP*–HOTag6, and each tetrameric SH2–EGFP*–HOTag6 recruits four indicating that FAK–SPARK reports FAK activity in the expected locations
FAKsub–EGFP–HOTag3. This multivalent interaction-driven crosslink- (Extended Data Fig. 3b). Together, our data using CA–FAK demonstrate
ing and phase separation lead to phase separation of EGFP, forming that FAK–SPARK enables imaging of FAK activity with spatial resolution
intensely bright droplets (Fig. 1b). in living cells. On the other hand, because the reporter droplets would no
To demonstrate and characterize FAK–SPARK, the reporter was longer have interaction with the active FAK, in locations with intracellular
expressed in HEK293 cells, which showed punctate fluorescence fluid flow, the droplets may drift away from the active FAK, resulting in
in the cytoplasm (Fig. 1c). This is in contrast to the homogeneous loss of spatial resolution. Thus, we suggest that, in general, the formation
fluorescence from a mutant reporter, in which the Y397 was mutated of FAK–SPARK droplets should be monitored to determine the location
to phenylalanine (F) so that it cannot be phosphorylated by FAK. of active FAK (see Fig. 2b and related description).
Because of multivalent interactions in the droplets, this would likely Next, we characterized temporal resolution of FAK–SPARK using
perturb EGFP fluorescence lifetime. Indeed, using fluorescence a small molecule-inducible tagging of CA–FAK to the nuclear HOTag1
lifetime imaging microscopy (FLIM), we found that the EGFP fluo- condensates (Extended Data Fig. 3c). In particular, we fused Frb to the
rescence lifetime in the droplets is shorter than that in the diffuse mKO3-labeled nuclear HOTag1 condensates (HOTag1–NLS–mKO3–Frb)
state (that is homogenously distributed GFP; Supplementary Fig. 1; and fused FKBP to a near-infrared fluorescent protein 2 (IFP2)-labeled
Methods). These data suggest that the multivalent interactions CA–FAK that located in the cytosol (FKBP–IFP2–CA–FAK)14–16. Upon
induced EGFP droplet formation and that FAK–SPARK reports FAK addition of rapamycin 17, CA–FAK translocated into the nucleus
activity in the cells. To further confirm this, we incubated the cells and localized to the HOTag1 condensates (Extended Data Fig. 3c).
with FAK inhibitor PF-562271, which inhibited punctate fluorescence. Time-lapse imaging showed that near-infrared fluorescent droplets
In comparison, incubation with dimethyl sulfoxide (DMSO) did not appeared in the nuclear condensates at around 5 min after the addi-
inhibit the punctate structure of FAK–SPARK. To analyze the imag- tion of rapamycin (Extended Data Fig. 3d). FAK–SPARK formed drop-
ing data in a quantitative way, we defined ‘SPARK signal’, which is the lets at around 6–7 min after addition of rapamycin. Furthermore, the
ratio of summarized fluorescent droplets’ pixel intensity divided near-infrared CA–FAK droplets were colocalized with the red HOTag1
by summarized cells’ pixel intensity (Fig. 1d). Thus, ‘SPARK signal’ condensates, suggesting that CA–FAK were indeed translocated to the
measures the percentage of GFP in the droplet form over the total. HOTag1 condensates as expected via rapamycin-mediated FKBP and
SPARK signal-based analysis confirmed that the FAK inhibitor blocked Frb interaction (Extended Data Fig. 3e). The green FAK–SPARK droplets
droplet formation, and the droplet formation was dependent on Y397 were colocalized with near-infrared CA–FAK droplets, indicating that
phosphorylation (Fig. 1d). Next, we characterized that the reporter FAK–SPARK reports active CA–FAK with spatial resolution (Extended
droplets were reversible upon the addition of the FAK inhibitor. Data Fig. 3e). Quantitative analysis of the time-lapse imaging data
Time-lapse imaging showed that half-to-maximum time (T1/2) was showed that SPARK signal of near-infrared CA–FAK appeared first,
~30 min (Fig. 1e and Supplementary Fig.2). We further showed recov- ~1 min to 2 min earlier than the SPARK signal of the green FAK–SPARK,
ery of reporter droplets after washing out the inhibitor (Extended while SPARK signal of the HOTag1 droplets was relatively stable over
Data Fig. 1). Thus, FAK–SPARK is reversible in reporting FAK activity. time (Extended Data Fig. 3f and Supplementary Video 1). Thus, FAK–
We also confirmed that FAK–SPARK could detect FAK activity in vari- SPARK achieves a temporal resolution of 1–2 min in reporting FAK
ous cancer cells (Extended Data Fig. 2). activation by the CA–FAK.
Finally, we demonstrated that FAK–SPARK was able to detect FAK We next applied it to visualize the dynamics of FAK activity
activity in living vertebrate animals. First, we created a transgenic during cell spreading. We transfected FAK–SPARK into cells, with
zebrafish that is referred to as ‘Tg(UAS–FAK–SPARK)’ in which the co-expression of a red fluorescent protein mApple-fused paxillin that
FAK–SPARK reporter is under the UAS promoter. Next, we crossed it labels FAs18. To induce and visualize cell spreading, we detached and
with the transgenic zebrafish TgBAC (ΔNp63:Gal4FF)la213 so that the reseeded cells 24 h after transfection. Approximately 1–2 h after reseed-
reporter is expressed in the basal epithelial cells in which mCherry is ing, the cells re-attach and spread19. We then conducted time-lapse
also expressed by UAS–mCherry9. Imaging in the tail bud of zebrafish imaging of the cell-spreading process (Fig. 2a). First, cells increased
at 48 h postfertilization showed bright green fluorescent droplets, area outward over time, indicating cell spreading (Fig. 2b). Meanwhile,
whereas mCherry showed homogenous red fluorescence (Fig. 1f). small FAK–SPARK droplets were generated from the leading edge, and
In contrast, the transgenic zebrafish expressing the Y397F variant moved inward, likely driven by the retrograde flow. Zoom-in images
reporter that does not respond to FAK showed homogeneous green showed that the small droplets were produced in the FAs and moved
fluorescence in the same region. These data indicate that FAK–SPARK toward the center of cells (Fig. 2b, inset). Quantitative analysis of the
reports FAK activity in the live and intact zebrafish. Therefore, FAK– imaging data showed that the area of the spreading cell increased
SPARK will be a useful tool for imaging FAK activity in animals. over time and that FAK–SPARK signal also increased. Furthermore, the

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1459

Article https://doi.org/10.1038/s41589-023-01353-y

a FAK–SPARK construct: Y Y = METDDYAEIIDE

EGFP HOTag3 2A SH2 EGFP* HOTag6

b PY
P
Y Phosphopeptide P Y
Y P
Y Y
P Phosphate P
Y P P
P Y Y YP
EGFP Y Y
FAK Y Y Y Y YP
Y P P P P P
HOTag3 (hexamer) Y P
Y
Y YP
Y
SH2 P YP
Y Y
P YP
EGFP*(=EGFP/Y66F) Y PY
Y
HOTag6 (tetramer) PY P

c d
FAK–SPARK Mutant (Y397F) FAK inhibitor (PF-562271) DMSO ΣIdroplets (pixel)
SPARK
ΣIcells (pixel)

*** ***
0.6

SPARK
0.3
10 10 10 10
Norm. fluo.

0
5 5 5 5

1
K

SO
39 K

27
FA PAR

/Y AR
7F

62
DM
0 0 0 0

SP
S

-5
K–

K–
0 30 60 0 30 60 0 40 80 0 40 80

PF
FA
Distance (pixel) Distance (pixel) Distance (pixel) Distance (pixel) FAK–SPARK

e f FAK–SPARK mCherry FAK–SPARK/Y397F mCherry

PF-562271 DMSO

0.6
0.8
***
0.6

SPARK
SPARK

0.4
0.3
0.2

0
0

Y3 K
F
97
AR
0 20 40 60 80

SP
K–
Time (min)

FA
Fig. 1 | Phosphorylation-induced GFP phase separation-based FAK reporter biological replicates, two-sided nonpaired t-test. P = 9.65 × 10−11 between ‘FAK–
visualizes the endogenous activity of FAK in living cells and vertebrates. SPARK and FAK–SPARK/Y397F’ and P = 4.66 × 10−6 between ‘DMSO and PF-562271’.
a, Schematic representation of construct of FAK reporter FAK–SPARK. e, FAK–SPARK is reversible upon the addition of FAK inhibitor, n = 3 biological
b, Working mechanism of FAK–SPARK. c, Top: two left panels,fluorescence replicates. f, Left: schematic representation of zebrafish. Middle: fluorescence
images of cells expressing FAK–SPARK, FAK–SPARK mutant Y397F that cannot be images of transgenic zebrafish expressing FAK–SPARK or FAK–SPARK/Y397F. Tg
phosphorylated by FAK; two right panels, fluorescence images of cells expressing (UAS–FAK–SPARK or FAK–SPARK/Y397F) was crossed with Tg (∆Np63:GAL4) and
FAK–SPARK incubated with FAK inhibitor or DMSO. Bottom: histogram along Tg (UAS–mCherry). Right: quantified SPARK signal, n = 4 biological replicates,
the dashed line. This experiment was repeated three times independently with two-sided nonpaired t-test. P = 0.0008. Data are mean ± s.e.m; ***P < 0.001. Scale
similar results. d, Quantified SPARK signal in cells with various conditions, n = 5 bars, 20 μm (c,f).

increase of FAK–SPARK signal is well correlated with the increase of cel- FAK activity precedes FA disassembly
lular area (Fig. 2b, right). This suggests that during the cell spreading, Because previous studies showed that FAK activity is required for the
FAK is activated, which mainly occurred in the leading edge at early disassembly of FAs4, we imaged FAs and FAK activity in living cells.
spreading (Supplementary Video 2). Time-lapse imaging revealed that before the disassembly of FAs in a
FAK–SPARK droplets were not only produced in the leading edge stationary cell, FAK–SPARK droplets appeared first in the single FAs
(Supplementary Video 3) but also produced during the assembly of (Fig. 3a and Supplementary Video 6). Quantitative analysis showed that
FAs in the area away from the leading edge (Supplementary Video 4). FAK was inactive and became activated before FA disassembly (Fig. 3a,
Time-lapse imaging revealed that following assembly of single FAs, right, and Supplementary Video 7). Our data, thus, indicate that FAK
which were visualized by mApple–paxillin, FAK–SPARK droplets formed activation precedes disassembly events of FAs, which is consistent with
within single FAs (Fig. 2c). Quantitative analysis of the FA growth and previous studies showing essential roles of FAK in FA disassembly20,21.
FAK–SPARK signal within this single FA region showed that FA growth FA ‘sliding’ has also been reported previously22,23, and we observed
precedes the FAK–SPARK signal (Fig. 2c, right), indicating that FAK such events in the spreading cells. FAK–SPARK droplets were produced
activation follows FA assembly. Furthermore, the green droplets were in the sliding FAs (Fig. 3b and Supplementary Video 8). Quantitative
generated in the single FA in a temporal order following the growth analysis indicated that FAK was active during FA sliding events (Fig. 3b,
direction of the single FA (Fig. 2c, time series insets at 108 min and right). At the late stage of cell spreading, the cellular area had little
118.5 min, and Supplementary Video 5). change, but FAK–SPARK droplets were still generated in the edge of cells

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1460

Article https://doi.org/10.1038/s41589-023-01353-y

a i) Transfect FAK–SPARK and ii) 24 h after iii) 1–2 h after

mApple–paxillin into cells transfection, reseeding, cells iv). Time-lapse
imaging
detach and reseed reattach and spread

b 0 min 10 min 12.5 min 90 min 0.6 3

Cell area (a.u.)

FAK–SPARK
mApple–paxillin
FAK–SPARK +

SPARK
0.3 2

Cell area
0 1
Spreading Retrograde flow 0 20 40 60 80
Time (min)
c

Time (min)
FAK–SPARK + 0 min FAK–SPARK mApple–paxillin

98 69.5
mApple–paxillin
0.6
4

FA (a.u.)
SPARK
0 min
0.3
2

108
0 0

118.5
117.5 min 70 80 90 100 110 120
Time (min)

Fig. 2 | FAK–SPARK detects FAK activation during cell spreading and during FA assembly and FAK activation within FAs (middle). Right: normalized
visualizes FAK activity within single FAs after FA assembly. a, Experimental SPARK signal and FA over time. Three independent repetitions of b and c had
procedure. b, Left: time-lapse images during cell spreading. Right: normalized similar results. Scale bars, 10 μm (b); 10 μm (left), 5 μm (middle) and 2 μm
SPARK signal (green) and cell area (red) over time. c, Left: time-lapse images (right) (c).

(Supplementary Video 9). Time-lapse imaging in the periphery of cells These FAK–SPARK droplets’ flow speed is ~0.5 μm min−1 to 1.5 μm min−1,
showed that FAs assembled (Fig. 3c, yellow arrows) and disassembled which is consistent with the reported actin retrograde flow ~0.6 μm min−1
(Fig. 3c, blue arrows) over time and that FAK–SPARK droplets were to 1.8 μm min−1 (ref. 24). Furthermore, these droplets became larger
generated in the FAs undergoing turnover (Fig. 3c and Supplementary on their way toward the cell center because the smaller droplets fuse
Video 10). Therefore, our imaging data showed that FAK–SPARK visual- together during their movement.
ized FAK activity during FA dynamic turnover in cells. We also verified that FAK was distributed along the entire FA using
We also examined whether the expression of the additional sub- a near-infrared IFP2-tagged FAK (Fig. 4e,f)13,15,25. We further used immu-
strate from FAK–SPARK would perturb FA dynamics and cell behavior. nostaining against the endogenous FAK, which indicated that the
Here we characterized the dynamics of FAs, including the FA assem- endogenous FAK was distributed along the entire FA (Supplementary
bly and disassembly rate in FAK–SPARK sensor-expressing cells, and Fig. 4). In contrast, newly formed FAK–SPARK droplets were found in
compared them to the control cells expressing GFP. We found that the distal tip of the newly formed FAs in the leading edge of a migrating
the FA assembly and disassembly rates are similar with no significant cell. Statistical analysis of over 40 single FAs showed that the formation
difference between the FAK–SPARK-expressing cells and the control of FAK–SPARK droplets occurred in the distal tip of these newly formed
(Extended Data Fig. 4). This suggest that FAK–SPARK expression does FAs in the leading edge of migrating cells (Fig. 4g). Furthermore, immu-
not perturb FA assembly and disassembly. To examine if FAK–SPARK nostaining with an antibody against FAK pY397 showed that FAK–SPARK
expression perturbs cell behavior, we characterized cell migration droplet signal colocalized with peak signal of FAK pY397, demonstrat-
rate using a wound-healing assay and found that cells expressing the ing that the sensor reports polarized FAK pY397 signal (Supplementary
reporter showed a similar migration rate in comparison to the control Fig. 5). Thus, our data suggest that although FAK is distributed in both
cells (Extended Data Fig. 5). Taken together, our data, thus, suggest the distal and proximal tip of FAs, the active FAK is mainly found in the
that expression of the reporter does not perturb FA dynamics and cell distal tip of FAs in the leading edge.
behavior such as migration. Based on our results and current understanding of FAK activation
mechanisms, we propose a model (Fig. 4h) to interpret the polarized
Polarized FAK activity in newly formed FA at the leading edge FAK activity in single FAs at the leading edge (Discussion). Briefly,
After late spreading, cells start to migrate19. We imaged FAK activity membrane protrusion is induced by actin polymerization in the lead-
at the leading edge of the migrating cells (Fig. 4a). Briefly, we trans- ing edge of a migrating cell. Actin contraction generates forces, which
fected FAK–SPARK and mApple–paxillin into HeLa cells, which were are likely stronger when closer to the membrane. Through the FAs and
then reseeded and allowed to re-attach and spread. FAK–SPARK visu- integrin–extracellular matrix (for example, fibronectin (FN)) interac-
alized green droplets, which were produced in the distal tip of newly tion, the actin contractibility-generated force is transmitted to FAK,
formed FAs in the leading edge (Fig. 4b–d). Quantitative analysis of which, on the one hand, is connected to actin filament via actin–vincu-
the fluorescence intensity confirmed that FAK–SPARK droplets were lin–paxillin–FAK26 by binding to paxillin via its focal adhesion targeting
distributed in the distal tip of single FAs (Fig. 4d, right). These droplets (FAT) domain, or via actin–Arp2/3–FAK27,28, and on the other hand, is
then dissociated from the FAs and moved toward the cell body due to anchored to the cell membrane via interaction of FERM (F for 4.1 pro-
retrograde flow (Supplementary Video 11). This is likely because, after tein, E for ezrin, R for radixin and M for moesin) with phosphatidylino-
FAK-induced phosphorylation of the reporter’s substrate peptide, sitol 4,5-bisphosphate (PI(4,5)P2)29,30. The tension on FAK dissociates
the phosphorylated substrates will no longer interact with FAK. Thus, the inhibitory FERM domain from the kinase domain, initiating FAK
after phosphorylation, the FAK–SPARK droplets will no longer inter- activation. Our model suggests that the closer an FA is to the membrane
act with FAK, and with the retrograde flow in the leading edge, they protrusion, the stronger the actin contractibility-induced force and
moved toward the cell body. This is also partly due to the relatively the stronger the tension on the FA (and FAK), resulting in polarized
long lifetime of the droplets, which is ~20 min (Supplementary Fig. 3). FAK activity along single FAs at the leading edge of a migrating cell.

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1461

Article https://doi.org/10.1038/s41589-023-01353-y

a FAK–SPARK + mApple–paxillin mApple–paxillin FAK–SPARK + mApple–paxillin FAK–SPARK mApple–paxillin

0 min 0.4 0

0 min 54 min 0 min 54 min

FA (a.u.)
SPARK
0.2 2

30 min 72 min 30 min 72 min

0
4
0 30 60 90 120
46 min 119 min 46 min 119 min
Time (min)

b FAK–SPARK + mApple–paxillin Time (min)

0 11 13.5 17 20 FAK–SPARK mApple–paxillin
0.2 2

FA (a.u.)
SPARK
0.1 1

0 0
0 4 8 12 16 20
Time (min)

c FAK–SPARK + mApple–paxillin Time (min)

5.5 9 16 21.5 22.5
5.5 min 22.5 min

Fig. 3 | FAK is activated before the disassembly of FAs or during sliding and images showing FAK activation during assembly and disassembly of FAs in a late
turnover. a, Left: fluorescence images showing FAK activation within single FAs spreading cell. Three independent repetitions of a–c had similar results. Scale
before disassembly in a stationary cell. Right: normalized SPARK signal and FAs bars, 10 μm (left) and 2 μm (right) (a); 5 μm (left) and 2 μm (right) (b); 10 μm (left)
over time. b, Left: fluorescence images showing FAK activation within single FAs and 3 μm (right) (c).
during sliding. Right: normalized SPARK signal and FAs over time. c, Fluorescence

Polarized FAK activity depends on inhibitory FERM domain blocked FAK activity32. These data demonstrate that FAK activity does
According to the above model, the inhibitory FERM domain is one depend on integrins, which is consistent with the proposed model.
critical component for the polarized activity of FAK in single FAs at Next, we perturbed the actin cytoskeleton and examined whether
the leading edge. To test this, we decided to target CA–FAK to the FAs this affects FAK activity. Incubation with blebbistatin, a myosin II inhibi-
at the leading edge and examine whether FAK–SPARK droplets would tor33, reduced FAK activity (Fig. 5e). Actin polymerization inhibitor,
be generated without polarization. We knocked down endogenous cytochalasin D30, also blocked FAK activity. Furthermore, Rho-kinase
FAK by short hairpin RNA (shRNA) against the c-terminal region of FAK inhibitor, Y27632 (ref. 34), also inhibited FAK activity. We also con-
(between 854 and 860 amino acids; Supplementary Fig. 6). Knockdown firmed the observed reduction of FAK activity by determining FAK
of FAK abolished FAK–SPARK droplets in cells (Supplementary Fig. 6). pY397 levels using western blot analysis, which showed that pY397 level
We then fused CA–FAK to mApple–paxillin and expressed the fusion was largely reduced upon treatment with the drugs (Supplementary
protein, which rescued FAK activity in cells (Supplementary Fig. 6). Fig. 9). These data indicate that actin contraction contributes to FAK
Time-lapse confocal microscopic imaging of single FAs revealed that activation, which is consistent with the previous studies18,30 and the
FAK–SPARK droplets were generated at proximal, center, and distal tip proposed model (Fig. 4h).
of FAs at the leading edge of migrating cells that expressed FA-targeted Finally, we examined several FAK variants that were previously
CA–FAK (Fig. 5a and Supplementary Video 12). Statistical analysis of reported to have altered activity or interaction with the plasma
~40 single FAs showed that FAK activity is no longer polarized at the membrane. We blocked basal FAK activity by co-expressing a
distal tip (Fig. 5b). Furthermore, we confirmed that re-expression of dominant-negative FAK that is comprised of the FRNK domain35,36. Then
the wild-type full-length FAK (wtFAK) in these cells showed polarized we expressed various FAK variants and FAK–SPARK. First, expression
activity of FAK (Supplementary Fig. 7). Interestingly, CA–FAK enhanced of the wild-type (WT) FAK rescued FAK activity as expected. Second,
FA assembly and disassembly rate compared to the wild-type full-length expression of FAK variant Y180A/M183A, which has reduced affinity
FAK (Supplementary Fig. 8). Taken together, our results indicate that between FERM and the kinase domain29, showed higher FAK activity
the FERM domain is critical for the polarized FAK activity, which is than WT FAK, which is consistent with the consensus that the FERM
consistent with the proposed model (Fig. 4h). domain interacts and inhibits the kinase (Fig. 4f). Third, the KAKTLRK
motif at the FERM domain has been characterized to be responsible
FAK activity depends on integrin and actin contractibility for binding to PI(4,5)P2 via the positively charged lysine and arginine
To further support the proposed model, we next examined whether the because mutation of these basic residues to alanine blocked FAK inter-
detected FAK activity is dependent on integrins that bind FN. α5β1 is action with PI(4,5)P2 (ref. 29). Expression of this FAK variant showed
one of the main integrins in HeLa cells that also express αvβ5 and αvβ3 little FAK activity, which is consistent with our model (Fig. 4h). Finally,
(ref. 31). Incubation of cells with α5β1 inhibitors reduced FAK activity expression of the kinase domain from 355 to 690 amino acids (that
in a dose-dependent manner (Fig. 5c,d)32. Antibody against α5β1 also is CA–FAK) showed the strongest FAK activity among the variants

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1462

Article https://doi.org/10.1038/s41589-023-01353-y

a i) Transfect FAK–SPARK and ii) 24 h after iii) 3–5 h after reseeding,

mApple–paxillin into cells transfection, cells reattach, iv) Time-lapse
imaging
detach and re-seed spread and migrate

b mApple–paxillin c mApple–paxillin mApple–paxillin + FAK–SPARK

0 min 0 min 34.5 min 34.5 min 43.5 min

al
im
ox
Pr
al
st
Di

d 39 min 39.5 min 41 min 41.5 min FA

e mApple–paxillin IFP2-FAK Merge/FAK–SPARK

Center Distal (FA)

Norm. fluo. (a.u.)

10
f Proximal tip of FA FA center Distal tip of FA

42 min 42.5 min 43 min 43.5 min 5

Total FAK
mApple–paxillin P Y Y
P

0 FAK–SPARK Active FAK Y

P
Y P P
Y Y P
Y

0 1 2 3 4 P
Y Y
P
Y Y
P
Y
P P
P

Distance (µm)
Y
Y

g h
P Y P
Y P

Newly formed FAs at leading edge Newly formed FAK–SPARK droplet Cell Y

mem
P
P Y

Proximal tip of FA FA center Distal tip of FA bran

e
Retrograde flow
Single FAs (mApple–paxillin)

Protrusion
F-actin

Vinculin etc Actin

Paxillin
FAK FAT P
e as
Kin
Talin

FER
M
Integrin PIP2
α β
0 1 2 3 4 5 6
Distance (µm) ECM (fibronectin)

Fig. 4 | FAK activity is polarized at the distal tip of newly formed FAs in FAK and activated FAK within single FAs. g, Distribution of newly formed FAK–
the leading edge. a, Experimental procedure. b,c, Time-lapse images of cells SPARK droplets in single FAs at the leading edge. h, Proposed model showing
expressing mApple–paxillin and FAK–SPARK. d, Left: fluorescence images of the polarized FAK activation in single FAs at the leading edge of a migrating cell.
boxed area in c. Right: normalized fluorescence intensity over distance along the Three independent repetitions of b–e had similar results. Scale bars, 10 μm (b),
single FA (dashed box at 43.5 min). e, Fluorescence images. f, Distribution of total 5 μm (c), 1 μm (d) and 2 μm (e).

(Fig. 5f). Addition of the FERM domain to this CA–FAK, that is, 1–690 Figs. 10–13). When the hairpin is in a closed state, the fluorophore is
amino acids, substantially reduced FAK activity to a level similar to quenched. Upon hairpin opening, induced by integrin–ligand tension,
that of WT FAK (Fig. 5f). the fluorophore is dequenched and becomes 10–20 times brighter39.
MTFM has a spatial resolution of conventional fluorescence micros-
Integrin–ECM ligand tension drives FAK activation copy and can be conveniently imaged in tandem with the FAK–SPARK
We first showed that integrin–ECM ligand tension precedes FAK activity. signal37. We focused on investigating the initial 20–60 min period of
To investigate the relationship between integrin–ECM mechanical ten- cell-substrate spreading because this involves the formation of nascent
sion events and FAK activity, we needed to simultaneously image integ- FAs in which paxillin is recruited and integrin–substrate traction force
rin tension while also imaging FAK–SPARK activity. We used molecular begins to mount39.
tension fluorescence microscopy (MTFM), which was developed in our DNA hairpin tension probes are threshold reporters and generate
group and has been broadly used to study mechanotransduction37–40. signal when the applied force exceeds the F1/2, which is defined as the
MTFM enables the mapping of piconewton (pN) tension events between equilibrium force that leads to a 50% probability of unfolding41. We
individual integrin and their ligands. We used DNA-based MTFM probes designed tension probes with F1/2 = 19 pN (Supplementary Fig. 12 and
that are highly tunable and are the most sensitive class of force sen- Supplementary Table 2), given that this magnitude of tension is associ-
sors39. These probes are comprised of a folded DNA hairpin structure ated with integrin activation and FA maturation42. Probes presented the
that is tagged with a Cy3B fluorophore and BHQ2 dark quencher pair cyclic Arg-Gly-Asp-d-Phe-Lys (cRGD) FN mimetic peptide that primarily
positioned at the base of the stem region (Extended Data Fig. 6a)41. engages αvβ3 and α5β1 integrins (Extended Data Fig. 6a and Supple-
The probe is assembled from the following three strands: a ligand mentary Fig. 10)43. When IFP2–paxillin and FAK–SPARK transfected
strand that carries the fluorophore, an anchoring strand that presents HeLa cells were plated on MTFM-functionalized surfaces, cells rapidly
the quencher and a hairpin strand with arms complementary to the spread on the surface as observed from reflection interference contrast
ligand and anchor strands (Extended Data Fig. 6a and Supplementary microscopy (RICM). Fluorescence imaging at 20 min after plating cells

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1463

Article https://doi.org/10.1038/s41589-023-01353-y

a 6 min 6.5 min 7 min 7.5 min FA

c 0.4

Proximal Center Distal tip *

SPARK
10 mApple–

Norm fluo. (a.u.)

paxillin– 0.2
CA–FAK *** ***
8 min 8.5 min 9 min 11 min
***
5 ***
*** ***
FAK–SPARK 0
0 –1 –1 –1
nM nM nM ml ml ml
0 1 2 3 4 50 0 0
10 50 0 µg 0 µg 0 µg
Distance (µm) 1 5 10
DMSO α5β1 inhibitor α5 antibody
b CA–FAK: Newly formed FAs at leading edge Newly formed FAK–SPARK droplet
Proximal tip of FA FA center Distal tip of FA FN No FN

d DMSO α5β1 inhibitor

Single FAs (mApple–paxillin)

f Co-expression of dnFAK
0 1 2 3 4 5 6
CA–FAK
Distance (µm)
***

3A
0.2

18
e

/M
DMSO Blebbistatin Cytochalasin D Y27632

0A
0.2

AA
SPARK

K Y18
DMSO

TL
tin

32
nD

AA
76
ista

FA

→A
***
asi
Y-2
SPARK

RK
0.1
bb

FAK
hal

TL
** FAK
Ble

0.1

K KAK
(1–690)
toc

*** *
Cy

FA
Vector
*** *
0
0

Fig. 5 | Polarization of FAK activity depends on FERM domain, integrin and FAK–SPARK, incubated with DMSO or integrin inhibitor. This experiment was
actin cytoskeleton. a, Left: fluorescence images showing no distal polarization repeated five times independently with similar results. e, Left: normalized SPARK
of FAK activity when CA–FAK is fused to paxillin and targeted to FAs in cells in cells incubated with inhibitors of myosin II, actin polymerization and ROCK.
with endogenous FAK knocked down by shRNA against FAK (details in text). Right: representative images, n = 5 biological replicates, two-sided nonpaired
Right: normalized fluorescence over distance along the single FA (boxed area at t-test between DMSO and other treatments, exact P values are (from left to right
11 min). This experiment was repeated three times independently with similar bars) = 0.00015, 9.29 × 10−6 and 0.0017. f, Normalized SPARK in cells expressing
results. b, Distribution of newly formed FAK–SPARK droplets in single FAs at various FAK variants. The cells co-expressed dominant-negative FAK (dnFAK)
the leading edge. c, Normalized SPARK signal in cells incubated with integrin and FAK–SPARK, n = 5 biological replicates, two-sided nonpaired t-test between
inhibitors or antibodies. Cells were grown on fibronectin, n = 5 biological pcDNA and other treatments, exact P values are (from left to right bars) = 0.021,
replicates, two-sided nonpaired t-test between DMSO and other treatments, 3.34 × 10−5, 0.034, 0.034 and 1.80 × 10−5. Data are mean ± s.e.m. *P < 0.05,
exact P values are (from left to right bars) = 0.00086, 0.00018, 9.01 × 10−5, 0.018, **P < 0.01 and ***P < 0.001. Scale bars, 1 μm (a), 10 μm (d) and 15 μm (e).
0.00040, 6.18 × 10−5 and 0.00055. d, Representative images of cells expressing

showed the accumulation of IFP2–paxillin signal and tension signal at to determine the temporal relationship between paxillin recruitment,
the cell edge, confirming that FAs generated tension signal as noted FAK–SPARK activity and integrin tension (Methods). Briefly, we iden-
in prior studies (Extended Data Fig. 6b)39,44,45. The FAK–SPARK channel tified t = 0 min when tension puncta appeared and then recorded
also showed green droplets accumulated at the cell edge near integrin the time delay for the appearance of a FAK–SPARK signal. This analy-
tension and paxillin recruitment (Extended Data Fig. 6b). Our previous sis indicated an average delay time of 21.0 ± 9.0 min between initial
studies and appropriate controls confirm that the tension signal indeed tension recruitment and FAK–SPARK droplet formation. Interest-
measures tension and does not result from some artifact at sites of ingly, the time delay was nearly identical when the analysis was per-
integrin binding to the RGD–DNA substrate (Supplementary Fig. 14)37,39 formed for paxillin recruitment and FAK–SPARK signal (Extended Data
and that the tension signal does not result from spectral bleed-through Fig. 6d,e). This temporal delay exceeds the measured delay of 1–2 min
between channels (Supplementary Fig. 14). To investigate the dynam- (Fig. 2f) between the onset of FAK phosphorylation and the appearance
ics of FAK activation, we next collected time-lapse microscopy images of the FAK–SPARK signal using CA–FAK. Thus, our data show that FAK
in the tension, paxillin and FAK–SPARK channels and subjected them phosphorylation occurs at the sites of FAs following the transmission
to kymograph analysis (Supplementary Fig. 15 and Supplementary of mechanical tension to integrins.
Video 13). These videos identified newly forming adhesions as indi- Next, to further investigate the causal relationship between inte-
cated by paxillin recruitment and appearance of tension. Interestingly, grin tension and FAK activation, we designed a set of experiments that
proximal FAK–SPARK droplets near the sites of FA formation and ten- controlled integrin force levels and then recorded FAK signaling out-
sion accumulation then appeared at a later timepoint (Extended Data comes. Specifically, we anchored the RGD ligands to the tension gauge
Fig. 6c,d). The kymograph analysis was used (Extended Data Fig. 6c,e,f) tether (TGT), which is a double-stranded DNA duplex that dissociates

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1464

Article https://doi.org/10.1038/s41589-023-01353-y

from the substrate at different magnitudes of force, described as the the nucleus, which enables detection of FAK activation dynamics during
tension tolerance (Ttol; Fig. 6a and Supplementary Fig. 21)42,46–48. The FA turnover including assembly and disassembly. On the other hand,
Ttol can be tuned from ~12 pN to 56 pN by altering the geometry of the the long time delay between tension build-up and FAK–SPARK signal
TGT. When the RGD ligand on the top strand is at the same end of the suggests that the temporal resolution of FAK–SPARK likely depends on
duplex as the anchoring group on the bottom strand, then forces lead several factors, including the time for FAK activation and accumulation
to unzipping that is facile and requires ~12 pN of force. Conversely, of sufficient FAK activity that leads to SPARK droplet formation. Taken
when the RGD ligand is at the opposite end of the duplex compared to together, FAK–SPARK provides an excellent tool for spatiotemporal
the surface anchoring group, then applied force re-orient the probe imaging of FAK activity within single FAs in living cells and animals
into a shearing geometry and dehybridization requires large forces using appropriate imaging approaches.
of ~56 pN (Fig. 6a)42,48. The top and bottom strands are modified with Our data suggest that FAK is active where FAs are most dynamic
quencher and fluorophore, respectively, such that mechanical dena- including FA turnover (assembly and disassembly). Previous studies
turation of DNA is quantified by fluorescence (Fig. 6a). Conveniently, show that FAK enhances cell spreading53. Indeed, we observed that FAK
we used total internal reflectance fluorescence (TIRF) microscopy to activation follows FA assembly during cell spreading and migration and
map DNA dehybridization, which is a readout of traction force history. that FAK activation is proportional to cell spreading. We also found
HeLa cells transfected with FAK–SPARK and IFP2–paxillin were FAK activation during FA disassembly, suggesting that FAK activity
plated on TGT surfaces of 12 pN and 56 pN Ttol for 20 min, and then we is likely required for FA disassembly. This is consistent with previous
imaged these three fluorescence channels in single cells to quantify studies that in FAK-deficient mice, cells show larger FAs, in addition to
FA formation, TGT rupture and FAK–SPARK signaling. Consistent with reduced cell motility4. FAK activation and potential phosphorylation
prior reports46, cell spreading and FA formation were limited on the of its substrate proteins in FA may trigger downstream events, result-
12 pN probes, but the TGT signal was greater for these 12 pN probes. ing in FA disassembly. And FAK activation might be caused by tension
Interestingly, FAK–SPARK droplets were more abundant on 56 pN TGTs resulted from the contraction of actin cytoskeleton.
compared to 12 pN TGTs (Fig. 6b,c). We also found that while the FA area The present study showed that FAK activity is polarized along the
was approximately twofold smaller in cells seeded on 12 pN TGT sub- newly formed single FAs at the leading edge of a migrating cell. Interest-
strate than those on 56 pN TGT, the FAK–SPARK activity normalized by ingly, several previous studies suggest that FAK is a tension-activatable
FA area was higher in cells seeded on 56 pN TGT substrate than those on kinase and that the FERM domain of FAK interacts with and inhibits the
12 pN TGT, demonstrating that FAK activity is proportional to the ampli- kinase domain, and that tension-induced pulling dissociates the FERM
tude of the tension even when the activity is normalized to the FA area from the kinase domain, initiating FAK activation29,54–56. Furthermore, a
(Supplementary Fig. 22). This difference in activity was sustained over previous study reported that the traction force on single FAs is skewed
80 min of cell spreading as quantified by image segmentation analysis to the distal tip of FAs (toward the leading edge)57 and that paxillin is
(Supplementary Fig. 23; Methods) demonstrating the persistent influ- more phosphorylated at the distal tip18.
ence of ligand Ttol on FAK activity. Time-lapse imaging of FAK–SPARK Combining these with our observations, we propose a model to
droplets showed retrograde flow for both 12 pN and 56 pN TGTs, and the explain the observed polarization of FAK activity (Fig. 4h). In the lead-
differential FAK signaling and dynamics were sustained over 80 min of ing edge of a migrating cell, actin and actomyosin contraction gener-
imaging (Supplementary Fig. 23 and Supplementary Videos 14 and 15). ates forces, which are likely stronger when closer to the membrane
Moreover, we also quantified phospho-FAK Y397 levels for cells on (Fig. 4h). Transmission of this actin contraction-generated force from
12 pN and 56 pN TGTs. These immunostaining experiments confirm actin cytoskeleton to FAK (via actin–vinculin–paxillin–FAK26) depends
the fidelity of the FAK–SPARK biosensor and showed substantially on the following: (1) interaction of FAK’s FAT domain with paxillin within
enhanced FAK pY397 signal per cell for the 56 pN TGT compared to single FAs that are connected to actin filament; (2) the integrin–ECM
that of the 12 pN probes (Supplementary Fig. 24; Methods). Increased interaction, which anchors the FAs to the substratum and (3) the FERM
FAK–SPARK activity for cells on 56 pN TGTs demonstrates that FAK interaction with the plasma membrane via PI(4,5)P2 (refs.29,30). Com-
activity is mechanosensitive and specifically integrin–ligand forces bining these factors, the tension on FAK is likely stronger toward the
>12 pN lead to enhanced FAK signaling. leading edge, as shown by the previous observations of polarized
tension within single FAs toward the leading edge57. This may lead to
Discussion differential degree of FERM dissociation from the kinase, resulting in
Our phase separation-based FAK activity reporter FAK–SPARK has polarized FAK activities within single FAs, consistent with polarized
several advantages compared to previous FAK reporters30,49,50. First, paxillin phosphorylation18. On the other hand, FAK activity enhances
FAK–SPARK achieves large dynamic range, simple signal pattern and actin polymerization via regulating Rho-family GTPases58. The polar-
high brightness, which enable robust detection of FAK activity in living ized FAK activity may thus direct actin polymerization and contract-
cells and animals. As a comparison, previous genetically encoded FAK ibility via actomyosin, generating positive feedback with the actin
activity reporters such as FRET-based reporters have a small dynamic contractibility-induced tension on FAK and its activation, resulting
range due to weak fluorescence change of the donor and acceptor in membrane protrusion by actin polymerization and directed cell
fluorophores upon FAK activation, which makes them difficult to detect migration. Therefore, our results, together with previous studies, sug-
FAK activation. Second, FAK–SPARK achieves a spatial resolution in gest that polarized tension and FAK activity in single FAs at the leading
single FAs and even within a single FA. This is consistent with another edge may guide cell migration.
GFP phase separation-based reporter for imaging ATM kinase activity To support the proposed model, we conducted various studies
upon DNA damage51,52. In contrast, previous FAK reporters suffer from including mutagenesis and DNA tension probe-based approaches.
poor spatial resolution. For example, they require genetic targeting of We showed that the FAK activity we observed does depend on integrin
FAs to detect FAK activity in FAs30. We emphasize that to achieve the spa- because inhibition of integrin by small molecule inhibitors of α5β1 or
tial and temporal resolution of FAK activity, formation of FAK–SPARK antibodies against α5 abolished FAK activity. Second, inhibitors of actin
droplets should be monitored because of the relatively long lifetime of polymerization, myosin II and Rho-associated protein kinase (ROCK),
the droplets and that they may drift away if there is intracellular fluid reduce FAK activity. These data indicate that FAK activation does
flow, such as actin retrograde flow. Third, FAK–SPARK reporter does depend on the actin cytoskeleton. Our results are consistent with the
not require exogenous expression of FAK, which is advantageous to the current consensus that the actin cytoskeleton propagates force to FAs
other reporters49,50. Fourth, FAK–SPARK achieves fast temporal resolu- via FA connection with actin filament and integrin18. We also verified that
tion within 1–2 min based on the data with expression of active FAK in FAK activity is dependent on the FERM domain’s interaction with PI(4,5)

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1465

Article https://doi.org/10.1038/s41589-023-01353-y

Vinculin etc Paxillin

a

FA

Kinase
FE
T

RM
Talin

TGT rupture
FAK–SPARK activation
12 pN Ttol unzipping
BHQ2 F > 12 pN
αβ

Cy3B

Biotinylated
coverslip
FAK–SPARK

Kinase
P
FAT

FE
RM
56 pN Ttol shearing
αβ

F > 56 pN TGT rupture

FAK–SPARK activation

b TGT rupture IFP2–paxillin FAK–SPARK c 1.5

FAK–SPARK ***
droplets
12 pN

Droplets/FA area (µm–2)

1.0

0.5
56 pN

0
56 pN 12 pN

Fig. 6 | FAK–SPARK droplet formation scales with integrin ligand Ttol. a, times independently with similar results. c, Number of identified FAK–SPARK
Diagram of 12 pN TGT geometry and TGT force induced fluorescence and droplets normalized by FA area in each cell for 12 pN and 56 pN TGTs. n = 10 cells
FAK activation mechanism. b, FAK–SPARK droplet (green outlined in yellow) (four biological replicates), two-sided nonpaired t-test, P = 0.0003. Data are
fluorescence micrographs of HeLa cells transfected with FAK–SPARK and on mean ± s.e.m. ***P < 0.001. Scale bar, 5 μm (b).
12 pN and 56 pN TGTs after 30 min incubation. This experiment was repeated four

P2 because disruption of FERM::PI(4,5)P2 interaction by mutagen- recruitment and integrin-mediated traction force on FN substrates,
esis does reduce FAK activity. And we further showed that FAK inhibi- demonstrating that FAK has a downstream influence on FA traction
tion is dependent on the FERM domain’s association with the kinase force61. Recent single-molecule force spectroscopy measurements
domain of FAK because the reduction of this interaction by mutating key showed that FERM-kinase rupture is associated with a force peak,
residues that mediate this interaction increases FAK activity. suggesting that FAK kinase activity could be modulated by mechani-
It is currently debated as to whether integrin-mediated traction cal tension because the FERM domain is known to inhibit the kinase
force is directly sensed by FAK and triggers its activation, or whether activity and its dissociation could lead to kinase activation62. Perhaps
FAK signaling exclusively promotes the generation of integrin traction the strongest evidence to support that FAK is a mechanosensor comes
forces. For example, it is thought that substrate stiffness does not from a study63 showing that FAK phosphorylation (pY397) and total FAK
affect the rate at which FAK is recruited to FAs59. It was hypothesized correlate with the amplitude of the applied force.
that allosteric changes induced by cell membrane interactions are Our results strongly support that FAK is a mechanosensor that
responsible for the priming of FAK, which fits with a model where FAK detects piconewton integrin force magnitude. Although we do not
activation occurs by force60. Furthermore, multiple studies support provide evidence of the mechanism of mechanosensation, the litera-
the model that FAK is a mechanosensor. FAK undergoes changes in ture suggests that forces drive a conformational change of FAK out of
autophosphorylation in response to changing substrate stiffness for the autoinhibited state, which is akin to the mechanism of how vinculin
FN-coated surfaces30. Micropillar array deflection experiments showed senses tension within FAs. Our data demonstrate that FAK activation
that inhibition of FAK reduces FA paxillin phosphorylation, vinculin occurs within FAs in less than 20 min following the mounting of tension

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1466

Article https://doi.org/10.1038/s41589-023-01353-y

and recruitment of FA markers. Moreover, capping the forces at FAs 18. Case, L. B. & Waterman, C. M. Integration of actin dynamics
using the TGT to values <12 pN shows dampening of FAK activity, thus and cell adhesion by a three-dimensional, mechanosensitive
directly demonstrating a causal link between integrin tension and FAK molecular clutch. Nat. Cell Biol. 17, 955–963 (2015).
activity. Taken together, our data are consistent with a model where 19. McGrath, J. L. Cell spreading: the power to simplify. Curr. Biol. 17,
FAK directly senses integrin tension and responds by phosphorylation R357–R358 (2007).
of its kinase. 20. Tomar, A., Lim, S.-T., Lim, Y. & Schlaepfer, D. D. A
FAK-p120RasGAP-p190RhoGAP complex regulates polarity in
Online content migrating cells. J. Cell Sci. 122, 1852–1862 (2009).
Any methods, additional references, Nature Portfolio reporting sum- 21. Ezratty, E. J., Partridge, M. A. & Gundersen, G. G.
maries, source data, extended data, supplementary information, Microtubule-induced focal adhesion disassembly is mediated
acknowledgements, peer review information; details of author con- by dynamin and focal adhesion kinase. Nat. Cell Biol. 7,
tributions and competing interests; and statements of data and code 581–590 (2005).
availability are available at https://doi.org/10.1038/s41589-023-01353-y. 22. Ballestrem, C., Hinz, B., Imhof, B. A. & Wehrle-Haller, B.
Marching at the front and dragging behind. J. Cell Biol. 155,
References 1319–1332 (2001).
1. Kechagia, J. Z., Ivaska, J. & Roca-Cusachs, P. Integrins as 23. Laukaitis, C. M., Webb, D. J., Donais, K. & Horwitz, A. F. Differential
biomechanical sensors of the microenvironment. Nat. Rev. Mol. dynamics of alpha 5 integrin, paxillin, and alpha-actinin during
Cell Biol. 20, 457–473 (2019). formation and disassembly of adhesions in migrating cells. J. Cell
2. Burridge, K. Focal adhesions: a personal perspective on a half Biol. 153, 1427–1440 (2001).
century of progress. FEBS J. 284, 3355–3361 (2017). 24. Gardel, M. L. et al. Traction stress in focal adhesions correlates
3. Mitra, S. K., Hanson, D. A. & Schlaepfer, D. D. Focal adhesion biphasically with actin retrograde flow speed. J. Cell Biol. 183,
kinase: in command and control of cell motility. Nat. Rev. Mol. Cell 999–1005 (2008).
Biol. 6, 56–68 (2005). 25. Feliks, M., Lafaye, C., Shu, X., Royant, A. & Field, M. Structural
4. Ilic, D. et al. Reduced cell motility and enhanced focal adhesion determinants of improved fluorescence in a family of
contact formation in cells from FAK-deficient mice. Nature 377, bacteriophytochrome-based infrared fluorescent proteins:
539–544 (1995). insights from continuum electrostatic calculations and molecular
5. Parsons, J. T., Horwitz, A. R. & Schwartz, M. A. Cell adhesion: dynamics simulations. Biochemistry 55, 4263–4274 (2016).
integrating cytoskeletal dynamics and cellular tension. Nat. Rev. 26. Le Coq, J., Acebrón, I., Rodrigo Martin, B., López Navajas, P. &
Mol. Cell Biol. 11, 633–643 (2010). Lietha, D. New insights into FAK structure and function in focal
6. Seong, J. et al. Detection of focal adhesion kinase activation at adhesions. J. Cell Sci. 135, jcs259089 (2022).
membrane microdomains by fluorescence resonance energy 27. Swaminathan, V., Fischer, R. S. & Waterman, C. M. The FAK-Arp2/3
transfer. Nat. Commun. 2, 406 (2011). interaction promotes leading edge advance and haptosensing by
7. Chung, C.-I., Zhang, Q. & Shu, X. Dynamic imaging of small coupling nascent adhesions to lamellipodia actin. Mol. Biol. Cell
molecule induced protein–protein interactions in living cells with 27, 1085–1100 (2016).
a fluorophore phase transition based approach. Anal. Chem. 90, 28. Serrels, B. et al. Focal adhesion kinase controls actin assembly via
14287–14293 (2018). a FERM-mediated interaction with the Arp2/3 complex. Nat. Cell
8. Zhang, Q. et al. Visualizing dynamics of cell signaling in vivo Biol. 9, 1046–1056 (2007).
with a phase separation-based kinase reporter. Mol. Cell 69, 29. Goñi, G. M. et al. Phosphatidylinositol 4,5-bisphosphate triggers
334–345.e5 (2018). activation of focal adhesion kinase by inducing clustering and
9. Schepis, A. et al. Protease signaling regulates apical cell conformational changes. Proc. Natl Acad. Sci. USA 111,
extrusion, cell contacts, and proliferation in epithelia. J. Cell Biol. E3177–E3186 (2014).
217, 1097–1112 (2018). 30. Seong, J. et al. Distinct biophysical mechanisms of focal adhesion
10. Zouq, N. K. et al. FAK engages multiple pathways to maintain kinase mechanoactivation by different extracellular matrix
survival of fibroblasts and epithelia: differential roles for paxillin proteins. Proc. Natl Acad. Sci. USA 110, 19372–19377 (2013).
and p130Cas. J. Cell Sci. 122, 357–367 (2009). 31. Orgovan, N. et al. Dependence of cancer cell adhesion kinetics
11. Mastop, M. et al. Characterization of a spectrally diverse set of on integrin ligand surface density measured by a high-throughput
fluorescent proteins as FRET acceptors for mTurquoise2. Sci. Rep. label-free resonant waveguide grating biosensor. Sci. Rep. 4,
20, 11999 (2017). 4034 (2014).
12. Tsutsui, H., Karasawa, S., Okamura, Y. & Miyawaki, A. Improving 32. Sundaram, A. et al. Targeting integrin α5β1 ameliorates severe
membrane voltage measurements using FRET with new airway hyperresponsiveness in experimental asthma. J. Clin.
fluorescent proteins. Nat. Methods 5, 683–685 (2008). Invest. 127, 365–374 (2017).
13. Yu, D. et al. Rational design of a monomeric and photostable 33. Pasapera, A. M., Schneider, I. C., Rericha, E., Schlaepfer, D.
far-red fluorescent protein for fluorescence imaging in vivo. D. & Waterman, C. M. Myosin II activity regulates vinculin
Protein Sci. 25, 308–315 (2015). recruitment to focal adhesions through FAK-mediated paxillin
14. Yu, D. et al. A naturally monomeric infrared fluorescent protein for phosphorylation. J. Cell Biol. 188, 877–890 (2010).
protein labeling. Nat. Methods 12, 763–765 (2015). 34. Labouesse, C., Verkhovsky, A. B., Meister, J.-J., Gabella, C. &
15. Yu, D. et al. An improved monomeric infrared fluorescent Vianay, B. Cell shape dynamics reveal balance of elasticity
protein for neuronal and tumour brain imaging. Nat. Commun. 5, and contractility in peripheral arcs. Biophys. J. 108, 2437–2447
3626 (2014). (2015).
16. Shu, X. et al. Mammalian expression of infrared fluorescent 35. Petridou, N. I., Stylianou, P. & Skourides, P. A. A dominant-negative
proteins engineered from a bacterial phytochrome. Science 324, provides new insights into FAK regulation and function in early
804–807 (2009). embryonic morphogenesis. Development 140, 4266–4276 (2013).
17. Banaszynski, L. A., Liu, C. W. & Wandless, T. J. Characterization of 36. Zheng, Y. et al. FAK phosphorylation by ERK primes Ras-induced
the FKBP.rapamycin.FRB ternary complex. J. Am. Chem. Soc. 127, tyrosine dephosphorylation of FAK mediated by PIN1 and
4715–4721 (2005). PTP-PEST. Mol. Cell 35, 11–25 (2009).

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1467

Article https://doi.org/10.1038/s41589-023-01353-y

37. Blanchard, A. et al. Turn-key mapping of cell receptor force 51. Li, X. et al. ATM-SPARK: a GFP phase separation-based activity
orientation and magnitude using a commercial structured reporter of ATM. Sci. Adv. 9, eade3760 (2023).
illumination microscope. Nat. Commun. 12, 4693 (2021). 52. Shu, X. Imaging dynamic cell signaling in vivo with new classes of
38. Glazier, R., Shinde, P., Ogasawara, H. & Salaita, K. Spectroscopic fluorescent reporters. Curr. Opin. Chem. Biol. 54, 1–9 (2020).
analysis of a library of DNA tension probes for mapping cellular 53. Wakatsuki, T., Wysolmerski, R. B. & Elson, E. L. Mechanics of cell
forces at fluid interfaces. ACS Appl. Mater. Interfaces 13, spreading: role of myosin II. J. Cell Sci. 116, 1617–1625 (2003).
2145–2164 (2021). 54. Bell, S. & Terentjev, E. M. Focal adhesion kinase: the reversible
39. Glazier, R. et al. DNA mechanotechnology reveals that integrin molecular mechanosensor. Biophys. J. 112, 2439–2450 (2017).
receptors apply pN forces in podosomes on fluid substrates. Nat. 55. Zhou, J. et al. Mechanism of focal adhesion kinase
Commun. 10, 4507 (2019). mechanosensing. PLoS Comput. Biol. 11, e1004593 (2015).
40. Stabley, D. R., Jurchenko, C., Marshall, S. S. & Salaita, K. S. 56. Zhou, J., Bronowska, A., Le Coq, J., Lietha, D. & Gräter, F. Allosteric
Visualizing mechanical tension across membrane receptors with regulation of focal adhesion kinase by PIP2 and ATP. Biophys. J.
a fluorescent sensor. Nat. Methods 9, 64–67 (2011). 108, 698–705 (2015).
41. Zhang, Y. et al. Platelet integrins exhibit anisotropic mechano­ 57. Plotnikov, S. V., Pasapera, A. M., Sabass, B. & Waterman, C. M.
sensing and harness piconewton forces to mediate platelet Force fluctuations within focal adhesions mediate ECM-rigidity
aggregation. Proc. Natl Acad. Sci. USA 115, 325–330 (2018). sensing to guide directed cell migration. Cell 151, 1513–1527 (2012).
42. Wang, X. & Ha, T. Defining single molecular forces required 58. Pollard, T. D. & Cooper, J. A. Actin, a central player in cell shape
to activate integrin and notch signaling. Science 340, and movement. Science 326, 1208–1212 (2009).
991–994 (2013). 59. Stutchbury, B., Atherton, P., Tsang, R., Wang, D.-Y. & Ballestrem, C.
43. Pfaff, M. et al. Selective recognition of cyclic RGD peptides of Distinct focal adhesion protein modules control different aspects
NMR defined conformation by αIIbβ3, αVβ3, and α5β1 integrins. of mechanotransduction. J. Cell Sci. 130, 1612–1624 (2017).
J. Biol. Chem. 269, 20233–20238 (1994). 60. Acebrón, I. et al. Structural basis of focal adhesion kinase
44. Liu, Y. et al. Nanoparticle tension probes patterned at the activation on lipid membranes. EMBO J. 39, e104743 (2020).
nanoscale: impact of integrin clustering on force transmission. 61. Panagiotakopoulou, M. et al. Cell cycle-dependent force
Nano Lett. 14, 5539–5546 (2014). transmission in cancer cells. Mol. Biol. Cell 29, 2528–2539 (2018).
45. Liu, Y., Yehl, K., Narui, Y. & Salaita, K. Tension sensing 62. Bauer, M. S. et al. Structural and mechanistic insights into
nanoparticles for mechano-imaging at the living/nonliving mechanoactivation of focal adhesion kinase. Proc. Natl Acad. Sci.
interface. J. Am. Chem. Soc. 135, 5320–5323 (2013). USA 116, 6766–6774 (2019).
46. Rashid, S. A. et al. DNA tension probes show that cardiomyocyte 63. Zhou, D. W. et al. Force-FAK signaling coupling at individual focal
maturation is sensitive to the piconewton traction forces adhesions coordinates mechanosensing and microtissue repair.
transmitted by integrins. ACS Nano 16, 5335–5348 (2022). Nat. Commun. 12, 2359 (2021).
47. Pérez, L. A. et al. An outside-in switch in integrin signaling caused
by chemical and mechanical signals in reactive astrocytes. Front. Publisher’s note Springer Nature remains neutral with regard to
Cell Dev. Biol. 9, 712627 (2021). jurisdictional claims in published maps and institutional affiliations.
48. Jo, M. H., Cottle, W. T. & Ha, T. Real-time measurement of
molecular tension during cell adhesion and migration using Springer Nature or its licensor (e.g. a society or other partner) holds
multiplexed differential analysis of tension gauge tethers. exclusive rights to this article under a publishing agreement with
ACS Biomater. Sci. Eng. 5, 3856–3863 (2019). the author(s) or other rightsholder(s); author self-archiving of the
49. Ritt, M., Guan, J.-L. & Sivaramakrishnan, S. Visualizing and accepted manuscript version of this article is solely governed by the
manipulating focal adhesion kinase regulation in live cells. J. Biol. terms of such publishing agreement and applicable law.
Chem. 288, 8875–8886 (2013).
50. Cai, X. et al. Spatial and temporal regulation of focal adhesion © The Author(s), under exclusive licence to Springer Nature America,
kinase activity in living cells. Mol. Cell. Biol. 28, 201–214 (2008). Inc. 2023

Nature Chemical Biology | Volume 19 | December 2023 | 1458–1468 1468

Article https://doi.org/10.1038/s41589-023-01353-y

Methods 1 μM biliverdin was added for imaging with IFP2 (FAK–SPARK/tension

Plasmid construction and lentivirus production probe experiments).
All plasmid constructs were created and purified by standard molecular
biology techniques and confirmed by exhaustively sequencing the Generation of FAK knockdown HeLa cell line
cloned fragments. To create FAK–SPARK, DNA fragments encoding Lentiviral vector with shRNA for targeting human FAK (CCGATTG-
ETDDYAEIIDE were inserted to replace FKBP in the pcDNA3 FKBP–EGFP– GAAACCAACATATA) was a generous gift from V. Weaver. Lentivirus
HOTag3 construct7,8. DNA sequence encoding SH2 domain was inserted was produced and used to infect HeLa cells. After puromycin selection,
upstream of a nonfluorescent variant of GFP (EGFP* containing Y66F cells were maintained for further use.
mutation) followed by HOTag6. A ‘self-cleaving’ 2A (T2A) sequence was
further added to 3′-downstream of FAK–EGFP–HOTag3 via PCR, and the Zebrafish maintenance and imaging
resulting FAK–EGFR–HOTag3–T2A fragment was then subcloned into Zebrafish were maintained and handled in compliance with the stand-
5′-upstream of SH2–EGFP*–HOTag6 to generate the full FAK–SPARK ard protocols (http://zfin.org) and the University of California San
expressing vector. The tyrosine was changed to phenylalanine (ETDD- Francisco Institutional Animal Care and Use Committee protocols.
PAEIIDE) for the mutFAK–SPARK via designed DNA oligo-synthesis. FAK– WT strains used were EKW. UAS–FAK–SPARK and UAS–mutSPARK
SPARK–NLS was generated via inserting the NLS sequence (PAAKRVKLD) transgenic zebrafish were made with Tol2 system. UAS-zebrafish were
before EGFP and EGFP* in FAK–SPARK. For the lentiviral vector, the DNA mated with TgBAC (ΔNp63:Gal4FF)la213 for skin expression. Imaging was
fragment encoding FAK–SPARK was cloned into pHR–SFFV–GFP1-10 carried out 2 d postfertilization (2 dpf) using Nikon Eclipse Ti inverted
(GFP1-10 fragment) to replace GFP1-10. Lentivirus production followed microscope under ×20 objective.
the standard protocol of ‘Improve Lentiviral Production Using Lipo-
fectamine 3000 Reagent’ from Thermo Fisher Scientific. Phospho-FAK Y397 immunofluorescence microscopy
HeLa cells were incubated for 90 min on TGT surfaces in a full cul-
Cell culture ture medium. Then, the medium was removed and replaced with 4%
The HEK293, 293FT, HeLa, U2OS, MDA–MB-231 and MEF cells were (vol/vol) formaldehyde in PBS and incubated for 15 min to allow fixa-
passaged in Dulbecco’s modified Eagle medium, while Kelly and SHEP tion. Special care was taken to reduce the force fluid shear flow dur-
cells were passaged in Roswell Park Memorial Institute (RPMI)-1640 ing wash steps so as not to dissociate adhered cells from the surface.
medium supplemented with 10% FBS, nonessential amino acids, peni- Formaldehyde solution was then removed, surfaces were washed
cillin (100 units per ml) and streptomycin (100 μg ml−1). All culture with three well volumes of PBS and then incubated with 0.1% (vol/vol)
supplies were obtained from the University of California San Francisco Triton-X in PBS, washed again, then incubated in 100 mg ml−1 bovine
Cell Culture Facility. serum albumin (BSA) for 1 h. Cells were then incubated in 1 μg ml−1
anti-FAK (phospho Y397) antibody EP2160Y in PBS, rabbit monoclonal
Live cell imaging (Abcam, ab81298) for 12 h at 4 °C. Cells were then washed with PBS and
Cells were grown on Nunc Lab-Tek II chambered coverglass for imag- stained with 3 μg ml−1 goat antirabbit IgG Alexa Fluor-488 (Abcam,
ing experiment. To express FAK–SPARK, HEK293 cells were transiently ab150077) in PBS and 50 μl nuc-blue (Invitrogen, R3760; to confirm cell
transfected using calcium phosphate transfection reagent with 50 ng presence via nuclear stain) in PBS for 1 h and then imaged.
FAK–SPARK plasmid. HeLa cells were transiently transfected using
Lipofectamine 3000 transfection reagent with 100 ng FAK–SPARK and Mass spectrometry (MS)
50 ng paxillin–mApple. HEK293 and HeLa cell were imaged 1 d after For synthetic intermediates products (products 1 and 2; Supplemen-
transfection. U2OS, MDA–MB-231, Kelly, SHEP, SKNAS and MEF cells tary Fig. 4), matrix-assisted laser deabsorption ionization (MALDI)
were infected with a medium containing ATM–SPARK lentivirus and time-of-flight (TOF) MS was carried out on an Applied Biosystems 4700
imaged 3 d after infection. Proteomics Analyzer MALDI–TOF mass spectrometer by dissolving the
All imaging was carried out on Nikon Eclipse Ti inverted micro- oligonucleotide at a concentration of 0.1–20 μM in a saturated solution
scope equipped with Yokogawa CSU-W1 confocal scanner unit of 3-hydroxypicolinic acid dissolved in of 49.4% (vol/vol) acetonitrile,
(Andor), digital complementary metal oxide semiconductor cam- 0.1% trifluoroacetic acid and 49.4% water; 5 mg ml−1 ammonium citrate
era ORCA-Flash4.0 (Hamamatsu) and ASI MS-2000 XYZ automated or saturated solution of α-cyano-4-hydroxycinnamic acid in 49.4%
stage (Applied Scientific Instrumentation). Imaging was performed (vol/vol) H2O, 49.4% acetonitrile and 0.1% trifluoroacetic acid. For syn-
in the environmental control unit incubation chamber (In Vivo Sci- thetic products 3–6 (Supplementary Fig. 4), the molecular weight of the
entific) maintained at 37 °C and with 5% CO2. Fluorescence images products was evaluated with an electrospray ionization method using
were acquired using Nikon CFI Plan Apochromatic ×20 dry (numerical a Thermo Fisher Scientific LTQ Orbitrap. The samples were prepared
aperture (NA): 0.75) objective or CFI apochromatic TIRF ×60 oil objec- in the mixture of 70% (18.2 MΩ) nanopure water and 30% methanol
tive (NA: 1.49) against GFP, Cherry and IFP per experiment settings. containing 10 µM ethylenediaminetetraacetic acid, 0.0375% trieth-
ylamine and 0.75% of 1,1,1,3,3,3-hexafluoro-2-propanol and recorded
HeLa cell transfection and reseeding the spectra with negative charge mode eluted with the same solution.
HeLa cells (~80% confluency in a 24-well plate) were transfected with The main peak of obtained electrospray ionization–MS spectrum (m/z)
pcDNA–FAK–SPARK (1 μg) and paxillin–mApple (0.2 μg) or paxillin– was then deconvoluted to obtain the average molecular weight for the
mApple-–CA–FAK or paxillin–IFP2 with lipofectamine. Five hours later, oligonucleotides64.
the medium was changed with fresh. Prepare FN-coated image chamber
by incubating the chamber with 10 μg ml−1 FN at 37° for 1 h followed DNA strand hybridization
by 3× PBS washing. Twenty-four hours after transfection, transfected To anneal MTFM and TGT DNA strands, the respective oligonucleotides
HeLa cells were resuspended with 0.05% trypsin, then washed and col- were suspended in 20 µl PBS at a concentration of 1 μM in a 0.2-ml micro-
lected in a CO2-independent medium (with 10% FBS). After incubation, centrifuge tube. Oligonucleotides were denatured at 95 °C for 5 min and
cells were checked for time-lapse imaging. For drug treatment during subsequently annealed by returning to 25 °C at a rate of 1.3 °C min−1.
reseeding, 1652 (dual inhibitor of α5β1 and αvβ1), P1D6 (anti-integrin
α5 antibody), cytochalasin D, blebbistatin and Y27632 were added Fabrication of 19 pN hairpin MTFM surfaces and imaging
together with resuspended cells into FN-coated chamber and imaged Modified DNA strands were purchased from Integrated DNA Technolo-
for FAKS–SPARK signal at an indicated time with a microscope. In total, gies and functionalized as previously reported (Supplementary Table 1

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

and Supplementary Fig. 13)65. To prepare MTFM probe-modified cov- Microscopy

erslips, 25 mm no. 2 diameter borosilicate glass coverslips (VWRCIR2- Cells were seeded onto functionalized coverslips at a density of
25MM) were mounted on a fluorocarbon rack in a 40 ml beaker and 2–3,000 cells per cm2 and incubated at 37 °C in a 5% CO2 atmosphere
rinsed three times in 40 ml of ethanol. Coverslips were then sonicated for 20–30 min before fluorescence microscopy imaging. Fluorescence
in ethanol for 30 min and then dried in an oven at 100 °C for 20 min. microscopy was carried out using a Nikon Eclipse Ti driven by the NIS
Surfaces were then functionalized with biotin, streptavidin and then an Elements Advanced Research software. The microscope objective
MTFM probe (Supplementary Fig. 13). The cleaned slides were treated used was CFI Apo ×100 (NA: 1.49) objective (Nikon). The optical system
with fresh piranha solution (3:1 vol/vol H2SO4:H2O2) for 10 min. CAU- includes a TIRF variable mirror launcher and a Nikon Perfect Focus
TION: piranha solution is highly corrosive and may violently explode System, an interferometry-based device that corrects the z-drift of
if mixed with organic solvents. The coverslips were then rinsed six the stage. Fluorescence excitation of samples through the optical sys-
times with 40 ml of 18.2 MΩ resistance purity (nanopure) water and tem was carried out with 488 nm (50 mW) and 561 nm (50 mW) lasers.
three times with pure ethanol. Subsequently, slides were incubated Fluorescence micrographs of IFP2–paxillin were acquired using a light
in a solution of 3% (vol/vol) (3-aminopropyl)triethoxysilane (Sigma, emitting diode source passed through a Cy5 fluorescence filter cube
440140-100ML) solution in ethanol for 1 h. Then, the coverslips were (Chroma 49006 filter set). The camera used to collect images was an
washed 3× with ethanol and dried in an oven at 100 °C for 30 min. The Andor DU-897 X-9319 camera imaging at 50–500 ms exposure times.
dry coverslips were then incubated overnight with 2 mg ml−1 NHS-biotin FLIM images were acquired using a Nikon Ti Eclipse Inverted Laser
(Thermo Fisher Scientific, 20217) in anhydrous DMSO in a petri dish at Scanning confocal microscope with a Plan Apo Lambda ×60/1.40 oil
room temperature sealed with parafilm. Biotin-modified coverslips objective equipped with PicoQuant time-correlated single photon
were stored at 4 °C and used within 2 weeks of preparation. On the day counting (TCSPC) upgraded with SymPhoTime 64 2.1.3813 TCSPC.
of FAK–SPARK/tension probe experiments, substrates were rinsed with The average fluorescence lifetime per pixel was determined using the
ethanol, mounted in an Attofluor cell chamber (Thermo Fisher Scien- SymPhoTime Fast FLIM algorithm.
tific, A7816) for microscopy and washed with 3 × 1 ml chamber volumes
of PBS (pH 7.4). The mounted surfaces were incubated with 0.1% BSA in FAK–SPARK particle delay time quantification
PBS for 30 min, washed with 3 ml PBS and then covered with a 50 μg ml−1 Fluorescence microscopy time-lapse videos of the FAK–SPARK and
solution of streptavidin (Rockland, S000-01) in PBS. After incubating IFP2–paxillin transfected cells plated to 19 pN hairpin MTFM surfaces
the coverslips for 1 h at room temperature, they were rinsed with 3 ml were collected. Newly forming adhesions of 19 pN fluorescence ten-
of PBS to remove the unbound streptavidin. The surfaces were then sion signal region of interests were manually selected for line-scan
covered with 0.5 ml of 60 nM annealed MTFM probe DNA in PBS for analysis using the Fiji ImageJ software package (NIH). Each line scan
1 h. Unbound DNA was washed away with 3 ml PBS and 2 ml of normal includes the newly formed 19 pN tension area in the early timepoint
culture media before plating cells on the surfaces. as well as the FAK–SPARK droplet forming in the later timepoint and
Our previous studies demonstrated that the tension signal indeed averages a two pixels width along the direction in the line. Orientation
measures tension and does not result from some artifact at sites of of the line scan is positioned to include both tension-defined FA and
integrin binding to the RGD–DNA substrate. First, our previous study FAK–SPARK droplet at the timepoint of the initial appearance of both
showed that the sensor is evenly distributed by generating probes FA and droplet, and also the droplet and tension area is closer than
that lack the quencher with a ±7% coefficient of variation37,65. Second, 2.5 μm in distance (Extended Data Fig. 6a). Images were then subjected
tension signal shows high coincidence with various FA markers (Sup- to rolling ball background subtraction with a ball radius of 50 pixels to
plementary Fig. 14)65,66. Furthermore, there are areas where adhesions increase contrast and minimize background noise. Kymographs of the
form, but no significant tension has been built up (Supplementary line scans were then generated with length on the y axis, time on the
Fig. 14; paxillin signal is high but without tension). Thus, tension signal x axis and kymograph pixel intensity, the average pixel intensity, of
does not result from spectral bleed-through between channels despite the 2-pixel width of the linear ROI (Extended Data Fig. 6b). Individual
the high level of colocalization between tension and paxillin. kymographs of FAK–SPARK, tension and paxillin were generated and
processed using a python script with the Anaconda package manager.
Fabrication of 12 pN and 56 pN TGT surfaces The maximum intensity on the y axis (distance) of the kymograph is
TGTs of 12 pN and 56 pN were synthesized (Supplementary then plotted versus the x axis (time) of the kymograph (Extended Data
Figs. 15–21)47. Glass coverslips were cleaned with piranha solution as Fig. 6c). The maximum intensity values of the plot of tension, FAK–
above for 30 min and washed with six volumes of nanopure water. SPARK and IFP2–paxillin are then normalized to a scale of 0–1. The time
Then coverslips were etched in 0.5 M KOH on ice in a sonicator for delay is calculated by subtracting the point on the FAK–SPARK trace
30 min. Slides were then washed with six volumes of water and then where the normalized maximum intensity reaches 75% of its maximum
four volumes of ethanol. Coverslips were then dried at 90 °C in an value from that of the tension signal.
oven for 5 min. To reduce nonspecific binding of cells to surfaces,
150 mg ml−1 biotin–PEG–silane (molecular weight (MW) 5 kDa; Broad- FAK–SPARK droplet production quantification
pharm, BP-24037) in anhydrous DMSO was added to the coverslips and To quantify relative FAK–SPARK droplet production on TGTs, an algo-
baked at 90 °C in a glass Petri dish for 15 min67. Coverslips were then rithm using the Fiji ImageJ software package assisted by a macro script
washed with three volumes of nanopure water, dried under a stream was used. Time-lapse micrographs of TIRF488 fluorescence were cap-
of nitrogen gas, stored at 4 °C and used within 2 weeks of fabrication. tured to track FAK–SPARK droplet production using 200 ms exposure
To functionalize surfaces with TGTs, the protocol for MTFM probes time with 100% laser intensity settings every 30 s for 50 min (Supple-
was followed as above but without blocking with BSA. TGT DNA probes mentary Videos 13–15).
were annealed as described above using the synthesized TGT DNA These were subjected to primary background subtracted subtract-
strands. Each replicate cell was plated and imaged in a 96-well round ing half of the median pixel intensity value of the image frame from
bottom microtiter plate (Grace Bio Labs, 204969) with functional- each frame. Each frame was then subjected to rolling ball background
ized coverslips applied to the adhesive bottom. Calculation of F1/2 of subtraction with a radius of 15 pixels. Timelapses were then corrected
DNA hairpin is outlined in Supplementary Fig. 12. The F1/2 threshold for stage drift using the Selective Plane Illumination Microscopy reg-
at which the DNA hairpins have a 50% probability of unfolding and istration69 ImageJ plugin and by selecting regions of immobile fluores-
generating MTFM fluorescence signal is based on assumptions and cence intensity to generate an image a drift-corrected image. A manual
measurements discussed in ref. 68. pixel intensity threshold was then applied then the ‘convert to mask’

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

function was applied to create a binary mask image of regions above the Data availability
threshold intensity. The watershed image segmentation algorithm was All data are available in the article, including the source data section.
applied as described below followed by the ‘analyze particles’ image Source data are provided with this paper.
segmentation function with a maximum particle area threshold that
was varied as described below. The number of identified particles for References
each frame was then recorded. 64. Hail, M. E., Elliott, B. & Anderson, K. High-throughput analysis of
To verify that the identified number of FAK–SPARK droplets oligonucleotides using automated electrospray ionization mass
identified by particle analysis over time does not change drastically spectrometry. Am. Biotechnol. Lab. 22, 12–14 (2004).
for cells for 12 pN and 56 pN TGT. Three permutations of the droplet 65. Zhang, Y., Ge, C., Zhu, C. & Salaita, K. DNA-based digital tension
identification algorithm were employed. One with a low 8 μm2 maxi- probes reveal integrin forces during early cell adhesion. Nat.
mum particle area threshold without applying the watershed image Commun. 5, 5167 (2014).
segmentation operation (settings 1) to the binary mask image result- 66. Liu, Y., Galior, K., Ma, V. P.-Y. & Salaita, K. Molecular tension
ing from intensity thresholding, one with an 8 μm2 maximum area probes for imaging forces at the cell surface. Acc. Chem. Res. 50,
and applying the watershed function (settings 2) and a third using 2915–2924 (2017).
a higher 24 μm2 maximum particle area (settings 3). Modifying the 67. Gidi, Y., Bayram, S., Ablenas, C. J., Blum, A. S. & Cosa, G.
particle identification algorithm settings changed the number of par- Efficient one-step PEG-silane passivation of glass surfaces for
ticles identified for some FAK–SPARK TIRF fluorescent micrographs single-molecule fluorescence studies. ACS Appl. Mater. Interfaces
(Supplementary Fig. 8). We found that FAK–SPARK droplet count did 10, 39505–39511 (2018).
not change substantially for cells over time regardless of the droplet 68. Woodside, M. T. et al. Nanomechanical measurements of the
identification algorithm (Supplementary Fig. 8) through comparing sequence-dependent folding landscapes of single nucleic acid
the number of droplets produced using setting by the unpaired t-test hairpins. Proc. Natl Acad. Sci. USA 103, 6190–6195 (2006).
with Welch’s correction. 69. Preibisch, S., Saalfeld, S., Schindelin, J. & Tomancak, P. Software
for bead-based registration of selective plane illumination
Phospho-FAK Y397 fluorescence quantification microscopy data. Nat. Methods 7, 418–419 (2010).
To measure cell spread area, the RICM outline of the cell was manually
created by creating an ROI of the cell edge and measuring its area. Acknowledgements
For measurements of fluorescence intensity, cell micrographs were We would like to thank S. Narum for carrying out FLIM and analysis,
subjected to background subtraction, bleed-through of fluorescence W. Weiss for sharing cell lines, W. Degrado and D. Sheppard for sharing
dye correction and camera detector background subtraction using a integrin inhibitors and antibodies and O. Weiner and D. Sheppard
Python script. To measure TGT fluorescence, the signal intensity in the for constructive comments. Funding for this work was provided by
average intensity in the RICM ROI in the 561 nm channel was measured. NIH NIGMS R35GM131766 to X.S. and NIH NIGMS R01GM131099 and
To measure phospho-FAK Y397 fluorescence intensity, the average R01GM124472 to K.S.
intensity in the 488 nm channel was measured within the RICM ROI
was measured. Author contributions
X.S. initiated the project. X.L. and X.S. designed the experiments and
Western blot analyzed the data. X.L. conducted the experiments. D.C. and K.S.
Cells were washed three times in cold PBS and lysed in 1× lysis buffer designed the tension experiments and analyzed the data. X.L., D.C.,
(with protease inhibitor and phosphatase inhibitor added) after K.S. and X.S. wrote the paper.
drug treatment. The lysate was further incubated at 4 °C for 40 min
and centrifuged at 15,000g at 4 °C for 20 min. Supernatant was col- Competing interests
lected and mixed with NuPAGE LDS sample buffer (4×). Sample was The authors declare no competing interests.
further resolved in NuPAGE 4–12% Bis–Tris protein gels. After trans-
ferring, the membrane was blocked in PBST containing 5% nonfat Additional information
milk and incubated with first antibodies (anti-pFAK EP2160Y; Abcam, Extended data is available for this paper at
ab81298 or anti-FAK antibody; Cell Signaling Technology, 3285; both https://doi.org/10.1038/s41589-023-01353-y.
600× dilution) at 4 °C overnight. On the following day, membrane
was washed three times in PBST and then incubated in horseradish Supplementary information The online version
peroxidase-conjugated second antibodies (1:1,000) for 1 h at room contains supplementary material available at
temperature. After washing three times with PBST, signal was visual- https://doi.org/10.1038/s41589-023-01353-y.
ized using standard enhanced chemiluminescence substrate on film.
Blots were quantified with ImageJ. Correspondence and requests for materials should be addressed to
Xiaokun Shu.
Statistics
All data were processed with ImageJ (1.53q), GraphPad Prism 8.0, Micro- Peer review information Nature Chemical Biology thanks Daniel
soft Excel (v2208) and the SymPhoTime Fast FLIM algorithm. Lietha, Adam W. Smith and the other, anonymous, reviewer(s) for their
contribution to the peer review of this work.
Reporting summary
Further information on research design is available in the Nature Port- Reprints and permissions information is available at
folio Reporting Summary linked to this article. www.nature.com/reprints.

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

Extended Data Fig. 1 | FAK-SPARK forms droplets after removal of the FAK inhibitor. Data are mean ± SEM (n = 3 biological replicates). Scale bar, 10 μm.

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

Extended Data Fig. 2 | FAK-SPARK is applicable to various cancer cells in for MDA-MB-231, p = 0.00056 for Kelly, p = 0.0026 for MEF, p = 2.64 × 10−8 for
imaging FAK activity using a lentivirus expressing FAK-SPARK. Data are SHEP, p = 0.0019 for SKNAS and p = 8.17 × 10−6 for U2OS). **: p value < 0.01. ***:
mean ± SEM, n = 5 biological replicates, two-sided nonpaired t-test. p = 5.52 × 10−7 p value < 0.001. Scale bar, 10 μm.

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

Extended Data Fig. 3 | FAK-SPARK visualizes FAK activity with FAK activity in the nuclear HOTag1 condensates that are tagged with CA-FAK
spatiotemporal resolution by targeting active FAK into specific locations. (left panels). FAK activity is absent the CA-FAK-absent HOTag1 condensates. a
a Fluorescence images showing that FAK-SPARK visualizes FAK activity in the and b were repeated three times independently with similar results. (c) Cartoon
centrosome when constitutively active FAK (CA-FAK) is fused to aurora A kinase showing experimental procedure of measuring FAK-SPARK temporal resolution.
that is located in the centrosome (left panels). FAK activity is absent in the d Time-lapse images after addition of rapamycin. This experiment was repeated
centrosome without CA-FAK (right panels). b Fluorescence images showing three times with similar results. e Normalized fluorescence along the dash line in
that nucleus-located FAK-SPARK-NLS (nuclear localization signal) visualizes (D). f Normalized SPARK signal over time. Scale bar, 5 μm (a, b, d).

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

Extended Data Fig. 4 | FAK-SPARK expression does not perturb dynamics disassembly rate calculation. f, g, Assembly rate and disassembly rate in HeLa
of focal adhesion assembly and disassembly. a, b, Representative images cells expressing FAK-SPARK or GFP. two-sided nonpaired t-test, p = 0.12 and 0.10
of HeLa cells expressing mApple-paxillin + FAK-SPARK and mApple-paxillin + for f and g respectively. Data represent mean ± SEM (n indicates FA numbers, 305
GFP, respectively. a and b were repeated for three times independently with and 298 for GFP and FAK-SAPRK in f, 331 for both GFP and FAK-SPARK in g). Data
similar results. c, d, Tracking of FAs using Focal Adhesion Analysis Server represent mean ± SEM. n.s., not significant. Scale= 10 μm.
{Steenkiste:2021iv, Berginski:2013dj} in a & b. e, A typical trace of assembly and

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

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

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

Extended Data Fig. 5 | FAK-SPARK expression does not perturb cell migration days (day 2) after scratching. The red line marks cell boundary. a, b, c and d were
during wound healing using wound scratch assay. Wound was induced in repeated three times independently with similar results. e, Quantification of
HeLa cells expressing FAK-SPARK using a scratcher. a, b, Representative images wound healing, n = 3 biological replicates. two-sided nonpaired t-test, p = 0.65
of HeLa cells expressing FAK-SPARK or GFP, respectively. c, d, Typical images and 0.46 for Day 1 and Day 2. Data represent mean ± SEM. n.s., not significant.
showing wound healing right after scratching (day 0), one day (day 1) and two Scale bar, 20 μm for A & B and 200 μm for c & d.

Nature Chemical Biology

Article https://doi.org/10.1038/s41589-023-01353-y

Extended Data Fig. 6 | Integrin-ECM ligand tension precedes FAK activity. fluorescence micrographs of yellow inset in b at different timepoints, yellow
a Diagram of 19 pN DNA-based MTFM probes and FAK-SPARK activation dotted line denotes the linear ROI used for kymograph analysis in (c). Scale bar,
mechanism. b Representative cell RICM (gray) and fluorescence micrographs 2 μm. e Plot of normalized maximum fluorescence over time derived from the
of 19 pN hairpin tension(red), IFP2-Paxillin(magenta), and FAK-SPARK(green) at kymograph, with threshold points used to derive the time delay denoted with
t = 0.0 min. Scale bar, 5 μm. c Overlayed kymographs, yellow line in d, of tension, black circles f Histogram of time delay measurements yielding 21.0 ± 9.0 min
FAK-SPARK, and paxillin channels. Yellow arrows denote the point of tension and average delay between 19 pN integrin tension and FAK-SPARK droplet formation.
paxillin recruitment and FAK-SPARK droplet formation respectively. d Individual n = 19 focal adhesions from 4 cells on 4 different surfaces (4 biological replicates).

Nature Chemical Biology