2022 Liu Nature

Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

Article

Clathrin-associated AP-1 controls


termination of STING signalling

https://doi.org/10.1038/s41586-022-05354-0 Ying Liu1,4, Pengbiao Xu1,4, Sophie Rivara1,4, Chong Liu1, Jonathan Ricci1, Xuefeng Ren2,
James H. Hurley2,3 & Andrea Ablasser1 ✉
Received: 17 February 2022

Accepted: 15 September 2022


Stimulator of interferon genes (STING) functions downstream of cyclic GMP-AMP
Published online: 19 October 2022
synthase in DNA sensing or as a direct receptor for bacterial cyclic dinucleotides and
Open access
small molecules to activate immunity during infection, cancer and
Check for updates immunotherapy1–10. Precise regulation of STING is essential to ensure balanced
immune responses and prevent detrimental autoinflammation11–16. After activation,
STING, a transmembrane protein, traffics from the endoplasmic reticulum to the
Golgi, where its phosphorylation by the protein kinase TBK1 enables signal
transduction17–20. The mechanism that ends STING signalling at the Golgi remains
unknown. Here we show that adaptor protein complex 1 (AP-1) controls the
termination of STING-dependent immune activation. We find that AP-1 sorts
phosphorylated STING into clathrin-coated transport vesicles for delivery to the
endolysosomal system, where STING is degraded21. We identify a highly conserved
dileucine motif in the cytosolic C-terminal tail (CTT) of STING that, together with
TBK1-dependent CTT phosphorylation, dictates the AP-1 engagement of STING.
A cryo-electron microscopy structure of AP-1 in complex with phosphorylated STING
explains the enhanced recognition of TBK1-activated STING. We show that
suppression of AP-1 exacerbates STING-induced immune responses. Our results
reveal a structural mechanism of negative regulation of STING and establish that the
initiation of signalling is inextricably associated with its termination to enable
transient activation of immunity.

After activation and exit from the endoplasmic reticulum (ER), mechanism that explains how STING-dependent immune signalling
STING engages two bifurcating cellular effector pathways. The first is ended.
pathway diverges along STING’s transition to the Golgi and enables We tracked activated STING (phosphorylated at S366; hereafter pST-
autophagy, an ancestral antiviral function of STING; by contrast, ING)) (ref. 17) along its intracellular trafficking route by high-resolution
the second pathway, which begins at the Golgi, promotes the tran- confocal and Airyscan microscopy in HeLa cells. Extending previ-
scriptional activation of innate immune genes—an evolutionarily ous findings20,21, after stimulation with the small-molecule agonist
more recent functional adaptation22,23. Both pathways eventually diABZI-C3 (hereafter, diABZI) (ref. 3), pSTING rapidly (around 0.5 h)
converge at the lysosome, where STING is degraded21,23–25. The initia- appeared at the trans-Golgi network (TGN) and then moved to LAMP1+
tion of STING’s downstream transcription cascade is controlled by a endolysosomal compartments before disappearing (Extended Data
multi-step process: it begins with the recruitment of TBK1, continues Fig. 1a–d). pSTING transited RAB7+ late endosomes, but did not colocal-
with the phosphorylation of STING by TBK1 and results in the engage- ize with EEA1+, a marker of early endosomes (Extended Data Fig. 1c,d).
ment of interferon regulatory factor 3 (IRF3) by phosphorylated Membrane traffic between the TGN and endocytic organelles can
STING17,20,26,27. STING-bound IRF3, in turn, is phosphorylated by TBK1, involve clathrin-coated transport vesicles (CCVs)29. Of note, STING
and translocates to the nucleus to regulate gene expression jointly and pSTING were enriched in CCVs obtained from diABZI-stimulated
with NF-κB and other transcription factors. Through this cascade of cells, but not in those obtained from unstimulated cells (Fig. 1a).
molecular events, STING triggers a broad range of effector functions, Furthermore, pSTING colocalized with clathrin heavy chain, a defin-
most notably the expression of type I interferons (IFNs), proinflam- ing component of CCVs, at TGN46+ compartments in activated cells
matory cytokines and co-stimulatory molecules. Owing to these (Fig. 1b,c and Extended Data Fig. 1e). Stimulated emission deple-
favourable immunostimulatory properties, STING agonists are being tion (STED) super-resolution microscopy confirmed that pSTING is
developed for use as immunotherapeutic agents3,4,28. However, to incorporated into small (around 100-nm diameter) CCVs (Fig. 1d and
optimize robust immune activation while preventing immunopathol- Extended Data Fig. 2a). In correlated light and electron microscopy
ogy, STING responses require strict regulation. Here, we looked for a (CLEM) experiments, we observed that, upon activation, characteristic

Global Health Institute, Swiss Federal Institute of Technology Lausanne (EPFL), Lausanne, Switzerland. 2Department of Molecular and Cell Biology and California Institute for Quantitative
1

Biosciences, University of California, Berkeley, Berkeley, CA, USA. 3Helen Wills Neuroscience Institute, University of California, Berkeley, Berkeley, CA, USA. 4These authors contributed equally:
Ying Liu, Pengbiao Xu, Sophie Rivara. ✉e-mail: andrea.ablasser@epfl.ch

Nature | Vol 610 | 27 October 2022 | 761


Article
a b Merge Merge pSTING CHC TGN46

t
en
m
at

I
BZ
tre
A
o
(kDa)

di
N
50 Flag–
STING
100 pTBK1
50 pSTING CCV
15 AP-1σ

180 CHC c d Merge pSTING CCVs

Manders’ coloc. coeff.


50 Flag– 1.0

(pSTING on signal)

Region 1
STING
100 pTBK1
WCL
0.5
50 pSTING

Region 2
35 GAPDH
0

CHC

TGN46
e

f g 1,600

pSTING area (px2)


siNC
siNC siAP-1 siAP-1
1,200
diABZI: – + – +
(kDa) 800
40 pSTING
STING 400
100 pTBK1
CCV 0
AP-1σ 0 3 6 9 12 15 18 21 24
20
diABZI (h)
180 CHC
h siNC siAP-1
100 pTBK1 diABZI (h) 0 2 4 0 2 4
(kDa) 40 pSTING
40 pSTING
40 STING WCL 40 STING

AP-1σ Ratio: 1.0 0.6 0.6 1.0 0.8 0.7


15
20 AP-1σ
120 Vinculin
40 GAPDH

Fig. 1 | AP-1 loading of STING into CCVs at the TGN. a, Whole-cell lysate (WCL) microscopy (left) or electron microscopy at different Z-heights (three right
and CCV fractions from HeLa cells expressing Flag-tagged STING (HeLaSTING) panels). White arrows indicate CCVs. Scale bars, 0.5 µm. f, WCL and CCV
were analysed by western blot. Clathrin heavy chain (CHC) and GAPDH were fractions from HeLa cells that were treated with non-targeting control (NC)
used as loading and processing controls. b, Airyscan imaging of HeLaSTING cells small interfering RNA (siRNA) or AP-1 siRNA and stimulated with diABZI were
stimulated with diABZI. One representative cell of n = 7 cells. White arrows analysed by western blot. CHC was used as a loading control. g, Quantification
point at occurrences of pSTING. Scale bars, 4 µm (left) or 1 µm (magnified of the area of pSTING in bright-field fluorescent microscopy images of
panels). c, Quantification of the colocalization of pSTING with CHC described HeLaSTING cells that were treated with siRNAs and stimulated with diABZI.
in b, using Manders’ colocalization coefficients. Mean ± s.e.m. of n = 7 cells. Mean ± s.e.m. of n = 3 independent experiments with 99 fields of view per
d, STED images showing pSTING enclosed in CCVs from cells transfected with condition. h, HeLa cells transfected with siRNAs and stimulated with diABZI
mCherry–clathrin and Flag–STING and stimulated with diABZI. Regions 1 and 2 were analysed by western blot. Ratios of STING versus loading control (GAPDH)
are magnified from a large-field-of-view STED image (see Extended Data normalized to the 0-h time point of each condition. One representative
Fig. 2a). Scale bars, 200 nm. e, CLEM of HeLaGFP–STING cells stimulated with example of three (a–c,h) or two (f) independent experiments is shown.
diABZI. The images depict box 3 of Extended Data Fig. 2b in Airyscan

STING perinuclear foci localized to areas that contain multiple CCVs in a marked reduction in the levels of STING, pSTING and pTBK1 within
(Fig. 1e and Extended Data Fig. 2b,c). Together, these results show that CCVs, whereas the overall cellular levels of STING and pSTING were
CCVs can function as transport vehicles of activated STING. increased (Fig. 1f–h and Extended Data Fig. 3a). Crucially, STING traf-
ficking from the ER to the TGN was not affected in AP-1-depleted cells
(Extended Data Fig. 3b). We confirmed that a knockdown of the σ1 subu-
AP-1 sorts STING into CCVs nit alone or a genetic knockout of the µ1 subunit in mouse embryonic
The formation of CCVs relies on heterotetrameric adaptor protein fibroblasts (MEFs) affected the degradation of STING, and that rein-
complexes, which physically connect clathrin with transmembrane troducing the missing subunit restored the decay of activated STING
cargo proteins29–31. The most prominent adaptor protein complex that (refs. 33,34) (Extended Data Fig. 3c,d). These findings suggest that AP-1
is involved in cargo shuttling from the TGN to endolysosomes is AP-1, has a role in controlling activated STING by gating its loading into CCVs.
which consists of the four subunits β1, γ, µ1 and σ1 (ref. 32). Notably, a Adaptor-protein-dependent sorting requires direct engage-
knockdown of three AP-1 subunits (AP1G1, AP1S1 and AP1B1) resulted ment between the multimeric complex and its cargo on target

762 | Nature | Vol 610 | 27 October 2022


a Merge Merge pSTING AP-1γ TGN

Manders’ coloc. coeff.


1.0

(pSTING on signal)
0.5

TGN46
AP-1
b diABZI treatment (h) d
siNC siAP-1
(kDa) 0 1 2 4 6 7
diABZI (h) 0 0.5 1.0 1.5 2.0 4.0 0 0.5 1.0 1.5 2.0 4.0
15 AP-1σ
(kDa) 90 pTBK1
110 AP-1γ IP: Flag
Ratio: 1.0 1.7 1.1 1.0 0.2 2.1 2.8 2.6 2.6 2.0
50 Flag–
90 TBK1
STING
60 pIRF3 60 pIRF3
Ratio: 1.0 1.5 0.8 0.5 0.3 3.2 4.5 4.1 4.0 1.8
15 AP-1σ
WCL 60 IRF3
110 AP-1γ
40 pSTING
50 Flag–
STING Ratio: 1.0 1.1 0.9 0.8 0.6 2.5 2.7 3.2 5.8 3.8
40 pSTING
c
H ZI ent

STING
P
AM
AB tm

Ratio: 1.0 0.8 0.5 0.4 0.4 0.3 1.0 0.8 0.7 0.6 0.7 0.7
G
di rea

90 -1
2′ er
-c
SV
m
t

3′
o
N

(kDa) 110 AP-1γ


110 AP-1γ
IP: Flag 120 Vinculin
Flag–
40 STING

AP-1γ
110
WCL f
Flag– siNC siAP-1
40 STING diABZI (h) 0 2 4 0 2 4
e P = 0.0003
(kDa) 90 pTBK1
P < 0.0001 200 P < 0.0001 120 2,000 P < 0.0001
250
IFNB1 (fold change)

60 pIRF3
IFIT1 (fold change)

IFIT3 (fold change)


IFIT2 (fold change)

200 100
150 1,500
80 40 pSTING
150
100 60 1,000
100 30 STING
40
50 500 Ratio: 1.0 0.8 0.3 1.0 0.9 0.5
50 20
0 0 0 0 15 AP-1σ
diABZI: – + – + diABZI: – + – + diABZI: – + – + diABZI: – + – + 40
GAPDH
siNC siAP-1 siNC siAP-1 siNC siAP-1 siNC siAP-1

Fig. 2 | AP-1 binding directs STING degradation to limit immune activation. diABZI and analysed by western blot. Vinculin was used as a loading control.
a, Airyscan imaging of HeLaSTING cells that were stimulated for 2.5 h with diABZI. e, Induction of IFNB1, IFIT1, IFIT2 and IFIT3 expression was assessed by
The colocalization of pSTING with AP-1γ is quantified by Manders’ quantitative PCR with reverse transcription (RT–qPCR) in HeLa cells
colocalization coefficients. One representative cell is shown, and the transfected with siRNAs and treated with diABZI for 3 h. Ratios of IFNB1, IFIT1,
quantification is the mean ± s.e.m. of n = 7 cells from one out of four IFIT2 and IFIT3 mRNA versus GAPDH mRNA normalized to the untreated groups
independent experiments. White arrows point at occurrences of pSTING. Scale of each condition. Data are mean ± s.d. of three technical replicates. P values
bars, 4 µm (left) or 1 µm (magnified panels). b, HeLaSTING cells were stimulated were obtained by two-tailed Student’s t-test. f, WI-38 human fibroblasts
with diABZI. After immunoprecipitation (IP) with anti-Flag antibody, cells were transfected with siRNAs for three days were stimulated with diABZI and
analysed by western blot. c, HeLaSTING cells were stimulated with diABZI for 2 h, analysed by western blot. GAPDH was used as a loading control. One
infected with HSV-1 (multiplicity of infection (MOI) = 10) for 6 h or transfected representative example of three (a,d–f) or two (b,c) independent experiments
with 90mer dsDNA (1 µg) for 3 h or 1 µM 2’3’-cGAMP for 6 h. After is shown. Ratios of target proteins versus loading control normalized to the 0-h
immunoprecipitation with anti-Flag antibody, samples were analysed by time point of each condition (d,f).
western blot. d, HeLa cells transfected with siRNAs were stimulated with

membranes30,31. Accordingly, pSTING colocalized with AP-1 at the of TBK1 in AP-1-depleted cells, consistent with a model in which
TGN, and STING robustly bound the γ and σ1 subunits of AP-1 after AP-1 functions as the initiator of signalling shutdown upstream of
activation (Fig. 2a,b and Extended Data Fig. 4a). STING associated lysosomes (Extended Data Fig. 4b). Cells that were depleted of AP-1
with AP-1 in response to diverse cellular activators of STING, includ- exhibited an increased type I IFN response, which correlated with a
ing double-stranded DNA (dsDNA), cyclic GMP-AMP (cGAMP) and more potent inhibition of HSV-1 replication (Fig. 2e and Extended
the physiological activator herpes simplex virus 1 (HSV-1), providing Data Fig. 4c). By contrast, suppression of AP-1 mildly decreased type
evidence that AP-1 recognition is an integral element of the cell bio- I IFN responses elicited by triphosphate RNA, a trigger for RIG-I-like
logical regulation of STING (Fig. 2c). helicases, showing that AP-1-mediated negative regulation of innate
immune activation is a specific feature of STING signalling (Extended
Data Fig. 4e). Analysis of various cells, stimuli and read-outs as well
AP-1 restricts STING signalling as single subunit knockdowns showed that AP-1-dependent restric-
We next determined how the disruption of AP-1 affects tion of type I IFN signalling is a universal mechanism for controlling
STING-dependent immune signalling. A knockdown of AP-1 increased the activity of STING (Fig. 2f and Extended Data Figs. 4f–i and 5a–d).
the levels of pTBK1 and pIRF3, in addition to pSTING, and prolonged Stronger STING-dependent type I IFN responses were also appar-
signalling when compared to stimulated control cells (Fig. 2d). ent in µ1A-deficient MEFs as compared to their µ1A-reconstituted
Blocking lysosomal acidification further boosted the activation counterparts (Extended Data Fig. 4j,k). In STING-associated

Nature | Vol 610 | 27 October 2022 | 763


Article
a hSTING b Flag–STING c Flag–STING
TM LBD CTT

ΔCTT

T
LI

LI
W

W
WT

ELI
diABZI – – + +

LI
(kDa)

E
1 134 157 336 379
110 AP-1γ (kDa) 100 pTBK1

MSQEPELLISGMEKPLPLRTDFS 50 Flag–
IP: Flag
Flag–
CCV
50 STING
IRF3 STING
AP-1 TBK1
180 CHC
110 AP-1γ
WCL 100 pTBK1
50 Flag–
STING
50 Flag– WCL
STING
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
N C

120 CHC
d GST AP-1 + – – – + + – + + – + + – + + – e
ARF1 – + – – + – + + – + + – + + – + siNC siTBK1 siIRF3
STING WT – – + – + + + – – – + + + – – – diABZI – + – + – +
STING ELI – – – + – – – + + + – – – + + +
(kDa) (kDa) (kDa) 100 AP-1γ
Flag– IP: Flag
180 180 40
130 130 STING
GST–AP-1γ 100 100
His–AP-1β 70 70 100 TBK1
AP-1μ 55 55
SUMO STING 40 40 60 IRF3
GST tag 35 35
25 25 100 AP-1γ WCL
ARF1
AP-1σ Flag–
15 15
40
10 10 STING
40 GAPDH
M Individual M Input M Pull-down
protein

f WT TBK1 KO IRF3 KO g 0.25


600 nM
300 nM 0.25
STING response (nm)

150 nM
diABZI (h) 0 1 2 4 0 1 2 4 0 1 2 4 0.20 75 nM 0.20
(kDa) 100 pTBK1 0.15 0.15

0.10 0.10
100 TBK1 Kd = 393 nM
0.05 0.05

55 pIRF3 0 0
0 300 600 900 0 200 400 600
IRF3 Time (s) Concentration (nM)
55 0.4 0.4
pSTING response (nm)

40 pSTING 0.3 0.3

0.2 0.2
40 STING 600 nM
300 nM
0.1 150 nM 0.1 Kd = 63 nM
Ratio: 1.0 0.6 0.7 0.4 1.0 1.1 1.0 1.0 1.0 0.7 0.7 0.5 75 nM
0 0
120 Vinculin
0 300 600 900 0 200 400 600
Time (s) Concentration (nM)

Fig. 3 | TBK1-dependent phosphorylation controls the binding of STING to NC siRNA or siRNAs against TBK1 or IRF3 were treated with or without diABZI.
AP-1. a, Schematic diagram of the CTT of human STING (hSTING) and sequence After immunoprecipitation with anti-Flag antibody, samples were analysed by
logo of the CTT as indicated from 50 species. b, HeLa STING-knockout (KO) western blot. GAPDH was used as a loading control. f, HeLa wild-type cells,
cells transfected with Flag-tagged STINGΔCΤΤ (Δ1–341), wild-type (WT) STING, HeLa TBK1 KO cells and HeLa IRF3 KO cells stimulated with diABZI for 0, 1, 2 or
STINGE(E360A), STINGLI(L364A/I365A) or STINGELI(E360A/L364A/I365A) were 4 h were analysed by western blot. Ratios of target proteins versus loading
treated with diABZI for 2 h. After immunoprecipitation with anti-Flag antibody, control normalized to the 0-h time point of each condition. Vinculin was used
samples were analysed by western blot. c, WCL and extracted CCV fractions as a loading control. g, Bio-layer interferometry binding studies of LBD-STING
from HeLa STING KO cells reconstituted with Flag-tagged wild-type STING or (top) or TBK1-phosphorylated LBD-STING (pSTING) (bottom) with AP-1 ΔμCTD.
STINGLI(L364A/I365A) and treated or not with diABZI were analysed by western The right graphs show the binding affinity of STING (top) and pSTING
blot. CHC was used as a loading control. d, Glutathione sepharose pull-down (bottom). One representative example of at least three (b,d,f) or two (c,e,g)
assays of wild-type LBD-STING or LBD-STINGELI by glutathione S-transferase independent experiments is shown.
(GST)-tagged AP-1 core with or without ARF1. e, HeLaSTING cells transfected with

vasculopathy with onset in infancy (SAVI), gain-of-function alleles


of STING1 cause constitutive type I IFN responses12,13. Depletion of Direct engagement of STING by AP-1
AP-1 in fibroblasts from patients with SAVI with distinct hyperac- Adaptor protein interactions depend on the recognition of linear sort-
tive STING variants12,35 led to exaggerated type I IFN signalling and ing signals in the cytosolic tail of transmembrane cargo proteins36–38.
the accumulation of STING (Extended Data Fig. 5e–j). Collectively, Inspection of the cytosolic parts of STING revealed a highly conserved
these results establish AP-1 as a crucial and conserved mediator acidic dileucine-based consensus motif [D/E]XXXL[L/I] (in which X
in balancing STING responses in various biologically relevant denotes any residue) located in the CTT (amino acids 336–379), jux-
contexts, and suggest that the recruitment of AP-1 to STING is the taposed to the PLPLRT/SD binding motif that is necessary for TBK1
initiating event in the termination of immune signalling. recruitment and partially overlapping with the ɸLXIS recruitment

764 | Nature | Vol 610 | 27 October 2022


ARF1
a b
AP-1μ AP-1γ

AP-1σ

90° pSTING tail

AP-1–ARF1 AP-1β

STING
ARF1

c ARF1 d e

AP-1μ AP-1γ I103


R15 L65
V98
E360 P361 F67
L364 L363 V88
I365

AP-1σ E360
G367
pSTING tail E362
Q359 H85
AP-1β pS366
N
C I365
L364

ARF1

f g h Flag–STING

A
66
T
HA–AP-1σ

LR
S3
W
LI
(kDa)
Flag–

8D
V8 S
40 STING
03
T
KR
W

I1
(kDa) CCV
20 AP-1σ
15 HA
pS366
IP: Flag 180 CHC
Flag–
50
STING
Flag–
40 STING
R61 K60 15 HA
WCL 20 AP-1σ WCL
50 Flag–
STING 40 GAPDH

i j Clathrin–AP-1–STING
6A

A
66

cluster formation
W /36

/3

Clathrin
S3 A
S3 A

S3 A
S3 A
58
66
58

58
66
58
T

T
S3

S3
W

AP-1
diABZI – – – – + + + + IRF3
(kDa) 110 AP-1γ T
CTT CCVs
Lysosomal
50 Flag– IP: Flag G
STING degradation
STING

110 AP-1γ TBK1-mediated Immune Signal


WCL phosphorylation activation termination
50 Flag–
STING

Fig. 4 | Structural basis for phospho-regulation of STING recognition by R61A), σ(I103S) and σ(V88D) were stimulated with diABZI for 2 h. Cell lysates
AP-1. a, Three-dimensional (3D) reconstructions of the complex in two different were extracted, immunoprecipitated with anti-Flag antibody and analysed by
orientations. The atomic models of dimeric human LBD-STING (Protein Data western blot. h, WCL and CCV fractions from untreated and diABZI-treated
Bank (PDB) code: 4KSY) and the AP-1 complex (PDB 6DFF) were fitted into the HeLa STING KO cells reconstituted with Flag-tagged wild-type STING and the
maps (grey) through rigid-body docking. b, High-resolution 3D reconstruction indicated STING mutants STINGLI, STINGLR(L374A/R375A) or STING(S358A)
from focused refinement on the AP-1 core and pSTING tail at 2.34-Å resolution were analysed by western blot. CHC and GAPDH were used as loading controls.
contoured at 3σ. c, Ribbon representation of the AP-1 pSTING complex i, HeLa STING KO cells reconstituted with Flag-tagged STING and the indicated
structure. d–f, Detailed views of the binding interface. e, Potentialhydrophobic STING mutants STING(S358A), STING(S366A) or STING(S358/366A) were
interactions around the EXXXLI motif are indicated with dotted lines. The stimulated with diABZI for 2 h or left untreated. After immunoprecipitation
density map (grey mesh) of pSTING tail is contoured at 3σ. f, Potential hydrogen with anti-Flag antibody, samples were analysed by western blot. j, Schematic
bonds around pS366 indicated with dotted lines. The density map (grey mesh) diagram of the function of AP-1 in the termination of STING signalling. One
of the pSTING tail is contoured at 3σ. g, HEK293T cells transfected with representative example of at least three (g–i) independent experiments is
Flag-tagged STING and HA-tagged wild-type AP-1σ or AP-1σ mutants σKR (K60A/ shown.

motif (in which ɸ denotes a hydrophilic residue) for IRF3 (refs. 17–19,27) by stronger IRF3 phosphorylation and more potent inhibition of HSV-1
(Fig. 3a). Of note, mutations of the two hydrophobic residues disrupted replication (Extended Data Figs. 4d and 6b). Given that the isoleucine
the binding of STING to AP-1 in activated cells, whereas mutation of the motif at position 365 (EXXXLI) is also required for the recruitment of
acidic residue alone had a negligible effect (Fig. 3b). A STING mutant IRF3 onto STING (ɸLXIS)17,27, mutagenesis of this position abolished the
devoid of the entire CTT also did not bind AP-1 (Fig. 3b). As expected activation of IRF3 (Extended Data Fig. 6c). Analysing natural variants
from the requirement of AP-1 for cargo loading, disruption of the LI of STING, we identified a missense mutation, p.LL363LF, located in the
motif abolished the incorporation of STING into CCVs and compro- EXXXLI motif that has been reported for human cancer39. Reconstitu-
mised subsequent degradation (Fig. 3c and Extended Data Fig. 6a). tion experiments revealed that the L364F substitution increased the
Moreover, neutralizing the hydrophobic leucine residue at position binding of STING to AP-1, accelerated STING degradation kinetics and
364 (EXXXLI) to alanine exaggerated type I IFN signalling, as shown strongly impaired STING signalling activity (Extended Data Fig. 6d–f).

Nature | Vol 610 | 27 October 2022 | 765


Article
Thus, unique substitutions within the EXXXLI motif influence type I AP-1 (Fig. 4f and Extended Data Fig. 10a). Binding studies validated that
IFN activation thresholds of human STING. substitution of K60 and R61 selectively reduced the AP-1 binding of
To directly analyse the recognition of the STING dileucine motif by pSTING, but not that of native STING, consistent with the importance
AP-1, we reconstituted AP-1 target binding in vitro using the AP-1 core of these two acidic residues in specifying the recognition of pSTING
complex40. Recombinant STING encompassing the ligand-binding (Fig. 3g and Extended Data Fig. 10b). In cells, expression of a K60/R61
domain (LBD) including the CTT (LBD-STING) robustly interacted with mutant or substitutions of residues within the σ subunit that are essen-
ARF1–GTP-activated AP-1 in pull-down assays40, whereas a mutation tial for dileucine recognition abolished interaction between STING and
of the STING dileucine motif (LBD-STINGELI) was unable to bind AP-1 AP-1 (refs. 43,44) Fig. 4g and Extended Data Fig. 10c). Reciprocally, muta-
(Fig. 3d). Wild-type LBD-STING, but not LBD-STINGELI, also bound a tion of S366, but not S358, on STING compromised binding between
version of the AP-1 core in which the µ C-terminal domain is truncated; STING and AP-1 and reduced the sorting of STING into CCVs (Fig. 4h,i).
this variant mimics the open conformation of AP-1 and readily permits Together, these results show how a unique phosphorylation event
interaction with dileucine motifs in the absence of allosteric activation within the C terminus of STING confers differential recognition by AP-1,
by ARF1 (refs. 41,42) (Extended Data Fig. 6g). Together, these results reveal and establish a generalizable mechanism for refining AP-1-based cargo
a direct interaction between STING and the ‛unlocked’ activated AP-1 selection through phospho-regulation.
complex, which is mediated by recognition of the STING dileucine
motif.
Discussion
Our results provide insight into a detailed mechanism that underlies
Phospho-regulation of the binding of STING to AP-1 the negative-feedback control of STING and, together with previous
Owing to the positioning of the EXXXLI motif relative to the signalling work17–19,21, support a model for TGN-compartmentalized regulation of
elements for TBK1 and IRF3 (see above; refs. 17–19), simultaneous interac- STING-dependent immune responses (Fig. 4j). First, STING oligomers
tions between STING, TBK1, IRF3 and AP-1, respectively, are physically bind TBK1, which results in robust phosphorylation of STING CTTs to
impossible. Thus, we considered whether IRF3 and TBK1 could each create a platform for interaction with downstream factors. Whereas
interfere with the AP-1 recognition of STING. Whereas depletion of IRF3 signal activation is dictated by repeating cycles of IRF3 recruitment,
had no effect on the binding of STING to AP-1, depletion of TBK1 strongly phosphorylation and nuclear translocation, signal inactivation is con-
reduced binding, which suggests that rather than weakening, TBK1 trolled by AP-1 recognition, which regulates the sorting of activated
enforces the interaction between STING and AP-1 (Fig. 3e). In agree- STING into nascent clathrin coats. Thus, initiation and shutdown of
ment with this finding, a truncated version of STING that is defective signal transduction at the Golgi are biochemically coupled, thus provid-
for TBK1 recruitment (STINGLR)18,19 was unable to bind AP-1 (Extended ing a molecular strategy to self-limit the immune activation induced
Data Fig. 7a). In addition, TBK1-knockout cells showed compromised by STING.
degradation of activated STING, and reconstitution studies and chemi- Our structural analysis revealed how remodelling of the primary
cal inhibition of TBK1 by BX795 revealed that the effect of TBK1 on STING dileucine binding motif by phosphorylation enables preferential rec-
degradation depends on intact kinase activity (Fig. 3f and Extended ognition of the activated state of STING by AP-1. Coat formation is a
Data Fig. 7b,c). We therefore hypothesized that TBK1-mediated phos- highly cooperative process that depends on the clustering of a large
phorylation might dictate the interaction between STING and AP-1. number of discrete ‘adapted’ cargos to efficiently cross-bridge the
Quantifying in vitro binding between STING and AP-1 showed that clathrin triskelion and facilitate vesicle budding. Thus, even a rela-
although unmodified LBD-STING bound to the open AP-1 core, in vitro tively modest gain of affinity in the sorting interface, as provided by
phosphorylation of STING by TBK1 (pLBD-STING) markedly increased phosphorylation, can markedly affect cargo selection at the cell level.
the apparent affinity (Fig. 3g and Extended Data Fig. 7d). As expected, Notably, previous work in the context of HIV-1 M-Nef has shown that a
LBD-STINGELI showed no binding, whereas a phospho-mimetic STING single phosphorylation event can function to repress dileucine motif
mutation18 increased the binding affinity to a level similar to that seen and tetherin downregulation by AP-1 (ref. 41). This, when taken together
for in-vitro-phosphorylated STING (Extended Data Fig. 7e,f). Thus, with our findings, highlights a considerable level of adaptability, in
TBK1-dependent phosphorylation is crucial for regulating binding which phosphorylation can be harnessed in opposing directions to
enhancement between STING and AP-1, which in the context of cells refine dileucine binding and enable AP-1 to discriminate between dis-
is important for efficient cargo recognition. tinct cargo states.
Despite high conservation in animals, the CTT only emerged in ver-
tebrate STING to enable IFN-based immunity, which complements
Structure of the pSTING–AP-1 complex primordial antiviral effector functions, including autophagy and
To define the mechanism that underlies the enhanced recognition NF-κB signalling22,45,46. In a similar manner, the AP-1-mediated traffick-
of pSTING by AP-1, we next determined the structure of the complex ing process that we describe here, which leads to the elimination of
between the ARF1-activated AP-1 core and TBK1-phosphorylated IFN-inducing STING, complements the autophagy-mediated degrada-
LBD-STING (Fig. 4a, Extended Data Figs. 8 and 9 and Extended Data tion of STING (refs. 24,25). The notable conservation of the AP-1 recruit-
Table 1). A 2.3-Å-resolution cryo-electron microscopy (cryo-EM) ment motif in the STING CTT (Fig. 3a) suggests that, in vertebrates,
reconstruction revealed that STING makes multiple contacts with AP-1 acquisition of the type I IFN signalling module might have evolved
through its C-terminal unfolded loop (residues 359–367) (Fig. 4b–f). At together with its own integrated regulatory system.
the C-terminal part of the loop, STING L364 and I365 of the EXXXLI sort- In summary, by revealing a mechanism of negative feedback at the
ing motif engage L65, F67, H85, V88 and V98 of the σ subunit through TGN, our work fills a gap in knowledge to provide a more complete
hydrophobic interactions (Fig. 4e). At the N-terminal end of the loop, understanding of STING-dependent immunity, and may offer a concep-
the carboxyl group of E360 contacts R15 of the γ subunit through an tual strategy to tune the immunogenic effects of STING for therapeutic
electrostatic interaction (Fig. 4e). Together, these contacts anchor interventions.
STING to AP-1 in a manner similar to that described previously for adap-
tor protein–cargo peptide complexes41,43. Notably, the phospho-moiety
on S366 makes hydrogen bonds with a conserved basic patch formed Online content
by K60 and R61 of the σ subunit, which can function cooperatively with Any methods, additional references, Nature Research reporting sum-
the dileucine motif interface to enforce binding between pSTING and maries, source data, extended data, supplementary information,

766 | Nature | Vol 610 | 27 October 2022


acknowledgements, peer review information; details of author contri- 29. Kirchhausen, T., Owen, D. & Harrison, S. C. Molecular structure, function, and dynamics
of clathrin-mediated membrane traffic. Cold Spring Harb. Perspect. Biol. 6, a016725
butions and competing interests; and statements of data and code avail- (2014).
ability are available at https://doi.org/10.1038/s41586-022-05354-0. 30. Owen, D. J., Collins, B. M. & Evans, P. R. Adaptors for clathrin coats: structure and
function. Annu. Rev. Cell Dev. Biol. 20, 153–191 (2004).
1. Ishikawa, H. & Barber, G. N. STING is an endoplasmic reticulum adaptor that facilitates 31. Traub, L. M. & Bonifacino, J. S. Cargo recognition in clathrin-mediated endocytosis. Cold
innate immune signalling. Nature 455, 674–678 (2008). Spring Harb. Perspect. Biol. 5, a016790 (2013).
2. Burdette, D. L. et al. STING is a direct innate immune sensor of cyclic di-GMP. Nature 478, 32. Heldwein, E. E. et al. Crystal structure of the clathrin adaptor protein 1 core. Proc. Natl
515–518 (2011). Acad. Sci. USA 101, 14108–14113 (2004).
3. Ramanjulu, J. M. et al. Design of amidobenzimidazole STING receptor agonists with 33. Zizioli, D. et al. Early embryonic death of mice deficient in γ-adaptin. J. Biol. Chem. 274,
systemic activity. Nature 564, 439–443 (2018). 5385–5390 (1999).
4. Corrales, L. et al. Direct activation of STING in the tumor microenvironment leads to 34. Meyer, C. et al. μ1A-adaptin-deficient mice: lethality, loss of AP-1 binding and rerouting of
potent and systemic tumor regression and immunity. Cell Rep. 11, 1018–1030 (2015). mannose 6-phosphate receptors. EMBO J. 19, 2193–2203 (2000).
5. Pan, B. S. et al. An orally available non-nucleotide STING agonist with antitumor activity. 35. Lin, B. et al. Case report: novel SAVI-causing variants in STING1 expand the clinical
Science 369, eaba6098 (2020). disease spectrum and suggest a refined model of STING activation. Front. Immunol. 12,
6. Chin, E. N. et al. Antitumor activity of a systemic STING-activating non-nucleotide cGAMP 636225 (2021).
mimetic. Science 369, 993–999 (2020). 36. Ohno, H. et al. Interaction of tyrosine-based sorting signals with clathrin-associated
7. Wu, J. et al. Cyclic GMP-AMP is an endogenous second messenger in innate immune proteins. Science 269, 1872–1875 (1995).
signaling by cytosolic DNA. Science 339, 826–830 (2013). 37. Doray, B., Lee, I., Knisely, J., Bu, G. & Kornfeld, S. The γ/σ1 and α/σ2 hemicomplexes of
8. Ablasser, A. et al. cGAS produces a 2'-5'-linked cyclic dinucleotide second messenger clathrin adaptors AP-1 and AP-2 harbor the dileucine recognition site. Mol. Biol. Cell 18,
that activates STING. Nature 498, 380–384 (2013). 1887–1896 (2007).
9. Gao, P. et al. Cyclic [G(2',5')pA(3',5')p] is the metazoan second messenger produced by 38. Chaudhuri, R., Lindwasser, O. W., Smith, W. J., Hurley, J. H. & Bonifacino, J. S.
DNA-activated cyclic GMP-AMP synthase. Cell 153, 1094–1107 (2013). Downregulation of CD4 by human immunodeficiency virus type 1 Nef is dependent on
10. Diner, E. J. et al. The innate immune DNA sensor cGAS produces a noncanonical cyclic clathrin and involves direct interaction of Nef with the AP2 clathrin adaptor. J. Virol. 81,
dinucleotide that activates human STING. Cell Rep. 3, 1355–1361 (2013). 3877–3890 (2007).
11. Decout, A., Katz, J. D., Venkatraman, S. & Ablasser, A. The cGAS–STING pathway as a 39. Konno, H. et al. Suppression of STING signaling through epigenetic silencing and
therapeutic target in inflammatory diseases. Nat. Rev. Immunol. 21, 548–569 (2021). missense mutation impedes DNA damage mediated cytokine production. Oncogene 37,
12. Liu, Y. et al. Activated STING in a vascular and pulmonary syndrome. N. Engl. J. Med. 371, 2037–2051 (2018).
507–518 (2014). 40. Ren, X., Farías, G. G., Canagarajah, B. J., Bonifacino, J. S. & Hurley, J. H. Structural basis for
13. Jeremiah, N. et al. Inherited STING-activating mutation underlies a familial inflammatory recruitment and activation of the AP-1 clathrin adaptor complex by Arf1. Cell 152, 755–767
syndrome with lupus-like manifestations. J. Clin. Invest. 124, 5516–5520 (2014). (2013).
14. Steiner, A. et al. Deficiency in coatomer complex I causes aberrant activation of STING 41. Morris, K. L. et al. HIV-1 Nefs are cargo-sensitive AP-1 trimerization switches in tetherin
signalling. Nat. Commun. 13, 2321 (2022). downregulation. Cell 174, 659–671 (2018).
15. McCauley, M. E. et al. C9orf72 in myeloid cells suppresses STING-induced inflammation. 42. Jia, X. et al. Structural basis of HIV-1 Vpu-mediated BST2 antagonism via hijacking of the
Nature 585, 96–101 (2020). clathrin adaptor protein complex 1. eLife 3, e02362 (2014).
16. Chu, T.-T. et al. Tonic prime-boost of STING signalling mediates Niemann–Pick disease 43. Kelly, B. T. et al. A structural explanation for the binding of endocytic dileucine motifs by
type C. Nature 596, 570–575 (2021). the AP2 complex. Nature 456, 976–979 (2008).
17. Liu, S. et al. Phosphorylation of innate immune adaptor proteins MAVS, STING, and TRIF 44. Mattera, R., Boehm, M., Chaudhuri, R., Prabhu, Y. & Bonifacino, J. S. Conservation and
induces IRF3 activation. Science 347, aaa2630 (2015). diversification of dileucine signal recognition by adaptor protein (AP) complex variants.
18. Zhao, B. et al. A conserved PLPLRT/SD motif of STING mediates the recruitment and J. Biol. Chem. 286, 2022–2030 (2011).
activation of TBK1. Nature 569, 718–722 (2019). 45. Kranzusch, P. J., Lee, A. S., Berger, J. M. & Doudna, J. A. Structure of human cGAS reveals a
19. Zhang, C. et al. Structural basis of STING binding with and phosphorylation by TBK1. conserved family of second-messenger enzymes in innate immunity. Cell Rep. 3,
Nature 567, 394–398 (2019). 1362–1368 (2013).
20. Mukai, K. et al. Activation of STING requires palmitoylation at the Golgi. Nat. Commun. 7, 46. Margolis, S. R. et al. The cyclic dinucleotide 2'3'-cGAMP induces a broad antibacterial and
11932 (2016). antiviral response in the sea anemone Nematostella vectensis. Proc. Natl Acad. Sci. USA
21. Gonugunta, V. K. et al. Trafficking-mediated STING degradation requires sorting to 118, e2109022118 (2021).
acidified endolysosomes and can be targeted to enhance anti-tumor response. Cell Rep.
21, 3234–3242 (2017). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
22. Margolis, S. R., Wilson, S. C. & Vance, R. E. Evolutionary origins of cGAS–STING signaling. published maps and institutional affiliations.
Trends Immunol. 38, 733–743 (2017).
23. Gui, X. et al. Autophagy induction via STING trafficking is a primordial function of the
Open Access This article is licensed under a Creative Commons Attribution
cGAS pathway. Nature 567, 262–266 (2019).
4.0 International License, which permits use, sharing, adaptation, distribution
24. Konno, H., Konno, K. & Barber, G. N. Cyclic dinucleotides trigger ULK1 (ATG1)
and reproduction in any medium or format, as long as you give appropriate
phosphorylation of STING to prevent sustained innate immune signaling. Cell 155,
credit to the original author(s) and the source, provide a link to the Creative Commons license,
688–698 (2013).
and indicate if changes were made. The images or other third party material in this article are
25. Prabakaran, T. et al. Attenuation of cGAS–STING signaling is mediated by a p62/SQSTM 1‐
included in the article’s Creative Commons license, unless indicated otherwise in a credit line
dependent autophagy pathway activated by TBK1. EMBO J. 37, e97858 (2018).
to the material. If material is not included in the article’s Creative Commons license and your
26. Balka, K. R. et al. TBK1 and IKKε act redundantly to mediate STING-Induced NF-κB
intended use is not permitted by statutory regulation or exceeds the permitted use, you will
responses in myeloid cells. Cell Rep. 31, 107492 (2020).
need to obtain permission directly from the copyright holder. To view a copy of this license,
27. Zhao, B. et al. Structural basis for concerted recruitment and activation of IRF-3 by innate
visit http://creativecommons.org/licenses/by/4.0/.
immune adaptor proteins. Proc. Natl Acad. Sci. USA 113, E3403–E3412 (2016).
28. Corrales, L., McWhirter, S. M., Dubensky, T. W. Jr & Gajewski, T. F. The host STING pathway
at the interface of cancer and immunity. J. Clin. Invest. 126, 2404–2411 (2016). © The Author(s) 2022

Nature | Vol 610 | 27 October 2022 | 767


Article
Methods 5 min and added on top of the well. The next day, the culture medium
was replaced and cells were put under puromycin (5 µg ml−1) selec-
Cell culture tion for three days. Surviving cells were expended in antibiotic-free
HeLa (CCL-2) cells were obtained from Sigma-Aldrich. HEK293T cells medium and sorted by fluorescence-activated cell sorting (FACS) into
were a gift from D. Trono, originally purchased from ATCC (SD-3515). single clones three days later. Growing clones were characterized by
THP-1 cells and WI-38 cells were obtained from ATCC. HaCaT cells were western blotting and selected for the absence of TBK1. HeLa cGAS/
obtained from CLS. Primary human alveolar epithelial cells (epithelial STING KO cells were generated using CRISPR–Cas9 technology. HeLa
cells) were obtained from a commercial supplier (Cell Biologics). MEFs cells were transfected with a pX459-sgcGAS plasmid for 24 h and then
(μ1 KO cells and μ1 KO cells reconstituted with μ1A) were a gift from selected with puromycin for three days. Surviving cells were expanded
P. Schu. Primary fibroblast cells from three patients with SAVI were in antibiotic-free medium and then transfected with the pX458-sgSTING
provided by R. Goldbach-Mansky. HeLa, HEK293T, WI-38, HaCaT and plasmid in the same way for three days and sorted by FACS. The cells
primary fibroblast cells were cultured in Dulbecco’s modified Eagle’s expressing GFP were maintained as HeLa cGAS/STING double KO cells.
medium (DMEM, Thermo Fisher Scientific, 41965039) supplemented Growing clones were characterized by western blotting and selected
with 10% (v/v) heat-inactivated fetal bovine serum (FBS) (Thermo Fisher for the absence of cGAS and STING.
Scientific, Gibco SKU, 10270106), 100 IU ml−1 penicillin–streptomycin
(BioConcept, 4-01F00-H), 2 mM l-glutamine (Thermo Fisher Scientific, Transfection
25030024) and 1 mM sodium pyruvate (BioConcept, 5-60F00-H) at For plasmid transfection, cells were transfected with plasmids and
37 °C and at atmospheric O2 and 5% CO2. THP-1 cells were cultured in Lipofectamine 2000 reagent (Invitrogen, 11668019) (for all imaging
RPMI 1640 medium (Thermo Fisher Scientific, 21875091) supplemented experiments) or GeneJuice transfection reagent (Millipore, 70967)
with 10% FBS, 1× penicillin–streptomycin–l-glutamine (Corning, (for all other experiments) following the manufacturer’s respective
30-009-Cl) and 1× 2-mercaptoethanol (Gibco) at 37 °C and at atmos- protocols. For the siRNA knockdown, 3 × 104 cells were transfected with
pheric O2 and 5% CO2. MEFs were cultured in DMEM (Thermo Fisher Lipofectamine RNAiMAX transfection reagent (Invitrogen, 13778075)
Scientific, 41965039) supplemented with 15% (v/v) heat-inactivated fetal and 40 pmol siRNA, following the manufacturer’s protocol, followed
bovine serum (FBS) (Thermo Fisher Scientific, Gibco SKU, 10270106), by three days of incubation. Medium containing transfection reagents
100 IU ml−1 penicillin–streptomycin (BioConcept, 4-01F00-H) and was replaced with fresh medium 6 h after transfection. Silencer select
1 mM sodium pyruvate (BioConcept, 5-60F00-H) at 37 °C and at atmos- predesigned siRNAs (4390847), siAP1G1 (s1143), siAP1B1 (s1141), siAP1S1
pheric O2 and 5% CO2. Primary human alveolar epithelial cells (epithelial (s3115) and siAP1S3 (s43490) were purchased from Thermo Fisher Scien-
cells) were cultured in complete human epithelial cell medium (Cell tific; siIRF3 and siTBK1 were synthesized by Mircosynth. The sequences
Biologics, H6621), according to the supplier’s instructions. Cell lines of siRNAs are provided in Supplementary Table 1.
were repeatedly tested for mycoplasma by PCR. No method of cell line
authentication was used. Stimulation of cells
Cells were treated with 2.5 μM diABZI (Selleckchem, S8796) and
Plasmids collected at the indicated time points. For cGAMP, 90mer and IVT4
For CRISPR–Cas9 plasmids, single-guide RNA (sgRNA) targeting TBK1, stimulation, 0.1 μM cGAMP (Invivogen), 0.2 μg 90mer or 0.5 μg IVT4
IRF3, cyclic GMP-AMP synthase (cGAS), AP-1σ1 and STING were designed was transfected using Lipofectamine 2000 (Invitrogen, 11668019)
using the web tool CRISPOR47. sgRNAs targeting TBK1, IRF3, cGAS and according to the manufacturer’s protocol, and cells were incubated
AP-1σ1 were cloned in a pSpCas9(BB)-2A-Puro (PX459) V2.0 plasmid for 3 h. The DNA sequences of 90mer is provided in Supplementary
(Addgene, 62988), whereas sgRNAs targeting STING were cloned in a Table 1. Poly(I:C) (Invivogen) was added to the cell medium at a final
pSpCas9(BB)-2A-GFP (PX458) plasmid (Addgene, 48138). Both plasmids concentration of 10 μg ml−1 for 24 h. After infection with the HSV-1 KOS
were gifts from F. Zhang48. The pEF-Bos-based STING truncations (1–341 strain (MOI of 10), infected cells were incubated for 6 h. Pretreatment
and 1–317) and mutations (E360A, LI364/365AA, ELI360/364/365AAA, with BX795 (MedChemExpress) was performed at 2 µM for 24 h and pre-
LR374/375ΑΑ, L364A, L364F and I365A) were obtained by site-directed treatment with bafilomycin A1 (Baf A1; Sigma) was performed at 20 nM
mutagenesis. pEF-Bos-human TBK1-Flag-His was a gift from S. Cer- for 1 h. MEFs treated with 5 μg ml−1 or 40 μg ml−1 DMXAA (Invivogen)
boni. TBK1(S172A) was generated by single-amino-acid mutation. were collected 2 h or 3 h after stimulation. For STING inhibition by H-151,
pCDNA3-HA γ-adaptin 1(AP1G1) (Addgene, 10712) was purchased H-151 (2 µM) was added into cells every 24 h for three days before cells
from Addgene. pCDNA3-HA-AP1S1 was generated by inserting the were examined by RT–qPCR.
coding sequences of AP1S1 flanked by 5’ BamHI and 3’ XhoI sites into
the pCDNA3 vector. AP1G1(R15E), AP1S1(I103S) and AP1S1(V88D) were Antibodies
obtained by single-amino-acid mutagenesis. The primers used for plas- Primary antibodies used: mouse monoclonal anti-vinculin (hVIN-1)
mid constructions and sgRNA sequences are provided in Supplemen- (Sigma-Aldrich, V9264, immunoblot 1:5,000), rabbit monoclonal
tary Table 1. Plasmids for NF-κB-Luc (Promega, E8491) were purchased anti-GAPDH (14C10) (Cell Signaling Technology, 2118, immunoblot
from Promega and those for pIFNβ–GLuc were previously described8. 1:3,000), mouse monoclonal anti-Flag (M2) (Sigma-Aldrich, F1804,
All constructs were confirmed by DNA sequencing. immunoblot 1:3,000), rabbit monoclonal anti-human phospho-STING
(Ser366) (D7C3S) (Cell Signaling Technology, 19781, immunob-
Stable cell lines lot 1:3,000), rabbit monoclonal anti-phospho-TBK1/NAK (Ser172)
HeLa STING KO cells were obtained from F. Martinon49. HeLaSTING cells (D52C2) (Cell Signaling Technology, 5483, immunoblot 1:1,000),
and HeLaGFP–STING cells were generated from HeLa STING KO cells by infec- rabbit monoclonal anti-phospho-IRF3 (Ser386) (EPR2346) (Abcam,
tion with a pTJ lentiviral vector carrying Flag–STING or GFP–STING, ab76493, immunoblot 1:1,000), rabbit monoclonal anti-TBK1/NAK
respectively, and a puromycin resistance gene. Cells were selected (D1B4) (Cell Signaling Technology, 3504, immunoblot 1:1,000), rabbit
with puromycin (2 µg ml−1). HeLa TBK1 KO cells, HeLa RIF3 KO cells polyclonal anti-TMEM173/STING (Proteintech, 19851-1-AP, immunoblot
and HaCaT σ1 ΚΟ cells were generated using CRISPR–Cas9 technology. 1:1,000), rabbit monoclonal anti-IRF3 (D6I4C) (Cell Signaling Technol-
In brief, HeLa cells were plated in six-well culture plates at about 80% ogy, 11904, immunoblot 1:1,000), rabbit monoclonal anti-clathrin
confluency and transfected. Per well, 3.5 µl Lipofectamine 2000 (Life heavy chain (P1663) (Cell Signaling Technology, 2410, immunob-
Technologies, 11668019) and 1 µg plasmid-DNA were each diluted in lot 1:500), mouse anti-clathrin heavy chain monoclonal antibody
125 µl OptiMEM (Life Technologies, 31985047), mixed, incubated for (X22) (Thermo Fisher Scientific, MA1-065, immunofluorescence (IF)
1:100), rabbit polyclonal anti-AP1S1 (Thermo Fisher Scientific, PA5- with 1× loading buffer for 10 min. Samples of 20 μl were loaded into gel
63913, immunoblot 1:1,000), rabbit polyclonal anti-AP1G1 (Thermo after a short centrifugation, and this was followed by SDS–PAGE and
Fisher Scientific, PA5-65290, immunoblot 1:1,000), rabbit polyclonal immunoblotting analysis.
anti-AP1B1 (Sigma-Aldrich, HPA065226, immunoblot 1:1,000), rabbit
polyclonal anti-AP1M1 (Proteintech, 12112-1-AP, immunoblot 1:1,000), Luciferase assay
mouse monoclonal anti-HSV-1 ICP0 (11060) (Santa Cruz, sc-53090, HEK293T cells were plated into 96-well plates and transfected with
immunoblot 1:500), mouse monoclonal anti-HA.11 epitope tag (16B12) non-targeting control or the combination of siRNAs targeting AP1G1,
(Biolegend, MMS-101R, immunoblot 1:2,000). mouse monoclonal AP1S1 and AP1B1 for three days. Cells were transfected using Gene-
γ-adaptin (AP1G1) (100/3) (Sigma-Aldrich, A4200, IF 1:100), mouse Juice transfection reagent (Millipore) with an IFNβ promoter–reporter
monoclonal EEA1 (E9Q6G) (Cell Signaling, 48453, IF 1:100), mouse plasmid (pIFNβ-GLuc) in combination with a STING-expressing plas-
monoclonal LAMP1 (H4A3) (Abcam, ab25630, IF 1:100), rabbit mono- mid (pEF-Bos-Flag-STING). Sixteen hours after transfection, cells
clonal anti-human phospho-STING (Ser366) (D8K6H) (Cell Signaling were stimulated with fresh medium containing 2.5 μM diABZI for 6 h.
Technology, 40818, IF 1:100, STED 1:50), mouse monoclonal RAB7 Gaussia luciferase activity was measured in the supernatants using
(E9O7E) (Cell Signaling Technology, 95746, IF 1:100) and sheep poly- coelenterazine (PJK GmbH) as substrate. For the measurement of
clonal human TGN46 (Bio-Rad, AHP500G, IF 1:200). HRP-conjugated NF-κB promoter luciferase activity, cells were transfected with siR-
secondary antibodies used: donkey anti-rabbit IgG (H+L)-HRP ( Jackson NAs as described before and then transfected using GeneJuice with a
ImmunoResearch, 711-036-152, immunoblot: 1:5,000) and donkey NF-κB promoter–reporter plasmid (NF-κB-Luc). After 16 h, cells were
anti-mouse IgG (H+L)-HRP ( Jackson ImmunoResearch, 715-036-151, stimulated with fresh medium containing 2.5 μM diABZI for 18 h. The
immunoblot: 1:5,000). Fluorescence-conjugated secondary antibod- promoter activity was determined using the Bright-Glo Luciferase
ies used: goat anti-mouse IgG2a cross-adsorbed secondary antibody, Assay System (Promega). The expression of proteins was confirmed
Alexa Fluor 647-conjugated (Invitrogen, A-21241, IF 1:800), donkey by immunoblotting. For determining the number of viable cells in
anti-sheep IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor culture, cells were seeded into a 96-well plate and were measured with
488-conjugated (Invitrogen, A-11015, IF 1:800), goat anti-rabbit IgG CellTiter- Glo Luminescent Cell Viability Assay System (Promega) every
(H+L) cross-adsorbed secondary antibody, Alexa Fluor 568-conjugated 24 h following the producer’s instructions.
(Invitrogen, A-11011, IF 1:800) and goat anti-rabbit IgG F(ab) ATTO647N
(H+L) (Hypermol,2318, IF 1:500). Antibody details are provided in CCV extraction
Supplementary Table 1. CCV extraction was performed according to a previously described pro-
tocol50. In brief, cells from one confluent 500-cm2 dish were treated with
RT–qPCR 2.5 μM diABZI for 2 h and then rinsed with PBS twice. Cells were scraped
Cells were lysed in the RLT buffer (Qiagen). RNA was extracted following into 5 ml buffer A (0.1 M MES, pH 6.5, 0.2 mM EGTA, 0.5 mM MgCl2) and
the manufacturer’s protocol (Qiagen RNeasy Plant Mini Kit). RNA was homogenized by pipetting up and down more than 25 times using a 5-ml
reverse-transcribed using the RevertAid First Strand cDNA synthesis syringe with a 22-G needle attached. Cell lysates were centrifuged at
Kit (Thermo Fisher Scientific) and analysed by RT–qPCR in triplicate 4,100g, 4 °C for 32 min. Supernatants were moved into new tubes and
or quadruplicate using the ChamQ Universal SYBR qPCR Master Mix treated with 50 μg ml−1 ribonuclease A on ice for 30 min followed by
(Vazyme). The qPCR reactions were run on a QuantStudio 7 Real-Time centrifugation at 50,000 rpm, 4 °C for 30 min using a Type 70 Ti rotor
PCR system (Thermo Fisher Scientific). GAPDH was used as a house- (Beckman Coulter). Pellets were resuspended in 300 μl buffer A and
keeping gene for normalization. Primer sequences are provided in mixed with an equal volume of buffer B (12.5% (w/v) Ficoll, 12.5% (w/v)
Supplementary Table 1. sucrose in buffer A), followed by centrifugation at 20,000 rpm, 4 °C
for 25 min. Supernatants were transferred into new tubes and diluted
Western blotting and immunoprecipitation with four volumes of buffer A, and centrifuged at 40,000 rpm, 4 °C for
Cells were collected, quickly rinsed with 1× phosphate-buffered saline 30 min to obtain the CCV-enriched fraction.
(PBS) and lysed in lysis buffer (20 mM Tris pH 7.4, 0.5% Triton X-100,
150 mM NaCl, 1.5 mM MgCl2, 2 mM EGTA, 2 mM DTT and 1× cOmplete Sample preparation for confocal microscopy
Protease Inhibitor Cocktail (Roche)) on ice for 30 min and centrifuged HeLaSTING cells were plated in CellCarrier-96 Ultra Microplates (Perki-
at 12,000 rpm, 4 °C for 10 min. Supernatants were boiled with 4× load- nElmer, 6055302) at a density of 10,000 cells per well and left for at
ing buffer (200 mM Tris pH 6.8, 8% SDS, 40% glycerol, 0.4 M DTT, 0.4% least 5 h to adhere. Cells were then stimulated by adding diABZI at 1 µM
bromophenol blue) for 10 min. Proteins were resolved by SDS–PAGE (MedChemExpress, HY-103665) for the indicated amount of time. In
using SurePAGE precast gels (GenScript) and transferred to nitrocel- time-course experiments, cells were stimulated in a sequential man-
lulose membranes using the Trans-Blot Turbo RTA Midi Nitrocellulose ner and fixed all at the same time. At the end of the stimulation, the
Transfer Kit (Bio-Rad) following the manufacturer’s instructions. Mem- wells were washed once with PBS, and then cells were fixed by add-
branes were blocked with 2% bovine serum albumin (BSA) + 1% milk in ing paraformaldehyde 4% in CBS buffer (10 mM MES pH 6.9, 138 mM
PBST (PBS + 0.05% Tween-20) at room temperature for 1 h and then KCl, 2 mM MgCl2 and 2 mM EGTA) for 5–10 min at room temperature.
incubated with the primary antibody (diluted in PBST) at 4 °C overnight. Cells were then washed three times for at least 5 min in PBS before
After washing in PBST, membranes were incubated with the secondary blocking for 1–2 h at room temperature with PBS supplemented with
antibody at room temperature for 1 h. Membranes were washed with 0.1% (v/v) saponin and 5% (v/v) heat-inactivated FBS and later incubat-
PBST, visualized with western blotting detection reagent (Advansta), ing overnight at 4 °C with the primary antibodies diluted in staining
and imaged using the ChemiDoc XRS Bio-Rad Imager and Image Lab solution (PBS supplemented with 0.1% (v/v) saponin) and 1% (w/v) BSA
Software. Band intensities were quantified using Fiji software (NIH). For (Sigma-Aldrich, A7906)). Antibodies and concentrations are listed in
immunoprecipitation, cells were seeded into six-well plates and were Supplementary Table 2. The cells were then washed with PBS three
transfected with the indicated plasmids. Sixteen hours after transfec- times for 5 min and incubated for 1.5 h at room temperature in second-
tion, cells were lysed in lysis buffer on ice for 30 min and centrifuged ary antibodies diluted in staining solution. From then on, the plate was
at 12,000 rpm at 4 °C for 10 min. Supernatants were transferred into protected from light. Cells were washed twice more (5 min each) in PBS
new tubes and mixed with anti-Flag M2 magnetic beads (Sigma-Aldrich, and incubated for 30–60 min in Hoechst 33342 (Sigma-Aldrich, B2261)
M8823) at 4 °C overnight on a rotator. After three to six washes with lysis 0.2 µg ml−1 in PBS. Cells were then put back into 100 µl PBS per well and
buffer and one to two washes with cold 1× PBS, the beads were boiled either imaged directly or kept at 4 °C until imaging. For experiments
Article
with KD of AP-1, HeLaSTING cells were plated (60,000 cells per well in a performed on Fiji using the BIOP JACoP plug-in with Otsu thresholding
six-well plate) and leftt to adhere for 6–16 h. They were then transfected for all channels. Data were further processed with KNIME (v.4.3.2) to
with siRNAs using Lipofectamine RNAiMax reagent (Thermo Fisher combine them and remove cells with threshold values lower than 500
Scientific) according to the manufacturer’s instructions. Silencer select for pSTING (considered background), and the results were then plotted
predesigned siRNAs were purchased from Thermo Fisher Scientific: with GraphPad PRISM 9 (v.9.3.1). The panels of images were assembled
negative control (4390847), or siAP1G1 (s1143), siAP1B1 (s1141) and using OMERO (v.5.11.0) (ref. 51).
siAP1S1 (s3115); see details in Supplementary Table 1. A total of 2 µl
at 10 µM of siRNAs (equally split between the three siRNAs for the CLEM
siAP-1 wells) combined with 7 µl of lipofectamine RNAiMax reagent HeLaGFP–STING cells were plated in glass-bottomed Petri dishes (MatTek,
in 400 µl OptiMEM total was used per well. The medium was changed P35G-1.5-14-CGRD) with an alpha-numeric grid pattern at a density of
after 6–16 h. Three days after siRNA treatment, cells were replated into 100,000 cells per dish and left to adhere overnight. They were stimu-
microscopy plates (10,000 cells per well) and the experiments were lated by adding diABZI at 1 µM (MedChemExpress, HY-103665) for
continued as described above for non-siRNA-treated cells. 2.5 h. They were then chemically fixed with a buffered solution of 1%
glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate buffer at
Imaging and analysis for confocal microscopy pH 7.4. The dishes were then screened with light microscopy to identify
Fixed and stained 96-well plates were then imaged on two microscopes: cells of interest, which were imaged with both transmitted and fluo-
a confocal Leica SP8 inverted microscope equipped with an HC PL rescence microscopy (AiryScan mode as described above) to record
APO 63×/1.40/oil (magnification/N.A./immersion) objective and HyD their position on the grid. The cells were then washed thoroughly with
detectors, operated with the Leica LAS X software; and a PerkinElmer cacodylate buffer (0.1 M, pH 7.4), and post-fixed for 40 min in 1.0%
Operetta CLS operated with the PerkinElmer Harmony software and osmium tetroxide with 1.5% potassium ferrocyanide and then 40 min
equipped with an Andor Zyla 5.5 camera, and with an LD C Apochromat in 1.0% osmium tetroxide alone. They were finally stained for 40 min
objective with magnification/N.A./immersion of 63×/1.4/water. Image in 1% uranyl acetate in water before being dehydrated through increas-
analysis and quantification were performed by combining PerkinElmer ing concentrations of alcohol and then embedded in Durcupan ACM
Harmony (v.4.9) and Fiji (v.2.3.0), and data were further processed with resin (Fluka). The dishes were then filled with 1 mm of resin and this
KNIME (v.4.3.2) and GraphPad PRISM 9 (v.9.3.1). The panels of images was hardened for 18 h in a 65 °C oven. Cells of interest were then identi-
were assembled using OMERO (v.5.11.0)51. In more detail, all depicted fied according to their position on the alpha-numeric grid, cut away
cell images were acquired with the Leica SP8 confocal microscope from the rest of the material and glued to a blank resin block. Ultra-thin
(63×), except for the images confirming that pSTING still accumulates (50-nm-thick) serial sections were cut through the entire cell with a
at the TGN after treatment with siRNA against AP-1 (compared to siNC), diamond knife (Diatome) and ultramicrotome (Leica Microsystems,
which are images captured on the Operetta (63×). Quantifications of UC7), and collected onto single slot grids with a pioloform support
the total area of pSTING or AP1G1 intensity were calculated in Harmony film. These sections were further contrasted with lead citrate and uranyl
software on the basis of Operetta (63×) images. acetate and images taken in a transmission electron microscope (FEI
Company, Tecnai Spirit) with a digital camera (FEI Company, Eagle). To
Airyscan microscopy correlate the light microscopy images with the electron microscopy
HeLaSTING cells were plated in µ-Slide eight-well chamber slides (ibidi, images and identify the exact position of the Centrin-1:GFP foci, fluo-
80826-IBI) at a density of 10,000 cells per well and left to adhere over- rescent images were overlaid onto the electron micrographs of the
night. Cells were then stimulated by adding diABZI at 1 µM (MedChem- same cell using Photoshop (Adobe).
Express, HY-103665) for 0, 150 or 360 min (Fig. 1b depicts the 150-min
time point and Extended Data Fig. 1e the 0- and 360-min ones) or only STED
0 and 150 min (Fig. 2a and Extended Data Fig. 1c,d). In time-course HeLa cGAS/STING double KO cells were plated in 3.5-cm glass-bottom
experiments, cells were stimulated in a sequential manner and fixed all Petri dishes (FluoroDish, WFD35-100) at a density of 100,000 cells
at the same time. At the end of the stimulation, the wells were washed per dish and left to adhere overnight. They were stimulated by add-
once with PBS, and then cells were fixed by adding paraformaldehyde ing diABZI at 1 µM (MedChemExpress, HY-103665) for 2.5 h. They
4% in CBS buffer (10 mM MES pH 6.9, 138 mM KCl, 2 mM MgCl2, 2 mM were then transfected with plasmids containing Flag-hSTING and
EGTA) for 5–10 min at room temperature. Cells were then washed three mCherry-clathrin (both in pEF-Bos mammalian expression plasmid)
times for at least 5 min in PBS before blocking for 1–2 h at room tem- using Lipofectamine 2000 Reagent (Invitrogen, 11668019) accord-
perature with PBS supplemented with 0.1% (v/v) saponin and 5% (v/v) ing to the manufacturer’s instructions. One microgram per plasmid
heat-inactivated FBS and later incubating overnight at 4 °C with the and 4.5 µl of lipofectamine in 250 µl of OptiMEM (Life Technologies,
primary antibodies diluted in staining solution (PBS supplemented 31985047) were used for one dish. The medium was replaced with fresh
with 0.1% (v/v) saponin) and 1% (w/v) BSA (Sigma-Aldrich, A7906)). culture medium six hours after transfection. The next day, the cells were
Antibodies and concentrations are listed in Supplementary Table 2. The stimulated by adding diABZI at 1 µM (MedChemExpress, HY-103665)
cells were then washed with PBS three times for 5 min and incubated for for 2.5 h. They were then fixed and stained for pSTING as described for
1.5 h at room temperature in secondary antibodies diluted in staining the AiryScan microscopy samples. STED images were acquired using a
solution. From then on, the plate was protected from light. Cells were Leica SP8 STED 3X (Leica Microsystems) equipped with a pulsed white
washed twice more (5 min each) in PBS and incubated for 30–60 min light laser (WLL) as an excitation source and a 775-nm pulsed laser as
in Hoechst 33342 (Sigma-Aldrich, B2261, blue on the depicted images) a depletion light source both for mCherry and ATTO647N. Samples
0.2 µg ml−1 in PBS. Cells were then put back into 200 µl PBS per well were imaged with a 100× objective (Leica, HC APO CS2 100×/1.40/
and either imaged directly or kept at 4 °C until imaging. Imaging was oil) using the LAS X software (Leica Microsystems). For excitation of
performed with a Zeiss LSM 980 Inverted microscope (multi-purpose the respective channels, the WLL was set to 587 nm for mCherry and
confocal with 32 Channels AiryScan, tPMT, widefield and bright-field 647 nm for ATTO647N. Hybrid spectral detectors were used to acquire
capability) using the Plan-Apochromat 63×/1.40/oil (magnification/ the images with a final pixel size of 9.2 × 9.2 nm. The detector time
N.A./immersion) objective and the AiryScan mode and images were gates were set to 1.5–7.5 ns for mCherry and 0.5–6 ns for ATTO647N.
processed in the ZEN software using embedded AiryScan processing Images were acquired as single planes of 1,392 × 1,392 pixels, 600 lines
(3D-mode and 'Normal' resolution). Image analysis and quantifica- per second, 32× line averaging for mCherry and 16× line averaging for
tions were performed with Fiji (v.2.3.0). Colocalization analysis were ATTO647N. Deconvolution of STED images was done with Huygens
Remote Manager v.3.7, using Good’s roughness maximum likelihood column (Cytiva 29323952) equilibrated in PBS with 2% glycerol at
estimation with 60 iterations and a signal-to-noise ratio equal to 2 until pH 7.5, to remove the TEV protease as well as free GST. When used for
it reached a quality threshold of 0.03. GST pull-down, TEV cleavage was skipped and the pool of GST-AP-1
eluate was directly loaded on the size-exclusion chromatography
Protein expression and purification column. AP-1 ΔμCTD was expressed and purified in the same way as
6His-TEVsite-Hs ARF1 (17-181)-Q71L in pHis2 vector was expressed AP-1 core.
in BL21 (DE3) bacteria (Sigma-Aldrich, CMC0014). A single colony
was inoculated in a culture flask with 100 ml LB with Ampicil- In vitro phosphorylation of SUMO–STING 139–379 by TBK1
lin (100 µg ml−1) and incubated with shaking (200 rpm, Infors-HT The recombinant SUMO–STING stock protein was diluted in assay
Multitron) at 37 °C overnight as preculture. Large-scale expres- buffer containing 20 mM Tris, 25 mM MgCl2, 2 mM EDTA, 4 mM EGTA
sion of the protein was started the next day by pouring 100 ml of and 1 mM DTT at pH 7.5, supplemented with phosphatase inhibitor
preculture in a 5-l Erlenmeyer flask containing 2 l LB with ampicil- cocktails and protease inhibitors. The pH was controlled before and
lin (100 µg ml−1). The cells were grown until the optical density at after addition in the sample of 10 mM ATP. TBK1 (MRC PPU Reagents,
600 nm reached 0.7. Expression was then induced by adding iso- DU12469) was added at a ratio of 1:20 (w/w) TBK1:STING. The reac-
propyl β-d-1-thiogalactopyranoside (IPTG) to a final concentration tion was performed at 4 °C overnight. The sample was loaded on a
of 0.5 mM while transferring the culture to an 18 °C shaking incuba- Superdex 200 Increase 10/300 GL size-exclusion chromatography
tor overnight. The bacteria were then collected by centrifugation, column (Cytiva, 28990944) to purify phosphorylated SUMO–STING
solubilized in HisTrap buffer A (20 mM HEPES, 500 mM NaCl, 20 M from the other reagents. The phosphorylation assay was monitored
imidazole, 1 mM DTT and 5% glycerol, pH 7.5) supplemented with by liquid chromatography electrospray ionization mass spectrometry
4-(2-aminoethyl)-benzolsulfonylfluoride-hydrochloride (AEBSF) and (LC/ESI-MS).
cOmplete protease inhibitors (Roche), lysed by sonication, cleared
by centrifugation at 20,000g and then passed through a 5-ml nickel GST pull-down assay
immobilized metal-affinity chromatography column (Cytiva, HisTrap For AP-1–ARF1 pull-down, 30 μg GST-tagged AP-1 complex, 10 μg
HP, 17524802) on an fast protein liquid chromatography (FPLC) sys- ARF1 and 10 μg LBD-STING were mixed together or individually in
tem. The protein of interest was eluted with buffer B (20 mM HEPES, 40 μl pull-down buffer (PBS supplemented with 2 mM MgCl2, 1 mM
500 mM NaCl, 500 mM imidazole, 1 mM DTT and 5% glycerol, pH 7.5). GTP and 2 mM TCEP). The mixture was incubated overnight on ice.
ARF1 was then purified by size-exclusion chromatography through a Thirty microlitres of glutathione sepharose beads (Cytiva) were incu-
Superdex 75 Hiload 16/600 column (Cytiva 28-9893-33). The cDNA of bated with the mixture for 30 min at 4 °C. Excess proteins were washed
human LBD-STING (139–379) was cloned into a pET-28 vector with an off the beads using 200 μl pull-down buffer each time for four times.
N-terminal His6-SUMO tag. LBD-STING was expressed in Escherichia Twenty microlitres of 5× SDS loading buffer was added to the resin and
coli BL21 (DE3) with 0.4 mM IPTG induction overnight at 16 °C. The the mixture was boiled for 5 min. The samples were then centrifuged
cell pellet was lysed by sonication and purified on a Ni-NTA column briefly. Five microlitres of supernatant was analysed by SDS–PAGE.
in 50 mM Tris at pH 8.0, 350 mM NaCl, 20 mM imidazole and 0.5 mM The protein bands were visualized by Coomassie blue staining. For
phenylmethanesulfonyl fluoride (PMSF). The protein was eluted with AP-1 ΔμCTD pull-down, 30 μg GST-tagged AP-1 ΔμCTD complex and
buffer containing 50 mM Tris at pH 8.0, 350 mM NaCl and 300 mM 10 μg LBD-STING were mixed together or individually in 40 μl PBS sup-
imidazole, and then loaded onto a Superdex 75 HiLoad 16/600 column plemented with 2 mM TCEP. The mixture was incubated overnight on
(Cytiva 28-9893-33) in PBS. The sample fractions were pooled and pro- ice. Thirty microlitres of glutathione sepharose beads (Cytiva) was
teins were quantified by molar absorption measurements. All mutants incubated with the mixture for 30 min at 4 °C. Excess proteins were
were generated using a PCR-based technique with appropriate prim- washed off the beads using 200 μl PBS each time for four times. Twenty
ers and confirmed by DNA sequencing. The mutant STING proteins microlitres of 5× SDS loading buffer was added to the resin and the mix-
were expressed and purified in the same way as the wild-type STING. ture was boiled for 5 min. The samples were then centrifuged briefly.
His-tagged AP-1β 1–584, GST-tagged AP-1γ 1–595, AP-1μ 1–423 and Five microlitres of supernatant was analysed by SDS–PAGE. The protein
AP-1σ 1–154 were cloned into a pST44 vector and referred to as AP-1 bands were visualized by Coomassie blue staining.
core. His-tagged AP-1β 1–584, GST-tagged AP-1γ 1–595, AP-1μ 1–142 and
AP-1σ 1–154 were cloned into a pST44 vector and referred to as AP-1 Bio-layer interferometry
ΔμCTD. The AP-1 core complex in the pST44 vector was expressed in Bio-layer interferometry analyses were performed at 25 °C using a
BL21 (DE3) (Sigma-Aldrich, CMC0014). A single colony was inoculated GatorPrime biosensor system (GatorBio) with streptavidin probes
in a culture flask with 400 ml LB with ampicillin (100 µg ml−1) and incu- (GatorBio, 160002). STING, STING mutants or pSTING were mixed
bated with shaking (200 rpm, Infors-HT Multitron) at 37 °C overnight with biotin (EZ-Link-NHS-LC-LC-Biotin, Thermo Fisher Scientific) at a
as preculture. Large-scale expression of the complex (8 l in total: 2 l molar ratio of three biotin molecules to one STING and incubated at
in four 5-l Erlenmeyer flasks) was started the next day by pouring room temperature for 30 min, and then excess biotin was removed by
100 ml of preculture into 2 l of Auto Induction Media Terrific Broth using a desalting column (PD 10, Cytiva). Biotinylated STING (10 μg ml−1)
(Formedium, AIMTB0210) with ampicillin (100 µg ml−1). Flasks were was immobilized onto the streptavidin biosensor (GatorBio, 18-5019)
incubated with shaking at 37 °C for 6 h, then incubated at 18 °C over- for 1 min. The tips were washed with PBS buffer for 2 min to obtain a
night. The cells were then collected by centrifugation (4,000g, 15 min). baseline reading, then the biosensors were dipped into wells contain-
A cell pellet of 2 l expression culture was transferred in a Falcon 50-ml ing the various concentrations of AP-1 ΔμCTD or its mutant for 5 min,
tube. The 2-l expression cells were solubilized in PBS with 1 mM DTT, which was followed by a 10-min buffer wash to allow the dissociation of
1 mM EDTA and 2% glycerol at pH 7.5, supplemented with AEBSF and molecules from the sensor. Data analysis was performed with GraphPad
cOmplete protease inhibitors, then lysed by sonication. The cell lysate PRISM 9 using a standard 1:1 binding model. Two independent experi-
was clarified by centrifugation followed by 0.45-µm filtration. The ments were performed for each sample.
supernatant was first purified on a glutathione sepharose 4B 5-ml col-
umn (Cytiva 28401748) on a FPLC system (Cytiva Aktä Pure). After TEV Cryo-EM data acquisition
cleavage at 4 °C overnight in a 3,500 molecular weight cut-off dialysis Two milligrams of GST-tag cleaved AP-1 core complex was incubated
tubing against PBS with 5% glycerol at pH 7.5, the sample was passed with 2 mg ARF1 for 30 min at room temperature in PBS supplemented
through a Superose 6 HiLoad 16/600 size-exclusion chromatography with 2 mM MgCl2, 1 mM GTP and 2 mM TCEP. Two milligrams of pSTING
Article
was then added, and the mixture was incubated on ice overnight. Excess
ARF1 and pSTING were removed with a Superose 6 increase 10/300 GL Reporting summary
column (Cytiva) in PBS. The AP-1–ARF1–pSTING complex fraction was Further information on research design is available in the Nature
collected and concentrated to 0.8 mg ml−1. Aliquots of 3 μl of AP-1– Research Reporting Summary linked to this article.
ARF1–pSTING complexes were loaded onto glow-discharged holey
carbon grids (Electron Microscopy Sciences, Q250AR1.3, Quantifoil,
Au, R 1.2/1.3, 300 mesh). Grids were blotted for 4 s and plunge-frozen Data availability
in liquid ethane using a Vitrobot at 4 °C and with 100% humidity. Grids Full scans for all western blots and the in-gel fluorescence images are
were screened for particle presence and ice quality on a TFS Glacios provided in Supplementary Fig. 1. The 3D cryo-EM density map is depos-
microscope (200 kV), and the best grids were transferred to TFS Titan ited into the Electron Microscopy Data Bank under accession number
Krios G4. Cryo-EM data were collected using a TFS Titan Krios G4 trans- EMD-14312. The coordinate is deposited in the PDB with accession
mission electron microscope (TEM), equipped with a Cold-FEG on a number 7R4H. Source data are provided with this paper.
Falcon IV detector in electron counting mode. Falcon IV gain refer-
ences were collected just before data collection. Data were collected 47. Concordet, J.-P. & Haeussler, M. CRISPOR: intuitive guide selection for CRISPR/Cas9
genome editing experiments and screens. Nucleic Acids Res. 46, W242–W245 (2018).
with TFS EPU v.2.12.1 using aberration-free image shift protocol (AFIS), 48. Ran, F. A. et al. Genome engineering using the CRISPR–Cas9 system. Nat. Protoc. 8,
recording eight micrographs per ice hole. Movies were recorded at a 2281–2308 (2013).
magnification of 270,000×, corresponding to the 0.45 Å pixel size at 49. Di Micco, A. et al. AIM2 inflammasome is activated by pharmacological disruption of
nuclear envelope integrity. Proc. Natl Acad. Sci. USA 113, E4671–4680 (2016).
the specimen level, with defocus values ranging from −0.8 to −1.8 μm. 50. Hirst, J. et al. Distinct and overlapping roles for AP-1 and GGAs revealed by the
Exposures were adjusted automatically to 60 e− Å−2 total dose, resulting “knocksideways” system. Curr. Biol. 22, 1711–1716 (2012).
51. Allan, C. et al. OMERO: flexible, model-driven data management for experimental
in an exposure time of approximately 3 s per movie. In total, 30,004
biology. Nat. Methods 9, 245–253 (2012).
micrographs in EER format were collected.
Acknowledgements We thank N. Jordan for technical assistance; members of the A.A.
Cryo-EM data processing laboratory for discussions; P. Schu, University Medical Center Göttingen (GER), for sharing
µ1A-deficient MEFs and their complemented counterparts; R. Goldbach-Mansky, National
Motion correction was performed on raw stacks without binning
Institute of Allergy and Infectious Diseases (NIAID), National Institutes of Health (NIH) (US), for
using the cryoSPARC implementation of motion correction. A total providing fibroblasts from patients with SAVI; the EPFL BioImaging and Optics Core Facility, in
of 1,701,051 particles were template-based automatically picked and particular N. Chiaruttini, R. Guiet, T. Laroche and C. Stoffel, for support with imaging and data
processing; and the EPFL Biological Electron Microscopy Core Facility, in particular G. Knott,
particles were binned by a factor of 4. Two rounds of two-dimensional
S. Rosset and J. Blanc, for performing the electron microscopy and alignments of the CLEM
(2D) classification were performed, resulting in a particle set of 539,684 experiment. Cryo-EM data acquisition was performed at the Dubochet Center for Imaging,
particles. Two-dimensional classification of particles showed that Lausanne and we are grateful to A. Myasnikov, B. Beckert, S. Nazarov and E. Uchikawa.
The work was funded by grants to A.A. from the Swiss National Science Foundation
the relative orientation between STING and AP-1 was highly varia-
(310030_188759), the Dr. Josef Steiner Cancer Research Foundation, the European Union’s
ble. Selected particles resulting from the 2D classification were used Horizon 2020 Research and Innovation program grant agreement (grant no. 804933,
for ab initio reconstruction. After two rounds of ab initio reconstruc- ImAgine), the Leenaards Foundation, the Fondation Acteria and the US National Institutes of
Health grant R01 AI 120691 to J.H.H. P.X. and C.L. are supported by EMBO Postdoctoral
tion, 326,772 particles were selected on the basis of STING densities.
Fellowships (ALTF 184-2021 and ALTF 88-2022).
The particles were re-centred and re-extracted by a binning factor
of 2. The particles were subjected to iterative CTF refinement and Author contributions The experiments were designed, conducted and analysed by Y.L., P.X.,
S.R., C.L. and J.R. Cell experiments and confocal and Airyscan microscopy studies were
non-uniform refinement in cryoSPARC to 2.34 Å. The reported reso- performed by Y.L. and S.R. Cryo-EM structural experiments and analysis were performed by
lutions are based on the gold-standard Fourier shell correlation 0.143 P.X. STED image acquisition and analysis was performed by C.L. Biochemical experiments
criterion. Local-resolution variations were estimated using cryoSPARC. were performed by P.X. and J.R. X.R. and J.H.H. shared materials for in vitro reconstitution of
adaptor protein complexes and provided advice for biochemical studies. A.A. conceived and
supervised the work and wrote the manuscript. All authors contributed to editing of the
Model building and refinement manuscript, and support its conclusions.
The AP-1–ARF1–pSTING model was generated using a published AP-1–
Competing interests A.A. is a scientific co-founder of IFM Due. J.H.H. is a co-founder and
ARF1–tetherin Nef structure (PDB 6DFF). The AP-1–ARF1 model after shareholder of Casma Therapeutics and receives research funding from Casma Therapeutics,
removing the tetherin Nef ligand was docked into the cryo-EM map in Genentech and Hoffman-La Roche. The remaining authors declare no competing interests.
Chimera and fine-tuned by manual adjustment with Coot. The pSTING
Additional information
tail was docked against the cryo-EM map in Coot and the whole model Supplementary information The online version contains supplementary material available at
was refined in PHENIX. Several loop regions of AP-1, ARF1 and pSTING https://doi.org/10.1038/s41586-022-05354-0.
were manually adjusted to fit into the map using Coot. The model was Correspondence and requests for materials should be addressed to Andrea Ablasser.
Peer review information Nature thanks Søren Paludan, Sandra Schmid and the other,
refined in real space again in PHENIX. All structure figures were made anonymous, reviewer(s) for their contribution to the peer review of this work.
using UCSF Chimera, UCSF ChimeraX and PyMOL. Reprints and permissions information is available at http://www.nature.com/reprints.
Extended Data Fig. 1 | See next page for caption.
Article
Extended Data Fig. 1 | Intracellular trafficking of pSTING. a,b, Confocal Colocalization of pSTING with the indicated marker is quantified by Manders’
imaging (a) or quantification of pSTING area in bright-field fluorescent colocalization coefficients (d). One representative cell is shown, and
microscopy images (b) of HeLaSTING cells stimulated for 0, 0.5, 1, 2, 3, 6, 9, 12 or quantification is the mean ± s.e.m. of n = 8 cells examined in 1 of 4 independent
24 h with 1 µM diABZI. Cells were fixed and stained for TGN46 (trans-Golgi), experiments. Scale bars, 4 µm in larger left panel, 1 µm in zoomed-in panels.
pSTING and LAMP1 (endolysosomal compartment) as well as with Hoechst dye e, Airyscan imaging of HeLaSTING cells stimulated for 20, 150 (see Fig. 1b) or
(nuclei, in dark blue). Mean ± s.e.m. of n = 3 independent experiments with at 360 min with 1 µM diABZI. Cells were fixed and stained for TGN46 (trans-Golgi),
least 84 fields of view each per condition. Scale bars, 4 µm. c,d, Airyscan pSTING and CHC (clathrins) as well as with Hoechst dye (nuclei, in dark blue).
imaging (c) of HeLaSTING cells stimulated for 2.5 h with 1 µM diABZI. Cells were One representative cell is shown of at least n = 6 cells from 3 independent
fixed and stained for TGN46 (trans-Golgi, not shown here), pSTING and the experiments. White arrows point at occurrences of pSTING. Scale bars, 4 µm in
indicated marker as well as with Hoechst dye (nuclei, in dark blue). larger left panel, 1 µm in zoomed-in panels.
Extended Data Fig. 2 | See next page for caption.
Article
Extended Data Fig. 2 | Super-resolution imaging of STING in CCVs. a, Large- stimulated for 2.5 h with 1 µM diABZI (b) or left untreated (c). For the stimulated
field-of-view STED image showing the colocalization of pSTING and CCVs. HeLa cell (b), some regions with accumulation of high GFP–STING intensity (yellow
cGAS/STING double KO cells were transfected with mCherry-clathrin and Flag- boxes) were re-imaged by electron microscopy (EM) in higher resolution at
STING. One day later they were stimulated with 1 µM diABZI for 2.5 h and then relevant Z-heights. Zoom-in and higher-resolution electron microscopy slices
fixed and stained for pSTING. Two representative events highlighted in the from box 3 is shown in Fig. 1e. LM – light microscopy (Airyscan). White arrows
boxes were magnified (see Fig. 1d). Scale bar, 2 μm. The image depicts n = 1 cell indicate CCVs. Scale bars in (b), 2 µm (un-zoomed top images) and 0.5 µm
out of 62 cells imaged over the course of 2 independent experiments, including (zoom-in box images). Scale bars in (c), 1 µm. n = 1 out of 3 stimulated cells (b)
25 cells showing clear colocalization and 3 exhibiting clathrin ring structures. and n = 1 out of 2 non-stimulated cells imaged from one sample per condition
b,c, CLEM of HeLa STING KO cells stably reconstituted with GFP-hSTING prepared for CLEM.
Extended Data Fig. 3 | AP-1 associates with STING after activation. each per condition. Scale bars, 4 µm. c, HeLa cGAS KO cells incubated with NC
a, Western blot showing levels of AP-1 subunits after NC siRNA or AP-1 siRNA siRNA or siRNAs targeting AP-1σ1 and AP-1 σ3 for 3 days and reconstituted with
transfection in HeLa cells. GAPDH was used as a loading control. b, Bright-field empty vector or HA-tagged AP-1σ1 were treated with 2.5 µM diABZI for 0, 2, 4 h
fluorescent microscopy images and corresponding quantification of AP-1γ before analysis by western blot. Vinculin was used as a loading control. d, AP-1
signal intensity of HeLaSTING cells treated for 3 days with NC siRNA or AP-1 siRNA μ1 KO MEFs and μ1 KO + μ1A MEFs were treated with 0.5 μg/mL DMXAA for 2 h
and then stimulated for 0, 0.5, 1, 2, 3, 6, 9, 12 or 24 h with 1 µM diABZI. Cells were and analysed by western blot. Vinculin was used as a loading control. One
then fixed and stained for TGN46 (trans-Golgi), pSTING and AP-1γ as well as representative of three (b,c) or at least two (a,c,d) independent experiments is
with Hoechst dye (nuclei, in dark blue). Images shown here are from the 2-h time shown. Ratios of target proteins versus loading control normalized to the
points. Mean ± s.e.m. of n = 3 independent experiments with 99 fields of view untreated sample of each condition (c,d).
Article

Extended Data Fig. 4 | See next page for caption.


Extended Data Fig. 4 | AP-1 depletion boosts STING-dependent immune assessed by RT–qPCR in HeLa cells transfected with NC siRNA or AP-1 siRNA for
activation. a, HEK293T cells transfected with FLAG-tagged STINGWT were 3 days and then transfected with 1 μg 90mer. h, Induction of IFNB1, IFIT1, IFIT2,
treated with 2.5 µM diABZI or left untreated for 2 h, immunoprecipitated (IP) IFIT3 expression was assessed by RT–qPCR in HeLa cells transfected with NC
with anti-FLAG antibody, and analysed by western blot. b, HeLa cells siRNA or AP-1 siRNA for 3 days and then transfected with 1 μM 2’3’-cGAMP.
transfected with NC siRNA or AP-1 siRNA for 3 days were treated with 20nM Baf i, Induction of IFIT1 expression was assessed by RT–qPCR in HeLa cells
A1 and 2.5 µM diABZI for 2 h before analysed by western blot. GAPDH was used transfected with non-targeting control (NC) siRNA or AP-1σ, AP-1β and AP-1γ
as a loading control. c, HeLa cells incubated with NC siRNA or AP-1 siRNA for 3 siRNA for 3 days and then treated with 1 µM diABZI for 3 h. j, mRNA levels of
days were infected with HSV-1 (MOI = 5) for 14 h and analysed by western blot. mouse ifi44 in MEFs cells treated with DMSO or 2 µM H-151 for 3 days were
GAPDH was used as a loading control. d, HeLa cGAS/STING double KO cells assessed by RT–qPCR. k, mRNA levels of mouse Cxcl10, Isg15, ifi44, ifnb1 in
transfected with FLAG-tagged STINGWT, STINGL364A were treated with 2.5 µM MEFs cells treated with 40 µg/mL DMXAA for 3 h were assessed by RT–qPCR.
diABZI for 2 h were infected with HSV-1 (MOI=5) for 14 h and analysed by One representative of three (b,e–k) or at least two (a,c,d) independent
western blot. GAPDH was used as a loading control. e, mRNA levels of IFNB1, experiments is shown. For RT–qPCR experiments, ratios of IFNB1, IFIT1, IFIT2,
IFIT1, IFIT2 and IFIT3 in HeLa cells transfected with NC siRNA or AP-1 siRNA for IFIT3 mRNA versus GAPDH mRNA normalized to the untreated groups of each
3 days and treated with 0.5 μg/mL IVT4 were assessed by RT–qPCR. f, NF-kB and condition. Mean ± s.d. of three (e–k) technical replicates. P values based on
IFN-β luciferase assay in HEK293T cells incubated with NC siRNA or AP-1 siRNA two-tailed Student’s t-tests (e–i,k). *P < 0.05, **P < 0.01, ***P < 0.001,
for 3 days, followed by transient expression of STING and stimulation with ****P < 0.0001 (k). Ratios of target proteins versus loading control normalized
diABZI (2.5 µM). g, Induction of IFNB1, IFIT1, IFIT2, IFIT3 expression was to the 0 time point of each condition (b).
Article

Extended Data Fig. 5 | AP-1 depletion boosts STING-dependent immune N154S mutation (e,f), H72N (g,h) or V147M (i,j) were incubated with NC siRNA
activation in different cell types. a,b, HaCaT cells (a) or THP-1 cells or AP-1 siRNA for three days and then analysed by western blot. Vinculin was
(b) transfected with NC siRNA or AP-1 siRNA for 3 days were treated with 2.5 µM used as a loading control in f,h,j. e,g,i, Induction of IFNB1, IFIT1, IFIT2, IFIT3
diABZI for indicated time and analysed by western blot. GAPDH was used as a expression was assessed by RT–qPCR in HeLa cells transfected with NC siRNA
loading control. c,d, Human primary epithelial cells incubated with NC siRNA or AP-1 siRNA for 3 days and then stimulated with 1 µM diABZI. Mean ± s.d. of
or AP-1 siRNA for 3 days were treated with 2.5 µM diABZI and were analysed by three (c,e,g,i) technical replicates. P values based on two-tailed Student’s
RT–qPCR (c) and western blot (d). Vinculin was used as a loading control. t-tests (c,e,g,i).
e–j, Fibroblast cells derived from patients with SAVI characterized by STING1
Extended Data Fig. 6 | AP-1 binds to STING through a dileucine motif. analysed by western blot. e, HeLa cGAS/STING double KO cells transfected with
a, HeLa cGAS/STING double KO cells transfected with FLAG-tagged STINGWT, FLAG-tagged STINGWT or STINGL364F were treated or not with 2.5 µM diABZI for
STING1–317 or STINGLI(L364A/I365A) were treated or not with 2.5 µM diABZI for 14 h 14 h and analysed by western blot. GAPDH was used as a loading control. f, IFN-β
and analysed by western blot. Vinculin was used as a loading control. b, HeLa luciferase assay in HEK293T cells transfected with plasmids expressing indicated
cGAS/STING double KO cells transfected with FLAG-tagged STINGWT or STING constructions and stimulated with diABZI (2.5 µM). g, Glutathione
STINGL364A were treated or not with 2.5 µM diABZI for 14 h and analysed by Sepharose pull-down assays of LBD-STINGWT or LBD-STINGELI(E360A/L364A/I365A)
western blot. Vinculin was used as a loading control. c, HeLa cGAS/STING by GST-tagged AP-1 ΔμCTD core. One representative of three (a–c,e–f) or two
double KO cells transfected with FLAG-tagged STINGL363A, STINGL364A or (d,g) independent experiments is shown. Ratios of target proteins versus
STINGI365A were treated or not with 2.5 µM diABZI for 14 h and analysed by loading control normalized to the untreated sample of each condition (a–c,e).
western blot. GAPDH was used as a loading control. d, HEK293T cells Mean ± s.d. of three (f) technical replicates. P values based on two-tailed
transfected with FLAG-tagged STINGWT, STINGL374A or STINGL374F were treated Student’s t-tests (f).
with 2.5 µM diABZI for 2 h, immunoprecipitated with anti-FLAG antibody, and
Article

Extended Data Fig. 7 | See next page for caption.


Extended Data Fig. 7 | STING–AP-1 interaction depends on TBK1-mediated loading control. One representative of at least two (a–c) independent
phosphorylation. a, HeLa STING KO cells transfected with FLAG-tagged experiments is shown. Ratios of target proteins versus loading control
STINGWT or STINGLR(L374A/I375A) were treated with 2.5 µM diABZI for 0, 1 or 2 h, normalized to the untreated sample of each condition (b,c). d, Mass
immunoprecipitated with anti-FLAG antibody, and analysed by western blot. spectrometry detected molecular weight of SUMO LBD-STING and TBK1-
b, HeLa TBK1 KO cells reconstituted with an empty plasmid or with plasmids phosphorylated LBD-STING (pSTING). e, Bio-layer interferometry binding
expressing TBK1WT or enzyme-dead TBK1S172A were treated with 2.5 µM diABZI studies of LBD-STINGELI(E360A/L364A/I365A) with AP-1 ΔμCTD. f, Bio-layer
for 2 h and analysed by western blot. GAPDH was used as a processing control. interferometry binding studies of LBD-STING3S(S355D/S358D/S366D) with AP-1 ΔμCTD.
c, HeLa cells pretreated with DMSO or 2 µM BX795 for 24 h were stimulated with One representative of at least two (e, f) independent experiments is shown.
2.5 µM diABZI or not (2 h) and analysed by western blot. Vinculin was used as a
Article

Extended Data Fig. 8 | Cryo-EM analysis of pSTING in complex with AP-1. pSTING complex, coloured according to the local resolution. g, 3D Fourier shell
a, Purification and SDS–PAGE analysis of AP-1 core in complex with pSTING and correlation of final 3D reconstruction of the AP-1 pSTING complex.
ARF1. b, Purification and SDS–PAGE analysis of AP-1 core in complex with h, Corrected Gold-standard Fourier shell correlation curves of the AP-1 pSTING
STING3S (S355D/S358D/S366D). c, Representative micrograph of AP-1 pSTING complex complex for the 3D electron microscopy reconstruction. i, Angular
in vitrified ice from 30,004 raw images. Scale bar, 30 nm. d, 2D class averages of distribution of the AP-1 pSTING particles included in the final
AP-1 pSTING complex particles. Box size, 27 nm. e, Flow chart of data reconstruction. One representative of at least two (a,b) independent
processing; see Methods for details. f, Final 3D reconstruction of the AP-1 experiments is shown.
Extended Data Fig. 9 | Density maps and structural models of the AP-1–pSTING complex. The density maps (grey mesh) of AP-1 and pSTING contoured at 3σ.
The protein structures fitted into the density map are shown by the stick models.
Article

Extended Data Fig. 10 | pS366 of pSTING binds to a basic patch of AP-1σ. FLAG-tagged STINGWT and HA-tagged AP-1γWT, γR15E, σWT, σI103S or σV88D,
a, Sequence alignment of AP-1σ in different species and human σ subunit respectively, were treated with 2.5 µM diABZI for 2 h, immunoprecipitated (IP)
isomers. b, Bio-layer interferometry binding studies of LBD-STING or with anti-FLAG antibody and then analysed by western blot. One representative
pLBD-STING with AP-1 ΔμCTD σKR. c, HEK293T cells transfected with of two (b,c) independent experiments is shown.
Extended Data Table 1 | Cryogenic-electron microscopy data summary table

Details of cryo-EM data collection, refinement and validation statistics.


γ

μ μ μ

You might also like