Nature | Vol 578 | 13 February 2020 | 303most efficient at pH 6.0 and was impaired under higher pH conditions
(>7.0) (Extended Data Fig. 2f, g), consistent with PAS formation under
starvation conditions that result in acidification of the cytoplasm of
budding yeast^15 to a pH around 6.0.
It is known that liquid–liquid phase separation of proteins is induced
by two distinct mechanisms: the first involves nonspecific weak interac-
tions between IDRs, whereas the second is brought about by multiple
specific interactions between proteins that possess numerous binding
modules^2. Atg13 possesses both an Atg17-binding region (17BR) and
an Atg17-linking region (17LR), which bind to distinct regions in Atg17
(Fig. 2f). Formation of scaffold droplets was severely impaired by the
F430A or F375A mutation in Atg13 and the D247A or P393A mutation in
Atg17 (Fig. 2g, Extended Data Fig. 2h), which attenuate the Atg17–17BR
and Atg17–17LR interactions that are essential for PAS formation and
autophagy^6 ,^16. Thus, phase separation of the Atg1 complex is facilitated
by multiple specific interactions between Atg13 and Atg17, both in vitro
and in vivo. The 1:1 stoichiometry of Atg13 and Atg17–Atg29–Atg31
is optimal for phase separation, which is impaired by excess Atg17–
Atg29–Atg31 (Extended Data Fig. 2i). In line with this, overexpression
of Atg17–GFP impaired droplet liquidity (Extended Data Fig. 2j). Col-
lectively, these data suggest that the Atg1 complex undergoes phase
separation to form a liquid droplet through two-site binding between
Atg13 and Atg17, an essential mechanism of PAS formation in vivo^6.
Regulation of phase separation
Under nutrient-rich conditions, Atg13 is highly phosphorylated by
TORC1, which inhibits the formation of the Atg1 complex and the
PA S^16 ,^17. Some Atg proteins accumulate en masse at the PAS when the
kinase activity of Atg1 is inhibited^10. These observations suggest that
phosphorylation events negatively regulate the formation of the PAS.
When Atg13 was incubated with TORC1 purified from yeast (Extended
Data Fig. 3a) and ATP, Atg13 was hyperphosphorylated, including on
Ser428 and Ser429, whose phosphorylation impairs the interaction
of Atg13 with Atg17 as well as PAS formation^16 (Fig. 3a). Phosphoryl-
ated Atg13 lost the ability to form droplets with Atg17–Atg29–Atg31
(Fig. 3b), indicating that TORC1 inhibits phase separation by directly
phosphorylating Atg13, especially at Ser428 and Ser429.
The kinase activity of Atg1, which is essential for autophagy pro-
gression^17 , is activated upon starvation. Activation of Atg1 requires
autophosphorylation of the kinase domain at Thr226, which is mark-
edly enhanced following rapamycin treatment^18. Clustering of Atg1 by
a selective autophagy cargo or by its targeting to the vacuole acceler-
ates the autophosphorylation^19 ,^20. We monitored phosphorylation of
Thr226 using a phosphorylation-specific antibody for Thr226^18 when
Atg1 alone, phase-separated Atg1 complex or an Atg1 complex in which
phase separation was inhibited by an F430A mutation in Atg13 were
incubated with ATP (Extended Data Fig. 3b). There was no increase in
autophosphorylated protein for Atg1 alone, but a mild increase was
observed in the Atg1(F430A) complex, and a more marked increase was
observed in the phase-separated Atg1 complex (Fig. 3c, Extended Data
Fig. 3c). These data suggest that phase separation of the Atg1 complex
facilitates activation of Atg1 kinase, possibly by increasing collisions
between individual Atg1 molecules.
Next, we studied the effect of Atg1 kinase activity on droplet for-
mation. Addition of ATP induced phosphorylation of Atg1, Atg13 and
Atg29 within 10 min in the Atg1 complex, but not in the Atg1(D211A)
kinase-dead complex, indicating that these three proteins were phos-
phorylated by Atg1 (Extended Data Fig. 3d, e). Incubation with ATP dis-
solved the droplets of the wild-type Atg1 complex, but not those of the
Atg1(D211A) complex, in a similar time frame (Extended Data Fig. 3f ),
confirming that Atg1-mediated phosphorylation, but not the activity of
ATP as a hydrotrope^21 , inhibited phase separation of the Atg1 complex.
In contrast to the in vitro observation that Atg1-complex droplets
dissolve within minutes in the presence of Atg1 kinase activity, the
PAS continues to exist for several hours despite its activation of Atg1
kinase. The PP2C phosphatases Ptc2 and Ptc3 have been reported to
promote PAS formation and autophagy by dephosphorylating Atg1
and Atg13^22. This suggests that the balance of phosphorylation and
dephosphorylation of Atg13 may be important for the maintenance
of the PAS. We studied the effect of Ptc2 on phase separation of Atg1
complex (Fig. 3d). Addition of ATP promptly dissolved Atg1-complex
droplets. Further addition of recombinant Ptc2 (Extended Data Fig. 3g,
h) gradually regenerated the droplets, indicating that Ptc2-mediated
dephosphorylation promotes phase separation of the Atg1 com-
plex. Upon ATP addition, phosphorylation of Atg1 at residue Thr226,
and shortly afterwards, of Atg13 at Ser428 and Ser429 (Ser428/429)
occurred. Thr226 and Ser428/429 were then dephosphorylated after
addition of Ptc2 (Fig. 3e). As Ser428/429 of Atg13 is the most critical
phosphorylation site for inhibiting PAS formation^16 , these results sug-
gest that Atg1 and Ptc2 regulate the reversible phase separation of the
Atg1 complex through phosphorylating and dephosphorylating Atg13
Ser428/429. Notably, dephosphorylation at Atg13 Ser428/429 is faster
than that at Atg1 Thr226 (Fig. 3e, 30–70 min). We thus conclude that
Ptc2-mediated dephosphorylation is one mechanism by which phase
separation of the PAS is maintained while simultaneously retaining a
subpopulation of Atg1 in an activated state.Dynamic structure of scaffold droplets
We next performed structural analysis of the scaffold droplets using
high-speed atomic force microscopy (HS-AFM). Molecules with an
S-shape, a distinctive feature of Atg17^23 , are distributed irregularly00.20.40.60.81.0010203040506070T226-P
S428/9-P00.010.020.030.040.050.06010203040506070a bTime (min)Absorbance at 350 nmATPPtc20102060ATP Ptc20 5 10 15 20 25 30 35 40 45 50 55 60 65 70
T226-P
S428/9-PATP Ptc2
Time (min)Phos-tag
SDS–PAGE
CBB staining
cedDroplet area (%)
Time (min)100100ATP+–kDaTORC1 + +Normal
SDS–PAGE
CBB staining
S428/9-PAtg13P-Atg13051015202505101520250510152025
T226-P
FlagAtg1 only Atg1 complexAtg13(WT) Atg1 complexAtg13(F430A)- 150
- 150
Time (min) kDaAtg13 P-Atg13Time (min)Time (min)Relative band intensityATPPtc200.20.40.60.805101520Atg13
P-Atg13Fig. 3 | Phase separation of the Atg1 complex is dynamically controlled by
phosphorylation-dependent regulation of Atg13. a, Phosphorylation of
Atg13 (P-Atg13) by TORC1. CBB, Coomassie brilliant blue (CBB) staining. b, Left,
impairment of phosphorylated Atg13 phase separation on mixing with Atg17–
Atg29–Atg31. Scale bar, 30 μm. Right, graph shows time-course analysis of
droplet area. Data are mean ± s.d. (n = 3 independent experiments) (b, d, e). c,
Enhancement of Atg1 Thr226 phosphorylation on phase separation in vitro. d,
Left, effect of ATP and Ptc2 on phase separation of the Atg1 complex. Right,
graph shows absorbance at 350 nm as an indicator of droplet formation. Scale
bar, 10 μm. e, Left, phosphorylation of Atg1 at Thr226 and Atg13 at Ser428/429
as assessed by western blot analyses. Right, relative band intensity quantified
from blots. Experiments were repeated independently three times with similar
results (a, c). For gel source data, see Supplementary Fig. 1 (c, e).