Nature - USA (2020-02-13)

(Antfer) #1
Nature | Vol 578 | 13 February 2020 | 299

proteasomes were rapidly exchanged into and out of the foci, further
supporting the idea that proteasome foci are liquid droplets (Fig. 3c).


p97 and RAD23B regulate proteasome foci


A mass-spectrometry-based proteomic screen for proteasome-
interacting proteins identified ubiquitin-selective chaperone p97
(also known as VCP), substrate-shuttling factor RAD23B and ubiq-
uitin ligase UBE3A (also known as E6-AP)^1 (Source Data of Fig. 4a).
These proteins extensively colocalized with the proteasome at the
foci (Fig. 4a, Extended Data Fig. 6a, b). Inhibition of p97 activity by
NMS-873 caused an increase of around 50% in foci size, suggesting a
positive role of p97 for degradation of ubiquitylated substrates in pro-
teasome foci (Fig. 4b). Indeed, NMS-873, like b-AP15, increased the size
of RPL29 condensates (Extended Data Fig. 7, Supplementary Videos 5,
6). Knockout (KO) of UBE3A caused a reduction of around 30% in the
number of foci, suggesting that UBE3A regulates the proteasome itself
or ubiquitylation of substrate proteins in the foci^22 (Fig. 4b, Extended
Data Fig. 8a). Notably, formation of proteasome foci was markedly


attenuated in RAD23B-KO cells. No significant effects were observed
on siRNA-mediated knockdown of ubiquitin-like proteins (UBQLNs)
and RAD23A, other shuttling factors, or XPC, which functions with
RAD23 proteins in the DNA nucleotide excision repair pathway^1 ,^23
(Extended Data Fig. 6c).
We initially predicted that ubiquitylated proteins would form foci
before the RAD23B-dependent recruitment of proteasomes. How-
ever, we found that neither the proteasome nor ubiquitin formed
the foci in RAD23B-KO cells (Fig. 4c, Extended Data Fig. 8b). This
was not due to a decrease in the level of ubiquitylated proteins
(Extended Data Fig. 8a). RAD23 family proteins have a ubiquitin-
like (UBL) domain and two ubiquitin-associated (UBA) domains that
bind the proteasome and ubiquitin chains, respectively^1 ,^24. Adding
back mutant RAD23B to RAD23B-KO cells revealed that formation of
ubiquitin-positive foci was UBA domain-dependent (Extended Data
Fig. 8b). Because overexpression of RAD23A rescued RAD23B-KO
cells, the difference in ability to form foci can be explained by dif-
ferences in the expression levels of the two proteins^25 ,^26 (Extended
Data Fig. 8c, d).

Hyperosmotic stress
ĺ Cell volume reduction
(molecular crowdingĹ)
ĺ Nucleolar stress
(orphan RPsĹ)

a

PSMB2

DAPI

p97RAD23B UBE3A

PSMB2–eGFP
0.2 M Sucrose, 30 min

b
****

***
29 ± 17

20 ± 13

27 ± 18

± 2.61.2

******
****

ControlNMS-873
RAD23B-KOUBE3A-KO

0

60

120

Foci number per cell
ControlNMS-873
RAD23B-KOUBE3A-KO

Foci diameter (

μm)

0

0.5

1.0

1.50.4
± 0.2

± 0.20.6

± 0.10.3

± 0.20.4

PSMB2 K48Ub PSMB2K48Ub DAPI

PSMB2–eGFPRAD23B-KO
0.2 M sucrose, 30 min

c

d In vitro

WT

RAD23B or A

K48Ub

DIC

Cy3-RAD23B

Cy5–K48Ub

ΔUBL ΔUBA1/2 Cy3–RAD23A

RAD23B(WT)RAD23B(

ΔUBL)

RAD23B

(ΔUBA1/2)RAD23A

0

0.5

1.0

Droplet circularity

RAD23B(WT)RAD23B(

ΔUBL)

RAD23B

(ΔUBA1/2)RAD23A

0

3

5

Droplet diameter (

μm)

1

2

4

± 0.060.99± 0.170.89 ± 0.060.99

ND ND

± 0.300.63

****± 0.841.17

± 0.250.51

**** ****

e

0 s7 s 16 s 21 s 28 s 35 s

Cy5–K48Ub + Cy3–RAD23B

DIC

RAD23BK48Ub

In vitro

(^005101520253035)
5
10
15
20
25
30
35
RAD23B(μM)
K48Ub
(μ 4
M)
Phase separation No phase separation
(^01234)
10
20
40
Ubiquitin chain length
Concentration (
μM) 30
f
g
p97
Proteasome foci formation
Foci clearance
by proteasomal degradation
MG-132
b-AP15
NMS-873
UBL UBAUBA
Ub
UbUb
Ub
RAD23BUBUBLL UBAUBAUBAUBA
Polyubiquitin
chain
Ubiquitylated proteins
Ub
UbUb
Ub
LLPS
p97
Proteasome
RAD23B
Peptides
Ub
Proteasome
Ub
Ub Ub
Fig. 4 | Polyubiquitin chain and RAD23B induce liquid–liquid phase
separation of the proteasome. a, p97, R AD23B and UBE3A localized to
proteasome foci. PSMB2–eGFP cells were stimulated with 0.2 M sucrose and
observed by immunof luorescence with indicated antibodies. Scale bars,
10 μm. b, PSMB2–eGFP cells treated with p97 inhibitor NMS-873 (1 μM,
1 h prior), PSMB2–eGFP cells lacking R AD23B or UBE3A were stimulated with
0.2 M sucrose for 30 min. The graph indicates the number of proteasome foci
per cell and the diameter of individual foci. n represents cell numbers (control,
176 cells; NMS-873, 168 cells; R AD23B-KO, 112 cells; UBE3A-KO, 108 cells). Data
are mean ± s.d., *P < 0.0001, P < 0.0002, P < 0.0021 by Kruskal–Wallis
with Dunn’s test. c, R AD23B-KO PSMB2–eGFP (PSMB2–eGFP/R AD23B-KO) cells
were stimulated with 0.2 M sucrose for 30 min, and endogenous K48Ub was
detected with K48-ubiquitin antibody. Scale bars, 10 μm. d, Top, liquid droplets
formed 90 min after mixing of 20 μM Cy5–K48Ub chains with 20 μM Cy3–
R AD23B(WT), Cy3–R AD23B UBL deletion mutant (ΔUBL), Cy3–R AD23B UBA1
and UBA2 deletion mutant (ΔUBA1/2) or Cy3–R AD23A in 3% PEG and 200 mM
sodium chloride. Scale bars, 5 μm. Bottom, quantified circularity and
quantified diameter of individual droplets (R AD23B(WT), n = 691 droplets;
ΔUBL, n = 230 droplets; R AD23A, n = 491 droplets). Data are mean ± s.d.,
**
P < 0.0001 by Kruskal–Wallis with Dunn’s test. e, Fusion of liquid droplets
containing Cy3–R AD23B and Cy5–K48Ub in vitro. The sample was identical to
the one used in d (15 min after mixing in 10% PEG). Scale bars, 1 μm. f, Left, phase
diagram of R AD23B and K48-tetraubiquitin (K48Ub 4 ) at different
concentrations. Right, phase diagram of R AD23B (10 μM) and ubiquitin chains
with different lengths. Red dots indicate phase separation and blue dots
indicate no phase separation. g, Current model of formation and clearance of
proteasome-containing liquid droplets. In a, c–e, representative results from
two independent experiments.

Free download pdf