612 | Nature | Vol 585 | 24 September 2020
Article
are flexible (Fig. 3a, b, Extended Data Fig. 8a–c). This is consistent with
data from hydrogen–deuterium exchange mass spectrometry (HX-MS)
showing that αD, αF and the ASL undergo high exchange of hydrogen
and deuterium after DNA-dependent activation of PARP1^23 (Extended
Data Fig. 8d). In our structure, delocalization of these helices and the
ASL loop generates a large opening between the PARP2 HD and ART
catalytic subdomains (Fig. 3b, Extended Data Fig. 8e). In the activated
conformation, the modelled NAD+ is buried, which implies that either
the complex needs to dissociate before the next ADP-ribose can be
added, or the complex cycles through conformational changes that
allow for NAD+ binding and product release. In the state with delocal-
ized ASL, αD and αF, the large opening could be used for NAD+ binding
and exchange (Fig. 3b, Extended Data Fig. 8e, f ). The ability of PARPs
to exchange NAD+ without having to release the substrate would make
PARPs processive enzymes that can add chains of ADP-ribose while
remaining bound to the substrate.
The model of the H3 tail bound to the activated PARP2–HPF1 complex
shows that the H3 tail could access this conformation and catalysis
could occur; however, the product containing the large ADP-ribose
would be buried between PARP2 and HPF1 (Extended Data Fig. 9a). We
solved a structure of another distinct intermediate state in which we
observe flexibility of two different helices in the PARP2 HD subdomain,
αB and the N-terminal part of αD, as well as two HPF1 helices, α7 and α8
(Fig. 3a, c, Extended Data Fig. 9b–d). The delocalization of PARP2 helices
αB and αD is also in agreement with HX-MS data for PARP1^23 (Extended
Data Fig. 8d). This structural change generates a large opening between
the PARP2 HD subdomain and HPF1 that could be used for the product
release (Fig. 3c, Extended Data Fig. 9e, f ), whereas the NAD+ exchange
site remains closed (Extended Data Fig. 9d). The ability to release
the product would enable PARP–HPF1 to modify multiple substrates
without needing to disassemble.
These structures suggest that PARP molecules could cycle between
states in which they could exchange NAD+, perform catalytic reaction
and release the product. This would allow the PARP–HPF1 complex to
bind and modify multiple substrates while remaining bound to, and
thus protecting, the chromatin lesion (Extended Data Fig. 10a). The
enzymatic cycle we describe is consistent with previous biophysical
data obtained for PARP1^23 and might be applicable to all DNA-dependent
PARPs and perhaps to other ADP-ribosylating enzymes.
Auto-PARylation dissociates the complex
To examine the effects of PARylation on interactions with the chromatin,
we initiated the reaction by adding NAD+ to the assembled PARP2–
HPF1–nucleosome complex and observed that PARP2 modified histone
H3, HPF1 and itself (Fig. 4a, b). This resulted in dissociation of PARP2–
HPF1 from the nucleosome and produced PARylated nucleosomes
(Fig. 4a, b). The kinetic of the reaction showed that H3 is PARylated
very rapidly, and longer incubation does not substantially increase
H3 PARylation, indicating that most H3 tails are modified immediately
after activation (Fig. 4c, d). Notably, the complex remains stably
bound to the nucleosome at this point (Fig. 4c, d). By contrast, PARP2
auto-PARylation increases with time and eventually destabilizes the
entire complex (Fig. 4a–d). These data show that PARP2–HPF1 rapidly
modifies histones that then serve as a recruitment platform for otherPARP2–HPF1 activated (3.9 Å) PARP2–HPF1 open state 1 (6.7 Å)αB αFαE αE
αB
Flexible ASL,
αF and αDPARP2 CATPARP2 CATHPF1NAD+ NAD+ASL NAD+ entrya Substrate entry and exitαFFlexible
αB and αDPARP2–HPF1 open state 2 (6.3 Å)NAD+PARP2 CATFlexible
α7 and α 8HPF1HPF1α (^8) α 7 α (^8) α 7
bc
αE
αD
Fig. 3 | PARP2 catalytic domain rearranges to open NAD+ and substrate-
binding sites. a–c, Views at the active site formed by PARP2 and HPF1 in the
nucleosome-bound activated conformation (a) and nucleosome-bound open
states 1 (b) and 2 (c). In open state 1, PARP2 helices αF and αD and the active site
loop (ASL) are f lexible, which might open the active site for NAD+ binding.
In open state 2, dislocation of PARP2 helices αD and αB and HPF1 helices α7
and α8 might open the substrate-release pocket. NAD+ (yellow) was modelled
based on an alignment with PDB 6BHV.
PARylated
HPF1
PARylated
H3
5 15 Time (min)
PARylated
PARP2
Anti-pan-ADP-ribose
Time (min)
Nuc + PARP2
+HPF1
PAnucRylated
Nuc
Marker
DNA
Nuc + PARP2 + HPF1 + NAD+
0515 30 60
c d
H3
H2A/H2BH4
90
PARP2
HPF1
40
25
Nuc + PARP2
- HPF1
DNA Anti-H3
Nuc + P
ARP2 + HPF1
Nuc +
PARP2 + HPF1 + NAD
NucMarker
70
25
Anti-pan-ADP-ribose^15
25
15
H3
PARP2
HPF1
ab
0.5
2.0
1.5
1.0
Marker (kb)NucNuc + PARP2 + HPF1Nuc + P
ARP2 + HPF1 + NAD
NucNuc +
PARP2 + HPF1
Nuc +
PARP2 + HPF1 + NAD
NucNuc + P
ARP2 + HPF1
Nuc + P
ARP2 + HPF1 + NAD
PAnuc Rylated
PAPARylatedRP2
Anti-H3 Anti-PARP2
Nuc + PHPF1 + NADARP2 ++
MarkerProtein
s
0 30 60
Fig. 4 | PARylated PARP2–HPF1 dissociates from the chromatin. a, EMSA
analysis of assembled PARP2–HPF1–nucleosome complex incubated with and
without NAD+. Note the shift in the position of PARylated nucleosomes with
respect to the unmodified nucleosomes. PARylated PARP2 does not co-migrate
with PARylated nucleosomes, indicating dissociation. b, SDS–PAGE and
immunoblotting of the reaction in a. Complex activation leads to
ADP-ribosylation of H3. c, EMSA analysis of the changes in the PARP2–HPF1–
nucleosome complex during the PARylation reaction. Initial PARylation does
not lead to complex dissociation; however, longer PARylation dissociates the
complex, generating PARylated nucleosomes. d, Extent of the PARylation
reaction, followed over time by PAGE and immunoblotting. At first, H3, HPF1
and PARP2 are PARylated to similar extents; extended reaction (60 min) leads
to increased PARP2 auto-PARylation. Numbers at left refer to size markers. One
representative of at least three independent biochemical experiments is
shown for all assays. For gel source data, see Supplementary Fig. 1.