Captured snapshots of PARP1 in the active state reveal the mechanics of PARP1 allostery

SUMMARY

PARP1 rapidly detects DNA strand break damage and allosterically signals break detection to the PARP1 catalytic domain to activate poly(ADP-ribose) production from NAD⁺. PARP1 activation is characterized by dynamic changes in the structure of a regulatory helical domain (HD); yet there are limited insights into the specific contributions that the HD makes to PARP1 allostery. Here, we have determined crystal structures of PARP1 in isolated active states that display specific HD conformations. These captured snapshots and biochemical analysis illustrate HD contributions to PARP1 multi-domain and high-affinity interaction with DNA damage, provide novel insights into the mechanics of PARP1 allostery, and indicate how HD active conformations correspond to alterations in the catalytic region that reveal the active site to NAD⁺. Our work deepens the understanding of PARP1 catalytic activation, the dynamics of the binding site of PARP inhibitor compounds, and the mechanisms regulating PARP1 retention on DNA damage.

Graphical Abstract

eTOC blurb

Rouleau-Turcotte et al. determined crystal structures representing the active state of PARP1, a DNA break sensing enzyme. The structures illustrate the allosteric conformational changes that are driven by PARP1 interaction with DNA strand breaks and that are required for PARP1 catalytic production of poly(ADP-ribose), a DNA damage-signaling posttranslational modification.

INTRODUCTION

PARP1 is a multi-domain nuclear enzyme with diverse cellular functions. PARP1 participates in multiple pathways of DNA repair, such as single-strand break (SSB) and double-strand break (DSB) repair, but it also plays a role in the stabilization of DNA replication forks and the modification of chromatin structure (Ray Chaudhuri and Nussenzweig, 2017). PARP1 is the founding member of the diphteria toxin-like ADP-ribosyltransferases family (ARTD) with 17 members, which includes PARP2 and PARP3 that are also involved in DNA repair (Luscher et al., 2021). PARP inhibitors (PARPi) have been the subject of extensive clinical studies in the past years given their ability to provoke synthetic lethality in cancer cells bearing DNA damage repair defects (Bryant et al., 2005, Farmer et al., 2005). However, the potential use of PARPi has recently broadened to other diseases such as neurodegenerative disorders that feature PARP1 hyperactivity (Thapa et al., 2021). Thus, there is a continued need to fully understand the mechanisms that regulate PARP1 activity.

PARP1 domain structure, from N- to C-terminus, comprises three zinc-fingers (Zn1, Zn2, and Zn3), a BRCT domain, a WGR (Trp-Gly-Arg) domain, and a catalytic (CAT) domain that is itself composed of two subdomains: the autoinhibitory helical domain (HD) and the ADP-ribosyl transferase fold (ART) (Fig. 1). PARP1 DNA binding domains (Zn1–3 and the WGR) collectively recognize DNA strand breaks, triggering an allosteric signal that relieves the autoinhibition performed by the HD on the ART (Dawicki-McKenna et al., 2015, Eustermann et al., 2015, Langelier et al., 2012). The ART catalytic pocket is then revealed, allowing PARP1 to robustly convert NAD⁺ into poly(ADP-ribose) (PAR). The local burst in PAR production at DNA damage sites recruits repair factors to process the DNA break and promotes the release of PARP1 from the damaged site (D’Amours et al., 1999, Prokhorova et al., 2021).

Figure 1. PARP1 allostery and HD dynamics.

PARP1 domains shown schematically with HD closing the catalytic pocket in the absence of DNA damage (top). PARP1 binding to DNA damage organizes the domains and allosterically leads to a dynamic HD (middle). NAD⁺ binding requires an open HD conformation, and NAD⁺ and type I PARP inhibitors strengthen PARP1 interaction with DNA (bottom).

The stepwise activation of PARP1 highlights that allosteric communication between domains is essential for DNA damage-dependent catalytic activity, as the HD does not contact the DNA break (Langelier et al., 2012). However, the specific mechanisms that underlie PARP1 HD alterations remain unclear. This gap in structural information could be the result of PARP1 domains capacity to “breath” on DNA damage, thus representing a challenging structural target. A crystal structure of PARP1 bound to a DSB using a combination of the essential domains (Zn1, Zn3, and WGR-CAT domains; PDB code 4DQY) (Langelier et al., 2012) revealed a distorted HD when compared to structures of the isolated CAT domain, yet the active site remained inaccessible to substrate NAD⁺. Hydrogen-deuterium exchange coupled with mass spectrometry (HXMS) analysis of PARP1 dynamics hinted that the PARP1/DNA crystal structure 4DQY did not represent the HD in its completely activated conformation (Dawicki-McKenna et al., 2015). The HXMS data indicated massive local changes in HD dynamics following DNA binding that were interpreted as specific HD helices rapidly sampling an unfolded conformation (Dawicki-McKenna et al., 2015), yet these helices were folded in 4DQY. Interestingly, certain PARPi impact PARP1 allostery through HD contacts (Ogden et al., 2021, Zandarashvili et al., 2020). It is currently difficult to appreciate HD dynamics, including with PARPi, due to a lack of structural information about the fully active HD state.

The HD sterically blocks the PARP1 active site in the absence of a DNA break. This autoinhibitory HD function necessitates that the tight interactions formed between the HD and the ART must weaken following detection of DNA damage (Dawicki-McKenna et al., 2015). The HD has also been implicated in PARP1 interaction with DNA damage. A non-hydrolyzable analog of substrate NAD⁺ (BAD; benzamide adenine dinucleotide) demonstrated that PARP1 affinity for DNA strand breaks increases following substrate binding in the active site, and this increase in DNA binding affinity required the presence of the HD (Langelier et al., 2018). The increase in PARP1 affinity for DNA has also been observed with EB-47, a PARPi that mimics NAD⁺ structure (Jagtap et al., 2004, Zandarashvili et al., 2020). BAD and EB-47 have been classified as Type I PARPi based on their ability to influence HD structure and dynamics and induce allosteric changes that tighten DNA damage interactions (Zandarashvili et al., 2020). Type I compounds can be viewed as agents that weaken the ability of the HD to interact with the ART. These results imply that the HD has two separate roles: blocking the active site by closely interacting with the ART, and participating in PARP1 allostery in a manner that increases DNA binding affinity (Fig. 1).

We reasoned that the HD alternates between interacting with the ART and forming contacts with the regulatory domains on DNA, in particular the WGR domain. The PARP1 CAT domain crystallized in its apo form would therefore represent the HD free of the influence of the WGR domain (i.e. a completely closed conformation). In contrast, the PARP1/DNA complex 4DQY likely represents an intermediate in PARP1 catalytic activation, as the HD is under the influence of both ART and WGR interactions. We sought to reduce the dynamics and breathability of PARP1 on DNA damage in order to capture its active state. We took advantage of mutants that favor the opening of the HD and conducted X-ray crystallography to capture snapshots of isolated PARP1 states that represent an important and currently missing step of PARP1 activation.

RESULTS

We tested whether PARP1 with a complete deletion of the ART fold (ΔART) would free the HD and allow its full interaction with WGR and other regulatory domains, expecting that ΔART would exhibit higher affinity for DNA damage, similar to PARP1 with BAD or EB-47. A fluorescence polarization (FP) DNA binding assay was conducted to assess the affinity of ΔART for a DSB (Fig. 2A). The K_D measured for ΔART is ~5-fold lower than wild-type PARP1 (WT) (12.1 nM versus 59.7 nM, respectively)(Table 1), highlighting an unambiguous increase in DNA damage binding affinity. ΔART binding affinity was comparable to WT in the presence of EB-47, suggesting that deleting the ART does indeed mimic the effects of EB-47 on PARP1. As expected, ΔART binding affinity was not affected by the addition of EB-47 (Table 1). These results suggest that deleting the ART fold has favored an HD conformation similar to its open state and allowed its full contributions to binding DNA damage.

Figure 2. ART deletion increases PARP1 DNA binding affinity and retention on DNA damage.

(A) Representative FP DNA binding assay showing that ART deletion results in higher DNA binding compared to WT. The fitted binding constants are listed in Table 1. (B) Representative SPR sensorgrams showing an increase in response units as PARP1 WT or ΔART (with or without EB-47) bind to a DNA SSB immobilized on the biosensor. The lower response units of PARP1 ΔART are attributed to its smaller size compared to WT. Binding constants resulting from experiments performed at multiple protein concentrations are listed in Table 1 (see also Fig. S1). (C) DNA competition assay showing that PARP1 ΔART exhibits increased retention on DNA.

Table 1.

PARP1 binding affinities for DNA strand breaks

Fluorescence polarization (FP) DNA binding assay
PARP1 full-length	KD (nM)	KD (nM) + EB-47
WT	59.7 +/− 9.2	8.4 +/− 1.0
ΔART	12.1 +/− 4.6	15.4 +/− 1.8
L698A/L701A	64.4 +/− 13.4	15.3 +/− 7.8
L713F	21.3 +/− 3.5	6.2 +/− 3.1
Y569A	78.3 +/− 15.7	20.6 +/− 1.4
Y569L	164.2 +/− 13.9	40.8 +/− 2.1
G558E	118.5 +/− 23.2	75.6 +/− 21.7
G558V	66.7 +/− 7.4	10.2 +/−1.5
G558L	57.2 +/− 5.8	7.6 +/− 1.2
ΔV687-E688	9.4 +/− 1.3	8.0 +/− 1.9
M686G/V687P	7.3 +/− 2.3	4.8 +/− 1.4
V687A/E688A	71.2 +/− 10.7	10.3 +/− 3.6
PARP1 domains added to Zn1-Zn3		KD (nM)
WGR-CAT		>800
WGR-CAT + EB-47		42.8 +/− 5.7
WGR-HD		7.0 +/− 2.4
WGR-HD ΔV687-E688		31.5 +/− 19.1

Surface plasmon resonance (SPR)
PARP1 full-length	ka (M−1s−1)	Kd (s−1)	KD (M)
WT	7.25 ± 0.65 E+6	1.93 ± 0.04 E−3	2.66 ± 0.20 E−10
WT + EB-47	8.02 ± 0.37 E+6	2.79 ± 0.95 E−4	3.51 ± 1.35 E−11
ΔART	5.73 ± 0.04 E+6	2.71 ± 0.36 E−4	4.71 ± 0.59 E−11

K_D, equilibrium binding constants; k_a, rate of association; k_d, rate of dissociation.

We employed a surface plasmon resonance (SPR) biosensor to investigate ΔART binding kinetics on a surface-immobilized SSB. The K_D of WT measured in the presence of EB-47 was ~6-fold lower than WT alone (0.035 nM versus 0.266 nM, respectively), largely due to a decreased rate of dissociation (Fig. 2B, Fig. S1A, B and Table 1). An equivalent increase in affinity was seen with ΔART compared to WT (Fig. 2B, Fig. S1C and Table 1), consistent with FP assay results. As a control, we confirmed that EB-47 had no effect on ΔART in the SPR assay (Fig. 2B). It appears that WT (in the presence or absence of EB-47) and ΔART all have similar association rates (k_a) on DNA damage (Table 1). However, the dissociation rates (k_d) of ΔART and WT with EB-47, although similar to one another, are significantly slower than WT alone (~7-fold). This behaviour of ΔART is thus reminiscent of the Type I PARPi allosteric effect that triggers PARP1 retention on DNA damage (Zandarashvili et al., 2020).

We also tested whether the observed increase in persistence on DNA would prevent ΔART from interacting with competitor DSBs. Briefly, the domains of PARP1 are dynamically bound to DNA damage and exhibit a “breathing” behavior that translates into rapid release when exposed to competitor DSBs that engage the domains as they breath (Rudolph et al., 2018). ΔART replicated the phenotype seen with WT in the presence of EB-47, with a reduced propensity to exchange to competitor DNA (Fig. 2C). These results support that deleting the ART has freed the HD to occupy an open and active conformation and that ΔART recapitulates the effects of EB-47 on PARP1.

We sought to crystallize PARP1 ΔART by taking advantage of a “divide and conquer” strategy that was used to obtain a complex of PARP1 essential domains bound to a DSB. Briefly, only four PARP1 domains are absolutely necessary for DNA damage-dependent catalytic activity (Zn1, Zn3, WGR, CAT); the Zn2 and BRCT domains can be fully deleted from PARP1 and still maintain catalytic activity (Langelier et al., 2012). We purified the first and third zinc fingers of PARP1 attached by an engineered linker (Zn1-Zn3), as well as the remaining essential domains of PARP1 ΔART: WGR-HD. We conducted additional FP DNA binding assays to confirm a functional reconstituted ΔART composed of PARP1 fragments (Fig. S2A, S2B). The DSB binding affinity of reconstituted ΔART was higher than WT and agreed with binding assays performed with PARP1 as a single polypeptide (Table 1).

We also used isolated Zn1 and Zn3 as opposed to the linked version of Zn1-Zn3. Crystallization trials were attempted using both forms of zinc fingers (Zn1 + Zn3, or Zn1-Zn3), WGR-HD, and different lengths of duplex DNA. Two crystal structures of a ΔART/DNA complex were obtained. The first contained linked Zn1-Zn3 domains and a 10-base pair duplex at a resolution of 3.3 Å (Fig. 3A; 1 complex per asymmetric unit), and the second non-linked zinc fingers Zn1 and Zn3 and a 12-base pair duplex at a resolution of 3.6 Å (Fig. S2C; 4 complexes per asymmetric unit) (Table S1).

Figure 3. Crystal structures of PARP1 ΔART bound to a DSB.

(A) The ΔART complex structure determined at 3.3 Å resolution (1 complex per asymmetric unit). Zn1 (red), Zn3 (purple), WGR (dark blue), and HD (green) domains are shown binding a DSB (orange backbone). The HD domain in light blue was extracted from 4DQY and overlaid to compare the WGR-HD interfaces. (B) Close-ups of the WGR-HD interfaces for all ΔART complexes are overlaid in the top left subpanel and shown individually in subpanels I to V as labeled. L698 and L701 are modeled as ball-and-sticks. (C) Comparison of WGR-HD interfaces highlighting the switch loop (i) in contact with αB (7AAA in teal), (ii) in an intermediate state (4DQY in blue), and (iii) in contact with the hydrophobic pocket of WGR (ΔART in green). The lower inset gives a second view of these structural changes. The structure comparison also highlights a reconfiguration of αB in ΔART (right inset). The relative positions of Y689 and Q694 are noted to illustrate the structural change. (D) Close-up view of αF from ΔART (green) and 7AAA (teal) as viewed from the ART fold. E763 approximates the center of ΔF rotation. (E) SDS-PAGE activity assay for the indicated PARP1 mutants in the absence or presence of DNA for the indicated time points. Figure 3 is related to Figure S2.

Examination of PARP1 ΔART crystal structures revealed HD conformations that have not been observed in previous PARP1 structures. In particular, the residues linking helix αB to helix αD adopted a novel conformation (Fig. 3A). Each of the 5 ΔART/DNA complexes displayed the same conformation for the αB-αD linker containing L698 and L701 (Fig. 3B; overlayed), whereas the length of αB varied among the complexes, with two complexes not exhibiting a structured αB helix (Fig. 3B; panels II and V). The ΔART/DNA complex displayed in panel III of Fig. 3B was used in further analysis given that its αB is the most complete and it did not engage in crystal contacts. Compared to apo PARP1 CAT domain structures, such as PDB 7AAA (Ogden et al., 2021), the HD changes displayed by ΔART are quite profound (Fig. 3C). Residues that normally adopt an extended coil structure following αB instead switch to a helical conformation and become an extension of αB in two of the ΔART/DNA complexes (Fig. 3B; panels I and III, and Fig. 3C, right panel). The remaining residues that link the extended αB to the unchanged αD in ΔART have moved closer to the WGR domain, hinting that in 4DQY the αB-αD linker represents an intermediate state between 7AAA and ΔART (Fig. 3C, lower left). This observation suggests that 4DQY captured PARP1 transiting between its inactive and active state, likely because the WGR and ART were both influencing the HD. The 5 ΔART/DNA complexes and the PARP1/DNA complex 4DQY otherwise all display a similar assembly of domains on DNA (Fig. S2D), with important interactions with the DSB and between PARP1 domains maintained. The 5 ΔART/DNA complexes also display the WGR-HD contact points that were first observed in 4DQY and they remain unchanged following the ART fold deletion (Fig. S2D).

The structural switch in the αB-αD linker brings residues into close proximity to the WGR domain (Fig. 3C), suggesting that these HD-WGR contacts contribute to the multi-domain assembly of PARP1 on DNA damage, thus influencing DNA binding affinity and allosteric signaling. Inspection of the WGR-HD interface revealed a hydrophobic pocket on WGR that the αB-αD linker appears to fill in the ΔART structures, notably with L698 and L701. In contrast, these residues contribute to the HD hydrophobic core in the PARP1 inactive state. As such, mutating both L698 and L701 to alanines (L698A/L701A) in the full-length enzyme increased PARP1 DNA-independent activity (Langelier et al., 2012), implying that L698A/L701A destabilized the HD closed conformation and thereby impaired its autoinhibitory function (Fig. 3E). However, in the PARP1 active state, L698 and L701 now appear important for the WGR-HD interface. We hypothesized that L698A/L701A would weaken interdomain communication following HD opening. Indeed, the L698A/L701A mutation exhibited a strong deficiency in DNA-dependent PARP1 catalytic activity, substantially reducing PAR amounts over time (Fig. 3E), although its DNA binding affinity remained similar to WT PARP1 (Table 1 and Fig. S2E). These results highlight a decoupling between the two events that the HD influences (i.e. binding DNA damage and catalytic activity). We interpret that the L698A/L701A mutation is sufficient to disrupt allosteric communication and thus DNA-dependent catalytic activity, but too conservative to disrupt HD contributions to DNA binding.

L698 and L701 contributions to the HD hydrophobic core in the closed state resemble that of L713, a residue in the center of the HD fold. Mutating L713 to a phenylalanine (L713F) also destabilizes the HD and leads to increased DNA-independent PARP1 catalytic activity (Langelier et al., 2012). However, in contrast to L698A/L701A, L713F displays increased DNA binding affinity (Table 1 and Fig. S2E) and DNA-dependent activity (Fig. 3E), suggesting that HD destabilization in this case diminishes the autoinhibitory role but still maintains allosteric signalling. The different outcomes for L713F and L698A/L701A mutations on DNA-dependent activity further indicates that L698 and L701 are contributing to the WGR-HD interface.

We also interrogated the WGR side of the new interface by mutating Y569 to an alanine (Y569A) and leucine (Y569L). Both Y569A and Y569L exhibited lower DNA binding affinity (78.3 nM and 164.2 nM, respectively versus 59.7 nM for PARP1 WT; Table 1 and Fig. S2E) and reduced DNA-dependent activity relative to WT (Fig. 3E). However, Y569A consistently displayed a milder phenotype than Y569L. While Y569A showed a mild pro-retention phenotype in the presence of EB-47, Y569L appeared resistant to the effects of EB-47 (Fig. S2H), suggesting that a branched aliphatic chain at this position is more disruptive. We also mutated G558 to valine (G558V) and leucine (G558L). Both G558V and G558L displayed similar DNA binding affinity and catalytic activity compared to PARP1 WT (Table 1, Fig. S2E and Fig. 3E). The more disruptive mutation of G558 to glutamate (G558E) resulted in lower DNA binding affinity (118.5 nM versus 59.7 nM for PARP1 WT; Table 1 and Fig. S2E) and virtually no DNA-dependent activity relative to WT (Fig. 3E). Moreover, G558E DNA binding affinity did not increase in the presence of EB-47 (Table 1 and Fig. S2G). These results suggest that G558E has disrupted HD contributions to DNA binding affinity, and the ability to maintain the HD in an open and active conformation. The intermediate phenotypes seen with Y569A and Y569L also support the importance of the αB-αD linker rearrangement in establishing the WGR-HD interface. The PARP1 mutations analyzed did not introduce major changes in thermal stability relative to WT, and the mutants with the largest changes in thermal stability were those designed to disrupt HD structure (e.g. L698A/L701A) (Fig. S3A).

In addition to the structural switch in the αB-αD linker, all PARP1 ΔART complexes also displayed a rotation of αF around a pivot point roughly located at residue Glu763 toward the N-terminal end of αF (Fig. 3D). Of note, αF normally faces the ART and acts as a regulator of active site availability to NAD⁺ (Langelier et al., 2018).

The ΔART structures indicated HD rearrangements that underlie PARP1 activation, in particular changes in αB and αF. We turned to the Va1687/Glu688 deletion (ΔVE) mutant of PARP1 as a way to view the HD active conformation in the context of the ART. PARP1 ΔVE is a two-residue deletion in αB identified in a high-density CRISPR-Cas9 mutagenesis screen aimed at discovering cellular PARP1 mutants resistant to the toxicity of PARPi talazoparib (Pettitt et al., 2018). Similar to other mutations that disrupt HD structure, ΔVE exhibited elevated levels of catalytic activity in the absence of DNA damage. However, ΔVE activity was strongly stimulated in the presence of DNA damage, indicating that PARP1 allosteric catalytic activation is still intact (Fig. S3B). FP DNA binding assays indicated that ΔVE has an increased affinity for DNA damage, reminiscent of PARP1 ΔART and WT PARP1 in the presence of EB-47 (Table 1 and Fig. S2E). Moreover, ΔVE also displayed an enhanced capacity to resist competitor DSBs as seen in the DNA competition assay (Fig. S2F). These results indicated that the ΔVE mutation favors the HD adopting the active conformation in the presence of DNA strand breaks, and can thus serve as a model for understanding PARP1 activation.

We produced the WGR-CAT fragment bearing the ΔVE deletion and crystallized it using the reconstitution strategy using PARP1 essential domains. FP DNA binding assays confirmed that reconstituted ΔVE binds DNA damage with elevated affinity (Table 1). Using Zn1-Zn3, WGR-CAT ΔVE, and a 5-base pair DNA duplex, we obtained crystals that diffracted up to 3.2 Å resolution (Table S1). The asymmetric unit contained two complexes and the assembly of PARP1 DNA binding domains within each complex is similar to 4DQY (Fig. S3C). Inspection of the WGR-HD interface revealed that ΔVE structure displays essentially the same αB-αD linker rearrangement observed in ΔART, with leucine residues approaching the hydrophobic pocket on WGR and a helical segment added to one end of αB (Fig. 4A, B). The opposite end of αB connects to αA and engages the base of αF. Notably, the orientation and position of this αB-αF interaction is distinct from the αB-αF interaction observed in the inactive state of PARP1 catalytic domain (PDB 7AAA), and this structural difference is particularly evident in the region where αB methionine residues M682 and M686 approach αF (Fig. 4C). Between the contact points at both ends of αB (WGR on one end, αA and αF at the other end), the helix structure is “broken” exactly at the point where the VE double deletion occurs. We anticipate that in WT PARP1, the contacts at both ends place a strain on αB that disrupts the helical structure, or places αB in a highly flexible state. The VE deletion likely relieves this stress by allowing the ends of αB to interact with both regions simultaneously, while retaining its distinctive αB-αD linker state that contributes to the multi-domain assembly on DNA.

Figure 4. Crystal structure of the ΔVE complex bound to a DSB.

(A) The ΔVE structure (salmon) was determined at 3.2 Å resolution (one of the two complexes in the asymmetric unit is shown). The ΔART structure (green) is superimposed. (B) ΔVE displays a WGR-HD interface similar to that observed in ΔART, except for a break in αB at the point where residues 687 and 688 are deleted. (C) Side-by-side close-up view of HD from the PARP1 CAT apo structure (7AAA in teal, left) and the ΔVE structure (in salmon, right). M686 and M682 on αB contribute to the HD hydrophobic core in PARP1 CAT apo by enveloping L698 (left panel). The HD hydrophobic core is modified in the ΔVE structure bound to DNA, with M686 and M682 now enveloping Y775 on αF (right panel). (D) DNA competition assay showing mutant M686G/V687P increases DNA retention relative to WT, while mutant V687A/E688A releases slightly faster from DNA damage. (E) Microirradiation assay monitoring PARP1 kinetics at sites of laser-induced DNA damage in cells. ΔVE and ΔART show increased retention overtime. (F) Microirradiation assay monitoring PARP1 kinetics at sites of DNA damage in cells in the presence of PARPi talazoparib. PARP1 WT and mutants show on average the same retention behavior. Figure 4 is related to Figure S3.

We reasoned that “breaking” αB, using a strategy other than the double deletion, should reproduce the ΔVE phenotype. We introduced a mutation in the same region of αB by mutating M686 and V687 to glycine and proline residues, respectively, with the goal of disrupting helix-forming potential. M686G/V687P showed an unambiguous increase in DNA binding affinity compared to WT (~8-fold) (Table 1, Fig. S3E). In fact, M686G/V687P exhibited a stronger DNA retention propensity compared to ΔVE (Fig. 4D), as well as elevated DNA-independent catalytic activity (Fig. S3B). These results suggest that M686G/V687P not only reproduces but rather enhances the ΔVE phenotype.

We also mutated V687 and E688 to alanines (V687A/E688A) to rule out the possibility that these residues participate in interactions that maintain PARP1 in a closed conformation, and that deletion of the residues in ΔVE removed these contacts. V687A/E688A displayed a similar DNA binding affinity as PARP1 WT and was not resistant to competitor DNAs, as seen in a DNA competition assay (Table 1 and Fig. 4D). Thus, the ΔVE phenotype cannot be attributed to the abrogation of specific interactions mediated by V687 and E688, but instead to the propensity of the deletion to “break” helix αB.

The increased DNA damage retention propensity of PARP1 ΔVE and ΔART was also assessed in cells through a microirradiation assay (Pettitt et al., 2018). After initial recruitment to sites of laser-induced DNA damage, PARP1 WT and V687A/E688A appeared to quickly release from the DNA damage overtime (Fig. 4E). In contrast, ΔVE and ΔART showed a delay in recruitment compared to WT, but displayed increased retention on sites of DNA damage, consistent with results obtained with the competition assay in vitro (Fig. 4D). In the presence of PARPi talazoparib, PARP1 WT and V687A/E688A displayed a delay in recruitment combined with an increased pro-retention capacity (Fig. 4F), mimicking ΔVE and ΔART kinetics in the absence of talazoparib. PARP1 ΔVE kinetics remained largely unchanged in the presence of PARPi, although later time-points mildly suggested that ΔVE remains at sites of DNA damage even longer.

The PARP1 ΔVE structure also indicated a rotation of αF relative to the inactive state (7AAA), but to a larger degree than observed in the ΔART structures (Fig. 5A, left panel). The αF rotation is centered around the same Glu763 pivot point, but the angle separating 7AAA and ΔVE is approximately 30°, compared to ~16° in the ΔART complexes. Unlike the HD, the ΔVE ART domain and catalytic pocket maintain the same structure as the inactive state, with an RMSD of less than 1.0 Å (Fig. S3D). However, the HD conformational changes are propagated to the ART in the form of a concerted rotation of the entire ART domain relative to the rest of PARP1 (Fig. 5A, right panel and framed inset). The rotation centers on a pivot point located around Glu763, the same pivot point as αF. Although αF and ART undergo the same overall rotation, the two regions are sheared apart such that αF is displaced from the ART. To analyze whether the displacement of αF observed in our structure would accommodate substrate binding, we soaked ΔVE crystals with the NAD⁺ mimic EB-47. The ΔVE/EB-47 crystals diffracted to 3.1 Å (Table S1) and the ΔVE/EB-47 structure is very similar to ΔVE crystallized in apo form (Fig. 5B). The ΔVE/EB-47 structure supports the hypothesis that ΔVE represents PARP1 in its open and active state since the CAT domain can effectively accommodate EB-47 without any clashes from the HD. In further support, alignment of the ΔVE and 7AAA CAT domains with a non-hydrolyzable NAD⁺ analog superimposed (Langelier et al., 2018; 6BHV) indicates that αF of 7AAA clashes with the adenine moiety, while αF of ΔVE provides enough space to accommodate NAD⁺ (Fig. 5C).

Figure 5. A concerted ART rotation opens the catalytic site and expands the ligand entry path.

(A) (left panel) Superposition of ΔVE (salmon), ΔART (green) and the apo CAT 7AAA (teal) highlights the major shift in the structures of αF and αB, as viewed from the WGR domain. The Cα of Y689 serves as a reference point to highlight the movement of αB. (right panel) ΔVE and apo CAT superposition viewed looking toward the WGR. ART rotation is highlighted by vector arrows depicting the displacement of equivalent atoms from each structure. The framed inset includes vectors for more equivalent atoms and only αF helices are shown. (B) Superposition of ΔVE structure (salmon) and ΔVE structure in complex with EB-47 (raspberry), showing no major changes in PARP1 structure with bound ligand. (C) The ΔVE/EB-47 structure (raspberry) modeled with the compound BAD (yellow) in place of EB-47. αF from 7AAA (teal) is overlaid to highlight the αF displacement in ΔVE when bound to an NAD⁺ analog. (D) (panel I) Alignment of the ART of ΔVE/EB-47 (raspberry) and 7AAA (teal). The NAD⁺ entry path is delimited by the ASL and αJ (both orange), αF (light blue), and αD (green). BAD (in spheres) is shown in the position of EB-47. (panel II) Surface representation of 7AAA colored as described in panel I. (panel III) Surface representation of ΔVE/EB-47 colored as described in panel I, showing a larger substrate entry path. Figure 5 is related to Figure S3.

To further understand how the ART and αF rotation can impact PARP1 catalytic activity, we compared the “back” of the NAD⁺ binding pocket in ΔVE and 4DQY (Fig. 5D I–III). This region of the binding pocket is lined on one side by the ART (active site loop, ASL and αJ) and on the other side by the HD (αF and αD), and potentially represents an entry point for NAD⁺ to reach the catalytic pocket (I). In 7AAA, αD protrudes from one side and appears to obstruct the opening to the catalytic pocket (II). However, in ΔVE the ART/αF rotation has altered the opening, given that αD has not undergone the same rotation. This HD alteration leads to a wider opening (III), suggesting easier access to the catalytic pocket. Thus, we predict that the concerted rotation of the ART fold and αF can play a role in PARP1 substrate availability, but there could also be other facets of the multi-step process of producing poly(ADP-ribose) that could benefit from the ART rotation motion that we observe.

DISCUSSION

Despite several studies aimed at characterizing PARP1 catalytic activation, the destabilization of the HD and subsequent opening remain poorly understood. A recent study characterized HD dynamics using the isolated PARP1 CAT domain in the presence of inhibitors and mutations that promote PARP1 catalytic activation (Ogden et al., 2021). HD changes due to mutants and inhibitors did not fully recapitulate changes seen in HXMS studies and therefore did not appear to represent PARP1 in its fully active state. As the HD bridges PARP1 DNA-binding domains and the catalytic domain to relay the activation signal across PARP1, it appears that the isolated CAT domain cannot capture the full range of HD dynamics. Thus, a structure of the HD in its open conformation represented a key missing piece of information for the puzzle of PARP1 activation.

We have demonstrated that a deletion of the PARP1 ART fold frees the HD to engage in PARP1 assembly on DNA (Fig. 2). This observation builds on previous results that hinted at the HD being essential to the allostery induced by Type I PARPi, such as BAD and EB-47, which lead to enhanced DNA binding. Type I PARPi are known to support a destabilized HD (Zandarashvili et al., 2020). Our work further supports that the HD has two distinct roles (i.e. closing the active site through autoinhibition, or responding to active site occupancy and the multi-domain assembly on DNA), since these roles depend on the HD adopting two different conformations and deleting the ART fold has seemingly triggered the open conformation.

While the ART deletion represents a very striking “mutation” that could arguably complicate the analysis of its effect, a point mutant was recently found with a similar behavior as ΔART. Mutating the PARP1 catalytic triad (H-Y-E) histidine to an aspartic acid (H862D) showed little to no catalytic activity, consistent with its role in catalysis (Shao et al., 2020, Steffen et al., 2013, Vyas et al., 2014). However, H862D also displayed DNA retention and increased DNA damage binding affinity, akin to PARP1 WT with BAD, hinting at an HD preferentially in an open conformation (Shao et al., 2020). H862D expressed in the ART fold fragment was not efficiently produced in E. coli and accumulated in inclusion bodies, suggesting that the mutation perturbed the global stability of the ART fold. It is therefore possible that H862D mimics an ART deletion given that its unstable ART fold is likely unable to achieve efficient contacts with the HD.

We determined two crystal structures of PARP1 ΔART bound to a DSB (Fig. 3 and Fig. S2). Despite ΔART displaying increased DNA binding affinity and elevated DNA damage retention, all complexes contact the DSB similarly to 4DQY. This observation is consistent with HXMS results of PARP1 WT bound to DNA damage that showed further protection of pre-existing interdomain and DNA-protein interactions in the presence of EB-47, effectively tempering down PARP1 breathability on DNA damage (Zandarashvili et al., 2020).

All structures of ΔART displayed a rearranged αB-αD linker contributing to a hydrophobic pocket with the WGR, thereby likely stabilizing the multi-domain assembly of PARP1 on a DNA break. Mutants in the hydrophobic pocket, Y659A, Y569L and G558E all lowered PARP1 DNA-dependent catalytic activity and DNA binding affinity. In addition, Y569A, Y569L and G558E all displayed various responsiveness to EB-47, further suggesting that they impaired PARP1 allostery. Another distinctive trait concerning all ΔART structures is the rotation of αF (Fig. 3D). This rotation does not seem to be driven by allosteric clashes since αB either moves away from αF or does not adopt any particular structure in the majority of ΔART complexes (Fig. 3B).

Apart from the rearrangement seen in the αB-αD linker and the αF rotation, all ΔART complexes displayed additional changes in their αB since most complexes did not support a complete helix, while other complexes showed an elongated αB (Fig. 3B). Previous work concerning the PARP1 active conformation has focused on HD destabilization as a hallmark of the open/active enzyme. The fact that some ΔART complexes display a relatively well formed and stable HD might be surprising. However, this apparent structural stability might be a result of the ART deletion, which has freed the HD from the restraints it is usually experiencing on both sides (from the WGR and the ART at the same time).

The ΔVE crystal structure revealed many similarities with ΔART (Fig. 4A and 4B). The αB-αD linker of ΔVE also contributes to the WGR hydrophobic pocket and the αF rotation is even more pronounced. Indeed, the double deletion has allowed the HD to fully interact with the WGR while connected to the ART and M686G/V687P, a double mutation highly unfavorable to a helical conformation, was found to enhance the ΔVE phenotype. PARP1 ΔVE kinetics at sites of DNA damage was compared to WT and ΔART via microirradiation of cell nuclei. Both ΔVE and ΔART exhibited a delay in recruitment to DNA damage compared to WT. Since PARP1 catalytic activity mediates the recruitment of additional PARP1 molecules to sites of DNA damage (Mortusewicz et al., 2007), it appears that the delay seen with ΔART could be due to a lack of catalytic activity. The ΔVE delay, however, could be the result of its increased DNA-independent activity and potential pre-PARylation, which could lower ΔVE affinity for DNA damage. Despite the delay in recruitment, both ΔVE and ΔART displayed an increased DNA retention propensity, which is consistent with results obtained with in vitro biochemical experiments and thus provide confirmation in a cellular context.

The most striking feature of the ΔVE structure is undoubtedly the concerted rotation of the ART fold, which keeps an almost identical structural arrangement but rotates in the same direction as αF (Fig. 5A). As seen in the ΔVE/EB-47 structure, the ART can accommodate EB-47 without any clashes from a triad of acidic residues on αF known to regulate the active site availability (E763/D766/D760). Indeed, a triple mutation of these residues to alanines (E763A/D766A/D760A) dramatically increases DNA-independent PARP1 activity (Langelier et al., 2018). It is expected that the concerted rotation and movement of αF away from the ART fold will also allow access to substrate NAD⁺.

The rotation of αF and ART in the same direction was unexpected, as one might expect a rotation of these elements away from each other to more efficiently remove clashes from the acidic residues on αF. Instead, it appears that keeping αF in close proximity to the catalytic pocket is needed to drive communication from the active site back to the DNA binding domains (i.e. “reverse” or bi-directional allostery). The proximity of αF, under a scenario where NAD⁺ is bound and provokes reverse allostery, could further stabilize PARP1 on DNA while performing catalysis. Keeping αF in close proximity could also facilitate the control of PARP1 catalytic activity (i.e. rapidly closing the active site after NAD⁺ hydrolysis and release).

Comparing the ΔVE structure to 7AAA or 4DQY did not highlight any major structural differences in the ART fold, hinting that despite its rotation, the ΔVE ART fold has kept a similar structural arrangement, as confirmed by RMSD obtained through alignments (Fig. S3D). This observation is consistent with previous results obtained with HXMS, which have not identified substantial HX differences in the ART fold upon DNA damage or inhibitor binding (Dawicki-McKenna et al., 2015, Langelier et al., 2018, Zandarashvili et al., 2020).

PARP1 interacts with different factors capable of modulating its catalytic activity and DNA damage binding. For example, histone PARylation factor 1 (HPF1) is a cofactor that directly contributes to the PARP1 catalytic site and thereby switches the specificity of PARP1 from glutamate/aspartate to serine residues (Suskiewicz et al., 2020). The cryo-EM structure of a PARP2/HPF1 complex bound to a DNA break between two nucleosomes (PDB 6X0L) compared to the PARP2 apo CAT domain (PDB 1GS0) shows that a rearrangement of the HD is necessary in order to allow HPF1 binding (Bilokapic et al., 2020, Oliver et al., 2004). The αB-αD linker of PARP2 moves closer to the WGR in 6X0L and a portion of the loop extends helix αB. The HD rearrangements of PARP2 to accommodate HPF1 thus appear similar to the ones seen in ΔART and ΔVE (Fig. S4A). A crystal structure of PARP1 bound to HPF1 has been determined (PDB 6M3I); however, this complex only includes the PARP1 ART fold, thereby not providing any information on the state of the HD in the complex (Sun et al., 2021). Our modeling indeed indicates that PARP1 ΔVE can accommodate HPF1 binding, and we confirmed that HPF1 can regulate PARP1 ΔVE reactions, permitting the modification of serine residues (Fig. S5A). We also confirmed that HPF1 can regulate other mutants like V687A/E688A, M686G/V687P, L698A/L701A and L713F, as these HD mutations could potentially have modified the HD structure in a manner that makes the HPF1 binding site more accessible. The presence of DNA and EB-47 alter HD structure such that PARP1 interaction with HPF1 increases over gel filtration (Suskiewicz et al., 2020) (Fig. S5B). We have confirmed that our PARP1 mutants with HD alterations also require DNA and EB-47 for interaction, indicating that the HD mutations alone do not stimulate HPF1 interaction (Fig. S5B-E). However, we noted that the HPF1/PARP1ΔVE interaction in the presence of DNA/EB-47 does appear more robust than with PARP1 WT (Fig. S5C), suggesting that the preferred open conformation of ΔVE favors the interaction with HPF1 and could therefore influence the dynamics of PARP1/HPF1 regulation.

The 6X0L PARP2 structure does not display a rotation of the ART fold, perhaps because its αB has taken on an extended and unbroken helix (as seen in the ΔART structures), thereby maintaining the ART domain in place. However, HXMS analysis of PARP1 has clearly indicated that αB does not adopt a stable helix in the active state (Dawicki-McKenna et al., 2015). Cryo-EM analysis of the PARP2/HPF1/nucleosome complex indicated a high level of PARP2 mobility, thus it is possible that the rotated ART conformation was not significantly represented in analysis. Indeed, this underlines our strategy of using mutant PARP1 versions that promote the active state. The PARP2 catalytic pocket in the cryo-EM structure was expected to accommodate NAD⁺ binding, but the structure was not determined in the presence of a ligand and the authors stated that a bound NAD⁺ would be trapped on all sides in the catalytic pocket (Bilokapic et al., 2020). The ART rotation observed in our ΔVE structure likely represents a movement to create an entry path by which the ligand can gain access to the catalytic pocket (as discussed in Fig. 5D). Thus, the ΔVE structure has allowed us to capture a PARP1 conformation that is likely to be transient in the WT enzyme.

A crystal structure of PARP2 bound to an activating 5’-phosphorylated DNA (PDB 7AEO) also displays a conformational change in the αB-αD linker that mimics ΔART and ΔVE (Fig. S4B) (Obaji et al., 2021). However, 7AEO does not show an ART rotation, but rather a complete displacement of the ART from the rest of the complex (Fig. S4C). With the ART displaced in such a way, it is difficult to envision how efficient allosteric communication can be achieved from the active site to the DNA binding regions, and the cryo-EM study of the PARP2/HPF1 complex bound to a DSB did not report this ART displacement. The PARP2 crystal structure highlights the mobility of the ART following DNA damage interaction, and it could also indicate differences in how PARP1 and PARP2 structures respond to DNA damage.

Our results allow us to present a mechanism for PARP1 allostery (Fig. 6; see also Video S1). First, PARP1 DNA binding domains arrange on the DNA strand break and create a collective HD binding platform, centered on WGR. The WGR-HD interface is first established through an interaction observed in 4DQY and the present study (Fig. 6A, 1), and this docking of the HD induces a “switch” in the αB-αD linker conformation that promotes a second WGR-HD contact site (Fig. 6A, 2). The structural switch involves the addition of helical segments to cap the end of αB (Fig. 6A, 3). Together, these changes modify the local HD structure and drive αB downward to a position that pushes on αA and thereby rotates the ART fold and αF (Fig. 6A, 4), shearing αF away from the catalytic pocket (Fig. 6A, 5) and opening access to NAD⁺ (Fig. 6B, 1). αB is strained through maintaining both the WGR-HD interface and the rotated conformation of αF and ART, and the αB-αD linker is prone to switch back to the off state in which αF contacts the ART (Fig. 6B, 2 and 3 reverting back to panel a). Binding of substrate NAD⁺ or a Type I PARPi, such as EB-47, to the active site prevents αF from engaging the ART fold, and thereby locks the HD in the open conformation (Fig. 6C). The ΔVE mutation stabilizes the open conformation by relieving tension on αB, thus allowing the WGR-HD interface and the rotated conformation to be stabilized (Fig. 6D). Thus, both the ΔVE mutant and Type I inhibitor binding impact PARP1 allostery, supporting the HD open conformation and strengthening PARP1 binding to a DNA break.

Figure 6. Mechanism for the HD dynamics that drive PARP1 allostery.

(A) Schematic representation of PARP1 in the inactive state, with the contacts and HD structural transitions noted: (1) first WGR-HD contact point is established, (2) the αB-αD linker repositions toward the WGR hydrophobic pocket, (3) αB is elongated and moves downward, (4) concerted rotation of ART and αF, and (5) αF is displaced from the catalytic pocket. (B) Schematic representation of PARP1 in the active state: (1) an open catalytic pocket results from the ART and αF being sheared apart, and (2, 3) tension in αB in the active state favors the αB-αD linker reverting to the inactive state. (C) Substrate NAD⁺ (or a Type I inhibitor) can lock PARP1 in the active state by supporting the HD conformation in which ART and αF are sheared apart. (D) The ΔVE mutation favors PARP1 in the active state by relieving the stress on αB. See discussion for further details.

Our structures of PARP1 captured in the active state have shed light on PARP1 allosteric communication and highlighted novel details concerning HD conformational changes. Snapshots of PARP1 in the active state have also revealed movement in the ART fold that can render the catalytic pocket accessible to substrate NAD⁺ in the multi-turnover reaction to produce poly(ADP-ribose). These novel insights further complete the puzzle of PARP1 activation and will help the development of PARPi aimed at promoting or disabling allosteric communication. For example, new PARPi with specificity toward PARP1 over PARP2 appear to engage specific residues on PARP1 αF (Johannes et al., 2021), and knowledge of HD dynamics can inform the most stable and targetable aspects of HD structure.

LIMITATIONS OF THE STUDY

The crystal structures determined in this study (PDB 7S68, 7S6H, 7S6M and 7S81) and the reference structure 4DQY do not include PARP1 Zn2 and BRCT domains. These structures represent the minimal arrangement of PARP1 domains needed to trigger catalytic activity, and do not capture potential contributions from the Zn2 and BRCT domains. These structures represent the most complete view of PARP1 currently available; there is no published structure for full-length PARP1.

STAR METHODS

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, John M. Pascal (john.pascal@umontreal.ca).

Materials Availability

All unique/stable reagents generated in this study are available from the lead contact with a completed material transfer agreement.

Data and Code Availability

• The atomic coordinates and structure factors have been deposited in the Protein Data Bank (https://www.rcsb.org) and are publicly available as of the date of publication (see the key resource table). Accession codes are 7S68 (ΔART bound to a 10-bp palindromic DNA duplex), 7S6H (ΔVE bound to EB-47 and a 5-bp DNA duplex), 7S6M (ΔVE bound to a 5-bp DNA duplex), and 7S81 (ΔART bound to a 12-bp palindromic DNA duplex). Unprocessed SDS-PAGE gels have been deposited in Mendeley Data (http://dx.doi.org/10.17632/cx5dcgnmtp.1) and are publicly available as of the date of publication.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
E. coli strain BL21 (DE3) Rosetta2	Millipore	Cat#71400
DH5-alpha electrocompetent E. coli cells	Goldbio	Cat#CC-203-TR
Biological samples
Chemicals, peptides, and recombinant proteins
EB-47 (hydrochloride)	Cayman Chemical	Cat#34684
Talazoparib	SelleckChem	Cat#S7048
Veliparib (ABT-888, hydrochloride)	Cayman Chemical	Cat#11505
Hydroxylamine (NH2OH)	Sigma-Aldrich	Cat#467804
Critical commercial assays
Lipofectamine 3000	Thermofisher	Cat#L3000001
Deposited data
PARP1 ΔART bound to a 10-bp palindromic DNA duplex	This paper	PDB: 7S68
PARP1 ΔVE bound to EB-47 and a 5-bp DNA duplex	This paper	PDB: 7S6H
PARP1 ΔVE bound to a 5-bp DNA duplex	This paper	PDB: 7S6M
PARP1 ΔART bound to a 12-bp palindromic DNA duplex	This paper	PDB: 7S81
Structure of Human PARP-1 bound to a DNA double strand break	(Langelier et al., 2012)	PDB: 4DQY
Crystal structure of the catalytic domain of human PARP1 (apo)	(Ogden et al., 2021)	PDB: 7AAA
Mendeley Data (uncropped gel images) http://dx.doi.org/10.17632/cx5dcgnmtp.1	This paper	http://dx.doi.org/10.17632/cx5dcgnmtp.1
Experimental models: Cell lines
CAL51 PARP1−/−	(Pettitt et al., 2018)	N/A
Experimental models: Organisms/strains
Oligonucleotides
10 nt palindromic DNA	IDT	N/A
12 nt palindromic DNA	IDT	N/A
5 nt DNA	IDT	N/A
48 nt DNA carrying an internal 6-FAM group (dumbbell)	IDT	N/A
48 nt DNA (dumbbell)	IDT	N/A
18 nt DNA with a 6-FAM group on the 5’ end	IDT	N/A
Recombinant DNA
PARP1 ΔART	This paper	N/A
PARP1 L698A/L701A	(Langelier et al., 2012)	N/A
PARP1 G558E	This paper	N/A
PARP1 ΔV687-E688	(Pettitt et al., 2018)	N/A
PARP1 M686G/V687P	This paper	N/A
PARP1 V687A/E688A	This paper	N/A
PARP1 V687A	This paper	N/A
PARP1 E688A	This paper	N/A
PARP1 L713F	(Langelier et al., 2012)	N/A
PARP1 Y569A	(Langelier et al., 2012)	N/A
PARP1 Y569L	This paper	N/A
PARP1 G558L	This paper	N/A
PARP1 G558V	This paper	N/A
PARP1 Zn1-Zn3 ΔZn2	(Steffen et al., 2015)	N/A
PARP1 Zn1	(Langelier et al., 2010)	N/A
PARP1 Zn3	(Langelier et al., 2008)	N/A
PARP1 WGR-CAT	(Langelier et al., 2012)	N/A
PARP1 WGR-HD	This paper	N/A
PARP1-GFP	(Pettitt et al., 2018)	N/A
Software and algorithms
XDS	(Kabsch, 2010)	https://xds.mr.mpg.de
PHASER	(McCoy et al., 2007)
CCP4	(Winn et al., 2011)	https://www.ccp4.ac.uk
COOT	(Emsley et al., 2010)	https://www2.mrc-lmb.cam.ac.uk/personal/pemsley/coot/
REFMAC5 (CCP4 suite)	(Murshudov et al., 2011)	https://www.ccp4.ac.uk/html/refmac5.html
Phenix	(Liebschner et al., 2019)	http://www.phenix-online.org
phenix.refine (Phenix suite)	(Afonine et al., 2012)	https://phenix-online.org/documentation/reference/refinement.html
SBGrid	(Morin et al., 2013)	https://sbgrid.org
Matlab	MathWorks	https://www.mathworks.com/products/matlab.htm
TraceDrawer	Ridgeview Instruments	https://tracedrawer.com

Experimental Model and Subject Details

DH5-alpha electrocompetent E. coli cells (Goldbio) were transformed with the product of QuickChange mutagenesis. Cells of Rosetta 2 (DE3) strain of E. coli (Milipore) were transformed with expression plasmids. CAL51 PARP1^−/− cells were maintained in DMEM medium (Gibco) supplemented with 10% foetal bovine serum (FBS) and 1×penicillin–streptomycin (Sigma-Aldrich).

Method details

Expression constructs and mutagenesis.

The following PARP1 constructs were expressed from a pET28 expression vector (Novagen) with an N-terminal hexahistidine tag: PARP1 WT (residues 1-1014) and mutants, Zn1 domain (residues 1-96), Zn1-Zn3 (residues 1-366 Δ97-206) and WGR-HD (residues 527-786 Δ663-676). PARP1 WGR-HD includes a deletion in between the WGR and the HD to remove an alpha helix (αA) belonging to the ART fold. The Zn3 domain (residues 216-366) and WGR-CAT (518-1014) fragments were expressed from a pET24 expression vector (Novagen) with a C-terminal hexahistidine tag. All mutations and deletions were performed using the QuickChange protocol (Stratagene) and verified by automated Sanger DNA sequencing.

Protein expression and purification.

PARP1 WT and mutant proteins were expressed and purified as described previously using three chromatography steps: Ni²⁺-affinity, heparin and gel filtration (Langelier et al., 2011, Langelier et al., 2017). Selenomethionine-containing WGR-HD was expressed in Escherichia coli grown in defined medium (Van Duyne et al., 1993), and purified following the same three chromatography steps. We grew WGR-HD with incorporated selenomethionine (SeMet) to aid/confirm the interpretation of the protein register in electron density maps.

Crystallization and data collection of the PARP1/DNA complexes.

The first PARP1 ΔART/DNA complex was formed by mixing Zn1-Zn3 and selenomethionine-containing WGR-HD at 300 μM each with a 10-bp palindromic DNA duplex at 165 μM (5’ GCCTGCAGGC 3’) in the following buffer: 25 mM HEPES pH 8.0, 150 mM NaCl, 1mM EDTA, and 0.1 mM TCEP. Crystals were grown by sitting drop vapor diffusion at 20°C by mixing the PARP1 ΔART/DNA complex with an equal amount of 10% PEG 8000, 14% ethylene glycol, and 100 mM HEPES pH 7.5. Crystals were rapidly transferred to a solution of 10% PEG 8000, 14% ethylene glycol, 100 mM HEPES pH 7.5, 20% sucrose, and 0.1 mM TCEP prior to flash cooling in liquid nitrogen.

The second PARP1 ΔART/DNA complex was formed by mixing Zn1, Zn3, and selenomethionine-containing WGR-HD at 300 μM each with a 12-bp palindromic DNA duplex at 165 μM (5’ ATGCGGCCGCAT 3’) in the following buffer: 25 mM HEPES pH 8.0, 150 mM NaCl, 1mM EDTA, and 0.1 mM TCEP. Crystals were grown by sitting drop vapor diffusion at 20°C by mixing the PARP1 ΔART/DNA complex with an equal amount of 25% tert-butanol and 100 mM Tris pH 8.5. Crystals were rapidly transferred to a solution of 25% tert-butanol, 100 mM Tris pH 8.5, 20% glycerol, and 0.1 mM TCEP prior to flash cooling in liquid nitrogen.

The PARP1 ΔVE/DNA complex was formed by mixing Zn1-Zn3 and WGR-CAT ΔVE at 300 μM each with a 5-bp DNA duplex at 165 μM (5’ CGACG 3’) in the following buffer: 25 mM HEPES pH 8.0, 150 mM NaCl, 1mM EDTA, and 0.1 mM TCEP. Crystals were grown by sitting drop vapor diffusion at 20°C by mixing the PARP1 ΔVE /DNA complex with an equal amount of 11% PEG 6000 and 100 mM MES pH 6.5. Crystals were rapidly transferred to a solution of 10% PEG 6000, 100 mM MES pH 6.5, 25% sucrose and 0.1 mM TCEP prior to flash cooling in liquid nitrogen. To obtain EB-47 bound to a PARP1 ΔVE/DNA complex, EB-47 (in water, final concentration of 625 μM) was spiked into wells containing PARP1 ΔVE/DNA crystals. After 3 days, crystals were rapidly transferred to a solution of 10% PEG 6000, 100 mM MES pH 6.5, 25% sucrose, 0.1 mM TCEP and 700 μM EB-47 prior to flash cooling in liquid nitrogen.

X-ray diffraction data for both PARP1 ΔART/DNA complexes and ΔVE/DNA complex with EB-47 was collected at beamline 8.3.1 of the Advanced Light Source while x-ray diffraction data for apo ΔVE/DNA complex was collected at beamline 08B1-1 of the Canadian Light Source. The data was processed using XDS (Kabsch, 2010).

Structure determination of PARP1/DNA complexes.

All PARP1/DNA complex structures were determined by molecular replacement (MR) using the program PHASER as implemented in the CCP4 suite (McCoy et al., 2007, Winn et al., 2011). The structure of PARP1 bound to DSB (4DQY) was modified and used as search model. Briefly, one copy of Zn1 and Zn3 bound to DNA were used as a search model, with the DNA trimmed to the right length for each complex (i.e. 5-bp, 10-bp or 12-bp). Following MR the WGR, HD and ART domains (when required) were positioned into MR-based electron density maps with the aid of the structure 4DQY. PARP1 ΔVE/DNA complexes required subsequent rigid body refinement to successfully position the ART fold in the electron density. The anomalous signal from data collected on selenomethionine-containing crystals (both PARP1 ΔART/DNA complexes) at the selenium absorption edge was used to guide the rebuilding of αB and the αB-αD linker in the HD. All models were constructed using COOT and refined in REFMAC5 (CCP4 suite) and/or phenix.refine (Phenix suite) using TLS, geometric restraints, and NCS (when applicable) (Afonine et al., 2012, Emsley et al., 2010, Liebschner et al., 2019, Murshudov et al., 2011).

Fluorescence polarization.

The DNA competition assay and the DNA binding assays were essentially performed as described previously (Langelier et al., 2012, Langelier et al., 2018) and as further described in greater detail in this section. All fluorescence polarization experiments were performed at least three times. Figures associated to each experiment display representative curves.

For the DNA competition assay, 40 nM PARP1 full-length WT or mutants were incubated with 20 nM of dumbbell DNA with a central nick carrying an internal fluorescent FAM group (5’-GCT GAG C/FAMT/T CTG GTG AAG CTC AGC TCG CGG CAG CTG GTG CTG CCG CGA-3’) for 30min at room temperature in 12 mM HEPES pH 8.0, 60 mM KCl, 8 mM MgCl₂, 4% glycerol, 5.7 mM beta-mercaptoethanol, 0.05 mg/ml bovine serum albumin with or without EB-47 (25 μM). A competitor unlabeled DNA of the same sequence was added at 100 nM and FP was measured over time on a VictorV plate reader (Perkin Elmer).

For the DNA binding assay using full-length enzymes, increasing concentrations of PARP1 WT or mutants were incubated for 30 min at RT with 5 nM of an 18-bp DNA duplex (5’-GGGTTGCGGCCGCTTGGG-3’) that carried a fluorescent FAM group on the complementary 5’ terminus in the following buffer: 12 mM HEPES pH 8.0, 250 mM NaCl, 4% glycerol, 5.7 mM beta-mercaptoethanol, 0.05 mg/ml bovine serum albumin with or without EB-47 (20 μM).

DNA binding assays with PARP1 fragments were performed as such: increasing concentrations of PARP1 Zn1-Zn3, WGR-HD, WGR-CAT or mutants were incubated for 30 min at RT with 5 nM of an 18-bp DNA duplex (described above) in the following buffer: 12 mM HEPES pH 8.0, 60 mM KCl, 8 mM MgCl₂, 4% glycerol, 5.7 mM beta-mercaptoethanol, 0.05 mg/ml bovine serum albumin. WGR-CAT and WGR-HD showed no apparent binding since these domains rely on the zinc fingers to establish efficient DNA binding. Following preliminary tests to establish saturating concentration of Zn1-Zn3, DNA affinity measurements were performed as follow: increasing concentrations of PARP1 WGR-HD, WGR-CAT or mutants were incubated for 30 min at RT with 3 μM Zn1-Zn3 and 5 nM of an 18-bp DNA duplex (described above) in the absence or presence of EB-47 (20 μM) in the following buffer: 12 mM HEPES pH 8.0, 60 mM KCl, 8 mM MgCl₂, 4% glycerol, 5.7 mM beta-mercaptoethanol, 0.05 mg/ml bovine serum albumin. FP for both DNA binding assays was measured on a VictorV plate reader (Perkin Elmer) and a 1:1 binding model was fit to the data using Matlab (MathWorks).

Surface plasmon resonance assay.

SPR was performed on a Reichert 4SPR biosensor. The system buffer was 10 mM HEPES pH 8.0, 250 mM NaCl, 0.1 mM TCEP, and 0.05% Triton X-100. Streptavidin-coated chips (Reichert) were used to capture on three different channels a DNA SSB bearing a biotin group (5, 10 and 20 nM respectively). PARP1 WT or mutants were flowed over the biosensor at various concentrations (from 0.63 nM to 10 nM) in the absence or presence of EB-47 (5 μM) (association phase, ~20–120 s). Buffer was passed over the biosensor during the dissociation phase (~120–380 s). All data was double-referenced to buffer and a control channel that did not contain the immobilized binding partner. PARP1 WT or mutant titration onto DNA SSB was processed and fitted with a 1:1 binding model using TraceDrawer (Reichert).

SDS-PAGE PARP1 activity assay.

The SDS-PAGE PARP1 activity assay was performed essentially as described (Langelier et al., 2008). PARP1 full-length WT and mutants (0.8 μM) were preincubated with 5 μM of an 18-bp DNA duplex for 10 min at RT. 5 mM NAD⁺ was added to the reaction and the mixture was incubated for various time before quenching with addition of SDS loading buffer containing 0.1 M EDTA. PARP1 DNA-independent activity was similarly assessed by omitting the DNA duplex from the reaction. The samples were resolved on SDS-PAGE and stained with Imperial Protein Stain (Pierce).

The SDS-PAGE PARP1 activity assay in the presence of HPF1 was performed as described (Langelier et al., 2021). PARP1 full-length WT and mutants (0.4 μM) were preincubated with 0.4 μM HPF1 and 4 μM dumbbell DNA with a central nick. NAD⁺ (180 μM) was added and reactions were incubated for 5 min before stopping the reaction with PARP inhibitor veliparib at 180 μM. Where indicated, reactions were treated with 0.4 M hydroxylamine (NH₂OH) for 1h at room temperature, then quenched with 0.12% HCl. SDS-PAGE loading buffer was added to the reaction prior to resolution on SDS-PAGE and stained with Imperial Protein Stain (Pierce).

Analytical gel filtration.

Protein co-elution by analytical gel filtration was carried out using an ÄKTA pure (GE Healthcare) FPLC. PARP1 WT and mutants (20 μM) were incubated with HPF1 (50 μM) or HPF1 (50 μM), a dumbbell DNA with a central nick (20 μM) and EB-47 (20 μM) for 20 min at room temperature. The complexes were passed over a 10/300 Superdex 200 Increase (GE healthcare) gel filtration column equilibrated in 25 mM HEPES pH 8.0, 150 mM NaCl, 1mM EDTA, and 0.1 mM TCEP.

Differential scanning fluorimetry (DSF) analysis.

DSF experiments were performed as described (Langelier et al., 2012) using 5 μM protein. Experiments were performed on a Roche LightCycler 480 RT-PCR in the following buffer: 25 mM HEPES pH 8.0, 150 mM NaCl, 1 mM EDTA and 0.1 mM TCEP. Experiments were performed in triplicate and a Boltzmann sigmoid was fit to the data to determine the TM values using Matlab (MathWorks).

Cell line transfection.

Cells were transfected with Lipofectamine 3000 (ThermoFisher, #L3000001) according to the manufacturer’s protocol. Briefly, cells were plated in 35-mm glass-bottom culture dishes (MaTek, #P35G-0.170-14-C) and transfected with 1 μg PARP1-GFP plasmid DNA the following day. The cells were imaged 24 hours post transfection.

Microirradiation assays.

Transfected cells were maintained in 10% FBS DMEM media at 37°C and 5% CO₂ in an incubation chamber mounted on the microscope. Imaging was carried out on an Andor Revolution system, ×100 water objective with micropoint at 365 nm. Each cell was irradiated at a spot with 1 μm diameter in the nucleus. For each experimental condition 15-20 individual cells were imaged, and all the experiments were repeated independently on different imaging days. Talazoparib was added to 100 nM final concentration 30 min prior to imaging.

Quantification and Statistical Analysis

Data from fluorescence polarization DNA affinity measurement assays were fitted to a 1:1 binding model using Matlab (MathWorks). Binding constants (K_D) reported in Table 1 were obtained by averaging the results of at least three independent experiments. The reported errors are the calculated standard deviations.

Data from SPR assays was processed using TraceDrawer (Reichert). Briefly, titration sensorgrams were buffer and reference subtracted, and fitted to a 1:1 binding model. Association (k_a), dissociation (k_d) and binding (K_D) constants reported in Table 1 were obtained by averaging the results of two separate titration sensorgrams (see Figure S1 for representative titration sensorgrams). The reported errors are the calculated standard deviations.

Data quantification of microirradiation assays. For each microirradiated nucleus the intensities of two areas were measured, the microirradiate spot and a neighbouring area of the same nucleus. Subsequently the intensity of the nuclear spot was subtracted from the intensity of the microirradiated spot. The maximum intensity difference for the whole time-lapse was set to 1. For each imaging day and experimental condition, 15-20 individual cells were analysed, and the data was presented as mean and standard deviation.

Supplementary Material

3

Supplementary Video S1. Mechanism for HD dynamics that drive PARP1 allostery. Related to main Figure 6. Cartoon representation of PARP1 apo CAT domain (PDB 7AAA in teal) morphing into PARP1 WT/DNA (PDB: 4DQY in blue) and PARP1 ΔVE/EB-47 (salmon). EB-47, in purple, is represented as spheres in the catalytic pocket. Important residues, such as L698, L701, E763, D766 and D770 are modeled as sticks. PARP1 DNA-binding domains (Zn1, Zn3 and WGR) are derived from PARP1 ΔVE/EB-47 and are represented as being static.

Highlights.

PARP1 crystal structures capture the active state of the dynamic helical domain (HD)

The active HD conformation reverses PARP1 autoinhibition and strengthens DNA binding

A new HD contact with WGR domain drives ART domain rotation to permit NAD⁺ binding

HD helix F senses ART active site occupancy with substrate NAD⁺ or PARP inhibitors