Domain architectures and phylogenetic relationships of PARP1/2/3
AtPARP1 is very similar to HsPARP1 in terms of its domain architecture (Additional file 1: Figure S1A). HsPARP1 is thought to be the founding member of the PARP family in humans [1, 3, 11]. Both HsPARP1 and AtPARP1 have five important domains with known functions arranged from the N to the C terminus: three N-terminal zinc fingers responsible for DNA damage detection, the BRCA-1 C-terminus (BRCT) domain for phospho-protein binding, the WGR domain with the conserved Trp-Gly-Arg (WGR) motif for nucleic acid binding, the PARP regulatory domain (PRD) or helical subdomain (HD) believed to regulate PAR-branching, and the C-terminal characteristic PARP domain with catalytic activity [3, 33] (Additional file 1: Figure S1A). Both HsPARP2 and AtPARP2 have no zinc fingers or BRCT domain but do have two SAF/Acinus/PIAS motif (SAP) domains in the N-terminal region that confer DNA-binding activity [11, 19, 32]. HsPARP3 has only three domains, i.e., the WGR, PRD/HD and PARP domains, and catalyzes mono (ADP-ribosyl) ation (mART), which attaches only one ADP-ribose unit to the target protein [34], whereas AtPARP3 has a long N-terminal region with unknown functions and bears BRCT, WGR, PRD/HD and the PARP catalytic domain.
The sequence analysis revealed that both AtPARP1 and AtPARP2 have a typical H-Y-E catalytic triad, whereas AtPARP3 has an alternative histidine-valine-glutamic acid (C-V-E) triad in its catalytic core (Additional file 1: Figure S1B). To understand the evolutionary relationships of the PARP family in different species, we constructed a phylogenetic tree using PARP1/2/3 homologs identified from twenty-eight organisms, including eighteen angiosperms, two gymnosperms, one lycopod, one moss, four metazoans and two fungi, using the PARP1 subfamily as an outgroup (Additional file 2: Figure S2). The phylogenetic tree topology revealed that the PARP1/2/3 subfamilies form two clades in eukaryotes, each with metazoan members, which suggests that the clade containing plant PARP2 and the clade with plant PARP1/3 originated from gene duplications before the common ancestor of eukaryotic organisms. The close homologs of plant PARP1 and PARP3 form separate phylogenetic groups, each involving lycopod and moss members, which indicates that these resulted from duplications before the divergence of extant land plants but after the separation of plants from metazoans and fungi.
AtPARP3 does not show auto-ADP-ribosylation activity in vitro
Because the activity of AtPARP3 is unknown, we mainly focused on AtPARP3. We first examined the biochemical activity of recombinant AtPARP3 protein expressed in E. coli. To avoid the possible influence of a fusion tag on protein function, we used two different expression vectors, pGEX-4 T-1 and pET-32a(+), which carry a glutathione S-transferase (GST) and a thioredoxin and histidine (TRXH) tag fused to the N terminus of the recombinant protein, respectively. These two tags reportedly facilitate the correct folding of the target proteins, particularly the cysteine-rich proteins. AtPARP1 was expressed by the pET-32a(+) vector as a positive control for the activity assay because it shows stronger in vitro activity than AtPARP2, as demonstrated in our previous studies [21]. The tag proteins were also expressed using the empty vectors for use as negative controls. We assessed their activities through a standard PARP activity assay based on the auto-modification feature of PARP proteins [35]. A sensitive ADP-ribose detection reagent that can detect both mono- and poly-ADP-ribose, called anti-pan-ADP-ribose binding reagent, was used to detect the ADP-ribose moiety on proteins (based on the certificate of analysis provided by Merck). It has been successfully used in a prior study on humans [36]. As shown in Fig. 1a, a strong PAR signal was generated by AtPARP1 within 1 min, and the band exhibited an upward smear typical of poly (ADP-ribosyl) ation, consistent with previous observations that PARP1 is a robust enzyme synthesizing PAR in seconds [35, 37]. The signal was abolished by the PARP inhibitor 3-AB, which indicated that it was a real PAR signal. However, no ADP-ribose signal was detected in the samples containing either GST-AtPARP3 or TRXH-AtPARP3. To exclude the possibility that AtPARP3 might need a longer reaction time, we extended the incubation time to 30 min, and no PAR signal was detected (Fig. 1b).
The AtPARP3 catalytic domain exhibits no activity
The PARP catalytic domain is important for PARP activity. The N-terminal DNA-binding domain can interact with and activate the C-terminal PARP catalytic domain after binding to DNA [33, 35]. Although AtPARP3 has BRCT, WGR, PRD/HD and PARP catalytic domains, it has no recognized DNA-binding domain in the N-terminal region (Additional file 1: Figure S1); in addition, a C-V-E motif replaces the classical H-Y-E triad in the catalytic core. To assess whether this domain has function in catalyzing PAR formation, we exchanged the catalytic domain of AtPARP1 with that of AtPARP3, as shown in Fig. 2a. The chimera proteins AtP1-P3 and AtP3-P1 were produced in the same expression system as AtPARP1 and AtPARP3. In contrast to the strong activity of AtPARP1, the AtPARP1 protein carrying the AtPARP3 catalytic domain (AtP1-P3) showed no PARP activity (Fig. 2b), which indicated that the AtPARP3 catalytic domain failed to form PAR even in the presence of other functional domains from AtPARP1.
However, the chimera PARP3 protein AtP3-P1, in which most domains of AtPARP3 were retained and only the catalytic domain was replaced by that of AtPARP1 (Fig. 2a), showed constitutive activity, and the PAR signal was detected at all time points, even at the time point “0” in the absence of exogenously supplemented NAD+ and DNA (Fig. 2b). 3-AB only slightly reduced the signal, which indicated that the signal had been produced in E. coli cells prior to purification, and NAD+ and DNA supplementation during the assay only mildly enhanced its activity. These results indicated that the catalytic activity of the AtPARP1 domain in AtPARP3 was constitutively “switched on.”
Other structural elements beyond the catalytic triad also determine activity
To understand whether the loss of activity of AtPARP3 is solely due to the change in “H-Y” of the catalytic triad, we mutated the “H-Y” of AtPARP1 to “C-V” (AtPARP1M) by point mutation and changed the “C-V” of AtPARP3 back to “H-Y” (AtPARP3M) (Fig. 3a), and then examined the activities of the AtPARP1 and AtPARP3 mutant proteins as well as those of their wild-type controls. No PAR signal was observed for either the AtPARP1M or AtPARP3M proteins (Fig. 3b and c), which indicated that mutation of the first two amino acids of H-Y-E in AtPARP1 is sufficient to eliminate the PARP enzymatic activity; however, H-Y-E is not the only important motif for determining the activity of AtPARP3 because even a typical H-Y-E triad failed to regenerate the PARP activity in AtPARP3.
Several other motifs are also considered to be important for PARP activity [31], and they varied in AtPARP3. In addition to the catalytic triad motif, AtPARP3 carried altered motifs 1 and 2 in the NAD+ fold (Additional file 1: Figure S1B). Histine-glycine-serine (H-G-S) in motif 1 and tyrosine-phenylalanine-alanine (Y-F-A) in motif 2 were replaced by cysteine-glycine-serine (C-G-S) and valine-phenylalanine-alanine (V-C-S), respectively, in AtPARP3, and these two motifs might provide a microenvironment for NAD+ binding.
AtPARP3 loses the ability to bind to NAD+
Substrate recognition is the prerequisite for an enzyme to perform activity. To understand the possible structural differences among Arabidopsis PARP enzymes, we first simulated the structures of Arabidopsis AtPARP1 and AtPARP3 using the resolved crystal structures of HsPARPs as models. Representative images illustrating the binding of the PARP catalytic domain (green) to the NAD+ molecule (white) are presented in Fig. 4a. The results indicated that the C and V residues in AtPARP3 failed to correctly orientate the NAD+ molecule relative to those of human and Arabidopsis PARP1. To investigate whether AtPARP3 retains the capability of binding to NAD+, we calculated the binding affinities between PARP proteins and NAD+ molecule using AutoDock software [39]. The binding affinity of AtPARP3 to NAD+ was significantly lower than those of other known active PARP proteins. Among the PARP proteins, AtPARP1, AtPARP2, HsPARP1 and HsPARP2 were verified as poly (ADP-ribose) polymerases; HsPARP3 is a mono-ADP-transferase; and HsPARP5a and HsPARP5b mediate oligo (ADP-ribosyl) ation [32, 40]. All of these proteins can bind to NAD+. The lower binding affinity between AtPARP3 and NAD+ implies that AtPARP3 might have lost the capacity to bind to NAD+.
To examine the binding activity of NAD+ to AtPARP3, we spotted the purified proteins onto a polyvinylidene fluoride (PVDF) membrane, using the irrelevant protein bovine serum albumin (BSA) and the fusion tag protein TRXH as negative controls, and incubated the membrane with biotinylated NAD+. If NAD+ bound to the proteins on the membrane, the biotin tag on NAD+ would allow detection of the NAD+ signal by streptavidin/HRP (Fig. 5a). An inner control anti-His antibody was used to visualize the amount of protein spotted on the membrane (Fig. 5b). The results showed that the active enzyme TRXH-AtPARP1 and the chimera TRXH-AtP3-P1 protein were able to bind to NAD+, whereas the protein with the AtPARP3 catalytic domains TRXH-AtPARP3 and TRXH-AtP1-P3 could not. Moreover, the inactive protein TRXH-AtPARP3M also showed NAD+-binding activity, which suggested that the reverse mutation of C-V-E to H-Y-E recovered the NAD+-binding activity of AtPARP3 but failed to recover its PAR-generating activity, and this finding further supported the notion that other motifs around the H-Y-E triad are also important for PARP enzymatic activity. Surprisingly, the AtPARP1 protein with the H-Y-E to C-V-E mutation still maintained NAD+-binding activity even though it was inactive, but the underlying molecular basis remains unknown.
AtPARP3 is not responsible for PAR formation in seeds
PARP modifies itself and other proteins by PARylation; thus, the PAR level in vivo reflects the cellular PARP activity directly. To investigate whether AtPARP3 is physiologically active, comparison of the PAR levels in wild-type plants and loss-of-function mutants of AtPARP3 is crucial. By examining the PAR levels in parp3 mutants and in wild-type plants, we could determine whether AtPARP3 is active in vivo.
We ordered T-DNA insertion mutants of AtPARP3 from TAIR and named them parp3–1 and parp3–2. Quantitative RT-PCR (RT-qPCR) and immunoblotting analyses confirmed that they were both null mutants (Additional file 3: Figure S3A-C). We also included multiple mutants of both AtPARP1 and AtPARP2 for comparison. For each gene, at least two different mutants were used, and all these mutants have been used in other studies [4, 20, 22,23,24, 27, 29]. The T-DNA insertion sites of the mutants for the AtPARP1 and AtPARP2 genes are shown in Additional file 3: Figure S3D and E.
Because AtPARP3 is abundantly expressed in seeds but is poorly expressed in other tissues [30] (Additional file 3: Figure S3B and C; Additional file 4: Figure S4A and B), it would be easier to detect its activity in seeds if AtPARP3 had PARP enzymatic activity. AtPARP3 is presumed to play an important role in repairing DNA damage caused by seed dehydration or storage prior to the re-initiation of cell division during germination [30, 31]. As such, it might be an active PARP enzyme in seeds. However, in dry or germinating seeds, no PAR signal could be detected under normal conditions, neither after treatment with genotoxin. The RT-qPCR data also showed that AtPARP1/2/3 expressions in the early seed germinating stages (within 24 h) were not responsive to zeocin or methyl methanesulfonate (MMS) treatment, which mainly induce double- and single-strand DNA breaks, respectively [41, 42] (Additional file 4: Figure S4C and D). Therefore, we modified our assay to a more robust one. Exogenous NAD+ and broken DNA were supplemented in the protein extracts because NAD+ is a known rate-limiting factor for PARP activity. In the presence of sufficient substrate (NAD+) and activating DNA, any residual PARP activity would be detectable. Surprisingly, we detected PAR formation in the seeds of wild-type, both parp2 mutants and both parp3 mutants, but not in the seeds of both parp1 mutants (Fig. 6a). The PARylated proteins displayed as large smeared bands on the SDS-PAGE gel due to the sufficient provision of NAD+ substrate, and the PAR chains were obviously long. The PAR signal strength detected in the two parp3 mutants was close to that of the wild-type, which indicated that AtPARP3 made no or an undetectable contribution to the PAR signal in seeds. In the parp1 parp2 (p1 p2) double mutant, no PAR signal was detected, which confirmed that AtPARP3 produced no PAR in vivo. Interestingly, in all mutants with the AtPARP1 mutation, such as the parp1, parp1 parp2, parp1 parp3 (p1 p3) and parp1 parp2 parp3 (p1 p2 p3) mutants, the PAR signal was undetectable, which indicated that even in seeds, AtPARP1 remains the main PARP responsible for PAR formation, although the expression level of AtPARP1 in seeds was much lower than that of AtPARP3 (Additional file 4: Figure S4A). To confirm the validity of the detected PAR signal in seeds, we added the competitive inhibitor 3-AB to the reaction samples. 3-AB eliminated the signal almost completely (Fig. 6b), which indicated that the detected signal was a real PAR signal.
AtPARP1 has predominant PARP enzymatic activity in response to zeocin and MMS treatments in seedlings
The results obtained in seeds aroused our interest to investigate whether AtPARP1 also plays the predominant PAR formation role in seedlings. In humans, HsPARP1 contributes over 90% of the total PARP activity in vivo and is considered the key member of the PARP family [32, 43]. However, in Arabidopsis, AtPARP2, instead of AtPARP1, is reportedly the predominant PARP involved in both DNA damage and biotic stress responses in seedlings [20, 22]. We first examined the expression levels of three PARP genes in Arabidopsis seedlings and found that AtPARP1 expression was highest while AtPARP3 expression was lowest under normal conditions (Additional file 4: Figure S4B). AtPARP3 could not be induced by zeocin and MMS in both seeds and seedlings (Additional file 4: Figure S4C, D, E, and F), whereas AtPARP1 and AtPARP2 transcriptions were strikingly induced by both zeocin and MMS in seedlings (Additional file 4: Figure S4E and F).
We then used the same parp mutants as those used in other laboratories to investigate PAR formation in response to the DNA-damaging agents zeocin and MMS. As expected, PAR was gradually induced in wild-type seedlings after treatment with genotoxin (Fig. 7a-d). All PAR signals originated from physiological PARP activity because the assay was performed without exogenous NAD+ and activating DNA. Overall, the PAR signal generated under zeocin treatment was stronger than that under MMS treatment (Fig. 7a-d), which indicated that the stronger double-strand breakage agent zeocin induced more PAR formation than the single-strand break-inducing agent MMS. Longer exposure times and higher concentrations of genotoxins also generated stronger PAR signals. Interestingly, little or no PAR signal was observed in the parp1–1, parp1–2 and parp1–3 mutants, and neither did we detect the automodification activity of AtPARP2 in the three parp1 mutants within 24 h after genotoxin exposure (Fig. 7e and f), whereas AtPARP2 was well induced at these time points in wild-type seedlings (Fig. 7g). Weaker PAR signals were detected in the parp2–1, parp2–2 and parp2–3 mutants than in Col-0, which indicated that the mutation of AtPARP2 decreased AtPARP1 activity. When grown on plates with MMS, all parp1 mutants displayed a mildly more sensitive phenotype than the parp2 mutants (Fig. 8c and e, please see Additional file 5: Table S1 for the source data); while on zeocin plates, although all seedlings could not grow up, the parp1 mutants had an averagely lower chlorophyll content than the parp2 mutants (Fig. 8d and f). Taken together, our biochemistry and genetic data support the conclusion that AtPARP1, instead of AtPARP2, makes the greatest contribution to PAR formation in Arabidopsis and plays a dominant role in DNA damage response.
In response to zeocin and MMS treatments, AtPARP1 is first activated, and then AtPARP2 is activated
The time-dependent results of PAR formation in response to DNA-damaging agents (Fig. 7a-d) indicated that a blurred band first appeared around the size of AtPARP1, which is approximately 113 kD. This blurred band disappeared in all parp1 mutants (Fig. 7e and f), which suggested that it was correlated with AtPARP1. AtPARP1 activation upon genotoxin attack was consistent with the observation in humans that HsPARP1 is rapidly activated by genotoxin [44]. At approximately 48 h after treatment with 200 μg/ml zeocin and 72 h after treatment with 200 μg/ml MMS, a second band, which was the size of AtPARP2 (72 kD), started to appear (Fig. 7b and d), and this band disappeared in the parp2 mutant (Additional file 6: Figure S5), which indicated that it likely corresponds to AtPARP2. As shown in Fig. 7g, before 48 h AtPARP2 was already highly induced by zeocin (Fig. 7g), but no PAR signal was detected on AtPARP2, which implied that although AtPARP2 was produced, it might not undergo auto-PARylation at this moment.
To further test whether AtPARP2 was modified by AtPARP1 or by itself in vivo, we examined the changes of the AtPARP2 protein by immunoblot analysis. It was difficult to observe PARylation-induced band shift under physiological conditions when detected by anti-PARP2 antibody, likely due to the negative charges of PAR which weakened the binding of the antibody to the modified proteins, or due to too limited PARylation of AtPARP2 to be observed. To amplify the smear signal typical of PAR formation, we added NAD+ to the samples; a very striking upward-blurred band was detected using anti-PARP2 antibody (Fig. 9a), and the upward traces were eliminated by 3-AB, which indicated that AtPARP2 was indeed poly (ADP-ribosyl) ated in the presence of exogenous NAD+. To assess whether the catalysis activity originated from AtPARP1 or AtPARP2 itself, we used the parp1, parp2, p1 p2 double mutant and parg1 mutant to observe the homeostasis of the PAR smears. The PAR signal was stronger in the parg1 mutant because PARG1 is mainly responsible for PAR hydrolysis in vivo, and the mutation of PARG1 allowed the easy detection of PARylated proteins (Fig. 9b). In the parg1 mutant, we observed very obvious PARylated AtPARP2 at 48 h after zeocin treatment; however, the smearing PAR signal on AtPARP2 was absent in the parp1 mutant, which indicated that most of the PAR signal on AtPARP2 was generated by AtPARP1, and not by itself. AtPARP2 might not undergo or only undergo very limited auto-PARylation under this condition. This result, together with the previous results that under physiological conditions (without exogenous NAD+) the PARylation of AtPARP2 in parp1 mutants was not detected (Fig. 7 and Additional file 6: Figure S5), suggested that under genotoxic stress, AtPARP1 is first activated and AtPARP2 is subsequently activated, and AtPARP2 is modified by AtPARP1 in vivo.