Arabidopsis thaliana PrimPol is a primase and lesion bypass DNA polymerase with the biochemical characteristics to cope with DNA damage in the nucleus, mitochondria, and chloroplast

PrimPol is a novel Primase–Polymerase that synthesizes RNA and DNA primers de novo and extents from these primers as a DNA polymerase. Animal PrimPol is involved in nuclear and mitochondrial DNA replication by virtue of its translesion DNA synthesis (TLS) and repriming activities. Here we report that the plant model Arabidopsis thaliana encodes a functional PrimPol (AtPrimPol). AtPrimPol is a low fidelity and a TLS polymerase capable to bypass DNA lesions, like thymine glycol and abasic sites, by incorporating directly across these lesions or by skipping them. AtPrimPol is also an efficient primase that preferentially recognizes the single-stranded 3′-GTCG-5′ DNA sequence, where the 3′-G is cryptic. AtPrimPol is the first DNA polymerase that localizes in three cellular compartments: nucleus, mitochondria, and chloroplast. In vitro, AtPrimPol synthesizes primers that are extended by the plant organellar DNA polymerases and this reaction is regulated by organellar single-stranded binding proteins. Given the constant exposure of plants to endogenous and exogenous DNA-damaging agents and the enzymatic capabilities of lesion bypass and re-priming of AtPrimPol, we postulate a predominant role of this enzyme in avoiding replication fork collapse in all three plant genomes, both as a primase and as a TLS polymerase.

www.nature.com/scientificreports/ orthologues are present in all members of the kingdom Plantae, including the green algae Micromonas pusilla and Bryophytes like Marchantia polymorpha (Fig. S3), suggesting the presence of this gene before the divergence between plants and animals. Throughout the divergence between plant and animal PrimPols have conserved the motifs to interact with replication protein A (RPA) (Fig. S2), implying that this interaction is conserved in the plant kingdom 35 . Although AtPrimPol and HsPrimPol exhibit a similar domain architecture, these enzymes present two notable differences in their catalytic core: (1) the linker region between motifs II and III of the AEP domain is 50 amino acids shorter in AtPrimPol 26 , and (2) the interdomain region between the AEP and Zn ++ finger binding domains is 18 amino acids longer in AtPrimPol (Fig. S2).
AtPrimPol is a DNA polymerase and a DNA primase. To investigate the predicted DNA polymerase and primase activities of AtPrimPol, we use recombinantly expressed AtPrimPol lacking its putative organellar targeting peptide (residues 1 to 86) that we dubbed wild-type, a mutant that deletes its Zn ++ finger subdomain (AtPPol∆Zf) (residues to 387-614), and a mutant defective in metal coordination at the AEP domain (D209A/ E211A) (AtPPol∆pol) using heterologous expressed proteins ( Supplementary Fig. S4A,B, S13). To investigate the putative primase activity of AtPrimPol, we used a single-stranded oligonucleotide that harbors the preferred template sequence for HsPrimPol, 3′-T (15) GTCCT (30) -5′, in which the underlined G is cryptic (necessary for recognition but not used as a template) 21 . In all enzymatic assays, we used a mixture of 10 mM MgCl 2 and 1 mM MnCl 2 , as these are the optimal conditions to assay the DNA polymerase activity of AtPrimPol (Supplementary Figs. S4C, S13). We did not use Mn ++ as a sole divalent cation because AtPrimPol recognizes alternative recognition nucleotide sequences under those conditions (data not shown). DNA primases from the DnaG and AEP superfamilies (i.e. bacterial DnaG and eukaryotic PriS/Prim1, respectively) synthesize short RNA primers for their cognate DNAPs 36 . In contrast PrimPol has a remarkable ability to synthesis DNA primers using dNTPs instead or NTPs 21,37 . Because of the traditional role of primases in synthetizing RNA primers, we decided to investigate if AtPrimPol was able to use solely NTPs as a substrate. In reactions labeled with [γ-32 P] ATP, in the presence of GTP and ATP, full-length AtPrimPol is the sole construct able to synthesize ribo-oligonucleotides. AtPrimPol is able to synthesize primers of 33 ribo-oligonucleotides, that correspond to the maximum length of the template substrate starting from the thymidine adjacent to the cryptic guanosine (Fig. 1B, lanes 1 to 3, Fig. S8). Both AtPPol∆Zf and AtPPol∆pol mutants are unable to carry out primer synthesis indicating a prevalent role of the Zn ++ finger in primer synthesis and the necessity of metal coordination for nucleotide incorporation (Fig. 1B, lanes 4 to 9, Fig. S8). To investigate DNA polymerization activity by AtPrimPol, we hybridized a 19-mer radiolabeled primer to a 28-mer template. At a 50-fold excess of AtPrimPol, this enzyme synthesizes products that reach the end of the template after an incubation period of 40 min (Fig. 1C, lanes 1 to 9, Fig. S8). AtPPol∆Zf efficiently performs template-dependent dNTP incorporation and at higher enzyme concentrations, this mutant performs the addition of one or two non-templated nucleotides (Fig. 1C, lanes 10 to 18, Fig. S8). As expected, the AtPPol∆pol mutant is unable to perform nucleotide incorporation 26 (Fig. 1C, lanes 19 to 27, Fig. S8). HsPrimPol preferentially starts de novo primer synthesis on single-stranded templates that provide a cryptic guanosine 21,27 . In order to test if AtPrimPol selectively uses a cryptic guanosine during priming synthesis we used a template in which the identity of the cryptic 3′-base was systematically modified as follows: 3′-ATC-5′, 3′-CTC-5′, 3′-GTC-5′, and 3′-TTC-5′ ( Fig. 2A, Fig. S9). In the presence of [γ-32 P] ATP and dGTP or rGTP, AtPrimPol executes 3pAdG synthesis in all substrates and as is the case for all primases, the initial nucleotide harbors a triphosphate group on the or 5′ of the growing chain. However, AtPrimPol only efficiently synthetizes the trinucleotide 3pAdGrA on the 3′-GTC-5′ substrate, indicating a moderate preference for the cryptic guanosine ( Fig. 2A, lanes 4 to 11).
In order to investigate the preference for the initial 5′ nucleotide, we executed an experiment in which the identity of the base that templates for the 5′ nucleotide was either adenine, cytidine, guanine or thymine and the initial 5′ nucleotide was either a ribonucleotide, deoxyribonucleotide or a mixture of ribo and deoxyribonucleotides (10:1 and 1:10) (Fig. 2B, lanes 1 to 17, Fig. S9). In this experiment, AtPrimPol was unable to start a primer with a 5′-UTP or 5′-dTTP (Fig. 2B, lanes 2 to 5), whereas moderate incorporation was observed when 5′-GTP or 5′-dGTP were used as initial nucleotides (Fig. 2B, lanes 6 to 9). When 5′-GTP or 5′-dGTP are used as substrates, they are competing with the labeled 32 P-dGTP decreasing the amount of labeled product. Even in large excess of competitor or the radioactive label, reactions incubated with AtPrimPol using 5′-GTP or 5′-dGTP generate elongated primers suggesting that 5′-GTP or 5′-dGTP can be used as initial nucleotides. 5′-dCTP or 5′-CTP were effectively used for AtPrimPol (Fig. 2B, lanes 10 to 13). Although, 5′-dCTP or 5′-CTP were used as initial nucleotides, the greater amount of primer formation was observed with ATP or dATP at the 5′ position, indicating that as HsPrimPol, AtPrimPol preferentially uses ATP or dATP as its initial nucleotide. A comparison between reactions incubated solely with ATP or dATP and labeled dGTP shows a similar amount of elongated primers but more nascent products of 2 and 3 nt are observed when using ATP (Fig. 2B, lanes 14 and 15). Our results suggest that AtPrimPol prefers ATP as initiating nucleotide over dATP because more primer products are synthesized in the presence of ATP 21,38 . Di and trinucleotides are more abundantly synthesized when ATP, but not dATP, is solely or majoritarily present in the reaction (Fig. 2B, lanes 14 and 16).
AtPrimPol localizes at the nucleus, mitochondria and plastids. In order to evaluate the sub-cellular localization of AtPrimPol, we constructed a protein fusion containing a GFP fused to the C-terminal of AtPrimPol under the control of 35S promoter (35S::PRIMPOL-GFP; Fig. 3). A close view of primary root shows that AtPrimPol-GFP localizes into the nucleus of cells at the root tip (Fig. 3A) www.nature.com/scientificreports/ of the same seedlings, we observed that AtPrimPol-GFP (Fig. 3B) colocalizes with the autofluorescence signal of chlorophyll from the chloroplasts of these cells (Fig. 3B). Finally, a close-up view of root hairs of seedlings stained with Mitotracker, shows a co-localization between AtPrimPol-GFP and mitochondria. (Fig. 3C). With the aim to corroborate the plastid and mitochondrial localization of AtPrimPol, we crossed independently plants from 35S::PRIMPOL-GFP lines with of 35S::COX4-CFP and 35S::NrRbcS-CFP lines. The former is a specific marker for mitochondria, while the latter is a marker for chloroplasts 39 . F3 seedlings from such crosses were analyzed and we confirmed that AtPrimPol-GFP colocalizes with COX4-CFP at the mitochondria in both hypocotyl cells AtPrimPol efficiently bypasses DNA lesions with low fidelity. PrimPols from H. sapiens and T.
brucei bypass DNA lesions like abasic sites (AP) and 8-oxo guanosine (8oxoG) 21,22,40 . Given the presence of reactive oxidative species and metabolic byproducts in organelles that potentially damage mitochondrial and plastid genomes, we investigated if TLS is a conserved property in AtPrimPol. We tested TLS on 8-oxoguanine (8oxoG), an abasic site (AP), and thymine glycol (Tg) using a 40-fold excess of enzyme over primer-template and AtPolIB exo − as a positive control (Fig. 4A, Fig. S10). Under these experimental conditions, AtPrimPol is able to bypass an abasic site, however with limited efficiency in comparison to an undamaged template (Fig. 1C, 4A lanes 1 to . Reactions were carried out in the presence of ATP and GTP or dGTP. Primer synthesis was mainly observed in reactions contained dGTP and longer products were only synthesized on a template containing the 3′-GTC-5′ motif. A control reaction showing primer synthesis (pAC and pACA products) by T7 Primase is used as molecular weight marker (lane 3). (B) AtPrimPol prefers rATP in its 5′ position. Primer formation was evaluated on four templates containing the 3′-GAC-5′, 3′-GCC-5′, 3′-GGC-5′, and 3′-GTC-5′ sequences. Each of these templates were assayed with U/T, G/dG, C/dC and A/dA respectively at 100 μM, and 100:1 μM. Primer synthesis on templates containing the CAG and CCG sequences was inefficient, whereas templates containing the 3′-GGC-5′ and 3′-GTC-5′ sequences readily primed and extended beyond the length of the template. Original data used to compose this figure is present in Fig. S9. www.nature.com/scientificreports/ 5). The length of the extension product synthetized by AtPrimPol is one nucleotide shorter that the expected full extension product of 28 nts synthesized by the control AtPolIB exonuclease deficient (exo − ) 14 (Fig. 4A, lanes 1 to 6). This suggests that the ability to skip an AP site is a conserved feature in PrimPols 38,41 . AtPrimPol synthesizes a full-length product at an 8oxoG template and with greater efficiency than an AP site ( Fig. 4A, lanes 7 to 12) as is the case for its human counterpart 38,41 . Thymine glycol (Tg) is a strong block or replicative DNAPs 42 . Two independent groups have shown that HsPrimPol is blocked at a Tg using Mg ++ as cofactor 22,43 . However, HsPrimPol bypasses a Tg lesion using Mn ++ (41). In our hands, AtPrimPol is able to bypass both (5R, 6S) and (5S, 6R) Tg isomers and perform full-length primer extension (Fig. 4A, lanes 13 to 24). HsPrimPol executes TLS across a cyclobutane pyrimidine dimer (CPD) and a (6-4) pyrimidine pyrimidines (6-4 PP) by skipping the lesion 24 . To investigate if this property is conserved in AtPrimPol, we tested TLS across CPD and 6-4 PP using the same substrates previously used for HsPrimPol 24 . AtPrimPol bypassed both CPD and 6-4 PP, however the main product was 6 nts shorter than the expected full-length product of 30 nts 38 . The abilities of HsPrimPol and AtPrimPol to performs TLS at UV-generated lesion are related to primer reannealing to microhomologous template regions beyond the lesion, mediated by PrimPol 38 .

Scientific Reports
As TLS at an AP generates a product that is one nucleotide shorter than the predicted full-length product, we decided to investigate the identity of the nucleotide that is incorporated when AtPrimPol encounters an abasic site. We performed a reaction using individual nucleotides at an AP (Fig. 4B, lanes 1 to 5, Fig. S10). When we used a primer paired just before the AP site, we found that AtPrimPol incorporates dAMP, dTMP, and dGMP with similar efficiency and is not able to incorporate dCMP (Fig. 4B, lanes 1 to 5). When we repeated this experiment using another template sequence harboring the same AP site but a different template sequence (Figs. S6A,B, S14), we noticed that adenine and guanine were incorporated with the same efficacy, suggesting that AtPrimPol follows either the A-rule (inserting dA opposite the abasic site 44 ) or copies the next templating base (dC), generating a -1 deletion product. Thus, AtPrimPol may use template bases in a primer realignment mode to direct primer templated synthesis or direct synthesis according to the context of the template sequence 38 . As previously demonstrated by Martínez-Jiménez and collaborators HsPrimPol bypasses AP sites by using the next available www.nature.com/scientificreports/ template as a substrate 38 . In contrast to the non-instructive character of an AP site, AtPrimPol preferentially incorporates dAMP or dCTP opposite an 8oxoG (Figs. 4B, lanes 6 to 10, S10). AtPrimPol incorporates all dNTPs opposite a Tg lesion. A detailed observation of TLS abilities at a Tg lesion show the presence of a 24 nts product and a full-length product of 30 nts. The appearance of the 24nts product is related to primer realignment in which different bases direct the character of the incoming nucleotide ( Fig. 4B, lanes 11 to 20, Fig. S10). The fact that AtPrimPol incorporates dNTPs with similar efficiency at an abasic site and that the length of the product is one nucleotide shorter than the full-length product on a canonical or an 8oxoG template, prompted us to investigate the relative efficiencies by which AtPrimPol skips the abasic site or uses the next template base. To investigate this question, we took advantage that more than 95% of an AP in E. coli is repaired before it can be used as a template 45,46 . In this experiment, we used a primer hybridized to a template harboring a modified coding sequence for eGFP. Primer extension was executed on a template containing an ATG codon that codes for the Tyr chromophore or in a template in which this codon is replaced by an A(AP)G-like codon. If a DNA polymerase incorporates a dAMP, accordingly to the "A-rule", opposite to the AP site, the A(AP)G-like codon would code for Tyr and fluorescence will be established. In a parallel experiment, we used a primer hybridized to a templated harboring A(AP)TG-like codon. If a dAMP or other base is inserted opposite the AP site, the chromophore will not be assembled, and white colonies would be observed. However, if the AP site is skipped and a dAMP is incorporated across the template, the ATG codon would be restored, and green colonies would be observed (Fig. S6C,D). An A(AP)G containing oligonucleotides amplified by AtPolIA exo − yield 90% of green colonies (355 out of 394) and 2% of emerald colonies (9 out of 394), in contrast oligonucleotides extended by AtPrimPol only yield 5% of green colonies (14 out of 295) (Fig. 4C). An experiment using an A(AP)TGlike codon with containing oligonucleotides amplified with AtPolIA exo − results in only 2% of green colonies, whereas its amplification by AtPrimPol resulted in 62% of green colonies. Overall the data indicates that AtPolIA exo − preferentially inserts an Ade opposite and abasic site and has little propensity to skip this lesion, on the other hand AtPrimPol preferentially skips an abasic site and use the next available base (T) as a template. The observed 38% of white colonies may correspond to the repaired A(AP)TG-like codon that is converted to an ATTC-like codon that shifts the open reading frame of eGFP (Fig. 4C). A control experiment in which the A(AP) www.nature.com/scientificreports/ G templated, is annealed with an oligonucleotide harboring the TAC sequence results in the appearance of 99% of green colonies, indicating that the AP is readily repaired and that AP site is substituted by a T (Fig. S6E) and as expected a control experiment in with the ATG codon is replaced by ATT-codon (Ile) annealed to a "TAA" yields only white colonies. An experiment in with the ATT-codon (Ile) and an oligonucleotide harboring the TAC sequence are annealed produces 49% and 51% of green colonies indicating that both DNA strands are independently replicated. This data resembles the 60% of polymerization errors observed in a β-galactosidase derived system using a single-stranded M13 template 47 . To investigate the fidelity during nucleotide incorporation by AtPrimPol we used four primer templates in which the identity of the next template bases changes using Mg ++ or Mn ++ as cofactors. AtPrimPol preferentially incorporates dNMPs following the Watson-Crick rules, however, AtPrimPol also misincorporates specially in the presence of Mn ++ (Figs. S7, S15).
AtPrimPol executes MMEJ with limited strand-displacement activities. In order to investigate a possible role of AtPrimPol in MMEJ we assayed primer-templates resembling partially resected DSBs (pss-DNA) 48 . These substrates contain microhomologous regions of 2, 4, 6 and 8 base pairs. These substrates have annealing temperatures higher than 37 °C and if annealed they can synthesize products of 54 and 58 nts 15,48 .
AtPrimPol and the AtPPol∆Zf mutant are able to use a pssDNA with 2 nts of microhomology as a template (Fig. 5A, Fig. S11). As is the case for other DNA polymerases, AtPrimPol uses pssDNA substrates more efficiently as the length of their microhomologous region increases (Fig. 5A, Fig. S11) 15,48 . AtPrimPol incorporate few nucleotides at the double-stranded junction of the annealed pssDNA, indicating that the strand-displacement properties of this enzyme are insufficient to completely remove the annealed non-template strand (Fig. 5A,  Fig. S11). This phenomenon is better observed on templates with 6 and 8 nts of microhomology as extension products of few nucleotides beyond the microhomologus region are synthesized specially in reactions incubated with the AtPPol∆Zf mutant (Fig. 5A, lanes 8,9,11,12).
To further corroborate the moderate (or minor) strand-displacement activities of AtPrimPol and the AtPPol∆Zf deletion mutant, we used a primer-template arrangement in which the primer strand and the blocking oligonucleotide are separated by 1, 6, and 16 nts and assayed extension reactions side by side using AtPolIB exo-, a DNA polymerase that exhibits strong strand-displacement (Fig. 5B, Fig. S11) 17 . AtPrimPol and the AtPPol∆Zf deletion mutant fill the gap of the 1-nt gapped substrate before the blocking oligonucleotide. AtPrimPol is unable to displace the blocking oligonucleotide, whereas AtPPol∆Zf displaces the lesion albeit with limited efficiency (Fig. 5B, lanes 2 to 5). This limited efficiency contrast with the strong strand-displacement exhibited by AtPolIB exo−, that efficiently synthesized a full-length product after displacing the blocking nucleotide (Fig. 5B, lanes  6 and 7). This pattern is maintained on gapped substrates with gaps of 6-nts and 16-nts (Fig. 5B, lanes 8 to 19), corroborating the limited efficiency of full-length AtPrimPol in displacing a blocking oligonucleotide and that the removal of the Zn ++ finger increases the strand-displacement abilities of this enzyme. The deletion of the Zn ++ finger subdomain in AtPrimPol promotes a limited or minor strand-displacement activity, as this modified enzyme is able to add few nucleotides on templates with microhomologous sequences and templates that have an oligonucleotide directly blocking the advance of the polymerase (Fig. 5A,B). We attributed that the added nucleotides are consequence of limited strand-displacement and not to enhanced DNA polymerization activity, as the AtPrimPol and AtPPol∆Zf present similar nucleotide incorporation activities on a primer-template without a blocking oligonucleotide (Fig. 1C).
AtPrimPol executes MMEJ preferentially on single-stranded templates. As AtPrimPol is deficient in strand-displacement, we asked if AtPrimPol would use a single-stranded template with a microhomologous region as a substrate. Both AtPrimPol and the AtPPol∆Zf are able to use a single-stranded substrate with a microhomologous region of 8-nts (Fig. 5C, Fig. S11). HsPrimPol is recruited to nuclear and mitochondrial DNA by RPA and mtSSB, respectively, albeit is not clear how human mtSSB interact with HsPrimPol (23,30). Besides two canonical mtSSBs, plant mitochondria harbor two different families of single-stranded binding proteins enzymes dubbed Organellar Single-Stranded Binding protein (OSB) and Whirlies [49][50][51][52][53] . In order to study a possible interaction between AtPrimPol and plant SSBs, we investigated if plant mitochondrial SSBs (AtmtSSB1 and AtmtSSB2), plant mitochondrial OSB2 (AtOSB2) and Whirly 2 (AtWhy2) physically interact with AtPrimPol by thermophoresis. All mtSSBs at a concentration of 2 μM were unable to interact with AtPrimPol (data not shown). Under our experimental conditions AtPrimPol converts 100% of the MMEJ-annealed template into a full-length product of 110 nts. As mtSSBs and AtPrimPol do not interact, mtSSBs may modulate MMEJ by preventing the access of AtPrimPol to single-stranded DNA. The addition of AtmtSSB1 and AtmtSSB2 does not have an effect on the MMEJ abilities of AtPrimPol. In contrast, both AtOSB2 and AtWhy2 at concentrations greater than 100 nM impinge MMEJ (Fig. 5D, Fig. S11). This inhibition can be related to higher binding affinity of AtOSB2 and AtWhy2 to ssDNA in comparison to AtmtSSB1 and AtmtSSB2 15 . The regulation of AtPrimPol supports a role of mitochondrial SSBs in reducing genomic rearrangements in plant organellar DNA 35,54 . AtPrimPol generate primers that are elongated by plant organellar DNA polymerases. We were curious to test if AtPrimPol could synthesize DNA primers for plant organellar DNA polymerases, AtPo-lIA and AtPolIB-The latter as an alternative to the synthesis RNA primers by the plant organellar primasehelicase, AtTwinkle, during the assembly of a T7-like replisome in plant organelles 16,55 . Aiming to investigate if AtPrimPol synthesizes primers that are elongated by AtPolIs, we used a single-stranded circular M13 in the presence of fixed concentrations of AtPrimPol and AtPolIs (Fig. 6A, Fig. S12). AtPrimPol, AtPolIA and AtPolIB alone are unable to synthesize a large polymerization product (Fig. 6A, lanes 2, 3, 4). However, AtPrimPol in combination with either AtPolIA (Fig. 6A, lanes 6 to 9) or AtPolIB (Fig. 6A, lanes 11 to 14) synthesize an initial product of approximately 7.2 kbp. In the case of AtPolIA, this initial product reaches nearly 20 kbs and in the www.nature.com/scientificreports/ case of AtPolIB it reaches 10 kbs. These phenomena correlate with the preferred extension of primers generated by AtTwinkle by AtPolIA in comparison to AtPolIB 14 and the proposed role of AtPolIA in DNA replication and AtPolIB in DNA repair 56 . The primer activity of AtPrimPol has the potential to create new origins of replication, transversions and deletions. Thus, its activity has to be regulated. Plant organellar single-stranded DNA binding proteins, mtSSBs, Whirlies, and OSB decrease the formation of rearrangements at microhomologous sequences and short repeats 56 AtPrimPol activity may be regulated by blocking its access to single-stranded DNA and alternatively, plant organellar single-stranded DNA binding proteins may block primer extension by AtPolIs. Here we measured the formation of elongation products on a single-stranded circular M13 by AtPrimPol and AtPolIA in the presence of increasing concentrations of AtmtSSB1, AtOSB2, and AtWhy2 using protein concentrations that rank from 0.5 to 5 μM (Fig. 6B, Fig. S12). Assuming that all proteins bind to a segment of 10-pb of single- www.nature.com/scientificreports/ stranded DNA, a concentration of 0.72 μM of these proteins would be sufficient to fully coat the single-stranded M13 template. All single-stranded binding protein AtmtSSB1, AtOSB2, and AtWhy2 impinge the synthesis of long DNA products by the combined action of AtPrimPol and AtPolIA. This blockage is specially observed in AtOSB2 and AtWhy2 that bind ssDNA with a nanomolar affinity. The residual DNA extension activity observed in reactions incubated by AtmtSSB1 could be due to its weaker binding to single-stranded DNA with respect to AtOSB2 and AtWhy2 15 .

Discussion
Photosynthetic organisms, like algae and flowering plants, are constantly exposed to the sun and therefore to UV irradiation. Plant harbor TLS DNAPs like DNA polymerase η and DNAP ζ that cope with UV-mediated DNA damage, especially in stem cells 8,57,58 . PrimPol is the most recent DNA polymerase identified in metazoans. Its evolutionary relationships with the AEP superfamily of primases agrees with its unique role in repriming stalled replication forks by synthesizing "de novo" DNA primers 21,21,[28][29][30][31][32]59 . According to the postulate "PrimPol is an ancient eukaryotic gene" 21 , herein we show that his enzyme was present in eukaryotes before the divergence between plants and animals. HsPrimPol is important in maintaining replication forks in the nucleus and avoiding sensitivity to DNA lesions like UV light. Its role in the nucleus is specially visible in the absence of DNA polymerase η 60 . Metazoan PrimPols harbor both nuclear and mitochondrial localization and PrimPols in plants have added a targeting peptide for mitochondrial and chloroplastic localization (Fig. 1, Fig. S2). Plant and metazoan PrimPols share a high degree of similitude in their APE and Zn++ finger subdomains and their main differences in primary structure are located at the C-terminal RPA binding region. In humans RPA is a heterotrimer composed of the subunits dubbed RPA70, RPA32 and RPA14 61 . Human PrimPol binds to RPA70 via two binding motifs dubbed RPA-binding motifs A and B, in which RPA-binding motif A plays a predominantly role using hydrophobic and negatively charged amino acids 35 . Plant PrimPols shows conservation in the RPA-binding  [13][14][15][16][17], showed that this protein efficiently blocks or inhibit the replication from 620 nM where the half of product was reduced. In the AtOSB2′ s reactions (lanes 8-12) the fall of de product is more slowly, even at the highest concentration some product was produced. Whereas the reactions with AtmtSSB1 (lanes 3-7) showed that AtPrimPol-AtPolI were able to replicate the ssDNA vector while the concentration of AtmtSSB1 was increasing. Original data used to compose this figure is present in Fig. S12. www.nature.com/scientificreports/ motif A, but not the RPA-binding motif B (Fig. S3). The lack of conservation between RPA binding motifs may be related to numerous events of gene duplication of RPA70 in Arabidopsis that contains five RPA70 paralogs 62 . The duplication events of RPA32 and RPA14 could produce up to 20 RPA heterotrimers in Arabidopsis 63 . AtPrimPol as a DNA polymerase bypasses strong-blocking lesions like AP site, CPD, and 6-4 PP by skipping the lesion. Using a GFP reporter we show that AtPrimPol almost exclusive skips an abasic site using the next templated (or templating) base. (Fig. 4A). The proposed MMEJ abilities of AtPrimPol were corroborated using pssDNA substrates (Fig. 5). As DNA polymerases θ and λ, AtPrimPol is able to use a pssDNA containing only 2 nts of homology as substrates. This reaction is more effective in polymerases lacking the Zn ++ finger domain, suggesting the possibility that alternative splicing may regulate MMEJ in vivo, as two of the three putative AtPrimPol isoforms lack its Zn ++ finger. Although AtPrimPol is able to use pssDNA substrates, this enzyme is not able to displace a blocking oligonucleotide, but could be stimulated by other accessory proteins like in HsPrimPol. Arabidopsis plants lacking a functional AtPolIB in combination with insertional lines in the genes encoding for AtWhy1 and AtWhy3 increase the formation of MMEJ re arrangements in chloroplast 56 . Interestingly this phenomenon is not observed in insertional lines of AtPolIA. The reduced fidelity during nucleotide incorporation of AtPolIB in comparison to AtPolIA, suggests a predominant role of AtPolIB in DNA repair rather than in replication 13 . As AtPolIA and AtPolIB have similar efficiency for MMEJ 15 , it is plausible that in plant organelles in the absence of AtPolIB, AtPrimPol is overexpressed leading the establishment of MMEJ that are further elongated by AtPolIA.
Human PrimPol is primordial to promote mitochondrial DNA replication, even in the absence of DNA damage 59,60,64 . Plant mitochondrial genomes suffer extensive and rapid rearrangements that could lead to the evolution of new genes in a phenomenon often associated with cytoplasmic male sterility (CMS) [65][66][67][68][69] . These genes are primarily originated by homologous recombination; however, it is clear that MMEJ and non-homologous endjoining (NHEJ) play a role in their formation 70,71 . Given that AtPrimPol can use single and pssDNA substrates with limited microhomology and its mitochondrial localization, we were curious to test if mitochondrial singlestranded binding proteins (mtSSBs) specific to plant mitochondria could modulate its MMEJ activity. In contrast to metazoan mitochondria, plant organellar harbor three different families of mtSSBs: (1) a duplicated copy of canonical SSBs (AtmtSSB1 and AtmtSSB2, (2) four paralogs of a protein dubbed Organellar Single-Stranded Binding protein (OSB) and (3) three paralogs of a protein dubbed Whirly 50,51,72 . We tested the effect AtmtSSB1, AtmtSSB2, AtOSB2) and AtWhy2, as these proteins are located in mitochondria, in regulating MMEJ promoted by AtPrimPol. Both OSB2 and Why2 hamper MMEJ by AtPrimPol, a phenomenon that correlates with the role of these proteins in avoiding re-arrangements at microhomologous regions 49,52,56 . The ability AtOSB2 and AtWhy2 in avoiding primer extension mediated by AtPrimPol also correlates with the role of these proteins in avoiding DNA re-arrangements in plant organelles and the ability of AtPrimPol to synthesize primers that are elongated by organellar DNA polymerases made it suitable in avoiding replication fork collapse.
In sum, our data suggests that AtPrimPol plays a predominant role in securing correct DNA replication in plant genomes, especially in organelles that lack nucleotide excision repair pathway necessary to remove bulky adducts like CPD of 6-4 PP.

Materials and methods
Cloning and purification. The coding region of AtPrimPol was synthetically optimized for its expression in bacteria (Biomatik). The coding region corresponding to residues 87 to 614 was PCR amplified and subcloned between Nde I and Bam HI restriction sites of a modified pET19b vector. Codon-optimized AtPrimPol was recombinantly induced expressed at 14 °C in an Escherichia coli BL-21 (DE3) strain supplemented with a pKJE7 plasmid. For protein purification, one liter of cell culture was harvested and resuspended in 30 ml of lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 10% glycerol, 5 mM imidazole and 1 mM PMSF) supplemented with 0.5 mg/ml lysozyme, and incubated at 4 °C for 30 min. Cells were lysed using two cycles of freeze-thawing. The lysate was centrifuged at 12,000 rpm for 45 min at 4 °C. The cell-free extract was purified by immobilized metal affinity chromatography (IMAC) using a 1 ml pre-packed column. The clarified lysate was loaded onto the IMAC column and the column was extensively washed with 50 ml of lysis buffer using an imidazole gradient (from 10 to 150 mM). AtPrimPol was eluted with 3 ml of lysis buffer supplemented with 500 mM imidazole and dialyzed in 50 mM Tris-HCl pH 8, 50 mM NaCl, 1 mM DTT, 10% glycerol and 1 mM PMSF. The dialyzed protein was loaded onto a heparin column equilibrated with dialysis buffer and washed with 100 volumes of a NaCl gradient (from 20 to 300 mM). AtPrimPol was eluted with 2 ml of dialysis buffer supplemented with 1 M NaCl. The eluted protein was dialyzed in 50 mM Tris HCl pH 8.0, 500 mM NaCl, 1 mM DTT, 1 mM PMSF, and 50% glycerol and stored at − 20 °C. The AtPrimPol ΔZnF mutant, that corresponds to residues 87 to 387 as this deletion eliminates the linker region, Zn++ finger and RPA binding domain, was subcloned using BamHI and NdeI restriction sites into the pCri-1b vector. AtPrimPol ΔZnF mutant was expressed and purified as the wildtype AtPrimPol.  www.nature.com/scientificreports/ 5′-CAC TGA CTG TAT GAT G-3′ was annealed with a 30-mer 5′-CTC GTC AGC ATC TXC ATC ATA CAG TCA GTG-3′, where X indicate 5R-Tg or 5S-Tg, and for CPD or 6-4 PP photoproducts the same 16-mer radiolabeled primer was annealed with a 30-mer 5′-CTC GTC AGC ATC XXC ATC ATA CAG TCA GTG-3′. TLS reactions contained 10 nM of dsDNA substrate were mixed with 400 nM of AtPrimPol or 20 nM of AtPolIB exo − and pre-incubated for 5 min at 37 °C. Reactions were initiated by addition of 200 μM dNTPs and stopped at 5, 10, 20 and 40 min and for AtPolIB exo − was a single time of 10 min. Reaction mixtures were run on a 17% denaturing polyacrylamide gel and visualized by phosphor imagery.
Fidelity reactions. First, we assayed nucleotide fidelity incorporation using four undamaged dsDNA template, where the base A, T, G or C in the template is immediately after 3-OH end primer. A typical nucleotide insertion reaction was initiated by addition of 100 μM of dATP, dGTP, dTTP or dCTP in presence of 5 mM MgCl 2 and stopped at 10 min by the addition of an equal amount of stop buffer. The same assay was performed in presence of 1 mM MnCl 2 . Single nucleotide insertion reactions were also performed using 10 nM of dsDNA substrate (AP site, 8-oxoG, 5R-Tg or 5S-Tg) mixed with 200 nM of PrimPol full-length in a buffer containing 10 mM Bis-Tris-Propane-HCl pH 7, 10 mM MgCl 2 , 1 mM DTT, 1 mM MnCl2 and pre-incubated for 5 min at 37 °C. Individual reactions were initiated by addition of 100 μM of dATP, dGTP, dTTP or dCTP and stopped at 10 min.
In vivo GFP approach fidelity assay. To test the nucleotide incorporation opposite an AP site, we carried out a typical polymerization assay, mixing AtPolIA exo − or AtPrimPol with a DNA substrate constructed with a 38-mer primer annealed to a 69-mer template harboring a modified coding sequence for eGFP. The oligonucleotide template containing an ATG codon that codes for the Tyr chromophore or another template in which this codon is replaced by an A(AP)G-like codon. To test if AP site is skipped by AtPrimPol, we used a 38-mer template annealed to a 70-mer where we put a T beside AP site resulting the A(AP)TG-like codon. Reactions were stopped by heating at 95 °C for 5 min. Products purified were digested with Spe I and Kpn I and each one of them were individually ligated into the modified plasmid pET19b harboring an eGFP sequence previously digested by the same enzymes to removing the chromophore zone. All constructions were transformed into the E. coli strain BL21 GOLD (DE3) and plated in LB media containing 100 μg/ml of ampicillin and incubated at 37 °C for 20 h and stored at 4 °C for 4 days to observe the green color in colonies overexpressing eGFP protein.
The percentage of green, emerald or white colonies was determined. Sequencing of individual colonies was used to determine the nucleotide inserted opposite the AP site, or if the AP site was omitted.
Microhomology and strand-displacement assays. All reactions were performed in Tris-HCl pH 7.4, 10 mM NaCl, 1.5 mM DTT, 0.1 mg/ml BSA, 10% glycerol, and 10 nM of 5 radiolabeled substrate. Enzymes were preincubated by 10 min at 37 °C and the reactions was initiated using 200 μM of dNTPs and 10 mM of MgCl 2 and 1 mM MnCl 2. The reactions were stopped using a stop buffer (95% formamide, 10 mM EDTA, 0.1% bromophenol blue and 0.1% xylene cyanol). All reactions were loaded onto a 17% polyacrylamide denaturing gel and analyzed by phosphorimaging.
Coupled AtPrimPol-AtPolI reactions. AtPolIA, AtPolIB and plant organellar single-stranded binding proteins were recombinantly expressed and purified to homogeneity as previously reported 14,15 . A coupled primase-DNA polymerase reaction containing 200 nM of AtPrimPol and 200 nM of AtPolIA or AtPolIB was performed as follows: the reactions were incubated at 37 °C during 0, 10, 20 40 and 60 min in the presence of dNTPs and labeled with 2 μCi [α-32P]-dGTP (3000 Ci/mmol) using as template a ssDNA M13 (1 nm) vector, reactions were stopped with EDTA and loaded in a 0.6% agarose gel. M13 ssDNA was incubated with increasing concentrations (0, 620, 1250, 2500 y 5000 nM) of organellar single-stranded binding proteins AtmtSSB1, AtOSB2 and AtWhy2 at RT during 30 min, afterwards a mixture of 200 nM of AtPrimPol and AtPolIA was added and the reaction was incubated at 37 °C for 40 min. Reactions were stopped with EDTA and loaded in a 0.6% agarose gel.

Generation of organelle marker lines. Mitochondrial and chloroplast markers lines plasmids used in
this study were obtained from the Arabidopsis stocks center (http:// www. arabi dopsis. org) and described in an article published previously 39 . For the co-localization analysis, transgenic Arabidopsis plants were generated using 3 week-old plants with 35S::AtPrimPol-GFP construct and transformed with plasmids based on a by the floral dipping method. Transgenic plants were screened on MS medium and glufosinate ammonium, the survived plants were transferred to soil in a greenhouse to harvest seeds. Homozygous lines were isolated at T3 generation. Fluorescence signals and images were taken using a confocal laser scanning microscope.
Statement on the use of plants. All