Mitochondria and chloroplasts (plastids) both harbour extranuclear DNA that originates from the ancestral endosymbiotic bacteria. These organelle DNAs (orgDNAs) encode limited genetic information but are highly abundant, with multiple copies in vegetative tissues, such as mature leaves. Abundant orgDNA constitutes a substantial pool of organic phosphate along with RNA in chloroplasts, which could potentially contribute to phosphate recycling when it is degraded and relocated. However, whether orgDNA is degraded nucleolytically in leaves remains unclear. In this study, we revealed the prevailing mechanism in which organelle exonuclease DPD1 degrades abundant orgDNA during leaf senescence. The DPD1 degradation system is conserved in seed plants and, more remarkably, we found that it was correlated with the efficient use of phosphate when plants were exposed to nutrient-deficient conditions. The loss of DPD1 compromised both the relocation of phosphorus to upper tissues and the response to phosphate starvation, resulting in reduced plant fitness. Our findings highlighted that DNA is also an internal phosphate-rich reservoir retained in organelles since their endosymbiotic origin.
Mitochondria and chloroplasts (plastids) originate from the endosymbiosis of ancestral alphaproteobacterium and cyanobacterium, respectively, approximately 1.5 billion years ago1. Reflecting this endosymbiotic origin is the retention of their own DNA genomes and transcription/translation machineries. However, during the evolution of eukaryotic cells, most genes from these endosymbionts have been transferred to the nucleus and only a small proportion of the ancestral genes remain within each organelle2,3,4. In the model plant Arabidopsis thaliana, for example, only 87 proteins are synthesized in chloroplasts, whereas all other constituent proteins are encoded in the nuclear genome5. Thus, present eukaryotes require the coordinated regulation between mitochondria or chloroplasts and the nucleus to fulfil organelle functionality6,7,8,9.
In contrast to their limited genetic capacity, organelle genomes of relatively small size are known to be highly abundant, with multiple copies in each organelle. A striking example is leaf mesophyll cells, in which chloroplast DNA (cpDNA) accounts for about 30% of cellular total DNA, with an estimated >1,000 copies per cell10,11,12,13. Typically, an A. thaliana mesophyll contains ~80 chloroplasts, resulting in >10 copies per chloroplast on average. Plastid DNA (ptDNA) copy numbers seem to vary in different species and in different plastid types, and they reach up to ~10,000 in developing leaves. As a consequence of the abundant DNA and protein synthesis, plastids contain a substantial amount of nucleic acids, which constitute a major pool of total cellular phosphorus in leaves14,15. Reportedly, chloroplast ribosomal RNAs account for the largest organic phosphorus pool, making up approximately half of the total nucleic acids pool15. The multiple copies of ptDNA represent a considerable phosphorus pool. Excess ptDNA can be dispensable without affecting organelle functionality or cell viability, potentially providing a source of organic phosphorus for relocation when degraded. However, whether the amount of cpDNA or ptDNA is controlled by degradation in mature leaves has long been unclear16,17. Little is known about the enzymatic degradation mechanism and its possible effect on the efficient use of the internal phosphorus pool in endosymbiotic organelles.
In reproductive organs, several nucleases targeted to the endosymbiotic organelles have been reported to digest DNA18,19. In animals, mitochondrial EndoG nuclease expressed during male gametogenesis has been reported, which secures maternal inheritance of mitochondrial DNA (mtDNA)20. In a green alga, uniparental disappearance of organelle DNA (orgDNA) during mating occurs, although the nuclease responsible remains unclear21. In flowering plants, we reported that the exonuclease DPD1 degrades orgDNA in male gametophytes22. However, DNA degradation by DPD1 per se does not contribute to maternal inheritance. Thus, we postulate that DPD1 has functions other than the control of maternal inheritance. In this study, we demonstrated that, in addition to its role in pollen, DPD1 degrades orgDNA in leaves undergoing senescence in which nutrients are recycled through various macromolecule degradation mechanisms. DPD1 presents a determinate mechanism of orgDNA degradation conserved in plants. Moreover, this orgDNA degradation was shown to affect the efficient use of phosphate positively when exposed to starvation conditions. We discuss a novel aspect of orgDNA, probably sensing phosphate availability and acting as an internal reservoir of phosphate, through degradation mediated by DPD1 in seed plants.
Exonuclease activity of DPD1 confined to DNA but not RNA
We have shown that DPD1 is conserved in angiosperms, but it was not detected in mosses or green algae, suggesting that it emerged during the evolution of flowering plants22. Our search in the PLAZA database allowed us to isolate 43 DPD1 homologues from 35 plant species (Supplementary Fig. 1). Consistently, no DPD1 homologues were present in microorganisms or bryophytes, although its presence extended to gymnosperms (coniferous plants such as Pinus, Picea, Pseudotsuga and Gnetum) (Supplementary Fig. 1), supporting this specific emergence of DPD1 in seed plants (spermatophytes).
DPD1, which exhibits exonuclease activity and is targeted to both mitochondria and plastids, is unique in that most of the cell death-associated nucleases identified previously in plants are S1-type or staphylococcal endonucleases23. Because these endonucleases digest both RNA and DNA when single stranded, we first tested whether DPD1 has substrate specificity. Our in vitro nuclease assay, conducted using a purified DPD1 carboxy-terminally fused to a histidine tag and synthesized oligonucleotides as substrates (Fig. 1a), demonstrated that DPD1 degraded only DNA and not RNA, irrespective of whether it was single stranded or double stranded (Fig. 1b). This activity depended on Mg2+ (Fig. 1c) and was inhibited when a substrate 3′ end labelled with a fluorescent dye (6-FAM) was used (Fig. 1b–d). We inferred that DPD1 is a 3′ to 5′ Mg2+-dependent deoxyribose-exonuclease, whose activity can be detected in physiological conditions equivalent to chloroplast stroma24 (Mg2+ concentration of >0.02 mM, temperature of 22 °C and pH 7–8; Supplementary Fig. 2). Given its heterogeneous forms, DPD1 alone can degrade at least a portion of orgDNAs processively, if they have a free 3′ end.
CpDNA degradation during leaf senescence
Our earlier survey of Arabidopsis transcriptome data predicted that DPD1 transcripts accumulate in senescing leaves as well as in pollen23. To examine whether DPD1 plays a role in vegetative tissues, detached Arabidopsis leaves were subjected to dark-induced senescence (see Methods), and orgDNA degradation was monitored by quantitative PCR (qPCR). CpDNA levels in wild-type (ecotype Columbia (Col)) leaves declined apparently as senescence proceeded during incubation in the dark (Fig. 2a). When normalized using haploid nuclear DNA levels, we estimated the cpDNA copy number before the onset of senescence as approximately 400–600, which was similar to that reported previously13 (Fig. 2b). After 5 days in darkness, DNA levels declined substantially to less than 100 copies. Concomitant with cpDNA level decline, our quantitative PCR with reverse transcription (qRT-PCR) analysis showed that DPD1 is upregulated (Fig. 2c), similarly or slightly earlier than senescence-related genes (Supplementary Fig. 3). Importantly, cpDNA levels did not decline in a dpd1 mutant and tended to stay constant (Fig. 2b). Retention of cpDNA in dpd1 mutants was verified using digital PCR (Supplementary Fig. 4) and cytological observations of senescing leaves (Fig. 2d). Taken together, we concluded that DPD1 degrades cpDNA during leaf senescence.
Although the mechanism for maintaining ptDNA quantity remains unclear, a defect in DNA replication has been shown to affect ptDNA copy number adversely25. DNA polymerase in plant organelles is a bacterial-type Pol I26. In Arabidopsis, two isoforms have been reported, Pol I-a and Pol I-b, of which Pol I-a plays the major role in ptDNA replication. Introduction of pol I-a2 into dpd1 seemed to decrease the copy number of cpDNA, whereas cpDNA levels stayed high during senescence (Fig. 2e). Thus, our results revealed an epistatic effect of DPD1-mediated cpDNA degradation over DNA synthesis.
MtDNA degradation during leaf senescence
We next examined whether mtDNA levels also declined in senescing leaves. The results showed a similar trend to cpDNA; mtDNA levels declined in Col leaves as senescence proceeded, although they tended to stay constant in dpd1 leaves. Thus, we concluded that DPD1 also degrades mtDNA during leaf senescence. However, the estimated copy number was found to be very low, ranging from around a few copies per nuclear DNA even before dark incubation (Supplementary Fig. 5a). To address whether mtDNA levels decreased during leaf maturation, we examined 2-week-old seedlings grown in Murashige and Skoog (MS) plates to estimate mtDNA copy number. The result showed that approximately 20 copies of mtDNA were detected in young seedlings (Supplementary Fig. 5b). We observed a slight increase of the mtDNA copy number (approximately 25) in dpd1 compared to Col seedlings, consistent with previous reports22. These results indicated that the mtDNA copy number declines before leaf maturation, which is independent of DPD1. Although the mechanism responsible for this mtDNA degradation remains unclear, the estimated copy number of mtDNA was consistent with previous reports describing that only a limited amount of mtDNA is detectable in mature leaves27,28. In contrast to this shortage in mtDNA, plant mitochondria are known to undergo active fusion and fission8,29. This dynamic behaviour of mitochondria might account for the proposed sharing of genomic information between each mitochondrion. We concluded that orgDNA degradation proceeds in both organelles, but the majority occurs in chloroplasts.
A weak stay-green phenotype in dpd1
A careful examination of senescing leaves revealed that dpd1 leaves displayed more greenness than Col leaves and a dpd1 line complemented by DPD1 (G31) (Fig. 3a,b and Supplementary Fig. 6). This stay-green phenotype defines DPD1 as a factor accelerating senescence and cell death. Conversely, prolonged chloroplast functionality might be detectable in the dpd1 mutant. To address this, we first assessed the expression levels of chloroplast genes. qRT–PCR analysis revealed that the senescence-dependent decline in chloroplast-encoded transcripts was retarded in dpd1 mutants (Fig. 3c and Supplementary Fig. 7). We inferred that the more abundant transcripts in dpd1 plants partly explained the stay-green phenotype and that cpDNA degradation with DPD1 resulted in a concomitant reduction in the chloroplast RNA pool. Subsequently, we tested whether the stay-green phenotype in dpd1 mutants prolonged chloroplast functionality. Photosynthetic activity, as measured by the carbon dioxide assimilation rate in the same attached leaves grown for 2 weeks, seemed to be maintained for longer in dpd1 leaves than in Col leaves (Fig. 3d,e). Together, these results confirmed that DPD1 accelerates senescence, although senescence still proceeds without DPD1.
Growth defect of dpd1 plants in phosphate starvation conditions
The synergistic action of DPD1 on leaf senescence led us to postulate that cpDNA degradation is associated with nutrient availability. As a trade-off between leaf longevity and nutrient deficiency, dpd1 prolongs photosynthesis, but it might impair growth in conditions with limited inorganic compounds. To test this possibility, we established hydroponic culture to grow Col and dpd1 plants (Supplementary Fig. 8) and observe the subsequent response to nitrogen or phosphate deprivation30. First, based on our standard hydroponic conditions (1/4 MS medium), we found that Col plants grew better than dpd1 plants as estimated from the weight of aerial parts (Fig. 4a). This result was unexpected because no apparent difference was observed previously when they were grown in soil. Supplementing the hydroponic media with additional phosphate rescued the defective growth of dpd1 plants (Fig. 4a), suggesting that dpd1 is specifically compromised in phosphorus availability. Additional phosphate did not have a significant effect in Col plants.
We next examined how Col and dpd1 plants respond to nitrogen and phosphate deprivation in our standard hydroponic culture (see Methods). In principle, both starvation conditions attenuated plant growth by reducing the weight of aerial parts (Supplementary Fig. 9a,b). However, dpd1 plants seemed to differ from Col plants in responding to phosphate deprivation. In nitrogen-deprived conditions, both lines showed a pale colour with slight anthocyanin accumulation, but no phenotypic difference was detectable. By contrast, phosphate-deprived conditions produced a profound growth defect in dpd1 plants, which was characterized by reduced growth and substantial accumulation of anthocyanin31 (Fig. 4b). Such typical symptoms of phosphate-deprived conditions were not observed in Col plants. These results suggested that the efficient use of exogenous phosphate was compromised in dpd1 mutants by the lack of orgDNA degradation. When exposed to nitrogen-deprived or phosphate-deprived conditions, qPCR demonstrated that cpDNA levels in Col leaves underwent degradation upon nitrogen or phosphate deprivation similarly to dark-induced senescence, whereas degradation was inhibited in dpd1 leaves (Fig. 4c,d).
Low fitness and phosphorus relocation of dpd1 in phosphate starvation conditions
To ascertain the compromised response to phosphate deprivation, we measured seed production in Col and dpd1 plants. First, seed numbers in plants grown in soil or hydroponic culture were counted. The hydroponic culture reduced seed production rate in Col plants, even in control conditions (1/4 MS), to approximately 80% of that grown in soil (Fig. 4e and Supplementary Fig. 9b). We also observed that the seed production rate was lower in dpd1 than in Col plants. Remarkably, phosphate deprivation reduced seed production, even in Col plants, to approximately 50% of that in the control conditions, whereas dpd1 plants consistently showed a greater reduction in seed set than Col plants. Although our data indicated that dpd1 plants exhibited lower fitness in phosphate-deprived conditions than Col plants, it was possible that this resulted from the delayed senescence in dpd1 plants, as evidenced by its weak stay-green phenotype. To compare this effect, we also measured seed set in nitrogen-deprived conditions. The results showed no significant difference in seed set between Col and dpd1 plants (Fig. 4e). Thus, nutrient starvation per se did not alter the sink capacity, instead dpd1 plants had a lowered fitness confined to phosphate-deprived conditions.
To examine whether the reduced fitness in dpd1 plants resulted from altered phosphorus remobilization, we measured the phosphorus content in leaves of Col and dpd1 plants in phosphate-deprived conditions. Assuming that degradation products of orgDNA contribute to seed set by relocating the catabolic products to reproductive organs, we expected to have lower phosphorus levels in the lower leaves of Col than in dpd1 plants. Inductively coupled plasma-mass spectrometry measurement of the total phosphorus concentration indeed showed that Col leaves at 2 weeks after phosphate deprivation relocated a significantly greater amount of phosphorus from the lower leaves (leaves pre-existing before phosphorus deprivation) to upper leaves (leaves that emerged after the start of phosphate deprivation treatment) (Fig. 4f). By contrast, no significant phosphorus relocation was detected in dpd1 plants. The adverse effect of phosphorus redistribution between Col and dpd1 plants in phosphate-deprived conditions was, as expected, shown to correlate with our fitness results. We concluded that orgDNA degradation contributes to efficient phosphorus relocation, particularly when plants face phosphorus-limited conditions.
Global response to nutrient starvation in dpd1 mutants
We next investigated global changes in the transcriptome in phosphate-deprived conditions using RNA sequencing (RNA-seq) (Fig. 5 and Supplementary Figs. 10 and 11). RNA was isolated from Col and dpd1 leaves either subjected to continuous growth in 1/4 MS or to phosphate deprivation (n = 3; the data set is presented as Supplementary Tables 1 and 2). Comparison of the gene expression profiles revealed that the response to phosphate starvation was profoundly altered between Col and dpd1 leaves (Fig. 5a). The number of genes that was differentially expressed upon phosphate deprivation treatment was 766 in Col plants, of which 655 genes were upregulated. By contrast, only 114 genes were differentially expressed in dpd1 plants; 96 genes were upregulated (Supplementary Fig. 10a). Col and dpd1 plants shared 99 genes, among which 86 upregulated genes had Gene Ontology terms related to phosphate starvation, photosynthesis, flavonoid biosynthesis and dephosphorylation.
To investigate these genes further, we specifically examined a set of genes that was categorized as related to the phosphate starvation response (PSR), mainly connected by a limited supply of inorganic phosphate in the root environment32 (Supplementary Table 3). Of 193 genes specified as being involved in PSR, we were able to extract 192 genes; among these, 123 and 40 genes were shown to be upregulated in Col and dpd1 plants, respectively, at the significance level of false discovery rate (FDR) < 0.05 (Fig. 5c). We also specifically examined the gene set that was reported as being under the control of the phosphate starvation response 1 (PHR1) transcription factor (PHR regulon)32 (Supplementary Table 4). Of 161 genes, 74 and 39 genes were upregulated in Col and dpd1 plants, respectively (Supplementary Fig. 10b). Based on these results, we inferred that the global gene expression in response to phosphate-deprived conditions was compromised in dpd1 plants. It was conceivable that orgDNA degradation and PSR are mutually interconnected and that proper PSR requires orgDNA degradation.
We scrutinized the PSR genes that were differentially expressed between Col and dpd1 plants (Supplementary Table 3 and Supplementary Fig. 10c). Upregulated genes in Col plants included those encoding phosphate transporter (PHT1;9, PHT5, PHT3;2, PHT2 and PHT1;4), purple acid phosphatase (PAP23, PAP7, PAP2, PAP22, PAP25, PAP17, PAP24, PAP14 and PAP12), enzymes for lipid biosynthesis (MGD2, DGD2, MGDC, SQD1 and SQD2) and RNase (RNS1), with which phosphate uptake or utilization is shown to be maximized in phosphate-deprived conditions. It was noteworthy that, in dpd1 plants, upregulation of the phosphate transporter genes is limited to PHT1;9 and PHT5, although most of the purple acid phosphatase genes (PAP22, PSP25, PAP2, PAP23, PAP12, PAP17 and PAP24) are also upregulated. In contrast to the genes preferentially upregulated in Col plants, we found several genes with expression levels that were higher in dpd1 than in Col plants (PPCK2, FHL and PLDZETA2). Overall, our RNA-seq analysis revealed that the response to phosphate-deprived conditions was disturbed severely by the loss of orgDNA degradation. Suppression of transporter genes in dpd1 plants, but not the PAP gene, implied that the effect of orgDNA degradation in PSR is complex and is correlated with intracellular and intercellular phosphate relocation.
To assess whether the altered transcriptome in dpd1 plants was rather specific to PSR, we also performed RNA-seq with the plants exposed to nitrogen-deprived conditions (n = 3; the data set is presented as Supplementary Tables 5 and 6). Both Col and dpd1 plants presented many genes that were differentially regulated (1,768 for Col and 961 for dpd1; Supplementary Fig. 11a), suggesting that nitrogen-deprived conditions generally affected a broad range of genes (Fig. 5b). To investigate the nitrogen deprivation response specifically, we focused on two sets of genes that have been reported previously to respond to nitrogen-deprived conditions33,34 (Fig. 5d, Supplementary Fig. 11b and Supplementary Tables 7 and 8). Comparison of these transcriptomes indicated that both Col and dpd1 plants displayed similar expression profiles based on the values at the median and upper/lower quartile, although dpd1 plants had a slightly reduced number of the genes compared to Col plants (Fig. 5d). These results were consistent with the growth defect and fitness observed in nitrogen-deprived conditions (Supplementary Fig. 9a). Taken together, we concluded that dpd1 plants were compromised in PSR and that orgDNA degradation acts in the efficient use of phosphate.
OrgDNA degradation mediated by DPD1 in natural conditions
To verify the role of orgDNA degradation, we questioned whether it occurs in the natural environment. Seasonal remobilization of nutrients such as nitrogen and phosphorus from senescing leaves has been documented as being important in deciduous trees35. In Populus alba, we have shown previously that about 60% of phosphorus in leaves was remobilized before the autumn leaf fall36,37. CpDNA degradation has also been reported in a tree38. Thus, we considered that P. alba is suitable to test whether DPD1-mediated orgDNA degradation coincides with phosphorus remobilization. We conducted leaf sampling from a field-grown P. alba tree (Supplementary Fig. 12), every month from the stage of bud break (April) up to complete leaf fall (November) in 2014, 2015 and 2016 (Fig. 6a). Estimation of cpDNA copy number by qPCR revealed that, in general, cpDNA was more abundant in spring and decreased gradually in autumn (Fig. 6b). Similarly to the case in Arabidopsis, mtDNA levels were much lower throughout the season (Fig. 6c). A small spike in orgDNA levels detected in the summer was probably due to leaf regeneration, which was consistent with phosphorus measurements in our previous study36. Cytological observation of cpDNA was consistent with qPCR, showing holistic disappearance of cpDNA in autumn leaf samples (Fig. 6d).
qRT–PCR analysis of these samples, designed based on the available reference sequence from Populus trichocarpa, demonstrated that a poplar DPD1 homologue was highly upregulated toward leaf fall, with the highest level observed in November (Fig. 6e). We obtained these data from two consecutive seasons (2015 and 2016), which all showed upregulation of DPD1 that accompanied concomitant upregulation of senescence-related genes (Supplementary Fig. 14). We also confirmed DPD1 upregulation in laboratory conditions, which mimicked natural seasonal changes and leaf fall with three defined growing conditions36 (Fig. 6f); DPD1 was specifically upregulated at stage 3, corresponding to autumn/winter (Fig. 6g and Supplementary Fig. 15). Thus, all of these experiments confirmed the contribution of the DPD1 system during natural leaf fall, during which phosphate is redistributed.
CpDNA degradation in leaf tissues has been documented for more than two decades39,40. However, whether DNA is degraded nucleolytically has been controversial, partly because of technical limitations of qPCR, variation within species and tissues, and a lack of mechanistic insights16,17,41. Our studies with DPD1 uncovered the prevailing degradation mechanism among seed plants. We focused initially on male gametophytes (pollen) because the disappearance of orgDNA in male germ cells is often related to maternal inheritance42,43,44, as evidenced in animal EndoG18,20. Although we identified DPD1 exonuclease through forward-genetic mutant screening22,45, orgDNA was shown to be degraded mainly in pollen vegetative cells, which deliver sperm cells to ovules but do not contribute to fertilization. In fact, we did not observe a contribution of DPD1 to the maternal inheritance mode of mtDNA22, which led us to reconsider the physiological role of orgDNA degradation mediated by DPD1. Here, we demonstrated that in both an annual plant (Arabidopsis) and a deciduous tree (P. alba), the DPD1 system operates on orgDNA degradation in vegetative tissues, towards the final stage of the leaf lifespan for phosphate availability. These results revealed that the primary role of the DPD1 system is associated with the efficient use of phosphate rather than orgDNA inheritance.
During leaf senescence, relocating internal macronutrients to the upper and reproductive tissues plays a critical role in maximum fitness46,47. Catabolism of macromolecules stored in chloroplasts is well described, including Rubisco, photosynthetic antenna protein, and pigments, which act mainly in relocating nitrogen48,49. Our finding adds orgDNA to the macromolecules undergoing degradation23,41. It is noteworthy that a substantial portion of orgDNA resides in chloroplasts of fully expanded leaves, whereas only a limited number of mtDNAs exist10,13,28. Although DPD1 is dual targeted to both organelles, its dominant role seems to be in chloroplasts or plastids. Conceivably, orgDNA degradation is beneficial in pollen vegetative cells because male gametophytes, once formed, are isolated from other parts of tissues, hampering their ability to receive external phosphorus efficiently. Conservation of the DPD1 system even in evergreen coniferous species (Supplementary Fig. 1) implicates that DPD1 has emerged during the evolution of microsporophytes, rather than the evolution of leaf senescence.
Several lines of evidence were presented to demonstrate the correlation of orgDNA with phosphate starvation. First, dpd1 plants showed defective growth in our standard hydroponic culture, which was then rescued by supplementing with additional phosphate (Fig. 4a). Second, dpd1 plants showed reduced fitness as well as typical symptoms in phosphate-deprived conditions (Fig. 4e). Third, these deficiencies in dpd1 plants were not detected in nitrogen-deprived conditions but were rather specific to phosphate-deprived conditions. Finally, RNA-seq analysis in phosphate-deprived conditions indicated that dpd1 mutants had compromised accumulation of PSR genes (Fig. 5). All of these results suggested that orgDNA degradation participates in the efficient use of phosphate. The most likely model to explain these results is that orgDNA itself acts as a phosphorus pool and is subjected to degradation for the redistribution of phosphate (Supplementary Fig. 15, model 1). Consistent with this, upregulation of DPD1 during autumn leaf fall in P. alba coincided with the relocation of phosphorus from leaves (Fig. 6), which accounts for 60% of total phosphorus36. Although whether orgDNA serves as a substantial pool of internal phosphorus awaits further investigation, we inferred that the lower fitness in dpd1 plants could be explained by this reservoir model. The alternative model is that some orgDNA degradation product(s), such as nucleotides or their catabolic components, act as a positive sensor of PSR (Supplementary Fig. 15, model 2). A lack of these products may prevent plants in phosphate-deprived conditions from responding to phosphorus deficiency properly, which leads to lower fitness. Although we cannot exclude these two possibilities mutually, our data revealed an interconnection between orgDNA degradation and efficient phosphate use.
In leaves, nucleic acids constitute the most abundant phosphate esters along with phospholipids15,50. Breakdown of nucleic acids and/or enzymes for the biosynthesis of galactolipids and sulpholipids to remodel phospholipids are induced as a part of PSR, along with purple acid phosphatases that hydrolyse phosphate monoesters51,52. We confirmed these PSR in our RNA-seq data32. Based on the results presented in this study, we considered that DNA degradation in endosymbiotic organelles also contributes to PSR. The nucleic acid phosphorus pool, representing 40–60% of the total internal organic phosphorus, consists of RNA and DNA, with rRNA as the largest pool15,50. To degrade these large phosphorus pools, endonucleases are upregulated. RNS1 and RNS2 are the major ribonucleases that supposedly degrade cytosolic or extracellular RNAs53. Bi-functional nuclease 1 (BFN1) has been reported to be upregulated during leaf senescence and to degrade single-stranded DNA or RNA54,55. These findings imply RNA as a major phosphorus pool for relocation, whereas DNA has been considered as a minor phosphorus pool because of its indispensability. By contrast, DPD1 is unique in that it is confined to plastids and mitochondria and degrades ‘dispensable’ orgDNA. Given that total DNA represents 20–30% of total nucleic acids in leaves7,56, we estimated that orgDNA comprises 6–9% of the total nucleic acid pool. Although minor, the dispensable orgDNA pool may serve as a safe guard of the phosphate reservoir, consequently giving an advantage in phosphate-deprived conditions. In principle, extra internal phosphate is considered to be stored in vacuoles57,58. Although transporting free phosphate out of vacuoles plays a critical role in phosphate deprivation management59, the contribution of chloroplasts remained elusive. During leaf senescence, dismantling of chloroplast compounds through enzymatic degradation and/or autophagic processes is recognized to be crucial48,60. Our finding reinforces the importance of chloroplasts for relocating macronutrients, particularly for phosphorus.
Lack of orgDNA degradation in dpd1 mutants caused a weak stay-green phenotype, but leaf senescence proceeded almost normally. Thus, we considered that orgDNA degradation is not a decisive factor controlling the onset of leaf senescence. As a consequence of higher levels of cpDNA being retained (Fig. 3c), chloroplasts showed prolonged functionality because of the retarded decline in chloroplast transcripts. We inferred that orgDNA indirectly determines leaf lifespan, by balancing a trade-off between prolonged photosynthesis and nutrient demand (Supplementary Fig. 15). In general, leaf senescence is associated with nutrient starvation, and an overlap between PSR and senescence-induced genes has been reported50. Given the fact that orgDNA levels decline in response to both nitrogen deprivation and phosphate deprivation but PSR is predominantly compromised in dpd1 mutants (Fig. 5), we considered that the primary role of orgDNA degradation is probably to maximize phosphorus availability in leaves. One possibility of orgDNA contributing to leaf senescence could be a ‘point of no return’, which is proposed to define the stage that the senescence process cannot be reversed47. It is known that senescence is reversible up to a certain point by providing additional nutrients. Conceivably, senescence is no longer reversible when cpDNA is completely lost.
DPD1 is homologous to DnaQ, an epsilon proofreading subunit of Escherichia coli DNA polymerase III22,41. Given that orgDNA replication adopts Pol I26, it remains unclear how DPD1 emerged during evolution. In principle, DPD1 alone can degrade orgDNA given their heterogeneity, as advocated by Bendich: many orgDNAs are nicked, linearized and have a free 3′ end10,61. Whether algae or mosses have other types of exonucleases remains elusive, although the salvage function of orgDNAs was postulated earlier for Chlamydomonas62. TREX1 is a mammal DPD1 homologue63, which has been shown to be associated with inflammatory disease. Unlike TREX1, which degrades foreign pathogenic DNA, DPD1 has evolved to degrade endogenous DNA for salvage. In agriculture, the use of excess nitrogen and phosphorus fertilizers has drawn considerable attention, owing to the fact that over-fertilization of crop fields disturbs the environment, and there is concern over whether the mining of phosphorus fertilizers will compromise their availability in the future. Our findings highlight orgDNA as a potential source of phosphorus storage and future engineering for the efficient use of phosphorus in crop production.
Arabidopsis growth conditions and sampling
A. thaliana ecotype Col was used as the control throughout this study. dpd1 mutants (dpd1-1 and dpd1-5) and a G31 transgenic line (dpd1-1 complemented with the DPD1 genomic sequence) were described previously22. For growing plants, surface-sterilized seeds were placed on 0.8% (w/v) agar plates supplemented with MS medium (Sigma) and 1% (w/v) sucrose for 3 days at 4 °C, followed by further growth in MS plates for 18 days at 23 °C, at a photoperiod of 10-h light and 14-h darkness. Seedlings were then transplanted to soil and were grown for a further 4–5 weeks. Dark-induced leaf senescence was induced in these mature plants, from which we excised all leaves. We placed the leaves in darkness in a sealed chamber containing wet paper to maintain humidity.
Hydroponic culture of Arabidopsis plants was performed as described by Conn et al.30 with a slight modification: the device used in the culture is shown in Supplementary Fig. 8 along with a detailed description in the legend. We used 1/4 MS medium as the hydroponic medium. Sterilized and cold-treated seeds were germinated on top of 1.5-ml microtubes immersed with 1/4 MS liquid medium (rack culture). After continuous growth for 1 month, whole plants with the microtubes were transferred and plugged into a new 15-ml tube filled with 1/4 MS media (tube culture). Growth was conducted in phosphate-depleted or nitrogen-depleted conditions by replacing the medium with medium lacking the corresponding elements 10 days after initiating tube culture. We prepared medium lacking potassium dihydrogen phosphate for phosphate deprivation and lacking ammonium nitrate and potassium nitrate for nitrogen deprivation. Hydroponic culture medium was exchanged with fresh medium every week. Phenotypes and responses to phosphorus or nitrogen deprivation were examined 2 weeks after plants were subjected to nutrient deprivation.
Periodic sampling of leaves from a white poplar tree (located at Uji Campus, Kyoto University, Kyoto, Japan, 34° 91′ N, 135° 80′ E, altitude 24 m above sea level; see Supplementary Fig. 12) was conducted during April–November in 2015 and 2016. Sampling was done every month between 13:30 and 14:30. Leaves in the area between 1 and 2.5 m from the ground were collected randomly. For each sampling, three sets were prepared, consisting of five leaves, which were subjected to RNA isolation followed by qRT–PCR. Meteorological data were acquired from the database of the Japan Meteorological Agency (http://www.jma.go.jp/jma/menu/menureport.html).
For sampling of leaves from a shortened seasonal cycle system in growth chambers, poplar plants were cultivated initially from potted cuttings (shoots with five leaves at a height of 10 cm) with subsequent incubation at stage 1 (1 month at 25 °C in a 14 h/10 h light/dark cycle), stage 2 (1 month at 15 °C in an 8 h/16 h light/dark cycle) and stage 3 (2–3 months at 5 °C in an 8 h/16 h light/dark cycle). Stages mimic spring/summer, autumn and winter, respectively, in natural field conditions. Other growth conditions were similar to those reported previously36. For sampling, the fifth to seventh leaves from apical meristems on the respective plants were collected. Three sets were prepared and subjected to qRT–PCR.
For studies of Arabidopsis, total DNA was isolated as described previously22. For qPCR of organelle genes, primers were designed as listed in Supplementary Table 9. The reactions were performed using a kit (Thunderbird SYBR qPCR Mix; Toyobo) and Light Cycler 2.0 software (Roche Diagnostics) with 40 cycles of denaturation (95 °C for 5 s) and extension (60 °C for 30 s). Quantitative data were obtained from at least three biological replicates and were analysed using LightCycler version 4.0 software (Roche Diagnostics). To normalize qPCR data from orgDNA over nuclear DNA, DPD1 was used as a control of single-copy nuclear DNA, except that 18S rRNA was used in Fig. 2a to follow Zoschke et al.13 (see corresponding figure legends). For studies on Populus, we conducted the same experiment as described above, but the primers were designed specifically based on the whole-genome sequence of P. trichocarpa (taxid: 3694, Phytozome 12, ver. 3.1) for nuclear genes, chloroplast genome sequence of P. alba (taxid: 43335, accession: AP008956) for chloroplast genes and mitochondrial genome sequence of Populus tremula (taxid: 113636, accession: KT337313) for mitochondrial genes. To minimize the amplification of mtDNA or ptDNA sequences included in the nuclear genome, we selected rpoC1 for ptDNA, and matR and cox3 for mtDNA as a reference. As a control nuclear gene, popular CAD gene (Potri.009G095800.1) was selected as a single-copy gene.
For qRT–PCR, total RNA was isolated using an RNeasy Plant Mini Kit (Qiagen), followed by reverse transcriptase and PCRs with a ReverTra Ace qPCR RT Kit (Toyobo) in accordance with the manufacturer’s instructions. For Arabidopsis, we used the histone variant gene H3.3 as an internal control, as described previously16. For P. alba, we used ACTIN2 (Potri.001G309500.1) as an internal control, to measure the expression levels of DPD1 (Potri.005G020600.1), SAG12 (Potri.004G055900.1) and SGR1 (Potri.003G119600.1).
For digital PCR, a QuantStudio 3D Digital PCR System (Thermo Fisher Scientific) was used. DNAs were labelled using the Taqman probe method with primers designed accordingly (Supplementary Table 9). We used PsbA and DPD1 to measure the respective levels of cpDNA (FAM labelled) and nuclear DNA (VIC labelled), and adopted 40 cycles of denaturation (98 °C for 30 s) and extension (60 °C for 2 min) for the PCR. Post-reaction chips were subjected to a QuantStudio 3D Digital PCR System. DNA levels were quantified using AnalysisSuite Cloud Software.
All primers used for PCR analyses in this study are listed in Supplementary Table 9, with the accession numbers of the corresponding genes. All quantitative data included at least three biological replicates and are presented with the standard deviation in the graphs (statistical analysis is indicated in the corresponding figure legends).
The recombinant DPD1-His protein was purified as described previously22 with a slight modification. Overexpression of proteins was conducted at 28 °C. Ni2+-affinity purification was performed with Ni-NTA agarose (GE Healthcare). After purification, the imidazole-containing buffer was exchanged for 2× DPD1 storage buffer (100 mM Tris-HCl (pH 7.5) and 200 mM NaCl) using a gel-filtration column midiTrap G-25 (GE Healthcare). The obtained fractions containing the desired protein were subjected to centrifugation with AmiconUltra-4 (10 K) (Millipore) to concentrate the recombinant protein to >2.0 µg µl−1. The protein solution was diluted to adjust the concentration to 2.0 µg µl−1 and was subsequently mixed with an equal amount of glycerol to make a 1.0 µg µl−1 stock solution. Stock solution aliquots were stored at −30 °C until use. Either aliquots of soluble proteins extracted from E. coli cells or purified recombinant proteins were solubilized by incubation at 75 °C for 5 min in the presence of 2% SDS and 0.1 M dithiothreitol. The protein samples were centrifuged for 1 min at >20,000g and were then subjected to SDS–PAGE with 12.5% (w/v) polyacrylamide gels. The proteins in the gel were subsequently visualized by staining (CBB Stain ONE; Nacalai Tesque).
For the in vitro nuclease assay, we used 6-FAM-labelled oligonucleotides purchased from Hokkaido System Science as substrates. Oligonucleotides of all types (double-stranded DNA, single-stranded DNA and single-stranded RNA) were designed based on the sequence (5′-CGAACACATACTTCACAAGC-3′) derived from one primer used earlier for amplifying a ptDNA fragment (ndhI gene). The nuclease assay was performed in a 12.5 µl reaction mixture that consisted of 40 mM Tris-HCl (pH 7.5), 2 mM MgCl2, 1.6 µM oligonucleotides and 17.5–175 ng purified DPD1-His protein. Each reaction was terminated by the immediate addition of stopping buffer (1% (w/v) SDS, 50% (v/v) glycerol and 0.05% (w/v) bromophenol blue). After each reaction, the digestion products were separated electrophoretically on 20% (w/v, acrylamide: bis = 29:1) polyacrylamide gels. For double-stranded DNA, reaction mixtures with no treatment were loaded on a polyacrylamide gel. Reaction mixtures containing single-stranded DNA or RNA were supplemented with an equal amount of denaturing buffer (TBE buffer containing 10 M urea, 20% (v/v) glycerol and 0.1% (w/v) bromophenol blue) and were then heated at 65 °C for 5 min. Subsequently, the samples were subjected to denaturing PAGE in the presence of 7 M urea. The separated fragments were detected using an image analyzer (LAS4000; Fuji).
For observing DNA with 4,6-diamido-2-phenylindole (DAPI), leaves were simultaneously fixed and stained with 1 µg ml−1 DAPI (3% (w/v) glutaraldehyde). The leaves were observed directly using a microscope (BX61; Olympus Optical) equipped with a disc scan unit. When necessary, sections were prepared using a vibratome VT 1200S (Leica Biosystems), with samples embedded in either 4% (w/v) gelatin or 1.5% (w/v) agarose, a setting blade speed of 0.4 mm s−1, blade vibration of 1.5 mm, thickness of 70–100 µm and a blade angle of 12–15°.
Protein sequences homologous to DPD1 were obtained from the PLAZA database (https://bioinformatics.psb.ugent.be/plaza/). Multiple alignment of the extracted homologues was performed using MUSCLE software with the MEGA7 database. An unrooted tree was constructed using the maximum likelihood method based on the JTT matrix-based model with the default settings in MEGA7.
Photosynthetic activity measurement
Photosynthetic activity of Col and dpd1 leaves of plants grown in soil was measured as the transpiration rate (LI-6400XT; Li-Cor). The same leaves were subjected to measurement to estimate the decline in photosynthetic activity at 1 and 2 weeks after the initial measurements. CO2-dependent photosynthesis curves were obtained at a light intensity of 1,000 μmol m−2 s−1. For each measurement, the relative moisture of the chamber was adjusted to 60–70%.
Arabidopsis RNA-seq analysis
For RNA-seq in Arabidopsis, total RNA was isolated from leaves either in phosphorus depletion or control conditions as described above. RNA-seq was conducted using a HiSeq 2500 or 4000 Illumina sequencing platform and outsourced (Macrogen), including DNA library preparation using a TruSeq RNA sample Prep Kit v2 and sequencing reaction with a TruSeq rapid SBS kit, TruSeq SBS Kit v4 or TruSeq 3000 4000 SBS Kit v3. Sequences were obtained as pair-end reads. At least 4 billion reads were obtained for each sample (n = 3). Mapping of the obtained sequences was performed using the Quas/R package. The gene expression levels were detected by edge/R after normalization with the TCC package. Volcano plots were constructed using the ggplot2/R package with the data set of all of the differentially expressed genes (Supplementary Tables 1 and 2 for phosphate deprivation and Supplementary Tables 5 and 6 for nitrogen deprivation). Box plots were constructed using boxplot and beeswarm/R packages with the data set of the selected genes (Supplementary Tables 3 and 4 for phosphate deprivation and Supplementary Tables 7 and 8 for nitrogen deprivation), which was reported earlier as phosphorus responding32 or nitrogenresponding33,34, respectively.
Measurement of total phosphorus contents
Plants grown in hydroponic culture, with 1/4 MS or in phosphate-deprived conditions, were subjected to phosphorus measurement. Before phosphorus deprivation, all leaves were marked as lower leaves, whereas newly emerged leaves in the phosphate-deprived condition (2 weeks) were designated as upper leaves. Samples (n = 6) were dried in an oven at 60 °C for at least 1 day. Dried samples were then digested with 60% (w/v) nitric acid at temperatures as high as 180 °C. The concentration of phosphorus in the digested solution was ascertained using inductively coupled plasma-mass spectrometry (7500CX; Agilent Technologies).
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Accession numbers of the genes used in this study are listed in Supplementary Table 9. Precise P values calculated by statistical tests in this study are listed in Supplementary Table 10. The raw data used to construct graphs in this study are presented as Supplementary Dataset. The raw transcriptomic data are deposited in the DDBJ with the accession number DRA007138, under the BioProject with the accession number PRJDB7233. All transcriptomic data used in Fig. 5 and Supplementary Figs. 10 and 11 are available in Supplementary Tables 1–8.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank R. Hijiya (Institute of Plant Science and Resources, Okayama University, Kurashiki, Japan) for technical support, H. Kanegae (Graduate School of Agricultural and Life Sciences, The University of Tokyo, Tokyo, Japan) for assisting mtDNA sequence alignment in Populus species and K. Baba (Research Institute for Sustainable Humanosphere, Kyoto University, Kyoto, Japan) for supporting poplar leaf sampling. This work was supported by KAKENHI grants from JSPS (16H06554 and 17H03699 to W.S.) and from the Oohara Foundation (to W.S.).