For screening rpt5a suppressor mutants, approximately 20,000 rpt5a-4 and rpt5a-6 seeds were treated with 0.2% ethyl methanesulfonate (EMS) for 16 h; then, those seeds were sown and grown on rockwool. The harvested seeds were cultured to produce M₂ seeds that were subsequently harvested. Approximately 20,000 M₂ seedlings were grown for 10 days in a 3 mM B MGRL media (Fujiwara et al., 1992) supplemented with 1.0% sucrose and solidified using 1.5% gellan gum (Wako, Japan). Plants that exhibited longer roots than those of the parental mutant were selected and transferred to 30 µM B plates for incubation. After 2–3 days of incubation, the selected plants were transferred to a rockwool, and the M₃ seeds were eventually harvested. The phenotypes of the selected suppressor lines were confirmed with the seedlings of the M₃ generation.

To obtain the nac103-3 single mutant, the rpt5a-6 nac103-3 mutant was backcrossed with the wild type to create a selfed-F₂ population. The nac103-3 mutation was detected by a CAPS marker after performing amplification of the genome using the primers nac103-3_F and nac103-3_R followed by digestion using HinfI; the resultant band sizes were as follows: wild type: 234; nac103-3: 66, 168 bp. The rpt5a-6 mutation was detected by a dCAPS marker after performing amplification using the primers rpt5a-6_BfuAI_F and rpt5a-6_R followed by digestion using BfuIA; the resultant band sizes were as follows: wild type: 254, 297, 401; rpt5a-6: 254, 698 bp.

To obtain mutant combination of rpt5a-6, nac103-3, and sog1-101, rpt5a-6 nac103-3 double mutant was crossed with sog1-101 mutant, and homozygous lines for the genotype of interests were established from their F₂ generation. sog1-101 mutation was detected by PCR using the specific primers as described previously (Ogita et al., 2018).

For the growth tests, seeds were surface sterilized using 70% ethanol aq. For 5 min followed by 99.5% ethanol aq. For 1 min. After rinsing off the ethanol, the seeds were sown on a sterile MGRL medium supplemented with 1.0% sucrose and solidified using gellan gum (Wako). Boric acid concentrations in the medium were modified from the original concentration for each specific test. After vernalization at 4°C for 2 days, the medium-containing plates were placed vertically in incubators at 22°C under a 16-h light (fluorescent lamps, 100–160 μmol photons m⁻² s⁻¹) and 8-h dark cycle.

The positional identification of the causal mutation gene was conducted by utilizing the polymorphism between the two ecotypes, Col-0 and Ler. To obtain the F₂-segregating generation, rpt5a-6 sup1 (Col-0 background) mutant was crossed with Ler. The F₁ was self-pollinated, and the obtained F₂ seeds were used for mapping of the gene responsible for the altered phenotype in root growth under excess boron condition. The F₂ seedlings were grown under excess boron conditions (3 mM boric acid), and seedlings with short hairy roots similar to the rpt5a-6 phenotype were selected for genotyping. The presence of rpt5a-6 mutations in the mapping population was confirmed by the CAPS marker, as described above. For genotyping, a single-sequence length polymorphism (SSLP) marker between Col-0 and Ler was used. A list of genetic markers near the candidate region used for mapping is shown in Table S1 .

For the genetic complementation tests, the NAC103 genomic sequence without the termination codon was cloned from the Col-0 genomic DNA using the primers NAC103ATG-2k_F and NAC103CDS-TAA_R with a PrimeSTAR^® GXL DNA Polymerase (Takara). The amplified fragment contained an upstream region of 1,589 bp in size beginning at the annotated transcription start sites. The fragment was cloned into pENTR D-TOPO (Invitrogen) according to the instructions in the manufacturer’s manual. The obtained entry vector was designated as pSOT1. The entire insert sequences in the entry vector were verified by sequencing analysis. Sequences of primers are listed in Table S1 . The vector for GFP fusion protein expression in plants was obtained by LR reaction between pSOT1 and pMDC107 (Curtis and Grossniklaus, 2003) using the gateway LR clonase enzyme mix (Invitrogen). The resultant expression vector carrying proNAC103-NAC103-GFP construct was designated as pSOT2. In a similar way, for NAC103 overexpression, the coding sequence of NAC103 was cloned from the cDNA derived mRNA isolated from Col-0 roots into pENTR D-TOPO using the primers NAC103_CDS_F and NAC103_CDS_R and then transferred into the expression vector, pMDC32 (Curtis and Grossniklaus, 2003). The resultant expression vector carrying the pro35S-NAC103 construct was designated as pSOT3.

For the stable transformation of the experimental plants, the binary vectors were introduced into the Agrobacterium tumefaciens strain GV3101(Bechtold and Pelletier, 1998), and the target plants were transformed with this culture using the floral dipping method (Clough and Bent, 1998). The transformed plants were selected on a half-strength MS medium containing 20 µg/mL hygromycin B (Wako) and 250 µg/mL Claforan (Sanofi), which was solidified using 0.5% agarose. Expression of the transgene was confirmed by qRT-PCR.

The superoxide was visualized by nitroblue tetrazolium (NBT) staining. Roots were incubated with an NBT staining solution (2 mM NBT in 20 mM MES, pH 6.1) for 5 min at room temperature in the dark. The roots were washed with water after staining and then mounted in chloral hydrate solution (a mixture of 4-g chloral hydrate, 1-mL glycerol, and 2-mL water). These roots were observed under a bright field microscope.

For RNA extraction, 6-day-old wild-type seedlings were grown under control conditions (30 µM boric acid) and were either subsequently subjected to excess boron conditions (3 mM boric acid) or maintained under control conditions for an additional 3 days. The whole roots were harvested and subjected to RNA extraction using a RNeasy Plant Mini Kit (Qiagen). For qRT-PCR analysis, approximately 500 ng of total RNA was reverse transcribed using a Takara Prime Script RT Master Mix according to the manufacturer’s instructions at a 10-μL scale. Real-time PCR was performed using a Takara Thermal Cycler Dice TP800 and Takara SYBR^® Premix Ex Taq™ II (Tli RNaseH Plus) according to the manufacturer’s instructions. For amplification, the PCR cycle was set at 95°C for 30 s, 40 cycles of (95°C for 5s, 60°C for 30s^*) for amplification, and then 95°C for 15s, 60°C for 30s, and 95°C for 15s^* for dissociation curve analysis. The fluorescence intensity was recorded at the steps marked with asterisks, and the primers used are shown in Table S1 .

For observation of the root meristems, approximately 5-mm regions from the root tips were excised and mounted in 10-µg/mL propidium iodide (PI) aq. For visualizing cell walls. The prepared slides were immediately observed under a confocal microscope (FV1000, OLYMPUS), which is equipped with a 40x objective lens, and subjected to excitation at 556 nm; the resulting signal intensity was detected between 570–670 nm.

Root images were captured using a digital camera, and root length was measured using the “Segmented Line” function in FIJI software (Schindelin et al., 2012).

For the comet assay, nuclei were extracted from 100 primary root tips (1 cm from the tip) for every single replicate. Roots were chopped in 300 µL of nuclei extraction buffer (50 mM EDTA in PBS) using a razor. The buffer was collected and passed through a 30-µm membrane filter (CellTrics^®, Sysmex Partec) to remove large cellular structures. Comet assay was performed using CometAssay^® Kit (25 x 2-well slides; Trevigen) by following the instructions in the kit’s manual and performing slight modifications: electrophoresis was performed at 1 V/cm for 6 min in an ice-cooled TBE buffer. For DNA staining, SYBR^® Gold Nucleic Acid Gel Stain (10,000x Concentrate in DMSO; Thermo Fisher) was used after diluting it by 30,000 times using TE buffer (pH 8.0). The prepared slides were observed under a fluorescent microscope (BX50WI, OLYMPUS), which was equipped with a U-MWIBA3 mirror unit (460–495 nm excitation, 510–550 nm absorbance, and a 505-nm dichroic mirror). Image quantification was performed by using default settings of CASP software (Końca et al., 2003).

For root length and elongation measurements, primary roots of 9-24 individuals were measured. For qRT-PCR, roots of 10 seedlings were sampled as one biological replicate, and three biological replicates were analyzed. For comet assay, nuclei were extracted from 100 seedlings as one sample, and 125 nuclei were measured for each sample. Dunnett’s test was applied for multiple comparison against control or wild-type. Tukey-Kramer’s test was applied for multiple comparison among all groups.

It has been reported that mutants of a subunit of the 26S proteasome, the RPT5A gene, are sensitive to excess boron (Sakamoto et al., 2018). Under excess boron conditions, mutants exhibit a severely inhibited root elongation phenotype, which is restored under normal boron conditions. To obtain insights into the role of the RPT5A gene with regard to excess boron tolerance, we screened for its suppressor mutants from M₂ population of mutagenized rpt5a-6, a non-sense mutant, and rpt5a-4, a knock-out mutant. From the M₂ rpt5a-6 population, we isolated a suppressor line, rpt5a-6 sup1, the root elongation of which under conditions of boron stress (3 mM boric acid) was greater than that of the original mutant rpt5a-6 ( Figure 1A ). Under normal conditions, roots of rpt5a-6 sup1 was significantly longer than that of rpt5a-6, but shorter than the wild type, suggesting that sup1 mutation suppresses not all the growth defects in rpt5a-6 ( Figure 1B ). To evaluate excess boron sensitivity, apart from basic growth defects, we compared relative root growth (excess boron/control) among the genotypes. Relative root growth of rpt5a-6 sup1 was higher than rpt5a-6 and comparable to the wild type, suggesting that sup1 mutation suppresses the excess boron sensitivity of rpt5a-6 to the same levels as the wild type.

Previously, rpt5a-6 had been found to exhibit an altered root morphology with short and squeezed root apical meristems and frequent cell death in the meristem even under normal boron conditions; these features become more evident by excess boron stress (Sakamoto et al., 2018). In rpt5a-6 sup1, the abnormal root meristem shape under excess boron stress was recovered and indistinguishable from the wild type ( Figure 1C ). Visualization of cell walls and dead cells using propidium iodide (PI) stain revealed that cell death is evident in stele cells in the meristem of rpt5a mutants after 40 h of excess boron treatment, and that the cell death was suppressed in rpt5a-6 sup1 ( Figure 1D and Supplementary Figure 1 ). From these results, we concluded that rpt5a-6 sup1 completely suppresses the root growth sensitivity observed in rpt5a-6 mutants to excess boron.

Crossing between the rpt5a-6 sup1 and rpt5a-4 mutants suggested that the rpt5a-6 sup1 mutation is semi-dominant because the F1-progeny exhibited intermediate root growth between rpt5a-6 sup1 and rpt5a-4 ( Supplementary Figure 2 ). Map-based cloning revealed that the causal gene for the suppression of excess boron sensitivity in rpt5a-6 sup1 is located on chromosome 5, between 25.4 Mb and 25.8 Mb ( Figure 2A ). Within the chromosomal region, whole genome sequencing identified only a single homozygous mutation. The mutation was in the NAC transcription factor family, NAC103 (AT5G64060), and is predicted to produce a premature stop codon in its coding region. We designated the mutation as nac103-3 ( Figure 2B ). In addition, we isolated other 11 independent suppressor lines from the M₂ rpt5a-4 population ( Supplementary Figure 3 ). These lines exhibited phenotypes similar to those exhibited by rpt5a-6 nac103-3. We determined the NAC103 genomic sequence in the 11 rpt5a-4 mutant suppressor lines and found that seven lines harbored a distinct mutation in the NAC103 coding sequence, designated as nac103-4, 5, 6, 7, 8, 9, and 10 ( Figures 2B , S2 ). The presence of several alleles exhibiting similar phenotypes strongly suggests that NAC103 is the gene responsible for the suppression. Because the phenotype was comparable among the abovementioned alleles, we focused on rpt5a-6 nac103-3 for further analysis.

To further confirm that the mutation in NAC103 is responsible for the suppression of the rpt5a-6 phenotype, we conducted complementation tests. A construct carrying proNAC103-NAC103-GFP was introduced into rpt5a-6 nac103-3, and two independent homozygous transgenic lines were established. A qRT-PCR analysis confirmed that those lines accumulated higher or comparable NAC103 mRNA compared to rpt5a-6. Under both normal boron concentration (control) and excess boron concentration, root elongation of these transformants was found to be significantly reduced in comparison with that of rpt5a-6 nac103-3 ( Figure 2C ). To examine the effect of expressing the NAC103 gene without rpt5a-6 mutation, we also introduced the same construct into the nac103-3 single mutant. The nac103-3 single mutant was generated by backcrossing rpt5a-6 nac103-3 with the wild type. The root elongation phenotype of the established nac103-3 single mutant was indistinguishable from that of the wild type ( Figure 2C ). In contrast to the rpt5a-6 nac103-3 proNAC103-NAC103-GFP lines, the three transformant lines of nac103-3 proNAC103-NAC103-GFP exhibited root growth indistinguishable to that of nac103-3, thereby indicating that growth defects caused by the introduction of NAC103 is specific to an rpt5a-6 background. Given that the expression of NAC103-GFP in the rpt5a-6 nac103-3 mutant resulted in the inhibition of root elongation in the rpt5a-6 mutant, we concluded that NAC103 is at least partly responsible for the phenotype of the rpt5a-6 suppressor mutant.

To investigate the reaction of the NAC103 gene to boron stress, we examined the accumulation of NAC103 mRNA in the rpt5a-6 mutant in response to conditions of excess boron. RNA was extracted from whole roots of seedlings treated with excess boron (3 mM boric acid) and from those of seedlings grown under control conditions (30 µM boric acid) for 3 days. In the wild type, excess boron significantly induced the NAC103 mRNA level by more than 10-fold ( Figure 3A ). mRNA accumulation was much higher in rpt5a-6 plants grown under both conditions, having more than 50-fold increase in comparison with that in the wild-type plants grown under control condition. These results suggest the possibility that the overaccumulation of NAC103 mRNA is responsible for root growth inhibition under conditions of excess boron in the rpt5a-6 mutant.

We hypothesized that an over accumulation of NAC103 mRNA is responsible for root growth inhibition under excess boron stress. To test the hypothesis, we generated NAC103 overexpression lines to observe their phenotypes under excess boron stress. For the NAC103 overexpression lines, the coding sequence of NAC103 was cloned from cDNA and introduced into the wild type under the control of the cauliflower mosaic virus 35S RNA promoter. Three independent transformant lines were established. A qRT-PCR analysis confirmed that those lines accumulated higher NAC103 mRNA than the wild type as well as the rpt5a-6 mutant, with a tendency to reduce under excess boron conditions for unknown reasons ( Figure 3A ). However, root growth in the overexpressing lines was indistinguishable from that in the wild type under both the excess boron and control conditions ( Figure 3B ). These results suggest that growth defects caused by excess boron are not explained by NAC103 mRNA over accumulation alone.

The NAC103 gene has been reported to play a role in the transcriptional regulatory cascade of the unfolded protein response (UPR) (Sun et al., 2013). They have reported that NAC103 expression is induced by endoplasmic reticulum (ER) stress, which is caused by an accumulation of unfolded proteins via direct transcriptional activation by an ER stress signaling regulator bZIP60, and transmits the signal by upregulating the downstream UPR genes. To examine the effects of the nac103-3 mutation on tolerance to unfolded protein accumulation and root growth, we treated rpt5a-6 and rpt5a-6 nac103-3 mutant seedlings with the reagent L-canavanine, which induces protein unfolding via its incorporation into the protein; L-canavanine is an analog of arginine and induces protein misfolding (Kurepa et al., 2003). Treatment with 0.5 µM L-canavanine reduced root elongation in the rpt5a-6 single mutant but not in the rpt5a-6 nac103-3 double mutant ( Figure 4A ), suggesting that the nac103-3 mutation suppressed the sensitivity of rpt5a-6 to the accumulated unfolded proteins.

To assess the involvement of the downstream UPR pathway in excess boron sensitivity with regard to the rpt5a-6 mutant, we examined the expression levels of the already known NAC103-downstream UPR-related genes: calnexin (CNX), calreticulin (CRT), protein disulfide-isomerase 5 (PDI5), and ubiquitin conjugase 32 (UBC32) (Sun et al., 2013). A qRT-PCR analysis detected no significant difference in the expression of the four abovementioned genes and proteins between the rpt5a-6 and rpt5a-6 nac103-3 mutants ( Figure 4B ), suggesting that the transcriptional regulation of these genes by NAC103 is not responsible for excess boron sensitivity of the rpt5a-6 mutant.

It has been reported that excess boron stress induces DNA damage in Arabidopsis root meristem (Sakamoto et al, 2011), and that the extent of the DNA damage is more severe in the rpt5a-6 mutant than that in the wild-type plant (Sakamoto et al., 2018). In addition, treatment of rpt5a-6 with zeocin, a reagent that causes DNA damage, results in a development of a phenotype similar to that of the wild type under excess boron stress conditions (Sakamoto et al., 2018). These results suggest that the susceptibility to DNA damage is a critical factor for excess boron sensitivity in rpt5a-6.

To investigate the role of NAC103 in this context, we examined the extent of DNA damage in the rpt5a-6 nac103-3 mutant under the excess boron stress. To estimate the accumulation of DNA damage in roots, the expression levels of genes known to respond to DNA damage, namely RAD51, BRCA1, and PARP2(AT4G02390), were quantified by qRT-PCR analysis ( Figures 5A–C ). Among all genotypes, expression of these genes tended to be higher under excess boron conditions compared to the control condition. Under excess boron conditions, expression of BRCA1 and PARP2 was significantly higher in rpt5a-6 than in wild type, but in rpt5a-6 nac103-3 it was comparable to wild type. These results suggest that the nac103-3 mutation suppresses DNA damage accumulation in the rpt5a-6 mutant. To obtain direct evidence for this, we assessed the extent of DNA damage in root tips using a comet assay ( Figure 5D ). The comet assay revealed that DNA damage accumulation was significantly lower in the rpt5a-6 nac103-3 mutant than in the rpt5a-6 mutant, regardless of boron concentration. These results also reveal that the nac103-3 mutation ameliorated the extent of DNA damage in rpt5a-6 mutant plants.

The rpt5a-6 mutant has been reported to be more susceptible to DNA damage and, more specifically, that caused by radiation or chemical reagents (Sakamoto et al., 2018). To investigate the involvement of NAC103 in sensitivity to DNA damage, we assessed the root growth of the rpt5a-6 nac103-3 mutant after treatment with zeocin ( Figure 6 ). Treatment with zeocin reduced root elongation in a dose-dependent manner among all tested genotypes ( Figure 6A ). Consistent with previous results, the rpt5a-6 mutant was more sensitive to zeocin than the wild type, with the former showing a higher root elongation inhibition rate and a more severe morphological alteration, including cell death in root meristems ( Figure 6C ). However, none of these defects were evident in the rpt5a-6 nac103 double mutant, suggesting that the NAC103 gene is responsible for the growth defects caused by DNA damage in the rpt5a-6 mutant.

The accumulation of DNA damage motivated us to focus on reactive oxygen species (ROS) because oxidation by ROS is known to be a cause of DNA damage (Amor et al., 1998). To observe ROS accumulation in the mutants, we stained control and excess boron-treated roots with nitroblue tetrazolium (NBT), rendering a purple stain to the superoxide ( Figure 7 ). Under excess boron conditions, the stain was more intense in rpt5a-6 mutant roots than that in the wild-type roots. However, rpt5a-6 nac103-3 showed lesser staining intensity than both the rpt5a-6 and the rpt5a-6 nac103-3 mutants, but it was comparable to that associated with the wild type. These results suggest that NAC103 is responsible for ROS accumulation in rpt5a-6 under excess boron stress.

As NAC103 has been suggested to function downstream of SOG1 transcription factor, a master regulator of DNA damage response genes (Ryu et al., 2019), we examined whether the suppression of excess boron sensitivity by nac103 mutation is related to SOG1 pathway. Examination of root elongation revealed that excess boron sensitivity of rpt5a-6 sog1-101 was not significantly different from rpt5a-6, suggesting that sog1-101 mutation does not alleviate excess boron sensitivity of rpt5a-6 ( Figure 8A ). In addition, excess boron sensitivity of rpt5a-6 nac103 sog1-101 triple mutant was similar to that of rpt5a-6 nac103 double mutant. To examine whether induction of NAC103 by excess boron stress depends on SOG1, we examined NAC103 mRNA abundance in sog1-101 mutant. Compared to rpt5a-6 mutant, rpt5a-6 sog1-101 exhibited even higher NAC103 mRNA abundance under excess boron conditions, indicating that upregulation of NAC103 under excess B condition does not require SOG1 ( Figure 8B ). Taken together, these results suggest that contribution of NAC103 to excess boron sensitivity is likely due to its involvement in pathways other than the SOG1 pathway.

We were able to limit the positional identification of the suppressor mutation in rpt5a-6 sup1 to the chromosomal region of 353 kb ( Figure 2A ). Whole-genome sequencing revealed DNA sequences in this region; within that region, we found only a single homozygous mutation in the NAC103 gene. Through independent experiments, we identified 11 mutants exhibiting phenotype similar to that of rpt5a-6 sup1. Examination of the NAC103 sequences revealed that 7 alleles out of the 11 rpt5a-4 lines carried distinct mutations in NAC103 ( Figure 2B ). These results suggest that NAC103 is the gene responsible for the suppression of the rpt5a mutant phenotype, but it is fair to consider the possibility that NAC103 is located on so-called hot spot of mutagenesis by EMS. In other projects apart from boron toxicity, we also sequenced other eight mutants, which were also mutagenized in a manner similar to the rpr5a-6 sup1 mutant. As none of the eight mutants carried a mutation in NAC103, the mutations in NAC103 in rpt5a suppressor mutants should not be attributed to a mutagenesis hot spot. In addition, all the NAC103 mutations in the rpt5a suppressor mutants were located within the N-terminus half of the gene, which was annotated as the NAC domain, a conserved DNA binding domain ( Figure 2 ). Thus, we suggest that the mutations can be attributed to the functionality of the proteins rather than chromosomal location. In the complementation tests, the introduction of the proNAC103-NAC103-GFP construct partially reversed the phenotype of the rpt5a-6 sup1 mutant to that of the rpt5a-6. The incomplete recovery could be attributed to possible effects GFP-tag fused to C-terminus of NAC103, rather than low expression levels of the transgene, because NAC103 mRNA levels in the transgenic lines were comparable or higher than in rpt5a-6 ( Supplementary Figure 4 ). However, given that the changes in phenotype were observed, the transgene is at least partially functional, and NAC103 gene is likely to be involved in the root growth defects in the rpt5a-6 mutant under boron stress. Taken together, we conclude that the mutation in NAC103 is responsible for the suppression of the excess boron-sensitive phenotype of rpt5a mutant.

The NAC family is one of the largest families of plant-specific transcription factors (Ooka et al., 2003). NAC103 has been characterized in the context of its expression in response to unfolded proteins levels (Sun et al., 2013). They showed that the NAC103 protein is a substrate of the 26S proteasome, and that it is maintained at a low concentration under normal conditions; furthermore, it is stabilized under unfolded protein stress.

The upregulation of NAC103 in turn activates the transcription of downstream unfolded protein response genes such as molecular chaperones [namely, calnexin (CNX), calreticulin (CRT), and protein disulfide-isomerase 5 (PDI5), and ubiquitin conjugase 32 (UBC32)]. In our study, mRNA accumulation of those NAC103 downstream genes was not significantly different between the rpt5a-6 and rpt5a-6 nac103-3 mutants ( Figure 4B ), suggesting that these genes were not responsible for the suppression of the rpt5a-6 phenotype in rpt5a-6 nac103-3 mutant plants. In addition, our previous study demonstrated that excess boron treatment does not affect accumulation of poly-ubiquitinated proteins, which include unfolded proteins to be degraded by proteasome (Sakamoto et al., 2018) Thus we can deduce that excess boron treatment does not induce unfolded protein significantly. These together suggest that it is not unfolded protein response pathway that is critical for excess boron response.

It has been reported that NAC103 interacts with VND, a master regulator of xylem vessel differentiation that functions as a transcription activator (Yamaguchi et al., 2015). According to their yeast two-hybrid assay, NAC103 can form hetero-complexes with various other NAC transcription factors, such as VND1, VND2, VND3, VND7, NAC1, CUC2, and VNI1. This suggests that NAC103 can play multiple roles in different cellular processes through its interaction with different counterparts. They also indicated that NAC103 may regulate cellular stress response by binding with other NAC domain proteins. This point of view provides possible explanation for our observation that overexpressing NAC103 alone did not affect root growth under excess B conditions. If NAC103 requires counterpart(s) to be functional under excess B conditions, overexpressing NAC103 could affect the phenotype only when enough number of counterpart proteins are available. Identification of the potential counterpart in future study would clarify this point.

Assuming that the mutations in NAC103 are loss of function, our results suggest that accumulation of NAC103 has negative effects on root growth, because nac103 mutations suppressed the growth defects of rpt5a mutants. Considering that NAC103 is a substrate of the 26S proteasome and that RPT5A is suggested its involvement in recognition of poly-ubiquitinated proteins in the Ub–proteasome pathway (Lam et al., 2002), it can be speculated that control of NAC103 protein accumulation was altered in rpt5a mutants. This should be confirmed by observation of NAC103-GFP in rpt5a-6 nac103-3. However, under our experimental conditions, we could not detect GFP fluorescence in transgenic plants introduced with the proNAC103-NAC103-GFP construct, even after treatment with MG132, which stabilizes the NAC103 protein (Sun et al., 2013). Could be because we did not have enough expression to observe fluorescence as the construct was driven by NAC103 own promoter, not by a strong constitutive promoter.

In Brassica napus, homologue of NAC103, BnaNAC103 was identified with 75.4% identity and 76.5% similarity at the nucleotide level (Niu et al., 2014). Furthermore, they revealed that in Nicotiana benthamiana leaves, overexpression of BnaNAC103 causes ROS accumulation and induces cell death. They mentioned in the report that whether NAC103 in A. thaliana (AtNAC103) regulates cell death is unknown yet. However, it is likely that AtNAC103 has a similar negative function considering that ROS accumulation and cell death in the rpt5a-6 mutant under excess boron stress was suppressed in rpt5a-6 nac103-3 ( Figure 7 ). Although how NAC103 could induce ROS remains unclear, the simplest hypothesis should be that NAC103 may upregulate ROS producing enzyme, or negative regulator of ROS scavenger, considering that NAC103 is a transcription factor.

The observed ROS accumulation and cell death under excess boron conditions can be interpreted in two ways: ROS induces cell death, or, cell death releases ROS from cells. Although further careful investigation is needed to dissect this cause-effect relationship, our results allow us to infer it. After three days of excess boron treatment, strong NBT staining was observed throughout most of the root area in rpt5a mutant, suggesting that ROS is accumulated in whole root ( Figure 7 ). On the other hand, under the same treatment condition, cell death was observed only in stele cells in the meristem ( Supplementary Figure 1 ). This suggests that, at least in most of the root tissue, ROS accumulation does not require preceding cell death. Thus, we deduce that it is more likely that ROS is the cause of cell death, rather than ROS accumulation is the result of cell death.

Several independent microarray analyses have revealed that NAC103 expression is induced during conditions of DNA damage caused by γ-ray exposure or treatment with hydroxyurea or bleomycin (Culligan et al., 2006; Yoshiyama et al., 2009; Cools et al., 2011; Yi et al., 2014). Ryu et al. (2019) reported that NAC103 is induced by SOG1, a NAC transcription factor known to regulate DNA damage response gene, and is partially responsible for DNA damage response phenotypes. They revealed that NAC103 interacts with the putative promoter regions of Rad51, PARP2 (AT4G02390, which is referred to as PARP1 in Ryu et al., 2018), RPA1E, BRCA1 and At4g22960, suggesting that NAC103 is a putative SOG1-dependent DNA damage response regulator. However, our results indicate that NAC103 has some functions in SOG1-independent pathway, because NAC103 induction under excess boron condition was SOG1-independent and sog1 mutation did not suppress excess boron sensitivity of rpt5a mutant ( Figure 8 ). In our study, the nac103-3 mutation suppressed both DNA damage and ROS accumulation in the root tips of rpt5a-6 mutant plants in the presence of excess boron. This indicates that irregular accumulation of NAC103 in the proteasome mutant rpt5a-6 induces DNA damage and ROS accumulation under excess boron stress, suggesting the role of NAC103 in DNA damage induction or suppression of DNA repair and/or ROS production or radical scavenging activity.

In summary, our study sheds light on the mechanisms underlying excess boron-induced DNA damage accumulation and growth reduction ( Figure 9 ). Our previous study revealed that excess boron causes histone acetylation and BRM-dependent chromatin opening (Sakamoto et al., 2018). This is thought to increase the susceptibility to DNA damaging factors. On the other hand, our study suggested that NAC103 is involved in a downstream pathway, which is likely responsible for the accumulation of a DNA damaging factor, ROS. Excess boron stress induces NAC103 mRNA in SOG1-independent manner. In the wild type, NAC103 protein level is likely maintained at low level by the 26S proteasome, which would ameliorate excessive ROS accumulation. However, an alteration in the 26S proteasome of the rpt5a-6 mutant can lose this maintenance, which may in turn result in accelerated ROS accumulation. This would result in increased DNA damage, which could be the direct cause for root growth inhibition. rpt5a-6 mutation also induced NAC103 mRNA, but considering that the induction was observed even under control conditions, this mRNA induction itself is likely due to a boron-independent mechanism.