Arabidopsis response to the spider mite Tetranychus urticae depends on the regulation of reactive oxygen species homeostasis

Reactive oxygen species (ROS) are molecules that play a prominent role in plant response to numerous stresses, including plant interactions with herbivores. Previous findings indicate that Arabidopsis plants showed an increase in H2O2 accumulation after Tetranychus urticae infestation. Despite its importance, no information has been reported on the relationships between ROS-metabolizing systems and the spider mite-triggered plant-induced responses. In this work, four ROS-related genes that were differentially expressed between the resistant Bla-2 and the susceptible Kon Arabidopsis accessions were selected for the analysis. These genes encode proteins putatively involved in the generation (BBE22) and degradation (GPX7 and GSTU4) of H2O2, and in the degradation of ascorbate (AO). Overexpressing BBE22 and silencing GPX7, GSTU4 and AO resulted in higher leaf damage and better mite performance relative to the wild-type plants. Minor effects on H2O2 accumulation obscure major effects on the expression of genes related to ROS-metabolism and JA and SA signaling pathways, and on ROS-related enzymatic activities. In conclusion, the integration of ROS and ROS-related compounds and enzymes in the response of Arabidopsis to the spider mite T. urticae was confirmed. However, the complex network involved in ROS signaling makes difficult to predict the impact of a specific genetic manipulation.


Results
Genes involved in ROS homeostasis are differentially expressed in response to spider mite feeding. We have previously highlighted the natural genetic variation of resistance between Arabidopsis accessions to T. urticae, identifying Bla-2 (resistant) and Kondara (susceptible) as accessions at the opposing ends of the spectrum 29 . Besides, we made a gene ontology enrichment analysis of the 1109 and 993 differential expressed genes between Bla-2 and Kon reported upon T. urticae infestation or in absence of spider mite herbivory, respectively 29 . Mining the dataset coming from these analyses, several functional categories related to ROS homeostasis appeared over represented in transcriptomic comparisons related to T. urticae infestation or to resistance/susceptibility of Arabidopsis accessions (Table 1). To further explore the defensive role of single genes directly related to ROS homeostasis, four genes differentially expressed in Bla-2 and Kon upon T. urticae infestation were selected ( Table 2). These candidate genes were: AT4G39830 (AO), AT4G31870 (GPX7), AT2G29460 (GSTU4) and AT4G20860 (BBE22). The putative implication of these proteins in ROS homeostasis may be deduced from the schematic metabolic network regarding hydrogen peroxide production and elimination showed in Supplementary Fig. S1. BBE22 would be involved in the generation of H 2 O 2 from reduced compounds; AO would have the role in the degradation of ascorbate to (mono)dehydroascorbate, and GPX7 and GSTU4 in the degradation of H 2 O 2 to water using glutathione as electron donor. RT-qPCR analysis carried out in several Arabidopsis tissues showed that mRNA expression levels were detected in most of the tissues for the four genes, which include leaves and rosettes ( Supplementary Fig. S2).
To confirm the inclusion of these genes in Arabidopsis responses to T. urticae feeding and their differential induction in Arabidopsis accessions, we infested Col-0, Bla-2 and Kon plants with spider mites and monitored levels of BBE22, AO, GPX7 and GSTU4 using the RT-qPCR assays (Fig. 1). BBE22 expression was induced     The expression of AT4G39830, AT4G31870, AT2G29460 and AT4G20860 genes was quantified at 1, 3, 6, 12 and 24 h post-infestation by RT-qPCR assays. Gene expression referred as fold change (2 −ddCt ). Data are means ± SE of three replicates. Different letters indicate significant differences (P < 0.05, One-way ANOVA followed by Student-Newman-Keuls test). Characterization of T-DNA insertion lines in differentially expressed ROS related genes. To investigate the role of the candidate ROS homeostasis related-genes in plant defense against T. urticae, available T-DNA insertion lines for these genes were characterized. The location of the T-DNA insertion in each gene is shown in Supplementary Fig. S3. The expression levels of the targeted genes were quantified in homozygous lines (Fig. 2a). In the AO lines (Salk_070852 and Salk_061411), T-DNA insertions were in coding regions (with the exception of the isoform 4 of this gene, in which an insertion was in a putative intron). As the expression of the gene was not totally abolished, these lines were considered as knock-down lines (AO KD1 and AO KD2).
The T-DNA insertions in the GPX7 lines were also in coding regions. Whereas the insertion of the T-DNA into exon 4 (Salk_023283 line) abolished the expression of GPX7 (GPX7 KO, knock-out), the T-DNA insertion in exon 2 (Salk_072007 line) only resulted in a reduced GPX7 expression (GPX7 KD, knock-down). For GSTU4, one T-DNA insertion was located in the second exon, at the end of the coding region, and the other inserted within the 3′UTR. In both lines, the expression of the gene was considerably reduced but still quantifiable. Thus, these lines were considered knock-down lines (Salk_1439280, GSTU4 KD1; Salk_125181, GSTU4 KD2). Finally, T-DNA insertions in the BBE22 gene were located in the promoter region. Interestingly, these insertions had opposite effects. Whereas the insertion closest to the 5'UTR caused a complete lack of gene expression (Salk_118723, BBE22 KO), the T-DNA insertion further upstream resulted in a strong over expression of the gene (Salk_044265, BBE22 OE).

Effects of ROS homeostasis related genes on plant resistance and mite performance.
To investigate the role of the candidate ROS homeostasis related-genes in plant defense, T-DNA insertion lines and non-transformed Col-0 controls were infested with mites and plant damage (chlorotic area) was visualized and quantified four days after mite feeding (Fig. 2b, (Fig. 3b). Besides H 2 O 2 accumulation, the activity of several enzymes involved in its metabolism was quantified (Fig. 4). Whereas catalase (Cat) and ascorbate peroxidase (APX) activities are directly related to H 2 O 2 degradation, dehydroascorbate reductase (DHAR) and glutathione reductase (GR) are involved in the regeneration of ascorbate and reduced glutathione, compounds that can be used by ascorbate peroxidases or glutathione peroxidases and S-transferases to reduce H 2 O 2 ( Supplementary Fig. S1). The reduced expression of H 2 O 2 -synthesyzing enzyme BBE22 in BBE22 KO line did not affect activities of APX, Cat, GR or DHAR enzymes in presence or the absence of mite feeding. These enzymes had similar activities to these observed in the Col-0 WT plants (Fig. 4). The overexpression of BBE22 in BBE22 OE line led to increased basal activity of APX and Cat that are directly related to H 2 O 2 degradation. Mite feeding did not affect the APX activity, but led to the increased Cat activity in these lines relative to the WT. APX quantification showed a higher basal activity in all T-DNA lines with reduced expression of genes encoding enzymes involved in the H 2 O 2 catabolism (AO, GPX7 and GSTU4) relative to Col-0 WT. However, while the APX activity increased upon mite feeding in WT plants; it did not change further in AO, GPX7 and GSTU4 T-DNA lines. The increase in basal activity was also observed for the Cat activity in these mutant lines. After mite attack, catalase activity did not vary in Col-0 and was reduced to Col-0 levels in all T-DNA lines that disrupt genes acting in the H 2 O 2 catabolism. Basal GR activity was similar for all lines. After mite infestation, this activity significantly increased only in GPX7 KO lines. Finally, GPX7 lines also showed higher basal DHAR activity while the rest of the lines had activity levels comparable to the WT plants. Whereas the activity after T. urticae feeding was reduced in the GPX7 KO lines, in WT Col-0 and AO KD the DHAR activity increased and it remained at the basal level in the GSTU4 KD lines. In summary, the overexpression of BBE22 that correlated with the increased generation of H 2 O 2 ( Fig. 3) resulted in the increased activity of APX and Cat enzymes that are directly related to H 2 O 2 degradation. The alteration of genes encoding enzymes involved in H 2 O 2 catabolism resulted in more complex changes in the enzymatic activities of APX, Cat, GR and DHAR, suggesting the existence of compensatory relationship among enzymes involved in H 2 O 2 catabolism that is consistent with the overall unperturbed changes in levels of H 2 O 2 in these mutant plants (Fig. 3).
To test this assumption, the expression of BBE22, AO, GPX7 and GSTU4 was analyzed by RT-qPCR in the T-DNA insertion lines ( Fig. 5 and Supplementary Table 1). AO expression was repressed in all T-DNA lines with the exception of GPX7 KO, in which its expression was induced. GPX7 expression was repressed in one GSTU4 KD line and in the BBE22 KO line. GSTU4 transcriptional levels were reduced in the GSTU4 KD lines and the GPX7 KO line and were increased in the BBE22 OE line. Finally, the expression of BBE22 was induced in the for AO and GPX7 KD lines. Thus, the expression of any one of the tested genes involved in the ROS homeostasis is affected by the perturbation of the expression of other genes within this process.

Effects of genetically modified ROS related genes on defense-related hormonal pathways.
Previously described effects of the perturbation in the expression of ROS related genes on mite-inflicted plant damage and mite performance suggest that these genes could also affect the signals that trigger hormone responsive pathways. As JA and SA are the main hormones involved in plant defense against spider mites, the expression of several genes controlled by these hormones was checked by RT-qPCR (Fig. 6). Non-expresser of PR gene1 (NPR1) involved in SA-triggered transcriptional regulation and Pathogenesis Related (PR1) genes were chosen as markers of SA responses. Basal NPR1 expression was quite similar in most lines. The BBE22 KO and OE lines had higher expression that was stronger in the BBE22 KO line (Fig. 6a). Upon mite feeding, the significant increase of NPR1 expression was detected in AO KD and BBE22 OE lines. The basal PR1 expression levels were elevated in lines with reduced activity of H 2 O 2 -catabolism related enzymes (Fig. 6b) Fig. 6d,e). While the reduced expression of BBE22 led to the high expression of the VSP2, the BBE22 overexpression favored the expression of the PDF1.2 upon mite feeding. Overall, alterations in the expression of genes involved in the ROS homeostasis were reflected in an alteration of JA and SA signaling pathways in response to mite feeding (Fig. 6f).

Discussion
Multiple genes encoding ROS-processing enzymes, including peroxidases, reductases, and dehydrogenases are present in plants.
A key challenge is to elucidate the physiological importance of specific members in terms of their potential influence on biotic stress resistance programs. The comparison of differentially expressed genes between the resistant Arabidopsis accession Bla-2 and the susceptible Kondara permit us to identify individual genes potentially related to ROS homeostasis in response to mite feeding. BBE22, AO, GPX7 and GSTU4 genes were also differentially induced upon the massive infestation of Arabidopsis rosette leaves with hundreds of mites for 1 h, treatment that was expected to enhance early responses at the feeding site 29 . Consistent with the expression patterns observed in the microarray experiment, we were able to replicate differential expression of these genes in Bla-2 and Kon accessions. We showed that BBE22, AO and GSTU4 are induced early upon mite feeding, while GPX7 was induced only after 24 h (Fig. 1). Furthermore, we previously showed that levels of H 2 O 2 strongly increased after mite infestation in leaves from WT and mati plants, while they accumulated at moderate levels in the more resistant MATI over-expressing plants 28 . This circumstantial evidence prompted us to investigate the role of BBE22, AO, GPX7 and GSTU4 in the establishment of Arabidopsis responses to mite herbivory. BBE22, GPX7, GSTU4 and AO have opposing functions in the maintenance of ROS homeostasis. BBE22 belongs to the Berberine bridge enzymes (BBE), which are proteins that oxidize different molecules using FAD as a cofactor to produce H 2 O 2 33 . In Arabidopsis, there are 28 BBE-like genes, with a physiological function mostly  Data represent means ± SE from at least six replicates. Different letters indicate significant differences among genotypes within control or mite infested lines (P < 0.05, One-Way ANOVA followed by Student-Newman-Keuls test) and asterisks indicate significant differences among control and infestation within each genotype (P < 0.05, t-student test). Arabidopsis glutathione transferases. GSTU4 is induced by H 2 O 2 38 and was shown to have GSH-conjugating and GSH-dependent peroxidase activities 39 . Finally, ascorbate oxidases (AO) catalyze oxygen reduction to water using ascorbate as the electron donor. AO activity could affect the overall redox state and could contribute to creation of a hypoxic microenvironment 40 . Recently, low leaf AO activities have been associated with the resistance against aphids, associated with the more reduced redox state of the apoplast 41 . We challenged the involvement of these genes in the establishment of the Arabidopsis-mite interaction by comparing the leaf damage measured as higher surface area of chlorotic spots when mites fed on the WT, loss-of-function (for BBE22, GPX7, GSTU4 and AO) and overexpression (for BBE22) lines. As perturbation of ROS homeostasis is expected to affect the formation of chlorotic spots, we also monitored mite fecundity as a measure of mite performance in the Arabidopsis-mite interaction (Fig. 2). Reduced expression of genes involved in H 2 O 2 catabolism, as well as overexpression of BBE22 (encoding a H 2 O 2 synthesizing enzyme), resulted in increased leaf damage and increased mite fecundity. These data clearly established the involvement of these genes in the establishment of the Arabidopsis-mite interaction. However, contrary to our expectations, the expression of these genes and the effects they had on the Arabidopsis-mite interaction did not correlate with the H 2 O 2 levels (Fig. 3). Several factors may contribute to this discrepancy. First, the genetic modification of any ROS-related gene was accompanied by transcriptional changes in other ROS-related genes (Fig. 5) and variations in ROS-related enzymatic activities (Fig. 4). An example of transcriptional reprogramming of ROS-related genes comes from the BOTRYTIS-INDUCED KINASE1 (BIK1) gene, which was also induced in the massively mite-infested Bla-2 and Kon plants 29 , and encodes an Arabidopsis receptor-like cytoplasmic kinase involved in ROS modulation. It has been previously shown that loss of BIK1 mutant exhibited elevated basal expression of ROS-generating and -responsive genes, but not ROS-metabolizing genes, rendering the bik1 mutant more resistant to M. persicae than wild-type plants 42 . Figure 7 shows the complexity of the network constructed from the functional interactions found among the ROS-related enzymes, activities and compounds analyzed in this work. Multiple connections are depicted for each node, which implies a tight interrelation among the different players to deal with the dual detrimental effect/signaling role of ROS. Thus, the perturbations of the expression of the ROS-associated genes have more global effect on gene expression, modifying the final physiological output of the initial perturbation.
Second, there is a cross-talk between the ROS and hormonal signaling pathways (Fig. 7). Extensive characterization of plant-herbivore interactions has demonstrated the integration between ROS and hormonal signaling in plant defense 20,43 . ROS and hormonal pathways have been connected through glutathione status. The SA-activated transcription factor NPR1 is known to be regulated by modulation of thiol status 44,45 , and glutathione has been suggested as a factor determining basal JA gene expression 46 . Previous results in Arabidopsis plants with different levels of MATI expression support the link between ROS, glutathione and the hormonal response of Arabidopsis plants to spider mites 28 . As the activity of most ROS-related enzymes likely has an impact on glutathione levels, genetic manipulation of ROS-related genes could have a subsequent effect on hormonal signaling pathways. Signaling effects caused by MATI support a fine-tuning of the expression of genes involved in the hormonal network leading to final responses through the JA and SA pathways 28 . Silencing AO, GPX7 and GSTU4 lines showed an expression pattern of SA and JA related genes that mostly resemble that observed in the wild-type. However, several differences could be appreciated. AO lines have an apparent enhance of the SA pathway. GPX7 KO has a strongest induction of PDF1.2 after mite treatment. GSTU4 lines showed a lower induction of PR1 after mite infestation together variable differences in VSP2 and PDF1.2 expression. In contrast, BBE22 had a profound effect on JA responses. Whereas loss of BBE22 function favored MYC2 regulated responses that are marked with the VSP2 gene, the overexpression of BBE22 shifted the JA-responses to the antagonistic ERF1/ORA59 regulated pathway that was monitored by the expression of PDF1.2 gene. Therefore, the variations in the hormonal signaling pathways would modify the resistance/susceptibility status of the T-DNA insertion lines.
Third, as H 2 O 2 levels were quantified at the whole leaf level after 24 h of mite feeding, we cannot discard transient or compartmentalized changes in ROS signals that could affect plant response, including hormonal signaling. In fact, the interaction ROS-hormones may have complex and variable effects on plant susceptibility to the attacker. The oral secretions of Manduca sexta induced ROS by a rapid elicitation of the NADPH oxidase Narboh D expression in Nicotiana attenuata. Silencing Narboh D plants were more vulnerable to the lepidopteran S. littoralis after M. sexta elicitation but showed no difference in the induction of jasmonic acid 47 . In contrast, fatty-acid amides induced ROS in Arabidopsis and suppress insect related defenses, most likely by attenuating jasmonic acid responses. NADPH oxidase rbohD/F mutant plants were impaired in ROS production and were more resistant to the herbivores Spodoptera exigua and Trichoplusia ni 48 . These examples reinforce the importance of ROS in the establishment of the plant-herbivore interaction, but also point to an unpredictable final effect of manipulation of ROS-related genes.
In conclusion, the integration of ROS and ROS-related compounds and enzymes in the response of Arabidopsis to the spider mite T. urticae has been confirmed. The challenge is the elucidation of specific molecular interactions that induce plant response. Genetic manipulation of ROS-related genes permits us to deep into the pathways associated to the induction/repression of ROS enzymes. However, the complex network that includes ROS-related genes hinders the prediction of specific effects exerted by specific genes. Furthermore, additional variables, such are the participation of other enzymes/enzymatic activities in the production/elimination of ROS or the compartmentation of ROS molecules leading to transient and subcellular ROS bursts, are expected to modulate ROS-responses and should be kept in mind while interpreting the herbivore-plant ROS signals.

Gene expression analyses by real time PCR (RT-qPCR). A. thaliana rosettes from Col-0, Bla-2 and
Kon accessions were sampled after different times of mite infestation (1, 3, 6, 12 and 24 h). A. thaliana Col-0 flowers, roots, siliques, leaves from stems, and 1, 2 and 3 week-old rosettes were also collected. In addition, Arabidopsis entire plants from selected T-DNA insertion lines and from the non-transformed Col-0 controls were collected. Total RNA was extracted as previously described 49 and reverse transcribed using Revert AidTM H Minus First Strand cDNA Synthesis Kit (Fermentas). RT-qPCR was performed for triplicate samples as previously described 30 using a SYBR Green detection system (Roche) and the CFX Manager Software 2.0 (Bio-Rad). Ubiquitin was used as housekeeping gene for normalization. Gene expression was referred as relative expression levels (2 −dCt ) or fold change (2 −ddCt ) 50 . Specific primers were designed through the PRIMER 3 program (http:// bioinfo.ut.ee/primer3-0.4.0/). Primer sequences are indicated in Supplementary Table S2.
Spider mite maintenance and fitness analyses. A colony of T. urticae, London strain (Acari: Tetranychidae) provided by Dr. Miodrag Grbic (UWO, Canada), was reared on beans (Phaseolus vulgaris) and maintained in growth chambers (Sanyo MLR-350-H) at 25 °C ± 1 °C, >70% relative humidity and a 16 h/8 h day/ night photoperiod. Mites were synchronized by inoculating 100 adult females (random age) on one leaf of bean confined in a closed system under water-soaked cotton. After one day, adult females were removed, and 10 days after, population on the leaf was synchronized. Spider mite fecundity was recorded on T-DNA insertion lines and non-transformed Col-0 controls. T. urticae fecundity assay was performed on detached leaves from 3 week-old plants. The newest emerged leaf (about 1 cm long) from each plant was placed in special dishes and infested with 12 synchronized adult females. After 36 h of infestation, the number of eggs was counted. Eight replicates were used for each plant genotype.
Plant damage determination. Quantification of plant damage after spider mite feeding was done on Arabidopsis entire plants from selected T-DNA insertion lines and non-transformed Col-0 controls. Three week-old plants were infected with 20 T. urticae female adults per plant. After 4 days of feeding, leaf damage was assessed by scanning the entire rosette using a hp scanjet (HP Scanjet 5590 Digital Flatbed Scanner series), as previously described 27 . Leaf damage was calculated in mm 2 , using Adobe Photoshop CS software. Six replicates were used for each genotype.

H 2 O 2 determination in leaf extracts. Three week-old Arabidopsis plants from T-DNA insertion lines
and non-transformed Col-0 controls were infested with 20 mites and incubated for 24 h. The H 2 O 2 concentration of crude extracts was determined by spectrofluorometry, basically as previously described 51 . All operations were performed at 0-4 °C. Leaves (0.1 g) from rosettes were homogenized in 300 µL 25 mM H 2 S0 4 and supernatants collected by centrifugation at 15,890 g for 5 min. The pH of leaf extracts was adjusted to 7.0 with NaOH and these extracts were used to measure the H 2 O 2 concentration. The reaction mixtures (300 µl) contained 50 mM Hepes buffer, pH 7.6, 50 mM homovanillic acid and 10 µL of sample. The reaction was started by adding 4 µM horseradish peroxidase and the fluorescence produced was measured in a Varioskan ® LUX (Thermo Fisher Scientific) microplate reader at excitation and emission wavelengths of 315 and 425 nm, respectively. The H 2 O 2 concentration was determined from a calibration curve of H 2 O 2 (Sigma) in the range 0.1-1000 µM. Three replicates were done by genotype and treatment.
The accumulation of H 2 O 2 was visualized using the 3,3-diaminobenzidine tetrachloridehydrate (DAB) substrate (Sigma-Aldrich) which produces a brown precipitate after oxidation in the presence of H 2 O 2 52 . Col-0 Arabidopsis leaf disks (1 cm diameter) from T-DNA insertion lines and the non-transformed Col-0 controls were infested with 10 mites and incubated for 24 h. Infested and non-infested control disks were stained with DAB 53 SCIENTIfIC REpORtS | (2018) 8:9432 | DOI:10.1038/s41598-018-27904-1 and observed under a Leica fluorescence stereoscope. DAB staining specificity was confirmed in presence of the H 2 O 2 scavenger, ascorbic acid (10 mM).
Antioxidant enzyme activity measurements. Three week-old Arabidopsis plants from T-DNA insertion lines and the non-transformed Col-0 controls were infested with 20 mites and incubated for 24 h. Ascorbateglutathione recycling enzymes were extracted from rosette leaves into a pre-cooled extraction buffer consisting of 1.5 ml of 0.1 M NaH 2 PO 4 , 1 mM EDTA (pH 7.5), 1% (w/v) insoluble PVP, 1 mM ascorbate. 1 M ascorbate was freshly prepared and added to the extraction medium with an approximated ratio of 1.5 ml/150 mg FW. The homogenate was centrifuged at 14,000 g for 10 min at 4 °C and the supernatant was used immediately for enzyme activities. All assays were performed in a final volume of 0.2 ml per well in a UV-microplate at 25 °C. Samples, controls and blanks were analyzed in triplicate. The microplate reader was a Varioskan ® LUX (Thermo Fisher Scientific) spectrophotometer. The SkanIt RE 4.0 software was used to check the reader and to analyze enzymatic reactions. Ascorbate peroxidase (APX, EC 1.11.1.11), dehydroascorbate reductase (DHAR, EC 1.8.5.1) and glutathione reductase (GR, EC 1.6.4.2) activities were quantified as previously described 54 with slight modifications. Catalase (CAT, EC 1.11.1.6) activity was measured by determining the degree of H 2 O 2 decomposition at 240 nm for 2 min 55 . The reaction buffer consisted of 0.1 M NaH 2 PO 4 , 1 mM EDTA (pH 7.5). Then, 2 M H 2 O 2 was added to each well (40 mM) and absorbance was recorded. The reaction was initiated by the addition of 10 µl of crude extract. All specific activities were calculated using the extinction coefficients and referring the results to the milligrams of protein in the extract (determined according to the method of Bradford 56 , with bovine serum albumin as standard). For APX, (ɛ 290 = 2.8 mM −1 cm −1 ), the results indicate the µmol of ascorbate oxidized per minute. Cat activity (ɛ 240 = 40 M −1 cm −1 ) is expressed as H 2 O 2 mmol removed in a minute. For DHAR (ɛ 265 = 14 mM −1 cm −1 ), results represent mmol of DHA reduced per minute. In the case of GR (ɛ 340 = 6.22 mM −1 cm −1 ), data show mmol of NADPH oxidized in a minute. At least six replicates were done by genotype and treatment.
Statistical analysis was performed using One-Way ANOVA for gene expression in Arabidopsis accessions, damage analysis and spider mite bioassays. Two-Way ANOVA was used for H 2 O 2 accumulation. For the enzymatic activities and gene expression studies in T-DNA lines two analysis were performed, One-Way ANOVA to compare the effect of the genotype within basal or infested conditions and T-Student to compare within each genotype the effect of the infestation. ANOVA analyses were followed by the Student-Newman-Keuls multiple comparison test. In figures, significant differences (P < 0.05) are reported with different letters or asterisks depending on the analysis performed.