Title: Quantitative Genetic Analysis of Thermal Dissipation in Arabidopsis

Feedback de-excitation is a photosynthetic regulatory mechanism that can protect plants from high light stress by harmlessly dissipating excess absorbed light energy as heat. To understand the genetic basis for intraspecies differences in thermal dissipation capacity, we investigated natural variation in Arabidopsis thaliana . We determined the variation in the amount of thermal dissipation by measuring nonphotochemical quenching (NPQ) of chlorophyll fluorescence in Arabidopsis accessions of diverse origins. Ll-1 and Sf-2 were selected as high NPQ Arabidopsis accessions, and Col-0 and Ws-2 were selected as relatively low NPQ accessions. In spite of significant differences in NPQ, previously identified NPQ factors were indistinguishable between the high and the low NPQ accessions. Intermediate levels of NPQ in Ll-1 x Col-0 F 1 and Sf-2 x Col-0 F 1 compared to NPQ levels in their parental lines, and continuous distribution of NPQ in F 2 indicated that the variation in NPQ is under the control of multiple nuclear factors. To identify genetic factors responsible for the NPQ variation, we developed a polymorphic molecular marker set for Sf-2 x Col-0 at approximately 10 cM intervals. From quantitative trait locus (QTL) mapping with undistorted genotype data and NPQ measurements in an F 2 mapping population, we have identified two high NPQ QTLs ( HQE1 and HQE2 ) on chromosomes 1 and 2, and the phenotype of HQE2 was validated by analysis of near isogenic lines. Neither QTL maps to a gene that had been identified previously in extensive forward genetics screens using induced mutants, suggesting that quantitative genetics can be used to find new genes affecting thermal dissipation.


INTRODUCTION
Plants require light energy, by definition, to drive photosynthesis.
However, too much light causes photo-oxidative damage in plants ( Barber and Andersson, 1992). Thus plants have diverse defense mechanisms against high light stress (Niyogi, 1999). For example, chloroplasts can move to absorb less light energy (Kasahara et al., 2002), and light-harvesting antenna size in chloroplasts can be reduced (Anderson, 1986). Plants also can harmlessly dissipate excess absorbed light energy as heat (Müller et al., 2001), and they have alternative electron transport pathways to relieve overreduction of electron transport components under stress conditions (Niyogi, 2000;Ort and Baker, 2002). Thermal dissipation is mediated by a mechanism called feedback deexcitation. Feedback de-excitation dissipates excess absorbed light energy as heat, thereby protecting plants from high-light stress (Horton et al., 1994;Niyogi, 1999). The amount of feedback de-excitation can be quantified by measuring nonphotochemical quenching of chlorophyll fluorescence (NPQ) (Müller et al., 2001;Baker, 2008). NPQ is induced by appearance of high light and is relaxed following disappearance of the high light. Based on its relaxation kinetics, NPQ can be divided into at least three components: energy-dependent quenching (qE), state-transition quenching (qT) and photoinhibitory quenching (qI) (Maxwell and Johnson, 2000;Müller et al., 2001). Among them, qE is generally the major component in plants (Maxwell and Johnson, 2000). Biochemical and molecular genetics studies have shown that a pH gradient across the thylakoid membrane (Briantais et al., 1979;Munekage et al., 2001Munekage et al., , 2002, the xanthophyll cycle (Demmig-Adams et al., 1990;Niyogi et al., 1998), and the PsbS protein (Li et al., 2000) of photosystem II (PSII) are important factors involved in controlling the induction and/or extent of NPQ. Based on the semidominance of loss-of-function mutations (Li et al., 2000;Li et al., 2002) and overexpression of the psbS gene (Li et al., 2002) in Arabidopsis, the expression level of the PsbS protein has been 6 suggested as an important factor in determining the qE (and total NPQ) capacity of plants.
Naturally occurring variation in NPQ capacity has been observed in different plant species (Johnson et al., 1993;Demmig-Adams and Adams, 1994;Demmig-Adams, 1998). The variation of saturated NPQ values ranges from 2.5 to 4.5 in British plant species, and plants grown in open habitats tend to have larger NPQ capacity (Johnson et al., 1993). Sun-acclimated plants contain up to four times as much NPQ capacity as low light-acclimated plants of the same species (Osmond et al., 1993;Ruban et al., 1993;Brugnoli et al., 1994;Demmig-Adams and Adams, 1994;Demmig-Adams et al., 1995;Demmig-Adams, 1998;Roberts et al., 1998). In Monstera deliciosa, for example, sun-acclimated leaves showed higher NPQ than low light-acclimated leaves (Demmig-Adams and Adams, 1994), and this difference in NPQ is correlated with changes in PsbS protein levels (Demmig-Adams et al., 2006). Although it has been suggested that there may be a genetic basis for the variation (Horton et al., 1994), this possibility has not yet been analyzed.
This kind of natural variation in plant traits, in most cases, shows continuous variations that are under the control of polygenic factors, and quantitative genetic studies are required to understand the genetic basis of the variation (Alonso- Blanco and Koornneef, 2000). Arabidopsis thaliana has become a model system for plant QTL mapping (Alonso-Blanco and Koornneef, 2000;Maloof, 2003;Tonsor et al., 2005), because it has considerable trait variations among accessions, advanced molecular biological tools for efficient genotyping with molecular markers, and a fully sequenced genome. In addition, development and improvement of statistical tools for QTL mapping facilitate analyses of quantitative traits (Lander and Botstein, 1989;Zeng, 1994;Sen and Churchill, 2001). Using Arabidopsis, a number of quantitative traits have been analyzed, however QTL mapping is still underutilized for photosynthesis-related traits, in spite of advantages for quantitative genetic studies such as simple ways for quantification of photosynthetic parameters (Krause and Weis, 1991;Laisk et al., 2002;Long and Bernacchi, 2003;Baker, 2008). 7 In this article, we report natural variation of NPQ among Arabidopsis accessions and test the hypothesis that the variation between a high NPQ accession (Sf-2) and a low NPQ accession (Col-0) is related to the PsbS protein.
We measured induction and relaxation of NPQ in Arabidopsis accessions and divided them into high and low NPQ accessions. Biochemical and molecular biological experiments did not associate the NPQ differences with PsbS or other previously identified NPQ factors. Genetic analyses revealed that the differences are controlled by polygenic nuclear factors. To identify these factors, we performed QTL mapping using Sf-2 x Col-0 F 2 progeny as a mapping population and identified two high NPQ QTLs. The significance of NPQ variation and possible roles for these QTLs in thermal dissipation are discussed.

NPQ in Arabidopsis accessions
To study the natural variation of NPQ capacity, we initially measured total NPQ in 62 Arabidopsis accessions (Supplemental Table S1). The tested accessions originated from diverse growth conditions, and parental lines of recombinant inbred lines (RILs) that are available or will be available in the near future were also included (http://www.inra.fr/internet/Produits/vast/RILs.htm). To focus on the major qE component, NPQ was measured for 10 min during actinic illumination with 1500 µmol photons m -2 s -1 , followed by 5 min in the dark to determine its relaxation. All NPQ measurements were done between 1 PM and 5 PM to exclude any possible effects of circadian control of NPQ. The results showed that there are variations in NPQ capacity among Arabidopsis accessions (Supplemental Table S1), and that most of the NPQ is qE because it relaxed very fast in the dark (data not shown). Based on these measurements, the accessions were classified into high NPQ and low NPQ Arabidopsis accessions.
Ll-1, Sf-2, Ts-1, Van-0, Mz-0, Mr-0 and Kin-0 are high NPQ Arabidopsis accessions containing total NPQ capacity higher than 3.0, whereas Col-0, Ler 8 and Ws-2, so-called lab accessions, are low NPQ Arabidopsis accessions having total NPQ capacity below 2.5 (Supplemental Table S1). Interestingly, the three highest accessions, Ll-1, Sf-2 and Ts-1, all originated from Spain. Except for Ll-1, the other high NPQ accessions have been used as parental lines of RILs, but the RILs were not yet available at the time these experiments were initiated. We decided to use Ll-1 and Sf-2 as high NPQ accessions and Col-0 and Ws-2 as low NPQ accessions for further measurement and characterization.
After growing these selected accessions together in identical growth conditions, we measured their NPQ in the next generation. Ll-1 and Sf-2 showed significantly higher total NPQ than Col-0 and Ws-2 after 3 min in actinic illumination (P < 0.01), and the NPQ differences became larger with time (P < 0.001) ( Figure 1).

Characterization of NPQ factors in selected accessions
Photosynthetic efficiency and previously identified NPQ factors were compared between these high and low NPQ accessions. First of all, quantum yield of PSII (Φ PSII ) measuring the proportion of energy used for photochemistry (Maxwell and Johnson, 2000) was calculated ( Figure 2A). The levels of Φ PSII in these accessions were very similar to each other, although Ll-1 had a slightly lower value than the others. In addition, the maximum quantum yield of PSII (F v /F m ) of the accessions was also indistinguishable (data not shown). Ll-1 and Sf-2 showed higher NPQ than Col-0 and Ws-2 in increasing light intensities, however the differences were not statistically significant up to the highest light intensity applied ( Figure 2B), possibly due to the longer measuring time and greater contributions of other NPQ components (e.g. qI).
To check differences in xanthophyll cycle activity, the contents of each xanthophyll cycle pigment were determined. The de-epoxidated xanthophylls, antheraxanthin (A) and zeaxanthin (Z), are generated from violaxanthin (V) by V de-epoxidase (VDE) in high light (Yamamoto et al., 1962). In low light (150 µmol photons m -2 s -1 ), Ll-1, Sf-2, Col-0 and Ws-2 contained a very small proportion of 9 de-epoxidated forms (A and Z). After being exposed to high light (1700 µmol photons m -2 s -1 ) for 4 min and 40 min, all accessions showed very similar levels of de-epoxidation ( Figure 2C). These results indicated that the high and low NPQ accessions contain functional VDE and that there are no differences in VDE activation by high light between the two groups. Additionally, the chlorophyll a/b ratios reflecting the PSII antenna size were similar to each other ( Figure 2D).
Third, the amount of the PsbS protein was compared. PsbS is required for qE, and qE (and total NPQ) is proportionally increased as the PsbS protein level increases (Li et al., 2000;Li et al., 2002;Li et al., 2002). To determine whether PsbS is responsible for the NPQ variation, the amount of PsbS in each accession was estimated by immunoblot analysis. In these accessions, the relative PsbS protein levels were very similar to each other, compared to D1 protein levels, although the D1 protein level in Col-0 was slightly lower than the others ( Figure 3A). We also determined the PsbS genomic DNA sequences and then compared their predicted amino acid sequences, because changes in amino acid residues could affect PsbS function (Li et al., 2002). Interestingly, we found DNA sequence polymorphisms between the high and the low NPQ accessions: Ll-1 and Sf-2 have nearly identical PsbS genomic DNA sequences, and the Col-0 and Ws-2 sequences are identical to each other ( Figure 3B). However, all polymorphisms in exons are silent; therefore, it was predicted that there are no amino acid differences. Taken together, these high and low NPQ accessions contained nearly identical levels of previously identified NPQ factors including VDE activity and PsbS, suggesting that the differences in NPQ may be controlled by a novel factor or factors.

NPQ of Sf-2 x Col-0 F 1 and F 2 progenies
To understand the genetic basis of the NPQ variation, we crossed Ll-1 and Sf-2 to Col-0, and measured induction and relaxation of NPQ in F 1 and F 2 progenies. Col-0 was used as a low NPQ parental line, because the genomic NPQ levels in the F 1 were significantly lower than those of Ll-1 and Sf-2, and were slightly higher than that those of Col-0, although multiple comparison tests (Duncan) indicated that F 1 from both crosses were in the same group as Col-0 ( Figure 4). F 1 plants from reciprocal crosses showed similar levels of NPQ and were statistically grouped together ( Figure 4). These observations indicated that the high NPQ is controlled by nuclear rather than cytoplasmic factor(s). To determine how many nuclear factors are involved, we measured NPQ in 80 Sf-2 x Col-0 F 2 progeny, and analyzed the frequency distribution of NPQ every 2 min during illumination with actinic light. At each time point, NPQ levels were widely distributed ( Figure 5), and a normality test (Shapiro-Wilk) showed that NPQ levels follow a normal distribution (P < 0.01), indicating that more than one factor is responsible for the NPQ variation. Between these two high NPQ accessions, Ll-1 is very similar to Ll-0 that is being used as a parental line for generation of Ll-0 x Ler RILs. Therefore, we decided to use Sf-2 and Col-0 as high NPQ and low NPQ accessions, respectively, to locate QTLs responsible for high NPQ.

Determination of genotypes
To determine genotype in the F 2 mapping population, we developed a polymorphic DNA marker set (Supplemental Table S2). Using the polymorphic markers, genotypes at each marker position were determined in 72 F 2 progeny.
The genotyping was almost complete, with only a few missing genotype data.
From the genotype data, genotype frequencies at marker positions were calculated (Supplemental Figure S1). The G-test (Dytham, 2003) indicated that overall, no significant segregation distortion was observed in our F 2 mapping population (Supplemental Figure S1). With the genotype data, we also confirmed the order of markers on each chromosome and estimated the distance between each two consecutive markers using MAPMAKER/EXP 3.0 (Lander et al., 1987). The re-estimated marker order was the same as the marker order on the Arabidopsis physical map (Supplemental Table S2). The average distance between flanking markers was calculated as 10.7 cM. The longest interval was 31.9 cM between SO392 and A44575 on chromosome 1, and the shortest one was 2.2 cM between C6 and NGA126 on chromosome 3 (Supplemental Table   S2).

QTL mapping
To find chromosome regions that are responsible for the high NPQ phenotype of Sf-2, we conducted QTL mapping using the R/qtl program (Broman et al., 2003). The QTL mapping with 10 min NPQ identified two high NPQ QTLs, High qE 1 (HQE1) and High qE 2 (HQE2) ( Figure 6J). The LOD scores of both QTLs were higher than the threshold LOD score of 3.2 that was estimated after 1000 permutations with 0.05 significance (Churchill and Doerge, 1994). HQE1 is 35 cM from the first marker on chromosome 1, and the nearest marker is NF19G9. HQE2 is 67.5 cM from the first marker on chromosome 2, and the nearest marker is C4H (Table I) (Table I). Among total variance, the portion of variance explained by these two QTL was 37.8%. In addition, there was little evidence supporting interaction between HQE1 and HQE2 (data not shown).
We also did QTL mapping using NPQ values measured at each minute and found interesting differences between HQE1 and HQE2 ( Figure 6A-I). The LOD score of HQE1 became significant after 3 min in actinic illumination ( Figure   6C), and thereafter HQE1 remained as a significant high NPQ QTL during the 10 min measurement. At 6 min, HQE1 alone explained 27.9% of total NPQ variance ( narrower than the interval at 10 min (Table I). In contrast, the response of HQE2 to the actinic light was relatively slower than that of HQE1. HQE2 became a significant QTL after 7 min in the light ( Figure 6G), and it remained significant until the end of the 10 min measurement. Therefore, from 3 min to 6 min, only HQE1 seems to be responsible for the high NPQ, and from 7 min in the actinic light, both HQE1 and HQE2 may control the high NPQ phenotype.
To confirm our NPQ QTL mapping, we generated near isogenic lines, NQE2, harboring the Sf-2 allele of HQE2 in the Col-0 background. F1 plants of Sf-2 and Col-0 cross were backcrossed to Col-0 four times, and lines were selected based on genotyping. NPQ measurements indicated that NQE2 showed higher NPQ than control lines containing the Col-0 allele of HQE2 ( Figure 7).

DISCUSSION
In our survey of natural variation of NPQ capacity in Arabidopsis, we tried to find some accessions that have much less NPQ than other accessions. In such cases, it might be relatively straightforward to determine the genetic basis for NPQ difference, as found for other traits (Maloof et al., 2001). However, so far, we could not find Arabidopsis accessions having very low NPQ. That might be because NPQ is an important photoprotection mechanism without which plants would suffer more oxidative stress (Li et al., 2002) and show much lower fitness estimated by lifetime seed production (Külheim et al., 2002).
Several accessions exhibited relatively high NPQ compared to the standard lab accession, Col-0. Why do the high NPQ accessions have a higher capacity for thermal dissipation than other accessions? We found a significant, but very weak correlation between latitude and NPQ among European accessions (data not shown). The latitude of geographical origin has been shown to have a significant relationship with variation of other traits in Arabidopsis (Li et al., 1998;Maloof et al., 2001;Stinchcombe et al., 2004).
Instead of latitude, environmental parameters in the natural habitat may be more 13 critical for NPQ capacity, because NPQ is critical to deal with sudden changes in light intensity (Müller et al., 2001;Külheim et al., 2002). To better understand the relationship between NPQ and natural habitat, habitat information needs to be quantified (Johnson et al., 1993).
Another possible explanation for the weak relationship between latitude and NPQ is that mechanisms other than NPQ may play a major role for photoprotection in some accessions (Johnson et al., 1993). Although NPQ capacity in Cvi and Ler, for example, was very similar (Supplemental Table S1), Cvi showed higher photo-oxidative stress tolerance than Ler as expected from Cvi's origin, a tropical zone (Abarca et al., 2001). Interestingly, Cvi contains a different allele of chloroplastic Cu/Zn-superoxide dismutase 2 (CSD2) from that of Ler in predicted amino acid sequences and in mobility in a protein gel, and it was suggested that the Cvi-specific CSD2 is one of the factors responsible for the higher tolerance of Cvi (Abarca et al., 2001).
Strong relationships were observed between de-epoxidation state and NPQ capacity among plant species as well as among plants of the same species grown in different light conditions (Johnson et al., 1993;Brugnoli et al., 1994;Demmig-Adams and Adams, 1994;Demmig-Adams et al., 1995;Demmig-Adams, 1998;Adams III et al., 1999). In addition, it has been known that the PsbS protein level can determine the NPQ level and that certain amino acid residues in PsbS play an important role in NPQ (Li et al., 2002;Li et al., 2002). However, the de-epoxidation states of the xanthophyll cycle in Ll-1, Sf-2, Col-0 and Ws-2 were similar to each other ( Figure 2C), and Ll-1 and Sf-2 contained almost identical levels of PsbS compared to Col-0 and Ws-2 without any polymorphisms in predicted amino acid sequences (Figure 3). These results suggest that the QTLs identified here, HQE1 and HQE2, do not act directly or indirectly through NPQ factors that have been found previously in extensive biochemical and genetic analyses of NPQ.
The response of HQE1 to photosynthetically active light was faster than that of HQE2 ( Figure 6A-J). After being exposed to actinic illumination for 3 min, HQE1 and HQE2 were responsible for the high NPQ ( Figure 6G). These results indicated that early NPQ induction might be affected by HQE1 alone, and saturated NPQ levels might be explained by both HQE1 and HQE2 without any interaction between HQE1 and HQE2. Therefore, it could be speculated that sudden changes in light intensity would be dealt with by NPQ controlled by HQE1 and that, if the high light intensity continues, HQE2 and HQE1 together increase NPQ capacity to cope with the high light. Although HQE1 was the only identified significant QTL between 3 and 6 min, the NPQ measured during that period time was widely distributed ( Figure 5).
In conclusion, we were able to identify high and low NPQ Arabidopsis accessions and map QTLs that influence NPQ in two of these accessions. This information provides a guideline for selection of appropriate RILs and experimental design for future NPQ QTL mapping. We have also developed DNA molecular markers showing polymorphisms between Sf-2 and Col-0 that can be used for genotyping of Sf-2 x Col-0 RILs. Although we had hypothesized that the difference in NPQ between the accessions would be related to PsbS, no differences in known NPQ factors were found. Analysis of other accessions might reveal cases in which NPQ variation is related to PsbS. Identification of the high NPQ QTLs described here will eventually make it possible to clone new genes responsible for high NPQ.

Plant materials and growth conditions
Arabidopsis thaliana seeds were obtained from the Arabidopsis Biological Resource Center (ABRC) and from the seed collection of the Niyogi laboratory.
Before planting on soil, the seeds were kept in water at 4°C for 1 week. dark) for 5 to 6 weeks in a light (150 µmol photons m -2 s -1 ) and temperature (22°C)-controlled growth chamber before chlorophyll fluorescence measurements and leaf tissue harvests.

Chlorophyll fluorescence measurements
Plants were dark-adapted for 2 h before NPQ measurement. The measurement began at 1 PM, and all measurements were completed by 5 PM.
Chlorophyll fluorescence parameters were measured on attached rosette leaves PCR conditions for previously available markers were obtained from the Arabidopsis Information Resources (TAIR), and restriction enzymes used for each marker are listed in Supplemental Table S2. For NQE2, CZSOD2 and T27E13 were used to determine genotypes between ER and C4H, and COP1 and VE017 were utilized for genotyping between C4H and NGA168. The newly PCR reactions of SNP397C and SNP299C were annealed at 58°C and polymerized for 1 min, and then the PCR products were digested with HpyCh4 IV and Dde I, respectively. Amplified A44575 DNA following PCR at 59°C annealing and 1 min polymerization was digested with Hinf I. PCR for D12320 was done at 59°C annealing and 30 sec polymerization, and then Dde I-digested fragments were separated on 3% agarose gels.
For genotyping, genomic DNA was extracted as follows: frozen small leaves were ground in a 1.5 ml tube and then the fine powder was thawed in 10 µl of alkali solution (0.5 N NaOH) (Klimyuk et al., 1993)

Statistical analysis and QTL mapping
Test statistics including ANOVA, Student's t-test and multiple comparison tests were calculated using Microsoft Excel (Seattle, WA) and XLSTAT (New York, NY). Following the genotyping, a G-test was used to test whether the genotype ratio at each marker position fitted to the expected F 2 segregation ratio (Supplemental Figure S1) (Dytham, 2003). Intervals between neighboring markers were estimated by using MAPMAKER/EXP 3.0 (Lander et al., 1987).
The first marker on each chromosome was set at 0 cM, and each marker position was calculated from accumulated intervals between markers.
QTL mapping was performed using the R/qtl program (Broman et al., 2003). The LOD significance threshold was estimated after permutation tests that had been replicated 1000 times (Churchill and Doerge, 1994). LOD support interval was calculated using the "lodint" function with 1.5 LOD unit drop to form the support interval. Additive and dominance effects of each QTL were calculated using estimations of the means of the three genotype groups that were in the output of "scanone" function with the EM method in the R/qtl program (Lander and Botstein, 1989;Tanksley, 1993). Estimation of percent variance explained by each QTL was obtained from a drop-one-term analysis of results in an additive model. The additive model was calculated using the "fitqtl" function with imputation method in the R/qtl program (Sen and Churchill, 2001).

Supplemental Table S1. List of NPQ values in Arabidopsis accessions
Supplemental Table S2 PSII and NPQ values were estimated by applying a pulse of saturating light to attached rosette leaves that had been exposed to the corresponding light intensity for 5 min. Each data point represents the mean ± SD (n=3). (C and D) Total xanthophyll cycle pigments (V: violaxanthin; A: antheraxanthin; Z: zeaxanthin) relative to chlorophyll a and chlorophyll a/b ratio were determined in low light (LL: 150 µmol photons m-2 s-1) and after being treated in high light (HL: 1700 µmol photons m-2 s-1) for 4 and 40 min. Each data point represents the mean ± SD (n=2~3).           I  II III IV V  I  II III IV V  I  II III IV V   5   4   3   2   1   0   I  II III IV V  I  II III IV V  I  II III IV V   I  II III IV V  I  II III IV V  I  II III   . NPQ levels of near isogenic lines for HQE2 (NQE2). NPQ levels were measured in near isogenic lines for HQE2 (NQE2), which contain the Sf-2 allele of HQE2 (black) in a Col-0 background (white bar). The breakpoints of the Sf-2 introgression are between the markers CZSOD2 and T27E13 on one side and between the markers VE017 and NGA168 on the other side. Segregants containing the Col-0 allele of HQE2 (CON) were used as control lines. The significance of differences between NQE2 and CON was tested using Student's t-test (*: P<0.05; **: P<0.01).