A New F131V Mutation in Chlamydomonas Phytoene Desaturase Locates a Cluster of Norflurazon Resistance Mutations near the FAD-Binding Site in 3D Protein Models

The green alga Chlamydomonas reinhardtii provides a tractable genetic model to study herbicide mode of action using forward genetics. The herbicide norflurazon inhibits phytoene desaturase, which is required for carotenoid synthesis. Locating amino acid substitutions in mutant phytoene desaturases conferring norflurazon resistance provides a genetic approach to map the herbicide binding site. We isolated a UV-induced mutant able to grow in very high concentrations of norflurazon (150 µM). The phytoene desaturase gene in the mutant strain contained the first resistance mutation to be localised to the dinucleotide-binding Rossmann-likedomain. A highly conserved phenylalanine amino acid at position 131 of the 564 amino acid precursor protein was changed to a valine in the mutant protein. F131, and two other amino acids whose substitution confers norflurazon resistance in homologous phytoene desaturase proteins, map to distant regions in the primary sequence of the C. reinhardtii protein (V472, L505) but in tertiary models these residues cluster together to a region close to the predicted FAD binding site. The mutant gene allowed direct 5 µM norflurazon based selection of transformants, which were tolerant to other bleaching herbicides including fluridone, flurtamone, and diflufenican but were more sensitive to beflubutamid than wild type cells. Norflurazon resistance and beflubutamid sensitivity allow either positive or negative selection against transformants expressing the mutant phytoene desaturase gene.


Introduction
Chlamydomonas reinhardtii provides an attractive model organism to dissect processes unique to photosynthetic eukaryotes [1]. Forward genetic screens for gain-of-function mutations that confer herbicide resistance provide a powerful approach to investigate herbicide mode of action [2]. These studies proceed rapidly using algal and plant models containing herbicide target sites, which are often absent in animals and fungi. Resistance resulting from a mutation in a target protein can identify an herbicide's binding site, which is normally revealed by identifying and sequencing the corresponding mutant gene. Identifying genes responsible for herbicide resistance is facilitated by the availability of complete sequences of the C. reinhardtii nuclear and organelle genomes [3,4]. Forward genetics provided a powerful tool to locate the binding sites of triazine herbicides in the D1 protein of photosystem II by sequencing mutant psbA genes encoding D1 from herbicide tolerant C. reinhardtii strains [5,6,7]. The locations of the amino acid substitutions in the D1 protein identified the herbicide target site which overlapped with the Q B quinone binding site in photosystem II [8,9].
Norflurazon (CID 33775) is a bleaching herbicide that blocks the synthesis of carotenoids by inhibiting the activity of phytoene desaturase (PDS) [10]. The first committed step of the carotenoid synthesis pathway produces phytoene in a reaction catalysed by phytoene synthase. PDS catalyses the next step of the pathway involving the two-step dehydrogenation of phytoene to produce zeta-carotene via a phytofluene intermediate [11,12]. Carotenoids protect chloroplasts from excess light energy [13,14] and their absence results in chlorophyll bleaching and eventual cell death [15]. In C. reinhardtii, PDS is encoded by the single nuclear PDS1 gene (Cre12.g509650) located on chromosome 12 [4]. The C. reinhardtii and A. thaliana PDS3 (At4G14210.1) proteins are homologous and share 66% amino acid identity. PDS mutations conferring norflurazon resistance have been described in algae, plants and cyanobacteria. A C. reinhardtii mutant tolerant to 2 mM norflurazon has been described [16] but the mutation (s) responsible remain uncharacterised. Substitutions at arginine 304 (R304S, R304T, R304H, R304C) of the 580 amino acid PDS protein from the aquatic flowering plant Hydrilla verticillata conferred resistance to norflurazon and the related bleaching herbicide fluridone [17,18]. A L516F substitution in the 558 amino acid PDS protein of the green alga Chlorella zofingiensis prevented bleaching in media containing 0.25 mM norflurazon; wild type (WT) cells were bleached by 0.05 mM norflurazon [19]. In the cyanobacterium Synechococcus PCC 7942, mutant PDS proteins conferring resistance to norflurazon contain R195P, L320P, V403G and L436R substitutions [11]. The herbicide lethal dose was 40 mM and 70 mM for strains harbouring the R195P and L436R substitutions, respectively, and 0.5 mM for WT Synechococcus [11]. The 474 amino acid Synechococcus and C. reinhardtii PDS proteins share 62% identity over 448 amino acids. In Synechocystis PCC 6803, R195P and R195S substitutions in PDS gave higher levels of norflurazon resistance than a R195C substitution [20]. The relationship between these resistance mutations located at distant sites of the plant, algal and cyanobacterial PDS proteins and the norflurazon target site in the folded protein is not known.
C. reinhardtii mutants tolerant to norflurazon have been previously isolated [16,21] but the mutant genes responsible for herbicide tolerance have not been isolated and sequenced. Down regulation of PDS RNA levels by about 80% did not appear to affect carotenoid levels in C. reinhardtii [22]. Whilst there are no previous reports of using forward genetics to isolate and sequence mutant PDS alleles conferring norflurazon resistance in C. reinhardtii, substitutions that impair PDS function resulting in light-induced bleaching have been identified [23]. Selection with relatively high concentrations of herbicide has the potential to isolate a new pds1 allele encoding an extremely resistant mutant protein that could act as a marker gene for nuclear transformation [24]. New missense mutations located in regions of the primary sequence not previously associated with herbicide resistance are particularly valuable for mapping the norflurazon binding site. Here we describe the isolation of a new mutant pds1 allele containing the first herbicide-resistance mutation to be localised to the dinucleotide-binding Rossmann-like domain. The mutant allele confers resistance to 150 mM norflurazon allowing direct herbicide-based selection of nuclear transformants. We mapped this new mutation, and other amino acid substitutions conferring norflurazon resistance in homologous proteins, from algae, cyanobacteria and plants, on 3D models of C. reinhardtii PDS to identify the target site of norflurazon.

Isolation of a Mutant Strain Highly Resistant to Norflurazon
After exposure of C. reinhardtii nic7 arg7 y1 mt+ cells to high energy UV light (254 nm), colonies were selected by their ability to grow on TAP medium containing 33 mM norflurazon. Norflurazon resistant mutant strain 1 (NFR1) was chosen for further analysis because it grew much more strongly under 33 mM herbicide selection. The parent nic7 arg7 y1 mt+ strain giving rise to NFR1 was sensitive to norflurazon and henceforth will be referred to as wild type (WT). The lowest concentration of norflurazon tested (1 mM) reduced WT cell densities by about 70% after four days of growth in liquid TAP medium (Fig. 1A). Growth of the WT strain was prevented by 5 mM norflurazon. The effective concentration of norflurazon inhibiting WT growth by 50% (Ec50) after a four day period was 0.760.1 mM. In contrast, strain NFR1 grew to similar densities in the presence or absence of 5 mM norflurazon after four days growth (Fig. 1B). The NFR1 culture densities were highest at four days when exposed to herbicide and then gradually declined. The Ec50 for the NFR1 strain was calculated to be 108 mM norflurazon at four days, which is over 150-fold higher than the WT Ec50. NFR1 grew in media containing 150 mM norflurazon, the highest concentration tested, and after four days reached cell densities that were about 30% of those reached in the absence of herbicide (Fig. 1B). Coloured carotenoid content was measured by absorbance at 470 nm. The coloured carotenoid content of WT cells was reduced by 79% and 99% following exposure to 2 mM and 10 mM norflurazon, respectively ( Fig. 2A). By comparison 2 mM and 10 mM norflurazon reduced coloured carotenoids in NFR1 cells by 13% and 44%, respectively (Fig. 2B).

Genetic Analysis of Norflurazon Resistance
Segregation analysis of resistance to norflurazon was performed by crossing NFR1 mt+ with WT norflurazon-sensitive strains S1D2 mt-or CC621 mt-and scoring haploid progeny dissected from the meiotic products of diploid zygotes. Of the 179 colonies derived from random haploid spores dissected from tetrads, 81 were norflurazon resistant. This data fits a 1:1 segregation pattern for a single Mendelian locus better than the 3:1 segregation ratio predicted for a resistance trait encoded by two unlinked genes, where only 25% of progeny would be resistant to norflurazon. The mutation conferring herbicide resistance was localised to a pds1-nfr1d allele encoding a mutant PDS (see below). Diploid (2 n) cells [25] were used to determine whether the mutant allele in the NFR1 strain was dominant or recessive to the corresponding WT gene. Norflurazon resistant strain NFR1a (pds1-nfr1d, arg7 mt+) was mated to a paraquat (PQR) resistant strain PQR20 (PDS1, nic7, pqr mt2) previously isolated in our laboratory (unpublished) to produce diploid cells. Complementation of the arg7 and nic7 loci in diploid cells resulted in prototrophic growth. All diploid colonies isolated were tolerant to norflurazon and paraquat. This indicated that the mutant genes responsible for norflurazon and paraquat resistance were dominant or semidominant to their corresponding WT alleles. Dominance or semidominance of the pds1-nfr1d allele over the WT PDS1 gene was confirmed by its use in transformation experiments (see below). DNA blot analysis was used to show that the cells tolerant to both herbicides were indeed diploid rather than recombinant cells. Hind III digests of DNA from diploid strains gave rise to a pattern of Gulliver transposable elements [26] representative of all the elements present in the parental PQR20 and NFR1a strains. An example is shown in Figure 3. NFR1a mt+ possesses two Gulliver bands (D and N in Figure 3) not present in the PQR20 mt2 parent, whereas PQR20 contains Gulliver H, which is not present in DNA digests from strain NFR1a. These monosomic D, H and N single copy Gulliver bands unique to one of the parents were less intense in the diploid cells relative to the duplicated Gulliver bands that were common to both parents. This additive pattern of Gulliver elements in diploid strains was consistent with their derivation from the combined genomes of the two parental strains.

A Base Substitution in the pds1-nfr1d Allele Results in a F131V Missense Mutation
The PDS enzyme is a well characterised target of norflurazon [10,11,18,27]. The PDS1 gene (Cre12.g509650) located on chromosome 12 is comprised of six exons (Fig. 4A) which are spliced together into a continuous coding region of 1692 nucleotides terminated by a TAA stop codon. All the pds1-nfr1d exons from strain NFR1 genomic DNA were amplified into PCR products and sequenced directly. Two mutations were found in the pds1-nfr1d allele compared to the C. reinhardtii WT PDS1 sequence in GenBank (Acc. No. XM_001690807). A guanine to adenine transition in exon three (base 987 of the CDS) was a silent mutation that does not change glycine 329 in the protein; base one is the adenine in the start ATG codon. A thymine to guanine transversion in exon 2, located at base 476 of the gene (base 391 of the coding region), changed the phenylalanine TTC codon at amino acid position 131 to a valine GTC codon (Fig. 4A). This F131V missense mutation resides within the Rossmann-like domain (see below) in a region showing a high degree of conservation between cyanobacterial, algal and plant PDS proteins (Fig. 4B).

Use of the Mutant pds1-nfr1d Allele as a Selectable Marker for Transformation
To show that the F131V mutation conferred norflurazon resistance, the mutant pds1-nfr1d gene was cloned in an E. coli plasmid vector and used to transform C. reinhardtii cw92, which is a norflurazon sensitive cell-wall deficient strain [28]. The 59 region of the pds1-nfr1d gene was isolated as a 3.2 kbp PCR product (Fig. 4A, primers 1017 and 1114) and joined to the 39 region of the WT gene amplified as a 3.0 kbp PCR product (Fig. 4A, primers 1113 and 1065). The two cloned PCR products were combined at a ClaI site inserted in the third intron and ligated to pBluescript to assemble the 9121 bp pNFR1 plasmid (Fig. 5A). The ClaI site was created by inserting an ATC before the GAT trinucleotide at base 130 of the third 612 bp intron. In order to include 59 and 39 expression elements, the coding region was flanked by 1468 bp of upstream and 1390 bp of downstream sequences. As a control construct, the equivalent 59 region was cloned from the PDS1 WT gene and fused to the 39 end of the PDS1 WT gene to produce the pWTPDS1 plasmid. The only difference between the pNFR1 and pWTPDS1 vectors was the thymine to guanine transversion in exon 2 of the pds1-nfr1d gene (Fig. 4A) in the pNFR1 vector, which was confirmed by sequencing both plasmids. pNFR1 and pWTPDS1 were used to transform the cell wall deficient and norflurazon sensitive strain cw92 using the glass-bead vortexing method [29]. Herbicide resistant colonies were isolated by direct selection on 5 mM norflurazon medium, followed by several rounds of continued selection on 5 mM norflurazon media. We found selection on 2.5 mM norflurazon gave rise to a high Accumulation of the PDS protein in transformants was compared to non-transgenic strains cw92 and cc621 by protein blot analysis (Fig. 5D). A PDS-specific antibody showed variable accumulation of a PDS band in the four transgenic lines. Transgenic strains A10 and 2A1 accumulated over two-fold higher levels of PDS protein (Fig. 5D, lanes 3, 5 and 6) than the untransformed cw92 and cc621 strains (Fig. 5D, lanes 7-8). Transgenic strains A2 and 1A1 (Fig. 5D, lanes 2 and 4) accumulated similar levels of PDS compared to the non-transgenic cw92 and cc621 strains. A,50 kDa band present in all lanes serves as a loading control and indicates similar amounts of protein were loaded per lane. Differences in PDS levels between transformants can be explained by position effects resulting from random integration of foreign genes into the C. reinhardtii nuclear genome by illegitimate recombination.

Transgenic pNFR1 Strains were Highly Resistant to Norflurazon
Growth of pNFR1 transformants and the recipient nontransgenic cw92 strain were compared in TAP media containing increasing concentrations of norflurazon (Fig. 6A). Growth of cw92 was severely reduced by 1 mM norflurazon (67% reduction) and prevented by 5 mM norflurazon after seven days growth. The growth of all transformants showed marked resistance to norflurazon. The highest levels of resistance were shown by strains 2A1 and A10 (Fig. 6A), which accumulated more PDS protein (Fig. 5D). All transgenic strains grew in media containing 150 mM norflurazon. When grown in TAP media without herbicide, the transformants grew at similar rates to WT (Fig. 6B). This indicates that the expression of pds1-nfr1d in transgenic strains does not have a marked impact on cell growth. Growth of transformant A2 plateaued earlier than the other strains and reached a final density that was 70% of the WT density at day six. The reduced cell density observed for one transformant (transformant A2) relative to cw92 is likely to reflect factors that are not directly related to the presence of transgenic DNA containing pds1-nfr1d, such as the transgene integration site. In the presence of 150 mM norflurazon, transformant 2A1 grew to the highest densities (Fig. 6C).

F131 Maps to a Rossmann-like Domain and Clusters with other Norflurazon-resistant Mutations in 3D Protein Models
A linear scheme of the 564 amino acid precursor PDS protein from C. reinhardtii containing the chloroplast transit peptide is shown in Figure 7A. F131V and previously described resistance mutations in homologous PDS proteins identify amino acids that are likely to affect norflurazon binding. The high conservation amongst PDS proteins allowed us to map substitutions conferring norflurazon resistance in homologous proteins (Fig. 8) onto the C. reinhardtii PDS sequence (Fig. 7A). The F131V mutation reported here lies in the N-terminal quarter, whereas previously reported amino acids substitutions conferring norflurazon-resistant mutations map closer to the middle (R268, L388) or C-terminus (V472, L505, L517) of the 564 amino acid protein (Fig. 7A). The F131V is the first herbicide resistance mutation localising to the dinucleotide binding Rossmann-like domain [30,31,32] spanning amino acids Y91 to L144 (Fig. 7A). This organisation of Rossmann-like domain amino acids contributing to a FAD-binding domain is characteristic of pyridine nucleotide-disulphide oxidoreductases [31]. A previously described substitution (E143K) in this domain reduces function and is suppressed by intragenic suppressor mutation K90M [23].
The crystal structure of a plant PDS has not been determined. Whilst the structure of a CRTI-type PDS enzyme present in bacteria and fungi is available [33], it is distantly related to the plant PDS enzyme and may have an independent evolutionary origin [34]. In order to map the F131V mutation and the positions of other norflurazon resistance mutations onto the structure of PDS, 3D models of the C. reinhardtii protein were constructed using four programs: 3DLigandSite [35] which uses the Phyre prediction system [36], MODELLER [37,38] via the BioSerf service [39], SwissModel [40] (Fig. 7B), and 3D-Jigsaw [41]. The predicted PDS structures are shown in Figure 7B and Figure S1. Each program used the crystal structures of different proteins to predict the folded structure of PDS. Phyre used human monoamine oxidase type B [42], MODELLER used an oxidoreductase from Methanosarcina mazei [43] and SwissModel used putrescine oxidase from Arthrobacter aurescens [44]. C.reinhardtii PDS shares: 15% identity (32% similarity) with human monoamine oxidase chain B over aligned PDS amino acids 76-537; 16% identity (30% similarity) with M. mazei oxidoreductase for PDS amino acids 76-526; 18% identity (32% similarity) with A. aurescens putrescine oxidase for PDS amino acids 76-553.
Norflurazon inhibits PDS activity by competing with the enzyme's cofactors rather than the phytoene substrate [46]. The purified plant enzyme is a flavoprotein containing FAD [47]. In bacterial CRT-I type PDS, FAD appears to be the only co-factor required for phytoene desaturation [33]. Following the desaturation reaction the reduced FADH 2 produced is re-oxidised to FAD by plastoquinone either directly or indirectly [48]. The 3DLi-gandSite program [35] predicts the location of the FAD binding domain [32,49,50] in PDS. The amino acids contributing to the predicted FAD binding domain in the C. reinhardtii PDS enzyme are shown in Fig. 7C. These include amino acids that are also part of the Rossmann-like domain (Fig. 7C, shaded yellow). Of the amino acids whose substitution confers norflurazon resistance, V472 (pink) is part of the predicted FAD-binding domain, whilst F131 (red) and L505 (green) lie adjacent to FAD-binding domain residues F130 and A506 (Fig. 7C). A magnified view of the relative positions of F131, V472 and L505 and FAD in the 3D-ligand model of PDS is shown in Figure 7D. The close proximity of F131, V472 and L505 to the predicted FAD binding site is consistent with norflurazon inhibiting the function of this co-factor.

Cross-tolerance Conferred by the pds1-nfr1d Gene to other Bleaching Herbicides
The highest levels of norflurazon resistance were exhibited by pNFR1 transformants 2A1 and A10. Four different bleaching

Discussion
We used forward genetics to isolate a new F131V substitution in the C. reinhardtii PDS protein that confers a high level of resistance to norflurazon. Whilst growth of WT cells was prevented by 5 mM norflurazon, cells containing the mutant pds1-nfr1d gene were able to grow in media containing 150 mM norflurazon. The Ec50 values to norflurazon were 0.7 mM and around 100 mM, for WT and mutant NFR1 cells, respectively. The norflurazon-resistance mutation conferred cross-tolerance to the bleaching herbicides fluridone, flurtamone, and diflufenican but enhanced sensitivity to beflubutamid, relative to WT cells. The mutant pds1-nfr1d allele is dominant or semi-dominant to the WT gene allowing the cloned pds1-nfr1d gene to be used as a marker for nuclear transformation with direct norflurazon selection for resistant transgenic cells.
The F131V substitution is the first norflurazon resistance mutation to be mapped to the dinucleotide binding Rossmann-like domain of PDS. F131V and three other resistance mutations, which correspond to substitutions at R268, V472 or L505 in the C. reinhardtii protein (Fig. 8), are located at distant sites of the primary sequence but cluster together in the predicted tertiary structure. Moreover, three of these amino acids lie adjacent to (F131, L505) or are part of (V472) the predicted FAD binding domain (Fig. 7C). These results, based on locating amino acid substitutions conferring norflurazon resistance on 3D-models of PDS, predict the herbicide's target region on the protein and suggest the herbicide's mode of action involves inhibition of FAD function. Substitutions of amino acids L388 and L517 also confer norflurazon resistance but these residues do not map close to the F131-R268-V472-L505 cluster of resistance mutations on 3D models. To resolve whether L388 and L517 interact with norflurazon directly or affect binding of norflurazon indirectly by affecting the folding of the protein will become clearer if a crystal structure becomes available.
Use of the pds1-nfr1d allele as a positive selection marker for nuclear transformation of C. reinhardtii avoids the use of antibiotic resistance genes [51,52,53,54,55], which may raise regulatory concerns for commercial applications of transgenic algae. The transformation procedure used the mutant pds1-nfr1d allele with its own promoter and terminator and 5 mM norflurazon, which was substantially higher than the concentration required to kill the majority of non-transformed cells (Ec50 of 0.7 mM). In a recent study, 0.5 mM norflurazon was used to select C. reinhardtii transformants expressing a cDNA encoding a mutant PDS protein in which the L505F substitution was introduced by site-directed mutagenesis [24]. The RbcS2 promoter [52] was used to express the L505F mutant gene resulting in elevated PDS transcripts [24]. Here we found that 5 mM norflurazon was required for effective selection of transformants in the light regimes used. Use of 2.5 mM norflurazon on selection plates led to a high background of nontransformed colonies. Previous herbicide resistant marker genes [24,56] and a mutant CRY1 gene providing resistance to the translational inhibitors emetine and cryptopleurine [57] have used 59 and 39 regulatory elements from the RbcS2 gene to allow the isolation of nuclear C. reinhardtii transformants [24,56,57]. The use of herbicide resistance genes with native regulatory elements [58] including the mutant pds1-nfr1d gene (this work) as marker genes overcomes the possibility of RbcS gene silencing and titration of transcription or translation factors needed to express the native RbcS genes [59] in C. reinhardtii.
Transformants containing the pds1-nfr1d gene were 4.8 fold less resistant to beflubutamid than WT cells. By comparison, transformation of the WT Hydrilla PDS protein into Arabidopsis increased seedling sensitivity to diflufenican [18]. Arabidopsis seedlings expressing a Hydrilla norflurazon resistant PDS protein with a R304C (corresponds to position R268 in the C. reinhardtii protein) mutation, were more sensitive to both diflufenican and beflubutamid [18]. These results show that changes in the region identified by the cluster of amino acid substitutions conferring norflurazon resistance (Figs. 7B, 7D) are critical in determining the action of herbicides and provide a step towards the rational design of new herbicides targeting this protein. The enhanced susceptibility of pds1-nfr1d transformants to beflubutamid allows negative selection. This enables pds1-nfr1d, which is a native algal gene, to be used for either positive or negative selection, which is conditional on the selection agent used. Norflurazon enables positive selection whereas beflubutamid allows negative selection. Because the transformants contain both the WT and norflurazonresistant PDS enzymes, this suggests that beflubutamid sensitivity is dominant or semi-dominant to resistance conferred by the WT enzyme. Raised PDS levels in the A10 and 2A1 transformants relative to WT cells was consistent with higher levels of the beflubutamid-sensitive mutant PDS enzyme in transgenic cells.
Dual and conditional positive/negative selection markers are powerful tools for manipulating genomes and controlling the growth of transgenic cells [60,61,62]. The high norflurazon resistance levels exhibited by pds1-nfr1d transformants suggests the coding region flanked by appropriate regulatory elements may serve as a marker for transformation of a wide range of algae and plants susceptible to norflurazon and related bleaching herbicides. The PDS protein is localised to chloroplasts, which raises the possibility of using a modified coding sequence with chloroplast regulatory elements as a new marker gene for the chloroplast transformation toolbox [63]. Negative selection can be used to restrict the spread of transgenic cells, influence recombination events such as marker excision, and in mutant screens. The pds1-nfr1d allele confers sensitivity to beflutamid and is a candidate for a new negatively selectable marker for algal genetics.

Conclusions
A phenylalanine to valine substitution at position 131 in the dinucleotide binding Rossmann-like domain of phytoene desaturase increased the 5 mM norflurazon concentration needed to restrict growth of WT C. reinhardtii cells to over 150 mM in mutant cells. The F131V substitution clusters with two other substitutions conferring norflurazon resistance close to the predicted FADbinding site in 3D-protein models.

C. reinhardtii Strains, Media, Mutagenesis and Genetic Crosses
All C. reinhardtii strains were from the Chlamydomonas Center (University of Minnesota). Strains were maintained at 20uC to 25uC on TAP medium [1]. Transformants were propagated on TAP medium with 5 mM norflurazon (Sigma-Aldrich, Poole, UK). Large 25 ml-200 ml liquid cultures were grown at 25uC in a New Brunswick Scientific Innova 4340 illuminated orbital shaker at 120 rpm. Arginine (Sigma-Aldrich) was used at a final concentration of 50 mg L 21 and nicotinic acid (Sigma-Aldrich) at 25 mg L 21 .
For mutagenesis 5 ml of 1610 7 cells ml 21 of C. reinhardtii strain nic7 arg7 y1 mt+ were irradiated for 20 seconds with a 254 nm XX-15 BlackRay germicidal UV lamp (UVP, Upland, California) at 2.42 mW cm 22 in 9 cm petri dishes with gentle rocking. Irradiated cells were kept in the dark for 12 h, and then concentrated by centrifugation at 6006g at room temperature. Three norflurazon resistant cell lines (NFR1-3) were selected on media containing 33 mM norflurazon at 25uC with a light intensity of 60 mmol m 22 s 21 in a 12 h light cycle. Only NFR1 survived several rounds of serial propagation on 33 mM norflurazon selective plates. Two additional resistant mutants were isolated using UV irradiation of 2.86 and 3.62 mW cm 22 for 20 seconds but these grew very slowly and could not be propagated.

Nucleic Acid Methods
Total DNA from C. reinhardtii was prepared according to Day et al 1988 [66], except an additional DNA precipitation was included by adding 0.6 volumes of 20% PEG 8000, 1.5 M NaCl (Sigma-Aldrich) to the total DNA solution. RNA was purified using the RNeasy plant mini kit (Qiagen) according to the manufacturer's instructions. Genomic DNA for PCR screening was isolated according to Cao et al. (2009) [67]. Hybridisation probes were comprised of a Gulliver [26] internal fragment and a 1.5 kbp PDS1 cDNA from C. reinhardtii strain CC-503 cw92 mt+ prepared using nested RT-PCR with primers 1297 and 1298 followed by primers 1328 and 1329. Probes were labelled with [a-32 P] dCTP (PerkinElmerNEN, Seer Green, UK) using the High Prime labelling kit (Roche, Mannheim, Germany). Digests of 2 mg of total DNA digests were fractionated on 0.7% W/V agarose gels for DNA blot analysis [66]. Membranes were washed at 65uC in 0.16SSC buffer and 0.1% w/v SDS. Table S1 lists the oligonucleotides used. PCR screening of transformants was performed using primer pairs 1322+1323, and 1324+1325, which amplified the regions spanning the PDS-vector junctions in pNFR1 and pWTPDS1 transformants. PCR reactions with actin gene primers 1127 and 1351 were used as positive controls to verify the quality of genomic DNA. PCR screening was performed using BioMix PCR master mix (Bioline, London). BigDye terminator v3.1 (Applied Biosystems) was used for cycle sequencing and the ladders fractionated using a 48 capillary ABI 3730 DNA Analyser.
Isolation and Sequence Analysis of the pds1-nfr1d Gene PCR products from NFR1 genomic DNA using Vent DNA polymerase (New England Biolabs, Massachusetts, USA) were sequenced directly or following cloning into pGEM-T Easy

Construction of pNFR1 and pWTPDS1
The pWTPDS1 plasmid was constructed by joining 59 (3235 bp; primers 1017+1114) and 39 (2965 bp; 1113+1065) PCR products (cloned in pGEM-T Easy) amplified from CC503 cw92 mt+ total DNA via a Cla I site located in primers 1113 and 1114. The WTPDS1 sequences in pGEM-T Easy were excised with Not I and Cla I (3235 bp sequence) and Cla I and EcoR1 (2965 bp sequence) and cloned into pBluescript KS- [68]. To construct pNFR1, the 3.2 kbp 59 WT PDS1 Not I -Cla I fragment was replaced with the equivalent 59 pds1-nfr1d fragment. The complete inserts in both plasmids were sequenced to exclude PCR errors.
Nuclear Transformation of C. reinhardtii Chlamydomonas cell wall deficient strain CC-503 cw92 mt+ was transformed according to Kindle, 1990 [29]. 25 ml of cell culture (,5610 6 cells/ml) was sedimented at 5006g. The cell pellet was resuspended in 2.5 ml of TAP media and 300 ml mixed with 0.3 g of sterile 0.4 mm glass beads (BDH, Poole, UK), 2 mg circular (mainly supercoiled) plasmid DNA and 100 ml of 20% (W/V) polyethylene glycol (M W 8000). The mixture was vortexmixed (Rotamixer, Hook and Tucker Instruments Ltd., Croydon, UK) at maximum setting for 30 seconds, 700 ml of TAP media was added, and the cells spread onto three 9 cm selection plates containing solid (1% W/V agar) TAP media with 5 mM norflurazon. Cells were allowed to recover for 36 h in dim light (10 mmol m 2 2 sec 21 ) and then transferred to high-light (250 mmol m 22 sec 21 ). Colonies became visible after 21 days and were transferred to fresh selection media.

The Effects of Herbicides on Cell Cultures
Growth curves for the NFR1 strain were prepared by inoculating cells to a concentration of 1610 5 cells ml 21 , growing them for 7 days in 50 ml of TAP media in a 12 h light cycle (60 mmol m 22 s 21 ) at 25uC with shaking (120 rpm) with increasing concentrations of norflurazon and measuring optical density (OD) at 750 nm. Transformant herbicide growth assays were carried out in continuous light (260 mmol m 22 s 21 ) at 25uC without shaking and culture ODs measured daily in a microplate reader (Biotek Synergy, Potton, UK). Transformant cells were inoculated in 1 ml of TAP media to an OD750 nm of 0.05 containing increasing concentrations of norflurazon in 24-well plates and grown for 7 days. Resistance curves for other herbicides were obtained by growing cells in 200 ml TAP media with increasing concentrations of herbicides in 96-well plates. Fluridone was used at 5, 10, 50 and 150 nM concentrations. Befublutamid and flurtamone were utilised at 25, 50, 75 and 100 nM concentrations. Diflufenican was used at 25, 50, 100 and 300 nM concentrations. All experiments were performed in triplicate. Herbicides were from Sigma-Aldrich.
For estimation of coloured carotenoids, cultures were grown for 7 days in TAP media in a 12 h day (60 mmol m 22 s 21 ) with shaking (120 rpm) at a starting density of 1610 5 cells ml 21 . Carotenoids in sedimented cells (1.1610 8 cells) were extracted twice with 4 ml of 80% acetone for 20 minutes. Carotenoids were partitioned into 8 ml of 50% diethyl ether with the addition of 2 M NaCl to promote phase separation. The organic phase was washed twice with 5 ml of 4 M NaCl and then dried under nitrogen. The dried carotenoids were dissolved in diethyl ether and coloured carotenoids were estimated from the absorbance at 470 nm [1].

Supporting Information
Figure S1 Space-filling model of C.reinhardtii phytoene desaturase predicted by 3D Jigsaw [41,45]. Amino acids F131, R268, L388, V472, L505 and L517 whose substitution are associated with norflurazon resistance are shown (see Figs. 7 and 8). Also shown are E143 and K90. Mutation of E143 impairs function and this is suppressed by a mutation at K90 [23]. The F131, R268, V472 and L505 cluster of amino acids associated with norflurazon resistance is magnified on the right.
Table S1 Sequences of oligonucleotides used as primers for PCR and sequencing reactions.