Broad 4-hydroxyphenylpyruvate dioxygenase inhibitor herbicide tolerance in soybean with an optimized enzyme and expression cassette.

With an optimized expression cassette consisting of the soybean (Glycine max) native promoter modified for enhanced expression driving a chimeric gene coding for the soybean native amino-terminal 86 amino acids fused to an insensitive shuffled variant of maize (Zea mays) 4-hydroxyphenylpyruvate dioxygenase (HPPD), we achieved field tolerance in transgenic soybean plants to the HPPD-inhibiting herbicides mesotrione, isoxaflutole, and tembotrione. Directed evolution of maize HPPD was accomplished by progressively incorporating amino acids from naturally occurring diversity and novel substitutions identified by saturation mutagenesis, combined at random through shuffling. Localization of heterologously expressed HPPD mimicked that of the native enzyme, which was shown to be dually targeted to chloroplasts and the cytosol. Analysis of the native soybean HPPD gene revealed two transcription start sites, leading to transcripts encoding two HPPD polypeptides. The N-terminal region of the longer encoded peptide directs proteins to the chloroplast, while the short form remains in the cytosol. In contrast, maize HPPD was found almost exclusively in chloroplasts. Evolved HPPD enzymes showed insensitivity to five inhibitor herbicides. In 2013 field trials, transgenic soybean events made with optimized promoter and HPPD variant expression cassettes were tested with three herbicides and showed tolerance to four times the labeled rates of mesotrione and isoxaflutole and two times the labeled rates of tembotrione.

For nearly 2 decades, maize (Zea mays), soybean (Glycine max), and cotton (Gossypium hirsutum) farmers have relied on glyphosate and glyphosate-resistant crops for weed control. While favored for its efficacy, economy, and convenience, the onset of glyphosate-resistant weeds, now numbering 29 species (Heap, 2014), signals a need for new herbicide and trait systems as effective as glyphosate (Green and Castle, 2010;Duke, 2012). Unfortunately, prospects for novel herbicide chemistries are not encouraging (Duke, 2012).
Our interest is in the development of crops tolerant to the 4-hydroxyphenylpyruvate dioxygenase (HPPD; EC 1.13.11.27) inhibitors that disrupt pigment production in plants (Fig. 1). Disruption of the conversion of Tyr, through 4-hydroxyphenylpyruvate (HPP), to homogentisate leads to failure to produce plastoquinone essential for photosynthetic electron transfer, leading to the loss of carotenoid pigments and photosynthesis failure (Moran, 2014). Plastoquinone is an essential cofactor for phytoene desaturase in the carotenoid biosynthesis pathway (Ruiz-Sola and Rodriguez-Conception, 2012). Plants lacking carotenoids cannot protect themselves from the radicals generated by the light activation of chlorophyll, causing bleaching, necrosis, and death (Fig. 1C). Further evidence shows that in the absence of carotenoids, the photosynthetic apparatus is not properly assembled (Sozer et al., 2010), explaining why only the new leaves in the plant in Figure 1C are white. Although plastoquinone is also requisite for tocopherol biosynthesis, tocopherol deficiency is not associated with an herbicidal phenotype (Maeda and DellaPenna, 2007). HPPD inhibitors are most effective on broad-leaf weeds but control some grasses as well. Currently, HPPD herbicides are selective for use in maize, while soybeans and other dicot crop species are sensitive.
A broad-spectrum HPPD tolerance trait in soybeans, used in combination with glyphosate tolerance and other traits or selective herbicides, will prolong the positive impact of the glyphosate systems and slow the appearance of resistant weeds. Weed resistance to HPPD-inhibiting herbicides was unknown until two cases involving Amaranthus spp. were reported recently (Heap, 2014). Two transgenic soybean crop products, one for tolerance to isoxaflutole (APHIS, 2009) and the other to mesotrione and isoxaflutole (APHIS, 2012), are in the U.S. Department of Agriculture deregulation approval process. The former uses a Pseudomonas fluorescens HPPD with a single amino acid change (Matringe et al., 2005), while the latter uses HPPD from oat (Avena sativa) with a single amino acid change (Hawkes et al., 2010). Our goal was to develop soybean lines with tolerance to a broad spectrum of HPPD inhibitors.
To develop a robust and stable HPPD tolerance trait in soybean, we sought first to understand the basic biology of the process, including expression of the native gene and localization of native and transgenic proteins, and to evaluate the efficacy of low to moderate expression of a desensitized protein. Data regarding the subcellular location of HPPD are ambiguous, perhaps resulting from species diversity. Early work with organelle fractions attributed most HPPD activity to the chloroplast in spinach (Spinacia oleracea; Fiedler et al., 1982) or Lemna gibba (Löffelhardt and Kindl, 1979). Organelle targeting can be conjectured from the observation that the N-terminal sequence of mature HPPD isolated from maize leaf begins at either Ala-17 (Fritze et al., 2004) or Ala-23 (Yang et al., 2004) with respect to the translated full-length gene. Two HPPD genes identified from EST libraries prepared from cotton tissue were 98.6% identical, each with a 23-amino acid sequence deemed likely to function in chloroplast targeting by analysis with the ChloroP prediction program (Moshiri et al., 2007). In the same study, tomato (Solanum lycopersicum) but not Brassica spp. HPPD was predicted to have a chloroplast-targeting peptide (CTP). Subcellular fractionation supported a cytosolic location of the carrot (Daucus carota) cell enzyme (Garcia et al., 1997). The same authors determined that after PSORT analysis failed to identify a targeting signal within the Arabidopsis (Arabidopsis thaliana) HPPD amino acid sequence, native Arabidopsis HPPD heterologously expressed in tobacco (Nicotiana benthamiana) was located exclusively in the cytosol (Garcia et al., 1999). In this study, fluorescent protein markers and microscopy allowed us to clearly visualize the subcellular localization of the native protein or that expressed from HPPD promoters, revealing the localization of HPPD exclusively in the chloroplasts of maize and in both chloroplasts and the cytosol of soybean.
Immunolocalization electron microscopy of maize tissue sections was carried out to detect native HPPD localization. Gold labeling was observed mainly in bundle sheath chloroplasts (Supplemental Fig. S1). Particles were also found in mesophyll chloroplasts, but in no other structures. To functionally corroborate the localization of maize HPPD, a binary plant transformation vector was constructed in which the portion of the maize HPPD gene coding for the N-terminal 50 amino acids was fused to the gene coding for Discosoma spp. RED FLUORESCENT PROTEIN2 (DsRed2), all under the control of the maize Rubisco activase promoter. A positive control vector was identical except that the DsRed2 insert was fused to the CTP of maize Rubisco activase, while a negative control was DsRed2 with no targeting sequence. The plasmids were transformed into Agrobacterium tumefaciens strain AGL-1, and agroinfiltration was used to introduce the constructs into plant cells for transient expression. Leaves of 3-week-old maize seedlings were infiltrated with the A. tumefaciens and examined by fluorescence microscopy of hand sections 2 d later. When DsRed2 was fused to Rubisco activase CTP, red fluorescence was seen in discrete packets in a pattern resembling perinuclear chloroplasts, as expected ( Fig. 2A). A similar pattern was seen when DsRed2 was fused to the N-terminal 50 amino acids of maize HPPD (Fig. 2B). Without targeting, fluorescence was diffuse, with some concentration in the nucleus (Fig. 2C). In another experiment confirming these results, maize leaf tissue was cobombarded with DNA from both the DsRed2-containing test plasmids and a plasmid encoding untargeted cycle 3 GFP (C3GFP). Transformation of guard cells with vectors encoding either Rubisco activase CTP-DsRed2 or the N-terminal 50 amino acids of maize HPPD fused to DsRed2 clearly resulted in plastid targeting of the DsRed2 reporter, whereas untargeted C3GFP showed no overlap with the DsRed2 signal (Supplemental Fig. S2).
To determine the length of the functional CTP for maize HPPD, vectors were constructed in which the portion of the maize HPPD gene coding for the N-terminal 0, 10, 20, 30, 40, or 50 amino acids was fused to the gene coding for DsRed2 and evaluated with transient expression following the agroinfiltration of maize leaves. Microscopy revealed that 50 amino acids of the maize HPPD N terminus effectively targeted DsRed2 to plastids ( Fig. 2F), but 40 amino acids or fewer failed to do so, with DsRed2 fluorescence visible only in the cytoplasm (Fig. 2,D and E). This result indicates that more than 40 amino acids of the N terminus are required for chloroplast localization and that 50 amino acids are sufficient for targeting. To investigate if the maize HPPD protein is unique, several monocot sequences were compared. The rice (Oryza sativa) HPPD N-terminal sequence effectively targeted DsRed2 to maize plastids (data not shown). Furthermore, 50 amino acids of the maize HPPD N terminus delivered DsRed2 to sorghum (Sorghum bicolor) chloroplasts. Figure 2I provides an N-terminal alignment of monocot HPPD proteins having identities of 59% to 85%. A consensus monocot HPPD CTP sequence (Fig. 2I) targeted the DsRed2 protein to maize chloroplasts. These data strongly suggest that monocot HPPD proteins share a homologous chloroplast-targeting motif and function.
The function of the monocot HPPD sequence in dicots was explored by transient expression in bush bean (Phaseolus vulgaris), tobacco, and soybean cells. The sequence encoding the first 50 amino acids of the maize HPPD protein was fused to a gene encoding Aequorea coerulescens GFP1 (AcGFP1) and inserted into a binary expression vector under the control of the Arabidopsis UBIQUITIN10 promoter. A positive control vector was identical except that the AcGFP1 coding region was fused to the 6H1 synthetic CTP (Lassner and Wilkinson, 2008), while a negative control was AcGFP1 with no targeting sequence. The first 50 amino acids of maize HPPD were sufficient to drive the chloroplast import of AcGFP1 in epidermal cells of bush bean, although some green fluorescence remained in the cytoplasm (data not shown). In tobacco, AcGFP1 remained in the cytoplasm, with none apparent in the chloroplasts (data not shown). Results by bombardment in soybean showed AcGFP1 in both plastids and cytoplasm (Fig. 2G). This shows that the maize HPPD CTP is recognized in some dicot plant species but may be inefficiently processed. Stably transformed soybean plants expressing the 50-amino acid N-terminal CTP fused to AcGFP1 showed strong fluorescence in the chloroplasts (Fig. 2H), suggesting that the inefficient translocation observed in transient expression conditions may be due to the high template concentration.

Bioinformatic and Functional Localization Analyses of Soybean HPPD
Soybean is a polyploid derived from two evolutionary genome duplications (Schmutz et al., 2010) and as such appears to have multiple paralog HPPD genes. As some of the annotated genes appear incomplete or have rare ESTs, we focused here on glyma14g03410.1 on chromosome 14. This soybean HPPD protein has been annotated previously as a 449-amino acid protein with N-terminal sequence MPIPMCNEIQ (Cahoon and Coughlan, 2007; sequence 36) and as a 443-amino acid protein with N-terminal sequence MCNEIQAQAQ (GenBank accession no. ABQ96868). Our analysis of genomic and EST data revealed that an in-frame N-terminal extension of the previously annotated coding region exists, adding 41 amino acids to produce a 490-amino acid full-length HPPD protein, as shown in Figure 3A.
59 RACE revealed two major transcripts for the native soybean HPPD gene. Cloning and sequencing of these PCR products revealed a shorter transcript starting at position +1 able to support translation of the 449-amino acid protein. The longer transcript began at position 2237 and comprised the full 490-amino acid open reading frame (ORF; Fig. 3A). Linked in vitro transcription and translation indicated that both mRNAs were translated and made the predicted sized proteins (truncated for convenience in Fig. 3B). The +1 transcript was translated to produce a single protein, designated the short protein.
The 2237 transcript yielded two proteins, with the lower M r band at the same molecular weight as the single protein produced from the +1 transcript, suggesting that translation initiated at both positions on the long transcript. The protein produced from the 2237 transcript is designated the long protein.
Bioinformatic evaluation of the shorter soybean HPPD sequence did not predict a chloroplast or other targeting sequence. However, a chloroplast-targeting function is predicted for the N-terminally extended form by ProtComp 9.0 (http://linux1.softberry.com), WoLF PSORT (Horton et al., 2007), and PCLR (Schein et al., 2001). TargetP (Emanuelsson et al., 2000) suggests chloroplast localization but gives a higher score to other, while MultiLoc2 (Blum et al., 2009) predicts a cytoplasmic localization for both the first 42 and the first 86 amino acids of the long protein. Functional analysis was required to clarify the predictions. Transient (Fig. 3C) or stable (Fig. 3, D and E) expression experiments indicated that the long HPPD protein is imported to chloroplasts, while the short protein remains in the cytosol. Plant expression cassettes were constructed by fusing portions of the N terminus of the soybean HPPD to either DsRed2 or AcGFP1. The first fusion contained amino acid residues 1 to 42 of the long protein, while the second fusion contained residues 1 to 45 of the short HPPD protein (corresponding to residues 42-86 of the long protein). The third fusion contained the full 1 to 86 amino acids of the long HPPD protein. These cassettes, under the control of the Arabidopsis UBIQUITIN10 promoter, were used to transfect soybean seedling leaves. As shown in Figure 3, C to E, fluorescence was clearly visible in the chloroplasts of infected cells only when the marker gene was fused to amino acid residues 1 to 86. When the fusion was made with the shorter sequences of both proteins, fluorescence was visible only in the cytoplasm. The pattern of dual localization seen with the reporter constructs was confirmed by immunolocalization electron microscopy (Supplemental Fig. S3). Thus, the N-terminal 42 amino acids of the long protein are predicted to be a CTP, but to be functional, a longer sequence is required.
Although the N-terminal sequences of maize and soybean HPPD differ considerably (Fig. 3F), they are both able to target proteins in the heterologous species. Transient expression indicated that the long (1-86) HPPD protein N terminus targeted the marker protein to maize cell chloroplasts, while the protein variant with the shorter N termini delivered the protein to the cytosol (data not shown). Thus, the dicot dual-targeting region of soybean HPPD is able to function in monocot cells.
Because the reported HPPD sequences were misannotated for soybean, we examined available sequence data for Arabidopsis and another legume, Medicago truncatula. The Arabidopsis chromosome 1 HPPD gene NC_003070.9 (GenBank) encodes a 418-amino acid ORF with no upstream ORFs. The lack of N-terminal extension is punctuated by two in-frame stop codons 15 nucleotides upstream of the ATG start. On the other hand, a search of the M. truncatula genome returned a predicted HPPD gene located on chromosome 5 (NC_016411.1; GenBank), wherein the annotated predicted protein of 437 amino acids is preceded by a long ORF that could initiate at Met residues, adding 74 or 115 amino acids. The 74-amino acid N-terminal extension shows a very strong chloroplast localization score with PCLR (Schein et al., 2001), while the 115-amino acid extension did not. Functional analysis of the M. truncatula gene was not attempted. Dicot species appear to be either chloroplast targeted or not in a species-specific pattern that is still not clear. However, legumes may be unique in using the dual targeting approach.

Directed Evolution of the Maize HPPD Protein for Increased Insensitivity to Herbicide Inhibitors
All herbicidal inhibitors of HPPD are competitive inhibitors (Secor, 1994) and form a tight complex with the enzyme by the dual mechanisms of coordination to the active-site iron atom through a pair of keto oxygens and a p stack of the aromatic ring of the inhibitor between a pair of active-site Phe residues (Neidig et al., 2005). As a result, conventional 50% inhibition constant determinations are not able to distinguish differences in Figure 3. Genomic analysis of soybean HPPD. A, Soybean HPPD protein sequence with in-frame N-terminal extension (underlined). B, Protein polyacrylamide gel of in vitro transcription-translation products of the two (39 truncated) HPPD transcripts. Predicted protein mass from 2237 mRNA, 30.6 kD; +1 mRNA, 26 kD. C, Transient expression of the N-terminal fragment 1 to 42 fused to DsRed2 in soybean. D, Stable expression of the N-terminal fragment 42 to 86 fused to AcGFP1 in soybean. E, Stable expression of the N-terminal fragment 1 to 86 fused to AcGFP1 in soybean. Bars in C to E = 10 mm. F, Alignment of maize and soybean HPPD proteins. Asterisks indicate identity; the colon shows the fusion point of soybean to maize-insensitive HPPD proteins.
binding affinity among various forms of HPPD and the inhibitor. All values will similarly approximate 50% of the enzyme concentration. To devise a parameter for detecting changes in inhibitor-binding affinity (K D ), we utilized the relationship between K D and the rates of binding and release of inhibitor (I) to and from the enzyme (E): At equilibrium, the rates of binding and release are equal. Thus: Written as a dissociation, the equation can be rearranged to: Higher K D (reduced affinity or increased insensitivity) can be attained with a numerically smaller k ON , a larger k OFF , or both. To detect and quantify changes in k ON and k OFF , observations of the time course of an HPPD reaction as inhibitor binds to and inactivates the enzyme (k ON ) or is released from a preformed enzyme-inhibitor complex (k OFF ) were expressed as ON and OFF rate ratios, as described in Supplemental Figure S4. The ON and OFF rate ratios were multiplied, creating an insensitivity parameter that is a surrogate for K D . K D is expressed as k OFF /k ON , but because the ON rate ratio is an inverse parameter (the higher the value of the ON rate ratio, the lower the value of the represented phenomenon, k ON ), it is appropriate to multiply ON and OFF rate ratios. Another parameter, which quantifies the combined catalytic and insensitivity properties of the enzyme, is k cat /K m 3 ON rate ratio 3 OFF rate ratio, which is termed enzyme fitness. This parameter is intended to have a meaning similar to the commonly used expression k cat /K m 3 K i , where k cat is the rate constant of the catalyzed reaction and K i is the dissociation constant of the enzyme-inhibitor complex.
Directed evolution is an iterative process whereby beneficial diversity is discovered and iteratively recombined to evolve variants that achieve a desired protein engineering goal. Our goal was to create an HPPD enzyme that retained kinetic activity yet was insensitive to most or all HPPD-inhibiting herbicides. Characterization of the maize and soybean HPPD proteins showed that the former was 40-fold less sensitive to mesotrione than the latter (Table I). Therefore, we chose to improve the maize enzyme for use in soybean plants. Final evaluation of shuffled HPPD enzymes was carried out with multiple HPPD herbicides to select the most robust variant.
In the first six rounds, shuffled gene variant libraries were made based on the maize HPPD protein template using techniques including family shuffling, single-gene shuffling, back-crossing, and semisynthetic and synthetic shuffling (Zhang et al., 1997;Crameri et al., 1998;Ness et al., 2002). The amino acid diversity for those libraries originated from phylogenetic sequence diversity, random mutagenesis, site-directed mutagenesis, and structural features based on crystal structures of proteins in the Protein Data Bank (PDB; www.pdb.org/pdb/home/ home.do). Typically, 4,000 to 5,000 variants per library were screened for kinetic and insensitivity parameters. Genes coding for shuffled variants of HPPD were expressed in Escherichia coli. Colonies with active HPPD enzyme cause the medium to turn brown due to the conversion of homogentisate to a brown ochronotic pigment (Zannoni et al., 1969). Proteins with the ability to turn the medium brown in the presence of 100 mM mesotrione in rounds 1 to 6 or tembotrione in rounds 7 and 8 were subjected to detailed analysis. Kinetic characteristics of k ON , k OFF , k cat , and K m were determined (Table I).
Saturation mutagenesis was used to discover functional diversity not available in the sequence database or accessed by other methods. As in all libraries, substitutions in the chloroplast-targeting amino terminus Table I. Kinetic activity and insensitivity parameters of purified evolved HPPD enzymes compared with the wild types from maize and soybean Insensitivity and fitness measurements are for the herbicidal compound mesotrione. Amino acid sequences are provided in Supplemental Figure S5. The diversity of substitutions found in the improved HPPD enzymes of this study is listed in Supplemental Table S1. In many cases, a single substitution was identified, while at other positions, many neutral or beneficial substitutions were identified. Figure 4A illustrates the progression of desensitization to mesotrione from the wild type to the best variant that emerged from the round 8 library. Overall improvements of 12-and 35-fold for mesotrione and tembotrione, respectively, were achieved (Table I; Fig. 4B).
Because there are diverse chemotypes among the registered herbicidal inhibitors of HPPD, a commercially valuable trait should confer tolerance to most or all of these registered chemistries. To ensure that directed evolution met that criterion, parents for each successive library were tested for insensitivity to the other inhibitors shown in Figure 1B, using the same procedures as for mesotrione. The result was that improved insensitivity, and hence enzyme fitness, obtained by selection with mesotrione or tembotrione was accompanied by improved insensitivity to all other inhibitors (Fig. 4B).

Isolation and Characterization of the Soybean HPPD Promoter
To mimic the pattern of expression with regard to growth stage, tissue and subcellular location, and uncharacterized signaling cues, the transgene promoters were based on the native soybean HPPD promoter. Attempts to directly isolate or measure native HPPD protein accumulation indicated low abundance in leaves (data not shown). This was confirmed by EST analysis. Our premise was that preserving the native expression pattern for HPPD while substituting in the improved insensitive maize HPPD variant genes would confer robust herbicide tolerance with only moderately higher expression of the transgene relative to the native.
A 1,225-bp genomic sequence at the 59 end of the soybean HPPD gene was isolated, sequenced, and subjected to promoter analysis using Promoter REAPer and Promoter Delineator (Simmons and Navarro Acevedo, 2010). Five putative TATA boxes, GTATA-AATAA(TATA1), CCAATATATG(TATA2), CCTTA-TATATC(TATA3), TATATAATAA(TATA4), and GAA-TATAAG(TATA5), were identified as shown in Figure  5A. The ORF starting from the first ATG after TATA3 encodes the 449-amino acid short protein. TATA2 is within the long protein transcript, and TATA3 is embedded in the coding region for the long protein. Based on the known transcripts, TATA5 and TATA3 appear to be best situated for the long and short transcripts, respectively. Each is located approximately 35 nucleotides from the transcription initiation, as expected for eukaryotic promoters (Smale and Kadonaga, 2003). ORF analysis of the long transcript indicates a short ORF encoding seven amino acids starting at position 2195. Downstream of this, the long ORF begins at position 299 and continues in frame comprising the short protein, which starts at +25. A schematic of the soybean HPPD promoter, transcripts, and translation products is shown in Figure 5B.
To evaluate the predicted TATA boxes for activity, 39 deletions and various mutations of the promoter region were constructed and the fragments fused with DsRed2 for expression analysis using transient expression in bush bean leaf tissue. Deletion of a 613-bp DNA fragment upstream of the short protein transcription start site, including all five putative TATA boxes, resulted in the elimination of promoter activity. When TATA3 and the short transcript region were deleted with the DsRed2 initiating at the same position as the long protein, 90% of the full promoter region accumulation of DsRed2 was measured, showing the contribution of the upstream transcript. To eliminate translation of the long protein and leave only the contribution of the short protein, a stop codon was introduced upstream of the DsRed2 Figure 4. A, Progression of enzyme fitness for mesotrione with rounds of shuffling. Bars indicate ON and OFF rate ratios, the insensitivity parameters that are functions of the rate constants for association and dissociation of enzyme and inhibitor, respectively. Fitness is the product of k cat /K m and insensitivity parameters (ON rate ratio and OFF rate ratio) and approximates k cat /K m 3 K i . B, Selection for insensitivity to mesotrione or tembotrione resulted in cross-insensitivity to other herbicidal inhibitors. The dashed line is drawn at the fitness level that provided tolerance in soybean to twice the labeled field rate (for maize) of tembotrione in 2013 field trials.
protein, which replaced the short protein. Only the short transcript DsRed2 would be translated in this construct. The same amount of DsRed2 accumulated as with the previous vector. Thus, it appears that both transcripts are capable of supporting translation activity, but this activity is not additive. Total HPPD product from the native promoter could be regulated via competition for transcription and/or translation. Upstream ORFs have been shown to reduce both transcription and translation in plant genes (Saul et al., 2009). However, in this case, a mutation to eliminate the ATG start codon of the HPPD upstream ORF had no effect on DsRed2 expression compared with the unaltered promoter construct.
To create transgene expression cassettes for HPPD inhibitor herbicide tolerance in soybean, cassettes were constructed with the intention of maintaining the HPPD regulatory information while modestly increasing the expression level. In some cases, either TATA3 or TATA5 was replaced with a synthetic core promoter (SynII; Bowen et al., 2000) to drive the short or long transcripts at an increased level. DsRed2 fusion constructs with SynII exhibited 2-to 3-fold increased protein accumulation. Adding the 35S enhancer region (Ow et al., 1987) at the 59 end of the altered promoter fragments or adding an enhancing region called Rsyn7, containing transcription factor motif sequences (Bowen et al., 2000), further increased expression. Plant transformation constructs were designed either with full-length shuffled maize HPPD in combination with promoters giving rise to only one transcript (e.g. Synthetic HPPD Promoter [SHP] 206:HPPD 9075) or with the N-terminal 86 amino acids of the long chloroplast targeting sequence of soybean followed by the conserved mature shuffled maize HPPD protein starting at amino acid 50 (e.g. SHP110e:HPPD Gm86:9075; Fig. 3F). The latter was designed to provide dual targeting of a highly insensitive HPPD enzyme. Final functional validation of the promoter/CTP/variant HPPD cassettes was by herbicide tolerance efficacy testing of transgenic soybean plants.

Herbicide Tolerance Conferred by Insensitive Maize HPPD in Soybean
Transgenic soybean events were created using particle bombardment of elite cultivar embryogenic callus cultures and selection with chlorsulfuron. T0 regenerated plants were acclimated to greenhouse conditions in soil Figure 5. A, Upstream 1,225-bp soybean promoter sequence and N-terminal regions of the long and short HPPD proteins. Underlined bold denotes predicted TATA sequences, red boldface denotes Met initiation sites for long and short proteins, red background denotes actual transcription start sites at positions 2237 and +1, and the gray box denotes the upstream ORF in the long transcript. B, Schematic of the soybean HPPD promoter and coding region. C, Schematics of example constructs SHP110e: HPPD Gm86:9075 and SHP206:HPPD 9075 used to create stably transformed soybean events. The amino acid sequence of round 6 HPPD 9075 is provided in Supplemental Figure S5.
for 2 weeks and sprayed with twice the labeled maize field rate (23) of mesotrione. When sibling regenerants from the same event were available, the second plant was sprayed with 23 tembotrione. For each construct, approximately 30 events were created with one or two copies of the HPPD insert at a single locus. Events that showed 30% or less injury to 23 mesotrione and 23 tembotrione (on sibling plants) were retained. T0 plants were sampled for DNA analysis to verify single-copy inserts and for protein analysis to measure HPPD accumulation in young leaves. Plants were grown to maturity, and the T1 seeds obtained were sown in the greenhouse or the field. T1 plants were treated with 43 mesotrione and screened by PCR to identify homozygotes. Tolerant (at or below 30% injury) homozygous lines were grown and harvested for seed increase.
For positive control comparisons, the strong synthetic virus-based promoters SUP and SCP1 (defined in Bowen et al., 2000) were used to drive HPPD variants. Repeated occurrences of wrinkled-leaf and chlorotic-leaf phenotypes along with occasional sterility in the SUP and SCP1 events suggested that these promoters were not optimal for event production. Measurements of HPPD protein accumulation in young leaves of T0 plants revealed that the strong synthetic promoters accumulated HPPD protein at or above 10 mg g 21 of total soluble protein. The promoters derived from the soybean native HPPD promoter led to the accumulation of HPPD protein in tolerant events at levels in the range of 0.2 to 1 mg g 21 total soluble protein. All localization scenarios resulted in the production of some herbicide-tolerant events, of varying efficacy. Neither protein accumulation nor construct properties alone predicted greenhouse T0 efficacy results. In all cases, however, tolerance to mesotrione at the maize field rate was more easily achieved than tolerance to tembotrione.
Field tests in 2013 with homozygous lines expressing the round 6 desensitized HPPD variant 9075 showed that several events created were highly tolerant to mesotrione, tembotrione, and isoxaflutole at 23 or 43 treatment levels. Some transient injury was apparent, but by 14 d after treatment, the plants were almost completely recovered, as shown in Figure 6. Tabulated results for two different events are shown in Table II. Additional testing of these events and round 8 variant events in a variety of environments and in several genetic backgrounds is in progress.

DISCUSSION
In addition to a steadily increasing inventory of glyphosate-resistant weeds, there are documented cases of weeds, such as Amaranthus spp., that are resistant to multiple herbicides, including acetolactate synthase inhibitors, atrazine, protoporphyrinogen oxidase inhibitors, HPPD inhibitors, and glyphosate (Heap, 2014). With no new broad-spectrum herbicides on the horizon, it will be essential for farmers to adopt best management practices, such as knowing which weed species are in the fields and which herbicides will control them and using multiple herbicides with differing modes of action together or in the same season to prevent resistant weeds from reproducing (Norsworthy et al., 2012). The availability of robust herbicide tolerance traits to complement glyphosate will facilitate the practice of using multiple herbicides to control the emergence and spread of resistant weeds.
Our first step in developing HPPD-tolerant soybean was to understand the physiological nature of the HPPD pathway in soybean (Fig. 1A). HPPD catalyzes the conversion of HPP and molecular oxygen to homogentisate. Current evidence supports a pathway structure in which HPP is produced by an aminotransferase from Tyr, which in turn is produced from prephenate via arogenate (Tzin and Galili, 2010). Extensive genetic and biochemical data indicate that a complete pathway for the biosynthesis of the three aromatic amino acids exists in chloroplasts (Maeda and Dudareva, 2012). Likewise, the enzymes that convert homogentisate to plastoquinone and tocopherol are also plastidic (Ruiz-Sola and Rodriguez-Conception, 2012).
Given a chloroplast localization of Tyr synthesis and homogentisate utilization, one might anticipate that Figure 6. Field trial with herbicide-tolerant transgenic soybean events treated with 23 the maize field rate of tembotrione 14 d after treatment. A, T3 generation homozygous plants expressing promoter SHP206 and full-length maize HPPD variant round 6 9075. B, T3 generation homozygous plants expressing 35S-enhanced promoter SHP110 with HPPD Gm86:9075. C, Nontransgenic unsprayed control. D, Nontransgenic sprayed control. Treatment was as described in Table II.
Tyr aminotransferase and HPPD would be found in the chloroplast. Data on Tyr aminotransferase are sparse, but it was shown recently that Arabidopsis gene locus At5g53970 encodes a specific and catalytically proficient Tyr:2-ketoglutarate aminotransferase that contributes a major but not exclusive portion of that activity (Riewe et al., 2012). Localization was not determined in that study. However, a petunia (Petunia hybrida) paralog of At5g53970 codes for a cytosolic enzyme that catalyzes the transamination of Tyr to HPP with phenylpyruvate as the acceptor substrate, providing for the special need for Phe in the petals as a precursor for aroma compounds (Yoo et al., 2013). HPP can also be made directly from prephenate, catalyzed by prephenate dehydrogenase (PDH). Although lacking in 10 nonlegume species tested, PDH activity was detected in various legumes, including soybean (Siehl, 1999). PDH activity in soybean (and Medicago spp.) was recently confirmed, genes identified, and the enzyme shown to be located in the cytosol (Schenck et al., 2014).
Our results with HPPD add to what appears to be a species-dependent variety of localization patterns. It was reported previously that Arabidopsis and Brassica spp. do not have predicted CTPs, while cotton and tomato do. Our analysis of the maize sequence did not predict a CTP but did predict a CTP for a previously unrecognized N-terminal extension of HPPD from soybean. Plant targeting peptides are very heterogenous, and different prediction programs often give conflicting results (Bruce, 2001;Jarvis, 2008). Thus, functional analysis was needed to validate the in planta localization of HPPD for each species. In the case of the maize HPPD, the N-terminal 50 amino acids were necessary and sufficient for chloroplast targeting in monocot and dicot cells. The soybean HPPD localization was more complex. We found two major ESTs with the long transcript giving rise to a protein both predicted and validated as capable of being translocated to the chloroplast, while the short transcript supported translation of a shorter protein that accumulated in the cytosol.
Why would the HPPD enzymatic step ever be localized in the cytosol? In theory, soybean and other legumes have a cytosolic minipathway that includes the cytosolic isozymes of chorismate mutase, PDH, and the short translation product of HPPD (Fig. 1A). While PDH bypasses prephenate aminotransferase, arogenate dehydrogenase, and Tyr aminotransferase, one still must postulate that the substrate (chorismate) and product (homogentisate) of the cytosolic minipathway must be transported across the chloroplast membrane. It is possible that the dual localization of HPPD is unique to legumes. Analysis of the M. truncatula genome sequence around the HPPD gene revealed that this legume species also has an N-terminal ORF extension that contains a predicted CTP. What is not clear is why Arabidopsis and some other nonlegumes would encode only cytosolic HPPD (Garcia et al., 1997(Garcia et al., , 1999. Perhaps the movement of homogentisate into the chloroplast serves as a regulatory step for the plastoquinone and tocopherol pathways. HPPD is the second step in the Tyr degradation pathway. Genes for all the subsequent pathway steps have been identified in Arabidopsis, shown to be expressed, and the encoded enzymes shown to be fully functional (Dixon and Edwards, 2006). Interestingly, just as deficiency in fumarylacetoacetate hydrolase, the ultimate step in the pathway, is a lethal disorder in animals, disruption of the same enzyme in Arabidopsis causes spontaneous cell death (Han et al., 2013). Although not investigated, overloading the Tyr degradation pathway may be a causal contributor to the negative phenotype seen when HPPD is expressed far above native levels.  Instead of relying on high overexpression of a transgene, we sought to enhance efficacy by creating highly insensitive HPPD variants. Despite the availability of structural analysis and rational design, the greatest gains in evolving the HPPD enzyme for inhibitor insensitivity were achieved with the introduction of primarily conservative diversity present in related sequences. Introduction of nonnaturally occurring diversity through saturation mutagenesis revealed novel beneficial mutations that were not available in the natural diversity. Assembling them in random combinations provided a further 3-fold improved fitness compared with the best round 6 variant. The variant with the highest fitness parameter, R8-1973, had 26 amino acid substitutions relative to native maize HPPD. The substitutions were distributed throughout the amino acid sequence and did not involve an amino acid with any reported role in catalytic function (Raspail et al., 2011). The evolution of improved function appeared to occur through the accumulated effect of many small changes outside of the active site that influence the catalytic properties of the enzyme in a manner that is not obvious upon inspection of the structure. We reached a similar conclusion after evolving a microbial acetyltransferase for activity with glyphosate (Siehl et al., 2007). Although desensitization to herbicidal inhibitors was associated with a drop in k cat from approximately 200 min 21 to 70 to 80 min 21 , nominal catalytic efficiency (k cat /K m ) was preserved due to a drop in K m from 6 mM to 2 to 3 mM (Table I).
A component of the effective management of herbicideresistant weeds is the use of compounds with various modes or sites of action within a program of best management practices. Adding HPPD inhibitors to the panel of available herbicides will allow growers to expand the choices of weed control chemistry they can use in soybean and other dicot crops. We believe that this work provides a viable candidate for a robust HPPD tolerance trait that should provide functional resistance to a broad range of HPPD inhibitor chemotypes.

Genomic and Transcript Analysis
A synthetic maize (Zea mays) wild-type HPPD complementary DNA (NCBI reference sequence NM_001112312.1) was assembled from commercially synthesized oligonucleotides. During the synthesis of the gene, an NcoI restriction site was engineered into the start of the sequence to facilitate cloning. The change of codon ATGCCC to ATGGGT resulted in the substitution of Gly for Pro at position 2. An EST (sgc5c.pk001.j9) coding for soybean (Glycine max) HPPD was identified from a DuPont-Pioneer proprietary soybean EST database using conventional bioinformatic tools including BLAST of the HPPD sequence described by Cahoon and Coughlan (2007). Using this soybean HPPD coding sequence as query, Pioneer Unigene PSO409914 was identified. An approximately 2-kb genomic sequence upstream of the EST was then identified from the genome assembly database. PCR primers (forward primer 59-GCAAGTATTTCAATACAATAGC-39 and reverse primer 59-GTTATCTGATATGATGTTGC-39) were used to amplify the HPPD locus from genomic DNA isolated from an elite soybean variety and the common Jack variety, following protocols for the isolation of plant genomic DNA (Qiagen). PCR reaction parameters were as follows: cycle 1, 94°C for 2 min; cycles 2 to 30, 94°C for 30 s, 65°C for 1 min, and 72°C for 5 min; cycle 31, 72°C for 10 min. A proofreading DNA polymerase, pfu Turbo (Stratagene), was used for PCR amplification. A 4,306-bp fragment and a 4,310-bp fragment were obtained from elite and Jack, respectively. These fragments were cloned into Zero blunt TOPO PCR cloning vectors (Invitrogen) and fully sequenced. Each locus comprises 39 sequences, the HPPD coding region, and upstream genomic sequence. The loci are highly conserved, with an overall 99% sequence identity at the nucleotide level.
59 RACE was used to validate transcription start sites for the native HPPD promoter using total RNA extracted from young elite soybean leaves and the First Choice RLM-RACE kit (Ambion) according to the manufacturer's protocol. Total RNA was isolated with the Qiagen RNeasy mini kit. Linked transcription-translation was performed with a Promega wheat germ kit, using 39 truncated soybean HPPD transcription vectors, and the products were analyzed by SDS gel electrophoresis (8%-12% gradient; Invitrogen).

Expression Vectors for Localization Microscopy and Promoter Analysis
Monocot transient expression vectors were made with DNA encoding N-terminal fragments of monocot HPPD proteins, the soybean HPPD protein, or the synthetic monocot consensus peptide fused to the gene coding for DsRed2 and C3GFP (Clontech) and inserted into a binary expression vector under the control of the maize Rubisco activase promoter (Liu et al., 1996) or the Arabidopsis (Arabidopsis thaliana) UBIQUITIN10 promoter (Norris, et al., 1993) and terminated with the potato (Solanum tuberosum) proteinase inhibitor II terminator region ) with a hygromycin selection cassette. The vector also contained an untargeted Zoanthus sp. Green cassette to provide cytoplasmic contrast and a kanamycin selection cassette. All genes were between the left and right border sequences of Agrobacterium tumefaciens transfer DNA. A positive control vector was identical except that the insert was DsRed2 fused to the CTP of Arabidopsis Rubisco activase, while a negative control was DsRed2 with no targeting sequence.
A second set of fluorescent vectors were similarly made with an AcGFP1 gene and inserted into a binary expression vector under the control of the Arabidopsis UBIQUITIN10 promoter (Norris et al., 1993) and terminated with the soybean Kunitz trypsin inhibitor 3 terminator region (NCBI accession no. S45092). Such vectors were used for either stable or transient gene expression in dicot plant cells. Fusions for testing the soybean CTP function contained the sequence-encoding amino acid residues 1 to 42, 42 to 86, or 1 to 86 of the long soybean HPPD protein. A positive control vector was identical except that the AcGFP1 coding region was fused to sequences encoding the 6H1 synthetic CTP (Lassner and Wilkinson, 2008), while a negative control was AcGFP1 with no targeting sequence.
Cassettes for testing the transient expression of promoters were similar to the DsRed2 N-terminal test vectors except that soybean HPPD promoterderived fragments replaced the Rubisco and UBIQUITIN10 promoters. A 2,061-bp fragment comprising the native HPPD promoter of elite soybean was created from the upstream genomic sequence. The PCR with forward primer hp0234 (59-GTTTT CCGCGGG TGTTGATCC-39) and reverse primer hp2296 (59-TCATT GGTACC TGGTGTGGTGTGATGCTGC-39) introduced SacII and KpnI restriction sites. This fragment was digested with restriction enzymes SacII and KpnI and ligated to a DsRed2 marker gene to form the native soybean HPPD promoter expression cassette. Introduction of the KpnI site between the promoter fragments and DsRed2 resulted in the introduction of a two-amino acid linker (GGTACC, Gly-Thr) in the fusion proteins containing the long protein CTP region.
The 1,225-bp genomic sequence at the 39 end of the promoter fragment was subjected to in silico promoter analysis using Promoter REAPer and Promoter Delineator (Simmons and Navarro Acevedo, 2010). Genomic sequences from other species including Arabidopsis, Medicago truncatula, Populus trichocarpa, Brassica rapa, Vitis vinifera, and the monocot sorghum (Sorghum bicolor) were included for comparison in this analysis. With the program Promoter REAPer, regions were identified in the soybean HPPD promoter that are predicted to be important for its activity based on the sequence conservation of a set of DNA motifs across the seven plant species. To evaluate the predicted TATA boxes in promoter activity, a deletion series was created with the 2,061-bp template using PCR with forward primer hp0234 (59-GTTTTCCGCGGGTGTTGATCC-39) and the reverse primers hp2154 (59-AGCATGGTACCTTGCGTCTGGGTTGAG-39), hp2048 (59-ATCTGGTACCTGATGTTGATGCGGC-39), hp1962 (59-AGGAGG-TACCGTCAAATCCACCTAG-39), hp1791 (59-AGCCTGGTACCTTGTGTG-TAAAAAAGATAAGAC-39), and hp1663 (59-TCCTTGGTACCTGATGCACTA-TATAACG-39). Single or triple mutations in the putative TATA boxes were created using site-directed mutagenesis (QuickChange; Stratagene). The resultant deletion and mutated promoter fragments were fused with DsRed2 to create cassettes for the analysis of expression activity in agroinfiltrated leaf tissues.
To create synthetic promoters based on the soybean HPPD promoter, synthetic elements based on the SynII core promoter and the Rsyn7 transcription factor-binding sequence (Bowen et al., 2000), each flanked by restriction sites XhoI and KpnI, were synthesized and ligated with the 39 ends of HPPD promoter-derived DNA fragments. SHP110e contains a cauliflower mosaic virus 35S enhancer (Ow et al., 1987) at the 59 end of the promoter fragment and a partial SynII core sequence inserted in place of TATA5. This promoter allows dual targeted proteins to be created with soybean-maize HPPD fusion proteins. SHP206 has point mutations in each of TATA1, TATA2, and TATA3. After the mutated TATA3 are the Rsyn7 enhancer and the SynII core followed by the 45-bp 59 untranslated region and the transcription start site of the native soybean gene. An in-frame stop codon prevents read through from upstream transcripts.

Marker Gene Expression and Fluorescence Measurements
All seedling plants were grown in growth chambers with 16 h of light at 375 to 450 mmol m 22 s 21 , 26°C day, and 22°C night. Expression plasmids for localization studies were transformed into A. tumefaciens AGL-1 via electroporation according to Shen and Forde (1989), and agroinfiltration (Kapila et al., 1997) was used to introduce the constructs into plant cells. A. tumefaciens cultures were grown overnight in Luria-Bertani medium with 40 mg L 21 kanamycin and a working suspension normalized to an optical density (OD) at 600 nm of 1 in 10 mM MgSO 4 , 400 mm acetosyringone, and 1 mM dithiothreitol (DTT). Leaves of 3-week-old maize or sorghum seedlings were infiltrated with the A. tumefaciens and examined by epifluorescence microscopy of hand sections 2 d later (Nikon Eclipse 80i microscope, with a 403/0.75 Plan Fluor Ph2 DLL objective through a TRITC HQR-Nikon 96321 filter) to visualize DsRed2. Images were captured with a Nikon Digital DS-5M camera and ACT-2U version 1.11.22.156 image-capture software. Exposure times of 1.5 to 3 s were used for all fluorescent image captures with a sharpness setting of 3. Infiltrated leaf samples were derived from plants of uniform developmental stage grown under the same conditions. Leaves of 4-week-old tobacco (Nicotiana benthamiana), 8-d-old bush bean (Phaseolus vulgaris variety Shade), and 10-d-old soybean seedlings were infiltrated with the A. tumefaciens and examined by epifluorescence microscopy of hand sections 4 to 5 d later using a Nikon Eclipse 80i microscope and a narrow band-pass GFP filter set.
Qualitative assessment of protein localization via bombardment with DsRed2 was determined by visually inspecting treated samples with a stereofluorescence microscope (Leica Microsystems; M165 FC with DsRed2 filter set no. 10447412) and acquiring images (Leica Microsystems; DFC300 FX R2) of representative examples at fixed exposure times. Initial examination of maize tissues expressing C3GFP or soybean tissues expressing AcGFP1 was conducted at approximately 24 h after bombardment with a Lumar fluorescence stereomicroscope (Carl Zeiss) equipped with both UV light-exciting (Zeiss Set 01) and red light-emitting (Zeiss Set 43 HE) filter sets to image the GFP and DsRed2, respectively. GFP fluorescence was captured using a 488-nm argon laser for excitation laser and a 500-to 550-nm band-pass emission filter. DsRed fluorescence was imaged using a 561-nm diode laser for excitation and a 575-to 615-nm band-pass emission filter. Chlorophyll fluorescence was captured by combining a 561-nm excitation and a 650-to 710-nm band-pass emission filter. AcGFP expression in stably transformed soybean seedlings was imaged with the confocal microscope in an identical fashion to bombarded tissues.
A. tumefaciens strains expressing plasmids for promoter characterization were infiltrated into bush bean leaves. Relative promoter strength was determined by quantitative measurement of the red fluorescence generated from expressed DsRed2 protein in infiltrated leaf tissues. Red fluorescence from 50 mg of protein extracted from infiltrated leaf discs was quantified using a Typhoon Trio+ Variable Mode Imager. Infiltration experiments for each construct were repeated at least three times. The background red fluorescence detected in leaves infiltrated with the negative control vector consisting of the double-enhanced version of the mirabilis mosaic virus promoter (Dey and Maiti, 1999) driving a GUS reporter was subtracted for data normalization. The DsRed2 readouts were used to calculate the level of DsRed2 from each construct relative to the DsRed2 expressed from the full native promoter construct (2,061 nucleotides 59 to the ATG of the short protein), which was set to 100%. Up to 30 leaf discs infected with the same culture were pooled for analysis. Each pool of infiltrated leaf samples (about 260 mg fresh weight) represented tissue equally derived from 15 plants of uniform developmental stage.

HPPD Expression, Purification, and Kinetic Analysis
Wild-type and shuffled variant genes were cloned into a modified version of pET24a(+) (Novagen), which places six His residues at the N terminus of the expressed protein. Vectors were electroporated into Escherichia coli host strain BL21(DE3). Cells were grown at 30°C in rich medium such as 23 yeast extract tryptone containing the selection antibiotic kanamycin. At OD 600 of about 0.6, isopropylthio-b-galactoside was added to 0.2 mM, the temperature was reduced to 16°C, and growth was continued for another 24 h. Cells were harvested by centrifugation and stored at 280°C. Cell pellets were lysed in BPER (Thermo Scientific) protein extraction reagent containing 0.2 mg mL 21 lysozyme, 1 mM DTT, bacterial protease inhibitor cocktail (Sigma-Aldrich), and endonuclease. Insoluble cellular debris was removed by centrifugation. HPPD protein was purified from the soluble protein solution by affinity chromatography on the nickel form of nitrilotriacetic acid resin (Qiagen). Protein concentrations were determined by the Bradford method, as supplied by Bio-Rad.
HPPD activity was detected through conversion of its product homogentisate to maleylacetoacetate, catalyzed by homogentisate dioxygenase from Pseudomonas aeruginosa (Amaya et al., 2004;generously provided by Dr. Graham Moran). The extinction at 320 nm was measured at 14.7 OD M 21 cm 21 . The homogentisate dioxygenase gene was expressed in E. coli strain BL21(DE3), and the protein was produced by methods similar to those described for HPPD. The purified enzyme had a K m for homogentisate of 23 mM and a k cat of 100 s 21 , parameters suitable for its use as a coupling enzyme.
HPPD activity was measured by placing a 6-mL aliquot of the substrate HPP at 50 times the desired final concentration into the wells of a low UV light-absorbing assay plate. Reactions were started by adding 294 mL of a mixture containing 25 mM HEPES, pH 7, 2 mM ascorbate, 10 mM FeSO 4 , 1 to 100 mM HPP, 50 nM homogentisate dioxygenase, and 5 to 240 nM HPPD. A 320 was monitored continuously in a plate-reading spectrophotometer (Spectramax; Molecular Devices). Catalytic performance was assessed by determining the substrate saturation parameters. The Michaelis-Menten kinetics protocol of the Spectramax software was customized for HPP concentrations ranging from 3.33 to 100 mM, or up to 1 mM if required. The software returned values of K m and V max using the Lineweaver-Burk transformation of the Michaelis-Menten equation.
To detect changes in ON and OFF rates, we observed the time course of an HPPD reaction as inhibitor binds to and inactivates the enzyme (k ON ) or is released from a preformed enzyme-inhibitor complex (k OFF ). In practice, a quantitative indicator of k ON was obtained by monitoring the time courses of HPPD reactions containing 60 or 120 nM HPPD and 100 mM HPP in the presence and absence of 4 mM inhibitor (e.g. mesotrione). The ratio of the reaction rate with inhibitor to that without inhibitor during the 70-to 90-s interval of the reaction was termed the ON rate ratio.
A quantitative indicator of k OFF was obtained by observing the time course of an HPPD reaction as inhibitor is released from a preformed enzyme-inhibitor complex. HPPD and inhibitor were incubated together at concentrations of 7.2 and 8 mM, respectively. Incubations with the same concentrations of enzyme but no inhibitor were done in parallel. After 1 h at room temperature, 5 to 10 mL of the enzyme-inhibitor complex was dispensed into the wells of the assay plate. Reactions were started with the addition of 290 to 295 mL of 25 mM HEPES, pH 7, 2 mM ascorbate, 10 mM FeSO 4 , 100 mM HPP, and 50 nM homogentisate dioxygenase. The reactions were monitored at 320 nm for 12 min. Reaction velocity accelerated as inhibitor was released from the enzyme until a steady state was reached, during which the reaction velocity was constant. The ratio of the steady-state rate in mixtures containing mesotrione (or other herbicidal inhibitor) to the initial velocity of mixtures lacking inhibitor was termed the OFF rate ratio.
The diketonitrile form of isoxaflutole was synthesized by DuPont chemists. All other inhibitors were obtained from Sigma-Aldrich.

HPPD Protein Detection in Planta
HPPD protein concentration in young soybean leaves was measured using liquid chromatography-tandem mass spectrometry according to Hu and Owens (2011) with some modifications. Leaf samples were collected from the youngest expanded trifoliate leaf. Total extracted protein samples were normalized to 0.8 mg mL 21 . Sample extracts (50 mL) used for analysis and purified HPPD protein standards were spiked into 50-mL negative tissue extracts for standard curves. Samples were heated for 15 min at 95°C, and 80 mL of digestion buffer (100 mM ammonium bicarbonate with 0.05% [v/v] Tween 20) was added. The samples were reduced with 6 mL of 0.25 M DTT at 50°C for 30 min and then alkylated with 6 mL of 0.3 M iodoacetamide at room temperature, in the dark, for 30 min. Samples were digested with trypsin (1 mg in 10 mL) and acidified. Internal standard (10 mL of peptide SIEDYEK, where K was labeled with stable isotopes 13 C and 15 N) was added. HPPD protein (10-mL injection) was quantified by its signature tryptic peptide SIEDYEK using a Waters ultra-performance liquid chromatography device coupled with the AB SCIEX QTRAP 4000 or 5500. Liquid chromatography was run with a linear gradient of 5% to 16% (v/v) 0.1% formic acid in acetonitrile over 2.5 min. Fragmentation transitions of 882.6 to 682.3 D and 890.6 to 690.3 D were monitored for peptide SIEDYEK and its internal standard peptide, respectively. Chromatogram peak area ratios to the corresponding internal standard were plotted against protein concentrations. The lower limit of quantification for HPPD by liquid chromatography-tandem mass spectrometry was 25 mg g 21 of total extracted proteins.

Stable Soybean Transformation and Herbicide Testing
Soybean plants expressing HPPD transgenes or marker genes were produced using the method of particle gun bombardment (Klein et al., 1987) using a DuPont Biolistic PDS1000/He instrument. A selectable marker used to facilitate soybean transformation was a chimeric gene composed of the S-adenosyl-Met synthase promoter (Falco and Li, 2010) from soybean, a highly resistant allele of soybean acetolactate synthase (Bedbrook et al., 1997), and the native soybean acetolactate synthase terminator region. The selection agent used during the transformation process was chlorsulfuron. HPPD cassettes comprised a synthetic constitutive promoter (Bowen et al., 2000) or the soybean HPPD-derived promoters, an insensitive HPPD variant, and an Arabidopsis UBIQUITIN3 gene terminator (Callis et al., 1995). Bombardments were carried out with linear DNA fragments purified away from any bacterial vector DNA. The selectable marker gene cassette was in the same DNA fragment as the HPPD cassette. In some cases, the characterized CTP sequences 6H1 (Lassner and Wilkinson, 2008) and the Arabidopsis Rubisco large subunit CTP (Lee et al., 2006) were fused to a truncated maize HPPD variant. Bombarded soybean embryogenic suspension tissue was cultured for 1 week in the absence of selection agent and then placed in liquid selection medium for 6 weeks. Putative transgenic suspension tissue was sampled for PCR analysis to determine the presence of the HPPD gene. Putative transgenic suspension culture tissue was maintained in selection medium for 3 weeks to obtain enough tissue for plant regeneration. Suspension tissue was matured for 4 weeks using standard procedures; matured somatic embryos were desiccated for 4 to 7 d and then placed on germination induction medium for 2 to 4 weeks. Germinated plantlets were transferred to soil in cell pack trays for 3 weeks for acclimatization. Plantlets were potted to 10-inch pots in the greenhouse for evaluation of herbicide resistance. Metal halide lights with 160 mE m 22 s 21 photosynthetically active radiation supplemented natural intensity during a 16-h photoperiod when light intensity was less than 500 mE m 22 s 21 . Day temperature was 28°C 6 2°C, and night temperature was 22°C 6 2°C. Relative humidity ranged from 50% to 90%.
T0 plants with HPPD transgenes were grown to the V2 to V8 growth stage and then sprayed in a chamber with commercial mesotrione (Callisto; Syngenta) formulation at a rate of 210 g ai ha 21 (two times the labeled rate for maize in the field). All mesotrione treatments were applied with 0.25% nonionic surfactant (Induce; Helena Chemical) and 1% ammonium sulfate in a spray volume of 374 L ha 21 . Individual plants were compared with untreated plants of similar genetic background, evaluated for herbicide response at 8 d after treatment, and assigned a visual response rating from 0% (no effect) to 100% (dead plant). Sibling plants were similarly evaluated with tembotrione (Laudis; Bayer Crop Science) at 93 g ai ha 21 . In the T0 generation, plants that had improved tolerance compared with controls based upon low injury scores (30% or less) were advanced for more extensive herbicide testing. T1 plants were evaluated for zygosity via quantitative PCR. Homozygous single-locus transgenic plants and their corresponding null segregants were identified, and T2 true-breeding seed was obtained from each. Plants were grown to the V1 to V2 growth stage and then sprayed with the commercial formulation of mesotrione at 420 g ai ha 21 as described for the T0 stage.
Homozygous T3 generation seeds were planted in short rows in Johnston, IA, at a DuPont Pioneer regulated field location. Herbicides were applied at the V4 stage with a backpack sprayer in a spray volume of 140 L ha 21 at two or four times the labeled rates for maize in the field, which are as follows: mesotrione, 118 g ai ha 21 ; tembotrione, 90 g ai ha 21 ; and isoxaflutole, 67 g ai ha 21 . All treatments included 0.25% (v/v) nonionic surfactant and 9.6 g L 21 ammonium sulfate. As negative controls, rows of nontransformed plants of the same variety were sprayed at the same rates. Positive control rows received no spray. Visual injury was rated at 3, 7, and 14 d after treatment.
The sequence of the N-terminally extended soybean HPPD gene was deposited in GenBank with the accession number KM460829.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Immunolocalization of native HPPD in maize leaf shows label only in chloroplasts.
Supplemental Figure S2. The N-terminal 50 amino acids of maize HPPD directs fluorescent protein to the chloroplasts of maize leaf guard cells.
Supplemental Figure S3. Immunolocalization electron microscopy of HPPD in soybean cotyledons shows label in chloroplasts and cytoplasm.
Supplemental Figure S4. Indirect measurement of the rates of association and dissociation of inhibitors and HPPD.
Supplemental Figure S5. Amino acid sequences of the wild type and shuffled variants of maize HPPD characterized in Tables I and II and Figures 4 and 6.
Supplemental Table S1. Chart of diversity contributing to the improvement of insensitive maize HPPD.