Physcomitrella PpORS, Basal to Plant Type III Polyketide Synthases in Phylogenetic Trees, Is a Very Long Chain 2′-Oxoalkylresorcinol Synthase*

Background: Physcomitrella PpORS is an ancient member of the plant type III polyketide synthase (PKS) family. Results: PpORS, produced in nonprotonemal moss cells, synthesizes pentaketide 2′-oxoalkylresorcinols using a unique substrate binding site. Conclusion: PpORS is a novel very long chain 2′-oxoalkylresorcinol synthase. Significance: This is the first step toward understanding the co-evolution of the type III PKS family and land plants. The plant type III polyketide synthases (PKSs), which produce diverse secondary metabolites with different biological activities, have successfully co-evolved with land plants. To gain insight into the roles that ancestral type III PKSs played during the early evolution of land plants, we cloned and characterized PpORS from the moss Physcomitrella. PpORS has been proposed to closely resemble the most recent common ancestor of the plant type III PKSs. PpORS condenses a very long chain fatty acyl-CoA with four molecules of malonyl-CoA and catalyzes decarboxylative aldol cyclization to yield the pentaketide 2′-oxoalkylresorcinol. Therefore, PpORS is a 2′-oxoalkylresorcinol synthase. Structure modeling and sequence alignments identified a unique set of amino acid residues (Gln218, Val277, and Ala286) at the putative PpORS active site. Substitution of the Ala286 to Phe apparently constricted the active site cavity, and the A286F mutant instead produced triketide alkylpyrones from fatty acyl-CoA substrates with shorter chain lengths. Phylogenetic analysis and comparison of the active sites of PpORS and alkylresorcinol synthases from sorghum and rice suggested that the gramineous enzymes evolved independently from PpORS to have similar functions but with distinct active site architecture. Microarray analysis revealed that PpORS is exclusively expressed in nonprotonemal moss cells. The in planta function of PpORS, therefore, is probably related to a nonprotonemal structure, such as the cuticle.

of polyketide natural products in plants and microorganisms (1). Type III PKSs are homodimers of 42-45-kDa subunits, and each subunit has its own active site that is shaped by highly conserved residues including the signature Cys-His-Asn catalytic triad. These enzymes iteratively condense a starter CoA substrate with acetate units derived from malonyl-CoA and cyclize the linear polyketide intermediates to produce compounds with unique ring structures. Different type III PKSs produce a diverse range of products. This diversity is the result of differences in the selection of the starter CoA substrate, the number of condensation steps, and the cyclization mechanism. For example, CHS, a plant-specific type III PKS that catalyzes the first step of the biosynthesis of flavonoids, condenses a phenylpropanoid-CoA starter substrate (e.g. p-coumaroyl-CoA (1k)) with three malonyl-CoA extender molecules, and then cyclizes the tetraketide intermediate by Claisen acylation to give a chalcone 2 (Fig. 1A). Several type III PKSs from microorganisms and plants utilize long chain fatty acyl-CoA esters as the starter substrate, although they differ in the cyclization mechanism used. Azotobacter vinelandii ArsC and Mycobacterium tuberculosis PKS18 produce triketide and tetraketide alkyl-2-pyrones (2, 3) (Fig. 1B). Conversely, A. vinelandii ArsB and Sorghum bicolor alkylresorcinol synthases (SbARSs) produce tetraketide alkylresorcinols (3,4), whereas Neurospora crassa 2Ј-oxoalkylresorcylic acid synthase (NcORAS) and Oryza sativa alkylresorcylic acid synthase (OsARAS) produce pentaketide and tetraketide alkylresorcylic acids, respectively (5-7) (Fig. 1B).
After extensive studies in the past few decades, we now know much about the catalytic mechanisms, gene regulation, and biological functions of type III PKSs (1,8). Although type III PKSs are found in bacteria (9) and fungi (5,10), they are more widely distributed in green land plants (Embryophyta). Each plant has a set of taxon-specific type III PKSs, which produce metabolites that are involved in UV protection (flavonoids), antimicrobial defense (stilbenes, bibenzyls, and alkylresorcinols), flower pigmentation (anthocyanins), spore/pollen protection (hydroxyalkyl pyrones), pollen tube growth (flavonols), and legume nodulation (isoflavonoids). The overall significance and scope of their roles suggest that type III PKSs have successfully co-evolved with land plants. This led us to investigate what major contributions ancestral type III PKSs might have made during the early evolution of land plants, especially during the colonization of land by ancestral plants. Our approach to gain insight into this question was to study the enzymatic properties of a modern-day plant type III PKS thought to closely resemble ancestral plant type III PKSs.
Bryophytes, comprising liverworts, mosses, and hornworts, are the simplest and earliest diverging lineages of land plants and offer unique windows into the early evolution of land plants. The model moss Physcomitrella patens is currently the only bryophyte whose genome has been sequenced (11), and its genome contains at least 17 putative type III PKS genes (12). Among them, PpORS (formerly PpCHS11) was shown to be basal to all plant type III PKS genes in phylogenetic trees. Both ORS and ARAS consume four molecules of malonyl-CoA. ORS catalyzes decarboxylative aldol cyclization, whereas ORAS catalyzes aldol cyclization. In contrast, ArsC condenses a long chain acyl-CoA ester with two or three molecules of malonyl-CoA to produce triketide or tetraketide 2-pyrone. ORS also produces triketide and tetraketide pyrones depending on the CoA ester substrate used. ARS and ARSA produce tetraketide alkylresorcinols and alkylresorcylic acids, respectively. Therefore, it was proposed to encode an extant enzyme that might closely resemble the most recent common ancestor (MRCA) of plant type III PKSs (13). In this study, we cloned PpORS and characterized the enzymatic properties of the recombinant PpORS to demonstrate that PpORS is a 2Ј-oxoalkylresorcinol synthase with substrate preference for very long chain fatty acyl-CoA esters. We then identified putative active site residues by performing structure modeling and mutagenesis studies. We also investigated the expression patterns of PpORS and Phypa126819, a P. patens PKS gene closely related to PpORS, by expressed sequence tag (EST) abundance and microarray analyses, and carried out phytochemical analysis in an attempt to learn about in planta function of PpORS. These studies should help us to understand the roles that type III PKSs may have played during early evolution of land plants.
Cloning of PpORS-P. patens (Hedw.) Bruch and Schimp subspecies patens, strain Gransden2004, was cultivated on sterile peat pellets (Jiffy-7; Jiffy Products International AS, Kristansand, Norway) for 1-1.5 months at 25°C under continuous light. Upper halves of gametophores without gametangia were collected with scissors. A full-length cDNA library was prepared by the oligo-capping method (16), and the cDNAs were cloned into the DraIII sites of pME18S-FL3 vector (AB009864). The full-length cDNA database in PHYSCObase was searched using the sequence of Contig1663 (17) as a query, and five corresponding clones (pph23j16, ppsp2a16, ppsp2k15, ppsp2n21, and ppsp12p08) were obtained. The coding region of PpORS was amplified by PCR from the ppsp2a16 clone using the primers shown in supplemental Table S1. The PCR products produced under standard PCR conditions were digested with restriction enzymes and subcloned into pET32a and pET28a expression vectors (Novagen) to give pET32-PpORS and pET28-PpORS, respectively.
Heterologous Production and Purification of Recombinant Proteins-Protein production and purification by Ni 2ϩ -chelation chromatography were performed as described previously (18) except that purification buffer was 20 mM potassium phosphate (KP i , pH 7.6) containing 200 mM NaCl. The enzyme solution was buffer-exchanged to 0.1 M KP i buffer, pH 7.6, using a 10DG column (Bio-Rad) for functional assays.
Enzyme Assay, Kinetic Analysis, and Product Determination-The standard assay mixture (100 l) contained purified enzyme (10 -20 g), 0.1 mM starter-CoA (e.g. C 24 -CoA), and 0.1 mM [2-14 C]malonyl-CoA (11 mCi/mmol) in 0.1 M KP i buffer, pH 7.6. Reactions to measure substrate preference and kinetic parameters were performed in 0.1 M KP i buffer, pH 7.6, containing 10% glycerol and 0.1% Triton X-100. After incubation at 30°C for 20 -40 min, the reaction was stopped by acidification (7.5 l of 1 N HCl), and the reaction products were extracted with ethyl acetate (200 l). The radioactive products were separated and quantified by thin layer chromatography (TLC) and phosphorimaging as described previously (15). Protein concentration was measured by an adapted Lowry method (Bio-Rad) with BSA as standard. The specific enzyme activity was expressed in pmol of the product produced s Ϫ1 mg Ϫ1 (picokatals mg Ϫ1 ).
Steady-state kinetic parameters of PpORS for C 10 -CoA and C 24 -CoA were determined in the presence of 0.1 mM malonyl-CoA and 9 M PpORS. The concentration of starter substrate varied from 5 to 80 M, and the reaction time was 10 min. K m and V max were calculated by fitting the data to the Michaelis-Menten equation using a nonlinear curve-fitting program (GraphPad Prism v.5.03).
Large scale reactions for product determination were performed with hexanoyl-CoA (C 6 -CoA), decanoyl-CoA (C 10 -CoA), or C 24 -CoA as the starter substrate under the standard assay conditions except that the concentration of malonyl-CoA was 0.2 mM and the reaction was run for 2 h. After standard work-up procedures, the reaction products were dissolved in methanol for MS analysis. Mass spectra were recorded using a Finnigan-Matt TSQ-700 mass spectrometer equipped with electrospray ionization and a Harvard syringe pump. Solutions were electrosprayed at 4.5 kV, with a capillary temperature of 75°C. Flow rate was varied between 1 and 10 l/min, and tube lens voltage was varied between Ϫ40 V and Ϫ80 V, depending on the compound. 2-Pyrone products of wild-type and mutant PpORS were compared with those produced by ArsC (2) or P. patens CHS (PpCHS) (17) by co-spotting on aluminum-backed silica 60 TLC sheets (EMD). Enzyme products were detected by staining with Fast Blue B salt (0.1% in H 2 O) (19).
Structure Modeling and Site-directed Mutagenesis-The structure of PpORS was first modeled with I-TASSER, which utilizes an ab initio multiple-threading approach (20). The quality of the model was further improved by a 10,000-step minimization in NAMD with AMBER ff99SBildn force fields, explicit solvation in TIP3P water, and Particle Mesh Ewald (21,22).
Characterization of PpORS Paralogs, EST Abundance, and Expression Analysis-Two putative genes homologous to PpORS were identified by tblastn search against the JGI Physcomitrella_patens.1_1 database with PpORS as the query sequence. Genomic sequences of these putative genes were manually translated into amino acid sequences based upon exon-intron architecture and homology to other type III PKSs. EST abundance of the two putative PpORS paralogs, Phypa126819 and Phypa72618, was then examined by blastn searches against the NCBI EST database, and the EST profile of PpORS was obtained by examining corresponding ESTs (Ppa.5302) in individual NCBI P. patens UniGene libraries. The expression patterns of PpORS and Phypa126819 were determined with whole genome microarrays (CombiMatrix, Mukileto, WA) based on all gene models v1.2 (11). RNA samples were obtained from protonema from liquid cultures, juvenile gametophores grown on solid medium (23), and freshly isolated protoplasts (24). The microarray experiments were done in biological triplicates. Data analysis with the Expressionist software (Genedata, Basel, Switzerland) was performed as described previously (25).
Phytochemical Analysis-Plants were grown on solid medium with (protonemata) and without (gametophores) ammonium tartrate as described previously (12). Dried and ground tissue (protonema, gametophore, or wheat bran, 0.5 g each) was extracted with 10 ml of acetone for 3 h with a wrist shaker, and the extract was filtered and vacuum-dried. The residue was dissolved in 6 ml of methanol. A portion (0.5 ml) was made basic (pH ϳ10) by the addition of 0.15 ml of 0.1 M KOH and incubated at 40°C for 4 h. The resulting hydrolyzed solution was acidified to pH ϳ2 with 10 l of 6 N HCl, and partitioned with hexanes (0.4 ml). The organic layer was vacuumdried, and the residue was dissolved in 30 l of methanol. Extracts were analyzed before and after alkali treatment by silica TLC (toluene/acetone/acetic acid 75/25/1, v/v/v), and stained with Fast Blue B salt (0.1% in H 2 O).
Phylogenetic Analysis-Phylogenetic analysis with the Bayesian inference method was performed using the MrBayes program (v. 3.2-cvs) (26), as described previously (13) with some modifications. The search was initialized at a user-defined tree, which was generated from the amino acid sequences by the default slow/accurate option in ClustalW. The Markov Chain Monte Carlo analysis was run for one million generations with four chains, and trees were sampled after every 100 generations. After all trees sampled during the first 250,000 generations were discarded, a consensus tree was constructed based on the remaining trees and displayed using MEGA4 (27).

Cloning and Heterologous Production of PpORS-
The fulllength coding region of PpORS was obtained from the moss gametophore cDNA library. Attempts to clone the gene from protonema of the moss were unsuccessful, suggesting that PpORS is not expressed during the protonema stage (see below). As discussed earlier (12), the third ATG codon among the four in-frame candidate start codons was assumed to be the translation initiation site of PpORS and used to produce the recombinant PpORS. The enzyme was produced both as a thioredoxin (Trx)-His 6 -tagged protein (Trx-PpORS, 61 kDa) and as a His 6 -tagged protein (PpORS, 44 kDa). The Trx tag increased the stability of the recombinant enzyme and had little effect on the product profile. Thus, Trx-PpORS was used for the large scale reactions. The deduced amino acid sequence of PpORS (ABU87504) contains the conserved catalytic residues, Cys 185 , His 323 , and Asn 356 (28) and the G 385 FGPG loop (29). The sequence identity of PpORS to other type III PKSs was generally low, and it was 20% to NcORAS, 34% to SbARS1 (4), and 36% to PpCHS and Medicago sativa CHS (MsCHS). Sequence alignments of PpORS with other type III PKSs are shown in supplemental Fig. S1.
In Vitro Analysis of PpORS Activity-We first tested the substrate preference of PpORS. PpORS produced a single major product (3j-3e) when the chain length of the starter fatty acyl-CoA substrate was C 6 to C 16 (Fig. 2). PpORS also produced the same type of compounds (3d-3a) from C 18 -to C 24 -CoA substrates, as evidenced by the progressively increasing R F value of the product with increasing chain length of the starter substrate (Fig. 2). From C 18 -to C 24 -CoA substrates, PpORS produced two additional types of compounds, one with lower R F values (4d-4a) and another with higher R F values (5d-5a). Compounds 5c-5a were the major products from C 20 -to C 24 -CoA substrates (Fig. 2). Based on ESI-MS (negative mode) and TLC data, the products, 3j-3a, were determined to be triketide alkylpyrones (Fig. 1B). ESI-MS analysis of 3h obtained from C 10 -CoA yielded a molecular ion peak [MϪH] Ϫ at m/z 237. Minor peaks were also observed at 273 [MϩCl] Ϫ and 475 [MϩMϪH] Ϫ , providing further confirmation of the mass. Compound 3h co-migrated on TLC with the reaction product of ArsC from the same starter substrate (C 10 -CoA), and both products were stained yellow-orange with Fast Blue B salt, similarly to 4-hydroxy-6-methyl-2-pyrone ( max ϭ 469 nm in . PpORS also produced additional minor products from C 8 -to C 16 -CoA substrates. The yields of these minor products were 3-7% of those of the major products (3j-3e) and remained uncharacterized. methanol) (supplemental Fig. S2, A and B). ArsC was shown to produce triketide alkylpyrones from the starter C 6 -to C 12 -CoA substrates and produces both triketide and tetraketide alkylpyrones from C 14 -to C 22 -CoAs (2). Therefore, we concluded that 3h is 4-hydroxy-6-nonyl-2-pyrone (Fig. 1B).
Compound 5a was stained violet with Fast Blue B salt, similarly to olivetol ( max ϭ 500 nm in methanol) (supplemental Fig. S2, A and C), suggesting that 5a contains a resorcinol ring. However, compounds 5a-5c exhibited lower R F values compared with the tetraketide alkylresorcinols produced by ArsB from the same substrates. After reduction with NaBH 4 , 5a and 4a were converted to polar compounds with lower R F values, whereas 3a remained intact, suggesting that 5a contains an oxo group as does 4a (data not shown). Positive mode ESI-MS of TLC-purified 5a yielded a molecular ion peak at m/z 950 corresponding to a protonated dimer [MϩMϩH] ϩ of 5-(2Јoxo)pentacosylresorcinol (supplemental Fig. S3). Based on these results, we concluded that 5a-5d are pentaketide 2Ј-oxoalkylresorcinols.
Enzymatic Properties and Kinetics of PpORS-We next tested the effects of different reaction conditions on the PpORS activity and product profile. The optimal activity for production of the pentaketide 2Ј-oxoalkylresorcinol (5a) from C 24 -CoA was observed at pH 7.5. The ratio of 5a to 3a decreased progressively as the pH increased, and the ratio was 2.3 at pH 6.5 and 0.67 at pH 8.0 (Fig. 3A). Similarly, the ratio of 5a to 3a varied depending on the concentration of the extender substrate (malonyl-CoA). More 5a was produced at higher concentrations of malonyl-CoA; however, overall activity decreased at 200 M malonyl-CoA, possibly due to substrate inhibition (Fig. 3B). Incubation time had no effect on the product profile, and the production of 5a, 3a, and 4a from C 24 -CoA increased steadily as incubation time increased up to 1 h (data not shown).
The steady-state kinetic parameters of PpORS for two representative substrates, C 10 Fig. S1). Because these unique substitutions might play roles in the differential enzyme activity of PpORS, these three residues were mutated individually and in combination to the corresponding MsCHS residues by sitedirected mutagenesis.
Among the seven single and multiple mutants studied, only the A286F mutant and the V277G/A286F double mutant exhibited activity, and the rest were inactive with all starter substrates examined. Compared with the wild-type enzyme (Fig. 2), it was evident that both mutants lost the ability to produce pentaketide oxoresorcinols and instead produced triketide alkylpyrones (3a-3j) from C 6 -to C 24 -CoA starter substrates (Fig. 5). Compound 3j produced by the A286F mutant co-migrated on TLC with 4-hydroxy-6-pentyl-2-pyrone ([MϪH] Ϫ at m/z 181) produced by PpCHS from C 6 -CoA. Both mutants exhibited similar substrate preference in that C 8and C 16 -CoA esters were most preferred. However, the V277G/ A286F double mutant was better than the A286F mutant at accepting C 22 -and C 24 -CoA esters as the starter substrate. Specific activity for the formation of 3a was 0.019 and 0.072 picokatals mg Ϫ1 for the A286F and V277G/A286F mutants, respectively. Tetraketide alkylpyrones (4a-4c) were also produced by both mutants at lower levels. Like the wild-type enzyme, neither mutant accepted p-coumaroyl-, cinnamoyl-, C 2 -, and C 4 -CoA esters as starter substrate.
Expression Profile of PpORS and Phypa126819-Based on homology to PpORS, two gene models, Phypa72618 and Phypa126819, were identified as PpORS paralogs in the Physcomitrella genome. Deduced amino acid sequences of the two gene models were 46 -48% identical to the PpORS sequence, assuming three putative introns at conserved positions. Proposed gene structures and sequence alignments of the deduced amino acids of the two gene models are provided in supplemental Fig. S5. The open reading frame of Phypa72618 contained two in-frame stop codons toward its 3Ј-end, and no corresponding EST was found in the NCBI EST database, suggesting that it is a pseudogene. On the other hand, the presence of a single Phypa72618 EST clone (BY962703) provided support for our deduced sequence of its gene product and also suggested that Phypa126819 may be a functional gene.
To investigate the expression pattern of PpORS, we first analyzed EST abundance within the EST libraries prepared from different Physcomitrella tissues. As compiled in the NCBI Uni-Gene database (Fig. 6A), PpORS ESTs were found in Physcomitrella EST libraries prepared both from gametophytes and sporophytes. EST counts were particularly high in the libraries of ppls (upper half part of gametophores) and ppaa (gametangia, shoot tip with antheridia and archegonia) but were lower in ppgs (green sporophytes) and ppsp (sporophytes with surrounding archegonia) libraries. In a sharp contrast, no EST was found in any of the libraries prepared from protonema and regenerated protoplasts. The results obtained from our microarray analysis agreed with the expression pattern inferred from the EST counts. Thus, PpORS was expressed in gametophores at mid-level compared with all transcripts, but its expression was not detected in protonema and freshly prepared protoplasts (Fig. 6, B and C). Phypa126819 was not expressed above the detection limit in all three tested tissue types in the microarray analysis.
Phytochemical Analysis-To examine whether PpORS products exist in planta either in monomeric or in esterified forms, we attempted to detect putative PpORS products from the moss gametophore before and after alkaline treatment. Wheat bran extracts, a positive control, yielded a major band on TLC, which stained violet with Fast Blue B salt. The extracts were determined to contain 5-nonadecylresorcinol ([MϪH] Ϫ at m/z 375) and 5-heneicosayresorcinol ([MϪH] Ϫ at m/z 403), in agreement with the literature (30) (supplemental Fig. S6). However, we failed to detect any resorcinol derivatives from the extracts of the moss gametophores as well as protonemata. No band that responded to the dye in the characteristic manner of resorcinol derivatives was detected. These results led us to conclude that the moss gametophore tissues do not contain (oxo)alkylresorcinols, either as monomeric or esterified forms. Alternatively, the amounts of (oxo)alkyresorcinols present are below the detection limit of this study (ϳ200 ng/g of tissue estimated based on the sensitivity of the dye staining).
Phylogenetic Analysis-Expanding previous phylogenetic analyses of type III PKSs (4, 13, 31), a phylogenetic tree was constructed with PpORS, the two P. patens paralogs, and other long chain acyl-CoA-utilizing type III PKSs including the gramineous ARSs (SbARS1, SbARS2, OsARAS1, and OsARAS2) (Fig. 7). The tree shows the expected progressive evolution from bacterial to fungal to plant enzymes. PpORS and its two moss paralogs form their own clade at the base of the plant clade. Thus, they are direct descendants of the MRCA of the plant type III enzyme family. The rest of the plant enzymes, in turn, form a sister clade to the PpORS clade. They themselves are divided into two sister clades, one made of anther-specific chalcone synthase-like enzymes (ASCLs) (15) and the other made of non-ASCLs. The gramineous ARSs belong to one of the two sister clades of the non-ASCL clade, reflecting their close evolutionary relationship among themselves. More importantly, PpORS and its moss paralogs are clearly separated from the gramineous ARSs, indicating that the gramineous

DISCUSSION
Pentaketide 2Ј-oxoalkylresorcinols were previously detected among the reaction products of NcORAS and long chain fatty acyl CoA substrates (5). However, based on a careful time course study, the authors concluded that NcORAS instead produced pentaketide 2Ј-oxoalkylresorcylic acids and that the detected 2Ј-oxoalkylresorcinols were indirect products formed by nonenzymatic (or enzymatic (6)) decarboxylation of the resorcylic acids. We also examined the possibility that PpORS may produce 2Ј-oxoalkylresorcylic acids. We monitored the PpORS reaction with C 24 -CoA and [2-14 C]malonyl-CoA at different time intervals up to 1 h and observed a steady increase of the formation of 5a and no evidence for the formation of a 2Ј-oxoalkylresorcylic acid. The same products were produced at different reaction pH values and substrate concentrations. Furthermore, we have demonstrated that 6-tridecyl-␤-resorcylic acid, an alkylresorcylic acid, is stable up to several hours in 0.1 M KP i buffer (pH 7.8) (14). Based on these results, we conclude that PpORS is a pentaketide 2Ј-oxoalkylresorcinol synthase. Thus, PpORS condenses a very long chain fatty acyl-CoA with four molecules of malonyl-CoA and cyclizes the pentaketide intermediate to produce 2Ј-oxoalkylresorcinol, through an aldol reaction accompanied by decarboxylation (Fig. 1B).
PpORS produces different major products from starter substrates of different chain lengths; triketide alkylpyrones from the C 6 to C 16 substrates and pentaketide oxoresorcinols from the C 20 to C 24 substrates. The shift is not abrupt, and from C 18 -CoA, the enzyme produced significant amounts of a triketide alkylpyrone (3d) and a pentaketide oxoalkylresorcinol (5d) along with a tetraketide alkylpyrone (4d). This "substratedirected product specificity" is not uncommon for type III PKS catalyzed reactions. Notably, NcORAS produces triketide alkylpyrones from C 4 to C 8 starter substrates, tri-and tetraketide alkylpyrones and tetraketide alkylresorcinols from C 10 to C 14 substrates, and tetra-and pentaketide alkylresorcylic acids from C 16 to C 20 substrates (5). Also, OsARASs produce alkylresorcylic acids and do not produce tetraketide alkylpyrones when the starter substrate is longer than C 14 -CoA (7). This raises a question as to the chemical nature of the in planta products of these enzymes. It has been postulated that enzymes that produce more than one product are advantageous in secondary metabolism because they generate chemical diversity at low cost (32). PpORS may well produce in planta alkylpyrones and 2Ј-oxoalkylresorcinols from fatty acyl-CoA esters of varying chain lengths for similar or different functions. In that case, substrate availability will determine the type of products made by the enzyme in planta. On the other hand, triketide and tetraketide pyrones are produced by most type III PKSs when nonphysiological substrates are given (1). Even with physiological substrates, most type III PKSs produce pyrones as in vitro derailment products. For example, CHS produces bisnoryangonin (a triketide pyrone) and coumaroyltriacetic acid lactone (a tetraketide pyrone) in addition to a chalcone (33). Furthermore, type III PKS mutants often produce pyrones when their active sites are compromised. The two PpORS mutants (A286F and V277G/A286F) in which putative active site residues were mutated failed to produce 2Ј-oxoalkylresorcinols, but still produced alkylpyrones (Fig. 5). These findings suggest that alkylpyrones produced by PpORS in vitro might be derailment products due to nonphysiological substrates or suboptimal reaction conditions. In that case the in planta products might be very long chain 2Ј-oxoalkylresorcinols.
Long chain alkylresorcinols have been found in higher plants including gramineous cereals. They are particularly abundant in the bran layer of cereal grains and are thought to exert antifungal activity (19). Long chain (C 19 -C 25 ) 2Ј-oxoalkylresorcinols were found as minor components in wheat and rye grains and etiolated rice seedlings (34,35). All plant (2Ј-oxo)alkylresorcinols identified to date are extractable monomers. To the best of our knowledge, no extractable alkylresorcinols have been detected in mosses, and the absence of alkylresorcinols in Sphagnum mosses is well documented (36). In A. vinelandii, monomeric alkylresorcinols and alkylpyrones produced by ArsB and ArsC from C 20 -and C 22 -CoA esters are the major lipid components of the protective cyst coat (2). PpORS is unique in that it produces exclusively 2Ј-oxoalkylresorcinols but does not produce alkylresorcinols. In addition to presenting an interesting mechanistic problem for future study, it might also bear significant implications for in planta function of PpORS because it implies important roles for the oxo group in the products. PpORS is expressed in gametophores, and its expression was not largely affected either by light/dark cycle (12) or by UV-B exposure (31). Moreover, PpORS is not expressed in moss protonemata and protoplasts (Fig. 6). Taken together with our failure to detect monomeric or esterified resorcinol derivatives from the moss gametophore extracts, these data suggest that 2Ј-oxoalkylresorcinols produced by PpORS might be constituents of gametophore-specific materials, such as a cuticle (37) or lignin-like materials (38). The plant cuticle is a waxy covering that protects plant from desiccation. In these materials, chemical components could be bound, at least partly, through alkaline-resistant linkages such as ether bonds. In this context, it is worthwhile to note that the ASCL-produced tetraketide 2Ј-oxoalkylpyrones have been proposed to be reduced by tetraketide ␣-pyrone reductases before being incorporated into sporopollenin, a biopolymer found in the pollen and spore walls (39). The resultant hydroxyl group of the hydroxyalkylpyrones might then form ether or ester linkages in the sporopollenin polymer. The oxo group in 2Ј-oxoalkylresorcinols might also be reduced in a similar manner in planta.
Thr 197 , Gly 256 , and Phe 265 (numbering of MsCHS) that are highly conserved in CHS and many other type III PKSs are uniquely replaced with Gln 218 , Val 277 , and Ala 286 in PpORS, respectively. All three residues are situated at the opposite side of the active site cavity from the nucleophilic Cys residue to which the growing polyketide chain is attached during catalysis (Fig. 4). Numerous studies have shown that Thr 197 and Gly 256 play critical roles in determining both substrate preference and the extent of condensation reactions by controlling the size and shape of active site cavity (reviewed in Ref. 1). For example, substitutions of Thr 197 , Gly 256 , and Ser 338 of MsCHS with the corresponding residues found in 2-pyrone synthase were sufficient to convert the T197L/G256L/S338I triple mutant of MsCHS to a functional 2-pyrone synthase (40). However, only a few mutational studies have been done on Phe 265 , which sits at the entrance to the active site. Whereas the F233A mutant of a bacterial type III PKS, RppA, was devoid of enzymatic activity (41), the F265V mutant of MsCHS exhibited similar substrate selectivity as the wild-type enzyme (42). We could not properly access the functional role of the Gln 218 residue of PpORS because the CHS-like substitution of the Gln 218 to a Thr apparently disturbed the folding processes and made the mutant protein insoluble when produced in E. coli BL21(DE3) cells. Replacement of Val 277 with a Gly had similar detrimental effects on protein structure, suggesting that Gln 218 and Val 277 are critical for proper folding and structural integrity of PpORS. Substitution of Ala 286 with a bulky Phe had dual effects on enzyme activity. First, the A286F mutant completely lost the ability to form pentaketide 2Ј-oxoalkylresorcinols. Second, the ability of the mutant to accept the starter substrate gradually decreased as the chain length grew longer than C 16 . Both effects, although the former was more drastic, could be explained by the constriction of the active site cavity by the bulky Phe residue. The V277G/A286F double mutant exhibited a slightly higher activity than the A286F mutant in accepting C 22 -and C 24 -CoA substrates. Introducing a smaller Gly residue in place of a Val could have expanded the active site cavity to better accommodate very long chain substrates, but it failed to allow the mutant to form the pentaketide product.
PKS18, NcORAS, SbARS, and OsARAS also accept long chain fatty acyl-CoA esters as starter substrates. X-ray crystallography and mutagenesis studies of PKS18 indicated that Thr 144 , Cys 205 , Ala 209 , Ile 220 , His 221 , and Cys 275 residues are situated along the long acyl binding tunnel and crucial in accepting the long chain acyl moiety of starter substrates (3). The corresponding residues in PpORS are Thr 154 , Ser 215 , and Ala 219 , Thr 234 , Leu 235 , and Ala 286 , respectively (supplemental Fig. S1). For NcORAS, Cys 120 , Thr 121 , Ser 186 , Met 189 , Phe 210 , and Ser 340 were proposed to be involved in shaping the long acyl binding tunnel (6). The corresponding residues in PpORS are Ser 153 , Thr 154 , Ser 215 , Gln 218 , Phe 238 , and Ser 358 , respectively. On the other hand, for the gramineous ARSs, Tyr 140 (numbering of SbARS1), Ala 145 , Ala 205 (Cys in OsARAS) and Met 265 were proposed to be important in determining their preference for long chain acyl-CoA substrates (4), and the corresponding residues in PpORS are Ser 153 , Thr 158 , Gln 218 , and Val 277 (supplemental Fig. S1). Although caution should be taken in the absence of x-ray structures, PpORS appears to be more similar to NcORAS than to the gramineous ARSs with regard to the construction of the acyl binding site. This is unexpected because the overall sequence of PpORS is more similar to SbARS (34% identity) than to NcORAS (20% identity).
The apparently different mechanisms for controlling substrate preference in PpORS and the gramineous ARSs can be understood when the evolutionary relationships of these enzymes are considered. In phylogenetic trees constructed in this (Fig. 7) and other studies (13,31), PpORS belongs to the basal clade along with its two moss paralogs, whereas the gramineous enzymes appear to have evolved later, independently from the PpORS lineage. In other words, PpORS and the gramineous enzymes appear not to be orthologs despite their similar activity. We searched available plant genome sequences for putative PpORS orthologs that belong to the PpORS clade, but were unable to find any. This suggests massive gene loss of PpORS orthologs in tracheophyte lineages. We have previously proposed an evolutionary scenario for the plant type III PKS family, in which ancestral acyl-CoA utilizing type III PKS genes have undergone repeated gene duplication-loss and functional diversification throughout evolution (13). Taken together, PpORS and the gramineous ARSs appear to represent an interesting example of enzyme evolution in which the same functions (the common use of long chain acyl-CoA starter substrates and the same type of cyclization) have evolved more than once within an enzyme family by adopting different structural strategy (different architecture of the binding sites).
In this study, PpORS was characterized to be a very long chain 2Ј-oxoalkylresorcinol synthase. A unique set of putative active site residues (Gln 218 , Val 277 , and Ala 286 ) were identified through structure modeling and sequence alignments. Replacement of the Ala 286 with a bulky Phe affected both the cyclization mode and the substrate preference. Expression profiling and phytochemical studies suggested that in planta products of PpORS might not exist in monomeric form and could be components of a polymeric structure that is specific to nonprotonemal tissues, such as the cuticle. More genetic and phytochemical studies are warranted to understand how the type III PKS family and land plants successfully co-evolved. This study represents the first step of such efforts.