Seeds of Cannabis sativa, Skunk and Fourway drug-type varieties (The Sensi Seed Bank, Amsterdam, the Netherlands), and Kompolti fiber-type variety (Dr. D. Watson, HortaPharm, Amsterdam, the Netherlands) were germinated and 9-day-old seedlings were planted into 11 LC pots with soil (substrate 45 L, Holland Potgrond, Van der Knaap Group, Kwintsheul, the Netherlands) and maintained under a light intensity of 1930 lux, at 26 °C and a 60.02 ± 7.43% relative humidity (RH). After three weeks the small plants were transplanted into 10 L pots for continued growth until flowering. To initiate the flowering, two-month-old plants were transferred to a photoperiod chamber (12 h light, 27 °C, and 37.0 ± 11.6% RH). 3-month-old male plants were used to pollinate female plants. 5-day-old seedlings, young leaves from 13-week-old plants, female flowers in different stages of development, fruits from 18-day-old plants and male flowers from 4-month-old plants were harvested. Roots from 4-month-old female plants were harvested and washed with cold water to remove residual soil. In addition, cones of H. lupulus at different stages of development were collected in September 2004 from the Pharmacognosy gardens (Leiden University). All vegetal material was weighed and stored at -80 °C.

Six grams of frozen female flowers containing 17-, 23-, 35-, and 47-day-old glandular trichomes from cannabis plants were removed by shaking frozen material through a tea leaf sieve and collecting it in a mortar containing liquid N₂, where it was immediately used for RNA extraction. The effectiveness of this method is comparable to the method reported by Yerger et al. (1992), which consists of shaking the tissues with powdered dry ice and sieving. For lupulin glands, frozen cones of hop were ground in liquid nitrogen using a frozen mortar and pestle only to long enough to separate the bracteoles and then were shaken using the same system as for cannabis glandular hairs.

For total RNA isolation from flowers, leaves, roots, seedlings, fruits, glandular hairs, glandular lupulins, and hop cones, frozen tissues (0.1-0.5 g) were ground to a fine powder in a liquid-nitrogen-cooled mortar, suspended and vortexed in 0.5 mL of extraction buffer (0.35 M glycine, 0.048 M NaOH, 0.34 M NaCl, 0.04 M EDTA, and 4% SDS) and 0.5 mL of water-saturated phenol. The suspension was centrifuged at 1,400 rpm for 2 min to separate the phenol and water phases. The RNA was precipitated in 1/3 volume 8 M LiCl at 4 °C overnight. The RNA was collected by centrifugation at 14,000 rpm for 10 min and resuspended in 0.1 mL of H₂O. The suspension was heated at 60 °C for 20 min and centrifuged. A total of 5 μL of 3 M Na-acetate (pH 4.88) was added to the supernatant to initiate the precipitation with 0.25 mL of 100% EtOH at -20 °C for 30 min and centrifuged at 14,000 rpm for 7 min. The pellet was washed with 250 μL of 70% EtOH, centrifuged for 2 min at 14,000 rpm, dried at 60 °C for 15 min, dissolved in 50 μL of H₂O, and incubated at 50 °C for 10 min.

Alternatively, Micro-Fast Track 2.0 kit and Trizol Reagent (Invitrogen, Carlsband, CA, USA) were also used for mRNA and total RNA isolation following the manufacturer's instructions. Isolated RNA was stored at -80 °C.

Degenerated primers, HubF (5'-GAGTGGGGYCA RCCCAART-3'), HubR (5'-CCACCIGGRTGWGYAAT CCA-3'), CHSF (5'-GGYTGYIIIGSYRGTGGMA-3'), CHSR (5'-CCIGGYCCRAASCCRAA-3'), STSF (5'-GGITGCIIIGCIGGIGGMAC-3'), and STSR (5'-CCIGGI CCRAAICCRAA-3') (Biolegio BV, Malden, the Netherlands) were designed based on CHS, STS, and stilbene carboxylate synthase (STCS) sequences from H. lupulus, A. hypogaea, Rheum tataricum, Pinus strobus, V. vinifera and H. macrophylla. For primers HubF, HubR, CHSF, and CHSR, the conserved regions were from CHS and VPS (GenBank accession nos. AJ304877, AB061021, AB061022, AJ430353, and AB047593), while for STSF and STSR they were from STS and STCS (GenBank accession nos. AB027606, AF508150, Z46915, AY059639, AF456446). RT-PCR was performed with total RNA or mRNA as a template using different combinations of primers. Reverse transcription was performed at 50 °C for 1 h followed by deactivation of the ThermoScript Reverse Transcriptase (Invitrogen) at 85 °C for 5 min. The PCR conditions were: 5 cycles of denaturation for 45 s at 94 °C, 1 min of annealing at 40 °C, 1 min of DNA synthesis at 72 °C, followed by 5 cycles with annealing at 41 °C and 5 cycles with annealing at 43 °C, and ending with 35 cycles with annealing at 45 °C. A Perkin Elmer DNA Thermal Cycler 480 and a Taq PCR Core kit (QIAGEN, Hilden, Germany) were used. A final extension step for 10 min at 72 °C was included. The PCR products were separated on 1.5% agarose gel, visualized under UV light, and recovered using Zymoclean Gel DNA Recovery kit (Zymo Research, Orange, CA, USA) or QIAquick PCR Purification kit (QIAGEN) according to the manufacturer's instructions.

For generation of 5' and 3' end cDNAs, we used total RNA, gene specific primers, and a SMART RACE kit (ClonTech, Palo Alto, CA, USA). The cycling parameters were: 94 °C for 1 min followed by 35 cycles at 94 °C for 35 s, annealing temperature for 35 s and 72 °C for 3 min. A final elongation step for 10 min at 72 °C was included. Gene-specific, amplification, and sequencing primers and annealing temperatures are shown in Table S1 (Supplementary Material). The PCR products were separated on 1.5% agarose gel and visualized under UV light. For generation of complete sequences, total RNA and amplification primers were used. Nested amplifications were made with gene-specific primers to select PKS sequences for sequencing. PKS full-length cDNAs were re-sequenced with sequencing primer in order to confirm that the ORFs of the sequences were correct. The corresponding amplification products were ligated into pGEM-T vector and cloned into JM109 cells according to the manufacturer's instructions (Promega, Madison WI, USA). Plasmids containing the inserted fragment were sequenced (BaseClear, Leiden, the Netherlands).

The PKS 3D models were generated by the web server Geno3D (Combet et al., 2002), using as template the X-ray crystal structures of M. sativa CHS2 (Protein Bank accession nos. 1BI5.pdb, 1CHW.pdb, and 1CMl.pdb). The models were based on the sequence homology of residues Arg5-Ile383 of the PKSs PKSG1, PKSG2, PKSG4, PKSG5, and PKSF3. The VPS model was based on the sequence homology of the residues Val4-Val390. The corresponding Ramachandran plots confirm that the majority of residues grouped in the energetically allowed regions. All models were displayed and analyzed by the program DeepView - Swiss-PdbViewer (Guex and Peitsch, 1997).

RNA isolated from glandular hairs of cannabis flowers and plant tissues was used as a template for reverse transcription-polymerase chain reaction (RT-PCR) amplification of segments of PKS mRNAs. RNA from hop tissues was used as a positive control. The degenerated primers corresponded to conserved regions surrounding Gln 119, the catalytic domain around Cys 164, a region surrounding His 303, and the C-terminal region of the selected protein sequences from CHS, STS, and STCS. Amplification with different primer combinations yielded products of expected size (Figure S1, Supplementary Material). Nucleotide sequence analysis showed open reading frames (ORFs) encoding for PKS proteins. Amplifications derived from mRNA of leaves, seedlings, glandular hairs, and female and male flowers showed 76%-78% homology with CHS3 from H. lupulus and 66%-69% homology with the known cannabis CHS-type PKS (Raharjo et al., 2004). This CHS-type PKS was also identified in female and male flowers and in glandular hairs. Partial sequences of VPS and CHS2 from the hop cone's secretory glands (also called lupulin glands) were obtained. It is known that VPS and CHS1 are expressed in lupulin glands (Okada and Ito, 2001; Matousek et al., 2002a,b) and the presence of a gene family of VPS, as well as one of CHS, has been suggested. Supplementary Figure S2 shows the strategy to obtain the full-length cDNAs of the likely PKS gene(s) differing from the earlier CHS-type PKS gene.

Several full-length PKS cDNAs containing ORFs of 1158 bp were obtained (Table 1). The main difference among these PKS cDNAs was the size of untranslated regions (UTRs). Several studies suggest that the untranslated regions (UTRs) are important for the control of gene expression in plants at the post-transcriptional level. Stability (Feldbrugge et al., 2001; Schwartz et al., 2006; Vaucheret, 2006), transport (Li and Hunt, 1997; Siomi, 2000), and translation (Klaff et al., 1996; Guo et al., 2000; Geslain and Ribas de Pouplana, 2004) of RNA depend on the 5'-UTR, the 3'-UTR, or both. We believe that variation in the size of UTRs from these PKS cDNAs could be the result of alternative transcription initiation and polyadenylation sites from post-transcriptional processing of PKS pre-mRNAs (Dean et al., 1986; Joshi, 1987; Rothnie, 1996). The nucleotide sequence data were deposited at GenBank database with the GenBank accession numbers EU551163 (PKSG1), EU551164 (PKSG2), EU551162 (PKSF3), EU551165 (PKSG4), and EU551166 (PKSG5).

The ORFs encode proteins of 385 amino acids with a calculated MW of approximately 42 kDa and a pI ranging from 5.98 to 6.09 (Table 2). The predicted amino acid sequences have more than 97% homology (Figure 1). In the plant H. lupulus, three CHS 1 mRNA sequences were isolated that shared more than 99% and 98% identity on nucleotide and amino acid levels, respectively. They are clearly homologous to the original CHS 1 sequence (AJ304877), forming a CHS 1 oligofamily (Matousek et al., 2006). The presence of this oligofamily could promote differences in concentration of prenylflavonoids in different varieties of H. lupulus (De Keukeleire et al., 2003; Matousek et al., 2005, 2010).

According to the percentage of identity at amino acid level (Table S2), our five PKSs showed to have more homology with the CHSs 3, 4, and VPS from H. lupulus than other PKSs (70%-73%). Conserved amino acid residues present in type III PKSs are also preserved in the amino acid sequences from our PKSs (Figure 1). The catalytic triad (Cys157, His297, and Ans330), the “gatekeeper” phenylalanines (Phe208 and Phe259), and Met130, which ties one catalytic site to the other one on the homodimeric complex, as well as Gly250, which determines the elongation cavity volume of the active site, are strictly preserved when compared to CHS2 from alfalfa (Ferrer et al., 1999; Jez et al., 2000a; Jez et al., 2001). The GFGPG loop, which is important for the cyclization reactions in CHS/STS-type PKSs (Suh et al., 2000), is also preserved on our PKSs. In the starter substrate-binding pocket, the amino acid residues Ser126 and Ser332 are also preserved as on alfalfa CHS2, but Glu185, Thr187 and Thr190 are replaced by an Asp, a Met and a Leu, respectively. Only in PKSG2 is the residue Thr187 preserved as on alfalfa CHS2. In the PKS 2-pyrone synthase (2PS), the amino acid residue Thr190 is replaced by a Leu. All these amino acid residues are important for the selectivity of the starter substrate. In alfalfa CHS2, the catalytic efficiency of the p-coumaroyl-CoA-binding pocket was affected by replacement of these residues (Jez et al., 2000b). The replacement of Thr197 by Leu slightly reduced its catalytic efficiency for the substrate p-coumaroyl-CoA. However, it was increased for the substrate acetyl-CoA. It was found that the change of three amino acid residues (Thr197Leu, Gly256Leu, and Ser338Ile) converts a CHS activity to 2PS activity. In our PKSs, the substrate-binding pocket could be slightly different from that of the alfalfa CHS2 by changes from polar to nonpolar amino acid residues (Thr187Met and Thr190Leu) and from length and bulkiness of the side chain residues (Glu185Asp185). Although the residues that shape the geometry of the active site (Pro131, Gly156, Gly160, Asp210, Gly256, Pro298, Gly299, Gly300, Gly329, Gly368, Pro369, and Gly370) are preserved on alfalfa CHS2 (Ferrer et al., 1999), Leu209 is replaced by the amino acid Ile.

The CHS-based homology modeling predicted that our cannabis PKSs would have the same three-dimensional overall fold as alfalfa CHS2 (Figure S3). A schematic representation of the residues that shape the geometry of the active site of cannabis PKSs is shown in Figure 2. The models suggest small differences in the local reorientation of the residues that shape the active site of the cannabis PKSs. The substrate and product specificity of the enzyme reaction can be affected by the steric modulation of the active-site architecture (Ferrer et al., 1999; Jez et al., 2000a,b, 2001; Suh et al., 2000). It can be inferred from these models that cannabis PKSs could have differences in substrate specificity or in catalytic efficiency (k Figure 2

The relative orientation of the side chains of the active site residues from M. sativa CHS with the 3D models of C. sativa PKS. The corresponding side chains in alfalfa CHS are shown in yellow backbones and are numbered.

We characterized five PKS cDNAs, four from glandular hairs (PKSG1, PKSG2, PKSG4, and PKSG5) and one from seedlings (PKSF3). The last one was also identified in male and female flowers by RT-PCR and sequencing, while the expression of PKSG2 was also detected in leaves. Although a low expression of the known cannabis CHS-type PKS was reported in female flowers, glandular hairs, leaves, and roots (Raharjo et al., 2004), we detected by RT-PCR that it is also expressed in male flowers. Southern blot analyses of C. sativa genomic DNA showed that at least four homologous PKS genes are present (Raharjo et al., 2004). A phylogenetic analysis (Figure 3) of our cannabis PKSs revealed that they group together with other non-chalcone and non-stilbene forming enzymes and appear to be most closely related to the CHSs 2, 3, 4, and VPS from H. lupulus, while the known cannabis CHS-type PKS groups with chalcone-forming enzymes and is most closely related with H. lupulus CHS1, of which expression is highly specific in the lupulin glands during the cone maturation, but also it can be detected on all the plant (Matousek et al., 2002a). CHS2 has been detected from hop leaf and lupulin fractions and CHS4 and VPS have been detected from the glandular tissue of hop cones, although no expression for CHS3 has been found on any hop tissue (Okada and Ito, 2001; Novak et al., 2003; Okada et al., 2004). In regard to enzyme activities, CHS1 shows CHS activity with p-coumaroyl-CoA. VPS, CHS2, and CHS4 use isovaleryl-CoA and isobutyryl-CoA, but CHS2 and CHS4 reactions form byproducts. No reaction products have been detected for CHS3 enzyme activity (Okada et al., 2004; Novak et al., 2006). Probably, the right substrates have not been identified for this enzyme yet.

A comparison of the 3D model of our PKSs, VPS, and alfalfa CHS2 predicted variations in the orientation of the active site residues which suggests a different substrate specificity regarding VPS (Figure 4).

The isolation and identification of PKSs with different enzymatic activity in one plant species has been reported, as well as the occurrence of PKS gene families in a species (Rolfs and Kindl, 1984; Zheng et al., 2001; Samappito et al., 2002; Matousek et al., 2006). A number of points suggest the participation of several PKSs in the secondary metabolism of this plant. These are: the CHS- and STS-type, and olivetol-forming PKS activities from protein crude extracts from C. sativa (Flores-Sanchez and Verpoorte, 2008a), the expression and partial characterization of a PKS cDNA from leaves with CHS-type activities (Raharjo et al., 2004), the characterization of four PKS cDNAs generated from mRNA of a glandular hair mixture and one from mRNA of seedlings, which is also expressed in female and male flowers (this study), and the small gene family of PKS detected in genomic DNA (Raharjo et al., 2004). Recently, the crystallization of a cannabis PKS, called hexanoyl triacetic acid lactone (HTAL) or olivetol synthase (OLS), condensing malonyl-CoA and hexanoyl-CoA to form hexanoyl triacetic acid lactone or olivetol, was reported (Taguchi et al., 2008; Marks et al., 2009; Taura et al., 2009). It has been proposed that pyrones or polyketide free acid intermediates undergo spontaneous cyclization to yield alkylresorcinolic acids or stilbenecarboxylic acids (Akiyama et al., 1999); or that post-PKS modifying enzymes are required to form them (Austin and Noel, 2003; Eckermann et al., 2003; Flores-Sanchez and Verpoorte, 2009). The homology of this protein with our PKSs was more than 97%. Although the differences in the amino acid residues from our PKSs and HTAL/OLS are small (Figure 1), probably because of the varieties of cannabis plant used, a complete biochemical characterization of the proteins encoded by PKSG1, PKSG2, PKSF3, PKSG4, and PKSG5 is required to study and understand their function and diversity, as well as to learn more about signals or factors that could control their transcription and translation.

It is interesting that the cDNAs PKSG1, PKSG2, PKSG4, and PKSG5 were generated from a combination of trichomes at different development stages. Under our greenhouse conditions, the beginning of the development of the trichomes on the perigonial bracts of female flowers was observed from 15 to 18 days after transferring the plants to a photoperiod regime. A full development of the glandular trichomes with presence of resin was observed from 30-35 days after initiation of the photoperiod regime. In addition, for the gland trichome isolation, flowers from two varieties of drug type (Skunk and Fourway) were used. Mahlberg et al. (1984) reported a glandular secretory system formed by three different forms of glandular trichomes on the epidermis of the outer surface of bracts from female flowers in C. sativa. The identification of non-glandular trichomes was also reported. A higher content of cannabinoids was detected in capitate-stalked glands than in capitate-sessile glands and appeared to be related to the gland age and type of cannabis plant. Bulbous glands are the smallest and there is no direct evidence for the presence of cannabinoids in them yet.

Olivetolic acid, an alkylresorcinolic acid, is the first precursor in the biosynthesis of pentyl-cannabinoids, and the identification of methyl- (Vree et al., 1972), butyl- (Smith, 1997) and propyl-cannabinoids (Shoyama et al., 1977) in cannabis plants suggests the biosynthesis of several alkylresorcinolic acids with different lengths of side-chain moiety (Figure S4). It is known that the activated fatty acid units (fatty acid-CoAs) act as direct precursors that form the side-chain moiety of alkylresorcinols (Suzuki et al., 2003). Probably, more than one PKS-forming alkylresorcinolic acid or pyrone co-exist in cannabis plants. An analysis of cannabinoid content from our plant material showed the presence of THCA, a pentyl-cannabinoid, and THVA, a propyl-cannabinoid, in female flowers (Flores-Sanchez and Verpoorte, 2008a). As in H. lupulus, the presence of these PKSs could yield differences in the concentration of cannabinoids into becoming different varieties of C. sativa. Thus, the biochemical characterization of PKSG1, PKSG2, PKSG4, and PKSG5 will be carried out in the future in order to determine which PKS(s) is (are) involved in the pentyl-cannabinoid and/or propyl-cannabinoid pathways.