A sequence similarity network (SSN) was generated using the EFI-EST tool (Gerlt et al., 2015) using the protein sequence from C. purpurea cloA (Cpur) as the input and the number of retrieved sequences was limited to 5,000 sequences. An alignment score of 160 corresponding to 50% sequence identity was chosen for the generation of the SSN, and the SSN was visualized using Cytoscape, a software platform for visualizing complex networks (Shannon et al., 2003). Nodes were filtered by using text filters “Clavine” and “cloA” and “Claviceps,” followed by expanding the network to include the node’s first neighbors. The network was further refined, by removing uncharacterized nodes and nodes sharing identical amino acid sequences. A total of 15 nodes were selected for gene synthesis.

Selected orthologs of wild-type CloA were codon optimized for expression in S. cerevisiae, synthesized (GenScript, United States), and cloned into a pYES2/CT yeast expression vector (Invitrogen, United States) between the BamHI and XhoI restriction sites. Chimeric mutants of cloA and alanine scanning mutants were PCR amplified using primers listed in Supplementary Table S1 and assembled into pYES2/CT yeast expression vectors, between the BamHI and XhoI restriction sites using Gibson assembly (New England Biolabs, United States) following the manufacturer’s directions. Cloned expression vectors harboring the chimeric cloAs were sequenced using GAL1 Promoter and CYC1 Terminator primers.

Expression vectors with the cloA genes were transformed into the strain BJ2168 (ATCC® 208,277) using a modified method that was previously described (Gietz and Schiestl, 2007). Individual colonies of transformed cells were inoculated into 4.5 ml of Synthetic Complete-Uracil (SC-Ura) 0.1% (w/v) glucose and incubated at 30°C for 30 h. After 30 h, yeast cells were induced with 500 μL of 20% (w/v) galactose supplemented with 10 mM FeCl₃ and 10 mM 5-aminolevulinic acid (5-ALA) to a final concentration of 2% galactose (w/v), 1 mM FeCl₃ and 1 mM 5-ALA. Cells were incubated at 25°C for another 24 h for protein expression. Post-induction, 998 μL of induced cells were transferred to a 96 deep well block and 2 μL of 2.1 mM agroclavine (Chiron AS, Norway) was added to a final concentration of 4.2 μM. Cells were incubated at 25°C for another 24 h. Cell suspension was spun down at 4000 rpm for 10 min and 200 μL of the supernatant were filtered through a 96 well MultiScreen® Solvinert filter plate.

Samples were separated using an Agilent InfinityLab Poroshell 120 EC-C18 column with the dimensions of 2.1 mm × 100 mm, 1.9 μm particle size coupled to an Agilent 1290 Infinity liquid chromatography system. All solvents used were LC-MS grade, mobile phases used consisted of (A) water with 0.1% (v/v) formic acid and (B) acetonitrile with 0.1% (v/v) formic acid at a flow rate of 0.2 ml∙min⁻¹. Sample injection volume was 5 μL, and separation was carried out using a linear gradient starting from a pre-equilibrated solvent ratio of 95:5 (A:B) to a solvent ratio of 5:95 (A:B) over 6 min, the column was washed with a solvent ratio of 5:95 (A:B) over 2 min before equilibrating to a solvent ratio of 95:5 (A:B) for another 1 min.

Mass spectrometry was carried out using an Agilent Technologies 6550 iFunnel Q-TOF with a Dual AJS electrospray ionization source with the following conditions: gas temperature of 200°C, drying gas 15 L∙min⁻¹, nebulizing gas pressure of 50 psi, sheath gas temperature of 350°C, a sheath gas flow of 11 L∙min⁻¹, capillary voltage of 2,500 V, and a nozzle voltage of 2,000 V.

Targeted MS/MS was performed using fixed polarity (positive), from the eluent beginning from 1 min into the run. Instrument parameters were set to run at; source gas temperature and flow of 200°C and 10 L min⁻¹, sheath gas temperature and flow of 350°C and 10 L min⁻¹, nebulizer pressure at 50 psig. Capillary and nozzle voltages were set to 4,000 V and 0 V, respectively. MS1 was set to a mass range of 40–1,000 m/z, at a scan rate of 3 spectra/s. MS2 was set to a mass range of 40–1,000 m/z, at a scan rate of 6 spectra s⁻¹ using a fixed collision energy of 20 eV. The targeted masses were set for: (1) 239.1543 m/z, narrow isolation width (1.3 amu); (2) 269.1285 m/z, narrow isolation width (1.3 amu); and (3) 255.1492 m/z, narrow isolation width (1.3 amu).

Determination of compound identities was performed by the comparison of retention time and MS/MS product ion spectrum against commercially obtainable standards where available (D-lysergic acid; Chiron; agroclavine; Chiron/Toronto Research Chemicals; elymoclavine; Toronto Research Chemicals; lysergol; Toronto Research Chemicals).

To probe the mechanism of the two-step oxidation reaction of CloA (Figure 1), an SSN was generated using the EFI-EST sequence BLAST function, using the protein sequence from Cpur CloA as the input and the number of retrieved sequences was limited to 5,000 sequences. In an SSN, protein sequences that are highly similar share a high degree of interconnectivity and form clusters. A user can infer the function of an uncharacterized enzyme by associating it with enzymes of known function in the same cluster (Gerlt et al., 2015).

The full SSN of CloA comprised 4,998 nodes and had 37,424 edges; text filters “Clavine” and “Claviceps” were used to select nodes that contained these descriptions. We expanded the selection to include the node’s first neighbors resulting in four clusters comprising of 77 nodes and 1,042 edges. Uncharacterized and redundant sequences were removed to refine the network to 34 nodes and 306 edges (Figure 2). Out of the 34 nodes, 15 genes with the following UniProt annotations were selected for study: M1WEN7 (Cpur), G8GV80 (Cpas), S5SWI2 (Nlo), E9F392 (392), E9EAT5 (AT5), A8C7R4 (C7R4), R9VXN6 (XN6), R9W253 (253), G9FM49 (9Hypo), T0KET3 (ET3), T0KX90 (X90), M7UEF9 (BOTF1), E9DSJ7 (SJ7), M1W6N5 (W6N5), and M1VZB1 (ZB1). A detailed description of the orthologs can be found in Supplementary Table S2.

A major concern for the overexpression of functional recombinant enzymes, such as cytochrome P450s, is the depletion of essential cofactors, such as heme and iron, which are essential components of the active site. To examine the effect of these cofactors on the quality of CloA expression, we performed a small-scale expression screening with Cpur CloA. We supplemented the medium with 5-aminolevulinic acid (5-ALA), the first intermediate in the biosynthesis of heme, and/or iron (III) chloride (FeCl3) during induction with galactose. Prior to screening the full complement of CloA orthologs, we wanted to determine if these supplements affected the activity of CloA.

Supplementing the media with FeCl₃ alone, did not improve production levels of Cpur CloA and showed comparable production levels of elymoclavine and lysergic acid with samples induced with galactose alone (Figure 3; Table 1). The addition of 5-ALA to the induction reaction mixture almost doubled the lysergic acid yield from 109.3 ± 7.8 μg∙L⁻¹ OD₆₀₀⁻¹ in samples induced with galactose alone to 195.6 ± 7.3 μg∙L⁻¹ OD₆₀₀⁻¹. The addition of 5-ALA also increased the yield of elymoclavine from 18.5 ± 2.8 μg∙L⁻¹ OD₆₀₀⁻¹ in samples induced with galactose to 121.6 ± 7.7 μg∙L⁻¹ OD₆₀₀⁻¹. A further increase in lysergic acid production was observed when both FeCl₃ and 5-ALA were supplemented together into the induction mix to yield a concentration of 327.3 ± 33.5 μg∙L⁻¹ OD₆₀₀⁻¹, almost three times more than induction with galactose alone.

The observed improvement of CloA activity upon the addition of both 5-ALA and FeCl₃ during induction of expression suggested that both 5-ALA and FeCl₃ are limiting during the expression of CloA in S. cerevisiae. Heterologous expression of CloA under the control of the strong inducible GAL1 promoter, can impose cellular stress by rapidly depleting intracellular heme levels, leading to increased protein misfolding and increased proteasomal degradation (Table 2).

Heme depletion has been identified and attributed to the overexpression of P450 monooxygenases in S. cerevisiae (Michener et al., 2012). An increase in the amount of theophylline produced from caffeine when caffeine demethylase P450 was expressed with heme biosynthesis enzymes. This suggested that heme depletion was a limiting factor that affected the activity of heme containing enzymes when expressed in S. cerevisiae. Increased levels of proteasomal activity were observed in cultures expressing P450 from a high copy plasmid, and proteasomal activity decreased when the same P450 was expressed, with genes that increase heme biosynthesis (Michener et al., 2012). Taken together, this suggested that without the necessary components to form a holoenzyme (with the prosthetic group), the apoenzymes (without the prosthetic group) will be degraded, leading to a lower enzyme activity.

While the proteasomal activity of strains expressing CloA was not examined in this study, the expression of CloA under a strong promoter without supplementation could result in increased enzyme degradation. This could explain the lower production levels of Cpur CloA activity when induced with galactose alone, as well as the increased production levels in the presence of 5-ALA and FeCl₃. Expression of CloA orthologs will be induced together with these supplements, for the rest of this study.

Applying the optimized expression conditions, we screened the 15 selected orthologs (Supplementary Figures S1, S2) and found that 8 of the 15 displayed activity on agroclavine. More interestingly, of the 8 active orthologs, 5 were able to produce lysergic acid and elymoclavine, while the remaining three were limited to producing elymoclavine. To gain insight into the differences in product profiles of the eight active CloA orthologs, we created a multiple sequence alignment. Pairwise sequence alignment of these sequences, which showed activity on agroclavine, revealed sequence identities ranging from 58% to 99% (Figure 4A). The highest pairwise sequence identity, between a single oxidation reaction ortholog and a double oxidation reaction ortholog, was between AT5 CloA and 9Hypo CloA. These two orthologs shared a 73% sequence identity. Notably, a gap of 11 amino acid residues was observed in the single oxidation producing orthologs AT5 CloA when its sequence was aligned to the other active CloAs (Figure 4B).

A chimeric AT5 CloA was constructed by inserting the sequence of 9Hypo CloA between the conserved residue pairs tryptophan-leucine (WL) and isoleucine-proline (IP). This chimeric AT5 CloA was termed AT5 9Hypo CloA and was screened for activity in the presence of the substrate agroclavine. The chimeric AT5 9Hypo CloA was now able to catalyze the second oxidation reaction (Figure 4C). The [M + H]⁺ mass of 255.15 m/z, which corresponded to the elymoclavine standard, and the [M + H]⁺ mass of 269.13 m/z, which corresponded to the lysergic acid standard, was observed in the extracted ion chromatograms of AT5 9Hypo CloA (Supplementary Figure S2).

The production levels of lysergic acid by this chimeric AT5 9Hypo CloA were measured to be 2.62 ± 0.235 μM∙OD₆₀₀⁻¹. This production level of lysergic acid was about 15 times higher compared to the wildtype 9Hypo CloA, which produced 0.174 ± 0.014 μM∙OD₆₀₀⁻¹ of lysergic acid (Figure 4C; Table 2).

The production level of elymoclavine in AT5 9Hypo CloA was 0.113 ± 0.005 μM∙OD₆₀₀⁻¹. This production level was lower compared to wildtype AT5 CloA (2.75 ± 0.098 μM∙OD₆₀₀⁻¹) and wildtype 9Hypo CloA (0.174 ± 0.014 μM∙OD₆₀₀⁻¹). This decrease in elymoclavine production observed in the chimeric AT5 9Hypo CloA is likely attributed to the gain of function of the enzyme, as the chimeric enzyme was now able to utilize elymoclavine produced from the first oxidation reaction, as the substrate for the second oxidation reaction to produce lysergic acid.

There are no known CloA structures deposited into the PDB. A protein modeling approach was used as an attempt to understand why the inserted sequence in AT5 9Hypo CloA enabled the enzyme to catalyze the second oxidation reaction. The Phyre2 web server is a combination of several software suites that aims to provide the user with a friendly interface for protein modeling and bioinformatic analysis (Kelley et al., 2015). The Phyre2 webserver was used to generate a list of homology models of CloA that are based on existing structures in the PDB.

The top 10 templates of AT5 9Hypo CloA are summarized in Supplementary Table S3 and each template is accompanied by two important information, the confidence level, and the sequence identity, respectively. The confidence level (0% to 100%) represents the probability that the query is homologous to the template while the sequence identity is the proportion of the query residues that are identical to the template.

It was observed that the top 10 templates of AT5 9Hypo CloA share a very high confidence level (>90%), but very low sequence identity (<20%). This suggested that the core of the model or the relative positioning of the secondary structures are modeled with a high degree of confidence (RMSD 2 to 4 Å). However, the low sequence identity suggested that the models may show substantial deviations in the loops and non-core regions and any structure–function relationships derived from these templates would still require experimental validation.

Of the models generated by Phyre2, the models of AT5 9Hypo CloA that was based on the rat mitochondrial P450 24A1 (PDB ID: 3K9V) template stood out and showed (1) a clear substrate access channel and (2) a cleft on the surface of the enzyme bounded by the 11 inserted residues in AT5 9Hypo CloA (Supplementary Figure S5). These observations were not apparent in the other models of PDB ID: 1TQN and 4FDH, which each shared a 19% sequence identity and PDB ID: 3NA0 which shared an 18% sequence identity to AT5 9Hypo CloA. As such, the homology models of both wildtype and AT5 9Hypo CloA were based on PDB ID: 3K9V and chosen for downstream analysis and experimental validation.

In the Phyre2 model structures of AT5 9Hypo CloA, residues Y244 to P248 form a short loop, while residues M249 to L253 form a short G’ helix. Examination of the surface representation of the inserted region in both the models of AT5 CloA and AT5 9Hypo CloA, revealed a cleft on the surface of AT5 9Hypo CloA. This cleft was bounded by the residues in the F-G loop and a tyrosine residue Y134 that was orientated toward the exterior surface of the enzyme (Figure 5). The Autodock Vina function of Chimera was used to dock elymoclavine into this region and it was observed that elymoclavine fits into this cleft with a score of-7.3 kcal∙mol⁻¹, suggesting a possible elymoclavine binding site at the surface of the enzyme (Figure 5-inset).

An alanine mutagenesis was performed on Y134 for both AT5 CloA and AT5 9Hypo CloA, respectively, to examine its role in the two-step oxidation of agroclavine. Interestingly, an alanine substitution in this residue in both AT5 CloA and AT5 9Hypo CloA resulted in a drastic change in the oxidation profiles. In AT5 CloA, the alanine mutant showed no turnover of agroclavine to elymoclavine, while the alanine mutant in AT5 9Hypo CloA showed a complete loss of lysergic acid production and only produced elymoclavine (Figure 6; Table 3).

The region between the F and G helices has been extensively observed in crystal structures and in molecular dynamics simulations to interact with substrates and the lipid bilayer membrane (Šrejber et al., 2018). In structural studies of camphor bound CYP101D2, the electron densities corresponding to multiple molecules of camphor were found in the camphor-soaked crystal structure. The electron density of one of these camphor molecules was found in a cavity on the F-helix side of the F-G loop, suggesting a multi-step binding mechanism whereby the substrate binds to a recognition site on the protein surface in the proximity of the F-G loop before entering the active site (Yang et al., 2010).

We hypothesize that Y134A in AT5 CloA could be essential for forming the early interactions with agroclavine at the surface of the enzyme, and loss of the Y134 side chain could result in a weaker interaction and a mutant that is unable to recognize or productively bind agroclavine within the active site of the enzyme. It is also hypothesized that Y134A in both AT5 CloA and AT5 9Hypo CloA may play a similar role in substrate recognition and binding. The difference, however, is that the extended F-G loop formed by the 11 inserted residues in AT5 9Hypo CloA may form interactions that is sufficient for the recognition of agroclavine but insufficient for the recognition of elymoclavine, which may help to explain the loss of the second oxidation reaction in this mutant. Further studies performed with the experimentally-determined structure of CloA would help to provide greater insight into this observation.

To examine if the interaction is sequence specific, we further screened additional chimeric AT5 CloA mutants by inserting the 11 residues from the other CloA sequences onto the same AT5 CloA scaffold (Figure 7A). These new AT5 CloA chimeric mutants were tested for activity in the presence of agroclavine.

Interestingly, it was observed that these new AT5 CloA mutants were also able to perform a two-step oxidation reaction to produce lysergic acid (Figure 7B). The yields of each chimeric AT5 CloA mutant are summarized in Table 4. The best producers of lysergic acid between the chimeric AT5 CloAs were AT5 Cpas CloA (2.678 ± 0.143 μM∙OD₆₀₀⁻¹) and AT5 9Hypo CloA (2.62 ± 0.235 μM∙OD₆₀₀⁻¹). Compared to the best lysergic acid producer Cpur CloA (2.23 ± 0.350 μM∙OD₆₀₀⁻¹), these chimeric mutants exhibited an increase in the percentage of lysergic acid to the overall detected ergot alkaloids (Figure 7C): AT5 Cpas CloA produced 98.03% of lysergic acid and AT5 9Hypo CloA produced 95.86% of lysergic acid as compared to 94.57% of lysergic acid produced by Cpur CoA. While the reasons for this observation are unclear, one possible explanation is that the residues of Cpas and 9Hypo CloA could form a surface topology that forms a more efficient interaction with elymoclavine at the enzyme surface. Solving the crystal structures of both wildtype and chimeric CloAs with substrates or analogs that fully occupy the enzyme’s putative binding sites would provide important insights to better explore this hypothesis.

A second possible explanation is the interaction between the residues in this region and the membrane. Eukaryotic cytochrome P450s are usually membrane-bound enzymes that form complex relationships with both the membrane, and other membrane-bound redox partners (Šrejber et al., 2018). Using a combination of molecular dynamics simulations supported by experimental data, the F-G loop of CYP101D2 has been identified as one region that interacts with the membrane (Baylon et al., 2013). This interaction helps to modulate the depth of membrane insertion and the uptake of substrates from the membrane.

To further examine the effect of loop deletions on catalysis, loop deletion mutants (DeLoop) in wildtype CloA orthologs were created and assayed for activity (Supplementary Figure S6; Supplementary Table S4). It was observed that all the loop deletion mutants of wildtype CloA orthologs were unable to catalyze the oxidation of agroclavine. A corollary expectation is that modifications of the F-G loop region in CloA could affect its association with the membrane and modulate access of the substrate into the active site of the CloA.