Two bi-functional cytochrome P450 CYP72 enzymes from olive (Olea europaea) catalyze the oxidative C-C bond cleavage in the biosynthesis of secoxy-iridoids - flavor and quality determinants in olive oil

Olive (Olea europaea) is an important crop in Europe, with high cultural, economic, and nutritional significance. Olive oil flavor and quality depend on phenolic secoiridoids, but the biosynthetic pathway of these iridoids remains largely uncharacterized. We discovered two novel, bi-functional cytochrome P450 enzymes, catalysing the rare oxidative C-C bond cleavage of 7-epi-loganin to produce oleoside methyl ester (OeOMES) and secoxyloganin (OeSXS), both through a ketologanin intermediary. Although these enzymes are homologous to the previously reported Catharanthus roseus Secologanin Synthase (CrSLS), the substrate and product profiles differ. Biochemical assays provided mechanistic insights into the two-step OeOMES and CrSLS reactions. Model-guided mutations of OeOMES changed the product profile in a predictable manner, revealing insights into the molecular basis for this change in product specificity. Our results suggest that, in contrast to published hypotheses, in planta production of secoxy-iridoids is secologanin independent. Notably, sequence data of cultivated and wild olives, points to a relation between domestication and OeOMES expression. Thus, the discovery of this key biosynthetic gene suggests a link between domestication and secondary metabolism, and could potentially be used as a genetic marker to guide next-generation breeding programs.

weight of the fruit (Amiot et al., 1986;Ryan et al., 1999). Oleuropein strongly influences the 48 quality of olive products because of its bitterness, which is a desirable trait in olive oil, though 49 table olives must be subjected to a lengthy curing process to degrade this compound (Garrido-50 Acquity UPLC® C18 column (2.1x50 mm, 1.7 µm, 100 Å; Waters) at 40 °C with a gradient 137 from water (A) to acetonitrile (B), both modified with 0.1% formic acid, using two different 138 programs, both at a flow of 0.3 mL/min and redirecting to waste the first 90 seconds. For 139 exploration of reaction products, an 11 minute gradient was used, starting at 1% B for one 140 minute, then linearly increasing to 20% B in five minutes, then to 40% B in two minutes, and 141 finally to 99% B in one minute, and kept for two minutes at 99% B. The column was 142 equilibrated in 1% B for another two minutes before the next injection. For the kinetic 143 experiments, the gradient was shortened to 4.5 minutes, starting at 1% B for 0.5 minutes, then 144 linearly increasing to 20% B in 2.5 minutes, then 25% B in 0.5 minutes, and immediately 145 changing to 99% B, which was kept for one minute. The column was equilibrated in 1% B for 146 one minute. 147 Acquisition of MS data was performed using the same conditions for exploration and kinetics, 148 but the latter with no fragmentation. Ionization was performed via pneumatic assisted Electro 149 Spray Ionization in negative mode (ESI-) with a capillary voltage of 3.5 kV, an end plate offset 150 of 500 V, and a nebulizer pressure of 3 bar. Nitrogen gas was used as frying gas at 350°C and 151 a flow of 11 L/min. For the kinetics experiments, acquisition was performed at 2 Hz with no 152 fragmentation, at a mass range from 250 to 1000 m/z. For exploration experiments, acquisition 153 was done at 12 Hz following a mass range from 250 to 1000 m/z, with data dependent MS/MS 154 without active exclusion. Fragmentation was triggered on an absolute threshold of 400, and 155 acquired on the most intense peaks using a target intensity of 2x10 4 counts, with a MS/MS 156 spectra acquisition of 12 Hz, and limited to a total cycle time range of 0.5 s, and a dynamic analysis, as it was the only one with mappings of more than 70% (Fig. S13). Of the 112 202 samples, 103 had a mapping greater than 70 %, having a total of 2x10 9 mapped reads, an 82% 203 mapping total (Fig. S13). 204

Iridoid standards 205
Loganin and secologanin standards were purchased from Sigma-Aldrich, and secoxyloganin 206 was obtained by incubating secologanin and CrSLS microsomes overnight, with five NADPH 207 equivalents. 8-epiloganin as well as oleoside methyl ester were purchased from AnalytiCon 208  showed high co-expression with the olive Iridoid Synthase (r=0.76 and r=0.95, respectively). 237 OeSLS1-3 were cloned, heterologously expressed in yeast and the subsequent yeast 238 microsomes were assayed for biochemical activity. During amplification of these genes from 239 olive cDNA, it became evident that OeSLS2 had a variant, OeSLS2A, with 99.8% identity to 240 OeSLS2. Subsequent enzyme assays indicated that OeSLS2A had identical biochemical 241 activity as OeSLS2, and was not investigated further. 242

Biochemical functions of SLS homologues 243
Oxidative cleavage of the iridoid scaffold is required to generate the secoiridoids. OeSLS3 turned over ketologanin; OeSLS2 oxidatively cleaved ketologanin to oleoside methyl 268 ester (OME) and OeSLS3 oxidatively cleaved it to secoxyloganin (Fig. 3a). We thus renamed 269 OeSLS2 Oleoside Methyl Ester Synthase (OeOMES) and OeSLS3 Secoxyloganin Synthase 270 Given that CrSLS can perform the sequential oxidation of loganin to secologanin to 272 secoxyloganin, we next investigated whether the olive SLS homologues were also capable of 273 performing two sequential oxidation steps. We synthesized a mixture of loganin diastereomers 274 where peaks corresponding to OME and secoxyloganin, respectively, were detected (Fig. S3). 277 Ketologanin was also detected in OeOMES and OeSXS reactions, and also, interestingly, when 278 CrSLS was reacted with 7-epi-loganin. In these enzymatic reactions, an isomer of ketologanin 279 was observed along with the ketologanin product (Fig. S3), although this product could not be 280 isolated in sufficient quantity for characterization. OeSLS1, as with the other substrates, was 281 largely inactive, but did produce trace amounts of ketologanin. These enzyme assays were then 282 repeated with purified 7-epi-loganin, which confirmed the observation that OeOMES and 283 OeSXS produce OME and secoxyloganin, respectively, while CrSLS produces ketologanin 284 (Fig. 3b). Although only trace amounts of ketologanin were detected in long incubations with 285 7-epi-loganin and OeOMES and OeSXS, its presence in the early time points of the kinetic 286 reactions supports its role as an intermediary. 287 We next tested 7-deoxyloganin to assess whether these enzymes could catalyze the three step 288 oxidation: 7-deoxyloganin to 7-epi-loganin to ketologanin, and then finally to either OME or 289 SXS. When incubated with 7-deoxyloganin, CrSLS produced loganin, subsequently oxidizing 290 it to secologanin and secoxyloganin (Fig. S4). Notably, C. roseus has a dedicated hydroxylase 291 (7DLH) that converts 7-deoxyloganic acid to loganic acid that has been validated by gene 292 silencing (Salim et al., 2013), so it is not clear whether CrSLS turns over 7-deoxyloganin in 293 planta. OeOMES oxidized 7-deoxyloganin to a product that could not be isolated in sufficient 294 quantity for characterization. However, since only traces of ketologanin and OME were 295 observed in this reaction, we assume that 7-deoxyloganin is not on pathway (Fig. S4). OeSXS was also found to produce this molecule, and OeSLS1 showed no activity. No activity was 297 observed when the microsomes were incubated with 7-deoxyloganic acid, 7-deoxy-8-298 epiloganic acid, and 7-deoxy-8-epi-loganin. 299

Steady-state kinetics of OeOMES and OeSXS 300
Kinetic parameters were calculated for OeOMES and OeSXS, incubated with either 301 ketologanin and 7-epiloganin (Fig. S5). If enzyme regeneration is faster than the oxidation 302 reaction, conditions that should be fulfilled by adding excess NADPH, the kinetic parameters 303 for the ketologanin substrate should follow Michaelis-Menten kinetics (Michaelis & Menten, 304 1913). The exponential fit obtained (Fig. S5a,c) suggests that this assumption is reasonable, 305 and can be used as a starting point to measure the kinetics of the consecutive reaction from 7-306 epiloganin. The subject of the dissociation of intermediates in consecutive reactions catalyzed 307 by cytochrome P450 enzymes remains contentious, but the kinetics of either case do not 308 practically differ under a quasi-steady-state assumption (Notes S1 Eq.S1.7 and Notes S2 Eq. used to calculate Fig. S5b,d, can be seen in Fig. S6 for OeOMES (Fig. S6a) and OeSXS (Fig.  318 S6b). 319 The relative turnover  (Table 1). These kinetic analyses indicate that OeOMES 321 is more specific towards 7-epi-loganin and the Vmax for oxidation of epiloganin to ketologanin 322 is twice as fast than the oxidative cleavage to OME. This explains the non-Michaelis-Menten 323 kinetics seen in the individual velocities (Fig. S6a), where as there is more substrate presence higher than the rate of ketologanin formation (Fig. S6b). 331

Mutational analysis of OeOMES 332
Structural information for plant cytochrome P450s is limited, which hampers rational 333 mutational analysis. Nevertheless, we decided to explore the effects of amino acid substitution, 334 in an attempt to identify the molecular basis of substrate and/or product specificity. Since 335 OeOMES can produce both secoxyloganin (from secologanin) and oleoside methyl ester (from 336 ketologanin), we used OeOMES as a template to introduce amino acid mutations. Using the substrate and iron, named A-E, we chose the amino acids that were not conserved among 347 OeOMES, OeSXS, CrSLS and CAC CYP72A612. We substituted all these amino acids in each 348 region of OeOMES with the corresponding residues of OeSXS, yielding six mutant designs 349 (Table S1). Each of these mutants was expressed in yeast, and the resulting microsomal 350 fractions were incubated with ketologanin, 7-epiloganin and secologanin ( Fig. 4; Fig. S8). 351 Wild type OeOMES produces OME from 7-epi-loganin and ketologanin, with only trace 352 amounts of secoxyloganin being formed. Production of secoxyloganin from ketologanin and 353 7-epiloganin increased in three mutants (Fig. 4): mutants in region B (4 amino acids switched), 354 which is close to the Fe-ligand interaction; mutants in the region C (4 amino acids), a loop and 355 sheet close to the substrate; and the fully switched mutant (22 amino acids). The latter had the 356 highest production of secoxyloganin of the mutants consistently producing 75% as much 357 secoxyloganin as OME from both 7-epiloganin and ketologanin, along with a decrease in OME from 7-epiloganin (Fig. 4b), which is ameliorated when fed ketologanin (Fig. 4a). Similarly, 364 the mutant in region E (3 amino acids), a loop physically close to region A and the substrate, 365 when fed ketologanin produces as much OME as the wild type, but not when incubated with 366 epi-loganin (Fig. S8). Mutations in regions D (6 amino acids), the heme-coordinating region, 367 respectively, had no effect in any of the tested activities (Fig. S8, S9). Secologanin was also 368 tested as a substrate to monitor overall oxidative activity. All mutants, regardless of the 369 observed product profile with ketologanin, efficiently produced secoxyloganin from 370 secologanin (Fig. S9). 371

Analysis of expression data of olive iridoid pathway 372
To further explore the in planta properties of OeOMES and OeSXS, we analyzed the extensive 373 publicly available expression data for olive. We separated the data in two sub groups: a study 374 with OeOMES (r = 0.67) and OeSXS (r = 0.58; Fig. S10). Interestingly, OeISY is negatively 379 correlated (r = -0.77) with OeSLS1, providing further evidence that OeSLS1 does not have a 380 role in iridoid related metabolism (Fig. S10). The studies showed that the only stressors that 381 increase OeOMES expression are in roots (Picual cultivar) 24h after wounding 382 (PRJNA256033), and in roots (Frantoio cultivar) two days after a challenge with Verticillium 383 dahliae (Fig. S11). Detailed experiments have measured expression in inflorescence tissue of leading to the major secoiridoid scaffold in olive, oleoside methyl ester (OME), the penultimate 394 step of oleuropein biosynthesis. Previously, reported feeding studies in olive failed to identify 395 the biosynthetic intermediates that lead to OME. These biochemical studies now show that 396 OeOMES converts 7-epi-loganin via a two-step oxidation to form OME. We also identified a 397 variant of this enzyme that catalyzes secoxyloganin biosynthesis, a metabolite that is also 398 present in olive. Unexpectedly, secoxyloganin is also derived from 7-epi-loganin, and in O. 399 europaea is not synthesized via secologanin, as we anticipated. This discovery highlights that 400 secologanin synthase homologues have evolved to catalyze several oxidative cleavage 401 reactions, as summarized in Fig. 6. Co-expression analysis shows that OeISY, OeOMES and 402 OeSXS are co-regulated, while OeSLS1 is negatively correlated to OeISY. Along with the fact 403 that we detected no activity for OeSLS1, this supports the notion that OeSLS1 is not involved 404 in iridoid biosynthesis. The kinetic experiments have shown that the oxidative cleavage of 405 ketologanin by OeOMES is slightly slower, compared to ketologanin formation from 7-406 epiloganin by the same enzyme. Overall, however, the catalytic efficiencies for the two 407 enzymes for both substrates are similar. Since only Oleaceae and select members of the 408 7-epiloganin to ketologanin. Notably, although the olive enzymes can turnover secologanin, 426 we found no olive cytochrome P450 that generates secologanin, and there are no reports of 427 secologanin in olive. Unexpectedly, OeOMES, which synthesizes the oleoside methyl ester, is 428 able to oxidize secologanin. Given the lack of secologanin in olive tissues, it is likely that the 429 ability of OeOMES and OeSXS to turnover secologanin to secoxyloganin is not a 430 physiologically relevant activity. Similarly, neither 7-deoxyloganin, 7-epiloganin nor 431 ketologanin have been detected in C. roseus, so the capability of CrSLS to turnover 7-432 epiloganin and 7-deoxyloganin is also likely not relevant in planta. 433 Although not all of these oxidation reactions may be physiologically relevant, they can shed 434 light into the mechanism of these enzymes. We speculate that the stereochemistry of the 435 hydroxyl group in position 7 of loganin determines the orientation of the substrate in the CrSLS 436 binding site, such that the hydrogen of C10 can be abstracted by the iron cofactor, which 437 ultimately leads to secologanin (Yamamoto et al., 2000;Fig. 7a). Formation of secoxyloganin 438 by OeSXS could follow a similar mechanism (Fig. 7b), with the abstraction of the hydrogen of 439 C10 ultimately leading to the ring opening. Conversely, when 7-deoxyloganin binds to CrSLS, 440 we hypothesize that the substrate binds such that the hydrogen of carbon 7 reacts with the iron 441 cofactor, leading to the formation of loganin (7S-OH) (Fig. S12b). Similarly, it appears that 7-442 epiloganin (7R-OH) also binds with the C7 H available to the iron cofactor, resulting in 443 formation of ketologanin (Fig. S12c). 444 It is likely that the oxidative cleavage of ketologanin by OeOMES follows a different 445 mechanism than CrSLS. One mechanistic scenario would involve the abstraction of the 446 hydrogen from Carbon 8 followed by rearrangement of the radical to Carbon 9, which could 447 then open to form OME. Alternatively, the hydrogen of C9 could be abstracted directly (Fig.  448   7c). Notably, some of the mutants accumulated a side product, which although could not be 449 isolated in sufficient quantity, displayed an m/z corresponding to a hydroxylated ketologanin 450 derivative. We hypothesize that in these mutants (regions B and C, Fig. 4) one of these radical 451 intermediates is quenched with a hydroxyl moiety to form this side product. 452 The mutational study of OeOMES provides some insights into which regions of the enzyme 453 control substrate binding. Swaps of region A and region E led to a significant decrease in the 454 oxidation, so these may be the major regions responsible for substrate coordination. The data that support the findings of this study are publicly available or available from the 494 corresponding author upon reasonable request. 495 Tables 677 Table 1. Kinetic parameters of OeOMES and OeSXS. Values are shown as the calculated 678 regression coefficient plus/minus the standard error. The relative turnover is the ratio of kcat and 679 the relative specificity is the specificity constant (kcat/KM) of epiloganin with respect to 680 ketologanin.