3.1 Gene dosage study
Major studies in expression of CYP725A4 in microbial chassis have so far focused on modifications to the P450 and cognate reductase sequences for improved solubility and expression (6), tandem P450-reductase expression (14) and constructing P450-reductase chimeras using strong eukaryotic reductases from human, rat liver or Stevia rebaudiana and cytochrome b5 (7). However, it remains necessary to optimise the CYP725A4 functionality in S. cerevisiae cell factories and therefore, the first question we sought to answer was to determine the rate-limiting gene through a gene dosage study.
The effect of increased CYP725A4 or Taxus POR gene copies was evaluated in a three-day run where BN1 strain, expressing an extra copy of CYP725A4, improved the diterpenoid I concentration significantly (p = 2.5e-02), proposed to be a potential isomer to T5α-ol in our earlier study (14). Also, BN2, expressing an extra copy of POR, showed a significant impact on OCT (p = 8.2e-03), iso-OCT (p = 0.02) and diterpenoid I (p = 6.6e-07) formation. BN3, expressing one copy of CYP725A4 with a low-glucose inducible promoter, pHXT7, also showed significant improvement in OCT (p = 0.0001), iso-OCT (p = 0.002), and diterpenoid I formation (p = 1.3e-03). However, factoring in the impact of interaction between the production time and the strain revealed more detailed results. In more detail, the interaction between strain type and growth time was found to significantly affect the taxadiene (p = 2.6e-08), OCT (p = 0.0015), taxadienediol (dioxygenated taxane; p = 0.003), diterpenoid I (p = 3.9e-08) and iso-taxadiene (p = 7.4e-09) titres. However, this interaction was found insignificant towards T5α-ol and iso-OCT titres (p = 0.06 and 0.06, respectively).
As depicted in Fig. 2, doubling the CYP725A4 gene copy numbers in BN1 did not improve the final T5α-ol in comparison to its parent strain, LRS6. However, OCT concentration was slightly improved by 1.2-fold (2.5 ± 0.3 mg/L, p = 1). Interestingly, diterpenoid I remained the major product in all the new strains with 2.6- and 1.9-fold improvement in its titre in BN1 and BN2, respectively, despite this improvement being insignificant (p = 7.8e-02 and 8.7e-01).
On the other hand, none of the strains showed any enhancement in taxadiene consumption and its titre was increased by 1.5- and 2.4-fold, in BN1 and BN2. Based on the data from Fig. 2, it was clear that the overexpression of CYP725A4 enzyme neither favoured the taxadiene conversion nor it caused any remarkable increase in T5α-ol production, while it improved the production of side oxygenated taxane products. Furthermore, BN2, with one more copy of Taxus POR, showed a decreased final biomass accumulation by 52% (OD600:16 ± 1.7) relative to LRS6. Similar improvement in the side oxygenated taxanes concentrations was initially observed for BN2, however, the final taxadiene concentration in this strain showed 1.5-fold improvement to that of BN1 strain, being 4.6 ± 0.2 mg/L. Despite this, due to the poor performance of BN2 in comparison to BN1, particularly its instability in later experiments and no growth, the next strains were created with focusing on P450 overexpression. In line with our results, the increased POR expression also reduced the CYP725A4 oxygenated taxanes concentration in Biggs et al. (6) study using E. coli, the phenomenon which authors attributed to resource competition, elevated burden and inefficient use of NADPH. They also noted that increasing CYP725A4 expression to POR resulted in higher accumulation of oxygenated taxanes.
Similar to our previous study (14), we did not detect any acetylated taxane formed due to Taxa-4(20),11(12)-dien-5α-olO-Acetyl Transferase (TAT) expression. Despite this, to account for any product changes due to TAT expression, BN3 was created with the 1st copy of CYP725A4 under the control of a different strong constitutive promoter (pHXT7) using taxadiene-producing LRS5. HXT7 gene is known to be repressed during high glucose availability (29), and since the carbon used in this study was galactose, it was deemed that it would produce a stronger effect on overall taxanes and T5α-ol formation. In comparison to BN1, T5α-ol and OCT titres were further increased by around 1.4-fold and 2.0-fold, respectively, regardless of around 1.8-fold increase in taxadiene titre obtained at 72 hours. Despite this, all these changes were found to be insignificant (T5α-ol: p = 1, OCT: p = 0.35 and taxadiene: p = 0.47) compared to BN1 at the end of the cultivation. However, comparing the oxygenated taxanes production by LRS6 and BN3, showed that despite using different promoters, TAT activity likely led to the consumption of all the major oxygenated taxanes products, with 1.4-, 2.3- and 2.3-fold lower T5α-ol, OCT and iso-OCT titres by LRS6, compared to BN3, without TAT expression cassette. Hence, it is possible that the side oxygenated products also play intermediary roles in Taxol® biosynthesis pathway that need to be further investigated using the same promoter and terminator sets for the overexpression. Also, the final biomass accumulation of BN3 was reduced to 17.3 ± 1.5, compared to BN1 and LRS6 with final biomass accumulations of 28.14 ± 2.5 and 22.2 ± 0.7, respectively. Hence, it was hypothesised that potentially the taxanes and/or their producing enzymes were inducing the oxidative stress in the cell, lowering the biomass accumulation.
Using the TRX2 gene promoter, which has been found to be responsive to hydroperoxide stresses (30), BN4 strain was constructed to integrate the 2nd copy of CYP725A4. This strain displayed no growth and therefore, no productivity. Biggs et al. has also separately reported that the higher production of CYP725A4 diterpenoids in E. coli resulted in loss of productivity (6). In their recent follow-up study (31), they attributed this phenomenon to potential membrane stress. It was therefore postulated that the increased abundance of the heterologous products together with the overexpressed P450 were cytotoxic. Furthermore, since the constructed strains did not favour the production of T5α-ol and efficient taxadiene conversion, the approach of multiplexing POR and CYP725A4 was deemed inefficient. The chromatograms of these strains are available in Additional Fig. 2 and the representative mass spectra are presented in Additional Fig. 3.
3.2 Primary screening of self-sufficient strains
The production of the oxygenated taxanes, other than T5α-ol, could be viewed from four different aspects; First, the possibility of inefficient electron coupling and P450-reductase interactions that result in the side products as the electron transfer occurs in two steps (32). Indeed, the optimal electron transfer is one of the most important parameters for improved P450 activity and decreased cellular toxicity due to the electron waste in reactive oxygen species (ROS) formation (33). As all the previous studies had only used eukaryotic PORs and ignored the prokaryotic counterparts, we selected the reductase genes from several natural self-sufficient P450 enzymes, known for their very high turnover number as a result of optimal electron transfer and coupling, and minimal electron leakage (34). As a control, a strain expressing CYP725A4 only (BN5) was also constructed. Interestingly, the strain only expressing the P450 enzyme still produced T5α-ol, OCT and iso-OCT, although at minor quantities of 0.11 ± 0.02, 0.51 ± 0.2 mg, and below the limit of quantification detection for iso-OCT, which represented a reduction of 77.4- and 15.4-fold for T5α-ol and OCT, respectively, in comparison to the strain BNF-1, expressing Taxus POR (Fig. 3). This denoted that the yeast native redox metabolism contributes to the production of oxygenated taxanes.
As shown in Fig. 3A, interestingly, among the self-sufficient strains, only BNF-1, BNF-2 and BNF-7 showed significant differences to BN5 strain in OCT and iso-OCT production. However, except BNF-1, no other strains depicted any significant different results to BN5 in terms of T5α-ol formation. Also, taxadiene and iso-taxadiene titre significantly improved in all CYP725A4-expressing strains opposed to BNF-1 and LRS5 (Fig. 3B); In comparison to taxadiene-producing strain, LRS5, surprisingly, there was a significant increase in taxadiene production by BNF-2 (adjusted p = 0.002), BNF-3 (adjusted p = 0.001), BNF-4 (adjusted p = 0.0004), BNF-5 (adjusted p = 0.002), BNF-7 (adjusted p = 0.04) and BNF-8 (adjusted p = 0.0012). Similarly, iso-taxadiene was significantly improved in BNF-2, BNF-3, BNF-4, BNF-5 and BNF-8 representing 12.3-, 13.9-, 14.7-, 14.8- and 13.9-fold improvement to that of LRS5. Despite this, the final biomass did not significantly change in all strains, except in BNF-1 with 1.45-fold reduction relative to LRS5. By conducting Tukey's HSD post hoc test, a measure of difference across the whole strain groups were revealed. Only BNF-1, BNF-2 and BNF-7 were grouped as significantly different in terms of OCT production, while BNF-1 stood as the most significant in T5α-ol and iso-OCT production.
Interestingly, OCT remained the major product of CYP725A4 in all strains as illustrated in representative chromatograms in Additional Fig. 4. While BNF-2 was constructed using non-codon-optimised E. coli FldA and Fpr genes, these genes have been shown to support the microsomal P450 activity at 1:1 ratio, while being less efficient than S. cerevisiae POR (35). Despite this, the other reason for the relatively poor activity of BNF-2 could be the absence of an effective domain for connecting the Fpr and Fld for optimal NADPH oxidation. Also, it must be noted that the E. coli reductases are neither classified as self-sufficient proteins nor reported to have efficient electron coupling power. On the other hand, solubilising Taxus POR through truncating its transmembrane sequence also did not help in oxygenated taxanes production in BNF-8, since BNF-7 with whole POR sequence produced 85% more oxygenated taxanes. This result was in line with (36), where they proposed that the soluble (truncated) POR could not interact with the P450 and the impact of artificial linker in P450 activity was highlighted. However, this was also in contrast to Biggs et al. study, where they proposed that decreasing POR lipophilicity enabled better CYP725A4-POR interactions and electron transfer in E. coli (6).
Of note, the final biomass accumulations of these self-sufficient strains were not found to be significantly different despite their overall poor performance. However, the taxadiene production was found to be higher by at least 3.9-fold as in BN5 to 6.2-fold in BNF-4 compared to taxadiene-producing LRS5 strain, without significant difference in final biomass accumulation. On the contrary, around 1.2-fold reduction in taxadiene formation was noticed in BNF-1 compared to LRS5. Therefore, BNF-1 expressing Taxus POR was found to be the best performing strain as more products were obtained, clearly showing a lower concentration of substrate taxadiene in the end of the experiments. As the linkers for these strains were selected from previous studies (37–41), it can be argued that the homology modelling for Taxus CYP725A4-POR proteins pair would give better insights on efficient linker design to optimise the enzymatic activity and the substrate consumption. However, except BNF-1, BNF-2 and BNF-7, the use of linker-containing and linker-free constructs both did not improve the production of the oxygenated taxanes. Also, except BNF-1 (p = 0.008), there was no significant difference between final taxadiene concentration of CYP725A4-expressing strain (BN5) and the self-sufficient strains. Hence, it is more likely that sub-optimal expression of prokaryotic reductases with strong coupling capacity (42, 43) caused the suboptimal performance of these strains.
3.3 Tandem expression of Taxus CYP725A4 and POR and the effect of other media on the product spectrum
In the previous section, we showed that the superior coupling capacity of prokaryotic reductase did not favour the CYP75A4 activity in yeast. Therefore, another strain, BN6, was constructed using Taxus cuspidata reductase (POR). As LRS6 strain was constructed using GAL promoters for both CYP725A4 and POR genes, to minimise any delay in electron transfer, POR was expressed under a strong consecutive promoter (pTDH3), while CYP725A4 was expressed using the strong galactose promoter (pGAL1) for BN6 strain. However, to keep CYP725A4 expression stronger than that of POR, stronger terminator (RPL41Bt) and weaker terminator (ICY2t) (44) were used for CYP725A4 and POR, respectively. In YPG medium, BN6 generated the highest amounts of OCT, iso-OCT and T5α-ol at the final concentrations of 14.6 ± 1.4, 12 ± 1.4 and 10.6 ± 1.8 mg/L, respectively, which represented 1.8-, 1.7-, and 1.2-fold improvement compared to BNF-1, expressing Taxus POR fused to CYP725A4. However, taxadiene was almost doubled to be 20.3 ± 0.9 mg/L and the final biomass slightly reduced to 26.5 ± 3.2 by the end of 72 hours of cultivation (Fig. 4). This denotes the cumulative negative effect of Taxus reductase expression and oxygenated taxanes as well as possibly higher allocation of NADPH for non-growth associated tasks in this strain. The higher oxygenated taxanes production compared to self-sufficient strains also denotes that potentially the uncoupling promoted the formation of the oxygenated taxanes. Hence, it is possible that besides the molecular oxygen, the formation of oxygenated taxanes by CYP725A4 was also mediated by other reaction intermediates which acted as the terminal electron and oxygen acceptor(s) (45). On this matter, the 1.73-fold improved taxadiene titre in BN6 and 3.9-fold in BN5 compared to LRS5 might be due to activation of oxidative stress response (31) by uncoupling, and as a measure to combat the GGPP stress (46). This proposal is also justified by slightly lower (0.84-fold) taxadiene accumulation in BNF-1 compared to LRS5. Hence, it is plausible to propose that the increased oxygenated taxanes was due to increased taxadiene production, where the P450-POR reaction intermediates induced the higher taxadiene production, without fully converting it to oxygenated taxanes. Then, that can also be explained by the slow uncoupling rate that led to 1.64-fold lower oxygenated taxanes productivity in BNF-1 (0.55 mg/L/OD600) compared to BN6 (0.90 mg/L/OD600), a phenomenon also reported for other P450-reductase fusions (47). Thus, it can be concluded that the uncoupling is the promoting agent in the oxygenated taxanes production which occurs efficiently through Taxus CYP725A4 and POR interaction. This is also supported by the epoxidation theory (15, 22), where the radical and cationic intermediates favour the formation of epoxides and hydroxides (48). From another perspective, the presence of POR was the key to increased oxygenated taxanes production, as its electron transfer capability led to 38.1-fold higher oxygenated taxanes titre compared to BN5, only expressing CYP725A4. However, as of now, it is unclear how the interaction between CYP725A4 and POR can be optimised to promote the uncoupling for higher oxygenated taxanes production whilst maximising the taxadiene consumption.
Interestingly, the T5α-ol: OCT: iso-OCT ratio remained very similar, 1:0.89:0.81 and 1:1.4:1.1 for BNF-1 and BN6, respectively, by the end of 72-hour cultivation in YPG medium. Since uncoupling is an auxiliary mechanism of activity, its effect on product distribution cannot be inferred through in vivo studies. Instead, as results suggests, and due to unknown reasons, T5α-ol is the later product formed when taxadiene consumption is high and possibly the least affected by the decreased uncoupling rate, given the electron transfer to CYP725A4 is established. Consequently, more optimal fusion protein design can be exploited to improve the P450-POR interaction for higher taxadiene conversion and oxygenated taxanes production. Despite this, due to the superior capability of BN6 for higher oxygenated taxanes production, this strain was used for all the other experiments performed later to answer our other hypotheses.
Our second formulated hypothesis was the cumulative effects of endogenous, underground metabolism (49) on the product distribution of the CYP725A4. Our previous study showed a correlation of taxadiene production with biomass accumulation in LRS5 (12). The latest study by Walls et al. (13) showed that the increased biomass resulted in improved total taxanes concentration. However, a lower dependence of oxygenated taxanes on the biomass proved that the growth and oxygenated taxanes production were not tightly correlated as they reported (Fig. 4). On the other hand, an earlier study by Edgar et al. (15) had illustrated the effect of two rich and minimal media types on the distribution of oxygenated taxanes from CYP725A4 in different microbial systems. Since the potential effect of media type on the product profile was not examined in detail in S. cerevisiae yet, we postulated that selecting a variety of microbiology media might be influential in a change in CYP725A4 product spectrum in S. cerevisiae.
The results clearly indicated that using more enriched medium like TBMG, while promoting the biomass accumulation of BN6 (Fig. 4H) to 52.3 ± 4.4, almost doubled the titre ratio of OCT + iso-OCT to T5α-ol. However, using minimal media like SDG and LBG did not result in any difference in this ratio and the use of BHIG medium only increased this figure by around 1.3-fold. One-way ANOVA analysis also revealed that the medium type only had significant effect on taxadiene (p = 2.6e-05), diterpenoid I (p = 3.4e-06), T5α-ol (p = 0.025) and taxadienediol (p = 6.42e-06) titres, and biomass (p = 4.5e-08) (Figs. 4B, E, F, G, H), but no significant impact on iso-taxadiene, OCT and iso-OCT were detected (Figs. 4A, C, D). Tukey post hoc test further clarified that LBG had the significantly highest T5α-ol titre (14 ± 1.5 mg/L), while using TBMG medium led to the highest concentration of another potential isomer of T5α-ol, here called diterpenoid I (28.3 ± 5.9 mg/L). Interestingly, diterpenoid I was not found in SDG media (Additional Fig. 5). In all media, taxadiene was still produced, where TBMG resulted in its maximal production (40.3 ± 4.8 mg/L) and SDG produced the least taxadiene (15 ± 1.6 mg/L). However, there was no significant difference between LBG, BHIG and YPG in terms of taxadiene and taxadienediol production. Also, in terms of biomass accumulation, YPG, BHIG and SDG were grouped together, while LBG resulted in the least biomass accumulation (13 ± 1.5 mg/L). Comparing based on the productivity, the total OCT, iso-OCT and T5α-ol productivity in BHIG, LBG, SDG, TBMG and YPG were 1.9, 3.8, 1.8, 0.9 and 1.4 mg/L/OD600, highlighting that the promotion of biomass is not tightly correlated to increased oxygenated taxanes production. Despite this, these ratios changed to 1, 2, 0.7, 0.8 and 0.8 mg/L/OD600 for taxadiene, showing that the taxadiene production was coupled to biomass production (12). As SDG is composed of complete synthetic mixture medium, which is enriched in amino acids, it can be postulated that the cellular NADPH was saved in favour of P450 activity. Supportive of this hypothesis, TBMG which contains a double amount of the yeast extract to tryptone (carbon to nitrogen ratio (C/N): 2:1) opposite to YPG (C/N: 1:2) promoted the formation of carbon-containing taxanes backbone. However, the lower nitrogen availability potentially directed the flux towards essential cellular tasks like amino acid biosynthesis than heterologous products. Regardless of the type of medium, OCT remained the major product in all media (Fig. 4C; Additional Fig. 5), and SDG was found superior to YPG for increased oxygenated taxanes production.
3.4 Definitive screening design interpretation
While S. cerevisiae was shown capable of producing early oxygenated taxanes, varying the electron transfer partner or reductase gene dosage did not necessarily help with higher CYP725A4 activity as shown in Figs. 2 and 3. Therefore, as the third hypothesis, it was postulated that the other enzymatic parameters might be important to control, while considering that the improvement of P450 expression does not guarantee a parallel improvement in the product titre (50). Built on our previous results in the gene dosage study, which strengthened the assumption on the promoting effect of uncoupling on oxygenated taxanes formation, it was therefore hypothesised that the induced toxicity might necessitate exogenous sources of cofactors to assist with replenishing the lost resources. Among these factors, the components of the prosthetic groups of both Taxus CYP725A4 and reductase proteins, being heme and flavins, respectively, were selected to be supplemented exogenously in the BN6 growth medium.
Despite our incomplete understanding of the underlying mechanism for heme prosthetic group incorporation into the apocytochrome form of CYP725A4, it was critical to investigate if the flavins were the controlling elements for CYP725A4 optimal activity as a result of improved POR activity. In parallel, to screen for the potential effect of each parameter on the product distribution, a three-level definitive screening design (51) was performed in two-steps. The tested variables included the addition of ready heme (hemin) and iron precursor (δ-Aminolevulinic acid (ALA)) as well as flavin adenine dinucleotide (FAD), flavin mononucleotide (FMN) and their precursors, riboflavin. Cross-validation was then used to divide the data points into multiple subsets for training and validating the data and for improving the model accuracy. Then, the stepwise forward selection procedure was carried out to determine the predictors for each of OCT, iso-OCT, T5α-ol and biomass, where a 5% significance level was used to enter and eliminate each variable at first-order, followed by applying the same filtering procedure for the interaction terms. The prediction of the models and the raw data are included in Additional Figs. 6 and 7. While all these parameters are related in biological pathways, a correlation analysis depicted that although FMN, FAD and Hemin were weakly, but statistically significant correlated, in general, these variables could independently predict the response factor to strengthen our final conclusions (Fig. 5B). The linear model formulas for each response factor are listed in Additional Table 7 and the respective effect plots are illustrated in Fig. 5A.
For all the oxygenated taxanes, ALA was a significant negative predictor (OCT: p-value = 5.30e-05; iso-OCT: p = 2.2e-09; T5α-ol: p = 3.6e-10). However, hemin, which is the heme ferric chloride species (or ferrous protoporphyrin IX) (52), was found to significantly negatively influence the iso-OCT (p = 2.8e-06) and T5α-ol formation (p = 5.17e-10), as a separate factor, but it had a promoting effect in interaction with ALA (Fig. 5A). The negative impacts of iron factors were opposite to their remarkable effect on P450 activity reported in earlier works (52–55). While yeast can regulate the heme biosynthesis effectively, the heme depletion reduces the total active P450 enzymes and increases the protein misfolding, leading to decreased growth rate (50, 54). Additionally, under heme depletion, the cell expression profile resembles that of anaerobic growth, rendering it unfavourable to CYP725A4 monooxygenase, which is in addition to heme synthesis repressing impact of glucose and galactose during the fermentation (56). While hemin is generally found to exert less positive effect on P450 product titre than ALA (36), no further improvement in product titres by this factor might be due to expression not being rate-limiting or the inactive portion of P450 enzyme being higher (50). Alternatively, the supplemented iron might have been used by other iron-sulfur assembly systems in iron trafficking processes, like by iron-sulfur containing enzymes involved in DNA repair and replication to combat the oxidative stress (57). It is argued that the hemin iron complex and ALA can cause lipid peroxidation (58) and ROS generation, respectively (59), resulting in metalloproteins inhibition (60), labile heme release (52), and P450 aggregation (61). This is supported by the model predicted for biomass, which illustrated the negative effect of ALA on final biomass (Fig. 5A). However, our end-point measurement of ROS did not identify any of iron sources being significant predictors for ROS, rendering the cellular-level ROS decreasing the oxygenated taxanes production a weak phenomenon up to now or the iron factors were not involved in a change in the ROS level, at their selected concentrations.
Among flavins, FAD and Riboflavin were found as significant factors (OCT: p-value = 2.4e-08, 1.1e-05; iso-OCT: p = 0.0001, 4.2e-09; T5α-ol: p = 2.6e-05, 5.2e-10) (Fig. 5A). While iron and flavin metabolisms are connected, their relative binding domains in the P450 and reductase proteins interact in electron transfer and oxygen activation (33). It is known that P450s interact with POR through a combination of FMN-domain binding motifs in a specific isoform and FMN domain is a determining factor for P450-POR docking in endoplasmic reticulum (ER) membrane (62). However, we did not find any significant contribution from FMN towards the oxygenated taxanes titres. Despite this, FAD was found influential in improving the final titre of all oxygenated taxanes, particularly at concentrations more than 5 mM, as suggested by the models’ terms coefficients (Additional Table 7). Interestingly, the interaction between FAD and Riboflavin was found as a significant predictor (p = 5.0e-06) for ROS formation, which was in line with previous study by Chen et al. (2020), where FAD was found to reduce the toxicity in S. cerevisiae. We noted that riboflavin exerted a negative impact on oxygenated taxanes production while it had a positive impact on increased ROS level, despite being known for its protecting effect for glutathione redox cycle improvement (63). Due to the shared predictors, it is therefore possible that ROS is likely to have some impact on the overall diminished performance of yeast for oxygenated taxanes production. In line with this conclusion, the positive and negative impact of FAD and riboflavin on biomass accumulation, respectively, is also evident in the biomass model (FAD: p = 5.1e-07; Riboflavin: p = 3.4e-16).
To expand our previous models, it was next essential to understand the importance of each predictor on the overall response parameter (ALA: negative for all diterpenoids; hemin: negative for all diterpenoids; riboflavin: negative for all diterpenoids and biomass; FAD: positive for all diterpenoids and biomass). Hence, a variation partitioning study was performed to examine the contribution of each predictor to the overall variance in each response. As a result, each variable was found significant at p = 0.001 by the ANOVA test, regardless of the effect of other significant predictors. Hence, the importance of each predictor reported in each model was again validated (Fig. 6A-D). Similar to what the model predicted, ALA and hemin interaction were found to only explain a negligible percentage of total variance for both iso-OCT and T5α-ol (0.3%). In this study, we did not observe any strong correlation (OCT: -0.28, p = 4.5e-05; iso-OCT: -0.17, p = 0.01; T5α-ol: -0.17, p = 0.01) between total cellular ROS level and oxygenated taxanes to explain the results, despite them sharing the same predictors of FAD and riboflavin in their regression models. Our variation partitioning model also illustrated that while only two variables could explain the ROS dataset, up to 58% of response variance could not be explained by the defined parameters of the study (Fig. 6E), highlighting the potential impact of other factors that could not be controlled in our study. Therefore, while we do not reject the impact of ROS on oxygenated taxanes production, evidently, further studies are required to quantify ROS by P450-POR (uncoupling) to untangle its contribution to diterpenoid formation more accurately. However, obviously, the molecular-level impact of cofactors on CYP725A4-POR performance is undeniable.
3.5 Effect of increased salt and antioxidant concentrations on CYP725A4 activity
Based on the initial conclusions drawn from the previous sections, our fourth hypothesis was that the Taxus CYP725A4-POR enzymes both induce oxidative stress through uncoupling, whilst protecting against it (64) through mechanism-based inactivation that results in more reactive metabolites (65) from the substrates. These hypotheses were therefore tested by supplementing the BN6 strain growth medium with ascorbic acid antioxidant at different concentration levels, as it was shown to reduce the uncoupling-induced ROS production by cytochrome P450 enzyme (47, 66). Similarly, as ionic strength has been shown to increase the uncoupling rate (67) and to potentially improve the P450 enzyme activity due to enhanced electron transfer (68), similarly, sodium chloride salt was also supplemented at different concentrations in BN6 growth medium (YPG) to study the changes in oxygenated taxanes production.
A one-way ANOVA test revealed that except T5α-ol and taxadienediol, all other factors including OCT (p = 2.6e-05), iso-OCT (p = 0.003), taxadiene (p = 1.04e-05), diterpenoid I (p = 0.02) and biomass (p = 0.01) were significantly affected by the addition of ascorbic acid to the medium (Fig. 7A). The pairwise t-test with the control showed that OCT and taxadiene titres both significantly decreased upon increasing the concentration of ascorbic acid to 2.55 mM and higher (adjusted p < 0.05). Also, iso-OCT concentration was decreased upon adding 3.8 mM ascorbic acid in the medium (adjusted p < 0.05), except that at 5 mM, the iso-OCT concentration was not different to that of control. While we did not observe any significant drop in T5α-ol and taxadienediol titres, the biomass accumulation was not different at any level except when ascorbic acid reached final concentration of 10 mM (adjusted p = 0.03). As OCT was the most affected oxygenated product, it can be concluded that it might be the initial product of CYP725A4 enzyme as mentioned earlier. Since the biomass was not significantly affected by the ascorbic acid addition except at 10 mM concentration, the drop in taxadiene and oxygenated taxanes in other levels cannot be attributed to reduced biomass accumulation as stated earlier. Therefore, at least at 5 mM level, the 1.4-fold decrease in total oxygenated taxanes production (16.7 ± 4.2 mg/L) could be a result of potentially poor enzymatic performance and lower substrate conversion (47). The presence of uncoupling events might support the activity of Taxus CYP725A4 and POR interaction in yeast as in Kells et al. (69) study, where the resulting reactive metabolites potentially act as the oxygen and electron donors to the P450 enzyme in addition to the molecular oxygen (45).
As illustrated in Fig. 7B, NaCl salt did not exert any significant impact on the final concentration of T5α-ol and taxadienediol. However, all other parameters including OCT (adjusted p = 0.001), iso-OCT (adjusted p = 0.013), taxadiene (adjusted p = 0.012), diterpenoid I (adjusted p = 0.03) and biomass (adjusted p = 3.8e-4) were all significantly affected by the addition of sodium chloride salt to the medium. A pairwise t-test comparison to the control level only depicted the significant decrease in OCT titre only at 10 mM NaCl (p = 0.01), and taxadiene concentration at 2.55 mM (adjusted p = 0.03) and 3.8 mM (adjusted p = 0.005) of NaCl in growth medium, without any significant impact at higher levels. Interestingly, except at 3.8 mM, in all other treatment groups, the biomass significantly decreased compared to control level (adjusted p < 0.05). While Tukey post hoc test revealed that the final biomass decreased as a result of increased ionic strength in as little as 1 mM, the overall oxygenated taxanes production per OD600 unit was found to increase by around 1.4-fold upon supplementing the exogenous NaCl concentration up to 5 mM, while a sudden drop to 0.81 ± 0.07 mg/L/OD600 was observed at 10 mM NaCl. The previous studies have reported on improved P450-POR interaction and intra-reductase electron transfer due to enhanced charge pairing and/or protein conformations, as a result of which the P450s performances were improved at increased ionic strengths (35, 36, 62, 70, 71). Alternatively, the ionic strength might have improved the POR structural changes which favoured its electron transfer to CYP725A4 (67, 72, 73). Therefore, the possibility of electrostatic forces also being determining in oxygenated taxanes production of CYP725A4 is likely as it was shown for E. coli cell-surface immobilised human CYP1A2 and POR (36). Despite this, as proposed earlier in section 3.3, the uncoupling seems to play a minor effect on T5α-ol formation. However, further experiments are needed to provide a deeper insight on this point.
3.6 Resting cell assay
Besides the potential uncoupling event by Taxus CYP725A4 and the reductase (POR), the lipophilic nature of taxadiene and its derivatives might increase their intracellular residence, leading to cytotoxicity due to their metabolism, by which the endogenous mechanisms promote their functionalisation and conjugations to render them more polar (65). Therefore, our fifth hypothesis was formulated based on the possibility that the endogenous pathways could contribute to overall CYP725A4 activity in yeast. In detail, it was hypothesised that some transient oxidised metabolites were formed and degraded quickly before being released into dodecane during yeast biphasic culture or the slow activity of P450 enzyme yielded different product profiles and limited the taxadiene conversion rate (14, 74). In response to this hypothesis, we did not detect any leftover of the oxygenated taxanes in the cell pellets of the control group from the previous sections and only an insubstantial amount of taxadiene was found to be present (0.37 ± 0.05 mg/L). Since our study about medium type effect was conducted while the yeast was growing, we postulated that conducting the resting cell assays might be a better tool for improving our insights about CYP725A4 activity in S. cerevisiae.
Therefore, to illustrate all the metabolites of CYP725A4, BN6 resting cell assays in 50 mM phosphate buffer at approximate pH of 6–8 were performed to keep the effect of endogenous pathways on the enzyme activity to the minimum. Our initial results revealed that pH 6 was the least optimal for total oxygenated taxanes production (p = 0.04). Also, the yeast biomass increased after 115 hours in pH = 8, while showing an almost double overall side oxygenated taxane: T5α-ol (22.5 ± 4) at time point one relative to pH = 7. However, the overall difference between pH = 7 and pH = 8 was insignificant (adjusted p = 0.69) (Additional Fig. 8). Also, it has been reported that the neutral pH is optimal for CYP725A4 activity, where this enzymes shows half maximal activity at more acid or basic environments (75). Despite this, the optimal acidity for Taxus POR has not been reported yet.
The subsequent experiments were then performed using 50 mM phosphate buffer at pH = 7. Two galactose induction schemes were then designed. In the first one, which we called pre-rest, the galactose was used for both growth and taxanes production and this carbon source was therefore excluded from the buffer during the resting stage. For the other one, which we called post-rest, glucose was used to only grow the taxane-producing BN6 strain to obtain sufficient biomass. Then, the buffer was supplemented with galactose (2% (w/v)) to induce the taxanes production during the resting stage. As the biomass accumulation did not change over time at pH 7, it can be inferred that the higher NADPH availability together with optimal pH favoured higher and more diverse oxygenated taxanes production, where a relatively higher total oxygenated taxanes titre at the last end point (115 hours) of pre-rest galactose induction (354.6 ± 51.1 mg/L, 8.9 ± 1.3 mg/L/OD600) (Fig. 8B) was achieved, compared to that of post-rest galactose induction (220.5 ± 34.2, 5.51 ± 0.9 mg/L/OD600) (Fig. 8A). This could be due to sub-optimal performance of metabolite trafficking and galactose metabolism in the inactive cells from the post-rest galactose induction set. Also, the titre of all taxanes, including taxadiene, increased over 115 hours in both resting cell assays. Interestingly, by the end of the experiment, the major products of the post-rest induction were iso-OCT (50.4 ± 5.9 mg/L), OCT (43.5 ± 5.4 mg/L) and T5α-ol (30.2 ± 4.3 mg/L) (Fig. 8A). However, the major products of the pre-rest induction were in order of OCT (54.6 ± 11.7), T5α-ol (38.1 ± 8.4) and iso-OCT (35.7 ± 7.8) (Fig. 8B). Therefore, OCT remained as the major product, while both iso-OCT and T5α-ol were the late products of the CYP725A4 enzyme (Additional Fig. 9). Surprisingly, having a peak of 288 (m/z), which corresponds to the molecular weight of monooxygenated taxanes (76), taxadiene was also found to be hydroxylated. Hence, it is called oxygenated taxadiene in this study. This denotes that the availability of cofactors and optimal resources do not necessarily result in the improved production of major CYP725A4 product. Since we did not observe any change in the yeast biomass at the endpoint, the increase in taxadiene could be justified by enzyme folding events, or the P450 uncoupling, which renders the substrate presence a less important factor for the enzymatic catalysis (61), or the mechanism-based inactivation of the taxadiene into more reactive metabolites by CYP725A4 which eventually hinders the proper activity of the CYP725A4 enzyme. However, the later seems unlikely as the P450-POR were still active until the experiment end point. Aside from these points, the upstream pathways producing the taxadiene precursor, geranylgeranyl diphosphate (GGPP), might have also become more active during the resting stage and increased taxadiene production flux. Hence, although the titres of oxygenated taxanes and taxadiene isomers had increasing trends over 115 hours, the low substrate conversion did not seem to be solely the result of the slow CYP725A4 activity.
Despite this, it is possible that the OCT and iso-OCT are the intermediates to the downstream pathways in Taxus sp., or they are more toxic than T5α-ol, being degraded by the plant underground metabolism. This hypothesis becomes stronger as Edgar et al. (15) showed that only a minor amount of OCT was observed in Taxus plant cell culture. Therefore, it could be postulated that all these three oxygenated taxanes are the major products of CYP725A4, while the other diterpenoids with lower quantities might have been formed as a result of oxidation, which has a higher energy barrier in comparison to the epoxidation process (65). As such, some of these oxygenated compounds could be transient metabolites (metabolons) that degrade into more stable compounds (15), which were not metabolised at the resting stage or are the late products due to higher NADPH and cofactor availability for taxanes production, or are formed due to structural arrangements (77). As we found at least ten additional other oxygenated taxanes in both resting assays with 288 (m/z) (Additional Figs. 10–12), while having only five time points, we conducted a Granger causality test to find the correlation and causality relationships among them including taxadiene isomer(s) (78). To ensure the results were not impacted by the yeast growth, only pre-rest results were subjected to Granger test. According to the test results, there was a significant two-way association between OCT and iso-OCT (p = 0. 04 and 0.04). This was in line with a previous report, where iso-OCT was reported to be potentially formed by the structural rearrangement of OCT (79).
Focusing on the late oxygenated taxanes, labelled with roman numbers due to their novelty, iso-taxadiene was found as Granger-causal of diterpenoid III (p = 0.04), and oxygenated taxadiene and potentially taxadiene that might have co-eluted with it, was found as the Granger-causal of diterpenoid IX (p = 0.02) and diterpenoid VII (p = 0.04). Interestingly, diterpenoid IX was also found to be associated with diterpenoid III (p = 0.045), and taxadienediol, which is a double oxygenated taxane (304 (m/z)), was Granger-caused by diterpenoid VII (p = 0.03). While the complete identity of this taxadienediol is not unknown, Edgar and coworkers (80) have proposed two different taxadienediols, the first one being taxadien-5α-10β-diol which is one of the primary dihydroxylation products of CYP725A4 enzyme and another being 5(12)-oxa-3(11)-cyclo-taxan-10-ol, which was justified to be formed upon epoxide hydroxylation as a result of change in the enzyme selectivity. Hence, it is possible that any of the new dioxygenated diterpenoids be either the hydroxylated product of an oxidised or epoxidised intermediate.
Interestingly, it was also found that diterpenoid III unidirectionally Granger-caused iso-taxadiene (p = 0.03) and similarly, diterpenoid VII and IX Granger-caused the oxygenated taxadiene (p = 0.03 and 0.05, respectively). Therefore, it is also possible that these diterpenoids stimulate the higher production of taxadiene isomers. Overall, the concentration of total taxanes during post-rest stage (361.4 ± 52.4 mg/L) in a 10-mL scale was remarkable and comparable to the previous study, reporting the highest total taxanes concentration by whole S. cerevisiae culture at 1-L scale with optimised cultivation medium (13).