Each of the three genes was cloned into separateepisomal vectors, transformed into A. nidulans, and the resulting transformants were grown on CD-ST agar plates . Following 5 days of growth in CD-ST, the sample media were extracted and analyzed by liquid chromatography-mass spectrometry . To obtain compounds for structural determination, we first optimized the culturing conditions to get high titers from shake flask culture. By examining the organic extracts from cells and media separately, we determined that most of the compounds were secreted into the media. We also observed that a spore inoculum size of 104 spores/mL led to the highest titers of the four compounds whereas inoculum sizes of 108 spores/mL and higher gave low production of target compounds . Notably, when the molecules were produced at high titers, A. nidulans adopted a morphology of globular pellets. As the titer drops upon increased inoculum size, A. nidulans grew as dispersed filaments . The unnatural pairing can compromise the performance of overall biosynthesis by a “ratelimiting” component such as Ti_OvaBC. Aside from the molecular recognition of the acyl chain by the SAT domain active site, a successful acyl chain transfer also requires complementary protein-protein interactions between HRPKS ACP domain and NRPKS SAT domain.Previous studies have shown that although for some non-cognate HRPKS-NRPKS pairs, acyl chain could occur and new products emerged without any protein engineering efforts,4×4 flood tray for others replacement of SAT domain is necessary to compensate for the otherwise undermined inter-domain communication.While the sequence identity between ACPs of Ti_OvaA and Ma_OvaA is 62%, the identities between NRPKS SAT of Ti_OvaB and Ma_OvaB is lowered to 48%.
Such moderate sequence identity between the SAT domains implies that the recognition sites between ACP from noncognate HRPKS and SAT can be weakened or even abolished. Therefore, to generate a combination that exclusively produces 3 at a higher titer, protein engineering endeavors that improve the compatibility between unnatural HRPKS and NRPKS enzymes are necessary. Alternatively, further genome mining of related clusters may lead to one that can produce olivetolic acid robustly in a heterologous host. In summary, we have discovered a novel platform to produce OA and its analogs from filamentous fungi. The platform consists of an HRPKS and an NRPKS, known to produce resorcylic acid moieties in tandem, and a separate TE enzyme. This platform represents a new strategy to produce these cannabinoid precursors in microbes without relying on the OAS and OAC found in Cannabis sativa.Small quantities of olivetolic acid are produced in Cannabis sativa and chemical synthesis of the compound has proven to be difficult. Therefore, we hypothesized utilizing a Design of Experiments approach on our novel platform can further increase titer and thereby effectively solve the issue of low production. With regards to increasing production of fungal secondary metabolites and enzyme expression, DOE has proven to be an effective tool. DOE has been used for increased secondary metabolite production in bacteria, increased lipase production in fungi, increased xylanase production in fungi, and increased lignocellulolytic production in fungi, amongst other uses. We therefore sought to utilize a DOE approach to optimize the olivetolic acid and olivetolic acid analogs’ titer produced by our novel platform. DOE has proven to be more effective than one-factor at a time analysis because it greatly reduces the number of experiments needed and considers the interactive effects that factors can have with each other. The DOE approach begins at screening and ends at optimization and predictive modeling. First, screening is done to identify the factors that are most significant in the parameter response.
We decided to focus our DOE approach on the media that our A. nidulans strain producing olivetolic acid and its analogs is cultured in. Since our DOE approach is focused on media, we tested the effects that nine different components have on the production of sphaerophorolcarboxylic acid, olivetolic acid, and the analogs. Utilizing JMP statistical software, we performed an initial screening run of 24 experiments in order to identify the 3-5 most significant facts. To perform the screening experiments, we had to decide between a variety of screening platforms provided by JMP. JMP offers two types of screening designs: classical screening designs which include fractional, factorial, Plackett-Burman, regular fractional factorial, Cotter, and mixed-level designs and main effects screening designs: screening designs focused on measuring main effects with negligible interactions between the factors. We opted for a classical screen design because we wanted to include the interactions between factors and considered the different types from there.Two-level full fractional factorial designs account for all the combination of the factor levels. Two-level refers to a high value and a low value for the factor tested. The total number of runs for this design is the product of the factor levels. For example, for a two level full fractional design, the total number of runs would be 2n where n is the number of factors tested. The design also estimates that all of the effects are uncorrelated and that there are no interactions between the factors, i.e., the factors are orthogonal to each other.From a screening point of view, we determined the full fractional factorial design to be inefficient and cumbersome to do since it would require us to generate 29 types of media to test. The next classical screening design to consider was two-level regular fractional factorial.Two-level regular fractional factorial designs are similar to full fraction factorial designs; however, instead of the number of runs being equal to 2n where n is the number of factors tested, regular fractional factorial design runs are equal to 2n-k where k<n. In other words, a two-level regular fractional factorial is just a fraction of the full two-level fractional factorial design and therefore like the full factorial design, considers all the factors to be orthogonal to each other.
Since we could determine the fraction we would use, this was seen as an adequate design to choose; however, we wanted to consider some interactions between factors we determined to not follow through with this design and next looked at Cotter designs. Cotter designs are useful due to the ability to test a large number of factors in a small number of runs and are also useful if one is interested in the interaction between factors. Cotter designs are upheld by what is known as the principle of effect sparsity. These designs operate under the assumption that if one of the components of the sum of factors has an active effect , the sum of the factors will display the response.However, this can be potentially misleading and lead to false negatives if for example, one factor has a positive effect on the response, while the other factor has a negative effect, therefore totaling the sum of the factors to be zero/near zero, and therefore failing to show an effect. We determined not to go through with this design due to the false negative risks. We proceeded then briefly to mixed level designs. Mixed level designs are typically used when screening categorical or discrete factors containing varied factor levels.For example, one can be screening for the effect that light and a four-level media component have on the titer of a metabolite. Since we desired to keep the screen simple with just two levels for our factors since and we did not have any qualitative factors,hydroponic tray we proceeded to our last design, the Plackett-Burman design. Plackett-Burman designs are somewhat like regular fractional factorial designs except the total number of runs are a multiple of four rather than a power of two. Additionally, interactive effects between factors in a Plackett-Burman design are only partially confounded by the main effects which differs from regular fractional factorial where the interactive effects are completely confounded by the main effects and are therefore indistinguishable from each other.Plackett Burman designs are typically utilized when testing for the main effects among a variety of factors and so with this in mind, we chose to utilize the Plackett-Burman design as our screening design. We implemented the two-level Plackett-Burman design and generated 24 runs to test, with each run being a different media composition. For the two levels , we determined these values to input: temperature , pH , starch , NaCl , dextrose , yeast extract , casein acidic digest , trace elements , and 20x nitrate salts . We cultured the strain in these different medias, assaying the titer in media sets of 4 with CD-ST as the control media in each run. We inputted the data into the JMP software and performed analysis of variance and generated a Pareto chart from ANOVA and noted the results. ANOVA is a widely used tool to determine if there are statistical differences between means of different groups.
It is a collection of different statistical models and is used to determine whether the variance of a specific effect or factor interaction is statistically significant.The Pareto chart puts the ANOVA data in a simple to understand form, displaying whether the factor or factor interaction has exceeded the t- value limit and Bonferroni limit. The t-value refers to the value of the difference relative to the variation of the data tested. It is a value that represents the ratio of the difference between the estimated value of factor and its hypothesized value to its standard error.126 The Bonferroni limit is the value from the Bonferroni method that answers which factors means are significantly different from each other. Factors and factor interactions above the Bonferroni limit indicate that they are statistically significant and have a great effect on the parameter response, factors and interactions between the t-value limit and Bonferroni limit are indicated as potentially significant, and factors and interactions below the t-value limit are noted as insignificant.Based on the Pareto chart, we noted that increased temperature, addition of dextrose, addition of yeast extract, and addition of NaCl all had values above the Bonferroni limit indicating that these factors significantly affected the titer. However, three of values were all labeled blue indicating a negative result. Increased nitrate salts which had a negative effect on the media based on its blue distinction had a t-value greater than the t-limit but lower than the Bonferroni limit whereas increased casein acidic digest and increased starch which were also labeled blue had t-values lower than the t-limit indicating that these factors are not statistically significant. For factors having a positive effect on the titer labeled orange, NaCl salt had a tvalue greater than the Bonferroni limit indicating its statistically significance and increased nitrate salts had a t-value lower than the t-limit indicating it is not significant. From the screening data, we proceeded with two factors for our optimization experiments: NaC1 and nitrate salts. Although it was tabulated as not significant, we chose to include nitrate salts mainly because we did not want to optimize just one factor and since nitrate salts were already included in the media. We also included addition of MgSO4 since that was noted in the literature to increase metabolite production.Once we obtained the three factors, we utilized the response surface methodology optimization approach for optimization of the media, seeking to add these factors to our CD-ST media containing starch, trace elements, and casein acidic digest as the base since these factors had no statistical significance towards the titer. RSM is a widely used method for modeling/predictive modeling.RSM optimizes the factors correlating to a response with the inclusion of the effects interactions between the factors have. RSM has proven to be just as effective in modeling as a 3-level full factorial design, but its advantage is that RSM greatly reduces the number of experiments needed to form an accurate model. As an example, in a 3- level factorial design utilizing 3 factors, one would need to do 27 experiments to get an accurate model as opposed to the 15 experiments one would need for a central composite design RSM approach. As the number of factors increase, the difference between the number of experiments needed for a 3-level full factorial design vs a central composite design RSM approach greatly increases, which is why RSM is the advantageous approach. Reports of the effectiveness of increasing secondary metabolite production in fungi with the RSM approach have been recorded. Talukdar et al. observed a 7-fold increase in antibiotic production of Penicillium verruculosum MKH7 utilizing an RSM approach on the media used.