Impact of heterologous expression of Cannabis sativa tetraketide synthase on Phaeodactylum tricornutum metabolic profile

Abstract

Background

Pharmaceutical safety is an increasing global priority, particularly as the demand for therapeutic compounds rises alongside population growth. Phytocannabinoids, a class of bioactive polyketide molecules derived from plants, have garnered significant attention due to their interaction with the human endocannabinoid system, offering potential benefits for managing a range of symptoms and conditions. Traditional extraction from cannabis plants poses regulatory, environmental, and yield-related challenges. Consequently, microbial biosynthesis has emerged as a promising biotechnological alternative to produce cannabinoids in a controlled, scalable, and sustainable manner. Developing diatom-based biofactories represent a crucial step in advancing this biotechnology, enabling the efficient production of high-valued compounds such as cannabinoids.

Results

We engineered the diatom Phaeodactylum tricornutum, a unicellular photosynthetic model organism prized for its naturally high lipid content, to produce olivetolic acid (OA), a key metabolic precursor to most cannabinoids. The genes encoding tetraketide synthase and olivetolic acid cyclase from cannabis were cloned onto episomal vectors and introduced using bacterial conjugation in two separate P. tricornutum transconjugant lines to evaluate enzyme activity and OA production in vivo. Both genes were successfully expressed, and the corresponding enzymes accumulated within the transconjugant lines. However, despite testing the cell extracts individually and in combination, OA accumulation was not detected suggesting potential conversion or utilization of OA by endogenous metabolic pathways within the diatoms. To investigate this further, we analyzed the impact of CsTKS expression on the diatom’s metabolome, revealing significant alterations that may indicate metabolic flux redirection or novel pathway interactions.

Conclusions

Our study demonstrates the successful expression of cannabinoid biosynthetic genes in P. tricornutum but highlights challenges in OA accumulation, likely due to endogenous metabolic interactions. These findings underscore the complexity of metabolic engineering in diatoms and suggest the need for further pathway optimization and metabolic flux analysis to achieve efficient cannabinoid biosynthesis. This research contributes to advancing sustainable biotechnological approaches for cannabinoid production.

Graphical abstract

Background

Cannabinoids are a class of molecules notorious for their recreational properties. They are naturally produced in highest abundance in the glandular trichomes of the female flowers of the Cannabis genus. The plant of cannabis was used for millennia for its anti-inflammatory, antiemetic, and as a pain relief. Those properties are due to the ability of cannabinoids to interact with the mammalian endocannabinoid system involved in several metabolic functions [1,2,3]. Today, more than a hundred terpenophenolic compounds in the glandular trichomes of Cannabis sativa (C. sativa) have been identified as cannabinoids. Among them, Δ⁹-tetrahydrocannabinol (THC) and cannabidiol (CBD) are the most well-known and abundant [4, 5]. Recently, the interest in cannabinoids focused on their potential in fighting neurodegenerative disease such as Alzheimer’s and Parkinson’s diseases [1, 6]. The pathway to biosynthesize THC and CBD starts from hexanoyl-CoA and malonyl-CoA moieties, the tetraketide synthase (TKS) condensates hexanoyl-CoA with three malonyl-CoA to produce 3,5,7-trioxododecanoyl-CoA (TdCoA) compound which is then cyclized in olivetolic acid (OA) by the olivetolic acid cyclase (OAC). An aromatic prenyltransferase (APT) fuses OA with the prenyl group of geranyl pyrophosphate (GPP) to produce cannabigerolic acid (CBGA). Cannabidiol acid synthase (CBDAS) and tetrahydrocannabinol acid synthase (THCAS) catalyze the conversion of CBGA into cannabidiolic acid (CBDA) and Δ⁹-tetrahydrocannabinolic acid (THCA), respectively. As a result, cannabinoids are classified as terpenophenols, characterized by a cyclic monoterpene moiety derived from GPP and an alkylated diphenol backbone originating from OA [7]. Upon combustion or heat exposure, CBDA and THCA undergo decarboxylation, yielding the bioactive compounds cannabidiol (CBD) and-Δ⁹-tetrahydrocannabinol (THC), respectively (Fig. 1) [8,9,10]. The increasing recognition of cannabinoids for their therapeutic and commercial value has led to a surge in cannabis cultivation, with the global legal market projected to exceed US$400 billion by 2030 [11,12,13]. However, cannabis cultivation faces several challenges, including stringent regulatory frameworks, the limited yield of specific cannabinoids due to their restricted production in female flowers, and complex extraction and purification processes. These limitations have driven the search for alternative biotechnological production systems [14, 15].

Fig. 1

Scheme of the proposed cannabinoids synthesis in Cannabis sativa from hexanoyl-CoA and malonyl-CoA. Arrows represent enzymatic reactions. Enzymes’ names are in bold

Bioengineering of model organisms has increasingly advanced, both to confirm biological knowledge, and as an alternative way to produce compounds of interest. It does not only allow to circumvent all the constraints associated with cultivation, but also has a higher organism mass volume ratio. Attempts of producing cannabinoids in E. coli have been performed on purified TKS using synthetic N-acetylcysteamine as a substrate and leading to the production of olivetol. However, attempts of producing THCA never succeeded due to the complex folding of THCAS therefore considering the bacteria unsuitable for the production of cannabinoids synthase [16]. To overcome the limits of bacteria, the production of cannabinoids has been achieved in yeast with the limit of reaching poor yields despite supplementation [9, 17]. Algae are promising heterologous hosts, as the number of available tools and data gathered have dramatically increased over recent years [18,19,20]. Compared to conventional platforms, algae offer additional advantages, such as the ability to fix CO₂ through photosynthesis. In 2019, a patent reported that cannabinoids can be heterologously produced and harvested in microalgae Chlamydomonas reinhardtii and Synechococcus elongatus [21]. Compared to other microalgae, a unique advantage of diatoms is their estimated contribution to around 20% of Earth’s oxygen production [22, 23]. Among diatoms, Phaeodactylum tricornutum has emerged as valuable platform to produce fatty acid-derived molecules, due to its fast-growing rate, high lipid content, and natural production of omega-3 fatty acids like eicosapentaenoic acid [24, 25]. Efficient protocols have been developed for the extraction of apolar compounds such as cannabinoids [26, 27]. P. tricornutum also naturally produces malonyl-CoA [28], GPP [29], and is predicted to produce hexanoyl-CoA [30], all essential precursors for cannabinoid biosynthesis. Together, these features suggest that P. tricornutum is a suitable model organism for the heterologous biosynthesis of cannabinoids. Indeed, the introduction of Streptomyces sp. aromatic prenyltransferase NphB [31], and CBDAS [32] into P. tricornutum, has led to the in vitro production of CBGA and CBDA, respectively. Additionally, co-expression of a fused construction of the first two enzymes of the pathway—TKS and OAC from C. sativa—in P. tricornutum lead to the transient production of trace amounts of OA [33]. The reason for the loss of OA accumulation over time remains unclear. We hypothesized that expressing TKS and OAC separately in P. tricornutum strains, rather than as a fused construct, might enhance enzyme functionality and stability, thereby improving OA synthesis. Therefore, in this study, we aimed to produce olivetolic acid in P. tricornutum by inserting TKS and OAC enzymes in different transconjugants and test their activity individually or pulled together. In order to observe the extend of CsTKS impact on the metabolite content and deepen our understanding on the potential compatibility of P. tricornutum to produce cannabinoids, using UPLC–qTOF-MS we assessed the impact of CsTKS accumulation on transconjugants metabolite content.

Material and methods

Microbial strains and growth conditions

Escherichia coli strains Epi300 (TransforMax Epi300 Biosearch™ Technologies, Novato, CA, USA) and 10β (New England Biolab^®, ON, Canada) were grown in Luria Bertani broth (LB) at 37 °C with 220 rpm agitation when in liquid medium, supplemented with chloramphenicol (35 μg mL⁻¹) (Thermo Fisher Scientific, Ottawa, ON, Canada) and gentamicin (50 μg mL⁻¹) (Thermo Fisher Scientific, Ottawa, ON, Canada) when containing pTA-MOB.

Saccharomyces cerevisiae VL6-48 strain was grown following [33, 34]. Cultures in rich medium were done in yeast extract peptone dextrose YEPD while cultures in minimum media were done in complete minimal (CM) medium lacking histidine and uracil or tryptophan and uracil (Teknova, Hollister, CA, USA) at 30 °C.

The P. tricornutum strain (CCAP 1055/1, Culture Collection of Algae and Protozoa, RRID:NCBITaxon_2850) was grown in L1 medium without silica at 18 °C under white fluorescent lights (75 μE m⁻² s⁻¹) with a photoperiod of 16 h of light and 8 h of darkness under the agitation of 130 rpm for liquid cultures with zeocin (50 μg mL⁻¹) (Invitrogen, Burlington, ON, Canada) when cultivating transconjugant cells containing the Sh Ble resistance gene [35].

Plasmid constructions and insertion in Phaeodactylum tricornutum

Recombinant plasmids, named pTKS for CsTKS and pOAC for CsOAC, were assembled through yeast assembly using S. cerevisiae strain VL6-48 (ATCC MYA-3666: MATα his3-Δ200 trp1-Δ1 ura3-52 lys2 ade2-1 met14 cir⁰) exactly as described in [34]. The cassette containing the CsTKS under the control of 40SRPS8 promoter and fcpA terminator was inserted in the pPtGE30 plasmid, named EV hereafter, composed of a pCC1BAC (RRID:Addgene_62862) backbone, containing a bleomycin resistance gene (Sh Ble) as a selection marker, and a P. tricornutum region contributing to plasmid stability.

Assembled pTKS and pOAC episomes were transformed into the E. coli Epi 300 strain containing pTA-MOB plasmid. The cargo plasmid was then transferred in the diatoms via conjugation as described in the literature [36]. Transconjugants of P. tricornutum were then selected following the protocol of [31] leading to a liquid culture of P. tricornutum clones in L1 supplemented with zeocin. Conjugations were verified by colony PCR. Positive clones were selected for plasmid extraction then sent for Next Generation Sequencing (NGS).

Growth kinetics

Optical density of the clones was measured at 730 nm (OD_730nm) every two days for 14 days. For each of them, 250 µL was added to a Microplate 96 Well, PS, F-Bottom Chimney Well µCLEAR^®, black (Greiner Bio-One, Kremsmünster, Austria) and absorbance was recorded using Synergy H1 BioTek microplate reader (Agilent, Santa Clara, CA, USA). Curves were obtained using GraphPad Prism software (RRID:SCR_002798). A t-test was performed using CGGC permutation test from Walter and Eliza Hall Institute of Medical Research with 10,000 permutations [37].

Plasmid rescue

For each transconjugant lines, 5 mL were centrifuged at 4700 g for 10 min after 7 days of culture. After supernatant removal, 235 µL of PDL1 Large Plasmid Extraction Kit (Geneaid™, New Taipei City, Taiwan), together with 5 µL Lysozymes (100 mg mL⁻¹) and 5 µL hemicellulase (25 mg mL⁻¹) were added to resuspend the pellet. Samples were incubated at 37 °C for 30 min. 250 µL of PDL2 solution was added to the mix before mixing by inverting ten times, then 375 µL of PDL3 was added before mixing. The rest of the procedure was performed according to the manufacturer protocol until elution in 50 µL of elution buffer preheated at 80 °C. DNA concentration was measured with a N60/N50 Implen NanoPhotometer^®. Purified plasmid DNA was then used to transform E. coli 10β strains from New England Biolabs (Whitby, ON, Canada) following their protocol. Bacteria were plated on LB with chloramphenicol (35 mg mL⁻¹). After inoculating 5 mL LB with identical concentration of chloramphenicol, the liquid culture was incubated over night at 37 °C under agitation. Plasmid DNA was extracted by miniprep (Geneaid™, New Taipei City, Taiwan). Purified plasmid DNA concentration was measured with the NanoPhotometer and the quality of the DNA sample was verified by digestion with restriction enzymes. The plasmid DNA sample was then used for NGS at Massachusetts General Hospital DNA Core.

Western blot

Approximately 160 mg of wet mass was resuspended in 1X Laemmli in PBS at a ratio of 1 mL:750 mg⁻¹, heated at 95 °C for 10 min, and centrifuged 5 min at 14000g at 4 °C. At least 500 µg of protein was loaded on a 10% SDS–polyacrylamide gel and migrated at 80 V for 25 min, then at 120 V for 1 h. The gel was then transferred to a 0.2 μm polyvinylidene fluoride (PVDF) membrane, previously activated in methanol and then equilibrated in transfer buffer (25 mM Tris, 192 mM glycine and 20% methanol). Transfer was carried out at 100 V, 400 mA, for 2 h in transfer buffer on ice. The membrane was blocked in 5% milk in Tris Buffer Saline (TBS) with 0.1% Tween 20 for 50 min at room temperature. Blots were washed three times for 10 min with TBST before being incubated with 6x-His monoclonal antibody (Invitrogen, Canada, RRID:AB_557403) (dilution 1:1 000) in 3% bovine serum albumin overnight at 4 °C. After three washes in TBST, blots were incubated for 1 h with a 1:20000 dilution, in 5% milk of Immun-Star Goat Anti-Mouse (GAM)-HRP from Bio-Rad (Ontario, Canada, RRID:AB_11125753). After three more washes in TBST protein detection was performed using Clarity Max, 10 ng of Multiple Tag (GenScript, Brockville, Ontario, Canada) were used as positive control. Red ponceau was applied for 2 min at room temperature and then rinsed with distilled water. Western ECL Substrate-Luminol solution from Bio-Rad. Images were acquired using ChemiDoc Imaging System (RRID:SCR_019037) with Image Lab™ Software (RRID:SCR_014210) (Bio-Rad, Hercules, CA, USA).

Protein extraction

For each sample, 500 µL of cell buffer (0.75 mM SDS, 10% Glycerol, 51.4 mM, Tris HCl pH8, 0.02 mM EDTA) was added to 125 mg of pellets. PMSF and protease inhibitor were added at a ratio of 1/10 and 1/100, then proteins were extracted by sonication with the following parameters: amplitude 35%, time 3 min, ON 30 s, OFF 25 s (Fisherbrand™ Model 505 Sonic Dismembrator). Extracts were then spun for 40 min at 4 °C at 14000g and the supernatant was removed and spun a second time in the same conditions. The new supernatant was flash frozen and kept at − 80 °C until use.

Enzymatic assay

Hexanoyl-CoA and malonyl-CoA were purchased from Millipore Sigma Canada Ltd (Oakville, ON, Canada) and the olivetol and olivetolic acid used were purchased from Santa Cruz biotechnologies (Dallas, TX, United states). Protein concentration was measured by colorimetric RC DC™ Protein Assay (Bio-Rad, Hercules, CA, USA). Every test was done in triplicate in 4-(2-hydroxyethyl)−1-piperazineethanesulfonic acid (HEPES) buffer 0.05 M pH 7, DTT 0.005 M, with 400 µg of TKS and/or OAC protein extract (800 µg total protein extract when pulled together), 0.2 mM hexanoyl-CoA, and 1 mM malonyl-CoA, completed to 100 µL with water. The reaction was performed at 25 °C for 16 h and was stopped with trichloroacetic acid 2% v/v. 300 µL of methanol was added to the reaction mix then was filtered on PTFE 0.2 µm filter and transferred into new tubes. The samples were dried using a vacuum concentrator (SpeedVac Thermo Savant SPD 2010), then resuspended in 100 µL methanol before analyzing olivetolic acid production by high-performance liquid chromatography with diode-array detection (HPLC–DAD).

HPLC–DAD analysis

An InfinityLab Poroshell 120 EC-C18 (4.6 × 100 mm, 2.7 mm; Agilent Technologies, Mississauga, ON, Canada) column, maintained at 30 °C, was used for separation. For each sample, 10 μL was injected into the HPLC device. The mobile phase used was composed at 30% of (A) formic acid 0.1% v/v in milli-Q water and 70% of (B) formic acid 0.1% v/v in methanol. Samples were processed at a flow rate of 1 mL min⁻¹ following 0 min, 70% B; 1.0 min, 70% B; 6.0 min, 77% B; 15.0 min, 90% B; 15.1 min, 70% B and 18.0 min, 70% B. The total run time was 18.5 min allowing the reconditioning of the column for the next injection. D2 (deuterium) lamps acquired wavelengths from 190 to 400 nm. The analyses were done using UV wavelength at 200 nm. The maximum absorption wavelength and the retention times of each detected signal in samples were compared to standards to allow their identification.

Untargeted metabolomic

Metabolite extraction

Metabolites were extracted from three biological replicates for each clone. Each replicate was inoculated at OD₇₃₀ 0.1 in 200 mL, homogenized, then separated into four 50 mL volumes in 250-mL flasks to optimize the growth rate. Cultures were grown for an additional week, then for each replicate, the 50 mL cultures were pooled together. Cultures were then centrifugated for 10 min at 1500g at 4 °C and pellets were weighed. 1 mL of ethanol 95% (Fisher Chemicals, Pittsburgh, PA, USA) per 100 mg of wet biomass was added to the pellet. Samples were mixed by vortexing and incubated overnight at − 20 °C. The following day, the samples were centrifuged at 4000g for 10 min at 4 °C and supernatants were filtered using polytetrafluoroethylene (PTFE) 0.2 µm filter. All filtrates were dried using a vacuum concentrator (SpeedVac Thermo Savant SPD 2010) without heating, for 3 h at maximum ramp.

Data acquisition

Metabolomic extracts that were analyzed by reversed-phase chromatography (RP) were resuspended in 100 µL of pure water, while the ones analyzed using hydrophilic interaction liquid chromatography (HILIC) were reconstituted in 100 µL of mobile phase (ammonium formate 10 mM in acetonitrile/water (95:5), pH 3.8). Samples were vortexed and sonicated for 2 min, followed by a 5 min 16000g centrifugation. Supernatants were collected and diluted 1:100 in their initial solvents followed by a filtration using 0.45 μm AcroPrep filters in a 5 mm Hg vacuum. Acquisition was done with an Eksigent μU HPLC (Eksigent, Redwood City, CA, USA) in conjunction with an ABSciex TripleTOF 6600 (ABSciex, Foster City, CA, USA) system with an electrospray interface. For data processing, acquisition and instrument control, the Analyst TF version 1.8 software was used. Data acquisition was performed in Data Dependent Acquisition (DDA) and SWATH mode for positive ionization mode. The source voltage was set to 5.5 kV, temperature at 220 °C, curtain gas at 50 psi, ion source gas 1 at 40 psi, and ion source gas 2 at 50 psi. For all methods, the mobile phase was composed of: A (water with 0.1% formic acid and water with 10 mM ammonium fluoride) and solvent B (acetonitrile with 0.1% formic acid and acetonitrile: water (95:5) with 10 mM NH4F). Columns Luna Omega Polar 100 × 2.1 1.6 µm (RP positive) and BEH Z-HILIC 100 × 2.1 Particle 1.7 µm (HILIC) from Waters were kept at 50 °C. The gradient began at 0% B for reversed-phase, and 100% for HILIC.

Bioinformatics analysis

Raw mass spectra files in profile mode were aligned, deconvoluted and annotated with MS-Dial version 5.1.230912 (RRID: SCR_023076) [38,39,40]. The parameters were set as follows for all three analyses: retention time range of 0–10 min; mass range of 50–1 250 Da; MS1 and MS2 tolerances, for deconvolution and alignment, of 0.01 Da and 0.025 Da, respectively; minimum peak height and width of 1000 amplitude and 5 Da, respectively; the smoothing method used was linear-weighted moving average, with smoothing level of 3 scans and mass slice of 0.1 Da; and the sigma window value was set to 0.5. For analyte identification, the following parameters were set: accurate mass tolerance 0.01 Da for MS1 and 0.05 Da for MS2; retention time tolerance was set to 0.1 min; adduct ions were set to [M + H]⁺ and [M + NH₄]⁺ for ESI⁺ and [M-H]⁻ for ESI⁻. Deconvoluted spectra were compared to MassBank ESI⁺ and ESI⁻ [41] and MS-Dial’s “ESI⁺-” and “ESI⁻-MS/MS from authentic standards” databases version 19 (all databases were obtained as MSP files from MS-Dial’s webpage on March 2024) for annotation. These comparisons were done with the default parameters, except for MS1 and MS2 mass tolerances, set to 0.01 and 0.001 Da, respectively. Annotations marked as “reference match” by MS-Dial were validated by checking that main peaks from reference were present in the current experiment’s spectra.

Peak height tables exported from MS-Dial were then used for comparative analyses in R version 4.3.3. RP ESI⁺ and ESI⁻ datasets included three blank samples, while HILIC ESI⁺ dataset included two blank samples. For each analyte, the abundance estimate was defined as

$${\text{peak}\_\text{height}}_{\text{replicate}}-(3*\text{max}\left({\text{peak}\_\text{height}}_{\text{blank}\_\text{samples}}\right))$$

If the abundance estimate was ≤ 0, the analyte was considered absent. For each strain and for both quality control replicates (pooled samples), only analytes present in all replicates were considered present. Finally, only analytes considered present in at least one strain were selected for further analysis. The abundance estimation table was prepared for MetaboAnalystR version 4.0 (RRID:SCR_016723) [42], which was used for replacing missing values, filter analytes (parameters: $\text{qc.filter}={\text{T}},\text{ rsd}=25,\text{ var.filter }={\text{"none"}}, \text{ var.cutoff }= -1, \text{int.filter}={\text{"mean"}}, \text{ int.cutoff}=0$) and normalize data (functions $\text{PreparePreNormData}$ and $\text{Normalization}(\text{"GroupPQN"}, \text{ transNorm}=\text{"LogNorm"}, \text{ scaleNorm}=\text{"ParetoNorm"}, \text{ ref}=\text{"QC"}, \text{ ratio}=\text{F, ratioNum}=20)$). Next, MetaboAnalystR’s filtered datasets were used to find significant differences in analyte abundances between pPtGE30 and AC9 strains using the function sam (parameters: $\text{method}=\text{"d.stat"},\text{ B}=200,\text{ med}=\text{F,use.dm}=\text{T, var.equal}=\text{ T,R.fold}=1$) from the R package siggenes version 1.76.0 [43]. Fold change was calculated as follows:

$${\text{log}}_{2}\frac{\text{mean}\left({\text{abundance}}_{\text{Ac}9\text{ clone}}\right)}{\text{mean}\left({\text{abundance}}_{\text{pPTGE}30}\right)}$$

If p < 0.01 and |fold change|> 1, the analyte was considered deregulated in the corresponding AC9 strain. Finally, principal component analysis (PCA) was performed using the combined normalized data from all three analyses (RP ESI⁺, RP ESI⁻ and HILIC ESI⁺) with MetaboAnalystR function PCA.Anal.

Results

Heterologous accumulation of CsTKS and CsOAC in P. tricornutum

Two expression cassettes were designed to independently characterize CsTKS and CsOAC expression in P. tricornutum. The plasmid referred to as pTKS, contained the CsTKS coding sequence (CDS) under the control of the endogenous 40SRPS8 promoter, tagged with a 6xHis marker, and terminated with the FcpA terminator. The plasmid referred to as pOAC, included CsOAC CDS under the control of the FcpD promoter, tagged with a myc marker, and terminated with the FcpD terminator (Fig. 2). These plasmids along with the empty vector (EV), were introduced into P. tricornutum wild-type strains via interkingdom conjugation with E. coli. The plasmids harbored the Sh ble gene, conferring zeocin resistance, and transconjugants were selected on L1 agar medium supplemented with zeocin.

Fig. 2

Schematic representation of pPtGE30 expression vector and insertion of CsTKS and CsOAC cassettes. The pTKS plasmid includes a construction that contains the 1971-bp cassette with CsTKS under control of native constitutive 40SRPS8 promoter, 6xHis marker and an FcpA terminator. The pOAC plasmid harbors a construction with the 1265 bp cassette CsOAC under control of FcpD promoter, marked with a Myc tag, and ended with an FcpD terminator. The Cen-His-Ars region is necessary for episomal replication while the OriT permits the transfer of the plasmid in P. tricornutum during interkingdom conjugation. Finally, the Sh ble and CmR coding sequence allows for zeocin resistance when expressed in P. tricornutum and chloramphenicol resistance in E. coli, respectively

Three transconjugants of each construct were selected for plasmid rescue, and the extracted DNA was subjected to restriction digestion analysis before sequencing (Additional file 1: Fig. S1). Transconjugants showing the expected digestion profiles, were further analyzed by Next Generation Sequencing. While the pTKS sequences exhibited minor substitutions, including one in the zeocin resistance gene, two in the 40SRPS8 promoter, and one in the backbone (Additional file 1: Table S1), the CsOAC CDS remained intact. Despite the substitutions, all transconjugants displayed growth on zeocin-supplemented media, indicating that the resistance gene remained functional. Importantly, the heterologous expression of these enzymes did not impair the growth of the transconjugants (Additional file 1: Fig. S2).

Heterologous protein production was then verified by western blot. All pTKS-containing transconjugants exhibited a band at the expected size of 42 kDa, absent in EV along with an additional smaller band likely corresponding to a truncated version of the protein (Fig. 3). For the pOAC transconjugants, two (pOAC-5 and pOAC-7) displayed a band at the expected size of 12 kDa, while one transconjugants (pOAC-6) did not. Further sequencing revealed a frameshift in the fcpD promoter of pOCA-6, which disrupted the myc-tag and explained the absence of a signal in the western blot (Fig. 3).

Fig. 3

Detection of TKS and OAC enzymes in P. tricornutum transconjugants by western blot analysis. a Western blot for CsTKS detection in P. tricornutum pTKS transconjugant extracts using anti-6His antibodies. b Western blot for CsOAC detection in P. tricornutum pOAC transconjugant extracts using anti-Myc antibodies. In both panels, protein extracts from transconjugants are compared to those from the empty vector (EV) and wild-type P. tricornutum. Protein ladder sizes are indicated on the left while the expected sizes of the proteins of interest and their respective tags are shown on the right of each blot. The lower panels display red Ponceau (RP) staining to confirm protein loading

To validate the added enzymatic activity from each of the recombinant proteins, total protein extracts from pTKS and pOAC were tested. Enzymatic assays were conducted using the extracts individually or by combining the CsTKS and CsOAC extracts. The resulting reactions were analyzed by HPLC–DAD, but none of the detected peaks corresponded to the OA standard (Additional file 1: Fig. S3). To rule out potential inhibitory effects from compounds present in the crude extract, protein purification was attempted. While TKS was successfully purified, purification OAC was not achieved.

Heterologous expression of CsTKS induces metabolic changes in P. tricornutum

The production of a tetraketide is a critical initial step in cannabinoid biosynthesis. To assess the impact of CsTKS activity on the diatom metabolome, we compared the metabolic profiles of pTKS transconjugants with those of the EV control. Metabolite extracts were separated using either hydrophilic interaction liquid chromatography (HILIC) or reversed-phase chromatography (RP) followed by tandem mass spectrometry. HILIC-separated metabolites were analyzed in positive ionization mode, while RP-separated metabolites were analyzed in both positive and negative ionization modes (referred to hereafter as HILIC positive, RP positive and RP negative).

HILIC positive and RP negative analyses detected over 3000 analytes each, with 19 and 8 metabolites annotated, respectively. RP positive analysis was more comprehensive, detecting over 7000 analytes, of which 29 were annotated (Fig. 4a). Among the metabolites detected, 12,348 compounds have been found in all replicates of each transconjugant, 160 analytes were uniquely present in CsTKS transconjugants and absent in the EV control. Each CsTKS transconjugant presented a different amount of exclusive analytes ranging from 48 to 120. A notable number of 278 analytes were present in all clones but pTKS-3 while only 66 excluded from pTKS-1 and 57 from pTKS-2. A total of 57 annotated analytes have been identified with 49 shared among all clones. No exclusive analyte from a clone was annotated. Three were exclusively shared by the pTKS clones: pyocyanin, a phenazine pigment derived from the shikimate pathway in Pseudomonas aeruginosa; 2-benzyl-4-chlorophenol (chlorophene), a phenolic halogenated compound with an unidentified biosynthetic route; and abietic acid, a diterpene biosynthesized from GPP. Notably, 18 analytes detected in the EV transconjugants were absent in all three pTKS transconjugants (Fig. 4b).

Fig. 4

Untargeted metabolomic analysis identifies specific metabolites unique to P. tricornutum pTKS transconjugant lines. a Proportion of analytes, annotated or not, detected using three different HPLC–MS/MS analytical methods: reversed-phase chromatography with positive ionization (RP ESI⁺), reversed-phase chromatography with negative ionization (RP ESI), and hydrophilic interaction chromatography with positive ionization (HILIC ESI⁺). b Total analytes detected (left) and annotated (right) for each strain, combining data from all three analytical modes. Among the annotated analytes absent in EV negative control, the following were consistently detected in pTKS transconjugant strains: pyocyanin, chlorophene and abietic acid were found in all pTKS lines; decyl dimethyl amine oxide and tri-isobutylphosphate were detected in pTKS-2 and pTKS-3; and 3-hydroxy-C6-homoserine lactone was identified in pTKS-1 and pTKS-3

The second parameter analyzed was the quantitative impact of CsTKS expression on the diatom metabolome (Fig. 5a). Analytes were heterogeneously deregulated among the three pTKS transconjugants compared to the three replicates of EV transconjugants, with fold changes ranging from − 10.3 (analyte with retention time (RT) = 3.768 min and with a m/z of 771.15744 in pTKS-3) to 11.2 (RT = 1.449 min; m/z = 431.29585 in pTKS-1). However, 48 analytes were consistently deregulated across all pTKS transconjugants triplicates (Fig. 5b).

Fig. 5

Differential metabolite abundance highlights consistent deregulation in P. tricornutum pTKS transconjugants compared to the empty vector. a Overview of differentially abundant analytes detected using three analytical methods (Reversed-phased liquid chromatography coupled with ESI+/ESI− MS/MS and HILIC coupled with ESI + MS/MS) compared to the empty vector. The y-axis lists all detected metabolites, showcasing the variability in abundance across pTKS transconjugant replicates. b Focused analysis of 48 analytes consistently deregulated in all pTKS transconjugants compared to the empty vector. Only analytes with a (|fold change|) greater than 1 and a significant threshold of p < 0.01 were included. These analytes exhibit consistent under- or overabundance across all pTKS transconjugant replicates

Among these, the most deregulated analytes were detected at RT = 2.921 min; m/z = 577.33159 (fold change between − 5.1 in pTKS-3 and − 6.4 in pTKS-1) and RT = 2.842 min; m/z = 581.31317 (fold change between 3.4 in pTKS-1 and 4.9 in pTKS-2). Among the 48 deregulated analytes, 26 were annotated (Fig. 6). Two of these annotations corresponded to 1-methyladenosine and two to thiamine. Among the 24 unique annotations, 13 were aromatic metabolites, including pinacidil, guanine, and worenine, the latter being downregulated in pTKS-2 transconjugants compared to the negative control. A single annotated metabolite, norvaline betaine, was deregulated in all pTKS transconjugants. Additionally, nine analytes were downregulated in all triplicates of at least one pTKS transconjugant. While analytes uniquely detected in the presence of CsTKS (Fig. 5b) were not consistently deregulated across all pTKS transconjugants due to variability in peak intensities among replicates, some showed significant upregulation. Among these, 3-hydroxy-C6-homoserine lactone, decyl dimethyl amine oxide, and proline betaine were the most upregulated analytes in the entire experiment.

Fig. 6

Presence of CsTKS coincides with altered abundance of cyclic metabolites in P. tricornutum transconjugants. Heatmap illustration the deregulation of annotated metabolites in each pTKS transconjugant replicates compared to the negative control. Only analytes with |fold change|> 1 and p < 0.01 were considered as deregulated. “Coumarin derivative [1]” corresponds to methyl 2-[(4-methyl-2-oxo-2H-chromen-7-yl)oxy]propanoate and “Isoflavonoid C-glycoside [1]” to Puerarin. Ala-Phe refers to alanylphenylalanine

Annotated metabolites deregulated by CsTKS in Phaeodactylum tricornutum

To investigate the impact of CsTKS on metabolism, we analyzed the structure and biosynthetic pathways of the nine annotated metabolites deregulated in at least two transconjugant strains (Table 1). Similar metabolites or pathways were highlighted in the table to identify shared metabolic routes. Of the nine detected and annotated analytes, eight possess aromatic structures, including two dipeptides (norvaline betaine and alanylphenylalanine), two alkaloids of quinolizine and quinolizidine nature (quinoline and thermopsine), and thiamine, which was detected as both phosphorylated and not phosphorylated. The two dipeptides originate from distinct biosynthetic pathways. While the pathway for norvaline betaine biosynthesis in P. tricornutum remains unresolved, norvaline is a modified valine expected to derive from pyruvate metabolism via either the (2S)−2-isopropyl malate or (S)−2-acetolactate branches, both of which converge at 2-oxoisovalerate, a precursor to valine [44]. The betaine moiety present in norvaline betaine is likely biosynthesized through the oxidation of choline to trimethyl-glycine [45, 46]. Coumarin and alanylphenylalanine are both derived from the shikimate pathway, though the specific biosynthesis of methyl 2-[(4-methyl-2-oxo-2H-chromen-7-yl)oxy] propanoate (a coumarin derivative) remains uncharacterized. 2-Hydroxyacetophenone, alanylphenylalanine, and the coumarin derivative share a benzene ring, with the latter also containing a pyran structure. Meanwhile, guanine and thiamine are derived from the nucleotide metabolism, with guanine from the purine metabolism, whereas the pyrimidine and thiazole moieties of thiamine are biosynthesized separately prior to be combined.

Table 1 Impact of heterologous CsTKS on P. tricornutum’s metabolism

Discussion

In this study, we successfully produced the C. sativa tetraketide synthase and olivetolic cyclase in P. tricornutum. Importantly, the heterologous expression of these enzymes did not impair the growth of the transconjugants (Additional file 1: Fig. S2). Due to the unavailability of 3,5,7-trioxododecanoyl-CoA, both as a substrate and as a standard, coupled with its high molecular weight of 991.22 g mol⁻¹, we were unable to directly test the enzymatic activities of CsTKS and CsOAC. Alternative strategies were therefore employed to evaluate their functionality. Analysis by HPLC–MS/MS (High pressure liquid chromatography with tandem mass spectrometry) of protein extracts containing CsTKS and CsOAC, either assessed alone or combined, did not yield a peak corresponding to olivetol or olivetolic acid standards. Several hypotheses could explain the absence of olivetolic acid in the extracts. Western blot analyses confirmed that both CsTKS and CsOAC were accumulating as soluble proteins in P. tricornutum, suggesting that protein misfolding is unlikely to account for the lack of product detection [57]. Instead, the failure to detect olivetol or olivetolic acid may be attributed to the utilization of key substrates or intermediates by competing endogenous pathways. Another plausible explanation is that CsTKS did not dimerize correctly in P. tricornutum cellular environment, as proper dimerization is critical for its activity [58].

To investigate the broader impact of CsTKS expression on the diatom metabolome, we performed an untargeted metabolomic analysis. This revealed significant differences in the metabolite profiles of pTKS transconjugants, with each replicate containing 50 to 120 exclusive metabolites that were absent from the control EV. A previous research study has suggested that clone consistency is rare in P. tricornutum [59] supporting the variability in metabolite content and abundance observed in this study. Furthermore, 160 analytes were consistently detected in all three CsTKS-expressing replicates but were absent from the control EV. While only a fraction of these metabolites could be annotated, it is plausible that many of the unidentified analytes include polyketide-derived metabolites influenced CsTKS activity. Previous untargeted metabolomic studies in P. tricornutum using Fourier-transform ion cyclotron-resonance mass spectrometer resulted in the detection of several hundred of annotated analytes [60]. However, the lower resolution of the UPLC-qTOF-MS used in our study limited the number of annotated metabolites. Despite these limitations, clear patterns emerged in the metabolite abundance data. Specifically, 48 metabolites consistently deregulated across all CsTKS-expressing transconjugant replicates, indicating a potential influence of CsTKS on intracellular metabolite content. These results suggest that CsTKS expression induces metabolic shifts in P. tricornutum, possibly altering the biosynthesis or accumulation of metabolites.

The structure of each annotated metabolites was analyzed to assess the potential involvement of TKS in their biosynthesis (Table 1). Polyketides are a diverse class of specialized metabolites characterized by the presence of C=O–CH₂ units or derived from such molecules, which complicates their structural identification. To address this challenge, we relied on structural information corroborated by literature evidence. Among the deregulated metabolites, eight out of nine that were altered in two or more replicates and nine out of 16 altered in only one replicate were aromatic (Table 1 and Additional file 1: Table S3). In C. sativa, OAC catalyzes the formation of a phenolic structure from trioxododecanoyl-CoA. In P. tricornutum, it is plausible that a different enzyme takes over the biosynthesis of unidentified compounds from a polyketide-derived products generated by the CsTKS enzyme.

Pyocyanin, one of the three exclusive annotated metabolites, was detected in all pTKS transconjugant replicates and was significantly upregulated in pTKS-1 transconjugants (Figs. 4, 6). Pyocyanin is a well-characterized virulence factor in Pseudomonas aeruginosa [61, 62] and is classified as a phenazine pigment produced from chorismic acid [63]. Given that culture conditions were carefully controlled and remained stable over time and across transconjugants, the observed upregulation of pyocyanin is likely attributable to a direct or indirect modification in its biosynthetic pathway as PKS are capable of synthesizing pigments, rather than being driven by environmental factors, such as changes in light conditions [64,65,66].

As previously mentioned, most of the detected metabolites exhibit a phenolic structure. Among them, a coumarin derivative, methyl 2-[(4-methyl-2-oxo-2Hchromen-7-yl)oxy]propanoate, was upregulated in two pTKS transconjugant lines. While its precise biosynthetic pathway in vivo remains unclear, it likely originates from coumarin or coumaroyl-CoA, intermediates produced via the phenylpropanoid pathway. The formation of coumarin core is thought to involve condensation reactions between phenolic and carbonyl metabolites, which may also contribute to the biosynthesis of other aromatic metabolites detected in this study [67]. The potential involvement of a polyketide synthase such as CsTKS in its synthesis remains uncertain. It has been reported that 4-coumaric acid can serve as substrate for type three polyketide synthase (PKSIII), which can catalyze the production of stilbene or chalcone metabolites [68]. Although CsTKS is a PKSIII, it is not expected to produce chalcone products, and no stilbene metabolites were detected among the 16 upregulated annotated metabolite. This is unexpected, as CsTKS is hypothesized to generate stilbene through linear polyketides synthesis [58]. Additionally, 3-hydroxy-C6-homoserine lactone, which is known to play a role in quorum sensing, was identified. While its structure resembles that of polyketides, its classification as a polyketide has not been confirmed [69, 70]. Another, metabolites, alanylphenylalanine, is a modified amino acid that could serve as a substrate for polyketide extension [71]. It is plausible that the activity of CsTKS stimulates the synthesis of alanylphenylalanine. Overall, we observed a general increase in metabolites derived from the shikimate pathway, including bergenin, pyocyanin, phenylalanine and coumarin. This suggest that CsTKS may either influence the shikimate pathway or its products may affect light perception, thereby increasing the synthesis of metabolites capable of light absorption [64]. It has not been reported in literature that heterologous production of enzymes in P. tricornutum has by itself an impact on the activity of the shikimate pathway, although few of those study perform metabolomic analysis.

The only analyte consistently upregulated across all transconjugants in this study is norvaline betaine. The biological role and biosynthetic pathway of this modified amino acid remains largely unknown. Additionally, a significant upregulation of proline betaine was observed in TKS-1 transconjugant strains (Fig. 6), suggesting that the presence of CsTKS might enhance betaine biosynthesis. Betaine, in the form of trimethyl-glycine, along with proline betaine and carnitine (upregulated in pTKS-1 and pTKS-2 lines, respectively, as shown in Additional file 1: Table S3), are well-known quaternary ammonium compounds. These metabolites function as osmolytes and are typically upregulated in saline stress [72]. Beyond their osmoprotective properties, betaines are also involved in epigenetic regulation and can act as methyl donors, although the in vivo mechanism of amino acid betainization is still unclear [46, 73]. The observed upregulation of glycine could be associated with increased thiamine production, though the underlying causal relationship remains speculative.

Quinolizine, thermopsine, and worenine, which belong to the quinolizine family, were also identified. The biosynthesis of quinolizine in vivo is not well-characterized, though a synthetic pathway has been successfully achieved using a PKSIII enzyme from Huperzia serrata to produce 2-hydroxy-4H-quinolizin-4-one [50]. This raises the possibility that CsTKS may contribute to quinolizine biosynthesis in P. tricornutum. However, thermopsine was downregulated in two pTKS transconjugant strains and worenine, an isoquinoline, was downregulated in one strain (Additional file 1: Table S3). This suggests that while CsTKS might be involved in the production of quinolizine-related metabolites, its precise role and the regulatory mechanism influencing these metabolites remains unclear.

Hydroxyacetophenones, such as karanjin, are suspected to be produced via the polyketide pathway [19]. If the biosynthesis of 2-hydroxyacetophenone follows a similar process in P. tricornutum, its reduced concentration in this study could be due to substrate channeling by the heterologous CsTKS, which may utilize precursors typically employed by native polyketide synthases.

Conclusions

This study highlights the inherent metabolic variability among P. tricornutum transconjugants, reinforcing the challenge of achieving consistent metabolite production in engineered strains. Significant differences were observed among transconjugant clones, with distinct metabolic profiles and varying analyte compositions. The presence of metabolites exclusive to CsTKS-expressing transconjugants, supports the metabolic activity of the heterologous enzyme, yet no annotated analytes were unique to a single clone. Among the metabolites detected exclusively in pTKS transconjugants, compounds such as pyocyanin, abietic acid, and chlorophene were identified, though they are unlikely to originate from the polyketide pathway. These findings emphasize the complexity of P. tricornutum metabolism and highlight the need for a deeper understanding of its metabolic network to enable controlled and sustainable olivetolic acid production.