Expression of cocoa genes in Saccharomyces cerevisiae improves cocoa butter production

Background Cocoa butter (CB) extracted from cocoa beans (Theobroma cacao) is the main raw material for chocolate production, but CB supply is insufficient due to the increased chocolate demand and limited CB production. CB is mainly composed of three different kinds of triacylglycerols (TAGs), 1,3-dipalmitoyl-2-oleoyl-glycerol (POP, C16:0-C18:1-C16:0), 1-palmitoyl-3-stearoyl-2-oleoyl-glycerol (POS, C16:0-C18:1-C18:0) and 1,3-distearoyl-2-oleoyl-glycerol (SOS, C18:0-C18:1-C18:0). In general, Saccharomyces cerevisiae produces TAGs as storage lipids, which consist of C16 and C18 fatty acids. However, cocoa butter-like lipids (CBL, which are composed of POP, POS and SOS) are not among the major TAG forms in yeast. TAG biosynthesis is mainly catalyzed by three enzymes: glycerol-3-phosphate acyltransferase (GPAT), lysophospholipid acyltransferase (LPAT) and diacylglycerol acyltransferase (DGAT), and it is essential to modulate the yeast TAG biosynthetic pathway for higher CBL production. Results We cloned seven GPAT genes and three LPAT genes from cocoa cDNA, in order to screen for CBL biosynthetic gene candidates. By expressing these cloned cocoa genes and two synthesized cocoa DGAT genes in S. cerevisiae, we successfully increased total fatty acid production, TAG production and CBL production in some of the strains. In the best producer, the potential CBL content was eightfold higher than the control strain, suggesting the cocoa genes expressed in this strain were functional and might be responsible for CBL biosynthesis. Moreover, the potential CBL content increased 134-fold over the control Y29-TcD1 (IMX581 sct1Δ ale1Δ lro1Δ dga1Δ with TcDGAT1 expression) in strain Y29-441 (IMX581 sct1Δ ale1Δ lro1Δ dga1Δ with TcGPAT4, TcLPAT4 and TcDGAT1 expression) further suggesting cocoa GPAT and LPAT genes functioned in yeast. Conclusions We demonstrated that cocoa TAG biosynthetic genes functioned in S. cerevisiae and identified cocoa genes that may be involved in CBL production. Moreover, we found that expression of some cocoa CBL biosynthetic genes improved potential CBL production in S. cerevisiae, showing that metabolic engineering of yeast for cocoa butter production can be realized by manipulating the key enzymes GPAT, LPAT and DGAT in the TAG biosynthetic pathway. Electronic supplementary material The online version of this article (10.1186/s12934-018-0866-2) contains supplementary material, which is available to authorized users.

limited and insufficient, and developing other stable and sustainable sources of CB supply is of interest [1].
TAG biosynthesis is mainly catalyzed by glycerol-3-phosphate acyltransferase (GPAT), lysophospholipid acyltransferase (LPAT) and diacylglycerol acyltransferase (DGAT), which add acyl chains from acyl-coenzyme A (acyl-CoA) to the sn-1, sn-2 and sn-3 position of the backbone glycerol, respectively [10]. Therefore, besides sufficient C16 and C18 supply, modulation of yeast GPAT, LPAT and DGAT activity is essential for higher CBL production. There are two GPATs (Gpt2p and Sct1p), two LPATs (Slc1p and Ale1p), one DGAT (Dga1p) and one phospholipid:diacylglycerol acyltransferase, PDAT (Lro1p) that participate in TAG synthesis in S. cerevisiae ( Fig. 1) [11][12][13][14]. The simultaneous deletion of either the two GPAT genes or the two LPAT genes (SLC1 and ALE1) is lethal to S. cerevisiae, but double deletion of DGA1 and LRO1 genes does not affect yeast growth [11,15]. Single deletion of one GPAT gene or one LPAT gene alters yeast fatty acid profiles. For example, Sct1p has a preference towards C16 fatty acids and sct1∆ cells showed a different distribution of fatty acids in the phospholipids with less C16 fatty acids [11,15]. Thus, disruption of some genes of the yeast TAG biosynthetic pathway and introducing corresponding heterologous genes would likely alter yeast CBL production. Overexpression of several synthesized cocoa TAG biosynthetic genes increased the CBL content in S. cerevisiae, but CBL production of these S. cerevisiae strains was still low [9]. As there are thirteen GPAT and nine LPAT genes in the cocoa genome [16,17], deep and global characterization of cocoa GPAT and LPAT genes might reveal optimal cocoa genes responsible for CB biosynthesis, and expressing them in S. cerevisiae could improve CBL production [9].
Here, we cloned and expressed ten cocoa CB biosynthetic genes individually in S. cerevisiae, and compared the lipid production of the engineered yeasts. We also expressed some of the cloned cocoa genes together with two previously characterized cocoa DGAT genes in order to increase yeast CBL production. Moreover, we improved CBL production by metabolic engineering of the TAG biosynthetic pathway of S. cerevisiae.

Strains, plasmids, and media
The cloning host in this study was Escherichia coli strain DH5α. The Lipomyces starkeyi strain DSM 70296 was purchased from the culture collection of the DSMZ (Braunschweig, Germany). The S. cerevisiae strain CEN. PK 113-11C (MATa MAL2-8c SUC2 ura3-52 his3-Δ1), which was kindly provided by P. Kötter [18] and S. cerevisiae strain IMX581 (MATa ura3-52 can1∆::cas9-natNT2 TRP1 LEU2 HIS3) [19] were used in this study. S. cerevisiae CEN.PK 113-11C was used to express cocoa genes and homologous recombination was used to construct S. cerevisiae CEN.PK 113-11C-derived strains (Additional file 1: Figure S1). The plasmids pMEL10 and pMEL13 [19] were used for construction of strains derived from IMX581. All primers used to construct the yeast strains are listed in Additional file 1: Table S1, and all yeast strains constructed and used in this study are listed in Table 1. The strains YJ0 and SYJ0 were the same, but they were constructed and tested at different time as the control strain (Table 1).

Cocoa sample collection
The cocoa fruit samples were collected from the same tree in the Gothenburg Botanical Garden in a nearly ripe (May 26, 2015) and unripe (October 12, 2015) state, respectively (Additional file 1: Figure S2). One ripe cocoa fruit and several unripe cocoa fruits including small and big fruits were collected, respectively. After collection, the cocoa samples were cut into small pieces and immediately put into liquid nitrogen. The samples were kept at − 80 °C before use.

RNA preparation and cDNA synthesis
Cocoa fruits were grinded to fine powder in liquid N 2 using mortar and pestle, and the fine powder was divided into 100 mg aliquots in 1.5 ml Eppendorf tubes. Subsequently, 0.5 ml cold (4 °C) PureLink plant RNA reagent (Life Technologies) was added to each tube, mixed briefly by vortexing until the sample was thoroughly resuspended, and then the tubes were incubated for 5 min at room temperature. The resulting solution was cleared by centrifuging at 12,000g in a microcentrifuge for 2 min at room temperature. Next, the lysate was transferred to a QIAshredder spin column placed in a 2 ml collection tube and the instructions of the RNeasy plant mini kit (Qiagen) were followed to obtain total RNA. For extraction of RNA from L. starkeyi, the yeast was cultivated in YPD medium for 48 h at 30 °C and 200 rpm, and the biomass was collected by centrifugation at 12,000 rpm for 1 min. The yeast RNA was extracted using the Qiagen RNeasy mini kit. The cocoa and yeast RNA was transcribed into cDNA by reverse transcriptase using the Qiagen Quanti-Tect Reverse Transcription Kit.

Phylogenetic analysis of cocoa GPAT, LPAT and DGAT genes
The GPAT and LPAT gene sequences of T. cacao were downloaded from the Genbank database. Reference GPAT and LPAT genes sequences of Arabidopsis thaliana, Homo sapiens and S. cerevisiae were downloaded from the KEGG database [23]. Amino acid sequences of GPATs, LPATs or DGATs were aligned using the MAFFT online version [24], and the multiple alignment results were used to create phylogenetic trees using the MEGA 7.0.21 software [25]. The Neighbor-Joining method with Poisson correction was used to create the tree and the bootstrap confidence values were based on 1000 replicates. Moreover, gaps in the alignment of GPAT and LPAT sequences were treated with the pairwise deletion option.

Plasmid and yeast strain construction
Cocoa genes encoding GPATs and LPATs were amplified using primers described in Additional file 1: Table S1 from cocoa fruit cDNA using the PrimeSTAR HS DNA polymerase (Takara) according to the manufacturer's instruction. The one GPAT and one LPAT sequences amplified from cocoa cDNA, which were different from the available annotated genes, were deposited at the Gen-Bank database under the accession numbers MF352000-MF352001. The primers used to amplify cocoa genes, promoters and terminators are listed in Additional file 1: Table S1 and some primers used for cocoa gene combination expression were the same as described before [9]. The LsGPAT gene was cloned from L. starkeyi cDNA using PrimeSTAR HS DNA polymerase (Takara) and the sequence was the same as described before [26]. The cocoa gene expression cassettes were verified by sequencing and illustrated in Fig. 2. Gibson assembly (NEB) was used to construct cocoa gene expression plasmids by ligation of the gene expression cassettes and the amplified linear backbone fragment of plasmid pBS01A, and were further verified by PCR and Sanger sequencing (Additional file 1: Table S2). The constructed plasmids were used to transform S. cerevisiae to construct the strains listed in Table 1.

Shake flask cultivation and lipid analysis
Shake flask fermentations were carried out in 20 ml minimal medium and cultivated at 30 °C and 200 rpm for 24 h. [9,27]. Each strain was cultivated in three replicates, and the fatty acid methyl ester (FAME) and lipid profiles were analyzed using a microwave-assisted method [28,29]. In order to obtain sufficient lipids for TAG analyses, 5 l shake flasks containing 1 l NLM medium were used for yeast biomass collection [9,27]. The lipids extracted from each strain were used to analyze yeast TAG profiles by UPLC using RI detection [30]. The TAG compositions of each strain were expressed in relative area percentages [27,30]. CB standards and TAG composition sequences were completed by AAK, and TAG standards were purchased from Larodan by AAK.

Seven cocoa GPAT and three cocoa LPAT were cloned from cocoa cDNA
Usually, there are many GPAT and LPAT genes in one plant species, and 13 genes were annotated as GPAT genes and nine genes were annotated as LPAT genes (TcLPAT10 was removed as a result of standard genome annotation processing) in T. cacao [16,17]. To characterize potential GPAT and LPAT genes, we designed primers to clone all the possible GPAT and LPAT genes except two GPAT (TcGPAT1 and TcGPAT2) and two LPAT (TcLPAT1 and TcLPAT2) genes, which had been synthesized and characterized in yeast before [9]. A total of seven GPAT and three LPAT genes were cloned from cocoa fruit cDNA. Most of the gene sequences were consistent with the genes described in the genomic data, except two genes, TcGPAT12-4 and TcLPAT3-2, which were different from the published corresponding gene sequences (Fig. 3). TcGPAT12-4 and TcLPAT3-2 were 231 and 114 bp shorter than TcGPAT12 and TcLPAT3, respectively, this might be due to the alternative splicing in plants [31]. Though we tried several different PCR conditions to clone more cocoa GPAT and LPAT genes, no other genes were further cloned from cocoa cDNA samples.
All cocoa GPAT sequences and their reference sequences can be divided into five clades in a phylogenetic tree, and the cocoa GPAT sequences were distributed in four clades except Clade 4 which contained yeast GPAT genes (Fig. 3a). The highest level of identity between the amino acids sequence of any cocoa GPAT gene and the yeast GPAT genes is less than 15%. At the same time, the identities between yeast GPAT genes and the Lipomyces starkeyi GPAT gene are 45% (Gpt2p and LsGPAT) and 41% (Sct1p and LsGPAT), respectively (Fig. 3a). We successfully cloned TcGPAT3, TcGPAT4, TcGPAT5, TcGPAT9, TcGPAT10, and TcGPAT12. TcG-PAT2 had been characterized before, so that each clade had at least one representative cloned or synthesized cocoa GPAT sequence (Fig. 3a). The LPAT sequences can be divided into four clades, and the cocoa genes TcLPAT3, TcLPAT4 and TcLPAT5 all formed part of LPAT clade 1 (Fig. 3b). The two cocoa DGAT genes, TcDGAT1 and TcDGAT2, have been characterized before. Additional annotated DGATs (TcDGAT3-TcDGAT11) were similar Fig. 3 Phylogenetic analysis of GPAT (a) and LPAT (b) amino acid sequences with an unrooted tree. All neighbor-joining trees were constructed using the MEGA 7.0.21 software (bootstrap values: 1000) with the peptide sequences. Cocoa genes are marked with rhombuses, genes marked with red rhombuses are genes cloned from cDNA of T. cacao in this study, genes marked with green rhombuses are genes synthesized in a previous study, genes marked with black rhombuses are genes which were not cloned or synthesized in this study; yeast genes are marked with purple triangles; genes of A. thaliana and H. sapiens are not marked. The bootstrap values are marked above the nodes and the scale bar is indicated under each tree with wax ester synthase genes of A. thaliana, hinting they would not participate in TAG biosynthesis [9].

Expression of single cocoa genes in S. cerevisiae changed its total fatty acid production
The cloned cocoa genes were assembled in expression cassettes with strong constitutive promoters and ligated into plasmid pBS01A as described in Fig. 2 and Additional file 1: Table S2. The empty plasmid pBS01A and 10 other plasmids harboring cocoa genes were introduced into S. cerevisiae CEN.PK 113-11C, resulting in the control strain YJ0 and another 10 yeast strains, respectively (Table 1). Fatty acid analysis showed that the most abundant fatty acids in each strain were C16 and C18 fatty acids and some yeast strains harboring cocoa genes produced more fatty acids than the control strain YJ0 (Fig. 4). Especially, yeast strains YJ-G03, YJ-G04, YJ-L03 and YJ-L04 displayed a significant increase of the total fatty acid amount compared to YJ0, indicating that TcG-PAT3, TcGPAT4, TcLPAT3-2 and TcLPAT4 were active in the yeast and might be engaged in cocoa CB biosynthesis.
The neutral lipids steryl esters (SE) and TAG are the two main storage lipids in S. cerevisiae, whose production is affected by the expression of GPAT, LPAT and DGAT genes [14,32,33]. The SE content analysis showed that SYJ-341, SYJ-342 and SYJ-441 displayed a significant increase over the control strain SYJ0, while the TAG content analysis showed that SYJ-331, SYJ-341, SYJ-342 and SYJ-441 displayed a significant increase over the control (Fig. 5b), suggesting expression of some cocoa gene combinations were beneficial for yeast storage lipid production. Among the eight strains, SYJ-341 and SYJ-441 produced 150 and 320% more SE than SYJ0, respectively, and they also produced 210% and 290% more TAG than SYJ0, respectively.

Cocoa gene expression partially restored neutral lipid production in a S. cerevisiae mutant with a partial TAG biosynthesis pathway
To further verify if the cloned cocoa GPAT and LPAT genes functioned in S. cerevisiae, we deleted parts of the yeast TAG biosynthetic pathway, including the GPAT gene SCT1, the LPAT gene ALE1, the DGAT gene DGA1 and the PDAT gene LRO1, to construct new yeast strain Y29 and expressed TAG biosynthetic genes in it (Table 1). Compared with the control strain SYJ0, the SE content was reduced, but specifically the TAG content drastically decreased (Figs. 5b, 6b). With regards to total fatty acids, Y29-TcD1 and Y29-441 displayed a significant difference with the control Y29-pBS01A, and the concentration of C16:0 in Y29-TcD2 and Y29-331 increased and showed significant differences with the control (Fig. 6a).
Concerning the neutral lipids, SE and TAG in Y29-331 and Y29-441 displayed significant increase over the control strain. Besides, SE in Y29-TcD2 also showed significant increase compared to the control strain (Fig. 6b). Moreover, we also attempted to delete one yeast GPAT gene (SCT1 or GPT2) and replace the other gene with one cocoa GPAT gene, but we failed to obtain viable colonies. However, when we used LsGPAT of L. starkeyi to replace yeast SCT1 in the yeast strain carrying the GPT2 deletion, it succeeded and yeast fatty acid and lipid production changed, indicating that a single of the cloned or previously synthesized cocoa GPAT gene cannot replace the function of yeast GPAT genes (Additional file 1: Figure S3). Unfortunately, the CBL production of yeast strain YJW09 was lower than the wild-type strain, though the CBL content of L starkeyi is higher than that of S. cerevisiae (Additional file 1: Figure S3B) [27].

Expression of cocoa gene combinations led to an increase in potential CBL production
We compared TAG production of four different yeast strains: control strain SYJ0, SYJ-331, which displayed a significantly higher fatty acid and TAG content compared to SYJ0, SYJ-342, which had an increased SE and TAG content, and SYJ-441, which displayed the highest amounts of total fatty acids, SE and TAG of the eight strains. The TAG compositions changed after introducing the cocoa genes in S. cerevisiae (Additional file 1: Figure S4A). Though all the four yeast strains produced more than 20 different kinds of TAGs, most of the TAGs only accounted for less than 5% of the total TAGs (Additional file 1: Figure S4A). Concerning the relative CBL production, SYJ-331 (4.7%) and SYJ-441 (6.9%) produced 64 and 140% more potential CBL than the control strain SYJ0 (2.87%), respectively, but SYJ-342 produced less CBL than the control (Fig. 7a). The potential POP and POS of SYJ-331 and SYJ-441 had increased by at least 30% and by at least 130%, respectively. The potential SOS (C18:0, C18:1, C18:0) proportion of SYJ-441 displayed a significant difference compared with SYJ0 (Fig. 7a). In fact, the potential POP, POS and SOS proportions of SYJ-441 had increased 143, 130 and 164%, respectively, compared with SYJ0. Considering that SYJ-441 produced 288% more TAGs than SYJ0, its potential CBL production increased more than eightfold compared with SYJ0, showing this cocoa gene combination of TcGPAT4, TcLPAT4 and TcDGAT1 not only increased TAG production, but also helped S. cerevisiae accumulate more potential CBL.

Expression of cocoa genes in different yeast strains changed fatty acid profiles and TAG composition
The C16 and C18 fatty acids were the main fatty acids in the TAGs of four yeast strains (SYJ0, SYJ-331, -342 and -441) and their proportions were similar, which is consistent with the total fatty acid composition analyses (Figs. 5a, 6a). Generally, saturated fatty acids in the TAGs of SYJ-331 and -441 increased, but decreased in the TAGs of SYJ-342 (Fig. 8a). Compared with SYJ0, for SYJ-331, the C16:0 portion in the TAGs increased, while other C16 and C18 fatty acids in the TAGs were equal or decreased; for SYJ-342, the C16:1 fraction in the TAGs increased, while other C16 and C18 fatty acids in the TAGs were equal or decreased; for SYJ-441, both C16:0 and C18:0 fatty acids in the TAGs increased with a significant difference, while other C16 and C18 fatty acids in the TAGs were equal or decreased. Though C16 and C18 fatty acids were the main fatty acids in the TAGs of three Y29 strains, the C16 fatty acid faction decreased in Y29-331 and Y29-441 and C18 fatty acids showed an increase in the TAGs of Y29-331 and Y29-441 compared with Y29-TcD1 (Fig. 8b). Besides, saturated fatty acids in the TAGs of Y29-331 and Y29-441 decreased compared with Y29-TcD1. In details, C14:0, C16:0, C16:1 fractions in the TAGs of Y29-331 and -441 decreased, while C18:0 and C18:1 increased. Besides, the C18:2 and C20:0 content in the TAGs of Y29-441 increased. However, the content of unknown TAGs was higher in these three Y29 strains (Y29-TcD1, Y29-331 and Y29-441) than in the four yeast strains (SYJ0, SYJ-331, -342 and -441) and the TAG profiles of Y29-derived strains were different from the other four yeast strains, further suggesting that cocoa GPAT and LPAT enzymes were active and altered the TAG profiles in the yeast strains (Fig. 8).

Discussion
CB is mainly extracted from cocoa beans, but the genes of its biosynthetic pathway in cocoa tree are unknown, hampering their application for developing microbial based CBL production [1,16,17]. 13 GPAT, 9 LPAT and 11 DGAT genes have been identified in the T. cacao genome [9,16,17]. Of the DGAT genes, only TcDGAT1 and TcDGAT2 are similar to the yeast DGAT genes, the others are similar to LRO1 of S. cerevisiae or wax ester synthase genes, and they might not participate in CB production. Two GPAT (TcGPAT1 and TcGPAT2), two LPAT (TcLPAT1 and TcLPAT2) and two DGAT (TcD-GAT1 and TcDGAT2) genes of T. cacao have been synthesized and characterized before [9]. In this study, we cloned another seven GPAT genes and three LPAT genes from seeds of T. cacao. Expression of some of these genes (TcGPAT3, TcGPAT4, TcLPAT3-2 and TcLPAT4) in yeast resulted in an increased total fatty acid production, indicating that these cocoa genes might play roles in cocoa CB production. Furthermore, expression of cocoa gene combinations including these genes plus the previously characterized DGAT genes (TcGPAT3, TcG-PAT4, TcLPAT3-2, TcLPAT4, TcDGAT1 and TcDGAT2) increased fatty acid, neutral lipid and CBL production significantly, suggesting that cocoa GPATs and LPATs functioned in yeast and contributed to a CBL production increase in this heterologous host. The esterification of acyl-CoAs to the glycerol backbone during TAG biosynthesis is catalyzed by GPAT, LPAT and DGAT [13]. Our results showed that SE and TAG production of strain Y29 decreased drastically, demonstrating that deletion of yeast TAG biosynthesis genes altered total neutral lipid production [11,15]. Compared with Y29-derived strains harboring either an empty plasmid or only one of the cocoa DGAT genes (Y29-TcD1 and Y29-TcD2), expression of cocoa gene combinations (Y29-331 and Y29-441) restored part of yeast neutral lipid production and CBL production, suggesting cocoa GPAT and LPAT contributed to the yeast CBL production increase. Furthermore, both GPAT genes of S. cerevisiae were replaced by LsGPAT of L. starkeyi, showing that the GPAT genes of S. cerevisiae can be replaced by homologues from the same clade (Clade 4 in Fig. 3a). However, replacement of both yeast GPAT genes with one cocoa GPAT gene (at least under the conditions used in this study) failed, suggesting that overexpression of one cocoa GPAT gene cannot complement the loss of both GPAT genes in S. cerevisiae. Moreover, neutral lipid production of strain Y29 harboring one cocoa GPAT gene, one cocoa LPAT gene and one cocoa DGAT gene was much lower than the wild-type yeast strains, further demonstrating cocoa TAG biosynthetic gene expression cannot compensate the gene loss in S. cerevisiae. The low identities between cocoa GPAT genes and yeast GPAT genes might be the reason why cocoa GPAT genes cannot replace yeast GPAT genes (Fig. 3a). As the model plant Arabidopsis thaliana contains several genes encoding enzymes with GPAT activities, the situation for the cocoa tree might be similar, pointing to that several genes/ enzymes may function simultaneously in vivo [34].
YJ0 and SYJ0 are supposed to have the same genotype, but when we cultivated these two strains in the same conditions (but not in the same experiment), the fatty acid production of YJ0 and SYJ0 was different. This might be due to the fact that YJ0 and SYJ0 were constructed at different time points and/or that they were in a different growth state when sampled. Overexpression of cocoa TAG biosynthetic genes increased the CBL content from 2.87% in the wild-type yeast strain SYJ0 to 6.9% in SYJ-441. In addition, overexpression of cocoa TAG biosynthetic genes in Y29 increased the CBL content, and the highest CBL content of Y29-441 was 6.0%. With regard to the less than 5.4% CBL content in the yeast strain YJ-221 harboring specifically selected cocoa TAG biosynthetic genes [9], this demonstrates that combination of phylogenetic analyses and experimental verification can lead to identification of effective enzymes (TcGPAT4 and TcLPAT4) for CBL production. The relative CBL content in SYJ-441 is still low (only 6.9%) and the relative CBL content in YJW09 harboring LsGPAT is even lower than in the wild-type S. cerevisiae strain, hinting that other metabolic engineering strategies should be attempted to increase yeast CBL production, including an improvement of specific fatty acid supply and the identification of other functional TAG biosynthetic genes from plants [35][36][37][38]. Considering that the C16:1 proportion in the yeast TAGs is more than 40% and that the C16:1 content in CB is very low [1], some strategies should be further implemented to alter the total fatty acid profiles of the yeast, such as screening for heterologous desaturases which can decrease C16:1 production and increase C18:1 production. Besides, the C16:1 content in the TAGs of some of the Y29-derived strains harboring cocoa gene combinations was lower than in the control strain of Y29-TcD1 (Fig. 8b), which is another indication that these cocoa genes are potential CB biosynthetic genes. As several GPAT, LPAT or DGAT genes might be responsible for CB production in vivo simultaneously, overexpression PK 113-11C-derived (a) and Y29-derived (b) strains. The error bars represent the standard deviation of two biological replicates. Shown is the peak area of the respective TAG in comparison to the summed peak areas of all TAGs. Asterisks (*) indicate significant differences in comparison to reference strain SYJ0 (a) or Y29-TcD1 (b), respectively; *p < 0.05; **p < 0.01. The p values are calculated based on paired t tests corrected for multiple comparisons of different cocoa TAG biosynthetic genes in one yeast strain would be a potential strategy to increase CBL production further. In the future, also overexpression of the identified genes in oleaginous yeasts represents a possibility to obtain a high-level of CBL production [27].

Conclusions
Ten different cocoa TAG biosynthetic genes were cloned from cDNA of T. cacao and characterized in yeast. Expression of some cocoa genes in a wild-type S. cerevisiae strain increased CBL production more than eightfold (the relative CBL content is 6.9%) over the wild-type strain, and expression in yeast mutant strain with a reduced TAG pathway could increase CBL production 134-fold (the relative CBL content is 6.0%) over the control strain Y29-TcD1. Besides, our results demonstrate that several cocoa GPAT, LPAT and DGAT genes might function simultaneously in T. cacao for CB production.   Figure S4. Relative TAG content (except potential CBL) of CEN.PK 113-11C-derived strains (A) and Y29-derived strains (B). (A) Relative TAG content (except potential CBL) of CEN.PK 113-11C-derived strains harboring cocoa genes. The error bars represent the standard deviation of two biological replicates. Shown is the peak area of the respective TAG in comparison to the summed peak areas of all TAGs. Asterisks (*) indicate significant differences (p-values are based on paired t-tests corrected for multiple comparisons) in comparison to control SYJ0. * indicates p < 0.05; ** indicates p < 0.01. (B) Relative TAG content (except potential CBL) of Y29-derived strains harboring 3 cocoa genes or 1 cocoa gene. Asterisks (*) indicate significant differences (p-values are based on paired t-tests corrected for multiple comparisons) in comparison to reference strain Y29-TcD1. * indicates p < 0.05; ** indicates p < 0.01. Table S1. List of primers used in this study. Table S2. List of plasmids constructed in this study.