P. pastoris X-33 (wild type), GS115 (his4) (both from Invitrogen, Carlsbad, CA, USA), P. guilliermondii DSM 70051

Escherichia coli Top10 (Invitrogen), E. coli NovaBlue (Novagen)

If not stated otherwise, standard procedures were used for DNA manipulation. PCR was carried out with KOD XL Polymerase supplied by Novagen, according to the Novagen user protocol. For oligonucleotide primers used in this work see Table 1.

Transformation of P. pastoris by electroporation was performed following the standard protocol provided by the Pichia manuals from Invitrogen. Transformants were selected on 100 μg/mL Zeocin and 500 μg/mL Geneticin (Invitrogen) respectively.

All the restriction enzymes, Calf Intestine Phosphatase, T4 DNA Polymerase and T4 DNA Ligase were supplied by New England Biolabs. For the overexpression of the RIB genes the expression vectors: pGAPZB (Invitrogen) and pGAPHIS 15 were employed. The PCR products derived from genomic P. pastoris X-33 DNA were amplified with the corresponding primers shown in Table 1. The expression vectors and RIB gene PCR products were digested with the restriction enzymes SfiI and NotI and ligated with T4 DNA Ligase to gain the desired RIB gene expression vectors. For the construction of the promoter replacement cassettes (PRC) the pSTBlue-1 standard cloning vector (Novagen) was used. The stepwise assembly of the PRC started with the blunt cloning of the ZeoLox cassette into the EcoRV site of pSTBlue-1. For construction of the ZeoLox cassette, the Zeocin resistance cassette was cut from pGAPZB and cloned into pUG6 16, thereby replacing the kanMX cassette between the loxP sites. The pSTBlue_ZeoLox vector was SacI digested and blunt ended with the T4 DNA polymerase and subsequently NotI digested. This procedure gave the acceptor vector for the GAP promoter fragment including the multiple cloning site of pGAPZB (Invitrogen) which was BglII digested blunt ended and subsequently NotI digested. This cloning step gave pSTBlue_ZeoLox_GAP_MCS in which the RIB genes were SfiI/NotI inserted as described above. The PRC was completed by the BamHI cloning of the promoter region PCR products into the unique BamHI site upstream of the ZeoLox cassette. The P. pastoris clones gained by transformation with the expression vectors and the PRC are listed in Table 2. For construction of the transient Cre-recombinase expression vector pKTAC-CRE, the kanMX4 cassette from pFA6-kanMX4 17 was cut BglII/SacI and inserted with blunt ends into the plasmid pYX022 (R&D Systems), opened AatII/KpnI, blunt ended. The resulting vector was cut with BglII, blunt ended and combined with the ARS/CEN cassette from pYC131 18, cut FseI, blunt ended. The cre-recombinase from pSH47 19 was inserted into the MCS of this vector. The ARS/CEN confers sufficient stability in P. pastoris for transient expression of the recombinase, but curing of the strains from the plasmid is easily possible by further cultivation without antibiotic.

Shake flask experiments were carried out in 250 mL baffled shake flasks on a Multitron II shaker (Infors, Switzerland) at 28°C with 180 rpm. Defined culture medium was composed of 1.34% (w/v) Difco™ Yeast Nitrogen Base without Amino Acids (Becton, Dickinson and Company, Sparks, MD, USA), 5% (w/v) glucose, 100 mM potassium phosphate pH 6.0, 4 × 10^-5% (w/v) biotin, 4 × 10^-3% (w/v) L-histidine, 5 × 10^-3% (w/v) L-glutamic acid, L-methionine, L-lysine, L-leucine, and L-isoleucine, 0,3% (w/v) CaCO₃.

A preculture of the respective P. pastoris strains incubated with shaking at 28°C for 24 h on YPG (per liter: 10 g yeast extract, 10 g peptone, 10 g glycerol) was used to inoculate the starting volume (1.75 liters of batch medium) of the bioreactors to a starting optical density at 600 nm of 1.0. Fermentations were carried out in 5.0-liter working volume bioreactors (Minifors, Infors, Switzerland) with a computer-based process control. Fermentation temperature was controlled at 25°C, pH was controlled at 5.0 with addition of 25% ammonium hydroxide, and the dissolved-oxygen concentration was maintained above 20% saturation by controlling the stirrer speed between 600 and 1,200 rpm, whereas the airflow was kept constant at 100 liters h^-1.

The batch medium contained (per liter) 2.0 g citric acid, 12.4 g (NH₄)₂HPO₄, 0.022 g CaCl₂·2H₂O, 0.9 g KCl, 0.5 g MgSO₄·7H₂O, 46.5 g glycerol, and 4.6 ml PTM₁ trace salts stock solution. The pH was adjusted to 5.0 with 25% HCl. The glucose fed-batch solution contained (per liter) 550 g glucose·1 H₂O, 10 g KCl, 6.45 g MgSO₄·7H₂O, 0.35 g CaCl₂·2H₂O, and 12 ml PTM₁ trace salts stock solution. The PTM₁ trace salts stock solution contained (per liter) 6.0 g CuSO₄·5H₂O, 0.08 g NaI, 3.0 g MnSO₄·H₂O, 0.2 g Na₂MoO₄·2H₂O, 0.02 g H₃BO₃, 0.5 g CoCl₂, 20.0 g ZnCl₂, 65.0 g FeSO₄·7H₂O, 0.2 g biotin, and 5.0 ml H₂SO₄ (95 to 98%). All chemicals for PTM₁ trace salts stock solution were from Riedel-de Haën (Seelze, Germany), except for biotin (Sigma, St. Louis, MO) and H₂SO₄ (Merck Eurolab).

After approximately 50 h, the batch was finished and the glucose fed batch with a feed rate of 16 g h^-1 was started for a period of about 160 h. The cultivations were prolonged between 24 and 50 h (without feed) to analyse if further riboflavin was accumulated. Samples were taken frequently and processed as described previously in 20.

Flow cytometric analyses were performed on a FACSCalibur instrument (Becton Dickinson, Franklin Lakes, N.J.). The cells were excited by using a 15 mW, 488 nm air-cooled argon ion laser, and the fluorescence emission was measured through a 530 ± 15 nm band-pass filter (FL1). Threshold settings were adjusted so that the cell debris were excluded from the data acquisition. A total of 10,000 cells were measured for every sample. Data analysis was performed afterwards with WinMDI 2.8 software, version 1.0 21.

Cells were grown on defined culture medium as described above. Fluorescence microscopy was performed by using a LEICA TCS SP2 microscope (excitation 488 nm, emission 494–541 nm).

Riboflavin concentrations were determined by HPLC using a Sigma Nucleosil C18 (10 mm * 4.6 mm ID, 5 μm) guard column and a Sigma Nucleosil C18 (150 mm * 4.6 mm ID, 5 μm) column with an isocratic flow of 1 mL/min running buffer (50 mM NaH₂PO₄-H₃PO₄ pH = 3; 1 mM tetramethyl ammonium chloride; 12% acetonitrile (v/v)) 22.

Culture supernatant was achieved by centrifugation in eppendorf 2 mL tubes for 1 minute at 13,000 rpm. Prior to injection on the HPLC column the samples were mixed with a 2-times concentrated running buffer in an equal ratio and filtered over PVDF Durapore Syringe Driven Filter Unit (Millipore) with a pore size of 0.22 μm. 100 μL of the final samples were loaded on the HPLC column and quantified by the peak height of absorption, detected at 223 nm and 445 nm respectively.

Previously published research on the formation of riboflavin in different riboflavin overproducing organisms indicates that the enzymatic activity of GTP-cyclohydrolase II (RIB1, Figure 1) plays a key regulatory role in the riboflavin biosynthetic pathway. Consequently, the overexpression of this gene was the first step for deregulation of riboflavin biosynthesis in P. pastoris. Since initially, the genome sequence of this yeast was not available to us we decided to express the GTP-cyclohydrolase II gene of S. cerevisiae (ScRIB1) 23.

The respective gene was PCR-amplified from genomic DNA and cloned into the P. pastoris expression vector pGAPZB, from where the gene is expressed under control of the constitutive glyceraldehyde-3-phosphate dehydrogenase promoter (P_GAP). Transformation of P. pastoris X-33 led to yellow colonies, clearly indicating the accumulation of a flavin. Furthermore, flow cytometric analysis revealed a significantly increased (auto-) fluorescence signal of the clones expressing ScRIB1 in comparison to the wild type (wt) strain (Figure 2). The wavelength of FL1 (530 nm ± 15 nm) corresponds to the fluorescent emission of riboflavin pointing to an increased riboflavin content of the cells. Intracellular riboflavin is localized in the vacuole (data not shown), as previously described for A. gossypii 24.

The production of riboflavin of wt P. pastoris X-33 and X-33 ScRIB1 was compared to the flavinogenic yeast P. guilliermondii in shake flasks cultures using minimal medium and glucose as carbon source. Riboflavin was proven to be stable under these conditions in presence of yeast cells for at least 10 days (data not shown). While the wt P. pastoris X-33 did not produce any significant amount of riboflavin, X-33 ScRIB1 accumulated 4 mg/L and P. guilliermondii produced 8 mg/L of riboflavin. This result clearly indicates that the single gene overexpression of RIB1 changed the phenotype of P. pastoris to flavin overproduction. Interestingly, both, P. pastoris X-33 ScRIB1 and P. guilliermondii accumulated riboflavin in the culture supernatant in the late stationary phase, when the glucose in the culture broth was completely consumed. Hence, flavin production is at least partly uncoupled from growth in the naturally flavinogenic yeast as well as the recombinant P. pastoris strain.

In order to test whether the upstream pathways of riboflavin synthesis, particularly the provision of GTP, are the next rate limiting steps, precursors for GTP synthesis (as glycine, threonine, and glutamine 2) were fed to shake flask cultures of P. pastoris X-33 ScRIB1. However, no positive impact on riboflavin synthesis could be detected. Similarly, UV mutagenesis followed by selection on deoxyglucose (for a deregulated glycolytic flux or uptake) or tubercidin (7-deaza-adenosine, for deregulated GTP formation) did not yield any positive effect on riboflavin production. These approaches have been successfully applied to enhance riboflavin synthesis in A. gossypii 2 and Candida famata 25. The fact that they do not have an effect on this P. pastoris strain indicates that further limiting step(s) have to be found downstream, within the riboflavin pathway itself. However, the data do not allow the conclusion that GTP synthesis would not limit riboflavin production at all, especially at higher rates.

Five known enzymes are located downstream of RIB1 (Figure 1), being potentially limiting for riboflavin production. Individual overexpression of each of these genes was examined in view of riboflavin production of the resulting recombinant strain. Since the P. pastoris genome sequence became available to us, we decided to proceed with the overexpression of the homologous P. pastoris genes instead of heterologous genes from S. cerevisiae. All RIB genes from P. pastoris were PCR-amplified from genomic DNA and cloned into the P. pastoris expression vector pGAPZB. Expression of PpRIB1 in P. pastoris X-33 led to yellow colonies as has been noted for overexpression of ScRIB1. However, none of the other genes showed any effect concerning neither the colony color nor the production of riboflavin in shake flask cultures when overexpressed from pGAPZB in P. pastoris X-33. This indicates that constitutive overexpression of RIB1 is an essential prerequisite for riboflavin overproduction in P. pastoris.

Consequently, for the identification of further pacemaker enzymes within the pathway, the other RIB genes had to be individually co-expressed in combination with RIB1. PpRIB1 was cloned into the GAPHIS expression vector and integrated into the HIS4 locus of a P. pastoris GS115 strain. The riboflavin production of the P. pastoris GS115 PpRIB1 was tested in shake flask cultures and an equal production of riboflavin compared to X-33 PpRIB1 was observed (Table 3). GS115 PpRIB1 was further transformed with the other RIB genes in pGAPZ, resulting in the strains: GS115 PpRIB1+empty, GS115 PpRIB1+1, GS115 PpRIB1+2, GS115 PpRIB1+3, GS115 PpRIB1+4, GS115 PpRIB1+5, and GS115 PpRIB1+7.

A combination of RIB1 with RIB3, RIB7 or a second copy of RIB1 enhances the riboflavin production in shake flask experiments as shown in Table 3. While a strain with a single copy of RIB1 accumulates 3 mg/L of riboflavin, a strain overexpressing RIB1 and RIB7 accumulates 5 mg/L. Two copies of RIB1 lead to an accumulation of 7 mg/L and the combination of RIB1 with RIB3 results in an accumulation of 10 mg/L of riboflavin, which is more than P. guilliermondii accumulates under comparable conditions. Co-expression of any other gene from the riboflavin pathway (RIB2, RIB4 or RIB5, coding for the last enzymes of the metabolic pathway) did not show any increase of riboflavin production. The bottleneck for riboflavin production appears therefore to be in the beginning of the pathway as the first three genes show an impact on final product concentration, but not the genes located in the final part of the pathway.

It is interesting to note that the addition of a second copy of RIB1 leads to a significantly more intense yellow color of the colonies compared to a strain overexpressing only one copy of RIB1. This is in contrast to colonies of a strain combining RIB1 and RIB3, which produces even more riboflavin in shake flask cultures then the double RIB1 strain. The addition of RIB3 does not change the color of the colonies significantly. However, the agar around the colonies becomes intensely yellow colored. This phenomenon does not occur for any of the other strains. It appears therefore, that the double RIB1 clones accumulate more riboflavin intracellularly, whereas, the riboflavin from the RIB1/RIB3 clones diffuses more efficiently into the agar. We have currently no explanation, which role RIB3 plays for diffusion or export of riboflavin. Further research has to be dedicated to this question.

The next step for the construction of an improved riboflavin producing P. pastoris strain was the combination of the early genes of the pathway and subsequently to test if the late genes have any impact at all for riboflavin production in P. pastoris.

If conventional techniques are to be used, six markers are required for the overexpression of six different genes – as was the goal of this work. However, P. pastoris strains carrying six auxotrophies have not been constructed yet and the number of possible dominant markers is also limited. Consequently, only the rescue and repeated use of the same marker allows the introduction of a higher number of recombinant DNA constructs. The use of the cre/lox system allows the easy rescue of any desired marker 26 and 27. However, a disadvantage is that the constructs have to be introduced with sufficient distance into the genome to make sure that the lox recombination eliminates the desired marker cassette but does not destabilize other parts of the genome.

In order to fulfill these requirements we opted for a replacement of the native promoter of every single RIB gene with the strong and constitutive P_GAP of P. pastoris, instead of introducing further copies of the genes. Figure 3 shows a schematic representation of the respective promoter replacement cassettes (PRC). The cassettes comprise 500 bp of homologous sequence on both sides to allow for targeted integration into the genome. Integration of the cassettes leads to a deletion of 200 bp directly upstream of the ATG, which putatively comprise the core elements of the native promoters. These 200 bp are replaced with the Zeocin resistance cassette flanked by two loxP sites and the desired P_GAP promoter element, providing a strong constitutive expression of the RIB gene.

Transformation of the RIB1 PRC into P. pastoris X-33 changed the colony color to yellow like it was observed for all strains overexpressing ScRIB1 or PpRIB1. Marker removal by transient expression of the Cre-recombinase allowed the next round of transformation to target the next gene. An interesting notion relates to the screening of a first round of clones that had been transformed with a cassette comprising the complete RIB1 coding sequence and not only the first 500 bp. It turned out that the yellow color of the selected colonies was significantly reduced after marker rescue. By PCR we proved that the clones with the most intense yellow color integrated the expression cassette more than one time. The cre/lox recombination eliminated these cassettes and only one remained. By using only 500 bp of the coding sequence for the recombination cassettes we made sure that even if multiple cassettes are introduced by the first transformation only one functional copy of the gene is generated, avoiding any phenotypic difference of the clones before and after cre/lox recombination.

Table 4 shows the results of shake flask cultures of the P. pastoris strains with sequential deregulation of the RIB genes. As expected, the overexpression of only RIB1 leads already to a significant accumulation of riboflavin in the culture supernatant (7 mg/L). X-33 PRC RIB1+3 and X-33 PRC RIB1+3+7 produce 9 mg/L or 10 mg/L respectively. As presumed the addition of RIB3 leads to an increased accumulation of the vitamin, but subsequent addition of RIB7 appears to have only a minor effect.

To avoid possible metabolic problems caused by the accumulation of intermediate products of the riboflavin synthesis we decided to increase the catalytic turnover of the last enzyme in the pathway as the following step. The resulting strain X-33 PRC RIB1+3+7+5 accumulates 13 mg/L riboflavin. Addition of RIB2 has a further slightly positive effect (15 mg/L) and closing the last gap by further overexpression of RIB4 has the most pronounced effect on riboflavin accumulation. The strain X-33 PRC RIB1+3+7+5+2+4 accumulates 20 mg/L of riboflavin in shake flask cultures.

While product accumulation is steadily increasing with the addition of further genes, the autofluorescence of the cells measured by flow cytometry is not (data not shown). This indicates that the amount of riboflavin retained inside of the cells reaches an upper limit and any further produced vitamin is secreted (or diffusing) into the culture supernatant.

The sequential increase of riboflavin accumulation with the addition of further RIB genes shows clearly that the regulation of riboflavin production is distributed over and dependent on more genes and enzymes of the entire pathway and not on just one single pacemaker enzyme alone. RIB1 is the first bottleneck – without increase of RIB1 no increase in the flux is possible. However, at least RIB3, RIB5, RIB2 and RIB4 play significant roles in controlling the pathway too. Addition of glycine, a precursor for GTP has no or only a slight impact on riboflavin production of all strains (data not shown), indicating that still the pathway itself is limiting and not the supply of precursors.

To assess the future potential of the constructed strains and to verify that the trend seen in the shake flask experiments holds true also for high cell density cultivations, the strains X-33 PRC RIB1 and X-33 PRC RIB1+3+7+5+2+4 were grown in fed-batch mode in bioreactors. Figure 4 shows the corresponding results. While the wt strain does not accumulate riboflavin in high cell density cultures grown on glucose as carbon source (data not shown) the strain overexpressing only RIB1 accumulates 33 mg/L in the culture broth, which corresponds to 50 mg/L in the supernatant. (The difference is due to the volume of the cells, which cannot be neglected in high cell density cultures). The strain overexpressing all of the genes from the riboflavin biosynthetic pathway (X-33 PRC RIB1+3+7+5+2+4) accumulates under the same conditions more than 125 mg/L in the culture broth, which corresponds to 175 mg/L in the supernatant. Confirming the indications from shake flask experiments a major part of the riboflavin accumulation takes place towards the end of the cultures, when the growth rate of the cells is already significantly reduced. Riboflavin accumulation is therefore at least partly uncoupled from growth. This is particularly surprising as the promoter driving the expression of the genes is most active in growing cells as is the supply of the precursors GTP and ribulose-5-phosphate.