A novel methanol-free Pichia pastoris system for recombinant protein expression

Background As one of the most popular expression systems, recombinant protein expression in Pichia pastoris relies on the AOX1 promoter (PAOX1) which is strongly induced by methanol. However, the toxic and inflammatory nature of methanol restricts its application, especially in edible and medical products. Therefore, constructing a novel methanol-free system becomes necessary. The kinases involved in PAOX1 activation or repression by different carbon sources may be promising targets. Results We identified two kinase mutants: Δgut1 and Δdak, both of which showed strong alcohol oxidase activity under non-methanol carbon sources. Based on these two kinases, we constructed two methanol-free expression systems: Δgut1-HpGCY1-glycerol (PAOX1 induced by glycerol) and Δdak-DHA (PAOX1 induced by DHA). By comparing their GFP expression efficiencies, the latter one showed better potential. To further test the Δdak-DHA system, three more recombinant proteins were expressed as examples. We found that the expression ability of our novel methanol-free Δdak-DHA system was generally better than the constitutive GAP promoter, and reached 50–60 % of the traditional methanol induced system. Conclusions We successfully constructed a novel methanol-free expression system Δdak-DHA. This modified expression platform preserved the favorable regulatable nature of PAOX1, providing a potential alternative to the traditional system. Electronic supplementary material The online version of this article (doi:10.1186/s12934-016-0578-4) contains supplementary material, which is available to authorized users.


Background
Methylotrophic yeast refers to a limited number of yeast species which are able to utilize methanol as the sole carbon and energy source for cell growth. Pichia pastoris, Hansenula polymorpha, Candida boidinii and Pichia methanolica are the most typical examples [1,2]. In order to metabolize methanol, these yeast species express an alcohol oxidase, named Aox in P. pastoris, Mox in H. polymorpha and Aod in C. boidinii [3]. Pichia pastoris has two alcohol oxidase coding genes, AOX1 and AOX2. The strength of AOX1 promoter (P AOX1 ) is much stronger than P AOX2 under methanol induction, therefore AOX1 is the major source of methanol-oxidizing activity [4]. Besides, in methanol cultured P. pastoris cells Aox protein level could reach 30 % of total soluble proteins [5]. Pichia pastoris has been exploited as an excellent heterologous protein expression system in 1980s [1]. So far, over 5000 recombinant proteins have been successfully expressed in P. pastoris including insulin, α-interferon and hepatitis B antigen [6] (http://www.pichia.com/). The increasing popularity of this particular expression system could be attributed to the following reasons [1,7,8]: (1) The P. pastoris genome has been completely sequenced and a lot of genetic manipulation tools are available; (2) The culture condition is simple, and cells can do highdensity culture with high levels of protein expressed at the intra-or extra-cellular level; (3) As a eukaryote, P. pastoris is able to perform special modifications such as glycosylation.

Open Access
Microbial Cell Factories In most cases, recombinant protein expression is driven by P AOX1 in P. pastoris. P AOX1 is induced only by methanol and repressed by other carbon sources such as glucose, glycerol and ethanol [4]. This special induction and repression feature functions as a switch which turns recombinant protein expression on and off under different culture conditions. This is beneficial especially when expressing proteins that are toxic towards cell growth. However, this system also has limitations. Since P AOX1 induction requires methanol, this toxic and inflammable material needs special handling and is not suitable for producing edible and medical products [8,9]. In addition, the by-product hydrogen peroxide (H 2 O 2 ) of methanol metabolism brings oxidative stress, which may result in the degradation of recombinant proteins [10,11].
One way to solve the problem is to develop a methanolfree expression system, which does not rely on methanol to induce the AOX1 promoter. Since P AOX1 is activated by methanol and repressed by glucose and glycerol, interrupting the glucose/glycerol repression pathway, or activating the methanol activation pathway will be a good strategy. The activation or repression by carbon molecules towards P AOX1 is not direct, but rather through complicated signaling pathways which have not been fully understood yet. So far several protein factors have been reported to be involved in the alcohol oxidase gene promoter regulation. One example lies in the hexose transporter and sensor family. Gcr1 in H. polymorpha [9,12] and Hxt1 in P. pastoris [8] are hexose transporters, and their mutation result in a de-repression of the alcohol oxidase promoter in glucose. As for hexose sensors, H. polymorpha Hxs1 [13] mutation and P. pastoris Gss1 mutation [14] cause de-repression of alcohol oxidase under glucose culture. Other existing studies focus on transcription factors. As shown in Table 1, a few transcription activators and repressors have been identified in methylotrophic yeasts. However, how the induction or repression signals are transduced from carbon molecules to these transcription factors is still largely unknown. Exploring these elements will be helpful to reveal more potential targets for constructing the methanol-free expression system.
Kinases always play an important role in cell signaling, since phosphorylation and de-phosphorylation processes are crucial for many biological activities. However, few kinases involved in P AOX1 activation/ repression have been identified so far. Therefore we performed a kinase screening and identified two kinases named GUT1 and DAK. By analyzing the phenotypes of the knocked out strains under different carbon sources, we constructed two strains whose AOX1 promoter could be activated by glycerol or dihydroxyacetone (DHA) as sole carbon source. Then we discussed and tested the possibility for each of these strains to become a novel methanol-free system by expressing several recombinant proteins as examples.

The Δgut1 and Δdak strains have abnormal Aox activity or growth rates under different carbon sources
In the genome of P. pastoris, 152 genes were annotated as kinase coding genes [15]. In order to screen for kinases involved in the P AOX1 activation/repression pathways, we knocked out 92 kinase genes separately and examined strain phenotypes under different carbon sources. For each knockout strain, a colorimetrical assay was used to measure the alcohol oxidase activity while cell growth was checked by the spotting assay. Among these 92 kinase genes, two members attracted our attention: PAS_chr4_0783 and PAS_chr3_0841. PAS_chr4_0783 encodes a glycerol kinase which converts glycerol to glycerol-3-phosphate, and PAS_chr3_0841 is the gene of dihydroxyacetone kinase which converts DHA to dihydroxyacetone phosphate (DHAP) (Fig. 1a). Therefore PAS_chr4_0783 is named PpGUT1 or GUT1 here, and PAS_chr3_0841 is named PpDAK or DAK. As shown by Fig. 1b, Aox in the wild-type strain GS115 was induced by methanol and strictly repressed by glucose and glycerol. However, Aox expression was de-repressed in glycerol cultured Δgut1 strain although cell growth was also largely restricted. The Δdak strain did not have any abnormality in Aox activity, but its growth under methanol was severely impaired (Fig. 1b).
The Δgut1-HpGCY1 strain has the potential to be developed to a novel methanol-free expression system Usually glycerol could be metabolized through two pathways in yeast, including the glycerol kinase mediated phosphorylation pathway and the glycerol dehydrogenase mediated oxidation pathway (Fig. 1a). As the first Mut3 Activator [32] Mpp1 Activator [33] P. pastoris Nrg1 Repressor [20] Mit1 Activator [34] Prm1 Activator [34] Mxr1 Activator [35] C. boidinii Mig1 Repressor [36] Trm1 Activator [37] Trm2 Activator [38] step of phosphorylation or oxidation pathway, glycerol is converted to glycerol 3-phosphate or DHA, respectively. Then both of them are converted to DHAP by different enzymes. Different methylotrophic yeasts prefer different pathways. For example, C. boidinii NO. 2201 utilizes the phosphorylation pathway and Hansenula ofunaensis prefers the oxidation pathway, while H. polymorpha has both [16]. As for P. pastoris, since Δgut1 showed significant Aox activity on glycerol but impaired cell growth (Fig. 1b), we considered that P. pastoris might preferentially use the phosphorylation pathway for the initial step of glycerol usage, and metabolites in this pathway may be repressing signals against Aox expression. These repressing signals are likely localized upstream of DHAP, since DHA and DHAP are common metabolites of both methanol and glycerol utilization pathways. Besides, DHA was an inducible carbon source and it supported Aox expression (Fig. 2a). Therefore, introducing the glycerol oxidation pathway into Δgut1 may be a good way to construct a methanol-free expression system. By converting glycerol directly to an inducible carbon source DHA, repressing signals generated by the phosphorylation pathway could be circumvented.
To test this idea, we introduced the glycerol dehydrogenases gene (GCY1) of S. cerevisiae and H. polymorpha into the Δgut1 strain separately, constructing the Δgut1-ScGCY1 and Δgut1-HpGCY1 strains. As shown by Fig. 2b, HpGCY1 was able to rescue cell growth on glycerol while ScGCY1 could not. Besides, the growth rate of the Δgut1-HpGCY1 strain increased with the elevated initial glycerol concentration, with a heavier final cell biomass at the stationary phase than WT. There results suggested that HpGCY1 works in P. pastoris to metabolize glycerol through the oxidation pathway while ScGCY1 does not. The reason why ScGCY1 did not work in P. pastoris may be due to the difference in conserved domains (Additional file 1: Figure S1). Usually, ScGCY1 works better in extreme conditions, such as high osmolality and micro-aerobic conditions [17,18] while HpGCY1 works in both common and high osmolality conditions [19].
Then we checked the Aox activity in glycerol cultured Δgut1-HpGCY1 strain (Fig. 2c). Aox activity could be detected in different glycerol concentrations from 0.5 to 6 %, suggesting P AOX1 was at least partially de-repressed in the Δgut1-HpGCY1 strain. In addition, lower glycerol concentrations (0.5 and 1 %) supported longer Aox activity. Aox expression in the Δgut1-HpGCY1 strain was still repressed by glucose (Fig. 2c, bottom). Therefore, this modified expression platform preserved the favorable regulatable nature of P AOX1 .
The catabolism of methanol depends on both methanol utilization pathway (MUT pathway) and peroxisomes biogenesis [3]. Usually, the de-repression of Aox expression is accompanied by elevated activities of enzymes involved in MUT pathway and peroxisomes biogenesis [3,20,21]. These genes include AOX1, CAT, DAS1, DAS2, FDH and FLD in MUT, and PEX3, PEX5, PEX10, PEX14, PMP20 and PMP47 in peroxisome biogenesis. Therefore we checked the transcriptional levels of these genes in glycerol cultured WT, Δgut1, and Δgut1-HpGCY1 strains (Fig. 2d). Compared with the glycerol cultured WT strain, the transcriptional levels of these genes in Δgut1 and Δgut1-HpGCY1 strains were much higher, especially AOX1 and FDH. Significant amount of Aox protein could be detected in glycerol cultured Δgut1 and Δgut1-HpGCY1 strains (Fig. 2e).
Taken together, these results indicate that P AOX1 derepression under glycerol is at least partially achieved in the Δgut1-HpGCY1 strain, and it has the potential to be developed to a novel methanol-free expression system. Here we named this system Δgut1-HpGCY1-glycerol. The Δdak strain has the potential to be developed to a novel methanol-free expression system as well Another interesting target revealed from our kinase screening is DAK. The Δdak strain showed similar Aox activity profile as WT, but impaired cell growth under methanol (Fig. 1b). In 1998, Luers et al. deleted this gene in P. pastoris PPY4, and the strain growth was abolished by methanol but supported by DHA [22]. Since DHA is an inducible carbon source for P AOX1 in WT cell (Fig. 2a), we checked the effect of DHA on P. pastoris Δdak strain. In agreement with previous studies, the Δdak strain recovered growth on DHA and its growth rate increased with elevated DHA concentration (Fig. 3a).
The Aox activity in the Δdak strain was then measured by the colorimetrical assay (Fig. 3b). The deep red color suggested that DHA was able to induce AOX1 expression, and lower DHA concentrations (0.2 and 0.5 %) showed better Aox activities than higher DHA concentration, especially at the later growth stage. Aox induction here could also be switched on and off easily by changing carbon sources since it was strictly repressed by glucose (Fig. 3b, bottom). Therefore, this modified expression platform also preserved the inducible nature of P AOX1 .
Again we examined the transcriptional levels of genes involved in MUT pathway and peroxisome biogenesis in DHA cultured WT and Δdak strains. The transcriptional levels of these genes in DHA cultured Δdak strain were much higher than that in DHA cultured WT strain (Fig. 3c). Being consistent with the colorimetrical assay, significant amount of Aox protein could be detected in DHA cultured Δdak strain, which were comparable with that in methanol culture WT strain (Fig. 3d).
Taken together, these results suggested that the Δdak strain also has the potential to be developed to a novel methanol-free expression system in which DHA functions as an inducible carbon source instead of methanol. This system is named Δdak-DHA system.

Compare the two potential novel methanol-free expression systems using GFP as a reporter
In order to test the abilities of the two potential systems in recombinant protein production, we expressed GFP under P AOX1 in Δgut1-HpGCY1-glycerol system and Δdak-DHA system. Green fluorescence intensity was measured to represent the GFP expression level, and GFP intensity in methanol cultured WT strain was used as a reference here. As shown by Fig. 4a, the fluorescence intensity of the Δgut1-HpGCY1-glycerol system was only 20-25 % of that from WT strain grown on methanol, while the Δdak-DHA system showed 80-90 % (Fig. 4b).
These distinct phenotypes could be traced to different growth rates between Δgut1-HpGCY1 and Δdak strains. Both glycerol and DHA are three-carbon molecules, however, the final cell density of glycerol cultured Δgut1-HpGCY1 strain was much higher than that of DHA cultured Δdak strain (Figs. 2b, 3a). Therefore, it is possible that Δgut1-HpGCY1 converted more carbon sources into biomass, while Δdak converted them into protein more efficiently. As a summary here, the Δdak-DHA expression system seemed to function better than the Δgut1-HpGCY1-glycerol system, thus the former one was then selected for further study.

Examine the recombinant protein production efficiencies in the Δdak-DHA system
In order to further elucidate the potential of the Δdak-DHA system, we expressed three more heterologous proteins and compared the expression levels with that in methanol induced WT strains. These heterologous proteins were amylase (Amy) from Geobacillus sp. 4j, glucose oxidase (God) from Aspergillus niger and hepatitis B small surface antigen (HBsAg) from human. Among them, Amy and God were secretory proteins while HBsAg was intracellular. Genes of these recombinant proteins were inserted after the AOX1 promoter. In order to eliminate the influence of gene copy number, single copied expression cassette was selected for all strains. The widely used constitutive promoter P GAP was also examined here as another control.
As measured by enzyme activities, the expression levels of three recombinant proteins in the Δdak-DHA system reached 50-60 % of methanol induced WT system, and became comparable (Amy and God) or even higher (HbsAg) than the constitutive P GAP system ( Fig. 5; Table 2). As shown by enzyme activity to biomass (U/ OD 600 ), the Δdak-DHA system worked generally better than the constitutive P GAP expression system, and showed about 50-60 % expression ability of the traditional methanol induced system.

Discussion
In this study we constructed and tested two modified P. pastoris expression systems on the basis of two kinase mutants. In both Δgut1-HpGCY1-glycerol and Δdak-DHA systems, P AOX1 could be induced by non-methanol Fig. 4 Evaluation of P AOX1 strength in Δgut1-HpGCY1-glycerol (a) and Δdak-DHA (b) systems through GFP reporter assay. GFP intensity in methanol cultured WT strain is used as a reference here. Carbon source concentrations are shown in the brackets by the mass/volume percentage. M methanol; G glycerol; DHA dihydroxyacetone carbon sources (glycerol or DHA) and repressed by glucose. Between them, the Δdak-DHA system showed better expression capacity. Exemplified by three typical recombinant proteins, its protein expression ability generally exceeded the constitutive P GAP system, and reached 50-60 % of the traditional methanol induced system.
It should be noted that further optimization could be done to improve the protein expression efficiency in the Δdak-DHA system. Since several transcription activators and repressors for P AOX1 are already identified, a combination strategy by overexpressing activators and knocking down repressors in the Δdak-DHA system is worth trying. Besides, optimization of expression conditions and parameters in the Δdak-DHA system will be likely to further increase the protein expression levels.
As the simplest ketose, DHA is always used as supplements in cosmetics, medicine and food industry [23].
DHA is non-toxic towards human and environment, and was added to FDA's list of approved cosmetic ingredients in the 1970s. This novel methanol-free system will help to broaden the application of P. pastoris mediated recombinant protein expression, especially in producing medical and edible products.
Another interesting question attracted our attention is why the Δdak strain grows on DHA but not methanol, since DHA is an intermediate in methanol metabolism (Fig. 1a). As one of the steps in methanol assimilation pathway, formaldehyde and xylulose 5-phosphate (Xu-5-P) are converted to one molecule of DHA and one molecule of glyceraldehyde 3-phosphate (GAP) (Fig. 1a). If DHA cannot be phosphorylated to DHAP, DHA will accumulate and cannot contribute to the regeneration of the C 1 -acceptor molecule Xu-5-P, which will not be sufficient for a continued function of the Xu-5-P cycle.  Since methanol and its first step product formaldehyde are toxic to cells, delayed dissimilation may cause growth arrest. We found that addition of xylose recovered cell growth in methanol cultured Δdak strain (Additional file 1: Figure S2A) and promoted methanol utilization (Additional file 1: Figure S2B). It is likely that Xu-5-P generated in xylose metabolism served as acceptor molecule for formaldehyde fixation and then made the cell recover the growth in methanol.
In order to examine why the Δdak strain growth was supported by DHA, we examined the activities of several related enzymes (Additional file 1: Table S2). Dak activity was totally abolished in the knockout, suggesting that P. pastoris does not have any additional isozymes. Both WT and DAK mutant showed significant DHA reductase (Dhar) activity (Additional file 1: Table S2, last column). These results indicate that in DHA cultured DAK mutants, DHA may be reduced to glycerol first and then be metabolized through the phosphorylation pathway. In order to test this, we constructed a double mutant Δgut1Δdak in which glycerol phosphorylation were blocked by GUT1 mutation. As expected, this strain failed to grow on both glycerol and DHA (Additional file 1: Figure S2C).

Conclusions
Our results indicate that the Δdak-DHA system is a novel methanol-free P. pastoris system for recombinant protein expression. The AOX1 promoter in this system is induced by non-methanol carbon source DHA and repressed by glucose. The protein expression ability of this novel system generally exceeds the constitutive P GAP system, and reaches 50-60 % of the traditional methanol induced system. Therefore, this modified expression platform has solved limitations caused by methanol usage and preserved the regulatable nature of P AOX1 , making a potential alternative to the traditional system. Future studies are still needed to further increase the protein expression efficiencies in this system.

Quantitative real-time RT-PCR (qPCR) analysis
The WT, Δgut1, Δgut1-HpGCY1 and Δdak cells were pre-grown in YPD to OD 600 of 2-8 and washed three times with sterile water. The washed cell pellets were transferred to YNG and YNDHA media. After cultured at 30 °C for 2.5 h, cell pellets were harvested and subsequently used to mRNA isolation. Genomic DNA was removed and cDNA was synthesized using ReverTra Ace qPCR RT Kit (TOYOBO). qPCR was carried out as described previously [8] using primers (Additional file 1: Table S1) designed by Beacon designer 7.9.

Cell extract preparation and western blot analysis
To prepare cell extracts, 30-50 OD 600 units of cells were harvested by centrifugation at 6000g for 3 min, washed twice with ice-cold 50 mM potassium phosphate buffer (pH 7.0), and then frozen at −20 °C. Cells were thawed and re-suspended in 1 ml lysis buffer [50 mM potassium phosphate buffer (pH 7.0), 1 mM phenylmethylsulfonyl fluoride (PMSF)]. Aliquots of 1 ml were mixed with 1.8 g glass beads (Biospec Products, Bartlesville, OK, USA) in a 2.0 ml screw-cap tube followed by disruption with a bead disrupter (Mini-BeadBeater-8; Biospec Products) for 8 cycles (1 min vibrating and 1 min resting in ice for each cycle). The lysate was centrifuged at 20,000g for 30 min, the pellet was discarded, and the supernatant was utilized for western blotting. The protein concentration was determined with a Bradford protein assay kit (Tiangen, Shanghai, China).
Each lane was loaded 10 μg total proteins for SDS-PAGE and then transferred onto a polyvinylidene difluoride (PVDF) membrane using the electrophoretic transfer method with rabbit anti-Aox antibody (a kind gift from Suresh Subramani, University of California, San Diego, USA) as the primary antibody and peroxidaseconjugated goat anti-rabbit immunoglobulin G as the secondary antibody.

Construction of Δgut1-HpGCY1 and Δgut1-ScGCY1 strains
As Δgut1 has used Sh ble selection marker, we need new marker to construct the GCY1 expression strains. We used primers ScaI-GAP/GAP-BamHI to amplify the GAP promoter from pGAPZA. pPIC3.5K was digested by restriction enzyme ScaI/BamHI to remove the AOX1 promoter and then ligated with the GAP promoter which was digested by the same restriction enzymes. Finally, we got a new plasmid which contains the GAP promoter and the geneticin selection marker. HpGCY1 was amplified from H. polymorpha genome by using primers BamHI-HpGcy1/HpGcy1-NotI. ScGCY1 was amplified from S. cerevisiae genome using primers BamHI-ScGcy1/ScGcy1-NotI. After using restriction enzymes BamHI/NotI to digest the fragment HpGCY1, ScGCY1, and the above plasmid, the two fragments were ligated into the plasmid respectively. Then we obtained two plasmid P GAP -HpGCY1 and P GAP -ScGCY1. The two plasmids were linearized by SalI and transformed into Δgut1 strain by electroporation. The positive transformants Δgut1-HpGCY1 and Δgut1-ScGCY1 were selected with histidine self-synthesis ability.

Construction of WT-GFP, Δgut1-HpGCY1-GFP, and Δdak-GFP strains
The primers 5-PBR-AOXTT/PBR-AOXTT-3 were used to amplify the fragment PBR-AOXTT [including three parts: E. coli origin of replication pBR322, Ampicillin resistance gene, and green fluorescent protein (GFP) expression cassette] from the plasmid pP-GFP. The primers 5-hph/hph-3 were used to amplify the hygromycin B phosphotransferase expression cassette from the plasmid pAG32, which was kindly provided by Prof. Suresh Subramani. These two fragments were ligated by using ClonExpress MultiS One Step Cloning Kit and then transformed into E. coli TOP 10 to screen correct plasmid. After verified by sequencing, correct plasmid was linearized by SacI and transformed by electroporation into GS115, Δgut1-HpGCY1, and Δdak respectively. The single copy strains of GFP expression cassette were screened according to the previously described method [24].

Assays of yeast growth, Aox activities and GFP expression
The strains were pre-grown in YPD media to OD 600 of 2-8. The cells were harvested by centrifugation at 3000g for 5 min, washed three times with sterile water, and resuspended with initial OD 600 of 1.0 in 50 mL YNB media supplemented with various carbon sources. At suitable intervals, OD 600 was measured for growth curve, 1 mL aliquot of culture media was removed, and cells were harvested by centrifugation and then stored at −80 °C for colorimetrical assay of Aox activities or measurement of GFP.
For measuring GFP, frozen cells were thawed, washed twice with sterile water, and transferred into 96-well plates with diluting to about OD 600 = 1. OD 600 and GFP were measured by enzyme-labeled instrument (BioTek) with three biological replicates.

Construction of three heterologous proteins expression strains
The GOD ORF was amplified from plasmid RINA1297-GOD (kindly provided by Juan Zhang, Jiangnan University) with primers SnaBI-GOD/GOD-NotI or KpnI-GOD/GOD-NotI. The fragment was digested by SnaBI/NotI and ligated into vector pPIC9K opened with the same restriction enzymes to yield the expression vector pPIC9K-GOD. With KpnI/NotI, the GOD ORF was ligated into pGAPZαA to yield the expression vector pGAPZαA-GOD. The vector pPIC9K-GOD was linearized with PmeI and transformed by electroporation into WT and Δdak. The vector pGAPZαA-GOD was linearized with BlnI and transformed into WT. In order to measure the recombinant proteins expression ability of the three expression systems, and eliminate the influence of copy number, single copy strain of God expression cassette was screened according to the previously described method [24]. The three single copy God expression strains were called WT (P AOX1 -GOD), Δdak (P AOX1 -GOD) and WT (P GAP -GOD), respectively.
The construction process of Amy expression strains was just similar to strains WT (P AOX1 -GOD), Δdak (P AOX1 -GOD) and WT (P GAP -GOD). The three single copy Amy expression strains were called WT (P AOX1 -AMY), Δdak (P AOX1 -AMY) and WT (P GAP -AMY), respectively.
The HBsAg sequence was synthesized by Suzhou GENEWIZ biotech Co., Ltd., China. Primer pairs BamHI-HBsAg/HBsAg-NotI were used to amplify the sequence and the 680 bp product was digested with BamHI/NotI. It was then ligated into pPIC3.5K opened with the same restriction enzymes to yield the expression vector pPIC3.5K-HBsAg. With BspT104I/NotI, the HBsAg was ligated into pGAPZαA to yield the expression vector pGAPZ-HBsAg. By using BspT104I/NotI, the α-Factor secretion signal in pGAPZαA can be removed. The vector pPIC3.5K-HBsAg was linearized with SalI and transformed by electroporation into WT and Δdak. The vector pGAPZ-HBsAg was linearized with BspHI and Zeocin was used to select the positive transformant. The three single copy HBsAg expression strains were called WT (P AOX1 -HBsAg), Δdak (P AOX1 -HBsAg) and WT (P GAP -HBsAg), respectively.

Production and activity assays of three recombinant proteins
The strains WT (P AOX1 -GOD), Δdak (P AOX1 -GOD) and WT (P GAP -GOD) were pre-grown in YPD media at 30 °C, 200 rpm. When OD 600 reached 2-8, cells were harvested by centrifugation at 3000g for 5 min, washed three times with sterile water, and resuspended with initial OD 600 1.0 in 50 mL BMMY, BMDHAY and BMDY, respectively. The initial concentration of carbon source in three media was 0.5 % methanol, 0.2 % DHA, and 2.5 % glucose. Every 24 h after the shift, 1 mL aliquot of culture media was removed, and cells were separated by centrifugation (2 min at 8000g). PMSF was added to the culture supernatants to the final concentration of 1 mM to inactivate proteases, and samples were stored frozen at −20 °C for subsequent analysis. Cells biomass was also monitored during the course of cultivation. Methanol and DHA were fed every 24 h to keep the concentration around 0.5 and 0.2 %, respectively. The strains WT (P AOX1 -GOD) and Δdak (P AOX1 -GOD) were induced for 120 h. The strain WT (P GAP -GOD) was batch culture for 120 h. God activity was measured using the coupled o-dianisidineperoxidase reaction. It was determined as in Bankar's manuscript [26].
The culture condition of Amy expression strains followed the same procedure of God. Amy activity was measured using the DNS method [27].
The culture condition of HBsAg expression strains followed the same procedure of God. The difference is that cells were harvested instead of the culture supernatants. HBsAg relative concentration was measured by HBsAg ELISA kit (Kehua, Shanghai, China).

Enzyme activities assays for Dak, Gcy1 and Dhar
Assays were performed as described [28,29]. Enzyme activities were examined by either NADH production or consumption. NADH level change was measured by UV absorbance at 340 nm.