Systems biotechnology offers new strategies for yeast strain engineering . In this study, a DNA microarray based data mining from experiments that showed an increased productivity of a complex foreign protein in P. pastoris has led to the identification of novel target genes or pathways for the improvement of heterologous protein production.
Engineering of the protein folding and quality control machinery is a very common and successful strategy to increase the secretory capacity of yeasts. For instance, overexpression of the unfolded protein response (UPR) proteins Pdi1p and Hac1p, the ubiquitin Ubi4p, and the chaperones BiP/Kar2p, Jem1p and Cne1p have been extensively shown to improve protein secretion (all reviewed in ).
The protein disulfide isomerase (Pdi) recycling assistant Ero1p, an ER membrane-resident protein and key component of the oxidative folding machinery, has also been reported as helper factor to increase the yields of human serum albumin in K. lactis  and of the human Fab 2F5 antibody fragment in P. pastoris . It is particularly for this reason that we expected to observe a similar effect in this study. At this point, however, it has to be stated that although the model protein in this study was identical to that used by Gasser and co-workers , the experimental design differed in three pivotal points: firstly, the strain was a protease deficient SMD1168 (pep4 mutant); secondly, the expression vector carrying the gene ERO1 was based on histidine selection, 3 times bigger in size and integrated in a different locus (HIS4); and thirdly, the ERO1 gene was derived from S. cerevisiae genomic DNA, while in this study the host specific P. pastoris ERO1 gene was amplified and over expressed. Despite these dissimilarities, the magnitude of the beneficial effect shown by ERO1 clone#9 was overlapping with that reported in Gasser et al. . In relation to the clonal variation observed amongst ERO1 clones (as well as for the rest of co-expressed genes), it has recently been observed that the gene dosage of PDI determines whether its co-expression has a beneficial or detrimental impact on foreign protein secretion . Besides, clonal variation could be also partially due to genetic differences (mutations) generated during the transformation event, as previously suggested .
Unlike ERO1, WSC4 has not been described as a helper factor for increased protein secretion so far. Wsc4p is a component of the ER and plays a role in the translocation of soluble proteins as well as the insertion of proteins targeted to the ER membrane. A contribution of the WSC family to enhanced environmental stress resistance has also been suggested [35, 36]. Interestingly, WSC4 is the closest of the four S. cerevisiae homologues to the TSR1 gene of Yarrowia lipolytica, whose gene product assists in the signal recognition particle (SRP) - dependent translocation of proteins through the ER and also interacts with the UPR-regulator BiP . The potential link to the UPR and its role in protein sorting through the ER membrane may explain the beneficial effect of WSC4 overexpression on Fab secretion. Consistent with these results is the finding that a S. cerevisiae null mutant of WSC4 (the gene is termed YCH8) accumulated soluble protein precursors, indicating defects in protein trafficking . A further link between the WSC family and the secretory pathway includes the activity of Wsc1p in secretory defective cells, where it is required for the repression of genes that make up the protein-synthetic machinery . Although this observation would contradict the herein proposed role of WSC4 on protein secretion (inferred from sequence similarities), it has to be stated that that the functions of the Wsc proteins overlap only partially. In fact, an implication in the signalling pathway that mediates the response to an interruption of the secretory pathway has been reported also for Wsc2p and Wsc3p, but not for Wsc4p . We therefore believe that WSC4 deserves closer attention as a potential target of the secretory system for engineering host cells.
During the identification of potential target genes significant changes were observed in the oxygen-dependent ergosterol and sphingolipid biosynthesis pathways. In addition to the strong transcriptional induction of these pathways in both strains under hypoxic conditions and, in particular, of oxygen consuming reactions therein, ERG25 was also significantly stronger expressed in the Fab producing strain. Considering that this stronger induction resulted not only from oxygen depletion but also from the additional charge of recombinant protein expression, we suggested a potential role of ergosterol or an intermediate during the protein secretion process. In order to investigate the capacity of this important component of the yeast membrane to influence protein secretion, its biosynthesis was perturbed by applying the antifungal agent fluconazole, which specifically inhibits the activity of Erg11p in the late steps of the ergosterol pathway (see Figure 3). Interestingly, the cumulative silencing of its synthesis by applying increasing amounts of fluconazole favoured protein secretion to its maximum at a concentration of 0.6 μg ml-1, while higher concentrations inhibited growth and therefore suggest a complete breakdown of ergosterol synthesis with fatal consequences for the cell. This partial silencing of the ergosterol synthesis by inhibiting Erg11p may resemble the ergosterol deprivation in hypoxic conditions. In fact, the ergosterol content was significantly decreased in fluconazole-treated cells, as shown in Table 2, thereby effectively mimicking the impact of hypoxia on the macromolecular composition of P. pastoris . We therefore suggested that the strong transcriptional induction of the pathway under oxygen scarcity might only reflect compensation for the intermediate substrate deficit provoked by reduced oxygen availability. In which way ergosterol depletion affects protein secretion has to be elucidated, but combining published data with the outcome of this work points to the plasma membrane as key player. Ergosterol and sphingolipids are highly abundant in the yeast plasma membrane and to a lesser extent in other cellular membranes, where they assemble to form so-called lipid rafts . These microdomains are composed of tightly packed sphingolipid acyl chains, attributing them detergent resistance (i.e. to non-ionic surfactants like Triton X-100). An important feature of such lipid rafts is their contribution to protein sorting, since the targeting of membrane protein cargo to the cell surface requires early raft association already in the ER . As a consequence, defective synthesis of sphingolipids and ergosterol has been shown to impair the trafficking and sorting of a raft-associated chimeric protein to the cell surface , and also resulted in missorting of the plasma membrane ATPase Pma1p to the vacuole for degradation [44, 45]. Bagnat and co-workers have shown that sphingolipids and ergosterol were required for incorporation of the cell wall protein Gas1p (a GPI-anchored α-1, 3-glucanosyltransglycosylase) to detergent resistant membranes (DRM), but not for vacuolar or secreted proteins . Consequently, ergosterol depletion may lead to reduced Gas1p in vivo incorporation to the cell wall and, therefore, increased cell wall porosity due to reduced ß-glucan crosslinking; this effect might facilitate the passage of heterologous proteins through the cell wall in a similar way as observed in GAS1 knockout strains . We assume that the defective transport of proteins destined to be incorporated to the plasma membrane might unbalance membrane fluidity by impairing the formation of lipid rafts. The consequences may include a less stringent control of the exchange of macromolecules between the cell and its proximate environment, possibly including secreted proteins whose translocation through the cell is not affected by ergosterol and sphingolipid depletion. It is also likely that membrane proteins that are not incorporated into the DRM, which usually takes place at the ER or Golgi level, are not incorporated into transport vesicles either, thus alleviating such transport compartments from cargo and facilitating the additional uptake and translocation of soluble proteins. These results, although preliminary, support evidence for a complex interaction between cellular membranes and protein secretion, implicating the plasma membrane as hitherto only marginally regarded, but promising target for strain engineering.
In addition, non-ionic surfactants including Tween80, Tween 20 and Triton X-100 were identified as potential enhancers of Fab secretion in P. pastoris with the highest (2 fold) increase in Fab productivity of cells grown in the presence of Tween 80 and a less pronounced increase after addition of Tween 20 or Triton X-100. This is not the first study that reports evidence for a stimulatory effect of Tweens or similar non-ionic surfactants on protein secretion. In bacteria, Tween 80-stimulated glycosyltransferase production correlated with alterations in membrane fluidity [26, 27]. The authors of these studies suggested that an increase in fluidity of the membrane lipids might facilitate the release of intracellular accumulated protein, thereby enhancing its rate of secretion. The beneficial effect of surfactants, including Tween, appeared to be valid also in other organisms, including cellulase secreting Trichoderma reesei , recombinant Schizosaccharomyces pombe  and recently also in recombinant P. pastoris [50, 51]. Apart from the more apparent explanation of a "leakier" plasma membrane that facilitates translocation of soluble proteins, speculations about the mechanisms underlying this effect also included fluctuations in the electrochemical gradient and enhanced membrane fusions of transport vesicles. This latter finding could also explain the observed increase in secreted Fab after the decrease of ergosterol biosynthesis by fluconazole treatment. A destabilization of the membranes by sterol deficiency could favour such membrane fusions and increase the volume of transport vesicles and consequently also the size of the cargo to be delivered to the surface.
In good accordance with these assumptions, in S. cerevisiae cultures grown under comparable conditions and expressing the same model protein, no hypoxic effect - i.e. favoured protein secretion by oxygen depletion - has been observed, probably because ergosterol synthesis did not seem to be affected on the transcriptional level either .
Our findings give therefore strong evidence for cellular membranes or membrane related genes and pathways as promising strain engineering targets.