Membrane proteins are drug targets for a wide range of diseases. Having access to appropriate samples for further research underpins the pharmaceutical industry's strategy for developing new drugs. This is typically achieved by synthesizing a protein of interest in host cells that can be cultured on a large scale, allowing the isolation of the pure protein in quantities much higher than those found in the protein's native source. Yeast is a popular host as it is a eukaryote with similar synthetic machinery to that of the native human source cells of many proteins of interest, while also being quick, easy and cheap to grow and process. Even in these cells, the production of human membrane proteins can be plagued by low functional yields; we wish to understand why. We have identified molecular mechanisms and culture parameters underpinning high yields and have consolidated our findings to engineer improved yeast host strains. By relieving the bottlenecks to recombinant membrane protein production in yeast, we aim to contribute to the drug discovery pipeline, while providing insight into translational processes.
- bioprocess control
- recombinant membrane protein production
- ribosome biogenesis
Membrane proteins as targets for recombinant protein production
Membrane proteins are central to many cellular processes: they are involved in the uptake and export of diverse charged and uncharged molecules, as well as mediating the interaction of cells with their environment. As a consequence, they are of prime importance as drug targets to pharmaceutical companies. In order to structurally or functionally characterize a given member of this important class of protein, a sufficiently stable active sample is required. In practice, this means the requirement for a regular supply of milligram quantities of purified membrane protein .
Few membrane proteins have evolved to be naturally abundant in their native membranes with notable exceptions including mammalian and bacterial rhodopsins, aquaporins and complexes involved in respiration and photosynthesis. Inevitably, these proteins were among the first to have their structures solved: the first high-resolution crystallographic structure of a membrane protein [that of the photosynthetic reaction centre from Blastochloris viridis (formerly Rhodopseudomonas viridis) isolated from natural sources] was published in 1995 . The fact that it remained the sole unique structure of a membrane protein for the next decade is indicative of the inherent challenges of working with this class of protein. The first such challenge is that recombinant production is required in order to secure the hundreds of milligram quantities necessary to complete a successful structural biology project .
In 1998, the first structures derived from recombinant membrane proteins were deposited. These were of the prokaryotic proteins MscL  and KcsA , both produced in Escherichia coli. Thereafter, the elucidation of unique structures derived from recombinant sources saw an exponential-like growth rate, with 30 deposited in 2010 . In contrast, the number of unique structures from native sources reached a plateau of around ten per year in 2002 . Of particular note are eukaryotic membrane protein structures, which represent the most challenging targets: they account for fewer than 20% of all the unique membrane protein structures deposited to date. The yeasts Pichia pastoris and Saccharomyces cerevisiae have made an important contribution to this count as they have been used to generate more than half of the recombinant eukaryotic membrane protein samples leading to structures, with the remainder coming from insect or mammalian cells . Specific examples derived from recombinant yeasts include the aquaporins SoPiP2;1  and AQP5 , the potassium channel Kv1.2  and the calcium ATPase SERCA1a (sarcoplasmic/endoplasmic reticulum Ca2+-ATPase 1a) .
Strategies for producing recombinant membrane proteins
The process of identifying a membrane protein target of interest to solving its structure at high resolution is complex, involving several potential bottlenecks . The development of validated protocols and technologies for recombinant membrane protein production that are less reliant on traditional trial and error approaches is therefore key to progress in this area.
Unfortunately, producing recombinant membrane proteins can still seem like something of an artform. There is no guarantee of success when using a previously successful experimental set-up, even for targets that are highly homologous with those that worked well in earlier trials . Suitable production hosts for this purpose include prokaryotic microbes such as bacteria, eukaryotic microbes such as yeast, higher eukaryotes such as insect and mammalian cells, and in vitro systems. All have their unique advantages and disadvantages . We and others take the view that yeast provides a good compromise, benefitting from the speed, ease of use and low cost of goods of a microbe while having the benefits of a eukaryote. As we have already seen, this position is strengthened substantially by its increasingly important contribution to the structural biology of membrane proteins .
Our own work in this area was prompted by an inability to produce Fps1, a member of the aquaporin family, despite the fact that we could routinely produce orthologues such as GlpF . We and others have therefore sought to understand the scientific principles that underpin how recombinant host cells can be used optimally for making membrane proteins. Some groups have chosen to tackle this problem by producing between 40 and 100 orthologues and then focusing on the protein that gives the best yields . This approach has led to structures, but it is labour-intensive and by its nature does not necessarily yield the specific protein of interest. Our focus has therefore been to understand yeast cells in order to make them more amenable to producing the target of choice . The strategy has been to increase the total volumetric yield of a production experiment, in particular by increasing the yield per cell or ‘specific productivity’.
Improving volumetric yields in S. cerevisiae by metabolic engineering
One of the key features of the respiratory yeast species P. pastoris is that it can be grown to very high cell densities. Attenuances (D595; widely referred to as ‘OD’ or ‘optical densities’) in the hundreds are achievable, which is a straightforward way of increasing volumetric yield. Conferring this property on respiro-fermentative S. cerevisiae, which sacrifices its biomass yield when producing ethanol, would be desirable as its genetics and molecular biology are much better established than those of P. pastoris. This strategy was therefore pursued by constructing a fully respiratory strain of S. cerevisiae: Otterstedt and colleagues replaced the endogenous hexose transporters in yeast with a single chimaeric transporter comprising the first six transmembrane domains of the low-affinity Hxt1 transporter and the last six transmembrane domains of the high-affinity Hxt7 transporter . The resultant TM6* strain was found not to exhibit the respiro-fermentative behaviour of a wild-type yeast, but rather a monophasic metabolism consistent with being respiratory. This is most likely to be due to a restricted glucose consumption and hence a reduced glycolytic rate . However, the important feature in this context is that TM6* was able to produce more protein than a wild-type strain by achieving a higher volumetric yield, albeit in the knowledge that its specific growth rate is approximately 67% that of the wild-type parent .
In the TM6* strain, we found that the yield per cell of Fps1 was the same as that for two wild-type strains . The increase in volumetric yield occurred, as anticipated, because the TM6* strain produces approximately 2–3-fold more biomass (determined by measurements of D595) in culture than a wild-type strain. When it was used to produce two GPCRs (G-protein-coupled receptors), the A2aR (A2a adenosine receptor) and the cannabinoid receptor 2, the Bmax and Kd values obtained were comparable with those from a wild-type strain. The increased volumetric yield was due to the fact that there was a doubling of D595 and, more relevantly, the total membrane protein yield in the TM6* strain (determined by protein assay) was 4-fold higher than in the wild-type strain .
Improving yields per cell of S. cerevisiae by strain engineering
Improving the biomass yield of a S. cerevisiae strain that also has an increased yield per cell could provide a route to maximal yields. We therefore looked at the global host cell response to producing a membrane protein . Using arrays, we examined two different growth conditions that both led to relatively low protein yields (35°C, pH 5, and 35°C, pH7) compared with normal conditions (30°C, pH 5) and looked for changes in mRNAs that occurred in the same direction in both sets. This highlighted 39 genes which were validated against the list obtained when comparing the single growth condition that led to higher protein yields (20°C, pH 5) with normal conditions (30°C, pH 5). In all but one case, genes that were down-regulated under low-yielding conditions were up-regulated under high-yielding conditions and vice versa .
In order to determine whether these genes influenced the yield per cell, we first looked at all non-essential genes that were down-regulated under high-yielding conditions in a screen of deletion strains. This revealed, perhaps surprisingly, that, in some cases, the deletion of a single gene could increase the yield of our test protein compared with the wild-type strain, as seen for SPT3, SRB5 and GCN5 (Figure 1). The three deleted genes are known to be components of the transcriptional SAGA (for GCN5 and SPT3) and Mediator (for SRB5) complexes. SAGA may have a role in the transcription of stress-induced genes in order to balance inducible stress responses with the steady output of housekeeping genes . The Mediator complex appears to be required for all transcriptional events and transmits regulatory signals from transcription factors to RNA polymerase II . We found that up to 9-fold improved yields of Fps1 over the corresponding wild-type control were not explained by changes in promoter activity or FPS1 transcript number. This suggested that the improvements that we observed were not due to changes at the transcriptional level, but that post-transcriptional events might regulate the production of recombinant Fps1 .
Next, we looked at the overexpression of genes that were up-regulated under high-yielding conditions by using the doxycycline-repressible tetO system. This revealed an additional three candidate genes that might have a role in increasing the yield per cell of our membrane protein (Figure 2). Most notable among them was BMS1, which is involved in ribosome biogenesis. Using qRT-PCR (quantitative real-time PCR), we noted a clear relationship between Fps1 yield and an increase in BMS1 transcript number compared with wild-type in all of our high-yielding host strains. In particular, we were able to tune the expression of BMS1 to maximize yields. When the promoter was fully on (with no doxycycline added) and the expression of BMS1 was fully induced, we saw relatively low expression of our target protein. This was also the case when the promoter was fully repressed (at >10 μg/ml doxycycline). High-yielding strains and conditions were found to be correlated with 0.5–0.7 copies of BMS1 mRNA per cell, with yield improvements of up to 70-fold. The use of this strain also doubled the yield of the GPCR A2aR after tuning with a different doxycycline concentration of 10 μg/ml .
Under high-yielding conditions, the cells grew more slowly and efficiently (indicated by a lower heat output rate as measured by online flow microcalorimetry) than under low-yielding or control conditions . This is in agreement with the fact that these cells consume glucose and make ethanol more slowly, which may enable them to accommodate the increased metabolic load of producing higher yields of protein. This phenotype is also seen for our three high-yielding deletion strains. These strains also show differences in their translation characteristics and polysome profiles: they have significant accumulation of the polysomes into 80S peaks, strongly hinting at a block in initiation. This indicates changes in their translational properties compared with wild-type cells and links to an overarching role for BMS1. Selective translation of specific proteins is well documented following stress when ribosome functions are usually stalled at the initiation stage and where the Gcn2–Gcn4 complex plays a pivotal role [19–21]. It would therefore be interesting to establish whether such a regulatory complex is involved in the direct regulation of membrane protein production.
Improving yields per cell of P. pastoris and S. cerevisiae by bioprocess control
In addition to genetic engineering, controlling culture parameters in both P. pastoris and S. cerevisiae has the potential to improve the yield per cell. Our initial work in P. pastoris was carried out on secreted GFP (green fluorescent protein), which is commonly used to tag membrane proteins . Since we wanted to find a simple method of identifying and scaling up promising production conditions (thereby minimizing trial and error) we adopted a DoE (design-of-experiments) approach, which is often used in industry, to examine the effect of three input parameters on the product yield. The DoE approach has several advantages since it reduces the number of experiments that need to be performed to examine the combined effect of the input parameters, allows any interactions between them to be interrogated and results in a predictive equation.
We used a Box–Behnken model for our DoE, with three input parameters (temperature, pH and dissolved oxygen) set at three levels. This statistical design requires only a subset of all possible combinations to be used to identify the optimal production condition; in this case, 13 experiments out of a possible 27. Our readouts were the D595 of the culture and its fluorescence. We therefore reported the yield per cell as the fluorescence per D595. Experiments were performed in a parallel mini-bioreactor system that can individually control pH, temperature and dissolved oxygen in each of its 24 vessels. This approach allowed us to develop a model that predicted the optimal yielding conditions for our culture (Figure 3), which we validated by comparing predicted with experimental yields for additional combinations of factors not used to construct the model. We demonstrated further that the model was scalable from 6 ml to 3 litres, although the relationship here suggested that it was under-predicting the yield. This led us to explore the induction regime, which is a critical component of a P. pastoris production experiment .
We noted that induction with 100% methanol did not result in a sustained production, and that, after 15 h, the total yield had typically reached a plateau. We therefore examined a mixed induction of 60% sorbitol and 40% methanol, and saw that production was sustained, and, in the case of our DoE-optimized conditions, the yield was dramatically improved. Gas chromatographic analysis of the residual methanol concentration showed that the best yielding conditions had residual methanol concentrations in the region of 3 g/l. At concentrations below this, yields were lower. This optimized feeding regime increased the yield per cell approximately 6-fold over standard conditions.
We have recently applied the DoE approach to the production of a GPCR in three strains of S. cerevisiae using three culture parameters as before. In addition, the design includes the absence or presence of additives that have been shown to improve the functional yields of some GPCRs. As predicted, the input parameters did not significantly affect the growth of any of the three strains. However, they had a clear effect on binding activity in the membranes, with each strain giving its highest values under different culture conditions. The complete dataset will be used to build a predictive model, which will then be validated at scale.
Yeast is a flexible recombinant host for eukaryotic membrane protein production. The conditions under which yeast cells are grown and harvested are critical in maximizing the total volumetric yield of functional protein. Moreover, differences in yields can be related to the differential expression of genes involved in transcription and translation, and these genes have provided the basis for developing new production strains that can improve yields per cell up to 70-fold. We have used these strategies to produce several membrane proteins in yields suitable for further study. Examples include three four-transmembrane domain proteins (CD81, CD82 and CLDN1), our initial target, Fps1, and the GPCR A2aR.
This work was supported by the European Commission [via contracts LSHG-CT-2006–037793 (OptiCryst) and HEALTH-F4–2007-201924 (EDICT: European Drug Initiative on Channels and Transporters)], Engineering and Physical Sciences Research Council [via a CASE (Co-operative Awards in Science and Engineering) studentship with Applikon Biotechnology] and Biotechnology and Biological Sciences Research Council (via CASE studentships with AstraZeneca Ltd and Glycoform Ltd, as well as a Targeted Priority Studentship in Ageing) to R.M.B.
Recent Advances in Membrane Biochemistry: Biochemical Society Annual Symposium No. 78 held at Robinson College, Cambridge, U.K., 5–7 January 2011. Organized and Edited by J. Malcolm East (Southampton, U.K.) and Frank Michelangeli (Birmingham, U.K.).
Abbreviations: A2aR, A2a adenosine receptor; DoE, design-of-experiments; GPCR, G-protein-coupled receptor
- © The Authors Journal compilation © 2011 Biochemical Society