Efficient Overproduction of Membrane Proteins in Lactococcus lactis Requires the Cell Envelope Stress Sensor/Regulator Couple CesSR

Background Membrane proteins comprise an important class of molecules whose study is largely frustrated by several intrinsic constraints, such as their hydrophobicity and added requirements for correct folding. Additionally, the complexity of the cellular mechanisms that are required to insert membrane proteins functionally in the membrane and to monitor their folding state makes it difficult to foresee the yields at which one can obtain them or to predict which would be the optimal production host for a given protein. Methods and Findings We describe a rational design approach to improve the lactic acid bacterium Lactococcus lactis as a producer of membrane proteins. Our transcriptome data shows that the two-component system CesSR, which senses cell envelope stresses of different origins, is one of the major players when L. lactis is forced to overproduce the endogenous membrane protein BcaP, a branched-chain amino acid permease. Growth of the BcaP-producing L. lactis strain and its capability to produce membrane proteins are severely hampered when the CesSR system itself or particular members of the CesSR regulon are knocked out, notably the genes ftsH, oxaA2, llmg_2163 and rmaB. Overexpressing cesSR reduced the growth defect, thus directly improving the production yield of BcaP. Applying this rationale to eukaryotic proteins, some of which are notoriously more difficult to produce, such as the medically-important presenilin complex, we were able to significantly diminish the growth defect seen in the wild-type strain and improve the production yield of the presenilin variant PS1Δ9-H6 more than 4-fold. Conclusions The results shed light into a key, and perhaps central, membrane protein quality control mechanism in L. lactis. Modulating the expression of CesSR benefited the production yields of membrane proteins from different origins. These findings reinforce L. lactis as a legitimate alternative host for the production of membrane proteins.


Introduction
Membrane proteins comprise up to 30% of the proteome of any organism [1] and, in humans, are the direct targets of 60% of all pharmaceuticals [2]. Although the biological and medical relevance of membrane proteins is quite clear, they have been largely neglected due to a number of technical constraints, such as the production and purification of appropriate quantities of these proteins in their native form. This notion is driving many initiatives and international consortia [3] and has led to the development of novel ways of producing membrane proteins, such as cell-free expression systems [4]. However, the most reliable method to produce membrane proteins still consists of using cells to produce and insert them in the membrane in their native form. The bacterium Escherichia coli is the standard prokaryotic protein production host but, since membrane proteins encompass molecules that greatly differ with respect to structure, sugar decoration, lipid requirements and folding-factors needed, a broad set of hosts may have to be screened to find one best suited for production of a certain protein [5].
In addition to selecting specific hosts, attempts have been made to understand the intricacies of protein overproduction in a given host in order to improve it. Recently, based on a characterization of the E. coli BL21(DE3)-derived C41(DE3) and C43(DE3) strains, which are known for their increased ability to produce membrane proteins [6], Wagner et al. [7] designed E. coli Lemo21(DE3), a strain with a wider applicability due to the tunable activity of the T7 RNA polymerase driving the production of recombinant proteins. The elucidation of the response of E. coli to the production of membrane proteins has been used to design and re-engineer strains with improved membrane protein production capacity [8][9][10].
The fact that the spectrum of membrane proteins that can be produced by Lactococcus lactis is broadening justifies the consideration of this bacterium as a host for the overproduction of membrane proteins [11,12]. Moreover, L. lactis is likely to provide a good membrane environment for proteins from the closely related pathogenic streptococci and enterococci. It is amenable to genetic manipulation and the paradigm for the broad clade of Lactic Acid Bacteria. Work with L. lactis profits of well-developed molecular biology protocols. High-and low-copy number stable plasmids as well as strong and tightly regulated promoter systems are available, allowing expression of toxic gene products [13]. Its relatively low proteolytic activity, single cytoplasmic membrane and an apparent failure to form inclusion bodies could make the production of membrane proteins more robust, easier to target and to purify than is the case for a number of other hosts, such as E. coli. The genomes of several L. lactis strains have been sequenced to completion: L. lactis subsp. lactis Il1403 [14], L. lactis subsp. cremoris SK11 [15], L. lactis subsp. cremoris MG1363 [16], L. lactis subsp. cremoris NZ9000 [17] and the plant-associated L. lactis subsp. lactis KF157 [18], enabling genome-wide studies such as transcriptome and proteome analyses. The small genome of the organism accounts for little redundancy, which facilitates complementation studies. Growth of L. lactis to high cell densities is rapid and does not require aeration, a feature that can save significant energy costs in industrial fermentations. L. lactis does not produce toxic/inflammatory compounds and, being Generally Regarded As Safe (GRAS) and widely used in food industry, should make it easier to market products of e.g., therapeutic interest made with L. lactis.
Despite these advantages, L. lactis is not yet an ideal host for (membrane) protein overproduction. To further improve L. lactis in this respect and to generate enough protein to proceed with e.g., structural studies, attempts should be made to rationally engineer improved protein production strains. Here, we present transcriptome data that reveal the response of L. lactis to the production of the endogenous cytoplasmic membrane protein BcaP, a branchedchain amino acid permease [19]. The involvement of the CesSR two-component system (TCS) and an analysis of its regulon suggest that this TCS monitors the integrity of the cell envelope and activates a response that facilitates the production of membrane proteins. The universality of this response is demonstrated in the accompanying paper, in which the transcriptome and proteome were analyzed in the context of stresses elicited by a range of membrane proteins [20]. This information was successfully used to rationally design L. lactis strains for improved functional production of (recombinant) membrane proteins of different types and origin, including eukaryotic ones.

Results
Functional overproduction of BcaP, BcaP-H6 and BcaP-GFP-H6 in Lactococcus lactis L. lactis BcaP (Llmg_0118) is an approximately 50-kDa integral membrane protein. The protein, previously branded as CtrA, was recently renamed due to a better characterization of its function: it is involved in the internalization of branched-chain amino acids (BCA) in a process driven by the proton motive force [19]. BcaP is predicted to contain 12 transmembrane domains with a cytoplasmic C-terminus [21], allowing the fluorescence read-out of BcaP-GFP-H6 to be used as an indicator of correct folding of the chimeric protein [22,23]. The functionality of the BcaP derivatives used in this study, carrying either a C-terminal hexahistidine (H6) or a coupled GFP-H6 tag, was confirmed by their ability to restore the growth of L. lactis MG1363DbcaP (Fig. S1) and MG1363DbcaPDbrnQ ( Figure 1) in chemically defined SA media [24] with free amino acids as the only nitrogen source. In particular, L. lactis MG1363DbcaPDbrnQ showed a severe growth defect in this medium as it cannot take up the essential BCAs due to the lack of the two dedicated transport systems, BcaP and BrnQ [19]. Complementation, in trans, with a copy of either chimeric version of BcaP restored the growth to wild type rates (Figure 1 and Fig. S1). All three BcaP variants were under the control of the nisin A-inducible P nisA promoter [13] and induction with nisin A in early exponential phase resulted in a complete overlap of the growth curves of the wild type and the complemented mutant strains.
When induced in a standard way, i.e. during the midexponential phase of growth (OD 600 = 0.5), production of BcaP-H6 was easily achieved in L. lactis NZ9000 and only a minor effect on bacterial growth was observed (Figure 2A). Production of BcaP-H6 was quite significant already 15 min after induction and continued to rise steadily throughout. Two hours after induction, the overproduced BcaP-H6 accounted for 21% of the membrane protein fraction ( Figure 2B). The overproduced BcaP-H6 was not clearly discernible using whole-cell-fractions on an SDS-PAA gel, suggesting that, at most, only a residual amount of the protein accumulated in the cytoplasm (data not shown).
Genome-wide transcriptomic response of L. lactis to the overproduction of BcaP-H6 To understand the response of L. lactis to overproduction of BcaP-H6, we examined the transcriptome of L. lactis (pNZ8048::bcaP-H6) one hour after induction of the cells with nisin A, using full-genome DNA microarrays [25]. L. lactis (pNZ8048) was treated similarly to serve as the empty vector control (Figure 2A). Induction for one hour was used as it allowed for considerable amounts of BcaP-H6 to accumulate in the membrane ( Figure 2B). Since growth was not significantly affected, the data is thought to mainly represent the direct response of L. lactis to the production and insertion of BcaP-H6 in the cytoplasmic membrane. Table 1 shows that the CesSR regulon [26] constitutes a clear subset of genes in the group of genes that are significantly up-regulated in the BcaP-H6 overproducing strain. This further validates that CesSR may act as a sensor of the integrity of the cell envelope and, perhaps specifically, the cytoplasmic membrane [26]. Indeed, the CesSR-regulon genes encompass many strategies to react to various sources of cell envelope stress and to minimize its consequences. Notably, CesR induces expression of genes of which the products might counteract protein misfolding and assist in membrane protein biogenesis, such as ftsH (membrane-bound protease), oxaA2 (pre-protein translocation component; YidC homologue) and ppiB (peptidyl-prolyl cis-trans isomerase).
Other significantly differentially expressed genes (Table 2 and 3) can be taken together as the stringent stress response orchestrated via the (p)ppGpp allarmone pathway and is discussed in the accompanying paper [20], which also underpins the transcriptomics data presented here with proteomics results.
The CesSR response is proportional to and follows, in time, the level of production of BcaP-GFP-H6 We characterized the kinetics of activation of CesSR in relation to the production and insertion of BcaP-GFP-H6 in the cytoplasmic membrane. Determining whole-cell-fluorescence enables directly estimating the amount of correctly folded BcaP-GFP-H6 [22,23]. Since fluorescence is dependent on the proper insertion of the entire chimeric protein in the cytoplasmic membrane, its quantification measures the production burden in terms of the amount of recombinant membrane proteins allocated in the membrane. L. lactis NZ9000 (pNZbcaP-GFP-H6; pAB0169) was used in these experiments. Plasmid pAB0169 carries a fusion of lacZ from E. coli to the promoter of llmg_0169, a gene that is activated by CesSR and serves as an indicator of CesSR activity [26]. Expression of the CesSR regulon was immediate; its level of up-regulation was highly correlated to the level of BcaP-GFP-H6 present in the cytoplasmic membrane at any given moment ( Figure 3). Similar data was obtained using a lacZ fusion to the promoter of llmg_2164, another CesSR regulon member (data not shown).
We also observed an increased growth defect, albeit relatively small, when more of the inducer nisin A was used (data not shown). Growth rate differences did not further activate CesSR, since using either 2.5 ng/ml or 25 ng/ml of nisin A yielded equivalent b-galactosidase activity profiles. Also, no increased fluorescence was detectable in those cells. This supports the conclusion that the level of activation of CesSR was exclusively a consequence of the amount of BcaP-GFP-H6 in the membrane and not, unlike many other stress responses, linked to a hampered growth rate.
The CesSR regulon comprises genes that are essential for membrane protein overproduction To examine whether genes induced after overproduction of membrane protein via CesR could constitute a mechanism(s) to cope with the added burden, in-frame and non-polar knock-out mutants of L. lactis NZ9000 were made for 11 genes in the regulon ( Table 4). As all mutants could be made, none of these genes were essential under the conditions used here. The only significant phenotypic difference observed was an approximately 20% decrease of the growth rate of L. lactis DrmaB in glucosesupplemented M17 medium ( Figure 4A).
The plasmid pNZbcaP-GFP-H6 was introduced in all mutants and production of BcaP-GFP-H6 was induced with nisin A. The growth rates of strains DftsH, DoxaA2, DcesSR and Dllmg_2163 were severely affected upon the induction, while the strains Dllmg_0165, Dllmg_0169, Dllmg_1115/6 and DrmaB were partially impaired ( Figure 4A). The remainder of the strains (Dllmg_1103, DspxB and Dllmg_1918) grew like the control strain NZ9000 (pNZbcaP-GFP-H6). Similar results were obtained when BcaP-H6 was overexpressed in the mutant strains, demonstrating that the GFP moiety of the fusion protein was not responsible for these observations ( Figure 4A). Adding nisin A to plasmid-free knock-out strains did not lead to changes in growth (data not shown), thus excluding the influence of the inducer, per se.
Production of BcaP-GFP-H6 in the knock-out strains was measured as whole-cell fluorescence using flow-cytometry. While no significant differences were observed for most mutants, the DftsH, DoxaA2 and DrmaB strains were greatly affected in their capacity to produce BcaP-GFP-H6 ( Figure 4B).
Overexpression of cesSR improves the capability of L. lactis to produce membrane proteins Given that some genes from the CesSR regulon are, apparently, critical in membrane protein production, we examined whether cell fitness could be improved by modulating expression of those genes, namely ftsH, oxaA2, cesSR, rmaB and llmg-_2163. The promoterless genes were cloned in pIL252 and pIL253 (low-and high-copy number vectors, respectively). Constitutive transcription of these genes in both plasmids was achieved by read-through from the plasmid replication genes [27]. The resulting plasmids were introduced in the corresponding L. lactis mutant strains, which already carried plasmid pNZbcaP-GFP-H6.
In trans complementation of the mutations with the matching pIL253 and pIL252 derivatives restored growth ( Figure 5A) and the capability to produce BcaP-GFP-H6 ( Figure 5B) to the wild type level for the complemented DcesSR and DrmaB strains, almost fully for Dllmg_2163, but only partially for the DftsH and DoxaA2 mutants. In all cases the use of the low copy-number pIL252 derivatives corresponded to an intermediate phenotype, showing that the genes were being expressed at different levels as a result of the different plasmid copy numbers.
Subsequently, the various pIL252 and pIL253 derivatives were introduced in L. lactis NZ9000 (pNZbcaP-GFP-H6). Using the standard induction concentration of 5 ng/ml of nisin A at an OD 600 = 0.5 did not yield any differences in the production of BcaP-GFP-H6 between these strains, as measured by flow-cytometry (data not shown). Although differences in the average fluorescence per cell were not observable, L. lactis NZ9000 (pNZbcaP-GFP-H6; pIL253::cesSR) showed a 30% improved growth rate ( Figure 6A).
Several other proteins, especially from eukaryotic origin, are produced to a much lesser extent in L. lactis compared to what we document here for BcaP [11]. We chose two relatively poorly produced proteins -a variant of presenilin missing exon 9 (PS1D9), which is linked to an increased susceptibility to Alzheimer [28], and a sucrose transporter from Solanum tuberosum (StSUT1) [29]. Both proteins were hexa-histidine-tagged at their C-termini. The construction of both nisin A-inducible overexpression constructs is described in the accompanying paper [20]. The presence of pIL253::cesSR in L. lactis strains in which PS1D9-H6 or StSut1-H6 overexpression was induced with nisin A resulted in over 3-fold (PS1D9-H6; Figure 6B) or over 1.5-fold (StSUT1-H6; Figure 6C) increase in the growth rates of the induced strains. The Table 1. Effect of producing BcaP-H6 in L. lactis NZ9000 on the transcription of genes putatively belonging to the CesSR regulon. A full-genome DNA microarray analysis was performed, comparing L. lactis NZ9000 (pNZbcaP-H6) and L. lactis NZ9000 (pNZ8048; empty vector control) (See Experimental Procedures). Cells were grown in GM17 and induced for 1 h with 5 ng/ml of nisin A once the OD 600 was 0.5. The first putative members of operons are underlined. The minus sign on expression ratios indicates down-regulation in the strain producing BcaP-H6. doi:10.1371/journal.pone.0021873.t001 Table 2. Up-regulated expressed genes in L. lactis NZ9000 overproducing BcaP-H6 (for members of the CesSR regulon see Table 1). presence of the lower copy pIL252::cesSR caused an intermediate growth improvement after induction, but still 2-fold for PS1D9-H6 ( Figure 6B). Co-expression of cesSR also directly influenced production of PS1D9-H6 and StSUT1-H6 (Figure 7). Strains carrying pIL253::-cesSR produced four times more PS1D9-H6 or two times more StSUT1-H6 than the control strain with the empty pIL253 vector. Despite not having any perceptible effect on the growth, pIL253::rmaB proved to be equally capable of aiding in the production of these eukaryotic proteins. Co-producing FtsH or OxaA2 also improved membrane protein production, although not to the same extent as CesSR or RmaB did, and only so for PS1D9-H6. We did not detect any significant improvement in PS1D9-H6 or StSUT1-H6 production with any of the pIL252 derivatives (data not shown).

Discussion
Studies on membrane proteins are greatly constrained by a number of technical obstacles, of which production of sufficient quantities in their native form is arguably the major one. To allow designing strains for the dedicated purpose of producing membrane proteins, we characterized the stress that L. lactis cells endure during the production of high amounts of these proteins.
L. lactis can efficiently produce the endogenous branched-chain amino acid permease BcaP and its derived tagged variants, BcaP-H6 and BcaP-GFP-H6. All three proteins accumulate in the cytoplasmic membrane to up to 20% of all membrane proteins in only two hours after induction of their production.
Transcriptome analysis of the response evoked on L. lactis, while producing BcaP-H6, revealed that CesSR is highly activated. This finding gives further credit to the idea that this two-component system (TCS) responds to cell envelope stresses. Previously, CesSR was shown to be triggered when cells are treated with lipid IIinteracting cationic polypeptides, while disruption of cesR led to an increased susceptibility to such membrane-active antimicrobials [26]. It has also been shown that L. lactis CesSR modulates resistance of peptidoglycan to hydrolysis via induction of spxB [30]. CesSR is homologous to Bacillus subtilis and Streptococcus mutans LiaRS [31,32] and Staphylococcus aureus VraSR [33,34], all of which are also involved in the response to cell envelope stresses. In these, and probably in many pathogenic bacteria, this response is linked to resistance to several antibiotics, such as b-lactams and vancomycin.
To the best of our knowledge, there has been no report of CesSR (or a homologous TCS) being activated during production of cytoplasmic proteins, a change in the culture pH, temperature, salt concentration or the redox state of the medium (Anne de Jong, personal communication), thus confirming the specificity of this response to cell envelope stresses.
In the apparent absence of stress response-dedicated sigma factors in L. lactis, some of their roles have most likely been taken over by TCSs. Eight putative TCSs have been identified in L. lactis MG1363 [16], six of which have been proven to be necessary for normal growth or to be involved in resistance to several kinds of aggressions [35]. Further support for the importance of CesSR in L. lactis comes from the relative size of its regulon. It contains more than 30 genes in L. lactis while only two promoters are known to be regulated by LiaRS in B. subtilis [36].
Our transcriptome data supports the previous description of the CesSR regulon with respect to its members as well as their relative induction levels [26], despite imposing a completely different stress on the cells. Our results give little support for llmg-_2420, encoding a putative glycosyl transferase, being regulated by CesRS. The genes ftsH (llmg_0021) and tgt (llmg_0164), both of which contain a putative CesR motif in their promoter region, were up-regulated in our study and are therefore likely members of the regulon. Also, our data indicates that the operon llmg_1115-llmg_1116 might also include llmg_1117 and that the operon llmg_1650-llmg_1646 (comprising cesSR) might also contain llmg_1645, llmg_1644 and llmg_1643.
It should be noted that the vast majority of genes regulated by CesSR are either putative membrane proteins or proteins directly acting on the membrane. Overall, it seems that this stress response provides L. lactis with a mechanism(s) to cope with cell envelope damage. Remarkably, some members of the regulon, such as ftsH, oxaA2 and ppiB, code for proteins that assist in membrane protein biogenesis and quality control, a strategy that might explain the high production yields of BcaP-H6 in L. lactis.
Other differentially-expressed general stress response genes, such as hrcA-grpE-dnaK, groES-groEL, clpE, clpP, clpB, sodA, cspE, are indicative of an increased number of mis-or unfolded proteins in the cell. These could be BcaP-H6 molecules that are not able to insert correctly into the membrane or, due to the high turnover, get misfolded. Additionally, the chaperones and translocation machinery used for BcaP-H6 production and insertion in the membrane might become overloaded at high production rates, leading to an increase of misfolding of other proteins.
A burden on the translocation pathway is reflected in the upregulation of ffh, ftsY, secG, secA, secE and the above mentioned ftsH and oxaA2. Apparently, the cell attempts to upgrade the capacity to insert proteins in the cytoplasmic membrane. The downregulation of secY seems to be in obvious contradiction. Possibly, L. lactis tries to adapt by manipulating the relative influence of specific translocation pathways, such as YidC and/or SecYEG independently, co-translationally or not. Translocation did not appear to be a putative bottleneck when L. lactis was induced to produce CFTR [37], a eukaryotic protein for which only trace amounts could be obtained. The up-regulation of members of the translocation pathway only when substantial amounts of recombinant membrane protein are produced suggests that translocation may represent a bottleneck in these situations.
BrnQ, a branched-chain amino acid transporter, and other peptide transport systems such as OppB2C2A2, OppDFBCA-PepO, MetN and DtpT were highly down-regulated, possibly as a result of the almost 100-fold up-regulation of bcaP-H6 relative to the endogenous bcaP in the control strain. It is likely that the native bcaP gene is also differentially expressed in the overproducing strain, although this could not be assessed as the dedicated DNA microarray probe does not distinguish between bcaP and bcaP-H6. Many other metabolic processes, such as the arginine, purine and pyrimidine biosynthetic pathways, were affected when L. lactis A full-genome DNA microarray analysis was performed, comparing L. lactis NZ9000 (pNZbcaP-H6) and L. lactis NZ9000 (pNZ8048; empty vector control) (See Experimental Procedures). Cells were grown in GM17 and induced for 1 h with 5 ng/ml of nisin A once the OD 600 was 0.5. The first putative members of operons are underlined. doi:10.1371/journal.pone.0021873.t002  overproduced BcaP-H6. Transcription, through rpoE, and translation, through the differential expression of many genes coding for ribosomal proteins, also seem to be affected in the overproducing strain. Some of these observations might have been caused by differences in the growth rate, however small, between the induced strain and its control.
The kinetics and intensity of the CesSR-mediated response to membrane protein overproduction was examined using lacZ fusions to the promoters of llmg_0169 and llmg_2164, while the fluorescent signal from single cells producing BcaP-GFP-H6 was used as a measure for the amount of correctly folded protein. The activation of the CesSR response was fully related to the amount of correctly folded proteins in the cytoplasmic membrane, even when induction started to visibly affect cellular growth. This observation is in accordance with the fact that other stresses, even the burden of overproducing cytoplasmic proteins, do not trigger CesSR (see accompanying paper [20] and Anne de Jong, personal communication).
Of 11 clean deletion mutants for CesSR regulon genes that were considered to be most relevant to the response, both based on their (putative) function and their fold up-regulation in our transcriptome data, none were affected under during normal growth conditions. Most of them, however, displayed growth problems under BcaP-H6 or BcaP-GFP-H6 overproduction stress. In particular, the growth of strains lacking ftsH, cesSR or llmg_2163 stopped completely under these conditions, a clear indication that CesSR activates a mechanism(s) that allows L. lactis to cope with the added burden. Moreover, L. lactis strains DftsH, DoxaA2 and DrmaB were greatly affected in their capacity to produce BcaP-GFP-H6. This observation indicates that all of these genes need to be expressed to some extent already at the moment of induction and in a CesSR-independent manner, since the DcesSR strain did not show the same phenomenon. OxaA2, homologous to YqjG from B. subtilis, proved to be essential only when cells are required to translocate unusually high amounts of protein. An L. lactis OxaA2 orthologue, Llmg_0143 (similar to B. subtilis SpoIIIJ) thus seems sufficient for the default translocation rate, provided that both proteins OxaA2 and Llmg_0143 are exchangeable, as has been shown to be the case in B. subtilis [38]. It remains to be examined whether an L. lactis Dllmg_0143 strain has a phenotype comparable to that of L. lactis DoxaA2. The role of FtsH in the quality control of membrane proteins explains the observed phenotype when L. lactis overproduces BcaP-GFP-H6. During normal growth of L. lactis DftsH, FtsH is not essential as cells probably have redundant mechanisms to correct normal levels of disorganization and misfolding of proteins in the cytoplasmic membrane. Unlike previously theorized for E. coli [8], L. lactis DftsH is handicapped when forced to produce high amounts of membrane proteins.
L. lactis RmaB is a putative regulator of the MarR family [16]. L. lactis DrmaB was the only mutant displaying a clear phenotype during normal growth conditions: it grew 20% slower than the wild type. Upon induction of BcaP-H6 overproduction, L. lactis   DrmaB displayed only a relatively small further increase in its growth defect but proved to be as limited in its capacity to produce BcaP-H6 as L. lactis DcesSR. These observations and the putative implication of RmaB in non-specific multiple antibiotic resistance, being a regulator of the MarR family, suggest that this regulator might contribute to the general fitness of the cell envelope. L. lactis strains DcesSR and Dllmg_2163 showed, per se, no perceptible handicap to overproduce BcaP-GFP-H6 and both genes were necessary only to sustain bacterial growth after induction. This shows that CesSR is required only when a stress response has to be triggered to cope with the added burden. This seems to be accomplished by activating and adjusting quality control mechanisms and translocation capacity. Llmg_2163, probably a transcriptional regulator since it contains a putative PspC DNA binding domain [39], might further enhance this cell envelope stress response.
By cloning ftsH, oxaA2, cesSR, rmaB and llmg_2163 into high-and low-copy number vectors, their influence on membrane protein production could be assessed under six different levels of expression (two different plasmids and an empty vector control in two different backgrounds i.e., the wild-type and deletion strains). As pIL252/pIL253 derivatives have a replicon that is different from that of pNZ8048 (see Table 5), the expression of the CesSR regulon genes could be examined simultaneously with the production of a membrane protein using a two-vector approach. Complementation of the mutant strains with their corresponding genes on pIL252 and pIL253 was obtained, but only partially for ftsH and oxaA2. Possibly, the native expression of these two genes is higher than what we were able to obtain with the plasmid constructs. In all cases, the pIL253 derivatives performed better at restoring the wild type phenotypes.
L. lactis NZ9000 (pIL253::cesSR; pNZbcaP-GFP-H6), overproducing BcaP-GFP-H6 in the presence of extra plasmid copies of cesSR, could be singled out for its improved growth rate. However, no significant improvement was detected for BcaP-GFP-H6 production per cell, probably as a consequence of the fact that the fusion protein, two hours after induction, is already produced at very high levels of over 20% of all proteins in the membrane fraction.
When L. lactis is forced to make eukaryotic proteins, e.g. PS1D9-H6 and StSUT1-H6, which are known to be very difficult to produce [20], pIL252::cesSR and pIL253::cesSR highly increased the growth rate of the producer strains, leading to a substantial increase in protein production. This result demonstrates the importance of the orchestrating role of CesSR, since overexpression of any other particular gene alone did not improve growth. Immunoblotting also showed that PS1D9-H6 and StSUT1-H6 were better produced during co-expression of cesSR. Although it did not improve growth, pIL253::rmaB was equally capable of improving the production of the two eukaryotic proteins. While the increased production capacity of the strain carrying pIL253::cesSR might be partially explained through an up-regulation of rmaB, the improved growth rate of that strain remains unexplained since none of the other vectors produced an equivalent result. It seems likely that cells overproducing CesSR benefit from the sum of all partial contributions of all, or a subset of, the members of the CesSR regulon.

Bacterial strains, plasmids and growth conditions
Strains and plasmids used in this study are listed in Tables 4 and  5, respectively. L. lactis was grown at 30uC in M17 medium (Difco laboratories, Detroit, MI) supplemented with 0.5% (wt/vol) glucose (GM17) as standing cultures or on GM17 agar plates containing 1.5% (wt/vol) agar. The chemically defined medium SA, in which free amino acids serve as a nitrogen source, was prepared and used as previously described [24]. When required, chloramphenincol (Cm), erythromycin (Em) or tetracycline (Tc) (all from Roche Molecular Biochemicals, Mannheim, Germany) were added to media at 5 mg ml 21 . When two antibiotics were used, the concentration of each was reduced to 2.5 mg ml 21 . All growth rates were calculated following OD 600 readings with an Infinite F200 microtiter plate reader (Tecan, Grödig, Austria). Induction with nisin A (Sigma-Aldrich, St. Louis, MO) was done with 5 ng/ml, unless stated otherwise.

DNA isolation and manipulation
Routine molecular cloning techniques were performed essentially as described by Sambrook and Russel [40]. Total chromosomal DNA from L. lactis was extracted according to Johansen and Kibenich [41]. Mini-preparations of plasmid DNA from L. lactis were done using the High Pure Plasmid isolation kit (Roche Molecular Biochemicals). Oligonucleotides were synthesized by Biolegio (Biolegio BV, Nijmegen, the Netherlands). PCR products were amplified using Phusion DNA polymerase (Finnzymes Oy, Keilaranta, Finland) according to manufacturer's instructions. PCR products were purified using the High Pure PCR product purification kit (Roche Molecular Biochemicals). Restriction enzymes and T4 DNA ligase were purchased from Fermentas (Ontario, Canada). Plasmid DNA was introduced into L. lactis by electrotransformation, using a Bio-Rad Gene Pulser (Bio-Rad Laboratories, Richmond, CA), as described before [42].

Construction of clean-deletion mutants and overexpression plasmids
Non-polar and in-frame deletion mutations were introduced into L. lactis NZ900 using, essentially, the method described by Solem et al. [43]. The pCS1966-derived vectors listed in Table 5 were constructed by consecutively cloning the two flanking regions of a given gene, obtained by PCR with the specific primer pairs P1/P2 and P3/P4, presented in Table S1. Since pCS1966 does not replicate in L. lactis, all intermediate cloning steps were performed in E. coli DH5a (grown in TY broth, with 120 mg ml 21 Em when required). Overexpression of ftsH, oxaA2, cesSR, rmaB and llmg_2163 was achieved via read-through from the repA promoter in pIL252 and pIL253 [27]. PCR fragments containing these genes and their native ribosome binding site, were obtained using the Forw/Rev primer pairs indicated in Table S1 and cloned into pIL252 and pIL253, yielding the respective derivatives depicted in Table 5.  Tris-HCl; pH 8.0; 1 mM EDTA) and transferred to 2 mL screwcapped tubes, after which 50 ml 10% SDS, 500 ml phenol/ chloroform, 500 mg glass beads (50-105 mm of diameter) and 175 ml macaloid suspension (Bentone, Hightstown, NJ) were added. The macaloid suspension was prepared as follows: 2 g macaloid were added to 100 ml TE, boiled during 5 min, cooled to room temperature, sonicated during short periods of time until a gel was formed, spun down and resuspended in 50 ml TE (pH 8.0). All reagents used for RNA work were treated with DEPC (diethyl pyrocarbonate; Sigma-Aldrich). Cells were disrupted by beadbeating twice during 45 s (with a cooling step on ice during 1 min in between) in a Mini-Bead Beater (Biospec Products, Bartlesville, OK). The cell lysate was cleared by centrifugation (10 min; 11,000 rcf; 4 uC) and 500 ml supernatant was extracted with 500 ml phenol/chloroform, and subsequently with 500 ml chloroform. Total RNA was isolated from the water phase using the High Pure RNA Isolation Kit (Roche Molecular Biochemicals, Mannheim, Germany) according to the manufacturer's protocol. RNA quality was verified in an Agilent Bioanalyzer 2100 using RNA 6000 LabChips (Agilent Technologies Netherlands BV, Amstelveen, the Netherlands) and RNA concentration was determined spectrophotometrically in a Nanodrop ND1000 (NanoDrop Technologies, Wilmington, DE). Copy DNA (cDNA) was synthesized using Superscript III Reverse Transcriptase (Invitrogen, Carlsbad, CA) and amino allyl-modified dUTP's in the nucleotide mix. Indirect Cy-3/Cy-5 labelling of cDNA was performed according to supplier's instructions (Amersham Biosciences, Piscataway, NJ). Hybridisation of Cy-labelled cDNA during 16 h at 45uC in a microarray hybridisation incubator ISO20 (Grant Boekel, Cambridgeshire, UK) was performed with Ambion Slidehyb #1 hybridisation buffer (Ambion Biosystems, Foster City, CA) on mixed amplicon and oligonucleotide spotted SuperAmine (ArrayIt, Sunnyvale, CA) glass slides. These array slides cover 2308 of the 2435 predicted ORFs from L. lactis subsp. cremoris MG1363. Slides were scanned using a GenePix Autoloader 4200AL scanner (Molecular Devices Corporation, Sunnyvale, CA). DNA microarray data from biological replicates were obtained with dye-swaps, to discard possible differences between the Cy-3 and Cy-5 labelling reactions. Slide images were analysed using ArrayPro 4.5 (Media Cybernetics, Silver Spring, MD). Slide data were processed and normalized using MicroPrep [44,45]. Differential expression tests were performed with the Cyber-T implementation of a variant of Figure 6. Production of eukaryotic membrane proteins in L. lactis NZ9000. Effect on growth at 30uC in GM17 with appropriate antibiotics of the simultaneous production of membrane protein (A. BcaP-H6; B. PS1D9-H6; C. StSUT1-H6) and members of the CesSR regulon, as indicated on the abscissa. The L. lactis NZ9000 strains carried 2 plasmids, one was a pNZ8048 derivative containing the nisin Ainducible gene of the membrane protein in question, the other was either a pIL252 (white bars) or a pIL253 (black bars) derivative from which the CesSR regulon gene was constitutively expressed (see Table 5). The membrane protein genes were induced with 5 ng/ml nisin A when the cultures had reached and OD 600 = 0.5. Control in all three panels: L. lactis NZ9000 (pNZ8048) carrying either pIL252 or pIL253. doi:10.1371/journal.pone.0021873.g006 Figure 7. Production of PS1D9-H6 and StSUT1-H6 in L. lactis NZ9000. Top, L. lactis NZ9000 (pNZPS1D9) strains. Bottom, L. lactis NZ9000 (pNZStSUT1) strains. Each strain carried, in addition to the indicated plasmid, a pIL253 derivative from which the protein specified at the top of the figure was constitutively co-expressed. Cells were grown at 30uC in GM17 and induced for 1 h with 5 ng/ml nisin A at an OD 600 = 0.5. The membrane protein fraction was obtained and separated on an SDS-(10%)PAA gel. The gel was immunoblotted and examined for the presence of PS1D9-H6 and StSUT1-H6 using anti-His antibodies. doi:10.1371/journal.pone.0021873.g007 Student's t-test [46]. Only operons in which at least one gene whose expression ratio was associated with a Bayesian p-value lower than 0.0001 were considered for further analysis.
All data is MIAME compliant and has been deposited in GEO [47].

Protein Techniques and Flow Cytometry
Membrane vesicles were prepared essentially as described by Poolman et al., [48], with the following modifications: 10 OD 660 units of an L. lactis culture was pelleted and resuspended in 1 mL of 50 mM potassium phosphate, pH 7.0; the suspension of cells was bead-beated twice (as described above) after which it was cleared by centrifugation (10 min; 11,000 rcf; 4uC). The supernatant was subsequently centrifuged (45 min; 80,000 rpm; 4uC) using a TLA-120.1 rotor in an Optima TLX Ultracentrifuge (Beckman, Fullerton, CA). The pellet was resuspended in loading buffer and used for SDS-(10%)PAGE analysis. Wet-blotting (16 h; 24 V; 4uC) of SDS-PAGE separated proteins was performed with Table 5. Plasmids used in this study.

Plasmid
Description Reference or Source pNZ8048 Cm r ; Expression vector with nisin A-inducible P nisA [52] pNGbcaP pNZ8048 containing bcaP downstream of P nisA [19] pNGbcaP-H6 pNZ8048 containing the bcaP-H6 gene downstream of P nisA ; for expression of BcaP with 6His-tag at the C-terminus [19] pNZcLIC pNZ8048 derivative that enables cloning by the LIC-VBEx procedure [53] pNZbcaP-GFP-H6 pNZcLIC derivative containing the bcaP-GFP-H6 gene downstream of PnisA; for expression of BcaP with GFP-H6 at the C-terminus [23] pNZPS1D9 pNZcLIC derivative containing the PS1D9-H6 gene downstream of PnisA; for expression of PS1 D9 with 6His-tag at the C-terminus [20] pNZStSUT1 pNZcLIC derivative containing the StSut1-H6 gene downstream of PnisA; for expression of StSUT1 with His-tag at the C-terminus [20] pAB0169 Tet r ; pPTL derivative carrying the promoter of llmg_0169 [26] pAB2164 Tet r ; pPTL derivative carrying the promoter of llmg_2164 [26] pCS1966 Em r ; oroP; 5-fluoroorotate selection/couterselection vector for chromosomal integration in L. lactis [43]  a Mini Trans-Blot (Bio-Rad Laboratories). Immunodetection was performed using Anti-His conjugated antibodies (Qiagen, Hilden, Germany) according to manufacturers' instructions. Visual quantification of protein bands on coomassie-stained SDS-PAA gels was performed with the Quantity One software (Bio-Rad Laboratories). Quantification of BcaP-GFP-H6 production was determined by whole-cell-fluorescence using an Epics XL-MCL flow-cytometer (Coulter, Fullerton, CA) using the WinMDI 2.8 software (http://facs.scripps.edu/software.html). b-galactosidase activity assays were performed on cell suspensions as described previously [49]. Figure S1 Functional production of BcaP and its derivatives in L. lactis MG1363DbcaP. Strains were grown at 30uC in SA medium with 5 mg ml-1 chloramphenicol, when required, and all 20 amino acids. BcaP (full squares), BcaP-H6 (full diamonds) or BcaP-GFP-H6 (full triangles) were induced in either strain with 5 ng/m of nisin A at an OD600 = 0.4. The uncomplemented L. lactis MG1363DbcaP is represented by a dotted line with empty circles and the uncomplemented wild-type L. lactis MG1363 by a dotted line with full circles.

Supporting Information
(TIF) Author Contributions