Bacillus subtilis is a sporulating rod-shaped Gram-positive bacterium (Fig. 1), which thrives in the soil. Like most of its closest relatives B. subtilis is non-pathogenic, and B. subtilis has even been awarded GRAS (Generally Recognized As Safe) status by the US Food and Drug Administration. The first known application of B. subtilis dates back more than a thousand years, when it was already used to produce natto, a Japanese food product consisting of fermented soybeans. Nowadays, B. subtilis is best known as a source of useful enzymes and fine biochemicals, and as an attractive host for the production of heterologous proteins. Many different enzymes, like proteases and amylases, originating from B. subtilis and related Bacillus species are being used in industry for a wide range of different applications 1234456. Importantly, B. subtilis is able to produce and secrete large quantities of proteins into the culture medium. Therefore, this organism is widely regarded as a prolific "cell factory" for industrial enzymes and biopharmaceuticals 17.

B. subtilis is genetically highly amenable as it develops genetic competence for DNA binding and uptake. This is one of the prime reasons why bacilli have been extensively used in both applied and fundamental scientific research for more than 50 years. In 1990 a European-Japanese research collaboration was started, with the aim to sequence the entire genome of B. subtilis strain 168. This has led to the publication of the entire annotated genome sequence in 1997 8. A subsequent international project has led to the identification of all essential genes in B. subtilis 9. Today, B. subtilis is one of the best understood of all living organisms, and it has become the paradigm for research on Gram-positive bacteria. Detailed data sets on the transcriptome 1011, proteome 12, secretome 13 and metabolome 14 of B. subtilis are available, representing a rich source of information for research on Bacillus species. Importantly, the relatively close relationships between B. subtilis and clinically relevant Gram-positive pathogens also make this organism highly relevant for research on potential targets for novel antimicrobials and anti-infectives.

Currently the most commonly used bacterial host for industrial production of heterologous proteins is Escherichia coli. Advantages of its use as a production host for proteins are that it can be grown easily in large fermentations, and that it is genetically amenable and able to produce large quantities of proteins. However, in E. coli the produced proteins usually accumulate within the cells where they have a high potential to aggregate, resulting in the formation of inclusion bodies. To acquire the protein, the inclusion bodies need to be separated from the cell and the proteins subsequently need to be recovered from the inclusion bodies. Moreover, the outer membrane of E. coli, and of Gram-negative bacteria in general, contains lipopolysaccharide (LPS or endotoxin), which is highly pyrogenic and needs to be totally removed before the produced proteins can be used for clinical purposes.

B. subtilis has excellent fermentation capacities that are equal to, if not better, than those of E. coli. In addition, B. subtilis is also capable of producing large quantities of proteins. However, in contrast to E. coli, B. subtilis lacks an outer membrane and is able to secrete proteins directly into the medium. Therefore, the secreted proteins can be purified easily from the medium in their active form, which simplifies the downstream processing considerably. Although most of the proteins that are commercially produced by B. subtilis are secreted into the medium, there are also successful examples of cytoplasmic protein production in B. subtilis 15.

Like all living organisms, B. subtilis has cellular quality control systems. These facilitate the production of high quality proteins, and remove misfolded and incompletely synthesized proteins 1617. Unfortunately, cellular quality control systems also represent significant bottlenecks in heterologous protein production 1819. This poses a fascinating challenge for cell factory engineering since inactivation of quality control systems can improve protein production yields, but these improved yields might be at the expense of product quality. Clearly, a reduced product quality would be an unwanted effect, especially if the product is a biopharmaceutical. Notably, the mechanisms underlying the degradation of secretory proteins in B. subtilis are nowadays fairly well-understood and, in many cases, the yields of "fragile" secretory proteins could be improved significantly by engineering of the cellular machinery for protein quality control 20.

Successful strategies for engineering of B. subtilis to improve protein production include the knockout of extracellular and/or intracellular proteases 212223, overexpression of chaperones and folding catalysts 24252627, overexpression of components of the secretion machinery, and/or modification of the cell wall microenvironment 2829. Besides engineering the host, also the expression system used to produce the protein can be modified in order to improve production and/or secretion, for example by the use of strong or inducible promoters 303132. Another strategy is to modify the protein that is being produced itself, for example by selecting an optimal signal peptide 3334, or by rendering the protein less sensitive for degradation through site-specific mutagenesis 35. The latter protein modification approach has the disadvantage that it can affect the functionality and folding of the protein.

Over all, it has become clear that a wide range of approaches for modification of B. subtilis can be applied to further improve this important cell factory for production of cytoplasmic and secretory proteins. Nevertheless, there are still many proteins that remain recalcitrant to such approaches. These include membrane proteins and proteins that are part of cytoplasmic or membrane-associated protein complexes.

To maintain the cellular homeostasis, the cytoplasmic membranes of bacteria are largely impermeable to ions, the majority of nutrients and signaling molecules. The vital communication between the cytoplasm, transmembrane compartments and the extracellular milieu is facilitated through membrane-embedded proteins. They typically account for about 30% of open reading frames in prokaryotic and eukaryotic genomes 36, and they carry out a diverse range of functions in vital processes such as cellular growth and division, maintaining cell integrity, energy transduction, signal sensing and transduction, cell-cell interactions, and transmembrane transport processes (Table 1). Membrane proteins are without any doubt the most important group of proteins in terms of current drug targets. Despite their functional and biotechnological importance, the study of membrane proteins has remained difficult due to their hydrophobicity. Accordingly, they generally require detergents to remain soluble upon extraction from the membrane. The presence of detergents, however, complicates the biochemical and structural analysis of membrane proteins. Consequently, high-resolution structural data is available for only very few membrane proteins, while X-ray crystal structures are available for daily increasing numbers of soluble proteins. To date, only the most abundant membrane proteins have been characterized in some detail.

Not only the analysis of the properties of individual membrane proteins is difficult, but the same applies even more so to complexes of membrane proteins as well as the entire membrane proteome. Thus, the analysis of membrane proteomes, in general, has so far been relatively unproductive compared to analyses of cytosolic proteomes, cell wall proteomes and exoproteomes. This also applies to the B. subtilis membrane proteome 133738. First studies to investigate the B. subtilis membrane proteome were undertaken by Bunai et al. 39 and by Dreisbach et al. 37. To this end, different methods for membrane protein solubilization were combined with gel-based, semi-gel-based and gel-free proteomics techniques (Fig. 2). More than 700 proteins were identified in the B. subtilis membrane; 122 of these proteins contain predicted N-terminal signal peptide-like sequences that may serve in membrane targeting. From the membrane proteins that were identified by Eymann et al., 268 proteins contain at least one potential membrane spanning domain 37, and 134 contained four or more potential transmembrane domains. Notably, most detected membrane proteins of B. subtilis are still of unknown function and this is in fact also true for a multitude of membrane proteins in other species. The functionally defined proteins are permeases and transporters, dehydrogenases, subunits of respiratory complexes, oxidoreductases, ATP-synthase components, two-component signal transduction proteins, penicillin-binding proteins, signal peptidases and proteins involved in cell motility, cell division, autolysis, chemotaxis, and osmoregulation.

Several of the identified B. subtilis membrane proteins have potential biotechnological applications. For example, cytochrome P450-like proteins can be exploited for the bioconversion of a wide range of substrates 40. Other identified B. subtilis membrane proteins, especially the essential ones, may represent potential new targets for the development of novel antimicrobial or anti-infective drugs.

Due to their roles in key cellular processes, membrane proteins represent interesting subjects for fundamental scientific research on their structure-function relationships. In addition, these proteins are crucially important from a pharmaceutical perspective, because they are drug targets that are relatively easy to address being exposed on the extracytoplasmic side of the membrane. Thus, pharmaceuticals acting on these proteins do not always need to enter the cell. Unfortunately, most membrane proteins are naturally present in relatively low amounts in the cells. This makes it difficult to obtain such proteins in sufficient amounts for functional and structural analyses. Moreover, achieving high-level expression of membrane proteins has turned out to be very difficult. Several high-throughput screens for membrane protein overproduction have been performed, most of them using E. coli as the expression host 414243. In some of these studies a clear correlation was found between the success of expression and the number of predicted transmembrane helices 4445. However, in contradiction with these results, a screen on overexpression of E. coli inner membrane proteins in E. coli indicated that there is no clear correlation between the ability to overproduce a membrane protein and protein size, the number of transmembrane helices or specific sequence characteristics 46. Remarkably, in yet another screen for expression of 49 E. coli membrane proteins it was even found that the majority of successfully expressed proteins had a high number of transmembrane helices 47. Furthermore, it was proposed that not only properties of the protein itself would determine whether a particular membrane protein could be overproduced successfully. Other important parameters were the E. coli strain used for overproduction, the type of detection/purification tag fused to the overproduced protein, and N-terminal or C-terminal fusion of such a tag to the overproduced protein 47. Another problem is that in most cases the produced proteins mainly accumulate in the cytoplasm and aggregate. In fact, there are only very few examples where it was shown that a large amount of the protein was inserted correctly in the membrane 414548. It is thought that the presence of high amounts of proteins in the membrane can affect the integrity of the membrane and thereby have a toxic effect on the cells. However, although cytoplasmic accumulation can prevent this toxicity, the purification of overproduced membrane proteins from the cytoplasm is not preferred for most applications, since the protein may either be folded incorrectly, inactive, or both. Notably, the overproduced membrane proteins can be hard to purify and the proteins can be readily lost during purification or subsequent crystallization for structural analyses. Therefore, there is a clear and generally recognized need for systems to overproduce correctly membrane-inserted membrane proteins in large amounts.

To date, it is unknown to what extent B. subtilis is exploitable for high-level membrane protein production. However, membrane protein biogenesis usually requires the same general secretion (Sec) pathway that is used by bacteria to direct the vast majority of exported proteins to extracytoplasmic cellular locations or the growth medium (reviews: 4950). This seems also to be true for B. subtilis 1338. Since the Sec pathway of B. subtilis has a huge capacity for protein secretion, often to commercially significant gram per liter levels, there is presently no reason to assume that the B. subtilis Sec pathway will be less effective in inserting proteins into the cytoplasmic membrane. Thus, it can be anticipated that B. subtilis will turn out a highly suitable host for membrane protein production at high levels.

For many years, B. subtilis has been a widely appreciated model organism in studies on basic cellular processes, such as cell division, DNA replication, and cell differentiation. Thus, this organism was a logical choice for subsequent functional genomics, transcriptomic and proteomics research on these cellular processes. As the next step in reaching a more global molecular understanding of cellular processes, new proteomics and systems biological methodologies are currently being explored for analyzing post-translational modifications, protein stability/degradation and protein interaction networks 51.

The proteome of any living organism is divided into structured protein interaction networks, all together known as the interactome. Such networks represent functional protein complexes (e.g. chaperones), molecular machines (e.g. the Sec translocase for protein translocation across the membrane), or highly dynamic cellular pathways (e.g. energy transducing systems). High-throughput approaches in yeast and other organisms have revealed that most proteins interact with only few other proteins. In contrast, relatively small numbers of proteins, the so-called "interaction hubs", have multiple interacting partners and thus seem to participate in multiple protein complexes or protein superstructures 525354. Furthermore, the available data indicates that flexible protein networks exist in which protein complexes are composed of core proteins and peripheral proteins that readily assemble and disassemble. Thus, the interaction hubs in protein networks can be divided into "date hubs" that are mainly involved in dynamic interactions, and "party hubs" that are involved in permanent interactions. Interestingly, the party hubs often seem to be connected, which suggests that they represent the cores of highly clustered functional modules 52. It should be noted that interactome studies have so far been predominantly focused on soluble proteins, and only little, if any, data is available for protein networks in membranes.

Interactome studies in B. subtilis have mainly involved small-scale protein networks that are related to DNA replication and chromosome dynamics 5556, cell division 56 and cell morphogenesis 57. These networks were defined by iterative cycles involving yeast two-hybrid screening ("interactome walking") 5558. The data has been deposited in the B. subtilis protein interaction database "SPiD" 5960.

One major challenge for postgenomic research is to produce protein complexes in sufficient amounts for biochemical and structural studies. Several studies have shown the feasibility of purifying endogenous complexes for structure determination, including RNA polymerase II 61 and the ribosome 6162. In addition, technical advances such as Tap-tagging have allowed easier purification of large multiprotein complexes 6163. This latter approach is currently restricted by the low abundance of many complexes within the cell. However, large scale functional characterization and structure determination of macromolecular complexes requires the purification of the different subunits in large quantities and their assembly into a functional entity. One way to obtain protein complexes from individual proteins in vitro involves producing highly purified and soluble proteins at high-concentrations and subsequent formation of protein complexes, which are suitable for further biochemical and crystallographic studies. This technique of in vitro reconstitution from separately purified components can be used to study small or mid-size assemblies. The major drawback of this technique is that it is relatively slow and often requires refolding steps. In many cases, proteins that form complexes in cells remain at least partially unfolded in the absence of their normal cellular partners in a heterologous expression system. Frequently, creation of a protein complex from individual proteins is not a simple task and carries along many technical problems. Firstly, over-expression of only one protein from a complex may be the cause for its insolubility 64. Secondly, posttranslational modifications can not be reproduced during such in vitro experiments and further studies may not be successful 65. In addition, it is often necessary to produce two or more proteins at the same time to obtain proper folding and/or interaction 66. To overcome some of these difficulties new methods for over-expression of two or more proteins in different hosts have been developed. Co-expression can be achieved by using two or more plasmids each of which bears a gene coding a subunit of a protein complex and a different selection marker. Another way is the introduction of several genes into one expression vector 64. In bicistronic vectors, despite the presence of ribosome binding sites for each gene, the expression of the second gene is usually much lower. Insertion of a promoter in front of the second gene may improve the yields of the second product 67. Additionally, construction of a plasmid bearing four genes coding for protein subunits was reported. This method uses LINK sequences and ligation-independent cloning (LIC), which avoids PCR. Thus, the generation of unwanted mutations can be avoided 64.

Efficient production of protein complexes requires suitable purification steps. In comparison with conventional methods like ion-exchange chromatography, size exclusion chromatography or hydrophobic interaction chromatography, affinity tags represent highly efficient tools for complex purification under mild elution condition 68. The use of different fusion tags can help to identify protein complexes. A clear disadvantage of this method is that the presence of a fusion tag may prevent the interaction with another protein of the complex. Furthermore, mass spectrometry is usually used for identification of proteins in the complex 69.

The new expression systems, from the common binary expression to the more complicated multi-expression systems for production of protein complexes, are well suited for structural proteomics high-throughput strategies as used for the SPINE (Structural Proteomics In Europe) and E-meP (European Membrane Protein consortium) projects. Structural proteomics projects are creating large amounts of data that has to be organized and archived. Recently, the Laboratory Information Management System (LIMS) for structural biology and genomics was developed 70. In addition, an integrated LIMS system, such as the Protein Information Management System (PIMS) is currently being developed in Europe 7172. This system can also handle complicated data, such as information on expression of protein complexes.

Most of the methods and techniques for production of protein complexes that have been mentioned above use as the host E. coli,Saccharomyces cerevisiae, baculovirus-infected insect cells or mammalian cells. Further development of efficient production systems for protein complexes seems to require the identification of new expression hosts with better characteristics. Clearly, B. subtilis is one of such candidate hosts with ample possibilities for improving the level and quality of protein complex production.

Membrane protein biogenesis in Gram-positive bacteria like B. subtilis is a largely unstudied field of research. The large majority of our knowledge on this process in Gram-positive bacteria is based on bioinformatic studies and comparisons with other organisms, while only a limited number of experimental studies exist. In general, it is believed that membrane protein insertion in Gram-positive organisms follows similar principles as resolved for the Gram-negative E. coli (Fig. 3).

In B. subtilis, as well as in all living cells, genes involved in a given cell function are activated at the time of execution of that function. Also, the genes encoding proteins that function in complexes are co-expressed, and temporal cascades of gene expression control multiprotein structure biogenesis. These multiprotein structures have a crucial role to direct complex processes during the cell life cycle. Studies on the formation of these protein superstructures require the most advanced technologies of molecular biology. In general, these multiprotein structures are built from more or less stable proteins and sub-complexes and some of them are amenable to purification, typically by affinity methods, and to subunit identification by mass spectrometry. One of the most extensively studied protein structures in B. subtilis is the divisome, a structure that is composed of division proteins and proteins involved in their biogenesis. Therefore, the divisome serves an important model function for studies on B. subtilis as a producer of protein complexes.

Cell division in bacteria is a complex process involving the coordinated participation of a group of proteins which assemble at the division site into a multiprotein complex called the divisome (for reviews see 91929394). This process has been best studied in two bacterial model systems: E. coli and B. subtilis. The earliest apparent event in cell division is the formation of an FtsZ ring (Z-ring) at the future septum site. In B. subtilis, the MinC and MinD proteins form a complex which blocks the formation of the Z-ring at the cell poles, whereas the nucleoid blocks the septation at mid-cell. The topological control of MinCD activity is provided by DivIVA in B. subtilis 9596 and by oscillating MinE in E. coli 979899. DivIVA can form oligomers which serve as building blocks in the formation of higher order assemblies giving rise to two-dimensional lattices in a time-dependent manner (see Fig. 4) 100. DivIVA is stably associated with the cell poles, to which it recruits MinCD, probably by direct interaction with MinD 101102. The initiation of septation is a complex process involving many proteins and their complexes as well as specific cell cycle conditions, such as DNA replication and segregation. The protein complexes respectively involved in these processes are known as the replisome and segresome. The FtsZ protein assembles into a cytokinetic ring on the inner surface of the cytoplasmic membrane at the place where division will occur 103104. The Z-ring structure provides the framework for the recruitment or assembly of about ten membrane and cytoplasmic proteins, uniquely required for cell division. Some of these are required for biogenesis of the new hemispherical poles of the two daughter cells. In E. coli, during cell division the proteins assemble in a defined order as follows: FtsZ, FtsA/ZipA, (FtsE, FtsX), FtsK, FtsQ, (FtsB, FtsL), FtsW, FtsI, FtsN, AmiC and EnvC, where the proteins in parentheses assemble simultaneously (Fig. 5A) 105. The assembly of the Z-ring depends on FtsA and/or ZipA, while the localization of the latter pair of proteins depends on FtsZ 106107. FtsK does not require any downstream proteins to assemble at the Z-ring. FtsL and FtsB localize in a co-dependent fashion 108. The localization of the last protein from this set, EnvC, depends on all of the other proteins. This hierarchical localization of division proteins in E. coli is likely to reflect a sequence of protein-protein interactions that lead to the assembly of the protein complexes of the divisome. B. subtilis has homologues of most of the E. coli division proteins, including FtsZ, FtsA, ZipA (possible functional homologue of B. subtilis EzrA), FtsL, YgbQ (DivIC in B.subtilis), FtsQ (DivIB in B. subtilis), FtsW (YlaO in B. subtilis), and Pbp3 (Pbp2B in B.subtilis) (reviewed in 91). However, SepF is only present in B. subtilis 109. In contrast to the hierarchical localization of division proteins in E. coli, in B. subtilis the equivalent division proteins are recruited in a more concerted manner (Fig. 5B) (reviewed in 91). DivIB, DivIC, FtsL, Pbp2B and probably YlaO are all completely interdependent in their assembly at the division site and depletion of FtsA, DivIC, FtsL or Pbp2B, abolishes the positioning of the other cell division proteins at mid-cell. The first three proteins from this division protein set, DivIC, FtsL and DivIB, seem to form one or more different oligomers 110. The possible role of FtsL is to stabilize DivIC through formation of a DivIC-FtsL complex 111 and DivIB has a role in FtsL turnover 112. The function of YlaO is closely connected to FtsL and likely to include targeting of cognate PBPs (penicillin binding proteins).

The dynamics of division protein complexes have been interrogated by mapping and mutation analysis. Many of these physical interaction studies have been complemented by genetic and phenotypic screens 91. Microtubule and cytoskeletal superstructures have also been subject to proteomic analyses. Genome-wide datasets and ever-more-complex networks built from such data seem to quickly overload human intuitive capacity. Inevitably, these vast amounts of information must be captured and processed in mathematical models, as has been done in the physical sciences for many years. User-friendly interfaces for simulation of biological systems have been created that will certainly be widely used to explore the manifestations of biochemical and genetic networks. The cell cycle has in fact long been subject to modelling efforts, which have become increasingly sophisticated and coupled to experimental tests of model predictions. A good example of intensive modelling is the oscillation of the MinCDE complex in E. coli which has been analysed using simple reaction-diffusion mathematical models 113114115 as well as more advanced mathematical models 116117.

The dynamics of division protein complexes have been interrogated also by mapping and mutation analyses. It is clear that a complete appreciation of the mechanism of divisome assembly in B. subtilis will require a much deeper understanding of the protein-protein interactions that take place between divisome proteins both before localization at the division site and during assembly of the divisome. Many of these physical interaction studies have been complemented by genetic and phenotypic screens 91. Some of the septasomal proteins were expressed, purified and used for biochemical and crystallographic studies (for reviews see 8992). Although a huge amount of data about the divisome exists, development of an efficient B. subtilis production system for protein complexes will be required for the in vitro reconstitution of this and other crucial cell cycle protein complexes. In turn, the knowledge thus generated is likely to lead to important insights and tools for removal of bottlenecks in the production of protein complexes in B. subtilis.

Molecular chaperones are cellular components, which assist folding processes of proteins by interacting with non-native polypeptide chains in a non-covalent manner. This definition excludes classical enzymes involved in catalyzing protein folding with covalent reactions, although in some folding factors these both types of activities can be distinguished; there may be an enzyme domain and a chaperone domain in the same component. Chaperones are typically cytosolic proteins or multiprotein complexes involved in protein folding assistance in various cellular processes. The roles of microbial molecular chaperones in protein folding assistance, aggregation prevention, protein quality control, chaperone-assisted protein degradation and the heat shock response have been extensively studied with cytosolic and secreted protein substrates and proteomes 118, but in the context of membrane protein folding and assembly, information on such functions of chaperones is scarce. Although most of the mechanistic studies of bacterial chaperones have been carried out with E. coli, the available information can also be applied to other bacterial systems including B. subtilis. When proteins are overexpressed in bacteria regardless of whether they are cytosolic, secreted or membrane proteins, a general problem is the formation of insoluble aggregates of predominantly misfolded proteins, so called inclusion bodies. Co-overexpression of chaperones may help in controlling the folding process of overexpressed proteins and thereby decrease aggregation.

B. subtilis has an extensive quality control system for protein production. This system can respond to the presence of misfolded or incompletely synthesized proteins by activating proteases that remove these proteins. The biotechnological advantage of the quality control system is that it enables the production of high quality proteins with few impurities of misfolded side-products. However, it can also represent one of the major bottlenecks for the production of especially heterologous proteins. Besides their role in protein quality control, proteases are also involved in the removal of cleaved signal peptides and in the processing of precursor proteins in order to acquire the active mature forms of these proteins. Furthermore, proteases are highly important for many regulatory processes within the cell.