Secretion and assembly of functional mini-cellulosomes from synthetic chromosomal operons in Clostridium acetobutylicum ATCC 824

Background Consolidated bioprocessing (CBP) is reliant on the simultaneous enzyme production, saccharification of biomass, and fermentation of released sugars into valuable products such as butanol. Clostridial species that produce butanol are, however, unable to grow on crystalline cellulose. In contrast, those saccharolytic species that produce predominantly ethanol, such as Clostridium thermocellum and Clostridium cellulolyticum, degrade crystalline cellulose with high efficiency due to their possession of a multienzyme complex termed the cellulosome. This has led to studies directed at endowing butanol-producing species with the genetic potential to produce a cellulosome, albeit by localising the necessary transgenes to unstable autonomous plasmids. Here we have explored the potential of our previously described Allele-Coupled Exchange (ACE) technology for creating strains of the butanol producing species Clostridium acetobutylicum in which the genes encoding the various cellulosome components are stably integrated into the genome. Results We used BioBrick2 (BB2) standardised parts to assemble a range of synthetic genes encoding C. thermocellum cellulosomal scaffoldin proteins (CipA variants) and glycoside hydrolases (GHs, Cel8A, Cel9B, Cel48S and Cel9K) as well as synthetic cellulosomal operons that direct the synthesis of Cel8A, Cel9B and a truncated form of CipA. All synthetic genes and operons were integrated into the C. acetobutylicum genome using the recently developed ACE technology. Heterologous protein expression levels and mini-cellulosome self-assembly were assayed by western blot and native PAGE analysis. Conclusions We demonstrate the successful expression, secretion and self-assembly of cellulosomal subunits by the recombinant C. acetobutylicum strains, providing a platform for the construction of novel cellulosomes.


Background
The development of cost-effective strategies to synthesise biofuels from renewable biomass has received renewed interest from both academic and industrial institutions. Consolidated bioprocessing (CBP) is one such strategy, potentially allowing effective conversion of cheap lignocellulosic materials into high value industrial products [1]. This process requires simultaneous enzyme production, saccharification and fermentation of biomass into valuable products, without added enzymes. Current efforts to generate an organism suitable for CBP fall into two main categories, the 'native' method and the 'recombinant' method [2]; the former involves the engineering of existing cellulolytic organisms to display improved cellulose degradation and solvent production, whereas the latter involves the modification of an existing solvent producer to grow on lignocellulose. To date, most studies have focused on the production of ethanol (for an overview of developments in various organisms, see recent reviews by Olson et al. [2] and la Grange et al. [3]). Recent examples of engineering for ethanol production via CBP via the native method include the production of 38.1 g/l ethanol from 92.1 g/l Avicel via a co-culture of modified Clostridium thermocellum and Thermoanaerobacterium saccharolyticum [4]; the 'recombinant' approach has recently generated strains of Saccharomyces cerevisiae [5] and Kluyveromyces marxianus [6] able to ferment hot water treated rice straw and Avicel respectively, although final ethanol titres were still low.
Biobutanol has been recognised as an alternative transportation fuel to fossil fuels; its properties of a high calorific value, high hydrophobicity, low freezing point, and low heat of vaporisation make it a more attractive fuel than ethanol [7]. The best known solvent-producing bacterium is the mesophile Clostridium acetobutylicum. It efficiently converts all major monosaccharides generated from plant cell wall depolymerisation [8], and some polysaccharides like starch, into acids and solvents during the ABE (Acetone-Butanol-Ethanol) fermentation process [9,10]. In addition, C. acetobutylicum secretes xylan-degrading enzymes [11], although it is unable to utilise this polysaccharide efficiently [12]. Although some strains of C. acetobutylicum can grow on cellobiose and amorphous cellulose [11,12], they are unable to grow on crystalline cellulose [13], as the small amount of cellulosome produced appears to be non-functional in this organism. The introduction of a functional cellulosome is one potential approach to engineering C. acetobutylicum to be a suitable CBP organism.
Functional cellulosomes are large, multienzyme complexes designed for efficient deconstruction of cellulose and hemicellulose, the two most abundant polymers on Earth [14,15]. The core of the complex is a modular, noncatalytic scaffoldin protein which binds the cellulosome to its substrate via a non-catalytic carbohydrate binding module (CBM) and integrates the catalytic enzymatic subunits into the complex via cohesin-dockerin interactions (see recent review by Fontes and Gilbert [15]). One of the best characterised and most efficient cellulosomes is that of C. thermocellum. Moreover, C. thermocellum presents one of the highest rates of cellulose utilisation known, and its cellulosome displays good activity against crystalline cellulose [15], even at 37°C [16].
The C. acetobutylicum strain ATCC 824 has been previously used for heterologous production and secretion of individual cellulosomal components and mini-cellulosomes from C. thermocellum and Clostridium cellulolyticum [17][18][19][20]. In these studies, all cellulosomal subunits, cellulases and scaffoldin proteins were expressed from multicopy plasmids, with their expression driven by either a weakened promoter or a promoter containing the lac operator, as cloning of the cellulolytic enzymes downstream of a strong promoter proved to be toxic to Escherichia coli cells [17][18][19][20]. In all of these systems, the levels of the secreted proteins, especially the larger family 9 and family 48 cellulases, were very low and the recombinant strains were unable to grow on crystalline cellulose. In the most recent study by Chanal et al. [20] the authors found that various scaffoldin modules, such as the family 3a carbohydrate binding module combined with the X2 modules from C. cellulolyticum or C. acetobutylicum, can trigger the secretion of larger cellulases when introduced to the cellulase as a carrier domain, but no further evidence for cellulose solubilisation was provided.
The expression of the cellulase genes from multi-copy plasmids raises potential issues regarding phenotype stability, especially if expression of the cellulases exerts a toxic effect on the cell as described by Mingardon et al. [19]. These issues were exemplified by the authors' description of the generation of a heterologous C. acetobutylicum strain capable of expressing Cel48F from C. cellulolyticum with a subsequent loss of the phenotype after retrieval from spore stocks. Moreover, as no clostridial plasmids exhibit complete segregational stability in the absence of antibiotic selection, their use in a CBP process at industrial scale would be impractical.
Until recently, no suitable genetic tools have been available for constructing C. acetobutylicum strains containing heterologous DNA. The development of Allele-Coupled Exchange (ACE) [21] has made possible the generation of stable and iterative integrations within a comparatively short timescale. By using homology arms of different lengths, this approach allows control of the sequence of recombination events. At the second recombination event, a plasmid-borne sequence is coupled to a chromosomal sequence, forming a new allele that is selectable. Accordingly, in the present study we set out to test the potential of ACE for the construction of strains of C. acetobutylicum producing cellulosomes from gene clusters that are stably integrated into the genome, as opposed to being localised to unstable autonomous elements.

Results
Expression of C. thermocellum derived CipA scaffoldin variants in C. acetobutylicum driven by the host's chromosomal thiolase promoter CipA is the primary scaffoldin of C. thermocellum and is thus responsible for the integration of the enzymes into the complex [15,22]. To establish the maximum size of CipA and thus the complexity of the cellulosome that can be efficiently expressed and secreted by C. acetobutylicum various forms of the scaffoldin proteins were constructed; the modular nature of C. thermocellum CipA [Cthe_3077, NCBI: NC_009012] allowed the design and construction by standard BioBrick2 (BB2) assembly of synthetic modular parts of a series of scaffoldin proteins with different numbers of cohesin modules as well as the reconstruction of full length cipA (Figure 1a).
All CipA variants contained the native CipA endogenous signal peptide and the family 3a CBM (CBM3a), which binds to crystalline cellulose. The native full length cipA gene was PCR amplified from C. thermocellum genomic DNA and used to compare the expression level of the native gene to two full length codon optimised variants, one synthesised as a single fragment, the other built by BB2 assembly and therefore carrying several Ala-Ser substitutions (assembly "scars") within the linker regions of the protein. All encoded CipA variants (synthetic and native) were fused at the carboxy-terminal with the Flag epitope tag, allowing the detection and quantification of the secreted scaffoldin proteins by western blot analysis.
All CipA variants were inserted into the C. acetobutylicum genome by Allele-Coupled Exchange (ACE) technology as described by Heap et al. [21], using the pMTL-JH16 integration plasmid. The CipA fragments were inserted downstream from the thiolase (thl) gene and upon integration were orientated to be under the transcriptional control of the strong, constitutive thl promoter (P thl ) from the host chromosome ( Figure 1b).
The chromosomal integration of all transgenes was confirmed by PCR and positive clones were selected and subjected to SDS-PAGE and FLAG® colorimetric western blot analysis. As each of the ten scaffoldin variants carries a CBM3a, the recombinant C. acetobutylicum culture supernatants were mixed with Avicel (crystalline cellulose) and proteins harbouring a carbohydrate binding module (CBM3a), including the heterologous CipA variants, were bound to Avicel ( Figure 2). Avicel-bound proteins were recovered by boiling the samples in SDS sample buffer. This resulted in cleavage of the CipA scaffoldin proteins, due to a labile aspartyl-proline (Asp-Pro) peptide bond within the second cohesin domain which has been shown previously to be sensitive to cleavage upon boiling and in the presence of SDS [23].
Western blot analysis indicated the successful expression and secretion of all CipA variants by the recombinant strains. Notably, the amounts of the secreted CipA variants were observed to decrease gradually with the addition of each cohesin domain. Using a Flag-tagged reference protein of known concentration (50 ng/lane) we estimated that the final concentration of cellulose bound CipA variants ranged from 0.001 mg/l to 0.1 mg/l. The intracellular levels of heterologous proteins were very low and almost identical in all samples (data not shown) suggesting that all the scaffoldin proteins were exported. The low yield of secreted scaffoldin proteins, compared to a previous report by Perret et al. [17], could be at least partially explained by the relatively lower gene copy number. Similar expression levels (at around 0.001 mg/l) were observed for all 3 full length CipA variants ( Figure 3) indicating that neither codon optimisation nor the BB2 assembly scars had a substantial effect on the heterologous protein expression levels. Enhanced expression levels of CipA scaffoldin variants by using an additional promoter inserted downstream of the chromosomal promoter Recombinant strains obtained by integrating heterologous genes downstream of the thl gene contain an artificial thl-ermB-inserted gene operon with the expression of all three genes dependent upon the chromosomal P thl promoter ( Figure 1b). This gene arrangement might not be ideal when high protein expression levels are desired, but allows the expression of heterologous genes which are lethal to E. coli cells when cloned downstream of a strong promoter. To test whether we could improve the expression levels of the scaffoldin variants, we inserted a second copy of the thl promoter upstream of the cipA coding regions; this insertion proved not to be lethal to E. coli and we were able to integrate the expression cassette into the same thl locus of C. acetobutylicum. (Figure 1c). We chose not to insert a transcriptional terminator upstream of the newly-inserted second promoter in order to maximise the number of cipA encoding transcripts by allowing read-through from the upstream native promoter. Western blot analysis indicated that the insertion of the additional thl promoter at least doubled the expression levels of the scaffoldin proteins ( Figure 4), from around 0.1 mg/l to 0.2 mg/l for the shortest CipA variant. The effect of the additional promoter seemed to be equally substantial on the shorter scaffoldin variants (two or three cohesin modules) and the full length variants. In later experiments, the shorter scaffoldin variants (miniscaffoldins) with two and three cohesin domains (CipA2 and CipA3) were chosen for co-expression with cellulolytic enzymes.
Expression of C. thermocellum-derived cellulases in C. acetobutylicum driven by the host's chromosomal thiolase promoter After successful expression of the various scaffoldin units, we aimed to introduce C. thermocellum-derived cellulases into the C. acetobutylicum genome, individually as well as in combination with CipA scaffoldin subunits. Although we found that the presence of an additional promoter increased the expression levels of the heterologous scaffoldin variants, it was not possible to clone the cellulases behind the strong thl promoter as this proved to be lethal to E. coli cells. Therefore, the cellulases, with their endogenous signal sequences, were cloned with a ribosome binding sequence, but without a promoter, to prevent expression in E. coli. Upon insertion into the chromosome of C. acetobutylicum, transcription will be directed by the strong chromosomal thl promoter. Similarly, as we were unable to observe any benefit of codon optimisation on the expression level of the scaffoldin genes when compared to the native scaffoldin, we chose to express the native cellulase genes from C. thermocellum. Four of the most abundant C. thermocellum cellulosomal cellulases [24,25], Cel8A (Cthe_0269), Cel9B (Cthe_0543), Cel48S (Cthe_2089) and Cel9K (Cthe_ 0412) were chosen for expression in C. acetobutylicum both individually and as an operon together with the shorter (two or three cohesin domain) CipA scaffoldin variants (mini-scaffoldins).
In order to confirm expression of the enzymes, western blot analysis was carried out on concentrated supernatants from recombinant C. acetobutylicum strains grown until late exponential phase (OD600 > 3), where the pH of the cultures dropped below 5 due to production of organic acids by the cells. Three (Cel8A, Cel9B and Cel9K) out of the four enzymes were expressed at similar but low levels (at around 0.05 mg/l) (Figure 5a), whereas Cel48S could not be detected at all. We observed that under these conditions, all other enzymes showed degradation products (Figure 5a), suggesting that Cel48S may be being completely degraded. In support of this, Cel48S could be detected in cultures grown in buffered media (pH 7, 40 mM MOPS) until early exponential phase (OD600 0.4-0.7) and before the pH of the culture dropped below 6 ( Figure 5b). Degradation of heterologous cellulases by C. acetobutylicum has been previously reported, most often apparently due to proteolysis of the susceptible C-terminal dockerin domain [19,26,27]. Such truncations would not be detected in the experiments described here, as the Flag tag would be removed.
Secretion of the C. cellulolyticum-derived family 48 enzyme (Cel48F) and some larger family 9 (Cel9G and Cel9E) enzymes in C acetobutylicum has been previously shown to be difficult and it has been suggested that the secretion of these enzymes might require the presence of specific chaperone(s) not present in C. acetobutylicum [20].

Expression of C. thermocellum-derived cellulosomal operons in C. acetobutylicum
To test whether our genome integration platform is suitable for expression of operons of various complexities from the chromosome of C. acetobutylicum, we built a series of cellulosomal operons comprising a miniscaffoldin (CipA2 or CipA3) and one or two cellulases. For this, we chose Cel8A and Cel9B, to represent families containing small and large enzymes, respectively. Cel8A belongs to the family of GHs with a small catalytic domain shown to be secreted efficiently, both in previous studies [19] and in this work, while Cel9B belongs to the family of enzymes with large catalytic domains similar to Cel9G and Cel9E from C. cellulolyticum. In earlier studies, the expression of GH9 cellulases in C. acetobutylicum proved to be either toxic to the cells [19] or their secretion was only triggered when a CBM3a, combined with the X2 modules, was grafted at the N terminus of the protein [20]. Nevertheless, in the chromosomal system described above (see previous result section), Cel9B was expressed and secreted at similar level to Cel8A.
In our initial experiments, we assembled operons containing the genes in the order they are found in simple cellulosomal systems, where the first gene of the operon is the scaffoldin gene followed by several hydrolytic enzymes. The first operons we tested contained a mini-scaffoldin (CipA2 or CipA3) and a hydrolase (Cel8A) integrated into the host's thiolase locus and driven by the chromosomal thiolase promoter. We observed that both CipA2 and CipA3 mini-scaffoldin proteins were expressed at a higher level than Cel8A. In addition, the expression of Cel8A as part of the operon was lower than when it was expressed on its own (data not shown). These observations prompted us to make two alterations to our initial strategy. Firstly, as the mini-CipA3 gene displayed a greater level of expression than observed for Cel8A, we placed the miniscaffoldin at the end of the operon, assuming that this would provide a better ratio of cellulase to scaffoldin. Secondly, as we had previously shown that adding an additional thl promoter leads to an increase in the level of expression of the scaffoldin variants, we placed all cellulases and cellulosomal operons under the control of an additional promoter. As it was not possible to clone any of the cellulase genes or cellulosomal operons downstream of the thl promoter in E. coli, we modified it to contain a symmetrical or 'ideal' lac operator sequence at its transcriptional start site (PthlOID promoter -nucleotide sequence can be found in Additional file 1), thereby allowing repression of transcription in an E. coli host harbouring lacI Q [17]. In this case, the lacI Q gene was incorporated into the vector backbone. In total, six constructs containing the thlOID promoter were assembled in E. coli and subsequently integrated into the genome of C. acetobutylicum ( Figure 6).
C. acetobutylicum recombinant strains expressing these cellulosomal operons were grown in buffered media (pH 7 using 40 mM MOPS) and tested at early exponential phase (OD600 around 0.7-0.8), before the pH of the cultures dropped below 6. Assaying the samples at early exponential phase ensured that the secreted heterologous proteins were intact; however their overall yield at the time of sampling was lower than expected (similar to that of strains expressing individual enzymes without an additional promoter, i.e. transcribed only from the upstream thl promoter, grown until late stationary phase). This is perhaps not surprising as the heterologous operons were placed under the control of a constitutively expressed promoter and the accumulation of secreted proteins is expected until late stationary phase. Western blot analysis (Figure 7) indicated that the expression levels of Cel8A (0.05 mg/l) and CipA3 (0.05 mg/l) were comparable when expressed as single entities, while Cel9B (0.01 mg/l) was synthesised at a much lower level. Furthermore, while the placement of cel8A before cipA3 in the operon does not appear to decrease the expression of CipA3, it is apparent that positioning cel9B upstream of cel8A (or theoretically cipA3 at the end of the three gene operon) has a significant negative impact on the expression of Cel8A. As Cel9B co-migrates with CipA3, we are unable to determine the level of expression of Cel9B or CipA3 from the western blot when these two proteins are co-expressed in the Cel9B-CipA3 or Cel9B-Cel8A-CipA operons.

In vivo mini-cellulosome assembly by Clostridium acetobutylicum
In order to confirm the formation of a mini-cellulosome in vivo, we carried out native PAGE analysis on samples concentrated by ultrafiltration. Figure 8 shows that the cellulosomal components readily form complexes; the predominant rapid migrating species of CipA3, Cel8A and Cel9B are absent in the lanes of culture supernatants containing mixtures of the scaffoldin variant and the enzymes. The multiple bands likely reflects different isoforms of the enzymes (different charged species through deamidation of aspartates and glutamates or weak oligomerisation), although the complex pattern may also reflect incomplete saturation of CipA3. Additionally, it is possible to contrast the complexes formed by the expression of the three different operons; the complexes formed by Cel9B in combination with CipA3 display faster migration than those formed by Cel8A in combination with CipA3, while the co-expression of all three subunits generates a pattern that is a mixture of the two banding patterns. This latter hybrid pattern indicates that some of the CipA3 molecules are bound to both Cel8A and Cel9B. It should be noted that in the lanes containing cohesin-enzyme combinations the antibody staining is much more intense, which may indicate that only a proportion of the enzymes, when not bound to CipA, migrate into the non-denaturing gels.

Enzymatic activity of the heterologous mini-cellulosomes secreted by C. acetobutylicum
To assess the activity of the heterologously expressed enzymes alone and in complex with mini-scaffoldin we assayed the activity of the early stationary phase culture supernatants (OD600~2.0) against cellohexaose (Cel9B, Cel8A and Cel9K) at 60°C, using HPAEC-PAD to monitor product release, and also 4-nitrophenyl-β-D-cellobioside (4-NPC) in the case of Cel9K. The data show that supernatants from wild-type C. acetobutylicum and cells expressing CipA3 alone display no detectable activity against the oligosaccharide or 4-NPC (Table 1). In contrast, cells expressing Cel8A from the thlOID promoter, either alone or with CipA3, display significant activity (Table 1), although the presence of the scaffoldin reduced the observed specific ac-tivity~2.5-fold. The activity of Cel8A expressed only from the upstream thiolase promoter was 20-fold lower than when transcribed from thlOID (Table 1), consistent with the differential expression of the scaffoldin variants under control of one or two promoters (Figure 4). The activity for all Cel9B-expressing cell supernatants was above background, but much lower than for Cel8A. Similarly, Cel9K activity was low but was clearly above background against both cellohexaose and 4-NP.

Discussion
The development of ACE technology [21] has made it possible to overcome the challenge of stably modifying the genome of C. acetobutylicum, and our results described here demonstrate the secretion, assembly and activity of mini-cellulosomes expressed from the chromosome of stable recombinant C. acetobutylicum strains, a major milestone in the engineering of this organism for butanol production via CBP. However, there are still several obstacles to be overcome. The first is the apparent impact of the gene arrangement within an operon on the expression of individual genes. As seen here and reported previously [28,29], the configuration of an operon greatly affects the expression of its genes. The rearrangement of genes within an operon results in changes in the sequence of the encoded mRNA and thus its secondary structure and its related properties. These changes will affect the rate of transcription and translation and the stability and/or degradation rate of the newly arranged mRNA molecules, Figure 6 Synthetic operons inserted into the C. acetobutylicum chromosome. Encoded genes are: thl, thiolase; ermB, erythromycin ribosomal methylase B, C. thermocellum derived cellulases cel9B and cel8A, mini-scaffoldin cipA3. All operons are preceded by the chromosomal thl promoter and the celllulosomal components by an additional synthetic thlOID promoter and followed by a transcriptional terminator (from Clostridium pasteurianum ferredoxin gene). ultimately resulting in altered protein levels. While we could potentially adjust the composition of the operon to identify the optimal configuration, it may be also possible to use multiple promoters to drive the expression of each component individually. The use of mono-cistronic expression constructs should also avoid a series of variations encountered when using operons. The use of BB2-standard components would easily provide assembly of cassettes containing the enzymes and/or scaffoldin under the control of different promoters. The use of promoters with different strengths may also allow a finer level of control over the proportions of each component in the secreted cellulosomes.
Another significant problem is the toxicity of heterologous cellulases when expressed in C. acetobutylicum [19]. The catalytic activity of the enzymes is unlikely to contribute to their lethality. A more likely explanation is that the secretion of the heterologous enzymes becomes stalled, and the resultant "blockage" causes a general defect in protein secretion leading to cell death. Thus, there is an urgent need to test this hypothesis and, if shown to be correct, approaches will need to be developed that ensure efficient export of the cellulases, possibly through the use of chaperones and appropriate secretion signals.
The third obstacle is to identify the cellulosome composition required for C. acetobutylicum growth on lignocellulose. The ability of cellulosomal hydrolases with different activities to act in synergy has been demonstrated; a trifunctional complex containing an exoglucanase (Cel48F), endoglucanase (Cel9G) and xylanase (XynZ) was shown to display greatly improved activity over bifunctional complexes [30]. While we have demonstrated the expression of both Cel8A and Cel9B, the recombinant strains developed are still unable to grow on a natural substrate, namely milled wheat straw (data not shown). Although C. acetobutylicum displays a small level of xylanase activity, it is possible that this is not enough to efficiently degrade the hemicellulose component of wheat straw and it may thus be necessary to introduce a xylanase as part of the enzyme cohort. Additionally, in order to best take advantage of the potential for enzyme synergy afforded by the use of a cellulosome, it may be important to also include an exoacting cellulase, such as Cel9K, in addition to an endoglucanase (Cel8A) and endoprocessive cellulase (Cel9B). Recently, Anderson et al. [31] created a strain of B. subtilis able to grow on untreated natural feedstocks through expression of C. cellulolyticum Cel5A, Cel9E, and Cel48F, an endoglucanase/xylanase, endoglucanase/exoglucanase, and a processive endoglucanase respectively, in combination with a cell wall-associated mini-scaffoldin. When grown on pretreated feedstocks, the rate of hydrolysis by Bacillus cells expressing three enzymes with the scaffoldin was greater than that of those expressing only two enzymes, with expression of only one enzyme giving the weakest activity. Finally, a fourth challenge may be the apparent difficulty of expressing longer scaffoldin variants. While we have demonstrated that the shorter scaffoldin variants (CipA2 and CipA3) are readily expressed, the expression levels of the longer variants are greatly reduced. While former studies [17] have examined expression of miniscaffoldin constructs in C. acetobutylicum, we are unaware of any attempted expressions of full-length heterologous scaffoldins in C. acetobutylicum to date. We hypothesise that there may be an issue with mRNA stability [32], either acetobutylicum strains. The culture supernatants of the recombinant strains (thlOIDCel9B, thlOIDCel8A, thlOIDCel9B_CipA3, thlOIDCel8A_ CipA3, thlOIDCel9B_Cel8A_CipA3, thlOIDCipA3) were concentrated 70-fold by ultracentrifugation and 10 μl of the concentrated samples loaded onto a native PAGE Tris-acetate gel. The presence of the cellulosomal subunits and the in vivo assembly of the minicellulosomes was detected using ANTI-FLAG M2 monoclonal antibodyhorseradish peroxidise conjugate (Sigma). The presence of multiple bands in the samples expressing mini-cellulosomes (thlOIDCel9B_ CipA3, thlOIDCel8A_CipA3 and thlOIDCel9B_Cel8A_CipA3) likely represents unsaturated complexes with cellulases binding different cohesin domains. These samples were loaded at two different concentrations: 10 and 5 μl (½) of the original samples.
due to the length of the construct, or the high level of homology between the different scaffoldin domains. Although degradation of lignocellulose has been shown to be possible with minicellulosomes containing only three enzymes [31], should we wish to increase the complexity of our minicellulosomes in the future, we may have to examine ways of increasing the stability, for example, by examining the effects of different 5′ mRNA leader sequences, or by engineering the sequence to reduce homology.

Conclusions
We have successfully demonstrated the expression, secretion and complex formation of three C. thermocellumderived cellulosome components inserted into the genome of C. acetobutylicum, providing a proof of principle of the utility of the ACE method to create stable C. acetobutylicum strains for the expression of a functional cellulosome. Generation of a strain able to utilise lignocellulose will require further improvement of the expression levels of the cellulolytic enzymes along with increased range of catalytic activities.

Design of synthetic DNA fragments
All DNA fragments were designed as standard, interchangeable "parts" based on BioBrick BB-2 RFC [12] (BB2) idempotent assembly proposed by Tom Knight [33]. This assembly means that any newly composed part will adhere to its assembly standard and can be used in future assemblies [http://dspace.mit.edu/handle/1721.1/45139]. Each part consists of a DNA fragment flanked by restriction sites, called prefix at the 5′ end and suffix at the 3′ end.
The BB2 prefix contains the EcoRI, NotI and SpeI restriction sites while the prefix contains the NheI/NotI/PstI sites. These sites allow parts to be assembled together, forming a new part which will maintain the same prefix and suffix and contain a scar (GCTAGT, corresponding to alanine and serine) where the cut and re-ligated restriction sites (SpeI and NheI) were fused together. We chose to use this method to build a range of synthetic cellulosomal units of various complexities.
The full length (5.6 kb) cipA gene [Cthe_3077] encoding the scaffoldin protein from C. thermocellum was codon optimised for usage in Clostridium acetobutylicum and the gene synthetically constructed by DNA 2.0 (Menlo Park, CA) as a single DNA fragment lacking the stop codon. To this fragment, the ribosome binding site of the strong constitutive thiolase promoter (Pthl) from C. acetobutylicum and the standard BioBrick2 prefix and suffix restriction sites were added, resulting in the BB2prefix_RBS_CipA_BB2suffix standard assembly part [deposited at GenBank, KC560744]. In parallel, the eleven scaffoldin subunits of the full length CipA, each encoding one of the nine cohesin modules, the carbohydrate binding module (CBM3a), or the X module, and flanked by the linker regions (where appropriate), were also synthesised by DNA 2.0 as standard BioBrick2 modular parts [deposited at GenBank: KC560733, KC560734, KC560735, KC560736, KC560737, KC560738, KC560739, KC560740, KC560741, KC560742, KC560743]. A list of plasmids containing all synthetic genes synthesised in BB2 format can be found in Additional file 2: Table S1.
This modular design allowed the standard assembly of a series of scaffoldin proteins with various numbers of cohesin modules and the reconstruction of the full length cipA [CipABB2, Genbank KC560745]. A list of plasmids created by BB2 assembly is listed in Additional file 2. As the assembly using the BB2 technology results in an alanine-serine (Ala-Ser) scar, the ten "breaking points" within the scaffoldin sequence were chosen to coincide either with an alanine (A) or a threonine (T) in the linker regions (in one case inside the second cohesin domain); the adjacent amino acid was omitted and, as a result of assembly, replaced by the serine (Ser) from the scar (The CipA amino acid sequence with highlighted features, including the BB2 assembly points can be found in Additional file 1).
The same design principles were applied to create all synthetic DNA fragments encoding the four GHs from C. thermocellum (Cel8A-Cthe0269, Cel9B-Cthe0543, Cel48S_Cthe2089 and Cel9K_ Cthe0412), except these genes were not codon optimised. The BB2 prefix and RBS from the thl promoter were inserted upstream of the coding sequence and the BB2 suffix inserted downstream resulting in standard BB2 parts (BB2prefix_RBS_ GH _BB2suffix).  The thlOID promoter, containing the lac operator sequence was designed as a BB2 standard assembly part, containing the lac operator sequence and lacking the original RBS of the promoter sequence (The sequence of the thlOID promoter can be found in Additional file 1).

Plasmid construction
All synthetic CipA scaffoldin fragments synthesised by DNA 2.0 were delivered in one of the pJ20x plasmids [Additional file 2: Table S1]. All four cellulases (Cel8A, Cel9B, Cel9K and Cel48S) were synthesised as standard BB2 parts by Biomatik (Canada) and delivered in one of the company's in-house vectors (pBMH) [Additional file 2: Table S1] The pJ202_Flag_2xSTOP vector, carrying the Flag tag and the stop codon, was created by annealing the Flag-Stop_Fw_CCCGGGGAATTCGCGGCCGCACTAGT GATTACAAGGATGACGACGATAAGTAATGAGCTAG CGCGGCCGCCTGCAGCCCGGG and Flag-Stop_Rev_ CCCGGGCTGCAGGCGGCCGCGCTAGCTCATTACT TATCGTCGTCATCCTTGTAATCACTAGTGCGGCCG CGAATTCCCCGGG primers and inserting the EcoRI/ PstI-digested fragment into the pJ202 backbone. All three pJ20x plasmid backbones were used for Biobrick2 standard assembly to create all desired DNA fragments for integration into the C. acetobutylicum genome [list of BB2 plasmids in Additional file 2].
The native CipA gene was amplified from the genomic DNA of C. thermocellum using the following primers: CipAtherm_Fw CACCATATGAGAAAAGTCATCAGTA TGCTCTTAGTTGTGGC and CipAtherm_Rev CACAC TAGTCTGTGCGTCGTAATCACTTGATG. Several cloning steps were necessary to create the native CipA variant in order to introduce the same RBS and Flag tag as the synthetic CipA genes. Firstly, the 5.6 kb PCR product was digested with NdeI/EcoRI and SpeI restriction enzymes (Fermentas) and two smaller fragments (a 4.5 kb and a 1.1 kb) were gel purified. The larger 4.5 kb fragment was ligated into an EcoRI/SpeI-cut BB2 backbone which carried the Flag tag (pJ202_Flag_2×Stop), generating the pJ202_truncCipA_Flagx2Stop plasmid. The truncated, Flag-tagged CipA was released from this plasmid by EcoRI/NheI digest and ligated together with the previously gel purified smaller (1.1 kb) PCR fragment into NdeI/ NheI-digested pMTL_JH16_RBS, resulting in thepMTL_ JH16_RBS_nativeCipA_Flag plasmid.
All BB2-assembled DNA fragments (synthetic and native cipA variants and all four cellulases) were fused at the carboxy-terminal with the Flag epitope tag (added as a BB2 fragment) and subsequently cloned as a NotI/NheI fragment into the pMTL-JH16 integration vector [21], resulting in pMTL-JH16_RBS_gene integration vectors [Additional file 2: Table S2]. Integration vectors which contained an additional thiolase promoter (Pthl) upstream of the CipA coding regions were generated by inserting the thl promoter into the pMTL-JH16_CipAx vectors as a NotI-NdeI fragment with the NdeI site overlapping the CipA start codon.
In order to achieve sufficient repression in E. coli cells of constructs containing cellulases and cellulosomal operons driven by the thlOID promoter, we modified the pJ201 Biobrick2 vector and the pMTL-JH16 integration vector to contain the lacI Q gene. The lacI Q gene sequence was amplified from the pGEX-6P-1 (GE Heathcare) vector using the FwlacIQ CCCGGGGTTTAAACGACACCAT CGAATGGTGCAAAACC and RevlacIQ CCCGGGGTT TAAACTTAATTGCGTTGCGCTCACTGC primers and subsequently cloned into the SmaI site of pJ201 and the PmeI site of the pMTL-JH16, resulting in the pJ201_lacIQ and pMTL-JH16_lacIQ plasmids [Additional file 2: Table S2].

Bacterial strains
Escherichia coli TOP10 (Invitrogen) strain was used for all BioBrick 2 cloning steps. E. coli cultures were grown in Luria-Bertani (LB) broth at 37°C with rotary shaking at 200 rpm, or on LB plates supplemented with the appropriate antibiotics (ampicillin 100 μg/ml, kanamycin 50 μg/ml, zeocin 25 μg/ml, chloramphenicol 25 μg/ml). The same E. coli strain containing the pAN2 [34] plasmid was used for in vivo methylation of plasmid DNA prior to Clostridium acetobutylicum ATCC 824 transformations.
E. coli BL21 (DE3) (NEB) cells were used for cloning most of the cellulases and cellulosomal operons behind the thlOID promoter. For cloning the more complex cellulosomal operons (thlOID_Cel9B_CipA3 and thlOID_ Cel9B_Cel8A_CipA3) into the modified integration vector (pMTL-JH16_lacIQ), E. coli XL10-Gold KanR ultracompetent cells (Agilent) were used. BL21 and XL10 E. coli cells with LacI Q -containing vectors were all grown at 30°C in order to further reduce the toxicity of the enzymesthis required an extra 24 h growth on plates and an extra 6 h in liquid culture.
Clostridium acetobutylicum ATCC 824 wild type strain and the subsequent recombinant strains [Additional file 2: Table S2] were grown in static culture at 37°C under an anaerobic atmosphere of N2:H2:CO2 (80:10:10, vol:vol: vol) in an anaerobic workstation (Don Whitley, UK) using 2×YTG media prereduced overnight under the same conditions. Methylated plasmids were transferred into C. acetobutylicum by electroporation as described previously [35]. Recombinant strains were selected on antibioticcontaining CGM plates [36] (primary transformants on 15 μg/ml thiamphenicol and double recombinants on 40 μg/ml erythromycin) and screened for chromosomal integration by PCR as described previously [21].
Clostridium thermocellum ATCC 27405 was grown at 60°C in air tight serum bottles containing pre-reduced GS-2 media [37] and used for preparation of genomic