The first cluster contains 134 genes highly expressed during exponential growth (hours 6 to 10; see Additional data file 2 for a list of the genes). This cluster characterizes highly motile vegetative cells (Figure 1b, I) and, given the minimal amount of knowledge on the genes responsible for motility and chemotaxis in clostridia, our analysis offers the possibility of identifying these genes at the genome scale 14. This cluster includes the flagella structural components flagellin and flbD, the main chemotaxis response regulator, cheY (CAC0122; responsible for flagellar rotation in B. subtilis  15), as well as several methyl-accepting chemotaxis receptor genes (CAC0432, CAC0443, CAC0542, CAC1600, CAP0048). COG analysis showed that genes related to cell motility (COG class N) and nucleotide transport and metabolism (COG class F) were overrepresented in this cluster (Additional data file 2). In order to investigate cell motility further, all genes that fell within this COG class were hierarchically clustered according to their expression profiles (see Additional data file 3 for Figure S2 and discussion). Interestingly, the two main cell motility gene clusters, the first including most of the flagellar assembly and motor proteins and the second containing most of the known chemotaxis proteins, clustered together and displayed a bimodal expression pattern (Figure S2). The genes were not only expressed during exponential phase but also during late stationary phase, around hour 38, which is consistent with the observation that a motile cell population was again observed in late stationary phase. Included in the category of nucleotide transport and metabolism are several purine and pyrimidine biosynthesis genes: a set of five consecutive genes, purECFMN, the bi-functional purQ/L gene, purA, pyrPR, pyrD, and pyrI. Two other purine synthesis genes (purH, purD) showed very similar profiles but were not classified within this cluster by the clustering algorithm. Vegetative cells, which correspond to this cluster, produce ATP through acidogenesis, whereby the cells uptake glucose and convert it to acetic and butyric acid. Because glucose is the main energy source, multiple genes for glucose transport were included within this cluster, including the glucose-specific phosphotransferase gene, ptsG, the glucose kinase glcK and CAP0131, the gene most similar to B. subtilis glucose permease glcP. The genes required for the metabolism of glucose to pyruvate did not show temporal regulation, suggesting that expression of these genes is constitutive-like (see Additional data file 3 for Figure S3 and discussion). Acetic acid production genes pta and ack were not temporally expressed, but butyrate production genes ptb and buk were. Though expressed throughout exponential phase, the expression of both ptb and buk slightly peaked during late exponential phase, as previously seen 7, and thus fall in the transitional (second) cluster. Analysis of the expression patterns of all the genes involved in acidogenesis, not just the differentially expressed genes discussed here, is included in Figure S3 in Additional data file 3. Finally, the expression patterns of the two classes of hydrogenases (iron only and nickel-iron) were investigated (Figure S3 in Additional data file 3). hydA, the iron only hydrogenase that catalyzes the production of molecular hydrogen, was expressed only during exponential phase, whereas the iron-nickel hydrogenase, mbhS and mbhL, was expressed throughout stationary phase.

The transitional phase is captured by 139 genes in the second cluster (Figure 1c,d; Additional data file 2). It is made up of genes that show elevated expression between hours 10 and 18 and is when solvent formation was initiated. This cluster characterizes the shift from vegetative cells to cells committing to sporulation and thus includes two important regulators of sporulation, abrB (CAC0310) and sinR (CAC0549), which are discussed in more detail below. Also characteristic of this shift from vegetative growth to sporulation was the overrepresentation of genes related to energy production and conversion (COG class C), since sporulation is an energy intensive process. Solvent production began in the transitional phase, though the genes responsible for solvent production fall in the next (third) cluster; the third cluster partially overlaps with this second cluster but is distinguished by a sustained expression pattern. In response to these solvents, C. acetobutylicum undergoes a change in its membrane composition and fluidity, generally decreasing the ratio between unsaturated to saturated fatty acids 161718. Consistent with this change, genes related to lipid metabolism (COG class I) were overrepresented in this cluster. To further investigate this COG class, all genes identified as COG class I were hierarchically clustered (see Additional data file 3 for Figure S4 and discussion). Seven genes that were upregulated just before the onset of sporulation fall within the same operon and are related to fatty acid synthesis. In contrast, many of the most characterized genes involved in fatty acid synthesis (accBC, fabDFZ, and acp) maintain a fairly flat profile throughout the timecourse (Figure S4 in Additional data file 3). Also within this cluster is the gene responsible for cyclopropane fatty acid synthesis (cfa), though classified in COG class M (cell envelope biogenesis) and not COG class I. Importantly, the ratio of cyclopropane fatty acids in the outer membrane has been shown to increase as cells enter stationary phase 1819, but the overexpression of this gene alone was unable to produce a solvent tolerant strain 19. Though not overrepresented in this cluster, all the genes within COG class M were also hierarchically clustered (see Additional data file 3 for Figure S5 and discussion). The transitional cluster also included several genes related to iron transport and regulation like the fur family iron uptake regulator CAC2634, the iron permease CAC0788, feoA, feoB, fhuC, and two iron-regulated transporters (CAC3288, CAC3290), which is consistent with the earlier, more limited data 7. Significantly, iron-limitation has been found to promote solventogenesis 20.

The third cluster (Figure 1c,d; Additional data file 2) of 175 upregulated genes represents the solventogenic/stationary phase as it contains all key solventogenic genes. This cluster characterizes the transcriptional pattern of clostridial cells, the unique developmental stage in clostridia and first recognizable cell type of the sporulation cascade, and exhibited a longer upregulation of gene expression than the previous two clusters. Indeed, its range overlapped the previous (second) and the next two (fourth and fifth) clusters. The clostridial form is generally recognized to be the form responsible for solvent production 821 and is distinguished morphologically as swollen cell forms with phase bright granulose within the cell 21. This cluster captures both of these characteristics with the inclusion of the solventogenic genes and several granulose formation genes. The solventogenic genes adhE1-ctfA-ctfB, adc, and bdhB were initially induced during transitional phase, the second cluster, but were expressed throughout stationary phase and were thus placed within this cluster. Two granulose formation genes, glgC (CAC2237) and CAC2240, and a granulose degradation gene, glgP (CAC1664), were included within this cluster. The other two granulose formation genes, glgD (CAC2238) and glgA (CAC2239), though not included in this cluster, displayed a similar expression profile to glgC and CAC2240. The concomitant requirement of NADH during butanol production drove the expression of three genes involved in NAD formation: nadABC. Expression of the stress-response gene hsp18, a heat-shock related chaperone, and the ctsR-yacH-yacI-clpC operon, containing the molecular chaperone clpC and the stress-gene repressor ctsR, also fell in this cluster and paralleled the expression of the solventogenic genes (see Additional data file 3 for Figure S6). Other important stress-response genes, groEL-groES (CAC2703-04) and hrcA-grpE-dnaK-dnaJ (CAC1280-83), mirrored this expression pattern, though were not differentially expressed according to the strict criteria employed for selecting the genes of Figure 2c,d (Figure S6 in Additional data file 3). Although genes encoded on the pSOL1 megaplasmid 22 represent less than 5% of the genome, they constitute 15% of genes in this cluster. pSOL1 harbors all essential solvent-formation genes and, importantly, some unknown gene(s) essential for sporulation 22. Besides the genes listed in this cluster, the vast majority of the genes located on pSOL1 were expressed throughout stationary phase, with most being upregulated at the onset of solventogenesis (see Additional data file 3 for Figure S7). Several key sporulation-specific sigma factors (σ^F, σ^E, σ^G) and the σ^F-associated anti-sigma factors in the form of the tricistronic spoIIA operon (CAC2308-06) belong to this cluster along with one of the two paralogs of spoVS (CAC1750) and one of three spoVD paralogs (CAP0150). The second spoVS paralog (CAC1817) did not meet the threshold of expression in 12 of the 25 timepoints; the other two paralogs of spoVD (CAC0329, CAC2130) were above the expression cutoff but did not show significant temporal regulation. Of unknown significance was the expression of a large cluster of genes involved in the biosynthesis of the branched-chain amino acids valine, leucine and isoleucine (CAC3169-74) coinciding with the onset of solventogenesis, as shown before 723, as well as the upregulation of several glycosyltranferases (see Additional data file 3 for Figure S8). The upregulation of valine, leucine, and isoleucine synthesis genes could be indicative of a membrane fluidity adaptation 7. In B. subtilis, these branched-chain amino acids can be converted into branched-chain fatty acids and change the membrane fluidity 24, and under cold shock stress, B. subtilis downregulates a number of genes related to valine, leucine, and isoleucine synthesis 25. Therefore, this upregulation may be another mechanism to change membrane fluidity, though the ratio of unbranched and branched fatty acids has not been reported in studies investigating membrane composition 16171826.

The fourth cluster (Figure 1c,d; Additional data file 2) of 84 genes represents a sharp induction of expression between 18 and 24 hours (early stationary phase). This cluster falls within the stationary (third) cluster described above. This is a compact group, with 70% belonging to one of three COG categories: carbohydrate transport and metabolism, transport and metabolism of amino acids, and inorganic ion transport and metabolism. A number of different carbohydrate substrate pathways, from monosaccharides (fructose, galactose, mannose, and xylose) to disaccharides (lactose, maltose, and sucrose) to complex carbohydrates (cellulose, glycogen, starch, and xylan), were investigated, and many exhibited upregulation during stationary phase, though only a few are highly expressed (see Additional data file 3 for Figure S9). The significance of this upregulation of non-glucose pathways is unknown, because sufficient glucose remains in the media (approximately 200 mM or about 44% of the initial glucose level). Of particular interest was the upregulation of several genes related to starch and xylan degradation (Figure S9 in Additional data file 3). The two annotated α-amylases (CAP0098 and CAP0168) along with the less characterized glucosidases and glucoamylase were all upregulated throughout stationary phase and a number were highly expressed, like CAC2810 and CAP0098. Also upregulated were the predicted xylanases CAC2383, CAP0054, and CAC1037, with CAP0054 and CAC1037 being highly expressed during stationary phase. Mirroring this pattern were CAC1086, a xylose associated transcriptional regulator, and the highly expressed CAC2612, a xylulose kinase. The genes related to glycogen metabolism are believed to be involved in granulose formation, as discussed earlier. Several genes for arginine biosynthesis (argF, argGH, argDB, argCJ, carB) were induced during this time, probably as a result of its depletion in the culture medium.

The fifth cluster (Figure 1c,d; Additional data file 2), representing the middle stationary phase, contains 120 genes mainly expressed between hours 24 and 36, and again falls within the stationary (third) cluster described above. Most of the genes in this cluster activate the sporulation-related sigma factors (σ^F, σ^E, σ^G) or are putatively regulated by them. These include spoIIE, the phosphatase that dephosphorylates SpoIIAA and results in the activation of σ^F, and the σ^E-dependent operons spoVR (involved in cortex synthesis), spoIIIAA-AH (required for the activation of σ^G), and spoIVA (involved in cortex formation and spore coat assembly). The σ^G-dependent spoVT gene has two paralogs in C. acetobutylicum (CAC3214, CAC3649); the transcriptional pattern suggests that CAC3214, included in this cluster, is the real spoVT. Sporulation-related genes included in this cluster are three cotF genes, one cotJ gene, one cotS gene, the spore maturation protein B, a small acid soluble protein (CAC2365), and two spore lytic enzymes (CAC0686, CAC3244). Though several sporulation-related genes are included in the next (sixth) cluster as well, most, beyond those listed here, are upregulated in mid-stationary phase (see Additional data file 3 for Figure S10 and discussion). Seven genes of the putative operon (CAC3157-63) encoding genes for tryptophan synthesis from chorismate and ten genes for histidine synthesis (CAC0935-43, CAC3031) were also included here.

The sixth cluster, representative of the late stationary phase, includes 162 genes mainly expressed after hour 36 (Figure 1c,d; Additional data file 2). This cluster captured the expression profiles of the forespore and endospore forms, free spores, and late-stage vegetative-like cells. The endospore form represents the last stage before mature spores are released, and therefore fewer sporulation-related genes are within this cluster than previous ones. The sporulation-related genes included in this cluster are two small acid-soluble proteins (CAC1522 and CAC2372), a spore germination protein (CAC3302), a spore coat biosynthesis protein (CAC2190) and a spore protease (CAC1275). Also within this cluster are the two phosphotransferase genes, CAC2958 (a galactitol-specific transporter) and CAC2965 (a lactose-specific transporter), another annotated cheY (CAC2218), various enzymes related to different sugar pathways (CAC2180, CAC2250, CAC2954), and two glycosyltransferases (CAC2172, CAC3049). Expression of these genes may be reflective of the late-stage vegetative-like cells observed during microscopy and demonstrate they have a different genetic profile compared to the early vegetative cells. Interestingly, this cluster is enriched in defense mechanism genes (COG class V) like a phospholipase (CAC3026) and multidrug transporters that may play a role in resistance to a variety of environmental toxins.

Two additional gene classes (cell division and ribosomal proteins), though not overrepresented in any of the six clusters described above, were investigated because of their importance in cellular processes and interesting expression patterns. COG class D (cell division and chromosome partitioning), besides important genes for vegetative symmetric division, includes ftsAZ, important for both symmetric and asymmetric cell division, and soj (a regulator of spo0J) and spoIIIE, important for proper chromosomal partitioning between the mother cell and prespore. These genes, along with several uncharacterized genes, were upregulated at the beginning of sporulation (see Additional data file 3 for Figure S11). Almost all the ribosomal proteins were downregulated as the culture entered stationary phase, and interestingly, about half of those downregulated genes were again upregulated in mid-stationary phase and remained upregulated until late-stationary phase (see Additional data file 3 for Figure S12). This upregulation is likely related to the late-stage vegetative-like cells seen.

Sporulation in bacilli is initiated by a multi-component phosphorelay 27, which is absent in clostridia, but the master regulator of sporulation, Spo0A, is conserved 113. Briefly, in B. subtilis, phosphorylated Spo0A promotes the expression of prespore-specific sigma factor σ^F and mother cell-specific sigma factor σ^E 28. σ^F is followed by σ^G, which is controlled by both σ^F and σ^E, and σ^E is followed by σ^K, which is controlled by σ^E and SpoIIID 28. sigH expression, in bacilli, is induced before the onset of sporulation and aids spo0A transcription 28. Here, sigH expression underwent a modest two-fold induction, relative to the first timepoint, during the onset of sporulation but never increased beyond three-fold, in contrast to all other sporulation factors (Figure 3a). spo0A expression also peaked during the onset of sporulation at over 12-fold and maintained a minimum of 3-fold induction until hour 36 (Figure 3a,b). Once phosphorylated, in bacilli and likely in C. acetobutylicum 29, Spo0A regulates the expression of the operons encoding sigF, sigE, and spoIIE 30, the latter of which acts as an activator of σ^F. sigF and sigE exhibited an initial 16- and 8-fold induction, respectively, at hour 12, the timing of peak spo0A expression, but a second higher level of induction, 46- and 66-fold, respectively, was reached later at hour 24 (Figure 3c) and confirmed with Q-RT-PCR (Figure 2). The plateau or decrease in expression of spo0A, sigF, and sigE coincided with the peak expression of two known repressors, abrB and sinR, of sporulation genes in B. subtilis (Figure 3b), the former repressing the expression of spo0A promoters and the latter directly binding to the promoter sequences of the spo0A, sigF, and sigE operons 3132. C. acetobutylicum contains three paralogs of abrB, among which CAC0310 exhibited the highest promoter activity and, when downregulated, causes delayed sporulation and decreased solvent formation 33. sinR (CAC0549) expression in C. acetobutylicum was previously reported 33 to be weak, but our data show a significant amount of expression and suggest a similar role as that in B. subtilis. In B. subtilis, Spo0A either indirectly (sinR) or directly (abrB) represses the genes of these two repressors 3234. The expression patterns of both genes did decrease after peak Spo0A~P deduced activity (Figure 4b; see below), indicating a similar regulatory network may be involved in C. acetobutylicum. sigF, sigE and sigG have very similar expression patterns (Figure 3c). Both sigF and sigE are activated by Spo0A~P, so similar expression profiles were expected. In B. subtilis, a sigG transcript is also detected early, but this transcript is read-through from sigE, located immediately upstream of sigG, and is not translated 3536. Translation of sigG occurs when the gene is expressed as a single cistron from a σ^F-dependent promoter located between sigE and sigG 3536. In C. acetobutylicum, sigE and sigG are also located adjacent to each other, but a σ^F promoter was not predicted between the two genes 37. Thus, it was predicted that sigG is only expressed as part of the sigE operon (consisting of spoIIGA, the processing enzyme for σ^E, and sigE). Our transcriptional data seem to support this prediction because all three genes, spoIIGA, sigE, and sigG, have very similar transcriptional patterns (Figure 3f), suggesting they are expressed as a single transcript, like the spoIIAA-spoIIAB-sigF operon (Figure 3e). However, from Northern blots probing against sigE-sigG, three separate transcripts were seen: one for spoIIGA-sigE-sigG, one for spoIIGA-sigE, and one for sigG 29. Unfortunately, the current data cannot resolve this issue definitively, since the microarrays only detect if a transcript is present or not.

We also desired to estimate the activity profiles for the key sporulation factors (σ^H, Spo0A, σ^F, σ^E, and σ^G; Figure 4). We did so by averaging the expression profiles of known or robustly identifiable canonical genes of their regulons 1. To adjust for differences in relative expression levels, expression profiles were standardized before averaging 7. This is a surrogate reporter assay, which we believe is as accurate as most reporter assays. For a detailed discussion of the genes used to construct the plots, see Additional data file 4. For all of the plots (Figure 4), peak activity took place after peak expression, as expected. Of all the factors, σ^H activity peaked first, during early transitional phase, and this was followed by a decrease in activity until stationary phase, when activity increased again (Figure 4a,f). Spo0A~P activity was the next to peak, during late transitional phase, and stayed fairly constant throughout the rest of the timecourse (Figure 4b,f). σ^F activity had an initial induction during transitional phase, but then stayed constant until 24 hours (Figure 4c,f). After 24 hours, the activity increased again and stayed fairly constant at this higher activity level for the rest of the culture. σ^E activity increased slightly during late transitional phase, but its major increase occurred after 24 hours during mid-stationary phase (Figure 4d,f). Like the previous sigma factors, σ^G activity increased throughout early stationary phase and early mid-stationary phase, but the major increase occurred after hour 30 (Figure 4e,f). The activity of all of the factors, except for Spo0A and σ^F, decreased during late stationary phase at hour 38. σ^G activity began to increase slightly again at hour 48 but did not peak again. Considering only major peaks in activity, the Bacillus model of sporulation is generally true with the peaks progressing from σ^H to Spo0A~P to σ^F to σ^E and finally to σ^G (Figure 4f).

In B. subtilis, the sigma factors downstream of Spo0A (σ^F, σ^E, and σ^G) are all regulated by a complex network of interactions 1. We desired to examine if our transcriptional data could be used to do a first test to determine whether the mechanisms employed in the B. subtilis model are valid for C. acetobutylicum. In B. subtilis, σ^F is held inactive in the pre-divisional cell by the anti-σ^F factor SpoIIAB. σ^F is released when the anti-anti-σ^F factor SpoIIAA is dephosphorylated by SpoIIE, resulting in SpoIIAA binding to SpoIIAB, which then releases σ^F. In C. acetobutylicum, spoIIAB (CAC2307) and spoIIAA (CAC2308) are transcribed on the same operon as sigF (Figure 3e), but spoIIE (CAC3205) is transcribed separately. The initial increase in σ^F activity during the transitional phase was not accompanied by an increase in spoIIE expression, but the peak in σ^F activity did occur after spoIIE upregulation (Figure 4c). Despite the sustained level of σ^F activity, sigF and spoIIE decreased in expression, though spoIIE expression did increase slightly again after 48 hours (Figure 4c). In B. subtilis, the pro-σ^E translated from the sigE gene undergoes processing from SpoIIGA, which must interact with SpoIIR in order to accomplish the σ^E activation. In C. acetobutylicum, SpoIIGA (CAC1694) is transcribed on the same operon as sigE (Figure 3f), and SpoIIR is coded by CAC2898. σ^E activity increased with the induction of spoIIR (Figure 4d), suggesting a similar mechanism as in B. subtilis. Finally, σ^G activation in B. subtilis is dependent upon the eight genes within the spoIIIA operon. Here, the second and larger increase in σ^G activity followed peak expression of the spoIIIA operon, but the early increase in σ^G activity was not characterized by a large induction of spoIIIA expression (Figure 4e). We tentatively conclude that the B. subtilis processing and activation model does generally hold true in C. acetobutylicum, but further investigation is needed to determine the exact timing and interaction of the various factors and their activators.

In B. subtilis, σ^K is formed by splicing together two genes (spoIVCB and spoIIIC), both under the control of σ^E and SpoIIID 38, separated by a skin element 39. In contrast, a single gene encoding σ^K has been annotated in C. acetobutylicum 13. The gene was initially identified using a PCR-approach 40 and was later detected by primer extension in a phosphate-limited, continuous culture of C. acetobutylicum DSM 1731 41. spoIIID, which controls sigK expression with σ^E in B. subtilis, reached peak expression at hour 30, which is consistent with it being under σ^E control (Figure 3d) 42. However, at no timepoint in this study did sigK exceed the cutoff expression criterion. Q-RT-PCR also showed a significantly lower sigK induction compared to the other sigma factors and suggests the transcript, if expressed, is at much lower levels than any other gene analyzed (Figure 2). The putative main σ^K processing enzyme, SpoIVFB (CAC1253), also did not exceed the cutoff criterion. To help determine if there is an active σ^K, we investigated two genes controlled by σ^K in B. subtilis. yabG (CAC2905), which encodes a protein involved in spore coat assembly, was upregulated mid-stationary phase and peaked at hour 30 (Figure 3d), and spsF (CAC2190), involved in spore coat synthesis, was not upregulated until late stationary phase, at hour 38 (Figure 3d). From these two genes, it is difficult to determine whether a functional sigK gene exists or not. Clearly they are both transcribed, but based on its expression pattern, yabG could fall under the control of σ^E instead of σ^K. spsF upregulation is late enough to possibly indicate σ^K regulation though. Ideally, more genes need to be investigated to draw firmer conclusions, but because few σ^K regulon homologs exist in C. acetobutylicum, we cannot currently determine if there is σ^K activity or not.

As discussed, phosphorylated Spo0A is responsible for initiating sporulation in both bacilli and clostridia along with solvent formation in C. acetobutylicum. In bacilli, Spo0A is phosphorylated via a multi-component phosphorelay 43, initiated by five orphan histidine kinases, KinA-E (kinases that lack an adjacent response regulator); this phosphorelay system is absent in all sequenced clostridia 1. Alternatively, Spo0A in clostridia may be directly phosphorylated by a histidine kinase, orphan or not, as was hypothesized in 17. This alternative was demonstrated in C. botulinum, where the orphan kinase CBO1120 was able to phosphorylate Spo0A 44. In C. acetobutylicum, five true orphan kinases have been identified with a sixth orphan, CAC2220, identified as CheA, which has a known response regulator 1.

A kinase that could directly phosphorylate Spo0A is expected to have a peak in expression before or during the activation of Spo0A, as the orphan kinases in B. subtilis do 454647. As a measure of Spo0A activity, the expression of the sol operon (CAP0162-64) was used, as before 7, because it is induced by Spo0A~P. The initial induction of the sol operon, almost 100-fold, occured at hour 10 (before spo0A reached it maximum expression), with detectable levels of butanol appearing before the second induction of the sol operon. This second induction, of another 10-fold, followed the peak in spo0A expression (Figure 5a). It is clear that some level of phosphorylated Spo0A exists at 10 hours; therefore, kinase candidates must display an increase in expression before 10 hours. Of the five orphan kinases (Figure 5b,c), CAC2730 displayed the earliest peak followed by CAC0437, CAC0903, and CAC3319. CAC0323 never displayed a prominent peak in expression either before or after sol operon induction (Figure 5b) and likely does not play a role in phosphorylating Spo0A. Of the remaining four, CAC0437 and CAC2730 peaked only once before the initial sol operon induction, while CAC0903 peaked before each induction of the sol operon (Figure 5b,c). CAC3319 expression slightly mirrored that of the sol operon, with an increase before initial induction followed by a plateau, and an increase in expression again until it peaked just after the sol operon peaked (Figure 5c). The proteins encoded by CAC0437 and CA0903 displayed the most similarity to the protein encoded by CBO1120, the orphan kinase in C. botulinum shown to phosphorylate Spo0A 44.

Though primarily interested in orphan kinases because of the similarity to the B. subtilis model, a two-component response system could also be responsible for the phosphorylation of Spo0A. The remaining 30 annotated histidine kinases were also investigated to determine if any displayed a peak in expression before the initial induction of the sol operon (Additional data file 5). Six kinases (Figure 5d,e) were found to have a peak in expression at 8 hours. CAC0290 and CAC3430 subsequently decreased in expression while CAC0225 and CAC0863 maintained expression at initial levels. Despite a dip in expression at hour 9, CAC1582 maintained an increased expression level from 8 hours on. CAC2434 peaked at hour 8, dropped back to initial levels, but then steadily increased with the second induction of the sol operon.

Loss of pSOL1 impairs sporulation at the level of spo0A expression 748, thus generating increased interest for sigma factors located on the pSOL1 plasmid as these may play a role in the regulation of sporulation. Two sigma factors, CAP0157 and CAP0167, are located on pSOL1 and are annotated as 'special sigma factor (σ^F/σ^E/σ^G family)' and 'specialized sigma factor (σ^F/σ^E family)', respectively. It was predicted that CAP0167 is putatively co-transcribed with CAP0166 from a promoter of the σ^F/σ^G family 37 and it displayed an expression pattern similar to that of spo0A, consistent with the computational prediction of an 0A box 29 and two reverse 0A boxes in its promoter region (Figure 6a). CAP0157 was expressed from an unidentified promoter late in the timecourse (40+ hours) and thus may be involved in late-stage sporulation, despite its low level of expression at hour 20 (Figure 6a). CAC3267, putatively the fourth gene in an operon starting with CAC3270 and ending with CAC3264 37, was mainly expressed during early exponential growth (Figure 6a), then decreased, and peaked again around 14 hours, after which expression decreased again. This pattern of expression suggests that it plays a role in vegetative growth and possibly early sporulation. CAC0550, putatively transcribed from a σ^A promoter as a single cistron 37, was mainly transcribed early with its expression ending after 20-24 hours (Figure 6b), suggesting that it is not involved in sporulation. CAC1766, expressed from an unknown promoter, displayed a unique pattern with a progressive buildup starting around hours 8-12 and a distinct peak around hour 22 (Figure 6b). CAC2052 is annotated as 'DNA-dependent RNA polymerase σ-subunit' and was putatively expressed together with CAC2053, a hypothetical protein, from a σ^A and/or a σ^F/σ^G promoter 37. Our data suggest that it is unlikely to be transcribed from a σ^F/σ^G promoter without any other effectors, as their transcription peaked at hour 16, when there was very little (if any) σ^F or σ^G activity (Figure 6b).

To help determine a possible function for these sigma factors, a phylogenetic tree was constructed of σ⁷⁰ sigma factors from ten species, including B. subtilis and all sequenced clostridial species. The resulting tree (Additional data file 6) contains eleven major branches, and of these, seven can be definitively classified based on known sigma factors within the branch. These categories are extracytoplasmic function (ECF), sporulation factors (sigF, sigE, and sigG), sigH, sigA (a basal sigma factor), sigD (regulates chemotaxis and motility), and sigB (a general response sigma factor). Two factors, CAC3267 and CAC1766, fell within ECF branches. CAC3267 fell within an ECF branch close to the B. subtilis σ^V, a sigma factor of unknown function, and σ^M, a sigma factor essential for growth and survival in high salt concentrations. CAC1766 fell within a different ECF branch close to B. subtilis σ^Z, a sigma factor of unknown function, and CAC1509, a sigma factor expressed for less than eight consecutive timepoints. The remaining four factors fell within clusters with other clostridial sigma factors of unknown function, though several could have possible ECF function.

Of the six expressed sigma factors of unknown function, CAP0157, CAP0167, CAC2052, and CAC1766 were chosen for further study because the timing and shape of their expression patterns suggested potential involvement in sporulation and/or solventogenesis. Since the two processes are coupled, phenotypic changes in differentiation may affect solvent production, as has been previously observed 46293349. Antisense RNA (asRNA) knock-down was chosen over knocking out the genes, because knockouts are still extremely difficult to produce in this and all other clostridia. Indeed, to date, only a handful of knockouts have been created 2950515253, and these have only been achieved after screening thousands of transformants 515253. Recently, a group II intron system has been developed for clostridia 54, but this system was not yet available when these experiments were carried out. In contrast, asRNA is relatively quick, has been shown to reduce gene expression by up to 90% 335556 and has been used to knock-down a large number of genes with a high level of specificity 33495556575859. asRNA constructs (see Additional data file 7 for specific sequences used) were designed against CAP0157, CAP0167, CAC2052, and CAC1766 along with CAC2053 and CAP0166, the first genes in the operons predicted to contain CAC2052 and CAP0167, respectively 37. Cultures of these strains were examined and compared against the wild type (WT) and plasmid control strain 824(pSOS95del) for cell morphology differences and metabolic changes.

Microscopy results from the asRNA-strain cultures revealed both novel morphologies and apparently altered differentiation (Figure 6d). Most notable were changes in strains asCAP0166, asCAP0167 and asCAC1766. Typical WT cultures display a predominately vegetative, symmetrically dividing population through 72 hours as evidenced by the thin, rod-shaped, phase dark cells (Figure 6d, I). By 72 hours, WT cultures exhibited only a small percentage of swollen, cigar-shaped clostridial forms and then a proportional population of free spores by 96 hours.

pSOS95del cultures exhibited clostridial forms by 48 hours, suggesting an accelerated differentiation compared to WT, as has been seen before in our laboratory (Figure 6d, II). Moreover, a greater percentage of clostridial forms and free spores compared to WT were observed at 72 and 96 hours, respectively. asCAP0166 cultures generated a large percentage of clostridial forms and endospores/free spores by hours 48 and 72, respectively (Figure 6d, III). This differentiation is accelerated in comparison to pSOS95del. By hour 96, asCAP0166 cultures exhibited predominately vegetative cells apparently derived from germinated spores (data not shown). asCAP0167 cultures also exhibited accelerated differentiation and displayed a novel (to our knowledge) form of cellular morphology that was most profoundly observable at 72 hours (Figure 6d, IV). This novel morphology has qualities of an excessively swollen clostridial cigar-form (which makes them look much shorter than normal clostridial forms), with what appears to be endospore formation occurring, but without the associated phase bright characteristics seen in the 72 hour asCAP0166 cultures. The asCAP0166 culture displayed cells in this novel morphological state as well, but to a lesser extent, although it is possible that because of its faster sporulation, such cell forms appeared prior to 72 hours. The asCAC1766 cultures also exhibited altered differentiation; most importantly, at 72 hours the majority of the cells exhibited a very swollen clostridial-form morphology similar to that in the asCAP0167 cultures at 72 hours, but slightly more elongated (Figure 6d, V).

To further characterize this novel cell form, transmission electron microscopy (TEM) and scanning electron microscopy images of cells were taken for strains asCAP0167 and asCAC1766. To determine morphological differences involved in differentiation, the TEM images were compared against cell images taken from the plasmid control strain (Figure 7). For both asRNA strains, the very swollen cell forms observed can be documented as approximately 2.5-4 μm long, and 1.1-1.3 μm in diameter, and should be compared to control or WT swollen clostridial forms, which are 3.5-6 μm long and 0.8-1 μm in diameter. Forespore and endospore forms of both asCAP0167 (Figure 7c,d) and asCAC1766 (Figure 7e,f) displayed a pinched end not seen in the plasmid control (Figure 7b). A slight pinching is seen in the clostridial forms of the plasmid control strain (Figure 7a), but this is probably indicative that an asymmetric division is about to occur. Rather, the pinched ends seen in the antisense strains occur after asymmetric division and while the spore is developing within the mother cell. These pinched ends are also noticeable in the scanning electron microscopy images (Figure 8). Though granulose is distinguishable in most of the TEM images (Figure 7c,d,f), it is not the characteristic electron translucent seen in typical clostridial, forespore, and endospore forms (Figure 7a,b). These differences were seen throughout the culture and additional TEM images of both the plasmid control and the antisense strains are included in Additional data file 8.

Glucose, acetone, and butanol concentrations from two to four biological replicates for each strain were averaged together, and the results are shown in Table 1. We averaged data from cultures that displayed similar characteristics; most cultures did so despite the fact that each culture was inoculated from a different colony for each strain. Acetone and butanol levels were typical for WT and control cultures, with the WT producing 90 mM of acetone and 150 mM of butanol and the plasmid-control strain producing 80 mM of acetone and 160 mM of butanol 60. By 192 hours, all strains had either produced comparable amounts of butanol to the WT and the plasmid control strain or had somewhat outperformed these two strains. The most significant differences were that all asRNA strains consumed higher levels of glucose and also had a delayed metabolism in terms of product formation. These metabolic changes, although preliminary, are consistent with and support the large changes in the kinetics of sporulation observed by microscopy.