The industrial anaerobe Clostridium acetobutylicum uses polyketides to regulate butanol production and differentiation

Polyketides are an important class of bioactive small molecules valued not only for their diverse therapeutic applications, but also for their role in controlling interesting biological phenotypes in their producing organisms. While numerous examples of polyketides derived from aerobic organisms exist, only a single family of polyketides has been identified from anaerobic organisms. Here we uncover a novel family of polyketides native to the anaerobic bacterium Clostridium acetobutylicum, an organism well-known for its historical use as an industrial producer of the organic solvents acetone, butanol, and ethanol. Through mutational analysis and chemical complementation assays, we demonstrate that these polyketides influence butanol production and act as chemical triggers of sporulation and granulose accumulation in this strain. This study represents a significant addition to the body of work demonstrating the existence and importance of polyketides in anaerobes, and showcases a novel strategy of manipulating the secondary metabolism of an organism to improve traits significant for industrial applications.


INTRODUCTION
encoded elsewhere on the genome. To determine the identity of any PKS-associated metabolites and probe the broader function of the pks locus, we performed a targeted in-frame deletion of the pks gene (ca_c3355) using an allelic exchange method previously developed for C. acetobutylicum 15 . The resulting mutant (Δpks) was confirmed by PCR analysis (Supplementary Fig. 2).

Identification of three polyketide metabolites through untargeted, comparative metabolomics
Quadruplicate batch fermentations with wild-type C. acetobutylicum and the mutant Δpks were performed and harvested at early stationary phase. Organic extracts from combined supernatants and cell pellets were obtained and analyzed via liquid chromatography-high resolution mass spectroscopy (LC-HRMS). Following untargeted metabolomic comparisons of the two strains using XCMS 16 , three major species with molecular formulas C 14 O 12 -enabled function of the KR domain (and the subsequent DH domain), leading to the production of the known shunt products tetraketide pyrone 5 and pentaketide pyrone 6 (Supplementary Figs. 8 and 9). Similar derailment in the normal programmed steps to yield the same shunt products has been observed in other iterative type I PKSs such as LovB 28 and ApdA 29 . These results demonstrated that ca_c3355 encodes a highly reducing type I PKS that functions iteratively, the first of its kind identified from an anaerobic organism. Furthermore, these results indicated that additional biosynthetic enzymes are needed for generating 1 and 3, and their encoding genes are located elsewhere on the genome of C. acetobutylicum.

Impact of the pks gene on ABE fermentation
To determine whether polyketide production influenced ABE fermentation, we compared the batch fermentation performance of wild-type C. acetobutylicum and Δpks (Fig. 2a-c;   Supplementary Fig. 10). While both strains displayed similar growth curves and the expected acidogenic and solventogenic phases, Δpks showed stronger butanol production with ~10% increases in both butanol titer and productivity relative to wild-type ( Fig. 2a and b). Perhaps related to the difference in butanol production, the butyrate concentration profiles also differed for Δpks and wild-type; in the Δpks culture, butyrate was produced more rapidly and reached a higher concentration during acidogenesis (0-20 hours), and was re-assimilated earlier during the transition to solventogenesis (20-23 hours) (Fig. 2c). This was also reflected in the pH profiles of the two strains, with the fall and rise of the culture pH (corresponding to the changes in metabolism) occurring earlier for Δpks (Fig. 2a). Although less apparent, small increases in acetone and ethanol production were also observed for Δpks relative to wild-type (Fig. 2b).
These results established a clear link between the pks gene and ABE fermentation, and the production of polyketides in wild-type seemed to negatively affect the metabolic switch between the acidogenic and solventogenic phases.
To further probe the relationship between the production of the three polyketides and the ABE fermentation profile, we obtained production time-course profiles of 1-3 for wild-type C. acetobutylicum (Fig. 2d). Maximum levels of 1 (clostrienoic acid) and 2 were observed during the same period as maximum butyrate/acetate concentrations (20-23 hours), while the production of 3 (clostrienose) and butanol/acetone initiated at approximately the same time (~16 hours) and increased for the remainder of the fermentation. These results suggest a direct, although currently unclear, link between polyketide production and the ABE fermentation phases, with the production of 1 and 2 associating with acidogenesis, and the production of 3 associating with solventogenesis. We propose that 1 and 2 are biosynthetic intermediates of the end product 3, and feeding studies with 1 and 2 showed that these compounds were readily converted to 3 in cultures of C. acetobutylicum Δpks.

Polyketides affect sporulation, granulose accumulation, and colony morphology
To better understand the broader biological impacts of polyketide production on C. acetobutylicum metabolism and physiology, we performed a transcriptome comparison of Δpks and wild-type strains using RNA-Seq. Analysis of samples taken at early stationary phase revealed a total of 392 genes that were differentially expressed (expression fold change > 2.0, pvalue < 0.003) between Δpks and wild-type, with 282 genes downregulated and 110 genes upregulated in Δpks. STRING network analysis 30 showed that the expression of genes related to four major cellular processes was downregulated in Δpks, including sporulation (33 genes), carbohydrate transport (21 genes), carbohydrate metabolism (33 genes), and carbohydrate transfer (11 genes), and the expression of genes related to four other major cellular processes was upregulated in Δpks, including sulfur metabolism (12 genes), cofactor biosynthesis (6 genes), stress response (8 genes), and amino acid transport (8 genes) ( Fig. 3; Supplementary Table 3).
It is notable that no significant difference in expression of the key solventogenic genes (including the crucial sol operon 31 ) was observed, suggesting that the difference in butanol production between Δpks and wild-type was not due to the direct transcriptional regulation of solventogenic genes by the polyketides. Consistent with the higher solvent production observed for Δpks, the major pathways upregulated in Δpks (particularly class I, III, and IV heat shock response machinery) have previously been associated with improved solvent tolerance in C. acetobutylicum 14,32,33 .
The sporulation genes downregulated in Δpks include those encoding late stage sporulation proteins such as spore coat and germination proteins, and the sporulation-specific sigma factor K (σ K ), one of the core regulators of sporulation in C. acetobutylicum 34 . Notably, transcription of the gene encoding the well-known master regulator of sporulation and solvent production, Spo0A 35 , was not significantly affected in Δpks. To determine whether reduced sporulation was an observable property of Δpks as suggested by the RNA-Seq analysis, sporulation assays in both liquid and solid media were performed for wild-type C. acetobutylicum and Δpks. Indeed, a significant decrease in sporulation was observed for Δpks, with spore formation decreasing by 3-4 orders of magnitude in liquid culture (Fig. 4a), and by ~2 orders of magnitude on solid media (Supplementary Fig. 11). Furthermore, the level of sporulation for Δpks was partially restored when Δpks culture broth was supplemented with clostrienose (3) (Fig. 4a). Since granulose biosynthesis and accumulation is related to the sporulation cycle, we then performed standard granulose accumulation assays using iodine staining. Following the trend observed for sporulation, a significant decrease in granulose accumulation was observed for Δpks, and the addition of 3 to Δpks culture completely restored granulose accumulation (even appearing to exceed levels observed for wild-type) (Fig. 4b).
These assays revealed that the polyketides are important, although not essential, for triggering both sporulation and granulose accumulation in wild-type C. acetobutylicum. Interestingly, pks inactivation also affected colony morphology of C. acetobutylicum. While wild-type colonies were relatively flat in elevation and featured a distinctive "spore center" in the middle of the colony, Δpks colonies were distinctively raised in elevation and featured a highly textured surface with no distinguishable spore center (Fig. 4c).

DISCUSSION
Clostridium is one of the largest bacterial genera, ranking second in size only to Streptomyces 36 .
While members of Streptomyces are known to be prolific producers of secondary metabolites 37 , only a handful of secondary metabolites have been discovered from Clostridium. However, recent genomic analysis has indicated that secondary metabolite gene clusters can be found among diverse members of this genus, prompting efforts to identify and characterize these 'cryptic' secondary metabolites 6 . In this study, we identified a suite of polyketides (clostrienoic acid and clostrienose) from C. acetobutylicum, a well-studied solvent-producing anaerobe with no previously associated natural products.
Clostrienoic acid and clostrienose are biosynthesized by a predicted type I single-module PKS. Through in vitro reconstitution of the activity of purified PKS, we demonstrated that this megasynthase functions as a highly reducing iterative type I PKS, the first known of its kind among anaerobic organisms. We propose that the PKS functions iteratively to generate a heptaketide intermediate, which is then modified by tailoring enzymes encoded elsewhere on the genome to yield clostrienoic acid. Two subsequent glycosylation events are proposed to install first the rhamnopyranoside group, followed by the galactofuranoside group (generating clostrienose). The proposed sequential biosynthesis of the three polyketides is consistent with the production timing of 1-3 observed in batch fermentation of C. acetobutylicum. Although galactofuranose has previously been detected in cultures of C. acetobutylicum 38 42 , few signaling small molecules have been reported from Clostridium. In addition to clostrubin which helps the pathogenic C. puniceum access aerobic territory 9 , a putative AI-2-like system was reported to be involved in stimulating toxin production in C. perfringens 43 . Furthermore, gene clusters with homology to the well-studied agr quorum sensing system have been shown to be important for cellular processes such as toxin production [44][45][46][47] , biofilm formation 44 , and sporulation 47,48 in several Clostridium species, although none of these studies identified the molecular structure of the presumed bioactive small molecules. This work provides solid evidence that morphological differentiation and solventogenesis can be regulated by secondary metabolites in Clostridium, and has revealed the molecular identity of the signal responsible for this behavior.
Solventogenic strains of Clostridium (such as C. acetobutylicum) were employed for industrial solvent production as early as 1916, and were eventually the source of 66% of US butanol production in 1945 10 . Although industrial operations of this process largely ceased during the 1950s due to the growth of the petrochemical industry, ABE fermentation has recently gained renewed interest given the wide range of agricultural feedstocks which can be converted to various commodity chemicals and potential biofuels using this process 49 . However, some drawbacks still exist that prevent the widespread use of ABE fermentation, such as low solvent titer/productivity due to solvent toxicity, and unfavorable cellular differentiation 50 . Rather than pursuing a traditional metabolic engineering strategy that focuses on the core metabolic pathway for solvent production, our work showcases an alternative approach by manipulating the secondary metabolism of the organism to improve traits significant for industrial ABE fermentation performance. In particular, given that both granulose biosynthesis and sporulation are undesirable traits for industrial fermentation (as granulose accumulation results in reduced solvent yields, and metabolically inactive spores do not contribute to solvent production) 50 , the reduced granulose accumulation and sporulation associated with Δpks represent improved industrial traits. Furthermore, although none of the solvent producing genes were upregulated in Δpks relative to wild-type, both butanol titer and productivity were increased in Δpks. This may be explained by the decreased commitment of cells to sporulation in Δpks (yielding a higher proportion of cells capable of solvent production), as well as the upregulation of cellular machinery related to butanol stress and adaptation as indicated by transcriptomic analysis.
In summary, we have discovered a family of novel polyketides that are biosynthesized by a highly reducing iterative type I PKS in C. acetobutylicum ATCC 824. In addition to the type II PKS-derived clostrubin, our work provides the second example of polyketide metabolites from a strictly anaerobic bacterium, and encourages continued efforts in exploring the uncharted terrain of secondary metabolites in the anaerobic world. We have further shown that the newly identified polyketides affect solvent production and are important for stimulating sporulation and granulose accumulation in C. acetobutylicum, adding to the extremely limited inventory of known signaling molecules used by Clostridium to control cellular physiology and metabolism.
Furthermore, this work has yielded an engineered strain of C. acetobutylicum with improved traits for industrial ABE fermentation, demonstrating a novel strategy of manipulating secondary metabolism as a means of improving this important renewable bioprocess.

Competing financial interests
The authors declare no competing financial interests.

Additional information
Supplementary information and chemical compound information are available in the online version of the paper. Reprints and permissions information is available online at http://www.nature.com/reprints/index.html. Correspondence and requests for materials should be addressed to W. Z. (wjzhang@berkeley.edu).
Plasmid construction. Oligonucleotides were synthesized by Integrated DNA Technologies. Supplementary Table 4 lists all relevant oligonucleotide sequences used. Phusion Polymerase (NEB) was used for all PCR amplifications. Genomic DNA isolation of C. acetobutylicum ATCC 824 was performed using a modified alkaline lysis method as previously reported 55 .
For constructing plasmid pKO_mazF_mod (which would later serve as a template for pKO_pks), primers pKON_Fo & pKON_Ro were used to PCR amplify a 5.0 kb region from the pKO_mazF template 15 . This step was necessary to remove a 677 bp region from the pKO_mazF backbone which we were unable to amplify via PCR. The 5.0 kb PCR product was gel purified, digested at the 5' and 3' ends using PshAI, ligated to form pKO_mazF_mod, and transformed into E. coli TOP10. Transformant clones were screened by purified plasmid test digestion, and Sanger sequencing was used to confirm the sequence of the final pKO_mazF_mod clone. For constructing plasmid pKO_pks (for deletion of the pks gene from C. acetobutylicum), a 3.8 kb region containing colE1, repL, bgaR, and mazF were PCR amplified from pKO_mazF_mod using primers pKO_F & pKO_R. For constructing plasmid pET24b_pks, the 5.4 kb C. acetobutylicum ATCC 824 pks gene (CA_C3355) was amplified from C. acetobutylicum genomic DNA using primers PKS_Fo and PKS_Ro. Doubly digested vector pET24b (EcoRI/XhoI digested) was ligated with doubly digested pks PCR product (EcoRI/XhoI digested), and the ligation product was transformed into E. coli TOP10. Transformant clones were screened by purified plasmid test digestion, and Sanger sequencing was used to confirm the sequence of the final pET24b_pks clone.

Electro-transformation of C. acetobutylicum.
Prior to transformation into C. acetobutylicum, vector pKO_pks was cotransformed with pAN3 into E. coli TOP10 via electroporation. This procedure permitted methylation of pKO_pks necessary for overcoming the native restriction-modification system active in C. acetobutylicum 56 . Plasmid purification of pKO_pks/pAN3 liquid culture was performed, and the resulting plasmid mixture was used for electroporations of C. acetobutylicum.
Electroporations of C. acetobutylicum were performed as previously described 56 , with the exception that all transformation cultures were started from heat shocked (80°C for 10 minutes) single colonies.
Deletion of pks gene (ca_c3355) and generation of genetic complementation strain. Targeted KO of the pks gene in C. acetobutylicum ATCC 824 was achieved using the previously published method 15 . Briefly, 5 µg of methylated pKO_pks/pAN3 plasmid mixture was transformed into C. acetobutylicum using the method referenced above. Following recovery in liquid 2xYTG medium for four hours, the cell pellets were collected by centrifugation, resuspended in 0.5 mL of fresh liquid 2xYTG, and 100 µL of the resuspended cell culture was plated on solid 2xYTG + 5 µg/mL Th + 40 mM β-lactose plates. Under these plating conditions, only cells which have undergone the desired double crossover homologous recombination event are expected to survive. Counterselection of the vector backbone is provided by the lactose-inducible promoter (P bgaL ) which drives the toxin gene mazF on the pKO_pks vector backbone. Following this plating procedure, roughly 10 colonies were observed on the 2xYTG + 5 µg/mL Th + 40 mM β-lactose plates. Of these 10 colonies, four were twice restreaked and subjected to colony PCR verification. Four sets of primers were used as the basis of colony PCR verification, as detailed in Supplementary   Fig. 2.

LC-HRMS metabolomic analysis.
For untargeted metabolomic comparisons of wild-type and ∆pks C. acetobutylicum, single colonies of each strain were heat shocked at 80°C for 10 min and used to inoculate 10 mL of liquid CGM (30 g/L glucose).
Following overnight incubation (stagnant) until reaching OD 600 ~ 1.0, these cultures were used to inoculate 10 mL of liquid CGM After combining the pooled fermentation culture (34 L) and removing the cell pellets via centrifugation, the cell-free culture supernatant was extracted using two volumes of ethyl acetate. Following isolation of the organic extract, the solvent was removed by rotary evaporation and the residue was redissolved in dichloromethane. The dark yellow oily residue (10.8 g) was subjected to silica gel column chromatography (60 Å, 220-440 mesh), and the compound-rich fractions were eluted with an ethyl acetate/hexane gradient system. Each of the compound-rich fractions were combined and concentrated to dryness (3.6 g of To determine the absolute stereochemistry of the secondary alcohol in 1, an additional 15 mg of 1 was purified and dissolved in 10 ml of absolute EtOH, and 0.75 mg of 10% Pd/C was added. The flask was purged with hydrogen, and the reaction was stirred at room temperature for 3 hours under positive pressure of hydrogen. Reaction progress was monitored by TLC. The  (d) Production of compounds 1-3 from batch fermentation of wild-type C. acetobutylicum ATCC 824. As indicated, the green region represents the acid-production phase of the fermentation (acidogenesis), while the blue region represents the solvent-production phase (solventogenesis). Error bars represent the standard deviation of values from duplicate fermentations. Experiments were repeated at least three times independently.

Figure 3 | RNA-Seq comparison of wild-type C. acetobutylicum and
Δpks. STRING network analysis of genes predicted to be transcriptionally downregulated (left) and upregulated (right) by RNA-Seq in Δpks fermentation culture relative to wild-type. Nodes represent differentially expressed genes, and lines represent predicted connections between genes including shared functional pathways of encoded proteins, chromosomal proximity, co-occurrence of genes in other organisms, co-expression of genes, and protein homology. For clarity, only genes with at least one predicted connection are shown. Four major concentrations of connected nodes were observed for each group (indicated by a-h), signifying the cellular pathways most affected by deletion of the pks gene. Detailed results from the RNA-Seq analysis are presented in Supplementary Table  3.