As subculturing has been employed to promote degeneration of solventogenic clostridia in the past (Kutzenok and Aschner, 1952; Hartmanis et al., 1986; Hayashida and Yoshino, 1989; Chen and Blaschek, 1999), we established our own subculturing protocol using wild-type (WT) C. beijerinckii NCIMB 8052, to promote degeneration of this strain and reproduce previously reported findings (see Supplementary Figure 1 for schematic). This protocol involved repeatedly subculturing WT C. beijerinckii at 24-h intervals, in a 1 in 10 dilution. Each subculture series was initially inoculated with a colony derived from a single heat-shock germinated spore to generate a genetically homogeneous population and remove any cells that had not formed functional spores (and therefore had already degenerated). We found that at each successive subculture, a consecutive decrease in the concentration of solvents (acetone, butanol and ethanol) produced by the culture was observed. This was mirrored by a steady increase in the production of organic acids (acetate and butyrate; Figure 1A). We found that after 15 subcultures, the total concentration of solvents had decreased 80% from a maximum of 7.23 ± 0.3 g/L at subculture 1 to 1.44 ± 0.26 g/L at subculture 15. Conversely in the same time frame, organic acids increased by 195% from 1.95 ± 0.22 g/L to 5.76 ± 0.32 g/L. Alongside a successive decrease in solvents, the number of viable heat resistant spores (i.e., functionally germinating spores) produced by the culture also decreased as the subculturing progressed, correlating with the decrease in solvents (Figure 1B). It should be noted that the lower solvent concentrations found at subculture 0 reflect the fact that each replicate tube is inoculated with a single colony compared to a larger inoculum from a full-grown culture which will affect the final solvents produced.

Much of the early studies into degeneration commented on aberrant colony morphologies emerging that would appear misshapen when compared to the WT morphology (Adler and Crow, 1987; Kashket and Cao, 1995). In the aforementioned subculturing experiment (Figure 1), we observed the colony morphologies of subcultured C. beijerinckii to assess whether colonies had undergone any morphological changes. We found four distinct colony morphologies emerged in the population which we named the round and dark (RD; Figure 1C), which is the WT morphology, the dark center with outgrowths (DCOG; Figure 1D), the caved in center (CIC; Figure 1E) and the flat and white (FW; Figure 1F). With four colony types established, we monitored their appearance throughout the subculturing process, which showed dynamic changes in their percentage of the whole culture (Figure 1G). In the early subcultures, the amount of observed RD within the population rapidly decreased with the advent of the DCOG at subculture 1. This type began to take over the population occupying around 70% of the observed CFU/ml by subculture 5. Although the DCOG quickly enriched within the culture, on average this increase was not maintained past the subculture 5 point as we observed, though less rapid, an incremental increase in the numbers of FW also from subculture 1. This type eventually outcompeted all others and by the end of the subculturing experiment (15 subcultures), occupied over 80% of the observed CFU/ml. We did on occasion for some replicates observe a sudden increase in the CIC at subculture 8, which appeared to promote DCOG competitiveness, however CIC CFU/ml was not maintained, and numbers of CIC decreased after this point. At the end of the subculturing, the final population consisted mostly of FW, with some DCOG still present. Neither the CIC nor the RD remained at this point, suggesting that the WT colony morphology was below the limit of detection.

Once we had established that four colony morphotypes existed through subculturing, we isolated a total of 17 RD, 17 DCOG, 18 CIC, and 19 FW in two separate rounds of isolation. Both rounds of isolation began with a single, WT C. beijerinckii colony that was subcultured 5 times. Single colonies were generated from germinated spores which were used to not only remove any variants unable to produce viable spores but also to provide the starting genomic profile which any subsequent mutations could be distinguished from. These single germinated spore colonies were then serially transferred five times and cultures were plated on CBMS agar to reveal the different colony morphotypes for isolation. We then phenotypically characterized each of the 71 variants based on their fermentation profiles (after 2 days of growth) and level of sporulation (after 5 days of growth) to ascertain the level, if any, of degeneration each variant had undergone (Supplementary Figures 2, 3). We found that generally, although there were some outliers, the RD, DCOG, and CIC fermentation profiles were similar as variants of these types produced solvents to a comparable degree to the WT (Supplementary Figure 2). The FW morphotype however showed the most drastic solvent phenotypes as only four of this type were able to produce comparable solvents to the wild type and the rest produced little to none (Supplementary Figure 2). We found however, far more homogeneity between morphotypes regarding their sporulation profiles as all variants, besides one RD, produced significantly less spores than the wild type even if their solvent profiles appeared normal (Supplementary Figures 2, 3). Again, the FW showed the lowest production of viable spores with only four variants capable of producing, albeit to a lower degree, viable spores (Supplementary Figure 3D). The profiles of each of the variants appeared to show that although solvent production and sporulation are often linked, normal levels of solvents can still be achieved for some mutants regardless if sporulation is impaired.

After phenotypic characterization of the variants, we genotypically characterized each variant by comparing their genome sequences to the parental WT they were originally derived from. Any mutations found within the parental WT were removed from mutations found in the variants due to their existence before subculturing. Genetic analysis revealed a total of 211 SNPs (single nucleotide polymorphism), 12 deletions and 2 insertions after five serial transfers (equating to approximately 16–17 generations starting from the subculture 0 population) for the 71 variants sequenced. The average number of mutations per degenerate was therefore 3.16 after the five serial transfers. The transversions G- > T and C- > A were over-represented as they made up 87% of the observed SNPs. A full table of the mutations found for each variant can be found in Supplementary Table 1 and provides a list of candidate genes for not only degeneration, but key regulatory elements in C. beijerinckii NCIMB 8052.

Four distinct “hot spot” regions emerged that contained significantly more mutations than elsewhere in the genome. These regions included a two-component sensor histidine kinase (Cbei_0017; hotspot region 1), the master regulator gene spo0A (Cbei_1712; hotspot region 2), a hybrid sensor histidine kinase/response regulator (Cbei_3078; hotspot region 3) and a final region that contained a hypothetical protein-encoding gene (Cbei_4884), a neighboring abrB (Cbei_4885) transcriptional regulator gene, as well as the intergenic region in between (hotspot region 4). To confirm that the mutations were present in our variants and not an artifact of our genome sequence analysis, we chose five random mutants from each hotspot region to PCR amplify chromosomal DNA at each hotspot region. All PCR product bands were subjected to Sanger sequencing which confirmed that the specific mutations were all present. Figure 2A shows the distribution of the mutations throughout the circular C. beijerinckii chromosome. Furthermore, the position of the mutations in the hotspots relative to the position in the gene and protein can be found in Supplementary Figure 4.

There was some link between colony morphotype and genotype. Out of 9 mutations found in Cbei_0017, 7/9 of those were of the RD morphotype. For 21 unique variants that had mutations in Cbei_3078, over half (12/21) were of the DCOG morphotypes and 12 out of the 16 spo0A mutants were of the FW morphotype. The CIC morphotype appeared to represent a less defined link between colony morphology and genotype as variants of this type had mutations in all the hotspots (although not all at once) and so may well embody a morphotype that is less strictly associated with mutations in individual genes.

We hypothesized that the two identified histidine kinases (Cbei_0017 and Cbei_3078) may be involved in Spo0A regulation in C. beijerinckii NCIMB 8052, as in the closely related species, C. acetobutylicum ATCC 824, direct phosphorylation by orphan histidine kinases has been shown to regulate Spo0A activity (Steiner et al., 2011). ClosTron (Heap et al., 2007, 2010) intron insertions were employed to functionally inactivate Cbei_0017, Cbei_3078 and as a control, spo0A, generating the mutants Cbei_0017::CTermB, Cbei_3078::CTermB and spo0A::CTermB, respectively. Furthermore, we subsequently complemented any mutants by inserting a functional copy, including the native promoter, of the inactivated gene back on the chromosome at the pyrE locus (Ng et al., 2013; Ehsaan et al., 2016) to confirm their role. As expected, we found that the spo0A intron insertions completely removed solvent and spore formation (Figure 2B). In contrast, the Cbei_0017 and Cbei_3078 insertion mutants were still capable of producing solvents at a similar concentration to the WT but did produce reduced spore counts (Figures 2B,C). Complementation by inserting intact copies of the genes at the pyrE locus on the chromosome restored lost phenotypes due to the ClosTron insertions (Figures 2B,C).

As with the naturally derived mutants, we observed an interesting link between colony morphology and genotype as the ClosTron mutants showed a remarkable similarity to the previously discovered colony morphologies. Cbei_0017::CTermB mutants appeared as the RD, and complementation of Cbei_0017 had no effect on morphology (Figures 2D,E). Cbei_3078::CTermB mutants however appeared like the DCOG morphotype, and complementation back onto the chromosome restored WT colony appearance (Figures 2F,G). spo0A::CTermB mutants appeared as the FW morphotype and again, complementation of the spo0A back onto the chromosome restored the colony morphotype back to normal (Figures 2H,I). These results further illustrated the role that these genes play for both physiological characteristics but also on colony morphology. The observed findings here support what we found for the subculture experiments as losing Cbei_0017 and Cbei_3078 activity had little effect on solventogenesis but severe effects on sporulation.

Ultra-deep sequencing was employed to track mutations within the degenerated populations to both understand the genetic dynamics of the subculturing process and to observe if and when our discovered hotspots gained or lost mutations within the whole culture. With this technology, it was possible to detect specific mutations when present within 4% or more of the population. Subculturing was again repeated in 5 independent parallel replicates and whole culture genomes, solvents, acids and spore samples were taken at each subculture point to provide an insight into both the genetic and phenotypic make-up of the culture at specific subcultures. Raw allele frequency (AF) numbers as determined for each subculture can be found in Supplementary Table 4 and were used to generate Figure 3.

Although mutations in hotspot regions 3 (Cbei_3078 operon) and 4 (the hypothetical protein and abrB region) were found, frequencies in these regions were less, relative to the colony sequence comparisons, as the frequency of mutations in these regions never reached above 50% (Figure 3.). This may in part be due to an overrepresentation of mutations in these regions at the colony sequencing stage when we initially selected colonies based on morphotype, rather than at random. Mutations of spo0A were found in significant frequencies and their appearance generally coincided with a decline in solvent formation by the cultures (Figure 3). Interestingly, no mutations in hotspot region 1 (Cbei_0017) were found, indicating that as with regions 3 and 4 it may have been overrepresented due to the morphotype selection bias during the single variant selection. It is clear however that the spo0A mutations dominated the subculturing process, and that the advent of these mutations appears to outcompete other mutations within the culture genomic pool. This is evident by the dynamic changes seen for some of the hotspot mutations that are lost by the end of the subculturing process, such as those seen for region 3 in replicate 2 (a similar trend as seen when observing the colony morphotypes). Interestingly, replicate 2 showed multiple spo0A mutations which we suspect all contributed to a loss in solvent formation, compared to a dominant single mutation seen for the other replicates. Elevated frequencies of mutations found in regions outside of the determined hotspots appear as carrier mutations that we suspect are found within the same mutants of spo0A, which are only enriched to high numbers due to increases in spo0A mutants.

Most of the mutations found were within genes, however, several mutations appeared to affect promoter or transcription factor binding regions (Figure 3). One notable promoter region mutation was found upstream of the abrB gene (Cbei_4885) at position 5,720,650 and 5,720,653. Mutations at position 5,720,653 have been reported previously in a variant of C. beijerinckii (BA105; Seo et al., 2017). This strain showed degeneration phenotypes as it made significantly more acids and less solvents compared to the WT. Sequence analysis of this region by Seo et al. (2017) revealed a putative 0A box with the sequence TGTCGAA (Ravagnani et al., 2000). The emergence of mutants in these regions suggests that not only mutations within genes, but mutations in promoter/0A box regions contribute to the emergence of degenerates.

To confirm that the mutations found from the ultra-deep sequencing were indeed dominating within the evolved populations, we took 10 random colonies from the final subculture for each of the 5 replicate cultures and whole genome sequenced these colonies to determine the mutations present at the end of the experiment. A full list of all the mutations for these colonies can be found in Supplementary Table 2. Indeed, we found that many of the individual colonies isolated from the terminal subculture of each replicate carried the same mutations as identified by ultra-deep sequencing, thus confirming the validity of our approach.

As we found that Spo0A, and the network that we suspect to regulate this protein, seem to play a key role in degeneration, we wanted to test whether loss of Spo0A activity gave a measurable fitness advantage. To do this we initially took a naturally derived spo0A mutant (named FW7) which contained two SNPs, one being a silent mutation within Cbei_2,920 and another within the specific DNA binding motif encoding part of spo0A⁷, rendering the protein nearly inactive as this strain produced less than 1 mM solvents and no spores (Supplementary Figures 2, 3). We subsequently mixed FW7 and the WT in various starting ratios, observing the CFU/ml of each at 0, 12, and 24 h after mixing. The relative fitness of FW7 to the WT was calculated based on Diggle et al. (2007; see section Materials and methods) with values above 1 indicating that the mutant is fitter than the WT. FW7 showed a high relative fitness, compared to the WT, when it occupied between 1% and 60% of the initial culture (Figure 4A). Relative fitness values were highest when the proportion of FW7 in the initial population was the least, and as the proportion of FW7 was increased, relative fitness values began to decrease for both the 12 and 24 h values. FW7 became as fit, or less fit, than the WT (relative fitness values of 1 or less) when the mutant occupied around 80% or more of the initial population (Figure 4A). These results are suggestive of negative frequency-dependent fitness (and perhaps some form of social cheating) as mutants are expected to do better when they are rare as there are more cooperating cells for them to exploit (Ross-Gillespie et al., 2007).

To further confirm that loss of spo0A can provide a measurable fitness increase, we repeated the same fitness experiment but used a ClosTron inactivated spo0A mutant (spo0A::CTermB) in place of FW7 (Figure 4B). Again, we found a very similar trend as low amounts of spo0A::CTermB in the initial population gave very high relative fitness values. These decreased as the initial population of the mutant was increased. Furthermore, spo0A::CTermB fitness become the same or less fit to the wild type when the mutant occupied above 80% of the population, the same as FW7. This further confirmed the considerable fitness benefit conferred by losing Spo0A functionality.

To test whether partial loss of spo0A activity would give similar results, we chose CIC4-1 as this variant had a mutation in the suspected phosphoacceptor domain of Spo0A (Brown et al., 1994), which allowed the mutant to produce some solvents but was unable to produce spores (Supplementary Figures 2, 3). This mutant had also gained SNPs within a non-coding region and an annotated galactose transporter (Cbei_3422) which we assumed did not play a significant role in degenerate phenotypes. Again, a similar frequency-dependent trend was seen for this mutant, mirroring what we observed for FW7 and spo0A::CTermB (Figure 4C). Interestingly however, CIC4-1 was able to sustain higher relative fitness values for longer than the other two mutants when the initial proportion was increased, showing values above 1 even when reaching 90% of the population.

We then tested the fitness between FW7 and CIC4-1 to observe whether full or partial loss would give the highest fitness values. We found that the relative fitness of FW7 was generally higher than that of CIC4-1 although there was not as clear evidence of frequency-dependent fitness as seen with the WT mixed cultures (Figure 4D). This still suggests however that complete loss of Spo0A activity does give the greatest fitness advantage, however, this advantage is less pronounced when grown against a strain carrying partially active Spo0A.

During the fitness assay tests, we observed a rapid decline in the CFU/ml of FW7 after more than 24 h of growth (Supplementary Figure 5). We hypothesized that although the mutant had gained considerable fitness compared the WT, this fitness only provided an advantage within the early stages of growth and could not be maintained as cell entered the stationary phase. We attempted to improve culture viability by repeating our standard subculturing experiment but changing the transfer time from 24 to 72 h. As before, we started with single, germinated WT spores which we inoculated into 10 ml of medium and subcultured every 72 h in a 1/10 dilution for five independent biological replicates. Using the 24 h transfers as a comparative control, solvent production was maintained throughout the whole subculturing process (Figure 5A), unlike what we observed previously (Figure 1). We did observe a drop in solvent production following subculture 3, however, after this point solvents stabilized at around 6 g/L. Furthermore, the number of spores produced by the cultures remained relatively stable throughout the whole experiment (Figure 5A). Another interesting observation was the maintenance of the RD type which remained consistently as the major colony type throughout the subculturing (Figure 5B). A small increase in the DCOG was seen at subculture 6, but this was not maintained. Adjusting the transfer time appeared to have a clear effect on the degeneration rate of the culture.

We also carried out a heat shock bottlenecking experiment to observe whether applying a bottleneck, one in which only spore-forming, un-degenerated cells could survive, to serve as a control to maintain culture stability. Subculturing was repeated at 24 h intervals for three independent replicates, however we heat shocked 1 ml of the transfer culture at 80°C for 10 min, thus inactivating all vegetative cells and inducing germination of any viable spores present. As to be expected, we found that culture stability was maintained throughout the 15 subcultures as solvent production was maintained at around 6 g/L (Figure 5C). Spore numbers were also maintained and did not drop below 10⁸ heat resistant CFU/ml throughout the experiment (Figure 5C). When we observed colony morphologies, the RD type was again the predominant type, however, the DCOG type existed in greater numbers than when the transfer time was adjusted (Figure 5D). This in part may be due to the DCOG type still capable of forming viable spores. There were little to none of the FW or CIC observed during this bottlenecking experiment.

In addition to phenotypic observations, we sequenced 10 colonies from three independent cultures found at the end of the heat shock experiment to observe whether this regime prevented the emergence of degenerate genotypes. Indeed, only one (replicate 2, colony 4) had gained a mutation linked to degeneration as this variant had mutation in spo0A (Supplementary Table 3). It is unclear whether this mutation was gained pre- or post-germination, however, the lack of significant degeneration genotypes exemplifies the effectiveness of this regime in preventing degeneration from occurring.

Bacterial strains utilized in this study are listed in Table 1. Clostridium beijerinckii NCIMB 8052 and generated mutants were grown at 37°C in an anaerobic cabinet (MG1000 Anaerobic Work Station, Don Whitley Scientific) containing an atmosphere of 80% nitrogen, 10% hydrogen and 10% carbon dioxide. The organism was routinely cultured in supplemented clostridial basal medium (CBMS), unless stated otherwise. CBMS was based on CBM as previously described (O’Brien and Morris, 1971) but contained glucose (60 g/L unless otherwise stated) and calcium carbonate (5 g/L) as a buffering agent. For agar plates, 15 g/L technical agar 1 (Oxoid, United Kingdom) was added. Escherichia coli TOP10 was grown in lysogeny broth (LB) at 37°C. Antibiotics were used at the following concentrations: erythromycin, 500 μg/ml for E. coli and 10 μg/ml for C. beijerinckii; spectinomycin 250 μg/ml for E. coli and 750 μg/ml for C. beijerinckii. Clostridium beijerinckii wild type was stored as a spore stock and generated mutants as 10% (v/v) glycerol stocks frozen at −80°C.

Plasmids and primers used in this study are listed in Tables 2, 3, respectively. Primers were synthesized by Sigma-Aldrich, United States. PCR amplifications were carried out using high fidelity Phusion polymerase (New England Biolabs, United States) or DreamTaq DNA polymerase (ThermoFischer Scientific, United States). Plasmid isolation carried out using Monarch Plasmid Miniprep Kit (New England Biolabs, United States) and genomic DNA preparations were carried out using the GenElute Bacterial Genomic DNA kit (Sigma-Aldrich, United States) following the manufactures instructions. Restriction enzymes were supplied by New England Biolabs and were used according to the manufacturers’ instructions. Electroporation of C. beijerinckii was performed as described previously (Oultram et al., 1988).

Clostridium beijerinckii NCIMB 8052 ClosTron mutagenesis was carried out as described by Heap et al. (Hayashida and Yoshino, 1989; West et al., 2006). ClosTron carrying plasmids were designed using the design tool available on http://www.ClosTron.com/ClosTron2.php and purchased from ATUM (formerly DNA 2.0). Numbers in the respective plasmid names (Table 2) indicate the retargeting site used within the disrupted gene. Genomic DNA from putative mutants was screened by PCR using primers that flanked the gene (see Supplementary Table 1) to establish whether the ClosTron-derived group II intron had inserted at the desired site. The generated PCR fragments were Sanger sequenced to obtain definite proof that intron insertion had occurred at the desired position.

Sanger sequencing was carried out by Source BioScience (Nottingham, United Kindom). Isolated genomic DNA was whole genome sequenced by MicrobesNG (Birmingham, United Kindom). Genome sequences were aligned and compared using CLC genomics workbench v.10 (Qiagen, Germany). Ultra-deep sequencing was carried out by Deepseq (Nottingham, UK.) and sequencing libraries were prepared using the KAPA HyperPlus PCR-free Kit (Roche; KK8513) and the KAPA Dual-Indexed Adapter Kit, Illumina Platforms (Roche; KK8722). Genomic DNA was quantified using the Qubit Fluorometer and the Qubit dsDNA BR kit (ThermoFisher; Q32853) and 500 ng of each sample were used for library preparation. Genomic DNA was fragmented for 10 min and the PCR-free workflow was followed, according to the KAPA HyperPlus Kit protocol (KAPA Biosystems; KR1145—v3.16). The final double-sided SPRI-based size selection was performed using 0.7/0.9 ratios of AMPure XP beads (Beckman Coulter; A63882). Libraries were quantified using the Qubit Fluorometer and the Qubit dsDNA HS Assay Kit (ThermoFisher: Q32854) and fragment length distributions were assessed using the Agilent 2,100 Bioanalyzer and the Agilent High Sensitivity DNA Kit (Agilent; 5067-4626). Libraries were pooled in equimolar amounts and final library quantification was performed using the KAPA Library Quantification Kit for Illumina (Roche; KK4824). The library pool was sequenced on the Illlumina NextSeq 500 using a NextSeq 500 High Output 300 cycle kit v2.5 (Illumina; 20024908), to generate 150-bp paired-end reads.

Variant calling: Reads were trimmed using Cutadapt (Martin, 2011) v2.10 with parameters -A “AGATCGGAAGAGCACACGTC” -a “AGATCGGAAGAGCACACGTC” --trim-n --nextseq-trim = 20 -m 30. Trimmed reads were aligned with BWA-MEM (Li and Durbin, 2009) v0.7.17 to C. beijerinckii reference (NCIMB 8052 ASM1696v1). Variant calling was performed using FreeBayes (Garrison and Marth, 2012) v1.3.2 and LoFreq (Wilm et al., 2012) v2.1.5. We performed Freebayes calling using parameters -J --no-population-priors -p 50 -F 0.005 -C 2 --min-base-quality 20 --use-best-n-alleles 4 --use-duplicate-reads. Lofreq calling was performed with default parameters. Multiallelic variant calls were split using bcftools (Danecek et al., 2021) v1.10.2 using parameters -m -any.

Joint-Called variant Filtering: SNPs called by FreeBayes with QUAL ≥ 100 and LoFreq with FILTER = PASS were kept. For each joint-called variant, we generated a pileup in first-forward and first-reverse orientation using RustyNuc v0.3.0 (Debebe, 2021) using parameters --min-reads 4 -q 20. Only calls with AF above 0.04 in both first-forward and first-reverse orientation were kept (Supplementary Table 4A).

Heatmap filtering: Heatmap was created using ComplexHeatmap v2.12.0 (Gu et al., 2016) and we excluded variants that (i) did not occur in at least three subcultures with AF > 0.08 in a single replicate, (ii) located in rRNA regions and (iii) already occurred early on in the starter culture (subculture 0 with AF > 0.5). In addition, we removed variant call at position 1,957,725 since it displayed similar dynamics across replicates, occurred in majority of subcultures and displayed correlation with sequencing depth across replicates (Supplementary Table 4C).

Genome Annotations: operon and transcription factor binding site locations were obtained from OperonDB (Pertea et al., 2009), CollecTF (Kiliç et al., 2014) and PRODORIC (Dudek and Jahn, 2022).

All raw data used for the ultra-deep sequencing has been deposited at the European Nucleotide Archive (ENA) with accession PRJEB51942 and can be accessed via https://lowfreqvar.github.io.

A full schematic of the subculturing protocol can be found in Supplementary Figure 1, but generally was performed as follows; 50 μl of N-WT spore suspension was heated at 80°C for 10 min in a heating block, serially diluted down to 10⁻³ in sterile PBS, plated on CBMS agar and left to germinate overnight at 37°C in the anaerobic work station (MG1000 Anaerobic Work Station, Don Whitley Scientific). Single colonies were then inoculated into 10 ml CBMS liquid medium and left for 24 h. 1 ml of culture was inoculated into fresh 9 ml CBMS liquid medium in a 1:10 dilution every 24 h for however many subcultures each experiment required.

For isolation of potential degenerates, a single germinated colony was mixed with 1 ml CBMS liquid medium and 100 μl of the suspension was distributed to 6 parallel tubes containing 10 ml CBMS liquid medium. The remaining liquid was left for 24 h and served as the parental wild type genomic profile. Parallel cultures were subcultured 5 times as previously described. Afterwards, 100 μl of culture was serially diluted 1:10 down to 10⁻⁵ with anaerobic PBS, plated on CBMS agar and incubated for 2 days to give single colonies. Colony variants were selected based on their colony morphology and inoculated into 10 ml CBMS liquid medium, left overnight and afterwards a 1 ml sample was taken and made into a 10% (v/v) glycerol stock for future use. Genomic DNA was then isolated for subsequent analysis.

To study culture dynamics with respect to culture productivity (organic solvents, acids and spores) and colony morphologies, the standard subculturing protocol as previously described was employed. 1 ml fermentation product samples were taken at the point of inoculation and 2 days after inoculation at each successive subculture. These were centrifuged at full speed for 1 min and the supernatant removed and frozen at −20°C for future analysis. 1 ml spore samples were taken after 5 days of growth and stored at 4°C until used in a spore assay. To observe colony morphologies at each subculture, after 1 day of growth (post inoculation), 200 μl samples were taken, serially diluted down to 10–5 and plated onto CBMS agar. Colonies were left to grow for 2 days and the CFU of each colony type was enumerated. Transfer times were changed from 24 to 72 h for one experiment. The 1 ml transfer culture was also heat shocked at 80°C before subculturing for one experiment.

Wild type spores were heat shocked as described for the above subculturing protocol and plated on CBMS agar and desired degenerated variants were streaked onto CBMS agar from a frozen stock. Each were left to grow for 24 h. Colonies were inoculated into 10 ml CBM 6% glucose broth and left to grow for 24 h. 24 h cultures of WT and the desired degenerate were added in varying volumetric ratios to make a total of 1 ml which was used to inoculate to 9 ml CBM 6% glucose liquid so that the final volume was 10 ml. For example, 100 μl degenerate variant +900 μl wild type added to 9 ml CBM-S broth 6% glucose. This would give a theoretical starting ratio of 1:10 (degenerate:WT) in the 1 ml inoculum, with exact ratios determined experimentally through enumeration of colony morphotypes (CFU/ml) following spread-plating. CFU/ml were enumerated at 0, 12, 24, 48, and 72 h using serial dilutions of the culture. Fitness of individuals was calculated using w = [x₂(1 – x₁)]/[x₁(1 – x₂)] from Diggle et al.⁴⁵ where x₁ is the initial proportion of mutant in the population and x₂ is the final proportion at a given time point.

Gas chromatography (GC) was used for acetone, ethanol, butanol, acetate and butyrate concentration analysis in which 1 ml fermentation samples were taken, centrifuged at 16,000 ×g for 1 min with the supernatant removed and stored at −20°C until needed. Extraction of fermentation products and their gas chromatographic analysis was carried out as described previously (Willson et al., 2016). Optical density at 600 nm was used as a measure of cell growth and determined using a Thermo Scientific Biomate 3 spectrophotometer with sterile liquid medium as a blank. If samples had OD600 above 0.8, samples were diluted 1:10 and re-measured.