Laboratory Evolution and Reverse Engineering of Clostridium thermocellum for Growth on Glucose and Fructose

C. thermocellum is an important candidate for sustainable and cost-effective production of bioethanol through consolidated bioprocessing. In addition to unsurpassed cellulose deconstruction, industrial application and fundamental studies would benefit from improvement of glucose and fructose consumption.

give a desired phenotype has been named reverse, or inverse, metabolic engineering (33,34). This can give insights into fundamental biology, such as phenotype-genotype relationships, as well as guide future metabolic engineering to improve the phenotype of the microorganism.
The aim of this study was to (i) reproducibly achieve growth of C. thermocellum on hexose sugars through laboratory evolution in chemostats with decreasing concentration of cellobiose and increasing concentrations of glucose or fructose in the medium, (ii) identify possible underlying genetic changes needed for growth on glucose and fructose through whole-genome sequencing of multiple single-colony isolates and reverse metabolic engineering, and through this (iii) elucidate whether mutational events or adaptation are needed for growth of C. thermocellum on hexose sugars by analysis of the role of the identified mutations in improving growth and reducing lag time on the monosaccharides.

RESULTS
Laboratory evolution for growth on glucose or fructose. To select for C. thermocellum growing on glucose or fructose, two sets of duplicate carbon-limited chemostats with decreasing concentration of cellobiose and increasing concentrations of glucose or fructose in the medium were used. Feeding was initiated with 4.7 g liter 21 cellobiose, and the first switch was to 2.4 g liter 21 cellobiose and either 2.4 g liter 21 glucose or 2.5 g liter 21 fructose. Thereafter, the feed was switched to 4.8 g liter 21 glucose or fructose as the sole carbon and energy source. After the first switch, the biomass concentration decreased approximately 50% (Fig. 2). During this period, the cellobiose fed was utilized while the hexose sugars accumulated. The time point at which the online optical density (OD) increased, indicating improved glucose consumption, was similar for the duplicate chemostats (42 and 43 h, Fig. 2A). In contrast, the onset of fructose uptake took longer (105 and 190 h, Fig. 2B). Biofilm formation was observed in the bioreactors upon switching to fructose-limitation, which made the online OD value estimations unreliable. During this phase, the residual fructose and offline biomass concentration (total organic carbon [TOC]) remained stable ( Fig. 2B; see also Fig. S1B in the supplemental material).
Biomass-specific consumption rates and extracellular product yields differed not only between the hexose sugars and cellobiose, but also between the two hexose sugars (Table 1). In order to maintain the same biomass-specific grow rate of 0.10 h 21 , glucose was taken up 60% faster than cellobiose (1.07 6 0.03 compared to 0.66 6 0.06 g g biomass 21 h 21 ). Consequently, and in line with known metabolism, the biomass yield on glucose was 40% lower (0.09 6 0.00 g biomass g substrate 21 ) than the biomass yield on cellobiose (0.15 6 0.01 g biomass g substrate 21 ). Extracellular product profiles showed a 40% increase in acetate yield on glucose compared to cellobiose and smaller changes in ethanol and formate (P , 0.05 based on Student's t test), while the estimated CO 2 product yield was the same. These changes significantly shifted the acetate/ethanol ratio from 1.2 to 2.0. Surprisingly, fructose was consumed 50% slower than glucose (0.55 6 0.02 g g biomass 21 h 21 ) resulting in a high biomass yield (0.18 6 0.00 g biomass g substrate 21 ). The fructose-limited cultures generally showed decreased product yields compared to the other two sugars. Formate was not significantly changed whereas acetate, ethanol, and CO 2 decreased (24, 32, and 30%, respectively, P , 0.05), maintaining the acetate/ethanol ratio constant. The observed residual fructose concentration was 0.5 g liter 21 , which was significantly higher than the residual 0.02 g liter 21 glucose and , 0.01 g liter 21 cellobiose concentrations (Table 1), indicating a low affinity for fructose in the evolved cultures.
Screening of growth rate and lag time on cellobiose, glucose, and fructose. To obtain single-colony isolates, a sample from the glucose-or fructose-limited cultures was plated on solid medium with the corresponding monosaccharide. After the first round of plating, colonies were picked, inoculated into liquid medium, and screened for growth characteristics on a plate reader. Based on specific growth rate and lag time, colonies spanning different characteristics were selected (indicated with arrow in Fig. S2 in the supplemental material). These candidates were plated in two subsequent rounds before single-colony isolates were saved as freezer stocks for further analysis. Four single-colony isolates were selected from each chemostat run, resulting in 16 strains in total (Table 2). From here on, the single-colony isolates from the two glucose-limited chemostats are referred to as the glucose isolates (G-strain number) and, similarly, the fructose isolates (F-strain number) represent the group of single-colony isolates from the two fructose-limited chemostats.
The specific growth rate and lag time of each single-colony isolate on each of the three sugars were subsequently quantified in microplate cultivations in a plate reader. Prior to a transfer to each of the three sugars, the inoculum was grown on the monosaccharide that the strain was isolated on, or on cellobiose in the case of wild type. As a reference, the wild-type strain grew on cellobiose with 0.50 6 0.07 h 21 ( Fig. 3D; see also Table S3 in the supplemental material), whereas on glucose the growth of wild type only commenced after 42 6 4 h at a lower growth rate of 0.25 6 0.04 h 21 (P , 0.01, Fig. 3E). No growth on fructose was observed for the wild type within the time frame of the experiment (80 h). To assess the impact of the laboratory evolution, both hexose isolates were investigated for growth on cellobiose. The fructose isolates as a group had a slightly lower specific growth rate on cellobiose (0.40 6 0.05 h 21 , n = 8, P , 0.01) than the wild type and demonstrated only minimal lag times (,3.6 h) (Fig.  3A). Among these isolates, F-LL1538 was an outlier with 0.51 6 0.04 h 21 and no lag time on cellobiose (see Table S4). In comparison to the fructose strains, the glucose isolates grew more variably on cellobiose (Fig. 3A). For instance, one strain (G-LL1521) did not grow on cellobiose within 80 h, while another strain (G-LL1519) grew similarly to the wild type at 0.50 6 0.04 h 21 (see Table S4). As a group, the glucose strains (excluding G-LL1521) showed an average specific growth rate on cellobiose of 0.34 6 0.09 h 21 and an average lag time of 15 6 11 h.
Next, growth and lag time of the hexose isolates on the sugar they evolved on, as well as the alternate hexose sugar, was investigated. On glucose, the glucose isolates as a group grew with 0.29 6 0.06 h 21 (Fig. 3B). On fructose, the fructose isolates as a group grew with 0.41 6 0.06 h 21 (Fig. 3C). Surprisingly, the fructose isolates also ) ,0.01 0.01 6 0.01 ,0.01 Carbon recovery (%) c 79 6 1 8 7 6 1 6 7 6 4 a Data are averages of four biological replicates on cellobiose and two biological replicates on glucose and fructose. b The CO 2 yield was calculated as the sum of the following molar yields: ethanol 1 acetate 2 formate. c Carbon recovery was calculated as the ratio between total carbon found in products and total carbon in consumed substrate (see Table S1 in the supplemental material).
demonstrated rapid growth on glucose at 0.21 6 0.04 h 21 with a lag time of 4.7 6 3.1 h, which was similar to that of the glucose isolates (Fig. 3B). This shows that both groups of isolates grew on glucose with significantly reduced lag times and growth rates comparable to those of the wild type. Interestingly, only one of the glucose isolates (G-LL1520) grew reproducibly on fructose after a 30 6 1 h lag time at 0.18 6 0.02 h 21 . Two of the glucose isolates (G-LL1517 and G-LL1519) showed growth on fructose for only three out of the five experimental replicates at a low growth rate of 0.05 6 0.01 h 21 (see Table S4). The remaining glucose isolates did not grow on fructose within the 80 h of the experiment (Fig. 3C). Constitutive growth on glucose or fructose from isolates. To investigate whether the isolates had adapted or evolved for growth on glucose and fructose, another platereader experiment was conducted where the isolates were grown for two transfers in media with 5 g liter 21 cellobiose. At midexponential phase, the cultures were used to inoculate a 96-well microplate with 5 g liter 21 glucose or fructose. The fructose isolates showed similar strongly reduced lag times for growth on glucose and fructose independent on whether the inocula carbon source was cellobiose or fructose ( Fig. 3E and F). Also, the glucose isolates showed a reduced lag time for growth on glucose upon transfer from cellobiose (Fig. 3E). Even though three glucose isolates-G-LL1518, G-LL1522, and G-LL1523-had a slightly increased lag time on glucose, from 1.7 6 0.6, 0, and 0 h when the inocula were grown on glucose to 13 6 0.4, 10 6 1, and 2.3 6 0.1 h (P , 0.01), respectively, these lag times are still more than four times shorter than those for the wild type on glucose (42 6 4 h) (see Table S3). These observations suggested that the isolates acquired mutations which allowed constitutive growth with no or short lag time on glucose and fructose upon transfer from cellobiose.
Whole-genome sequencing of isolates revealed targets for reverse engineering. The 16 single-colony isolates were whole-genome sequenced in order to identify mutations contributing to constitutive growth on glucose and fructose. The sequencing reads were mapped against the DSM 1313 reference genome to find single- nucleotide polymorphisms, insertions, and deletions. All identified mutations are listed in Data Set S1 in the supplemental material. Mutations found in the parent strain (LL1004) were not considered for reverse engineering. A mutation was also excluded for further study if it was found in (i) a hypothetical protein, (ii) a noncoding region far from coding regions, or (iii) genes unlikely contributing to growth on glucose and fructose based on the current genome annotation and knowledge of C. thermocellum metabolism.
The fructose and glucose isolates shared mutations in three genes (Table 3). First, the gene clo1313_1828 (cbpA [24]) was frequently mutated in both groups. The gene cbpA encodes a periplasmic sugar-binding protein belonging to an ATP-binding cassette (ABC) transport system, which is highly translated (20,26), and has binding affinity for cellotriose (24). Seven of eight glucose isolates had a single-point mutation g.2140693C.G (the number refers to the position in the genome) that resulted in the amino acid change A173P. The eighth isolate (G-LL1520) had a double mutation g.2140395G.A and g.2140503C.G, which resulted in the amino acid changes A272V and G236A, respectively. All fructose isolates shared a mutation in a different position in cbpA, g.2140767C.A, resulting in the different amino acid change G148V.
Second, the gene clo1313_1831 was mutated in both groups. Clo1313_1831 is annotated as a member of the Repressor-ORF-Kinase (ROK) family of transcriptional regulators and kinases (35). It has both a helix-turn-helix motif common for repressors and an ATP binding motif common for kinases (36). For seven fructose isolates and three glucose isolates, this mutation resulted in a frameshift or a premature stop codon, likely inactivating the gene product. For the other five glucose isolates (G-LL1516 to G-LL1519, as well as G-LL1523), an insertion sequence ISCth1 (see Fig. S3A) was inserted 137 bp downstream of clo1313_1831. The ISCth1 belongs to the IS982 family and contains a transposase within the IS1 protein family (37). While clo1313_1831 is located on the reverse strand, ISCth1 was on the forward strand.
Third, clo1313_1857 was mutated in the two sets of fructose isolates and in one glucose isolate. In three fructose isolates (F-LL1542 to F-LL1544) from the same chemostat, FIG 3 Specific growth rate and lag time on 5 g liter 21 cellobiose, glucose, and fructose of single-colony isolates from glucose-limited (blue squares) or fructose-limited (red triangles) chemostats. The isolates were grown in batch cultivations for up to 80 h in a microplate incubated in a plate reader. Inocula for the isolates in panels A to C were grown on the hexose in which they were isolated, whereas in panels D to F the inocula were grown on cellobiose. Wild type (black circle) was used as a control. The data are shown as averages 6 the standard deviations for 20 biological replicates for the wild type and three to five biological replicates for the remaining strains.   The remaining mutations were not shared by the two groups (Table 3). Half the glucose isolates had a frameshift or premature stop codon in cbpB (clo1313_1194), which is another highly translated sugar-binding protein belonging to an ABC-transport system (20,24,26,36). Another gene, clo1313_0289, was frameshifted in three fructose isolates from the same chemostat (F-LL1542 to F-LL1544), likely inactivating the protein. This gene is annotated as a putative transcriptional regulator within the CopG family (36). Since it was mutated in F-LL1542, which lacked a mutation in the clo1313_1831 ROK gene, it may have had a contributing role to growth on fructose.
Reverse engineering identified important genes for growth on glucose and fructose. Fructose isolates not only grew effectively on fructose but also grew similarly to the wild type on cellobiose and similarly to the glucose isolates on glucose. In view of this and the time-consuming nature of gene editing in C. thermocellum (38), only mutations found in the fructose isolates were selected for reverse engineering. Based on the whole-genome sequence analysis in the previous section, the following four mutations were selected to investigate their contributions to growth on fructose and glucose through reverse metabolic engineering: knockout of the genes clo1313_0289, clo1313_1831, and clo1313_1857 and introduction of the point mutation g.2140767C.A (G148V) into cbpA.
These mutations were introduced in the genetically tractable reference strain LL345 (Dhpt). Growth on cellobiose, glucose, or fructose, was characterized in a plate-reader experiment. This reference strain grew readily on cellobiose (0.48 6 0.05 h 21 ) and slower on glucose (0. 32 6 0.07 h 21 , P , 0.01, Fig. 4). The lag time on glucose was 45 6 4 h, and the strain did not grow on fructose within the time frame of this experiment (80 h). All this was similar to the previously used wild-type strain.
Introduction of three of the four individual mutations had no beneficial effects and was actually detrimental for growth on at least one of the sugars (Fig. 4). Knockout of clo1313_1857 (AVM012) resulted in a 53% longer lag time on glucose (69 6 4 h). The DcbpA (AVM011), cbpA G148V (AVM028) and Dclo1313_0289 (AVM009) mutants did not show growth on glucose within 80 h. In addition, Dclo1313_0289 (AVM009) grew 32% slower on cellobiose compared to LL345 (0.33 6 0.05 h 21 , P , 0.01). In contrast, knockout of clo1313_1831 (AVM013) maintained the high growth rate on cellobiose (0.49 6 0.03 h 21 ) and enabled growth on fructose, with a specific growth rate of 0.11 6 0.01 h 21 and a lag time of 12 6 1 h. The lag time on glucose was also significantly reduced to 2.6 6 0.5 h while maintaining slightly lower growth rate compared to LL345 (0.24 6 0.01 h 21 , P , 0.01). To test a hypothesis that phosphorylation is limiting growth on fructose, it was investigated whether deletion of the ROK protein Clo1313_1831 resulted in derepression of a fructokinase. Fructokinase activity was not detected in the parent strain LL345, neither ATP-nor GTP-linked (Table 4). On the contrary, the Dclo1313_1831 mutant (AVM013) grown on cellobiose showed 0.29 6 0.02 U mg protein 21 fructokinase activity with both ATP and GTP. The GTP-linked glucokinase activity did not show a similarly large increase (Table 4).
In comparison to the fructose isolates, growth of Dclo1313_1831 (AVM013) showed a longer lag time and lower growth rate on fructose. To investigate the contribution of additional mutations to this phenotype, the clo1313_1831 deletion was combined with the other three reverse-engineering candidates. Combinations of Dclo1313_1831 with Dclo1313_1857 and Dclo1313_0289, respectively, yielded the same growth rate and lag time on fructose compared to the parent strain (Dclo1313_1831) (Fig. 4). However, introduction of the single-point mutation cbpA G148V in the Dclo1313_1831 mutant (AVM027) more than doubled the growth rate to 0.32 6 0.03 h 21 on fructose and abolished the lag time. On glucose, there was no significant changes compared to the single knockout, while the growth rate on cellobiose was reduced by 30% to 0.34 6 0.01 h 21 (P , 0.01).

DISCUSSION
In this study, C. thermocellum was successfully evolved for constitutive growth on fructose and glucose. Single-colony isolates from glucose-or fructose-limited chemostats showed no or very short lag time on the respective sugar ( Fig. 3A to C). In line with a previous study on agar plates (17), this improved phenotype was constitutive and transgenerational ( Fig. 3E and F), suggesting that the chemostat cultures had acquired beneficial mutations. Whole-genome sequencing and subsequent reverse metabolic engineering identified beneficial mutations that are sufficient to enable fast and efficient growth on both glucose and fructose. However, in view of the reproducible short lag time of 42 6 4 h of the wild type on glucose in batch cultures (n = 20; Fig. 3E), which is difficult to reconcile with the stochastic nature of spontaneous mutations, additional contribution of physiological adaptation to growth of the wild type on glucose cannot be excluded. Growth on glucose and fructose was differentiated with respect to other features as well. In batch cultures, the lower maximum growth rate on glucose compared to both fructose and cellobiose (Fig. 3) could be due to a lower maximum uptake rate (12,23,30). Furthermore, glucose-limited cultures showed a lower biomass yield of 0.09 g biomass g substrate 21 compared to 0.15 g biomass g substrate 21 on cellobiose, which is similar to a previous report on strain LQRI (12), and reflects the bioenergetic benefit from growing on the disaccharide, in line with the current understanding of metabolism (8). The glucose-limited cultures also produced more acetate, which is stoichiometrically coupled to ATP-formation. Increased acetate production may be to compensate for the lower ATP-yield from glycolysis. In contrast, the biomass yield on fructose was 0.18 g biomass g substrate

21
, similar to a previous report of 0.17 g biomass g substrate 21 with ATCC 27405 at 0.11 h 21 (39), which might reflect different transport and phosphorylation mechanisms compared to glucose.
The genetic change needed in the wild type for constitutive and reproducible growth without a lag time on glucose was inactivation of the gene clo1313_1831. On fructose, a point mutation in cbpA on top of the clo1313_1831 deletion was necessary to fully abolish the lag time and achieve a high growth rate (Fig. 4). Inactivation of Clo1313_1831 was likely also necessary in the chemostat evolution, where it was mutated in most isolates. In the few isolates that had no mutation in clo1313_1831, its transcription may have been affected by a transposable element inserted downstream of the gene. Transcriptional interference can occur when a strong transposase promoter is inserted "face-to-face" with another weak promoter, in so-called convergent transcription (40,41). However, further studies into gene expression of those isolates would be needed to confirm such interference. Interestingly, in a previous study, this gene was secondarily mutated in an evolved and engineered strain of C. thermocellum (LL1210), which compared to the wild type no longer accumulated glucose (20). Although the contributing role of this mutation was not studied, it may play an important role in light of these results. Deletion of Clo1313_1831, which is a member of the Repressor-ORF-Kinase (ROK) family of transcriptional regulators and kinases, resulted in upregulation of fructokinase activity (Table 4). However, the impact of deleting this putative regulatory protein is likely broader, including for instance transport-related processes, as illustrated by the impact of this deletion for growth on glucose without affecting glucokinase activity. Investigation of the targets of the putative transcriptional regulator Clo1313_1831 by transcriptomics would be of relevance for follow-up studies. The point mutation in cbpA, which is suggested to be part of an ABC-transporter system (24) or its signaling cascade, may have increased the affinity of this protein for fructose. Such a modification has been shown in vitro in the arabinose-binding protein encoded by araF in Escherichia coli, where a single-point mutation in the hinge region increased affinity for galactose 20-fold (42). Transport studies complemented with transcriptomics and/or proteomics in mutant strains could guide further efforts in understanding the molecular basis for the contribution of this cbpA mutation to growth on the monosaccharides. Even though the remaining targets for reverse metabolic engineering (clo1313_1857 and clo1313_0289) did not positively contribute to growth on glucose or fructose in batch, these targets may have contributed to improved fitness (e.g., better affinity) in the continuous cultures. In general, regulation of carbon utilization is poorly understood in C. thermocellum and these findings, together with further expression studies that for instance might reveal activation of otherwise silent genes (43), might help increase the understanding of how the organism regulates its metabolism on a broader range of carbon substrates.
The present study demonstrated the genetic changes necessary for constitutive and reproducible growth on glucose and fructose without a lag time and can guide engineering of strains for industrial applications and fundamental studies. In industry, high product yields are of key importance for production of low-value high-volume products such as ethanol (21). Studies have shown that C. thermocellum accumulates glucose in batch cultivations with industrially relevant cellulose concentrations (18,19). Also engineered strains for ethanol production show this phenotype (20,44). In academia, a strain capable of growing on readily available, inexpensive, and highly soluble model substrates facilitates fundamental studies using, e.g., extensive continuous cultures, isotope-based labeling, or high-substrate loadings. Furthermore, these insights also suggest that interpretation of previous studies on C. thermocellum using hexose-based medium might be complicated by the occurrence of spontaneous mutations with possible broader regulatory consequences. Future studies of these mutations should include investigation of the effect of these genetic changes on the cellulolytic capability of C. thermocellum and on glucose accumulation in high-substrate batch cultivations. Together, these insights can guide engineering of strains for fundamental studies into transport and upper glycolysis, as well as maximized substrate utilization in industrial applications.

MATERIALS AND METHODS
bottom-hose outlet (Kimble KIMAX from Fisher Scientific, Pittsburgh, PA) and a rubber head-plate with stainless-steel tubes. It was autoclaved (at 121°C for 2.0 to 2.5 h) with water and then continuously purged with N 2 gas. The medium stock solutions were filter-sterilized into the feed vessel in alphabetic order. The feed vessel was connected to both bioreactors through a precalibrated peristatic pump (Watson-Marlow, Falmouth, Cornwall, England). The weights of both the feed vessel and collected effluent were recorded and used to correct the feed rate to ensure an accurate dilution rate. The headspace of the bioreactor was purged with 20% CO 2 and 80% N 2 for 2 to 4 h before inoculation. After inoculation, the culture was grown in a batch phase with no overhead purging until it reached an OD 850 peak (16 to 24 h). A medium feed of 30 ml h 21 was started, corresponding to a dilution rate of 0.10 h 21 , and the headspace was purged with 5 ml min 21 N 2 gas. At specific time points, the feed vessel was aseptically replaced with various carbohydrate mixtures. The feed medium at the start contained 4.7 g liter 21 cellobiose and was switched to a mixture of 2.4 g liter 21 cellobiose and either 2.5 g liter 21 fructose or 2.4 g liter 21 glucose, as indicated. Fructose cultures were switched after 240 h to facilitate collection of additional quantitative data. The final switch was to 4.8 g liter 21 fructose or glucose.
Sampling for determining rates and yields was done after at least four residence times on each combination of carbon sources with a smaller than 5% change in biomass concentration during one volume change. After withdrawal of large sample volumes, the reactor volume was brought back to the working volume level with an equal volume of fresh medium. Smaller samples (8 ml) were collected throughout the cultivation.
Single-colony isolation. The single-colony isolation consisted of three rounds of plating with one screening step in between rounds. In detail, a continuous-culture sample was transferred to an anaerobic chamber, plated in 10-fold dilution series on solid LC medium with glucose or fructose, and incubated until colonies were visible and distinct. Seven colonies from each bioreactor were picked and screened for growth rate and lag time on the different sugars. The colonies were grown in serum tubes with 4 ml of LC medium with glucose or fructose to an OD 600 of 0.4 to 0.6 and then used to inoculate a 96-well plate with 200 ml per well of LC medium with either 5 g liter 21 cellobiose, glucose, or fructose. The plate was incubated with shaking at 55°C, and the OD 600 was measured over 72 h in a modified BioTek PowerWave XS plate reader, as described by Olson and Lynd (47). After the screening, selected colonies were plated in 100-fold dilution series in two subsequent rounds to ensure colonies were based on single cells. After the last round, colonies were grown in liquid LC medium with glucose or fructose. At an OD 600 of 0.6 to 1.0, aliquots were stocked in 5-ml serum bottles and stored at 280°C. The pellet of the remaining culture was sent for whole-genome sequencing. The culture purity was checked by 16S rRNA sequencing, as well as by whole-genome sequencing. No contamination was detected in the single-colony isolates by either method.
Analytical techniques. The supernatant from chemostat samples was acidified by adding 35 ml of 10% sulfuric acid (wt/wt) to 700 ml of supernatant. After centrifugation, it was filtered (0.22-mm nylon) and stored at 4°C until analysis. Ethanol, acetate, formate, lactate, pyruvate, malate, glucose, cellobiose, and fructose in the supernatant were quantified by HPLC (Waters, Milford, MA) using an Aminex HPX-87H column (Bio-Rad, Hercules, CA) equipped with a refractive index detector and UV detector. The column was run with 0.5 ml min 21 of 2.5 mM H 2 SO 4 at 60°C. Cell dry weight samples were measured in triplicate by filtering 5-ml samples through preweighed 0.2-mm GTTP Isopore membrane filters (Merck Millipore, Ltd.), washing with an equal volume of ultrapure water, and drying at 100°C for 24 h before weighing. For analysis of pellet total nitrogen (TN) and total organic carbon (TOC), frozen cell pellets were thawed and washed twice with ultrapure water. The cell TN and TOC were analyzed in a Shimadzu TOC-Vcph TOC analyzer with a TN unit and ASI-V autosampler added (Shimadzu Scientific Instruments, Columbia, MD), using an acidified glycine standard as described previously by Holwerda et al. (45). The offline OD was measured at 600 nm in triplicate in a Thermo Scientific Genesys 335901 visible spectrophotometer.
Plate-reader cultivation and screening. In order to quantify the growth rate and lag time of isolates and engineered strains, plate-reader cultivations were performed with cellobiose, glucose, or fructose. A freezer stock was propagated in two serial transfers on 5 to 50 ml of LC medium. During exponential growth of the second transfer at an OD 600 of 0.5 to 1.0, the culture was brought into the anaerobic chamber and used to inoculate a preheated 96-well plate (sterile, clear, flat-bottom, non-surface-treated Eppendorf cell-culture plate with moat from Sigma-Aldrich, Stockholm, Sweden). The plates contained 200 ml of LC medium per well and were sealed with sterile adhesive SealPlate film (Sigma-Aldrich). A BioTek Epoch 2 plate reader from AH Diagnostics AB (Solna, Sweden) incubated the plate at 55°C with continuous double orbital shaking (425 cpm, 3 mm). The plate reader measured the OD at 600 nm every 5 min and was run for 80 h. Each strain was grown in at least five replicates on each carbon source with a distribution covering different parts of the plate. The strains were inoculated to the same starting OD (0.02). On each plate, at least six wells with media were used as negative controls.
A script to analyze the plate-reader OD data was written in Python 2.7 (from the Anaconda Distribution) and was deposited at Github (https://github.com/johannesyayo/plate_reader_analytics). Blanks were averaged at each time point and subtracted from each well. The standard deviation between the blanks was below 0.006. OD values below this threshold were considered as noise. Following this, the specific growth rate was calculated in two steps. In the exponential phase, the specific growth rate is constant and can be calculated as the slope of the natural logarithm of the OD. First, the maximum slope was found for a time frame of 5 h. Second, the exponential phase of the growth curve was expanded with 5-min steps if the new slope was $95% of the maximum slope and R 2 was above 0.995. The time until the start of this exponential phase was defined as the lag phase. Manual curation was applied to remove outliers. In case of diauxic growth due to carry-over of a different carbon Engineering C. thermocellum for Fast Growth on Hexoses Applied and Environmental Microbiology source from the inoculum, the lag time was manually adjusted as the time between the end of the initial growth phase and the beginning of the main exponential phase.
Resequencing. Whole-genome sequencing was used to verify strain construction and check for secondary mutations. Raw data are available from the NCBI Sequence Read Archive (see the accession numbers in Table 2). DNA was submitted to the Joint Genome Institute (JGI) for sequencing with an Illumina MiSeq instrument. Unamplified libraries were generated using a modified version of Illumina's standard protocol. First, 100 ng of DNA was sheared to 500 bp using a focused ultrasonicator (Covaris). The sheared DNA fragments were size-selected using SPRI beads (Beckman Coulter). The selected fragments were then end repaired, A tailed, and ligated to Illumina compatible adapters (IDT, Inc.) using KAPA Illumina library creation kit (KAPA Biosystems). Libraries were quantified using KAPA Biosystems' nextgeneration sequencing library qPCR kit and run on a Roche LightCycler 480 real-time PCR instrument. The quantified libraries were then multiplexed into pools for sequencing. The pools were loaded and sequenced using the appropriate Illumina reagent kit for a 2 Â 150-bp indexed run to generate pairedend reads.
Data were analyzed with CLC Genomics Workbench v12 (Qiagen). First, sequences were trimmed for quality (.0.001) and ambiguous residues (maximum of 1). Then, 2,500,000 reads were sampled to give an average read depth of about 105. Reads were mapped to the NC_017304.1 reference genome, using the default parameters, except that the similarity fraction was increased from 0.8 to 0.95 to reduce mapping errors. The preliminary alignment was improved by two passes of local realignment. Mutations were identified using the "basic variant detection" and "structural variant" tools. Single-nucleotide polymorphisms were filtered to eliminate those supported by fewer than 35% of reads or 10 total reads at a given position. Breakpoints were matched to known transposons within the C. thermocellum genome by BLAST (48).
Strain construction. Plasmids and primers used in this work are shown in Tables 5 and 6, respectively. The backbone plasmid for deletions and insertions was pDGO145 (49), which is derived from pDGO-68 (50) but with a low-copy E. coli origin of replication p15a instead of pUC19. Genomic DNA was extracted using a GeneJet kit from Thermo Fisher Scientific (Stockholm, Sweden). Primers (Invitrogen, Thermo Fisher Scientific) were designed to amplify fragments from genomic DNA with overhangs via PCR using Thermo Scientific Phusion High-Fidelity DNA polymerase. The primers used to amplify cbpA for plasmid pJY16 were designed with the specific point mutation g.2140767C.A (G148V). The PCR fragments were assembled with the backbone plasmid using Gibson assembly (51) in an E. coli BL21 derivate (New England Biolabs catalog number C2566I purchased from BioNordika AB, Solna, Sweden). Purification of PCR amplicons and plasmid DNA was done using commercially available kits from GeneJET (Thermo Fisher Scientific). Gene editing in C. thermocellum was performed as previously described by Olson and Lynd (38), through transformation by electroporation with a deletion or insertion plasmid, followed by several rounds of positive and negative selection using homologous recombination. For gene insertion or simultaneous gene deletion and insertion, the gene of interest was placed between the 59 flank and the 39 flank on the plasmid. For the deletion of clo1313_0289, pJY8 was designed with an extra 59 flank instead of an internal region due to the small gene size. As a first step in the strain construction, the genes clo1313_1857, clo1313_1831, clo1313_0289, and cbpA were deleted individually using plasmids pJY3, pJY5, pJY8, and pJY12, respectively, resulting in strains AVM012, AVM013, AVM009, and AVM011 (Table 2). In AVM011, the gene cbpA with point-mutation g.2140767C.A (G148V) was inserted using plasmid pJY16, resulting in strain AVM028. Second, Dclo1313_1831 was combined with the other gene modifications. The gene clo1313_1857 was deleted in AVM013 (Dclo1313_1831) using plasmid pJY3, resulting in strain AVM055. The single-point mutation was introduced into AVM013 (Dclo1313_1831) by simultaneous deletion of the native cbpA and insertion of cbpA G148V using pJY16, resulting in the strain AVM027. The double mutant Dclo1313_1831 Dclo1313_0289, named AVM058, was constructed from AVM009 (Dclo1313_0289) using plasmid pJY5. The culture purity (see below) and final genotype was confirmed by Sanger sequencing (Eurofins Genomics, Edersberg, Germany).
Culture purity. Culture purity was assessed in chemostats, isolates, and genetically modified strains, using 16S rRNA primers from Integrated DNA Technologies (IDT, Coralville, IA) (forward primer, 59-AGA GTT TGA TCC TGG CTC AG-39; reverse primer, 59-ACG GCT ACC TTG TTA CGA CTT-39). The sequenced PCR product was compared to the C. thermocellum DSM 1313 genome. Microscopy was also used during the continuous cultivations. The culture purity was also assessed by analysis of whole-genome  Engineering C. thermocellum for Fast Growth on Hexoses Applied and Environmental Microbiology