Improved growth rate in Clostridium thermocellum hydrogenase mutant via perturbed sulfur metabolism

Strains and culture conditions

Clostridium thermocellum DSM1313 and mutant strains C. thermocellum Δhpt ΔhydG (referred to as ΔhydG) and C. thermocellum Δhpt ΔhydG Δech (referred to as ΔhydG Δech) [11] were grown in CTFUD medium [6] and MTC minimal medium [15] prepared as described in [16]. CTFUD medium composition was (L⁻¹): 3 g sodium citrate tribasic dehydrate, 1.3 g ammonium sulfate, 1.43 g potassium phosphate monobasic, 1.8 g potassium phosphate dibasic trihydrate, 0.5 g cysteine-HCl, 10.5 g 3-morpholino-propane-1-sulfonic acid (MOPS), 6 g glycerol-2-phosphate disodium, 5 g cellobiose, 4.5 g yeast extract, 0.13 g calcium chloride dehydrate, 2.6 g magnesium chloride hexahydrate, 0.0011 g ferrous sulfate heptahydrate, and 0.0001 g resazurin, adjusted to pH 7.0. MTC medium consisted of (L⁻¹): 2 g sodium citrate dehydrate, 1.25 g citric acid monohydrate, 1 g sodium sulfate, 1 g potassium phosphate dibasic trihydrate, 2.5 g sodium bicarbonate, 1.5 g ammonium chloride, 2 g urea, 1 g magnesium chloride hexahydrate, 0.2 g calcium chloride dehydrate, 0.1 g ferrous chloride tetrahydrate, 1 g l-cysteine hydrochloride monohydrate, 5 g cellobiose, 0.001 g resazurin, 5 g MOPS, 20 mg pyridoxamine dihydrochloride, 1 mg riboflavin, 1 mg nicotinamide, 0.5 mg DL-thioctic acid, 4 mg 4-amino benzoic acid, 4 mg D-biotin, 0.025 mg folic acid, 2 mg cyanocobalamin, 4 mg thiamine hydrochloride, 0.5 mg MnCl₂·4H₂O, 0.5 mg CoCl₂·6H₂O, 0.2 mg ZnSO₄·7H₂O, 0.05 mg CuSO₄·5H₂O, 0.05 mg HBO₃, 0.05 mg Na₂MoO₄·2H₂O, and 0.05 mg NiCl₂·6H₂O.

Whole-genome resequencing

Genome resequencing was performed by the Department of Energy Joint Genome Institute (JGI, Walnut Creek, CA) using an Illumina MiSeq instrument. Genomic DNA was extracted using a Qiagen DNeasy kit (Qiagen, Valencia, CA), was sheared to 500 bp fragments using the Covaris LE220 ultrasonicator (Covaris), and size selected using AMPure XP SPRI beads (Beckman Coulter). The fragments were treated with end-repair, A-tailing, and ligation of Illumina compatible adapters (IDT, Inc) using the KAPA-Illumina library creation kit (KAPA Biosystems). The prepared libraries were quantified using KAPA Biosystem’s next-generation sequencing library qPCR kit and run on a Roche LightCycler 480 real-time PCR instrument. The quantified multiplexed libraries were pooled in sets of 10, and sequenced on the Illumina MiSeq sequencer using an indexed PE150 protocol with MiSeq V2 chemistry.

Resequencing data analysis was performed using QIAGEN Bioinformatics CLC Genomics Workbench (http://www.qiagenbioinformatics.com/products/clc-genomics-workbench), which incorporates a comprehensive set of analysis tools for Next-Generation Sequencing data. Paired-end reads were mapped to the reference genome [Genbank: CP002416] using the built-in Map Reads to Reference Tool. Further refinement of the reads mapping was performed by the Local Realignment Tool, which attempts to re-align each mapped read by exploiting the alignment information of other mapped reads. Realignment typically occurs in areas around insertions and deletions in the sample reads relative to the reference, resulting in more accurate mapping. Mapped reads were next analyzed by the built-in tools Basic Variant Detection Tool for putative SNV and MNV detection, and InDels and Structural Variants Tool for detection of putative structural variants. Variants occurring in <90% of the reads and variants that were identical to those of the parent Δhpt strain (e.g., due to errors in the reference sequence or mutations present at the beginning of strain construction) were filtered out. Raw data are available from the JGI Sequence Read Archive (JGI Project Id: 1053867 and 1053888).

Fermentation conditions

The inoculum for batch fermentation was prepared by growing the mutants in MTC medium overnight at 55 °C in an anaerobic chamber (COY Laboratory Products, Grass Lake, MI). The fermentation was carbon limited and carried out in 27 mL Balch tubes with 10 mL of MTC medium containing 5 g L⁻¹ of cellobiose as the carbon source, supplemented with 5 mM sodium acetate where noted, under a N₂ headspace sealed with butyl rubber stoppers. The tubes were inoculated with 0.5% v/v culture and incubated at 55 °C. The fermentation products were determined after 53 h of growth. Final cellobiose concentration was usually <0.5 mM, suggesting that fermentation activity was complete. Fermentations were performed at least two times with three independent biological replicates each. The “No Acetate” data were previously reported [11], which were generated simultaneously with the “Added Acetate” data reported here.

Analytical methods

Fermentation products, including ethanol, acetate, lactate, and formate, were analyzed on Breeze 2 High-Performance Liquid Chromatograph system (Waters Corp, Milford, MA) using an Aminex-HPX-87H column with a 5 mM sulfuric acid mobile phase. Sulfide was measured using an Orion silver ion selective electrode (Thermo Fisher Scientific, Waltham, MA) as previously described [17]. H₂ was measured using a 6850 Series II Gas Chromatograph (Agilent Technologies, Santa Clara, CA) using a thermal conductivity detector at 190 °C with a N₂ reference flow and a Carboxen 1010 PLOT (30.0 m × 530 µm I.D.; model Supelco 25467) column.

RNA isolation

The cells were grown to an OD of 0.3–0.4 in CTFUD medium, centrifuged at 4 °C for 5 min, and immediately flash frozen in liquid N₂. Pelleted cells were resuspended in 1.5 mL of TRIzol (Invitrogen, Carlsbad, CA). Glass beads (0.8 g of 0.1 mm glass beads; BioSpec Products, Bartlesville, OK) were added to the cell suspension and lysed with 3 × 20 s bead beating treatments at 6500 rpm in a Precellys 24 high-throughput tissue homogenizer (Bertin Technologies, Montigny-le-Bretonneux, France). Total RNA was purified using an RNeasy kit (Qiagen, Valencia, CA) with DNase I on-column treatment. RNA quantity was determined by NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific) and RNA quality was assessed with Agilent Bioanalyzer (Agilent Technologies). RNA (10 µg) was used as the template to generate ds-cDNA using Invitrogen ds-cDNA synthesis kit according to the manufacturer’s protocols (Invitrogen).

Microarray sample labeling, hybridization, scan, and statistical analysis of array data

The ds-cDNA was labeled, hybridized, and washed according to the NimbleGen protocols. Hybridizations were conducted using a 12-bay hybridization station (BioMicro Systems, Salt Lake City, UT) and the arrays dried using a Maui wash system (BioMicro Systems). Microarrays were scanned with a Surescan high-resolution DNA microarray scanner (5 µm) (Agilent Technologies), and the images were quantified using the NimbleScan software (Roche NimbleGen, Madison, WI). Raw data were log2 transformed and imported into the statistical analysis software JMP Genomics 6.0 (SAS Institute, Cary, NC). The data were normalized together using a single round of the LOESS normalization algorithm within JMP Genomics, and distribution analyses were conducted before and after normalization as a quality control step. An ANOVA was performed in JMP Genomics to determine differential expression levels between conditions using the False Discovery Rate (FDR) testing method (p < 0.05). Microarray data have been deposited in NCBI Gene Expression Omnibus (GEO) database under accession number (GSE54082). Data are average of three independent biological replicates.

Real-time quantitative-PCR (RT-qPCR) analysis

Microarray data were validated using real-time qPCR, as described previously [18]. Based on microarray hybridizations of C. thermocellum mutants, a set of 5 genes (Clo1313_0115, Clo1313_0147, Clo1313_0372, Clo1313_1559, and Clo1313_2243) representing a range of gene expression values was analyzed using qPCR from cDNA prepared for microarrays. Oligonucleotide sequences of the primers targeting the five genes selected for qPCR analysis are shown in Additional file 1: Table S1. Data are average of three independent biological replicates.

Proteomics analysis

Crude protein fraction of C. thermocellum cell pellet was processed and digested with trypsin, and peptides were eluted and analyzed over an 11-step MudPIT as described previously [19]. High mass accuracy was utilized for both MS1 (30 K resolution) and MS2 (7.5 K resolution; CID) scans (1 microscan each), with data-dependent acquisition settings as follows: 1 full scan followed by 20 MS/MS scans, isolation window = 2.1 m/z, dynamic exclusion window, duration, and max = −0.52/+ 1.02 m/z, 15 s, and 500, respectively. Peptides generated from C. thermocellum strain DSM1313 FASTA database concatenated with common contaminants and reversed entries were matched to MS/MS spectra using MyriMatch v. 2.1 [20]. Common sample prep-induced modifications, i.e., Cys + 57.0214 Da (alkylation; static), Met + 15.9949 Da (oxidation; dynamic), and N-terminus + 43.0058 Da (carbamylation; dynamic), were included in the search parameters. Matches were filtered and assembled using IDPicker v. 3.0 [21] using a minimum of two distinct peptides per protein identification and adjusting the minimum spectra count (SpC) per protein to achieve protein level FDRs < 5%, peptide-level FDRs < 1%, and PSM-level FDRs < 0.25%. Protein identifications with associated spectral counts (SpC) were tabulated, balanced, and normalized for semi-quantitative proteomics as previously described [18]. Normalized SpC (nSpC) were used as a proxy for protein abundance across individual samples. To assess differences in protein abundance, the top 99% of total assigned spectra (across all sample conditions) was log2-transformed and processed by ANOVA (JMP Genomics v. 4.1) to assess statistical significance. Proteins with significant differences in abundance (p value ≤0.01) and minimum of twofold change were identified and compared with transcriptomics data to identify proteins affected by the knock-out of hydrogenases as well as the addition of acetate to the culture. Data are average of three independent biological replicates. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://www.proteomexchange.org) via the PRIDE partner repository with the data set identifier PXD000777.

Addition of acetate increased growth rate of ΔhydG strains

Over the course of strain construction, we observed that growth of the C. thermocellum ΔhydG mutant was improved by the addition of exogenous acetate. Acetate was initially added based on the hypothesis that the addition of acetate could decrease flux through the acetate formation pathway [22], which could help prevent redox imbalances in the absence of H₂ production. Indeed, acetate was a component of the culture medium during C. thermocellum ΔhydG Δech strain construction, and we were only successful in deleting ech in C. thermocellum ΔhydG when the medium contained added acetate. To characterize this growth phenotype, wild type, ΔhydG and ΔhydG Δech were grown in medium supplemented with 5 mM acetate. Growth of the wild-type strain was unaffected by the addition of acetate (Fig. 1a; 0.26 and 0.3 h⁻¹, respectively, with nearly perfectly overlapping growth curves). However, in the presence of added acetate, the ∆hydG strain grew 33% more rapidly (0.16 h⁻¹) than without added acetate (0.12 h⁻¹) and attained a maximum optical density of 0.76 within 20 h (Fig. 1b), as opposed to 30 h in the absence of added acetate. While less pronounced, addition of acetate to the growth medium increased the growth rate of ∆hydG ∆ech as well, from 0.22 to 0.24 h⁻¹ (Fig. 1c). While the growth rate was improved, the fermentation product profile did not substantially change upon addition of acetate (Table 1), with an ethanol yield similar to that reported previously without added acetate [11]. Individual amino-acid concentrations are presented in Additional file 2: Figure S2 and are not substantially different from [11].

	Wild type		ΔhydG		ΔhydGΔech
	No acetatec	Added acetate	No acetatec	Added acetate	No acetatec	Added acetate
Ethanola	19.63 ± 2.98	22.88 ± 1.82	30.94 ± 1.61	31.43 ± 0.6	35.84 ± 1.0	36.06 ± 1.62
Acetateb	9.50 ± 1.9	8.90 ± 1.50	2.56 ± 0.07	1.10 ± 0.6	3.46 ± 0.36	1.76 ± 0.90
Lactate	3.84 ± 1.03	3.52 ± 1.56	0.11 ± 0.15	0.07 ± 0.2	0.12 ± 0.50	0.1 ± 0.98
Formate	6.54 ± 0.82	5.73 ± 1.17	4.33 ± 0.01	4.0 ± 0.01	6.8 ± 0.04	3.84 ± 0.03
Amino acids	4.1 ± 0.24	4.5 ± 0.09	2.83 ± 0.10	3.41 ± 0.04	2.14 ± 0.08	2.93 ± 0.13
Hydrogen	14.53 ± 3.36	14.04 ± 3.14	1.17 ± 0.37	0.96 ± 0.42	ND	ND

ND not detected

^aAll values are reported in mmol/L

^bNet acetate production

^cData previously reported [11]

To balance redox reactions, each molecule of synthesized acetate requires the concomitant production of two molecules of a more reduced compound (e.g., formate or H₂). Furthermore, there is a net production of NAD(P)H during microbial biomass formation from sugars [23]. Thus, one would expect more formate and H₂ to be produced than would be needed to balance the production of acetate alone. In wild type, 2 mM extra (formate + H₂) was synthesized relative to the amount that would be required to balance acetate production (Additional file 1: Table S2). The mutant strains, on the other hand, produced approximately zero additional (formate + H₂) beyond what is needed to balance acetate production, raising the possibility that an additional electron sink could be in use.

Sulfate metabolism genes are upregulated in ∆hydG with addition of acetate

To gain insight into the cellular response to elimination of H₂ formation and the addition of acetate, we explored changes in gene expression (Fig. 2; Additional file 3: Data set S1) and protein abundance (Additional file 4: Data set S2). A broad range of these gene expressions were also validated by RT-qPCR and a correlation coefficient of R ² = 0.98 between microarray and qPCR analysis confirmed differential expression (Additional file 2: Fig. S3). In the absence of added acetate, 204 genes were significantly differentially expressed in ΔhydG compared to wild type, including 79 increased and 125 decreased in ΔhydG (Additional file 5: Data set S3). Genes with increased expression in ΔhydG included genes in Clusters 7 and 8 (Fig. 2b) associated with CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) functions (Clo1313_2705-2713; Clo1313_2970-2976), glutamate biosynthesis (Clo1313_2035-2036), and cysteine metabolism (Clo1313_2325-2331). Genes with lowered expression included genes in Cluster 6 associated with sulfate uptake and assimilation (Clo1313_0115-0125) [24], and siroheme synthesis (Clo1313_0372-0375).

Surprisingly, addition of acetate to the fermentation medium had almost no effect on the transcriptomic profile of the wild-type strain, with only three genes significantly down regulated, including one iron sulfur cluster protein and two hypothetical genes (Fig. 2a), none of which were detected by proteomics (see below). Upon addition of acetate to the growth medium of ΔhydG, expression of 51 genes increased and five genes decreased. The two genomic regions of Cluster 6 in particular, consisting of total 27 genes, displayed dramatic increase in expression in the presence of added acetate. Proximal genes to these genomic loci were also upregulated in the presence of acetate and grouped in Cluster 1 (Fig. 2c). The first region from Clo1313_0108-0125 includes a predicted sulfate transport system and a complete sulfate assimilatory pathway, including genes for siroheme biosynthesis, which is a cofactor for sulfite reductase. The second genomic loci Clo1313_0372-0393 also included subunits of a putative pyruvate ferredoxin oxidoreductase, a putative carbon monoxide dehydrogenase, and two ABC transporters. Addition of acetate to ΔhydG Δech increased expression of cluster 6 as well, though not to the same degree as ΔhydG, since the basal (i.e., no added acetate) level of cluster 6 was not significantly different from wild type.

The upregulation of the Clo1313_0108-0125 and Clo1313_0372-0393 gene clusters in ΔhydG was further confirmed by proteomics using the same cultures harvested for transcriptomic analysis. About 1700 proteins were identified, out of which 21 were significantly upregulated and 37 downregulated in ΔhydG upon addition of acetate. Similar to the transcriptomics, acetate addition significantly increased the abundance of proteins from two clusters in ΔhydG strain: Clo1313_0110-0125 increased 78—709-fold and Clo1313_0373-0391 increased 2.7—408-fold (Additional file 4: Data set S2). Taken together, these data suggest that sulfur metabolism may be perturbed in C. thermocellum ΔhydG.

We also considered the possibility that differential protein acetylation could account for the difference in gene expression, but no differences in acetylation were detected in the proteomics data set (Additional file 4: Data set S4). The average number of detected acetylation events per sample was comparable between wild type (172), wild type with added acetate (155), ΔhydG (151), and ΔhydG with added acetate (152). Furthermore, no individual protein had statistically significant differences in acetylation.

Sulfur metabolism is altered in the C. thermocellum ∆hydG mutant

With the increased abundance of transcripts and proteins related to sulfur metabolism, we examined the effect of deletion of hydG on sulfate reduction and cysteine metabolism. The measurement of sulfate consumption from the medium was masked by excess sulfate ions produced by the thermal degradation of the MOPS buffer during incubation at 55 °C during fermentation (Additional file 2: Fig. S3). However, sulfate is reduced to sulfide prior to incorporation into cysteine. While the wild type released ~7.5 μM sulfide into the culture medium, ~18 μM sulfide was released by ΔhydG, with similar results when acetate was added to the medium (Fig. 3a). As C. thermocellum is known to divert a significant flux of carbon and electrons toward production of secreted amino acids, we further examined the abundance of amino acids in the supernatant. Cysteine is provided in the culture medium as a reductant to make the medium anaerobic. Interestingly, the wild-type strain removed ca. 1.6 mM of 2.4 mM cysteine in the medium. C. thermocellum ΔhydG and ΔhydGΔech, on the other hand, did not consume any cysteine. Acetate addition in these mutants did not substantially alter uptake of cysteine from the medium for any of the strains (Fig. 3b).

Genome resequencing of C. thermocellum ΔhydG and ΔhydGΔech

The improved growth of C. thermocellum ΔhydGΔech, even in the absence of added acetate, was surprising given that metabolism is more constrained by the deletion of the Ech hydrogenase. We, therefore, considered the possibility that additional, beneficial mutations occurred during strain construction, and that these hypothesized mutations could be influencing the phenotype. To address this, we resequenced the genomes of C. thermocellum ΔhydG and ΔhydGΔech (Additional file 1: Table S3). The parent Δhpt strain contained 27 mutations relative to the published genome sequence, which likely represents a combination of mutations that have accumulated during and since strain construction as well as possible errors in the published genome. The expected deletions in hydG and ech were identified in the respective strains, as was the previously identified adhE mutation. Nine additional mutations were identified in ΔhydG relative to the parent strain. In ΔhydGΔech, four additional mutations were identified relative to ΔhydG. Of these ΔhydGΔech mutations, two result in amino-acid changes, and both could be related to gene expression: a putative ribonuclease and a putative transcription factor.