Clostridium thermocellum LL1210 pH homeostasis mechanisms informed by transcriptomics and metabolomics

C. thermocellum LL1210 chemostat growth and lower pH limits in MTC medium

Clostridium thermocellum LL1210 was cultured in duplicate chemostats fed with defined medium (MTC) containing 3 g/L cellobiose for 1058 h (Fig. 1, Additional file 1). After initial batch fermentation, steady-state chemostat growth was achieved for pH 7.0 at a dilution rate of 0.1 h⁻¹. The culture pH was then reduced to pH 6.5 at hour 189.5 and stead-state growth was maintained. Culture pH was checked using a regularly calibrated external probe and reported in Additional file 1. A pH adjustment to pH 6.10 resulted in substantial decreases in cell densities (~ 308 h, Fig. 1), along with lower ethanol concentrations and concomitant increases in cellobiose concentrations (Additional file 1), which suggests the specific growth rate of strain LL1210 was lower than the dilution rate of 0.1 h⁻¹. Medium flow was temporarily halted for 37.5 h and the culture pH set to 6.25, which restored cellobiose consumption, ethanol production and cell density and permitted medium flow to be resumed at dilution rate of 0.1 h⁻¹. After the equivalent of three fermentor vessel volumes, the chemostats were adjusted from pH 6.25 to 6.15, which coincided with lower culture densities. The mean culture turbidity measurements nearest to the omics sampling points, as measured by optical density (OD_{600 nm}) and their standard deviations were 0.655 (0.02), 0.651 (0.04), 0.533 (0.06), and 0.143 (0.05) for pH values pH 7, 6.5, 6.25 and 6.1, respectively. A change in dilution rate to 0.05 h⁻¹ permitted growth at pH 6.10 and 6.00. At pH 5.90, culture turbidity steadily declined; the dilution rate was reduced to 0.01 h⁻¹ and the OD continued to decline for 144 h in one chemostat while the other oscillated, indicative of stress, and eventually recovered. A single chemostat was able to maintain growth at pH 5.90 at a dilution rate of 0.01 h⁻¹, although culture turbidity showed a steady decline at pH 5.80 over a 133 h period and was unable to recover when the flow was stopped for 84 h. Thus, growth-limiting pH for this system is approximately pH 6.25 using strain LL1210 in a chemostat with a dilution rate of 0.1 h⁻¹ and slower dilution rates are required for steady-state growth below this pH.

Samples at dilution rate of 0.1 h⁻¹ were withdrawn from each bioreactor at pH 7 (actual pH = 6.98), pH 6.5 (actual pH = 6.48), pH 6.25 (actual pH = 6.24), and at apparent washout pH 6.15 (actual pH = 6.12) for analysis by transcriptomics and metabolomics (Fig. 1). Additional metabolomics samples were taken at pH 5.80 at a lower dilution rate (0.05 h⁻¹) to help elucidate the physiological changes.

pH homeostasis mechanisms inferred by differential gene expression

A greater number of differentially expressed genes were observed as the pH diverged further from pH 6.98. Using pH 6.98 as a reference, there were 80, 469, and 536 differentially expressed genes, respectively, for cells growing at pH 6.48, 6.24, and pH 6.12 (Additional file 2). To more easily categorize gene expression responses to pH changes, differentially expressed genes were clustered (Fig. 2a), and then gene clusters were checked for enrichment of gene ontology (GO) terms in comparison to the frequency of GO terms for all genes in the C. thermocellum genome by Fisher’s exact test (Fig. 2b). Cluster 1 contained only one gene (a 2-isopropylmalate synthase; Clo1313_0857) and cluster 3 only contained tRNA genes and were omitted from enriched function tests. Cluster 2 contained genes for rRNA, tRNA, tRNA modification, purine metabolism, and polyamine transport but no enrichment of GO terms was found. Flagella biosynthesis and chemotaxis, sulfate transport and reduction, nitrate transport and nitrogen fixation, glutamate dehydrogenase (GLDH, Clo1313_1847), and fatty acid biosynthesis (clusters 4, 5, 6) were downregulated, while F₀F₁-ATP synthase genes (cluster 7) were upregulated at pH 6.24. Diguanylate cyclase genes Clo1313_1339, Clo1313_1404 (with TRP repeats), Clo1313_1813 (with putative polar amino acid sensor domains and a phosphodiesterase domain), and Clo1313_2478 (putative response regulator) were all found in cluster 4. Genes for MotAB (Clo1313_0056-7) were part of cluster 6. MotAB system has been show to generate flagellar torque using proton-motive force [32]. Upon being shocked with pH 6.12, cells increased expression of genes for translational machinery (cluster 8), while gene expression decreased for carbohydrate binding and polysaccharide catabolic processing enzymes (cluster 9) including those for cellulosomal proteins CipA, OlpB, Orf2p, OlpA, CelD, CelH, CelR, CelV, CprA, CtManF, LecA, XynA, and XynD (Additional file 2). Cluster 9 also contained a malate shunt gene, phosphoenolpyruvate carboxykinase (PEPCK, Clo1313_0415) [14]. All but one of the glutamate synthase (GOGAT) genes (Clo1313_2032-3, Clo1313_2035-6) were also in cluster 9.

Previous studies have identified genes commonly expressed during acid stress by Gram-positive neutrophilic bacteria [30], which includes genes for amino acid decarboxylases/antiporter systems, DnaK, GroEL, ClpP protease, urease, F1F0-ATPase, arginine deiminase (ADI), agmatine deiminase (AgDi), and cyclopropane fatty acid (CFA) synthase. Most of these gene systems were not differentially expressed in C. thermocellum and genes for ADI, AgDi, or CFA are not annotated in the genome. Rather than increasing in expression at low pH, several decarboxylases were downregulated at pH 6.24 and 6.12, including aspartate 1-decarboxylase (Clo1313_1318), diaminopimelate decarboxylase (Clo1313_1540), orotidine 5′-phosphate decarboxylase (Clo1313_1266), putative oxaloacetate decarboxylase (Clo1313_1523), and a putative sodium pump decarboxylase γ-subunit (Clo1313_1525). The only decarboxylase that was upregulated at any pH was the S-adenosylmethionine decarboxylase proenzyme (Clo1313_1509), which produces the aminopropyl substrate need for synthesis of the polyamines spermidine and spermine. However, expression of the agmatinase and spermidine synthase (Clo1313_1529-1530) that synthesize the spermidine and spermine were either not differentially expressed or showed decreased expression at lower pH. Then again, genes for a putative spermidine/putrescine ABC transporter (Clo1313_1471-5) were upregulated at pH 6.24 and 6.12 (Additional file 2). Because a polyamine transporter and biosynthetic gene were upregulated, and polyamines can decrease membrane permeability and may protect cells in acid environments [33], their potential to augment C. thermocellum LL1210 growth at lower pH values was evaluated. Medium was supplemented with 100 μM of the polyamines putrescine, spermidine, spermine, or metabolic precursor arginine and the terminal pH was determined after 144 h. C. thermocellum LL1210 fermentations with and without amendments initiated at pH 7 and 6.75 had similar growth profiles (Additional file 3).

Downregulation of the GLDH (Clo1313_1847) and GOGAT (Clo1313_2032-3, Clo1313_2035-6) genes were observed at pH 6.12, but no significant expression changes were observed for the Type I glutamine synthetase (GS, Clo1313_2031), two Type III glutamine synthetases (Clo1313_1357 and Clo1313_2038), or the glutamate synthetase (Clo1313_1849) (Additional file 2). The type III glutamine synthase (Clo1313_2303) was significantly upregulated at pH 6.24 (Additional file 2). To examine the role of nitrogen metabolism in relation to pH and physiology, deletion strains for GLDH, GOGAT, GS, GS–GOGAT, and an annotated NifH (Clo1313_2339) were characterized using unbuffered (MOPS free) carbon-replete medium (10 g/L) (Additional file 4). Consistent with an earlier study [18], the parental strain had a higher specific growth rate compared to strain LL1210 (Additional file 4a). The final pH for the GS-mutant- and LL1210 cultures were about pH 6.3 and pH 5.9, respectively, whereas the range for the other strains was ~ pH 5.6–5.8 (Additional file 4b). In the unbuffered system, the GS mutant achieved the highest culture turbidity and produced the second highest titer of ethanol after strain LL1210 (2.0 and 2.7 g/L, respectively) (Additional file 4c). The higher final ethanol titers and pH values for strains GS and LL1210 coincided lower levels of residual substrate.

Genes for two orphan histidine kinases previously predicted to be part of the sporulation cascade in C. thermocellum [34], a pro-σ^E processing protease and BofA, inhibitor of the stage IV pro-σ^K processing protease SpoIVFB, were upregulated in C. thermocellum LL1210 at pH 6.24 compared to 6.98 (Additional file 5). To better understand the role of sporulation in the pH response of C. thermcellum, growth and sporulation of strains LL1210, asporogenous strains M1726 (DSM1313 ∆hpt ∆spo0A1) and M1725 (DSM1313 ∆hpt ∆spo0A1 ∆ldh ∆pta), and the DSM1313 ∆hpt parental strain were examined. C. thermocellum sporulation was inefficient, consistent with an earlier study [35] and sporulation-deficient mutant strains had neither a growth advantage or disadvantage at lower pH values for the conditions used in this study (Additional file 5).

The molecular chaperone GrpE (Clo1313_0932) and heat shock protein Hsp 20 (Clo1313_0678) were upregulated (2.0- and 2.2-fold) at pH 6.12, which were previously found to be upregulated after furfural shock and heat shock in C. thermocellum ATCC 27405 [36]. Hsp33 (Clo1313_2544) and MutS (Clo1313_1982) were upregulated at pH 6.24 (2.2- and 2.0-fold, respectively). Hsp33 was previously found to be more highly expressed in cellulose-adhered C. thermocellum ATCC 27405 compared to planktonic cells [4]. At pH 6.12, downregulated genes included one of the annotated Clp protease genes (Clo1313_1116) (0.48-fold), an annotated DNA repair proteins RadC (Clo1313_1465) (0.36-fold), and UvrD (Clo1313_2004) (0.31-fold) and a small subunit of an exodeoxyribonuclease VII (Clo1313_1389) (0.50-fold).

Metabolite profiles for low pH conditions

Extracellular amino acid accumulation at growth-limiting pH

At pH 6.24, only valine, alanine, and threonine extracellular concentrations were significantly higher than when pH was 6.98 (α = 0.05) (Additional file 7). Asparagine was significantly lower at pH 6.24 compared to 6.98 (α = 0.05). Extracellular glutamate was not significantly different at pH 6.24 compared to 6.98, although it was significantly lower at pH 6.48 compared to 6.24 (α = 0.05). When cells were shocked with pH 6.12 compared to when they were growing at pH 6.98, there were significantly lower concentrations of all detectable extracellular amino acids except methionine (α = 0.05). The amino acids that had the highest average extracellular concentrations in chemostats were valine (1.5 mM at pH 6.24), asparagine (147 μM at pH 6.98), and alanine (104 μM at pH 6.24). The highest average concentration of glutamate was 32 μM (near the limit of detection) at pH 6.24 (Additional file 7). An ABC-type branch-chain amino acid transporter gene (Clo1313_0822) was upregulated at pH 6.24 and 6.12 and ABC-type polar amino acid transport genes (Clo1313_0794-5) were upregulated at pH 6.24 (Additional file 2).

Metabolic inhibition at a cellular level

To gain a global perspective on metabolic inhibition at growth-limiting pH values, at and below pH 6.24, key mechanisms and responses for pH homeostasis were summarized (Fig. 3). A key feature of the physiological response at lower pH values was changes in expression for use and conservation of ATP. At pH 6.24, LL1210 increased expression of its F₁F₀-ATPase (Fig. 2). The F₁F₀-ATPase likely pumps protons to maintain pH homeostasis by ATP hydrolysis as is the case of many other neutrophiles [30]. Competing uses for ATP included valine transport, chaperones protecting cytoplasmic macromolecules from acid damage, and assimilation of ammonium by GS, GOGAT, and GLDH (Table 1, Additional file 2). ATP is also consumed by fatty acid biosynthesis (Table 1, Fig. 2), but there are many metabolites and downregulated genes for this aspect of metabolism. Therefore, it was omitted from Fig. 3, with the exception of biosynthesis pathways for valine and leucine, which are precursors for biosynthesis of iso-fatty acids. The accumulation of fatty acids and iso- derivatives may improve the capacity cytoplasmic membrane to repel protons, but adjustments to membrane lipid composition in response to an increase in proton concentrations are diverse among bacteria and not well understood [37]. Though fatty acid adjustments may have reduced membrane permeability to protons, downregulation of fatty acid biosynthesis genes suggests that ATP conservation was more important. Other gene expression changes that decrease ATP consumption include downregulation of sulfate and nitrate ABC transporter and reduction genes and 80 putative flagella biosynthesis and motility genes (listed in Additional file 2). Reduction of proton influx may also have been achieved through downregulation of gene expression for proton-channeling motility proteins MotAB (Clo1313_0056-7) at pH 6.24 (Fig. 2) [32]. The putative proton-pumping PP_i-ase (Clo1313_0823) was constitutively and highly expressed at all pH values. Accumulation of intercellular phosphate at pH 5.80 (Table 1) may have been due to PP_i-ase activity, ATP hydrolysis or other cellular activities. Another possibility is that collapse of the membrane gradient would have made PP_i hydrolysis to free phosphate much more thermodynamically favorable via the membrane-bound PPi-ase. Both reduction of proton influx through MotAB and proton-pumping through the putative PP_i-ase would allow more ATP conservation.

Bacterial strains, media and growth

Clostridium thermocellum strains DSM1313 ∆hpt [10], LL1210 [18], M1726 (LL376) [48], M1725 (LL375) [48], and AG1329 [47] were used in this study. The GLDH (Clo1313_1847), GOGAT (Clo1313_2032-6), GS–GOGAT (Clo1313_2031-6), and a nifH (Clo1313_2339) strain were deleted in a C. thermocellum DSM1313 ∆hpt background to create AG1327, AG1328, AG2068, and AG2078, respectively. Mutants were constructed using the procedures described in detail [49, 50]. Strains are available upon request. Primers used to construct strain AG2078 and other strains used in this study as well as SNPs identified from genome sequencing are provided in Additional file 8. Strains were stored frozen at − 80 °C in a modified version of DSM 122 medium, referred to as CTFUD [9], that contains 50 mM of MOPS (morpholinepropanesulfonic acid) and 10 mM of sodium citrate. Actively growing cultures were transferred twice from freezer stocks prior to batch fermentation experiments and inocula were 2% of the final volume. Cells were routinely cultured at 55 °C in either 50 ml (160 ml serum bottles) or 10 ml (26 ml Balch tubes) of MTC medium as described earlier [51], except that MOPS was omitted in this study. The LL1210 strain was also cultured in 1 ml volumes in a 48-well plate reader (Biotek Eon, Winooski, VT, USA) inside a Coy anaerobic chamber kept at 55 °C with growth data collected at OD_{600 nm} every 15 min. Growth in MTC-containing 100 μM of spermine (AC132750010, Fisher Scientific, Hampton, NH, USA), spermidine (S0266, Sigma-Aldrich, St. Louis, MO, USA), putrescine (AC112120250, Fisher Scientific) or arginine (A5006, Sigma-Aldrich) was assessed in plate reader assays.

Strain LL1210 was grown at 55 °C in duplicate 750 ml (total vessel capacity 1.3 L) chemostat cultures using water-jacketed BioFlo110 bioreactors (New Brunswick Scientific, Edison, NJ, USA), essentially as described previously in another C. thermocellum chemostat study [52], except that the medium contained 3 g/L of cellobiose as the carbon source. Briefly, the MTC medium was MOPS free and fed at a dilution rate of 0.1 h⁻¹, except where noted, for 1058 h. Temperature, pH, and agitation speed were monitored and controlled during fermentation using the BioCommand software (version 2.62). Gel-reference pH electrodes Mettler-Toledo (Woburn, MA) calibration, accuracy checks and pH control were maintained as described earlier [23], except that 1 N HCl and 3 N KOH were used for acid and base additions, respectively. Recalibration of the bioreactor probes was performed at 546 h after pH drifted by 0.08 and 0.12. Samples for omics were taken in 50 ml volumes; then harvested and stored as described in [52]. For transcriptomics, two samples were taken from different chemostats at pH 6.48 and 6.12, and for pH values 6.98 and 6.24 biological duplicates and two technical replicates separated by at least one change of vessel medium were collected. For intracellular metabolomics, samples were taken immediately after transcriptomic samples, and additional biological duplicate samples were taken at pH 5.8 for metabolomics only. Culture turbidity was measured taking optical density readings at 600 nm using a Genesys 20 spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, USA).

Cells were enumerated using a modified version of the direct cell counting procedure described in [53]. Briefly, 1 ml of culture samples were preserved for up to 1 week in 0.025 mg/ml (final concentration) paraformaldehyde (Sigma-Aldrich,). Cells were diluted 1:1000 with 0.9% (m/v) saline solution and stained with 25 nM 4′,6-diamidino-2-phenylindole (DAPI) (final concentration), immobilized to a 1.5 cm diameter section of a black MilliQ filter, and then imaged with a Zeiss Axio Imager.M1 with a Plan Apochromat 63×/1.4 Oil Immersion objective and X-Cite Series 120 Exfo fluorescence power source (Carl Zeiss, Germany). 8-bit images were imported into ImageJ (http://imagej.nih.gov/ij/) for manual or automated counting. Manual counting was performed with the cell counter plugin, and automated counting was performed with JAVA scripts (Additional file 1) which assumed cells were visible at a % Area (saturation threshold) of 1% (± 0.1%) with a particle size greater than 5 pixels after binary images had been processed with Watershed. Spherical morphologies were assumed to have a particle size range of 20–200 pixels and circularity of 0.85–1.00 for conservative estimates, and a particle size range of 50–100 pixels and circularity of 0.95–1.00 for liberal estimates. Automated parameters were validated with manual counts of 24 images. Stepwise calculations of standard deviation of progressively more sample counts were performed to ensure the count variance plateaued. Highly refractive mature spores were also observed by light microscopy after heat fixing to 20 μl of stationary cells to glass slides.

Substrate and metabolite analyses

Extracellular fermentation product samples for high-performance liquid chromatography (HPLC) were collected and processed as described in [51]. HPLC data were generated using a Shimadzu Prominence LC-20A Series system (Shimadzu Scientific Instruments, Columbia, MD) fitted with a refractive index detector (model RID-20A) and an Aminex HPX 87H HPLC column (300 × 7.8 mm) (Bio-Rad, Hercules, Dallas, TX, USA) against known standards for cellobiose, glucose, ethanol, acetate, lactate, and formate. Intracellular metabolite fold changes and select metabolite concentrations were determined using a 5975C inert XL gas chromatography–mass spectrometer (Agilent, Santa Clara, CA) [22]. Extracellular amino acid concentrations were quantified by an Aracus Amino Acid Analyzer (membraPure, Berlin, Germany) [47]. Two-tailed paired T tests were performed to determine if amino acid concentrations were significantly different. Homoscedasticity was confirmed by performing Breusch–Pagan and Abridged White’s tests prior to performing T tests.

RNA isolation, cDNA library preparation, sequencing and RNA-seq analysis

Total RNA was isolated from 50 ml of chemostat sample using the procedures described in [54], and quantity and quality were determined using a Nanodrop instrument (ThermoFisher Scientific Waltham, MA). Depletion, library preparation and sequencing were completed at the Joint Genome Institute (JGI, Walnut Creek, CA). In short, 100 ng of total RNA was depleted of ribosomal material using a Ribo-Zero (TM) rRNA Removal Kit (Epicentre). The rRNA depleted RNA was fragmented and reversed transcribed using random hexamers and SSII (Invitrogen) followed by second strand synthesis. Stranded cDNA libraries were generated using the Illumina Truseq Stranded RNA LT kit, following the manufactures protocols, using ten cycles of PCR. qPCR was used to determine the concentration of the libraries. Libraries were pooled and sequenced on an Illumina HiSeq 2500 platform with version 2 chemistry in a 2 × 151 bp configuration (JGI, Walnut Creek, CA; Project ID 503106). Fastq files were verified for integrity by checksum and read quality checked using FastQC [55]. RNAseq reads were trimmed using Trimmomatic MAXINFO method with target length and strictness parameters set to 40 and 0.8, respectively [56]. Reads were mapped to C. thermocellum DSM1313 (NC_017304, last modified 2015/07/30) using Bowtie2 with the same parameters as the very-sensitive preset option expect the number of mismatches was set to 1 [57, 58], and reads were counted with HTSeq [59]. Principle component analysis (PCA) was used to check for sample variation using JMP Genomics 8 (SAS Institute, Cary, NC, USA). Differential expression analysis was performed with DESeq 2 [60]. Genes were considered significantly differential expressed when their adjusted p values were less than or equal to 0.05 and log2-fold change values were ≥ 1 or ≤ 1. Raw RNA-seq data have been deposited in NCBI SRA under project number PRJNA395926. PCA plots, raw counts, RPKM normalized gene counts, and differential expression are provided in Additional file 2. Genes found to be differentially expressed at pH 6.48, 6.23, and pH 6.12 compared to pH 6.98 were clustered by their expression patterns using the K-means cluster procedure in JMP Genomics 8. Fisher’s exact test was then performed on 7 of the 9 K-means clusters to identify enriched GO terms. Prior to this analysis, GO terms were identified for the entire C. thermocellum DSM1313 genome using the standard workflow in Blast2GO 4.1.5 [61].