Introduction

Diminishing petroleum reserves, unsustainable net rates of greenhouse gas emissions, and technological advancements have revived interest in the microbial formation of fuels and biochemicals from renewable resources (Stephanopoulos 2007; Bartley and Ronald 2009; Liu and Khosla 2010). Abundant hardwood, softwood, grasses, and agricultural residues contain up to 70 % cellulose and hemicellulose that are closely associated with lignin (Klinke et al. 2004). Pretreatment methods required to remove lignin and make the hemicellulose and cellulose fractions accessible for fermentation inevitably generate inhibitors, which negatively impact the microorganisms that generate the desired biochemical or fuel product (Sun and Cheng 2002; Klinke et al. 2004; Parawira and Tekere 2011).

Acetic acid (acetate) is one of the most abundant inhibitors generated through lignocellulosic pretreatment processes (Pettersen 1984; Sakai et al. 2007), and this acid affects the specific growth rate and biomass yield of microbes utilizing glucose or xylose (Luli and Strohl 1990; Roe et al. 1998, 2002; Takahashi et al. 1999). For example, 8 mM acetate (<0.5 g/L) decreased the growth rate of Escherichia coli on glucose by 50 % (Roe et al. 1998), and an increase in acetate from 34 to 68 mM led to a 60 % decrease rate of CO2 formation by Saccharomyces cerevisiae (Maesen and Lako 1952). A recent review details the tolerance of bacteria to acetic acid (Trĉek et al. 2015).

Several approaches have been proposed to remove acetate prior to microbial conversion, including adsorption with activated charcoal (Carvalho et al. 2006), liquid-liquid extraction (Um et al. 2011), and electrodialysis (Wong et al. 2009, 2010). A microbial method to remove acetate selectively from a mixture has also recently been reported (Lakshmanaswamy et al. 2011; Xia et al. 2012). This method is based on knocking out various sugar uptake systems to yield a strain which will consume only acetate and leave the sugars generated in lignocellulosic hydrolysates unaffected. Such microbial detoxification of acetate-containing lignocellulosic hydrolysates converts a portion of the acetate carbon into cellular metabolites that can be used in subsequent microbial conversion processes. The rate of acetate removal depends on the specific growth rate of the microorganism selected for acetate detoxification. Acetate “tolerance” (i.e., growth rate of cells when exposed to an elevated concentration) may be less important since under various continuous or semi-continuous modes of operation the microorganism is exposed to a very low acetate concentration regardless of its continuous consumption. Despite the broad interest in acetate “detoxification,” little is known about what genetic factors elevate microbial acetate utilization.

Continuous culture systems, such as chemostats, auxostats, and turbidostats, are frequently used for studying evolution and selection (Dykhuizen and Hartl 1983; Sauer 2001). For example, de Crécy et al. (2007) described a small volume turbidostat to adapt a mutant to higher growth rates, and more recently demonstrated the evolution of E. coli for thermophilic growth (Blaby et al. 2012). Using continuous cultures to select for increased growth is straightforward because under various operational modes slow growing members of the population are “washed out” of the system, selectively leaving members with the desired high growth rate phenotype.

The goal of this study was to compare strains of E. coli using acetate as the sole carbon source. The transcriptome of selected strains was compared to elucidate which genes were differentially expressed during growth on acetate. Additionally, the strain with the greatest growth rate was grown in a progressively increasing pseudo-steady-state (chemostat) culture for adaptive evolution, and the fastest growing isolate was sequenced to determine what mutations had occurred.

Materials and methods

Bacterial strains

Eighteen strains of E. coli were used in this study (Table 1), including many commonly used wild-type E. coli strains.

Table 1 E. coli strains used in this study
Full size table

Growth medium and conditions

A defined medium was used for all the shake flask (growth rate) experiments containing the following (per liter): 1.70 g citric acid, 13.30 g KH2PO4, 4.50 g (NH4)2HPO4, 1.2 g MgSO4·7H2O, 13 mg Zn(CH3COO)2·2H2O, 1.5 mg CuCl2·2H2O, 15 mg MnCl2·4H2O, 2.5 mg CoCl2·6H2O, 3.0 mg H3BO3, 2.5 mg Na2MoO4·2H2O, 100 mg Fe(III) citrate, 4.5 mg thiamine·HCl, 8.4 mg Na2(EDTA)·2H2O, and 85 mM acetate using Na(CH3COO)·3H2O. Additional experiments were conducted with the 85 mM lactic acid or 85 mM galactose instead of acetate. In each case, solutions were sterilized separately, combined, and the pH adjusted to 7.0 with a 20 % (w/v) NaOH solution. Cells were routinely cultured on Lysogeny Broth (LB) agar plates, transferred in 1:50 dilution to 50 mL defined medium in 250 mL shake flasks, grown to an optical density (OD) of 2.0, and then transferred again to two to three identical flasks used for growth rate determination and analysis. For each growth rate study, three to six flasks were incubated at 37 °C with an agitation of 350 RPM, and at least six samples examined for OD during the early exponential phase. Samples were stored at −20 °C for subsequent acetate analysis.

Adaptive evolution

Adaptive evolution using the fastest growing strain of E. coli was conducted in a chemostat with 800 mL defined medium in a 2.5-L bioreactor (Bioflo 2000, New Brunswick Scientific Co. Edison, NJ, USA). To prevent wall growth during the process, the glass vessel and the stainless steel parts of the system were coated with PEG-Silane. First, the fermenter was cleaned and hydroxyl groups were created on surfaces by exposure to hot piranha solution (7:3 1.2 M H2SO4:3.0 % H2O2). Silane was used to attach anti-biofouling polyethylene glycol (PEG) to the surface as previously described (Papra et al. 2001; Lan et al. 2005). The PEG-silane was coated onto the all surfaces by 3 days of exposure using 30 mM 2-[methoxy(polyethyleneoxy)propyl] trimethoxysilane (PEG-silane, Gelest SIM6492.7) in toluene. The coating on the stainless steel was confirmed by contact angle analysis, with an average contact angle of 34° indicating the surface presence of the hydrophilic PEG-silane. Air was sparged into the bioreactor at a flow rate of 1.0 L/min, the agitation was 500 rpm, the temperature was maintained at 37 °C, and the pH controlled at 7.0 using 20 % (w/v) NaOH and 20 % (v/v) H2SO4. The feed used for adaptive evolution contained 35 mM acetate, and all other nutrients at one quarter the concentrations used in the shake flask studies. When the dilution rate reached 0.50 h−1, the culture was plated onto LB plates, and three isolates designated MEC134, MEC135, MEC136 were selected for further study.

A spectinomycin-resistant mutant (rpsE) of MEC136 was obtained by plating 0.2 mL of a culture grown overnight on LB onto multiple LB agar plates containing 100 μg/mL spectinomycin. The rpsE mutant was obtained at a frequency of 1 in 109 cells. A P1 lysate was made from MEC136 and used to transfer the rpoA mutation back into ATCC8739 by selecting for spectinomycin resistance.

Analytical methods

The optical density at 600 nm (OD) (UV-650 spectrophotometer, Beckman Instruments, San Jose, CA) was used to monitor cell growth. Liquid chromatography with a refractive index detector and a Coregel 64-H ion-exclusion column (Transgenomic Ltd., Glasgow, UK) using a mobile phase of 4 mN H2SO4 was used for acetate analysis as described previously (Eiteman and Chastain 1997). For dry cell weight measurement, samples at different ODs were centrifuged (8400×g for 10 min), the pellets washed twice with DI water and dried on an aluminum weighing boat at 60 °C for 24 h.

Transcriptome analysis

Samples having OD600 of 0.5–0.6 were mixed with the RNA protect-reagent Bacteria Reagent (Qiagen Sciences Inc., Germantown, MD, USA) in the volumetric ratio of 1:2, incubated for 5 min at room temperature (20–25 °C) and centrifuged (5000×g for 10 min). Supernatant was discarded, and the pellet was stored at −20 °C. Two microliters of TE buffer containing 15 mg/mL lysozyme and 15 μL proteinase K was used as the lysis reagent. Addition of lysis reagent was followed by incubation in a water bath at 37 °C for 10 min and three freeze-thaw cycles (Lahtvee et al. 2009). Total RNA was extracted using QIAGEN RNeasy Mini kit and genomic DNA was removed by on-column DNase digestion using QIAGEN RNase-free DNase set. Single-stranded cDNA was synthesized from 5 μg total RNA in a 40 μL total reaction volume using superscript III first strand synthesis system (Invitrogen, Carlsbad, CA) following manufacturer’s instructions, including RNase H treatment. cDNA cleanup was carried out using QIAGEN Qiaquick PCR purification kit.

E. coli K-12 (4 × 72K format) microarrays were used in this study (Roche NimbleGen catalog #A6697-00-01). The labeling, hybridization, washing, and image scanning steps were performed at a core facility in Florida State University. The samples were labeled with cy3 random nonamers on a hybed on a 4 bay NimbleGen hybridization (Applied Biosystems, Foster City, CA). Scanning of the array was performed using a NimbleGen MS200 2 μm scanner (Roche NimbleGen, Madison, Wisconsin). Raw data extraction, image analysis, and normalization steps were performed using NimbleGen’s DEVA software. ArrayStar software was used for data analysis (DNASTAR, Madison, WI). For each strain, transcriptome analyses were performed in technical duplicates and on three independently grown shake flask cultures (n = 6).

Next generation sequencing (NGS) data analysis

Illumina paired end reads for the parent (ATCC8739) and the evolved strain (MEC136) were generated using the MiSeq platform by Research and Testing Laboratory (Lubbock, TX). The next generation sequencing data from this project has been made available through the NCBI BioProject/Sequence Read Archive data repository and can be accessed via the NCBI BioProject Accession ID: PRJNA314220 or SRA Accession ID: SRP071112. Individual raw data reads from this project can be accessed using SRA Accession ID: SRR3202030 (ATCC 8739) and SRA Accession ID SRR3190692 (MEC136). The parent strain library yielded a total of 2,373,109 sequences with 628,873,885 bases and the evolved strain library yielded a total of 2,373,109 sequences with 635,993,212 bases.

The sequencing reads were checked for quality using FastQC (Andrews 2014) and trimmed of adapters using trimmomatic (Bolger et al. 2014). The trimmed data for each strain was aligned to the E. coli ATCC8739 reference genome (NC_010468.1) using bowtie2 (Langmead and Salzberg 2012). Samtools was used to convert the resulting .sam files to .bam files, index the .bam data files, and generate variant statistics for all nucleotide positions based on an indexed genome file, which was also generated using samtools (Li et al. 2009). Bcftools was then used to generate the variant call files based on the variant statistics (Li et al. 2009). The indexed genome sequence file, gene annotation file for E. coli, indexed.bam file, and the variant call file were co-visualized using the Integrated Genome Viewer (Thorvaldsdóttir et al. 2013) to inspect the depth of coverage at variant call positions for each strain. Depth of coverage, which was evaluated using IGV, ranged from just over 100 to nearly 1900 for both libraries. The confirmed variant positions were then compared for the two strains and a single variant position that was present only in the evolved strain was identified.

Results

Growth rate on acetate

Eighteen strains were compared (three to six replicates each) for growth in defined medium containing acetate as the sole carbon source at a final concentration of 85 mM (Table 2). The C strain ATCC8739 had the greatest specific growth rate of 0.41 h−1, and SCS-1 had the lowest specific growth rate of 0.15 h−1. The five strains with the highest growth rate were all wild-type strains. A wide distribution of growth rates was observed, although 15 of the 18 showed growth rates between 0.26 and 0.37 h−1. Observed biomass yield for the strains was also calculated as the ratio of biomass generated to the acetate consumed, and no correlation was found between the biomass yield and specific growth rate (data not shown).

Table 2 Maximum specific growth rate (μ max) of E. coli strains (mean ± standard deviation of three to six independent studies)
Full size table

Transcriptional analysis

Three strains (ATCC8739, SMS-3-5, and BL21) were selected for transcriptional analysis using microarrays to understand why ATCC8739 grows fastest on acetate. These three strains differ in their maximum specific growth rate by approximately 0.05 h−1 from each other in the decreasing order ATCC8739, SMS-3-5, BL21. Seventy-one genes were 2-fold or more upregulated in ATCC8739 compared to SMS-3-5 and BL21 (Supplement, Table S1). The list includes prophage-related genes, phenylacetate metabolic genes, pseudo-genes, and transcriptional regulators. Ten genes (yqe and yge) belong to the type 3 secretion system pathogenicity island ETT2. The ETT2 gene cluster is present in a majority of E. coli strains (pathogenic or commensal) and has been subjected to varying degree of mutational attrition that render it unable to encode a functioning secretion system (Ren et al. 2004). Furthermore, over 90 % of these upregulated genes (66 genes) showed a greater expression level in SMS-3-5 compared to BL21. Thus for these genes, the expression level correlated with the maximum specific growth rate. One hundred twenty-eight genes were 2-fold or more downregulated in ATCC8739 compared to SMS-3-5 and BL21 (Supplement, Table S2). Of these, more than half (61 %, 77 genes) showed a lower expression with SMS-3-5 compared to BL21, and were correlated with maximum specific growth rate.

Adaptive evolution

The strain with the greatest μ MAX, ATCC8739, was evolved for growth rate by exposure to a 35 mM acetate feed at progressively greater dilution rates in a pseudo-steady-state chemostat. Specifically, after achieving steady-state at a growth rate of 0.25 h−1, the dilution rate was increased step-wise by 0.01 h−1 every 6 h, with the culture OD consistently maintained between 1.0 and 1.3. When the dilution rate reached 0.35 h−1, this step-wise increase was made every 8–10 h. After 185 h with a dilution rate of 0.47 h−1, the culture OD had decreased to about 0.4, and the effluent acetate concentration was about 17 mM. In order to ensure that the culture did not wash out, the dilution rate was reduced to 0.45 h−1 for about 10 h, then the step-wise increase in dilution rate continued until the dilution rate attained 0.50 h−1, at which point a sample was plated onto a Petri dish to isolate individual colonies. The experiment corresponded to about 450 generations.

Three isolates designated MEC134, MEC135, and MEC136 from the chemostat operating at 0.50 h−1 were selected to calculate (in triplicate) the maximum specific growth rate on a defined medium containing 85 mM acetate. Whereas the growth rate of ATCC8739 on acetate was determined to be 0.41 ± 0.01 h−1, the μ MAX of all three isolates was greater: MEC134 (0.50 ± 0.01 h−1), MEC135 (0.43 ± 0.01 h−1), and MEC136 (0.51 ± 0.01 h−1).

Transcriptional analysis of the evolved strain MEC136

Transcriptional analysis was performed on the evolved strain with the greatest growth rate on acetate, MEC136, and compared with ATCC8739 to elucidate what genes were differentially expressed in the two strains and might play a role in the elevated growth rate of MEC136 on acetate. Only one gene (ylcG) was upregulated 2-fold or more in MEC136 compared to ATCC8739, while eight genes were downregulated 2-fold or more in MEC136 (Table 3). Interestingly, ylcG was found to be downregulated in ATCC8739 compared to SMS-3-5 and BL21 (Supplement, Table S2), and would thus appear not to correlate with growth rate. Three genes were differentially expressed consistently between the four strains (Supplement, Table S2): mcrB, yihO, yjdQ. Each of these genes was downregulated in MEC136 compared to ATCC8739, and each of these genes was also downregulated in ATCC8739 relative to SMS-3-5 and BL21.

Table 3 Genes having two-fold or greater difference in expression level comparing MEC136 with ATCC8739
Full size table

Sequencing and reconstructing the evolved strain MEC136

Next generation sequencing as detailed in “Materials and methods” showed that MEC136 was caused by a single T to C transition mutation at base pair 454,696, which resulted in a S266P change in the α subunit of the RNA polymerase core enzyme coded by the rpoA gene and the presence of the mutation was confirmed by conventional sequencing. Interestingly, neither the evolved MEC134 nor MEC135 strain contained a mutation in rpoA. Because the rpoA gene is less than 4000 bp from the rpsE gene that codes for the 30S ribosomal subunit protein S5, which can be mutated to yield spectinomycin resistance, we expected that P1 lysates prepared from a spectinomycin-resistant mutant of MEC136 would allow us to transduce the rpoA mutation back into wild-type ATCC8739 thus recreating MEC136 at a frequency of about 90 %. The rpoA gene was sequenced in one such transductant (designated MEC537) and found to contain the S266P mutation. MEC537 was determined to have a specific growth rate of 0.52 ± 0.02 h−1 on a defined medium containing 85 mM acetate, statistically identical to the growth rate of MEC136 on the same medium.

Growth rates of the evolved MEC136 strain on other carbon sources

Because the altered RpoA protein could affect the utilization of other carbon sources besides acetate, we also determined the growth rates of MEC136 and MEC537, as well as the wild-type parental strain ATCC8739 and the wild-type strain MG1655 on another organic acid, lactate, and one carbohydrate, galactose (Fig. 1). For each of these three carbon sources, ATCC8739 and its derivatives, MEC136 and MEC537, attained a growth rate significantly greater than MG1655. MEC136 and MEC537 grew significantly faster than ATCC8739 on acetate. In contrast to growth on acetate, the growth rates observed for ATCC8739 were slightly greater than the growth rates observed for either MEC136 or MEC537 for both lactate and galactose. For all these carbon sources, the growth rate of MEC136 was indistinguishable from the growth rate of MEC537.

Fig. 1
figure 1

Comparison of the specific growth rate (h−1) of the four strains MG1655, ATCC8739, MEC136, and MEC537 on defined medium containing 85 mM acetate, lactate, or galactose as the sole carbon source. Data represent mean of three to six replicate experiments and standard error of the mean

Full size image

Discussion

Growth rate affects the rate of transcription or mRNA turnover, leading to changes in gene expression levels and protein synthesis. Pedersen et al. (1978) studied the expression of 140 individual E. coli proteins as a function of growth rates in five different media. The expression of nearly three quarters of these proteins (102) increased monotonically with growth rate. Similarly, Pease et al. (2002) demonstrated that the expression of pyridoxal 5′-phosphate synthesis genes in E. coli K-12 increased with growth rates, measured as the amount of mRNA synthesized and the rate of protein accumulation. These proteins are examples of growth rate-dependent regulation (Pearse and Wolf 1994; Tao et al. 1999).

In our study, 18 strains of E. coli were compared for growth on acetate. The transcriptional analysis of three strains was compared when each was growing at its own maximum specific growth rate on acetate. Of the many genes differentially expressed between ATCC8739, SMS-3-5, and BL21, none have any known relationship with acetate. A few of the genes can be attributed to changes in growth rate more generally.

For example, the folK gene encoding 7,8-dihydro-6-hydroxymethylpterin pyrophosphokinase (HPPK) was upregulated more than 2-fold in ATCC8739 compared to both SMS-3-5 and BL21. HPPK catalyzes the transfer of pyrophosphate from ATP to 7,8-dihydro-6-hydroxymethylpterin, leading to the biosynthesis of folate cofactors required for the syntheses of many essential nutrients including thymidylate, methionine, purines, serine, glycine, and pantothenate. Harvey (1973) observed that 0.1 μg/mL trimethoprim, an inhibitor of dihydrofolate reductase, reduced the growth rate of E. coli by 50 %. The highest intracellular folate levels in S. cerevisiae were observed during the highest growth rates in either batch or steady-state cultivation (Hjortmo et al. 2008), suggesting that greater demand for nucleotides and amino acids leads to a elevated folate pool. The increased expression of folK in ATCC8739 compared to SMS-3-5 and BL21 may thus merely reflect the increased pool for biosynthetic precursors in the strain growing at the highest growth rate.

Similarly, yafP was more than 10-fold downregulated in ATCC8739 compared to SMS-3-5 or BL21. YafP and PolIV proteins are hypothesized to form a toxin-antitoxin pair: in the absence of PolIV, YafP increased the cytotoxicity of 4-nitroquinoline-1-oxide (NQO) and nitrofurazone (Gutierrez et al. 2011). Lower expression of the yafP gene in ATCC8739 might thus merely indicate reduced cytotoxicity leading to an improved growth compared to the other strains. A few of the differentially expressed genes are related to transport functions. The glutamate, aspartate ABC transporter permease subunit (gltJ), the fructuronate transporter (gntP) and the L-idonate/D-gluconate transporter (idnT), a predicted N-acetylneuraminic acid transporter (yjhB), an uncharacterized sodium-dependent transporter (yidK), and an uncharacterized galactose-pentose-hexuronide transporter (yihO) were each downregulated in ATCC8739 compared with BL21 and SMS-3-5. Only yihO was also downregulated in MEC136. Although the function of YihO is not known, it is proposed to be involved in an ion-driven metabolite uptake system (Poolman et al. 1996). However, it is not clear how these genes might play a role in growth on acetate.

No difference was observed in the steady-state transcription of any metabolic genes associated with acetate consumption. Indeed, the absence of differentially expressed genes which can be specifically attributed to acetate suggests that high growth rate of ATCC8739 growing on acetate may simply be a general characteristic of this strain rather than particular to acetate as the sole carbon source. For example, neither genes for acetyl-CoA synthetase (acs) nor acetate kinase/phosphotransacetylase (ackA/pta) mediating the conversion of acetate to acetyl-CoA were differentially expressed. Genes expressing ICL (aceA), ICDH (icd), or ICDH kinase/phosphatase (aceK) were also not overexpressed. Similarly, no genes associated with acid stress response (Foster 2004) were differentially expressed, a result which may simply be a consequence of all three strains been exposed identically to acetate. The transcription profiles provide little insight as to why ATCC8739, BL21, and SMS-3-5 have different growth rates on acetate.

Using a chemostat, ATCC8739, the fastest growing strain on acetate that we tested, was adapted to yield MEC136, which had a growth rate of 0.51 h−1 on defined medium containing 85 mM acetate. ATCC8739 had a growth rate of 0.41 h−1 on the same media. While transcriptional analysis did not reveal any significant differences between MEC136 and ATCC8739, whole genomic sequencing showed that the MEC136 strain contained a single mutation, which resulted in a S266P change in RpoA, the α subunit of RNA polymerase.

RpoA facilitates the action of upstream-promoter (UP) elements as well as the activators and repressors that interact with these elements, including the well-characterized global transcriptional regulators, cAMP receptor protein (CRP), which regulates transcription in response to carbon availability, and fumarate and nitrate reductase (FNR), which regulates the transcriptional response to oxygen limitation (for a general review, see Grainger and Busby 2008). The C-terminal domain (αCTD) of RpoA is responsible for making contact with the UP elements and their regulators and several researchers have shown that amino acids 260–270 of RpoA are crucial for this interaction (Tao et al. 1995; Murakami et al. 1996; Yang et al. 1997; Savery et al. 1998, 2002).

The location of the single mutation that we observed at amino acid 266 of RpoA is within the key 260–270 amino acid region and suggests that this mutation affects the transcription of a gene or genes that increase the ability of MEC136 to grow on acetate as a sole carbon source. Interestingly, other studies have shown that RpoA is involved in the ability of E. coli to grow on acetate. Nègre et al. (1997) showed that the αCTD region of RpoA binds to the UP region of the promoter for the aceBAK operon, which encodes isocitrate lyase (aceB), malate synthase (aceA), and isocitrate dehydrogenase kinase/phosphorylase (aceK). Arnold et al. (2001) demonstrated that rpoA expression was repressed by 100 mM acetate. Beatty et al. (2003) found that acs, which codes for acetyl-coenzyme A synthetase, the high-affinity acetate-scavenging enzyme that converts acetate to acetyl-coenzyme, is activated by CRP via the aCTD region of RpoA. Collectively, these results support the likelihood that the rpoA S266P mutation that we observed in MEC136 is responsible for transcriptional changes in a gene or genes that enable E. coli to grow faster on acetate as a sole carbon source.