Isolation of σ⁷⁰ mutants

To obtain σ⁷⁰ mutants with high n-butanol tolerance, random mutagenesis was performed to construct a rpoD mutant library of around 10⁶. In the first round of screening, 483 mutants were selected from the mutagenesis library on agar plate containing 0.5 % (v/v) butanol. They were further inoculated into 24-well plate under butanol pressure. Among them, ten mutants with OD₆₆₀ of 1.3–2.0 were selected due to their advantageous growth under 0.5 % (v/v) butanol than other mutants (OD₆₆₀ < 1) (Fig. 1a). Then these mutants were cultured in medium containing 0.5–1.5 % (v/v) n-butanol (0.1 % gradient increasing). One mutant exhibited the highest tolerance (0.76 OD₆₆₀) at 1.2 % (v/v) n-butanol and was designated as D3, which was confirmed with two amino acid substitutions (I41L, P97Q).

To further improve the butanol tolerance, the second round of random mutagenesis was performed to construct a variant library using D3 as a template. In the second round of screening, eight mutants exhibited higher butanol tolerance (OD₆₆₀ > 2.0) than others (Fig. 1b). In further screening under 0.5–1.5 % (v/v) butanol (0.1 % gradient increasing), one mutant B8 with three substitutions (I41L, E57D, P97Q) exhibited excellent growth in the presence of 1.2 % (v/v) butanol, reaching 1.432 OD₆₆₀ compared with 0.778 of D3 and 0.247 of WT (Additional file 1: Figure S1). Then, mutants B8, D3 and WT were further cultured under higher n-butanol concentration (1.2–2.2 %) to assess their maximum butanol tolerance (Fig. 2a). The cell growth of WT was completely inhibited when butanol reached over 1.2 % (v/v), and mutant D3 could hardly grow in the presence of 1.6 % (v/v) butanol. Nevertheless, for mutant B8, OD₆₆₀ reached 0.372 after 8 h incubation with 2.0 % (v/v) butanol, suggesting a much higher butanol tolerance than D3 and WT, while no cell growth was observed for B8 at over 2.0 % (v/v) n-butanol. Finally, B8 was selected as the best candidate for further study.

Characterization of butanol-tolerant mutants

Cell morphology

The cell morphology of microorganisms can adapt to the environmental changes [26]. Both WT and B8 grew in the absence or presence of butanol. In transmission electron microscope observation (Fig. 3), mutant B8 cells were significantly longer than WT cells after butanol treatment. In the absence of n-butanol, the cell size was (1.72 ± 0.17) × 0.73 μm for WT (Fig. 3a) and (1.89 ± 0.22) × 0.87 μm for B8 (Fig. 3c), respectively. After 0.8 % (v/v) n-butanol treatment, the cell size of WT and B8 shifted to (2.18 ± 0.25) × 0.53 μm (Fig. 3b) and (2.94 ± 0.39) × 0.52 μm, respectively (Fig. 3d), based on the measurement of around 100 cells. The changes in cell size were confirmed to be statistically significant (p < 0.05). Although the cell morphology was observed based on 100 cells per strain, the cell sizes could also be affected by reasons unrelated to the tolerance effects. Notably, B8 cells were longer than WT when subjected to n-butanol stress. A similar observation on the cell size of ethanol-adapted Saccharomyces cerevisiae has also been reported [27]. Neumann and coworkers reported that the microorganism cells of larger size could be more advantageous over the smaller ones under stress conditions, because the ratio of surface area to volume (S:V = 4πr²:4/3πr³) of larger cells is relatively lower than that of the smaller ones [28]. Similarly, B8 cells with smaller S:V value exhibited greater butanol tolerance than WT. Meanwhile, it was observed that the cytoplasm of B8 shrank in the presence of butanol. Occasionally, invaginated bodies (Fig. 3d) appeared in B8 and the inner membrane of B8 cells was not broken or leaky under butanol treatment. The huddling cytoplasm in the inner membrane could be a self-protection mechanism which protects the cells from damage due to the solvent [29]. Aono and coworkers also reported a similar phenomenon in E. coli cell membrane or cytoplasm in the presence of n-hexane or cyclohexane [30]. Overall, butanol-tolerant mutant B8 demonstrated a stronger stress response capability. Our results suggest that microbial cells could adapt to solvent stress via morphology change.

DNA microarrays and data analysis

Organic solvent tolerance-related genes are usually correlated in a precise regulatory network. The global gene expression profile of mutant B8 and WT was analyzed by DNA microarray. Data were analyzed by Gene Spring Software (Santa Clara, CA, USA) to identify differentially expressed genes. For correlation of gene expression difference under solvent challenge, an over twofold change in gene expression was required with a p value <0.05 [31]. The results show that 329 genes (including 197 up-regulated and 132 down-regulated) exhibited differential expression (p < 0.05; FC ≥ 2) between B8 and WT under 0.8 % (v/v) butanol stress (Additional file 2: Figure S2). Supporting information on the detailed description of differentially expressed genes is provided in Additional file 3: Table S3. Furthermore, genes showing significant difference in transcription level were selected for clustering analysis, which helps to understand the relationships and discrepancy of samples more comprehensively and intuitively (Additional file 4: Figure S3). The same types of genes were gathered in a cluster with similar biological functions.

Identification of genes associated with butanol tolerance

Analysis of membrane-related down-regulated genes

Many genes have been confirmed to be related to the organic solvent tolerance of the E. coli strain. For example, overexpression of marA could enhance the function of AcrAB-TolC efflux pump, so that the toxic substances could be extruded more efficiently [21, 39]. Escherichia coli mutant △lon (cell envelop related gene) showed enhanced solvent tolerance level [40]. An E. coli mutant with gene disruptions in both proV and marR showed increased solvent tolerance due to their functions in regulation of osmotic pressure [22]. In this study, seven down-regulated genes (yibT, yghW, ymgI, yhcN, yrbL, ECs4086, ybjC) were selected for further study. They were rarely investigated previously and exhibited over sixfold changes in microarray analysis. All seven genes were annotated as predicted protein or hypothetical protein. The down-regulation folds of these genes are 46.72, 13.00, 16.02, 14.74, 14.07, 12.91 and 8.73, respectively. The knockout strains were constructed for further investigation. Among them, △yibT, △yghW and △ybjC exhibited higher n-butanol tolerance, while other knockouts showed similar growth to the control (Fig. 4). So far, there has been no report on the solvent tolerance-related functions of yghW and yibT. It was, however, noticed that these two predicted proteins are closely related to membrane proteins in stitch networks (http://www.stitch.embl.de/). Stitch is a database which contains interaction information for connected proteins and chemicals. It allows querying by genes name and metabolic pathways [41]. In stitch networks, gnsA and gnsB are predicted regulators of phosphatidylethanolamine synthesis and unsaturated fatty acids regulatory proteins, respectively, which are both linked to yibT. Gene yghW is linked to the inner membrane protein encoding genes ybjO and ybjM, as well as genes encoding the predicted lipoprotein or conserved protein in stitch networks. It is therefore presumed that modulation of membrane fatty acid compositions is one possible defense mechanism of yibT and yghW. It has been reported that the properties of membrane fatty acids, such as fatty acid chain length, branching pattern and unsaturation degree of fatty acids, could change when exposed to environment stress [42–45]. In this study, phospholipids of △yghW and △yibT were extracted and analyzed. Our results show that the main components of membrane fatty acids are C16:1, C16:0, C18:0 and C18:1, which account for over 70 % of the total fatty acids in all strains (Table 2) and are responsible for the integrity and fluidity of the membrane [46]. It was noticed that the proportion of unsaturated fatty acid (UFA) in total fatty acids was increased in both knockouts, especially palmitoleic acid (C16:1) and oleic acid (C18:1). Oleic acid (C18:1) has been regarded as the most important UFA in counteracting the toxic effects of the solvent by modulating plasma membrane fluidity [47]. Additionally, palmitoleic acid (C16:1) could influence the rigidity and integrity of membrane lipid bilayer [46, 48]. It is speculated that a higher proportion of oleic acid and palmitoleic acid in △yghW and △yibT contributes to a lower membrane fluidity, higher rigidity and membrane integrity, which might be a compensatory advantage of the membrane challenged by butanol. As a result of changes in membrane fatty acid composition, other physicochemical properties of the membrane, such as proton permeability and lipid–protein interactions could also be affected [49, 50]. Our findings suggest that the butanol tolerance mechanism of △yghW and △yibT are related to membrane fatty acid composition.

Major fatty acids	Fatty acid (%)a			Difference △yibT	p valueb	Difference △yghW	p valueb
	JM109	△yibT	△yghW
C11:0	0.20 ± 0.04	0.00 ± 0.00	0.13 ± 0.04	−0.20	***	−0.08	**
C12:0	0.69 ± 0.44	0.33 ± 0.04	0.45 ± 0.12	−0.36	*	−0.25	*
C13:0	0.00 ± 0.00	0.05 ± 0.07	0.07 ± 0.09	0.05	*	0.07	*
C14:0	2.24 ± 0.96	1.42 ± 0.27	1.41 ± 0.13	−0.82	*	−0.84	*
C15:0	0.65 ± 0.52	2.45 ± 0.05	3.03 ± 0.87	1.80	***	2.38	***
C15:1	0.00 ± 0.00	0.12 ± 0.16	0.31 ± 0.07	0.12	*	0.31	***
C16:0	26.96 ± 1.32	18.86 ± 0.33	15.48 ± 1.18	−8.10	***	−11.49	***
C16:1	9.06 ± 1.08	10.90 ± 1.05	11.08 ± 1.52	1.84	**	2.02	*
C17:0	1.52 ± 0.42	6.18 ± 0.77	5.67 ± 0.37	4.66	***	4.15	***
C17:1	4.69 ± 0.62	4.80 ± 0.39	6.35 ± 0.88	0.11	*	1.66	**
C18:0	9.56 ± 0.95	6.49 ± 1.44	5.36 ± 0.31	−3.07	**	−4.20	***
C18:1	21.03 ± 1.36	25.23 ± 0.73	27.05 ± 0.37	4.20	***	6.02	***
C18:2	0.03 ± 0.04	0.06 ± 0.08	0.42 ± 0.29	0.03	*	0.39	**
C18:3	0.38 ± 0.54	0.00 ± 0.00	0.04 ± 0.06	−0.38	*	−0.34	*
C19:1	1.22 ± 0.17	0.08 ± 0.11	0.07 ± 0.09	−1.14	***	−1.16	***
C20:0	0.23 ± 0.01	0.08 ± 0.11	0.16 ± 0.06	−0.15	**	−0.08	*
C20:1	1.07 ± 0.62	0.00 ± 0.00	0.00 ± 0.00	−1.07	**	−1.07	**
C22:1	0.13 ± 0.18	0.06 ± 0.08	0.05 ± 0.07	−0.07	*	−0.08	*
C25:0	0.06 ± 0.08	0.07 ± 0.09	0.00 ± 0.00	0.01	*	−0.06	*
SFA	42.09 ± 0.68	35.91 ± 0.08	31.73 ± 1.92	−6.18	***	−10.37	***
UFA	37.59 ± 0.51	41.22 ± 2.28	45.36 ± 3.34	3.63	**	7.76	***

^aThe values represent percentages of total fatty acids and are means ± standard deviations from three independent experiments

^bStatistical significance between knockout and control strain (E.coli JM109)

* p > 0.05, ** 0.01 < p < 0.05, *** p < 0.01

Moreover, the hydrophobicity, acidity and alkalinity of cell surface are important properties related to the solvent tolerance [51–53]. Using the MATS method, the adsorptions of three strains (△yghW, △yibT and JM109) to organic solvents with different hydrophobicities such as chloroform, hexadecane, ethyl acetate and decane were determined (Additional file 5: Figure S4). Our results show that the adsorptions of △yghW, △yibT and JM109 to chloroform were 1.89-, 1.86- and 1.61-fold of those to hexadecane, indicating that the alkaline strength of the knockout strains was higher than that of WT. The alkaline strength of the cell surface has been reported to be proportional to the adsorption ratio of chloroform adsorption to hexadecane [51]. Similarly, the adsorptions of △yghW, △yibT and JM109 to ethyl acetate were 1.81-, 1.79- and 1.65-fold of those to decane. It has also been reported that the acidity strength of cell surface was proportional to the adsorption ratio of ethyl acetate to decane [51]. It is supposed that the increase of surface acidity and alkalinity is due to the presence of proteins and charged chemical groups on the cell surface, such as PO₄ ²⁻ and COO⁻ [54], which may assist strains to counteract extracellular toxic substances. Additionally, the adhesion to hexadecane and decane reflects the hydrophobicity of the cell surface. In Additional file 5: Figure S4, the adhesion of △yghW and △yibT cells to hexadecane and decane was weaker compared with that of the control, suggesting that the surface hydrophobicity of the knockouts was higher than that of the control. The contact angle measurement (CAM) was also conducted to determine the surface hydrophobicity. Our results show that the contact angle of the control, △yghW and △yibT were 21.17 ± 1.78°, 36.25 ± 2.13° and 34.65 ± 2.04°, respectively (Additional file 6: Figure S5). Consistent with the results of MATS, CAM assay represents a higher surface hydrophobicity of △yghW and △yibT. Other studies suggest that the expansion of surface hydrophobic region could promote the interactions between the phospholipids and embedded proteins and bonds between cation and electronegative phospholipids in the membrane [55, 56]. It is therefore rational to conjecture that the cell surface of △yghW and △yibT is less permeable, which could help to prevent the intruding of toxicity compounds.

The gene (ybjC) encoding a predicted inner membrane protein was also investigated. To verify the location of YbjC and its tolerance-related mechanism, YbjC–GFP (green fluorescent protein) fusion protein was constructed. As shown in Fig. 5a, a clear fluorescence signal surrounding the membrane region was detected in E. coli cells expressing fusion protein YbjC–GFP. Differently, the entire cells were filled with green fluorescence when GFP was expressed alone (Fig. 5b). Here, a known membrane protein YidC fused with GFP was expressed as the positive control (Fig. 5c), and the microscopy pattern of YbjC–GFP fusion was similar to YidC–GFP. This further suggests that YbjC is a membrane protein. Unfortunately, the function of YidC related to the butanol tolerance mechanism is not clear yet.

Strains, plasmids and culture conditions

Escherichia coli JM109 was used as the host strain. Gene deletion strains were generated by Red-mediated recombination approach and overexpression strains were generated using pQE80L as an expression vector. Strains and plasmids are listed in Additional file 10: Table S1. Primers used in this study are listed in Additional file 11: Table S2. Plasmid pQE80L was purchased from Qiagen GmbH (Hilden, Germany). pHACM-rpoD ^WT was presented as a kind gift of Dr. Huimin Yu (Tsinghua University, China). Restriction enzymes and PrimeSTAR^®HS DNA Polymerase were purchased from Takara (Tokyo, Japan). The dam-methylated DNA-specific restriction enzyme DpnI was purchased from New England Biolabs (Ipswich, MA, USA). GeneMorph II Random Mutagenesis Kit was obtained from Stratagene (La Jolla, CA, USA).

All strains were grown in Luria–Bertani (LB) medium (tryptone 10 g/L, yeast extract 5 g/L, NaCl 10 g/L) at 37 °C, 120 rpm. Butanol was added as specified in each experiment. When necessary, antibiotics chloramphenicol (34 μg/mL), ampicillin (100 μg/mL) and kanamycin (50 μg/mL) were added to the media. For gene overexpression, 0.2 mM IPTG was added to the medium at around 0.3 OD₆₆₀.

Construction of random mutagenesis library

Using plasmid pHACM-rpoD ^WT as the template, random mutagenesis was performed by the GenemorphII Random Mutagenesis kit (Stratagene) with a mutation rate of approximately 4.5–9 mutations/kb. The error-prone PCR program was set as follows: 5 min at 95 °C, 30 cycles of 95 °C for 30 s, 57 °C for 1 min, followed by 72 °C for 2 min, and 10 min at 72 °C. After running the whole plasmid PCR, the PCR product mixture was digested with DpnI and then transformed into E. coli JM109. Escherichia coli transformants were plated on LB agar plates containing 34 μg/mL chloramphenicol and incubated at 37 °C. Then the colonies were scraped off to create a σ⁷⁰ mutant library for further butanol-tolerant phenotype selection. The total mutant library size was approximately 10⁶.

Phenotype selection of n-butanol-tolerant mutants

Solvent shock treatment

The overnight cell culture was inoculated (1 %) into a fresh LB medium. When OD₆₆₀ reached 0.8, 3 % (v/v) n-butanol was added into the culture. After incubation at 37 °C for 1.5 h, the cultures were serially diluted for 10⁵, 10⁴, 10³, 10² and 10-fold with aseptic water. Then, 10 μL of the each diluted culture was spotted onto LB/agar plates and further incubated at 37 °C overnight.

Cell morphology

The σ⁷⁰ mutant B8 and WT strains were cultured overnight and inoculated (1 %) into fresh LB/Cm liquid medium for incubation at 37 °C and 120 rpm for 8 h with or without 0.8 % (v/v) n-butanol. The cells were diluted and spread on LB/Cm agar plates. Single colonies were treated as described in the literature and observed using transmission electron microscope [30]. Briefly, single colonies were picked and fixed by immersion in 2.5 % glutaraldehyde at 4 °C for 3 h. The cell suspension was mixed once every half-hour. Then, cells were washed four times with 0.1 M phosphate buffer (pH 7.2), and the samples were diluted for cell morphology observation under Hitachi-H7650 transmission electron microscopy (Japan). The average cellular size of WT and B8 was counted on the electron microscope based on around 100 cells.

DNA microarrays

Escherichia coli strains harboring σ⁷⁰ mutant B8 and WT were cultured overnight and inoculated (1 %) into fresh medium. n-Butanol (0.8 %, v/v) was added at 0.8 OD₆₆₀ and further incubated for 1.5 h. Cells were harvested by centrifugation (8800 g, 4 °C). Total RNA was extracted using Qiagen RNeasy kit (Hilden, Germany) following the manufacturer’s instructions. Qualified total RNA was further purified by Qiagen RNeasy mini kit and RNase-Free DNase Set. Three biological replicates of RNA samples were stored in dry ice and subjected to further DNA microarray analysis. The microarray service was provided by Shanghai Biotechnology Co., Ltd. (Shanghai, China) using Agilent SurePrint E. coli 8 × 15 K slides, and the quality and integrity of RNA were examined before analysis.

Slides were scanned by Agilent Microarray Scanner (Santa Clara, CA, USA), and data were extracted with Agilent Feature Extraction software 10.7. Raw data were normalized by Quantile algorithm, Gene Spring Software 11.0 (Santa Clara, CA, USA). Differentially expressed genes were identified using the rank product method [61]. The MeV (TM4) software was used for clustering and other expression profile analysis [62]. The related metabolic pathway of differentially expressed genes was analyzed using Kyoto Encyclopedia of Gene and Genomics (KEGG) database [63]. The microarray data have been deposited at the gene expression omnibus (GEO) under the accession number GSE79305.

Assay of n-butanol sensitivity of knockout and overexpression strains

Gene knockout (for down-regulated genes) and overexpression (for up-regulated genes) strains were constructed based on the microarrays results. Spot assay as a tolerance confirmation method was conducted for knockout strains. n-Butanol (0.8 %, v/v) was added when the cell density reached 0.8 OD₆₆₀, then the cells were further cultured for 1.5 h. Cell culture was serially diluted at a tenfold gradient, and 10 μL of the diluted culture was spotted onto LB/agar plates for incubation at 37 °C overnight. For overexpression strains, the overnight culture was inoculated (1 %) into fresh medium and cultured at 37 °C and 120 rpm, and 0.2 mM IPTG was added when OD₆₆₀ reached 0.3. Then, 0.8 % (v/v) butanol was added at 0.8 OD₆₆₀ for further incubation at 30 °C for 8 h. The growth of the strains was monitored by measuring cell density at OD₆₆₀.

Lipid extraction

Lipid extraction was performed as described by Bligh [64]. Briefly, the stationary phase cells of E. coli JM109, △yibT and △yghW were collected by centrifugation and washed with 10 mM, pH 7.4 sodium phosphate buffer. The cell pellets (0.3 g wet weight) were suspended in a mixture consisting of buffer, methanol and chloroform in the proportion of 0.8:2:1 (v/v) and incubated at room temperature for 2 h, with brief shaking every half-hour. Then, the extract was centrifuged at 4 °C after diluting with methanol and chloroform (1:1, v/v). The lower chloroform phase (containing lipids) was added into the methanol solution containing 5 % (v/v) H₂SO₄. After 2 h of methylation at 70 °C, the mixture was cooled to room temperature and extracted three times with pentane. Then, the samples were air dried to remove pentane. Finally, the samples were reconstituted with n-hexane for GC–MS analysis.

Determination of cell surface hydrophobicity

MATS (microbial adhesion to solvents) analysis was performed following the MATH method developed by Bellon-Fontaine with slight modification [53, 65]. Briefly, the strains were cultured in LB medium, and butanol (0.8 %, v/v) was added when OD₆₆₀ reached 0.8. Cells were further incubated for 12 h at 37 °C and were collected by centrifugation (8800 g and 10 min at 4 °C), then washed with 100 mM, pH 6.0 potassium phosphate buffer and centrifuged again. The cell concentration was then adjusted to around 1.0 OD₄₀₀ (A₀) using the same potassium phosphate buffer. The cell suspension (4.8 mL) was mixed with 0.8 mL of chloroform, hexadecane, ethyl acetate and decane, respectively. The two-phase mixture was mixed by vortexing for 90 s and then incubated statically at room temperature for 15 min. The aqueous phase was removed and OD₄₀₀ was measured (A) to calculate the adhesion ratio [Adhesion % = (1 − A/A₀) × 100 %]. The above experiment was repeated for three times.

Contact angle measurements (CAM) was performed to measure the surface hydrophobicity [51]. Briefly, strains were cultured in the LB medium, and butanol (0.8 %, v/v) was added when OD₆₆₀ reached 0.8. Cells were further incubated for 12 h at 37 °C and were collected by centrifugation at 8800 g for 10 min at 4 °C and washed twice by 0.85 % saline. Then, the concentration of cell suspension was adjusted to 50 mg wet cells/mL, and 10 mL cells suspension was filtered through polyvinylidene difluoride membrane (0.22 μm, 50 mm in diameter) under vacuum. For each strain, three biological replicates were performed and measured independently. The contact angle was measured in three phases: the bacterial lawn, n-tetradecane and a droplet of distilled water using a contact angle meter (Dataphysics, Germany).

Analytical methods