Background

Since the first demonstration of microbial product formation more than a century ago [1], vitamins, antibiotics, nucleotides, amino acids and organic acids have been produced in ever increasing quantities. For example, about three million tonnes of sodium glutamate are produced each year as a small microbial molecule. Bacterial synthesis is increasingly also used for the production of small molecules not naturally made by bacteria, such as pharmaceutical intermediates [2, 15] resulted in a 260-fold coverage (Table S4 in Additional file 1). The genome sequence of strain K051 has been deposited at the European Nucleotide Archive under accession number HE802067. Within K051, 268 SNPs are manifest. They are unevenly distributed in the genome (Figure 5b). The number of SNPs is within the range observed for E. coli treated with MNNG [24]. All of the SNPs identified are transitions, as expected with this mutagen, the majority of them resulting in amino acid exchanges (Figure 5b; Table S1 in Additional file 2). In addition, NCgl0863, which carries the amino acid exchange G54D, was partially duplicated, with the variant copy placed 6,108 bp distant from NCgl0863 in an intergenic region.

We searched the mutations in K051 for genes known to increase L-lysine production and to participate in the pathway from glucose uptake up to L-lysine excretion (Figure 5a). Specific mutations in zwf and gnd in the pentose phosphate pathway are known to increase L-lysine formation due to an increased supply of NADPH [25]; K051 has mutations in devB and tal that could also be effective. K051 also has mutations in pck and gltA, genes encoding phosphoenolpyruvate carboxykinase and citrate synthase, where reduced activities are known to increase the supply of pyruvate and oxaloacetate for L-lysine synthesis [26, 27]. Also, mutations of branched-chain amino acid metabolism have been demonstrated to increase lysine formation, and K051 carries a mutation in ilvE, as well as in the Leu-tRNA synthetase LeuS. Of particular interest was the murE mutation (murE-G81E) in K051. This gene encodes UDP-N-acetylmuramyl-tripeptide synthetase, an enzyme that utilizes D, L-diaminopimelate as a substrate, as does the D, L-diaminopimelate decarboxylase, in L-lysine synthesis.

Influence of murE mutations on L-lysine synthesis

To determine whether the murE-G81E mutation identified could generate increased L-lysine formation, we introduced it by allelic replacement into DM1132, DM1728, DM1730, DM1800, and DM1933. The new strains were cultivated in parallel to their ancestor strains in shake flask cultivations and final L-lysine concentrations were determined after 48 h. As shown in Figure 6, the mutation caused strong L-lysine accumulation when introduced into the WT DM1132 and also DM1728, the strains that have few mutations and which form comparatively little L-lysine. Yet even with the best producer available, strain DM1933, a significant increase in L-lysine accumulation was determined. Given this finding, we sequenced murE in the remaining mutants isolated by our HT technology that had no identified mutation (Figure 4), and found murE-L121F in strain K055. Introduction of this specific mutation into the five defined L-lysine producers yielded increased L-lysine accumulation, too (Figure 6). Whether the increases with the two murE mutations identified were due to increased availability of D, L-diaminopimelate for L-lysine synthesis, or whether a global regulatory effect pushes synthesis of D, L-diaminopimelate remains to be studied.

Figure 6
figure 6

Effect of murE mutations on lysine accumulation. Lysine production by different strains modified to carry a chromosomal murE mutation. Color code: gray, ancestor strains; orange, strains carrying the amino acid exchange L121F in MurE; green, strains carrying the amino acid exchange G81E in MurE.

Full size image

Discussion

The key requirement for visualization of single cells with elevated concentrations of a small molecule of interest is the availability of suitable in vivo sensor systems with sufficient sensitivity and specificity. There are a large number of options for develo** customized reporters sensing intracellular metabolites. They are based on natural molecular recognition, allosteric switching, and gene regulation behavior of proteins and RNA. Every system has its own specific advantages and disadvantages, and the reader is referred to recent reviews on the numerous ideas and ongoing developments in the field [12, 28–33]. Whereas protein sensors based on periplasmic binding proteins and Förster resonance energy transfer (FRET) in principle enable concentration determinations in real time, use of TFs relies on expression of the reporter gene. This delay between ligand binding and the corresponding phenotypic change is not a disadvantage in develo** or characterizing recombinant cells since stable genetically encoded genotypes are sought. With respect to the use of TFs in metabolite sensing for screening purposes, the present work based on LysR of C. glutamicum is the first example where the responsiveness of the optical output to an existing intracellular metabolite concentration is given, and where a TF-based sensor is used in an HT screen applying FACS for the isolation of new bacterial small-molecule producers.

The responsiveness of TFs previously characterized is deduced from the external addition of the effector molecule and whole culture response. Although this may only be of limited significance for screening, it is disadvantageous for precise characterization since various processes such as active uptake, active export, diffusion and degradation of effector might result in a different cytosolic concentration than that present extracellularly. In the case of LysG-based pSenLys, we determined a detection range of 4 to 25 mM intracellular L-lysine. Sensor responsiveness is characterized by an analog-like response that, when fitted to the Hill equation, is described by napp of 3.19 ± 1.45. It enables the differentiation of WT from medium- and high-level producer cells (Table S2 in Additional file 1). As our intracellular determinations and the comparison of the isogenic strains with one copy and two copies of lysE revealed, the effective range of detection may be extended by altering export activity. This could be of relevance for further improvement of good producers. Sensor response and its usefulness will depend on the interplay between the cytosolic concentration of the small-molecule and export activity, as well as on the affinity of the sensor to the effector and target promoter site.

Three of the small-molecule sensors described in the present work are based on a LysR-type TF, and one on a ROK-type TF. Fortunately, the range of small molecules detectable by TFs is large. E. coli has more than 230 TFs, with many of them detecting small molecules. In bacteria, TFs have been found to sense sugars, sugar phosphates, vitamins, 2-oxoacids, ions, antibiotics, and acyl-CoA derivatives [9]. Moreover, TFs with new specificities can be generated [11]. An example is AraC, which has been switched from a natural L-arabinose sensor to a sensor detecting D-arabinose [34] or mevalonate [10], and the latter effector specificity has been used in a plate-based assay to screen for improved mevalonate producers. Other sensors that were given new specificities were developed from NahR or XylR for the detection of benzoic acid-related compounds [35], or TetR for structural derivatives of tetracycline [36]. Advances in the design of microbial-based molecular reporters and customizing ligand dependence derived from natural TFs have recently been reviewed [12]. Thus, sensors for a significant number of small molecules of biotechnological or pharmaceutical importance are within reach.

Whereas the WT of C. glutamicum does not excrete L-lysine, cytosolic sensing and FACS as an efficient screen enabled the rapid isolation of 185 new mutants accumulating L-lysine in the culture supernatant. The current number of genes where mutations cause increased L-lysine synthesis is about 12 [37, 38]. These mutations serve to increase flux through the L-lysine pathway itself, or to increase the pyruvate and oxaloacetate pool, or the NADPH supply. However, there are still unknown mutations to be discovered, since it is known that in an L-lysine-producing mutant developed over decades in classical screenings, many genes of biosynthesis pathways exhibit increased expression [39], and in a similarly derived L-arginine producer, arginine biosynthesis genes are highly expressed in a manner not achievable by plasmid-encoded expression [40]. Our approach provided alleles of known genes, and this is very useful for genomic reconstruction of producers where advantageous mutations are combined, and alleles may result in different productivity [2, 41]. The number of 268 SNPs present in K051 is too great to study their individual impact on product formation, but new possibilities might be offered when more genome sequences become available. Striking was the murE mutation present in K051. We suggest that the catalytic activity of UDP-N-acetylmuramoyl-L-alanyl-D-glutamate:meso-diaminopimelate ligase in MurE-G81E is reduced, with the consequence that more D, L-diaminopimelate is available for L-lysine synthesis. MurE of C. glutamicum is similar to MurE of Mycobacterium tuberculosis and E. coli, the crystal structures of which are known [42]. From these, it can be deduced that G81E is close to the nucleoside part of UDP-MurNAc-L-Ala-D-Glu, and L121F in the second mutant identified is close to the ATP-binding site. Thus, a reduced activity is meaningful, and in line with the increased L-lysine formation obtained with all strains when the murE mutations were introduced in their genomes. It is also in line with the reduced growth rates of these new recombinants (Table S5 in Additional file 1), since less D, L-diaminopimelate is channeled towards cell wall synthesis. An alternative to simple mass balance effects is that a lack of cell wall building blocks initiates a global response that has a positive effect on biosynthesis.

We applied one of our transcriptional sensors for HT screening of a mutant library with chromosomal mutations, but the same principle may also be explored for HT screening of cells carrying plasmid libraries. This is attractive, since many pharmaceuticals currently produced microbially, such as amorpha-4,11-diene, taxadiene and lycopene, use plasmid-encoded biosynthesis pathways, for example, in E. coli [2, 3, 13]. Use of an appropriate sensor combined with FACS-assisted screening may significantly accelerate the development of producers for such small molecules, too. The HT selection routine for mutant isolation closes the gap between HT generation of mutant libraries and HT sequencing technologies, and further applications of sensing small molecules in single cells are in progress, such as the verification of producer population homogeneity and time-lapse microscopy of C. glutamicum in microfluidic chips [43].

Conclusions

This work examines visualization of the intracellular concentration of small molecules at the single cell level by the use of specific TFs. It opens up various possibilities to characterize and analyze single cells in populations with respect to their cytosolic small molecule concentration. We have demonstrated that the visualization of L-lysine combined with HT sorting of genomic mutant libraries via FACS enables the isolation of new mutants. Together with whole-genome sequencing, this therefore establishes rapid access to new mutations to achieve more efficient product formation. In addition to the screening of cells with genomic mutations, the system is also suitable for screening cells with plasmid libraries to identify more efficient product accumulation.

Materials and methods

Sensor plasmid construction

The regulatory units were synthesized (LifeTechnologies GmbH, 64293 Darmstadt, Germany) and cloned into pJC1 using the restriction sites BamHI and SalI. An overview of the sensor plasmids is shown in Figure S2 in Additional file 1. The entire plasmid sequences were deposited at EMBL under the accession numbers HE583184 (pSenLys), HE583185 (pSenArg), HE583186 (pSenSer), and HE583187 (pSenOAS1).

FACS analysis and cell sorting

Cells were diluted to an optical density below 0.1 and immediately analyzed by a FACS ARIA II high-speed cell sorter (BD Biosciences, Franklin Lakes, NJ USA 07417) using excitation lines at 488 and 633 nm and detecting fluorescence at 530 ± 15 nm and 660 ± 10 nm at a sample pressure of 70 psi. Data were analyzed using BD DIVA 6.1.3 software. The sheath fluid was sterile filtered phosphate-buffered saline. Electronic gating was set to exclude non-bacterial particles on the basis of forward versus side scatter area. For sorting of Crimson- or EYFP-positive cells the next level of electronic gating was set to exclude non-fluorescent cells. Background was estimated using non-induced C. glutamicum for sorting of Crimson-positive cells. When sorting EYFP-positive cells, non-producing C. glutamicum cells were used.

Mutagenesis and library screening

C. glutamicum ATCC13032 carrying pSenLys was grown in 5 ml BHI complex medium (Difco Laboratories Inc., Detroit, MI 48201, USA) containing 25 μg ml-1 kanamycin to an optical density of 5 to ensure exponential growth. Whole-cell mutagenesis was done by the addition of MNNG dissolved in dimethyl sulfoxide (DMSO) to a final concentration of 0.1 mg ml-1 and incubation for 15 mintes at 30°C. The treated cells were washed twice with 45 ml NaCl, 0.9% (w/v), resuspended in 10 ml BHI and regenerated for 3 h at 30°C and 180 rpm. Mutant cells were stored at -30°C as cryostocks in BHI containing 40% glycerol (w/v). Of the initial cells, 46.2% survived the MNNG treatment and among the surviving cells approximately 16% were auxotrophs. For FACS screening, the mutant stock population containing 7.5 × 108 viable cells per ml was diluted 1:100 in 20 ml minimal medium containing 0.1 mM IPTG to induce expression of the far-red fluorescent protein Crimson, which was taken as an indicator of metabolically active cells. After 2 h of cultivation, 6.5 × 106 cells were analyzed by FACS and 2 × 106 Crimson-positive cells collected in fresh 20 ml minimal medium without IPTG. After cultivation for a further 22 h, 1.8 × 107 cells were screened and 350 EYFP-positive cells spotted on Petri dishes containing minimal medium. Colonies grown after 48 h at 30°C were further analyzed.

HT cultivation and culture fluorescence analysis

HT cultivation was done in 48-well Flowerplates (FPs; m2p-labs GmbH, 52499 Baesweiler, Germany) at 30°C, 990 rpm and a throw of ø 3 mm. The specific geometry of the FPs ensures high mass transfer performance and can be used together with the microcultivation system BioLector [44], allowing online monitoring of growth and fluorescence. The medium used for FP cultivations was the MOPS-buffered salt medium CGXII [45], with 4% glucose as substrate and 25 μg ml-1 kanamycin to select for maintenance of pSenLys. For offline cultivations, FPs were cultivated on a Microtron high-capacity microplate incubator operating at a shaker speed of 990 rpm, throw ø 3 mm (Infors AG, CH-4103 Bottmingen, Switzerland). Shake flask cultivations were used to compare the consequences of the murE mutations for L-lysine accumulation (Figure 4b); these were done in 500 ml baffled Erlenmeyer flasks with 50 ml medium. The medium was the same as used in FP cultivations, except that the phosphate concentration was reduced by half. Cells pregrown in CGXII medium were used as inocula for all cultivations.

Amino acid quantification

Amino acids were quantified as their o-phthaldialdehyde derivatives via high-pressure liquid chromatography using a uHPLC 1290 Infinity system (Agilent, Santa Clara, CA 95051, USA) equipped with a Zorbax Eclipse AAA C18 3.5 micron 4.6 × 75 mm and a fluorescence detector. As eluent, a gradient of 0.01 M Na-borate buffer pH 8.2 with increasing concentrations of methanol was used, and detection of the fluorescent isoindole derivatives was at λex = 230 nm and λem = 450 nm.

Determination of cytosolic amino acid concentrations and amino acid export rates

Cells were pregrown as for FP cultivations for 24 h. They were washed once with fresh CGXII medium at room temperature and transferred into new medium in FPs to give an initial optical density of 10, which corresponds to 3.0 mg (dry weight) ml-1. Cultures were incubated at 30°C on the Microtron high-capacity microplate incubator as above. Samples were processed at regular intervals to separate extra- and intracellular fluid by silicone oil centrifugation [46]. For the resulting fractions, amino acids were quantified as described above. The intracellular volume used to calculate the internal amino acid concentration was 1.6 μl mg (dry weight)-1. When peptides were added (Figure 1e; Figure S3 in Additional file 1) mixtures of di-peptides at a final concentration of 3 mM were used, such as 1 mM Arg-Ala plus 2 mM Ala-Ala, to ensure that a constant supply of Arg-Ala-derived Arg is present over time in the cytosol at the lower Arg-Ala concentrations.

Epifluorescence microscopic analysis

Fluorescence imaging was performed using a fully motorized inverted microscope (Nikon Eclipse Ti) equipped with a focus assistant (Nikon PFS), Apo TIRF 100× Oil DIC N objective, NIKON DS-Vi1 color camera, ANDOR LUCA R DL604 camera, Xenon fluorescence light source and standard filters for EYFP detection (λex = 490 to 510 nm; λem = 520 to 550 nm). Differential interference contrast (DIC) microscopy images as well as fluorescence images were captured and analyzed using the Nikon NIS Elements AR software package. Prior to analysis, cells were fixed on soft agarose-covered glass slides.