Distribution of ε-Poly-l-Lysine Synthetases in Coryneform Bacteria Isolated from Cheese and Human Skin

Every year, microbial contamination causes billions of tons of food wasted and millions of cases of illness. ε-Poly-l-lysine has potent, wide-spectrum inhibitory activity and is heat stable and biodegradable. It has been approved for food preservation by an increasing number of countries. ε-Poly-l-lysine is produced from soil bacteria of the genus Streptomyces, also producers of various antibiotic drugs and toxins and not considered to be a naturally occurring food component.

The microorganisms on and in cheese and their secondary metabolites play key roles for the quality, preservation, safety, and flavor of the final cheese products. In this study, we discovered a Pls from cheese-isolated bacteria, confirmed its activity by heterologous expression, and investigated its distribution.
We tested « -PL production from C. variabile using a two-stage culture method which was efficient in finding Streptomyces producers (18). However, no « -PL was detected in the culture. We reasoned that the cheese bacteria may have different regulation of « -PL biosynthesis from that of soil bacteria of the genus Streptomyces. Therefore, we cloned the C. variabile gene onto a plasmid with an inducible pBAD promoter. The recombinant plasmid was transferred into model organism Corynebacterium glutamicum. However, again « -PL production could not be observed in the cultures with or without arabinose induction. We checked the enzyme expression by whole-cell proteomic analysis. The Pls expression in the recombinant strain after induction was confirmed, with an 8-fold increase of normalized signal abundance over the uninduced sample (see Data Set S1 in the supplemental material). In the sample of C. variabile, the signal of Pls protein was not detected, while 1,647 of the 2,972 predicted proteins were detected (Data Set S2).
In Streptomyces, the promoter sequence is critical for « -PL production. It has been demonstrated that expression of pls in the native host S. albulus with an altered promoter did not lead to « -PL production, but the use of the original promoter resulted in « -PL production even in a heterologous Streptomyces host (19). Inspired by this, we cloned the C. variabile gene under the control of the S. albulus pls promoter and transferred the plasmid into Streptomyces coelicolor M145, which does not have an endogenous pls gene. Using this expression system, « -PL production was confirmed by ultrahighperformance liquid chromatography (UHPLC) and high-resolution mass spectrometry (MS) (Fig. 2). The titer in shaking flask cultures was determined to be 120 6 28 mg/liter by methylene blue agar diffusion assay (20).
We investigated the distribution of pls in a genome collection of 156 bacteria isolated from cheeses from Europe and the United States (21). As many microorganisms from surface-ripened cheeses can also be found in animal and human skin microbiota (22) and « -PL is also used in cosmetic products (1), we further included a genome collection of 124 microorganisms isolated from human skin (23). We used experimentally confirmed Pls protein sequences as queries to do BLASTP against the two collections with cutoffs of 40% sequence identity and 80% sequence coverage. Pls homologs were found to be concentrated in coryneform actinobacteria, including Corynebacterium, Brevibacterium, Arthrobacter, Microbacterium, Glutamicibacter, Rhodococcus, Micrococcus, and Dermacoccus. No hit was found in bacteria from other genera or phyla (Fig. 3A). No hit was found using the other three synthetases as queries with the same cutoffs.
Phylogenetic analysis shows that all the hits cluster together with the experimentally confirmed Pls from C. variabile, Kitasatospora, Streptomyces, and fungi, while the synthetases of the other three isopeptides are on more distant branches (Fig. 3B). Most importantly, NRPSpredictor2 (24) results show that the 10-residue substrate-recognizing pockets (25) of the coryneform bacterial proteins are identical or highly similar to that of the confirmed Pls proteins but substantially different from those of the other three synthetases (Fig. 3B), which strongly suggests that their substrate is lysine.

DISCUSSION
In this study, we confirmed that the cheese bacterium C. variabile DSM 44702 harbors a functional « -PL synthetase gene. However, we did not observe « -PL production by C. variabile under our artificial culturing conditions. This is likely caused by a regulatory mechanism that requires an unknown trigger signal, which was missing in our cultivations. Such tightly controlled biosynthetic pathways are very common in microbial secondary metabolite biosynthesis and thus have shaped the term "silent biosynthetic gene cluster" (26). Corynebacteria are related to Streptomyces and have many properties desired for industrial fermentation, like being nonfilamentous and having faster growth, a simpler life cycle, and a simpler secondary metabolism. Successful activation of « -PL synthesis in corynebacteria may provide the basis for a new « -PL production process. Furthermore, Pls were widely found in cheese-and skin-isolated coryneform bacteria. The majority of Brevibacterium and Corynebacterium isolates, which are among the most important microorganisms in cheese production and also commonly found on human skin, have Pls. It is possible that « -PL naturally exists in cheese and on human skin and may have a role in their ecologies. Other antimicrobial compounds, like bacteriocins, have been known to be produced in cheese and skin environments and modulate the microbiota compositions (27,28). The existence and quantity of « -PL on human skin and in different cheeses and different stages of the cheese making process require further study.

MATERIALS AND METHODS
Bacteria. Corynebacterium variabile DSM 44702 was obtained from the DSMZ. Streptomyces coelicolor M145 and Corynebacterium glutamicum MB001(DE3) were used as the heterologous hosts. Escherichia coli DH5a was used for DNA cloning.
Gene cloning. An expression vector pXJ0GC for corynebacteria was developed from shuttle plasmid pAL374 (29). An AraC-pBAD fragment was amplified from pBAD18 with primers xj336.1 and xj337. An rrnBT1T2 fragment was amplified from pBAD30 with primers xj338 and xj339. An aac(3)-oriT fragment was amplified from pRM4.3 with primers xj340 and xj341. A replication origin fragment was amplified from pAL374 with primers xj342 and xj343. An mScarlet-FDterminator fragment was chemically synthesized. The above-mentioned fragments were assembled by Gibson reaction into plasmid pXJ00. A pT7-pTrc-gfp-cmr fragment was amplified from pACY-gfp with primers julie11 and julie12. It was assembled with pXJ100 digested with SfaAI, HindIII, and NdeI, resulting in plasmid pXJ0GC. The C. variabile Pls gene was PCR amplified from C. variabile DSM 44702 genomic DNA with primers xj372 and xj373 and cloned onto the vector backbone amplified from pXJ0GC with primers bb0s and bb0a. The resulting plasmid, pXJ146, was used for gene expression in C. glutamicum. The promoter sequence of the S. albulus Pls gene was chemically synthesized and then amplified with primers xj420 and xj421. The C. variabile Pls gene was PCR amplified from C. variabile DSM 44702 genomic DNA with primers xj426 and xj427. A plasmid backbone was amplified from shuttle vector pRM4e with primers xj422 and xj423. The above-mentioned three fragments were assembled by Gibson reaction into plasmid pXJ155CV and used for gene expression in S. coelicolor. Primer sequences and DNA sequences chemically synthesized are listed in Tables 1 and 2.
Culture conditions. Streptomyces and Corynebacterium strains were maintained on ISP2 agar (BD Difco). They were assayed for « -PL production by a two-stage cultivation method (18). S. coelicolor strains were inoculated in M3G medium (30) at pH 6.8 for 24 h at 30°C, and then the pH was adjusted to 4.0 by HCl and culture was continued for another 3 days with shaking at 120 rpm. Corynebacterium strains were cultured similarly with GMPY medium (malt extract at 10 g/liter, peptone at 10 g/liter, and yeast extract at 0.1 g/liter, autoclaved, with glucose added at 10 g/liter as a carbon source), and 1% arabinose was used for induction of gene expression in recombinant strains. Extraction. A Bond Elut LRC-CBA column (Agilent; part number 12113037) was conditioned by washing with 5 ml of methanol and then 5 ml of water. Bacterial culture supernatant was adjusted to pH 8 with NaOH and loaded on the column with a speed of 3 ml per min. The column was washed with 5 ml of water and then eluted with 5 ml of methanol twice. The elution was dried in a rotary evaporator at 38°C, redissolved and collected with 1 ml of methanol, and then concentrated to 50 ml using a vacuum centrifuge.
UHPLC-MS analysis. UHPLC-MS analysis of the extract was performed on a Dionex Ultimate 3000 UHPLC system coupled to a high-resolution Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific, Waltham, MA) and a UV-visible (UV/Vis) diode array detector (DAD). Separate positive-and negative-ion-mode electrospray ionization (ESI) experiments were carried out with an MS scan range of 100 to 1,000 Da. Injections of 8 ml of each sample were separated using a Waters Cortecs T3 column, 150 by 2.1 mm (inside diameter [i.d.]) and 1.6-mm particle size, at a temperature of 35.0°C and a flow rate of 0.35 ml/min. Elution was performed with 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B) in a multistep program: 0% mobile phase B for 2.5 min, a linear gradient from 0% to 100% mobile phase B in 15 min, 100% mobile phase B for 2 min, and 0% mobile phase B for 2 min.
Methylene blue agar diffusion assay. The « -PL titer of the Streptomyces culture was determined by methylene blue agar diffusion assay as described in a previous paper (20). The agar plates were made with 0.75% agar and 0.002% methylene blue. A 100-ml sample was applied to the plate and incubated at 30°C for 5 h before the diffusion diameter was measured. A standard curve was made from six « -PL concentrations ranging from 50 to 1,000 mg/liter. A regression coefficient of 0.9948 was achieved.
Proteomic analysis. Cells were collected by centrifugation at 12,000 Â g for 10 min 48 h after the inoculation and stored at 220°C until analyzed. After thawing of the cells on ice, the samples were centrifuged again and any remaining supernatant was removed. The samples were added with two 3-mm zirconium oxide beads (Glen Mills, Clifton, NJ) and then moved away from ice and immediately added with 100 ml of 95°C guanidinium HCl solution [6 M guanidinium hydrochloride, 5 mM tris(2-carboxyethyl)phosphine, 10 mM chloroacetamide, and 100 mM Tris-HCl (pH 8.5)]. Cell disruption was performed in a mixer mill (MM 400; Retsch, Haan, Germany) set at 25 Hz for 5 min at room temperature, followed by 10 min in a ThermoMixer at 95°C at 2,000 rpm. Remaining cell debris was precipitated by centrifugation at 15,000 Â g for 10 min. A 50-ml volume of the supernatant was collected and diluted with 50 ml of 50 mM ammonium bicarbonate. The protein concentration was determined by bicinchoninic acid (BCA) assay; 100 mg of protein was subjected to tryptic digestion at constant shaking (400 rpm) for 8 h and then added with 10 ml of 10% trifluoroacetic acid (TFA). The samples were cleaned by stage tipping using C 18 resin (Empore; 3M, USA).
The proteomic analysis was carried out on a CapLC system (Thermo Scientific) coupled to an Orbitrap Q Exactive HF-X mass spectrometer (Thermo Scientific). Samples was first injected and carried at a flow rate of 10 ml/min on a precolumn (m-precolumn C 18 PepMap 100, 5 mm, 100 Å) and then at a flow of 1.2 ml/min on a 15-cm C 18 EASY-Spray column (PepMap RSLC C 18 , 2mm, 100 Å, 150 mm by 15 cm) for peptide separation. The mobile phase gradient increased from 4% to 76% acetonitrile in water over a 3 was used for analysis of the Thermo raw files with the following settings: fixed modifications, carbamidomethyl (C), and variable modifications, oxidation of methionine residues. First-search mass tolerance was 20 ppm, and tandem MS (MS/MS) tolerance was 20 ppm. Trypsin was used as the digestion enzyme, and one missed cleavage was allowed. The false-discovery rate (FDR) was set at 0.1%. The match-between-runs window was set to 0.7 min. Quantification was based only on unique peptides, and normalization between samples was based on total peptide amount. For the searches, a protein database consisting of the reference proteome in combination with the expressed target proteins was used.
Protein domain analysis was performed by InterProScan (10). Phylogenetic analysis was done by MEGA-X using Muscle for multiple-sequence alignment and Poisson model for UPGMA (unweighted pair group method using average linkages) tree building. NRPS A-domain substrate prediction was done by NRPSpredictor2 (24). For the cheese microorganism genome data set, 156 genomes were downloaded directly from the Data Set S1 file of reference 21, and the amino acid sequences were extracted from the GenBank format files using CLCgenomics (v.20.0). In addition to this, we found updated genomes for 47 of the 156 strains in NCBI. The genome accession numbers are listed in Data Set S3. Genes were downloaded from NCBI and included in the analysis. For the human skin microorganism genome data set from reference 23, 124 genomes were downloaded from the NIH Human Microbiome Project (https:// www.hmpdacc.org/hmp/catalog/grid.php?dataset=genomic). We downloaded the following Pls proteins to use as a reference: those with NCBI GenPept accession numbers BAG68864.1, BAH85292.1, WP_041630296.1, AZL89021.1, CCE28893.1, and BBU42014.1. We then used BLASTP (v.2.6.01) with the following parameter to identify putative Pls in the downloaded amino acid data sets: -evalue 0.000001. We then extracted hits with at least 40% identity and at least 80% coverage of the reference proteins.