Optimized Genetic Tools Allow the Biosynthesis of Glycocin F and Analogues Designed To Test the Roles of gcc Cluster Genes in Bacteriocin Production

The entire 7-gene cluster for the diglycosylated bacteriocin glycocin F (GccF), including the natural promoters responsible for gcc gene expression, has been ligated into the Escherichia coli-lactic acid bacteria (LAB) shuttle vector pRV610 to produce the easily modifiable 11.2-kbp plasmid pRV610gcc for the efficient production of glycocin F analogues. In contrast to the refactoring approach, chemical synthesis, or chemoenzymatic synthesis, all of which have been successfully used to probe glycocin structure and function, this plasmid can also be used to probe in vivo the evolutionary constraints on glycocin scaffolds and their processing by the maturation pathway machinery, thus increasing understanding of the enzymes involved, the order in which they act, and how they are regulated.

ABSTRACT

The emergence of multidrug-resistant pathogens has motivated natural product research to inform the development of new antimicrobial agents. Glycocin F (GccF) is a diglycosylated 43-amino-acid bacteriocin secreted by Lactobacillus plantarum KW30. It displays a moderate phylogenetic target range that includes vancomycin-resistant strains of Enterococcus species and appears to have a novel bacteriostatic mechanism, rapidly inhibiting the growth of the most susceptible bacterial strains at picomolar concentrations. Experimental verification of the predicted role(s) of gcc cluster genes in GccF biosynthesis has been hampered by the inability to produce soluble recombinant Gcc proteins. Here, we report the development of pRV610gcc, an easily modifiable 11.2-kbp plasmid that enables the production of GccF in L. plantarum NC8. gcc gene expression relies on native promoters in the cloned cluster, and NC8(pRV610gcc) produces mature GccF at levels similar to KW30. Key findings are that the glycosyltransferase glycosylates both serine and cysteine at either position in the sequence but glycosylation of the loop serine is both sequence and spatially specific, that glycosylation of the peptide scaffold is not required for export and subsequent disulfide bond formation, that neither of the putative thioredoxin proteins is essential for peptide maturation, and that removal of the entire putative response regulator GccE decreases GccF production less than removal of the LytTR domain alone. Using this system, we have verified the functions of most of the gcc genes and have advanced our understanding of the roles of GccF structure in its maturation and antibacterial activity.

IMPORTANCE The entire 7-gene cluster for the diglycosylated bacteriocin glycocin F (GccF), including the natural promoters responsible for gcc gene expression, has been ligated into the Escherichia coli-lactic acid bacteria (LAB) shuttle vector pRV610 to produce the easily modifiable 11.2-kbp plasmid pRV610gcc for the efficient production of glycocin F analogues. In contrast to the refactoring approach, chemical synthesis, or chemoenzymatic synthesis, all of which have been successfully used to probe glycocin structure and function, this plasmid can also be used to probe in vivo the evolutionary constraints on glycocin scaffolds and their processing by the maturation pathway machinery, thus increasing understanding of the enzymes involved, the order in which they act, and how they are regulated.

INTRODUCTION

Bacteriocins are antimicrobial ribosomally synthesized peptides, many of which have a narrow target range limited to closely related species (1). The use of narrow-spectrum bacteriostatic antimicrobials to treat disease in humans has at least two advantages that are only beginning to be understood. First, they have been shown to significantly limit disturbance to the natural microbial diversity of the gut, reducing the incidence of irritable bowel disease (2), Crohn’s disease, and ulcerative colitis (3). Second, because many of these substances do not kill pathogenic strains, but rather substantially inhibit their growth, target organisms are less likely to develop resistance to them (4).

The diverse bacteriocins secreted by the generally regarded as safe (GRAS) lactic acid bacteria have been grouped into two main classes, modified and unmodified (5). The glycocins are a subclass of modified bacteriocins that are characterized by posttranslational glycosylation (6, 7). So far, five glycocins (SunA [8, 9], GccF [10, 11], ASM1 [12, 13], enterocin 96 [14], and enterocin F4-9 [15]) have been isolated from bacterial cultures and biochemically characterized, while others have been identified though genome mining and synthesized chemoenzymatically (thurandacin [16]) or expressed in Escherichia coli using pathway refactoring systems (pallidocin [17], bacillicin CER074, geocillicin, bacillicin BAG2O, and listeriocytocin [18]). Glycocin F (GccF) is a 43-amino-acid bacteriocin, secreted by Lactobacillus plantarum KW30, that is modified by both O- and S-linked N-acetyl-d-glucosamines (GlcNAcs) (Fig. 1A) (10). It is active against a moderate range of Gram-positive bacteria, including several lactobacilli, enterococci, and streptococci, some of which are pathogens (19). Although it has been shown to be glycoactive, only the O-linked GlcNAc is absolutely essential for antimicrobial activity, with the S-linked GlcNAc increasing activity ∼53-fold (10, 20, 21).

FIG 1.

Glycocin F (GccF) structure, the gcc gene cluster, and putative gcc promoters. (A) Model of the GccF NMR structure (created from PDB 2KUY using the PyMOL molecular graphics system [version 1.8; Schrödinger, LLC]) and a schematic diagram showing the positions of the nested disulfide bonds and O- and S-linked N-acetylglucosamines. (B) Gene arrangement of the gcc cluster (GenBank accession no. GU552553). gccH encodes the immunity protein. gccA encodes a glycosyltransferase (GTase) with a family 2 domain (PF00535). gccB encodes an ABC transporter containing an N-terminal peptidase_C39 domain (PF03412), a transmembrane region (PF00664), and a C-terminal ATP-binding domain (PF00005). Both gccC and gccD are predicted to encode proteins containing a thioredoxin family domain (PF00085). gccE encodes a putative response regulator (RR) with a C-terminal LytTR DNA binding domain (PF04397). gccF encodes the 64-amino-acid (aa) GccF precursor peptide including a 21-aa double glycine leader peptide. (C) Putative overlapping divergent promoters in the gccHA intergenic region (numbering is from GU552553). Predicted −10 and −35 elements of P_gccH and P_gccA-F are boxed, and bent arrows indicate transcription start points. Ribosome binding sites are marked with asterisks. Bold blue letters indicate an imperfect heptanucleotide repeat separated by a 16-bp spacer. (D) The gccF promoter P_gccF, including the putative LytTR RR recognition elements (in blue).

GccF is encoded within a cluster of seven genes (Fig. 1B), but only those encoding the immunity protein GccH (12, 22, 23) and GccF (10) have had their functions experimentally verified. In contrast, there is no experimental evidence to support the functionality predicted by sequence analysis of the remaining five genes. gccA encodes a protein in which residues 42 to 152 are predicted to form a family 2 glycosyltransferase (GTase) domain (PF00535; E value, 1.1e⁻¹⁵) that is thought to be responsible for catalyzing the transfer of one or both GlcNAcs to the GccF peptide scaffold. Its sequence is similar to those of the GTase domains encoded by the sublancin (30% identity) and thurandacin (30% identity) gene clusters, and the corresponding GTases have been heterologously expressed as active enzymes in E. coli (16, 24). The ABC transporter (GccB) amino acid sequence is consistent with an N-terminal proteolytic domain that cleaves the double glycine leader peptide and a transmembrane domain that facilitates bacteriocin export. L. plantarum genomes typically encode a single ABC transporter of this type: PlnG, dedicated to the export of plantaricins PlnA, PlnEF, and PlnJK (25). gccC and gccD encode proteins belonging to the Trx-domain family and contain the characteristic CXXC motif (PF00085, GccC residues 37 to 122; E value, 5.4e⁻⁰⁸; GccD residues 64 to 154; E value, 1.9e⁻⁰³). GccC and GccD have only moderate sequence similarity to BdbA and BdbB, two proteins in the sublancin gene cluster that were originally thought to facilitate the formation of disulfide bonds in sublancin and are analogous to DbsA and DbsB, respectively (two enzymes involved in disulfide bond formation in a wide range of bacteria) (26). However, while BdbA has ∼30% sequence identity to both GccC and GccD, BdbB has no significant sequence similarity to either of these Gcc thioredoxin-like proteins. Interestingly, BdbA, the thioredoxin most similar to GccC and GccD, is dispensable for sublancin production (27), raising questions about the true role(s) of the Gcc “thioredoxins.” The putative response regulator (GccE) contains a C-terminal LytTR DNA binding domain, likely to be a regulator of gccF transcription (28), and an N-terminal domain with little similarity to any other protein (10). Further analysis and annotation of these genes can be found in the GenBank accession number GU552553.

Many attempts were made to knock out gcc genes in L. plantarum KW30 and to produce soluble Gcc proteins in both E. coli and yeast, but none of these were successful, necessitating the development of an alternative system to test gene function as well as to further investigate the maturation pathway of GccF. A strategy that involved cloning the whole gcc cluster (6,059 bp; gccH and gccA-F) into a shuttle vector that could be used to transform plasmid-free L. plantarum NC8 was therefore employed to produce soluble fully active GccF and to functionally characterize specific cluster genes. Because gccH and gccA-F are transcribed divergently (Fig. 1B and C), it was decided to rely on native promoters to drive the coordinated transcription of gcc genes. Analysis using the algorithm BPROM (29) predicted the presence of divergent promoters between gccH and gccA (Fig. 1C) that are responsible for constitutive gccH and gccA-F transcription seen in previous quantitative PCR (qPCR) studies (23). A third promoter, P_gccF (Fig. 1D), is thought to drive the production of greater quantities of GccF secreted by L. plantarum KW30 from late log phase onwards.

This strategy also provided insights into the complex interplay between GccF structure and the individual cluster enzymes and the maturation/secretion of GccF.

RESULTS

GccF produced in L. plantarum NC8 is indistinguishable from wild-type KW30 GccF.

The entire gcc gene cluster was amplified by PCR to introduce XmaI and XbaI restriction sites for ligation into the similarly digested shuttle vector pRV610 (30) to produce the 11.2-kbp pRV610gcc (see Fig. S1, S2, and S3 in the supplemental material). This plasmid was transformed into L. plantarum NC8, and biological activity assays were used to confirm that the transformed cells secreted an antimicrobial substance into the supernatant, using cells transformed with an empty vector as a negative control. These assays (Fig. S4) showed that only the cells containing pRV610gcc had activity against the indicator strain L. plantarum ATCC 8014. Recombinant GccF was then purified from the supernatant of an L. plantarum NC8(pRV610gcc) culture yielding the typical reversed-phase high-performance liquid chromatography (RP-HPLC) elution profiles (10) (Fig. S5). The GccF peak was collected and analyzed on a tricine SDS-PAGE 16% gel. Staining with Coomassie blue showed a broad band at the expected mass for GccF (10) (Fig. S6). L. plantarum NC8(pRV610gcc) GccF was shown to have a monoisotopic mass of 5,199.0468 Da (Fig. S7), consistent with that of GccF purified from the supernatant of its natural producer, L. plantarum KW30 (10), suggesting there had been no change to its structure in this expression system and that all posttranslational modifications were present.

Two alternative hosts were transformed with pRV610gcc; Lactobacillus sakei Lb790 produced GccF in similar amounts to NC8 (Fig. S8a), whereas Enterococcus faecalis JH2-2 transformed with pRV610gcc failed to produce detectable activity (results not shown) and remained susceptible to GccF (Fig. S8c), suggesting that the gcc promoters are not functional in JH2-2. Challenging L. sakei Lb790 with/without pRV610gcc with GccF showed that pRV610gcc conferred resistance to GccF, while a challenge with erythromycin confirmed the presence of pRV610gcc, which carries an erythromycin resistance gene (30). pRV610gcc did not decrease growth inhibition by streptozotocin, an antibiotic imported into L. sakei Lb790 by the N-acetylglucosamine phosphotransferase system (PTS) transporter (NagE) which also serves as the GccF receptor (31), indicating that NagE remained functional (Fig. S8b).

Activity of purified recombinant GccF.

The concentration of recombinant GccF required to inhibit the growth of L. plantarum ATCC 8014 by 50% (IC₅₀) was calculated to be 2.0 ± 0.3 nM compared to 2.2 ± 0.3 nM for wild-type GccF under identical assay conditions, indicating the molecule was correctly folded (Fig. S9).

Furthermore, the antimicrobial activity of recombinant GccF was attenuated by simultaneously exposing the indicator strain to 5 mM GlcNAc (Fig. S10), a property shared with the wild-type peptide (10). That the recombinant peptide behaves like the wild type in the presence of free GlcNAc strongly suggests that the two HexNAcs are GlcNAcs (10, 20).

“Indicator agar tube” assays are more sensitive for detecting GccF activity in CFS.

To detect GccF activity in the cell-free supernatant (CFS) following mutations resulting in a significant loss of activity, a more sensitive activity assay was developed that used small tubes (diameter, 6 mm) filled with indicator agar. Up to 50 μl of CFS could be applied to the surface, thus increasing the amount of GccF. Unlike the indicator agar plate method where the diameters of the zones of inhibition were compared, inhibition could be measured by the depth of the zone of inhibition. Figure 2 shows that the activity of purified GccF was detectable at concentrations as low as 0.16 μM. L. plantarum NC8(pRV610gcc) CFS had activity equivalent to 1.25 μM, showing the limit of detection of this assay is ∼12% of the normal recombinant activity.

FIG 2.

Calibrating the GccF activity assay. Twenty microliters of purified GccF diluted to the indicated concentrations in MRS broth was pipetted into tubes containing the L. plantarum ATCC 8014 indicator strain embedded in MRS agar. L. plantarum NC8(pRV610gcc) CFS (20 μl) was added to a single tube.

Is the glycosyltransferase GccA responsible for both O- and S-linked GlcNAc modifications?

Residues 40 to 267 of the gcc cluster GTase, GccA, have a high level of sequence similarity (E value, 4e⁻²⁹) to a subset of family 2 peptide glycosyltransferases and contain the DXD motif that is essential for activity (32). Rather than introducing a premature stop codon in gccA, which could have allowed an alternative start codon to be used, GccA activity was abolished by introducing two point mutations in the DNA encoding the DXD motif, creating GccA_D123N,D125N. This resulted in a loss of detectable GccF activity (Fig. 3), suggesting that Ser18, at least, had not been glycosylated.

FIG 3.

Using GccF activity to assess GccA activity. Forty microliters of L. plantarum NC8(pRV610gcc) wild-type or analogue GccF CFS was pipetted on MRS agar containing L. plantarum ATCC 8014, and the tubes were incubated overnight at 30°C.

To show that lack of activity was not due to aggregation because of the free thiol group on Cys43, analogue GccF_C43S was produced by NC8 containing pRV610gcc with either wild-type gccA or the mutated version. Figure 3 shows that CFS containing the GccF_C43S analogue glycosylated by wild-type gccA was slightly less active than wild-type GccF produced in the same background. After purification using an elongated RP-HPLC gradient (see the supplemental material), there were two species present (Fig. S11). Analysis by mass spectrometry showed that one was diglycosylated and the other monoglycosylated (Fig. S12). To show that the GlcNAc of the monoglycosylated GccF_C43S was located on Ser18, a tryptic digest of nonreduced peptide was analyzed by mass spectrometry, which confirmed that the HexNAc resided only on Ser18 (Fig. S13). These experiments showed that GccA not only is most likely responsible for the glycosylation of both Ser18 and Cys43 but also glycosylates a Ser at position 43, albeit with reduced efficiency.

To confirm that GccA was indeed solely responsible for glycosylating both Ser18 and Cys43, the possibility that unmodified GccF is secreted into L. plantarum NC8(pRV610gcc_GccA_D123N,D125N_GccF_C43S) culture medium was investigated. The supernatant from a 4-day 2-liter culture was collected, and GccF-type peptides were purified using the (standard) shorter RP-HPLC gradient as described in the supplemental material. Surprisingly, nonglycosylated GccF_C43S was isolated, and mass spectrometry showed it had both disulfide bonds but no HexNAcs (Fig. S14a). Furthermore, the helical structure of the wild type appeared to be retained (Fig. S14b).

These results raised the question of whether GccA would be able to glycosylate a Cys at position 18 or whether it is specific for Ser18. A GccF_S18C analogue was produced and purified. Mass spectrometry confirmed the presence of 2 disulfide bonds and 2 HexNAcs (Fig. S15). As the chemically synthesized analogue GccF_S18C was 2.5-fold more active than wild-type GccF (20), the finding that the CFS of L. plantarum NC8(pRV610gcc_GccF_S18C) showed a 30% reduction in activity compared to that of the wild type (Fig. 3) was somewhat surprising, although it is possible that this GccF analogue was produced in smaller quantities.

The loop region of GccF is of interest as it contains the O-linked GlcNAc that is essential for bacteriostatic activity. To further test possible structural requirements for recognition of the glycosylation site by GccA, several modifications were made to the loop that involved both substitutions of single amino acid residues and shifting the relative position of Ser18 within the loop as shown in Fig. 4. GccF_{G15del,S18→17,G19insG} was created in order to see if there was any change in GccF activity when the position of the GlcNAc was shifted one residue toward the center of the loop, retaining both the number of residues in the loop and the identity of the residues flanking Ser18. Unexpectedly, there was no detectable activity in NC8(pRV610gcc_GccF_{G15del,S18→17,G19insG}) CFS. Due to lack of time, no purification and mass spectrometry analysis were attempted; future work will determine if this analogue is synthesized and secreted but has low specific activity.

FIG 4.

GccF loop mutations. Wild-type GccF with amino acid substitution sites colored turquoise (left), and the GccF_{G15del,S18→17,G19insG} change with the shifted serine residue colored red (right). The red arrow shows the change in location of Ser18 (to Ser17) and its O-linked GlcNAc.

The conserved loop residues Tyr16 and Asp17 may be required for GccA to recognize/bind its peptide substrate (6). Conservative mutations, GccF_Y16F and GccF_D17N, were chosen to probe the role of these GccA substrate residues. GccF_D17N greatly reduced activity (Fig. S16), indicating that the charge on Asp17 might be important. In contrast, replacing Tyr16 with Phe had no apparent effect on GccF activity, indicating that the hydroxy group of the aromatic side chain is not required for GccA recognition or GccF activity.

GccB function.

GccB, strongly predicted to be an ABC transporter with an N-terminal C39 cysteine protease domain that recognizes the double glycine motif in the GccF leader peptide as a cleavage site (33), is a large multidomain protein with several potential internal translation initiation codons in the corresponding open reading frame (ORF). Rather than introducing a stop codon into gccB, a mutation designed to prevent cleavage of the signal peptide, and so prevent export, was introduced into gccF. Specifically, the codons for glycine residues at positions −1 and −2 were mutated to threonine codons, abolishing the double glycine cleavage motif (Fig. S17a). As expected, this mutation resulted in a loss of measurable GccF activity in the CFS (Fig. S17b). Presumably, this amino acid substitution (GG→TT) traps precursor GccF within the cytosol, as seen in similar experiments with other bacteriocins (34).

GccC and GccD are not required for in vivo synthesis of GccF by NC8 in MRS broth.

As the cytosol of any cell is highly reducing, it is unlikely the disulfide bonds of GccF are formed prior to its export. Upon export, the process of disulfide bond formation is thought to be facilitated by the thioredoxin-like proteins GccC and GccD, which are predicted to be membrane anchored outside the cell (23). The CXXC motif is a requirement for the activity of most thioredoxins. Although GccC and GccD both contain this motif, the possibility that these proteins are involved in another process could not be discounted based on sequence alone. Premature stop codons were introduced near the beginning of both ORFs. Sequence analysis showed that while the GccC_K3X ORF lacked an alternative start codon, translation of the GccD_L4X mRNA could be rescued by the codon for Met55 with retention of the CXXC motif; although without a transmembrane anchor, it should not be able to function as efficiently. When these mutations were introduced into the pRV610gcc constructs, both individually and in tandem, all constructs were still able to produce active GccF, albeit with attenuated activity (Fig. S18).

Role of GccE.

Homology studies of the LytTR DNA binding domain of GccE suggest that this protein is somehow involved in the regulation of GccF production. However, the N-terminal domain has no sequence similarity to any protein in the publicly available databases except for that of AsmE, the putative response regulator found within the ASM1 gene cluster (12). This suggests that the N-terminal domain of GccE has a novel structure and an unknown function. To investigate the role of GccE in the production of GccF by the L. plantarum NC8 expression system, two mutations were made to gccE. The first of these, GccE_L19X introduced a stop codon near the start of the ORF to create a severely truncated and nonfunctional protein. The result of this mutation was a reduction, but not a loss, of measurable activity (Fig. S19a). A second mutation, GccE_L148X, was designed to retain the N-terminal region while removing the LytTR DNA binding domain. In contrast to removing all but the first 18 amino acids, the GccE_L148X mutation appeared to stop the production of active GccF within the limits of detection of the assay (Fig. S19a).

To confirm that the decreased GccF activity seen in these mutations was due to decreased production and not to structural modification of GccF, mass spectrometry was used to analyze GccF purified from L. plantarum NC8(pRV610gcc_GccE_L148X) (Fig. S19b). The fact that mature GccF was able to be purified, albeit in very small amounts, suggests that GccE is a transcriptional activator of the gccF promoter responsible for the high levels of GccF produced by the natural producer strain KW30 in late log and early stationary phase (23).

DISCUSSION

We have successfully expressed GccF, an unusual posttranslationally modified antibacterial peptide, in a nonnative host and showed that the purified recombinant peptide has the same activity as the wild-type bacteriocin. The fact that L. plantarum NC8 and L. sakei Lb790 (but not E. faecalis JH2-2), strains previously shown to be susceptible to GccF (12), both secrete GccF and resist its effects when transformed with pRV610gcc indicates that all the genes required for GccF regulation, production, and immunity are functional in these hosts. The NC8 system is a good model for GccF production in L. plantarum KW30 and has enabled us to verify the functions of most Gcc proteins. Although the pRV610 plasmid allowed for selection with erythromycin in L. plantarum NC8 and L. sakei Lb790, this was not used for cultures greater than 50 ml, as it has been reported that pRV610 is very stable in lactobacilli, with 90% of L. sakei cells retaining it for 40 generations without selection (30). Although pRV610gcc is much larger than the original plasmid, and large plasmids are known to be less stable than smaller ones (35), plasmid stability did not appear to be an issue, probably because the plasmid provided host cells with immunity to GccF, in contrast to the wild-type strains (19). A summary of all mutations and their effects is shown in Table 1. It should be noted that the use of this plasmid allowed us to investigate the effects of specific amino acid changes on the in vivo maturation pathway of pre-GccF and showed unexpected differences in the activity of some recombinant GccF molecules compared to those synthesized chemically (20, 21).

TABLE 1.

Overview of amino acid substitutions in Gcc proteins and GccF analogues produced in this study compared with chemically synthesized analogues

Protein	aa change or ORF mutationa	CFS activity (mm)b	GccF structurec	Synthetic analogued	IC50 (nM)
GccF	Native protein			GccFNative	2.0 ± 0.2d
	Chemically synthesized			GccFSyn	1.13 ± 0.2d
	GccFRecomb	6			2.0 ± 0.3e
Leader peptide	G(−1)T, G(−2)T	ND
GccF loop modifications	Y16F	6
	D17N	ND
	S18C	5	C18 and C43 glycosylated	S18C	0.60 ± 0.1f
	G15del, S18→17, G19insG	ND
				ΔG13	57 ± 6.8d
				ΔG13ΔG15	2,480 ± 540d
				ΔG13ΔG15ΔG19	2,710 ± 370d
GccF tail modifications	C43S	4	∼50% C43S and 100% S18 glycosylated	C43S	12.1 ± 2.4f
GccA	D123N, D125N	ND	No HexNAcs
	D123N, D125N_GccFC43S		Disulfides intact
GccC	K3X	4	Disulfides intact
GccD	L4X	5	Disulfides intact
GccC/D	K3X, L4X	4	Disulfides intact
GccE	L19X	4
	L148X	ND

^aaa, amino acid; −1 and −2 refer to the residues immediately upstream from the pre-GccF cleavage site, Gly20 and Gly21 of pre-GccF. X refers to the introduction of a stop codon at this position.

^bTube assay: estimated depth of growth inhibition from surface. ND, not detected.

^cPeptide structural features that may or may not be affected by gccF mutation.

^dResults from Bisset et al. (20).

^eThis study.

^fResults from Amso et al. (21).

Experimental evidence showed not only that GccA is responsible for the glycosylation of both Ser18 and Cys43 but also that it can glycosylate a Ser at position 43 to add an O-linked HexNAc as well as Cys at position 18 to produce the more unusual S-linked HexNAc. It is probable that the decreased activity seen for GccF_C43S was due to a substantial fraction of prepeptide being processed to a monoglycosylated species that had previously been shown (using the chemically synthesized monoglycosylated analogue) to have an IC₅₀ approximately 55-fold less than that of native GccF (20). Two possible reasons for the production of monoglycosylated GccF_C43S are that (i) the less reactive Ser43 is not as efficiently glycosylated as the more reactive Cys at the end of a very flexible/mobile peptide chain, resulting in partial glycosylation at this position prior to export, and (ii) the monoglycosylated species is formed by host glycosidase-catalyzed hydrolysis of the more accessible (C-terminal) O-glycosidic bond (36). Glycosidase-catalyzed hydrolysis seems the most likely explanation for the 10-fold reduction in the IC₅₀ seen for the chemically synthesized fully glycosylated GccF_C43S analogue (21). The finding that the CFS of L. plantarum NC8(pRV610gcc_GccF_S18C) has an activity approximately 30% lower that of the wild-type CFS was unexpected, considering the results obtained for the chemically synthesized analogue which had an activity approximately 2-fold higher than that of wild-type GccF (21). Recombinant GccF purified from a 2-liter culture of L. plantarum NC8(pRV610gcc_GccF_S18C) had a mass of 5,215.0248 Da (see Fig. S12 in the supplemental material), indicating the presence of two Cys (S)-linked HexNAcs. A possible explanation for the reduction in activity is an attenuated production of fully active GccF, due to reduced glycosylation efficiency or compromised cellular processing. Regardless of these possibilities, it was shown that the Ser and Cys residues at positions 18 and 43 can be interchanged and still be glycosylated in vivo by GccA.

The ability of the cell to secrete nonglycosylated GccF_C43S was unexpected, as the current model for maturation predicts the GccF peptide is glycosylated within the cytosol prior to export (23). In retrospect this was not surprising, as other bacteriocins with similar (C-X₆-C)₂ scaffolds in which the loop contains no amino acids with modifiable hydroxy or thiol side chains have been identified in Bacillus cereus E33L. Although the associated gene cluster is similar to that which encodes sublancin, it does not contain a GTase gene, making it unlikely that these bacteriocin scaffolds are posttranslationally modified, consistent with the fact that for GccF_C43S, glycosylation is not an absolute prerequisite for export (37).

The position of the glycosylated Ser in the loop was also shown to be very important for GccF activity, in contrast to sublancin, where Cys22 could be moved to different positions in the loop and still be glycosylated with no significant change in antibacterial activity (24). Moving GccF Ser18 one residue toward the center of the GccF loop (i.e., to Ser17) without disrupting the sequence immediately preceding it appeared, within the limits of the tube assay, to abolish activity. There are at least two possible explanations for this lack of activity; the first is that this analogue scaffold cannot be processed by the GccF maturation machinery (e.g., Ser17 is not a suitable substrate for GccA), and the second is that glycosylation proceeds normally but the change in the precise presentation of the GlcNAc and the loop conformation prevents recognition of GccF by its receptor. As chemically synthesized, fully glycosylated GccF analogues in which the size of the loop was decreased to 7, 6, or 5 residues displayed moderately to severely attenuated activity (20), the fact that even more subtle changes in GccF loop structure could compromise receptor recognition seems plausible.

The remaining amino acid substitutions, chosen to investigate the roles that the loop residues may play in directing GccA to glycosylate Ser18, showed that while Tyr16 could be changed to Phe without any obvious effect on bacteriostatic activity, mutating Asp17 to its polar congener Asn appeared to completely abrogate activity, potentially due to the presumed absence of a GlcNAc on Ser18. Although both eukaryotic and prokaryotic protein O-glycosylation are now better understood, it is still unclear how Ser/Thr residues, often found in Ser/Thr-rich repeats, are selected for O-glycosylation (38, 39). An exception is protein O-glycosylation in Bacteroides fragilis, where the three-amino-acid glycosylation motif D-(S/T)-(A/I/L/M/T/V) seems to be required (40). In fact, mutation of the Asp to Ala or Glu led to a loss of glycosylation in the proteins tested (41). While we have not yet tested whether the D17N analogue is not glycosylated on Ser18, the results obtained with the B. fragilis GTase suggest that this may be the case. Examination of the genomic region of B. fragilis NCTC9343 involved in protein glycosylation (41) identified a protein containing a family 2 glycosyltransferase domain (BF9343_4192; residues 10 to 136; E value, 2.6e⁻²⁵). Comparing its amino acid sequence with that of GccA shows there is very weak sequence homology (E value, 1e⁻⁰⁸) over 95 residues that are part of the family 2 GTase domain of GccA (residues 10 to 104 of BF9343_4192 and 41 to 135 of GccA). Although the B. fragilis putative family 2 glycosyltransferase lacks a DXD motif, the less common DXE motif is found at residues 127 to 129. As family 2 peptide glycosyltransferases appear to be relatively rare, it is tempting to conclude that this group may indeed require the presence of an aspartic acid residue N-terminal to the Ser/Thr in order to be glycosylated.

Mutations designed to remove both Trx proteins showed that although GccF production was decreased, neither was essential. A report that neither of the SunA cluster thioredoxins is required for the biosynthesis of active sublancin, providing DsbA is produced by the cell (42), is consistent with these results. The L. plantarum NC8 genome encodes five named thioredoxins (TrxA1, TrxA2, TrxA3, TrxB, and TrxH), making it likely that one or more of these proteins are able to substitute for the inactivated putative thioredoxins GccC/D. Since the deletion of GccD seems to have the least effect on GccF activity (Fig. S18), it may be that this protein is only required for GccF maturation under conditions not used in these experiments.

As the coding sequence of gccD overlaps with the first 31 bp of the gccE ORF (Fig. S3), and the promoter for gccF extends into the 3′ end of gccE (Fig. 1D), it was not possible to completely delete gccE without affecting the flanking genes. The two mutations made to disrupt the function of GccE were expected to have similar phenotypes. Both mutations should allow transcriptional readthrough from P_gccA-F (located upstream of the gccA ORF) (Fig. 1C) that is responsible for the transcription of gccABCDEF (23). While the very low level of active GccF (undetectable in the CFS) purified from NC8(pRV610gcc_GccE_L148X) cultures is presumably due to such readthrough, it remains a puzzle as to why removal of the entire GccE protein should give higher levels of bacteriostatic activity (approximately 40% of wild type) and, indeed, how transcription is regulated in this cluster. It is interesting to note that two LytTR response regulators, PlnC (activator) and PlnD (repressor), control the transcription of genes in the chromosomal plantaricin locus. Good amino acid sequence similarity between GccE and PlnC/D is largely confined to the (C-terminal) LytTR domains of these three proteins (Fig. S20).

While these results have added to our understanding of factors affecting the maturation of GccF, there are still questions to be answered. The development of the easily modifiable plasmid pRV610gcc allows the production of GccF analogues with redesigned scaffolds, thus expanding glycocin diversity without using pathway refactoring (18) or synthetic approaches (20, 21). We hope that it may be applied to the development of more targeted glycocins that are less prone to develop resistance, thus expanding the armory needed to combat multidrug-resistant pathogens.

MATERIALS AND METHODS

Strains and culture conditions.

Transformed E. coli strains were grown at 37°C either in Luria broth (LB) containing 100 μg/ml ampicillin with shaking or on LB agar plates with the same selection. L. plantarum strains were grown at 30°C in de Man, Rogosa, and Sharpe (MRS) broth in static culture or on MRS agar plates. L. plantarum NC8 cells transformed with pRV610gcc were cultured under erythromycin selection (10 μg/ml) except when the culture was larger than 50 ml or when cell-free supernatant (CFS) was used to assess the inhibitory activity of secreted GccF. To test the ability of the cell machinery of other firmicute strains to process the gene cluster, L sakei Lb790 and Enterococcus faecalis JH2-2 were transformed with pRV610gcc and tested for both production of and resistance to GccF. Lb790 was also tested for resistance to erythromycin (the pRV610 plasmid contains an erythromycin resistance gene) and streptozotocin (a DNA-active analogue of GlcNAc). Details of all strains and plasmids used in this work are given in Tables S1 and S2 in the supplemental material.

Construction of pKS-gcc.

Genomic DNA was extracted from L. plantarum KW30 cells, purified by CsCl density gradient centrifugation, and digested to completion with SnaBI (NEB, Beverly, MA). The resulting blunt-end DNA fragments were fractionated by 0.6% agarose gel electrophoresis, and fragments in the 16- to 20-kbp range were extracted and purified (10). These fragments were then ligated into HpaI-digested pSMART-VC (Lucigen, WI), and transformed into E. cloni replicator electrocompetent cells (Lucigen, WI) to produce the vector pSMART-VC#18 containing a 17,952-bp insert that included the gcc cluster. pSMART-VC#18 was then digested with SacII (cuts 207 bp upstream of the “gccI” start codon) and AfeI (cuts 174 bp downstream of the gccF stop codon), and the resulting 6,838-bp fragment containing the gcc cluster was gel purified and ligated into SacII/AfeI-digested pKS-Arg (43), producing pKS-gcc. This high-copy-number E. coli-compatible plasmid contained the gcc cluster (gccH and gccA-F), and “gccI,” which is not part of the gcc cluster (12, 23).

pRV610gcc construction.

gcc forward and reverse primers (Table S3) were designed to add XbaI and XmaI restriction sites just outside the 5′ and 3′ boundaries of the gcc cluster DNA, respectively. PCR was used to amplify the gcc cluster from pKS-gcc or gcc PCR product templates (Biometra TRIO 48 [Biometra, Göttingen, Germany]) with Phusion polymerase (Thermo Scientific, MA). The following thermocycling program was used: initial denaturation, 98°C for 30 s; amplification, 30 cycles of 98°C for 10 s (denaturation), 60°C for 20 s (annealing), and 72°C for 4 min (extension); final extension, 72°C for 10 min. The QIAquick PCR purification kit (Qiagen, Hilden, Germany) was used to clean up the PCR product before it was ligated into the XbaI/XmaI-doubled digested shuttle vector pRV610 and transformed into E. coli EC100 cells. pRV610gcc was purified from E. coli EC100 using the High Pure plasmid isolation kit (Roche, Basel, Switzerland) and transformed into electrocompetent L. plantarum NC8 cells (31) to produce L. plantarum NC8(pRV610gcc). To confirm transformation of L. plantarum NC8, the plasmid was extracted using the High Pure plasmid isolation kit with the addition of a lysozyme (Sigma; MO) pretreatment step (44).

Mutagenesis of the pRV610gcc plasmid.

The site-directed ligase-independent mutagenesis (SLIM) method (45) was used to introduce mutations into pRV610gcc. Each mutation required two 40-plus-bp primers to introduce the desired mutation as well as two standard primers (Table S3). The mutant constructs created during this study were transformed into chemically competent E. coli EC100 cells, propagated, purified, and sequenced. Once the desired mutation in a plasmid had been verified, that construct was transformed into L. plantarum NC8 by electroporation. To confirm transformation, plasmids were extracted from cells as described above and subjected to restriction digestion analyses.

GccF purification.

GccF was purified from 1 liter of 3-day-old culture supernatant of L. plantarum NC8(pRV610gcc), as described in the supplemental material, and analyzed by mass spectrometry (Table S4). The concentration of GccF was determined by measuring the absorbance at 280 nm in a Cary 300 UV-visible spectrophotometer (Agilent Technologies Inc., Santa Clara, CA), using the predicted peptide-only molar absorptivity value of 18,700 liters mol⁻¹cm⁻¹ (46).

Biological activity assay.

Indicator plates were prepared by embedding GccF-susceptible L. plantarum ATCC 8014 cells in MRS agar (10). Once the plates had set, specific volumes (2 to 6 μl) of a known concentration of purified GccF, or CFS, were spotted on the surface and allowed to absorb into the agar. The plates were then sealed with Parafilm and incubated at 30°C overnight. Sample dilutions were made using either MRS broth or Milli-Q water. Alternatively, tubes containing indicator agar were prepared. Molten (40°C) MRS agar containing L. plantarum ATCC 8014 cells was carefully pipetted into 40- by 6-mm glass tubes to a depth of 15 to 20 mm. Once the agar had set, up to 50 μl of CFS was layered on top and left to absorb into the agar. The tubes were sealed with Parafilm and incubated at 30°C overnight. For IC₅₀ measurements, GccF was diluted in MRS, and 150 μl was pipetted into the wells of a flat-bottomed 96-well plate. Log-phase L. plantarum ATCC 8014 cells were then added to the wells to a final volume of 300 μl with an equal number of cells in each well. All assays were carried out in triplicate, and the optical density was measured at 600 nm on a MultiSkan GO plate reader (Thermo Scientific, MA) over 15 h at 30°C. The growth of inhibited and noninhibited cells was compared to produce plots of GccF inhibition over time. IC₅₀ values were calculated by plotting the maximum inhibition reached at each concentration of GccF and interpolating the concentration that would cause 50% growth inhibition.