The Abi Proteins and Their Involvement in Bacteriocin Self-Immunity

Abstract

The Abi protein family consists of putative membrane-bound metalloproteases. While they are involved in membrane anchoring of proteins in eukaryotes, little is known about their function in prokaryotes. In some known bacteriocin loci, Abi genes have been found downstream of bacteriocin structural genes (e.g., pln locus from Lactobacillus plantarum and sag locus from Streptococcus pyogenes), where they probably are involved in self-immunity. By modifying the profile hidden Markov model used to select Abi proteins in the Pfam protein family database, we show that this family is larger than presently recognized. Using bacteriocin-associated Abi genes as a means to search for novel bacteriocins in sequenced genomes, seven new bacteriocin-like loci were identified in Gram-positive bacteria. One such locus, from Lactobacillus sakei 23K, was selected for further experimental study, and it was confirmed that the bacteriocin-like genes (skkAB) exhibited antimicrobial activity when expressed in a heterologous host and that the associated Abi gene (skkI) conferred immunity against the cognate bacteriocin. Similar investigation of the Abi gene plnI and the Abi-like gene plnL from L. plantarum also confirmed their involvement in immunity to their cognate bacteriocins (PlnEF and PlnJK, respectively). Interestingly, the immunity genes from these three systems conferred a high degree of cross-immunity against each other's bacteriocins, suggesting the recognition of a common receptor. Site-directed mutagenesis demonstrated that the conserved motifs constituting the putative proteolytic active site of the Abi proteins are essential for the immunity function of SkkI, and to our knowledge, this represents a new concept in self-immunity.

Bacteriocins are ribosomally synthesized antimicrobial peptides and proteins produced by a wide variety of bacterial genera. The majority of bacteriocins from Gram-positive bacteria are classified into two groups: the class I lantibiotics, containing posttranslationally modified peptides with ring-forming lanthionine or methyllanthionine residues, and the nonmodified class II peptide bacteriocins (8, 33, 34). Class II bacteriocins are further subdivided into pediocin-like bacteriocins (class IIa), two-peptide bacteriocins (class IIb), and nonpediocin one-peptide bacteriocins (class IIc) (33). Bacteriocin-producing bacteria normally possess a mechanism of immunity to protect themselves from their own bacteriocins, and such self-immunity is often mediated by a dedicated protein (32). For a few bacteriocin systems, the mechanisms by which these proteins confer immunity have been elucidated. For instance, immunity to the lantibiotic nisin (class I) involves a combined action which includes (i) sequestering of bacteriocins on the bacterial cell membrane by a protein called NisI and (ii) removal of the bacteriocins from cells by a dedicated ABC transporter (NisFEG) (39, 44). On the other hand, proteins conferring immunity to pediocin-like bacteriocins (class IIa) as well as lactococcins A and B (class IIc) have been shown to bind directly to the bacteriocin receptor and thereby inhibit pore formation (13). Hitherto, no immunity mechanism is known for any class IIb two-peptide bacteriocins.

Recently, we reported that several bacteriocin loci encode proteins belonging to the Pfam Abi protein family (Pfam accession no. PF02517) (14). These loci include the plantaricin (pln) locus of Lactobacillus plantarum, encoding two two-peptide bacteriocins (12), the multibacteriocin pnc locus of Streptococcus pneumoniae (25), and the streptolysin S (sag) locus found in group A streptococci (35) (Fig. 1A). Some of the Abi proteins encoded in these loci (PlnI in L. plantarum, PncO in S. pneumoniae, and SagE in Streptococcus pyogenes) are probable bacteriocin self-immunity proteins on the basis of gene knockout studies (10, 25) and genetic organization (i.e., being closely associated with bacteriocin structural genes), while others (e.g., PlnP and PlnTUVW in L. plantarum and PncP in S. pneumoniae) have completely unknown functions.

FIG. 1.

Abi-associated bacteriocin and bacteriocin-like loci. (A) Three known bacteriocin loci containing Abi genes, including the pln locus of L. plantarum (14), the pnc locus of S. pneumoniae (25), and the sag locus of S. pyogenes (10). (B) Seven potential new bacteriocin loci identified by genome mining for bacteriocin-associated Abi genes. Abi genes are shown as black arrows, bacteriocin structural genes are shown in red, ABC transporter genes are shown in blue, regulatory genes are shown in green, and other genes are shown in white. Gene names or locus tags are shown below the arrows. The genes are drawn approximately to scale. The boxed arrows in the L. sakei 23K genome represent a disrupted histidine protein kinase gene.

The Abi family, also known as the CAAX amino-terminal protease family, consists of putative membrane-bound metalloproteases from both eukaryotes and prokaryotes (38), with the majority being bacterial proteins (90%). The Abi family is recognized by three highly conserved motifs (38): motif 1, consisting of two neighboring and invariant glutamate residues and a conserved arginine separated by three residues (EEXXXR, where X denotes any amino acid); motif 2, consisting of a conserved phenylalanine and a conserved histidine separated by three residues (FXXXH); and motif 3, with an invariant histidine. The three conserved motifs are thought to constitute the active site of the Abi protease, and their importance in proteolytic activity has been demonstrated by mutational analysis of the yeast Abi protease RCE1 (15). In eukaryotic cells, Abi family proteins are involved in membrane targeting of proteins harboring the C-terminal sequence CAAX (C, cysteine; A, aliphatic amino acid; and X, any amino acid) via a process known as prenylation, which consists of the following three sequential reactions (23): (i) a geranylgeranyltransferase/farnesyltransferase attaches a prenyl group (lipid anchor) to the cysteine in the fourth-to-last position, (ii) a CAAX protease of the Abi family cleaves off the three C-terminal amino acids (-AAX), and (iii) an isoprenylcysteine carboxyl methyltransferase then attaches a methyl group to the C-terminal cysteine.

Despite being widespread in prokaryotic genomes, the function of Abi proteins has not been investigated much in bacteria, with the exception of gene knockout experiments on the Abi genes sagE and pncO, which implies their involvement in bacteriocin self-immunity in streptococci (10, 25). Another Abi-like gene, prsW in Bacillus subtilis, has been studied to some extent (16). PrsW does not belong to any Pfam family but contains the same three motifs as the Abi proteins, with the exception of the conserved histidine in motif 2, which has been replaced by a glutamate in this protein. PrsW is a protease involved in response to antimicrobial peptides via a process known as regulated intracellular proteolysis (16). In this process, PrsW together with another protease (YluC) cleaves an anti-σ factor (RsiW) to release σ^W, which subsequently regulates gene expression in a manner that protects the cells from antimicrobial peptides. However, the PrsW target protein RsiW in B. subtilis does not harbor the C-terminal consensus sequence CAAX found for Abi target proteins in eukaryotes.

Pfam (http://pfam.sanger.ac.uk/) is a comprehensive collection of protein families and domains. For each protein family in Pfam, a profile hidden Markov model (profile HMM) is built from an alignment of sequences from nonredundant representatives of the family (seed sequences), and this profile HMM is then used in an iterative fashion to find new members of the protein family (18, 43). The Abi family in Pfam contains a large number of sequences (1,966 by September 2009). However, several proteins containing Abi or Abi-like motifs are somehow not detected by the current search tool. Examples of this include PrsW from B. subtilis and PlnL from L. plantarum, which apparently contain all three motifs but somehow do not fit into Pfam's Abi family. These observations suggest that the profile HMM for Abi may be based upon a slightly skewed sample of seed sequences, resulting in a low sensitivity. We provide here an updated profile HMM to detect Abi-like proteins in prokaryotes that are omitted in the present protein family. Furthermore, searches for novel Abi-associated bacteriocin loci were also performed in silico. Several potentially novel bacteriocin loci were identified, and one of them was assessed further by experimentation. The role of bacteriocin-associated Abi genes in self-immunity was also addressed by heterologous expression and site-directed mutagenesis.

MATERIALS AND METHODS

Construction of modified profile HMM for the Abi family.

The profile HMM for the Abi family (PF02517) in Pfam is based on 46 seed sequences (September 2009). A modified profile HMM was constructed here for more sensitive detection of prokaryotic members of the Abi family, as some proteins containing typical Abi motifs are not identified by the current algorithms (see below). This modified profile HMM was based on 49 seed sequences, including the 46 existing seed sequences as well as the Abi-like region sequences from the proteins SkkI (GenBank accession no. YP_395178), PlnL (NP_784203), and PrsW (CAB14210). First, the original profile HMM for the Abi family was downloaded from Pfam and aligned with each of the three new protein sequences to detect the region most similar to the existing Abi pattern. These regions from the new sequences were then added to the existing 46 seed sequences, making a list of 49 new seed sequences. From this extended collection of seed sequences, a multiple alignment was constructed, and a profile HMM was estimated from this multiple alignment, using the same model length (115 match states) as in the original Abi profile HMM. All computations were done in Matlab (MathWorks Inc.), using the profile HMM tools in the Bioinformatics Toolbox, which is fully compatible with Pfam. The Matlab script file is available upon request.

Assigning proteins to the modified Abi family.

To retrieve potential members of the Abi protein family, each of the 49 seed sequences was queried against the NR database of NCBI (ftp://ftp.ncbi.nih.gov/blast/db) by use of PSI-BLAST (2), with the maximum number of iterations set to 20 and with default parameters used otherwise. The hit sequences were then scored against the modified profile HMM. The score distribution showed a distinct bimodal shape, and a two-component Gaussian mixture model was fitted to separate high-scoring from low-scoring sequences. This was done by use of the mclust package (19) in R (www.r-project.org). Proteins with scores of >60 were assigned to the Abi family.

In silico identification of putative bacteriocin loci.

From the modified Abi family, we extracted the genes present in fully sequenced genomes (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi). Given the fact that genes involved in bacteriocin synthesis normally are clustered (32), Abi flanking genes (5 upstream and 5 downstream of each Abi gene) were evaluated for their resemblance to bacteriocin structural genes or other genes involved in bacteriocin biosynthesis. A gene qualified as a potential bacteriocin structural gene when several of the following criteria were met: the encoded prepeptide should be relatively short (50 to 85 amino acids [aa]) and contain an N-terminal secretion signal (double-glycine or sec-dependent leader), and the predicted mature peptide (20 to 65 aa) should be cationic (having an isoelectric point above 7). Furthermore, bacteriocin-related genes involved in immunity, transport (e.g., an ABC transporter and its accessory protein), and/or regulation (e.g., a histidine protein kinase and a response regulator) should be located in the vicinity.

Bacterial strains and growth conditions.

The bacterial strains used in this study are listed in Table 1. All lactobacilli were grown in de Man-Rogosa-Sharpe (MRS) medium (Oxoid) at 30°C without shaking. When appropriate, 5 μg ml⁻¹ erythromycin and/or 5 μg ml⁻¹ chloramphenicol was added to the growth medium for selection.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Relevant characteristicsa	Source or reference
Strains
L. plantarum 965	Strain sensitive to sakacin 23K	9
L. sakei Lb790(pSAK20)	Host strain for sakacin A (sap)-based expression system; carries pSAK20	4, 41
Plasmids
pSAK20	Camr; contains all genes necessary for transcriptional activation and processing of peptides with sap leader	4
pLPV111	Eryr; expression vector	4
pLG101	pLPV111 derivative containing genes for sakacin A (sapA) and sakacin A immunity (saiA) behind the sapA promoter	11
pSkkα	pLPV111 derivative containing genes for Skkα with sapA leader and for SkkI, behind the sap promoter	This study
pSkkβ	pLPV111 derivative containing genes for Skkβ with sapA leader and for SkkI, behind the sap promoter	This study
pMG36e	Eryr; expression vector containing P32 promoter	48
pSkkI	pMG36e with skkI cloned downstream of P32	This study
pSkkIE133A E135A	pSkkI with E133A E135A double mutation	This study
pSkkIE133Q E134Q	pSkkI with E133Q E134Q double mutation	This study
pSkkIH214D	pSkkI with H214D mutation	This study
pPlnI	pMG36e with plnI cloned downstream of P32	This study
pPlnLR	pMG36e with plnLR cloned downstream of P32	This study

^aCam^r, chloramphenicol resistance; Ery^r, erythromycin resistance.

Heterologous expression of sakacin 23K genes.

The genetic determinants for the putative two-peptide bacteriocin from Lactobacillus sakei 23K, here termed sakacin 23K (the individual peptides are called Skkα and Skkβ), were heterologously expressed in L. sakei Lb790 by use of the sap two-plasmid expression system (5). In this system, genes for transcriptional activation (sapRK) of sap promoters, as well as genes necessary for processing and export (sapTE) of peptides with the sap N-terminal double-glycine leader sequence, are located on one plasmid (pSAK20), while the other plasmid contains genes to be expressed downstream of a sapRK-dependent promoter (plasmid derived from pLPV111). The two genes encoding sakacin 23K were expressed separately as fusion genes, using the sap double-glycine leader sequence (4), and the fusion genes are called sap-skkA and sap-skkB (Fig. 2). To construct pSkkα, containing sap-skkA, the fragment containing the sap promoter and the sap leader was amplified from pLG101 (11), using primers dbd84F and mk189, and the fragment containing skkI was amplified using primers mk186 and mk187, with genomic DNA from L. sakei 23K as the template. The two overlapping primers mk189 and mk186 contain the skkA-encoding sequence (without a leader). The two PCR fragments were joined in a subsequent PCR step, using the outer primers dbd84F and mk187. The second plasmid, pSkkβ, containing sap-skkB, was constructed in a similar manner, using primers dbd84F and pr191 to amplify the fragment containing the sap promoter and the sap leader sequence and the primers mk194 and mk187 to amplify the fragment containing skkB (without a leader sequence) and skkI. Primer sequences can be found in Table S1 in the supplemental material. The final PCR products were cleaved with XhoI and HindIII and ligated into pLPV111 (4). The constructs were verified by DNA sequencing. Plasmids were transformed into L. sakei Lb790(pSAK20) by electroporation, as described previously (3).

FIG. 2.

sap-based expression system used for production of sakacin 23K. The sapA promoter (P_sapA) is located upstream of the bacteriocin fusion gene, skkA or skkB (gray arrow), and the immunity gene, skkI. Note that the fusion genes contain a sequence encoding the sap double-glycine leader (dashed bars) used for export. The restriction sites XhoI and HindIII were used for cloning.

Heterologous expression and site-directed mutagenesis of Abi immunity genes.

The skkI gene (GenBank accession no. LSA0564_c) was amplified using the primers dbd376 and dbd377, with genomic DNA from L. sakei 23K as the DNA template. The plnI gene and the plnLR gene pair were amplified from L. plantarum C11 genomic DNA by use of primers dbd371 and dbd372 for plnI and primers mk241 and dbd374 for plnLR. The PCR products were cleaved with restriction endonucleases NcoI and XhoI (New England Biolabs, Ipswich, MA) and ligated into pMG36e downstream of a strong, constitutive promoter (P32).

Site-directed mutagenesis of the conserved motifs in SkkI was performed using a two-step PCR approach (20). Three mutations were introduced into skkI, using dbd376 and dbd377 as outer primers and mutagenic primers mk206 and mk207 for the E133A and E134A mutations, mk204 and mk205 for the E133Q and E134Q mutations, and mk210 and mk211 for the H214D mutation. The mutated genes were cleaved and ligated into pMG36e as described above. All plasmids used in this study are listed in Table 1. All plasmids were verified by sequencing and were transformed into L. plantarum 965 by electroporation as described previously (3).

Bacteriocin assay.

The individual peptides of sakacin 23K were concentrated from culture supernatants by precipitation with 40% ammonium sulfate as described previously (13). The plantaricin peptides, PlnE, PlnF, PlnJ, and PlnK, were synthesized by GenScript (Piscataway, NJ). Bacteriocin activity was assessed using an agar diffusion assay in which the bacteriocin samples were spotted directly onto MRS soft agar containing the appropriate indicator strain. The plates were incubated overnight at 30°C, and bacteriocin activity was seen as clear inhibition zones. Alternatively, the bacteriocin activity was analyzed using a microtiter plate assay, where 1,000-fold-diluted overnight cultures of the indicator strain were exposed to 2-fold dilution series of the bacteriocins (1:1 molar ratio of each peptide) in a microtiter plate, with a total volume of 200 μl in each well. The plates were incubated for 12 h at 30°C before the inhibition was scored spectrophotometrically at 600 nm.

RESULTS

Reanalysis of the Abi protein family.

Each protein family in Pfam is based on a profile HMM which is constructed from a number of representative sequences called seed sequences, and a protein must score above a certain threshold value in order to be selected for the protein family (18, 43). By multiple sequence alignments, we observed that certain prokaryotic Abi-like proteins did not score above the threshold, indicating that the current profile HMM may be based on a skewed sample of seed sequences. To improve its sensitivity, we constructed a modified profile HMM, which was based on the Abi-like regions from PrsW (GenBank accession no. P50738), PlnL (CAA64196), and SkkI (YP_395178) in addition to the 46 seed sequences used for the Abi family in Pfam as of September 2009. By PSI-BLAST searches against NCBI's NR database, we obtained a total of 3,213 potential Abi proteins. These sequences were then scored against our modified profile HMM. The distribution of scores (histogram) is depicted in Fig. 3, where the curves illustrate a two-component Gaussian mixture model fitted to the histogram data. The optimal separation between low-scoring (green) and high-scoring (red) sequences is at a score of 48.5, but in order to reduce the number of false-positive results, we used a score of 60 as the threshold; hence, only proteins giving a score higher than the threshold value of 60 were qualified as members of the Abi family. With the modified algorithm, a total of 2,706 sequences were assigned to the Abi family, which is 740 more than the number found for Pfam's Abi family as of September 2009. After completion of this work, the Abi family in Pfam was revised (November 2009), using an updated profile HMM based on 155 seed sequences, and the size of this family has now increased from 1,966 to 4,443 members. Interestingly, although a large portion (95%) of the 740 new Abi sequences identified by our approach is included in the updated Abi family, a small but significant portion (37 sequences), including PlnL, SkkI, and PrsW, is still missing, demonstrating that the current profile HMM used for the Abi family is not sufficient.

FIG. 3.

Histogram of the bimodal score distribution obtained by scoring the 3,213 potential Abi proteins against the modified profile HMM. The overlaid curves illustrate a two-component Gaussian mixture model fitted to the histogram data to separate the low-scoring (green line) and high-scoring (red line) sequences. The threshold score for selecting proteins to the Abi family was set to 60 (black arrow).

In silico analysis of Abi flanking genes reveals seven novel bacteriocin-like loci.

Among the 2,706 Abi genes, 1,356 were found in 475 completely sequenced prokaryotic genomes. The number of Abi genes per genome was highly variable, e.g., with 15 in Bacillus anthracis but only a single or no Abi gene in many other genomes. In general, the number of Abi genes was much higher in the genomes of Gram-positive bacteria than in those of Gram-negative ones.

By examining Abi flanking genes in sequenced genomes, seven new bacteriocin-like loci were identified (Fig. 1B) in addition to the three known Abi-associated bacteriocin loci depicted in Fig. 1A. (A list of 1,356 Abi genes in sequenced genomes, as well as the functional annotation of 5 genes upstream and 5 genes downstream for each Abi gene, can be found in Table S2 in the supplemental material.) The vast majority of Abi genes were found in association with genes of various functions, and no predominant genetic organization involving the Abi genes was found, although some genetic links were often observed in a number of different bacterial species. For example, some Abi genes were often closely associated with ABC transporter genes (data not shown), suggesting that Abi proteins play an important but still unidentified role in many ABC transporter systems.

The seven new bacteriocin loci (Fig. 1B) were found in Gram-positive bacteria classified as Firmicutes, including five lactic acid bacteria which are either probiotic or used in food production. Four of the loci are from different species of Lactobacillus, one is from Leuconostoc mesenteroides, one is from Clostridium perfringens, and lastly, one locus is found in both Anoxybacillus flavithermus and Bacillus cereus. The seven new loci are diverse with respect to genetic organization, and they are also different from the three known bacteriocin loci containing Abi genes (pln, pnc, and sag) (Fig. 1A). A multiple sequence alignment of all bacteriocin-associated Abi proteins is shown in Fig. 4. Several of the putative bacteriocins encoded in the new loci display similarities with previously identified antimicrobial peptides (Table 2). The amino acid sequences and physiochemical properties of the putative bacteriocins are presented in Table 2.

FIG. 4.

Multiple sequence alignment of Abi proteins from the different bacteriocin loci shown in Fig. 1. Only the part containing the Abi motifs is shown, and the conserved residues in the three motifs are highlighted in bold. The alignment was constructed using T-COFFEE (www.tcoffee.org). The asterisks denote proteins (PlnL and SkkI) not classified as Abi proteins in the Pfam database.

TABLE 2.

Amino acid sequences of putative bacteriocins found in new loci

Peptide locus tag	Amino acid sequencea	Length (aa); pIb
Anoxybacillus flavithermus WK1 peptides
Aflv_2616d	MRMFQELNHVELQGIDGG-SWKSHVVNLVGVVSGFGTTGAVIGGSFGGPLGAAAGGFVGAHYGAVAYAIGVLLDSSNRRK	61; 9.91
Aflv_2615d	MEGVVFTMELMLEKNGSISFLSEEELKEIDGG-RGSWTNAVIGAGTLSPIVASAVRGAQQGVRFGRLGGPWGVVAGAVVGAVVG	58; 10.83
Aflv_2618	MKELKADELVSIDGG-ISACGKVSVAAFAHTAAFTGLMTFAGVSGPVGWVLGTFVGGAWLGASAAAGCLK	54; 8.90
Bacillus cereus BDRD-ST26 peptides
Bcere0013_55840	MEELKEFELENIDGG-SWKSHTINVVGNVATYGGIGTAICGPACGVVGAHYGAVAYGVGYLLDNK	49; 7.79
Bcere0013_55850d	MKLETNLNVVDLTNEELEINGG-GSWANASVGAGTGASVAVGALKGAQKGASWGSRVGPWGLAAGAVAGAAIGGYLAYD	56; 9.53
Clostridium perfringens SM101 peptide
CPR_1081d	MENLNLNQLENINGG-YWKTIWAVGPGLYQRDTETGKYRWIQTQDNLSYTTNVIANGWAGSAAGGYFSGR	54; 9.23
L. acidophilus NCFM peptides
LBA1791	MNKFKDLNELELSNIAGG-SNNIFWTRVGVGWAAEARCMIKPSLGNWTTKAVSCGAKGLYAAVRG	46; 10.06
LBA1797c	MEKLMVLNEEKLSYVIGG-GNPKVAHCASQIGRSTAWGAVSGAATGTAVGQAVGALGGALFGGSMGVIKGSAACVSYLTRHRHH	65; 10.32
LBA1800	MKKKVVKKTVLKEKELTKVVGG-KKAPISGYVGRGLWENLSNIFKHHK	25; 10.17
LBA1802	MKLRQEQLNRKELSQVIGG-RRDMILVALPHAVGPDGMPGSGRGGGAQMRAIGSIPPWRPNWWK	40; 8.33
LBA1803	MQEWKKTTLSDNELIDVIGG-SAKSYIRRLGPDGGYGGRESKLIAMADMIRRRI	33; 10.90
LBA1805	MKKLKVMNNGELEKVIGG-SLYEMKNSVPRLLGPDGMEGSMGGSTGGIQSFRHFPGFGR	44; 11.83
L. casei BL23 peptides
LCABL_25750	MISKEVGITLKQHDLVLIQGG-AKRRNKPSGCIVSTIGGAVAGAAGLNPFTTVAGAAIGLSLCLSTNYIHPA	50; 9.85
LCABL_25760d	MSYNYRQLDDFQLSGVSGG-KKKFDCAATFVTGITAGIGSGTITGLAGGPFGIIGGAVVGGNLGAVGSAIKCLGDGMQ	58; 8.82
L. johnsonii NCC 533 peptide
LJ0775bd	MSKFQQLTPEDLMETKGG-KIYHATPWQICNSKTHKCWADNAAIARTCGRVIVNGWLQHGPWGAR	46; 9.75
L. sakei 23K peptides
LSA0564_ad	MERISEYKVLNNNVLAGVQGG-KKKKGGFFWHYFGDPIVSFGKGFIGY	26; 9.88
LSA0564_bd	MNKKLDSFSSIEDDKLGLVIGG-RNNLAYGLGKLVRAGVDIGIAIGSKGRYKPRH	32; 11.03
Leuconostoc mesenteroides ATCC 8293 peptides
LEUM_0069d	MEKLSEQELAKVSGG-FPLLPIVVPIIAGGATYVAKDAWNHLDQIRSGWRKAGNSKW	41; 9.99
LEUM_0070	MDFKTQRNVLNSEKLMMISGG-STDDSWEGFGSGLHKTVNTVVYAGTTVARAHTRSHQRCFTGNKW	44; 9.18

^aThe dash indicates the predicted proteolytic cleavage site immediately after the double-glycine leader sequence.

^bLengths and pIs of the mature peptides (after removal of the leader sequence) were calculated using the Expasy Compute pI/Mw tool (http://au.expasy.org/tools/pi_tool.html).

^cThe underlined sequence shows 100% identity to the bacteriocin acidocin J1132αβ, a two-peptide bacteriocin from L. acidophilus (46).

^dPeptides displaying similarities with previously identified bacteriocins. Aflv_2615 and Aflv_2616 display 43.1% and 21.3% identity to peptides LF221B and LF221A of the bacteriocin acidocin LF221 from Lactobacillus gasseri (26), while Bcere0013_55850 displays 46.4% identity to the LF221B peptide. CPR_1081 is 33.3% identical to the class IIc lactococcal bacteriocin lactococcin A (21). LCABL_25760 displays 32.8% identity to thermophilin 13 from Streptococcus thermophilus (28). LJ0775_b is 32.6% identical with the class II sec-dependent bacteriocin hiracin JM79 from Enterococcus hirae (40). LSA0564_a and LSA0564_b are 50% and 25% identical to the β-peptide and α-peptide, respectively, of the two-peptide bacteriocin plantaricin S from L. plantarum (45). LEUM_0069 displays 34.1% identity with Orf4, a putative bacteriocin found in the plantaricin locus of L. plantarum J51 (31).

Sakacin 23K, a novel bacteriocin from L. sakei 23K.

In the putative bacteriocin operon of L. sakei 23K, the Abi gene is located immediately downstream of two bacteriocin-like genes (Fig. 1B), in a genetic organization identical to that of some known two-peptide bacteriocins, such as plantaricin JK and plantaricin EF in the pln system (Fig. 1A). To examine whether the two bacteriocin-like genes (called skkA and skkB) and the associated Abi gene (called skkI) in L. sakei 23K represent a novel bacteriocin system, these genes were selected for experimental analysis. L. sakei 23K has not been reported to produce any bacteriocins, and we could not detect any bacteriocin activity in the culture supernatant (data not shown). The regulatory histidine protein kinase gene in this locus appears to be disrupted by frameshift mutations, resulting in the formation of two open reading frames encoding different parts of the protein (Fig. 1B). Abolished bacteriocin production due to defects in the regulatory genes has been reported for Pediococcus pentosaceus ATCC 25745, and when this bacterium was complemented with an intact gene by cloning, normal bacteriocin production was restored (11). Similarly, to investigate whether the lack of active bacteriocins could result from no or low gene expression, the bacteriocin locus (GenBank accession no. LSA0564_abc) was cloned into a plasmid and overexpressed in the original host, L. sakei 23K. However, antimicrobial activity still could not be detected in the culture supernatant (data not shown), suggesting that a posttranslational process is probably defective.

To assess whether the two bacteriocin-like genes constitute a bacteriocin unit of class IIb, these two genes were then expressed separately as fusion genes in L. sakei Lb790, using the sap double-glycine leader for secretion (5). This strategy has previously proven successful for heterologous production of both one- and two-peptide bacteriocins (5, 11, 29, 36). The concentrated culture supernatants from the resulting clones were found to inhibit the growth of several Lactobacillus strains (Lactobacillus cellobiosus NCDO 928, Lactobacillus curvatus NCDO 2739, L. plantarum 965, L. plantarum LMGT 2379, L. plantarum LMGT 2389, L. sakei LMGT 2345, and L. sakei NCDO 2714) as well as a few strains of Listeria and Staphylococcus; the result for L. plantarum 965 is shown in Fig. 5A. Interestingly, both peptides (Skkα and Skkβ) displayed bacteriocin activity when assessed individually, and no major synergistic effect was seen when the two peptides were combined in different ratios. The antimicrobial activities of the two peptides were proteinase K sensitive and heat stable (intact activity after 10 min at 90°C).

FIG. 5.

Bacteriocin soft agar assay using L. plantarum 965 as the indicator bacterium. Antimicrobial activities of the peptides were assessed individually (Skkα or Skkβ) and in combination, at a ratio of 1:1 (Skkα and -β). L. plantarum 965 containing the empty expression vector was sensitive to both the individual and the combined peptides (A), but expression of the immunity protein SkkI rendered the cells immune (B). Site-directed mutations in the conserved Abi motifs, namely, E133A and E134A (C), E133Q and E134Q (D), and H214D (E), abolished the immunity function of SkkI. Bacteriocin sensitivity is seen as clear zones on the lawn of indicator cells.

Abi proteins confer bacteriocin immunity.

To examine whether the skkI gene, which is located just downstream of the two bacteriocin genes (skkAB) (Fig. 1B), could serve as an immunity determinant, this gene was cloned and expressed in the sakacin 23K-sensitive strain L. plantarum 965. As expected, expression of this gene rendered strain 965 immune to both peptides individually and in combination, confirming that skkI is an immunity gene (Fig. 5B and Table 3). In a similar manner, we also demonstrated that two other bacteriocin-associated Abi genes, plnI and plnL, could confer immunity to their cognate bacteriocins, plantaricins EF and JK, respectively (Table 3). These bacteriocin systems belong to the pln locus, in which plnI is located just downstream of the plnEF bacteriocin genes, while plnL is just downstream of the plnJK bacteriocin genes (Fig. 1A). Interestingly, while plnI functions alone as an immunity gene, plnL requires coexpression of the flanking gene plnR to confer immunity (data not shown). plnR encodes a putative protein of unknown function.

TABLE 3.

Immunity and cross-immunity by bacteriocin-associated Abi genes

Bacteriocin	MIC (AU) for straina
	965(pMG36e)	965(pSkkI)	965(pPlnI)	965(pPlnLR)
Sakacin 23K	1	>256	16	2
Plantaricin EF	1	>2	>2	1
Plantaricin JK	1	256	64	>1,024

^aOne arbitrary unit (AU) was defined as the minimum amount of bacteriocin required to produce 50% growth inhibition of L. plantarum 965(pMG36e) in a 200-μl microtiter plate. For the synthesized peptides plantaricins EF and JK, 1 AU corresponds to concentrations of 0.625 ng μl⁻¹ and 1.22 pg μl⁻¹, respectively.

Most known immunity genes investigated hitherto show strong specificity toward their cognate bacteriocins. For instance, the lactococcal immunity proteins LciA and LciB confer immunity only toward the cognate bacteriocins lactococcins A and B (class IIc), respectively, and no cross-immunity is found between them (47). Similar strong specificity has also been observed for most pediocin-like bacteriocins (class IIa) and their cognate immunity proteins (17). Surprisingly, we observed a relatively high degree of cross-immunity between the Abi bacteriocin systems (Table 3). The indicator strain L. plantarum 965 heterologously expressing skkI showed cross-immunity toward plantaricins EF and JK (>2- and 256-fold increased protection, respectively). Similarly, expression of plnI also protected against sakacin 23K (16-fold increased protection) and plantaricin JK (64-fold increased protection), in addition to the cognate, plantaricin EF. On the other hand, expression of plnLR showed less cross-immunity, as it gave no or very little protection against plantaricin EF and sakacin 23K.

Conserved motifs in the Abi proteins are essential for immunity function.

The immunity protein SkkI contains the three conserved motifs characteristic of the Abi protein family (Fig. 4). To our knowledge, the importance of these conserved motifs in the function of Abi proteins has not been studied in bacteria. By site-directed mutagenesis, we demonstrated that these motifs are indeed important for the immunity function. We found that the SkkI mutants in which the two invariant glutamate residues in motif 1 were changed to two alanine residues (E133A and E134A) (Fig. 5C) or two glutamine residues (E133Q and E134Q) (Fig. 5D) totally lost the ability to confer immunity. Likewise, replacement of the invariant histidine in motif 3 with an aspartate (H214D) (Fig. 5E) also had a detrimental effect on the immunity function of SkkI.

DISCUSSION

New bacteriocins have traditionally been identified by screening a large number of strains for bacteriocin production. However, the wealth of sequence information present today offers an invaluable source for genome mining of new antimicrobial peptides, and indeed, such in silico approaches have already proven successful for discovery of bacteriocin systems (6, 11). For instance, a novel two-peptide lantibiotic, lichenicidin, was identified following rational genome mining for LanM proteins, which are enzymes involved in posttranslational modification of lantibiotics (6). In the present work, we identified seven novel bacteriocin-like loci by genome mining for the bacteriocin-associated Abi genes, which encode putative metalloproteases also known as CAAX proteases.

Based on the observation that some clearly Abi-like proteins were left out of the Abi family in the Pfam database, we suspected a slightly skewed training set for this model. Extending the seed set from 46 to 49 sequences by including three Abi-like proteins, we re-estimated the HMM probabilities. By first using PSI-BLAST to search for Abi-like proteins and then filtering these through the new HMM, we ended up with 2,706 Abi proteins, which is over 700 more proteins than the number registered in the Pfam database for the Abi family by September 2009. We chose to use a conservative threshold value for assigning proteins to the Abi family, emphasizing specificity more than sensitivity. Based on the fitted mixture model, the threshold of 60 corresponds to an expected specificity of >95%, i.e., <5% false-positively detected Abi proteins (Fig. 3). Many of the new Abi proteins we found were recently assigned to the Abi family after an update of the Pfam database in November 2009. However, a number of Abi-like proteins (e.g., PrsW, PlnL, and SkkI) are still omitted from the family, suggesting that there are probably many more Abi proteins in prokaryotes than presently annotated in the databases. We suggest that our rationale should be taken into consideration in future updates of the Abi protein family.

Members of the Abi protein family are ubiquitous in the bacterial world, but surprisingly very little is known about their biological function(s) and how they take part in different pathways. In silico screening of Abi-containing genetic loci revealed that Abi genes were colocated with a wide variety of different genes, and this adds further to the notion that the Abi family appears to be a relatively promiscuous protein family involved in a diverse group of pathways. A small fraction of the Abi genes appear to be associated with bacteriocins, and among these, seven new bacteriocin-like loci were identified. All seven contain open reading frames encoding cationic peptides with an N-terminal double-glycine leader, features typical of bacteriocins (32). Like the case for the pln locus, some of these Abi genes are found just downstream of the bacteriocin structural genes (e.g., in L. sakei 23K), a genetic location resembling that of an immunity determinant, while others are found closely associated with genes involved in transport (e.g., in Leuconostoc mesenteroides ATCC 8293, A. flavithermus WK1, and B. cereus BDRD-ST26) or other functions (e.g., in Lactobacillus casei BL23). To our knowledge, none of the strains carrying the seven new bacteriocin-like gene clusters are known as bacteriocin producers, but this may be due to a lack of or suboptimal testing for bacteriocin production, or the bacteriocins may be nonactive, transcriptionally repressed, or possibly depleted of the machinery required for processing and export. Remnants of inactive bacteriocin loci are indeed relatively common in bacterial genomes (11, 30), and they represent an important source for discovery of novel bacteriocins. For instance, the genome of P. pentosaceus ATCC 25745 contains a bacteriocin locus which is inactive due to a naturally occurring frameshift mutation in the pheromone-encoding gene. When this bacterium was complemented with an intact pheromone gene by means of cloning, normal bacteriocin production was restored (11). The active bacteriocin itself (penocin A) was also produced when its gene was expressed in a suitable heterologous host (11). Similarly, we confirmed here that the bacteriocin structural genes skkA and skkB from the non-bacteriocin-producer L. sakei 23K were bacteriocinogenic, as they produced antimicrobial activity when expressed heterologously in a host that contains a complete regulatory network and a functional transport system. The two consecutive structural genes in the sakacin 23K locus suggest the presence of a two-peptide bacteriocin of class IIb whose optimal activity is, by definition, constituted by two different peptides (37). However, the Skkα and Skkβ peptides were active individually, and we could not detect any synergistic effects by combining them in different ratios, demonstrating that the sakacin 23K locus might contain two one-peptide bacteriocins. Still, we cannot exclude the possibility that these two peptides might act synergistically toward other bacterial strains. Moreover, potent antimicrobial activities of individual peptides in loci with a similar organization to that of sakacin 23K is not a unique feature, since such activity is found, for example, for the α-peptide from plantaricin S (22) and for LafA from the two-peptide bacteriocin lactacin F (1).

It has been shown that the Abi genes sagE and pncO are involved in self-immunity toward their cognate bacteriocins in streptococci (10, 25), and in the present work, three other Abi genes (skkI, plnI, and plnLR) were found to serve a similar immunity function. Most immunity determinants act specifically toward their cognate bacteriocins, and cross-immunity is a relatively rare phenomenon (8). Intriguingly, two of the three Abi immunity genes examined in the present study showed extensive cross-immunity. This fact might suggest that the encoded immunity proteins recognize and protect the same receptor(s) or pathway(s) in producer cells. Since these Abi genes encode putative metalloproteases, it is therefore tempting to speculate that Abi-mediated immunity arises from a proteolytic mechanism. This view is supported further by the fact that the conserved Abi motifs (Fig. 4), which are believed to constitute the active site of the enzyme (38), were found to be critical for the immunity function (Fig. 5). Based on these observations, it is reasonable to believe that the Abi proteases confer immunity by direct degradation of the bacteriocin, in a manner similar to that described for the extracellular metalloprotease gelatinase in Enterococcus faecalis (27, 42), or by modifying a receptor protein that can no longer be recognized by the cognate bacteriocin. It is also possible that Abi proteases confer immunity via complex transcriptional remodeling, such as occurs in the PrsW-mediated resistance to antimicrobial peptides in B. subtilis (7, 16). From some preliminary results, we could not observe any significant degradation of the bacteriocin when it was exposed to cell extracts from Abi-immune cells or to whole cells, thus disfavoring a bacteriocin degradation mechanism.

The critical role of the conserved motifs in the immunity function suggests that the mode of self-immunity mediated by Abi proteins is completely different from other known bacteriocin immunity mechanisms (e.g., immunity to nisin [44] or lactococcin A [13]). In several bacteriocin systems, self-immunity appears to be linked physically to the mechanism of receptor recognition. For example, class IIa and some class IIc bacteriocins employ membrane components of the mannose phosphotransferase system as target sites to form pores on sensitive cells, but in immune cells the immunity protein binds tightly to the target proteins to prevent the subsequent lethal pore formation by the bacteriocin (13, 24). Thus, studies of the immunity mechanism(s) and the protein(s) targeted by the Abi proteases could be a crucial step toward understanding target recognition and how the Abi-associated bacteriocins kill sensitive cells.