The Lactococcus lactis plasmidome: much learnt, yet still lots to discover

Abstract

Lactococcus lactis is used extensively worldwide for the production of a variety of fermented dairy products. The ability of L. lactis to successfully grow and acidify milk has long been known to be reliant on a number of plasmid-encoded traits. The recent availability of low-cost, high-quality genome sequencing, and the quest for novel, technologically desirable characteristics, such as novel flavour development and increased stress tolerance, has led to a steady increase in the number of available lactococcal plasmid sequences. We will review both well-known and very recent discoveries regarding plasmid-encoded traits of biotechnological significance. The acquired lactococcal plasmid sequence information has in recent years progressed our understanding of the origin of lactococcal dairy starter cultures. Salient points on the acquisition and evolution of lactococcal plasmids will be discussed in this review, as well as prospects of finding novel plasmid-encoded functions.

plasmid, bacteriophage, dairy fermentation, abortive infection, conjugation, transduction

Introduction

Lactococcus lactis is a member of a diverse bacterial group known as the lactic acid bacteria (LAB), which is a functional group of Gram-positive, micro-aerophilic coccoid and rod-shaped bacteria that produce lactic acid as the main end product of hexose fermentation (Makarova et al., 2006). Many LAB species are highly valuable due to the biotechnological properties they impart on fermented food products. These properties include organoleptic and rheological qualities, in addition to shelf-life extension. The latter is enabled by the ability of LAB to decrease the pH by lactic acid production and, in certain cases, the production of other antimicrobial substances such as bacteriocins (Cotter et al., 2005; de Vos, 2011).

Amongst the LAB, L. lactis is one of the most widely applied starter cultures in the dairy industry and for this reason has enjoyed extensive scientific scrutiny. The sheer quantity of fermented foodstuffs produced annually makes L. lactis one of the most economically important bacterial species, with recent estimates published in 2011 putting the collective economic value of cheese products alone (largely involving L. lactis strains) in the region of €55 billion (de Vos, 2011). Furthermore, the recent appraisal for the potential application of L. lactis strains in oral vaccine delivery may expand the importance of lactococcal investigations into the medical/pharmaceutical arena (Bermudez-Humaran et al., 2013).

Lactococcus lactis is found in many environments, although the original niche for L. lactis is now widely accepted to be plant based (Price et al., 2011; Siezen et al., 2011). Lactococcus lactis strains are readily isolated from fermenting plant material and minimally processed fresh fruit and vegetables, such as mung bean sprouts or corn, and in some cases have been presumed to represent the dominant element of the microbiota associated with these products (Kelly et al., 1998).

Lactococcus lactis is currently divided into several subspecies (Price et al., 2011). The classification of the two dominant subspecies, that is, subsp. lactis and cremoris, was originally based on industrially relevant phenotypic traits (Kelly et al., 2010). However, genetic analysis has highlighted that amongst dairy strains, two dominant genetic lineages (genotypes) of L. lactis strains exist, which are also appropriately termed lactis and cremoris (Samarzija et al., 2002; Rademaker et al., 2007; Passerini et al., 2010). In general, the genotypic distinction is matched by the phenotypic designations; however, atypical strains, whose phenotype and genotypes do not match, also exist (Wegmann et al., 2007).

Lactococci have unwittingly been exploited by humans for millennia for the production of an extensive range of dairy-based, fermented foodstuffs, such as cheese, and as a result enjoy a so-called generally recognised as safe (GRAS) status (Wegmann et al., 2007). Lactococcal strains that are used in the dairy industry appear to have undergone extensive adaptation to the nutrient-rich dairy environment through a process of reductive evolution (Bachmann et al., 2012), which, when compared to lactococcal strains isolated from plant material, appears to have resulted in a smaller genome size, a higher number of pseudogenes and acquisition of a much more extensive plasmid complement (Bolotin et al., 2004; Makarova et al., 2006; Kelly et al., 2010; Ainsworth et al., 2013). Plasmids are semi-autonomously replicating extrachromosomal DNA entities, which are normally dispensable for bacterial growth (Pinto et al., 2012). Plasmids typically confer traits that impart important niche-specific phenotypes (Siezen et al., 2011) and can often form the basis for individuality amongst strains (Kelly et al., 2010). The associated phenotypic traits may include industrially important survival strategies, metabolic capabilities, virulence factors and antibiotic resistances (Grohmann et al., 2003; Mills et al., 2006). Furthermore, genetic engineering of lactococcal plasmids has led to many highly efficient and successful biological tools (for reviews see Mierau & Kleerebezem, 2005; Mills et al., 2006; Morello et al., 2007), which has enabled many lactococcal functions and processes to be defined.

The following review aims to cover, amongst others, the most recently reported progress and discoveries related to lactococcal plasmids. In particular, the possible origins and evolution of lactococcal plasmids will be discussed, in addition to prospects for further discoveries of plasmid-associated traits.

Plasmid frequency

As mentioned above, the vast majority of dairy-associated lactococcal isolates carry extensive plasmid complements (estimated to be up to 14 per individual isolate; Kelly et al., 2010), commonly constituting more than 150 kb of extrachromosomal DNA and representing up to 9% of the total (i.e. chromosome plus plasmids) genetic material (Price et al., 2011; Ainsworth et al., 2013). These plasmids have sizes ranging from 2 kb to over 100 kb and have in several cases been shown to be mobilisable and transmissible by conjugation (Grohmann et al., 2003). The availability of complete plasmid complement sequences from several dairy-derived lactococcal strains has reaffirmed that key industrial traits, such as lactose metabolism, casein utilisation and many bacteriophage resistance systems, are plasmid-encoded, thereby demonstrating the extensive adaptation of such dairy strains to the milk environment (Siezen et al., 2005).

In contrast, plasmid profiling surveys of plant-derived lactococcal strains have highlighted that plasmids occur at a lower frequency (typically one or two, or no plasmids; Kelly et al., 2010) in plant isolates of L. lactis compared with their dairy counterparts (Kelly et al., 2010). This seems particularly true for plasmids of < 10 kb, which are common in dairy isolates (Siezen et al., 2005; Gorecki et al., 2011; Fallico et al., 2012; Ainsworth et al., 2013). Indeed, L. lactis KW2, a fermented corn isolate whose complete genome was recently sequenced, harbours no plasmids (Kelly et al., 2013). Nevertheless, plasmids of several plant and dairy isolates have been sequenced, and plasmids of lactococcal strains isolated from dairy samples, particularly strains isolated from raw milk products, hint at a plant-associated origin (Fallico et al., 2011) as will be discussed later.

There are currently more than 80 individual, completely sequenced lactococcal plasmids present in public databases, including the entire plasmid complement of five dairy isolates and a single human isolate (Table 1). Additionally, partial sequences of many lactococcal plasmids are also available. Due to the fact that, as mentioned above, lactococcal plant isolates harbour relatively few, if any, plasmids (Kelly et al., 2010), and because only a relatively small number of plant isolates have been sequenced, such plasmid sequences are clearly under-represented. Currently, DNA sequences of just two plant-associated lactococcal plasmids are publicly available (Tanous et al., 2007; Siezen et al., 2010). The complete plasmid complements of two human lactococcal isolates, L. lactis CV56 (three plasmids; Gao et al., 2011) and Lactococcus garviae 21881 (five plasmids; Aguado-Urda et al., 2012), have recently been made available, revealing many plasmid-encoded features that are common amongst the plasmids found in such human isolates and those present in their dairy-derived relatives.

Table 1

List of all sequenced Lactococcus lactis plasmids to date. Data obtained from the GenBank L. lactis annotation report (http://www.ncbi.nlm.nih.gov/genome/genomes/156?subset=plasmids&details=on&)

Name	Accession number	Size (Kbp)	GC%	Genes	Niche	Replication mode
pKF147A	NC_013657.1	37.51	32.4	29	Plant	Theta
pSK111	NC_008503.1	14.041	34.4	10	Dairy	Theta
pSK112	NC_008504.1	9.554	30.4	6	Dairy	Theta
pSK113	NC_008505.1	74.75	35.4	61	Dairy	Theta
pSK114	NC_008506.1	47.208	34.8	35	Dairy	Theta
pSK115	NC_008507.1	14.206	33.5	8	Dairy	Theta
pCV56A	NC_017483.1	44.098	32.1	39	Human	Theta
pCV56B	NC_017487.1	35.934	34.5	28	Human	Theta
pCV56C	NC_017484.1	31.442	32.5	29	Human	Theta
pCV56D	NC_017485.1	5.543	32.2	8	Human	Theta
pCV56E	NC_017488.1	2.262	33.8	3	Human	Theta
pQA504	NC_017497.1	3.978	37.8	3	Dairy	?
pQA518	NC_017495.1	17.661	37.4	12	Dairy	Theta
pQA549	NC_017493.1	49.219	35.1	42	Dairy	Theta
pQA554	NC_017496.1	53.63	34.9	69	Dairy	Theta
pCIS8	NC_019430.1	80.592	34	67	Dairy	Theta
pCIS7	NC_019431.1	53.051	32.4	41	Dairy	Theta
pCIS1	NC_019438.1	4.263	32	2	Dairy	Theta
pCIS5	NC_019432.1	11.676	34.1	11	Dairy	Theta
pCIS3	NC_019433.1	6.159	35.9	4	Dairy	Theta
pCIS2	NC_019434.1	5.461	30.1	5	Dairy	Theta
pCIS6	NC_019436.1	38.673	37.1	27	Dairy	Theta
pCIS4	NC_019437.1	7.045	38.4	8	Dairy	Theta
pMRC01	NC_001949.1	60.232	30.1	63	Dairy	Theta
pVF18	NC_015900.1	18.977	33.9	21	Dairy	Theta
pVF22	NC_015901.1	22.166	35.1	19	Dairy	Theta
pVF50	NC_015902.1	53.876	34.5	41	Dairy	Theta
pVF21	NC_015912.1	21.728	33.6	14	Dairy	Theta
pAW153	NC_017494.1	7.122	31.4	8	?	Theta
pIL6	NC_019308.1	28.434	33.6	25	Dairy	Theta
pAF04	NC_019347.1	3.801	32	4	Dairy	Theta
pAF07	NC_019348.1	7.435	36.4	6	Dairy	Theta
pAF12	NC_019349.1	12.067	33.3	11	Dairy	Theta
pAF14	NC_019350.1	14.419	34.1	11	Dairy	Theta
pAF22	NC_019351.1	22.388	34.9	23	Dairy	Theta
pLP712	NC_019377.1	55.395	37.4	44	Dairy	Theta
pIL105	NC_000906.2	8.506	29.8	7	Dairy	Theta
pNZ4000	NC_002137.1	42.81	33.3	45	Dairy	Theta
pAG6	NC_007191.1	8.663	33.7	8	?	Theta
pS7a	NC_004652.1	7.302	33.4	5	Dairy	Theta
pWV02	NC_002193.1	3.826	31.3	1	Dairy	Theta
pSRQ800	NC_004960.1	7.858	31.3	7	Dairy	Theta
pHP003	NC_004847.1	13.433	40.1	6	Dairy	Theta
pGdh442	NC_009435.1	68.319	35.1	63	Plant	Theta
pMN5	NC_004922.1	5.67	30.3	4	Dairy	RCR
pCI305	NC_002502.1	8.694	32.4	8	Dairy	Theta
pBL1	NC_004955.1	10.899	32.6	8	Dairy	Theta
pCL2.1	NC_004981.2	2.047	34	2	?	RCR
pDR1-1	NC_004164.2	7.412	33.7	6	Dairy	Theta
pAH82	NC_004966.1	20.331	34.4	17	Dairy	Theta
pSRQ900	NC_004959.1	10.836	31.1	11	Dairy	Theta
pKL001	NC_011610.1	6.068	32.9	4	?	Theta
pWV01	NC_002192.1	2.178	33.4	4	Dairy	RCR
pNP40	NC_010901.1	64.98	32.3	62	Dairy	Theta
pAR141	NC_013783.1	1.594	36.1	2	Dairy	RCR
pS7b	NC_004653.1	7.264	33.6	5	Dairy	Theta
pIL1	NC_015860.1	6.382	32.3	7	Dairy	Theta
pIL3	NC_015861.1	19.244	35.1	20	Dairy	Theta
pIL5	NC_015863.1	23.395	34.5	22	Dairy	Theta
pIL7	NC_015864.1	28.546	34.1	26	Dairy	Theta
pIL4	NC_015862.1	48.978	35.1	47	Dairy	Theta
pDBORO	NC_009137.1	16.404	35.2	15	?	Theta
pDR1-1B	NC_004163.1	7.344	33.7	6	?	Theta
pK214	NC_009751.1	29.871	32.4	29	?	Theta
pIL2	NC_017489.1	8.277	34.8	10	Dairy	Theta
pSK11L	NC_017478.1	47.165	34.8	40	Dairy	Theta
pCRL291.1	NC_002799.1	4.64	33.5	3	?	Theta
pCRL1127	NC_003101.1	8.278	34.8	7	?	Theta
pBM02	NC_004930.1	3.854	35.7	6	Dairy	RCR
pND324	NC_008436.1	3.602	33.4	3	?	Theta
pSRQ700	NC_002798.1	7.784	34.2	9	Dairy	Theta
pAH33	NC_002150.1	6.159	35.9	7	Dairy	Theta
pKP1	NC_016042.1	16.181	35.9	7	Dairy	Theta
pCIS3	NC_002138.1	6.159	35.9	3	Dairy	Theta
pSK11A	NC_017498.1	10.372	30.9	13	Dairy	Theta
pSK11P	NC_017500.1	75.814	35.4	61	Dairy	Theta
pSK11B	NC_013551.1	13.332	34.3	14	Dairy	Theta
pL2	NC_008594.1	5.299	32.5	5	Dairy	Theta
pWC1	NC_004980.1	2.846	29.5	1	Dairy	RCR
pCD4	NC_002748.1	6.094	33.4	5	Dairy	Theta
p1	NC_022587.1	4.094	30.0	6	Dairy	RCR

Presumed replication mode.? = unknown.

Replication of lactococcal plasmids

Lactococcal plasmids, like the majority of plasmids, replicate via two alternative modes of replication: rolling circle replication (RCR; Leenhouts et al., 1991) or theta-type replication (Kiewiet et al., 1993). Both replication mechanisms require certain functions from the DNA replication machinery of the host, and a brief outline of these two replication methods is given below (for a detailed review on lactococcal plasmid replication, see Mills et al., 2006).

The replicon of a typical rolling circle plasmid is composed of a replication protein (encoded by rep) and a double-stranded origin (dso) of replication, which contains a so-called nic site, composed of (an) inverted repeat(s), and a Rep-binding site, which is comprised of a set of two to three direct repeats or an inverted repeat (del Solar et al., 1993; Mills et al., 2006). The Rep initiates replication by introducing a single-stranded break into the nic site of the dso, generating a free 3′ hydroxyl group, which is subsequently used in leading strand synthesis by the host replication machinery. Replication displaces the parental ‘plus’ strand and continues until the newly reconstituted dso is reached. Lagging-strand replication on the displaced parental strand occurs from a noncoding region, which can generate a stem loop structure, termed the single-stranded origin (sso; del Solar et al., 1998). Rolling-circle-type plasmids always replicate in a unidirectional manner and represent the minority of sequenced (seven of 82 analysed lactococcal plasmids are known or predicted to replicate via the RCR mechanism) lactococcal plasmids (Siezen et al., 2005; Gorecki et al., 2011). Most lactococcal rolling-circle-type plasmids belong to the pWV01 family of replicons and exhibit a broad host range, being able to replicate in a range of Gram-positive and Gram-negative bacteria. However, they have a limited replicon size (< 10 kb) and are incompatible with other RCR plasmids (Leenhouts et al., 1991), limiting each strain to a maximum of just a single RCR replicon. These limitations, coupled to their intrinsic structural and segregational instability, are amongst the likely reasons for the under-representation of rolling circle replicons in currently sequenced L. lactis plasmids.

The majority (75 of 82) of lactococcal plasmids appear to replicate via the theta-type mechanism of replication (Seegers et al., 1994; Table 1). Large (80 kb) theta replicon plasmids have been sequenced (Ainsworth et al., 2013) reflecting their superior structural stability relative to RCR-type plasmids. They appear to be highly related and are described as pWV02-type replicons, being named after the plasmid that serves as the prototype of this group of replicons (Kiewiet et al., 1993). Unlike RCR plasmids, these replicons have a limited host range (Kiewiet et al., 1993). The typical replicon of a lactococcal theta-type-replicating plasmid comprises of a replication initiator protein, encoded by a gene that is often termed repB, and an origin of replication (ori) which is comprised of an AT-rich region and three and a half iterons of 22 bp in length (Fig. 2b). Additionally, two small inverted repeats overlap the -35 site of the repB promoter and the ribosome binding site (Kiewiet et al., 1993; Mills et al., 2006). DNA synthesis of theta plasmids can be uni- or bidirectional and may initiate from multiple origins; synthesis of the leading strand is continuous, while lagging-strand synthesis is discontinuous.

In contrast to the rolling circle plasmids, multiple theta replicons may coexist in the same lactococcal strain or even on the same plasmid (Seegers et al., 1994). The sequencing of complete plasmid complements has highlighted the extent to which highly related theta replicons can coexist within the same host strain or plasmid. For example, the plasmid complement of L. lactis SK11 contains six theta-type replicons across four plasmids, with two of such replicons being present on pSK11L and pSK11B (Siezen et al., 2005), while the L. lactis IL594 seven plasmid complement harbours nine theta-type replicons (Gorecki et al., 2011). These observations are in apparent contrast to the generally accepted notion that the more related two replicons are, the greater their degree of incompatibility. This phenomenon is well described (Seegers et al., 1994; Gravesen et al., 1997; Emond et al., 2001; Gorecki et al., 2011), and Gravesen et al. (1997) performed a screen for incompatible theta replicons, identifying two pairs of incompatible plasmids in their study. Analysis of these incompatible replicons suggested that the incompatibility determinant was contained within the 22-bp direct repeats and/or in the inverted repeat IR1 of the ori, which has been proposed to interact with a specific 13 amino acid region of RepB (Gravesen et al., 1997). Accordingly, analysis of this 13 amino acid region revealed nine variable residues amongst compatible plasmids, whereas incompatible plasmids were shown to harbour identical amino acid sequences within this region. Analysis of the nucleotide sequences of direct and inverted repeat regions of the nine IL594 replicons revealed that they all varied slightly in nucleotide sequence (Gorecki et al., 2011). The authors suggest that each ori region interacts uniquely and specifically with its corresponding RepB, allowing coexistence of several related replicons within one cell.

The key traits of dairy lactococci: lactose and casein utilisation

The ability to rapidly ferment lactose, a typical feature of LAB, is a plasmid-encoded and well-defined characteristic amongst dairy-associated lactococci (Cords et al., 1974). The genes encoding lactose acquisition and utilisation are located in a single operon, lacABCDFEGX, which is regulated by the repressor LacRii encoded by the divergently oriented lacRii gene. Lactose is acquired through the phosphoenolpyruvate-phosphotransferase system (PEP-PTS), which is encoded by lacEF, and which catalyses the synthesis of lactose-phosphate as part of lactose transport across the cell membrane. Once inside the cell, it is utilised by the tagatose-6-phosphate enzymes (encoded by lacABCD) through the action of phospho-β-galactosidase (lacG), which cleaves lactose-phosphate to galactose-6-phosphate and glucose (Mills et al., 2006).

Recently, the lactose PEP-PTS system has been implicated in galactose uptake and metabolism (Neves et al., 2010). A mutation in galP, encoding the high-affinity galactose permease, in L. lactis NZ9000 was shown not to abolish galactose uptake in the presence of the lactose PEP-PTS. Further investigation determined that the lactose PEP-PTS system possesses a low affinity for galactose. Once transported inside the cell, galactose can be metabolised via the tagatose pathway. However, this system alone cannot sustain growth on galactose, and it has been hypothesised that such growth is due to low fructose 1,6-bisphosphatase activity, which impedes gluconeogenic use of galactose (Benthin et al., 1994; Neves et al., 2010).

Loss of the lactose-metabolising phenotype is a well-documented occurrence (McKay et al., 1972). Sequencing of entire lactococcal plasmid complements from dairy isolates has shown that lactose utilisation is predominantly associated with one of the larger plasmids harboured by a given strain, where the same plasmid in most investigated cases also encodes proteins crucial for casein degradation (Siezen et al., 2005; Wegmann et al., 2012; Ainsworth et al., 2013). The absence of the lactose utilisation operon in the plasmid complements of human isolates L. lactis CV56 (Gao et al., 2011) and L. garviae 21881 (Aguado-Urda et al., 2012), and the plant isolate L. lactis KF147 (Siezen et al., 2010) is indicative of the specific, nondairy niche that such strains occupy.

The primary source of amino acids in the dairy environment is in the form of milk proteins, the majority of which is casein (Swaisgood, 1982). Casein hydrolysis by lactococcal proteases and peptidases is vital for desirable flavour development due to the resulting formation of aroma compounds from peptide/amino acid catabolism of casein breakdown products (Steele et al., 2013). The major L. lactis extracellular, cell wall-anchored protease, PrtP (for a recent review see Steele et al., 2013), is responsible for casein hydrolysis and is almost always plasmid-encoded (Savijoki et al., 2006). PrtP exhibits broad substrate specificity and can cleave caseins into over 100 different oligopeptides (Juillard et al., 1995; Savijoki et al., 2006). The extracellular hydrolysis products are then taken into the cell by the oligopeptide uptake and transport system, encoded by the opp operon (Yu et al., 1996), after which the internalised peptides are further digested by intracellular peptidases with varying substrate specificities (Savijoki et al., 2006; Steele et al., 2013). The opp operon is generally plasmid-encoded, and the opp operon, the prtP-specified proteinase and its associated plasmid-encoded maturase PrtM are required for the proteolysis phenotype (Yu et al., 1996). The recent availability of complete plasmid complements of individual strains has revealed that the opp operon and prtP are often located on separate plasmids, and such strains are unable to grow on casein as a sole source of nitrogen/amino acids if one of these plasmids is lost from a given strain (Siezen et al., 2005; Ainsworth et al., 2013). The dispersal of this essential (in a dairy environment) phenotype across two plasmids may thus act as a crude plasmid stability system amongst dairy lactococci.

In addition to PrtP and Opp, as described above, the proteolytic system of L. lactis also constitutes various peptidases which further act upon the internalised casein hydrolysis products (Steele et al., 2013). A genomic comparison of the proteolytic systems of 39 L. lactis strains using CGH analysis detected variability in the presence or absence of genes encoding the peptidases Pcp, PepO2, PepF2 and PepX2, which the authors hypothesised was directly due to the presence or absence of plasmids harbouring these genes (Liu et al., 2010).

In addition to characterised peptidase-encoding genes, plasmid sequencing has revealed several hypothetical proteins which possess peptidase-like domains (Siezen et al., 2005; Ainsworth et al., 2013). These uncharacterised proteases are often strain-/plasmid-specific and may be associated with specific organoleptic profiles of particular fermented dairy products (Steele et al., 2013). An example of a plasmid-associated gene specifying a peptidase-like domain is that harboured by L. lactis UC509.9 plasmid pCIS8. Bioinformatic analysis of the plasmid-borne open reading frame (ORF) uc509_p8059 shows that its protein product is a likely extracellular protein, containing a C-terminal hydrophobic region with a conserved tryptophan (ChW) repeat adhesion/cell adhesion domain and a transglutaminase-like protease. ChW domain-containing proteins appear to be almost exclusively limited to Clostridium acetobutylicum (Sullivan et al., 2007) and have been implicated in the degradation of polysaccharides and proteins (Nölling et al., 2001; Sullivan et al., 2007). The N-terminus of the predicted protein product of uc509_p8059 contains a C47 peptidase family domain, which is commonly associated with proteolytic activities encoded by Staphylococcus and Enterococcus species, and is predicted to possess a single cleavage site. The C47 domain contains 24% identity across the full length of the domain with the C47 domain of S. aureus virulence factor staphopain A, which is a broad-specificity protease (Dubin et al., 2001). Interestingly, a clear homolog of uc509_p8059 is present on the chromosome of the plant-derived L. lactis KF147.

Novel plasmid traits

Current industrial L. lactis strains that genotypically or phenotypically belong to the subspecies cremoris are thought to be the descendants of a relatively small number or genetically related lineages, which were widely disseminated around the globe during the early 20th century with the advent of industrial fermentations (Lawrence et al., 1978; Kelly et al., 2010). While these related strains may present distinct phenotypes, such as bacteriophage resistance (Ward et al., 2004), it has been suggested that large redundancies exist in academic and industrial collections (Kelly et al., 2010). Evidence for this low diversity is obvious from the high level of chromosomal homogeneity amongst industrial lactococcal dairy strains as compared to the more genetically diverse artisanal and plant strains (Rademaker et al., 2007; Kelly et al., 2010). Constant selective pressure as applied in dairy fermentation may nevertheless have caused industrial strains to develop distinctive traits, in terms of mobile elements, prophages and possibly plasmid complements (Le Bourgeois et al., 2000).

Due to the presumed low diversity of current starter strain collections (Kelly et al., 2010), new environmental strains with desirable traits, such as higher stress tolerance and new flavour or aroma development, are eagerly sought to be exploited for particular industrial processes (21, 22). Plant-derived and artisanal isolates exhibit phenotypes that may be suitable for industrial exploitation, such as increased stress tolerance (Nomura et al., 2006), novel flavour compound formation, food-grade markers and production of broad-spectrum bacteriocins (Kelly et al., 1998; Tanous et al., 2007; Fallico et al., 2011). Many of these traits are plasmid-encoded and therefore have the potential to be transferred to dairy strains (Fallico et al., 2011). Research investigating lactococci from diverse environments has also uncovered some interesting, although in cases less desirable traits, such as antibiotic resistance. Furthermore, sequencing has revealed potentially beneficial dairy adaptations, including stress resistance mechanisms and supplementary mineral uptake systems, such as Mg²⁺ and Mn²⁺ transporters (Seegers et al., 2000; Mills et al., 2006), which were previously unknown due to the absence of an easily attributable phenotype (Siezen et al., 2005).

Flavour development

New flavour and aroma development is a highly desirable trait in industry. Traditionally, L. lactis subsp. lactis biovar diacetalyactis strains have been exploited in dairy fermentations for their typical ability to metabolise citrate, which leads to desirable aroma and flavour development through the synthesis of diacetyl (Starrenburg & Hugenholtz, 1991). Citrate uptake is possible due to the presence of citP, encoding a citrate permease (Sesma et al., 1990), which is often a plasmid-encoded trait (Mills et al., 2006). More recently, glutamate dehydrogenase (GDH)-encoding genes have been observed on plasmid sequences derived from plant and raw milk isolate strains (Tanous et al., 2005; Fallico et al., 2011). GDH catalyses the reversible oxidative deamination of glutamate to 2-oxoglutarate and ammonia using NADP as a cofactor. As a result, amino acid to aroma compound conversion is stimulated through the supply of 2-oxoglutarate which is required for transamination (Tanous et al., 2002, 2007). In a plant environment, this ability is hypothesised to enable the (lactococcal) carrier to assimilate ammonia in an amino acid-deprived environment (Tanous et al., 2005).

Stress response

Dairy fermentations and cheese ripening impart many stresses upon lactococcal cells. These stresses include changes in pH, temperature and osmolarity (Duwat et al., 2000). Plasmid (complement) sequencing has revealed that L. lactis plasmids can encode multiple stress-associated genes, which could aid in obtaining strain robustness. Many of the identified plasmid-encoded stress proteins appear to be duplicates of genes encoded on the L. lactis chromosome (Siezen et al., 2005) and therefore are presumed to enhance survival through a gene-dosage effect.

The gene encoding the universal stress protein, uspA, has been detected in all currently sequenced dairy plasmid complements. The universal stress protein superfamily is conserved in many domains of life and is thought to play a role in protecting the cell from DNA-damaging agents (Kvint et al., 2003). In Escherichia coli, UspA is expressed in response to multiple environmental stress conditions, such as carbon and nutrient starvation, heat exposure and oxidation, amongst many others (Kvint et al., 2003). The plasmid complement of L. lactis SK11 encodes, in addition to a plasmid-specified UspA, carbon starvation and cold-shock proteins (Siezen et al., 2005), both of which may enhance strain survival during dairy fermentation. Two predicted cold-shock proteins are also encoded on plasmid pNP40, and transcriptional fusions of promoter regions from the two corresponding genes, cspC and cspD, of pNP40 have indeed demonstrated induction by cold shock (O'Driscoll et al., 2006). Furthermore, a (small but) significantly increased level of survival was observed for cells harbouring pNP40 undergoing freeze–thawing as compared to plasmid-free cells (O'Driscoll et al., 2006).

Another example of a plasmid-encoded stress response system is found as two predicted, individual copies of the stress response-induced, low-fidelity polymerase, encoded by umuC (Ainsworth et al., 2013) on L. lactis UC509.9 plasmid pCIS7. UmuC homologs in E. coli are known to function as a lesion-bypass DNA repair system as part of the SOS response (Maor-Shoshani et al., 2000). Predicted UmuC proteins in L. lactis are encoded on several plasmids, for example plasmid pKF147A, found in a lactococcal plant isolate, and plasmid pNP40, which originates from a dairy isolate. Expression of the UmuC homolog from pNP40 was monitored by transcriptional fusion postexposure of the host strain with mitomycin C, which resulted in a threefold transcriptional increase in umuC promoter activity. In addition to UmuC, pNP40 is predicted to encode two further DNA-damage repair functions. A homolog of RecA (Garvey et al., 1997; Lusetti & Cox, 2002), which functions in DNA repair, the SOS response and homologous recombination, is encoded by orf18, whereas orf25 encodes a protein with a putative UvrA excinuclease domain (Van Houten et al., 2005), implicating this protein in nucleotide excision repair. Finally, genes encoding hypothetical proteins with BRCT (BRCA1 carboxyl-terminal) domains are present in the plasmid complements of strains UC509.9, SK11, IL594, A76 and DPC3901. BRCT domain proteins are predicted to function in DNA-damage repair (Makiniemi et al., 2001) and may act as a further stress response mechanism.

Plasmid-encoded antimicrobial peptides and corresponding immunity

Bacteriocins are small, ribosomally synthesised antimicrobial peptides which generally induce cell death through disruption of membranes and cell wall biosynthesis. Their action can be narrow (targeting bacteria of the same species) or broad (targeting bacteria across genera; Cotter et al., 2012). Many bacteriocins are produced by LAB, although very few are commercially exploited in the food industry in food preservation and safety (Cotter et al., 2005). Bacteriocin production in L. lactis (and other LAB) is very well investigated, and several lactococcal bacteriocin production and immunity systems have been shown to be plasmid encoded (for reviews see (Cotter et al., 2005; Mills et al., 2006; Zendo et al., 2010; Cotter et al., 2012).

Most lactococcal bacteriocins appear to possess a narrow spectrum activity, thereby killing only closely related bacteria. Furthermore, strains producing bacteriocins must also encode an immunity determinant for self-protection. Immunity proteins are usually found in operons encoding the bacteriocin they are providing immunity against. However, plasmid sequencing has revealed that several lactococcal plasmids harbour solitary bacteriocin immunity genes. Examples include the nisin resistance determinant nsr found on plasmids pAF65 (Fallico et al., 2012) and pNP40 (O'Driscoll et al., 2006), harboured in L. lactis subsp. lactis biovar. diacetylactis strains DPC3758 and DRC3, respectively. The Nsr protein confers nisin resistance by proteolytic cleavage of the carboxy-terminal of the nisin tail (Sun et al., 2009). Additionally, located on pAF65 is lctFEG, which is responsible for synthesis of the immunity determinant for lacticin 481 (Rince et al., 1994; Fallico et al., 2012). Lactococcus lactis UC509.9 plasmid pCIS8 (Ainsworth et al., 2013), in addition to encoding a complete lactococcin A gene cluster, also harbours a gene encoding a putative EntA_Immun (PF08951) family protein, which is thought to confer broad-range class II bacteriocin immunity (Johnsen et al., 2005).

Plant niche-specific adaptations

The sequence of L. lactis SK11 plasmid pSK11P revealed the presence of a cluster encoding four genes, cscABCD, thought to encode cell surface proteins (Siezen et al., 2005). A subsequent bioinformatic investigation of these genes found that they are conserved amongst the LAB and apparently absent in many related pathogenic bacteria, such as streptococci and staphylococci, which lead these authors to suggest that they represent a niche-specific trait (Siezen et al., 2006). The Csc proteins are thought to be involved in acquisition of complex carbohydrates. Transcriptomic analysis has demonstrated that csc clusters are controlled by catabolite repression, which also suggests functional links with sugar metabolism. Additionally, some CscC proteins were found to contain lectin/glucanase domains, known to interact with specific complex carbohydrates.

In L. lactis, csc genes can be found both on the chromosome and on plasmids (Siezen et al., 2006). The chromosomally located csc loci are commonly flanked by IS elements and are therefore thought to be horizontally acquired and mobile. Their apparent confinement to the LAB suggests a niche adaption for plant polysaccharide utilisation. Their presence in dairy lactococci is perhaps a relic of their plant ancestral heritage, although they may still impart an as of yet unknown benefit upon their host in the dairy environment.

Genes whose protein products may modify plant cell wall polysaccharides have also been found on plasmid pVF18 of the raw milk isolate L. lactis DPC3901 (Fallico et al., 2011). This plasmid appears to harbour genes that are beneficial for colonisation of a plant environment as opposed to the dairy environment. Examples include orf11 from pVF18, whose deduced gene product belongs to protein families involved in hydrolysis of plant cell walls, such as chito-oligosaccharide deacetylase from Rhizobium, and orf25, which encodes a protein with a cupin domain. Enzymes harbouring cupin domains are functionally diverse and are associated with epimerases and isomerases, which are involved in modification of cell wall carbohydrates in bacteria and plants (Dunwell, 1998; Dunwell et al., 2004; Fallico et al., 2011).

Host cell surface alterations

Various LAB species are considered probiotic (Clarke et al., 2012), and muco-adhesive properties are thought to play a major role in the efficacy of a probiotic species (von Ossowski et al., 2010; Turroni et al., 2013a, b). Gastrointestinal adhesion aids maintenance of stable or extended colonisation and therefore may, for example, contribute to competitive exclusion of pathogenic species (Saxelin et al., 2005).

Lactococcus lactis is not considered to represent a natural inhabitant of the human gastrointestinal tract (Lukić et al., 2012). As a result, investigations into the probiotic potential of L. lactis have been sparse. Recently, investigations of plasmids from artisanal cheese and plant-derived lactococcal isolates have revealed novel lactococcal muco-adhesive traits (Le et al., 2013). The presence of such traits on mobile elements opens up possibilities of manipulation of these traits for functional food and pharmaceutical applications.

A recent investigation of a unique auto-aggregation characteristic displayed by an artisanal strain from semi-hard cheese, L. lactis subsp. lactis BGKP1, revealed that this phenotype was plasmid-encoded (Kojic et al., 2011). A single gene, designated aggL, located on plasmid pKP1, was predicted to encode a 200 kDa protein belonging to the collagen-binding superfamily, and this gene was shown to confer an aggregation phenotype upon various lactococcal and enterococcal hosts. The presence of plasmids encoding AggL appeared to result in a fitness burden to the cell, as a significantly increased doubling time was observed for cells containing pKP1 compared with pKP1-free strains. Subsequent investigation of AggL-expressing cells revealed that this protein significantly increases the hydrophobicity of the strain and that this appeared to correlate with its binding properties to the colonic mucus by means of nonspecific hydrophobic interactions (Lukić et al., 2012).

Additionally, bioinformatic analysis of pKP1 demonstrated the presence of another novel lactococcal plasmid-borne gene, named mbpL. The deduced MbpL protein is predicted to contain a MucBP-like (mucin-binding protein) domain, which led the authors to speculate a potential role for this protein in mucin interaction (Kojic et al., 2011). Further investigations determined that (compared with a control strain not expressing MbpL) strains expressing MbpL have a significantly enhanced adhesion ability on HT29-MTZ intestinal ileum cells, which predominantly excrete mucins MUC3 and MUC5AC (Lukić et al., 2012). These results therefore suggest that two distinct adhesion mechanisms with alternative adhesion specificities are encoded by pKP1, indicating that strains harbouring pKP1 are able to colonise different parts of the gastrointestinal tract.

Interestingly, plasmid-borne genes predicted to encode proteins required for pilus biogenesis have been identified in a plant-derived L. lactis strain TIL448 (Meyrand et al., 2013), allowing this lactococcal strain to bind to caco-2 intestinal epithelial cells. The plasmid, pTIL448 (Marie-Pierre Chapot-Chartier, personal communication) encodes the typical genetic biosynthetic machinery specifying sortase-dependent heterotrimeric pilus biosynthesis, including a major pilin, a class-C sortase which possesses pilin polymerase activity, a minor pilin and a tip pilin. Similar pili systems have been described in various gut commensals such as lactobacilli (von Ossowski et al., 2010, 2011; Lebeer et al., 2012) and bifidobacteria (O'Connell Motherway et al., 2011; Turroni et al., 2013a, b) and have been demonstrated to be involved in adhesion to and immunomodulation of the host. The lactococcal plasmid-encoded pili have been suggested to represent a niche adaption mechanism, as the tip of the pilus is found to harbour a lectin-like domain, which may function in adhesion of the strain to plant cell walls through specific recognition of a particular plant poly/oligosaccharide. Subsequent investigation into the mucin-binding properties of TIL448 was examined using atomic force microscopy. Results demonstrated a high number of specific adhesive events between TIL448 and pig gastric mucin (Le et al., 2013). Furthermore, blocking assays with fractions of pig gastric mucin, including O-glycans, demonstrated the role of neutral oligosaccharides in the interaction of TIL448 and pig gastric mucin, which is in agreement with the prediction of lectin-like domains being present at the pilus tip (Meyrand et al., 2013).

Antibiotic resistance and food safety

Several recent reports have highlighted the presence of antibiotic resistance in artisanal lactococcal isolates. A study of 94 lactococcal isolates from traditional Italian cheese found that 26 exhibited resistance to tetracycline, while a further 17 were shown to be resistant to both tetracycline and erythromycin (Devirgiliis et al., 2010). Southern blot analysis indicated that the genes encoding these functions were plasmid-associated. Furthermore, plasmid sequencing has demonstrated the presence of antibiotic resistance genes on plasmids from strains that had been isolated from raw milk (Florez et al., 2008; Fallico et al., 2011); therefore, the suitability of newly isolated plasmids for potential use in food production needs to be carefully assessed so as to avoid unwanted spread of such antibiotic resistance determinants.

Additionally, plasmid sequencing has revealed several plasmids, such as pVF18 (Fallico et al., 2011), harbouring genes encoding putative aminoglycoside 3-N-acetyltransferases (Siezen et al., 2005; Ainsworth et al., 2013), which have been implicated in aminoglycoside antibiotic resistance, amongst other functions. However, the predicted involvement of these genes in antibiotic resistance has so far not been confirmed experimentally (Fallico et al., 2011).

Bacteriophage resistance

Bacteriophages infecting industrial L. lactis strains can lead to slow or failed fermentations and loss of product (Kleppen et al., 2011) and are therefore of economic concern. As such, bacteriophage resistance mechanisms of lactococci, which appear to be predominantly a plasmid-encoded feature, have been extensively studied (for reviews see Chopin et al., 2005; Mills et al., 2006; Labrie et al., 2010). Below we discuss recent discoveries and advances regarding plasmid-borne phage resistance mechanisms in L. lactis.

CRISPR-Cas

CRISPR-Cas (clustered regularly interspaced short palindromic repeats-CRISPR-associated) loci specify complex bacteriophage defence systems, which provide acquired immunity against phages through targeting invading phage DNA in a sequence-specific manner (Horvath & Barrangou, 2010). Despite being widespread in nature and frequently found in other members of the LAB (Horvath et al., 2009; Makarova et al., 2011), there is an apparent absence of CRISPR-Cas bacteriophage defence systems in L. lactis. Recently, a highly bacteriophage-resistant industrial dairy isolate was isolated and was observed to harbour a conjugation-transmissible plasmid, termed pKLM encoding a novel type III CRISPR-Cas system (Millen et al., 2012). Transfer of this plasmid to the plasmid-free strain IL1403 conferred resistance against infections by certain 949, 936 and P335 type phages, and loss of the plasmid caused a reversion to the strain's original phage-sensitivity profile. Analysis of its spacer regions revealed sequences with high homology to lactococcal phages, suggesting that the system is active in Lactococcus. However, the described system appeared unable to acquire new spacers and therefore could only provide protection against phages which shared homology with existing spacers. CRISPR-Cas systems appear to be rare in L. lactis as only four of more than 400 industrial isolates screened by PCR were identified to contain the above system (Millen et al., 2012). The authors suggest that the paucity of CRISPR-Cas systems in L. lactis is due to the plasmid dependency of industrial dairy lactococcal strains and the ancillary activity of active CRISPR-Cas against plasmids. Screening of plant isolates of L. lactis, which appear to harbour a relatively small number of plasmids (Kelly et al., 2010), may therefore increase the frequency with which chromosomally encoded CRISPR-Cas systems may be detected in this species.

Abortive infection systems

Abortive infection systems (Abi) are altruistic bacteriophage resistance systems (Chopin et al., 2005). Their common defining feature is the prevention of phage proliferation by interfering with some critical aspect of the phage lytic cycle that also results in bacterial death. This limits the number of progeny produced, thereby protecting other members of a bacterial population susceptible to the Abi-sensitive phage (Labrie et al., 2010). Abortive infection systems are highly diverse and seem particularly abundant in L. lactis (Chopin et al., 2005), where they are often plasmid-encoded (Mills et al., 2006). There are currently over 20 described lactococcal Abi systems which affect different stages of phage multiplication (Table 2), such as DNA replication (in the case of AbiA, encoded on pTR2030; Hill et al., 1990), transcription (in the case of AbiG, encoded on pCI750; O'Connor et al., 1999) and major capsid protein production (in the case of AbiC, encoded on pTN20; Klaenhammer & Sanozky, 1985). Typically, they are encoded by a single gene, but multigene Abi systems, such as AbiE (encoded on pNP40; Garvey et al., 1995), AbiR (specified by pKR223; Twomey et al., 2000) and AbiT (encoded on pED1; Bouchard et al., 2002), have been described.

Table 2

Currently known lactococcal abortive infection systems and related features

Abi system	Discovered on plasmid	Demonstrated phage protection	Proposed mode of action	Phage mutations resulting in Abi-escape mutants	Additional features	References
AbiA	pTR2030	c2 (c2); p2, sk1 (936); φ31, φP013, φQ30, ul36, φQ33 (P335)	Prevents DNA replication, possibly through preventing genome circularisation	Virulent phages acquire point mutations in single-stranded DNA annealing proteins	Homolog of AbiK; heat sensitive; copy number effects level of resistance	Hill et al. (1990), Dinsmore et al. (1998) and Tangney & Fitzgerald (2002)
AbiB	pIL2101	bIL170 (936)	Causes decay of phage mRNA transcripts	Unknown	–	Parreira et al. (1996)
AbiC	pTN20	p2 (936); ul36 (P335)	Interferes with phage protein production	Unknown	–	Durmaz et al. (1992)
AbiD	pBF61	c2 (c2); sk1 (936)	Unknown	Unknown	–	McLandsborough et al. (1995)
AbiD1	pIL105	bIL67 (c2); bIL66, bIL170 (936)	Interferes with phage DNA packaging	936, ORF1 regulator of RuvC-like enzyme; c2, unknown	Toxic to host; activated at level of translation	Anba et al. (1995) and Bidnenko et al. (2009)
AbiE	pNP40	712 (936)	Unknown	Unknown	Two protein system, AbiEi and AbiEii	Garvey et al. (1995) and Tangney & Fitzgerald (2002)
AbiF	pNP40	c2 (c2); 712 (936)	Affects DNA replication	Unknown	–	Garvey et al. (1995) and Tangney & Fitzgerald (2002)
AbiG	pCI750	c2 (c2); 712 (936); φQ30 (P335)	Affects RNA transcription	Unknown	Two protein system, AbiGi and AbiGii	OConnor et al. (1996), O'Connor et al. (1999) and Tangney & Fitzgerald (2002)
AbiH	pLDP1	φ53 (c2); φ59 (936)	Unknown	Unknown	–	Prevots et al. (1996)
AbiI	pND852	c2 (c2); 712 (936)	Unknown	Unknown	Activity at 37 °C	Su et al. (1997)
AbiJ	pND859	712 (936)	Unknown	Unknown	Possible N-terminal DNA-binding helix-turn-helix	Deng et al. (1997)
AbiK	pSRQ800	c2 (c2); p2, P008 (936); P335 (P335)	Prevents DNA replication of P335 phages, possibly through preventing genome circularisation	Virulent phages acquire point mutations in single-stranded DNA-binding proteins; temperate phages recombine with host prophage sequences	Homolog AbiA; toxic to host; copy number effects level of resistance	Bouchard & Moineau (2000), Emond et al. (1997), Fortier et al. (2005) and Bouchard & Moineau (2000, 2004)
AbiL	pND861	c2 (c2); 712 (936)	Possibly translation	Unknown	Two protein system, AbiLi and AbiLii	Deng et al. (1999)
AbiN	Chromosomal	φ53 (c2); φ59 (936)	Unknown	Unknown	Chromosomal encoded; toxic to host	Prevots et al. (1998)
AbiO	pPF144	φ53 (c2); φ59 (936)	Unknown	Unknown	Toxic to host	Prevots & Ritzenthaler (1998)
AbiP	pIL2614	bIL66 (936)	Prevents DNA replication, affects temporal control of early phage transcripts	Homologous recombination with resistant 936 phage, exchange of gene eb7	Activity enhanced by adjoining upstream ORF	Domingues et al. (2004a, b)
AbiQ	pSRQ900	c21 (c2); p2 (936)	Toxin-antitoxin system		Heat stable; toxic to host; toxin-antitoxin system	Emond et al. (1998)
AbiR	pKR223	c2 (c2)	Targets DNA replication	Unknown	Heat sensitive; multicomponent protein system	Twomey et al. (2000)
AbiS	pAW601	jj50, p2, sk1 (936)	Unknown	Unknown	–	Holubová & Josephsen (2007)
AbiT	pED1	p2 (936); φQ30, ul36 (P335)	Affects phage DNA replication	Mutations in phage bIL170 e14, phage P008 orf41 and phages P008 and p2 major capsid proteins	Two protein system, AbiTi and AbiTii	Bouchard et al. (2002)
AbiU	pND001	c2 (c2); 712 (936); ul36 (P335)	Delays RNA transcription	Unknown	Two protein system, AbiU2 possibly regulates AbiU1	Dai et al. (2001)
AbiV	Chromosomal (integrant plasmid)	c2, bIL67, ml3, eb1 (c2); bIL170, jj50, P008, p2, sk1 (936)	Inhibits phage early transcripts	Mutations in the SaV early gene product	Dimeric protein; can be expressed at high levels; forms AbiV2-Sav2 complex	Haaber et al. (2008, 2009, 2010)
AbiZ	pTR2030	p2, sk1 (936); φ31, φQ33, ul36 and 13 other P335 species phages	Premature lysis of infected cells	P335 mutants arise through homologous recombination with prophage sequences encoding early and middle phage genes	–	Durmaz & Klaenhammer (2007)

Specific phage names are supplied, with phage species indicated in brackets.

While discovery of novel Abi systems appears to have slowed down in recent years (the only recently described novel plasmid-encoded Abi is AbiZ (Durmaz & Klaenhammer, 2007; see Table 2), advances in determining molecular triggers and modes of action of currently identified Abi systems have been made. Abi proteins display low (if any) levels of similarity to other proteins; therefore, their mode of action is usually difficult to predict (Chopin et al., 2005). The low cost and ease of genome sequencing has allowed the identification of molecular triggers, modes of action and possible interaction sites for several Abi systems, mainly through the examination of genomes of so-called Abi-escape mutants (Bidnenko et al., 2009).

The most recently described lactococcal Abi, AbiZ, is encoded on the conjugative plasmid pTR2030, which also encodes another Abi system, AbiA (Hill et al., 1990), and which is harboured by L. lactis ME2 (Durmaz & Klaenhammer, 2007). The AbiZ protein product, encoded by a single ORF, confers a bacteriophage protective phenotype against various P335 species phages, such as φ31. AbiZ appears to exert its phage resistance by causing premature lysis of the phage-infected host bacterium (Durmaz & Klaenhammer, 2007).

AbiZ is predicted to possess two transmembrane helices, and experiments expressing the phage holin and lysin suggest that AbiZ acts cooperatively with the holin to increase the bacterial cell membrane permeability. However, examination of φ31 escape mutants suggests a link between the phage's early or middle genes and AbiZ. These escape mutants of φ31 had undergone recombination with the host genome, exchanging an approximate 9-kb region which encompasses the presumed origin of replication and early genes. The exact nature of the escape mutations that confer insensitivity to AbiZ-mediated phage resistance is not known. The authors suggest that an as yet uncharacterised early or middle phage gene product may act as a regulator of the phage holin, thus explaining their observations (Durmaz & Klaenhammer, 2007).

Recently, AbiQ was reported to function as a toxin–antitoxin (TA) system, representing the first such system described in Lactococcus (Samson et al., 2013a, b). TA systems, first described in the 1980s as plasmid stability systems (Jaffe et al., 1985), are typically comprised of a stable toxin protein and an unstable antitoxin protein. As plasmid stability systems, they ensure faithful plasmid partitioning upon cell division, resulting in the death of any cell which has not received a copy of the harbouring plasmid. Since their identification, many systems have been identified in bacteria, archaea and possibly in unicellular fungi, and it has become clear that TA systems have a much wider biological role than just being involved plasmid stability (Yamaguchi et al., 2011; Schuster & Bertram, 2013). Originally discovered on L. lactis W-37 plasmid pSRQ900 (Emond et al., 1998), AbiQ represents a type III TA system, which induces cell death through the accumulation of nonmature forms of viral DNA, and which is effective against members of the so-called 936 and c2 phage species, as well as against certain rarely encountered lactococcal phages (Emond et al., 1998; Samson et al., 2013a, b). The AbiQ system is homologous to the Pectobacterium atrosepticum Abi termed ToxIN (Fineran et al., 2009). Both AbiQ toxin and antitoxin structures have been determined, representing the first Abi system to be characterised at a structural level (Samson et al., 2013a, b). AbiQ is transcribed and translated constitutively in noninfected cells, and the mechanism of its activation is currently unknown, although it was suggested that the antitoxin is sequestered by binding to a phage product (Samson et al., 2013a, b). Examination of AbiQ escape mutants of phage P008 suggests that orf210, encoding a putative DNA polymerase, plays a role in its activation.

AbiK was recently described as a novel reverse transcriptase (RT; Wang et al., 2011). AbiK is a self-priming polymerase, generating long ‘random’ DNA sequences > 500 nt. The N-terminal RT motif of AbiK is essential for phage resistance (Fortier et al., 2005). Previous mutational analysis of this DNA-binding, leucine-rich repeat motif of the AbiK homolog, AbiA, was similarly observed as being essential for conferring phage resistance (Dinsmore et al., 1998). Despite the obtained information as outlined above, the exact mechanism by which AbiA and AbiK provide phage protection to a bacterial culture remains unclear. However, as AbiK possesses RT activity with which it produces long random DNAs, and as AbiK escape mutants localise to saK, encoding a single-stranded DNA annealing protein, which facilitates strand exchange, the AbiA/K systems may inhibit specific homologous recombination events such as genome circularisation. This would explain the lack of DNA replication of sensitive phages in AbiA/AbiK-containing cells, as phage genome replication requires circularisation to produce concatemeric DNA via a RCR-type mechanism (Scaltriti et al., 2011; Wang et al., 2011).

AbiP is encoded on L. lactis IL420 plasmid pIL2617 and is an efficient Abi system that is active against particular members of the 936 species of lactococcal phages, conferring high-level phage resistance. While the AbiP phenotype is conferred by a single ORF (abiP), an adjacent upstream ORF (orf1) was identified that was shown to enhance AbiP's activity (Domingues et al., 2004a, b). Further examination of AbiP has revealed that the protein is associated with the bacterial membrane via an N-terminal transmembrane domain and that it possesses a cytosolic nucleic acid-binding domain. The cellular localisation of AbiP is important for its abortive infection phenotype, as phage resistance was not observed in AbiP mutants lacking the transmembrane domain. The nucleic acid-binding domain of AbiP has a 10-fold binding preference for RNA relative to ssDNA. However, AbiP does not appear to preferentially bind phage-originating nucleic acids. Due to its lack of toxicity towards host nucleic acids, the activation of AbiP is thought to be dependent on (an) unknown phage-specific factor(s) (Domingues et al., 2008).

Recently, the first abortive infection system activated at the level of translation (Bidnenko et al., 2009) was elucidated. AbiD1 was originally detected as being encoded on L. lactis IL964 plasmid pIL105 (Gautier & Chopin, 1987) and displays a resistance phenotype against 936 and c2 lactococcal phages. Experiments to understand the regulation of the AbiD1 system showed that abiD1 is transcribed in uninfected cells; however, abiD1-encompassing transcripts are unstable and produced at low levels. In addition, translation of AbiD1 is inefficient, which is probably due to a stable inverted repeat that may preclude the abiD1-associated translation initiation signals from efficient ribosomal recognition. Infection of abiD1-containing lactococcal cells by phage bIL66 activates AbiD1 mRNA stabilisation and enhances its translation through involvement of the N-terminus of phage bIL66 protein ORF1. ORF1 is believed to bind AbiD1 mRNA and act as a cofactor to increase translation. The tightly controlled and specific activation of the AbiD1 phage-defence mechanism, by a phage component whose C-terminus is absolutely required for infection, shows that AbiD1 is highly evolved in its function.

AbiT is one of the few Abi systems that requires two proteins, AbiTi and AbiTii, for its activity. The abiT genes are cotranscribed and are thought to impart resistance through inhibition of phage genomic DNA maturation. AbiT is active against members of the 936 and P335 phage species, causing efficiency of plaquing (EOP) reductions that range from 10⁻⁵ to 10⁻⁷. Recently, possible phage-specific targets of AbiT activation have been identified (Labrie et al., 2012). Examination of AbiT escape mutants of multiple different 936 phages highlighted three individual genes harbouring mutations. These were two separate early expressed genes (e14 and orf41, for bIL170 and P008, respectively) and orf6 for both P008 and P2, which was subsequently demonstrated to encode the major capsid protein for phage p2 (Labrie et al., 2012). The AbiT resistance-causing mutations in phage p2's capsid protein were localised to its C-terminus, which is incorporated into the phage virion. The potential mechanism for phage resistance conferred by AbiT is based on tentative similarities to the E. coli antiphage system PifA (Labrie et al., 2012). Escherichia coli phage mutants resistant to PifA were shown to require a mutation in a gene necessary for replication, or a double mutation in the gene specifying a capsid protein (Molineux, 1991; Cheng et al., 2004). Similar to PifA, AbiTi possesses a C-terminal hydrophobic region, likely resulting in its localisation to the bacterial membrane (Bouchard et al., 2002). AbiT, like PifA, also reduces phage replication and prevents genomic maturation. This is possibly due to the finding that AbiT alters bacterial membrane permeability, which results in the leakage of ATP and/or other molecules (Labrie et al., 2012).

Plasmid metabolic impact

Lactococcal plasmid exploration efforts have been justified because of their importance in developing new genetic tools and discovery of biotechnological relevant traits which can, for example, improve the robustness of dairy starter cultures. However, several studies have noted that while some plasmids may encode technologically desirable traits, such as bacteriophage resistance, the same plasmids may actually impede cellular fitness for two main reasons. Firstly, addition of new plasmid material will increase competition for the host's replication, transcription and translational machinery, thereby increasing the metabolic burden on the host cell (Lee & Moon, 2003). Secondly, plasmid-encoded factors may negatively impact upon host cell metabolism. For example, conjugational transfer of a 60 kb plasmid pMRC01 to a new host promotes cell permeability and autolysis (Fallico et al., 2009). Transconjugants exhibited lower specific growth rates and higher generation times as compared to their parental strains, although the acidification ability of the strain appeared unaffected. Extensive investigation has not yet been able to reveal the mechanistic action of such phenotypes (Fallico et al., 2009).

Plasmid pBL1 is a small (11 kb) plasmid which encodes the synthetic machinery for bacteriocin Lcn972 (Sánchez et al., 2000). Strains harbouring this plasmid grown under standard laboratory conditions are able to produce Lcn972 without displaying any impact on growth. However, a transcriptomics study of pBL1-containing cells revealed apparent reduced expression of the membrane transporter IIC of the cellobiose PTS, celB (Campelo et al., 2011). Further physiological investigations demonstrated that the presence of pBL1 restricts cellobiose uptake, modulating fermentation products towards that of a more mixed acid profile and altering intracellular glycolytic intermediate levels in L. lactis MG1363 cells grown on cellobiose as compared to glucose. The authors noted that these findings have practical consequences for strains being utilised for bacteriocin production, with media rich in dextrins or cellobiose to be avoided. Furthermore, they note that CelB has been shown to take part in lactose uptake in strains lacking lacEF; therefore, such strains harbouring pBL1 would be seriously compromised when used in dairy fermentations (Campelo et al., 2011).

L. lactis plasmid evolution

As the number of available lactococcal plasmid sequences has significantly increased in recent years, several authors have remarked on the apparent mosaic nature of larger lactococcal plasmids. For example, Gorecki et al. (2011) noted that plasmids pIL1 and pIL2 are highly similar to pCD4 and pCRL1127, respectively, whereas plasmid pIL6 appears to be a combination of particular genetic modules from plasmids pAH82 and pNP40. Also pLP712 appears to be a composite of modules present on pGdh442, pSK11L, pSK11P and several other lactococcal plasmids (Fig. 1; Wegmann et al., 2012). Furthermore, the large lactose/proteinase plasmids that are typical of the dairy L. lactis strains share extensive sequence identity with each other. Plasmids pLP712 (Wegmann et al., 2012), pSK11L, pSK11P (Siezen et al., 2005), pGdH442 (Tanous et al., 2007), pIL4, pIL5 (Gorecki et al., 2011), pVF50 and pVF21 (Fallico et al., 2011), although representing unique plasmids, appear to be closely related derivatives of each other. Interestingly, plasmid pGL6, isolated from a clinical isolate of L. garviae (Aguado-Urda et al., 2012), shares almost 50% of its coding sequence with the L. lactis IL594 plasmid pIL7, suggesting a horizontal transfer event between (ancestors of) these two species. Despite the apparent extensive similarity amongst sequenced lactococcal plasmids, plasmid individuality is also apparent, as 13% of thus far identified lactococcal plasmid genes have currently no known lactococcal homologs.

Figure 1

The mosaic nature of Lactococcus lactis plasmid pLP712. Arrows in the inner circle represent encoded open reading frames (ORFs) present on pLP712. Homology of discreet, coloured modules present on pLP712 is compared with homologous [amino acid identify (%)] modules located on alternative L. lactis plasmids. Blue: lac operon. Green: pepF. Purple: prtP/M. Orange: repB, parAB. Pink: module encoding a putative multicopper oxidase and D-lactate dehydrogenase, along with hypothetical ORFs. Red arrows: transposase-encoding ORFs.

The presence of multiple replicons and GC skew examinations of large lactococcal plasmids (Siezen et al., 2005; Gorecki et al., 2011) and the relative ease of isolation of plasmid derivatives which have undergone physical rearrangements relative to plasmids present in parental strains (Wegmann et al., 2012) suggests that the lactococcal plasmidome (which is the overall plasmid content in a given species or environment; Walker, 2012) is essentially fluid and actively evolving. Wegmann et al. (2012) hypothesised that it is plasmid-borne transposase-encoding genes, commonly found flanking genetic modules, that are directly involved in genetic reshuffling events that cause the generation of alternative plasmid forms. Spontaneously generated deletion derivatives of pLP712 varied in size and structure, which the authors suggest cannot be accounted for on the basis of RecA-dependent homologous recombination alone. The widespread dissemination of uspA, discussed earlier, amongst L. lactis plasmids is possibly due to its position within a predicted small conjugative transposon, which also specifies a putative Mn²⁺/Fe²⁺ cation transporter (Siezen et al., 2005). This putative conjugative transposon is present in all currently sequenced dairy plasmid complements and is also common amongst various Streptococcus and Enterococcus species, suggesting its genetic spread by horizontal transfer.

Transposable elements, such as conjugative transposons and insertion sequences, have long been known to reside on lactococcal plasmids and have been associated with important functional dairy phenotypes, such as bacteriocin production, carbohydrate metabolism and transport, and bacteriophage resistance (for a review, see Mills et al., 2006). Insertion sequence (IS) elements may comprise significant proportions of the strain-specific chromosome and can contribute to genome plasticity, pseudogenes (Makarova et al., 2006) and in some cases activate gene transcription (Bongers et al., 2003). Sequencing of lactococcal genomes and plasmid complements have revealed the extent of IS element abundance in Lactococcus. For example, the genetic information of the combined IS elements in MG1363 encompasses over 66 kb of DNA, while in L. lactis UC509.9, this increases to over 100 kb, in the latter representing 5% of its corresponding chromosome (Ainsworth et al., 2013). As previously noted (Mills et al., 2006), their abundance in lactococcal plasmids is likely a reflection of the importance of IS elements/transposons in promoting evolution of lactococcal plasmids under selective dairying conditions. Examination of all the proteins currently encoded by the lactococcal plasmidome suggests that copies of IS6, IS982 and ISLL6 are particularly abundant (Fig. 2) and collectively constitute 1 in every 10 genes encoded on lactococcal plasmids.

Figure 2

Comparative analysis conducted on seven complete Lactococcus lactis subsp. lactis and L. lactis subsp. cremoris plasmid complements (IL594, UC509.9, SK11, DPC3901, A76, CV56 and KF147). (a) Accumulated number of distinct genes in the generated pan-plasmidome, plotted against the number of employed strains. (b) Accumulated number of genes in core-plasmidome, plotted against the number of employed strains. (c) Total number of genes in the plasmidome, plotted against the number of employed strains. (d) Heatmap resulting from the hierarchical clustering analysis based on the presence/absence of gene families in plasmidome. Highlighted in red are the common plasmid genes with their predicted functions indicated.

Evidence from lactococcal plasmid sequences and experimental data suggests that lactococcal plasmid transfer is predominantly driven by conjugation and transduction. It has long been ascertained that many lactococcal plasmids are mobilisable (Coakley et al., 1997; O'Driscoll et al., 2006). As the process of conjugation is considered a natural DNA transfer system, representing a food-grade process, it has been exploited for the dissemination of technologically desirable phenotypes, such as bacteriophage resistance and bacteriocin production, to a wide range of industrial strains (Mills et al., 2006).

Conjugation is a process whereby plasmid material is passed from a donor cell to a recipient cell via a channel or pilus also known as the conjugation apparatus (Grohmann et al., 2003). Mobile plasmids are transmitted in a single-stranded conformation after it has been nicked at the AT-rich origin of transfer (oriT) region by a nicking enzyme. Molecular mechanisms of conjugation have been described in detail for many Gram-positive and Gram-negative species (Grohmann et al., 2003). Several sequenced lactococcal plasmids are mobilisable or appear to encode the apparatus to enable conjugal transfer (Mills et al. 2006; O'Driscoll et al., 2006; Millen et al., 2012); however, the majority of sequenced plasmids to date do not appear to be transmissible, and the precise molecular mechanisms of conjugation in L. lactis remain unclear.

Sequence analysis suggests that there are two types of nicking enzymes present amongst L. lactis plasmids, MobD and MobA. Both are members of the relaxase (PF03432) family of enzymes. However, MobD and MobA share only 24% amino acid identity across 74% of the length of the enzyme. Additionally, mobA is usually associated with mobB and mobC, which specify three proteins that are thought to form a relaxosome at the oriT (Hofreuter & Haas, 2002), while mobA is usually not associated with Tra-encoding genes (below), as is the case, for example, on plasmids pCIS8 and pSK11P (Ainsworth et al., 2013; Siezen et al., 2005). Alternatively, mobD seems to be associated with mobC and not with mobB, and this gene is often located in a large gene cluster encoding other (presumed) conjugation functions (below).

The mobile lactococcal plasmid pNP40 (O'Driscoll et al., 2006) and the apparently nonmobile pIL6 (Gorecki et al., 2011), which encodes a MobD-type nickase, both contain a 17–18 kb gene cluster for (presumed) conjugal transfer, encompassing approximately 20 open reading frames, which are fairly well conserved between the two plasmids. Human isolate L. lactis CV56 (Gao et al. 2011) harbours two similar gene clusters for putative conjugal transfer on separate plasmids, pCV56A and pCV56C, both of which appear dysfunctional due to transposon insertions. The majority of the open reading frames found in such operons are hypothetical, and it is not apparent if they have an essential or any role in the conjugation process. However, based on sequence similarity, several genes have been identified which are predicted to be involved in transfer and physical entry of conjugated DNA from the donor cell to recipient. A pNP40 open reading frame, termed traF, is a putative membrane-spanning protein, which O'Driscoll et al. (2006) suggested to be involved in channel formation. A TraF homolog is also encoded on pIL6. Genes termed traG and traE are well conserved across sequenced lactococcal conjugative operons and contain domains which suggest that they are involved in the formation of a conjugal pilus structure, similar to type IV secretion systems (O'Driscoll et al., 2006; Gorecki et al., 2011). Hypothetical proteins are encoded by the two conjugal transfer gene clusters, some of which contain potential CHAP and muramidase domains. Both O'Driscoll et al. (2006) and Gorecki et al. (2011) speculate that these proteins may be involved in localised cell wall degradation to aid cell-to-cell contact and promote plasmid delivery to the donor cell.

Despite many plasmids apparently being nontransmissible, they frequently contain (remnant) features of a plasmid transmission machinery. A clear example of this is the multicopy presence of the mobility protein MobC-encoding gene (23 copies being present amongst sequenced lactococcal plasmids). MobC is thought to act as a molecular wedge, aiding in the separation of complementary DNA strands at the oriT locus (Zhang & Meyer, 1997). The high frequency of mobC presence and other associated ‘orphan’ mobilisation genes across lactococcal plasmids may be a reflection of plasmid acquisition/transfer events by mobilisation.

Plasmid transfer via conjugation has been observed to lead to rearrangements and deletions in existing and introduced plasmids through homologous recombination events (Wegmann et al., 2012). Conjugational transfer of plasmids between L. lactis and members of several LAB genera, such as Lactobacillus (Langella & Chopin, 1989; Toomey et al., 2010) and Enterococcus spp. (Tuohy et al., 2002; Devirgiliis et al., 2010) has been demonstrated in in vitro and in vivo models (Toomey et al., 2009). These studies suggest that plasmid transfer via conjugation is a relatively common phenomenon in various naturally occurring environmental matrices, thereby ensuring a constant influx of genetic material for L. lactis and other LAB plasmidome augmentation. High frequency plasmid transfer by transduction in L. lactis has also been observed. In L. lactis NCDO712, transduction of small (< 5 kb) plasmids has been reported to occur at a frequency of 2.1 × 10⁻³ to 2 × 10⁻⁴ transductants per plaque-forming unit (PFU; Wegmann et al., 2012), whereas transduction of large (> 50 kb) plasmids has also been observed to occur at lower frequencies. However, the size of transferred plasmids appears to be limited by the physical packing space of the transducing bacteriophage. For example, Wegmann et al. (2012) observed phage-mediated transfer of the lactose-metabolising phenotype specified by plasmid pLP712; however, when examined, the transduced plasmids were similar in size to the transducing phage genome and smaller than pLP712. Recently, plasmid transduction between LAB species was observed to occur from Streptococcus thermophilus to L. lactis (Ammann et al., 2008), adding a further mechanism by which new genetic material can be incorporated into the L. lactis plasmidome from a nonlactococcal donor.

Bioinformatic analysis (Fig. 2), using reciprocal best-blast hits and MCL clustering algorithms, of the current lactococcal plasmidome corroborates the notion that the lactococcal plasmidome is fluid in nature and actively evolving. Extrapolation of the results, performed using a similar approach employed for analysing bacterial pan-genomes, further suggests that the lactococcal pan-plasmidome (defined as the collection of all distinct genes residing on currently available lactococcal plasmid sequences) is currently not yet fully defined (Fig. 2a and c), thus offering a high potential for the discovery of novel lactococcal plasmid-borne genes. This analysis also reveals that genes commonly present in the combined plasmid content of a given strain, excluding replication proteins, principally consist of mobilisation proteins, transposases and site-specific recombinases, further adding to the notion that their innate variability is driven by rapid evolution through mechanisms of module exchange and genetic rearrangements. Despite the apparent fluid nature of the lactococcal plasmidome, analysis by hierarchical clustering performed on plasmid complements of seven strains IL594 (Gorecki et al., 2011), UC509.9 (Ainsworth et al., 2013), SK11 (Siezen et al., 2005), DPC3901 (Fallico et al., 2011), A12 (Passerini et al., 2013), CV56 (Gao et al., 2011) and KF147 (Siezen et al., 2010) indicates a clear separation of the two L. lactis subspecies based on the presence/absence of plasmid genes. This result is may be simply due to the evolutionary relatedness amongst particular subspecies, yet may explain some of the phenotypic differences between the two subspecies (Fig. 2d).

The relatively low abundance of plasmids in plant isolates of L. lactis coupled with the rarity of lactose in the plant environment and extensive genome decay in dairy isolates suggest that L. lactis strains obtained the majority of its plasmids with dairy-associated phenotypes by horizontal acquisition to adapt to the dairy environment. The ancestral origins of these phenotypes are currently unclear. Lactose is a sugar primarily found in milk and is rare in the plant niche (Mills et al., 2006). This is consistent with the notion that lactose utilisation genes have not been detected in lactococcal plant isolates (Siezen et al., 2010). It is possible that the genetic material supporting this phenotype was first acquired from a mammalian commensal contaminant of milk, such as a member of the Lactobacilli or pathogenic Streptococci. The horizontal acquisition of such elements is consistent with the differential modal codon usage (i.e. the codon usage which matches most of the genes of the genome as opposed to the genome wide average (Davis & Olsen, 2010; Wegmann et al., 2012) observed between pLP712 and its host, L. lactis NCDO712 (the parental strain of L. lactis MG1363). In fact, lactococcal plasmid codon usage is closer to that of the Streptococcus agalactiae chromosome than that of the chromosome of the lactococcal host. The ability to degrade casein may not have been acquired by horizontal transfer, but through adaption of already present proteases, as PrtP-like proteases are found in both plant and dairy lactococcal isolates (Liu et al., 2010; Price et al., 2011). Price et al. (2011) have suggested that alterations in the substrate binding site of PrtP-like proteases may have allowed adaption of these proteins towards casein degradation.

Prospects for future research

Despite the recent increase in lactococcal genome sequencing, a significant proportion of lactococcal plasmid genes remain uncharacterised. For example, it was recently shown that the presence of a cryptic plasmid can modify the cell surface of L. lactis, increasing cellular adhesion properties through electron-donor/electron-acceptor interaction (Cavin et al., 2007), although the precise genetic cause of this phenomenon remains unknown. We expect that the continued pace of lactococcal genome sequencing and continuing interest in exploration of environmental L. lactis strains for industrial applications will undoubtedly result in the discovery of novel plasmid-encoded functions. We also expect that as a result of the fluidity of the lactococcal plasmidome, the discovery and/or deliberate construction by conjugation, mobilisation or transduction of L. lactis strains with novel plasmid combinations will create new scientific and commercial challenges and opportunities.

Acknowledgements

We would like to thank Marie-Pierre Chapot-Chartier for providing unpublished data. This research was funded by a Science Foundation Ireland (SFI) Principal Investigatorship award (Ref. No. 08/IN.1/B1909) to DvS.

Author notes

Editor: Grzegorz Wegrzyn