Large plasmidome of dairy Lactococcus lactis subsp. lactis biovar diacetylactis FM03P encodes technological functions and appears highly unstable

Background Important industrial traits have been linked to plasmids in Lactococcus lactis. Results The dairy isolate L. lactis subsp. lactis biovar diacetylactis FM03P was sequenced revealing the biggest plasmidome of all completely sequenced and published L. lactis strains up till now. The 12 plasmids that were identified are: pLd1 (8277 bp), pLd2 (15,218 bp), pLd3 (4242 bp), pLd4 (12,005 bp), pLd5 (7521 bp), pLd6 (3363 bp), pLd7 (30,274 bp), pLd8 (47,015 bp), pLd9 (15,313 bp), pLd10 (39,563 bp), pLd11 (9833 bp) and pLd12 (3321 bp). Structural analysis of the repB promoters and the RepB proteins showed that eleven of the plasmids replicate via the theta-type mechanism, while only plasmid pLd3 replicates via a rolling-circle replication mechanism. Plasmids pLd2, pLd7 and pLd10 contain a highly similar operon involved in mobilisation of the plasmids. Examination of the twelve plasmids of L. lactis FM03P showed that 10 of the plasmids carry putative genes known to be important for growth and survival in the dairy environment. These genes encode technological functions such as lactose utilisation (lacR-lacABCDFEGX), citrate uptake (citQRP), peptide degradation (pepO and pepE) and oligopeptide uptake (oppDFBCA), uptake of magnesium and manganese (2 mntH, corA), exopolysaccharides production (eps operon), bacteriophage resistance (1 hsdM, 1 hsdR and 7 different hsdS genes of a type I restriction-modification system, an operon of three genes encoding a putative type II restriction-modification system and an abortive infection gene) and stress resistance (2 uspA, cspC and cadCA). Acquisition of these plasmids most likely facilitated the adaptation of the recipient strain to the dairy environment. Some plasmids were already lost during a single propagation step signifying their instability in the absence of a selective pressure. Conclusions Lactococcus lactis FM03P carries 12 plasmids important for its adaptation to the dairy environment. Some of the plasmids were easily lost demonstrating that propagation outside the dairy environment should be minimised when studying dairy isolates of L. lactis. Electronic supplementary material The online version of this article (10.1186/s12864-018-5005-2) contains supplementary material, which is available to authorized users.

response to changing environmental conditions. This behaviour facilitates their distribution among bacteria occupying the same ecological niche. Depending on the environmental conditions, plasmids could have beneficial or adverse effects for the recipient strain. They could give the bacteria the ability to grow better on particular nutrients or survive better under harsh conditions, but at the same time they can be a metabolic burden by either replication of the plasmids or by expression of the plasmid-encoded genes [12].
In this study, the complete nucleotide sequences of twelve plasmids of the dairy isolate of L. lactis subsp. lactis biovar diacetylactis FM03P are presented, together with analysis of the putative biological functions that were assigned to them. Plasmid-cured variants were made by protoplast-induced curing to confirm some of the putative functions and to demonstrate the impact of the plasmids on growth.

Results and discussion
Sequencing Lactococcus lactis subsp. lactis biovar diacetylactis FM03P has been isolated from 10-week-old Samsø cheese. Its genome has been sequenced using a combination of an Illumina HiSeq2500 and PacBio RS instrument as previously described [13]. This revealed the complete sequence of the chromosome and 7 plasmids, designated pLd1 to pLd7. Subsequent next-generation sequencing attempts of L. lactis FM03-V1, a single colony isolate of a culture of L. lactis FM03P, using only Illumina revealed the sequence of 5 other plasmids, designated pLd8 to pLd12. The different sequencing attempts have been summarised in (Additional file 1: Figure S1). Using PCR and analysis by gel electrophoresis, we confirmed the presence of all 12 plasmids in parent strain L. lactis FM03P (Additional file 1: Figure S2). The obtained sequences of pLd8 to pLd12 were not present in the raw unassembled PacBio and Illumina reads obtained from strain FM03P demonstrating that these plasmids were already lost during propagation.

Replication
Plasmids replicate independently from the chromosome using plasmid-encoded Rep proteins. Plasmids in lactococci replicate using either rolling-circle replication (RCR) or theta-type replication [15]. The replication mechanism as well as the sequence of the replication protein and the origin of replication affect the stability of the plasmids, their copy number and their compatibility. RCR plasmids replicate via the synthesis of ssDNA intermediates, and these plasmids are usually small in size, have multiple copies and are incompatible with other RCR plasmids [16,17]. Sequence homology of the replication initiator protein and the presence of a double-stranded origin of replication (dso) indicated that pLd3 is the only RCR plasmid in L. lactis FM03P. On pLd2 we found a gene encoding a protein that had high similarity to the N-terminal part of RepB which is normally encoded on RCR plasmids of the type pMV158, but the C-terminal end was found to be missing.
Theta plasmids are more common in lactococci and a single bacterial cell can contain multiple theta plasmids [18]. Based on the homology of the replication initiator protein and structural motifs in the repB promoter, it was concluded that all plasmids except pLd3 replicate with a theta-type mechanism (Additional file 1: Figures  S3 and S4). On plasmid pLd8 two repB genes are located of which the second replicon (with the repB2 gene) seems to be functional as it is highly similar to the replicons of other plasmids. In contrast, in the promoter of repB1 the inverted repeat IRb is missing and the C-terminus of the predicted RepB1 protein is quite different from the other predicted RepB proteins (including RepB2). Both the repB genes on pLd10 seem to be functional. The DNA sequences of the repB genes on plasmids pLd2 and pLd9 and their promoters are 100% identical. This could cause incompatibility of these plasmids and could result in plasmid loss [15,19]. This may also explain why plasmid pLd9 was not found in the first sequencing attempt. We also observed spontaneous loss of plasmid pLd2 during continuous chemostat cultivation, while pLd9 was kept (data not shown).

Mobilisation
Plasmids can carry conjugation or mobilisation regions that increase their spread in the population via conjugation events. Conjugative plasmids are self-transmissible, while mobilisable plasmids are only transmissible in the presence of additional conjugative functions [20]. To determine if plasmids were transmissible, the plasmid sequences were searched for known conjugation and mobilisation regions. Plasmids pLd2, pLd7 and pLd10 showed a highly similar (> 97%) operon of 4 genes involved in mobilisation of the plasmid. The DNA sequence of these operons in pLd2, pLd7 and pLd10 is 96, 97 and 98% similar to the mobilisation region of pNZ4000, respectively, which was demonstrated to be functional [21]. Plasmid pNZ4000 was found in Lactococcus lactis and carries genes necessary for the production of exopolysaccharides. The oriT sequences, essential for plasmid mobilisation, of pLd2, pLd7 and pLd10 are highly similar to the functional oriT1 sequence of pNZ4000 (Fig. 2) and all plasmids carry a mobA gene encoding a relaxase that is involved in nicking at the nic sites of the oriT sequences. Plasmids pLd2, pLd7 and pLd10 also carry mobC which is present but not annotated in pNZ4000. The genes mobC and mobB, of which the start codon overlaps with the stop codon of mobA¸most likely encode accessory proteins for MobA [22,23]. The function of mobD, designated mobC in pNZ4000, remains to be elucidated. Recently, the nucleotide sequence of plasmid p229C of L. lactis 229 has been  published [24]. This plasmid is remarkably similar to plasmid pLd7 (> 99.9%), while the other plasmids show no or limited similarity. Both strains are isolated from the dairy environment and the mobilisation operon might have increased the transfer rate of this plasmid within the dairy environment. In addition to the mobilisation operons, pLd11 carries 3 genes (orf3, orf9 and mobC) encoding proteins which may be involved in mobilisation. Moreover, pLd8 carries a gene encoding a putative conjugal transfer protein (nickase) of the MobA-MobL family. In contrast to the described genes on pLd2, pLd7 and pLd10, these genes are not part of a large mobilisation gene cluster.

Plasmid-encoded functions
On the 12 plasmids, we identified and annotated in total 203 putative genes or fragments thereof, which corresponds to 7.4% of the total number of putative genes present in the entire genome (Table 1). Based on homology with other proteins, putative functions were ascribed to 74% of the plasmid encoded genes. Of all genes 10% were pseudogenes containing frameshifts, a premature stop codon or truncations. An overview of all the genes in the plasmidome of L. lactis FM03P is given in Fig. 3 and their putative functions are given in Table 2.
In the coming sections a selection of genes encoding functions that could enhance growth and survival of the bacteria and/or encode technological properties are described in more detail.

Substrate uptake and utilisation
To thrive in particular environments, microorganisms require specific transporters and metabolic pathways to take up substrates from the environment and use them for growth. Therefore, the presence of genes encoding particular transporters or metabolic enzymes indicate adaptation to a specific environments. To investigate if L. lactis FM03P was adapted to the dairy environment, we searched the plasmid sequences for genes involved in utilisation of the main carbon and energy sources found in bovine milk: citrate, lactose and proteins.
Citrate utilisation is characteristic for the biovariety diacetylactis of Lactococcus lactis that contains a plasmid-encoded citQRP operon. In L. lactis FM03P, this operon is located on plasmid pLd1, which is 99% identical to lactococcal plasmids pCRL1127 and pIL2. The citP gene is encoding a citrate permease enabling the host to take up divalent citrate from the environment [25]. Citrate utilisation results in the generation of a proton motive force in L. lactis [26] and at the same time increases the pH of the environment [27]. Citrate  utilisation also has been linked to the production of acetoin and diacetyl. These buttery aromas are important flavour compounds in dairy products. Both citQ and citR are involved in the regulation of expression if citP [28,29]. The gene citP is mainly expressed at low pH (around 5.5) when the abundance of divalent citrate is maximal [30,31], minimising the metabolic burden of maintaining this plasmid. Genes involved in lactose uptake and utilisation are found on plasmid pLd8, which contains the lacR-lacABCDFEGX operon for lactose uptake via a phosphotransferase system (PTS) and utilisation via the tagatose-6-phosphate pathway. As described for L. lactis IL1403 [32], L. lactis FM03P also contains the chromosomal-encoded Leloir pathway for lactose utilisation. The presence of both pathways in one strain might give this strain a competitive advantage by fast uptake and utilisation of lactose. Interestingly, the LacR protein might not be functional due to a 40 amino acids deletion at the C-terminus, most likely caused by an IS element insertion in the lacR gene. LacR is a transcriptional repressor of the lac operon and deletion of the lacR gene has been shown to increase the activity of the lac promoter both during growth on glucose and lactose [33,34]. Therefore, a non-functional LacR might increase the maximum lactose utilisation rate.
Genes involved in utilisation of proteins, or more specifically oligopeptides, are found on plasmid pLd8, which contains the pepO gene encoding a neutral endopeptidase and the complete oppDFBCA gene cluster encoding the oligopeptide permease (Opp) system [35]. All of these genes are also encoded on the chromosome with a high similarity in amino acid sequence (> 99%), except for oppA which only has 87% similarity to its chromosomal homolog. Plasmid pLd11 carries a pepE gene encoding a putative aspartyl dipeptidase E, which does not have a chromosomal homolog. The opp and pep genes are required, in combination with the extracellular protease PrtP, for utilisation of the milk caseins as nitrogen source [35,36]. The extracellular protease is often plasmid encoded by prtP and prtM [37]. Interestingly, the prtP and prtM genes are not found in L. lactis FM03P and this strain does not show a caseinolytic phenotype (data not shown). Presence of the pepE, pepO and oppDFBCA gene cluster and absence of the prtP and prtM genes could give the strain an advantage when growing in combination with a protease-positive (prt + ) strain in milk, in particular at high cell densities [38]. Only the prt + strain secretes the protease to hydrolyse the milk caseins, thereby investing energy in production of this protein, while the protease-negative (prt − ) strain can use the peptides generated by the protease without having the burden of protease expression. This combination of prt + and prt − strains is found in many dairy starter cultures, for instance in the Ur starter culture [39]. The presence of pepO and opp genes on both the chromosome and plasmid pLd8 could result in faster utilisation of the peptides depending on the copy number of the plasmid and the regulation of expression of the pepO and opp genes.

Phage resistance by restriction-modification systems
An important technological property that is often carried by plasmids is the resistance to bacteriophage infections via the type I restriction-modification system comprising of     three subunits. HsdS and HsdM are both necessary for methyltransferase activity, while HsdR is required in addition to the HsdS-HsdM complex for restriction of foreign' DNA that is not methylated [40]. The HsdS subunit contains two variable target recognition domains (TRDs) that determine the target sequence specificity of both the restriction and modification activities of the complex [40,41]. The variable domains are flanked by conserved regions required for specific associations with the other subunits and for maintaining the relative positions of the two TRDs [40]. In L. lactis FM03P the three subunits of this system (HsdR, HsdM and HsdS) are all encoded once on the chromosome. On 7 out of the 12 plasmids other HsdS subunits are found. The protein sequences of all HsdS proteins were aligned with MAFFT (Fig. 4). Most of the HsdS proteins contained two variable regions flanked by highly conserved regions as expected. However, the HsdS protein encoded on plasmid pLd1 is not complete. The hsdS gene carried by pLd1 is about half of the size, most likely due to a deletion of 500 to 600 nucleotides. Interestingly, the second variable region of the HsdS proteins of pLd2 and pLd5 are identical and also 97% similar to pLd7, while the first variable regions do not show significant homology to each other. Similarly, the first variable regions of the HsdS proteins of pLd2, pLd9 and pLd12 are identical, while the second variable region of pLd2 does not show significant homology to those of pLd9 and pLd12. The high similarities of one of the variable regions suggest homologous recombination events as also found for plasmids pAH33 and pAH82 [42]. This could result in new R/M specificities.
In addition to the many hsdS genes, on pLd7 also HsdR and HsdM subunits are encoded. These proteins have a low homology to their chromosomal counterparts (41% and 34%, respectively), although the C-terminus of HsdM is quite similar to its chromosomal counterpart. As Schouler and co-authors [43] described before, the Putative functions that could be beneficial in the dairy environment are shown in bold. Pseudogenes are indicated with an asterisk. The species that contains the best homolog is given of genes that are rarely found in L. lactis C-terminal parts of the HsdM subunits of different families have a common sequence that could be involved in the association of HsdM with HsdS. Therefore, the different HsdM and HsdR variants found on the chromosome and the plasmids in combination with the many different plasmid encoded HsdS subunits can form an effective recombination system for broadening the target specificity of the system. In addition to the type I R/M system, plasmid pLd8 harbours orf33, orf34 and orf35 encoding a putative type II restriction-modification system. A similar operon is found in Leuconostoc mesenteroides LK-151 (90%) and in the strains JM3, SK110 and N41 of Lactococcus lactis. orf34 is predicted to encode a type IIG restriction enzyme/N6-adenine DNA methyltransferase according to the restriction enzymes database REBASE [44], which contains both a methylase and recognition domain (pfam12950) according to the SMART database [45]. However, no restriction domain was found in the protein encoded by orf34. The restriction domain is most likely part of the protein encoded by orf33. This protein contains a putative phospholipase D (PLD) domain (pfam13091), a SNF2 ATPase domain (pfam00176), a DEAD-like helicase domain and a helicase C-terminal domain (pfam00271) according to the SMART database. The PLD domain is the metal-independent catalytic site in type IIS restriction endonucleases [46] and might also catalyse the restriction. Notably, the predicted domains in the proteins encoded by orf33 and orf34 are similar to the domains found in the recently identified Class I DIS-ARM system for bacteriophage defence in which the methylase was also predicted to be of the IIG type [47], although they do not share a similar organisation.
Finally, plasmid pLd9 carries an abi gene encoding a putative abortive infection system. The first 642 nucleotides of this gene are 100% identical to a gene encoding the characterised abortive infection system Abi-859 [48]. However, the last 207 nucleotides are missing in pLd9 due to an insertion of a mobile element, so this system might not be functional.

Exopolysaccharide (EPS) production
Exopolysaccharides are thought to have several functions including protection against low-moisture environments and toxic compounds (e.g. bile salts and hydrolysing enzymes, metal ions and antibiotics), colonisation and preventing phage infection [49]. However, in the dairy industry the most important function of EPS is that due to its water-binding capacity, it may improve the rheological properties of the fermented product by affecting the viscosity, syneresis, firmness and taste perception [50]. Plasmid pLd10 carries 18 eps genes encoding putative proteins involved in polysaccharide production via the Wzy-dependent pathway. A typical eps gene cluster in L. lactis consists of the 6 highly conserved genes epsR, epsX, epsA, epsB, epsC and epsD at the 5′ end, a variable region including genes encoding a polymerase (wxy)¸a flippase (wzx), one or more glycosyltranferases and/or other polymer-modifying enzymes and the conserved genes epsL and orfY at the 3′ end [50]. These genes are usually transcribed as a single mRNA [21]. Interestingly, in pLd10 the eps genes are distributed over two clusters and in between these clusters a replication and a mobilisation operon are located, which are flanked by IS elements (Fig. 1). The first cluster consists of: i) the epsR gene encoding an XRE family transcriptional regulator; ii) the conserved gene epsX with unknown function; iii) the phosphoregulatory module consisting of epsABC; iv) epsD encoding an undecaprenyl-phosphate galactosephosphotransferase which catalyses the first step in the assembly of the EPS basic repeating unit (i.e. addition of galactose-1-phosphate to the lipid-phosphate carrier); v) epsE encoding a putative glycosyltranferase; and finally (vi) epsL and orfY with an unknown and regulator function, respectively. The second cluster consists of 6 putative glycosyltransferase genes (epsH, epsI, epsJ, epsO, epsP, epsQ) of which epsI is most likely not functional, 2 putative flippases for export (wzxA and wzxB) and a putative polymerase (wzy). On the chromosome, two other eps clusters are found (Fig. 5). The first cluster has a typical organisation of L. lactis eps operons (epsRXABCD -epsL-orfY), while the second cluster has a typical Lactobacillus eps cluster. This operon starts with epsA encoding a LytR transcriptional regulator followed directly by the phosphoregulatory module (epsBCD) and does not contain epsX, epsL and orfY.

Cation transport
Cation transporters play an important role in growth and survival of microorganisms i) by protecting against toxic heave metals, like cadmium, ii) by providing essential metals, like magnesium, or iii) by providing metals that increase stress survival, like manganese. Several cation transporters were found on the plasmids of L. lactis FM03P.
Plasmid pLd2 carries genes encoding the proteins CadA and CadC, which are > 99.8 and 100% similar to CadA and CadC encoded on plasmids pAH82 of L. lactis DPC220 and pND302 of L. lactis M71 and on the chromosome of Streptococcus thermophiles 4134, respectively [51][52][53]. The cadCA genes have been shown to provide resistance towards cadmium and zinc in both S. thermophiles and L. lactis [51,53]. CadC is a transcriptional repressor which binds to its own promoter region [54]. At high cadmium concentrations, CadC is released from its promoter resulting in the production of cadCA transcript and the expression of cadA [51]. CadA is an ATPase of the P-type catalysing the efflux of cadmium and zinc [55]. The gene cadD, which is located on the chromosome, also encodes a cadmium transporter with 44% similarity in amino acid sequence to the plasmid encoded (pRW001) CadD of Staphylococcus aureus [56].  On plasmid pLd7 we found the corA gene encoding a CorA family transporter, which is expected to be the main Mg 2+ uptake system in bacteria but might transport other cations like cobalt or nickel as well [57]. The CorA protein on pLd7 is 99% identical to the CorA protein found on plasmid pSK11P of L. lactis SK11. On the chromosome three other CorA family transporters are found as well as two other Mg 2+ transporting ATPases. The best match of all these proteins only had 27% homology to CorA of pLd7. Kehres et al. [58] observed a high degree of diversity in sequences within the CorA family even within one species and argued that members within the CorA family might have functions other than Mg 2+ transport. This may also be the case for some of the CorA proteins found in the genome of L. lactis FM03P.
The third cation transporter on plasmids of L. lactis FM03P is the manganese transporter MntH, which belongs to the Nramp family of transporters for divalent metal ions. Mn 2+ ion is an important trace metal required for growth and survival of many bacteria [59]. Several species of lactic acid bacteria accumulate Mn 2+ to scavenge toxic oxygen species, especially superoxide radicals (·O 2 ). This enables the bacteria to survive oxidative stress conditions [59]. Therefore, it is expected that the presence and activity of this transporter could enhance growth in environments with a low manganese concentration and increase survival towards oxidative stress. Plasmid pLd7 and pLd9 encode similar MntH proteins which are 99% homologous and belong to class C of the MntH proteins [60]. The chromosome of L. lactis FM03P harbours three operons that encode manganese transporters: i) a mntH gene of class Cβ, which contains a 344 bp deletion compared to the mntH Cβ gene of L. lactis subsp. lactis ATCC 11454 [60], ii) a mntH gene of class B with 99% identity to mntH B of L. lactis subsp. lactis ATCC 11454 [60], and iii) a mtsBCA operon encoding an ABC transporter for manganese. The mntH genes on the plasmids most likely take over the role of the disrupted chromosomal mntH Cβ gene.

Stress resistance
Besides the potential oxidative stress resistance provided by MntH, there are universal stress proteins (UspA) encoded by genes on pLd7 and pLd9 and a cold shock protein CspC encoded on pLd2. This CspC protein is identical to the CspC protein of L. lactis MG1363 [61]. Furthermore, orf22 carried by pLd8 encodes a putative large-conductance mechanosensitive channel which acts as osmotic release valve in response to a hypoosmotic shock preventing cell lysis [62].

Antibiotic resistance
Plasmid pLd8 carries orf16 encoding a putative aminoglycoside 3-N-acetyltransferase, which catalyses the acetylation of aminoglycoside antibiotics at the 3-amino group and thereby this gene is potentially involved in resistance towards these antibiotics [63]. BLAST analysis revealed highly similar genes (> 99% similarity), which are often found on plasmids in L. lactis, for instance in plasmids pJM3B, pSK11L, pC43 pJM2C, pUC06B, pJM4E, pUC109B, p158C, pCIS8, pJM1A and pUC08A. All these plasmid have in common that they are large (> 47 kb) and carry the lacR-la-cABCDFEGX operon for lactose uptake and utilisation and the several oligopeptide transporters.

Miscellaneous beneficial functions
Other functions encoded by genes on the plasmids that might be beneficial for the host are the putative FAD-dependent D-lactate dehydrogenases on pLd7 and pLd9, which are both 99% similar to the plasmid encoded putative D-lactate dehydrogenase from L. lactis SK11. Siezen et al. [64] suggested that D-lactate dehydrogenase could play a role in D-lactate utilisation in aerobic cultures, which could increase the external pH and the conversion to acetate leads to ATP production. However, Tanous et al. [65] did not find D-LDH activity in pGdh442-containing strains, which carry a highly similar dld gene, nor were they capable of growing on M17 containing D-lactate as carbon source. Plasmid pLd8 carries orf18 encoding a C4-dicarboxylate ABC transporter, which function is to transport dicarboxylates such as aspartate, malate, fumarate, succinate and oxaloacetate. Finally, plasmid pLd7 and pLd9 both carry a gene encoding a transporter of the major facilitator superfamily.

New genes
The plasmids found in L. lactis FM03P carry several genes that are rarely found before in L. lactis, which could indicate horizontal gene transfer events. BLASTP analysis of the amino acid sequence of the predicted protein encoded by orf2 on plasmid pLd3 gave mainly hits from Lactobacillus reuteri (< 29% similar in amino acids) and contained a DUF3552 domain with unknown function. On pLd4 the fmtB gene, encoding a putative peptidoglycan-binding protein, has the highest homology to Streptococcus thermophilus strain B59671, although the homology only covers 60% of the gene. Both the orf2 and the fmtB gene encode large proteins (569 and 1217 amino acids, respectively), account for a large fraction of the plasmids (40.3 and 30.4%, respectively), and are located on the only two plasmids which do not carry other genes with known dairy functions. The genes orf33 and orf34 on pLd8 were homologous to genes of Leuconostoc mesenteroides LK-151 encoding a putative type IIG restriction-modification system. Many genes in the eps operon on pLd10 (especially all genes from tra to orf46; Table 2) were most similar to Lactobacillus species and not found in Lactococcus species. Finally, various small (pseudo) genes were found which had the highest homology to species other than L. lactis. These include orf13, orf14, repC and orf16 on pLd2, which are most similar (> 96%) to genes found in Streptococcus, Lactobacillus and Enterococcus and orf2 and orf3 on pLd4, which are similar to genes of Enterococcus faecalis (99 and 97%, respectively).

Protoplast-induced curing
To determine the effect of plasmid content on the growth performance, plasmid-cured variants were made by protoplast-induced curing. After cells were harvested, the cell wall was degraded with lysozyme. Subsequently, protoplasts were regenerated on plates and the plasmid content of random selected colonies was analysed by PCR targeting genes on each plasmids followed by gel electrophoresis. The obtained variants are shown in Table 3.

Plasmid loss in non-dairy environments
The plasmids carry several genes important for growth in a dairy environment, but in non-dairy environments the plasmids have to be maintained with limited benefits and plasmid-cured variants might arise. Some plasmids were already lost during a single propagation step in either M17 supplemented with glucose or chemically-defined medium containing lactose showing that these plasmids were segregationally unstable in laboratory conditions in the absence of a selection pressure (e.g. bacteriophages, lactose, peptides) ( Table 3). To determine the effect of the plasmid content on the growth of L. lactis in a non-dairy environment, the obtained plasmid-cured variants were grown in LM17 and the optical density at 600 nm was monitored. The growth curves significantly differed between the variants showing that the plasmid content did affect the growth in M17 (Fig. 6a). The highest maximum optical density was found for variant FM03-V1, which only lost plasmid pLd7 (OD 1.1). Variants 48, 49 and 50 had the same maximum growth rate (Fig. 6b) but reached a slightly lower maximum OD (between 0.91 and 0.96). These three variants all lost plasmids pLd4, pLd5 and pLd6, indicating that one of these plasmids might carry a gene which resulted in the higher OD. The growth curve of the parent strain FM03P was similar to variants 48, 59 and 50, but the maximum growth rate could not be estimated using the two-fold dilution method due to flocculation of this variant in M17. The remaining five variants reached a much lower maximum optical density and also grew slower (Fig. 6). Both variant 47 and 63 lost plasmid pLd8 carrying the lacR-lacABCDFEGX operon involved in lactose uptake and utilisation. Therefore, the drop in the growth rate in these variants at an optical density of approximately 0.25 could be caused by limited lactose utilisation and suggests that the lacR-lacABCDFEGX operon is functional when present. Variant 3, 5 and 51 were all growing much slower throughout the cultivation. Variant 3 and 5 are the only variants missing plasmid pLd10 carrying the eps genes, while variant 51 is the variant with the lowest number of plasmids since it has lost plasmids pLd4, pLd5, pLd6, pLd7, pLd8, pLd9, pLd11 and pLd12.
Additionally, growth performance of variants 48 and 50, the latter carrying plasmid pLd2 containing the cadCA operon, was determined in co-cultures in M17 supplemented with 0.5% glucose and 35 μM cadmium. Variant 50 dominated the populations after 27 generations (68%) which shows that pLd2 provided a competitive advantage in the presence of 35 μM cadmium. This is in line with the slightly higher maximal growth rates of variant 50 compared to that of variant 48 in the presence of cadmium (data not shown).

Conclusions
In the past, important industrial traits have been linked to plasmids in L. lactis. Carrying twelve plasmids, L. lactis FM03P has the biggest plasmidome of all completely Some of its plasmids were already lost during a single propagation step showing that the plasmids can be easily lost during propagation in a non-dairy environment. Because we directly sequenced the strain after it was isolated from cheese, the risk of plasmid loss was minimised. Examination of the twelve plasmids of L. lactis FM03P showed that 10 of the plasmids carry genes known to be important for the growth and survival in the dairy environment. These genes encode functions such as lactose and citrate utilisation, degradation and uptake of peptides, exopolysaccharide production, cation transport and bacteriophage and stress resistance. This shows that the plasmids play an important role in the adaptation of this strain to the dairy environment. Two plasmids, pLd3 and pLd4, did not carry any genes that are known to be linked to the dairy environment, but both do harbour a large gene with unknown function that has not been found before in L. lactis and may have a function relevant for dairy processing.

Strain and media
Lactococcus lactis subsp. lactis biovar diacetylactis FM03P, which has been isolated from 10-weeks-old Samsø cheese using Nickels and Leesment medium [66], was used in this study as well as several variants with a different plasmid content (Table 3). These variants were made by protoplast-induced curing or isolated from cultures of L. lactis FM03P. Phenotypic characterisation of the variants started always with streaking the variants on M17 agar plates [67] supplemented with either 0.5% (w/v) glucose or lactose. After incubation for 2 days at 30°C a single colony was inoculated in M17 supplemented with 0.5% lactose or glucose and grown overnight at 30°C.

Whole genome sequencing
Sequencing of FM03P L. lactis FM03P was plated on M17 agar plates supplemented with 0.5% lactose (LM17) and incubated for 2 days at 30°C. A single colony was inoculated in LM17 broth and incubated overnight at 30°C. Subsequently, genomic DNA was extracted using the Wizard® genomic DNA purification kit (Promega, USA), sequenced using an Illumina HiSeq 2500 and a PacBio RS instrument, do novo assembled into contigs and scaffolds and closed by PCR and Sanger sequencing. A more detailed explanation of the method can be found in the genome announcement of L. lactis subsp. lactis biovar diacetylactis FM03 [13]. In addition to the complete chromosomal sequence, 7 plasmids were found, which were defined as complete circular contigs with a origin of replication and which could be targeted in a PCR.
Sequencing of FM03-V1 and the plasmids pLd8, pLd9, pLd10, pLd11 and pLd12 L. lactis FM03-V1, a single colony isolate of a culture of FM03P, was pre-cultured as described above. Subsequently, genomic DNA was extracted using the DNeasy Blood & Tissue kit (Qiagen, Germany) according to the manufacturer's procedure, sequenced using an Illumina HiSeq 2500 instrument (total 2.3 million quality-filtered paired-end reads and average coverage of 211 times) and de novo assembled into contigs using VelvetOptimiser 1.1.0 (k-mer of 83). The contig sequences were compared to the obtained sequence of L. lactis FM03P, of which only plasmids pLd1 to pLd7 were known at that time. Unmapped contigs with a high coverage were further assembled into scaffolds and closed by PCR and Sanger sequencing. The obtained complete circular sequences are designated plasmids pLd8, pLd9, pLd10, pLd11 and pLd12. a b Fig. 6 Growth of plasmid-cured variants of L. lactis FM03P in LM17. a Growth curves of the plasmid-cured variants. Dotted lines represent the standard deviation of representative biological duplicates. The dashed line represents an OD600 of 0.2, which was used as threshold to determine the time to detection. b Maximum growth rate as function of the total plasmid size calculated using the two-fold dilution method. Numbers near the symbols correspond to the different plasmid-cured variants. The maximum growth rate of the parent strain FM03P could not be estimated using the two-fold dilution method due to flocculation affecting the optical density measurement and thus the time to detection and is therefore omitted. Error bars represent the standard deviation of biological triplicates

Phenotype testing Growth in LM17
The maximum growth rate of all variants in LM17 was determined with the two-fold dilution method in a Bioscreen C as described by Biesta-Peters et al. [69].
Briefly, maximum growth rates were determined with three biological replicates and two technical replicates and performing five dilutions per replicate. A single colony was inoculated in LM17 broth and grown overnight at 30°C. Subsequently, 500 μl of the overnight culture was transferred to 9.5 ml fresh LM17 and grown for 3 h at 30°C to have an exponentially growing culture. The exponential culture was diluted > 1000 times in LM17 medium and two-fold dilutions were made in a 100-well Honeycomb plate, which was incubated for 24 h at 30°C with measurements of the optical density at 600 nm every 5 min. Before each measurement, the plate was shaken for 15 s. The lowest dilution contained 10 4 CFU/ ml at the start of the incubation. The time to detect an optical density of 0.2 was determined for each dilution from which the maximum growth rate was determined as described by Biesta-Peters [69]. The maximum growth rates correspond to − 1/slope when plotting the natural logarithm of the inoculum concentration versus the time to detection. Representative wells of the highest inoculum concentration were used to compare the growth curves (Fig. 6a).

Growth performance experiments
The growth performance of variants 48 and 50, differing in plasmid pLd2, was assessed by sequential propagation in mixed cultures. The variants were grown overnight in M17 supplemented with 0.5% glucose (GM17) and mixed in a ratio 1:1 based on optical density measurements at 600 nm in GM17 supplemented with 0.035 mM CdCl 2 . Subsequently, the culture was incubated at 30°C and propagated every 48 h at 1% inoculation level (100 μl added to 10 ml fresh medium) for 4 times. To determine the ratio of the variants at the end of the propagation (~27 generations), appropriate dilutions of samples were plated on GM17 agar and incubated for 3 days at 30°C. Subsequently, the plasmid content was determined of approximately 20 single colonies in three steps: i) DNA extraction, ii) PCR, and iii) gel electrophoresis. DNA was extracted from single colonies by a lysis treatment followed by DNA extraction using the InstaGene™ Matrix (Bio-Rad, USA). Colonies were incubated for 30-45 min at 37°C in 200 μl lysis buffer consisting of 20 mM Tris-HCl, 2 mM EDTA and 1 mg/ml lysozyme (pH 8.0). Subsequently, samples were centrifuged at 13800×g for 2 min and washed with phosphate buffer saline. The pellets were then treated with the InstaGene™ Matrix (Bio-Rad, USA) according to the manufacturer's procedure using 100 μl matrix instead of 200 μl. PCR and gel electrophoresis were performed as explained for the screening of the plasmid content. We used pLd2 to distinguish variant 48 and 50 and pLd1 was used as positive control for a successful DNA extraction.

Additional file
Additional file 1: Figure S1. Summary of sequencing attempts. Figure  S2. PCR products confirming the presence of the 12 plasmids in L. lactis FM03P. Figure S3. Multiple sequence alignment of repB promoters of theta-type replication plasmids in L. lactis FM03P. Figure S4