For a full list of Lactococcus lactis and Lactobacillus plantarum strains used in this study see Table 1 and Additional File 1. The names of lactic acid bacteria, whose heme induced phenotypes are derived from other literature sources, are not included in Table 1, nor in the Additional File 1. Growth medium (MRS-broth or GM17-broth Merck, Amsterdam, the Netherlands 21) was supplemented, when indicated, with heme (hemin) (Sigma, stock solution: 0.5 mg/ml in 0.05 M NaOH) to a final conc. of 2 μg/ml and/or vitamin K₂ (menaquinone-4) (Sigma, stock solution: 2 mg/ml in ethanol) to a final conc. of 20 ng/ml. As a control the equivalent volume of 0.05 M NaOH or ethanol were added if no heme or menaquinone was added respectively.

Aerobic growth conditions were achieved in shake flask cultivations with a 1:10 medium/volume ratio, while shaking at 250 rpm. For high throughput (96-wells micro-titer plates) aerobic cultivations, microtiter plates were filled with 150 μl medium/well (well volume μl 320), covered with breathseals and incubated in a microtron, shaking at 1000 rpm (Greiner Bio-one, Germany). The conditions of cultures grown in stationary tubes were considered anaerobic (or micro-aerobic). All strains were grown for 48 hours at 30°C before measuring biomass (optical density at 600 nm) and acidification.

The following standard method was used for menaquinone measurement in cells; Overnight cultures were washed twice in phosphate buffer (50 mM K₂HPO₄, pH 5.0) and re-suspended to an OD600 of 10. Of this cell-suspension 2 ml was lysed by bead beating, using 0.1 mm silica-beads (Biospec products, Inc) in a Savant Bio 101 FastPrep FP120 and frozen till further use. Menaquinones were extracted by thoroughly mixing 500 μl of this lysed cell-suspension with 5 ml extraction buffer (90%hexane, 10% ethanol). The hexane layer was transferred to a new tube after centrifugation. This extraction procedure was repeated twice, and the combined hexane layers subsequently evaporated under nitrogen-gas. The precipitated menaquinones were re-dissolved in 300 μl ethanol. This menaquinone solution was analysed on a Thermo Finnigan (Waltham, MA) TSQ Quantum ms-ms system in combination with a Shimadzu (Kyoto, Japan) LC system. The samples were injected on a Synergi 4μ MAX-RP 80A 150 × 2 mm (Phenomenex) column where the compounds were eluted with a linear gradient, starting with 100% water/methanol 25/75 to 100% 2-propanol and detected with ms-ms. The TSQ Quantum ms system was equipped with an APCI (Atmospheric Pressure Chemical Ionization) source, set in the negative mode. A capillary temperature of 210°C was used with a vaporizer temperature of 300°C and a sheat gas pressure of 21psi. The collision energy for measuring in ms2 was set at 38 Volt for all compounds.

The presence of the cydABCD and the menaquinone biosynthesis genes were based on annotation of the respective complete genome sequences and by homology analysis with Lactococcus lactis MG1363 gene-products, using BLASTP 2.2.18 (basic local alignment search tool) 1222. Comparison of genomic local organization was performed using the KEGG genome map http://www.kegg.com/, and the pinned region function of ERGO http://ergo.integratedgenomics.com/ERGO/.

The identification of the genomic site of integration of the transposase gene, of the selected respiratory-defective, Lactococcus lactis B1157 mutants, was performed as described previously 23.

Each individual cydABCD gene product of Lactococcus lactis MG1363 was entered as a query to search for homologues in other lactic acid bacteria using the BLAST algorithm 22. Sequence entries found to be homologous were retrieved (june '08) from GenBank, and separately aligned using the MUSCLE algorithm 24. From the amino acid sequence alignments, bootstrapped neighbor joining trees were obtained using Clustal 25 with default settings, except for the number of iterations, which was set to 1000. Trees were analyzed in LOFT 26 and visualized in MEGA3 27. An identical exercise was carried out for 16S rRNA sequences from the organisms that were found to contain cyd-genes. Finally, the topology of the four trees, based on the cyd-gene products, was analyzed and compared to a standard phylogeny tree based on 16S rRNA.

Respiration in Lactococcus lactis is easily induced by addition of heme, and results in enhanced growth efficiency with an almost doubling of biomass (Fig 1). Although Lactococcus Lactis cells grown without heme are less robust, in the complex medium used, extensive cell-lysis after the onset of the stationary phase is not apparent 15. The final pH, at the end of the exponential growth phase tends to be higher in respiring cultures when compared to aerobic cultures (grown without heme). This higher final pH is a result of more progressive conversion of lactate to acetate during respiration 15. We screened a diverse range of lactic acid bacteria (both species and strains) for similar heme induced growth stimulation. Of the 29 different species of lactic acid bacteria that were aerobically cultivated, 6 were clearly stimulated by the presence of heme (Table 2 and 3). In addition to Lactococcus lactis, the presence of heme led to increased biomass formation in Carnobacterium divergens, Enterococcus casseliflavus, Lactobacillus paralimentarius, Streptococcus entericus, and Streptococcus uberis, although the pH did drop.

Menaquinones form a part of electron transport chains, facilitating membranous electron transfer 28. Many lactic acid bacteria are unable to produce menaquinones (more widely known as vitamin K₂). Addition of both heme and menaquinone to aerated cultures of Lactobacillus plantarum WCFS1 increased the final (stationary phase) biomass and pH, quite similar to the heme supplemented phenotype of Lactococcus lactis (Fig 1 and Table 3). This is in agreement with the absence of a complete gene-set, in Lactobacillus plantarum, for menaquinone biosynthesis http://www.kegg.com. The measured biomass (OD600) levels of Lactobacillus plantarum, grown with or without heme and menaquinone, remained relatively stable after the onset of the stationary phase (data not shown). In our selection of lactic acid bacteria we observed that Lactococcus garviae, Lactobacillus rhamnosus, Lactobacillus brevis and Enterococcus faecalis also require both heme and menaquinone for aerobic growth stimulation (Table 3). As reported in literature, a combination of these cofactors also stimulate biomass formation in Streptococcus agalactiae NEM316 20. Furthermore, heme-induced cytochrome formation in Leuconostoc mesenteriodes (NCIB 9917) and Enterococcus faecalis (V538) has also been reported 6192930.

Currently there are 62 sequenced genomes of lactic acid bacteria available in the NCBI database of which 45 are complete http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi. We examined the potential for respiration in this diverse set of lactic acid bacteria as indicated by the presence of the cydABCD genes in their genomes. The cydA and cydB encode for the structural subunits, and cydC and cydD are required for assembly of the bd-type cytochrome 31. Using BLASTP and available annotation, we found that in 22 of these genomes all four cyd-genes were present (Table 4) 22.

In the case of Lactobacillus brevis ATCC367 cydC has degenerated to a pseudo gene. Streptococcus agalactiae NEM316 has four open reading frames (gbs1787-1784) with a high similarity to cydABCD of Lactococcus lactis MG1363 and a similar operon structure. Lactobacillus casei ATCC334 possesses, besides a cydABCD- operon, a separate fusion gene of cydA and cydB. The genome of Streptococcus agalactiae H36B contains 3 genes that are annotated as cydA Of these, SAI_1857 lies in an operon together with cydBCD while SAI_1865 and SAI_1866 are next to each other on the chromosome. SAI_1865 and SAI_1866 are truncated versions of cydA and are unlikely to encode a functional subunit I of the bd-type cytochrome.

Although, there is some confusion in the annotation of the open reading frames as either cydC or cydD genes among the lactic acid bacteria, the presence or absence of all four if the cyd-genes could be unambiguously determined for the studied sequenced genomes.

Next, we constructed a phylogenetic tree for each of the cydABCD genes. For this purpose, we used the cyd-gene products (AA-sequences) found in the lactic acid bacteria, and also as found in many non-lactic acid bacteria. Overall, the topologies of the phylogenetic trees made for cydA, cydB, cydC and cydD, resembled each other and the canonical phylogenetic (evolutionary) trees based on 16S rRNA. In all cases the lactic acid bacteria cluster together, and are neighbored by representatives of Listeria and Bacillus, forming the Firmicutes branch (Fig 2).

We compared the responses to heme (and menaquinone) (both observed in this work and reported in literature) with the genomic presence of the cydABCD genes and menaquinone biosynthesis. For the sequenced species Lactococcus lactis MG1363, SK11, Lactobacillus plantarum WCFS1, Enterococcus faecalis V583, Streptococcus agalactiae NEM316 there is a match between genotype and phenotype. Heme enhanced aerobic growth (biomass) has also been observed for Oenococcus oeni (A. Gruss presented at the LAB9 congress, Egmond aan Zee, 31 Aug–4 Sep, 2008). These species show respiration behavior (either biomass stimulation or formation of cytochromes) and have cyd-genes present on their genomes. Furthermore, as mentioned, Lactococcus lactis is known to produces menaquinones, whereas both Lactobacillus plantarum WCFS1 and Streptococcus agalactiae NEM316 lack many genes typically involved in menaquinone biosynthesis http://www.kegg.com. This is thus also in line with our observations of the dependency on both heme and menaquinone. Also relevant, in this respect is that we did not observe that heme stimulated biomass formation in Lactobacillus sakei 23K, Lactobacillus delbrueckii B1799 (not sequenced) or in Pediococcus pentosaceus DSM 20333.

There are some discrepancies however, as the sequenced Leuconostoc mesenteroides ATCC 8239 has all four cyd-genes and menaquinone biosynthesis genes, but in our hands, did not show enhanced production of biomass, in the presence of heme (and menaquinone). Heme-induced cytochrome formation has, however, been reported for other strains of Leuconostoc mesenteroides 6. Furthermore, Enterococcus faecalis B145 required both heme and menaquinone, while the sequenced strain Enterococcus faecalis V538 appears to have a complete menaquinone biosynthesis pathway http://www.kegg.com.

The production of menaquinones by several lactic acid bacteria, such as Lactococcus lactis and Leuconostoc mesenteroides has been observed during non-respiratory (anaeorobic) conditions 10. We have cultivated the Lactococcus lactis MG1363, in both respiratory and non-respiratory conditions to study its influence on the composition of the menaquinone pool (Table 5). Menaquinones can vary in their side-chain length (or the number of isoprenoid residues) although species with more than 10 isoprenoid residues are not typically found in (grampositive) lactic acid bacteria 32. Assuming that the levels of menaquinone with 11 or more isoprenoid residues in their side-chain are negligible, we observed that, contrary to expectations, the total production of menaquinones was highest in anaerobic conditions by roughly 2-fold. In particular, the level of the menaquinone k₂(3) was elevated in anaerobic conditions. In aerobic conditions, the concentration of menaquinone-species with 9 and 10 isoprenoid residues (k₂(9), k₂(10)) were elevated.

The presence or absence of the cyd-genes in various species is remarkably consistent, as shown in table 4 and the corresponding text. The cyd-genes were not present in the (completely sequenced) genomes of any of the strains of Streptococcus pyogenes (13), Streptococcus pneumoniae (11), Streptococcus suis (2), Streptococcus thermophilus (3) and Lactobacillus delbrueckii (2), but consistently present in all Streptococcus agalactiae (8), Lactococcus lactis (3), Lactobacillus reuteri (2) and Oenococcus oeni (2) strains. Furthermore, in two separate studies 43 strains of Lactobacillus plantarum were genotyped and their genomic content compared to the genome of Lactobacillus plantarum WCFS1. In all these 43 genomes the cydABCD operon was determined to be present 33 (and Tseneva, personal communication). To determine if the respiratory-phenotype is as consistent as the genotype, we tested 88 strains of Lactococcus lactis (including SK11, MG1363 and IL1403) and 20 strains of Lactobacillus plantarum (including WCFS1). Heme (and vitamin K₂) induced the respiratory response (enhanced biomass formation) in 95% of the Lactococcus lactis strains, and in 80% of the Lactobacillus plantarum strains (data not shown). One of the 4 Lactococcus lactis strains that did not respond to heme was the sequenced strain IL1403. In this particular case, we observed that a transposase is situated directly in front of the cyd-genes on the genome, which could be responsible for this lack of heme-induced response. For the other 3 Lactococcus lactis strains, that did not respond to the presence of heme, no extensive genomic information is available. Both the genotypic data and the growth experiments indicate that respiration is characteristic for a diverse group of lactic acid bacterial species.

By using an insertion knock-out bank of Lactococcus lactis B1157, we show that high throughput screening, in 96-well plates, is possible to isolate heme-stimulated lactic acid bacteria 23. Approximately 8000 insertion mutants were aerobically incubated (with or without heme) and screened on biomass yield.

A total of 73 mutants were found which showed no significant biomass increase upon cultivation with heme. The position of integration of the insertion-knockout vector was determined in 13 of these 74 mutants by analysis of the surrounding genome sequence. The genomic region and the genes of Lactococcus lactis B1157 that were disrupted by the insertion event were identified by comparison with the Lactococcus lactis MG1363 genome sequence (Table 6).

Five of the isolated respiration mutants carried mutations in genes involved in menaquinone biosynthesis. Respiration mutants with disruptions of aroB and aroE are also likely to suffer from an impaired menaquinone biosynthesis, since these genes are involved in synthesis of chorismate, the menaquinone precursor molecule. Three mutants were found that carried disruptions in the cyd-genes that are obviously essential to synthesize the bd-type cytochrome. Furthermore, we have now experimental evidence that a disruption of noxA, annotated as a NADH-dehydrogenase is directly linked to a respiration negative phenotype. 11 out of 13 respiration mutants carry mutations in genes that can be readily explained and thus validate this high throughput method to screen for respiration in lactic acid bacteria.