Heme Uptake in Lactobacillus sakei Evidenced by a New Energy Coupling Factor (ECF)-Like Transport System

Lactobacillus sakei is a nonpathogenic bacterial species exhibiting high fitness in heme-rich environments such as meat products, although it does not need iron or heme for growth. Heme capture and utilization capacities are often associated with pathogenic species and are considered virulence-associated factors in the infected hosts. For these reasons, iron acquisition systems have been deeply studied in such species, while for nonpathogenic bacteria the information is scarce. Genomic data revealed that several putative iron transporters are present in the genome of the lactic acid bacterium L. sakei. In this study, we demonstrate that one of them is an ECF-like ABC transporter with a functional role in heme transport. Such evidence has not yet been brought for an ECF; therefore, our study reveals a new class of heme transport system.

ABSTRACT

Lactobacillus sakei is a nonpathogenic lactic acid bacterium and a natural inhabitant of meat ecosystems. Although red meat is a heme-rich environment, L. sakei does not need iron or heme for growth, although it possesses a heme-dependent catalase. Iron incorporation into L. sakei from myoglobin and hemoglobin was previously shown by microscopy and the L. sakei genome reveals the complete equipment for iron and heme transport. Here, we report the characterization of a five-gene cluster (from lsa1836 to lsa1840 [lsa1836-1840]) encoding a putative metal iron ABC transporter. Interestingly, this cluster, together with a heme-dependent catalase gene, is also conserved in other species from the meat ecosystem. Our bioinformatic analyses revealed that the locus might correspond to a complete machinery of an energy coupling factor (ECF) transport system. We quantified in vitro the intracellular heme in the wild type (WT) and in our Δlsa1836-1840 deletion mutant using an intracellular heme sensor and inductively coupled plasma mass spectrometry for quantifying incorporated ⁵⁷Fe heme. We showed that in the WT L. sakei, heme accumulation occurs rapidly and massively in the presence of hemin, while the deletion mutant was impaired in heme uptake; this ability was restored by in trans complementation. Our results establish the main role of the L. sakei Lsa1836-1840 ECF-like system in heme uptake. Therefore, this research outcome sheds new light on other possible functions of ECF-like systems.

IMPORTANCE Lactobacillus sakei is a nonpathogenic bacterial species exhibiting high fitness in heme-rich environments such as meat products, although it does not need iron or heme for growth. Heme capture and utilization capacities are often associated with pathogenic species and are considered virulence-associated factors in the infected hosts. For these reasons, iron acquisition systems have been deeply studied in such species, while for nonpathogenic bacteria the information is scarce. Genomic data revealed that several putative iron transporters are present in the genome of the lactic acid bacterium L. sakei. In this study, we demonstrate that one of them is an ECF-like ABC transporter with a functional role in heme transport. Such evidence has not yet been brought for an ECF; therefore, our study reveals a new class of heme transport system.

INTRODUCTION

Iron is an essential element for almost all living organisms (1), and heme, an iron-containing porphyrin, is both a cofactor of key cellular enzymes and an iron source for bacteria. Many bacteria encode the complete heme biosynthesis pathway to be autonomous for heme production and partly to guarantee their iron supply. However, some others lack heme biosynthetic enzymes and rely on the environment to fulfill their heme requirements. Lactococcus lactis and all known lactobacilli are heme-auxotrophic bacteria (2). Also, it is well established that lactic acid bacteria do not require iron to grow (3) and that their growth is unaffected by iron deprivation. Nevertheless, numerous lactic acid bacteria, such as L. lactis, Lactobacillus plantarum, or Enterococcus faecalis, require exogenous heme to activate respiration growth in the presence of heme (2).

Lactobacillus sakei is a nonpathogenic lactic acid bacterium frequently found on fresh meat. L. sakei is often associated with meat products and in particular with raw meat products stored at low temperature and under vacuum packaging (4). Interestingly, abundance of L. sakei has been shown to prevent growth of undesirable pathogens such as Listeria monocytogenes (5, 6), Escherichia coli O157:H7 in both cooked and minced meat (5, 7, 8), and spoilers such as Brochothrix thermosphacta (7, 8). Therefore, this species is often used as a bioprotective culture in meat products. Nevertheless, mechanisms of synergy and competition between species in such complex matrices are still poorly understood (9). Meat can be considered a growth medium naturally rich in iron and heme. Quantification of total iron content in raw meat reported a mean of 2.09 mg of total iron/100 g for four beef meat cuts in which 87% was heme iron (10). Although L. sakei has a tropism for meat and is known to possess a heme-dependent catalase (11), it is considered to be a bacterium that requires neither iron nor heme to grow (12).

First insights on iron/heme utilization by L. sakei came from its whole-genome analysis (13) with the identification of coding sequences of several iron transporters, regulators, and iron-containing enzymes. Later, microscopy analysis of L. sakei cells combined to spectroscopy methods showed that L. sakei is able to incorporate iron atoms from complexed iron such as myoglobin, hemoglobin, hematin, and transferrin (12). This suggested that L. sakei may display heme or heminic-iron storage ability, although the analytical method used was not quantitative and the precise amount of iron compound that L. sakei is able to store was not determined. Hematin did not show any effect on growth of L. sakei, but hematin has been shown to prolong bacterial viability in stationary phase (13). However, the mechanisms underlining L. sakei survival in the presence of heme need to be unraveled.

Heme acquisition systems have mainly been studied in Gram-negative and Gram-positive pathogens that acquire heme from host hemoproteins in a two steps process (for a review, see references 14, 15, and 16). First, cell surface or secreted proteins scavenge free heme molecules or complexed heme. Then, transmembrane transporters, generally ATP-binding cassette (ABC) transporters, carry the heme moiety into the intracellular space. Gram-positive bacteria rely mainly on surface-exposed receptors that shuttle heme through the cell wall and deliver it to an ABC transporter for subsequent transfer into the cytoplasm. Within Gram-positive pathogens, one of the most well characterized heme uptake system is the Staphylococcus aureus Iron surface determinant (Isd) system. The staphylococcal machinery is inserted into a 10-gene locus encoding cell wall-anchored proteins (IsdABCH), a membrane transport system (IsdDEF), a sortase (SrtB), and two cytoplasmic heme-oxygenases (IsdG and IsdI) (17, 18). IsdB and IsdH are responsible for binding host hemoproteins or heme. IsdA extracts heme from IsdB or IsdH and transfers it to IsdC. Funneled heme is finally transferred into the cytoplasm through the membrane by the IsdDEF ABC transporter, where it is finally degraded to release free iron by the heme oxygenases IsdG and IsdI. Several of these Isd proteins contain the NEAr iron Transporter (NEAT) domain, present only in Gram-positive bacteria, and specific to interact with hemoproteins and heme. The NEAT domain is a 150-amino-acid-residue domain that despite sequence variability displays a conserved β-barrel and a hydrophobic pocket involved in heme binding (19).

Thus far, heme acquisition systems in heme auxotrophic organisms have only been reported for streptococci (15, 20, 21). In S. pyogenes, the system involves the Shr and Shp NEAT domain proteins and the Hts ABC transporter (20, 22, 23). In Lactococcus lactis, heme homeostasis, especially heme efflux systems, has been deeply characterized (24, 25). Nevertheless, the acquisition of exogenous heme remains poorly characterized. Heme transport across L. sakei membrane is still unknown. In addition, bioinformatic analysis shows that the genome of L. sakei does not contain any NEAT domain (13), which suggests that heme transit could involve transport systems distinct from streptococci and S. aureus (14).

Regarding prokaryotic metal ion uptake transporters, comparative and functional genomic analysis have identified energy-coupling factor (ECF) transporters as a novel type of ABC importers widespread in Gram-positive bacteria and first identified in lactic acid bacteria (26). The studies identified genes encoding ABC-ATPases plus three or four membrane proteins within the same or adjacent operons, which were implicated in vitamin production or synthesis of metal-containing metalloenzymes (27). Their predicted role in cobalt or nickel ion uptake and delivery within the cell was demonstrated in Salmonella enterica and Rhodococcus capsulatus, respectively. Since then, ECF-coding genes have been evidenced in Mycoplasma, Ureaplasma, and Streptococcus strains. They were also shown to function as importers not only for transition metal ions but also for vitamins such as riboflavin and thiamine (27). Recently, several ECF systems have been characterized, among them folate and pantothenate ECF transport in Lactobacillus brevis and cobalt ECF in R. capsulatus (28–31). It was shown that ECF transporters constitute a novel family of conserved membrane transporters in prokaryotes, while sharing a four-domain organization similar to that of the ABC transporters. Each ECF displays a pair of cytosolic nucleotide-binding ATPases (the A and A′ components also called EcfA and EcfA′), a membrane-embedded substrate-binding protein (the S component or the EcfS), and a transmembrane energy-coupling component (the T component or EcfT). The quadripartite organization has a 1:1:1:1 stoichiometry. Notably, the S component renders ECF mechanistically distinct from ABC transport systems as it is predicted to shuttle within the membrane, when carrying the bound substrate from the extracellular side into the cytosol (see the recent review in reference 26). Accordingly, the S component solely confers substrate specificity to the uptake system (28). Until the 2000s, folate, riboflavin, and thiamine ECF importers were reported for L. lactis (32–34). Similarly, folate, hydroxyl pyrimidine, and pantothenate ECFs have been reported and structurally characterized for L. brevis (28, 30, 31), both Gram-positive rod-shape species of lactic acid bacteria.

Here, we mainly targeted L. sakei locus lsa1836-1840 encoding a putative ABC transporter, and demonstrated its role as a heme uptake system, combining in silico bioinformatics analysis with in vitro functional analysis. We showed that this system encodes the complete machinery of an ECF-like importer, including the extracellular proteins that initiate heme scavenging. In parallel, we quantified the heme-iron and heminic-iron storage properties of L. sakei and compared wild-type (WT) L. sakei with the Δlsa1836-1840 L. sakei deletion and overexpression mutants using an intracellular heme reporter gene and mass spectrometry quantification of iron-labeled heme. We were able to show in vitro that this five-gene locus plays an important role in active heme import.

RESULTS

Putative iron and heme transport systems in L. sakei.

Accurate analysis of the genome of L. sakei 23K (13), focused on heme/iron transport systems and heme utilization enzymes, previously led to the identification of six putative iron/heme transport systems and one heme-degrading enzyme (Table 1). First, two genes, lsa0246 and lsa1699, encoding proton motive permeases, which belong to the MntH family of manganese uptake, might be involved in iron or heme uptake. Notably, in L. lactis, a mntH mutant was impaired in Fe²⁺ transport (35).

TABLE 1.

Genes putatively involved in iron/heme transport and heme modification

Functional category and locus tag	Protein ID	Predicted protein function
Genes putatively involved in iron/heme transport
ABC transporters
lsa0399-0402	CAI54700–CAI54703	Ferric iron uptake (Fhu-like)
lsa1836-1840	CAI56143–CAI56147	Putative metal ion ABC transporter, cobalamin transporter
lsa1366-1367	CAI55670–CAI55671	Putative ABC exporter (heme efflux machinery)
Proton motive force transporters
lsa0246	CAI54546	Mn2+/Zn2+/Fe2+ transporter
lsa1699	CAI56006	Mn2+/Zn2+/Fe2+ transporter
Membrane proteins
lsa1194-1195	CAI55498–CAI55499	Uncharacterized proteins
Gene putatively involved in heme modification
lsa1831	CAI56138	Dyp-type peroxidase

Second, an operon, composed of the genes lsa1194 and lsa1195 (lsa1194-1195) coding for poorly defined membrane proteins of the CCC1 family, is putatively involved in iron transport. In yeast, CCC1 is involved in the manganese and iron ions transport from the cytosol to the vacuole for storage (36).

Third, two ABC systems homologous to the HrtAB and Pef heme detoxification systems present in L. lactis and Streptococcus agalactiae (24, 37) were also identified in L. sakei genome. These systems are encoded by the lsa1366-1367 and lsa0419-0420 genes, respectively. The sequencing of the lsa0419-0420 region has confirmed the presence of a frameshift and indicated that these genes are not expressed in L. sakei 23K strain. The lsa1366-1367 gene products are homologous to the L. lactis Llmg_0625-0624-encoded proteins. The L. lactis genes code for the HrtB and HrtA proteins, respectively (24). An in silico analysis of Lsa1367 and HrtB indicated that these proteins share 33% sequence identity and, accordingly, the same fold, as assessed by TOPPRED analysis (38). Particularly, the cytoplasm-exposed Y168 and Y231 amino acid residues, shown as important for HrtB-heme interaction in L. lactis (25), are also present in Lsa1367, which suggests that these genes might be homologous to the L. lactis heme export system.

Lastly, two iron or heme uptake ABC transporters were identified. Markedly, the operon lsa0399-0402 encodes an Fhu system, sharing homology with various orthologous genes and operons encoding complexed iron transport systems, possibly homologous to the Listeria monocytogenes HupCGD system. Also, L. monocytogenes shows that HupCGD and Fhu are involved in heme and ferrioxamine uptake, respectively (39).

Then, the ABC system encoded by the lsa1836-1840 genes was automatically annotated as being involved in cobalamin transport, and it shows some similarity to heme import systems described in Gram-positive bacteria (40–43). At first, we carried out a multiple alignment of all putative substrate-binding lipoproteins encoded in the L. sakei 23K genome and noticed that Lsa1839 was closely related to Lsa0399 from the Fhu system (data not shown), suggesting a possible link to iron/heme transport. Furthermore, if heme transportation would represent a specific fitness for growth in meat, we wondered whether other meat-borne bacteria would contain a similar cluster in their genome. As shown in Fig. 1, comparative genomic analysis revealed that the lsa1836-1840 gene cluster is present in several species known to harbor a tropism for meat. The most interesting observation is that species harboring the lsa1836-1840-like cluster also have in their genome a katA gene, encoding a heme-dependent catalase, while the other species lacking the cluster, such as Leuconostoc and Lactococcus spp., were shown to be deprived of the catalase-encoding gene. Although such cooccurrence could not constitute a proof of the role of the lsa1836-1840 cluster in heme transport, this analysis provided an additional argument consolidating this hypothesis.

FIG 1.

Gene synteny within and around the lsa1836-1840 gene cluster of L. sakei 23K with other Gram-positive species found frequently on meat products. Genes shown in gray are unrelated to this cluster and are not conserved between the different genomes. The names of the species and the strains used for analysis are depicted on the left. All of these species contain a katA gene (encoding a heme-dependent catalase) in their genomes. Other meat-borne species, including Leuconostoc, Lactococcus, and Vagococcus species, also found on meat are not shown due to the lack of both katA gene and lsa1836-1840 gene cluster.

lsa1836-1840 encodes an ECF-like transport system putatively involved in heme transport.

Due to the conservation of the operon lsa1836-1840, each of the five sequences was analyzed comprehensively using bioinformatics. It includes multiple sequence alignment, as well as the three-dimensional (3D) structure, the protein network, and export peptide predictions. Lsa1836 shows a sequence similarity of >30%, associated with a probability above 99% with an E value of 8 ⋅ e⁻¹⁵, to share structural homology with the membrane-embedded substrate-binding protein component S from an ECF transporter of the closely related L. brevis, as computed by HHpred (44). Accordingly, its sequence is predicted to be an integral membrane component with six transmembrane helices and a very high rate of hydrophobic and apolar residues, notably 11 tryptophan amino acid residues among the 230 residues of the full-length protein (Fig. 2A). HHpred analysis indicates that Lsa1837 shares more than 50% sequence similarity with the ATPase subunits A and A′ of the same ECF in L. brevis (Fig. 2A). With 100% of probability and an E value of 1 ⋅ e⁻³⁵, Lsa1837 includes two repetitive domains, at positions 9 to 247 and positions 299 to 531, where each refers structurally to one ATPase very close in topology to the solved ATPase subunits, A and A′ of ECF from L. brevis, respectively. Appropriately, the N-terminal and C-terminal ATPases are predicted to contain an ATP-binding site. Lsa1837 could correspond to the fusion of ATPase subunits A and A′. Protein Lsa1838 shows sequence similarity of above 30%, with a probability of 100% and an E value of 1 ⋅ e⁻³⁰, to share structural homology with the membrane-embedded substrate-binding protein component T from the ECF transporter of L. brevis (Fig. 2A). Interestingly, similar results from bioinformatic sequence analysis and structure prediction demonstrated that Lsa1839 and Lsa1840 share both 99.8% structural homology, and E values of 1 ⋅ e⁻²⁴ and 1 ⋅ e⁻²¹, with the β and α domains of human transcobalamin, respectively (Fig. 2A). Consistently, both proteins have an export signal located at their N-terminal ends. Taken together, these results predict with high confidence that the transcriptional unit encodes the complete machinery of an ECF, including the extracellular proteins that initiate the scavenging of iron-containing heme (Fig. 2A). Each protein compartment is predicted through the presence or absence of its signal peptide as being extracellular, embedded in the membrane, or cytosolic. Correspondingly, every protein sequence is associated with the appropriate subcellular location with respect to its predicted function. In line with that, the network computed by String for the set of proteins of the operon shows that they interact together from a central connection related to Lsa1837, which corresponds to the ATP-motor couple of ATPases (45).

FIG 2.

(A) Structural and functional bioinformatic assessment for each gene of the lsa1836-1840 operon. (B) Lsa1839 and Lsa1840. (Left) The binding interaction and affinity of the human haptocorrin with cyanocobalamin and bovine transcobalamin with cobalamin, respectively, are highlighted. They were used as a 3D template and positive control for the modeling of transcobalamin-like proteins Lsa1840 and Lsa1839. (Right) Best pose of iron-containing heme as computed by Autodock4 within the binding pocket formed at the interface of α and β subunits of homology-modeled Lsa1840 and Lsa1839, respectively. The polar and hydrophobic interactions between the heme and the α and β chains are highlighted as brown sticks.

The transcriptional unit also encompasses Lsa1839 and Lsa1840, which are highly homologous to the β and α subunits of transcobalamin, respectively, that are commonly hypothesized to initiate the scavenging of heme from the extracellular medium. To address the capacity of those subunits of the transcobalamin-like binding domain to bind a heme moiety, we homology modeled Lsa1839 and Lsa1840. We then assembled the biological unit composed of the heterodimer formed by β and α subunits, using the related 3D templates of corresponding subunit of haptocorrin and transcobalamin. Subsequently, an iron-containing heme moiety was docked into the groove, located at the interface of the complex formed by the two proteins. The docking highlights a heme binding through polar and hydrophobic interactions. Nevertheless, no particular π stacking could be detected (Fig. 2B). The redocking of cobalamin in haptocorrin and cyanocobalamin in transcobalamin shows binding energies of −17 and −12 kcal/mol, respectively (Fig. 2B). With a binding energy of −9 kcal/mol, the heme bound to the crevice formed by Lsa1839 and Lsa1840 displays an affinity in the same range as the endogenous ligands, and this finding emphasizes that the assembly composed of Lsa1839 and Lsa1840 could be compatible with the recognition and binding of a heme (Fig. 2B). Furthermore, Lsa1836-1840 represents a complete machinery that could be able to internalize a heme instead of or in addition to a cobalamin molecule. Importantly, this operon includes also the extracellular scavenging α- and β-like subunits of transcobalamin, which promotes the S component Lsa1836 as likely very specific for iron-containing heme. Markedly, the S component displays a closely conserved fold, and yet it does not show any of the strictly conserved residues known to bind specifically cobalt-containing cobalamin.

No heme synthesis enzymes are present in L. sakei genome; nevertheless, a gene coding for a putative heme-degrading enzyme of the Dyp-type peroxidase family, lsa1831, was identified in the L. sakei genome (Table 1). Its structure is predicted to be close to DypB from Rhodococcus jostii (46). Interestingly, residues of DypB involved in porphyrin binding, namely, Asp153, His226, and Asn246, are strictly conserved in Lsa1831 (47). Markedly, the lsa1831 gene is located upstream of the lsa1836-1840 operon putatively involved in the active heme transport across the membrane.

Our bioinformatic analysis allows the functional reannotation of the lsa1836-1840 genes into the complete machinery of an ECF, possibly dedicated to the transport of iron through the heme (Fig. 3). Consistently, the Lsa1831 enzyme, which is close to the lsa1836-1840 loci, could participate downstream to release iron from the heme once inside the cytoplasm.

FIG 3.

(A) Functional reannotation of the operon lsa1836-1840 from L. sakei 23K after serial analysis of 3D structure-function prediction for each gene of the operon. (B) Reconstitution of iron-containing heme transport, initially scavenged between the α and β subunits of the transcobalamin-like transporter, coded by lsa1839-1840 and then transported from the extracellular into the intracellular compartments through the complete ECF machinery coded by the lsa1836-1838 portion of the operon. Possibly, gene lsa1831 positioned in the vicinity of the loci lsa1836-1840 could code for a protein Dyp-type peroxidase that ultimately releases the iron from the heme.

Lsa1836-1840 is in vitro an effective actor of heme uptake in L. sakei.

To confirm the above transporter as being involved in heme trafficking across the membrane, an lsa1836-1840 deletion mutant was constructed by homologous recombination. The L. sakei Δlsa1836-1840 mutant was analyzed for its capacity to internalize heme using an intracellular heme sensor developed by Lechardeur et al. (24). This molecular tool consists in a multicopy plasmid harboring a transcriptional fusion between the heme-inducible promoter of hrtR, the hrtR coding sequence, and the lacZ reporter gene, the pP_hrthrtR-lac (Table 2). In L. lactis, HrtR is a transcriptional regulator that represses the expression of a heme export system, HrtA and HrtB, as well as its own expression in the absence of heme. Upon heme binding, the repression is alleviated, allowing the expression of the export proteins (24). Since L. sakei possesses the lacLM genes, it was necessary to construct the Δlsa1836-1840 mutant in the L. sakei RV2002 strain, an L. sakei 23K ΔlacLM derivative, yielding the RV4057 strain (Table 2). pP_hrthrtR-lac was then introduced in the RV2002 and RV4057 strains, yielding the RV2002 hrtR-lac and RV4057 hrtR-lac strains (Table 2). The β-galactosidase (β-Gal) activity of the RV4057 hrtR-lac strain, grown in a chemically defined medium (MCD) (48) in the presence of 0.5, 1, and 5 μM hemin, was determined and compared to that of the RV2002 hrtR-lac used as control (Fig. 4A). We showed that hemin reached the intracellular compartment as β-Gal expression was induced by hemin. The relative β-Gal activity of the RV4057 hrtR-lac mutant strain showed a slight increase compared to the WT at 0.5 μM heme, but a significant 2-fold reduction was measured at 1 μM heme and, further, a 40% reduced activity was shown at a higher hemin concentration. This indicates that the intracellular abundance of heme is significantly reduced in the RV4057 bacterial cells at 1 and 5 μM heme, while it is similar to the WT at low heme concentrations. The method described above did not allow us to quantify the absolute amount of heme incorporated by bacteria since only cytosolic heme may interact with HrtR. Therefore, we used hemin labeled with the rare ⁵⁷iron isotope (⁵⁷Fe-hemin) combined with inductively coupled plasma mass spectrometry (ICP-MS) to measure with accuracy the total heminic-iron content of cells. Quantification of ⁵⁷Fe was used as a proxy to quantify heme. The absolute number of heme molecules incorporated by the Δlsa1836-1840 mutant was also quantified by using ⁵⁷Fe-hemin. The Δlsa1836-1840 mutant was constructed in the WT L. sakei 23K genetic background to obtain the RV4056 strain (Table 2). Bacteria were incubated in the MCD, in the absence or presence of 1, 5, or 40 μM ⁵⁷Fe-hemin. ICP-MS quantification indicated that the ⁵⁷Fe contents of the two strains were similar at 1 μM ⁵⁷Fe-hemin. By comparison with the WT, a 5-fold reduction in the ⁵⁷Fe content of the RV4056 strain at 5 μM heme concentration and a 8-fold reduction at 40 μM heme were measured (Fig. 4B).

TABLE 2.

Strains and plasmids used in this study

Strain or plasmid	Characteristics	Source or reference
L. sakei strains
23K	Sequenced strain	13
RV2002	23K derivative, ΔlacLM	60
RV2002 hrtR-lac	RV2002 carrying pPhrthrtR-lac; Eryr	This study
RV4056	23K derivative, Δlsa1836-1840	This study
RV4056c	RV4056 carrying pPlsa1836-1840; Eryr	This study
RV4057	RV2002 Δlsa1836-1840	This study
RV4057 hrtR-lac	RV4057 carrying pPhrthrtR-lac; Eryr	This study
Plasmids
pPhrthrtR-lac	Plasmid carrying the PhrtRhrtR-lac transcriptional fusion	24
pRV300	Shuttle vector, nonreplicative in Lactobacillus; Ampr Ermr	61
pRV566	Vector used for complementation; Ampr Ermr	62
pRV441	pRV300 derivative, exchange cassette for lsa1836-1840	This study
pPlsa1836-1840	pRV566 carrying the promoter and the lsa1836-1840 coding sequences	This study

FIG 4.

Heme incorporation is reduced in the Δlsa1836-1840 L. sakei deletion mutant. (A) In vivo detection of intracellular heme content of the RV2002 and Δlsa1836-1840 (RV4057) mutant strains. Strains carrying pP_hrtRhrtR-lac were grown in hemin and β-Gal activity was quantified by luminescence (see Materials and Methods). For each experiment, the luminescence values obtained with no added hemin are subtracted, and the β-Gal activity of strains was expressed as the percentage relative to the RV2002 strain for each hemin concentration. Mean values are shown (n = 3). Error bars represent the standard deviations. (B) Quantification of the ⁵⁷Fe content of the WT (23K) and the Δlsa1836-1840 (RV4056) strains grown in the absence or presence of the indicated ⁵⁷Fe-hemin concentrations. The results represent the means and ranges from at least two independent experiments. (C) Quantification of the ⁵⁷Fe content of the WT (23K), the Δlsa1836-1840 (RV4056), and the Δlsa1836-1840(pPlsa1836-1840) (RV4056c) strains grown in the absence or presence of the indicated ⁵⁷Fe-hemin concentrations. The results represent the means and ranges of two independent experiments.

To confirm the major role of the lsa1836-1840 gene products in heme acquisition, we analyzed the ⁵⁷Fe content of the RV4056 strain harboring the pPlsa1836-1840, a multicopy plasmid that expresses the lsa1836-1840 operon under its own promoter, and compared it to the WT. Quantification of the ⁵⁷Fe atoms in the RV4056 pPlsa1836-1840 bacteria showed 1.3-fold and 7-fold higher iron contents at 5 and 40 μM ⁵⁷Fe-hemin, respectively, compared to measurements performed on WT bacteria (Fig. 4C). These experiments confirm that the Lsa1836-1840 system is involved in vitro in the active incorporation of heme in L. sakei.

Heme accumulates inside the L. sakei cytosol at low heme concentrations.

We then addressed the ability for L. sakei to consume heme or iron to survive. We knew from a previous study that L. sakei incorporates preferentially heminic compounds from the medium, probably as an adaptation to its meat environment (12). Data obtained previously showed that the incorporation of heme molecules are qualitatively correlated with both the concentration of heme in the growth medium and the survival properties of the bacteria in stationary phase, suggesting that L. sakei could use heme or iron for its survival (see Fig. S1 and S2 in the supplemental material). Nevertheless, heme incorporation could not be quantified with accuracy in the previous studies. To tackle that, the intracellular heme levels incorporated by L. sakei were quantified. The RV2002 hrtR-lac strain (Table 2) was grown in MCD in the presence of increasing concentrations of hemin, and the β-Gal activity of the cells was measured (Fig. 5A). We found that the β-Gal activity increased with the concentration of the hemin molecule in the growth medium. A plateau was reached when cells were grown in 0.75 to 2.5 μM hemin. Incubation of cells in higher hemin concentrations did not lead to further increases in β-Gal activity.

FIG 5.

Quantification of heme incorporation in L. sakei. (A) In vivo detection of intracellular hemin molecules through the expression of the lacZ gene. The L. sakei RV2002 hrtR-lac strain was grown for 1 h in the presence of the indicated concentrations of hemin. The β-Gal activity was quantified using the luminescence (see Materials and Methods). Mean values are shown (n = 7). Error bars represent the standard deviations. Conditions for which the β-Gal activities of the cells are different compared to the control condition (0 μM hemin) are indicated with stars. Significance is based on the Kruskal-Wallis test, followed by Dunn’s multiple-comparison test (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant). (B) Quantification of the ⁵⁷Fe content of the WT (23K) strain grown in the absence or presence of ⁵⁷Fe-hemin for 1 and 19 h. The mean values and ranges of two independent experiments are shown. RLU, relative light units.

Heme incorporation in L. sakei is rapid and massive.

The absolute number of heme molecules incorporated by L. sakei 23K was also quantified using ⁵⁷Fe-hemin (Fig. 5B). Cells were grown in MCD in the presence of labeled hemin. Measurements of the ⁵⁷Fe content of cells showed that the incorporation of ⁵⁷Fe-hemin is massive and rapid since bacteria are able to incorporate about 35,000 ⁵⁷Fe atoms of heminic origin within 1 h in the presence of 1 μM ⁵⁷Fe-hemin (Fig. 5B). The iron content of cells increased to 160,000 and 260,000 atoms on average, respectively, when bacteria were grown in media containing 5 and 40 μM ⁵⁷Fe-hemin. This indicates that increase in the ⁵⁷Fe content in L. sakei cells is correlated with the increase in the ⁵⁷Fe-hemin content in the medium. This increase was observed with concentrations of ⁵⁷Fe-hemin ranging from 1 to 40 μM in the medium. Measurements of the iron content of bacteria, growing in the presence of ⁵⁷Fe-hemin for an extended period of time (19 h), did not show any additional ⁵⁷Fe accumulation in the bacteria (Fig. 5B). Instead, the number of ⁵⁷Fe atoms associated with bacteria decreased over time, highlighting the fact that a massive incorporation of labeled hemin occurs rapidly after bacteria are in contact with the molecules.

DISCUSSION

Heme acquisition systems are poorly documented in lactic acid bacteria, probably because neither heme nor iron is mandatory for growth of these bacterial species, at least under nonaerobic conditions. However, the acquisition of exogenous heme allows numerous lactic acid bacteria, among them L. lactis and Lactobacillus plantarum, to activate, if needed, a respiratory metabolism, when grown in the presence of oxygen (2, 49, 50). This implies that heme has to cross the thick cell wall of these Gram-positive organisms and may require heme transporters. Thus far, heme acquisition systems in heme auxotrophic organisms have only been reported for streptococci (20, 21) and S. pyogenes, where they both involve Shr and Shp NEAT domain proteins and Hts ABC transporter (20, 22, 23). To our knowledge, in lactic acid bacteria, NEAT domains have been identified in several species of lactic acid bacteria, including 15 Lactobacillus, 4 Leuconostoc, and 1 Carnobacterium species (19), but no such functional heme transport has been identified so far, and our present study confirmed that L. sakei proteins are devoid of such domains.

In L. lactis, the fhuCBGDR operon has been reported to be involved in heme uptake since an fhuD mutant is defective in respiration metabolism, suggesting a defect in heme import (15). A genome analysis of several lactic acid bacteria has revealed that a HupC/FepC heme uptake protein is present in L. lactis, L. plantarum, L. brevis, and L. sakei (15). The latter in L. sakei 23K may correspond to locus lsa0399 included in an fhu operon. An IsdE homolog has also been reported in L. brevis genome, but the identity of this protein has not been experimentally verified (15).

Genome analysis of L. sakei 23K (13), when focused on heme/iron transport systems and heme utilization enzymes, led to the identification of several putative iron transport systems, heme transport systems, and heme-degrading enzymes. This heme uptake potential is completely consistent within the meat environment-adapted L. sakei. Similarly, the membrane transport system encoded by lsa1194-1195 genes, whose function is poorly defined, seems to be important for the bacterial physiology since an lsa1194-1195 deletion affects the survival properties of this strain (see Fig. S3 and S4 and additional discussion in the supplemental material).

Meanwhile, here, we report that the transcriptional unit lsa1836-1840 shows exquisite structure-function homology with the cobalamin ECF transporter, a new class of ATP-binding cassette importer recently identified in the internalization of cobalt and nickel ions (Fig. 2 and 3). Indeed, a comprehensive bioinformatics analysis indicates that the lsa1836-1840 locus codes for five proteins that assemble together to describe a complete importer machinery called energy coupling factor. Any canonical ECF transporter comprises an energy-coupling module consisting of a transmembrane T protein (EcfT), two nucleotide-binding proteins (EcfA and EcfA′), and another transmembrane substrate-specific binding S protein (Ecsf). Indeed, Lsa1836-Lsa1838 shows high structural homology with Ecf-S, EcfA-A′, and Ecf-T, respectively. Despite sharing similarities with ABC transporters, ECF transporters have different organizational and functional properties. The lack of soluble-binding proteins in ECF transporters differentiates them clearly from the canonical ABC importers. Nevertheless, lsa1839 and lsa1840 here code for proteins structurally close to β and α subunits, respectively, of the transcobalamin-binding domain. They are highly suspected to be soluble proteins dedicated to scavenge heme from the extracellular compartment, and we hypothesize that they could bind it and then transfer it to the Ecf-S component coded by lsa1836 (Fig. 3). In line with this, the heterodimer composed of Lsa1839 and Lsa1840, possibly β and α subunits, respectively, has been modeled in silico and was shown to accommodate, with high affinity, an iron-heme ligand at the binding site, located at the interface of the two proteins.

Internalization of the cobalt and nickel divalent cations through porphyrin moiety via this new class of importer has been demonstrated in lactic acid bacteria, such as L. lactis and L. brevis. However, nothing was known for the internalization/incorporation of iron-containing heme. A functional analysis of the lsa1836-1840 gene products was undertaken using Δlsa1836-1840 deletion mutant and a complemented strain. Our experiments indicate that the intracellular abundance of heme is significantly reduced in Δlsa1836-1840 mutant bacterial cells at 1 and 5 μM heme, while it is similar to the WT at low heme concentrations. Conversely, the mutant strain, in which lsa1836-1840 is expressed from a multicopy plasmid, showed an increase in the heme uptake. Taken together, these experiments confirm that the Lsa1836-1840 system is involved in vitro in the active incorporation of heme in L. sakei. To our knowledge, this is the first time that an ECF has been reported to be involved in heme incorporation. One could consider that such an ability to transport and accumulate heme/iron may represent an ecological fitness trait for surviving in the heme-rich meat ecosystem, wherein heme does not represent a limiting resource that would lead for competition strategies between species. This is probably true, not only for L. sakei but also for the other meat resident species, since our synteny analysis for this operon shows that this feature could be shared within several Gram-positive meat-borne bacteria.

In addition, we were able to quantify the amount of heme internalized in the three genetic contexts using isotope-labeled hemin and ICP-MS, as well as to evaluate the intracellular content of heme using a transcriptional fusion tool. We observed that the intracellular abundance of heme increases with the concentration of heme in the growth medium and can be detected with the intracellular sensor in the 0 to 2.5 μM heme range (Fig. 5A). The drop in β-Gal activity at higher heme concentrations may result from regulation of heme/iron homeostasis through exportation of heme, degradation of the intracellular heme, or storage of the heme molecules, making them unable to interact with HrtR and promoting lacZ repression. However, data obtained with the intracellular sensor at higher heme concentrations (5 to 40 μM) contrast with the microscopic observations (see Fig. S2) and ICP-MS measurements (Fig. 5B), which reported a higher heminic-iron content in cells grown in 40 μM heme than in cells grown 5 μM heme. Indeed, the β-Gal activity reflecting the abundance of intracellular heme was maximal when cells were grown in a medium containing 1 to 2.5 μM hemin (Fig. 5A), while ICP-MS measurements showed 4.5- and 8-fold higher numbers of ⁵⁷Fe atoms in bacteria growing in 5 or 40 μM ⁵⁷Fe-hemin, respectively, than in 1 μM ⁵⁷Fe-hemin (Fig. 4B). These data are in good agreement with the electron energy loss spectroscopy (EELS) analysis (Fig. S2), which strengthens the hypothesis that heme homeostasis occurs in L. sakei and that the incorporated heme molecules would be degraded while iron is stored inside iron storage proteins such as Dps, of which orthologous genes exist in L. sakei. Thus, iron is detected in L. sakei cells but not bound to heme and unable to interact with the intracellular heme sensor HrtR. Storage of heme inside membrane proteins is still an open question since L. sakei does not contain cytochromes nor menaquinones (12).

Further analysis is required not only to decipher the exact role of these proteins during the different steps of heme transport across the L. sakei membrane and the fate of heme inside L. sakei cells but also to better understand the molecular specificity of the Lsa1836-1840 machinery toward iron-containing heme versus cobalamin.

MATERIALS AND METHODS

Bacterial strains and general growth conditions.

The different bacterial strains used throughout this study are described in Table 2. L. sakei and its derivatives (RV2002, RV2002 hrtR-lac, RV4056, RV4056c, RV4057, and RV4057 hrtR-lac) were propagated on MRS (2) at 30°C. For physiological studies, the chemically defined medium MCD (3) supplemented with 0.5% (wt/vol) glucose was used. MCD contains no iron sources but contains possible traces of iron coming from various components or distilled water. Incubation was performed at 30°C without stirring. Cell growth and viability of cells in stationary phase were followed by measuring the optical density at 600 nm (OD₆₀₀) on a visible spectrophotometer (Secoman) and by determination of the number of CFU ml⁻¹ after plating serial dilutions of samples on MRS agar. When needed, media were supplemented with filtered hemin or hematin (Sigma-Aldrich) or with ⁵⁷Fe-hemin (Frontier Scientific) solutions resuspended in 50 mM NaOH.

E. coli K-12 strain DH5α was used as the host for plasmid construction and cloning experiments. E. coli cells were chemically transformed as previously described (4). L. sakei cells were transformed by electroporation as previously described (5). For routine growth, E. coli was propagated in Luria-Bertani medium at 37°C under vigorous shaking (175 rpm). The following concentrations of antibiotic were used for bacterial selection: kanamycin at 20 μg/ml and ampicillin at 100 μg/ml for E. coli and erythromycin at 5 μg/ml for L. sakei.

DNA manipulations.

Chromosomal DNA was extracted from LS cells with a DNA isolation kit for cells and tissues (Roche, France). Plasmid DNA was extracted from E. coli by a standard alkaline lysis procedure with a NucleoSpin plasmid kit (Macherey-Nagel, France). PCR-amplified fragments and digested fragments separated on 0.8% agarose gels were purified with kits from Qiagen (France). Restriction enzymes, Taq or Phusion high-fidelity polymerase (Thermo Scientific, France) and T4 DNA ligase (Roche), were used in accordance with the manufacturer’s recommendations. Oligonucleotides (Table 3) were synthesized by Eurogentec (Belgium). PCRs were performed in an Applied Biosystems 2720 Thermak thermocycler (ABI). Nucleotide sequences of all constructs were determined by MWG-Eurofins (Germany).

TABLE 3.

Oligonucleotides used in this study

Primer	Sequence (5′–3′)a	Restriction site
PHDU-lsa1836F	CATGGTACCGGTCGGCTCAATTATGAGT	KpnI
PHDU-lsa1836R	AATGAACTAGTTAGCGCTCGCAGCCTATATTGCGAGT
PHDU-lsa1840F	AGCGCTAACTAGTTCATTAGACTTCCGTCACTTGTGAA
PHDU-lsa1840R	CTGGAATTCATGCTGAGCGATGGTTTCT	EcoRI
PHDU-crblsa1840F	CGACAAGTCAACTCAGTGCTA
PHDU-crblsa1840R	GTGAACCGTAATCTTGAGTG
Lsa1836R	TTCCCGGGAACTTACAAAAGGCCACGC	XmaI
Lsa1840F	AAAAGCGGCCGCGCCTCCTTATAAAAACTG	NotI
566-F	GCGAAAGAATGATGTGTTGG
566-R	CACACAGGAAACAGCTATGAC

^aUnderlined sequences indicate the location of restriction sites, and italicized letters indicate complementary overlapping sequences used to join PCR fragments as described in Materials and Methods.

Bioinformatic analyses.

Analyses were performed in the sequenced L. sakei 23K genome (accession number CR936503) as described previously (13). Each fasta sequence of every gene of the operon comprised between lsa1836 and lsa1840 was retrieved from the UnitProtKB server (http://www.uniprot.org/uniprot) and then uploaded and analyzed using the HHpred server (44) that detects structural homologues. For Lsa1839, which shares strong structural homology with the β domain of human transcobalamin bound to co-β-[2-(2,4-difluorophenyl)ethinyl]cobalamin (PDB 5NSA, chain A) (51) and the β domain of human haptocorrin (PDB 4KKI, chain A) (52), and for Lsa1840, which shares strong structural homology with the intrinsic factor with cobalamin (PDB 2PMV) (53) and transcobalamin (PDB 2BB6, chain A) (54), homology modeling was performed using Modeler, vMod9v18 (55). The heterodimer was then formed with respect to the functional and structural assembly of α and β domains of the native haptocorrin (52). Upon dimer formation, the best poses for heme inside the groove, which is located at the interface of this heterodimer, were computed using the Autodock4 tool (56). The protocol and grid box were previously validated with the redocking of cyanocobalamin within human haptocorrin (4KKI) (42) and of cobalamin within bovine transcobalamin (2BB6). To compute the binding energy of every complex, the parameters of the cobalt present in the cobalamin and cyanocobalamin were added to the parameter data table, while the iron parameters of the heme were already noted in the parameter data table. The docking poses were then explored using the Lamarckian genetic algorithm and subsequently analyzed using PyMOL of the Schrödinger suite (57).

Comparative genomic analysis for conservation of gene synteny between meat-borne bacteria was carried out with the MicroScope genome annotation platform, using the Genome Synteny graphical output and the PkGDB synteny statistics (58).

Construction of plasmids and L. sakei mutant strains.

All the primers and plasmids used in this study are listed in Tables 2 and 3. The lsa1836-1840 genes were inactivated by a 5,118-bp deletion using double-crossover strategy. Upstream and downstream fragments were obtained using primer pairs PHDU-lsa1836F/PHDU-lsa1836R (731 bp) and PHDU-lsa1840F/PHDU-lsa1840R (742 bp) (Table 3). PCR fragments were joined by splicing by overlap extension (SOE) using the primers PHDU-lsa1836F/PHDU-lsa1840R, and the resulting 1,456-bp fragment was cloned between EcoRI and KpnII sites in pRV300, yielding pRV441 (Table 2). pRV441 was introduced into the L. sakei 23K and L. sakei 23K ΔlacLM (RV2002) strains by electroporation, as described previously (59). Selection was done based on erythromycin sensitivity. Second crossover erythromycin-sensitive (Erm^s) candidates were screened using the primers PHDU-crblsa1840F and PHDU-crblsa1840R (Table 3). Deletion was then confirmed by sequencing the concerned region, and the lsa1836-1840 mutant strains were named RV4056 and RV4057 (Table 2). To construct the RV2002 hrtR-lac and the RV4057 hrtR-lac strains, the pP_hrthrtR-lac (Table 2) was transformed by electroporation into the corresponding mother strains.

For complementation, a pPlsa1836-1840 plasmid (Table 2) was constructed as follows. A DNA fragment encompassing the promoter and the five genes of the lsa1836-1840 operon was PCR amplified, using the primer pair Lsa1836R/Lsa1840F (Table 3). The 5,793-bp amplified fragment was cloned into plasmid pRV566 at the XmaI and NotI sites. The construct was verified by sequencing the whole DNA insert using the 566-F and 566-R primers (Table 3), as well as internal primers. pPlsa1836-1840 was introduced into RV4056 bacteria by electroporation, and transformed bacteria were selected for erythromycin resistance, yielding the RV4056c complemented mutant strain.

β-Galactosidase assay.

Liquid cultures were usually grown in MCD to exponential phase corresponding to an A₆₀₀ equal to 0.5 to 0.8, followed by incubation for 1 h at 30°C with hemin at the indicated concentration. The β-galactosidase (β-Gal) activity was assayed on bacteria permeabilized as described. The β-Gal activity was quantified by luminescence in an Infinite M200 spectroluminometer (Tecan), using the β-Glo assay system as recommended by the manufacturer (Promega).

Intracellular ⁵⁷Fe determination.

The various strains were grown in MCD to an A₆₀₀ of 0.5 to 0.7 at 30°C, prior to the addition (or without addition) of 0.1, 1, 5, or 40 μM ⁵⁷Fe-labeled hemin (Frontier Scientific). Cells were then incubated at 30°C for an additional hour and overnight (19 h). The cells were washed three times in H₂O supplemented with 1 mM EDTA. Cell pellets were desiccated and mineralized by successive incubations in 65% nitric acid solution at 130°C. ⁵⁷Fe was quantified by ICP-MS (Agilent 7700X; Géosciences, University of Montpellier, Montpellier, France).

Statistical analysis.

To determine whether the differences in heme incorporation by L. sakei cells grown in the presence of increasing concentrations of heme, measured using the molecular reporter, were different from the control condition (cells grown in the absence of heme), a nonparametric Kruskal-Wallis test, followed by Dunn’s multiple-comparison test with a family-wise significance and a confidence level of 0.05, was performed using Prism version 8.4.2 for macOS (GraphPad Software, La Jolla, CA).