The MprF homolog LysX synthesizes lysyl-diacylglycerol contributing to antibiotic resistance and virulence

ABSTRACT

Lysyl-diacylglycerol (Lys-DAG) was identified three decades ago in Mycobacterium phlei, but the biosynthetic pathway and function of this aminoacylated lipid have since remained uncharacterized. Combining genetic methods, mass spectrometry, and biochemical approaches, we show that the multiple peptide resistance factor (MprF) homolog LysX from Corynebacterium pseudotuberculosis and two mycobacterial species is responsible for Lys-DAG synthesis. LysX is conserved in most Actinobacteria and was previously implicated in the synthesis of another modified lipid, lysyl-phosphatidylglycerol (Lys-PG), in Mycobacterium tuberculosis. Although we detected low levels of Lys-PG in the membrane of C. pseudotuberculosis, our data suggest that Lys-PG is not directly synthesized by LysX and may require an additional downstream pathway, which is as yet undefined. Our results show that LysX in C. pseudotuberculosis is a major factor of resistance against a variety of positively charged antibacterial agents, including cationic antimicrobial peptides (e.g., human peptide LL-37 and polymyxin B) and aminoglycosides (e.g., gentamycin and apramycin). Deletion of lysX caused an increase in cellular membrane permeability without dissipation of the membrane potential, suggesting that loss of the protein does not result in mechanical damage to the cell membrane. Furthermore, lysX-deficient cells exhibited an attenuated virulence phenotype in a Galleria mellonella infection model, supporting a role for LysX during infection. Altogether, Lys-DAG represents a novel molecular determinant for antimicrobial resistance and virulence that may be widespread in Actinobacteria and points to a richer landscape than previously realized of lipid components contributing to overall membrane physiology in this important bacterial phylum.

IMPORTANCE

In the past two decades, tRNA-dependent modification of membrane phosphatidylglycerol has been implicated in altering the biochemical properties of the cell surface, thereby enhancing the antimicrobial resistance and virulence of various bacterial pathogens. Here, we show that in several Actinobacteria, the multifunctional protein LysX attaches lysine to diacylglycerol instead of phosphatidylglycerol. We found that lysyl-diacylglycerol (Lys-DAG) confers high levels of resistance against various cationic antimicrobial peptides and aminoglycosides and also enhances virulence. Our data show that Lys-DAG is a lipid commonly found in important actinobacterial pathogens, including Mycobacterium and Corynebacterium species.

INTRODUCTION

The cell envelope is the first line of defense that protects bacteria from the stressors of the environment. Cells have evolved systems to adapt the constituents of their cellular envelope in response to changing environmental conditions (1). One of these systems utilizes aminoacyl-tRNA (aa-tRNA) as a donor of amino acids (aa’s) for the aminoacylation of the polar head groups of membrane lipids. The multiple peptide resistance factor (MprF), first discovered in Staphylococcus aureus, exhibits lysyl-phosphatidylglycerol synthase (LysPGS) activity that uses Lys-tRNA to aminoacylate the free hydroxyl group of phosphatidylglycerol (PG) to produce lysyl-phosphatidylglycerol (Lys-PG). The addition of Lys to membrane PG decreases the net negative charge of the cell surface, thereby increasing resistance to various cationic antimicrobial peptides (CAMPs) and enhancing bacterial evasion of the killing activities of neutrophils and macrophages (2). MprF is a bi-functional integral membrane protein comprising a GCN5-related N-acetyltransferase (GNAT)-like C-terminal domain supporting the aa-tRNA/phosphatidylglycerol transferase activity and an N-terminal flippase (i.e., translocase) domain, which flips neosynthesized lipids from the inner surface of the cytoplasmic membrane to the outer surface (3, 4).

Since the initial discovery of MprF in S. aureus, a family of orthologs has been uncovered, with members appearing in the genomes of most bacterial phyla, in archaea (5), and, most recently, in fungal pathogens (6 – 8). Members of this family of proteins exhibit differing specificities for both the aa’s and lipids that are used as substrates. Lipid modification by Lys and Ala (9, 10) has been described in various bacterial species, while Gly and Asp are substrates used by the fungal proteins (6 – 8). Likewise, some of these enzymes act on lipids that are distinct from PG, such as cardiolipin in proteobacteria (11, 12) and diacylglycerol (DAG) in Corynebacterium glutamicum (1 3). The more distant MprF orthologs found in fungi aminoacylate a sterol (i.e., ergosterol) instead of a glycerolipid (e.g., PG, DAG, or cardiolipin) (6 – 8, 14), demonstrating the substrate diversity among this family of proteins.

The phenotypes associated with Lys-PG are by far the best characterized. It is thought that lysylation of PG decreases the net electrostatic charge of the cytoplasmic membrane, thereby enhancing resistance to antibiotics and CAMPs, substances secreted by the innate immune system of infected hosts [for review, see references (15 – 18)]. Lys-PG increases bacterial virulence in cell cultures and animal models (11, 19, 20) and enhances resistance to both macrophages (20) and neutrophils (2, 21). These effects have been observed in the context of a variety of Gram positive human pathogens such as Staphylococcus aureus (2, 19, 21, 22), Bacillus anthracis (23), Listeria monocytogenes (11, 24, 25), and Mycobacterium tuberculosis (20). Similar effects have been described in Gram negative species such as Pseudomonas aeruginosa (5, 9, 26) and Rhizobium tropici (27).

MprF homologs in Actinobacteria are structurally and functionally distinct from the canonical proteins found in other bacterial species. For instance, an alanyl-diacylglycerol synthase (AlaDAGS) responsible for the synthesis of alanyl-diacylglycerol (Ala-DAG) was identified in Corynebacterium glutamicum (13). In Mycobacterium turberculosis, the gene lysX (Rv1640c) encodes a 1,176-amino acid long peptide consisting of an integral membrane domain and GNAT-like domain, characteristic of the MprF family of proteins, in addition to a C-terminal lysyl-tRNA synthetase (LysRS) domain used for synthesis of the Lys-tRNA^Lys necessary for formation of Lys-PG (20). LysX has been implicated as an important factor for antibiotic resistance and host-pathogen interactions in the M. tuberculosis pathogen. Cells harboring mutations in the protein are sensitive to various cationic antibiotics and peptides (20, 28) and show defective growth phenotypes in mouse and guinea pig lung models (20, 29). LysX was also shown to be important for bacterial survival in human monocytes and macrophages (30, 31).

Many Actinobacteria, notably species of the Corynebacterium, Mycobacterium, and Nocardia (CMN) group, exhibit two MprF paralogs (13), one homologous to C. glutamicum AlaDAGS and the other to M. turberculosis LysX. Here, we study the role of these paralogs in the model organism Corynebacterium pseudotuberculosis, an important pathogen responsible for caseous lymphadenitis and other chronic diseases in livestock animals and, occasionally, humans (32, 33). Using a combination of genetic, biochemical, and mass spectrometry (MS) approaches, we show that the LysX homolog in C. pseudotuberculosis is responsible for the synthesis of the modified lipid lysyl-diacylglycerol (Lys-DAG). Although Lys-DAG was first reported over three decades ago in Mycobacterium phlei, its role and biosynthetic pathway have since remained elusive (34, 35). Lys-DAG was the only amine-containing lipid (ACL) detectable by TLC (thin layer chromatography) in C. pseudotuberculosis membrane extracts; however, trace amounts of Lys-PG were also revealed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). To further explore the role of LysX in related organisms, the protein’s activity was interrogated in two additional mycobacterial species. We find that LysX homologs from Mycobacterium abscessus and Mycobacterium parafinicum also synthesize Lys-DAG. Our data suggest that the only product of LysX activity is Lys-DAG and the low amounts of Lys-PG in C. pseudotuberculosis may be the product of a secondary pathway that may use Lys-DAG as a substrate. We show that LysX in C. pseudotuberculosis is an important resistance factor against various cationic antibacterial agents, including the human pore-forming peptide LL-37, polymyxin B, and several positively charged aminoglycosides. Deletion of lysX disturbs cellular uptake of the cationic dye propidium iodine (PI) without dissipation of the membrane potential, demonstrating that lysX is essential for membrane impermeability to cationic compounds but not for maintaining a proton gradient. The lysX mutant strain also exhibits an attenuated virulence phenotype in a Galleria mellonella infection model.

RESULTS

MprF homologs from C. pseudotuberculosis synthesize two amine-containing lipids

The two MprF homologs from C. pseudotuberculosis share a high percentage of identity with the previously characterized LysX from M. turberculosis (42%) and AlaDAGS from C. glutamicum (45%). Both enzymes exhibit an integral membrane domain with six predicted transmembrane helices (based on the TOPCONS web server) and a GNAT-like domain, which catalyzes the aa-tRNA lipid transferase reaction (Fig. 1A). Both of these domains are characteristic of the MprF family of enzymes (18, 36 – 38). The LysX proteins from M. turberculosis and C. pseudotuberculosis share other architectural features as well. They both include a LysRS domain at their C-terminal ends. This domain includes a catalytic domain characteristic of a LysRS and an oligonucleotide-binding (OB) fold, which, in the context of a cytosolic LysRS, is used to bind the tRNA anticodon (39) (Fig. 1A). In M. tuberculosis, it was demonstrated that the LysRS domain of LysX synthesizes the Lys-tRNA^Lys substrate used for Lys-PG formation, which is catalyzed in a secondary reaction by the MprF GNAT-like domain (20).

FIG 1.

MprF paralogs and analysis of ACLs in C. pseudotuberculosis. (A) Architecture of MprF paralogs in C. pseudotuberculosis. The boundaries of the membrane domain, the GNAT transferase domain, and the LysRS domain were established using known structures of MprF homologs (36, 37, 39). LysX contains features characteristic of a class II LysRS, including an anticodon binding domain (OB-Fold) and an aa-tRNA catalytic domain. (B) TLC analysis of lipids from C. pseudotuberculosis. Total lipids were visualized using iodine vapors (left image), and ACLs were specifically revealed by spraying the same TLC with ninhydrin reagent (right image). Lipids from wt C. pseudotuberculosis (Cps), ΔalaDAGS, and ΔlysX strains are shown in the three left-most lanes in both images. The three next lanes show lipids isolated from C. glutamicum ΔpesTΔalaDAGS cells (CglΔ) (13), harboring either the expression vector pEKEx2 (pEK) alone or heterologously expressing alaDAGS or lysX from C. pseudotuberculosis. Control lipids, Ala-PG, Lys-PG, Ala-DAG, and PE, were obtained from C. glutamicum (Cgl) and E. coli C41, expressing the LysPGS from C. perfringens (10), and are shown in the two right-most lanes. ACL1 and 2, amine-containing lipids; o, TLC origin; f, TLC front.

To determine whether the activities of the C. pseudotuberculosis enzymes are identical to the previously characterized proteins, we carried out thin layer chromatography (TLC) analysis of total lipid extracts. Extracted lipids from C. pseudotuberculosis revealed two amine-containing lipids (ACL1 and ACL2; Fig. 1B). ACL1 migrated similarly to the control lipids alanyl-phosphatidylglycerol (Ala-PG) and phosphatidylethanolamine (PE), while ACL2 migrated like Ala-DAG. Lipids from ΔalaDAGS and ΔlysX strains were analyzed to determine which MprF homolog is responsible for the synthesis of each of the ACLs. The ΔlysX strain was unable to produce ACL1, and the ΔalaDAGS strain was unable to produce ACL2.

To confirm the correlation of the C. pseudotuberculosis homologs with the formation of ACL1 and ACL2, LysX and AlaDAGS from C. pseudotuberculosis were cloned into the vector pEKEx-2 and expressed heterologously in C. glutamicum. The C. glutamicum strain ΔpesTΔalaDAGS (deprived of the gene associated with the synthesis of Ala-DAG and Ala-PG) was used to avoid interference with the activities of the C. pseudotuberculosis enzymes. Expression of C. pseudotuberculosis lysX in C. glutamicum promoted synthesis of ACL1, while expression of the AlaDAGS homolog promoted synthesis of ACL2 (Fig. 1B). Notably, expression of lysX in C. glutamicum did not promote the synthesis of Lys-PG, consistent with the results obtained in C. pseudotuberculosis described above. Collectively, these data suggest that the AlaDAGS homolog is responsible for the synthesis of Ala-DAG, while LysX mediates the formation of an amine-containing lipid, ACL1, which shares similar chromatographic properties with Ala-PG or PE but not Lys-PG. ACL1 and ACL2 were found to be the main amine-containing lipids produced in wild-type (wt) C. pseudotuberculosis.

LysX catalyzes the tRNA-dependent lysylation of ACL1

To further define the substrates of LysX and the identity of ACL1, membrane extracts from wild-type C. pseudotuberculosis and the ΔlysX strain were assayed in vitro using [¹⁴C]-Lys in the presence or absence of total tRNA and LysRS from Escherichia coli. Figure 2 shows that no lipid lysylation was observed with membrane extracts from the ΔlysX strain. However, [¹⁴C]-Lys-labeled ACL1 was the main product of the reaction when a membrane extract from the wild-type strain was used. A minor product migrating at the same location as the control Lys-PG (above the TLC origin; Fig. 2) was also observed, suggesting that Lys-PG synthase activity may also be present in the membrane extract from C. pseudotuberculosis. It is worth noting that some radiolabeled material did not migrate and was detected at the origin of the TLC. Production of both ACL1 and the minor product was dependent on the addition of tRNA, as the absence of tRNA abolished their synthesis. ACL1 synthesis was independent, however, of exogenous LysRS, suggesting that the LysRS domain of LysX is functional in the assay and able to aminoacylate the tRNA used for lipid labeling. Altogether, these results suggest that C. pseudotuberculosis LysX, like its homolog in M. tuberculosis, utilizes lysine for lipid aminoacylation. However, contrary to the case in M. tuberculosis, the predominant recipient of lysine in the C. pseudotuberculosis LysX pathway appears to be a lipid other than PG.

FIG 2.

Activity of C. pseudotuberculosis LysX in vitro. (A) LysX activity was reconstituted in vitro using [¹⁴C]-Lys and membrane extracts from C. pseudotuberculosis wt (Cps) or ΔlysX (CpsΔlysX) cells. The reaction medium was supplemented with DAG and PG in the presence or absence of total tRNAs and LysRS from E. coli (EcLysRS), to interrogate the specificity of the reaction. [¹⁴C]-lysylated products were separated by TLC and visualized by autoradiography. (B) Control lanes. ACL1 was obtained from strain C. glutamicum heterologously expressing LysX from C. pseudotuberculosis and was revealed by spraying with ninhydrin reagent (Fig. 1). [¹⁴C]-Lys-PG and [¹⁴C]-Ala-PG were obtained from an in vitro reaction medium including membrane extracts from E. coli expressing the AlaPGS or LysPGS from Clostridium perfringens and the corresponding [¹⁴C]-Ala and [¹⁴C]-Lys, respectively (10). TLC front (f) and origin (o) are indicated.

Identification of C. pseudotuberculosis ACLs by LC-MS/MS

To determine the structure of ACL1, lipid extracts from wild-type C. pseudotuberculosis and deletion strains (ΔlysX and ΔalaDAGS) were analyzed by LC-MS/MS set to the positive mode. Four aminoacylated lipids were identified in the lipid extract from the wild-type strain. Ala-PG (34:1) and Lys-PG (32:1) were identified, as well as various species of Ala-DAG (34:0, 34:1) and Lys-DAG (32:1, 32:2, 34:1; Fig. 3). The predominant MS2 ion products corresponded to those of a protonated, dehydrated diacylglycerol moiety ([DAG-H₂O + H]⁺; Fig. 3B) (40). Ala-PG and Ala-DAG exhibited fragmentation patterns identical to those previously reported, with neutral losses of 243 and 89 amu, corresponding to the loss of the alanylated polar head groups of these lipids (alanyl-glycerophosphate and ala, respectively) (13). Identical fragmentation patterns with neutral losses corresponding to the polar head groups of Lys-PG and Lys-DAG (300 amu for lysyl-glycerophosphate and 146 amu for Lys) were also observed. A protonated ion fragment corresponding to the lysyl-glycerophosphate polar head moiety was also detected at 301 amu, consistent with previous observations for Lys-PG (23). Ala-PG and Ala-DAG were absent from lipids extracted from ΔalaDAGS cells, and Lys-DAG was missing from the ΔlysX strain (data not shown), further supporting the identity of these lipids and their associations with the corresponding enzymes. The LC-MS/MS chromatogram suggests that Lys-DAG is the main lysylated lipid in wild-type C. pseudotuberculosis, consistent with the TLC analysis of membrane lipids described above (Fig. 1). LC-MS/MS analysis of lipids isolated from C. glutamicum cells heterologously expressing lysX revealed several species of Lys-DAG (32:1, 34:1, 34:2, 36:1, 36:2), but no Lys-PG was detected (data not shown). These data suggest that Lys-DAG is the unique product of LysX in this species. Lys-PG may be produced by a secondary enzyme in a fashion similar to the pathway used for the synthesis of Ala-PG in C. glutamicum (13).

FIG 3.

LC-MS/MS analysis of aminoacylated lipids in C. pseudotuberculosis. (A) Base peak (m/z 200–2000) chromatogram of total lipids (left) and chromatographic deconvolution of the different species of aminoacylated PG and DAG (right). (B) Collision-induced dissociation of MS1 peaks at m/z 695, 666, 849, and 820, corresponding to Lys-DAG, Ala-DAG, Lys-PG, and Ala-PG, respectively. All data were acquired in the positive mode.

Confirmation of Lys-DAG by isotopic labeling

To our knowledge, the neutral losses described above, together with the observation of dehydrated DAG as an ion product (i.e., [DAG + H-H₂O]⁺ in Fig. 3), are not consistent with any other known lipids (40, 41), lending support to our determination of Lys-DAG as the product of LysX activity in C. pseudotuberculosis. However, the neutral loss of 146 amu, consistent with the loss of Lys, could also correspond to the loss of a rhamnosyl group, even though no known rhamnosylated lipids would yield [DAG + H-H₂O]⁺ as an ion product (42 – 44). Nevertheless, to confirm the identity of Lys-DAG in the system, isotopic labeling was performed in vitro using a membrane extract from C. pseudotuberculosis as a source of LysX, in the presence or absence of tetradeuterated L-Lys (D4Lys, Fig. 4), and with supplemental PG and DAG added to the reaction. The products of the reaction were analyzed by MS2. In the absence of D4Lys, a peak corresponding to Lys-DAG 34:1, which was pre-existing in the membrane extract, was detected at a m/z of 723, while the addition of D4Lys promoted synthesis of the labeled D4Lys-DAG 34:1 at a m/z value of 727.5. MS2 analysis of each MS1 peak demonstrated characteristic fragmentation of Lys-DAG 34:1, with a neutral loss of 150 amu for the loss of tetradeuterated Lys and 146 amu for the loss of the unlabeled amino acid. Altogether, these data confirm the identity of ACL1 as Lys-DAG and that this lipid is synthesized by LysX.

FIG 4.

Isotopic labeling of Lys-DAG in vitro. (A) C. pseudotuberculosis membrane extract was used in a tRNA-dependent lipid aminoacylation assay in the absence (top) or presence of D4Lys (bottom). (B) MS2 analysis of MS1 peaks at m/z 723 (Lys-DAG 34:1) and 727 (D4Lys-DAG 34:1).

LysX from M. abscessus and M. parafinicum produces Lys-DAG

Previous reports suggested that LysX in M. tuberculosis synthesizes Lys-PG instead of Lys-DAG (20, 29, 30). Based on our experimental evidence regarding the role of LysX in the synthesis of Lys-DAG and the observation that this lipid occurs in other mycobacterial species (i.e., M. phlei) (34), we decided to investigate LysX-mediated lipid aminoacylation in two additional mycobacterial species. LysX from M. parafinicum was cloned into pEKEx-2 and heterologously expressed in C. glutamicum lacking endogenous MprF homologs (i.e., the ΔpesTΔalaDAGS strain). LC-MS/MS analysis of lysylated lipids isolated from LysX-expressing cells revealed the presence of Lys-DAG 34:1 but not of Lys-PG (Fig. S1). To investigate the role of LysX in M. abscessus, the lysX gene was disrupted using the strategy developed by Viljoen et al. (45) to generate the knockout strain, lysX:pVT. LC-MS/MS analysis of lipids from wild-type M. abscessus revealed the presence of several species of Lys-DAG (32:0, 32:1, 34:0, 34:1, 36:1) but not Lys-PG. As expected, Lys-DAG was entirely absent from the lysX:pVT strain (data not shown). These results confirm that the product of LysX aminoacylation in these two mycobacterial species is Lys-DAG and not Lys-PG.

LysX increases resistance to positively charged antibiotics

Lysylation of PG or CL was reported to increase antimicrobial resistance against various CAMPs and other classes of antimicrobials such as penicillins and aminoglycosides. These observations were made in diverse bacterial genera such as Staphylococcus (2, 46, 47), Listeria (11), Bacillus (23), and Mycobacterium (20). To investigate the role of LysX-mediated lysylation in C. pseudotuberculosis, we first assessed the impact of lysX mutations on bacterial generation time (Fig. S2). ΔlysX cells grew significantly slower (by 46%) than the wild-type strain (P < 0.0001), while ΔalaDAGS cells were unaffected. To determine the effect of lipid aminoacylation on the antibiotic resistance of C. pseudotuberculosis, minimum inhibitory concentrations (MICs) of various inhibitors were measured for the wild-type and mutant strains. Disruption of lysX decreased the MIC of various positively charged, pore-forming antimicrobials (i.e., CAMPs and aminoglycosides; Table 1). The human CAMP LL-37 could not be added at high enough concentrations to determine the MIC with the wild-type strain, but the decrease in the MIC after deletion of lysX was estimated to be >80-fold. Mutation of lysX decreased the MIC for the highly charged antibiotic polymyxin B by 64-fold, and for the aminoglycosides gentamicin and apramycin, the MICs decreased 64- and 32-fold, respectively. There was an apparent correlation between the degree of positive charge born by the antimicrobials and the magnitude of resistance associated with lysX (Table 1). Thus, lysX conferred the strongest protection against CAMPs with five or six positive charges (such as those listed above). Lesser effects were measured for compounds exhibiting a lower number of positive charges, and no protective advantage was observed with negatively charged antimicrobials such as dermcidin and ceftriaxone. In contrast to the lysX gene, deletion of alaDAGS did not significantly alter bacterial resistance to any of the tested antimicrobials. Therefore, we did not determine the MIC for the ΔalaDAGS complementation strain (Table 1).

TABLE 1.

Minimum inhibitory concentrations against C. pseudotuberculosis wild-type, ΔalaDAGS, and ΔlysX mutants

Antibiotics	Charge a	MIC (µg/mL)				wt/ΔlysX (fold-change)
		wt	ΔlysX	ΔalaDAGS	ΔlysX complement
Peptides/glycopeptides/lipopeptides
Human LL-37 c	+6	>60	0.75	>60	>60	>80
Polymyxin B	+5	640	10	640	640	64
Nisin	+4	80	2.5	80	80	32
Melittin	+5	>40	10	>40	>40	>4
Human defensin HNP-1	+3	3	0.75	3	3	4
Vancomycin	+1	0.75	0.375	0.75	N.D. b	2
Bacitracin	0	6.4	3.2	6.4	N.D.	2
Dermcidin DCD-1L	−5	>50	>50	>50	N.D.	1
Penicillins, cephalosporins
Oxacillin	−1	2.5	1.25	2.5	N.D.	2
Ampicillin	0	1.25	1.25	1.25	N.D.	1
Azlocillin	−1	0.156	0.156	0.156	N.D.	1
Ceftriaxone	−2	0.625	0.625	0.625	N.D.	1
Macrolides
Tulathromycin	+3	2.5	1.25	2.5	2.5	2
Erythromycin	+1	0.156	0.156	0.156	N.D.	1
Polyketides
Tetracycline	−1	0.312	0.312	0.312	N.D.	1
Oxytetracycline	0	0.625	0.625	0.625	N.D.	1
Aminoglycosides
Gentamicin	+5	5	0.078	5	5	64
Apramycin	+5	20	0.62	20	20	32
Streptomycin	+3	20	1.25	20	20	16
Kanamycin	+4	10	1.25	10	10	8
Other
Spectinomycin	+1	2.5	1.25	2.5	2.5	2
Chloramphenicol	0	10	10	10	N.D.	1

^a Charge at pH 7, DrugBank (89).

^b N.D., not determined.

^c Human peptide LL-37 fragment (18–37).

Δlysx cells exhibit altered membrane permeability

Because lysX was found to be associated with resistance to many antibiotics, the membrane integrity of ΔlysX cells was assessed using the BacLight LIVE/DEAD Bacterial Viability assay (Invitrogen) and compared to that of the wild-type strain. This assay uses SYTO 9 and PI dyes, which both exhibit a net positive charge. When bound to cellular nucleic acids, these dyes fluoresce green and red, respectively (48). SYTO 9 is membrane-permeable and stains all bacterial cells, while PI only penetrates those exhibiting a compromised membrane. PI permeation displaces SYTO 9 from its binding sites and is often used as an indicator for dead cells. C. pseudotuberculosis strains were stained and analyzed by flow cytometry and compared to dead (heat-treated) bacterial cultures as a control (Fig. 5A). A cytometry gate for bacterial cells was set using forward and side scattering parameters and used for all samples (Fig. S3).

FIG 5.

Permeability and membrane potential of C. pseudotuberculosis wild-type and ΔlysX cells. (A) Density plots of C. pseudotuberculosis stained with SYTO 9 and PI. Dashed lines demarcate the positions of dead cells used as controls. The data were gated as shown in Fig. S3. Major and minor populations of cells can be distinguished according to PI permeability. (B) Histogram of PI fluorescence of the cells shown in A. A shift in the PI fluorescence was observed for the ΔlysX population of cells. (C) The size of the subpopulation exhibiting partial PI permeability (data shown in A, upper right quadrant) is represented by gray bars. The median PI fluorescence of the major cell population (data shown in B, lower right quadrant) is represented by red bars. Histograms with error bars represent means ± SD’s and were derived from three independent experiments (n = 3). * Indicates a significant difference with the wt strain, P < 0.0003. (D) Dot plot of C. pseudotuberculosis stained with the membrane potential-sensitive dye DIOC₂(3). Green (FITC-A) versus red fluorescence (Keima-Red-A) of DiOC₂(3) was measured in control bacteria (orange dots) and bacteria depolarized with CCCP (blue dots). The data were gated as shown in Fig. S4. A decrease in the red signal after addition of CCCP demonstrates the presence of a membrane potential in each of the strains tested (49).

Although deletion of lysX negatively affected the generation time of C. pseudotuberculosis (Fig. S2), it did not alter cell viability. Figure 5A shows that both wild-type and ΔlysX strains exhibited a low level of dead cells (upper left quadrant) and two predominant subpopulations of cells with different levels of PI fluorescence (right quadrants). The major population of cells (lower right quadrant) exhibited a lower PI fluorescence signal, characteristic of live bacteria impermeable to PI. The minor population in the upper right quadrant exhibited the same SYTO 9 signal as the major population but a PI signal that was intermediate between that of PI-impermeable cells (bottom right quadrant) and control dead cells (top left quadrant), suggesting that this population exhibits partial permeability to PI. This minor population was three times larger in the ΔlysX strain (33.2 ± 4.7%) than in wild-type cells (11.0 ± 2.1%, P = 0.0003), demonstrating that lysX deletion increases overall permeability to PI (Fig. 5C). PI fluorescence within the major population of cells also shifted upward along the y-axis for the ΔlysX strain (2.3-fold higher than wild-type; Fig. 5B and C, P < 0.0001), demonstrating that a loss of lysX increases permeability of the membrane to PI within this subpopulation as well.

Cell populations exhibiting partial permeability to PI have been reported before with various bacterial species cultured in standard or challenge conditions with antimicrobials (50, 51). These populations are often referred to as injured, compromised, or damaged (50, 52). To investigate whether the ΔlysX strain exhibits notable membrane damage, which could explain altered membrane permeability and sensitivity to antibiotics, membrane functionality as an impermeable barrier was qualitatively assessed using the membrane potential-sensitive dye DiOC₂(3) (Fig. 5D; Fig. S4). This dye emits green fluorescence in the monomeric state when adsorbed on the surface of cells and red fluorescence in the aggregated state after translocation inside the cell in a membrane potential-dependent manner. When cells are exposed to the proton ionophore CCCP, the proton gradient dissipates, and the membrane potential decreases, which in turn decreases DiOC₂(3) import and the red fluorescence signal (49). Figure 5D (see Fig. S4 for details) shows that in the absence of CCCP, C. pseudotuberculosis wild-type and ΔlysX cells exhibit a similar red fluorescence signal indicative of a membrane potential. These findings suggest that, although lysX deletion increased permeation of PI across the membrane in a subpopulation of cells, the bacterial membrane retained its capacity to maintain a proton motive force and therefore did not show evidence of substantial mechanical damage.

lysX increases C. pseudotuberculosis virulence in a G. mellonella infection model.

To determine whether lysX and alaDAGS play a role in C. pseudotuberculosis virulence, G. mellonella larvae were inoculated with the lysX- and alaDAGS-deficient strains and compared to a control group inoculated with the wild-type strain. The median survival time was determined for each group (Fig. 6). Larvae infected with the wild-type strain exhibited a median survival time of 69 ± 8 h (n = 4). The survival time of larvae infected with ΔlysX cells was increased by 75% (121 ± 1.3 h, P = 0.006, n = 4), while no significant difference was observed with larvae inoculated with the ΔalaDAGS mutant. These experiments support the hypothesis that lysX is a virulence factor in C. pseudotuberculosis in a G. mellonella infection model.

FIG 6.

G. mellonella survival after infection with C. pseudotuberculosis. (A) Kaplan-Meier survival curves of infected larvae. Larvae were infected with 2.5 × 10⁵ CFU of C. pseudotuberculosis strains (wild-type, mutant, or complementation) as described in Materials and Methods. (B) Median survival time of larvae infected with C. pseudotuberculosis strains. Histograms with error bars represent means ± SD’s derived from four independent experiments (n = 4). *Indicates a significant difference compared to the wt strain, P = 0.006.

DISCUSSION

Lys-DAG was reported in 1988 in Mycobacterium phlei, and its biosynthesis and biological significance have since remained unknown (34, 35). Here, we show that LysX from C. pseudotuberculosis and two mycobacterial species synthesizes Lys-DAG and not Lys-PG. Lys-DAG confers resistance to C. pseudotuberculosis against many positively charged antimicrobials, and resistance is proportional to the number of charges born by the antimicrobial. We show that Lys-DAG is a bacterial virulence factor (in C. pseudotuberculosis) using an insect infection model. Lys-DAG and Ala-DAG are abundant amine-containing lipids in C. pseudotuberculosis that is readily detected by TLC. LC-MS analysis also revealed low levels of Lys-PG in vivo that were not detectable by TLC analysis. The synthesis of Lys-DAG and Lys-PG both appear to be dependent on lysX. Heterologous expression of lysX in C. glutamicum did not yield any detectable Lys-PG, but only Lys-DAG. Reconstitution of LysX activity in vitro and determination of the reaction product with TLC or LC-MS only revealed the synthesis of Lys-DAG. This lipid specificity for DAG was also demonstrated for LysX orthologs originating from two Mycobacterium species. LC-MS analysis of lipid extracts from M. abscessus only revealed the presence of Lys-DAGs, and no Lys-PG was detected. Altogether, these results suggest that Lys-PG in C. pseudotuberculosis may be synthesized through a downstream pathway dependent on LysX, like the synthesis of Ala-PG in C. glutamicum (13). This would involve transferase activity to translocate Lys from Lys-DAG to PG. The presence or expression of this downstream pathway may be species-specific or may be induced under certain culture conditions. These possibilities would explain the presence of Lys-PG in M. turberculosis and its apparent dependence on lysX (20).

Similar to the role of Lys-PG in other bacterial species [for review, see references (15 – 18)], the synthesis of Lys-DAG provided C. pseudotuberculosis with increased resistance against various positively charged antimicrobial peptides and aminoglycosides. ΔlysX markedly impacted MICs, with a large effect of up to >80-fold. MIC breakpoints for antibiotics to treat Corynebacterium infection are scarce (53, 54) compared to those available for other more common pathogens (55). However, the susceptibility breakpoint for gentamicin reported for Corynebacterium spp. is 4 µg/mL (53). C. pseudotuberculosis wild-type and ΔlysX strains exhibiting MICs of 5 and 0.075 µg/mL, respectively, show that lysX alone could potentially affect the treatment outcomes of a C. pseudotuberculosis infection, which reaffirms the relevance of lipid aminoacylation systems as an important mechanism of antibiotic resistance. On the other hand, the synthesis of Ala-DAG does not affect growth, antibiotic resistance, or virulence, which corroborate earlier observations in C. glutamicum for this lipid (13). The biological significance of Ala-DAG remains to be uncovered.

Flow cytometry data presented here suggest that the wide-spectrum antibiotic sensitivity of the ΔlysX strain is the result of an alteration of the membrane permeability to cationic compounds rather than a membrane disruption. CAMPs and aminoglycosides must first ionically associate with the cytoplasmic membrane to produce their effects (56, 57), and it is likely that Lys-DAG decreases the binding of cationic compounds and their permeation through the membrane directly by electrostatic repulsion, as has been proposed for Lys-PG in other bacteria (2, 58 – 60). In addition, aminoacylated PG is known to establish salt bridges with neighboring phosphate groups, which stabilize membranes (61 – 63). Interestingly, C. pseudotuberculosis exhibits two populations of cells with different membrane permeabilities to PI. The size of the population of cells exhibiting high PI permeability is increased in the ΔlysX strain. This increased PI permeability is not the result of a compromised cell membrane because the membrane still supports membrane potential. PI permeability has been used for decades as a marker for dead cells or for cells exhibiting extensive membrane damage (64). However, various reports suggest that some bacterial species exhibit a transient and natural permeability to PI during their cell cycle or during some specific and reversible metabolic state (65, 66). The reason for this staining pattern is unknown, but microscopy and flow cytometry evidence in Corynebacterium and Mycobacterium species suggests that this phenomenon may be the result of a normal asymmetric (hyphae-like) cell division process, yielding one of the daughter cells with a transiently higher permeability to PI (65, 67). Together with the fact that the ΔlysX strain exhibits an increased generation time, it is possible that lysX may play a role in cell division in C. pseudotuberculosis. This hypothesis is supported by additional observations regarding the role of lipid lysylation in other bacteria. In M. tuberculosis, lysX deletion disturbs cell division (29). The septal localization of MprF (synthesizing Lys-PG) during division is also evident in several bacterial species (58, 68). In Enterococcus faecalis, septal localization of MprF protects negatively charged microdomains hosting the SecA/Sortase A secretory system against CAMPs during cell division (58). It has been proposed that the lack of an MprF homolog in Streptococcus pyogenes is responsible for its sensitivity against polymyxin B, which targets the negatively charged microdomain hosting the exportal secretory system adjacent to the septum (69). Among the CAMPs tested, polymyxin B had one of the most pronounced effects on the C. pseudotuberculosis ΔlysX mutant. It is possible that Lys-DAG may play a role similar to that of Lys-PG in other bacteria, modulating membrane affinity for CAMPs and protecting specific foci on the cell surface.

MATERIALS AND METHODS

Bacterial strains and growth conditions

C. pseudotuberculosis ATCC 19410 was obtained from the American Type Culture Collection and grown in BHI (BD Biosciences) at 37°C with aeration by shaking. Alternatively, cells were grown on plates of the same media containing 1.5% (wt/vol) agar and supplemented with 25 mg L⁻¹ of kanamycin, as necessary. M. abscessus was grown at 37°C on solid LB or in liquid BHI or Sauton media. Liquid media were supplemented with 10% (vol/vol) Oleic Albumin Dextrose Catalase (OADC), prepared as previously described (70).

C. pseudotuberculosis and M. abscessus competent cells and transformation

Electrocompetent C. pseudotuberculosis cells were prepared according to a protocol developed by Dorella et al. (71). Briefly, a starter culture, inoculated using a single colony isolated from BHI, was grown overnight at 37°C and subsequently used to inoculate 200 mL of BHI. Cells were harvested when the optical density at 600 nm (OD₆₀₀) reached 0.8 and washed four times with decreasing volumes (100, 80, 50, and 25 mL) of ice-cold 15% (vol/vol) glycerol. Then, they were resuspended to a final OD of 60 and stored as aliquots at –80°C. Electroporation was achieved with 45 µL of competent cells and 0.5 µg of plasmid DNA in a 1 mm electroporation cuvette. Electroporation parameters were 1,900 V, 200 Ω, and 25 µF. Cells were immediately resuspended in 800 µL of BHI, incubated at 37°C for 2 h, and plated on selective BHI. Transformants were recovered after 48 h of growth at 37°C. C. glutamicum cells were transformed as previously described (13).

Electrocompetent cells of M. abscessus were prepared according to the protocol described by Viljoen et al. (45). Briefly, single colonies grown on LB solid medium were used to inoculate 5 mL of Sauton + OADC and 0.025% (vol/vol) tyloxapol. Cells were grown for 6 days at 37°C with agitation until the OD₆₀₀ reached 5.0. Then, 1 mL fractions of the bacterial suspension were stored at −80°C. Frozen aliquots were used to inoculate 200 mL of BHI + OADC and grown for 23 h at 37°C with constant agitation (200 rpm) until the OD reached 0.8. Cultures were chilled on ice for 20 min, and cells were washed with decreasing volumes (50, 25, 25, and 5 mL) of ice-cold 10% (vol/vol) glycerol. Bacteria were resuspended in 2 mL of the wash solution (OD was typically around 70) and stored at −80°C in 200 µL of aliquots. For transformation, aliquots were thawed in the presence of 0.5 µg of plasmid DNA. Electroporation was achieved in 2 mm cuvettes at 1,900 V, 200 Ω, and 25 µF. Bacteria were recovered in 1 mL of Sauton + OADC for 4 h at 37°C with agitation and plated on selective LB agar. Transformants were collected after 7 days of incubation at 37°C.

General cloning strategy and expression in C. glutamicum

Bacterial strains, plasmids, primer sequences, and gene accession numbers are reported in Tables S1 and S5. Cloning was achieved using the FastCloning strategy described in reference (72). PCR was performed using Phusion High-Fidelity DNA Polymerase (ThermoFisher). Plasmid templates were removed by digestion with DpnI, and PCR products of inserts and plasmids were transformed into competent cells (DH5α) prepared using rubidium chloride as previously described (73). C. pseudotuberculosis genes were cloned into the expression plasmid pEKEx2-GFP (13, 74, 75). lysX from M. parafinicum was codon optimized for expression in C. glutamicum and cloned into pEKEx-2 (Genscript). Expression of C. glutamicum transformants was achieved by incubation overnight at 30°C in THB (BD Biosciences).

In-frame deletion of C. pseudotuberculosis

In-frame and markerless deletion of C. pseudotuberculosis genes was achieved using allelic exchange by homologous recombination in a classical two-step approach with the vector pK19mobsacB (13, 76). Briefly, the upstream and downstream sequences (1 kb each) flanking the regions targeted for deletion were PCR amplified using C. pseudotuberculosis ATCC19410 genomic DNA as a template (Fig. S5) and stitched together by overlap PCR using primers shown in Table S1 and Fig. S5. The resulting cassettes were cloned into the suicide vector pK19mobsacB. The C. pseudotuberculosis genes lysX and aladags were inactivated using the vectors pK19ΔlysX and pK19Δaladags, respectively, as previously described (13). Briefly, plasmid constructs were transformed into C. pseudotuberculosis ATCC19410 by electroporation, and recombinant cells were selected on BHI supplemented with 25 mg L⁻¹ of kanamycin. Plasmid insertion at the expected loci was confirmed by PCR using flanking primers (Fig. S5). To obtain markerless in-frame deletions, positive colonies were passaged three times in BHI without antibiotics, before being plated on BHI containing 10% (wt/vol) sucrose to counter-select bacteria harboring the recombinant plasmid with the counter-selectable marker SacB. Sucrose-resistant, kanamycin-sensitive clones, having lost the pK19mobsacB plasmid and containing the expected deletion, were confirmed by PCR.

LysX disruption in M. abscessus

The suicide vector pVT (nearly identical to pUX1) (45) and the plasmid pVT-lysX (used for the disruption of lysX in M. abscessus) were built using the primers shown in Table S1 and according to the map shown in Fig. S5. Upstream and downstream flanking regions of lysX were cloned into pVT to generate an in-frame, markerless deletion of the gene in M. abscessus as described above. The gene was inactivated after a single crossover event, leading to the insertion of the plasmid into the chromosome as described in reference (45). Positive transformants were selected on LB supplemented with 75 mg L⁻¹ of kanamycin and displayed a strong red fluorescence signal due to expression of the protein tdTomato (45). Plasmid insertion at the expected locus was verified by PCR using primers flanking the targeted region (primers 1212 and 1213; Fig. S5).

TLC analysis of lipids from Corynebacterium and M. abscessus

Lipids were prepared from biomass after the growth of C. glutamicum and C. pseudotuberculosis in 200 mL of THB overnight at 30°C or 37°C, respectively. Lipids were prepared using the Bligh-Dyer procedure (77), with modifications as described (78). For M. abscessus, bacteria were grown in 200 mL of LB at 37°C for 3 days (until the OD reached 3.0) and washed in 30 mM Tris-HCl, pH 8.0. Lipids were prepared according to methods previously described (79, 80), with the following modifications. Bacteria were resuspended in 2.5 mL of methanol in a glass tube and mixed with 5 mL of chloroform. The suspension was mixed by inversion and placed at 4°C overnight. The organic phase beneath the cellular debris was collected and dried. All lipid preparations (from Corynebacterium or Mycobacterium) were resuspended in chloroform:methanol (2:1), and any insoluble material was removed by centrifugation. Lipids were analyzed by TLC using a solvent system consisting of chloroform:methanol:water (14:6:1). Total lipids were visualized using iodine vapors and amine-containing lipids with ninhydrin.

tRNA-dependent lipid aminoacylation assay

tRNA-dependent lipid aminoacylation was measured according to established methods (81). Activity was assayed using enzymes present in a membrane extract isolated from C. pseudotuberculosis as previously described (81). Aminoacylation reactions contained 200 mM Hepes•NaOH, pH 7.2, 60 mM KCl, 20 mM MgCl₂, 4 mM ATP, 2 mg/mL total tRNA (from E. coli MRE600; Roche Applied Science), 20 µM ¹⁴C-L-Lys (75 Ci/mol; Perkin Elmer), 100 nM lysyl-tRNA synthetase from E. coli, 0.25 mg/mL membrane extract (as determined by Bradford assay; Biorad), and a mixture of lipids consisting of 2 mg/mL egg-PG and 0.5 mg/mL of DAG (Avanti Polar Lipids). Prior to adding to the reaction mixture, egg-PG and DAG were mixed, dried, and emulsified by low-power sonication in a buffer containing 100 mM Hepes, 30 mM KCl, and 10 mM MgCl₂. After incubation of the complete reaction mixture at 37°C, lipids were extracted using the Bligh and Dyer procedure (22), dried, and resuspended in 20 µL of chloroform:methanol (2:1). Reaction products were mixed with total lipids extracted from C. pseudotuberculosis and spotted on an HLF silica gel TLC plate (Analtech). TLCs were developed in chloroform:methanol:water (14:6:1), and [¹⁴C]-lysylated lipids were visualized by phosphorimaging. For MS analysis, [¹⁴C]-Lys was substituted with deuterated 4,4,5,5-D₄ L-Lys (ThermoFisher). After incubation of the reaction mixture at 37°C for 1 h, lipids were extracted in the presence of 120 mM ammonium acetate, dried, and resuspended in chloroform:methanol (2:1) before MS analysis.

Analysis of lipids by mass spectrometry

LC-MS/MS of total lipid extracts was performed on a liquid chromatograph Surveyor Plus system (Thermo Scientific) with the autosampler connected to an LTQ-Velos linear ion trap analyzer mounted with a heated electrospray ionization probe, HESI-II (Thermo Scientific). Lipid samples extracted from C. pseudotuberculosis were injected onto an Ascentis Express C18 HPLC column (10 cm × 2.1 mm, 2.7 µm; Sigma-Aldrich) at a temperature of 45°C as described in reference (82), with minor modifications. Elution was performed at a flow rate of 260 µL/min with a binary gradient, where A was 60:40 water:acetonitrile and B was 90:10 isopropanol:acetonitrile. Both solutions contained 0.1% formic acid and 10 mM aqueous ammonium formate. Elution was performed over 19 min with the following gradient conditions: for the first minute, B was maintained at 32%; from 1 to 2 min, B was increased to 62%; from 2 to 9 min, B was increased to 75%; and from 9 to 16 min, B was increased to 100%. Solution B was then maintained at 100% for 3 min. The instrument was tuned with egg phosphatidylglycerol in solvent B (Avanti Polar Lipids). The drying gas flow rate was 20 U, and the temperature of the ESI was 350°C. Full-scan spectra were collected in the 110–2000 m/z range (set to positive mode), and data-dependent collision-induced dissociation tandem MS (MS2) spectra were acquired for the 15 most intense MS1 peaks in the 300–1,900 m/z range. Lipids, labeled in vitro with deuterated lysine, were injected directly into the electrospray interface after lipid extraction. Chromatograms were computed using the ADAP algorithm (83) and deconvoluted using the local minimum search method with MZmine 2.53 software (84). Lipids were identified using a customized version of the MS-DIAL LipidBlast library (85).

Determination of MICs and generation time

MICs were measured in a 96-well plate using the two-fold serial microdilution method (86). Briefly, C. pseudotuberculosis was cultured in Mueller Hinton Broth (MHB) with 0.1% (vol/vol) Tween 80. Wells contained 150 µL of medium inoculated with 28,000 CFU of starter cells and 5 µL of inhibitory compounds (or water as blank). Growth (OD at 600 nm) was recorded in real time in an ELx808 BioTek microplate reader with constant shaking at 37°C. The MIC was defined as the lowest drug concentration inhibiting growth after 24 h of incubation. The reported MIC values represent the means of at least three independent determinations. The specific maximal growth rate (µmax) was determined using the log-linear model of the R package growthrates (87, 88), with data from at least 25 growth curves carried out in the absence of antibiotics. Generation times were calculated using the relationship Ln2/µmax.

Flow cytometry

Overnight cultures of C. pseudotuberculosis were grown at 37°C in BHI supplemented with 0.1% (vol/vol) Tween 80. Strains were re-inoculated in BHI (without Tween) at a final OD₆₀₀ of 0.1 and cultured at 37°C. At an OD₆₀₀ of 0.6, cells were washed and resuspended in Dulbecco’s phosphate-buffered saline (Thermo Fisher Scientific) at a final OD₆₀₀ of 0.16. For the BacLight LIVE/DEAD Bacterial Viability assay (Invitrogen), cells were stained using 3.3 and 20 µM of Syto 9 and PI, respectively. Dead cells (for controls) were obtained by incubating bacteria at 80°C for 10 min. For the BacLight Bacterial Membrane Potential assay (Invitrogen), 3,3′-diethyloxacarbocyanine iodide [DIOC₂(3)] was added to a final concentration of 30 µM, and uncoupling was achieved using 5 µM of the protonophore m-chlorophenylhydrazone (CCCP). Treated cells were incubated for 1 h in the dark at room temperature before counting. Cells were analyzed in a CytoFLEX S Flow Cytometer (Beckman Coulter), and the data were analyzed using the software CytExpert 2.3 (Beckman Coulter). DIOC₂(3) excitation was carried out at 488 nm, and fluorescence emission was measured at 610 (Keima red filter) and 425 nm (FITC filter). Excitation/emission wavelengths for Syto 9 and PI were 488/525 (FITC filter) and 561/610 (ECD filter), respectively.

Galleria mellonella infection

G. mellonella larvae were purchased from Timberline Fisheries (Marion, IL, USA). Overnight starters of C. pseudotuberculosis were grown overnight in BHI supplemented with 0.1% (vol/vol) Tween 80. Strains were re-inoculated at a final OD₆₀₀ of 0.1 in BHI without the addition of Tween and subsequently grown to an OD₆₀₀ of 1.5. Bacteria were washed and resuspended in a saline solution containing 150 mM NaCl. Groups of 12 larvae weighing between 250 and 350 mg were infected with C. pseudotuberculosis by injection in the right foremost leg of the insect with 10 µL of saline solution containing 2.5 × 10⁵ CFU. The larvae were incubated at 37°C, and their appearance and behavior were recorded using a webcam. Survival status was checked periodically, and larvae were counted as dead when they changed color to dark brown and stopped moving. Death was confirmed by unresponsiveness to gentle manipulation. The median survival time was derived from a Kaplan-Meier survival plot using GraphPad Prism 9.1 (GraphPad Software, San Diego, CA, USA). This value was calculated as the minimum time at which 50% of the population of infected worms had died.

Statistical analysis

Means and SDs were calculated from at least three replicates, as indicated. Means were compared using a one-way analysis of variance test with Bonferroni’s correction using GraphPad Prism 9.1. P-values below 0.05 were considered significant.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/spectrum.01429-23.

Table S1-S2, Figure S1-S5. spectrum.01429-23-s0001.pdf.

Supplemental Tables and Figures.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.