A cluster of residues in the lipopolysaccharide exporter that selects substrate variants for transport to the outer membrane

SUMMARY

Most Gram-negative bacteria assemble lipopolysaccharides (LPS) on their surface to form a permeability barrier against many antimicrobials. LPS is synthesized at the inner membrane and then transported to the outer leaflet of the outer membrane. Although the overall LPS structure is conserved, LPS molecules can differ in composition at the species and strain level. Some bacteria also regulate when to modify phosphates on LPS at the inner membrane in order to become resistant to cationic antimicrobial peptides. The multi-protein Lpt trans-envelope machine, which transports LPS from the inner to the outer membrane, must therefore handle a variety of substrates. The most poorly understood step in LPS transport is how the ATP-binding cassette LptB₂FG transporter extracts LPS from the inner membrane. Here, we define residue K34 in LptG as a site within the structural cavity of the Escherichia coli LptB₂FG transporter that interacts electrostatically with phosphates on unmodified LPS. Alterations to this residue cause transport defects that are suppressed by the activation of the BasSR two-component signaling system, which results in modifications to the LPS phosphates. We also show this residue is part of a larger site in LptG that differentially contributes to the transport of unmodified and modified LPS.

Graphical Abstract

ABBREVIATED SUMMARY

Gram-negative bacteria are naturally resistant to many antimicrobials because they cover their surface with the glycolipid LPS. It remains unknown how newly synthesized LPS is extracted from the cytoplasmic membrane by an ATP-binding cassette transporter so that it can be routed to the cell surface by the trans-envelope Lpt machine. This study identifies a substrate-binding domain in a membrane component of the Escherichia coli Lpt transporter that is crucial for LPS extraction.

INTRODUCTION

The impermeability of Gram-negative bacteria to small hydrophobic molecules poses a great challenge in the development of antimicrobials (Nikaido, 2003; Richter & Hergenrother, 2018). This impermeability results from the composition of the Gram-negative cell envelope, which contains two membranes, an aqueous cellular compartment termed the periplasm, and a peptidoglycan cell wall (Fig. 1A) (Silhavy, Kahne, & Walker, 2010). The inner membrane (IM) surrounds the cytoplasm and is a typical phospholipid bilayer that is permeable to hydrophobic compounds. The outer membrane (OM), in contrast, contains phospholipids in its inner leaflet and the glycolipid lipopolysaccharide (LPS) in its outer leaflet (Henderson et al., 2016; Kamio & Nikaido, 1976). The structure of LPS drives tight packing of LPS molecules, which decreases the fluidity of the OM and creates a hydrophilic, polyelectrolyte barrier against hydrophobic molecules at the cell surface (Nikaido, 2003; Silhavy et al., 2010). This barrier-like quality and the fact that LPS is essential in relevant Gram-negative pathogens make targeting LPS biogenesis a plausible strategy to treat Gram-negative infections (Zhang, Meredith, & Kahne, 2013).

Figure 1: PEZ model of LPS transport by the Lpt system.

A) LPS is extracted from the IM by the LptB₂FG transporter. LPS transverses the periplasm by sheltering its acyl chains in the periplasmic bridge formed by LptCAD, followed by translocation to the cell surface through the LptDE translocon. B) Structure of E. coli LPS with relevant possible modifications colored in gold (L-Ara4N on the 4′ phosphate) and red (PEtN on the 1 phosphate). Kdo = 3-deoxy-D-manno-oct-2-ulosonic acid, Hep = heptose, Gal = galactose, Glu = glucose.

LPS is a large amphipathic molecule composed of up to three regions: lipid A, an acylated glucosamine disaccharide that forms the outer leaflet of the OM; core oligosaccharide, composed of diverse sugars; and O antigen, a polysaccharide made of a repetitive oligosaccharide unit (Fig. 1B) (Raetz & Whitfield, 2002). Despite these conserved features, LPS structure differs among Gram-negative bacteria with respect to the amount and type of acyl chains on lipid A, the presence of chemical modifications on the glucosamines of lipid A, the identity and arrangement of sugars in the core oligosaccharide and O antigen, and the absence of O antigen (Raetz, Reynolds, Trent, & Bishop, 2007). In some bacterial species, these structural variations can be regulated by environmental signals. For example, positively charged L-4-aminoarabinose (L-Ara4N) and phospho-ethanolamine (PEtN) can be added to phosphates on lipid A (Fig. 1B) to increase resistance to antimicrobial cationic peptides by decreasing the binding of these compounds to LPS (Raetz et al., 2007; Trent, Ribeiro, Lin, Cotter, & Raetz, 2001).

Once synthesized and possibly modified at the IM, LPS molecules must be transported across the cell envelope to the OM by the Lpt (Lipopolysaccharide transport) proteins (Okuda, Sherman, Silhavy, Ruiz, & Kahne, 2016; Sperandeo, Martorana, & Polissi, 2017). LptAB₂CDEFG form a bridge that spans all compartments of the cell, allowing the coupling of energy, from ATP in the cytoplasm, to the transport of LPS across the envelope (Bos, Tefsen, Geurtsen, & Tommassen, 2004; Braun & Silhavy, 2002; Okuda, Freinkman, & Kahne, 2012; Okuda et al., 2016; Ruiz, Gronenberg, Kahne, & Silhavy, 2008; Sampson, Misra, & Benson, 1989; Sherman et al., 2014; Sherman et al., 2018; Sperandeo et al., 2007; Sperandeo et al., 2008; Sperandeo, Pozzi, Deho, & Polissi, 2006; Wu et al., 2006). In the PEZ model for LPS transport (Fig. 1A), LPS is initially extracted from the IM by the ATP-binding cassette (ABC) transporter LptB₂FG (Dong, Zhang, Tang, Paterson, & Dong, 2017; Luo et al., 2017; Narita & Tokuda, 2009; Okuda et al., 2012; Okuda et al., 2016; Sherman et al., 2018). Subsequently, LPS traverses the aqueous periplasm with its acyl chains sheltered inside the hydrophobic groove of the Lpt domains present in LptCAD, with each new molecule of LPS that is extracted from the IM pushing a stream of LPS molecules along this groove (Okuda et al., 2012; Okuda et al., 2016). Ultimately, LPS molecules in this stream reach the β-barrel of the LptDE translocon, which inserts them in the OM (Chng, Ruiz, Chimalakonda, Silhavy, & Kahne, 2010; Dong et al., 2014; Freinkman, Chng, & Kahne, 2011; Gu et al., 2015; Okuda et al., 2016). At present, it is unknown how the Lpt machinery transports different LPS variants. Furthermore, it is yet to be determined if post-synthesis, the Lpt system plays a role in regulating the type of LPS displayed at the cell surface in organisms that synthesize various forms of the glycolipid.

The most poorly understood step in LPS transport is its energetically unfavorable extraction from the IM by LptB₂FG. LptF and LptG, the transmembrane subunits of this ABC transporter, are thought to mediate this extraction and subsequent loading of LPS onto the periplasmic Lpt bridge (Dong et al., 2017; Luo et al., 2017; Narita & Tokuda, 2009; Okuda et al., 2012; Okuda et al., 2016; Sherman et al., 2018). In support of this model, recent crystal structures of the LptB₂FG complex have revealed that LptF and LptG form a cavity within the IM that might accommodate LPS during its extraction (Dong et al., 2017; Luo et al., 2017). In this study, we have identified a cluster of residues localized at the membrane-periplasm interface of this cavity that is critical for the transport of LPS. Based on data derived from structure-function and suppressor analyses, and in vitro biochemical reconstitution of LPS extraction in liposomes, we propose that a residue within this domain interacts with unmodified LPS through electrostatic interactions. Further, we propose this domain participates in an early step in the recognition of LPS prior to its extraction from the IM by the LptB₂FG complex. Our data also suggest that the E. coli LptB₂FG transporter binds to different LPS substrates in distinct manners.

RESULTS

Correlation between charges in lipid A and LptG suggests an LPS-interaction site in LptG.

Previous work by Hamad et al. identified a residue in LptG of Burkholderia cenocepacia that might interact with LPS (Hamad, Di Lorenzo, Molinaro, & Valvano, 2012). In bacteria such as E. coli K-12 and Salmonella, modification of the phosphates at the C1 and C4′ positions of lipid A with L-Ara4N and PEtN is regulated and usually present only at low levels under standard laboratory growth conditions (Raetz et al., 2007). In contrast, in B. cenocepacia, the L-AraN modification is constitutive and essential to viability, and its loss results in phenotypes resembling those reported in Escherichia coli cells lacking Lpt function (Ortega et al., 2007). Notably, Hamad et al. showed that substituting aspartate at position 31 (D31) in LptG of B. cenocepacia with histidine resulted in an LptG variant (LptG^D31H) that suppressed the lethality caused by the absence of L-Ara4N (Hamad et al., 2012). This suggested that in wild-type, residue D31 in LptG might interact with the L-Ara4N moiety in LPS, and that its replacement with histidine might facilitate the transport of unmodified LPS in an L-Ara4N-deficient mutant. In support of this idea, structures of the LptB₂FG transporter of P. aeruginosa and K. pneumoniae show that the equivalent residues in these bacteria are in a cavity formed by LptF and LptG that is thought to accommodate LPS during extraction from the IM (Fig. 2A) (Dong et al., 2017; Luo et al., 2017). In fact, the placement of this residue near the periplasmic interface of helix 1 of LptG could allow an interaction between this residue and moieties at the C1 and C4′ positions of lipid A.

Figure 2: Residues in the K34 region of LptG are important for LPS transport.

A) Location of relevant LptG residues in the LptB₂FG crystal structure from P. aeruginosa (PDB = 5×5y). B) Close-up of the relevant residues (labelled by their identities in E. coli) shown in panel A. C) Substitutions in the K34 region of LptG cause defects in LPS transport, as determined by enhanced sensitivity to bacitracin (Bacitracin^S) using a disc-diffusion assay. Data represents the diameter of the zone of inhibition of growth (in mm); the black dotted line marks the diameter of the antibiotic discs. WT refers to strain NR2761. Substitutions in LptG are shown; KK/DD refers to the K40D/K41D double substitution. More details in Table S1. D) LPS release activity of LptG^WT, LptG^K34D, and LptG^K41D complexes. LPS release was monitored at various time points by UV-dependent crosslinking to LptAI36Bpa in the presence and absence of ATP. The immunoblot in the top panel was probed with LPS antiserum to detect crosslinked LptA-LPS adducts (marked LptA x LPS), whereas the immunoblot in the bottom panel was probed with LptA antiserum to show uncrosslinked LptA.

Surprisingly, the suppressor B. cenocepacia strain producing LptG^D31H is viable regardless of whether the phosphates in lipid A are modified with L-Ara4N or not, and residue D31 is not conserved among LptG homologs (Hamad et al., 2012). We wondered whether this lack of conservation is the result of co-evolution with differences in the lipid A structure, given the chemical groups at the C1 and C4′ positions of lipid A vary among bacteria. Indeed, we found a correlation between the charges of moieties at the C1 and C4′ positions of the main type of lipid A produced by a diverse group of Gram negatives and the residue in LptG equivalent to D31 in B. cenocepacia (Fig. 3). Often, both groups on lipid A bear the opposite charge of that present in the relevant LptG residue, and the three positions never carry the same charge. This correlation could explain the results obtained by Hamad et al. in B. cenocepacia (Hamad et al., 2012). The lethality caused by the loss of L-Ara4N could result from a charge clash between one of the unmodified phosphates at the C1 and C4′ positions in lipid A and residue D31 in LptG; this clash between negative charges would be relieved in the suppressor strain producing LptG^D31H (Hamad et al., 2012). Thus, the correlation between the charges at this residue of LptG and the C1 and C4′ positions of lipid A suggests that these specific positions may interact and, as a result, have co-evolved differently among Gram negatives to facilitate interaction between the LptFG cavity and the different types of LPS they must transport.

Figure 3: Correlation between residue in the cavity of LptG and LPS structure.

A) Alignment of LptG proteins from various Gram-negative organisms. Alignment was performed using Clustal Omega (Li et al., 2015; McWilliam et al., 2013; Sievers et al., 2011) and the following Uniprot accession numbers: C. jejuni = Q0PBJ6, H. pylori = E1S717, D. acetophilus = D4H8C5, P. gingivalis = B2RKU5, G. forsettii = A0LYY5, C. cresentus = A0A0H3C7C4, F. tularensis = A0A0B6KiX2, A. baumannii = D0CB66, P. aeruginosa = Q9HXH5, V. cholerae = A0A0H3Q758, E. coli = P0ADC6, H. influenzae = P45332, B. cenocepacia = U1XMW3, N. meningitidis = A0A0G4BVQ8, S. azorense = C1DTN4, G. uraniireducens = A5G7X0, E. minutum = B2KAY2. Residues examined in this study are colored in the same scheme as figures in the main text. B) Correlation between a specific LptG residue equivalent to D31 in Burkholderia and moieties at the C1 and C4′ positions of lipid A. Residue equivalency was determined by protein alignment (panel A). Moieties on lipid A are based on previously published data and numbers in parentheses denote position on lipid A (C1 or C4′) (Cullen et al., 2011; De Soyza, Ellis, Khan, Corris, & Demarco de Hormaeche, 2004; Hankins et al., 2011; Korneev et al., 2015; Moran et al., 1991; Pelletier et al., 2013; Phillips, Schilling, McLendon, Apicella, & Gibson, 2004; Pupo, Hamstra, Meiring, & van der Ley, 2014; Rangarajan et al., 2017). In bacteria where LptG has a negatively (red text) charged amino acid at the relevant site, the C1 and/or C4′ positions in lipid A are frequently modified with positively (blue text) charged groups; in those in which the relevant amino acid is positively (blue text) charged, the C1 and C4′ positions in lipid A contain a negatively (red text) charged phosphate; and, in bacteria with an uncharged (black text) amino acid, the chemical groups at the C1 and C4′ positions in lipid A vary. Furthermore, in bacteria producing lipid A that has the same type of charge at both C1 and C4′ positions, we usually found the opposite, and never the same, charge in the relevant LptG residue. L-Ara4N = L-4-aminoarabinose.

The putative LPS-binding site in LptG is essential for LPS transport.

We next tested in E. coli the idea that the aforementioned sites in LptG and lipid A might interact. Notably, in E. coli, the relevant charges in LptG and lipid A are reversed with respect to those in B. cenocepacia. In E. coli K-12, LptG has a positively charged lysine residue (K34) at the position equivalent to D31 in LptG of B. cenocepacia, and the phosphates in lipid A are unmodified under normal growth conditions (Fig. 3). If charge clashes between these positions in LptG and lipid A were detrimental to LPS extraction from the IM, substituting residue K34 in LptG of E. coli with a negatively charged residue should result in phenotypes characteristic of Lpt defects. Because LPS transport to the OM is essential for viability in E. coli, and proper transport is required for innate resistance to hydrophobic antibiotics, such defects could result in any of the following phenotypes according to decreasing severity: lethality under all growth conditions; conditional lethality by which mutants can only survive in slow-growing conditions such as minimal medium; and increased sensitivity to hydrophobic antibiotics like bacitracin (Yao, Davis, Kishony, Kahne, & Ruiz, 2012).

Accordingly, we introduced mutations that substituted the positively charged K34 residue in LptG with negatively charged aspartate (K34D) or glutamate (K34E) residues. We generated plasmid-encoded lptFG alleles and tested their ability to complement a chromosomal deletion of lptFG (Simpson et al., 2016). Haploid strains generated from complementing lptG mutant alleles were also tested for defects in LPS transport by assessing their sensitivity to hydrophobic antibiotics. We found that LptG^K34E was stably produced, but not functional since lptG(K34E) did not complement a chromosomal deletion of lptFG (Fig. 2B, S1). This might explain why an LptG^K34E/R136E variant had been reported as non-functional (Dong et al., 2017). In contrast, lptG(K34D) complemented, but the resulting lptG(K34D) haploid strain was very sensitive to hydrophobic antibiotics (Figs. 2 and S1, and Table S1). These defects suggested that the negative charges on the side chain of the substituted amino acid and LPS might clash, and that this clash might occur more efficiently when position K34 is changed to a glutamate relative to a one-carbon shorter aspartate. To test if these severe defects in LPS transport were limited to changes that reverse the charge, we substituted LptG residue K34 with others that maintain the positive charge (K34R) or neutralize it and are either nonpolar (K34C and K34A) or polar (K34Q). Alleles encoding these substitutions complemented, and while the lptG(K34R) allele behaved similarly to wild-type lptG, the remaining alleles increased sensitivity to hydrophobic antibiotics, although not as severely as lptG(K34D) (Fig. 2C and Table S1). These results therefore suggest that the charge of residue K34 in LptG is important for LPS transport in E. coli.

We further explored the effect of the charge at position 34 of LptG in E. coli by exploiting the ability to modify LptG^K34C with cysteine-reactive small molecules that form adducts of different charges since there are no other free periplasmic cysteines in Lpt factors. Compounds 2-sulfonatoethyl methanethiosulfonate (MTSES) and 2-(Trimethylammonium)ethyl methanethiosulfonate (MTSET) cross the OM and react with free sulfhydryl groups in the periplasm to form structurally similar cysteine adducts that are either negatively or positively charged, respectively (Fig. 4A) (Butler, Davis, Bari, Nicholson, & Ruiz, 2013; Zhu & Casey, 2007). Since substituting K34 in LptG with cysteine conferred mild defects (Fig. 2C, Table S1), we thought that we could test the effect of adding differently charged adducts at this position using MTSES and MTSET, provided that the internal cavity formed by LptFG is accessible to the periplasmic environment. If a positive charge at position 34 is required for proper LptG function, we predicted we might be able to further decrease LPS transport, and thereby increase the OM permeability, in lptG(K34C) cells by treating them with negatively charged MTSES but not with positively charged MTSET. To test these predictions, we monitored the growth of wild-type and haploid lptG(K34C) cells in the presence of a sub-inhibitory concentration of the lysis-inducing hydrophobic antibiotic bacitracin, which cannot efficiently cross the wild-type OM, but can enter cells with defects in the Lpt system. We then continued to monitor their growth after the addition of either MTSES or MTSET. Under the conditions tested, the growth of wild-type cells was not significantly affected by the addition of either MTSES or MTSET (Fig 4B). However, as predicted, the addition of MTSES, but not that of MTSET, increased the sensitivity of lptG(K34C) cells to bacitracin, as reflected by the induction of lysis 1.5 h after treatment (Fig. 4B). These results further demonstrated that the positive-charge character of K34 is crucial for its function. In toto, the severe phenotypes observed by reversing the charge of K34 in LptG of E. coli, the previous data on its equivalent residue in B. cenocepacia (Hamad et al., 2012), and the placement of this residue within the internal cavity formed by LptFG support a model in which this site contacts opposite charges in lipid A during its extraction from the IM.

Figure 4: Charge requirement in cavity-facing residues in LptG for proper LPS transport.

A) Structures of MTSES, MTSET and their products after they react with free cysteine thiols. B) Shown is sensitivity of various residues in the K34 region of LptG to modification by a negatively charged moiety (MTSES) or positively charged moiety (MTSET) when grown in the presence of sub-inhibitory concentrations of bacitracin. Defects in LPS transport increase sensitivity to bacitracin, which results in a drop in culture turbidity (OD₆₀₀) as a result of lysis. Arrows indicate time of MTSET/MTSES addition. Refer to Experimental Procedures for details

The periplasmic end of transmembrane domain 1 in LptG constitutes a functional domain in LPS transport.

Having identified the positive charge of K34 in LptG as important for LPS transport, possibly by mediating an interaction with the phosphates of lipid A, we set out to investigate if other charged or polar residues in the same region of LptG are also important for function. The LptB₂FG structure (Dong et al., 2017; Luo et al., 2017) revealed two additional positively charged residues (K40 and K41) in helix 1 of LptG that are positioned two turns away from K34, and a negatively charged (D37) and a polar (Q38) residue one turn away from K34 (Fig. 2A-B). We thus generated LptG variants with substitutions at these sites and assessed their function and protein levels.

We found that alleles encoding an aspartate substitution at either K40 or K41 of LptG readily complemented and, unlike lptG(K34D), only caused a mild increase in OM permeability (Fig. 2, Table S1). We then tested for possible functional redundancy between these residues and K34 by constructing double and triple alleles. The lptG(K40D/K41D) allele yielded a haploid strain. This was not surprising, given that an lptG(K40E/K41E) allele has been reported to support growth of an LptFG-depletion strain under depleting conditions (Dong et al., 2017). However, we observed that the lptG(K40D/K41D) haploid strain was commensurably more sensitive to hydrophobic antibiotics than its parental single mutants (Fig. 2C, Table S1). Further introducing either a K34D or K34A change into the lptG(K40D/K41D) allele resulted in triple mutant alleles that were unable to complement the loss of chromosomal lptG, despite causing no notable decrease in protein levels (Fig. S1). This synthetic lethality and the proximal location of these residues within LptG suggest these residues function together in LPS transport.

To further evaluate the functional role of the positively charged residues in this domain, we prepared proteoliposomes containing LPS and LptB₂CFG complexes containing wild-type LptG (LptG^WT), LptG^K34D or LptG^K41D, and examined their ability to extract LPS. To monitor LPS extraction from these liposomes, we added purified LptA-I36pBPA-His₆, an LptA variant containing the UV-crosslinkable amino acid p-benzoyl phenylalanine (pBPA) (Okuda et al., 2012; Sherman et al., 2018). In this assay, LPS molecules that are extracted by LptB₂CFG complexes and transferred to LptA can be covalently trapped in LptA upon UV crosslinking. The resulting LptA-LPS adducts can be detected by a mass shift after immunoblotting for LPS. As previously reported, complexes transferred LPS to LptA in a time- and ATP-dependent manner (Fig. 2D) (Okuda et al., 2012; Sherman et al., 2018). Consistent with the sensitivity data discussed above, the ability to release LPS to LptA was reduced for both variant-containing complexes relative to the LptG^WT complexes, with LptG^K34D complexes releasing less LPS than those containing LptG^K41D (Fig. 2D).

We also examined the negative and polar residues (D37 and Q38) of this domain, and found that alleles encoding either alanine or cysteine substitutions at D37 or Q38 also complemented the deletion of chromosomal lptG, but the resulting haploid strains exhibited OM-permeability defects (Fig. 2C, Table S1). As before, we also took advantage of the ability to covalently modify cysteine residues with charged sulfhydryl-reactive reagents to investigate the effect of introducing negative and positive charges at D37 and Q38. Notably, for lptG(D37C) and lptG(Q38C) cells, we had to use a fourth of the concentration of bacitracin than that used for wild-type and lptG(K34C) strains, and observed effects on growth in a shorter time scale. We do not currently understand if this time difference is related to differences in the accessibility of these residues to the chemical probes or, more likely, to the fact that the lptG(K34C) mutant is less permeable to bacitracin than the lptG(D37C) and lptG(Q38C) mutants (Fig. 2C). Nevertheless, our results revealed that reversing the charge of D37 or adding a charge at Q38 conferred sensitivity (Fig. 4B), indicating that the negative charge of D37 and the uncharged polar nature of Q38 are important for LptG function. Together, these results show that charged and polar residues at the membrane-periplasm interface of helix 1 in LptG play a crucial role in LPS transport.

Suppressor analysis suggests that residue K34 in LptG interacts with the phosphates of lipid A.

To further investigate the role of K34 in LptG, we selected for suppressors of OM-permeability defects of a haploid lptG(K34D) strain. One suppressor mutation that increased resistance to several hydrophobic antibiotics (Table S1) generated the basS(L102Q) allele, which changes amino acid 102 in BasS from leucine to glutamine. BasS is the histidine kinase of the BasSR two-component system, which controls chemical modifications of LPS (Fig. 5A) (Raetz et al., 2007). Most notably, BasSR upregulates transcription of the arn and eptA operons, which encode the enzymes that modify the phosphates at C1 and C4′ in lipid A with the positively charged moieties L-Ara4N and PEtN, respectively (Figs. 1B, 5A) (Gibbons, Kalb, Cotter, & Raetz, 2005; Raetz et al., 2007; Trent et al., 2001).

Figure 5: BasSR-dependent activation of modification of LPS structure by EptA and ArnT suppresses LPS transport defects in lptG(K34D) mutants.

A) Model of activation of LPS modification by the BasSR system. Activation of the BasS kinase, results in phosphorylation of the BasR response regulator. BasR activates transcription of the eptA and arn operons, which encode enzymes that modify lipid A with PEtN and L-AraN, respectively. B) Polymyxin resistance conferred by basS^c as determined by measuring its minimal inhibitory concentration (MIC) for various strains. C) Bacitracin sensitivity (in mm) of the lptG(K34D) basS^c mutant in the presence and absence of eptA and arnT, as assessed by disc diffusion assay. Black line marks the width of the antibiotic discs. More details in Table S2.

We wondered if basS(L102Q) suppressed lptG(K34D) by constitutively activating the BasSR system, as this would lead to increased modification of the lipid A phosphates with positively charged moieties. First, to determine if the basS(L102Q) allele activates the BasSR system, we took advantage of the fact that modification of LPS with positively charged moieties results in resistance to the cationic antimicrobial peptide polymyxin B (Raetz et al., 2007). As we predicted, lptG⁺ and lptG(K34D) strains carrying basS(L102Q) were more resistant to polymyxin B than their respective basS⁺ parents (Fig. 5B), indicating that basS(L102Q) is a constitutive allele (basS^c). We next confirmed that this basS^c allele upregulates modifications in lipid A using MALDI-TOF mass spectrometry (MS) analysis of lipid A isolated from the wild-type strain, and from lptG(K34D) mutants carrying either the basS⁺ or basS^c alleles. MS analysis of the wild-type basS⁺ sample revealed primarily the expected unmodified hexa-acylated 1,4′-bis-phosphate lipid A (Fig. S2A) (Raetz & Whitfield, 2002; Rubin, Herrera, Crofts, & Trent, 2015). In contrast, the lptG(K34D) basS⁺ sample contained equal amounts of unmodified hexa-acylated 1,4′-bis-phosphate lipid A and hepta-acylated 1,4′-bis-phosphate lipid A (Fig. S2B). The presence of hepta-acylated LPS was expected since deficient LPS transport causes the mislocalization of phospholipids to the outer leaflet of the OM, which induces the enzymatic activity of the OM protein PagP to generate hepta-acylated LPS by transferring an acyl chain from a phospholipid to hexa-acylated LPS (Bishop et al., 2000; Ruiz et al., 2008). In contrast, the lptG(K34D) basS^c samples contained predominantly hexa-acylated lipid A modified with a single PEtN group and a smaller portion of hexa-acylated lipid A modified with two PEtN groups (Fig. S2C). We also detected a small amount (ca. 1%) of LPS modified with L-Ara4N (Fig. S2). The increase in PEtN modification was expected from the upregulation of eptA, while the reduction in the levels of hepta-acylated lipid A was expected from its ability to suppress LPS transport defects conferred by lptG(K34D).

We next tested whether suppression of lptG(K34D) by basS^c is mediated by the modification of the lipid A phosphates. Elimination of the PEtN modification by deleting eptA in the lptG(K34D) basS^c strain abolished most of the suppression (Fig. 5C, Table S2). Although we only detected approximately 1% of LPS modified with L-AraN in the MS analysis of the suppressor strain (Fig. S2), the residual suppression observed in the lptG(K34D) basS^c ΔeptA strain was eliminated by deleting arnT, which is needed for this modification (Fig. 5C, Table S2) (Raetz et al., 2007; Trent et al., 2001). The relative effect of EptA and ArnT on suppression is consistent with the results of the MS analysis of lipid A showing that the basS^c allele strongly induces the modification of lipid A with PEtN (Figs. 5A and S2C).

We next demonstrated that the basS^c suppressor allele did not change LPS levels (Fig. S1D) and, is not a general suppressor of Lpt defects. Specifically, basS^c cannot suppress other lpt alleles such as lptD4213, which alters the OM translocon (Braun & Silhavy, 2002), and lptF(E84A) and lptG(E88A), which alter the helices in LptFG that couple with LptB (Simpson et al., 2016) (Fig. 6A, Table S3). Taken together, our results and those from Hamad et al. indicate that electrostatic interactions between residue 34 in LptG of E. coli (residue 31 in Burkholderia) and moieties at the C1 and C4′ positions in lipid A are critical for LPS extraction from the IM.

Figure 6: Activation of LPS modification by basS^c specifically suppresses defects in the K34-K41 domain of LptG.

A) The basS^c allele does not suppress defects in the coupling helix of LptF [lptF(E84A) allele] and LptG [lptG(E88A) allele], or in the OM translocon LptD (lptD4213) as determined by bacitracin sensitivity (in mm) of the wild type and lpt mutants carrying either wild-type basS⁺ (grey/black) or basS^c (dark green). The black line represents the width of the antibiotic discs. B) OM-permeability defects of LptG variants with changes in the K34-K41 domain are suppressed by the basS^c allele. Shown is bacitracin sensitivity (in mm), as assessed by disc diffusion assay, of WT and LptG variants in basS⁺/basS^c backgrounds. Data for the basS^c strains (dark green) are superimposed on their basS⁺ counterparts. The black line represents the width of the antibiotic discs. More details in Tables S1 and S3.

Differential binding of modified and unmodified lipid A to the LptFG cavity.

Our results support the notion of an interaction between opposing charges at residue 34 of LptG and moieties at the C1 and C4′ positions in lipid A in both the wild-type and lptG(K34D) strains. However, we observed that the basS^c allele also suppressed LPS transport defects caused by changes of K34 in LptG to alanine, cysteine, and glutamine (Fig. 6B). Furthermore, the basS^c allele could also suppress defects conferred by changes in LptG residues D37, Q38, K40 and K41 (Fig. 6B, Table S1). Together, these results suggest that basS^c does not suppress by simply restoring a charge-charge interaction between residue 34 and lipid A. The simplest explanation of the results presented in this study is that residues K34, D37, Q38, K40, and K41 in LptG constitute at least a portion of a binding domain for LPS in the cavity formed by LptFG. Within this domain, residue K34 forms electrostatic interactions with unmodified phosphates on lipid A that are important for the extraction of LPS from the IM. In addition, the charges of residues D37, K40 and K41, and the polar character of residue Q38 are also important for the transport of unmodified LPS molecules. Lastly, our results also suggest that LPS molecules with positively charged modifications interact either with a different domain in the LptFG cavity or with the same domain in helix 1 of LptG but using a different binding mode to that of unmodified LPS.

DISCUSSION

LptF and LptG were identified a decade ago (Ruiz et al., 2008), but their role in LPS transport has remained largely obscure even after the recent elucidation of their structure (Dong et al., 2017; Luo et al., 2017). Our results provide the first evidence that in E. coli, the cavity formed by LptFG interacts with LPS to extract it from the IM. We have identified an LPS-binding domain in LptG containing a cluster of residues that differentially contributes to the binding of unmodified and modified LPS. This domain is located in the periplasmic end of helix 1 of LptG and contains a positively charged residue (K34) that interacts via electrostatic interactions with unmodified phosphates in LPS. Interestingly, residue K34 is not conserved among LptG homologs, likely because the structure of lipid A varies among bacteria.

How LPS is extracted from the IM remains unknown, but it has been proposed that it enters into a central cavity formed by LptFG and that, once inside, movement of the ATPase LptB in the cytoplasm is transduced to LptFG so that they undergo conformational changes that lead to LPS extraction towards the periplasmic Lpt domains of LptF and/or LptG, and, eventually, LptC (Fig. 1A) (Dong et al., 2017; Luo et al., 2017; Okuda et al., 2012; Okuda et al., 2016; Simpson et al., 2016). The domain we have identified to be critical for LPS transport is localized inside the LptFG cavity, right at the periplasmic interface of transmembrane 1 of LptG, where the hydrophilic portions of lipid A might be at the initial stages of extraction from the IM (Fig. 2A). Changing the positive charge of position K34 to a negative charge in this domain of LptG in E. coli, which normally produces bisphosphorylated lipid A, causes defects in LPS transport that can be ameliorated by modifying the phosphates on lipid A with positively charged moieties. These findings are in line with results previously described in Burkholderia (Hamad et al., 2012), and support a model stating that the charged side chain in this position of LptG interacts with the opposite charges at positions C1 and/or C4′ of lipid A molecules. Indeed, in Burkholderia, lipid A is constitutively modified with positively charged L-Ara4N and LptG has a negatively charged aspartate at the relevant position 31; inversely, in E. coli the phosphates in lipid A are normally unmodified and the relevant position in LptG, K34, is positively charged. However, this model alone cannot explain how, in wild-type E. coli, Lpt can transport both negatively and positively charged lipid A-containing LPS, since this bacterium can modify LPS with L-Ara4N and PEtN under certain conditions (Gibbons et al., 2005; Raetz et al., 2007; Trent et al., 2001). The aforementioned model predicts that such modifications would result in a clash between positive charges in lipid A and K34 in LptG that would be expected to result in LPS transport defects. Yet, this is not the case, even in strains that constitutively activate the BasSR-regulated modifications of lipid A (Fig. 6A). Moreover, altering the charge and polar nature of residues near K34 in LptG (D37, Q38, K40, and K41) also causes defects in Lpt function that can be suppressed by constitutively activating BasSR in the presence of native K34 in LptG (Fig. 6B).

To reconcile all these data, we propose that in wild-type E. coli: 1) residues K34, D37, Q38, K40 and K41 of LptG form at least a portion of a domain that mediates interactions between LPS and the cavity of LptFG; 2) this cluster of residues exhibits differential binding to unmodified LPS, with residue K34 mediating electrostatic interactions with a phosphate in lipid A that are crucial for the extraction of unmodified LPS from the IM; and, 3) modified LPS binds either to this domain in LptG in a different way to that of unmodified LPS, or contacts a yet-to-be-identified site in the LptFG cavity. According to this model, wild-type E. coli can transport both types of LPS because the LptFG cavity can interact with both forms of LPS, albeit differently. In contrast, in the lptG(K34D) mutant, unmodified LPS, which is the predominant form synthesized under normal growth conditions, clashes with the negatively charged aspartate, thereby causing severe defects in LPS transport. These defects can then be suppressed by constitutively activating the BasSR system because this activation shifts the type of LPS the mutant produces from negatively charged to mostly positively charged, PEtN-modified LPS. This shift in the type of LPS molecules that are synthesized favors a different mode of interaction of LPS with the LptFG cavity, thereby increasing LPS transport and suppressing OM permeability defects in the lptG(K34D) mutant.

Our findings, and the fact that these LptG residues are localized at the membrane-periplasm interface of the LptFG cavity, lead us to propose that this cluster of residues mediate interactions with LPS at an early step in the extraction process. As such, the identity of these residues could serve to dictate how different LPS variants interact with the LptFG cavity prior to extraction. Since modification of lipid A occurs before LPS engages with LptB₂FG, it is also possible that this cluster of residues in LptG could ultimately control the type and proportion of LPS being displayed at the cell surface. In fact, in Burkholderia, in which L-AraN modification of lipid A is essential and constitutive, LptFG may have evolved to only handle modified substrate (Hamad et al., 2012). This could imply that residue D31 of LptG in Burkholderia may only be able to directly interact with L-AraN-modified LPS through electrostatic interactions, much like residue K34 in LptG of E. coli interacts with the unmodified phosphates. Residue D31 of LptG in Burkholderia would also prevent transport of unmodified LPS by clashing electrostatically with the phosphates. Alternatively, D31 might only serve to discriminate against unmodified LPS while L-AraN-modified LPS interacts differently with the transporter, much like we propose modified LPS does in E. coli. Altogether, we propose that the cavity of LptFG and the structure of LPS have co-evolved in different bacteria to ensure efficient transport and possibly modulate the diversity of LPS at the OM.

EXPERIMENTAL PROCEDURES

Strains and growth conditions.

Strains (Table S5) were typically grown at 37°C in either lysogeny broth (LB) or M63 supplemented with 0.2% (wt/vol) glucose with aeration for liquid cultures and on 1.5% agar plates for solid media. When appropriate, media was supplemented with ampicillin (125 μg/mL), bacitracin (100 μg/mL), chloramphenicol (20 μg/mL), isopropyl-β-D-1-thiogalactopyranoside (0.16 mM), kanamycin (30 μg/mL), novobiocin (33 μg/mL), polymyxin B (2 μg/mL, 10 ug/mL), tetracycline (25 μg/mL), vancomycin (25, 75 μg/mL), and X-Gal (33 μg/mL).

Mutant construction.

lptG alleles were generated through site-directed mutagenesis PCR using the pBAD18LptFG3 plasmid (Simpson et al., 2016), or pCDFDuet.LptBFG (Okuda et al., 2012) as templates, PfuTurbo polymerase (Agilent Technologies, Inc.) or KOD Hot Start Polymerase (Novagen), and primers listed on Table S5. Complementation tests and construction of haploid strains bearing mutant lptG alleles was carried out using the segregation-defective pRC7LptFG plasmid system as described (Simpson et al., 2016). Mutant alleles of eptA, arnT, and basS were introduced into strains by P1_vir transduction either through linkage to nearby marker zjd2211::Tn10 (for basS^c) or by selecting for kanamycin resistance (for ΔeptA::kan and ΔarnT::kan). The kan marker was removed by pCP20-encoded FLP recombinase (Cherepanov & Wackernagel, 1995). Allele replacement was verified by PCR and the presence of the basS(L102Q) allele was confirmed by DNA sequencing.

Selection and mapping of suppressors.

A 100-μL aliquot of a culture of strain NR3673 [haploid lptG(K34D) strain] grown overnight in M63 with glucose was plated on LB solid medium supplemented with bacitracin (100 μg/mL). A single suppressor isolate was selected for mapping based on its robustly restored resistance to multiple antibiotics. The suppressor was mapped via P1_vir co-transduction using as donor a randomly generated mini-Tn10 library (Ruiz, Wu, Kahne, & Silhavy, 2006). Loss of suppression was monitored by the inability to grow in the presence of vancomycin (25 μg/mL). A mini-Tn10 transposon marker, referred to as tet2–2, ~25% linked to the suppressor allele was inserted at minute 94 of the E. coli chromosome. Sequencing of this region of the chromosome revealed a change in nucleotide 305 of basS from T to A, and thereby altering residue 102 of the BasS protein from L to Q. This allele was therefore named basS(L102Q).

Antibiotic sensitivity.

To assay OM permeability to bacitracin, novobiocin, rifampin, and erythromycin, 100 μL of overnight cultures of haploid strains were grown in M63 with glucose and subjected to disc diffusion assays as described previously (Sherman et al., 2014). MICs were determined by presence or absence of growth in liquid media in the presence of the compound at various concentrations in 96-well plates as described previously (Ruiz et al., 2006).

MTSES/MTSET sensitivity assays.

Strains were grown overnight in 5-mL cultures either in LB or M63 with glucose, and then diluted 1:500 or 1:250, respectively, in fresh LB containing sub-inhibitory concentrations of bacitracin (4 μg/mL for NR2761 and NR4077, 1 μg/mL for NR4518 and NR4232). These cultures were incubated at 37°C and their growth was monitored by measuring their optical density at 600 nm (OD₆₀₀) every 30 min for 2–3 h until an OD₆₀₀ of ~0.2 was reached, upon which 0.4 mM of MTSES or MTSET was added. After adding these cysteine-reactive reagents, we continued to monitor OD₆₀₀ every 30 min until a total elapsed time of 5 h.

Overexpression and purification of the Lpt inner membrane complex.

To overexpress His₆-LptBFGC, His₆-LptBFG-K34D-C, and His₆-LptBFG-K41D-C, KRX cells were transformed with pET22/42.LptC and either pCDFDuetHis₆LptBFG, pCDFDuet.LptBFG(K34D) or pCDFDuet.LptBFG(K41D). Overexpression and purification of each inner membrane complex variant was done as previously reported (Sherman et al., 2014). Cultures were grown at 37°C after diluting overnight cultures 1 to 100 into fresh LB broth supplemented with 50 μg/mL spectinomycin (Sigma) and 50 μg/mL carbenicillin (Teknova). Expression was induced with 0.02% L-rhamnose at OD₆₀₀ ~ 1, and cultures were grown for 16 h at 18°C. Cells were harvested by centrifugation at 4,200 × g for 20 min and resuspended in 50 mM Tris-HCl (pH 7.4), 300 mM NaCl, 2 mM MgCl_2, supplemented with 1 mM phenylmethanesulfonyl fluoride (PMSF, Sigma), 100 μg/mL lysozyme (Sigma), and 100 μg/mL DNase I (Sigma). The resuspended cells were passaged through an EmulsiFlex-C3 high-pressure cell disruptor three times. The cell lysate was centrifuged at 10,000 × g for 10 min to remove unbroken cells. Membranes were isolated by centrifugation at 100,000 × g for 1 h. Membranes were resuspended and solubilized in 20 mM Tris-HCl (pH 7.4), 300 mM NaCl, 5 mM MgCl₂, 10% (vol/vol) glycerol, 1% n-dodecyl-β-D-maltopyranoside (DDM, Anatrace), 2 mM ATP at 4 °C for 1 h, followed by centrifugation at 100,000 × g for 30 min. The supernatant was applied to Ni-NTA Superflow resin (Qiagen), and eluted with 20 mM Tris-HCl (pH 7.4), 300 mM NaCl, 10% (vol/vol) glycerol, 0.05% DDM, 100 mM imidazole. The eluate was concentrated with an Amicon centrifugation filter, 100-kDa molecular weight cutoff (MWCO, Amicon Ultra; Millipore), and then subjected to size exclusion chromatography on a Superdex 200 10/300 GL column (GE Healthcare) in 20 mM Tris-HCl (pH 7.4), 300 mM NaCl, 10% (vol/vol) glycerol, 0.05% DDM. Fractions were pooled and concentrated to ~5 mg/mL. All complexes were analyzed by SDS-PAGE to assess purity. Protein was flash frozen in liquid nitrogen and stored at −80°C until use.

Overexpression and purification of LptA.

LptA-I36pBPA-His₆ was overexpressed in the periplasm and purified by making spheroplasts, as previously reported (Okuda et al., 2012). BL21(λDE3) cells with pSup-BpaRS-6TRN and pET22b-LptA-I36pBPA-His6 were grown to OD600 ~ 0.5 in 500 mL LB broth supplemented with 50 μg/mL carbenicillin, 30 μg/mL chloramphenicol, and 0.8 mM pBPA. Protein expression was induced with 50 μM IPTG (Sigma) for 2 h at 37°C. The cells were harvested and converted to spheroplasts. The periplasmic fraction was incubated with Ni-NTA Superflow resin (Qiagen), washed with 20 mM Tris-HCl (pH 8.0), 300 mM NaCl, 20 mM imidazole, and then eluted with 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, 200 mM imidazole. The purified LptA-I36pBPA-His6 was concentrated with 10 kDa cut-off Amicon centrifugal concentrators (Millipore) to ~2.5 mg/mL, flash frozen with liquid nitrogen and kept at −80°C in 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, and 200 mM imidazole containing 10% glycerol.

IM complex proteoliposome preparation.

Inner membrane proteoliposomes were prepared as previously described (Sherman et al., 2018). Aliquots of E. coli polar lipid extract stock solution and LPS were thawed and sonicated briefly to homogenize. Proteoliposomes containing the Lpt inner membrane complex variants were prepared by a detergent dilution method (Brundage, Hendrick, Schiebel, Driessen, & Wickner, 1990). Prior to dilution, a mixture with the following final concentrations was prepared in 20 mM Tris-HCl (pH 8.0), 150 mM NaCl buffer: 7.5 mg/mL lipid stock, 0.5 mg/mL LPS, 0.25% DDM, and 0.86 μM purified inner membrane complex. To make this mixture, the DDM was first added to the lipid stock solution to make detergent-destabilized liposomes. LPS was added to this mixture, which was kept on ice for 10 min to allow for mixed detergent-phospholipid-LPS micelles to form. The protein complex was added, and the mixture was incubated on ice for 20 min. The mixture was then transferred to an ultracentrifuge tube and diluted 100× with cold 20 mM Tris-HCl (pH 8.0), 150 mM NaCl. After letting the diluted mixture sit on ice for 30 min, the proteoliposomes were pelleted by ultracentrifugation at 300,000 × g for 2 h at 4°C. The proteoliposomes were resuspended in 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, diluted 100× again, then pelleted by ultracentrifugation at 300,000 × g for 2 h at 4°C. For every 100 μL original mixture (prior to the first dilution step), 250 μL cold 20 mM Tris-HCl (pH 8.0), 150 mM NaCl, 10% glycerol was added. If the resuspended proteoliposomes were not used immediately, they were flash frozen in liquid nitrogen and stored at −80°C.

Reconstitution of LPS release to LptA.

The reconstitution of LPS release from proteoliposomes to LptA was conducted similarly to the ATPase assay described above. All assays were done in 50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 10% glycerol (final concentrations). Reactions contained 60% inner membrane proteoliposomes by volume (prepared as described above, thawed on ice). The remaining volume was composed of Tris-HCl, NaCl, and glycerol such that the final concentrations would be the above values. All assays were done in 50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 10% glycerol (final concentrations). Reactions contained 60% IM proteoliposomes by volume (prepared as described above, thawed on ice). The remaining volume was composed of Tris-HCl, NaCl, and glycerol such that the final concentrations were the above values. LptA-I36pBPA-His₆ was added to a final concentration of 2 μM prior to starting the assay. Assays were initiated at 30°C with the addition of ATP/MgCl₂ (final concentrations were 5 mM ATP, 2 mM MgCl₂). At specified time points, 25 μL aliquots were removed from the reactions and added to a microtiter plate, which was subsequently irradiated with UV light (365 nm) on ice for 3 min using a B-100AP lamp. Following UV-irradiation, samples were added to 225 μL cold 20 mM Tris-HCl (pH 8.0), 150 mM NaCl supplemented with 0.2% DDM. To each sample, 250 μL 20% TCA was added. The proteins were precipitated and gel samples prepared as described above. Samples were boiled and subject to SDS-PAGE followed by immunoblotting.

Preparation of lipid and LPS stock solutions.

For proteoliposome preparation, E. coli polar lipid extract (Avanti Polar Lipids, Inc.) was dissolved in water and sonicated for 30 minutes to make a 30 mg/ml aqueous suspension stock. The lipid stock solution was flash frozen in liquid nitrogen and stored at −80°C. A 2 mg/mL aqueous suspension stock of LPS from E. coli EH100 (Ra mutant, Sigma) was dissolved in water and sonicated for 30 minutes. The LPS stock solution was flash frozen in liquid nitrogen, and stored in aliquots at −80°C.

Isolation of lipid A from E. coli cells.

Cultures of each strain were grown in LB medium (200 mL) at 37°C. Cells were harvested when they reached OD ~ 1.0 and then washed with 25 mL phosphate buffered saline, pH 7.4. Lipid A was isolated from each sample using the Bligh-Dyer extraction method and acid hydrolysis (Henderson, O’Brien, Brodbelt, & Trent, 2013). Briefly, pellets were resuspended in 95 mL of a single-phase Bligh-Dyer mixture (chloroform:methanol:PBS (pH 7.4); 1.0:2.0:0.8 v/v). The mixture was allowed to sit at room temperature for 30 minutes for lysis. The sample was centrifuged at 2,000 x g for 20 minutes and then the pellet was washed with an additional 25 mL of single phase Bligh-Dyer mixture. After centrifugation, the pellet was suspended in 2.4 mL hydrolysis buffer (50 mL sodium acetate, pH 4.5, 1% SDS) and boiled for 30 minutes. Lipid A was harvested by extractions in two-phase Bligh-Dyer mixture (chloroform/methanol/water, 2:2:1.8 v/v). After extraction, the combined chloroform layers were dried by rotary evaporation and the sample was transferred using a 4:1 chloroform:methanol mixture. The dried sample was washed once with acidified ethanol (2% HCl in EtOH) and 2x with ethanol to ensure complete removal of SDS (Caroff, Tacken, & Szabo, 1988). Samples were then lyophilized and stored at −20°C.

Mass spectrometry of lipid A species.

Samples were analyzed using a Bruker Ultraflextreme MALDI-TOF/TOF mass spectrometer. An ATT matrix was used and it was prepared as follows. A saturated solution of 6-aza-2-thiothymine in 50% acetonitrile and a saturated solution of tribasic ammonium citrate were combined (20:1, v/v). Lipid A samples were dissolved in chloroform:methanol (4:1). The matrix (0.5 uL) was deposited on the sample plate and allowed to dry and then an equal volume of sample was added on top of the matrix. Each spectra represents the average of ~50 laser shots. Data were acquired in negative-ion mode with a reflectron analyzer and delayed extraction.

Immunoblotting.

Levels of LptG variants were assessed by immunoblotting with rabbit anti-LptG antisera and horseradish peroxidase-conjugated anti-rabbit antibody from goat (GE Healthcare Life Sciences) using whole cell lysates from overnight cultures that were normalized for cell density based on OD₆₀₀ readings, as described previously (Simpson et al., 2016). LPS levels were determined similarly, substituting anti-LPS antibody (Bio-Rad 4329–5004) for the primary antibody and horseradish peroxidase-conjugated anti-mouse antibody from sheep (GE Healthcare Life Sciences) for the secondary antibody. Quantitation of LPS levels was carried out with Imagelab 5.2.1 (Bio-Rad).