Advances in the Structural Biology, Mechanism, and Physiology of Cyclopropane Fatty Acid Modifications of Bacterial Membranes

SUMMARY

Cyclopropane fatty acid (CFA) synthase catalyzes a remarkable reaction. The cis double bonds of unsaturated fatty acyl chains of phospholipid bilayers are converted to cyclopropane rings by transfer of a methylene moiety from S-adenosyl-L-methionine (SAM). The substrates of this modification are functioning membrane bilayer phospholipids. Indeed, in Escherichia coli the great bulk of phospholipid synthesis occurs during exponential growth phase, but most cyclopropyl synthesis occurs in early stationary phase. In vitro the only active methylene group acceptor substrate is phospholipid bilayers containing unsaturated fatty acyl chains.

INTRODUCTION

This is an update of a review published in 1997 (1). This update is mainly focused on the enzymes that introduce cyclopropane rings into membrane phospholipids (Fig. 1 and 2), because these are very abundant in nature (>18,000 annotations in the NCBI Protein database including most bacteria and several plants) whereas the enzymes that introduce cyclopropane rings into the mycolic acids of mycobacteria are different and restricted to these bacteria, although some mycobacteria lack cyclopropane modification. Despite the 25 years that have elapsed, the 1997 review is still cited (26 citations in 2021), indicating the relevance of the review and its subject. That review covered the history of cyclopropane fatty acid (CFA) discovery, the regulation plus discovery of the E. coli cfa gene, and purification of the enzyme. Hence, these topics will not be revisited. However, in 1997 the E. coli CFA synthase had only recently been purified by a laborious procedure (this was prior to the advent of affinity tags) and crystal structures of the enzyme were not available. Recently, the first structures of CFA synthases that catalyze phospholipid modification have been reported (predated by several structures of mycolic acid cyclopropane modification enzymes). The first was that of E. coli (2), followed closely by that of Lactobacillus acidophilus (3), and most recently that of the extremophile Aquifex aeolicus (T. Lukk, J. E. Cronan, and S. K. Nair, submitted for publication). Despite originating from very diverse bacteria, the proteins have essentially the same structure (see below). This is a striking finding because A. aeolicus is thought to represent a very early stage in the development of metabolism (4), which argues that this postsynthetic modification of phospholipids arose very early in the development of metabolism. Another advancement is that the cfa genes of diverse bacteria have been inactivated, thereby blocking CFA synthesis. At the time of the prior review, that was true only of E. coli. To date, as first seen in E. coli, the loss of CFA synthesis in diverse bacteria does not block growth in the laboratory. However, there are now instances in bacterial pathogens where strains lacking cyclopropane modification are deficient in pathogenesis. Other phenotypes of CFA-lacking bacteria in response to stress have recently been uncovered, as have new aspects of cfa gene expression in E. coli.

FIG 1.

CFA synthase modification of phospholipid bilayers. The phospholipid shown is a phosphatidylethanolamine having an unsaturated acyl chain (palmitoleic acid) in the sn-2 position (the favored substrate of the E. coli CFA synthase). The SAM methyl group is colored blue with the “extra” H atom in red.

FIG 2.

Cartoon of the CFA synthase reaction with phospholipids. The dots represent the polar head groups, the lines represent acyl chains. and the triangles are cyclopropane rings. This cartoon is not fully accurate since the phospholipids generally contain only one or slightly more than one unsaturated acyl chain. Modified from a figure in the 2015 IGEM presentation of Santa Clara University (44) with permission.

CFA STRUCTURES

The crystal structures of the bacteria CFA synthases and those that form cyclopropane rings in the complex mycolic acid lipids of mycobacteria all show a bicarbonate ion in the active site (this was first seen in the mycobacterial enzymes) (5). The bicarbonate ion is essential for CFA synthase activity and the binding residues are conserved (see below). The essentiality has been demonstrated by inactivation of the E. coli CFA synthase by depletion of the bicarbonate ion (6) and by its replacement with borate ion. Borate ion mimics the planar trigonal structure of bicarbonate and is a competitive inhibitor (6). Mutagenesis of the E. coli CFA synthase residues that bind the bicarbonate ion (H266, Y317, and E239) inactivate the enzyme, although activity can be rescued by addition of 50 mM bicarbonate (7). Relative to the wild-type enzyme, bicarbonate addition increased the activities of the C139S and H266A mutant proteins to about 69% and 21%, respectively, whereas the completely inactive Y317F mutant protein was restored to an activity of about 12% of the wild-type protein. In these studies, two other mutant proteins, E239A and E239D, could not be rescued by bicarbonate addition at least at concentrations that did not inhibit the wild-type protein. Similar rescue experiments were done by others using a lower bicarbonate concentration (3 mM) with concomitantly lower degrees of activity rescue (6). However, these investigators reported very low (<1% of the wild type enzyme rate) bicarbonate rescue of the E239A and E239D proteins. The differences in the two studies can probably be attributed to the CFA synthase assays used.

The three reported phospholipid CFA synthase crystal structures are a 2.07 Å structure of the E coli enzyme (PDB 6BQC) (2), a 2.7 Å structure of the L. acidophilus enzyme (PDB 5Z9O) (3), and a 1.60 Å structure of the Aquifex aeolicus enzyme (PDB 7QOS) (Lukk et al., submitted). All three crystal structures were solved by molecular replacement using as the search models the crystal structures of mycolic acid cyclopropane synthases from M. tuberculosis. CmaA1 was the search model for the E. coli (2) and A. aeolicus CFA (Lukk et al., submitted) synthases, whereas the PcaA structure was used for L. acidophilus CFA synthase (3). It has been estimated that to obtain a meaningful structure solution by molecular replacement, the replacement model needs to cover at least 50% of the structure and the root mean square distance between the C_α of the search model and structure to be solved must be better than 2 Å (8). Thus, given their very disparate bacterial origins, it is remarkable that these enzymes have essentially the same structural fold and overall architectures that are sufficiently similar to achieve this demanding crystallographic task (Fig. 3).

FIG 3.

(A) Structural alignments of the CFA synthase crystal structures of L. acidophilus (PDB ID 5Z9O, cyan), E. coli (PDB ID 6BQC, purple), and A. aeolicus (PDB ID 7QOS, beige).The CFA synthase from A. aeolicus contains the full set of cofactors and a fatty acid ligand. (B) CFA synthase from A. aeolicus (beige) and mycolic acid synthases PcaA (PDB ID 1L1E, gold) and CmaA1 (PDB ID 1KPG, green) from M. tuberculosis aligned at 0.933 Å root mean square deviation across all atom pairs utilizing the MatchMaker function in UCSF Chimera (45). For clarity of the view of active site, the first 17 residues of the mycolic acid synthase were omitted from the figure.

The crystal structure of the CFA synthase from A. aeolicus represents two states in one crystal structure—the structure of a CFA synthase primed for methyl group transfer in one chain (chain B of 7QOS); and the structure of postmethylene group transfer in the other chain (chain A of 7QOS) (Lukk et al., submitted). This serendipitous occurrence is likely due to contamination of SAM with SAH since SAM is a notoriously unstable compound and SAH is a known breakdown product (9). Another possibility is that the bound SAM of chain A was used to modify a phospholipid. This unexpected bonus of two subunits crystallized at different stages of the reaction gives the most comprehensive view of the structural basis of the cyclopropanation reaction in a phospholipid CFA synthase (Fig. 4A). Although the E. coli and L. acidophilus structures contain a phospholipid occupying the λ-shaped binding pocket and a bicarbonate moiety, no detectable electron density for the SAM cofactor is present in either of these structures and the L. acidophilus workers were unable to bind SAM to their crystals in soaking and crystallization experiments, whereas the E. coli report did not address SAM binding (3).

FIG 4.

(A) The dimeric structure of A. aeolicus CFA synthase solved at 1.6 Å resolution (PDB ID 7QOS). In addition to the electron densities corresponding to the bicarbonate moiety and the λ-shaped fatty-acid ligand, the structure also contains S-adenosyl-homocysteine in chain A (green) and S-adenosyl-methionine in chain B (gold). (B) The electron density maps obtained for the A. aeolicus CFA synthase at 1.6 Å resolution are defined well enough to suggest the presence of a phosphoethanolamine containing lipid as the bound ligand in the active site. In the crystal structure, the sn-3 lipid position contains phosphoethanolamine, whereas sn-1 and sn-2 positions both contain a palmitate moiety. The corresponding maps calculated with Fourier coefficients (2F_obs-F_calc) are contoured at 1.2σ above background and curbed at a radius of 1.8 Å around the ligands. (C) Stereo view of the active site residues of A. aeolicus CFA synthase with bound bicarbonate ion, SAM, and a fatty acid ligand (FA).

No experimental evidence is provided for the identity of the ligand occupying the λ-shaped fatty acid binding pocket. However, the electron density in the A. aeolicus structure does suggest a phosphoethanolamine-containing phospholipid (Fig. 4B) similar to that described for the other two structures. The binding of the bicarbonate ion in the A. aeolicus CFA synthase is achieved through its coordination to His272, Glu245, the backbone amides of Cys137, and Gly242 and via hydrogen bonding to the p-hydroxyl group of Tyr323 (Fig. 4C). The positioned sn-2 fatty acyl chain is curved about the methylene group of SAM cofactor, which is positioned just 3.7 Å from the C10 atom of the fatty acid chain. The same C10 atom of the modeled palmitate chain is also positioned 3.8 Å from the oxygen atom of the bicarbonate moiety. Due to the rather heterogeneous electron density for the fatty acid in the sn-2 position, the model represents only an approximation of one of the possible conformations for that fatty acid chain. Nevertheless, the positioning of the C10 atom of the fatty acid relative to the bicarbonate oxygen and methylene group of SAM cofactor seem to visibly confirm the structural basis for the carbocation mechanism of cyclopropanation of this and other CFA synthases. Chain B of the A. aeolicus structure contains electron density of similar shape to SAM in the λ-shaped binding pocket, but there is no detectable electron density for a methyl group in the chain A; hence, SAH was modeled into the structure.

CFA synthases are a head to tail dimers, and each monomer has a small N-terminal domain connected to the larger C-terminal domain by a flexible linker of about 20 residues (Fig. 5). The roles of dimerization and of the two domains have been extensively studied in the E. coli enzyme (2). The dimer interface consists largely of antiparallel pairing between the terminal β sheet strands of the C-domain plus packing mediated by the preceding α-helices (2). Although the interface appears strong, loss of a single salt bridge between E308 and R361 by a E308Q mutation abolishes dimerization and CFA synthase activity (2). The two domains of the E. coli monomer interact strongly and remain associated following cleavage of the linker, although the split enzyme retained only 2% of the wild-type activity. A similar loss of activity resulted from duplication of the linker. Hence, the linker plays an essential role in enzymatic activity (2). Dimerization also plays an essential role: the E308Q monomeric protein is completely inactive (2). Loss of a basic patch on the N-terminus of the protein resulted in inefficient lipid binding, suggesting that the basic patch binds the negatively charged phospholipid substrates. Salt is known to release E. coli CFA synthase from substrate phospholipid vesicles (10), consistent with the proposed ionic interaction. The requirements for the linker and for dimerization argue that CFA synthases are highly dynamic enzymes.

FIG 5.

Active sites of A. aeolicus CFA synthase (PDB ID 7QOS). Chain A on the left (gold), containing the SAM cofactor with the bicarbonate moiety and methylene group of SAM primed for methyl group transfer. The C10 of atom of fatty acid at the sn-2 position of the modeled phospholipid is positioned at an almost equal distance to the methyl group and bicarbonate ion oxygen atom. Chain B on the right (green) represents the state of post methyl group transfer, with SAH modeled in as the postreaction cofactor. The corresponding maps calculated with Fourier coefficients (2F_obs-F_calc) are both contoured at 1.2σ above background and curbed at a radius of 1.8 Å.

MECHANISM OF THE CFA SYNTHASE REACTION

The 1997 review assigned a carbocation mechanism based on the 1969 proposal of Lederer for sterol C-alkylation (11). That mechanism (Fig. 6A) has not only stood the test of time but has been greatly strengthened by several lines of evidence. The competing mechanisms that involved a sulfur ylid or metal carbenoid formation have been eliminated. The lack of metals in pure E. coli CFA synthase (6, 12) precluded a metal-dependent mechanism, whereas the sulfur ylide mechanism was rendered unlikely by the stereochemistry of the reaction and the effects of substitutions of fluoride atoms for the double bond hydrogen atoms (13). The formation of a diverse set of products by a mutant enzyme in which the highly conserved G236 was converted to glutamate, the residue found in the Hma mycobacterial mycolic acid enzyme that produces chains containing hydroxymethyl groups, clearly demonstrates the carbanion mechanism. The G236E mutation resulted in sharply decreased CFA production in an E. coli Δcfa strain and production of phospholipids containing a slew of abnormal unsaturated fatty acids (13, 14). These acids contained methyl groups attached to the acyl chain, often to one of the carbon atoms of the double bond. Detected were 9-Me-16:1Δ10, 10-Me-16:1Δ9, 9-Me-16:1Δ9, either 9-Me-16:1Δ8 or10-Me-16:1Δ8, and perhaps 10-Me-16:1Δ10 plus a minor amount of CFA (12). Since these are the products predicted from the diverse resolution pathways of the carbanion intermediate, these data very strongly support the carbanion mechanism. The mechanism is also supported by studies of phospholipid substrate analogs (13).

FIG 6.

The carbocation mechanism and the role of bicarbonate ion. Panel A is the figure from the 1997 review recycled since it has stood the test of time (1). Panel B gives the proposed role of bicarbonate ion acting as a general base to form bicarbonate.

The most parsimonious role proposed for the bicarbonate ion is that the double bond attacks the SAM methyl group to form a carbocation intermediate (probably a protonated cyclopropane), which is then deprotonated by transfer of the “extra” proton to the bicarbonate ion, which acts as a general base (7) (Fig. 6B). The bicarbonate ion accepts the proton from the protonated cyclopropane and the cyclopropane ring forms. The fact that the bicarbonate ion can be removed and readily replaced argues against a structural role for this tiny ion (6, 7). The major argument against the bicarbonate ion acting as the base in the carbanion mechanism is the report that methyl transfer to the olefin substrate is rate limiting and thus the bicarbonate ion could play a secondary role perhaps in stabilization of the developing carbocation intermediate (15). However, a more recent report argues that methyl addition onto the double bond and the deprotonation step are both rate-limiting (13). In contrast to the earlier reports, these authors were unable to detect any deuterium incorporation into the product of reactions catalyzed by the E. coli CFA synthase in the presence of D₂O. These studies used ultra-resolution mass spectroscopy to obtain mass values to five decimal places, which allowed the natural abundance of ¹³C to be used as an internal standard (13). The first of the prior studies was done using in vivo labeling with deuterated methionine, raising the possibility of metabolic exchange of the methionine methyl group atoms (16). The second study used kinetic analyses of the effects of deuteration of the SAM methyl group, in which the data obtained were differences between two small values (15). The state-of-the-art mass spectroscopy data thereby supersede the prior reports and indicate that methyl transfer is not the sole rate-limiting step and shares that property with transfer of the “extra” proton to the bicarbonate ion.

The carbanion mechanism whereby the phospholipid unsaturated acyl chain is converted to a cyclopropane ring seems settled, and we must now turn to the considerably more difficult aspect of the reaction. How does CFA synthase recognize an unsaturated acyl chain in the phospholipid lipid bilayer, convert the double bond to a cyclopropane ring, and return the modified phospholipid to the membrane? Early work showed that E. coli CFA synthase binds only to phospholipid bilayers containing either unsaturated acyl chains or cyclopropane-modified acyl chains (9). The enzyme is unable to bind phospholipid bilayers that contain only saturated or terminally branched acyl chains. How does the protein recognize only phospholipids having unsaturated or cyclopropane acyl chains? Two properties that distinguish double bonds from saturated chains are the π orbitals and the 30° kink imparted by cis double bonds. A straightforward π orbital recognition mechanism would be to overlap the π orbitals of the double bond with the π orbitals of an aromatic amino acid side chain (Phe, Tyr, or Trp) in a coplanar arrangement. If π orbital overlap provides the needed recognition, we would expect to find an aromatic residue or residues close to and planar with the acyl chain double bond and the bicarbonate ion. The E. coli CFA synthase crystal structure, although of higher resolution than the L acidophilus CFA crystal structure, showed an active site containing a mixture of lipids presumably due to cocrystallization of protein molecules that had bound different phospholipid species. In contrast, the active site of L acidophilus CFA crystal structure seemed to contain only a single phospholipid species, although the low resolution obtained did not allow its definitive identification (3). The two groups each modeled a phosphatidylethanolamine molecule into the active site and obtained very similar models in which the molecules have a structure resembling the Greek lower-case lambda (λ) where the top of the λ is the polar head group and the arms are the acyl chains. As expected, the acyl chains are bound by hydrophobic residues. The L acidophilus report presented a detailed model that includes the specific hydrophobic residues. The active site was identified by the bicarbonate ion (attempts to obtain a SAM complex were unsuccessful). Strikingly, the residues adjacent to the bicarbonate ion and modeled acyl chain are largely aromatic (Tyr and Phe), whereas the other hydrophobic residues thought to contact the acyl chain are essentially all aliphatic. If this notion reflects reality, the π orbital overlap may provide double bond recognition. The release of the cyclopropane product may result from weaker π orbital overlap or the different shapes of the π orbitals (cyclopropane rings have aromatic character and thus π orbitals).

Recognition of the 30° kink due to the cis double bond of unsaturated acyl chains provides another plausible mechanism. Conversion to cis cyclopropane rings preserves the kink (17). Hence, as previously reported (10), phospholipids containing either species would be bound by CFA synthase. An argument against the kink hypothesis is the M. tuberculosis CmaA2 (cyclopropane mycolic acid synthase A2) enzyme, which introduces both cis and trans cyclopropane rings into mycolic acids (18). Fatty acids having double bonds of the trans conformation have only a tiny kink and are essentially a linear alkyl chain. It would seem difficult for an enzyme to recognize opposite double bond isomers. Note that E. coli CFA synthase demonstrates a strong specificity in vivo for unsaturated phospholipid acyl chains having the double bond located between 9 and 11 carbons from the carboxyl terminus (19, 20). To understand recognition in detail will require a high-resolution crystal structure containing a defined phospholipid species with an appropriate unsaturated acyl chain. It should be possible to prepare CFA synthase molecules that contain only a single phospholipid species. E. coli strains defective in unsaturated fatty acid synthesis grow when provided with non-native fatty acids that are incorporated into phospholipids but that are not CFA synthase substrates (19–21). When CFA synthase expression is carried out in such a system, the synthase molecules should lack phospholipids. Following purification, the protein could be exposed to phospholipid bilayers composed of a single pure phospholipid species and crystallized.

A plausible working model for the cyclopropane modification reaction is that the N-terminal basic patch identified in the E. coli CFA synthase work (2) binds the negatively charged headgroups of phospholipid substrates and this binding rotates the phospholipid molecule out of the bilayer and into the CFA synthase active site. This would involve breaking the hydrophobic interactions between the bilayer acyl chains, but the CFA synthase active site is also hydrophobic, so the overall energetic cost would seem minimal. Following modification, the phospholipid molecule would be returned to the bilayer presumably by movements of the protein structure. A precedent for such a process is found in the methylation enzymes that modify double stranded DNAs (22). The methylase enters the hydrophobic milieu of the base pairs and disrupts the base stacking interactions and the hydrogen bonds such that the base flips into the methylase active site for modification. Several methylase-DNA complexes with flipped-out bases have been observed by crystallography (22).

PHYSIOLOGICAL ROLES OF CFA MODIFIED PHOSPHOLIPIDS

As discussed in the prior review, discovery of the physiological role of CFA synthesis was elusive (1). Upon comparing wild-type E. coli K-12 strains with their Δcfa derivatives, the results were dependent on the strain background. The first robust phenotype originated from a clue in the food science literature in which the resistance of five diverse wild type E. coli strains to a rapid and drastic decrease in pH was tested and correlated with the level of CFA present in the cell membrane phospholipids of the strains (23). Those strains having high levels of CFA survived acid shock much more efficiently than strains with low CFA levels. Subsequently, this clue was tested in E. coli K-12 strains by measuring the relative viabilities of wild-type strains with their Δcfa derivatives upon acid shock from pH 7 to pH 3 (24). The Δcfa strains showed viabilities several orders of magnitude lower than those of the wild-type parental strains. Similar data were reported for S. enterica serovar Typhimurium (hereafter S. enterica) (25, 26). The E. coli results were shown due to the presence of CFAs in the phospholipids, because when Δcfa strains blocked in unsaturated fatty acid synthesis were fed exogenous CFAs, the ability to survive acid shock largely returned to wild-type levels (24). Moreover, it was recently reported that strains lacking the RydC or AarS small RNAs (discussed below) were also unusually sensitive to acid shock (lack of CpxQ had no effect).

The major role of CFAs is to decrease the permeability of phospholipid bilayers. In the case of acid shock, CFA in phospholipids were shown to decrease proton permeability and increase proton extrusion relative to phospholipids lacking CFA (27). This is consistent with a detailed molecular dynamics analysis of phospholipid bilayers containing either unsaturated acyl chains or cyclopropane acyl chains, which concluded that CFAs stabilize the membranes by increased ordering of the chains without sacrificing membrane fluidity (28).

ROLES OF CYCLOPROPANE MODIFICATIONS IN BACTERIAL PATHOGENESIS

Mycobacterium tuberculosis

The first report of reduced pathogenesis due to lack of a single cyclopropane modification is the loss of the PcaA enzyme of Mycobacterium tuberculosis (29). The gene was identified by an inability to form colonies with serpentine cords, a morphology called cording. The mycolic acids of this bacterium are composed of a number of molecular species. PcaA is responsible for introduction of the proximal cis cyclopropane ring on the longer chain of the α-mycolic acids that comprise at least 70% of the mycolic acids of the organism and contain two cyclopropane rings. Mycolic acids are composed of a longer β-hydroxy chain with a shorter α-alkyl side chain. These acids contain between 60 and 90 carbon atoms (the number of carbons varies with species). M. tuberculosis expresses several other cyclopropane modification enzymes, CmaA1, MmaA2, and CmaA2, which like PcaA contain an active site bicarbonate ion (Uma1 may also be a member of this group) (5). Note that CmaA1 synthesizes a trans cyclopropane ring whereas MmaA2 is responsible for the distal modification of the longer chain of the α-mycolic acids (18). PcaA is very specific for its substrate, whereas the others show wider specificity. There also is another homologous group of enzymes, MmA1, MmA3, and MmA4, that make methyl olefins, methyl ethers, and methyl alcohols and have acidic residues that replace the bicarbonate ion (5). All together M. tuberculosis encodes about 10 CFA synthase-like enzymes (6). In vitro biochemistry of these enzymes with their native substrates seems nearly impossible given their extremely insoluble substrates. However, some long chain unsaturated fatty acids can substitute, but these are very poor substrates (30). Low levels of methylene transfer to a possible intermediate in mycolic acid synthesis has also been reported (31).

M. tuberculosis pcaA null mutant strains replicated well in mice but were eliminated from the animals much more rapidly than the wild-type strain (29). A striking and unequivocal result was that all of the mice infected with the wild-type strain died, whereas all of those infected with the ΔpcaA strain survived. The difference seemed due to less lung damage in the ΔpcaA strain-infected animals (29). Note that although null mutants in the various mycolic acid cyclopropane modification genes are not lethal in the laboratory, combinations of null mutations seemed synthetically lethal. However, more recently, an M. tuberculosis strain lacking all cyclopropane modifications was constructed and found to be viable but severely defective in pathogenesis (32). The main property of cyclopropane modifications in M. tuberculosis seems to be avoidance of the host immune system (33, 34).

Bacteria Having Phospholipid Membranes

There are two reported examples of reduced pathogenesis in bacteria lacking the ability to form CFA in their membrane phospholipids. The first and more striking of these is the gastric pathogen Helicobacter pylori (35). E. coli and S. enterica strains lacking CFA are very sensitive to acid shock, and therefore it seemed likely that this would also be the case in H. pylori since the bacterium is found in the strongly acidic stomach environment. Indeed, H. pylori Δcfa strains showed 2 to 4 orders of magnitude lower viability upon acid shock, with the lower pH values giving the lower viability. This was true for both a laboratory strain and a clinical isolate (35). Moreover, the H. pylori Δcfa strains colonized mice much less efficiently (>4 orders of magnitude) than the parental wild-type strains (35). The lack of CFA significantly increased the susceptibility to the antimicrobial agents used to treat H. pylori infections. Cell membrane integrity was also compromised in the Δcfa strains, as shown by greater uptake of the hydrophobic fluorescent probe N-phenyl-1-naphthylamine (34).

Similar but less dramatic results were seen in S. enterica (25). S. enterica Δcfa mutants were more sensitive to extreme acid pH as described above. Relative to the wild-type strain the Δcfa strain was also more sensitive to the protonophore CCCP and to hydrogen peroxide. In addition, the Δcfa strain had reduced viability in murine macrophages. S. enterica Δcfa strains also showed attenuated virulence in mice. The probabilities of survival were about 2-fold greater for the wild type and the Δcfa strain complemented by a cfa plasmid than that seen in the Δcfa strain. Note that loss of CFA in Enterococcus faecalis was recently reported to have no effect on pathogenesis in mice (36).

NEW PLAYERS IN REGULATION OF E. COLI cfa GENE EXPRESSION?

At the time of the 1997 review, it was known that the E. coli cfa gene is transcribed from two promoters (37). Later work showed the same arrangement in the closely related S. enterica (26, 38). The more distal promoter is a standard σ70 constitutive promoter, whereas a second promoter closer to the coding sequence required RNA polymerase containing the RpoS sigma factor, σS, a complex often called the stationary phase RNA polymerase. Since RpoS accumulates during the entry into the stationary phase, this accounts for the large increase in CFA formation during entry into the stationary growth phase, where the σ70 promoter would be responsible for the low levels of CFA synthesis in log phase (37). Recently, a small RNA (sRNA) of ∼65 nucleotides called RydC was found to markedly stimulate transcription initiated by the σ70 constitutive promoter by binding the longer cfa mRNA and protecting it from the RnaseE endonuclease (38). S. enterica RydC is poorly expressed (∼4–16 copies/cell) such that it has little effect unless it is overexpressed from a plasmid (38). Upon overexpression, the levels of both the C17 and C19 CFA species increased about 4-fold in log phase cells (38). Similar results were obtained in E. coli, although no increase was reported in the C19 CFA (39). However, Figure 7a of that report seems in error since Table S3.1 in the dissertation of the first author (40) reports increased C19 CFA levels as previously seen in S. enterica (36).

Two other sRNA species, ArrS and CpxQ, also directly affected cfa expression at the posttranscriptional level (39). Like RydC, ArrS acts by masking an RnaseE cleavage site in the cfa mRNA 5′ untranslated sequence, whereas CpxQ binds to a different site in the cfa mRNA 5′ UTR and represses cfa expression (39). Although these studies were done in great detail and the data are strong, a major caveat is that regulation is only seen upon high level overexpression of the sRNAs. The general rationale for sRNA regulation is that a stress response increases the level of a given sRNA. However, the triggering stress is unknown, which precludes placing these findings into physiological contexts.

Note that prior work showed that upon dilution of late stationary phase cultures, little CFA synthesis is seen, indicating loss of the high levels of CFA synthase accumulated in the stationary phase cultures. This loss is due to proteolysis (41), although the responsible protease is unknown. FtsH has recently been suggested to play this role (2).

INHIBITORS OF CYCLOPROPANE SYNTHESIS REACTIONS

Screening of compound libraries and synthesis of substrate analogs have produced some inhibitory compounds (42). A number of these are SAM analogs and thus inherently nonspecific given the many SAM-dependent methyltransferases in cells. Phospholipids having a fluorine atom substituted for one of the double bond hydrogen atoms are not substrates but reversible inhibitors (13). Similar results were reported for a phospholipid in which an epoxide replaced the double bond (13).

Curiously, a rather prosaic compound, dioctylamine, has proven to be a very useful mimic of the methylene accepting substrate. The compound was first shown to be a potent inhibitor of the E. coli enzyme in vitro (42). Although no data on inhibition of E. coli CFA synthesis in vivo are reported, this may be due to the impermeability of the outer membrane to hydrophobic molecules. However, in other bacteria the compound has been very useful (35). In H. pylori, dioctylamine blocks CFA synthesis in vivo to extents approaching that given by deletion of the cfa gene (34). Dioctylamine seems a highly specific inhibitor of CFA synthesis in H. pylori because CFA synthase overproduction largely (but not completely) offsets inhibition by the compound (35). In mycobacteria dioctylamine seems to be a general inhibitor of cyclopropanation in that it inhibits the lipid modifications catalyzed by MmaA4, MmaA3, PcaA, MmaA2, and CmaA2 (31). In each case, overproduction of the responsible enzyme overcame dioctylamine inhibition of that specific modification. The CmaA2 protein was cocrystallized with dioctylamine, and the inhibitor was positioned with the nitrogen atom situated at the position of the substrate double bond to be modified (30). The two octyl aliphatic chains extend toward the entrance to the catalytic site. Moreover, other mycolic acid cyclopropane modification enzymes have been cocrystallized with compounds closely related to dioctylamine with similar results (30). Indeed, dioctylamine inhibits growth of M. tuberculosis, and this can largely be overcome by the combined overexpression of MmaA1, MmaA2, MmaA3, MmaA4, CmaA2, and PcaA (30).

CONCLUSIONS

It is satisfying to find that cyclopropane formation plays important roles in bacterial physiology and pathogenesis. The reaction consumes SAM, an unusually expensive molecule. SAM synthetase (methionine adenosyltransferase) is unusual in that the entire tripolyphosphate chain is cleaved from the ATP molecule and the tripolyphosphate is further degraded to pyrophosphate and phosphate. Hence, synthesis of each SAM molecule requires three ATP equivalents. Following methyl (methylene) transfer, the SAH product is cleaved to homocysteine (which can be converted to methionine) plus adenosine. In E. coli and S. enterica, the bulk of CFA formation occurs in early stationary phase where energy (ATP) is limiting due to low oxygen tension and nutrient limitation. It would seem reckless for a bacterium to expend three ATP equivalents to form each CFA molecule, unless the modification reaps strong benefits. This raises the question of why not synthesize CFA during exponential phase growth where ATP is not limiting? Why is E. coli CFA synthase proteolytically destroyed in stationary phase following phospholipid modification? Are CFA modified phospholipids deleterious to exponential phase cells? It does not appear so. Overproduction of CFA synthase from a plasmid in exponential cultures of E. coli resulted in increased CFA (essentially 100% modification of phospholipid unsaturated acyl chains) and synthase levels but had no discernible effect on growth (43). Perhaps the explanation for CFA synthase destruction is to conserve SAM for essential methylations of DNA, RNA, and other molecules that are rapidly synthesized in the exponential phase, where CFAs are not needed.

Are there conditions where high levels of CFA are needed in exponential growth? The sRNAs that act on the transcript from the promoter that expresses cfa in exponential phase argue this might be the case. However, the only difference in phenotype between wild type and ΔrydC strains is a very modest (∼2-fold) increase in sensitivity to acid shock, whereas strains deleted for arrS or cpxQ showed no increase in sensitivity to acid shock (39). Identification of the stresses that gives sRNA levels similar to those given by those expressed from plasmids would uncover new roles for CFA-modified phospholipids.

Biographies

John E. Cronan received his PhD in 1968 from the University of California, Irvine. He then spent two years as a postdoc with Roy Vagelos at Washington University before joining the Yale biochemistry faculty. After 8 years he moved from Yale to the University of Illinois, where he remains as Microbiology Alumni Professor and Professor of Biochemistry. Cronan is a bacterial geneticist/biochemist who has made fundamental contributions in several different fields. He is best known for his work on the synthesis of fatty acids and the related vitamin cofactors, biotin and lipoic acid. His lab was the first to determine the pathways for synthesis of lipoic acid, a key cofactor and the first pathways for synthesis of pimelic acid, the first intermediate in synthesis of the key cofactor biotin. Cronan is a Fellow of the American Academy of Microbiology and a member of the National Academy of Sciences, USA.

Tiit Luk obtained his PhD in 2009 with John A. Gerlt at the University of Illinois at Urbana-Champaign on the determination of enzyme function of the members of the enolase superfamily. Next, he joined the laboratory of Satish K. Nair at the same university as a postdoctoral research associate, to investigate among other topics the structural biology of lignin-modifying enzymes. In 2013 he joined the team of researchers at the Cornell High Energy Synchrotron Source (CHESS), where he served as a beamline scientist at the macromolecular diffraction facility (MacCHESS). In 2017, after returning to his native Estonia, he established his own research group at the Department of Chemistry and Biotechnology at Tallinn University of Technology (TalTech), focusing on a variety of technologies related to biomass valorization. He currently leads the core research laboratory of biomass valorization technologies at TalTech.