Untying the anchor for the lipopolysaccharide: lipid A structural modification systems offer diagnostic and therapeutic options to tackle polymyxin resistance

Currently, the gold standard recommended by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) for clinical diagnostics of polymyxin resistance used in clinical microbiology laboratories is minimal inhibitory concentration (MIC) (28). Although these diagnostic methods are the most reliable at the moment, they do not inform about the resistance mechanism involved. In clinical settings, this mechanism is generally interpreted through polymerase chain reaction (PCR) after phenotypic resistance has been established (7). Studies have shown that some bacterial strains, referred to as polymyxin heteroresistant, do not show resistance in MIC susceptibility testing but do acquire lipid A modifications along with the genes for polymyxin resistance (29–35). These strains raise concern over their potential to spread polymyxin resistance in clinics, as they usually are not tested for genetic polymyxin resistance. Yet, information about the exact polymyxin resistance mechanism that a bacterium has developed can help us decide the course of hospital treatment. For instance, a patient infected with a bacterium that has plasmid-encoded polymyxin resistance may require quarantine and more cautious behaviour of clinical staff. This is because plasmids can be transferred horizontally from one bacterial species to another. Bacteria that have colonised a hospital area and clinical staff might receive genes for polymyxin resistance by being in close contact with a patient infected by a bacterium carrying plasmid-encoded polymyxin resistance.

Researchers have been working on developing rapid clinical microbiology diagnostic assays which rely on the mass spectrometry (MS) of lipid A structural modifications. Prior to analysis, these bacterial samples require specific treatment to cleave a ketosidic bond in the LPS to release lipid A.

Clinical microbiology laboratories have increasingly been utilising matrix-assisted laser desorption/ionisation – time of flight – mass spectrometry (MALDI-TOF-MS) to identify microbes based on protein profiling. This includes profiling for lipid A structures to develop assays that rapidly determine mechanisms underlying polymyxin resistance in clinical settings (36–44). Molecules analysed by MALDI-TOF-MS need to be ionised prior to separation. Negative ionisation mode is optimal for lipid A because its phosphate groups can easily be ionised into anionic form. In view of the fact that lipid A structures are species-specific, the findings of MS analysis need to be interpreted carefully and individually. To illustrate the rationale of lipid A analysis for clinical diagnostics, we will show examples of lipid A structural modifications for some bacterial species which can be visualised by MALDI-TOF-MS.

Klebsiella pneumoniae has at least four native lipid A structures, which have the MS spectrum m/z values of 1824, 1840, 2063, and 2078 (37, 41, 45). Depending on the mechanism underlying polymyxin resistance, there can be an addition of L-Ara4N or/and pEtN, in which case the m/z values increase by +131 or +123, respectively. The same rationale in structural modifications has been observed in other Gram-negative bacteria. Therefore, chromosome-encoded resistance can be viewed as lipid A m/z values of 1971 and 2209, plasmid-encoded resistance can be viewed as lipid A m/z values of 1963 and 2201, and, if both mechanisms are involved, as lipid A m/z values of 1963, 2201, and 2209 (37, 41). Table 1 shows the proposed structures of native and modified lipid A for K. pneumoniae.

As for Pseudomonas aeruginosa, there is no evidence of pEtN addition to lipid A. What we know for now is that the acquired resistance arises from the addition of one or two L-Ara4N moieties to position 1 and/or 4’ phosphate in the native lipid A (39). The most common native lipid A structures can be viewed as m/z values of 1446 and 1462. Therefore, chromosome-encoded resistance can be viewed as lipid A m/z values of 1577, 1593, 1708, and 1724 (39). Table 2 shows the proposed structures of native and modified lipid A for P. aeruginosa.

Native lipid A structures of Acinetobacter baumannii have m/z values of 1728 and 1910. Chromosome-encoded resistance is based on a 4’ pEtN addition to lipid A and a +123 shift (38). Table 3 shows the proposed structures (38, 41). Some minor native lipid A peaks include tetraacylated and pentaacylated lipid A with m/z values of 1404 and 1530, respectively (these structures are not included in this review). Plasmid-encoded resistance has not been detected in A. baumannii yet.

The most common Escherichia coli native lipid A structure has the m/z value of 1796. Chromosome-encoded resistance is identified at m/z of 1927 (L-Ara4N), plasmid-encoded resistance at m/z of 1919, and both mechanisms at respective m/z values of 1919 and 1927 (40, 46). Table 4 shows the proposed structures (40, 41, 46).

Salmonella enterica has at least five native lipid A structures, which correspond to m/z values of 1796, 1812, 1876, 2034, and 2050 (Table 5). Chromosome-encoded resistance is identified as lipid A m/z values of 1927 and 2165 (corresponding to the L-Ara4N addition), and plasmid-encoded resistance as lipid A m/z values of 1919, 1935, 2157, and 2173 (corresponding to the pEtN addition) (36, 41).

Based on the findings from the Larrouy-Maumus research group at Imperial College London, the MALDIxin test, commercialised by Bruker Corporation under the name MBT Lipid Xtract^™ kit (47), shows great promise as indicator of polymyxin resistance to be used in clinical diagnostics.

Currently, polymyxins are the most promising agents for treating infections with extensively drug-resistant (XDR) Gram-negative bacteria. Therefore, we need to foresightedly work on polymyxin adjuvant therapeutics. Adjuvant compounds normally lack inherent antimicrobial activity as single agents, but in combination with a specific antibiotic they potentiate its effects. The idea of polymyxin adjuvant therapy is to reverse polymyxin resistance by inhibiting one of the resistance-driving mechanisms (mechanisms involved in lipid A modifications). Some of the most promising targets include enzymes involved in lipid A modification systems responsible for resistance [e.g. MCR-1, ArnT and diacylglycerol kinase A (DgkA)]. The interactions of colistin with lipid model membranes have been described thoroughly by structural biology studies (48), which have shown how these interactions change when pEtN is added to the phosphate groups of lipid A (49, 50) and how changes in LPS charge renders polymyxins ineffective. Therefore, in theory, for polymyxins to regain their efficacy these changes need to be prevented. In this context, we see another potential application of MALDI-TOF-based polymyxin resistance screening. For instance, analysing lipid A structure would indicate which inhibitor class would be useful for each patient.

MCR-1 is a pEtN transferase encoded by plasmid-borne mcr-1 genes. Xu et al. (51) have evidenced that MCR-1 [expressed, purified, and incubated with a natural substrate mimetic (nitrobenzodiazole-labelled glycerol-3-pEtN] transfers pEtN to phosphate groups of lipid A in Gram-negative bacteria, while nitrobenzodiazole-glycerol is released. The structure of MCR-1 soluble C-terminal domain has been elucidated through X-ray crystallography (52–56). Its active site is located in the soluble domain, coordinated by an essential Zn²⁺ ion, and has a highly conserved (through the pEtN family) T285 aminoacid residue that acts as a catalytic nucleophile in substrate modifications (52, 57, 58). These findings provide valuable starting point for computer-aided drug design (CADD) in the quest for novel MCR-1 inhibitors (59, 60). Since the discovery of MCR-1, some research groups have been screening for potential MCR-1 inhibitors from natural products (61–66) or developing de novo synthesised MCR-1 inhibitors (67). Interestingly, an FDA-approved drug, auranofin (used in the treatment of rheumatoid arthritis), has also shown MCR-1 inhibition (68). So far, this is the most studied lipid A-related target for potential adjuvants. As MCR-1 has been studied quite extensively, there is hope that these findings can be applied to other members of the MCR family. It would be ideal that a candidate drug inhibits all members of the MCR protein family.

As for chromosome-encoded polymyxin resistance, small molecules that downregulate TCS have been found to reverse resistance (69, 70). Another promising target related to chromosome-encoded resistance is ArnT. Ghirga et al. (71) performed a virtual screening for ArnT inhibitors to discover a potent colistin adjuvant BBN149, which can be used as a lead molecule for further development. Quaglio et al. (72) have shown that ent-Beyerane diterpenes inhibit ArnT thanks to their resemblance to L-Ara4N. This means that the structure of these compounds most likely has a similar binding mode as L-Ara4N. Another target which might be of interest is DgkA. As mentioned above, EptA is transferring pEtN to the phosphate groups of lipid A. EptA uses phosphatidylethanoloamine (PE) as its pEtN donor, which results in the formation of diacylglycerol (DAG) as a by-product (10). DAG is quickly phosphorylated by DgkA into phosphatidic acid, a precursor for PE synthesis (73), and DAG accumulation has been shown to inhibit EptA activity (15, 74). All this points to DgkA as a prominent target of polymyxin resistance adjuvant therapy.

Another potential option are eukaryotic kinase inhibitors, which have been found to re-sensitise bacteria resistant to colistin. Although the mechanism is not clear yet, they have been reported to lower the rate of lipid A modifications by L-Ara4N and pEtN (75). Sertraline, an FDA-approved selective serotonin reuptake inhibitor (SSRI) used for the treatment of depression, has been found to revert polymyxin resistance, as it seems to interfere with the eptA lipid A modification pathway (76).

All things considered, there is immense potential for the development of polymyxin adjuvants by targeting lipid A modification systems.