Comparativa de clasificadores

The recent surge in publicly available bacterial genome sequences has led to the identification of several cryptic gene clusters that potentially encode for interesting compounds. Several research groups have embraced the idea of lantipeptide genome mining, resulting in the identification of various compounds that would not necessarily have been found in an activity-based screening approach. A cryptic lantipeptide cluster can be of particular interest because the product may be related to potent known

compounds, because of a predicted novel lantipeptide scaffold, or even because the cluster is predicted to encode a novel biochemical mechanism of post-translational modification. Various high throughput tools will be required that will help to cope with the increasing amounts of bacterial genome data, but current techniques are already very promising. Lantibiotic detection by MALDI-ToF MS on whole cell samples can make compound screens more efficient since less handling is required (Hindré et al. 2003).

Bioinformatic mining of genomic data for gene clusters involved in production of lantipeptides and non-modified bacteriocins can be performed with the BAGEL2 software (de Jong et al. 2010). This section describes both biochemical and molecular genetic approaches aimed at the identification of a product from cryptic lantipeptide gene clusters.

I.4.6.a. Biochemical approaches to genome mining

One of the first successful genome mining attempts was the identification of the two-peptide lantibiotic haloduracin from a cryptic cluster found in Bacillus halodurans (McClerren et al. 2006). Both LanM enzymes were purified from a heterologous E. coli host and shown to modify the purified HalA1 and HalA2 precursor peptides in vitro. A structure was proposed for the resulting modified peptides and they were shown to display antimicrobial activity when both of them were used together, a common feature of two-peptide lantibiotics. A further structure-activity relationship study identified the Halα B ring as expendable and that the Halα cystine is not required for activity, but protects the peptide against proteolytic degradation (Cooper et al. 2008). Two Ser residues were shown to escape dehydration, requiring the previous structure to be revised. In vivo production of haloduracin was also shown by purification from a B.

halodurans fermentation and by mutational analysis (Lawton et al. 2007).

A similar biochemical approach was taken for a cryptic gene cluster in S. venezuelae, which led to the identification of venezuelin (Goto et al. 2010). This case provides an example of a gene cluster that was of interest because of the unusual modification enzyme VenL. Venezuelin does not appear to have antibacterial activity and in vivo production has not yet been demonstrated. The genetic analysis of venezuelin production is a subject of this thesis and is discussed in more detail in Chapter VII.

A bioinformatic search designed to find examples of bacteria that can produce multiple lantipeptides with the same LanM enzyme identified several clusters in marine planktonic cyanobacteria (Li et al. 2010). The genome of Prochlorococcus MIT9313 contains only one lanM homologue, but seven putative lanA genes in this cluster along with 22 additional lanA genes elsewhere in the genome. The prepropeptides display a remarkably high sequence identity in the leader region, but their propeptides (ranging from 13 to 32 amino acids in length) have very diverse sequences. An in vitro dehydration and cyclisation assay with purified ProcM and seventeen ProcA precursors showed that this enzyme was capable of efficient modification of this wide variety of substrates. Several of these, collectively called ‘prochlorosins’, were shown to be produced by Prochlorococcus MIT9313 in vivo (Li et al. 2010). This is a very nice example of how an organism uses a relatively simple biochemical system to generate a broad diversity of secondary metabolites.

I.4.6.b. Molecular genetic approaches to genome mining

The genome sequences of clinical isolates of S. aureus revealed the presence of a cryptic lantipeptide gene cluster (Daly et al. 2010). Reverse genetics and a mutational analysis of this cluster were performed to produce the Bsa (for bacteriocin of S. aureus) lantibiotic, which turned out to be identical to the previously identified, but not structurally characterised, bacteriocin staphylococcin Au-26 (Scott et al. 1992). Bsa is structurally related to gallidermin, but its immunity genes do not contribute significantly to gallidermin immunity.

An in silico screening strategy of the public databases identified 89 LanM homologs, including 61 in strains that were known as lantibiotic producers. A cryptic gene cluster in Bacillus licheniformis was shown to produce a two-peptide lantibiotic, lichenicidin, which exhibits antimicrobial activity against important pathogens such as MRSA, VRE and Listeria monocytogenes (Begley et al. 2009). Insertion inactivation mutants were generated for the two lanM genes, linking them to the production of lichenicidin (Dischinger et al. 2009). A lichenicidin cluster with identical structural genes was identified in a different B. licheniformis strain and the structure of both the Lchα and Lchβ peptides were determined by NMR (Shenkarev et al. 2010).

An elegant system was devised that employed the nisin biosynthetic machinery to produce the previously uncharacterised two-peptide lantibiotic pneumococcin (Majchrzykiewicz et al. 2010). This demonstrated that the Class I nisin biosynthetic machinery was capable of post-translational modification of Class II propeptides when fused to the nisin leader sequence. Although this was a very nice proof of principle, it was also rather adventurous, since the use of the original or more closely related biosynthetic machinery intuitively seems to have a greater chance of success for correct production of cryptic lantipeptides. Indeed, several versions of both peptides were detected, each with different number of dehydrated residues (Majchrzykiewicz et al.

2010). This could be indicative of inefficient modification, which in turn could have implications for correct (Me)Lan formation. The two modified pneumococcin peptides displayed antibacterial activity against Micrococcus flavus, but they were not found to act synergistically, as commonly observed for two-peptide lantibiotics. Another issue is the introduction of modifications that do not naturally occur in nisin, such as AviCys for example.

I.5. Microcins

Microcins are a class of low-molecular weight (< 10 kDa) ribosomal peptides produced by enterobacteria to kill off closely related competitors. Some microcins are unmodified and will not be discussed here. Others contain extensive post-translational modifications, resulting in a wide variety of structures (Figure I.1.).