In their review, Cotter et al.1 have provided an excellent account of contemporary research into bacteriocins of food-grade lactic acid bacteria (LAB). The bacteriocin subset of the antibiotics have been defined as those 'ribosomally encoded proteinaceous molecules released extracellularly by bacteria that inhibit the growth of other bacteria, typically including some that are closely related to the producing bacterium and to which the producer cell expresses a degree of specific immunity'2,3,4, Cotter et al. rightfully acknowledge that the term bacteriocin (coined >50 years ago by Jacob et al.5) is a more general descriptor for colicin-like antimicrobials produced by bacteria. However, it is problematic that they then proceed to re-define the bacteriocins as 'heat-stable peptides'1, since most colicins, being heat-labile proteins, would consequently not qualify as 'true' bacteriocins! This clearly fails to properly acknowledge the historical significance of colicins as the prototype bacteriocins. Later, the authors propose a very timely revision of the incumbent 'gold standard', but now outdated, LAB bacteriocin classification scheme devised by Klaenhammer in 1993 (Ref. 6).

The Klaenhammer classification scheme (Fig. 1a) comprised four major bacteriocin classes6. By contrast, the Cotter proposal has, as its basis, just two major divisions: Class I (the lantibiotics) and Class II (the non-lantibiotic bacteriocins) (Fig. 1b). Klaenhammer's classes IIc and IV were withdrawn, since they are unproven entities and the novel post-translationally modified cyclic peptides7,8 (recently recommended by Kemperman et al.9 to be assigned as 'Klaenhammer Class V') were relegated to Class IIc. It is however the proposal to withdraw, and rename as bacteriolysins, the large heat-labile Class III (and therefore colicin-like) proteins that causes us particular concern. All of the defining characteristics of bacteriocins (see above) are exhibited by the bacteriolytic proteins lysostaphin and zoocin A, including genes encoding specific immunity factors10,11. Moreover, several other Class III LAB bacteriocins such as helveticin J12 and the plasmid-encoded dysgalacticin13 (N.C.K. Heng et al., unpublished results) do not kill by lytic means and hence are clearly not classifiable as bacteriolysins.

Figure 1: Proposed classification scheme for bacteriocins.
figure 1

A 'universal' scheme is put forward that builds upon the foundations of (a) the original classification scheme for the lactic acid bacteria (LAB) bacteriocins by Klaenhammer6, and incorporating elements of (b) the revised scheme of Cotter et al.1 Although not specifically mentioned in the text, we also propose formalizing the subdivisions of the lantibiotics into Types Ia (linear), Ib (globular) and Ic (multi-component). The first documented division of the lantibiotics into Types A and B for linear and globular molecules, respectively, was by Jung14.

Full size image

So, to redress this apparently unjustifiable 'Class distinction' directed against the Class III colicin-like bacteriocins, we now advance an alternative classification scheme: one that builds upon the foundations of the Klaenhammer scheme and also incorporates aspects of the Cotter proposals, but which differs in that it makes provisions for all currently known antimicrobials that conform to the original 'Jacobian' definition of a bacteriocin.

In the scheme that we propose (Fig. 1c), in agreement with Cotter et al. the Klaenhammer classes IIc and IV are eliminated. However, in contrast to Cotter, Class III (large bacteriocins) is retained and now divided into Types IIIa (bacteriolytic) and IIIb (non-lytic), and the cyclic post-translationally modified bacteriocins (Class IIc) are now upgraded to Class IV. This classification scheme can be applied to most bacteriocins, if not all, regardless of the Gram status of the producer strain. Inevitably, the scheme that we have proposed will itself continue to evolve in parallel with our increased knowledge of these seemingly ubiquitous manifestations of microbial diversity.