For a long time, all L. monocytogenes isolates were regarded as strictly pathogenic at the species level, and were always related to disease. However, from the experimental data collected over recent years, it has become clear that L. monocytogenes demonstrates serotype/strain variations in virulence and pathogenicity rate . The population structure of 43 low-virulence strains was investigated with that of 49 virulent strains to estimate their diversity from virulent strains. We also investigated whether low-virulence strains formed a homogeneous subpopulation of L. monocytogenes or whether they originated from a random loss of virulence genes and thus diversified in multiple distinct directions.
We based our analysis on PFGE and different DNA-sequence-based approaches. The PFGE gave the greatest discriminatory power. Indeed PFGE gave profiles for different strains that by another way were grouped together in MSTrees. For example, ST2 (Figure 3) comprised low-virulence strains of the phenotypic Groups-I, -V, and -VI, which had different PFGE profiles. Similarly, the low-virulence strains AF105 and LSEA-99-23 exhibited the same MLST profile but had distinct profiles in PFGE. Interestingly, MSTree identified specific ST for half of the low-virulence strains belonging to lineage II.
Overall, we identified low-virulence L. monocytogenes strains in both lineages I and II. No hypothesis could be advanced for the lineage III/IV, as they were few strains studied here represented these lineages. Our population structure showed that low-virulence strains are linked firstly according to their lineage, then to their serotypes and after which, they lost their virulence suggesting a relatively recent emergence. MSTree analyses showed that low-virulence strains belonging to lineage II formed a tightly clustered, monophyletic group with limited diversity, in contrast to the low-virulence strains of lineage I. All our observations further supported the fact that some correlations existed between virulence level and point mutations, base substitutions inducing a stop-codon, or inactivation of different virulence proteins, rather than on horizontal transfer or gene loss [7, 8, 20]. A characteristic of lineage II low-virulence strains was that all strains had a point mutation in the virulence inlA gene. Interestingly, there was a strong correlation between the inlA mutation and the genotypic group which were based on the mutations responsible for the virulence lost. Moreover, all strains of ST31 had only two different inlA mutations, but only the strains with the mutation type 5, according to Van Stelten also have the PrfAK220T mutation . This observation suggested that the inlA mutation appeared before the prfA mutation. Regardless of the nature of mutations in inlA in the different low-virulence strains, there was clearly a link between their prevalence in food environments and the inlA mutations. Indeed, the inlA mutations were identified mainly in serotypes 1/2a and 1/2c from lineage II isolated from food and food-processing environments [17, 21]. As such, it is reasonable to hypothesize that variations within these groups have been shaped to a greater extent by selective constraints operating in food manufacturing-plants.
It is intriguing that InlA, and to a lesser extent PrfA, which are important bacterial factors for host colonization, were lost. This pattern could be explained either by relaxation of the selective constraint to maintain InlA and PrfA function or by a selective advantage provided by the loss of functional virulence proteins in the ecological niche occupied by these strains. Clonal families might be adapted to different niches, and their occurrence as mammalian pathogens may be of limited significance for their evolutionary success in the long term. Considering all altered factors, the low-virulence strains could represent over 50% of the L. monocytogenes strains . The fact that the growth of some low-virulence L. monocytogenes strains was impaired on selective medium suggests that the prevalence of these strains may be higher than that currently reported . Moreover, only a few L. monocytogenes strains isolated from the environment and/or food have been analyzed, in contrast to strains of human origin. Developing reliable and easy-to-perform virulence tests could be useful, particularly for risk analysis, where it is important to evaluate the risk associated with the consumption of food products contaminated with L. monocytogenes not only on the basis of levels of bacterial contamination but also on the virulence level of the strains.
In this complex diversity scheme, the case of the A23 strain is very intriguing. Indeed, it is still virulent in mice, despite non-functional major virulence genes, due to point mutations in inlA, inlB and plcA that characterize the genotypic Group-IIIa . This strain was found to be in the same cluster as the Group-IIIa strains using PFGE and MLST analyses, but to be in a specific ST using MSTree (ST 196 and 193, respectively). The fact that this strain has an additional mutation in mpl compared to Group-IIIa strains  suggests that it evolved from this group and thus reacquired virulence genes after initial virulence-gene loss. However, optical mapping does not support this hypothesis, since compared to the EGDe genome, specific fragments have been inserted in the genome of the Group-IIIa strains but not in strain A23, suggesting that the Group-IIIa strains have evolved from the latter. The complete sequencing of the genome of these strains should clarify this question.
This analysis corroborated the classification obtained for the phenotypic Groups-I and –III. Moreover the new detected low-virulence strains exhibiting the same phenotypes and harbouring the same mutations in the virulence genes, as previously observed, reinforced our observations. The new results allowed us to subdivide the former Group-IV into 3 new Group-IV, -V and –VI and to suggest different hypothesis concerning the population structure and diversity of the low-virulence strains compared to virulent strains.