The lysozymes from the three Caenorhabditis species are listed in Table 1 and 2. The genomic distribution of clustered genes is illustrated in Fig. 1. As a first step, we compared all complete lysozyme protein sequences from C. elegans with those from various vertebrates, invertebrates, protists, and one phage. For this purpose, a multiple sequence alignment was generated based on a hierarchical method, i.e. similar sequences are aligned first, followed by alignment of less similar sequences (see methods). We noted that the resulting alignment almost exclusively contained variable positions. Moreover, if we varied the settings of the alignment algorithm (e.g. gap opening, gap extension, or gap distance penalties), only few regions could be recovered in identical form. Therefore, positional homology may not be entirely reliable. Since the alignment was inferred from the hierarchical algorithm, it should still be informative as to the general phylogenetic position of the nematode lysozymes. In fact, its phylogenetic analysis highlighted that the nematode possesses two distinct lysozyme types (Fig. 2; inferred from the alignment obtained using default settings of the alignment programme), thus confirming previous observations. In particular, five C. elegans lysozymes group with the invertebrate-type lysozymes. These lysozyme genes are thus labelled Cel-ilys-1 up to Cel-ilys-5. The remaining ten C. elegans lysozymes are the previously labelled genes lys-1 up to lys-10. They fall into two separate lineages within the distinct clade of protist lysozymes.

For the more detailed phylogenetic analyses, we examined the two lysozyme types separately. For this purpose, we generated two new alignments and several subsets of these (see methods and below).

The genomes of C. briggsae and C. remanei contain two and three invertebrate-type lysozyme genes, respectively. They were named in consideration of their similarity and phylogenetic affinity to the C. elegans lysozymes (see below; Table 1). Three of the five invertebrate-type lysozymes from C. elegans are found in a single cluster and with the same orientation on chromosome IV (Fig. 1A). None of the other genes are present in clusters (Table 1). All invertebrate-type lysozymes could be reliably aligned to each other at both protein (alignment 2; Fig. 3) and DNA sequence level (alignment 3; position of indels is identical between the two alignments). Two genes show unusual properties in comparison to the others and thus they may be non-functional (i.e. they are pseudogenes). In particular, the gene Cbr-ilys-4 possesses an unusual amino terminus and it lacks a signal peptide. Cel-ilys-1 contains a large insertion, it shows many nucleotide differences to the other sequences, and it also lacks a signal peptide.

Phylogenetic analysis of protein and DNA sequences yielded essentially identical tree topologies (Fig. 4A). The only two differences refer to (i) the exact position of Cel-ilys-1, Cel-ilys-2, and Cel-ilys-3 in relation to each other, and (ii) the position of Cel-ilys-4 and Cbr-ilys-4 in relation to the monophylum of Cre-ilys-4.1 and Cre-ilys-4.2. These discrepancies are reflected by low bootstrap support for the respective branches in both protein and DNA trees, indicating lack of sufficient unambiguous phylogenetic information in the sequences. Otherwise, the inferred genealogy identifies two distinct clades, one with the ilys-4 genes and the other with all remaining genes. Both clades contain genes from all three taxa.

Our analysis did not reveal any indication for adaptive sequence evolution across sequences (likelihood ratio test [LRT] comparison between model M8 with either model M7 or M8a, P ≥ 0.5). However, we consistently identified two episodes of positive selection along the phylogeny, regardless of the analysis method (Table 3; Fig. 4). In both cases, adaptive sequence evolution associates with incidences of intra-lineage lysozyme radiations (in one case within the C. remanei and the other case within the C. elegans lineage; Fig. 4A). Most of the remaining branches have a d_N/d_S rate ratio well below 1, suggesting purifying selection (i.e. amino acid changes are selectively disfavoured).

The protist-type lysozymes are present with either seven (both C. briggsae and C. remanei) or ten genes (C. elegans; Table 2). Synteny is found for the genes lys-1, lys-2, and lys-3, which are clustered in all three species – in both C. elegans and C. briggsae on chromosome V and in C. remanei on supercontig 13 (Table 2; Fig. 1B). The gene lys-1 is always found in opposite orientation to the other two. In C. remanei, the lys-3 homologue is separated from the other two genes by approximately 10,000 nucleotides (and four open-reading frames) in contrast to both C. elegans and C. briggsae, where the three genes are directly adjacent to each other. In C. elegans, the lys-7 gene is additionally found on chromosome V, but in a different location than the three clustered genes (Fig. 1B). C. elegans contains a second well-defined cluster of protist-type lysozymes on chromosome IV, including Cel-lys-4, Cel-lys-5, and Cel-lys-6. In this case, there is no synteny in the other species. Interestingly, however, the C. briggsae chromosome IV contains a cluster that combines genes from the above C. elegans cluster (in this case the C. briggsae genes Cbr-lys-6.1 and Cbr-lys-6.2) with the gene Cbr-lys-10. The C. elegans orthologue of the latter gene, Cel-lys-10, is similarly present on chromosome IV but in a different location than the cluster (Fig. 1A). In C. remanei, two additional genes are found in relatively close physical proximity to each other: Cre-lys-8.1 and Cre-lys-8.2 are located on supercontig 9 (Table 2) separated by approximately 200,000 nucleotides and 56 open-reading frames.

The overall phylogeny of the protist-type lysozymes from nematodes and one outgroup taxon (Dictyestelium discoideum) was assessed with an alignment of the complete protein sequences (alignment 4; Fig. 5 and Additional file 1). This alignment was robust to variations of the settings of the alignment programme. In contrast, for the corresponding DNA sequences, several regions could not be recovered in identical form under similar variations. Therefore, it cannot be entirely excluded that these regions bear an increased risk of homoplasy. To reduce this risk for the detailed analysis of lysozyme evolution (i.e. inference of non-synonymous and synonymous substitution rates), we extracted five subsets from alignment 4. Of these, alignment 5 consists of the alignable part of all protist-type lysozyme DNA sequences from the Caenorhabditis nematodes (see vertical black lines with arrows below the alignment in Fig. 5/Additional file 1). Since alignment 5 considered only a comparatively short part of the genes, we additionally analyzed the clade 1 and 2 protist-type lysozymes separately. These separate analyses allowed us to include complete or almost complete genes and thus additional phylogenetic information as contained in the regions excluded in alignment 5. Here, analysis of clade 1 lysozymes was based on the alignable part of the genes (see vertical black lines with arrows above alignment in Fig. 5/Additional file 1; alignments 6 and 7 for protein and DNA sequences, respectively), while that of the clade 2 lysozymes included the complete protein or DNA sequences (alignments 8 and 9, respectively; Fig. 5/Additional file 1). We would like to emphasize that alignments 5–9 are subsets of alignment 4 as indicated in Fig. 5 and Additional file 1 (i.e. position of indels is identical between alignments). Cel-lys-9 was always excluded because it did not permit reliable alignment to the other lysozymes.

For all data sets, protein and DNA sequence alignments yielded essentially identical tree topologies. The only differences referred to (i) the exact position of Cel-lys-7, Cel-lys-8, Cbr-lys-8, and the clade containing Cre-lys-8.1 and Cre-lys-8.2 in relation to each other (alignments 4–7; Figs. 6 and 7), and (ii) the exact position of Cel-lys-5, Cel-lys-6, Cbr-lys-6.1, Cbr-lys-6.2, and Cre-lys-6 in relation to each other (alignments 4, 5, 8, 9; Fig. 6 and 8). Almost all of these differences are again associated with low bootstrap support, suggesting that the available sequences lack sufficient unambiguous phylogenetic information at these two levels. All other relationships were consistently identified, irrespective of the alignment used, indicating the availability of robust phylogenetic information in these cases. The phylogenetic analysis yielded the following information.

(i) The protist-type lysozymes fall into two distinct clades (clade 1 and 2), which diverged before separation of the three species (Fig. 6A).

(ii) Within clade 1, four distinct phylogenetic groups are identified (Fig. 7A). Three of them contain one orthologue per species, indicating duplication of genes before species separation. The fourth group includes one gene for C. briggsae, two monophyletic genes for C. elegans, and two monophyletic genes for C. remanei.

(iii) The inferred clade 2 topology shows less hierarchical structure than the clade 1 topology (Fig. 8A). Here, the lys-10 orthologues form a monophyletic group, which is most closely related to Cel-lys-4. The remainder of this clade shows a single gene from C. remanei, two monophyletic genes from C. elegans, and two monophyletic genes from C. briggsae.

The analysis of positive selection across sequence alignments yielded a single significant result. In particular, for the aligned clade 1 coding sequences (alignment 7, see methods and Fig. 5) model M8 differed significantly from both model 7 (LRT, 2ΔL = 8.786, P = 0.012) and model 8a (LRT, 2ΔL = 6.589, adjusted P = 0.005). A single alignment position was found to be subject to adaptive sequence evolution according to the Bayes empirical Bayes method (P = 0.99). The alignment position is found in the middle of the genes and it is highlighted in Fig. 5. For the other data sets, the comparisons were all insignificant

During assessment of adaptive sequence evolution along branches, the different data sets and the two methods for inference of statistical significance produced slightly different results. For instance for the clade 1 sequences, only a single branch was inferred to have a d_N/d_S rate ratio significantly above 1 by both methods (the branch leading to the lys-1 orthologues). In the analysis of the complete data set (including both clade 1 and 2), the same branch was found to be significant by only one of the two approaches. In spite of these variations, the results from all data sets and methods, taken together, consistently point to two main tree regions that are likely to be subject to positive selection: (i) the branch leading to the lys-10 orthologues (Tables 4 and 6; Figs. 6 and 8), and (ii) the branches associated with the early radiation of the lys-1, lys-2, and lys-7/8 orthologues (Tables 4 and 5; Figs. 6 and 7). The majority of the remaining branches yielded a d_N/d_S rate ratio that was clearly below 1, indicating purifying selection.