Assessing in silico the recruitment and functional spectrum of bacterial enzymes from secondary metabolism

Abstract

Background: Microbes, plants, and fungi synthesize an enormous number of metabolites exhibiting rich chemical diversity. For a high-level classification, metabolism is subdivided into primary (PM) and secondary (SM) metabolism. SM products are often not essential for survival of the organism and it is generally assumed that SM enzymes stem from PM homologs.

Results: We wanted to assess evolutionary relationships and function of bona fide bacterial PM and SM enzymes. Thus, we analyzed the content of 1010 biosynthetic gene clusters (BGCs) from the MIBiG dataset; the encoded bacterial enzymes served as representatives of SM. The content of 15 bacterial genomes known not to harbor BGCs served as a representation of PM. Enzymes were categorized on their EC number and for these enzyme functions, frequencies were determined. The comparison of PM/SM frequencies indicates a certain preference for hydrolases (EC class 3) and ligases (EC class 6) in PM and of oxidoreductases (EC class 1) and lyases (EC class 4) in SM. Based on BLAST searches, we determined pairs of PM/SM homologs and their functional diversity. Oxidoreductases, transferases (EC class 2), lyases and isomerases (EC class 5) form a tightly interlinked network indicating that many protein folds can accommodate different functions in PM and SM. In contrast, the functional diversity of hydrolases and especially ligases is significantly limited in PM and SM. For the most direct comparison of PM/SM homologs, we restricted for each BGC the search to the content of the genome it comes from. For each homologous hit, the contribution of the genomic neighborhood to metabolic pathways was summarized in BGC-specific html-pages that are interlinked with KEGG; this dataset can be downloaded from https://www.bioinf.ur.de .

Conclusions: Only few reaction chemistries are overrepresented in bacterial SM and at least 55% of the enzymatic functions present in BGCs possess PM homologs. Many SM enzymes arose in PM and Nature utilized the evolvability of enzymes similarly to establish novel functions both in PM and SM. Future work aimed at the elucidation of evolutionary routes that have interconverted a PM enzyme into an SM homolog can profit from our BGC-specific annotations.

Keywords: Enzyme design; Enzyme evolution; Primary metabolism; Secondary metabolism.

Fig. 1

Occurrence of EC subdivisions in PM* and SM* and their overrepresentation in BGCs. a A plot of f _PM *(EC_sd) -values versus f _SM *(EC_sd) -values. These are the normalized frequencies for the occurrence of EC subdivisions in the datasets enzymes _PM* and enzymes _SM*, respectively. b A plot of overrep(EC_sd)-values versus f _SM *(EC_sd) -values. Each overrep(EC_sd)-value is the ratio f _SM *(EC_sd)/f _PM *(EC_sd) that relates the abundance of a subdivision in enzymes _PM* and enzymes _SM*. For subdivisions with an overrep()-value > 20 (red symbols) and those most abundant in enzymes _SM* (green symbols), the EC_sd number is given

Fig. 2

Multifunctionality deduced from homologous PM*/SM* pairs and determined for EC classes. The nodes represent the six EC classes and arrows indicate the relation of functional difference PM* → SM*. For example, the arrow 6 → 2 signals that PM* homologs of SM* class 2 enzymes belong to EC class 6; this arrow is marked with a ♦ symbol. The width of the arrows represents the number of BLAST hits of enzymes from enzymes _SM* in enzymes _PM* and their color the mean E-value; hits were binned as indicated. In addition, for each class, the number of PM* BLAST hits is given and the rate of functional conservation fc, which is the fraction of PM* BLAST hits that belong to the same EC class as the SM* queries. The class EC 1 subsumes oxidoreductases that catalyze oxidation/reduction reactions and EC 2 transferases that transfer functional groups. EC 3 consists of hydrolases that catalyze the formation of two products from a substrate by hydrolysis and EC 4 contains lyases that catalyze the non-hydrolytic addition or removal of groups. The isomerases of EC 5 catalyze the intramolecular rearrangement within a single molecule and the ligases of EC 6 join together two molecules under consumption of ATP or similar triphosphates

Fig. 3

Multifunctionality deduced from homologous PM/SM pairs and determined for EC subdivisions. The nodes represent EC subdivisions and arrows indicate the relation of functional difference PM* → SM*. For example, the arrow 2.6.1 → 4.2.1 signals that PM* homologs of SM* subdivision 4.2.1 belong to EC subdivision 2.6.1; this arrow is marked with a ♦. The width of the arrows represents the number of BLAST hits of enzymes from enzymes _SM* in enzymes _PM* and their color the mean E-value; hits were binned as indicated. In addition, for each subdivision, the number of PM* BLAST hits is given and the rate of functional conservation fc, which is the fraction of PM* BLAST hits that belong to the same EC subdivision as the SM* queries. Subdivisions that do not occur in enzymes _PM* are indicated by a “-“

Fig. 4

KEGG genome map for the neighborhood of gene Sare_3902 from S. arenicola. The picture was created by using the “Genome map” function of the KEGG gene entry Sare_3902. KEGG uses the following color code to fill the arrows representing genes: amino acid metabolism (orange), metabolism of cofactors and vitamins (pink), energy metabolism (violet), environmental information processing (yellow), unclassified (white). Sare_3902 codes for a 4-oxalocrotonate decarboxylase; compare Table 6

Fig. 5

Software architecture for the determination of PM*/SM* homologs. This schema illustrates how the number and the enzyme function of PM* homologs was determined for SM* enzymes. For each of the 15 species listed in Table 1, the genome was downloaded from KEGG and the functional assignment was supplemented by using InterPro and other databases. Subsequently, a BLAST database (DB) was built for each of the genomes. The sequences of the SM* enzymes deduced from the chosen BGCs were then BLASTed against all 15 databases. All BLAST hits were stored in PM* → SM* hitlist that contained all PM* → SM* pairs. Based on the chosen E-value cutoff, our program Hitlist Analyzer selected those hits that both had assigned an EC number and compiled the output table EC hits-counts. This table contained for each EC number the number of PM* hits (TOT) and the number of PM* hits having assigned the same (SAME) and a different (DIFF) EC number. These raw data were further processed to determine frequencies and related parameters for EC numbers, subdivisions, and classes