2.1 KEGG metabolic pathway map

The KEGG metabolic pathway maps are graphical diagrams representing knowledge of enzyme-catalyzed reaction networks. Each map is manually drawn to capture the overall architecture of how main compounds are converted. The details of individual reactions involving all substrates and products can be examined in the KEGG REACTION entries linked from the map. It is also drawn as a generic map combining and summarizing experimental evidence in different organisms, so that it can be used for interpretation of any genome. This is accomplished by the KEGG Orthology (KO) system described below. Basic graphics objects in the KEGG metabolic pathway maps are boxes for enzymes and circles for chemical compounds (see, for example, http://www.kegg.jp/pathway/map00010). Each circle is identified by the chemical compound identifier (C number). Each box is given two types of identifiers: the reaction identifier (R number) and one or more KO identifiers (K numbers). Although the Enzyme Commission (EC) numbers are usually displayed in the boxes, they are not identifiers and are treated as attributes to KO identifiers. Note that the EC numbers may represent reaction classification of the EC system or gene/protein functional classification in the genome annotation. These two aspects of enzymes are clearly separated by the R number and the K number identifiers in KEGG, enabling the analysis of chemical networks and genomic networks in much better defined ways than using the EC numbers.

2.2 KEGG Orthology

The KEGG Orthology (KO) system is a collection of manually defined ortholog groups (KO entries) for all proteins and functional RNAs that appear in the KEGG pathway maps (both metabolic and non-metabolic) as well as in the KEGG BRITE functional hierarchies (ontologies). Whenever a pathway map is drawn based on experimental observations in specific organisms, an additional manual work is performed for generalizing gene information from those specific organisms to other organisms. This is done by assigning KO entries to the map objects (boxes) and, when necessary, by defining a new KO entry and creating a corresponding set of orthologous genes from available genomes. Each KO entry also represents a sequence similarity group. This allows computational assignment of KO identifiers in newly determined genomes and metagenomes by sequence comparison, which may then be used for KEGG pathway mapping (reconstruction) analysis. Note that the degree of similarity in each group varies significantly because each KO is defined in a context (pathway) dependent manner.

2.3 KEGG reaction class

The KEGG REACTION database contains all biochemical reactions that appear in the KEGG metabolic pathway maps together with the set of experimentally characterized enzymatic reactions in the Enzyme Nomenclature [9], i.e., those with the official EC numbers. Less than one half of the reactions in the KEGG pathway maps correspond to the Enzyme Nomenclature reactions, suggesting the difficulty of using EC numbers for a comprehensive analysis. In order to analyze chemical compound structure transformation patterns, the following processing is performed for all reactions both computationally and manually. First, reactant pairs are defined as one-to-one relationships of substrate-product pairs by considering the reaction type (as classified by the EC system) and the flow of atoms. Second, structure transformation patterns are computed, manually curated, and represented by the so-called RDM patterns of KEGG atom type changes [5-7]. Third, the identity of RDM patterns for the main reactant pairs, i.e., the reactant pairs that appear in the KEGG pathway maps, is used to define KEGG reaction class [8]. The resulting KEGG reaction class (identified by RC number) is like an ortholog group of reactions defined by localized structural changes and accommodating global structural differences of reactants.

2.4 KEGG module

Functional units of enzyme complexes and subpathways are often encoded in positionally correlated gene sets (operon structures) in prokaryotic genomes. When complete genome sequences first became available, a graph analytical method was used to extract enzyme gene clusters on the chromosome that encode consecutive reaction steps in the metabolic pathways [10]. Such functional units are now accumulated in the pathway module section of the KEGG MODULE database. Each KEGG module (identified by M number) is manually defined as a combination of KO identifiers. For example, the reaction sequence involving oxaloacetate + acetyl-CoA, citrate, isocitrate, and 2-oxoglutarate in the citrate cycle (map00020) is the KEGG pathway module M00010 named as “Citrate cycle, first carbon oxidation” and defined by:

where alternative enzymes are given in parentheses. The positional correlation of operon-like structures is not always observed, but when it exists, at least, in certain organism groups, as is the case for many KEGG pathway modules, it well supports the definition of functional units.

2.5 KEGG reaction module

An alternative way to define functional units in the metabolic pathways has been developed recently [8]. It relies only on the chemistry of reactions without using the information about genes and proteins. As mentioned, KEGG pathway nodes (boxes) are given both K numbers (gene/protein orthologs) and R numbers (reactions), where the latter can be converted to RC numbers (reaction class or reaction orthologs). While KEGG pathway modules are conserved subnetwoks of the K number network, different types of conserved subnetworks may exist in the RC number network. This is in fact the case, and conserved reaction sequences termed reaction modules can be extracted from known metabolic pathways [8]. Furthermore, reaction modules (also called RC modules) tend to correspond to KEGG pathway modules (also called KO modules) despite the fact that they are separately defined from different properties. A case in point is the RC module RM001, which exactly matches the KO module M00010, for the reaction sequence from oxaloacetate to 2-oxoglutarate. RM001 is named as “2-Oxocarboxylic acid chain extension by tricarboxylic acid pathway” and defined by:

where the notation is somewhat more complex because of the existence of three subtypes and multi-step reactions denoted by plus signs.

3.1 Reaction modules used in combination

The analysis of reaction modules has revealed the modular architecture of the metabolic network with two interesting aspects: the existence of chemical units containing chemical logic of organic reactions and the correspondence of chemical and genomic units [8]. The chemical units of reaction modules are used in combination as if they are building blocks of the metabolic network, generating different chemical substances in different pathways. A notable example is illustrated in Fig. 1 for 2-oxocarboxylic acid chain elongation and modification as part of the biosynthesis pathways of branched-chain amino acids (valine, leucine, and isoleucine) and basic amino acids (arginine and lysine). 2-Oxocarboxylic acids are an important class of precursor metabolites including pyruvate (monocarboxylic acid), oxaloacetate (dicarboxylic acid), and 2-oxoisovalerate (methyl-modified monocarboxylic acid). The increase of the chain length from these three is shown by vertical arrows in Fig. 1, where eight 2-oxocarboxylic acids are denoted by shaded (red) circles. All the vertical arrows correspond to the reaction module RM001 for increasing the 2-oxocarboxylic acid chain length by one using acetyl-CoA as a carbon source. This module consists of a series of characteristic reactions involving tricarboxylic acids.

Each of the six 2-oxocarboxylic acids, excluding pyruvate and 2-oxobutanoate, is followed by a single reaction step of reductive amination (RC00006 or RC00036) indicated by a horizontal line. The reaction modules for 2-oxocarboxylic acid chain modifications are shown by horizontal arrows. RM033 is for branched-chain addition and RM032 and RM002 are for carboxyl to amino conversion. There is an interesting distinction between RM032 and RM002 defined by:

RM032 for a shorter chain is a direct conversion via a phosphorylation step (RC00043), but RM002 uses a protective N-acetyl group that is added before and removed after (both RC00064) the core sequence of RM032. Furthermore, for a longer chain in lysine biosynthesis, the protective N-acetyl group is attached to a carrier protein.

A similar variation exists in the biosynthetic pathways of fatty acids, in which the acyl chain length is increased by two in one cycle of the four-step reaction sequence consisting of ketoacyl synthase (KS), ketoreductase (KR), dehydratase (DH), and enoylreductase (ER) reactions. Two slightly different reaction modules are identified for this sequence: RM021 for the major pathway and RM020 for the minor pathway in mitochondria. The minor pathway, which is essentially the reversal of beta oxidation, does not involve acyl carrier protein and uses acetyl-CoA as a carbon source. In contrast, the major pathway, which may be a more recent invention, involves acyl carrier protein and uses malonyl-CoA as a carbon source. These examples suggest that the reaction modules seem to contain design principles of a series of organic reactions including how to achieve an activated transition state (e.g., phosphorylation), how to introduce a protective group (e.g., N-acetylation), and how to increase specificity and efficiency (e.g., using carrier protein and switching from acetyl-CoA to malonyl-CoA). This type of chemical logic may have played roles in the chemical evolution of metabolic networks.

3.2 Correspondence of chemical and genomic units

Reaction modules are derived from purely chemical properties of substrate-product structure transformation patterns, but they are found to correspond to KEGG pathway modules defined as sets of enzyme orthologs in the genome. This is already mentioned for the correspondence of RM001 and M00010. In fact, as shown in Fig. 1 the chain elongation module RM001 corresponds to different enzyme units in different pathways: M00010 for oxaloacetate to 2-oxoglutarate in citrate cycle, M00433 for 2-oxoglutarte to 2-oxoadipate in lysine biosynthesis, M00535 for pyruvate to 2-oxobuanoate in isoleucine biosynthesis, and M00432 for 2-oxovalerate to 2-oxosicaproate in leucine biosynthesis. There are two characteristic features. First, the enzyme units are encoded in operon-like gene clusters, at least, in certain genomes. Second, the enzyme units are similar in the sense that they contain paralogous genes coding for similar amino acid sequences.

These features are commonly observed in the KEGG MODULE database. Here a few examples of RM001 are shown in Table 1 for the genomes of Lactococcus lactis IO-1 [11], Bacteroides fragilis YCH46 [12], Thermus thermophilus HB27 [13] and Saccharomyces cerevisiae. The numbering of gene identifiers (locus tags) in the three prokaryotic genomes indicates the proximity of genes on the chromosome. Among the three enzymes (or enzyme complexes) that constitute RM001, the second dehydratases and the third dehydrogenases clearly form paralogous gene groups with sequence identity ranging 35% to 45%. The first enzymes, citrate synthase, homocitrate synthase and 2-isopropylmalate synthase, are more distantly related; only the enzymes in the lysine and leucine pathways exhibit some sequence similarity. Thus, the same reaction sequence appears to be generated in different pathways by duplicating and slightly modifying enzyme clusters. The enzyme that catalyzes the first step of the reaction sequence tends to be more divergent, possibly reflecting the constraint of specific substrate recognition.

Other notable examples can be found in the microbial biodegradation pathways. Certain groups of bacteria are capable of metabolizing non-biological chemicals accumulated in the environment by acquiring or modifying gene sets for biodegradation. These gene sets are often encoded in plasmids and can be transferred within a bacterial community. Environmental pollutants such as endocrine disrupting compounds are mostly aromatic compounds, and the modular architecture of the biodegradation pathways consists of well-defined reaction modules for aromatic ring cleavage. They include preprocessing modules of methyl to carboxyl conversion on aromatic ring (RM004 and RM013) and the main modules of dihydroxylation and cleavage. Four types of dihydroxylation modules are defined distinguishing dioxygenase and dehydrogenase reactions (RM004), dioxygenase and decarboxylating dehydrogenase reactions (RM005), two monooxygenase reactions (RM006), and dealkylation and monooxygenase reactions (RM007). Two basic types of cleavage modules are defined for ortho-cleavage (RM008) and meta-cleavage (RM009). Not surprisingly, these reaction modules well correspond to enzyme gene clusters in operon-like structures defined as KEGG pathway modules. For example, the BTX (benzene, toluene, and xylene) degradation capacity is well represented by the corresponding sets of reaction modules and KEGG pathway modules: preprocessing of toluene to benzoate (RM003 and M00538) or xylene to methylbenzoate (RM003 and M00537), dihydroxylation of benzene to catechol (RM006 and M00548), dihydroxylation of benzoate to catechol (RM005 and M00551), meta-cleavage of catechol (RM009 and M00569), and ortho-cleavage of catechol (RM008 and M00568). These observations suggest a link between genomic diversity and chemical diversity. It should be emphasized again that the link is not simply between individual genes and reactions, but rather between genomic units and chemical units reflecting the modular architecture of the metabolic network.

3.3 Degree of modularity

The modular architecture of reaction modules and enzyme gene sets was most apparent in carboxylic acid metabolism (for 2-oxocarboxylic acids and fatty acids) and aromatics degradation. However, such modularity cannot explain the architecture of the entire metabolic network. Fig. 2 is an overview map for the biosynthesis of twenty amino acids (http://www.kegg.jp/pathway/map01230). Circles represent chemical compounds and lines connecting them are reactions (or sets of reactions). The twenty amino acids are shown in shaded (red) circles, the reaction modules RM001 and RM002 are shown in thick (blue) lines, and the KEGG pathway modules are shown as separate (red) lines with M numbers attached. This overall pathway may be viewed as consisting of the core part and its extensions.

The core part is the pathway module M00002 for conversion of three-carbon compounds from glyceraldehyde-3P to pyruvate, together with the pathways around serine and glycine. M00002 is the most conserved pathway module in the KEGG MODULE database and is found in almost all the completely sequenced genomes. The extensions are the pathways containing the reaction modules RM001 and RM002 for biosynthesis of branched-chain amino acids (left) and basic amino acids (bottom), and the pathways for biosynthesis of histidine and aromatic amino acids (top right). Note that no reaction modules are extracted by our method [8] for the biosynthetic pathways of histidine and aromatic amino acids because they are not shared in other pathways. However, they can be considered uniquely defined modular units because of the existence of enzyme gene clusters. It is interesting to note that the so-called essential amino acids that cannot be synthesized in human and other organisms generally appear in these extensions. Furthermore, the bottom extension of basic amino acids appears to be most divergent containing multiple pathways for lysine biosynthesis and multiple gene sets for arginine biosynthesis.

Fig. 2 shows only the pathways that are relevant to amino acid biosynthesis. What constitutes the core part of the entire metabolic network and how it has evolved would require more detailed analyses of the central energy metabolism in relation to diverse environmental conditions in which various organisms inhabit. The increasing amount of genome sequences and metagenome sequences, together with the accumulated knowledge of metabolism as represented in KEGG pathway maps, will enable such analyses to be performed.