Preferential attachment in the evolution of metabolic networks

Sara Light, P. Kraulis, Arne ElofssonBMC 2005 · DOI

Abstract

Here, we have found indications of preferential attachment in the metabolic network of E. coli. A possible biological explanation for preferential attachment growth of metabolic networks is that novel enzymes created through gene duplication maintain some of the compounds involved in the original reaction, throughout its future evolution. In addition, we found that enzymes which are candidates for horizontal gene transfer have a higher average connectivity than other enzymes. This indicates that while new enzymes are attached preferentially to highly connected enzymes, these highly connected enzymes have sometimes been introduced into the E. coli genome by horizontal gene transfer. We speculate that E. coli has adjusted its metabolic network to a changing environment by replacing the relatively central enzymes for better adapted orthologs from other prokaryotic species.

Introduction

Many biological systems, from protein interactions to the internet, exhibit "scale-free" architectures. These networks are characterized by a few highly connected "hubs" and a multitude of nodes with sparse connections. A leading theory for the emergence of such structures is preferential attachment: as a network expands, new nodes are more likely to link to existing nodes that already possess a high degree of connectivity—a "rich-get-richer" phenomenon.

While mathematically compelling, demonstrating preferential attachment in biological evolution is challenging. In metabolic networks—the intricate web of chemical reactions sustaining life—researchers sought to determine if growth follows this pattern. If so, a clear correlation should exist between the evolutionary age of an enzyme and its connectivity: the oldest enzymes should be central hubs, while newer additions remain peripheral.

Methodology: Mapping the Metabolic Network

To test this hypothesis, the metabolic network of Escherichia coli was modeled as a "reaction graph" using data from EcoCyc and KEGG. Unlike standard metabolic maps where metabolites are nodes, this study utilized an enzyme-centric approach to better reflect evolutionary changes in the proteome.

Comparison of common metabolic representations versus the enzyme-centric reaction graph used in this study.

In this representation:

  • Nodes: Represent enzymes (identified by EC numbers).
  • Edges: A connection is drawn between two enzymes if they share a metabolite (e.g., Enzyme A produces a compound that Enzyme B consumes).

To estimate enzyme "age," researchers analyzed their distribution across 163 organisms, spanning Eukaryotes, Archaea, and Bacteria. Enzymes were categorized into five groups ranging from "Ancient" (found in all three domains) to "Very Recent" (found only in close relatives of E. coli).

Phylogenetic distribution of organisms used to determine the evolutionary age of E. coli enzymes.

Key Findings

The Age-Connectivity Correlation

The analysis revealed a striking relationship: older enzymes are significantly more connected. "Ancient" enzymes (Group 1) exhibited an average connectivity approximately 40% higher than the most recent additions (Group 5). Statistical significance was confirmed via Z-scores calculated against 100,000 randomized network versions, ensuring the trend was not due to chance.

Evidence of Preferential Attachment

By comparing a reconstructed "ancient" core network to the modern E. coli network, the authors observed that new edges are added to highly connected enzymes at a significantly faster rate. This linear relationship ($r = 0.87$) between initial connectivity and the rate of new connection acquisition provides direct evidence for preferential attachment.

The Role of Horizontal Gene Transfer (HGT)

A surprising finding was that enzymes identified as candidates for HGT were more connected than average (mean connectivity $k = 42.2$ vs. $33.1$ for non-HGT enzymes). This suggests that E. coli does not just add peripheral genes via HGT but may "plug in" highly efficient enzymes from other species directly into its central metabolic hubs to adapt to environmental shifts.

Biological Mechanisms and Implications

The primary driver for this growth pattern is likely gene duplication. When a gene for a central enzyme duplicates, the new variant often initially interacts with the same substrates and products, inheriting the original enzyme's high connectivity instantly.

However, the study notes limitations. Data completeness, especially for Archaea, can introduce noise into phylogenetic groupings. Furthermore, "promiscuous" metabolites like water or ATP, which connect almost everything, had to be filtered out (top 15–40 compounds) to reveal the underlying network structure.

In conclusion, the E. coli metabolic network evolves through a combination of preferential attachment and strategic horizontal integration, ensuring that ancient, central pathways remain robust while allowing for rapid adaptation to new ecological niches.

Summary updated

Abstract

Here, we have found indications of preferential attachment in the metabolic network of E. coli. A possible biological explanation for preferential attachment growth of metabolic networks is that novel enzymes created through gene duplication maintain some of the compounds involved in the original reaction, throughout its future evolution. In addition, we found that enzymes which are candidates for horizontal gene transfer have a higher average connectivity than other enzymes. This indicates that while new enzymes are attached preferentially to highly connected enzymes, these highly connected enzymes have sometimes been introduced into the E. coli genome by horizontal gene transfer. We speculate that E. coli has adjusted its metabolic network to a changing environment by replacing the relatively central enzymes for better adapted orthologs from other prokaryotic species.
BibTeX
@inproceedings{light2005preferential,
  author = {Sara Light and P. Kraulis and Arne Elofsson},
  title = {{Preferential attachment in the evolution of metabolic networks}},
  booktitle = {BMC},
  year = {2005},
  doi = {10.1186/1471-2164-6-159},
}
Figure