The 14 glycolytic enzymes in
E. coli interact with at least 237 unique proteins in at least 339 interactions. These proteins also involve 14 proteins of unknown function (
Table S1). Both data and results are summarized in
Table 5.
3.1. Glycolytic Enzymes Interact Very Specifically with Non-Glycolytic Proteins
The 14 enzymes have very distinct interaction patterns with some proteins having almost none and some proteins having dozens of interactions (
Figure 2A). For instance, Eno has 93 interaction partners including five uncharacterized proteins (UPs). By contrast, GpmB (predicted phosphoglycerate mutase) has only one interaction. All other enzymes had interaction numbers in between these extremes (
Table 1). Overall, 16 uncharacterized proteins interacted with glycolytic enzymes and one of these interactions, PykA:YggR (a putative pilus biogenesis ATPase), has been found in two studies (
Table 4). One UP (YdjL; a putative zinc-type alcohol dehydrogenase) was found to interact with two glycolytic enzymes, namely with enolase and GapA. Hence, the number of unique uncharacterized proteins in our dataset is 14.
Six proteins have uncharacterized interaction partners, namely Pgk (1 interactor), GpmA (1), GpmI (2), PykA (2), GapA (4), and Eno (5). These numbers are roughly proportional to their total numbers of interactions (Eno: 90 unique interactors, GapA: 56, GpmA: 13 GpmI: 4).
Among the interactions reported previously, again Eno and GapA had the most interaction partners with 13 and 10 interactors, respectively (
Figure 2B). The other enzymes have from 0 to seven reported interactions in the literature.
Figure 2.
Interacting partners of glycolytic enzymes in
Escherichia coli and their main sources. The 6 different color codes in the plot correspond to the 4 large-scale PPI studies and 7 small scale data sets (“other”,
Table 3). Uncharacterized proteins are shown in black. Evidence-based interactions have support from text-mining (see methods).
Figure 2.
Interacting partners of glycolytic enzymes in
Escherichia coli and their main sources. The 6 different color codes in the plot correspond to the 4 large-scale PPI studies and 7 small scale data sets (“other”,
Table 3). Uncharacterized proteins are shown in black. Evidence-based interactions have support from text-mining (see methods).
3.1.1. Glycolytic Enzymes Interact with Both Enzymatic and Non-Enzymatic Proteins
We wondered if glycolytic proteins interacted with other enzymes or rather with non-enzymatic proteins, such as catalytically inactive regulators. As a proxy for enzymatic activity, we used the EC number provided by Uniprot. Overall, glycolytic enzymes interact with both enzymes as well as non-enzyme proteins at similar ratios (46.8% enzymes vs. 47.2% non-enzymes with 5.9% uncharacterized,
Figure 3A).
For instance, enolase (Eno), the most promiscuous enzyme in glycolysis interacts with 49 enzymatic protein partners and 39 non-enzymatic protein partners as well as 5 uncharacterized proteins. All others had between 0 (GpmB) and 27 (GapA) enzyme interaction partners. Notably, PykF was the only glycolytic enzyme which had more non-enzyme partners than enzymes. The preference for enzymatic partners is even more pronounced among the literature-curated interactions (
Figure 3B): 73.9% of glycolytic interactors are enzymes vs. 26.1% being non-enzymes. However, three enzymes (FbaB, TpiA, and PykF) had more non-enzymatic partners than the others.
3.1.2. Glycolytic Enzymes Interact with Both Metabolic and Non-Metabolic Proteins
Given that many glycolytic enzymes interact with non-enzymes, we wondered how many interactors were involved in other metabolic pathways or in non-metabolic pathways. The metabolic functionality of a protein was determined using Gene Ontology (GO) Biological Process annotations (from Uniprot). If the interactor had at least 1 GO term mentioning “metabolism”, we considered it to be a metabolic protein. Note that the GO classification has a broad definition of “metabolism” so that processes like DNA repair are also considered “metabolic” (because they are involved in “DNA metabolism”).
Overall, 50.2% of glycolytic interactors were involved in metabolism, with carbohydrate GO metabolic terms like glucose metabolic process [GO:0006006]/carbohydrate metabolic process [GO:0005975]- being the most common with 27 interactors. Remaining 210 interactors showed a wide range of GO metabolic terms including ribosomal large subunit assembly (translation), ATP biosynthetic process, L-proline biosynthetic process (amino acid metabolism) etc. There were 25 proteins which did not show any GO metabolic terms (
Figure 3C,D).
For instance, Eno had 35 metabolic protein interactors, 43 non-metabolic protein interactors, 7 interactors which did not show any GO terms and 5 uncharacterized proteins. The most-common metabolic GO terms among enolase’s interactors were cellular response to DNA damage stimulus [GO:0006974] and mRNA catabolic process [GO:0006402]. Interestingly, enolase had more non-metabolic interactors, and the most over-represented (high-level) GO terms dealt with regulation of cell shape [GO:0008360], peptidoglycan biosynthetic process [GO:0009252] and cell wall organization [GO:0071555] (43 interactors).
The evidence-based analysis reflected the pattern we saw in the enzyme vs. non-enzyme distribution, with most enzymes having metabolic interactors. FbaB was the only glycolytic enzyme that had only non-metabolic interactors, namely DnaJ (chaperone/heat shock protein J).
Overall, more metabolic than non-metabolic interactors were found in the glycolytic interactome, as expected, but the substantial number of non-metabolic interactors indicates an extensive crosstalk between glycolysis and other processes in bacterial cells.
We used the KEGG database to determine the pathway characteristics of the interactors (
Figure 3E,F). Our analyses showed 22 proteins were part of a carbohydrate metabolic pathway, 94 interactors were part of a non-carbohydrate metabolic pathway and 121 interactors were not involved in any metabolic pathway including 14 uncharacterized proteins. For instance, Eno had nine interactors involved in carbohydrate metabolic pathways, 40 interactors involved in non-carbohydrate metabolic pathways, 42 had no metabolic pathways, and five were uncharacterized proteins.
3.1.3. Glycolytic Enzymes Interact Primarily with Non-Essential Genes
While glycolysis is an important central pathway, it is not absolutely essential, as long as cells can obtain energy from sources other than glucose, such as amino acids. However, if glucose is the only carbon source, glycolysis becomes an absolutely essential pathway, hence it is conditionally essential. We wondered if glycolytic enzymes interact with other essential proteins or whether these interactors are dispensable.
We evaluated if a protein is encoded by an essential gene using the Database of Essential Genes (DEG). Then, we counted the number of essential genes and non-essential genes among the interactors of glycolytic enzymes. Uncharacterized proteins are not essential, as expected. These proteins did not have an EC number (except 1 protein: YegV), and did not show metabolic terms from GO or KEGG (metabolic pathways).
In the raw data analysis, glycolytic enzymes generally interacted with more non-essential than essential proteins, and Pgi, GpmI and GpmB had
only non-essential proteins as interactors (
Figure 3G). As the most promiscuous protein, enolase shows a fairly representative pattern with 29 essential gene interactors, 59 non-essential gene interactors and 5 uncharacterized proteins. In the evidence-based analyses, the pattern was similar, although there were no uncharacterized interactors (
Figure 3H).
3.1.4. Glycolytic Enzymes Interact Primarily with Metabolic Proteins
Once we had identified interactions with enzymes and metabolic proteins, we wanted to differentiate these proteins further, especially with regard to metabolic pathways. Hence, we defined 4 types of proteins based on the aforementioned results (
Table 6): Metabolic Enzyme (ME): An interactor that is an enzyme (based on EC number) AND a metabolic protein (from a metabolic GO term) or part of a metabolic pathway (from metabolic pathway KEGG term). Metabolic Protein (MP): An interactor that is NOT an enzyme (EC number) AND is a metabolic protein (GO) or part of a metabolic pathway (KEGG). Non-Metabolic Enzyme (NME): An enzyme (EC) AND NOT a metabolic protein (GO) AND NOT part of a metabolic pathway (KEGG). Non-Metabolic Protein (NMP): A non-metabolic protein AND NOT an enzyme (EC) AND NOT a metabolic protein (GO) AND NOT part of a metabolic pathway (KEGG).
Using this classification, approximately 70% of the all the interactors of 14 glycolytic enzymes are parts of metabolism (
Figure 4).