Integrated multi-omics unveil the impact of H-phosphinic analogs of glutamate and α-ketoglutarate on Escherichia coli metabolism

Abstract

Desmethylphosphinothricin (L-Glu-γ-P_H) is the H-phosphinic analog of glutamate with carbon-phosphorus-hydrogen (C-P-H) bonds. In L-Glu-γ-P_H the phosphinic group acts as a bioisostere of the glutamate γ-carboxyl group allowing the molecule to be a substrate of Escherichia coli glutamate decarboxylase, a pyridoxal 5'-phosphate-dependent α-decarboxylase. In addition, the L-Glu-γ-P_H decarboxylation product, GABA-P_H, is further metabolized by bacterial GABA-transaminase, another pyridoxal 5'-phosphate-dependent enzyme, and succinic semialdehyde dehydrogenase, a NADP⁺-dependent enzyme. The product of these consecutive reactions, the so-called GABA shunt, is succinate-P_H, the H-phosphinic analog of succinate, a tricarboxylic acid cycle intermediate. Notably, L-Glu-γ-P_H displays antibacterial activity in the same concentration range of well-established antibiotics in E. coli. The dipeptide L-Leu-Glu-γ-P_H was shown to display an even higher efficacy, likely as a consequence of an improved penetration into the bacteria. Herein, to further understand the intracellular effects of L-Glu-γ-P_H, ¹H NMR-based metabolomics, and LC-MS-based shotgun proteomics were used. This study included also the keto-derivative of L-Glu-γ-P_H, α-ketoglutarate-γ-P_H (α-KG-γ-P_H), which also exhibits antimicrobial activity. L-Glu-γ-P_H and α-KG-γ-P_H are found to similarly impact bacterial metabolism, although the overall effect of α-KG-γ-P_H is more pervasive. Notably, α-KG-γ-P_H is converted intracellularly into L-Glu-γ-P_H, but the opposite was not found. In general, both molecules impact the pathways where aspartate, glutamate, and glutamine are used as precursors for the biosynthesis of related metabolites, activate the acid stress response, and deprive cells of nitrogen. This work highlights the multi-target drug potential of L-Glu-γ-P_H and α-KG-γ-P_H and paves the way for their exploitation as antimicrobials.

Keywords: LC-MS; NMR; desmethylphosphinothricin; glutamate; multi-omics; multi-target drug; nitrogen starvation; phosphinates.

Figure 1

Structural formula of phosphinates.L-phosphinotricin (L-PT) and the natural tripeptides that derive from it (Bialaphos and Phosalacine) are shown along with the H-phosphinic compound L-Glu-γ-P_H (L-desmethylphosphinotricin) and its derivatives obtained either by enzymatic decarboxylation (GABA-P_H), followed by transamination (SSA-P_H) and oxidation (succinate-P_H), or by either transamination or oxidative deamination (α-KG-γ-P_H).

Figure 2

Univariate analysis of the 20 most abundant and significantly affected metabolites. The metabolites include those most abundant in the control group (cyan) whose concentration is significantly altered in the L-Glu-γ-P_H-(red) and/or α-KG-γ-P_H (green) treated cells. Data are plotted as box-and-whisker plot. Values above/below 3.5 times the interquantile range were considered outliers and excluded. Statistical analysis was performed with one-way ANOVA or Kruskal-Wallis’s test, and p-value was adjusted with the post hoc Tukey’s HSD test or Mann-Whitney U’s test (with the Bonferroni’s correction). p > 0.05 (ns); p ≤ 0.05 (∗); p ≤ 0.01 (∗∗); p ≤ 0.001 (∗∗∗); p ≤ 0.0001 (∗∗∗∗).

Figure 3

PCA, PLS-DA and VIP scores of the intracellular metabolites identified by¹H NMR.A, PCA and B, PLS-DA scores plot depicting the projection of control (n = 10, cyan triangle), and of L-Glu-γ-P_H (n = 9, red cross) and α-KG-γ-P_H (n = 10, green X) treated samples along PC1 and PC2 (A), or (B). Shaded ellipse areas indicate the 95% confidence interval based on the data points for individual groups. The letters in the score plots represent the biological replicates. Control: A, B, G, H, AE, AF, AM, AN, AO, AP; L-Glu-γ-P_H: C, D, I, Q, R, Z, AA, AG, AH; α-KG-γ-P_H: E, F, M, N, S, T, AB, AC, AI, AL. In PLS-DA, the two dashed-ellipse areas of the L-Glu-γ-P_H treated samples define the two sub-clusters. Q2 of first component is 0.92057, while Q2 of the second component is 0.93525. The permutation test value (p) is 0.0015. C, most influent metabolites ranked according to the VIP scores of the first component of the PLS-DA model. Only metabolites with a VIP score > 1 are reported.

Figure 4

Volcano plots and Gene Ontology (GO) enrichment analyses. The graphs illustrate differentially abundant proteins and most affected pathways upon exposure to L-Glu-γ-P_H or α-KG-γ-P_H. A and B, Volcano plots of all proteins identified in the proteome database when cells were challenged with L-Glu-γ-P_H (A) or α-KG-γ-P_H (B). Green points: differentially expressed proteins that were significantly downregulated (T-fold < −1.5; p < 0.05). Red points: differentially expressed proteins that were significantly upregulated (T-fold > 1,5; p < 0.05). Proteins whose T-fold is > 8 or < −8 are text-labelled. C and D, Bar charts showing the GO terms for biological process ranked by fold enrichment of cells exposed to L-Glu-γ-P_H (C) or α-KG-γ-P_H (D). Green bars: significant downregulated pathways (p < 0.05). Red bars: significant upregulated pathways (p < 0.05).

Figure 5

Vienn diagrams and Heatmap visualization illustrate differentially abundant proteins upon exposure to L-Glu-γ-P_Hor α-KG-γ-P_H. A and B, values in the Venn diagrams indicate the number of proteins that were upregulated (A) or downregulated (B) when cells were treated with L-Glu-γ-P_H or α-KG-γ-P_H. Numbers in the overlapping circles refer to proteins that were found to be co-upregulated (A) or co-downregulated (B) by both compounds. C, Heatmap visualization for the top 60 proteins based on the analysis of variance (ANOVA) test. Biological replicates (horizontal axis) and single proteins (vertical axis) are separated using a hierarchical clustering based on Euclidean distance.

Figure 6

Integratedprotein-metabolitepathway enrichment analysis. The integrated protein-metabolite pathway enrichment analysis of cells exposed to L-Glu-γ-P_H (A) or α-KG-γ-P_H (B) was performed versus the control group. The analysis was carried out using the joint pathway module from MetaboAnalyst (34) in order to integrate the results obtained from the combined proteomics and metabolomics studies performed under the same experimental conditions. The hypergeometric test was used for enrichment analysis, while the degree centrality (which considers the number of links that connect a node) was used for the measure of the topology. Moreover, as an integration method of proteomics and metabolomics data, the combine queries option was used. The database utilized was KEGG. Results were plotted as a function of impact within the pathway and significance. Labeled pathways have a pathway impact index ≥ 1.0 and p-value (FDR) < 0.05. Pathways labeled in green are those shared by (A) and (B). In both graphs, the pathway with the greatest impact was Novobiocin biosynthesis but, due to its low p-value, -log₁₀(p) < 2, it was not labeled.

Figure 7

The increment in NH₄⁺in EG medium improves the fitness of the cells when exposed to L-Glu-γ-P_Hand α-KG-γ-P_H. At sub-inhibitory concentrations, L-Glu-γ-P_H (3,5 μg/ml; A) and α-KG-γ-P_H (8.5 μg/ml, B) delay cell growth in EG medium pH 7.0 (red squares) compared to the untreated group (black dots). In the same experimental conditions, a 2-fold increase in the ammonium content (2X NH₄⁺ EG) mitigates the inhibitory effect of L-Glu-γ-P_H (A) and α-KG-γ-P_H (B) on cell growth (green rhombus) compared to the untreated group (brown dots). No significant NH₄⁺-related effect is observed if the untreated cells are grown in 2× NH₄⁺ EG (brown dots) compared to EG (black dots). Data, shown as mean ± sd, are the averages of at least three independent experiments.