Host-Microbe Co-metabolism Dictates Cancer Drug Efficacy in C. elegans

Abstract

Fluoropyrimidines are the first-line treatment for colorectal cancer, but their efficacy is highly variable between patients. We queried whether gut microbes, a known source of inter-individual variability, impacted drug efficacy. Combining two tractable genetic models, the bacterium E. coli and the nematode C. elegans, we performed three-way high-throughput screens that unraveled the complexity underlying host-microbe-drug interactions. We report that microbes can bolster or suppress the effects of fluoropyrimidines through metabolic drug interconversion involving bacterial vitamin B6, B9, and ribonucleotide metabolism. Also, disturbances in bacterial deoxynucleotide pools amplify 5-FU-induced autophagy and cell death in host cells, an effect regulated by the nucleoside diphosphate kinase ndk-1. Our data suggest a two-way bacterial mediation of fluoropyrimidine effects on host metabolism, which contributes to drug efficacy. These findings highlight the potential therapeutic power of manipulating intestinal microbiota to ensure host metabolic health and treat disease.

Keywords: 5-FU; C. elegans; E. coli; Keio; autophagy; cancer; chemical-genomics; co-metabolism; holobiont; nucleotide metabolism.

Copyright © 2017 The Author(s). Published by Elsevier Inc. All rights reserved.

Figures

Graphical abstract

Figure 1

Bacterial Activity Modulates Fluoropyrimidine Efficacy in C. elegans (A and B) Worms cultured on laboratory (A) and WT (B) bacterial strains show disparate responses to 5-FU. E. coli K-12 strains: BW25113, HT115, W3310, MG1665; E. coli B strains: E. coli B, OP50, BL21G; K-12/B hybrid: HB101. (C) Heat/UV treatment of E. coli impairs 5-FU action. (D) Fluoropyrimidine effects on worms are bacterial strain specific. (E and F) Bacterial growth (E) and bacterial sensitivity to 5-FU (F) do not correlate with 5-FU effects in worms. (G) PCA of metabolomics data for C. elegans and E. coli treated with 5-FU. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. See also Figure S1 and Table S1.

Figure 2

Chemical-Genomic Bacterial-Host Screens Identify Pathways for 5-FU Action Not Revealed by Bacterial-Drug Screens (A) 5-FU inhibits worm development on control (BW25113) but not mutant E. coli (Δupp). (B) Diagram of the three-way host-microbe-drug interaction screen. (C) Screen design and Venn diagram of biologically relevant hits. (D) 3D graph correlating effects of gene knockout on bacterial growth (x axis), effects of 5-FU on bacterial growth (y axis), and effects of knockout on worm growth inhibition by 5-FU (colored circles). Gray dashed fit line (correlation between 5-FU and knockout effects in bacteria) determines bacterial sensitivity to 5-FU (blue/green color gradient box). Error bars represent SD. (E) Venn diagram of E. coli sensitive/resistant hits (FDR <0.05) with C. elegans 5% top hits for 5-FU treatment. (F) KEGG (K) and EcoCyc (E) pathway enrichment for gene deletions, and their effects on C. elegans ranked by coverage. Knockouts with MIC >5 μM are hits. Violin plots display distribution of MIC values; Contour color, number of hits; interior color, pathway coverage. ∗FDR <0.05; ∗∗FDR <0.01. (G) Metabolic network between chorismate, one-carbon, and vitamin B6 metabolism based on screen results. See also Figure S2 and Table S2.

Figure 3

Pyridoxal-5-Phosphate Is a Key Cofactor for the Mediation of 5-FU Effects (A) The de novo (blue, E. coli) and salvage pathway genes (green, E. coli; red, C. elegans) of B6 metabolism. (B) Knockout of B6 de novo pathway enzymes in E. coli reduces 5-FU efficacy in worms. (C) E. coli B6 salvage pathway modulates the de novo pathway to regulate 5-FU effects on worms. (D and E) Supplementation of pyridoxal (PL) improves 5-FU efficacy in worms (D) and rescues bacterial B6 deficiency as measured by LC-MS/MS (E). Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. See also Figure S3. For statistics, see Table S3.

Figure 4

Vitamin B6 Acts in Concert with Glycine and Folate Metabolism to Mediate 5-FU Effects (A) Disruption of bacterial glycine and serine metabolism impairs 5-FU action. (B) B6 effects are mediated by the glycine cleavage system. (C and D) Disruption of bacterial folate metabolism (C) impairs 5-FU action in worms and alters folate homeostasis (D). DHF, dihydrofolate; THF, tetrahydrofolate; CH+-THF, 5,10-methenyl-THF; CH2-THF, 5,10-methylene-THF; CH3-THF, 5-methyl-THF; CHO-THF, 10-formyl-THF. Each metabolite is the ratio between the sum of the values for the different glutamate side chains (1–7) and the sum of all metabolites measured. (E) Disruption of glycine (ΔgcvP) or folate (Δfau) metabolism alters CH2-THF polyglutamylation. (F) Metformin impairs 5-FU action in worms fed OP50 but to a lesser degree on metformin-resistant strain OP50-MR. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. See also Figure S4. For statistics, see Table S4.

Figure 5

Bacteria Complement C. elegans Metabolically to Mediate 5-FU Effects (A) Diagram of the bioconversion experiment. (B) Pre-conversion of 5-FC by bacteria enhances drug effects on the host. (C) Control but not codA metabolize 5-FC and excrete 5-FU. BD-below detection in codA but not BW. (D) Diagram of bacterial (fluoro)pyrimidine metabolism. Dashed arrows, more than one reaction. (E) Opposite effects of salvage deoxyribonucleotide (Δtdk) and ribonucleotide (ΔuppΔudkΔudp) metabolism in 5-FU efficacy. (F) Bacterial de novo pyrimidine metabolism regulates the effects of 5-FO. (G and H) B6 deficiency (G) and PL supplementation (H) regulate 5-FU effects through bacterial ribonucleotide metabolism. (I) Bacterial conversion of 5-FU alters fluoropyrimidine profiles and availability in C. elegans. (J) Knockout of host umps-1 mediates drug effects on worm development only in the absence of bacterial conversion of 5-FO. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. See also Figure S5. For statistics, see Table S5.

Figure 6

Bacteria Regulate 5-FU Effects in C. elegans by Two Distinct Mechanisms (A) Bacterial ndk effects on the efficacy of 5-FU are independent of 5-FO bioconversion mediated by pyrE. (B) IC50 values for bacterial growth with 5-FU. (C and D) LC-MS/MS quantification of fluoropyrimidines in E. coli (C) and C. elegans (D) supplemented with 50 μM 5-FU. (E) Volcano plot of nucleotide metabolism of bacterial mutants. (F) Diagram of deoxynucleotide metabolism in E. coli. X = G or A. (G) Bacterial deoxyribonucleotide imbalance caused by Δndk and Δdcd improves 5-FU effects. (H) 5-FU alters folate metabolism homeostasis in embryos. Inset: effects of 5-FU on egg hatching of the analyzed samples. (I) Activation of autophagy (Lgg-1::GFP reporter) in embryos by 5-FU is dependent on bacteria. (J) Host ndk-1 is required for 5-FU-induced autophagy activation by bacterial deoxynucleotide imbalance. (K) Bacterial ndk effects on 5-FU efficacy are mediated by the host ndk-1 gene. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. See also Figure S6. For statistics, see Table S6.

Figure 7

5-FU Improves Survival and Reduces Germline Hyper-Proliferation of gld-1 RNAi Worms (A) upp and udk mediate the effects of 5-FU regardless of bacterial genetic background. (B and C) LC-MS/MS quantification of fluoropyrimidines in E. coli (B) and C. elegans (C) supplemented with 50 μM 5-FU. (D) Representative images of DAPI-stained hyperproliferative gonads of gld-1 RNAi worms. Images were rotated and aligned for ease of comparison. (E) 5-FU and 5-FO reduce tumor size in a bacterial-dependent manner. Tumor retraction, distance between gonad arms at midsection/distance between gonad loops. (F) 5-FU and 5-FO extends the lifespan of gld-1 RNAi worms when fed on control bacteria but not ribonucleotide mutants. (G) Diagram summarizing the effects of 5-FU on the C. elegans/E. coli holobiont. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. See also Figure S7. For statistics, see Table S7.

Figure S1

Effects of Fluoropyrimidines on Bacterial Growth and C. elegans Egg Hatching, Related to Figure 1 (A) C. elegans fed heat or UV-killed OP50 show increased resistance to 5-FU. MICControl = 2 μM; MICUV = 16 μM; MICHeat = 32 μM. (B–E) Fluoropyrimidines show varying efficacy on worms fed E. coli K-12/B hybrid HB101, K-12 BW25113, B OP50 and Comamonas aquatica DA1877. Response to 5-FO (B), FUdR (C), capecitabine (D) and 5-FC (E). (F and G) Effects of 5-FU on (F) laboratory and (G) wild-type bacterial growth in NGM broth. No significant changes in IC50 values were observed within bacterial strain serotype despite 8-fold changes in worm MICs (e.g., BW25113 versus HT115, Figures 1A, 2B, and 2F). Drug response curves were calculated using a log(inhibitor) versus response - variable slope (four parameter) model. C. aquatica grew poorly in NGM and was readily killed by 5-FU while E. cloacae grew the best on NGM with or without 5-FU (G), but both conferred low worm MICs (Figure 1B). (H and I) Effects of fluoropyrimidines on (H) BW25113 and (I) OP50 bacterial growth in NGM broth. Note that E. coli is remarkably resistant to growth inhibition by 5-FC but capable of modulating the pharmacodynamics of 5-FC in the worm hatching assays (E). (J) Rates of consumption of diverse bacterial strains by C. elegans over a period of 8 hr. Differences in bacterial consumption do not correlate with 5-FU efficacy. (K) Expression of worm genes involved in fluoropyrimidine metabolism does not correlate with bacterial-induced changes in drug efficacy. For example, increases in uridine monophosphate kinase (C29F7.3) and uridine phosphorylase (upp-1) expression for HB101-fed nematodes does not correlate with the reduced drug efficacy observed for this strain (Figures S1B–S1E) compared to OP50, BW25113 and C. aquatica. (L) Metabolite profiling in worms and E. coli, treated with 5-FU. BLOD = below level of detection. Only statistically significant changes in metabolite levels are displayed. (M) KEGG pathway enrichment analysis for metabolomics in E. coli and C. elegans. Metabolite concentration comparisons were made against control conditions in appropriate species. Grey indicates that enrichment is non-significant (p > 0.05), white – enrichment could not be estimated. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. For statistics see Table S1.

Figure S2

Chemical Genomic Bacterial Screens and Chemical Genomic Bacterial Host Screens, Related to Figure 2 (A) 5-FU inhibits development of worms fed control E. coli BW25113 but not Δupp mutants. This effect was reversed by gene complementation (pUpp). Worm size was obtained from fluorescence intensity measurements of the DA2123 worm strain. (B) Representative GFP fluorescence images of DA2123 worms from (A) illustrating the effects of 5-FU on worm size. (C) Cumulative distribution of C. elegans MIC values and variability (from 3 independent biological replicates) for 5-FU effects for each knockout with MIC > 5 μM. (D) Bacterial growth (OD600nm) of Keio knockouts in liquid NGM. Correlation between control and 50 μM 5-FU treatment. Linear fit (red) indicates general growth reduction by drug treatment (slope 0.44 < 1; p < 2 × 10−16). (E) Linear modeling applied for bacterial growth to determine antagonistic and synergistic hits in bacterial screen. Examples for WT, Δupp, ΔpaaF and ΔgcvA. Black arrows mark knockout effects in comparison to WT, purple arrows mark 5-FU treatment effect, gray bars indicate expected combined effect of knockout and 5-FU treatment. Significant interaction between knockout and 5-FU treatment is shown by green (antagonistic) and blue (synergistic) arrows. (F and G) Correlation between C. elegans MIC values and E. coli growth OD600nm in NGM (F) or NGM + 50 μM 5-FU (G). Linear regression fit and covariation ellipse are marked in red. Knockouts that exhibit significant interaction with 5-FU (FDR < 0.05) are color-coded green (resistant) and blue (sensitive). Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. For statistics, see Table S2.

Figure S3

Effects of Bacterial Vitamin B6 Metabolism on 5-FU Efficacy, Related to Figure 3 (A) Vitamin B6 pools are altered in ΔpdxH mutant E. coli measured by LC-MS/MS. PLP = Pyridoxal-5-Phosphate; PNP = Pyridoxine-5-Phosphate; PMP = Pyridoxamine-5-Phosphate; PL = Pyridoxal; PN = Pyridoxine; PM = Pyridoxamine; PA = Pyridoxic acid. (B) Media supplementation with 10 μM PL rescues 5-FU efficacy on worms fed ΔpdxJ bacteria to control levels (BW + PL = 0.25 μM versus BW = 0.25 μM, p = 0.779; ΔpdxJ + PL = 0.25 μM versus ΔpdxJ = 2 μM, p < 0.001; ΔpdxJ + PL = 0.25 μM versus BW = 0.25 μM, p = 1; ΔpdxJ + PL = 0.1 μM versus BW + PL = 0.1 μM, p = 1). (C) Inhibition of both B6 de novo and salvage pathways using a ΔpdxJΔpdxK double mutant decreases 5-FU efficacy which cannot be rescued by supplementation with 1 mM PN, PM or PL. (D) Efficient RNAi knockdown of the worm F57C9.1 gene, an ortholog of human PDXK. (E) Knockdown of F57C9.1 by RNAi does not alter worm responses to 5-FU upon supplementation with PL. (F) Gene expression levels by qRT-PCR of enzymes involved in the glycine cleavage system, folate and vitamin B6 metabolism in worms grown on ΔpdxJ bacteria. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. For statistics, see Table S3.

Figure S4

Bacterial Folate Metabolism Regulates 5-FU Action, Related to Figure 4 (A) Diagram of the chorismate and PABA biosynthetic pathways. (B) Disruption of chorismate and PABA biosynthesis impairs 5-FU action in worms. (C) Polyglutamylation profiles of detectable folate metabolites in control, ΔgcvP and Δfau mutants. No striking changes are observed for the majority of folates. (D) 5-FU treatment (50 μM) impairs OP50 growth, but not in the presence of metformin (50 or 100 mM). (E) 5-FU treatment (50 μM) impairs OP50 metformin-resistant strain (OP50-MR) growth regardless of metformin effects (50 or 100 mM). (F) Metformin effects on bacterial growth are antagonistic to 5-FU effects in OP50 but not OP50-MR. (G) Impairment of the folate cycle (Δfau) and the glycine cleavage system (ΔgcvP) in E. coli does not modulate folate metabolism in C. elegans. Each metabolite is the ratio between the sum of the values for the different glutamate side chains (1-7) and the sum of all metabolites measured. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. For statistics, see Table S4.

Figure S5

Bacterial Bioconversion of Fluoropyrimidines, Related to Figure 5 (A) Disruption of nucleobases (ΔuraA) and nucleosides import (ΔnupCΔnupG) reduces 5-FU efficacy. (B) Supplementation with 200 μM uridine impairs 5-FU effects in worms fed BW25113 by 320-fold but not the triple mutant ΔudpΔudkΔupp. (C) Supplementation with 200 μM orotate impairs 5-FU effects in worms fed BW25113. The effect is partially rescued when worms are fed ΔpyrE but not ΔpyrD mutant bacteria. (D) Supplementation with 200 μM thymidine improves 5-FU effects in worms fed Δtdk mutant bacteria but not control BW25113 or Δupp mutants. (E) Pre-conversion of 5-FO by WT bacteria enhances drug effects on the host. Note that incubation of 5-FO with ΔpyrE mutant bacteria fully abolishes bacterial-mediated effects on host metabolism. Thus, de novo nucleotide metabolism is the unique pathway for the conversion of 5-FO. (F) Vitamin B6 supplementation increases GFP expression in ndk, udk, and upp promoter reporter strains. (G) Disruption of bacterial vitamin B6 production impairs 5-FC bioconversion. (H) LC-MS/MS quantification of fluoropyrimidines in E. coli supplemented with 50 μM 5-FU. (I) LC-MS/MS quantification of fluoropyrimidines in C. elegans supplemented with 50 μM 5-FU. (J) LC-MS/MS quantification of fluoropyrimidines in E. coli supplemented with 50 μM 5-FU. (K) LC-MS/MS quantification of fluoropyrimidines in C. elegans supplemented with 50 μM 5-FU. (L) Developmental response to 5-FU is uniquely mediated by bacterial genotype (BW control or triple mutant ΔudpΔudkΔupp) but not host genotype (N2 Wild-type or umps-1(zu456)). Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. For statistics, see Table S5.

Figure S6

Bacterial Deoxynucleotide Metabolism Effects on 5-FU Efficacy, Related to Figure 6 (A) Effects of 5-FU on bacterial growth of sensitive ΔyjjG, resistant Δupp and neutral Δndk or Δdcd. Drug response curves were calculated using a log(inhibitor) versus response - variable slope (four parameter) model. (B) Supplementation with 200 μM and 2 mM uridine impairs 5-FU effects in worms fed ΔyjjG and Δndk mutants in a distinctive manner. (C) Heatmap of metabolite profiles for bacterial nucleotide metabolism mutants. Absolute metabolite levels shown by color scale, clustering done by Euclidean distance. (D) Difference in metabolite levels between Δndk (y axis) and ΔpyrE (x axis) mutants in comparison to BW. Color gradient shows Δndk/ΔpyrE logFC, indicating opposing/concordant effects of mutants on metabolite levels. Metabolites with significant changes in Δndk/BW and Δndk/ΔpyrE are labeled (See Table S6). (E) 5-FU does not alter folate metabolism homeostasis in whole worms at 5 and 20 μM. (F) LGG-1::GFP expression in embryos from DA2123 worms fed BW25113, UV-irradiated BW25113 and triple mutant ΔudpΔudkΔupp treated with 5-FU at various concentrations and supplemented with 200 μM uridine or thymidine. Uridine supplementation rescues 5-FU induction of autophagy while thymidine supplementation improves it in a bacterial dependent manner. (G) Increased resistance to 5-FU in the mismatch repair worm mutant msh-6(pk2504) is mediated by bacteria. Resistance to 5-FU by msh-6 mutant worms is only observed when fed on HT115 but not HT115(Δndk) mutant bacteria. (H) msh-6(pk2504) does not confer resistance to 5-FU compared to WT worms when fed BW25113 or Δndk bacteria. (I) Efficient knockdown of ndk-1 in worms by RNAi. Data are represented as mean ± SD. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001. For statistics, see Table S6.

Figure S7

5-FU Effects on Tumor Size and Survival of the Host Are Dependent on Bacterial Genotype, Related to Figure 7 (A) Supplementation with uridine and orotate (200 μM or 2 mM) impairs 5-FU effects in worms fed diverse bacterial strains. (B) Role of bacterial yjjG, udk, upp, and ndk in E. coli K-12 strains in the regulation of 5-FU effects on the host. (C) Role of bacterial yjjG, udk, upp, and ndk in E. coli B strains in the regulation of 5-FU effects on the host. OP50p refers to OP50 prototroph for uracil/uridine. (D) Representative 10x images of DAPI stained gld-1 RNAi whole worms treated for 4 days with 5-FU and fed control or Δupp mutant bacteria. Tumor retraction as measured by distance (A’ to B’)/(A to B). (E) 5-FU extends the lifespan of gld-1 worms when fed on BW25113 bacteria but not HB101 or HT115. (F and G) 5-FU extends the lifespan of gld-1 (F) and glp-1 (G) gain-of-function worms when fed on control bacteria, but not Δupp mutants. Data are represented as mean ± SD. For statistics, see Table S7.