ROS-based lethality of Caenorhabditis elegans mitochondrial electron transport mutants grown on Escherichia coli siderophore iron release mutants.

Caenorhabditis elegans consumes bacteria, which can supply essential vitamins and cofactors, especially for mitochondrial functions that have a bacterial ancestry. Therefore, we screened the Keio Escherichia coli knockout library for mutations that induce the C. elegans hsp-6 mitochondrial damage response gene, and identified 45 E. coli mutations that induce hsp-6::gfp We tested whether any of these E. coli mutations that stress the C. elegans mitochondrion genetically interact with C. elegans mutations in mitochondrial functions. Surprisingly, 4 E. coli mutations that disrupt the import or removal of iron from the bacterial siderophore enterobactin were lethal in combination with a collection of C. elegans mutations that disrupt particular iron-sulfur proteins of the electron transport chain. Bacterial mutations that fail to synthesize enterobactin are not synthetic lethal with these C. elegans mitochondrial mutants; it is the enterobactin-iron complex that is lethal in combination with the C. elegans mitochondrial mutations. Antioxidants suppress this inviability, suggesting that reactive oxygen species (ROS) are produced by the mutant mitochondria in combination with the bacterial enterobactin-iron complex.

Caenorhabditis elegans, like many nematode species, consumes bacteria, which supplies many nutritional needs. C. elegans in the laboratory consumes Escherichia coli, but in its natural habitat of rotting fruit, feeds on a diet of hundreds of bacterial species (1, 2). These diverse bacteria supply micronutrients to the animal such as vitamins and cofactors. The bacterial supply of such cofactors is so dependable that C. elegans is a heme auxotroph fully dependent on the bacteria it consumes to acquire this cofactor for many mitochondrial proteins in the electron transport chain (ETC) (3). Because many of the more than 1,000 eukaryotic nuclearly encoded proteins that localize to the mitochondria share a common ancestor with bacteria, we reasoned that other bacterial gene pathways that animals may depend upon may function for mitochondrial biogenesis or function. For example, E. coli mutations in the ETC cytochrome bo terminal oxidase A gene cyoA cause induction of a C. elegans mitochondrial unfolded protein response gene, hsp-6, in wild-type C. elegans (4). Other bacterial biosynthetic pathways on which the mitochondrion may still depend could be discovered using an animal reporter gene for mitochondrial dysfunction and a comprehensive bacterial collection of mutations. Therefore, we conducted a genome-wide screen for E. coli mutations that induce the C. elegans mitochondrial unfolded protein response (UPRmt) or affect the viability or growth of C. elegans. In this screen, we identified mutations in 45 E. coli gene disruptions, out of more than 4,000 gene disruptions tested, that strongly activate the mitochondrial unfolded protein response gene, hsp-6; 24 of these 45 mutations also slow C. elegans growth. Because each of the 45 E. coli mutations induce a mitochondrial stress response, we tested for genetic interactions between each of these E. coli mutant strains and a set of C. elegans mitochondrial mutant strains. We found that 4 E. coli mutations that affect the import of the iron siderophore enterobactin into the cytoplasm of E. coli or the retrieval of iron in the bacterial cytoplasm from the imported enterobactin bound to iron showed a dramatic synthetic lethality with a variety of mutations in C. elegans mitochondrial ETC components, which when grown on wild-type bacteria are viable. Because Fe(III) is insoluble in aerobic environments, bacteria produce siderophores to retrieve Fe(III). Enterobactin binds Fe(III) with a Km of 10−39 M; the enterobactin bound to Fe(III) is then retrieved from outside of the cell or periplasm by specific E. coli enterobactin::Fe receptors and the tightly bound Fe(III) is removed from enterobactin in the bacterial cytoplasm by a dedicated enterobactin esterase fes (5). E. coli mutations that disrupt enterobactin biosynthesis genes were not synthetic lethal with C. elegans mitochondrial mutants. In fact, E. coli double mutants defective for both enterobactin synthesis and enterobactin import, or for both enterobactin synthesis and Fe(III) removal, were no longer synthetic lethal with the C. elegans mitochondrial mutants. Thus, the production of enterobactin and a failure to remove the covalently bound iron from enterobactin is required for the toxic interaction between siderophore uptake or esterase mutant E. coli and C. elegans mitochondrial mutants. Antioxidants such as ascorbic acid or resveratrol strongly suppressed the inviability of C. elegans mitochondrial mutants grown on the E. coli enterobactin siderophore utilization or import mutants. We hypothesize that reactive oxygen species (ROS) produced by C. elegans mitochondrial mutations when combined with ferric-chelated enterobactin synergistically produce dramatically increased ROS and lethality. This toxicity of a siderophore bound to Fe(III) and variations in the ETC between organisms may be an evolutionarily selected feature of siderophores that selects against their expropriation by other microbes.

Results

To systematically identify bacterial mutations that may affect mitochondrial function, we fed wild-type C. elegans carrying a mitochondrial unfolded protein response chaperone gene hsp-6::gfp with individual bacterial mutant strains from the E. coli Keio collection library and screened for E. coli mutations that induce the mitochondrial unfolded protein response (UPRmt) and/or slow C. elegans development. The Keio library is ∼4,000 single-gene in-frame insertions of a kanamycin resistance cassette in the E. coli K12 BW25113 background (6). C. elegans hsp-6 encodes a mitochondrial matrix HSP70 chaperone that is specifically up-regulated by mutations or drugs that impair mitochondrial function and is a standard marker for the UPRmt (7). We inoculated each of the 4,000 Keio collection E. coli strains individually on the surface of an agar NGM C. elegans/E. coli growth agar, added about 30 larval stage 1 (L1) wild-type C. elegans carrying hsp-6::GFP animals, and screened for induction of hsp-6::GFP 48 h later using a dissecting scope fluorescence microscope. We also screened the same wells for slower than normal growth of the C. elegans strain. We identified 45 mutant E. coli strains that robustly induced hsp-6::gfp (Dataset S1 and ) mainly in the intestine. In the same screening wells observed the next day, we screened the same E. coli mutants again for a delay in C. elegans rate of development from L1 to adulthood. Twenty-four of these 45 E. coli mutants also slowed wild-type C. elegans growth (Table 1). All 24 of the E. coli mutants that cause growth delay also were retrieved in the set of 45 E. coli mutants that activated C. elegans hsp-6::gfp, suggesting that mitochondrial dysfunction is a major axis of bacterial/host nutritional interaction (Table 1). Surprisingly, none of the ∼4,000 viable E. coli Keio mutants screened failed to supply some essential micronutrient or macronutrient or produced a toxin that disallowed wild-type C. elegans growth. Thus, 45 E. coli gene disruptions out of 4,000 tested induce a C. elegans mitochondrial stress response, and one-half of these 45 E. coli mutations also slowed wild-type C. elegans growth. However, none of the 4,000 Keio mutant collection was missing any essential bacterial factor (such as heme) for C. elegans growth. Our screen would have easily detected such a lethal interaction. There are 303 E. coli Keio gene disruptions that are inviable and thus not easily tested for C. elegans genetic interactions.

Table 1.

Quantification of developmental delay phenotype in wild-type background and spg-7(ad2249) mutant background

	Wild-type C. elegans animals		spg-7(ad2249) mutant
	% Adults		% Adults
E. coli genotype	Day 3	Day 4	Day 4	Day 5
BW25113 (wild type)	100 ± 0 (244)	NA	100 ± 0 (195)	ND
fes	100 ± 0 (244)	NA	0 ± 0 (169)	0 ± 0 (224), L1-arrest
fepG	100 ± 0 (265)	NA	0 ± 0 (185)	0 ± 0 (246), L1-arrest
fepD	100 ± 0 (259)	NA	0 ± 0 (183)	0 ± 0 (255), L1-arrest
fepC	100 ± 0 (214)	NA	0 ± 0 (186)	0 ± 0 (247), L1-arrest
cyoD	9.8 ± 3.7 (221)*,†	83.2 ± 9 (236)	0 ± 0 (181)	64.6 ± 6.4 (244)
yafC	100 ± 0 (240)	NA	0 ± 0 (167)	60.6 ± 8 (256)
cyoB	0 ± 0 (233)†	59 ± 6.3 (233)	0 ± 0 (188)	66.6 ± 7.2 (247)
cyoA	0 ± 0 (256)†	69.4 ± 17.2 (227)	0 ± 0 (155)	77.1 ± 5 (235)
tap	83 ± 6.4 (251)*	98.2 ± 0.8 (238)	0 ± 0 (166)	78.3 ± 20 (226)
cpxR	0 ± 0 (249)†	69.2 ± 15.4 (233)	0 ± 0 (185)	80.7 ± 13.4 (235)
stfP	0 ± 0 (255)†	64 ± 8.4 (248)	0 ± 0 (149)	74.6 ± 10.9 (250)
mdoC	0 ± 0 (247)†	72.2 ± 11.7 (229)	0 ± 0 (156)	75.1 ± 9.1 (215)
ycfk	0 ± 0 (233)†	72.9 ± 17.4 (221)	0 ± 0 (181)	78.7 ± 7.7 (208)
yejL	100 ± 0 (260)	NA	0 ± 0 (209)	76.4 ± 12.3 (260)
yejK	100 ± 0 (210)	NA	0 ± 0 (182)	80.3 ± 6.4 (228)
allA	0 ± 0 (233)†	52.8 ± 14.1 (241)	0 ± 0 (172)	81.8 ± 4.6 (236)
ybjX	0 ± 0 (197)†	56.9 ± 15.4 (230)	0 ± 0 (178)	71.5 ± 14.8 (225)
yfaE	0 ± 0 (250)†	55.1 ± 13.2 (240)	0 ± 0 (184)	82.6 ± 3.8 (236)
ybaW	0 ± 0 (246)†	63.1 ± 9.5 (255)	0 ± 0 (192)	80.1 ± 8 (237)
ycfX	0 ± 0 (240)†	62.9 ± 8.9 (257)	0 ± 0 (192)	82.4 ± 10.3 (234)
ygbM	0 ± 0 (187)†	61.4 ± 12.8 (230)	0 ± 0 (200)	69 ± 5.5 (224)
ydaN	0 ± 0 (260)†	63.1 ± 7.9 (232)	0 ± 0 (194)	79.1 ± 1.6 (220)
ybiB	0 ± 0 (243)†	57.1 ± 15.5 (236)	0 ± 0 (199)	76.7 ± 5.9 (227)
mntR	100 ± 0 (264)	NA	0 ± 0 (135)	76.9 ± 9.8 (219)
yaaJ	100 ± 0 (228)	NA	0 ± 0 (159)	66.2 ± 16.4 (231)
yfaY	0 ± 0 (253)†	61.7 ± 15.6 (245)	0 ± 0 (188)	73 ± 18 (247)
yejB	0 ± 0 (228)†	60.2 ± 3.7 (251)	0 ± 0 (172)	66.4 ± 12.2 (233)
yfiB	0 ± 0 (234)†	59.8 ± 3.6 (250)	0 ± 0 (190)	75.1 ± 8.1 (233)
tar	100 ± 0 (242)	NA	0 ± 0 (179)	73.9 ± 4 (226)
yheT	0 ± 0 (252)†	64.6 ± 4.7 (234)	0 ± 0 (154)	77.7 ± 7.5 (233)
yaaU	100 ± 0 (257)	NA	0 ± 0 (196)	80.4 ± 7.1 (251)
aroH	100 ± 0 (231)	NA	0 ± 0 (172)	76.8 ± 4 (235)
aes	0 ± 0 (232)†	63.9 ± 5.3 (239)	0 ± 0 (181)	76.3 ± 2.1 (234)
cyaA	100 ± 0 (257)	NA	0 ± 0 (165)	77.6 ± 2 (223)
hlyE	100 ± 0 (227)	NA	0 ± 0 (159)	75.1 ± 3.4 (237)
yceF	0 ± 0 (238)†	72.47 ± 3.1 (236)	0 ± 0 (193)	78.2 ± 2 (252)
wbbI	0 ± 0 (236)†	63.5 ± 8.8 (239)	0 ± 0 (183)	69.7 ± 12 (253)
ssuE	100 ± 0 (244)	NA	0 ± 0 (195)	77 ± 6.5 (232)
cspH	100 ± 0 (270)	NA	0 ± 0 (163)	71 ± 6.6 (230)
yaiF	100 ± 0 (251)	NA	0 ± 0 (198)	75.5 ± 6.4 (236)
rhsD	100 ± 0 (255)	NA	0 ± 0 (169)	82.6 ± 0.65 (236)
rhsB	100 ± 0 (265)	NA	0 ± 0 (182)	80.4 ± 5.2 (244)
ydcC	0 ± 0 (240)†	67.7 ± 15.7 (239)	0 ± 0 (183)	67.7 ± 13.2 (246)
cspD	100 ± 0 (215)	NA	0 ± 0 (137)	77.8 ± 7.4 (229)
eutD	100 ± 0 (256)	NA	0 ± 0 (165)	78.4 ± 4.6 (242)

Synchronized L1 wild-type animals were inoculated onto NGM media plates seeded with individual Keio E. coli mutant strains and incubated at 20 °C. The number of adult animals and the total number of animals were counted on day 3 and day 4 of feeding wild-type animals on Keio E. coli mutant strains. Data from 3 independent trials were collected; the average of percentage of adults and the SD are shown. The total number of animals counted in all 3 independent trials is shown in parentheses. Boldface indicates L1-arrest phenotype was observed.

Compared to wild-type animals fed on BW25113 on day 3, unpaired t test, P < 0.0001.

Since there were no adults or only few adults on these plates, the plates were scored on day 4 for the presence of adults. While there were no adults on day3, many of the wells had animals that reached adulthood on day 4. NA, not applicable since all of the animals were adults on day 3 itself. For columns 4 and 5, synchronized L1 spg-7(ad2249) mutant were inoculated onto NGM media plates seeded with individual Keio E. coli mutant strains and incubated at 20 °C. The number of adult animals and the total number of animals were counted on days 4 and 5 of feeding spg-7(ad2249 animals on Keio E. coli mutant strains. Data from 3 independent trials were collected; the average of percentage of adults and the SD are shown. The total number of animals counted in all 3 independent trials is shown in parentheses.

To verify the C. elegans UPRmt induction, we showed that the 45 mutant E. coli strains that activate hsp-6::gfp, also activate hsp-60::gfp, another marker for the C. elegans mitochondrial UPRmt (Dataset S1). ATFS-1 is a transcription factor required for the activation of UPRmt genes, such as hsp-6 and hsp-60. We tested each of the Keio E. coli mutants on a C. elegans atfs-1(tm4525); hsp-60::GFP C. elegans strain and found that all 45 bacterial mutants failed to induce the hsp-60::GFP UPRmt in the atfs-1(tm4525) background (Dataset S1). Thus, these E. coli mutants activate the C. elegans UPRmt via the expected ATFS-1 transcription factor. The C. elegans stress response induced by the 45 E. coli Keio mutants is specific for mitochondria because none of the 45 E. coli mutants caused activation of hsp-4::gfp (Dataset S1), a reporter of the unfolded protein response in the endoplasmic reticulum (UPRer) (8). Similarly, the 45 E. coli Keio mutants did not activate pgp-5::gfp (Dataset S1), a reporter of a translational stress response (9).

Because each of these 45 E. coli mutants induce the C. elegans mitochondrial-unfolded response, we examined whether any of these E. coli mutants strongly interact with C. elegans mutations in nuclear-encoded mitochondrial proteins. C. elegans RNA interference screens have shown that many mitochondrial gene inactivations are lethal (10). However, viable reduction-of-function mutants have been characterized (11). We tested the viable mitochondrial membrane protease spg-7(ad2249) mutant (12) and the viable electron transport point mutations nduf-7(et19), isp-1(qm150), clk-1(qm30), gas-1(fc21), nduf-2.2(ok2397), nuo-6(qm200), mev-1(kn1), and ucr-2.3(pk732) for synthetic lethal interactions with the 45 E. coli mutations that induce C. elegans hsp-6::gfp. Dramatically, 4 of the 45 Keio E. coli gene disruptions that induce the C. elegans hsp-6 mitochondrial response gene in wild-type C. elegans caused a dramatic developmental arrest in spg-7(ad2249), nduf-7(et19), isp-1(qm150), or clk-1(qm30) C. elegans mutants: Δfes::kan, ΔfepD::kan, ΔfepG::kan, ΔfepC::kan mutants (Table 1 and ). For example, Δfes::kan feeding induces a highly penetrant and dramatic developmental arrest with C. elegans spg-7(ad2249), while wild-type C. elegans fed on Δfes::kan grow normally (Fig. 1).

Fig. 1.

E. coli mutations that affect enterobactin siderophore utilization or import are synthetic lethal with C. elegans mitochondrial mutations. (A) spg-7(ad2249) animals grown on Δfes::kan E. coli arrest as L1 larvae. (Scale bar, 1,000 μm.) (B) Operon structure of genes in the enterobactin pathway. (C) Diagrammatic representation of enterobactin biosynthetic pathway in E. coli.

Each of the 4 E. coli gene disruptions that cause a synthetic arrest with C. elegans mitochondrial mutant strains mediates steps in the retrieval of iron from the E. coli siderophore enterobactin. Enterobactin is a high-affinity iron siderophore produced by many Gram-negative bacteria, including E. coli (5, 13). E. coli produce enterobactin under iron deficiency via the proteins encoded by the entCDEBAH operon (Fig. 1 ). Enterobactin is secreted into the environment, the soil or in a cell, where it binds to iron, forming a complex that is imported into E. coli through the transporters located in the bacterial outer membrane (13) (Fig. 1). The ferric enterobactin complex is transported into the cytoplasm via the fepDGC ATP-binding cassette transporter (14, 15). FepB encodes a periplasmic protein that binds ferric enterobactin and directs it to inner membrane transporters, FepD and FepG, which pump it into the bacterial cytoplasm. FepC assists FepD and FepG in transporting ferric enterobactin from the periplasm into the cytoplasm (Fig. 1) (16). Fes encodes enterobactin esterase that catalyzes the hydrolysis of both enterobactin and ferric enterobactin. In the E. coli cytoplasm, Fes hydrolyses enterobactin to release iron and linearized 2,3-dihydroxy-N-benzoyl-l-serine trimer as well as the dimer and monomer (Fig. 1) (17–19). Genetic disruption of E. coli fes causes accumulation of the unhydrolyzed ferric enterobactin in the bacterial cytoplasm (20, 21), whereas genetic disruption fepDGC causes accumulation of the ferric enterobactin in the E. coli periplasm where the cytoplasmic Fes esterase cannot release iron from its nearly covalent bond in ferric enterobactin (14, 15) (). This complex enterobactin synthesis and import biology is not parochial to E. coli: Siderophores related to enterobactin are used across bacterial phylogeny, and the hijacking of iron-loaded siderophores produced by one species and taken up by another bacterial species is a key feature of the iron competition landscape (22, 23).

We analyzed the synthetic lethal interaction between E. coli enterobactin iron release mutations and C. elegans mitochondrial mutations in more detail. Based on the morphological features, animal size measurements, and the germline development stage, Δfes::kan feeding in spg-7(ad2249) mutant causes an L1 arrest (). C. elegans nduf-7(et19), isp-1(qm150), and clk-1(qm30) mutant animals grown on the E. coli Δfes::kan mutant also dramatically arrest as L1 larvae (). nduf-7(et19) is a partial loss-of-function mutation in nduf-7 (NADH-ubiquinone oxidoreductase Fe-S), a subunit of the mitochondrial ETC complex I (11, 24). nduf-7(et19) mutant animals strongly activate the UPRmt and have a reduced respiration rate and longer life span (24). isp-1 encodes a Rieske iron sulfur protein that functions in the cytochrome b-c1 complex III subunit of the mitochondrial respiratory chain (11). isp-1(qm150) mutants have lower oxygen consumption, decreased ROS production, and increased life span (25). clk-1 encodes an ortholog of the highly conserved demethoxyubiquinone hydroxylase that is necessary for ubiquinone biosynthesis (11). Mutations in C. elegans clk-1 cause slow development, reduced respiration, and increased life span when grown on an E. coli strain that is competent to produce bacterial ubiquinone that is identical except for a shorter isoprenoid chain that localizes this electron carrier to the membrane (26).

Not all C. elegans mitochondrial mutations were synthetic lethal with the E. coli enterobactin iron release mutations: The C. elegans mitochondrial mutants gas-1(fc21), nduf-2.2(ok2397), nuo-6(qm200), mev-1(kn1), and ucr-2.3(pk732) did not arrest when grown, for example, on the E. coli Δfes::kan mutant (Table 2 and ). Because spg-7, nduf-7, isp-1, and clk-1 mediate distinct steps in mitochondrial respiratory pathways, the arrest induced by feeding mitochondrial mutants on E. coli Δfes::kan mutant is likely due to a toxic interaction between the mitochondrial homeostasis induction (one aspect of which is hsp-6 expression) induced by the spg-7, nduf-7, isp-1, and clk-1 mutations [but not by the gas-1(fc21), nduf-2.2(ok2397), nuo-6(qm200), mev-1(kn1), and ucr-2.3(pk732) mutations] and the iron toxicity caused by E. coli enterobactin mutants (Table 3).

Table 2.

Table showing list of the mitochondrial mutants tested to potential interaction with Δfes mutant

		% Larval arrest (N)
Animal strain	Description	Fed on BW25113	Fed on Δfes-722
Wild type		0 ± 0 (130)	0 ± 0 (115)
fzo-1(tm1133)	Mitofusin/FZO1	0 ± 0 (80)	0 ± 0 (95)
drp-1(tm1108)	Dynamin-related protein 1 (DRP1)	0 ± 0 (77)	0 ± 0 (93)
nduf-7(et19)	NADH dehydrogenase [ubiquinone] Fe-S protein 7 (NDUF7) of complex I mutated in Leigh syndrome	0.7 ± 1.2 (88)	99.2 ± 1.2 (147)
gas-1(fc21)	NADH dehydrogenase [ubiquinone] Fe-S protein 2 (NDUF2) of complex I	0.5 ± 0.9 (101)	0 ± 0 (119)
nduf-2.2(ok2397)	NADH dehydrogenase [ubiquinone] Fe-S protein 2 (NDUF2) of complex I	0 ± 0 (100)	0 ± 0 (120)
nuo-6(qm200)	NADH dehydrogenase [ubiquinone] 1 beta subcomplex subunit 4 (NDUFB4/B15) of complex I	0 ± 0 (98)	0 ± 0 (99)
mev-1(kn1)	Succinate dehydrogenase subunit (SDHC-1) of complex II	0 ± 0 (96)	0 ± 0 (81)
isp-1(qm150)	Ubiquinol-cytochrome C reductase, Rieske Fe-S protein (UQCRFS1) of complex III	1.2 ± 1.1 (92)	100 ± 0 (166)
ucr-2.3(pk732)	Ubiquinol-cytochrome C reductase core protein 2 (UQCRC2) of complex III.	0 ± 0 (99)	0 ± 0 (95)
clk-1(qm30)	COQ7; essential for ubiquinone biosynthesis	0 ± 0 (94)	100 ± 0 (128)

Data from 3 independent trials were collected; the average of percentage of L1 animals and the SD are shown. The total number of animals counted in all of the 3 independent trials is shown in parentheses. Boldface indicates strains that arrested as L1-larve when fed on E. coli Δfes mutant.

Table 3.

Feeding ferric enterobactin induces developmental arrest in C. elegans mitochondrial mutants

		% Larval arrest (N)
Fed on	Description	Wild type	spg-7(ad2249)
BW25113	Wild type	0 ± 0 (179)	0 ± 0 (219)
ΔfepC726::kan	Ferric enterobactin transport; ATP-binding protein fepC	0 ± 0 (132)	100 ± 0 (178)
ΔfepG727::kan	Ferric enterobactin transport system permease protein FepG	0 ± 0 (195)	100 ± 0 (196)
ΔfepD728::kan	Ferric enterobactin transport system permease protein FepD	0 ± 0 (198)	100 ± 0 (181)
ΔfepB730::kan	Ferrienterobactin-binding periplasmic protein	0 ± 0 (231)	0 ± 0 (174)
ΔfepA721::kan	Ferrienterobactin receptor	0 ± 0 (159)	0 ± 0 (188)
Δfiu-777::kan	Catecholate siderophore receptor Fiu	0 ± 0 (188)	0 ± 0 (164)
ΔcirA782::kan	Colicin I receptor; postulated to participate in iron transport	0 ± 0 (201)	0 ± 0 (190)
Δfes-722::kan	Enterochelin esterase	0 ± 0 (151)	100 ± 0 (205)
ΔentF724::kan	Enterobactin synthase component F	0 ± 0 (142)	0 ± 0 (165)
ΔentC731::kan	Isochorismate synthase EntC	0 ± 0 (166)	0 ± 0 (179)
ΔentE732::kan	Enterobactin synthase component E	0 ± 0 (202)	0 ± 0 (173)
ΔentB733::kan	Enterobactin synthase component B	0 ± 0 (191)	0 ± 0 (172)
ΔentA734::kan	2,3-Dihydro-2,3-dihydroxybenzoate dehydrogenase	0 ± 0 (183)	0 ± 0 (162)

Synchronized L1 wild type or spg-7(ad2249) mutant was inoculated onto NGM media plates seeded with individual Keio E. coli mutant strains and incubated at 20 °C. The number of L1-arrest animals and the total number of animals were counted on day 3.

To determine whether other E. coli genes required for enterobactin production cause the dramatic L1-synthetic larval arrest in the spg-7 C. elegans mitochondrial mutant, we tested each of the E. coli enterobactin pathway gene mutants on the C. elegans spg-7 mutant. Feeding C. elegans spg-7(ad2249) mutants on ΔentC::kan, ΔentD::kan, ΔentE::kan, ΔentB::kan, ΔentA::kan, and ΔentH::kan, which are defective in the production of enterobactin, does not cause L1-larval arrest in spg-7(ad2249) (Table 3). This suggests that it is not lack of enterobactin or E. coli iron starvation that causes developmental arrest in the C. elegans spg-7 mitochondrial mutant, but that instead it is the accumulation of iron-chelated enterobactin in the cytoplasm or periplasm of E. coli or iron-chelated enterobactin that is transported from E. coli to C. elegans cells with electron transport mutations that causes arrest of the C. elegans mitochondrial mutant. Such a model explains why it is the E. coli mutants that deliver ferric enterobactin, but not E. coli mutants that make no enterobactin that are toxic for animals with the spg-7, nduf-7, isp-1, and clk-1 mitochondrial dysfunctions. Inactivation of fes causes accumulation of the unhydrolyzed ferric enterobactin in the E. coli cytoplasm (20, 21), whereas inactivation in FepDGC causes accumulation of the ferric enterobactin in the E. coli periplasm (14, 15), which is inaccessible for cytoplasmic Fes to act upon (). As C. elegans consumes the Fes or Fep mutant E. coli, the ferric enterobactin may be unaltered by C. elegans esterases, which may not recognize this nearly covalent iron chelation. This ferric enterobactin may be transported into C. elegans cells, where if mitochondria are defective, perhaps producing ROS (), this ferric enterobactin is now toxic.

fes is in an E. coli operon with ybdZ and entF, which are also involved in enterobactin biosynthesis. Because bacterial insertion drug-resistant cassette mutations in one gene in the operon are often polar on downstream genes, we excised the Keio collection antibiotic resistance cassette insertion in Δfes::kan to create an in-frame nonpolar deletion allele of the single fes gene in the operon, which we call Δfes. Feeding spg-7(ad2249) mutant animals on Δfes E. coli also caused developmental arrest, suggesting that that the arrest phenotype is due to the absence of Fes protein, not polar effects on downstream genes (Fig. 2 ). To test whether the developmental arrest of the C. elegans mitochondrial mutants is due to ferric enterobactin, we inactivated enterobactin biosynthesis in the Δfes mutant background by constructing Δfes entA::kan and Δfes entB::kan double mutants. While 100% of C. elegans spg-7(ad2249) mutants arrest when grown on Δfes mutant E. coli, 0% of spg-7(ad2249) mutants arrest when grown the Δfes entA::kan or Δfes entB::kan double-mutant E. coli that does not generate enterobactin and therefore does not generate ferric enterobactin ( and Fig. 2 ). Furthermore, to test whether ferric enterobactin is required within E. coli to produce the arrest in spg-7(ad2249) mutants, we constructed Δfes cirA782::kan, Δfes fepA::kan, and Δfes fiu::kan double mutants (Fig. 2). FepA is an outer membrane protein that binds and transports ferric enterobactin into the periplasm of E. coli (27, 28). Fiu is an outer membrane protein that mediates uptake of dihydroxybenzoylserine, a breakdown product of enterobactin (29). Cir is another outer membrane protein that mediates uptake of ferric enterobactin and other breakdown products of enterobactin (29). Although FepA, Fiu, and CirA mediate import of ferric enterobactin into the bacterial periplasm, FepA is the major transporter necessary for the uptake of ferric enterobactin into bacterial periplasm. While 100% of C. elegans spg-7(ad2249) mutants arrest when grown on Δfes E. coli, Δfes cirA::kan, or Δfes fiu::kan double-mutant E. coli, 0% of spg-7(ad2249) mutants arrest in the Δfes fepA::kan double-mutant E. coli (Fig. 2). This result suggests that blocking uptake of ferric enterobactin from the environment into the E. coli can suppress the L1-larval arrest of C. elegans mitochondrial mutants grown on Δfes E. coli mutant. Mixing experiments with Δfes E. coli and wild-type E. coli showed that even nongrowing Δfes E. coli mutant could supply the enterobactin–iron for the toxic interaction ().

Fig. 2.

Enterobactin is necessary for the developmental arrest of C. elegans mitochondrial mutants. (A) spg-7(ad2249) animals grown on wild-type E. coli BW25113 develop normally, while spg-7(ad2249) mutant grown in Δfes mutant arrest as L1 larvae. In contrast, spg-7(ad2249) animals grown on Δfes ΔentA::kan double mutants that do not synthesize enterobactin develop normally. (Scale bar, 1,000 μm.) (B) Table showing suppression of larval arrest in animals grown on Δfes ΔentA::kan, Δfes ΔentB::kan, and Δfes ΔfepA::kan mutants. Data from 3 independent trials were collected; the average of percentage of adults and the SD are shown. The total number of animals counted in all 3 independent trials is shown in parentheses.

Because the E. coli enterobactin uptake and iron retrieval mutants activate a mitochondrial unfolded protein response even in a wild-type C. elegans that grows at the normal rate (Table 1), we assessed the effect of these E. coli enterobactin–iron release mutants on C. elegans mitochondrial morphology. The mitochondria in the hypodermis of wild-type animals grown on wild-type E. coli BW25113 have tubular or slightly elongated morphology, but the mitochondria of wild-type animals grown on Δfes E. coli are fragmented as assessed by nonyl acridine orange (NAO), a membrane potential-independent mitochondrial dye (Fig. 3). We also used a transgenic strain that expresses mitochondrial outer membrane protein TOMM-20 tagged with mRFP in the body wall muscles of C. elegans. The mitochondrial morphology is tubular and elongated in animals grown on wild-type E. coli BW25113, whereas wild-type C. elegans grown on Δfes E. coli show a more fragmented mitochondria in the body wall muscle (Fig. 3). Consistent with the abnormal mitochondrial morphology, ATP levels are significantly reduced in wild-type C. elegans grown on Δfes E. coli compared to those grown on wild-type E. coli BW25113 (Fig. 3 ). These data show that the induction of the hsp-6 in wild-type C. elegans grown on the E. coli iron–siderophore retrieval mutants accurately reports their moderate mitochondrial defects.

Fig. 3.

Mitochondrial structure and function are disrupted in wild-type animals grown on Δfes mutant E. coli. (A) Wild-type animals grown on Δfes mutant E. coli display fragmented hypodermal mitochondrial morphology as assessed using NAO. (Scale bar, 100 μm.) (B) Transgenic animals expressing TOMM-20::mRFP in the muscles grown on Δfes mutant E. coli display a fragmented mitochondrial morphology. (Scale bar, 100 μm.) (C) Animals grown on Δfes mutant E. coli have decreased luciferase luminescence, a proxy for ATP levels, compared to animals grown on E. coli BW25113 as assessed using a strain that expresses luciferase in all tissues. Unpaired t test; ***P < 0.001. Mean ± SD of n = 10. (D) ATP levels in animals grown on Δfes mutant E. coli display are decreased compared to animals grown on E. coli BW25113. Unpaired t test; ***P < 0.001. Mean ± SD of n = 3. (E) Antioxidant treatments with l-ascorbic acid, NAC, or transresveratrol suppress the L1-larval arrest induced by feeding Δfes mutants to C. elegans mitochondrial mutants. cis-Resveratrol and cis-trimethoxyresveratrol fails to suppress the arrest. Data from 3 independent trials were collected; the average of percentage of adults and the SD are shown. The total number of animals counted in all 3 independent trials is shown in parentheses. (F) Mouse lipocalin addition suppresses the larval arrest of spg-7(ad2249) animals grown on Δfes mutants. Data from 3 independent trials were collected; the average of percentage of adults and the SD are shown. The total number of animals counted in all 3 independent trials is shown in parentheses.

Mutations in isp-1, nduf-7, and spg-7 cause increased ROS production by the mitochondria (12, 24, 30). We found that spg-7(ad2249) animals that were grown on Δfes E. coli mutant, but mixed with antioxidants such as ascorbic acid, N-acetyl cysteine (NAC), resveratrol, or trolox were completely suppressed for the arrest normally caused by E. coli Δfes mutant fed to C. elegans spg-7 (Fig. 3). The fact that several other antioxidants were able to suppress arrest suggests that ROS is the likely cause of the toxic interaction between bacterial mutants that accumulate ferric enterobactin and the C. elegans mitochondrial mutants. Elevated ROS in isp-1(qm150) contributes to its increased longevity because the longevity increase is suppressed by NAC (30). UPRmt activation in the nduf-7(et19) is also suppressed by NAC (24). Consistent with this observation, we found that the L1-larval arrest induced by feeding Δfes E. coli to the nduf-7(et19) mutant is suppressible by NAC (). When we fed C. elegans spg-7(ad2249) animals the E. coli Δfes pregrown in the presence of transresveratrol and washed off the transresveratrol, the animals arrest, suggesting that the site of action of the antioxidants is likely within the animals ().

Because hydrogen peroxide is one of the major forms of endogenous ROS, we monitored in vivo hydrogen peroxide levels of the interaction between the E. coli Δfes mutations and C. elegans, using HyPer, a peroxide-specific sensor protein. Transgenic wild-type animals that express the HyPer sensor (jrIs1) (31) were grown either on E. coli BW25113 or E. coli Δfes and peroxide levels were measured. Endogenous peroxide levels were not significantly different between wild-type C. elegans grown E. coli BW25113 or E. coli Δfes; however, addition of exogenous peroxide results in significantly higher peroxide levels in animals grown on E. coli Δfes compared to animals grown on E. coli BW25113 (). We hypothesize that the exogenous peroxide interacts with ferric enterobactin and via the Fenton reaction to generate more peroxide.

In the evolutionary “tug-of-war” for iron, animals rely on the enterobactin-binding protein Lipocalin-2 to counteract iron loss to microbes that secrete siderophores (32). Secreted lipocalin 2 sequesters the ferric enterobactin to limit bacterial growth (32). We tested whether exogenous lipocalin-2 can suppress the arrest of spg-7(ad2249) grown on the E. coli Δfes mutant. We found that spg-7(ad2249) animals that were grown on the E. coli Δfes in the presence of 0.4 μM recombinant mouse lipocalin-2 no longer arrest as L1-larvae (Fig. 3).

Discussion

By screening a comprehensive collection of E. coli mutations, we identified 45 E. coli mutations that activate the C. elegans hsp-6 mitochondrial stress response gene. Surprisingly, 4 of these E. coli mutations that affect the retrieval of iron from the enterobactin siderophore are synthetic lethal in combination with multiple C. elegans mitochondrial mutants. These E. coli siderophore iron retrieval mutations also caused a mild disruption of mitochondrial morphology in wild type and the induction of C. elegans hsp-6 (Fig. 3 ). However, disruption of the production of E. coli enterobactin did not cause this synthetic lethal interaction with a C. elegans mitochondrial mutants and could even suppress the synthetic lethality of E. coli mutants defective for enterobactin siderophore utilization or import into the cytoplasm, suggesting that the production of enterobactin and binding to iron is required for the toxic interaction between the iron-loaded E. coli siderophore and C. elegans mitochondrial mutants. The bacterial literature on ferric enterobactin supports this model: fes mutants accumulate ∼50 times more intracellular ferric enterobactin than wild type (18), a fes mutant grows slowly on iron-limited media with and without addition of purified enterobactin (33), and an E. coli K-12 Δfes mutant has a growth deficit when grown on iron-rich medium and this defect was suppressed by a deletion in enterobactin synthesis gene entB (34).

We found that most but not all of the mitochondrial mutations we tested were synthetic lethal with the enterobactin iron retrieval mutants. It may be important to the genetic interaction with the E. coli siderophore iron release mutations whether the ROS is generated in complex I or complex III, and how accessible it might be to the iron-loaded enterobactin. The detailed cryo-EM structure of the entire animal mitochondrial ETC points to a highly intricate system of 8 proteins bearing iron sulfur cluster redox centers that pass electrons with quantum mechanical precision, and efficiently couple to movements of transmembrane segments of other coupled complex I proteins (there are 14 core proteins in complex I) to pump protons against a proton gradient (35). The many FeS clusters that mediate the movement of electrons along the many protein subunits of complex I are the likely source of ROS in the complex I mutants. Interestingly, from the set of ETC mutants that are synthetic lethal with the 4 E. coli siderophore iron retrieval mutants, nduf-7 encodes an ortholog of human NDUFS7 (NADH:ubiquinone oxidoreductase core subunit S7), which has a 4Fe4S cluster. The complex III ETC mutation isp-1 also shows a strong synergistic interaction with the E. coli siderophore mutations. ISP-1 is an ortholog of human UQCRFS1 (ubiquinol-cytochrome c reductase, Rieske iron-sulfur polypeptide 1) and has 2 iron, 2 sulfur clusters and ubiquinol-cytochrome-c reductase activity. Again, this is an iron sulfur center mutant protein that interacts with ubiquinone that shows such a strong synergy.

The C. elegans spg-7 point mutation causes a mitochondrial defect and is strongly synthetically lethal with the E. coli siderophore Fe retrieval mutations. spg-7 does not encode an ETC component but encodes a protease that is the ortholog of the E. coli FtsH protease. This protease has a small number of client proteins, which includes the iron sulfur assembly factor iscS and the iron sulfur protein lpxC and fdoH (36). C. elegans mitochondrial proteins from the nuclear genome that are probable clients of SPG-7 are likely to be orthologs of the E. coli clients of ftsH: The C. elegans iron sulfur assembly factor iscu-1 is a prime candidate for causing the iron sulfur protein defect in the ETC (36).

Similarly, the clk-1/mammalian coenzyme Q7 mitochondrial 5-demethoxyubiquinone hydroxylase that is synthetic lethal with the E. coli siderophore iron retrieval mutants fails to synthesize ubiquinone with its normally 10 isoprene unit hydrophobic tail in C. elegans. This null allele is not lethal to C. elegans because E. coli ubiquinone is imported into C. elegans cells. However, the E. coli ubiquinone, while competent for viability, has a shorter bacterial isoprene tail and causes slow growth of C. elegans, probably due to inefficient transfer of electrons from complex I to E. coli ubiquinone or from E. coli ubiquinone to complex III. The most likely explanation for the synergistic lethality of the C. elegans clk-1 mutation with the E. coli siderophore mutations is that the efficiency of E. coli ubiquinone interaction with complex I or complex III is compromised so that ROS to synergize with the iron loaded siderophore is generated (37).

Of the ETC subunits that do not interact with the E. coli siderophore mutations includes gas-1 and mev-1. gas-1 is orthologous to human NDUFS2 (NADH:ubiquinone oxidoreductase core subunit S2) and does not have an iron sulfur cluster. mev-1 is an ortholog of human SDHC (succinate dehydrogenase complex subunit C) and is a heme protein. mev-1 is a component of complex II, which does not participate in the NADH to oxygen redox/proton pumping cascade and would not be expected to generate ROS under normal redox conditions.

If the enterobactin iron retrieval E. coli mutants and the C. elegans mitochondrial electron transport mutations were simply synthetic lethal, we could not generate a molecular model for the interaction. However, the synthetic lethality was strongly suppressed by addition of free radical scavengers, pointing to a reactive oxygen-based toxicity (). We propose that ROS production at particular sites in the ETC is the key to the genetic interaction with E. coli siderophore mutations. Such localized production of ROS may not be detectable when measuring ROS levels in extracts; but the localized ROS within complex I or III could explain the inviability. Because endogenous ROS production is minimal in wild-type C. elegans, whereas mitochondrial mutants have increased ROS production, we propose that the triggering ROS in the C. elegans mitochondrial mutant animals when combined with ferric-chelated enterobactin induces a much larger ROS burst via the Fenton reaction to cause toxicity to C. elegans. In wild-type C. elegans, there may not be sufficient ROS to initiate the Fenton reaction. Nevertheless, the ferric chelated enterobactin that enters wild-type C. elegans does activate the mitochondrial stress response (Dataset S1 and ).

There is biochemical support for an interaction between E. coli enterobactin and C. elegans mitochondrial proteins: An enterobactin-biotin fusion binds to the C. elegans ATP synthase F1 α-subunit ATP-1 that is the ortholog of bacterial atpA (38). A 21-amino acid peptide of the ATP-1 protein binds enterobactin (38) and is highly conserved in the bacterial kingdom as well as in eukaryotes (). This region of ATP synthase F1 α is cytoplasmic but could interact with the FeS clusters of the ETC, which we have found to genetically interact with E. coli enterobactin iron retrieval mutants.

This interaction between a bacterial siderophore iron retrieval mutations and C. elegans ETC components that we have discovered and the biochemical interaction between this siderophore and ATP synthase (38) may be grounded in the complex ecology of bacterial siderophores. Siderophores allow bacteria to retrieve the vanishingly small concentrations of soluble Fe(III) in the environment, including in cells (22). While many bacterial species synthesize siderophores, unrelated bacteria hijack the iron-bound siderophores from those siderophore-synthesizing species to usurp their iron, a limiting essential micronutrient. However, some bacterial species with siderophore biosynthetic pathways have evolved siderophore biosynthetic modifications to append antibiotics to the siderophore, inhibiting the growth of these siderophore-iron thieves (22). In fact, enterobactin has been engineered to carry antibiotics as a sort of Trojan horse for antibiotic delivery (23). Our data show that iron-loaded enterobactin itself is a weapon against a C. elegans mitochondrial mutant; this genetic interaction may be based on an evolved toxic function for iron-loaded enterobactin and other bacterial species. If E. coli enterobactin–Fe is retrieved by diverse bacteria, it may interact with a diversity of electron transport oxidants and reductants that produce ROS, or the usurping bacteria may not encode the proper enzymes (such as Fes) to remove the Fe so strongly bound to multiple catecholamines of enterobactin; this toxic role of enterobactin–Fe that we discerned in a C. elegans mitochondrial mutant screen may be a highly evolved siderophore role to target the diverse components of bacterial and eukaryotic mitochondrial ETCs that compete with E. coli.

Materials and Methods

C. elegans culture, E. coli cultures, staining of mitochondria, microscopy, and imaging were performed as described previously (4). Removal of kanamycin cassette was performed as described previously (6). ATP measurements were performed as described previously (31). Statistical analysis was performed with GraphPad Prism 7 software. Student’s t test was used for analysis. Details of the genome-wide screening protocol is described in . Details of materials and methods used are available in . Details about E. coli and C. elegans strains used are also available in .