Siderophore-mediated zinc acquisition enhances enterobacterial colonization of the inflamed gut

Zinc is an essential cofactor for bacterial metabolism, and many Enterobacteriaceae express the zinc transporters ZnuABC and ZupT to acquire this metal in the host. Unexpectedly, the probiotic bacterium Escherichia coli Nissle 1917 exhibited appreciable growth in zinc-limited media even when these transporters were deleted. By utilizing in vitro and in vivo studies, as well as native spray metal infusion mass spectrometry and ion identity molecular networking, we discovered that Nissle utilizes yersiniabactin as a zincophore. Indeed, yersiniabactin enables Nissle to scavenge zinc in zinc-limited media, to resist calprotectin-mediated zinc sequestration, and to thrive in the inflamed gut. Moreover, we discovered that yersiniabactin’s affinity for iron or zinc changes in a pH-dependent manner, with higher affinity for zinc as the pH increased. Altogether, we demonstrate that siderophore metal affinity can be influenced by the local environment and reveal a mechanism of zinc acquisition available to many commensal and pathogenic Enterobacteriaceae.


INTRODUCTION
In addition to STm, other Enterobacteriaceae can thrive in the inflamed intestine. One EcN), a strain that was first isolated in WWI from the stool of a soldier who did not develop 99 gastroenteritis during a Shigella outbreak 28 . Since then, EcN has proven to be effective 100 in the treatment and prevention of intestinal disorders including chronic constipation, 101 ulcerative colitis, and infantile diarrhea 29-32 . Despite being used as a probiotic for nearly 102 a century, the mechanisms through which EcN exerts its protective effects are not 103 completely understood.

105
Our previous work has demonstrated that EcN utilizes multiple iron uptake systems and 106 secretes antimicrobial proteins known as microcins to outcompete and reduce STm 107 colonization in mouse models of gastroenteritis 3,33 . As a follow-up to these studies, we 108 initially sought to investigate the contribution of high-affinity zinc transporters to gut 109 colonization by EcN. Unexpectedly, we found that an EcN strain lacking ZnuABC and 110 ZupT was still able to grow appreciably in zinc-limited media, leading us to hypothesize 111 that EcN expresses an additional means of acquiring zinc. After a genome search did not 112 yield any promising candidate transporters, we hypothesized that EcN produces and 113 secretes an unknown zincophore. Using our recently developed native spray 114 metabolomics approach 34 , we subsequently discovered that EcN produces the 115 siderophore yersiniabactin (Ybt), which is capable of binding zinc. Moreover, we 116 demonstrate that EcN utilizes Ybt, in addition to the zinc transporters ZnuABC and ZupT, 117 to effectively acquire zinc in vitro, to resist the antimicrobial activity of CP, and to colonize 118 the inflamed gut. E. coli Nissle is more resistant to calprotectin-mediated zinc sequestration than S. 122 Typhimurium. 123 We have previously shown that multiple iron uptake systems enable EcN to colonize the 124 inflamed gut and to compete with STm 33 . As zinc is also limited in the inflamed gut, we 125 hypothesized that EcN must also have robust mechanisms for acquiring this metal. We 126 thus compared the growth of EcN to the growth of STm in a rich medium supplemented 127 with CP, a host antimicrobial protein that sequesters zinc and limits its availability to 128 microbes 15,25 . To this end, we employed CP concentrations (150-250 μg/ml) comparable 129 to those found in the inflamed gut 15 . EcN and STm grew to the same density after 16 h 130 of culture without the addition of CP, or in the presence of 150 μg/ml CP (Fig. 1a). 131 However, we noticed that in media containing CP at 250 μg/ml, EcN grew ~8 times better 132 than STm (Fig. 1a). These results indicated that EcN is more resistant than STm to the of CP, but not in the presence a Site I/II knockout mutant CP (MU CP; lacks the ability to 146 bind zinc) 36,37 , or when ZnSO4 was added to the media (Fig. 1b). These  Puzzlingly, we observed that the EcN znuA zupT mutant grew almost 1,000-fold better 151 than the STm znuA zupT mutant in the presence of 150 μg/ml CP (Fig. 1b). Although 152 higher concentrations of CP (250 μg/ml) reduced the growth of the EcN znuA zupT mutant, 153 it was still 100-fold higher than the STm znuA zupT mutant (Fig. 1c) Fig. 1a), we posited that EcN is able to acquire zinc via an additional mechanism. for decades that some siderophores can bind other metals besides iron (reviewed in 38 ).

163
Among the siderophores produced by EcN, Ybt has been shown to also bind copper, 164 gallium, nickel, cobalt, and chromium 39 . Intriguingly, a product of the Ybt gene cluster 165 has been proposed to contribute to zinc acquisition by the pathogen Yersinia pestis 40,41 ; 166 however, its identity and mechanism are unknown, as two prior studies did not provide 167 evidence of direct zinc binding by Ybt 39,41 . We thus sought to determine whether, in To this end, we deleted the ybt cluster's irp2 gene that encodes the synthetase HMWP2, 172 thus rendering EcN unable to synthesize Ybt 42-44 . We also deleted the ybtX gene, which 173 encodes for an inner membrane permease that was found to promote zinc acquisition by 174 Y. pestis 40,41 . Of note, the first published genome sequence of EcN (GenBank 175 CP007799.1, 45 ) indicated that irp1 and irp2 were disrupted (frameshifted and insertion 176 sequence, respectively), although a recent sequencing effort of our lab's EcN strain 177 revealed these genes to be intact (GenBank CP022686.1), which is consistent with a prior 178 study showing that EcN produces Ybt 46 . Next, we tested the growth of EcN strains lacking 179 these genes, in addition to the znuA zupT genes, in metal-limiting conditions (M9 minimal 180 medium). Strains lacking znuA zupT and either irp2 or ybtX displayed a severe growth 181 defect in M9 minimal medium, growing 1,000-fold less than EcN wild-type and more than 182 10-fold less than EcN znuA zupT (Fig. 1e). Furthermore, growth of all mutants was 183 restored in ZnSO4-supplemented M9 minimal medium (Supplementary Fig. 1b) and in 184 LB broth without metal limitation (Supplementary Fig. 1c), confirming that the observed 185 growth defects were indeed due to zinc deficiency. Taken together, these results 186 suggested that a product of the ybt gene cluster contributes to zinc acquisition by the 187 probiotic EcN in zinc-limited media. We therefore hypothesized that the Ybt locus may 188 encode for the production of a zincophore. Yersiniabactin is a zincophore. 9 To identify whether the ybt gene cluster produces a zincophore, we cultured EcN wild-  (Fig. 2a, b). Furthermore, two peaks 208 were observed in both culture supernatant (Fig 2c) and pellets from EcN wild-type that 209 were absent in the irp2 mutant cultured in M9 minimal media (Fig. 2c). Feature-based 210 molecular networking using MZmine 2 in conjunction with Global Natural Products Social 211 (GNPS) Molecular Networking 47,48 allowed us to putatively identify these two peaks as 212 Ybt. Ybt is known to tautomerize at C10 (Fig. 3) and Ref. 49 . We confirmed that these 213 peaks were two diastereomers of Ybt by matching the retention time, exact mass, and 214 MS/MS spectra acquired from culture extracts to an authentic Ybt standard (Fig. 2e, f). 215 Post-LC pH neutralization and zinc-infusion revealed the zinc-bound Ybt species, 216 indicating that Ybt is indeed capable of binding zinc (Fig. 2d). To our surprise, we also 217 found that one of the diastereomers (at retention time = 4.0 min) binds zinc with higher 218 preference than the other (at retention time = 4.3 min) (Fig. 2d). Since Ybt was initially 219 discovered as an iron-binding molecule, and thus termed a siderophore, we next sought 220 to determine the preferential conditions for binding iron versus zinc. To assess the 221 competition between iron and zinc binding, we performed direct infusion mass 222 spectrometry competition experiments at multiple pH values. In these experiments, we 223 added equimolar amounts of zinc and iron to Ybt in ammonium acetate buffer adjusted to 224 pH 4, 7, and 10. While Ybt preferentially bound iron at low pH (pH 4), it exhibited a higher 225 preference for zinc at high pH (pH 10) (Fig. 2g). Intriguingly, at neutral pH (pH 7), Ybt 226 was observed bound to iron or zinc at roughly equal proportion (Fig. 2g).

228
To confirm the zinc-binding observed by native electrospray metabolomics, we monitored 229 a zinc-titration into Ybt by 1D 1 H NMR (Fig. 3). Although Ybt is in equilibrium between two 230 tautomers at C10 that have different relative affinities for zinc, only one set of signals is   apo-Ybt (Fig. 1e). Addition of the siderophore apo-enterobactin, which is not expected to 252 bind to zinc, did not significantly rescue the growth of either the znuA zupT irp2 mutant or 253 the znuA zupT ybtX mutant (Fig. 1e). Taken together, these results demonstrate that Ybt 254 binds to both iron and zinc, that metal binding preference can be influenced by pH, and 255 that Ybt can scavenge zinc for EcN in zinc-limited media. Next, we assessed whether Ybt 256 enables EcN to evade the host response.

260
In the host, zinc limitation is largely dependent on the antimicrobial protein CP 53 . We thus 261 tested whether Ybt-mediated zinc acquisition enhances EcN's growth in CP-262 supplemented rich media. Above, we demonstrated that when the ZnuABC and ZupT 263 transporters were deleted (znuA zupT mutants), EcN grew better than STm (Fig. 1b-d). 264 Moving forward, when either irp2 or ybtX were additionally deleted in EcN, growth of the 265 znuA zupT irp2 and the znuA zupT ybtX mutants were ~8-fold lower than the parental 266 EcN znuA zupT strain in the presence of 150 µg/ml CP (Fig. 1d). Although the growth of   After demonstrating that Ybt promotes EcN resistance to CP in vitro, we next sought to 283 investigate whether Ybt confers a growth advantage to EcN during inflammatory 284 conditions in vivo, where CP is highly expressed 1,15 and zinc is limited 15 . To induce 285 intestinal inflammation, we employed the dextran sodium sulfate (DSS) mouse colitis 286 model (Fig. 4a). After 4 days of DSS administration, we orally inoculated the mice with a EcN wild-type exhibited a significant competitive advantage over the znuA zupT mutant 291 beginning at day 1 post-inoculation, which increased to an average of ~28-fold by day 7 292 (Fig. 4b). These results indicated that ZnuABC and ZupT are needed for optimal 293 colonization of the inflamed gut. By contrast, EcN znuA zupT showed a significant 294 competitive advantage over both triple mutants, which increased over time up to ~20-fold 295 (znuA zupT ybtX mutant) and ~50-fold (znuA zupT irp2 mutant) ( Fig. 4b and 4c). In both 296 cases, the increased competitive advantage was due to the decreased colonization level 297 of the triple mutants, as the znuA zupT mutant colonized at similar levels 298 (Supplementary Fig. 3a-c). Of note, host antimicrobial gene expression levels (Lcn2, 299 S100a8, S100a9) were similarly upregulated in all DSS-treated mice (Fig. 4d), and all 300 DSS-treated mice developed similar levels of colitis, as shown by histopathology 301 evaluation of the distal colon (Fig. 4e, f). Collectively, these results indicate that both Ybt 302 production (via Irp2) and Ybt transport (via YbtX) enhance EcN colonization of the 303 inflamed gut. Because Ybt production and acquisition conferred a colonization advantage 304 14 to the znuA zupT mutant, these data support the idea that Ybt can scavenge zinc in vivo, 305 in zinc-limited conditions such as those found in the inflamed gut.

309
Next, we ascertained whether the zinc transport systems of EcN play a significant role in 310 the absence of gut inflammation. As EcN colonization levels decline over time in 311 conventional mice in the absence of inflammation, we used germ-free mice (Fig. 5a), in 312 which we previously observed high levels of EcN colonization for extended periods of 313 time 3 . When we inoculated germ-free mice with a 1:1 mixture of EcN znuA zupT and 314 znuA zupT ybtX (Fig. 5a), we recovered similar amounts of both strains from mouse feces 315 throughout the experiment ( Fig. 5b and Supplementary Fig. 3d). Whereas S100a8, 316 S100a9, and Lcn2 were highly expressed in the ceca of DSS-treated animals colonized 317 with EcN, these genes were only minimally upregulated (less than 10-fold) in germ-free 318 mice colonized with EcN (Fig. 5c). The absence of inflammation in EcN-colonized germ-319 free mice was also confirmed by cecal pathology (Supplemental Fig. 3f, left panel).

321
To further probe whether Ybt provides a means for EcN to evade CP-dependent zinc 322 depletion in vivo, we employed S100a9 -/mice (deficient in CP) treated with DSS, and 323 inoculated them with a 1:1 mixture of EcN znuA zupT and znuA zupT ybtX (Fig. 5d). 324 Although the mice developed intestinal inflammation (Supplementary Fig. 3f, right panel), We have previously shown that pathogenic STm and probiotic EcN evade lipocalin-2-340 mediated iron sequestration in the inflamed gut via the production of stealth siderophores 341 16,33 . As we have found that STm also evades CP-mediated zinc sequestration in the 342 inflamed gut 15 , we sought to investigate whether EcN also evades CP to acquire zinc 343 and thrive in the host. As EcN, akin to STm, expresses ZnuABC and ZupT, we initially 344 hypothesized that these zinc transporters mediate EcN resistance to CP. However, when 345 we found that an EcN znuA zupT mutant still grew up to 1,000-fold better than an STm 346 znuA zupT mutant in media containing CP (Fig. 1), we speculated that EcN must utilize 347 additional mechanisms to acquire zinc. In the work presented herein, we unexpectedly 348 discovered that EcN scavenges zinc with the siderophore Ybt. Ybt is a phenolate siderophore that was first discovered as being produced by Yersinia 351 enterocolitica 56 . The term siderophore has its origin in the Greek language and means 352 "iron carrier", as these molecules are widely characterized as being produced by 353 microorganisms in order to acquire iron. However, recent studies have proposed that at 354 least some siderophores may also bind to other metals. For example, the siderophore 355 ferrioxamine was shown to bind manganese 57,58 , and Ybt was shown to bind copper as at the C10 position (Fig. 3) into a racemic mixture 49 . Using post-LC pH neutralization and 382 metal infusion in a recently developed workflow termed native metabolomics, we found 383 that one isomer (retention time = 4.0 min) preferentially binds zinc (Fig. 2). The different 384 affinity of siderophore diastereomers for a metal is not unprecedented. Pyochelin, a 385 siderophore with a similar thiazoline core as Ybt and produced by Burkholderia cepacia 386 and several Pseudomonas strains, also exists as two diastereomers, only one of which 387 binds iron 68 . Moreover, although pyochelin was shown to bind both iron and zinc in vitro 388 69 , to our knowledge, the biological relevance of pyochelin-mediated zinc scavenging has 389 not been investigated. Similarly, only one of the Ybt isomers was shown to bind gallium 390 when the compound's structure was initially characterized 49 . We used 1D 1 H NMR 391 spectroscopy to confirm this observed zinc binding by Ybt (Fig. 3) Because Ybt is known to bind iron, we performed a competition assay with equimolar 398 amounts of zinc and iron. We observed that the metal-binding preference of Ybt is pH-399 dependent -Ybt preferentially binds to zinc in basic conditions (pH = 10), to iron in acidic 400 conditions (pH = 4), and exhibits similar preference for both at pH 7 (Fig. 2e-h). In contrast 401 to Ybt, the binding capacity of pyochelin to different metals is pH independent 69 . We  EcN more resistant than STm to zinc sequestration by CP (Fig. 1) in vitro prompted us to 420 investigate the function of Ybt during EcN colonization of the inflamed gut. We found that 421 EcN mutants lacking either Ybt or the putative inner membrane receptor YbtX, in addition 422 to lacking ZnuABC and ZupT, showed more severe colonization defects than the znuA 423 zupT mutant in mice with DSS-induced colitis (Fig. 4). As four other iron transport systems 424 (including the stealth siderophores salmochelin and aerobactin, as well as heme uptake) 425 are still present in these strains, it is unlikely that the in vivo phenotype of the mutants is 426 due to an inability to overcome iron starvation.

428
Together with the observations that Ybt contributes to optimal growth of EcN in zinc-429 limited conditions in vitro (Fig. 1), and that Ybt directly binds zinc at the pHs found in the 430 intestine (Fig. 2), the colonization defect of EcN znuA zupT irp2 and of EcN znuA zupT 431 ybtX in DSS-treated mice is consistent with the strains' limited ability to acquire zinc.

432
Moreover, the colonization advantage provided by Ybt is highly dependent on the state 433 of inflammation and presence of CP, as EcN znuA zupT and EcN znuA zupT ybtX 434 colonized to similar levels in S100a9 -/mice as well as in germ-free mice (which lack 435 inflammation and only express low levels of CP) (Fig. 5). These results are in agreement 436 with the in vitro results showing that Ybt and the putative receptor YbtX enable EcN to 437 acquire zinc in media supplemented with CP (Fig. 1).

439
Altogether, our work demonstrates that Ybt directly binds to zinc in a pH-dependent  Table 2). Strain names for the mutants are listed in Supplemental Table   465 1. To confirm integration of the resistance cassette and deletion of the target, mutant 466 21 strains and wild-type controls were each assayed utilizing PCR, and sequencing primers 467 (Supplemental Table 2) that flank the target sequence were used in conjunction with a 468 common test primer to test for both new junction fragments.      were made using a combination of 2D COSY (Supplementary Fig. 2b), 2D ROESY with 615 a mixing time of 300 ms (Supplementary Fig. 2c) dissolved in deuterated acetonitrile (CD3CN, 300 μl) then a baseline spectrum was taken.

629
A spectrum was recorded between each addition of ZnCl2. 0.5 equiv. of ZnCl2 dissolved 630 in CD3CN was added to the dissolved Ybt, followed by a second 0.5 equiv. (1 equiv. total), 631 another 1 equiv. (2 equiv. total), and finally another 3 equiv. (5 equiv. total). Once the 632 spectrum with 5 equiv. of ZnCl2 was recorded, an NaOD (in D2O) titration was performed 633 to determine the effect of increasing pH on the binding of Zn 2+ . Sequential additions of 634 0.5 equiv., a second 0.5 equiv., 1 equiv., and 3 equiv. were made and 1D 1 H NMR spectra 635 were recorded for each titration point.  Histopathology. Distal colonic tissues from wild-type mice, proximal colonic tissues from 667 S100a9 -/mice, and cecal tissues from germ-free mice were fixed in 10% buffered 668 formalin, then processed according to standard procedures for paraffin embedding. 5 µm 669 sections were stained with hematoxylin and eosin, then slides were scanned on a (mononuclear infiltration, edema, epithelial injury, and neutrophilic inflammation/crypt 673 abscesses) was determined as absent (0), mild (1), moderate (2), or severe (3).