Iron uptake pathway of Escherichia coli as an entry route for peptide nucleic acids conjugated with a siderophore mimic

Bacteria secrete various iron-chelators (siderophores), which scavenge Fe3+ from the environment, bind it with high affinity, and retrieve it inside the cell. After the Fe3+ uptake, bacteria extract the soluble iron(II) from the siderophore. Ferric siderophores are transported inside the cell via the TonB-dependent receptor system. Importantly, siderophore uptake paths have been also used by sideromycins, natural antibiotics. Our goal is to hijack the transport system for hydroxamate-type siderophores to deliver peptide nucleic acid oligomers into Escherichia coli cells. As siderophore mimics we designed and synthesized linear and cyclic Nδ-acetyl-Nδ-hydroxy-l-ornithine based peptides. Using circular dichroism spectroscopy, we found that iron(III) is coordinated by the linear trimer with hydroxamate groups but not by the cyclic peptide. The internal flexibility of the linear siderophore oxygen atoms and their interactions with Fe3+ were confirmed by all-atom molecular dynamics simulations. Using flow cytometry we found that the designed hydroxamate trimer transports PNA oligomers inside the E. coli cells. Growth recovery assays on various E. coli mutants suggest the pathway of this transport through the FhuE outer-membrane receptor, which is responsible for the uptake of the natural iron chelator, ferric-coprogen. This pathway also involves the FhuD periplasmic binding protein. Docking of the siderophores to the FhuE and FhuD receptor structures showed that binding of the hydroxamate trimer is energetically favorable corroborating the experimentally suggested uptake path. Therefore, this siderophore mimic, as well as its conjugate with PNA, is most probably internalized through the hydroxamate pathway.


Introduction
Iron is one of the essential nutrients necessary for bacterial growth, but in the environment, it occurs almost exclusively in an insoluble ferric form (Raymond et al., 2003;Schauer et al., 2008).Therefore, despite its overall abundance, the concentration of the Tsylents et al. 10.3389/fmicb.2024.1331021Frontiers in Microbiology 02 frontiersin.orgobtainable ferric iron is extremely low (10 −18 M).The iron access for pathogenic bacteria is restricted even further in the organism of the host (Drakesmith and Prentice, 2012;Johnson and Wessling-Resnick, 2012;Cassat and Skaar, 2013;Wilson et al., 2016;Neumann et al., 2017).In the human body, ferric iron is scarcely accessible (10 −24 M) since it is either kept intracellularly within hemoglobin in erythrocytes or is extracellularly bound by transferrin or lactoferrin.
Iron can be also chelated with lower affinity by other molecules such as albumin, citrate, or even amino acids (Drakesmith and Prentice, 2012;Cassat and Skaar, 2013).Additionally, during the infection, hypoferremic defense response, including hepcidin production by the liver, is activated (Drakesmith and Prentice, 2012;Cassat and Skaar, 2013).Hepcidin is a hormone responsible for the regulation of iron transfer by prohibiting iron cellular efflux and decreasing its extracellular concentration (Drakesmith and Prentice, 2012;Johnson and Wessling-Resnick, 2012;Cassat and Skaar, 2013).To surpass all these obstacles and meet physiological requirements for growth (10 −6 -10 −3 M of iron; Gumienna-Kontecka and Carver, 2019), bacteria developed several ways to compete for and obtain iron(III), including the production of siderophores (Cassat and Skaar, 2013;Neumann et al., 2017;Kim et al., 2021).Once secreted by the bacterial cell, siderophores capture the ferric iron from the natural environment but also compete for iron with other chelating compounds present in the infected host.
The binding of the siderophore to its specific outer-membrane receptor results in the unfolding of the Ton box, which otherwise remains sequestered (Figure 1; Noinaj et al., 2010).The exposure of the Ton box triggers the interaction between the TBDT and the TonB-ExbB-ExbD complex, which provides energy generated from the proton motive force (Noinaj et al., 2010).This energy is needed for the structural changes of the globular luminal domain (also termed a 'plug') required for the transport of the iron chelator across the outer membrane through the TBDT (Schauer et al., 2008).Next, to pass the inner membrane, siderophores are captured by a periplasmic binding protein.Then they are delivered to the inner-membrane-bound ATP-binding cassette of the ABC transporter.The latter transports them further to the cytoplasm, where appropriate reductases convert iron(III) into iron(II), which is released from the siderophore (Schauer et al., 2008;Delepelaire, 2019).A scheme of the TonB-dependent E. coli intake systems of catecholate (enterobactin, 2,3-dihydroxybenzoylserine-DHBS) and hydroxamate (ferrichrome, coprogen) siderophores (Andrews et al., 2003;Braun, 2003).Upon recognition by its specific TBDT, the ferric-siderophore is internalized into the periplasm.This is possible due to the TonB-ExbB-ExbD complex which transmits the energy to the Ton box of appropriate TBDT.This energy is generated by the proton-motive force across the inner membrane.Next, the siderophore binds its respective periplasmic binding protein (catecholate type siderophore -FepB and hydroxamate type -FhuD), which delivers it to the dedicated ABC transporter complex (FepCDG for catecholate and FhuBC for hydroxamate type siderophores) for transport across the inner membrane.Once in the cytoplasm, a dedicated reductase (Fes or FhuF) 'unpacks' the Fe 3+ iron by reducing it to Fe 2+ for which siderophores show lower affinity.Although iron is one of the key nutrients, its intracellular level has to be controlled to avoid iron accumulation and further oxidation.In E. coli this role is assigned to a ferric uptake regulator (Fur) that uses Fe 2+ as a cofactor to bind to DNA sequences named Fur boxes (fur-regulated promoters; Hantke, 2001;Noinaj et al., 2010).These DNA sequences repress the expression of multiple genes involved in iron metabolism, e.g., iron transporters and proteins involved in siderophore biosynthesis (Hantke, 2001;Noinaj et al., 2010).Thus, Fur regulates iron transport by controlling the expression of TBDTs required for ferric siderophores, transcription of the tonB gene, and the exbB-exbD operon (Althaus et al., 1999;Chen et al., 2007).Moreover, Fur can bind directly to the promoter region of fepA-entD, fecABCDE, fhuACDB, and cirA transport genes (Griggs and Konisky, 1989;Hunt et al., 1994;Angerer and Braun, 1998;Chen et al., 2007;Noinaj et al., 2010).The Fur protein contains a helix-turn-helix motif and two metal-binding sites occupied by Zn 2+ or other divalent cations (Hantke, 2001;Noinaj et al., 2010).However, when the Fe 2+ ion binds to the metal-binding site of the Fur protein, it promotes conformational changes and allows Fur binding to the respective Fur boxes (Noinaj et al., 2010;Pasqua et al., 2017).Therefore, when iron is limited, the Fur protein has a lower affinity to DNA and cannot prevent transcription of Fur boxes, leading to increased expression of these genes.Importantly, the Fur-deprived mutants lack iron regulation and do not limit the uptake of iron(III).
Peptide nucleic acid (PNA) is a synthetic nucleic acid analog.PNA has the same nucleobases as DNA, but they are connected via a peptide-like backbone consisting of N-(2-aminoethyl)-glycine units (Nielsen et al., 1991).This neutral backbone eliminates electrostatic repulsion and increases PNA affinity toward natural nucleic acids.As a result, PNA oligomers create stable complexes with either DNA or RNA.Importantly, since PNA structure differs from both peptides and nucleic acids, it is not recognized by proteolytic and nucleolytic enzymes and is resistant to both (Wojciechowska et al., 2020).These advantages highlight the PNA potential in gene targeting strategies, both in diagnosis and as antimicrobial agents [reviewed in Saarbach et al. (2019); Wojciechowska et al. (2020); Perera et al. (2021); Tsylents et al. (2023)].
The main drawback seriously limiting PNA applications is its poor membrane permeability.Another factor limiting PNA uptake in gram-negative bacteria is the lipopolysaccharide (LPS) layer, which forms a negatively charged structural barrier protecting the inner leaflet of the outer membrane and preventing passive diffusion of hydrophobic molecules (Good et al., 2000;Bertani and Ruiz, 2018).Therefore, the conjugation of PNA oligomers to carriers that would enable PNA transport across the membrane has been investigated (Lu et al., 2010;Volpi et al., 2021;Patil, 2022;Patil et al., 2022).One strategy is to use cell-penetrating peptides (CPPs), such as (KFF) 3 K, TAT or (RXR) 4 XB (Bendifallah et al., 2006;Barkowsky et al., 2019Barkowsky et al., , 2022;;Patil et al., 2019;Wojciechowska et al., 2020;Yavari et al., 2021;Tsai et al., 2023).Unfortunately, CPPs are not universal as their delivery efficiency depends on a bacterial strain (Goltermann et al., 2022) and they may also be hemolytic (Wojciechowska et al., 2020).The other unique candidate tested as a PNA carrier was the selfassembling DNA tetrahedron (Readman et al., 2017).This DNA structure bound and transported a short PNA oligomer inside a bacterial cell, but the transport mechanism requires further research (Readman et al., 2017).Another non-peptidic strategy, proposed by us, is to use the cell's active transport system by conjugating the PNA oligomer to vitamin B 12 , which is an essential nutrient required for the growth of most bacteria (Równicki et al., 2017(Równicki et al., , 2019)).Indeed, vitamin B 12 was found to carry PNA into E. coli and S. Typhimurium cells.Unfortunately, the concentrations of vitamin B 12 required by these bacteria are significantly smaller than the concentrations of PNA oligomer necessary to block mRNA translation and achieve the antibacterial effect (Równicki et al., 2019;Wojciechowska et al., 2020).
Following the successful use of vitamin B 12 as the PNA carrier, we turned to microbial iron regulation mechanisms, which also seem viable for the Trojan horse strategy.A similar approach (reviewed by (Negash et al., 2019;Zhanel et al., 2019)), where conjugation of siderophores with various antibiotics was attempted, resulted in several promising β-lactam-siderophore conjugates.Using siderophore mimics should allow us to take advantage of the active transport of a limited yet essential nutrient that is required in amounts similar to the PNA concentrations we would like to deliver.Therefore, we hypothesized that siderophore mimics are promising candidates for PNA carriers through the TonB-dependent transporters.In this work, we designed and synthesized hydroxamate-based iron chelators mimicking the natural ferrichrome and coprogen.We then verified their iron-binding properties, the interactions of iron with the siderophore at an atomistic level of detail, as well as the path of their uptake.Next, we tested the carrier potential of one of these siderophores by conjugating it with the PNA oligomer aimed at silencing the expression of the reporter gene in E. coli.

Design and synthesis of siderophore mimics
Based on the structures of the ferrichrome and coprogen, wellknown hydroxamate-type natural siderophores, recognized by, respectively, the FhuA and FhuE E. coli TBDT receptors (Figure 1), we designed two synthetic siderophore mimics -the linear (S L ) and cyclic (S C ).To synthesize these siderophores, we used a commercially available modified ornithine (N α -Fmoc-N δ -acetyl-N δbenzyloxycarbonyl-l-ornithine).This building block is suitable for solid-phase peptide synthesis (SPPS) and provides the N δ -terminal hydroxamate groups that are crucial for metal binding.These ornithine derivatives were used to obtain a linear N δ -acetyl-N δhydroxy-l-ornithine trimer (Orn(Ac,OH)) 3 to preserve the same number of metal-binding groups as in ferrichrome and coprogen and to achieve the optimal number of hydroxamate groups for iron coordination (Minnick et al., 1992;Boukhalfa and Crumbliss, 2002).
Ferrichrome is a siderophore forming a cyclic structure.Therefore, after the synthesis of the linear siderophore mimic S L (Figure 2), we obtained a cyclic analog using the head-to-tail cyclization strategy (Figure 3).Synthesis details are provided in the Materials and Methods section.Cyclization was attempted to increase the proteolytic stability and to add structural constraints to the peptide backbone.Restricting the flexibility of the hydroxamate groups could decrease their conformational freedom to make them adopt a conformation that would stably coordinate iron(III).Therefore, we proceeded to test the iron coordination properties of both linear and cyclic siderophore mimics.

Iron coordination by siderophore mimics
Upon binding iron, siderophores with three hydroxamate groups, where six oxygen atoms are used for Fe 3+ coordination, form the octahedral hexacoordinate complex.This specific structure has two optical isomers: Δ (right-hand propeller) and Λ (left-hand propeller), which are separate from geometric isomers (van der Helm et al., 1980).Natural siderophores with chelated iron(III) adopt the structure of one of the most optimal optical isomers.For example, the Fe 3+ -desferrichrome complex adopts the Λ configuration (van der Helm et al., 1980), but for ferric-coprogen, the Δ configuration is optimal although the Λ configuration is also present (Wong et al., 1983;Winkelmann, 1991).Importantly, it was shown that specific receptors responsible for the siderophore transport can recognize optical isomers (Winkelmann, 1979;Thulasiraman et al., 1998).
To elucidate the configuration of the siderophore-iron(III) complexes, we used circular dichroism (CD) spectroscopy where the minimum and maximum with Cotton effect are expected in the range of 350-490 nm (Winkelmann, 1991;Matsumoto et al., 2001).For the Λ configuration, the minimum band in the CD spectrum should be present before the maximum.Since the Δ configuration is a mirror image of the Λ configuration, the maximum band appears first (Winkelmann, 1991).In the presence of the Fe 3+ salt, the natural siderophores and S L solution changed from colorless to yellow after The synthesis scheme of azido-(Orn(Ac,OH)) 3 -named S L .several minutes.This indicates the presence of the siderophore-ferric iron complex and further confirms the metal-binding properties of these compounds (Winkelmann, 1991).Based on this observation, during the CD experiments, for every siderophore, we adjusted the incubation time at the optimal concentration, ranging from 40 to 90 min (see Supplementary Figure S1).CD spectra show that S L chelates ferric iron adopting the structure of the Λ isomer exhibiting the Cotton effect with the negative (360 nm) and positive (458 nm) bands (Figure 4).Similar positions of the minima and maxima in the Fe(III)-S L complex and ferrichrome spectra suggest that, while capturing iron, S L adopts a structure comparable to the cyclic ferrichrome (Table 1).However, the signal intensity for S L is not as notable as for ferrichrome even though the concentrations are equal.This suggests the presence of the Δ configuration since linear siderophores are flexible (Wong et al., 1983).On the other hand, S L is not as flexible as deferoxamine, which can appear as a racemic mixture of two optical isomers (Hossain et al., 1986).For deferoxamine and its complex with Fe 3+ , known as feroxamine, sometimes the Λ configuration can be assigned, and later the Δ isomer is more prominent, but the signal intensity could not be improved either with increasing concentration or incubation time.
The CD spectrum of S C with Fe 3+ does not resemble the spectrum of any natural siderophore and does not fit the description of the Δ or Λ isomers.The minimum is observed but the required Cotton effect is not present in the region between 360 and 450 nm.Additionally, no color change was observed for the mixture of S C and Fe 3+ salt solutions throughout the whole duration of the experiment.This strongly indicates that S C cannot bind ferric iron.This is probably due to the structural hindrance or rigidity of S C , which does not allow the proper arrangement of the hydroxamate oxygen atoms for iron coordination.Therefore, based on these findings, further experiments and simulations focused solely on the S L or its conjugates with PNA.

Growth recovery assays confirm S L is a functional siderophore recognized by the FhuE receptor
After confirming that S L binds iron(III), we tested if the S L molecule can be used by E. coli as a siderophore.E. coli uses several siderophore-mediated iron uptake systems (Andrews et al., 2003).Each of them consists of one or more outer-membrane receptors, periplasmic binding protein, inner-membrane transporter, and cytoplasmic reductase (Figure 1).The two most prominent iron uptake systems recognize catecholate-and hydroxamate-type siderophores.To identify the uptake path of S L and verify its siderophore function as the iron-delivering compound, we performed a growth recovery assay adopted from Zheng et al. (2012).In this assay, bacteria are cultured in conditions that limit their growth, which we ensured by the presence of iron-chelating agent -2,2′-dipyridyl (DP).Two parallel cultures were prepared, with and without the addition of the S L siderophore mimic.If S L works as a functional siderophore, then we should observe a growth increase in the culture that contains S L .If no growth recovery or increase is observed, then either the siderophore is non-functional or its use is prohibited, e.g., by the lack of a protein in the uptake system.Thus, we tested various E. coli K-12 mutants in which the genes coding for proteins of these transport systems were deleted (Table 2).
First, we tested mutants lacking the FepB and FhuD proteins, which are the periplasmic binding proteins in the catecholate-and CD spectra of the natural (ferrichrome and feroxamine) and synthetic (S L and S C ) siderophores with Fe 3+ salt solution in phosphate buffer (pH = 7.0).Optimal Λ (solid line) or Δ (dashed line) configuration was observed for siderophore-iron(III) complexes apart from S C (dasheddotted line).hydroxamate-type siderophore uptake systems, respectively (Figure 1).Growth recovery was observed for the FepB mutant, whereas no significant change in bacterial growth for the mutant lacking FhuD was detected (Figure 5).These results were further confirmed by testing mutants lacking cytoplasmic siderophore reductases for the catecholate (the Fes protein) or hydroxamate (the FhuF protein) systems.The output indicates that S L is internalized via the hydroxamate siderophore uptake system.This is consistent with our expectations because S L mimics the hydroxamate-type siderophore.Next, we proceeded to identify the outer-membrane receptor for S L .Two outer-membrane receptors belong to the hydroxamate uptake system: the ferrichrome receptor (FhuA) and the coprogen receptor (FhuE; Killmann and Braun, 1998; Figure 1).Mutants deficient in these receptors were tested in growth recovery assays showing statistically significant recovery for the FhuA mutant but not for the FhuE mutant (Figure 5).In summary, the growth recovery assays confirmed that S L is a functional siderophore for E. coli, and indicated that S L is recognized by the coprogen outermembrane receptor FhuE and internalized through the hydroxamate pathway.

2.4
The S L siderophore mimic transports PNA to Escherichia coli cytoplasm After we confirmed that S L is internalized by E. coli K-12, we tested if this siderophore mimic acts as a PNA carrier through the E. coli cell wall.For this aim, we conjugated S L with different PNA oligomers.Based on the CD experiments, only S L was used for conjugation.This siderophore mimic was synthesized with a linker possessing an azide moiety needed for its conjugation with PNA.The conjugation was performed using the copper-catalyzed azide-alkyne cycloaddition (CuCAAC) reaction resulting in the non-cleavable linkage via the triazole ring (Figure 6).
To test if S L can transport PNA oligomers to E. coli cells, we first performed an altered growth recovery assay in which instead of S L alone we used its conjugate with PNA.The sequence of the PNA was designed to hybridize with mRNA of the acpP gene (Figure 6) and as a consequence silence its expression (Good et al., 2001).The acpP encodes for acyl carrier protein involved in the fatty acid synthesis pathway and is one of the E. coli housekeeping genes (Nejad et al., 2021).We expected that silencing of the acpP gene would result in a measurable growth inhibition as observed previously for PNA anti-acpP conjugated to peptide carriers (Good et al., 2001;Hansen et al., 2016;Castillo et al., 2018).To our surprise, the endpoint experiment showed no significant growth differences in bacteria treated with S L -PNA anti-acpP conjugate (data not shown).To obtain better insight into this phenomenon, we performed a kinetic experiment during which bacterial growth was monitored every 30 min for 20 h of culture (Figure 7; Supplementary Figure S2).To ensure the conjugate was internalized into bacterial cells, we used the E. coli K-12 mutants ΔfepB, ΔfhuA, and Δfes for which we already confirmed the uptake (Figure 5).When treated with S L -PNA anti- acpP , no statistically significant growth differences were observed for the ΔfepB and Δfes mutants (Supplementary Figure S2).On the contrary, the ΔfhuA strain showed initial growth inhibition followed by later recovery (Figure 7).We suspect that this growth pattern may be the result of initial partial growth inhibition during the early stages of the culture when the ratio of the conjugates to bacterial cells is high.As the bacteria grow in numbers, this ratio drops which reduces the inhibition effect and allows growth recovery.Therefore, this experimental design may carry an internal flaw, namely the inability to separate the growth-promoting effect of the siderophore from the growthinhibiting effect of PNA anti-acpP .Growth recovery assay.Various E. coli mutants were cultured in iron-limiting conditions with or without S L (at a concentration of 16 μM).The growth of different strains was measured by OD 600 and normalized to OD 600 measured for a given strain in non-iron limiting conditions.The experiment was repeated three times on different days.The errors shown are SEM, n = 4-6.The statistical significance of the difference in growth was verified by a two-tailed t-test (**p < 0.01, *p < 0.05, ns: non-significant). 10.3389/fmicb.2024.1331021 Frontiers in Microbiology 07 frontiersin.org To overcome the above limitations of the acpP-based assay, we designed another experiment in which we cultured the E. coli K-12 strains expressing the red fluorescent protein (RFP) in iron-limiting conditions.We used the protocol based on our previous studies on PNA conjugates with vitamin B 12 (Równicki et al., 2017;Pieńko et al., 2021), but the culture was supplemented with the conjugate of S L and PNA.To silence the RFP expression, the PNA anti-rfp oligomer targeted the mRNA of the mrfp1 gene encoding RFP (Figure 6).A 14-mer long PNA sequence complementary to the start codon and covering five nucleotides of the ribosome binding site (RBS) in mRNA was used.This approach assumes that if the conjugate of S L and PNA reaches the bacterial cytoplasm, it would bind and sterically block the complementary mRNA fragment.As a result, RFP expression will be reduced, which would manifest itself in a fluorescence decrease.The change in fluorescence should not be affected by the suggested growth promotion effect of the siderophore, so the results obtained for the PNA anti-rfp sequence should provide clearer insight into the carrier potential of S L .Here, fluorescence was measured using the flow cytometer.We chose this method over the traditional plate-based assay because it enables insight into individual cells.Thus, it is less prone to noise allowing for more precise measurements (Ambriz-Aviña et al., 2014).
Initially, we tested the wild-type E. coli K-12 (Figure 8).As expected, we observed no reduction of RFP fluorescence when the culture was supplemented with S L , PNA anti-rfp , and S L -PNA control2 (the scrambled PNA sequence not fully complementary to the targeted mRNA transcript; Figure 6).Indeed, as found by us (Wojciechowska et al., 2014(Wojciechowska et al., , 2020;;Równicki et al., 2017) and other groups (Good et al., 2000;Bendifallah et al., 2006;Readman et al., 2017;Barkowsky et al., 2019), PNA itself is not able to penetrate cell membranes without additional help.The scrambled PNA control2 sequence in the S L -PNA control2 conjugate is only partially complementary to the RFP-coding transcript (with seven mismatches) and does not overlap with the start codon or ribosome binding site (RBS), so it served as a negative control.
Surprisingly, we did not observe statistically significant RFP silencing for the fully complementary S L -PNA anti-rfp .We hypothesize that although the S L -PNA anti-rfp molecule reaches the bacterial cytoplasm and hybridizes with the proper mRNA fragment, it lacks sufficient numbers to saturate enough binding sites to cause a detectable reduction in RFP fluorescence.Since PNA molecules are insensitive to intra-cellular enzymatic degradation it should be possible to increase the number of S L -PNA anti-rfp per cell by stimulating the siderophore uptake.Bacteria were already cultured in iron-limiting conditions, which should induce such uptake.However, as proved in the growth recovery assays, as S L enters bacterial cells it provides them with iron, which may reduce the further siderophore uptake over time.What is more, E. coli can produce enterobactin and acquire iron independently to S L , which in turn would reduce intake of S L even further.Thus, we decided to perform this experiment using the E. coli K-12 Δfur mutant (Figure 8).Fur is a global iron-dependent regulator of E. coli that controls the expression of more than 90 genes (Escolar et al., 1999).It works as a positive repressor, which means that in a high iron environment, Fur binds to iron and causes repression of genes involved in iron acquisition.If Fur is missing then bacteria keep acquiring iron no matter if its demand was already satisfied.Indeed, the flow cytometry data acquired for E. coli K-12 Δfur showed a substantial loss of RFP fluorescence for bacteria cultured in the presence of S L -PNA anti-rfp (Figure 8).No other compound showed a statistically significant difference to the positive control.
To further verify the results obtained for E. coli Δfur, we performed a dose-response experiment (Figure 9).PNA alone did not affect the observed RFP levels at any concentration tested.Unconjugated S L showed an increased RFP cell count, which probably reflects the 'extra-nutrition' effect provided by S L acting as a functional siderophore delivering iron.As expected, increasing the concentration of the S L -PNA anti-rfp conjugate reduced, in a dosedependent way, the RFP cell count.However, interestingly, S L -PNA control2 increased the RFP count at low concentrations (up to 2 μM) and decreased at higher concentrations similarly to the S L -PNA anti-rfp .This may be the result of an interplay between two contradictory effects, the first being the siderophore effect, and the second one -the PNA-dependent RFP silencing.Although PNA control2 sequence is only half-complementary to the mRNA fragment (with seven mismatched bases, including replacement of one pyrimidine by another in three separate positions) as compared to the PNA sequence, at concentrations of 4 and 8 μM, it seems to silence the RFP expression similarly as the dedicated PNA.However, for higher concentrations (16 μM) it is no longer as efficient.This leaves room for further investigation into the secondary structures and interactions between PNA anti-rfp and RNA molecules, as well as improvement of PNA sequence design.Altogether, our results indicate the possibility of using siderophore mimics, such as S L , to carry the PNA into the bacterial cells and to modify the expression of a particular gene.

Molecular dynamics of iron(III)
binding to the linear N δ -acetyl-N δ -hydroxy-l-ornithine trimer Coordination of iron(III) was detected by CD spectroscopy only for the S L siderophore.To understand the dynamics of this process at an atomistic level of detail, especially the flexibility of the siderophore mimic, we performed microsecond-long classical molecular dynamics simulations of various S L variants without and in the presence of the Fe 3+ ion (see Methods; Supplementary Figure S3).
Simulations of the free S L showed that this modified ornithine trimer in solution is internally flexible and does not adopt any preferential structure.The atomic root-mean-square fluctuations (RMSF) suggest that each S L monomer is equally mobile (see exemplary RMSF in Supplementary Figure S4).Also, the clustering analyses pointed to four clusters with occupancies between 20 and 30% each (Supplementary Figure S5).
The simulations with the iron(III) ion, initially randomly positioned in solvent away from S L , reproduced the association and binding of iron(III) to this siderophore.The hydroxamate-type siderophores bind iron(III) via electrostatic attraction coordinating up to six oxygen atoms belonging to the three -NOH-CO-groups.However, binding is connected with the dissociation of protons bound to the oxygen atoms, and with positions of the oxygen atoms of the hydroxamate groups in the cis conformation.
We monitored the distances of the iron(III) ion to the six oxygen atoms of the three hydroxamate groups.In eight simulations of the siderophore with deprotonated hydroxamate oxygens in the cis configuration we observed at least 4 oxygens interacting with iron(III).In two simulations, we found 6 oxygens close to Fe 3+ .The distances of iron(III) to the coordinating oxygens in one of the  The flexibility of this trimer allows for such positioning of the hydroxamate oxygens that could coordinate iron(III) by either four or six oxygens (Figure 11).The other oxygen atoms that are attracted by Fe 3+ in MD trajectories are from the amide terminal group or the peptide bond.However, these conformations result from classical MD trajectories, and for a detailed description of the final coordinated state quantum-mechanical approach would be necessary.This would require quantum-mechanical calculations to obtain force field parameters or quantummechanical molecular dynamics.However, we have shown that even with the classical molecular mechanics description of point charges, the distances of the iron(III) ion to these oxygens suggest that coordination of iron is possible and corroborates our CD experiments.The linear S L peptide-like molecule is flexible enough to achieve a structure in which 6 oxygens are closely positioned to iron.
2.6 Molecular docking of the linear N δ -acetyl-N δ -hydroxy-l-ornithine trimer-Fe 3+ complex and natural siderophore to the FhuD and FhuE proteins Experiments with E. coli mutant strains indicate that the hydroxamate siderophore uptake pathway is involved in the internalization of the S L siderophore mimic.To obtain further insight into molecular interactions of S L with FhuD and FhuE, the key members of the hydroxamate siderophore uptake pathway, we performed molecular docking (see Materials and methods).
After validating the robustness of the docking protocol by successfully re-docking crystallographic ligands (with the ligand heavy atom root-mean-square deviations of ≤0.75 Å; Supplementary Figure S6), we generated docking poses of S L bound to FhuD and FhuE.The resulting ligand poses largely coincide in terms of contacting amino acids with coprogen molecules present in the crystal structures of both proteins (Figure 12).They also possess The distances between the Fe 3+ ion and hydroxamate oxygens or the center of mass (CoM) of the siderophore as a function of the simulation time.The plot shows distances to the six hydroxamate group O-atoms (OH or OZ in the deprotonated cis form with the S L residue number in parenthesis).The distance to CoM is marked with red circles.The CoM was calculated for all non-hydrogen atoms.

FIGURE 11
Exemplary conformations in a ball and stick representation from MD simulations of S L in the presence of Fe 3+ with either two (A) or three (B) hydroxamate groups close to the Fe 3+ ion (FE is shown as a brown sphere).The insets show the corresponding systems in van der Waals sphere representations.The Na + ion is light pink.The distances between FE and oxygen atoms are between 2 and 4 Å.favorable docking scores with predicted binding free energies of −9.3 and − 9.6 kcal/mol in FhuD and FhuE, respectively.These scores are by 2-3 kcal/mol worse than the scores assigned to the best poses of re-docked coprogen, i.e., −11.3 and − 12.6 kcal/mol for FhuD and FhuE, respectively.However, these estimates are consistent with coprogen being a much stronger natural ligand of hydroxamate siderophore uptake pathway proteins than S L .However, the S L docking poses in FhuE show that the binding pocket is large suggesting that different binding modes for other siderophores may be possible and that FhuE could recognize and accommodate different ligand types.

Conclusion
Using modified ornithine we synthesized hydroxamate-type siderophore mimics and conjugated one of them with PNA oligomers.
We found that the linear hydroxamate trimer coordinates iron(III), but its cyclic analog does not.CD spectroscopy showed that the ferric-S L complex adopts the Λ configuration similar to ferrichrome.Molecular dynamics simulations confirmed that the S L siderophore is internally flexible and showed that either four or six hydroxamate oxygens surround and interact with the iron(III) ion.In some cases, the S L flexibility allows also the C-terminal amide and peptide bond oxygen to come at distances below 4 Å to the iron(III) ion.
The growth recovery assays on various E. coli mutants indicated that the uptake of the linear siderophore mimic occurs via the E. coli receptors recognizing the hydroxamate-type siderophores, and pointed to the FhuE outer-membrane receptor and FhuD periplasmic binding protein as necessary receptors.Molecular docking suggested that S L is a plausible binder to FhuE and FhuD, similar to the natural coprogen siderophore, transported via the same pathway.The experiment with the S L -PNA anti-acpP conjugate, targeting the essential acpP gene, suggested that two contradictory effects are present; the growth inhibition of the PNA anti-acpP is probably countered by the growth promotion coming from S L .To overcome this obstacle another assay, based on RFP fluorescence with PNA anti-rfp targeting the non-essential RFP gene carried on a plasmid, was introduced.These flow cytometry experiments showed that the conjugates of S L -PNA anti-rfp entered E. coli cells.They suggest that the whole conjugate is recognized by the E. coli receptors and S L transports PNA inside the cell.Therefore, the conjugation of PNA anti-rfp with S L via a non-cleavable triazole ring resulted in PNA silencing activity confirming that PNA reached the mRNA target in the cytoplasm.This study is a proof-of-concept that the Trojan horse strategy using the siderophore uptake pathway is possible for PNA, but to avoid the simultaneous growth promotion from the siderophore it needs to be optimized not to function as an iron(II) provider to bacteria.
4 Materials and methods
Conjugation reactions were monitored using reverse-phase highperformance liquid chromatography (RP-HPLC).Analytical chromatography was performed using the Knauer C18 columns (5 μm particles, 4.6 × 250 mm) in buffer A (0.1% trifluoroacetic acid (TFA) in acetonitrile) and buffer B (0.1% TFA in water).The following conditions were used for HPLC: flow rate of 1.5 mL/min for 30 min, room temperature, UV/vis detection at 220 nm for siderophores and conjugates or 267 nm for PNA oligomers and conjugates.Additionally, the identity of the synthesized products was confirmed by mass spectrometry (MS) using the Q-TOF Premier mass spectrometer.All chemicals were of analytical or reagent grade, and the buffers were prepared using distilled water from a Direct-Q Millipore system.

Synthesis of linear hydroxamate siderophore: azido-(Orn(ac,OH)) 3
The azido-hydroxamate ornithine peptide was synthesized by manual SPPS.Standard Fmoc chemistry (Figure 2) was used with a 3-fold molar excess of the Fmoc-protected amino acids and rinkamide TentaGel S RAM resin (amine groups loading of 240 μmol/g; this resin has a linker which yields a C-terminal amide upon TFA cleavage of the peptide).Fmoc-protected modified ornithines (Fmocl-Orn(Ac,OBz)-OH) and 6-azido-hexanoic acid (N 3 -Hx-OH) were assembled as active derivatives with the use of O-(7-Aza-1Hbenzotriazole-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HATU) and with the addition of 1-hydroxy-7-azabenzotriazole (HOAt) and collidine (1:1:2) using the dimethylformamide/Nmethylpyrrolidone (DMF/NMP; 1:1, v/v) solution for 60 min.Coupling efficiencies of amino acids were monitored using the Kaiser test (Wellings and Atherton, 1997).If the coupling reaction was not complete, the reaction was repeated with 1.5-fold molar excess of the amino acid for 45 min.The Fmoc deprotection was accomplished using 20% piperidine in DMF for 2 cycles (for 5 and 15 min).Benzoyl (Bz) protecting groups were removed from hydroxamate-ornithine derivative using 1 M solution of lithium hydroxide (LiOH) in methanol/tetrahydrofuran (MeOH/THF; 1:1; v/v) for 2 cycles (for 30 min each).Cleavage of the product from the resin was performed by treatment with a TFA/triisopropylsilane/water (95:2.5:2.5, v/v/v) mixture for 60 min.The obtained crude product S L was lyophilized and subsequently analyzed by MS and purified by RP-HPLC (see Table 3; Supplementary Figure S11).The mobile phase gradients with buffers A and B were 0 to 50% for 30 min.

Synthesis of PNA conjugates with hydroxamate siderophore
PNA-siderophore conjugates were synthesized using the CuAAC reaction (Figure 6).S L (2.0 mg, 0.003 mmol) was dissolved in 0.1 mL of DMF/H 2 O (1:1 v/v) solution and 60 μL of the solution was added to 0.001 mmol of a PNA oligomer (2.9 mg of alkyne-PNA anti-acpP or alkyne-PNA control , or 4.6 mg of alkyne-PNA anti-rfp or alkyne-PNA control2 ) dissolved 1.0 mg, 5 μmol) and tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA; 5.0 mg, 10 μmol) were dissolved in 0.5 mL of DMF/H 2 O (1:1 v/v) solution, stirred for 20 min and then added to the alkyne-PNA and azido-siderophore solutions.The reaction mixtures were stirred for 5 min before using ultrasounds for 30 min with the temperature of 50°C.The mixtures were centrifuged to remove the catalyst.The solutions, containing the crude products, were analyzed by MS and purified by analytical RP-HPLC.The mobile phase gradients with buffers A and B were 0 to 50% for 30 minutes unless stated otherwise (see Table 3; Supplementary Figures S13-S16).Siderophore-PNA solutions were prepared with distilled water.Concentrations of conjugates were determined by measuring the UV absorption at 260 nm (using Thermo Scientific Evolution 300 UV-Vis spectrophotometer) and calculated based on molar extinction coefficients provided by manufacturers.

CD measurements
CD spectra were recorded in aqueous buffer solution (100 mM phosphate buffer, pH 7.0), in the presence of 4 mM iron(III) salt solution (FeCl 3 in the phosphate buffer).To obtain optimal recording conditions, siderophore concentrations were varied in the range 0.24-1.0mM with an incubation time of 20-90 min (Supplementary Figure S1).In all CD experiments, the siderophore-Fe 3+ molar ratio of 1:1 was maintained.The spectra were collected using the Biokine MOS-450/AF-CD spectrometer equipped with the Xe lamp using a 0.1 cm CD cell.The acquisition duration time was 2 s with a resolution of 1 nm.The measurements were performed in 100 mM phosphate buffer, pH 7.0, room temperature, and wavelength range 300-500 nm.The presented CD spectra were smoothed with the Savitzky-Golay method using GraphPad and are the averages of three scans.Each CD experiment was conducted twice to confirm the repeatability of the spectra.

General microbiology methods
If not stated otherwise, bacteria were cultured either in the lysogeny broth (LB) medium (VWR) or 50% Mueller Hinton Broth (MHB) medium (Difco).Bacterial stocks were prepared by mixing 1 mL of overnight LB culture of a given strain with 0.5 mL of autoclaved 50% glycerol (POCH).Liquid bacterial cultures were incubated in Innova 44 Incubator Shaker (New Brunswick Scientific) at 240 RPM.Optical density at 600 nm (OD 600 ) of bacterial cultures was measured using SP-830 Plus Spectrophotometer (Metertech) with the clean medium as the reference.Wild-type E. coli K-12 was our in-house strain.All other E. coli K-12 mutants were acquired from the Keio collection (Table 2; Baba et al., 2006;Yamamoto et al., 2009).

Preparation of Escherichia coli strains expressing the red fluorescent protein
Bacterial strains expressing RFP were prepared as described previously (Pieńko et al., 2021).Briefly, chemically competent cells were transformed with pBBR1MCS5(rfp) plasmid using the Kushner method.Following transformations bacteria were selected against gentamycin (20 μg/mL).The RFP expression was confirmed by measuring red fluorescence on the UV illuminator.

Growth recovery assays
The growth recovery protocol was adapted from the literature (Zheng et al., 2012). 2 mL of LB medium was inoculated with a freezer stock of E. coli respective mutants and cultured overnight at 37°C with shaking.The overnight culture was then diluted 1:100 in 2 mL of fresh 50% MHB medium with or without the addition of 200 μM 2,2′-dipyridyl (DP, Merck).Bacteria were then cultured at 37°C with shaking until reaching OD 600 = 0.6, and diluted to OD 600 = 0.001.The diluted culture was aliquoted (90 μL per well) onto a 96-well, flat-bottom, fully-transparent plate.10 μL of 10x concentrated solution of the siderophore that was previously either preloaded or not with iron (1:1 molar ratio siderophore: FeCl 3 ) was added to each well.A series of Growth Controls (GCs) was prepared to allow compensation for additional iron added to the culture due to siderophore preloading.Plates were sealed with adhesive foil and incubated at 37°C with shaking for 19 h.Next, bacterial growth was measured as OD 600 using the Tecan Sunrise plate reader.Each well condition was prepared in duplicate, and three independent biological replicates of each assay were conducted on different days.Data analysis included subtracting

Flow cytometry
2 mL of LB medium was inoculated with a freezer stock of E. coli respective mutants and cultured overnight at 37°C with shaking.The overnight culture was then diluted 1:100 in 2 mL of fresh 50% MHB medium with 200 μM DP.Bacteria were then cultured at 37°C with shaking until reaching OD 600 = 0.5, and diluted to OD 600 = 0.001 in fresh 50% MHB medium with DP. 90 μL of such diluted culture was mixed with 10 μL of a tested compound dissolved in deionized water and cultured at 37°C with shaking for another 20 h.After that time bacteria were centrifuged at 5000 x g for 10 min and resuspended in 1 mL of FACSFlow buffer (BD Biosciences).Samples that required dead bacterial cells were then additionally incubated at 95°C for 5 min and cooled to room temperature.Next, 1 μL of SYTOX Red dead cell stain (Life Technologies) was added to each sample.After 30 min of incubation in darkness at room temperature samples were measured using LSRFortessa Cell Analyzer (BD Biosciences).For each sample, 10,000 events were recorded.Based on side-scatter and forward-scatter measurements, cells and within them single cells were identified.Within a single-cell population, only live cells were further analyzed.Gates for RFP-positive and RFP-negative cells were determined based on the measurements for non-treated bacteria expressing RFP and bacteria that do not express RFP, respectively (Supplementary Figures S17-S20).The experiment was repeated three times on different days.

Dose-dependent RFP fluorescence silencing
Overnight cultures were prepared by inoculating 2 mL of 50% MHB medium with 200 μM DP with a freezer stock of E. coli and incubated at 37°C with shaking.Overnight cultures were then diluted 200 times of fresh 50% MHB medium with 200 μM DP and mixed with deionized water solutions of tested compounds (90 μL of bacteria +10 μL of the compound).Subsequently, bacteria were incubated for 3 h at 37°C with shaking, centrifuged, and resuspended in 500 μL of FACSflow buffer (BD Biosciences) containing SYTOX Red dead cell stain (Life Technologies; 1:1000 dilution).Cells were then analyzed as described in the Flow cytometry section.

Molecular dynamics simulations
Since S L was found to both coordinate iron(III) and transport PNA inside E. coli cells, to understand the flexibility of this molecule and the dynamics of how iron is captured by the hydroxamate groups of S L , we performed molecular dynamics (MD) simulations.MD simulations were carried out for (Orn(Ac,OH)) 3 with the azide-linker at the N-terminus replaced with the COCH 3 group for neutrality.The hydroxyls of the hydroxamate groups were either in protonated or deprotonated states, and the positions of the oxygens were either in cis or trans (Supplementary Figure S3).Various combinations were simulated because iron coordination occurs via deprotonated oxygens and in MD the barriers between the cis and trans conformation were too high to observe the transition.The free molecule and in the presence of iron(III) were simulated.

Structure preparation and force field
The structures were drawn with molview.org and converted using Openbabel.The Amber ff14sb force field was used and the parameters for N δ -acetyl-N δ -hydfroxy-l-ornithine were derived using the AmberTools21 suite (Case et al., 2021).Atomic partial charges and force field atom types were assigned using antechamber with the AM1-BCC method for charge derivation (Jakalian et al., 2000), which was sufficient for our purposes.The Amber atom names, types, and partial charges are shown in Supplementary Figures S21-S22.The OPC3 (Izadi and Onufriev, 2016) parameters were used for water molecules, which were added to provide at least a 15 Å thick layer from the solute.In the case of simulations with ions, 11 Cl − ions and 11 Na + ions were included to ensure ionic strength of 150 mM NaCl.The systems formed a periodic cubic box with a size of 50.1 Å 3 (Supplementary Figure S23).For the Na + , Cl − , and Fe 3+ ions, the parameters of Li and Merz (12-6 normal usage set) were applied (Li et al., 2021;Sengupta et al., 2021).
The approaches to reproduce iron(III) binding included five series of simulations.Four water molecules at least 10 Å from the solute were randomly replaced with one Fe 3+ ion and 3 Cl − ions to achieve electroneutrality of the system.The starting conformations for the simulations including Fe 3+ were taken as cluster representatives from the simulations of the free system.In each case, three systems were generated that only differed by the initial position of the Fe 3+ ion. the percentage of variance explained, SSR/SST, the selected number of clusters was four (Shao et al., 2007).

Molecular docking
Molecular docking was carried out in the YASARA Structure (Ozvoldik et al., 2021).The following PDB structures were chosen as receptors: 1esz (Clarke et al., 2002) and 6e4v (Grinter and Lithgow, 2019) for FhuD and FhuE, respectively.Crystal water molecules were deleted.Hydrogens were added assuming a pH equal to 7.4, and the hydrogen bond network was optimized using the OptHyd command (Krieger et al., 2012).The binding site in each structure was chosen by placing a cubic simulation cell around the crystallographic ligand.To thoroughly explore available protein cavities, the box length of 25 Å for FhuD and 30 Å for FhuE structure was applied.The crystallographic ligand (coprogen) was deleted from the receptor model before S L docking.
Protonated coprogen molecules (PDB ligand IDs CPO and HWS for FhuD and FhuE structures, respectively; see Supplementary Figure S6) were used for validation of the docking protocol via re-docking.The representative starting conformations of the S L molecule were taken from MD simulations described above, with Fe 3+ well-coordinated by hydroxamate groups (Figure 11).For the ferric ligand to be handled properly by the docking protocol, dative metal coordination bonds preserving the distances between Fe 3+ and hydroxamate oxygen atoms were created.
Initial docking was carried out using the AutoDock Vina protocol (Trott and Olson, 2010) via the YASARA dock_run macro.The number of docking runs was set to 25, the ligand remained flexible, and exhaustiveness was set to 32.Subsequently, the results were optimized using dock_rescore macro with all previously obtained docking poses rescored with an energy minimization followed by 12 local docking runs each using the Vina local scoring method.Side chains were free during minimization.Point charges and dihedral barriers were obtained from the YASARA2 force field (Krieger et al., 2009).

FIGURE 6
FIGURE 6Schematic structure of the S L -PNA conjugate with the PNA sequences used.The PNA anti-acpP sequence was used to silence the expression of the acpP gene encoding the acyl carrier protein and PNA anti-rfp to silence the mrfp1 gene encoding the red fluorescent protein (see text).The control scrambled PNA sequences for PNA anti-acpP (PNA control ) and PNA anti-rfp (PNA control2 ) are shown with underlined mismatched bases.

FIGURE 7
FIGURE 7Kinetic measurement of the growth recovery assay for the ΔfhuA strain cultured in iron-limiting conditions without or with S L and its conjugate with PNA anti-acpP or PNA control (each at a concentration of 16 μM).The growth was normalized to the OD 600 measured without the addition of any compound.The experiment was performed in two biological replicates, and two technical replicates each.The errors shown are SEM, n = 4. Statistical significance was tested by the two-way ANOVA test with horizontal lines showing periods for which a significant difference (p < 0.05) was observed between S L -PNA anti-acpP and other compounds.

FIGURE 8 RFP
FIGURE 8 RFP fluorescence silencing in E. coli K-12 wild type and Δfur mutant.Bacteria were cultured in iron-limiting conditions and with the addition of various compounds (at a concentration of 16 μM).Bacterial cells were tested for RFP fluorescence using the flow cytometer with gates set up based on two control cultures: bacteria expressing RFP and bacteria lacking the RFP gene.The experiment was repeated three times on different days.Errors are SEM, n = 3. Statistical significance of the observed differences was determined by a two-way ANOVA test (****p < 0.0001, *p < 0.05, ns, nonsignificant).

FIGURE 9
FIGURE 9 Dose-dependent RFP fluorescence silencing in the E. coli K-12 Δfur mutant.Bacteria were cultured in iron-limiting conditions and with the addition of various compounds (at concentrations ranging from 0 to 16 μM) preloaded with iron.Bacterial cells were tested for RFP fluorescence using the flow cytometer with gates set up based on two control cultures: bacteria expressing RFP and bacteria lacking the RFP gene.Results were normalized to respective Growth Control values.The experiment was repeated three times on different days.Errors shown are SEM, n = 3.

FIGURE 12
FIGURE 12Top scoring poses of S L docked to FhuE (A) or FhuD (B) structures.Structures from docking are superposed with the crystallographic ligand (coprogen; green ball-and-stick model).Protein residues contacting S L , defined as residues within 4 Å from the ligand, are shown as gray sticks.The light blue rectangle overlayed on FhuE depicts an approximate position of the E. coli outer membrane.Hydrogen atoms are omitted for clarity.

TABLE 1
Summary of the CD spectroscopy experiments.

TABLE 2
Bacterial strains used in this study.

TABLE 3
Retention times (t R ) and molecular masses of the synthesized compounds.Obtained from Q-TOF Premier mass spectrometer; 2 estimated based on the a resin loading or b measuring the areas under the curves on the chromatograms.PNA oligomers include lysine at the C-terminus.For PNA sequences see Figure6.