Acquisition of regulator on virulence plasmid of hypervirulent Klebsiella allows bacterial lifestyle switch in response to iron

ABSTRACT Hypervirulent Klebsiella pneumoniae causes liver abscess and potentially devastating metastatic complications. The majority of Klebsiella-induced liver abscess are caused by the CG23-I sublineage of hypervirulent Klebsiella pneumoniae. This and some other lineages possess a >200-kb virulence plasmid. We discovered a novel protein IroP nestled in the virulence plasmid-encoded salmochelin operon that cross-regulates and suppresses the promoter activity of chromosomal type 3 fimbriae (T3F) gene transcription. IroP is itself repressed by iron through the ferric uptake regulator. Iron-rich conditions increase T3F and suppress capsule mucoviscosity, leading to biofilm formation and cell adhesion. Conversely, iron-poor conditions cause a transcriptional switch to hypermucoid capsule production and T3F repression. The likely acquisition of iroP on mobile genetic elements and successful adaptive integration into the genetic circuitry of a major lineage of hypervirulent K. pneumoniae reveal a powerful example of plasmid chromosomal cross talk that confers an evolutionary advantage. Our discovery also addresses the conundrum of how the hypermucoid capsule that impedes adhesion could be regulated to facilitate biofilm formation and colonization. The acquired ability of the bacteria to alternate between a state favoring dissemination and one that favors colonization in response to iron availability through transcriptional regulation offers novel insights into the evolutionary success of this pathogen. IMPORTANCE Hypervirulent Klebsiella pneumoniae contributes to the majority of monomicrobial-induced liver abscess infections that can lead to several other metastatic complications. The large virulence plasmid is highly stable in major lineages, suggesting that it provides survival benefits. We discovered a protein IroP encoded on the virulence plasmid that suppresses expression of the type 3 fimbriae. IroP itself is regulated by iron, and we showed that iron regulates hypermucoid capsule production while inversely regulating type 3 fimbriae expression through IroP. The acquisition and integration of this inverse transcriptional switch between fimbriae and capsule mucoviscosity shows an evolved sophisticated plasmid-chromosomal cross talk that changes the behavior of hypervirulent K. pneumoniae in response to a key nutrient that could contribute to the evolutionary success of this pathogen.

worldwide (3,4), has acquired a large virulence plasmid that greatly enhances its virulence (5,6). Hypervirulent K. pneumoniae can infect healthy individuals and is mostly known as the primary cause of monomicrobial Klebsiella-induced liver abscess (KLA) in Asia (7). However, it can systemically disseminate in the infected host causing pneumonia and endophthalmitis (2).
The rapid expansion and global dissemination of the hypervirulent K. pneumoniae CG23-I sublineage that first emerged in 1928 accounted for >80% of KLA infections (8). These strains possess a >200-kb low copy number large virulence plasmid (KpVP). Many hypervirulent K. pneumoniae biomarkers such as the rmpADC operon that controls hypermucoid capsule production (9,10), aerobactin (iuc) and salmochelin (iro) ironsequestering siderophores (11), and a putative membrane transporter (12) are encoded on the KpVP. It possesses its own plasmid maintenance systems and post-segregational killing of plasmid-lacking cells through toxin-antitoxin systems (1). Interestingly, KpVPs do not carry native antibiotic resistance genes or conjugation machinery but are highly stable in hypervirulent K. pneumoniae. This suggests that it provides a benefit to these strains and could contribute to the stability and global dissemination of the CG23-I sublineage.
K. pneumoniae thrive ubiquitously in environments from soil and water to animals (13). To survive in harsh conditions, bacteria form communities in biofilms after establishment on certain surfaces (14). Key factors that facilitate this include the capsule and surface adhesions such as the type 1 and 3 fimbriae (15,16). As hypervirulent K. pneumoniae infections in patients are often thought to originate from the gut, it is postulated that the bacteria are acquired from contaminated food sources and translocated through the gut into the bloodstream before systemic dissemination. In fact, CG23 hypervirulent K. pneumoniae was found on cucumber samples in China (17). However, its environmental reservoirs are undefined.
Current understanding of hypervirulent K. pneumoniae's ability to acclimatize in environmental habitats and human hosts is still lacking. While the KpVP was reported to be a major virulence factor in mice during intraperitoneal infection (18), its role in the bacterium's ability to thrive in extracellular habitats or other host niches is unknown. The basis of KpVP's contribution to mammalian virulence is attributed to the hypermucoid capsule and the major siderophores. However, the extent of their contribution is variable and sometimes contradictory, depending on the genetic profile of the isolates (19,20). Although the capsule is known to be important for systemic virulence (21,22), it is not clear how hypermucoid capsule is regulated in K. pneumoniae during infection such as to allow colonization, as the mucoviscosity of the capsule is known to impede cell adhesion (23) and yet shown to be important for virulence in a mouse pneumonia model (9).
In this study, we describe the discovery of a novel regulator designated IroP on the KpVP that represses the type 3 fimbriae (T3F). IroP's expression is inhibited by the presence of iron through the ferric uptake regulator (Fur). We propose that the acquisition of this genetic switch in CG23 K1 lineage allows inverse regulation of T3F and hypermucoid capsule production in a synchronized manner, allowing alternation between a hypermucoid phenotype with low T3F and a less mucoviscous capsule with high T3F phenotype to adapt to changing environments. This versatility could represent a powerful survival tactic for the dominance of the CG23-I. Our finding also addresses the conundrum of how hypervirulent K. pneumoniae regulates its capsule and fimbriae in relation to the environment to suit different stages of its life cycle.

The loss of KpVP upregulates T3F transcription
We examined global changes to the bacterial transcriptome after curing KpVP to determine how KpVP can potentially contribute to changes in bacterial phenotypes. We replaced the VagD toxin in the KpVP major toxin-antitoxin system of CG23-I proto typical strain SGH10, with a rhamnose-inducible RelE toxin cassette (24) to cure the plasmid. RNA-Seq analysis revealed that the 10 most upregulated genes included the T3F encoding mrk cluster (Table S1). RT-qPCR validation showed a significantly higher mrkA (major structural subunit) and mrkH (T3F transcriptional regulator) expression in SGH10 ΔKpVP (Fig. 1A). Transmission electron microscopy (TEM) imaging of SGH10 ΔKpVP showed surface filaments that were absent in SGH10 and SGH10 ΔKpVP ΔmrkA ( Fig. 1B through D). Flow cytometry validated the higher surface T3F protein expression in SGH10 ΔKpVP (Fig. 1E through L). This demonstrates that the KpVP encodes a T3F suppressive factor. As the capsule could conceal surface adhesions such as the type 1 fimbriae (15,23), we examined whether the thick hypermucoid capsule could also be masking T3F detection. Both the non-hypermucoviscous SGH10 ΔrmpA mutant and the capsule-null SGH10 ΔwcaJ mutant showed T3F levels comparable with SGH10. In contrast, only SGH10 ΔKpVP exhibits high levels of T3F ( Fig. 1M through P). This clarifies that the capsule is not masking detection and does not block the binding of the anti-MrkA/D antibody to T3F.
Salmochelin is formed from enterobactin by the action of a C-glucosyltransferase iroB on enterobactin (26). It is secreted through a periplasmic transport protein iroC (27). Uptake of iron-bound salmochelin is achieved through the receptor iroN (28), and iron is then released when the esterase iroD degrades salmochelin (29). We constructed singledeletion mutants of each gene and also deleted both iroB and its chromosomal homolog mceC (30,31). We validated the elimination of salmochelin synthesis by determining if the strains made functional microcin E492 (mccE492) that requires salmochelin to kill E. coli, as it is a siderophore microcin (31,32). Indeed, SGH10 ΔmceC ΔiroB no longer kills E. coli (Fig. S2A). However, SGH10 ΔiroB, ΔiroC, ΔiroD, and ΔmceC ΔiroB do not derepress T3F ( Fig. 2B through E). This shows that repression of T3F is independent of salmochelin biosynthesis.
Deletion of the iroN operon comprising both the promoter and the coding sequence (CDS) (SGH10 ΔiroN) derepressed T3F expression ( Fig. 2F and G). However, complementa tion of the iroN coding sequence (SGH10 ΔiroN/pMLBAD::iroN CDS) did not restore repression ( Fig. S2B through F), and deletion of the coding sequence (SGH10 ΔiroN CDS) did not fully derepress T3F to the same extent as the iroN operon deleted mutant (Fig.  2H). We found a 222-bp open reading frame (ORF) located 61 bp downstream of iroN which we named iroP. This ORF iroP is strongly controlled by the iroN promoter (Fig. S3A). Deletion of iroP upregulated T3F levels comparable with SGH10 ΔiroN (Fig. 2I). Comple mentation of iroP on pMLBAD in the ΔiroP mutant restored T3F suppression (Fig. S3B through G). Complementation of iroP in SGH10 ΔKpVP similarly restored T3F suppression ( Fig. 2J through O). This demonstrates IroP as the major, if not the only, T3F-suppressive factor on KpVP.

Fur represses IroP and capsule mucoviscosity in the presence of iron
Although Fur has been described as a transcriptional repressor, it has also been reported to positively regulate the mrkHI T3F regulatory loci (33,34). The salmochelin receptor IroN was shown to be repressed by Fur (35). Since iroP is controlled by the P iroN promoter, we examined the role of Fur in the regulation of T3F via iroP. We observed constitutive transcriptional expression of both iroN and iroP at the beginning of log phase of growth in SGH10 Δfur (Fig. 3A), validating Fur's role in suppressing iroN expression. We tagged a 1× FLAG sequence to native IroP in the KpVP to examine protein expression. Although IroP-FLAG was expressed in both SGH10 and SGH10 Δfur, we observed a significantly lower expression of IroP-FLAG in SGH10 during iron-rich growth ( Fig. 3B through E). This translates to a constitutively suppressed T3F in SGH10 Δfur even under iron-rich conditions; however, T3F levels are significantly upregulated in SGH10 under iron-rich conditions. When iroP is deleted, there is constitutively high expression of T3F regardless of iron availability ( Fig. 3F through M). This shows that Fur acts upstream of IroP in regulating T3F. Incidentally, the rmpADC operon is also regulated by Fur through a Fur binding box at the rmpADC promoter (36). Thus, we revalidated the role of Fur with regard to capsule overexpression and mucoviscosity mediated by the rmp operon. Although SGH10 could still produce capsule when grown under iron-rich conditions ( Fig. S4A through C), mucoviscosity decreased in SGH10 to levels comparable to SGH10 ΔrmpA and SGH10 ΔwcaJ controls. Deletion of fur resulted in a persistent mucoid phenotype regardless of iron concentrations (Fig. 3N). This validates Fur repression of rmp-mediated capsule mucoviscosity phenotype under iron-rich conditions.

Inverse regulation of T3F and hypermucoidy on bacterial phenotypes
Hypervirulent K. pneumoniae infection often originated within the patient's gut (37,38). The capsule hinders colonic epithelial cell adhesion but does not affect gut colonization (21). As the T3F is also involved in biofilm formation (16,39), we tested biofilm formation on abiotic surfaces and observed significantly higher biofilm formation by SGH10 under iron-rich conditions facilitated by the T3F (Fig. 4A). Under iron-rich growth, we observed a significantly higher adhesion of SGH10 to HT29 MTX-P8 colon epithelial cells, which was diminished when mrkA was deleted (Fig. 4B). These validate the importance of iron regulating the T3F to facilitate both biofilm and cell adhesion. Next, we investigated the role of IroP and hypermucoid capsule production in cell adhesion by infecting HT29 MTX-P8 cells with SGH10 versus various mutants. We observed significantly higher adhesion of SGH10 ΔKpVP to HT29 MTX-P8 cells. The lack of hypermucoid capsule mediated by rmpA did not increase adhesion to HT29 MTX-P8 colonic epithelial cells. Deletion of iroP slightly increased adhesion despite expressing high T3F, whereas the rmpA and iroP double mutant showed the highest adhesion comparable to the SGH10 ΔKpVP. Adhesion was attenuated in the ΔrmpA ΔiroP ΔmrkA triple mutant (Fig. 4C). This demonstrates that the hypermucoid capsule is impeding the T3F in colonic epithelial cell adhesion. Thus, the simultaneous suppression of the hypermucoid capsule and derepres sion of T3F can promote SGH10 binding onto different surfaces.

IroP regulation on bacterial virulence and gut colonization
We investigated whether IroP and its regulation of T3F play any role in bacterial virulence in a systemic model. By 48 h post infection, all mice succumbed to an infection dose at 10 5 CFU except for mice infected with SGH10 ΔKpVP or SGH10 ΔKpVP ΔmrkA ( Fig. 4D and E). This shows that KpVP is the major factor affecting systemic virulence. Aerobactin as encoded by the iuc operon on KpVP had been shown to be important in subcutaneous and intraperitoneal infection in mice (40). However, our aerobactin-deficient ΔiucCD mutant was as virulent as SGH10. The loss of virulence in KpVP is also not due to upregu lation of T3F as the IroP mutant as well as deletion of mrkA has no impact on virulence.
Our previous work showed that the rmpA-null mutant still demonstrated virulence in murine systemic infection (21). Thus, other factors on KpVP contribute to systemic virulence.
Since T3F was reported to be important in cellular adhesion (16,41), we tested singlestrain gut colonization in C57BL/6 mice via oral gavage. SGH10-infected mice showed (Continued on next page) Research Article mBio progressive decline in gut colonization by 20 d.p.i. There is a trend where the iroP mutant showed higher colonization than SGH10, the double mutant, and the triple mutant, but this did not reach statistical significance when all the groups were compared with one another ( Fig. 4F and G). The lack of an obvious effect with the rmpA mutation either suggests that hypermucoid capsule is irrelevant for gut colonization or that rmpA is already downregulated in SGH10 when in the gut. During day 13, when we compared by Student's t-test the difference between SGH10 ΔiroP and the iroP, rmpA, and mrkA triple mutant, we saw a small but statistically significant lower load in SGH10 ΔiroP ΔrmpA ΔmrkA triple mutant (Fig. 4H). Therefore, T3F is not an important contributor to gut colonization in this model.

IroP prevalence across K. pneumoniae isolates
In  (Table S2). Nevertheless, other capsular subtypes can also possess iro, such as three KLA KL5 isolates that possess iro1 and two KLA KL5 isolates that possess iro3 (Table S5). Identical or highly similar iroP sequences with at least 95% identity were found across all strains that have the iro cluster. However, two KLA strains were found to have a truncated iroP, with one having a 170-bp iroP with 100% identity to iroP in SGH10 and the other having 206 bp but 95% identity to SGH10's iroP ( Fig. 5A; Table S5). Prevalence of iroP is much lower in the BSI collection. However, strains with the iro cluster also possess highly similar or identical iroP, with one strain having a truncated iroP of 181 bp but 100% identity to SGH10's iroP sequence ( Fig. 5A; Table S2). There are some rare exceptions of strains having iroP but no iro and vice versa-there is one KLA isolate with iroP but no detectable iro cluster, whereas there are two BSI isolates with iro cluster but no iroP, and three BSI isolates having iroP but not the iro cluster. Our search for the complete sequence of iroP in bacteria other than Klebsiella species turns out to be limited (Fig. 5B). The iro cluster together with iroP was found in the chromosome of Salmonella enterica isolate PNUSAS065329 (accession number AACZPG010000001), as well as in an unnamed plasmid in a UTI E. coli strain 2020CK-00232 (accession number CP107259). IroP in these strains is annotated as hypothetical proteins. Among the KLA isolates, the K2 strains NUH04 and TTSH29 both possess iroP on their KpVPs. Nucleotide sequence variations were seen among the various species; however, mutations along the N-terminal end of the protein are synonymous, while missense mutations were found toward the C-terminal region (Fig.  5B). These results suggest hypervirulent K. pneumoniae co-acquired iroP together with iro via mobile genetic elements.
We then compared T3F expression in several KLA strains; SGH04 possesses iro1 just like SGH10, NUH04 and TTSH29 possess iro2, and SGH07 and TTSH21 possess iro3. Interestingly, T3F expressions varied across these strains. High iron increased T3F levels

IroP suppression of T3F in K. pneumoniae
We then examined whether the suppression of T3F could be due to IroP's suppression of the promoter activity of the mrkABCDF and mrkHI operons. We utilized mrkABCDF and mrkHI promoter (P mrkA and P mrkH ) fusions to superfolder green fluorescence protein (sfGFP) reporters. In E. coli, the induction of iroP did not significantly decrease sfGFP fluorescence of both P mrkA -sfGFP and P mrkH -sfGFP (Fig. 6A). When tested in the SGH10 ΔKpVP ΔwcaJ mutant, only P mrkA but not P mrkH was suppressed by iroP (Fig. 6B). This suggests that iroP alone may not be sufficient to act on the promoters and likely requires other co-factors found in K. pneumoniae but not in this lab strain of E. coli. A capsule-null mutant background was used as the absence of hypermucoid capsule allows reliable recovery of bacteria via centrifugation as well as prevent possible interference of sfGFP fluorescence signals. As our previous experiments with IroP were done in the CG23-I lineage SGH10, we examine whether the sufficiency of IroP in suppressing T3F is seen in other lineages. NUH29 is a non-hypervirulent K. pneumoniae strain without the KpVP or other hyperviru lence-associated factors (45,46). It shows high T3F expression without addition of iron. The exogenous introduction of iroP from SGH10 suppressed T3F in NUH29 grown under normal IM (Fig. 6C through H) as well as in the K5 iro3 strain TTSH21 when grown under iron-rich conditions (Fig. S5A through G). This shows the adequacy of IroP for T3F suppression in K. pneumoniae in strains that do not have the iroP gene or those with a different iro locus. However, when iroP was deleted in a K2 hypervirulent strain TTSH29 with negligible T3F expression, T3F was still undetectable (Fig. 6I through L). This points to heterogeneity in IroP regulation.
We further utilized AlphaFold (47,48) to generate a protein structure prediction of IroP. The predicted structure of IroP is visually represented using color coding based on the predicted local distance difference test (pLDDT) score. The pLDDT scores ( Fig. 6M; Fig.  S6A) indicated a high level of confidence in the predicted structure. Moreover, the predicted alignment error (PAE) score, which measures the positional error of amino acid residues in angströms (48), demonstrated a low error rate in the domain positions (Fig.  S6B). Prediction using DeepFRI (49) on the structure reveals a nucleic acid-binding function with a significant score of 0.58. IroP also has a different protein folding in comparison with the predicted structures of MrkH (Fig. 6N) and MrkI (Fig. 6O). This is further supported by protein structure alignment using TM-align (50), which showed a low score of <0.5 when IroP was compared with both MrkH and MrkI, representing

DISCUSSION
Plasmids can benefit bacterial hosts by gaining antibiotic or toxin resistance. They can also affect metabolism, nutrient transport, or virulence gene expressions. For example, the plasmid p24835-NDM5 can upregulate carbohydrate metabolic genes in hyperviru lent K. pneumoniae (51), whereas pAB04-1 and pAB3 suppress the type VI secretion system in Acinetobacter baumannii (52,53). In Rhodococcus equi, plasmid-encoded transcriptional regulators modify the expression of several chromosomally encoded virulence-associated traits to enhance survival of both the bacterium and the plasmid in macrophages (54). These examples are evidence that plasmid-encoded regulators may have important fitness consequences for the plasmid and the host cell (55). We have much to understand about the role of KpVP in K. pneumoniae fitness and virulence. Although some K. pneumoniae strains have chromosomally incorporated iuc, iro, and rmpA (56, 57), a majority of K1 and K2 hypervirulent K. pneumoniae retain the KpVP. Apart from the known rmp operon and siderophores, KpVP contains many other genes of unknown functions. Changes in transcriptomic profile due to KpVP loss provide strong indications of cross talk between KpVP and the chromosome, and could reveal regulatory pathways to circumvent fitness costs to the bacterial host. Interestingly, the biggest change we observed with the loss of KpVP is the upregula tion of T3F in the bacteria. T3F is an extended surface appendage found on Enterobacter iaceae and known to be an important factor in biofilm formation on abiotic surfaces and during in vitro cell adhesion by binding to collagen V (58,59). The main structure consists of a major subunit MrkA with an adhesion tip MrkD and is assembled through an MrkB chaperone and an MrkC chaperone usher protein (60,61). MrkF is likely part of the T3F assembly mechanism with unknown functions (62). The entire mechanism is regulated by iron and c-di-GMP, where the latter controls the activity of the positive transcriptional regulator MrkHI. MrkHI in turn binds to the mrkABCDF promoter (63). Fur had previously been postulated to regulate T3F through mrkHI and a hypothetical unknown repressor of mrkABCDF (34). Our findings that KpVP repression of T3F occurs via IroP support this hypothesis, and we believe IroP is that "hypothetical repressor. " However, what is surprising is that IroP is nestled within the iro cluster on the large virulent plasmid. Its coexpression with IroN allows its regulation by Fur and iron. We also demonstrated an inverse functional relationship between adhesions and capsule mucoviscosity in influencing biofilm and adhesion. So far, it is not known how hypermucoviscous K. pneumoniae navigate its interaction with the host through adhesion and colonization while having a thick and mucoid capsule. The assumption is that capsule must be regulated to expose fimbriae and adhesions for colonization to take place. This may stem from how both are organized on the bacterial surface as T3F interferes with hydrogen bonding between capsular polysaccharides (64). However, as we have shown, T3F is not constitutively expressed in CG23-I, as this would be futile during stages when capsule mucoviscosity is fully expressed. Therefore, the synchronization of the on-off switch of T3F and the degree of capsule hypermucoviscosity prevent the interference of the Research Article mBio hypermucoid capsule with T3F and allow bacterial lifestyle adaptation to a biofilm state, or to establish cellular adhesion and colonization. In K. pneumoniae ST258, selection pressure for persistence in the bladder in urinary tract infection was conferred by capsule-null mutations that resulted in increased invasion and biofilm formation in the bladder (65). Conversely, blood infections select for those with mutations in wzc that resulted in hypermucoviscous capsule that resists phagocytosis. For SGH10, alternation of these traits associated with each phase of infection could be accomplished by the convenience of the Fur-IroP switch instead of through mutant selection, which would confine the mutants to a particular niche. High biofilm formation in SGH10 under iron-rich condition is attributed to dere pressed T3F expression and downregulated capsule mucoviscosity. Interestingly, T3F overexpression on its own does not significantly improve biofilm formation (64). We propose that adhesion can only be achieved when there is an optimal balance between fimbriae levels and the capsule mucoviscosity, and both are mediated by iron availability. Siderophores can play an important role in iron uptake since it contributes to the majority of iron in P. aeruginosa and is important for initial establishment and biofilm formation (66). In UTI89 E. coli, siderophore iron uptake promotes the production of metabolites required for biofilm formation (67). K. pneumoniae strains also produce more siderophores during the planktonic phase than during the sessile state (68). Hyperviru lent K. pneumoniae strains produce more siderophores than classical strains and can form biofilms more robustly (69). We postulate that during the planktonic phase, iron is acquired leading to a gradual increase in intracellular iron and formation of Fur-Fe complexes. This will gradually repress hypermucoviscosity and increase T3F expression. At this point, bacteria then switch from the planktonic and likely disseminating phase to a sessile state as it colonizes a niche through surface adhesion and/or biofilm formation. This switch may take place in the host gut, given that diet, host factors, and gut micro biome can cross-regulate iron availability. Generally, free iron is extremely low (10 −24 M) in humans, and most of the iron is bound to high-affinity host proteins. However, higher concentrations of iron are expected to be in the colon as the majority of iron that is not absorbed in the duodenum ends up there (70). In addition, fecal iron excretion in adults who consumed a diet containing a normal concentration of iron at 6-8 mg/d was about 7.5 mg/d (1.3 × 10 −4 M/d) (71). It is possible that in the colon, there is enough iron from a normal diet to induce Fur activation in hypervirulent K. pneumoniae, resulting in high T3F and low capsule mucoviscosity, thereby favoring colonization.
However, despite the clear contribution of T3F and capsule mucoviscosity during in vitro cell adhesion and biofilm experiments, the difference in colonic colonization between bacteria with derepressed T3F and mutants with both derepressed T3F as well as deletion of rmpA was not seen. This is likely due to the already repressed rmpA expression in the colon. The difference contributed by the SGH10 iroP deletion mutant and SGH10 in colonization was also not statistically significant, and this would be expected if iroP was already suppressed by Fur in SGH10. There was a small difference between the ΔiroP and the ΔiroP ΔrmpA ΔmrkA triple mutant, showing that deletion of T3F results in poorer intestinal colonization at day 13, meaning that T3F would have been upregulated in that environment but its role in intestinal colonization remains limited. It is also conceivable that the advantage of this reversible switch is more evident when bacteria are transiting between different environments not tested in our study. We have not explored other niches, such as in the lung, liver, or in extracellular niches such as soil where iron sources could be abundant, especially in acidic soil (72). It would be The different lineages of iro loci found in K. pneumoniae, E. cloacae, or E. coli (42) suggest that the regulation of IroP differs across different iro loci. Indeed, T3F expression differs across hypervirulent K. pneumoniae sequence types as well as non-hypervirulent K. pneumoniae isolates. Furthermore, we do not see T3F upregulation under iron-rich condition in non-CG23-I hypervirulent K. pneumoniae that possess IroP as significantly as seen in CG23-I. Deletion of iroP also does not necessarily lead to T3F upregulation in strains such as TTSH29. This suggests that iroP is regulated differently in these strains from what is seen in the CG23-I sublineage. It is interesting to note that IscR, an ironsulfur cluster-containing transcriptional regulator, has been documented to control T3F and capsule production through the effect of iron in a K2 K. pneumoniae isolate CG43 (75). Apo-IscR activates MrkA transcription, whereas holo-IscR which is formed when IscR binds iron-sulfur cluster represses T3F via the direct repression of MrkHI promoter. Apo-IscR does not directly bind to the mrkA or mrkHI promoters (75). However, holo-IscR activates capsule synthesis genes (76). This means that under iron-rich conditions, T3F is repressed, whereas capsule synthesis is increased. This regulatory control is in direct contrast to the effect of Fur and IroP. The caveat is that Fur represses IscR (75) during iron-rich conditions and supersedes the effects of IscR. Perhaps only in iron-poor conditions, IscR could be opposing the effect of IroP because apo-IscR activates T3F whereas IroP would be derepressed and suppresses T3F. CG43 also possesses an identical IroP to SGH10. It is possible that these competing levels of regulatory control could lead to heterogeneity in T3F regulation by IroP. In our promoter-sfGFP fusion assays, we did not see IroP exerting repression on the mrkHI promoter, although it repressed the mrkA promoter. This is somewhat surprising as there was transcriptional upregulation of mrkHI in the KpVP mutant (Fig. 1A). One limitation of this work is that we performed the promoter-fusion assays in a capsule-null mutant background, and capsule's absence might have affected the repression on mrkHI. Alternatively, IroP mainly acts on the mrkABCDF operon, and the increase in mrkHI seen in Fig. 1A could be due to factors affected by the absence of KpVP. For example, there may be fewer holo-IscR resulting in an increase in mrkHI since the loss of KpVP can reduce iron levels due to the absence of iron uptake systems encoded on the plasmid.
Although we did not find any DNA-binding domains using domain and motif searches in IroP, the new DeepFRI platform based on graph convolutional networks on 3D structures shows a strong prediction of DNA-binding function (49). However, the dissimilarity of IroP protein folding compared with MrkH and MrkI suggested that IroP may bind to DNA differently than MrkH and MrkI. This may not be surprising as IroP suppresses, whereas MrkH and MrkI form a complex to upregulate the mrkA operon. Nevertheless, further work would be required to examine whether IroP, with or without cofactors, could interact with the promoter region of the mrkA operon to directly mediate repression. were shaded in dark blue, scores between 70 and 90 were shaded in lighter blue, low confidence scores between 50 and 70 were shaded in yellow, and those with the lowest confidence scores (<50) were shaded in orange.

Research Article mBio
Acquisition of salmochelin in hypervirulent K. pneumoniae might have evolutionarily yielded an advantageous adaptation to control T3F expression in relation to iron. IroP may be described as a "genome hitchhiker, " given that its function does not seem aligned to the rest of the genes in the salmochelin operon. While it is first identified in Salmonella, the iro locus is detected in hypervirulent K. pneumoniae, E. cloacae, and E. coli (42). The packaging of both iro and iroP genes into mobilizable genetic elements such as plasmids or pathogenicity-associated islands has the potential to accelerate the spread and acquisition of a T3F switch, which can provide pathogens with enhanced pathogen esis. In fact, hybrid virulence plasmids that acquired conjugation machinery have been reported to conjugatively transfer to E. coli from hypervirulent K. pneumoniae (77). A foodborne E. coli pathogenic strain EC1108 was also found to possess a virulence plasmid harboring a similar iuc-rmp-iro region as those found in hypervirulent K. pneumoniae (78). The recent release of an unnamed plasmid sequence that has the identical sequence of iroP as SGH10 in human urine E. coli isolate implies that the salmochelin cluster is capable of disseminating across different strains in the urinary tract. This could be of major concern as T3F is a well-known virulence factor during urinary tract infections. Rising reports of mobile elements carrying the iro locus could mean that other Enterobacter iaceae species apart from K. pneumoniae and E. coli can also acquire the salmochelin cluster. Interestingly, a recent report documented the wide distribution of rsm genes and its homologs on plasmids in diverse taxa (79). This is an example of a plasmid global translational regulator evolved to control the bacterial behavioral switch from a motile to a sessile lifestyle and bacterial metabolism through plasmid chromosomal cross talk. The authors propose that plasmids may commonly control bacterial lifestyle in the clinic, agricultural settings, and other environments.
In conclusion, we discovered a novel regulatory pathway mediated by the Fur-regula ted IroP found on the KpVP. IroP likely acts with other transcription factors to regulate the T3F in a plasmid-chromosomal cross talk. We propose that Fur-Fe-mediated regulation of T3F and hypermucoid capsule provides a reversible transcriptional switch for bacteria navigating through different environmental niches with variable iron availability. Iron limitation inactivates Fur, allowing expression of both rmpADC and iroP, which leads to T3F suppression, in turn impeding biofilm formation while facilitating hypervirulent K. pneumoniae dissemination. Iron repletion triggers Fur-Fe binding to the relevant promoters, suppressing iroP and rmpADC while upregulating mrkABCDF leading to high T3F production with low hypermucoid capsule and promoting biofilm formation, epithelial cell adhesion, and perhaps gut colonization (Fig. 7). During systemic circulation where free iron sources are limited, the phenotype of hypermucoid capsule and low T3F dominates to aid in dissemination. This regulatory network allows CG23 hypervirulent K. pneumoniae to alternate between a hypermucoid phenotype versus a phenotype with high T3F and a basal level of capsule production, modulating its adhesion to various surfaces in different niches. This presents another powerful example of how plasmid-encoded regulators can become tightly interwoven with chromosomal gene regulation through a complex cross talk that is coupled to a key and valuable nutrient iron. This regulation endows the bacteria with new-found abilities to deftly navigate through harsh conditions that likely contribute to their evolutionary success. Whether this IroP regulatory switch will spread and be integrated into more lineages and species through mobile genetic elements remains to be seen.

Deletion of KpVP in SGH10
Kanamycin resistance and rhamnose-inducible relE cassette from pSLC-217 (24) as well as vagD upstream and downstream flanking sequences were amplified using Q5 Polymer ase (NEB, Ipswich, MA, USA) and assembled in pR6KmobsacB zeo (R6K replication origin replaced pMB1 in pK18mobsacB (80) with zeocin resistance) using the HiFi Assembly Master Mix (NEB). Plasmids were electroporated into S17 E. coli, and colonies selected on LB agar with 100 µg/mL kanamycin and 50 µg/mL zeocin. Colonies were mixed at 10 E. coli : 1 SGH10 ratio for 24 h and plasmids carrying SGH10 were selected on LB agar plates + 100 µg/mL kanamycin + 100 µg/mL carbenicillin. A chosen colony was then grown for 24 h at 37°C in LB with 20% sucrose and plated on LB agar + 100 µg/mL kanamycin to counterselect for colonies with rhamnose inducible RelE cassette. Selected colonies were subsequently grown in LB with 2% rhamnose to induce RelE and then plated on normal LB agar plates. Cured KpVP mutant strains were screened for seven genes (parA1, copC, tRNA, coA, rmpA, terC, and dppC) by qPCR. The list of primers for amplification of relE cassette and qPCR primers are listed in Tables S2 and S3, respectively.

Construction of gene mutants
Gene deletion in SGH10 was performed as previously described (21). In brief, Q5 amplified upstream and downstream regions of the target gene together with Research Article mBio pR6KmobsacB were assembled and introduced via conjugation into SGH10 using E. coli S17 as donor. Selection was done with 100 µg/mL kanamycin and 100 µg/mL carbenicil lin. Negative selection was performed by growing in LB with 20% sucrose and spread on LB agar. Colonies were screened using PCR to validate gene deletion. All primers used are listed in Table S4.

RNA isolation, cDNA reverse transcription, and quantitative real-time qPCR
Overnight cultures were subcultured into new IM media and grown for 2 h at 37°C. Trizol-chloroform method was performed to isolate RNA using Purezol (Bio-rad, Hercules, CA, USA) and the Purelink RNA Mini Kit (ThermoFisher Scientific). cDNA was synthesized using the Maxima H Minus Synthesis Kit (ThermoFisher Scientific). Real-time qPCR was done using iQ SYBR Green Supermix (Bio-rad). Relative RNA transcripts were normalized against the wild-type control using the 2 -ddCt threshold cycle method (81) with the recA as the reference gene and rpoB as the internal housekeeping control. All qPCR primers are listed in Table S5.

TEM sample preparation and imaging
OD 600 of overnight bacterial cultures were measured and diluted to 10 8 CFU/mL. Cells were pelleted at 2,000 × g centrifugation for 7 min and fixed with 2.5% glutaraldehyde in 1× phosphate buffered saline (PBS) (pH 7.3) for 2 h at 4°C. Cells were washed twice with 1× PBS (pH 7.3) for 20 min on ice. Fixed cells were air dried on the transmission electron microscope formvar and carbon-covered grid for 5 min, followed by staining with 1% phosphotungstic acid for 1 min. Images were taken with the help of the electron microscopy unit.

Flow cytometry
A volume of 10 7 CFU/mL of overnight culture was stained on ice in the dark for 1 h with anti-MrkA/D rabbit primary antibodies at 1:650 µL ratio. Secondary goat anti-rabbit Alexa-488 conjugated antibodies (ThermoFisher Scientific) were added at 0.5:400 µL ratio in 1 × PBS and incubated on ice in the dark for 40 min. Three hundred microliters of 1% formaldehyde (Sigma, St Louis, MI, USA) were added dropwise to fix cells and incubated at room temperature for 15 min. Formaldehyde was removed, and each sample was resuspended in PBS for analysis. All samples were acquired for 100,000-200,000 events and gated for K. pneumoniae. Detection of target population was gated based on the rabbit IgG isotype control of the individual strains. The side-scatter area depicting detection of bacterial cells was plotted against Alexa-488 fluorescence.

Mucoviscosity test using low-speed centrifugation
Low-speed centrifugation test was conducted according to our protocol as previously described (21). In brief, 10 9 CFU/mL of overnight cultures was spun down at 2,000 × g for 10 min. Two hundred microliters of supernatant were aliquoted into a 96-well polystyrene plate, and absorbance was measured at OD 600 .

Bacterial cell adhesion
HT29-MTX P8 were seeded at a cell density of 0.25 × 10 6 on 24 well plates. Bacterial infection of cells was done at multiplicity of infection of 10 bacteria:1 cell and incubated for 30 min at 37°C. After washing thrice in 1× PBS, cells were lysed with 0.1%-0.25% Triton X-100 (Sigma). Appropriate dilutions were plated on lysogeny broth agar (LBA) and incubated at 37°C overnight.

Intraperitoneal mouse infection
A volume of 10 5 CFU in 100 µL of bacterial culture was injected into the intraperitoneal cavity of each mouse. Mice were monitored twice daily until day 7 post infection. Based on the body condition score system (5-healthy active mouse to 1-severely ill, not able to keep upright), mice that fell within body condition score of 2 were euthanized and harvested for organs. Organs were homogenized using 2.8-mm ceramic beads (Omni International, Kennesaw, GA, USA) in 1 × PBS using a Beadrupter (Omni International). Appropriate dilutions were plated on Klebsiella Selective Agar (KSA).

Gut colonization
Ampicillin pre-treatment of 2.5 mg in 100 µL was administered to each mouse via oral gavage for 5 d. After pre-treatment, all mice were infected via oral gavage with 10 6 CFU bacteria in 100 µL. Each mouse was individually separated into different cages to prevent coprophagy. Stools from all mice were collected every 2 d post infection and homogenized using 1.4-mm ceramic beads (Omni International) in 1 × PBS. Appropriate dilutions were spread plated on KSA plates.

Gene alignment using Pathogenwatch and BLAST
The sequence of iroP was obtained using the ORFfinder prediction (83). Sequences of the entire iro operon as well as iroP were queried using Pathogenwatch (84) and aligned using BLAST to databases obtained from the A-KLASS study comprising 133 wholegenome sequences (43) and the collection comprising 365 whole-genome sequences (44), which we refer to as "BSI. " Sequences of A-KLASS KLA isolates NUH11, NUH27, NUH29, SGH04 and SGH07 were previously uploaded to Genbank under Bioproject PRJNA351910. Sequences of the remaining KLA isolates were uploaded to Genbank under Bioproject PRJNA956314.

sfGFP promoter activity measurement
Overnight cultures, grown for 24 h, were spun down and diluted to OD 600 of 0.5 in 1 × PBS. Two hundred microliters of the culture were aliquoted into 96-well black plates with clear bottom (Corning, New York, NY, USA) with technical duplicates. GFP fluorescence intensity was measured at 485 nm emission and 510 nm excitation.

Protein structure prediction using AlphaFold
IroP protein structure was predicted via AlphaFold (v2.3.1) (47), using Docker on an in-house system. To obtain the pairwise confidence measure, multiple sequence alignments were performed against the full database (db_preset = full_dbs) and model_preset as monomer_ptm. The predicted structure was visualized using ChimeraX (85), while python script was utilized to generate pLDDT and PAE graphs.

Statistical methods
Student's t-test was used to compare means and standard deviation of two groups. For three groups or more, a one-way or two-way ANOVA with Dunnett's multiple comparison was performed to compare means and standard deviation relative to the wild type. For mouse colonization model, each group was individually compared with a second group at each time point, using Student's t-test to compare means and standard error of means. Survival curve was analyzed with Mantel-Cox log rank test. All graphical and statistical methods were performed using Prism Graphpad 9 (GraphPad Software, La Jolla, CA, USA). 10 Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore

DIRECT CONTRIBUTION
This article is a direct contribution from Yunn-Hwen Gan, a Fellow of the American Academy of Microbiology, who arranged for and secured reviews by Trevor Lithgow, Monash University, and Olaya Rendueles, Institut Pasteur.

DATA AVAILABILITY
All data needed to evaluate the conclusions in the paper are present in the paper and the supplemental material. Sequences of KLA isolates were uploaded to Genbank under Bioproject PRJNA956314.

ETHICS APPROVAL
All animal experiments utilized 7-to 8-week-old female C57BL/6 mice. The mice were pre-weighed upon arrival and allowed to proceed with experiment if weight was deemed healthy between 17 and 20 g. All animal care and infection protocols have been approved by the Institutional Animal Care and Use Committee (IACUC) in NUS under protocol R18-0252, under National Advisory Committee for Laboratory Animal Research (NACLAR) guidelines.

ADDITIONAL FILES
The following material is available online.