The Neisseria meningitidis iron acquisition protein HpuA moonlights as an adhesin and inhibits host cell migration

Neisseria meningitidis can cause meningococcal disease, a rapidly developing and potentially fatal infection. Despite this, it normally resides as a commensal in the nasopharynx of healthy individuals. The mechanisms by which meningococci access deeper tissues remain unknown. Epidemiological data suggest that mucosal disruptions increase the risk of meningococcal disease. We previously investigated whether meningococci inhibit host cell wound repair, enhancing invasive disease risk. Here, using genome sequencing and a collection of closely related household isolates that differ in their ability to inhibit host wound repair, we identify the responsible meningococcal factor. This protein, HpuA, has previously been characterized as part of a bipartite heme acquisition transporter. We constructed mutants to demonstrate that HpuA, but not HpuB, inhibits wound repair, acts as an adhesin for epithelial cells, and promotes cellular invasion. We showed this was not due to iron starvation resulting from the bacteria, differences in growth rate, or manipulation of host haptoglobin. Heterologous expression of HpuA in E. coli mediated adherence to 16HBE cells in an HpuA-dependent manner and conferred an aggregative phenotype onto E. coli, suggesting that HpuA may play a role in the formation of microcolonies on host cells. We also demonstrated that iron supplementation of meningococci restored the inhibition of wound repair in strains lacking HpuA (NZCM112, ΔhpuA mutant) to levels seen with the wild type. This was also seen with unrelated carriage strains previously shown not to inhibit wound repair. Iron supplementation also increased adherence and invasion of meningococci for strains lacking HpuA, while not affecting those that expressed HpuA. These findings suggest there may be a second meningococcal protein that inhibits wound repair. Together, these results suggest that HpuA is an important meningococcal virulence factor with multiple moonlighting functions, mediating adherence, invasion, inhibition of wound repair, and bacterial aggregation. Author Summary Neisseria meningitidis causes meningococcal disease, a potentially fatal and rapidly developing illness that most often occurs in children. Despite this, the bacteria are frequently carried harmlessly as part of the normal airway microflora in healthy people, only rarely causing invasive disease, which involves replication in the bloodstream or central nervous system. It remains unknown precisely how the bacteria reach the deeper tissues from the airways, though some epidemiological evidence suggests that wounds or disruptions to the airways may increase risk. Here, we show that a N. meningitidis protein, HpuA, moonlights from its usual job of acquiring nutrients from the host, to enable the bacteria to adhere to and invade host cells, as well as inhibiting wound closure. Furthermore, we also show that meningococci that lack HpuA acquire the ability to inhibit wound repair when they are supplemented with iron, suggesting that there are additional meningococcal proteins to be discovered that may inhibit wound repair.


Introduction
In the current study, we used closely related isolates that differed in their ability to inhibit 126 wound repair to identify the factor responsible. This group of isolates included one collected 127 from a patient and two similar nasopharyngeal carriage isolates from the patient's healthy 128 household contacts. The three isolates were indistinguishable by conventional laboratory 129 typing methods, but one of the carriage isolates had lost the ability to inhibit wound repair. 7 and host epithelial cells, suggesting there may be additional N. meningitidis factors affecting 144 wound repair and adherence to be discovered. These results are the first to demonstrate a 145 moonlighting function for a N. meningitidis iron acquisition protein and suggest that there are 146 likely others to be identified.

148
Closely related household contact isolates differ in their 149 ability to inhibit epithelial cell wound repair 150 As part of our efforts to identify the N. meningitidis factor responsible for the inhibition of 151 wound repair, we screened a collection of isolates that were derived from a household 152 contact study, carried out in Auckland, New Zealand, in the 1990s, during the peak of the 153 serogroup B N. meningitidis epidemic (36). Patient isolates were cultured from blood or 154 cerebrospinal fluid samples from the patient, while carriage isolates were collected from 155 healthy household contacts of the patients by nasopharyngeal swabs. For this study, we 156 focused on disease and carriage isolates that were indistinguishable by standard laboratory 9 hours of co-incubation, we asked whether the NZCM112 phenotype we observed could be 174 due to impaired growth under the assay conditions. While we confirmed that similar numbers Deletion of hpuA reduces inhibition of host cell wound repair 234 The genome sequence data revealed that an intact HpuA protein was not predicted to be 235 synthesized in NZCM112, while previous studies have shown that when the hpuA gene is in 236 the phase-variable "off" configuration, hpuB expression is lost or reduced (41). As the hpuA 237 gene was shown to be intact in NZ97/052, our aim was to generate an allelic replacement of 238 both hpuA and hpuB. The operon organization of these genes made it a challenge to ensure 239 that only hpuA was targeted, particularly since complementation of genetic mutations in N. 240 meningitidis is not easily done via plasmid. We therefore constructed three allelic 241 replacements in the NZ97/052 parental background: two single mutants in hpuA and hpuB 242 alone, and a double hpuA-hpuB mutant. While disruption of hpuA is likely to have a polar 243 effect on the expression of hpuB, the disruption of hpuB, the downstream gene, is not 244 predicted to alter hpuA expression (43). All three mutants were generated through the 245 insertion of an antibiotic resistance cassette into the targeted coding region. We then tested 246 the mutants to see if they recapitulated the phenotype of NZCM112.

248
The three mutants were first compared to the wild type in an in vitro wound repair assay. As

251
There was no defect in wound repair inhibition for the hpuB mutation. These results suggest NZCM112. As shown in Fig 2B, the hpuA mutant had a slight growth defect, compared to 256 the wild type parent, when grown in liquid culture over a time course; however, it did not 257 have the same defect as NZCM112. These results suggest that the additional SNP variants 258 seen in NZCM112 cause this isolate to grow more slowly during exponential phase, and to 259 reach stationary phase at a lower cellular concentration. The basis of this reduced growth 260 remains unknown but is likely due to additional factors than just the loss of hpuA. Because

261
HpuA is known to play a role in iron acquisition from the host, we also assessed growth in 262 the presence of epithelial cells, under conditions that mimicked those of the wound repair 263 assay. The bacterial strains were grown in co-culture with epithelial cells and the number of 264 viable colonies was assessed at the end point of 16 hours. As with the liquid culture, the 265 growth of NZCM112 was impaired relative to either the NZ97/052 wild type or the isogenic 266 hpuA mutant (Fig 2C). While the hpuA mutant may have had a slight growth impairment 267 relative to NZ97/052, this did not reach statistical significance.

269
These data suggested that the growth defect of NZCM112 was not due to the loss of HpuA, 270 but rather was the result of other SNPs in the genome. However, because of the role HpuA 271 plays in the acquisition of iron for the bacteria from the host, we asked if iron 272 supplementation could rescue the growth defect of NZCM112. Growth was measured by 273 OD600 in liquid standing culture over a time course ( Fig 2D) and in co-culture with epithelial 274 cells at an endpoint via CFU (Fig 2E). Both strains, NZ97/052 and NZCM112, reached 275 slightly higher cell density when the culture was supplemented with 20 µM FeSO4, but the 276 growth of NZCM112 was not rescued by iron supplementation in either case. As seen in Fig   277   2E, similar results were seen with both Fe 2+ (FeSO4) and Fe 3+ (FeCl3).

279
Together these data demonstrate that the N. meningitidis iron acquisition protein HpuA plays 280 a role in the inhibition of wound repair in epithelial cells, and that this effect is not due to 281 impaired growth of the isolate in the conditions tested.

292
in the absence of HpuB, detectably bind hemoglobin, haptoglobin-hemoglobin complex, or 293 apo-haptoglobin, HpuA is known to be critical for optimal recognition and use of ligands (46).

295
The majority of the iron in the human body is stored in hemoglobin and ferritin complexes, 296 and free iron is quickly sequestered by transferrin or lactoferrin. About two thirds of iron in 297 the human body stored in hemoglobin in erythrocytes. If it is released following lysis, 298 haptoglobin binds with high affinity, protecting the body from the oxidative properties of 299 heme and facilitating its removal by macrophages (47, 48). In addition to its role as an acute 300 phase blood protein, haptoglobin is recognized to have a role as an antioxidant and to play a 301 role in the response to infection (49). Haptoglobin has also been shown to have 302 bacteriostatic properties, particularly against Gram-negative pathogens, and to be induced 303 by microbial antigens, such as LPS (49, 50). It has also been shown to be expressed in 304 airway epithelial cells, and in one model system, haptoglobin was shown to increase the rate 305 of cellular migration, possibly by providing iron to migrating cells (49,51

327
where the highest haptoglobin concentration (1000 ng/ml) reduced bacterial growth for P. 328 mirabilis and P. aeruginosa to 50% compared to samples without haptoglobin, inhibition of 329 growth was not seen for any of the meningococcal isolates tested (Fig. 3B). The

337
We did not discover any role for HpuA in enhancing bacterial growth through iron the interactions of the wild type NZ97/052 parent and the hpuA mutant with host cells. We strain was calculated as a percentage of the initial inoculant, also measured by CFU 356 enumeration ( Fig 4D). The NZ97/052 isolate displayed about 4.6-fold higher levels of 357 association with 16HBE cells relative to NZCM112 (p < 0.0001). The NZ97/052 hpuA 358 isolate significantly decreased association with 16HBE cells relative to the wild type parent accessing nutrients or evading the immune response. We therefore carried out gentamicin 364 protection assays to determine whether the increased bacterial association led to increased 365 cellular invasion. The assay was carried out similarly to the cell association assay, but the 366 bacteria and cells were co-incubated for a longer period (4 hours, rather than 1 hour), after 367 which gentamicin was added to kill any extracellular bacteria. The NZ97/052 isolate

383
We wished to restore the hpuA gene to either NZCM112 or the NZ97/052 hpuA mutant, to 384 see if the adherent phenotype could be restored. However, we were unable to complement 385 the hpuA mutation by restoring it on a plasmid, due to a lack of access to replicating 386 plasmids for N. meningitidis. Instead, we opted to express HpuA in E. coli BL21(DE3), which observed that E. coli strains are toxic to bronchial epithelial cells, so we were unable to do 392 any experiments that involved a lengthy co-incubation (e.g., wound repair inhibition or 393 gentamicin protection assay). During initial experiments, the 16HBE cells were monitored via 394 microscopy to ensure that they retained integrity. The expression of HpuA was induced with 395 IPTG for one hour, as this was found to result in sufficient expression of HpuA with no

406
The clusters of NZ97/052 bacteria adherent to epithelial cells, which was not seen with the 407 hpuA mutant or the carriage CM112 strain (Fig 4A-C), led us to hypothesize that HpuA 408 could be playing a role in bacterial aggregation. We assayed the ability of HpuA expression 409 to confer an aggregative phenotype on the host E. coli strain, which normally does not auto-410 aggregate. Sedimentation profiles ( Fig 5C) showed that the sedimentation rate of the strain 411 expressing HpuA was significantly greater than the non-expressing E. coli control (p < 412 0.0001 at hours 5 and 6). This assay suggests that HpuA is a major determinant of the mediating adherence to host epithelial cells and aggregation. Cumulatively, these functions 418 enable the bacteria to adhere better to host cells and to form microcolonies that may 419 promote colonization and persistence in the airways.

438
Wound repair inhibition remained unchanged for NZ97/052, and the overall rate of wound 439 closure in the absence of bacteria did not change with iron supplementation. Instead, the 440 addition of iron appeared to switch the NZCM112 and NZ97/052 hpuA strains to a meningococcal strains that lack HpuA, not the host cells.

445
We tested a range of concentrations of FeSO4 supplementation to determine if there was a 446 dose-dependent effect. However, as seen in Fig 6B, it did not appear to be dose-dependent, 447 as no effect was seen with 0.1 µM FeSO4 added, while similar effects were seen with 448 supplementation with 1 µM and 5 µM FeSO4. This indicates more of an "on-off" transition 449 mechanism, rather than a dose-dependent response.

451
We asked whether this phenomenon was limited to our closely related isolates, or if it could 452 be observed in other, unrelated, meningococcal isolates that do not inhibit wound repair. A 453 range of unrelated carriage isolates (obtained from nasopharyngeal swabs from healthy 454 volunteers) that we had previously shown not be able to inhibit wound repair, were tested 455 with and without supplementation with 20 µM FeSO4 (35). These strains belong to different 456 sequence types and a range of serogroups (including serogroups B, C, and non-groupable).

457
As seen in Fig

486
Nutritional immunity, the withholding of key nutrients by the host, has long been known as an HpuAB or HmbR receptors, though strains with only HpuAB are more highly associated with 504 carriage (59). However, more than 90% of isolates from certain hypervirulent clonal 505 complexes were found to express both HpuAB and HmbR (60). In one case, the phase 506 variable hpuA-hpuB locus was found to have been in the "off" phase in the inoculating 507 isolate, but in the "on" phase following accidental human passage (61). HmbR is more 508 strongly associated with invasive disease in N. meningitidis and has been shown to be

619
Meningococcal aldolase also mediates adherence to human brain microvascular endothelial          The antimicrobial effect of human haptoglobin on N. meningitidis isolates was assessed 843 using a previously described method (50). Briefly, 16HBE cells were grown to confluence in Aldrich), ranging from 1 to 1000 ng/ml final concentration, were added to the co-cultures.

847
Viable bacteria from the co-cultures were enumerated immediately after the infection (to