1 Antibody-dependent enhancement of bacterial disease: Prevalence, mechanisms and 1 treatment. 2 3 Von Vergel L. Torres a , Carrie F. Coggon a , Timothy J. Wells a # 4 5 a The University of Queensland Diamantina Institute, The University of Queensland, 6 Woolloongabba, QLD 4102, Australia 7 8 Running Head: ADE of bacterial infections 9 10 #Address correspondence to Timothy J. Wells, timothy.wells@uq.edu.au 11 12 IAI Accepted Manuscript Posted Online 8 February 2021 Infect Immun doi:10.1128/IAI.00054-21 Copyright © 2021 Torres et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license. on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 2 ABSTACT 13 Antibody-dependent enhancement (ADE) of viral disease has been demonstrated for 14 infections caused by flaviviruses and influenza viruses, however antibodies that enhance 15 bacterial disease are relatively unknown. Over recent years, a few studies have directly linked 16 antibody in exacerbating bacterial disease. This ADE of bacterial disease has been observed 17 in mouse models and human patients with bacterial infections. This antibody-mediated 18 enhancement of bacterial infection is driven by various mechanisms and are disparate from 19 those found in viral ADE. This review aims to highlight and discuss historic evidence, 20 potential molecular mechanisms and current therapies for ADE of bacterial infection. Based 21 on specific case studies, we report how plasmapheresis has been successfully used in patients 22 to ameliorate infection related symptomatology associated with bacterial ADE. A greater 23 understanding and appreciation of bacterial ADE of infection and disease could lead to better 24 management of infections and inform current vaccine development efforts. 25 26 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 3 INTRODUCTION 27 Antibodies are a critical part of the human humoral immune response that aid in the 28 elimination of pathogenic microorganisms via a range of mechanisms (1). This depth in 29 function can be attributed to their Fragment antigen-binding (Fab) and Fully crystallizable 30 (Fc) structural domains joined by a flexible amino acid stretch termed the hinge region (2, 3). 31 To neutralize pathogens the Fab region binds to antigenic structures, typically localized at the 32 cell surface, preventing direct interactions with host cells. The Fab region can also bind 33 bacterial toxins, preventing their interaction with host cellular receptors and neutralizing their 34 activity. In addition to this, antibodies can recruit potent effector molecules through 35 interactions with their Fc regions, initiating a variety of clearance mechanisms precluding 36 pathogenic infection. The specific effector molecules recruited are dependent on class and 37 subclass of antibody involved (4). Despite these well characterized protective aspects, 38 antibodies that instead amplify disease have also been described, through phenomena such as 39 Antibody-Dependent Enhancement (ADE) and Vaccine-Associated Enhanced Respiratory 40 Disease (VAERD). 41 ADE is best known for its role in viral infections; such observations have been reported for 42 flaviviruses (5-7) and influenza viruses (8-10). Recently viral ADE of infection has been of 43 increased interest for its possible role in exacerbating coronavirus infections (11, 12). ADE of 44 viral infection occurs when sub-optimal antibodies (i.e. low affinity, concentration and/or 45 specificity) cannot completely neutralize the etiologic pathogen thereby resulting in increased 46 binding, uptake and replication leading to inflammation (11). Additionally, binding of these 47 antibodies naturally activates cytokine release or complement cascade for host protection, but 48 when unchecked may result in a fatal chain of events due to persistent tissue damage coupled 49 with chronic inflammation. 50 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 4 VAERD is a unique clinical syndrome first observed in the 1960s when whole-inactivated 51 viral vaccines for measles and respiratory syncytial virus (RSV) were administered in 52 children (13). Along with aberrant T cell responses (14), the use of a limiting dose or 53 conformationally incorrect antigen in vaccines can result in the production of non- 54 neutralizing antibodies (15). As mentioned previously, this generation of non-neutralising 55 antibodies runs the potential risk of binding pathogenic microorganisms - but not blocking or 56 clearing them - allowing them to replicate or permit formation of immune complexes 57 activating complement pathways associated with several inflammatory disorders (15). 58 In a few cases, antibody that enhance bacterial infection and disease have been described, 59 although the mechanisms underpinning this phenomenon are distinct from those for viral 60 ADE. In this review, we first cover research highlights of ADE of bacterial infection in 61 animal studies and human cohorts. We then discuss the proposed molecular mechanisms and 62 review other instances where these mechanisms have been reported. Finally, we examine 63 current therapies for ADE of bacterial disease. 64 65 ANTIBODY DEPENDENT ENHANCEMENT OF BACTERIAL INFECTION 66 The first hints of a specifically induced serum factor that may enhance bacterial infection 67 dates back to 1894. Pfeiffer and Issaeff found that naive guinea pigs given serum from guinea 68 pigs previously infected with Vibrio cholera , were more susceptible to intraperitoneal 69 infection by the Vibrio species (16). However, it was not until later that specific antibody was 70 confirmed to enhance some bacterial infections (Table 1). 71 Antibody that enhances bacterial infection has been demonstrated within a few in vitro and 72 mouse model studies (Table 1). For Streptococcus pneumoniae , bacteria-specific antibody 73 responses enhance bacterial attachment to respiratory cells (17). In a mouse model of 74 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 5 Acinetobacter baumannii pneumonia, the passive immunization of a monoclonal antibody 75 (mAb) specific for the capsule increased mortality and bacterial burden in the blood, lung and 76 spleen (18) The presence of these antibodies were found to increase adhesion of the bacteria 77 to resiraptory epithelial cells (18). In a mouse model of Neisseria gonnorrhoaea infection, 78 passive immunization with mAb towards the reduction modifiable protein (Rmp) abrogated 79 the protective effects of the 2c7 mAb (19), causing both increased bacterial burden and 80 duration of infection (20). Finally, pre-existing serum IgG specific for enteric bacteria 81 correlated with worse survival in a mouse model of sepsis induced by cecal ligation and 82 puncture (21). 83 Evidence of antibodies that enhance bacterial infection in humans have been more difficult to 84 discern. The anti-Rmp antibody responsible for enhancing mouse infecion of N. 85 gonnorrhoea , was also found to to increase susceptibility to N. gonnorrhoea in a study of 243 86 adult women (22). Based on findings from our work, we have demonstrated that the presence 87 of high-titers of LPS-specific IgG2 in serum is associated with worse lung function in 88 patients with bronchiectasis and chronic Pseudomonas aeruginosa infection (23). 89 Although there are only a few cases with direct evidence of bacterial ADE, the mechanisms 90 underlying these antibody-mediated pathologies have been described in a wide array of 91 bacterial infections. Here we will discuss the prevalance and potential mechanisms of 92 bacterial ADE of infection including antibody-mediated serum resistance, virulence 93 enhancement by antibody proteolysis, antibody enhancement of adhesion and antibody- 94 mediated protection from phagocytosis. 95 ANTIBODY-MEDIATED SERUM RESISTANCE 96 The complement system is a vital part of the immune response against bacterial pathogens, 97 consisting of mutiple soluble factors acting in a cascade to either promote 98 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 6 opsonophagocytosis or direcly kill bacteria through pore formation and cell lysis (24). The 99 complement system can be activated though three distinct pathways, i) the classical, ii) 100 alternate and iii) lectin, however all three converge to form the membrane attack complex 101 (MAC) which can insert into the outer membrane of gram-negative bacteria leading to cell 102 death. The classical pathway of complement-mediated killing requires the binding of specific 103 antibody to antigen on the cell surface of pathogens. Bound antibodies recruit the C1q 104 component, leading to the complement cascade and formation of the MAC (24). 105 Resistance to complement-mediated lysis, commonly known as serum resistance, is an 106 essential virulence trait of many pathogenic bacteria as it allows survival in the bloodstream 107 (25, 26). Bacteria have developed numerous virulence factors to assist them in resisting the 108 bactericidal effects of serum (27). Briefly, gram-positive bacteria are innately resistant to 109 MAC insertion due to the thick layer of peptidoglycan on their cell surface (28). For gram- 110 negative bacteria the production of polysaccharide capsule, O-antigen positive 111 lipopolysaccharide (LPS), or lipooligosaccharide (LOS) have been shown to be important for 112 serum resistance by providing a steric barrier protecting the outer membrane from insertion 113 of the MAC (29-31). In spite of this defined role, many O-antigen or capsule expressing 114 strains still remain sensitive to serum-killing (23, 32, 33). Recently, the role of other 115 proteins such as Lpp in E. coli has been shown to be vital for resistance (34). Some species 116 also reduce complement activation via recruitment of complement regulatory proteins within 117 the host to their cell surface (35). Antigenic variation and the shedding of surface-bound 118 complement factors also provide resistance by circumventing pathogen recognition 119 mechanisms (27). Finally, some species are capable of offensive manoeuvers in the form of 120 enzyme secretion which bring about degradation/inactivation of complement components or 121 inhibit the formation of MAC (36). 122 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 7 Antibody-mediated serum resistance seems paradoxical, due to its accepted role in classically 123 activated complement-mediated killing. However, in both N. gonorrhoeae and P. aeruginosa 124 infection , the antibody associated with enhanced infection or disease has been shown to block 125 or inhibit the complement-killing of bactericidal sera. These ‘blocking antibodies’ have now 126 been described for many different species including Neiserria spp., Brucella spp., Salmonella 127 sp., Proteus spp., and Klebsiella sp (Table 2) Descriptions of inhibitory sera date back to 128 1901 when the ‘Neisser -Wechsberg ’ phenomenon was described wherein deactivated serum 129 taken from a rabbit repeatedly infected with Vibrio metchntkovi increased the bactericidal 130 effect of normal serum of this strain in moderate amounts, but inhibited the bactericidal effect 131 when mixed in large amounts (37). This inhibiting factor was shown to be specific, as it could 132 be adsorbed out of the sera using the homologous strain (38). Below we detail the bacterial 133 diseases for which these antibodies have been described. 134 Meningitis: Neisseria meningitidis 135 In the 1940s the first specific description of a factor in humans that blocked serum-killing of 136 a bacterial strain was reported. Sera taken from patients two weeks after their initially 137 reported infection with Neisseria meningitidis was unable to kill a serum sensitive strain, in 138 direct contrast to sera taken during the acute phase of infection (39). These convalescent sera 139 were also able to block the killing of otherwise bactericidal normal human sera (NHS). A 140 similar blocking factor was found in multiple patient cohorts (40, 41) and eventually 141 identified as IgA1 binding to capsular polysaccharide of group C meningococcus (42). 142 Adsorbing out all IgA restored bactericidal killing (41), whereas adsorption of all antibody 143 specific for the capsule (42), removed both the inhibition and bactericidal activity, suggesting 144 anti-capsule IgG promoted cell-lysis (42). Blocking IgA specific for capsule of various 145 serotypes of N. meningitidis has since been reported in multiple human cohorts (43, 44). 146 Finally, evidence for an anti-capsule IgG antibody that inhibits alternative pathway 147 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 8 complement-mediated killing was found in two C2-deficient sisters. One of these sisters 148 when vaccinated with tetravalent meningococcal vaccine (containing capsular 149 polysaccharides from serogroups A, C, W-135 and Y) developed an antibody response that 150 blocked alternative pathway-mediated killing of serogroup W-135. IgG was identified as the 151 blocking antibody and was found to inhibit killing of the W-135 serogroup strain but not a 152 serogroup C strain (45). In contrast to the studies above, IgA was not shown to be involved in 153 the inhibition (45). 154 Antibodies that inhibit complement-mediated killing also target other outer membrane 155 structures of N. meningitidis . The serogroup B meningococcal capsule is not immunogenic, 156 however blocking antibodies have been identified against lipoproteins expressed by these 157 strains. The sera of 17 healthy individuals were tested for their ability to kill a serogroup B N. 158 meningitidis strain, finding variable killing throughout. Three of the four sera with low 159 bactericidal activity were able to block bactericidal killing, however they could not inhibit the 160 killing of the strain by all bactericidal sera. The inhibition was found to be IgG3 specific for 161 two lipoproteins: Lip/H.8 and Laz (46). These aforementioned lipoproteins contain almost 162 identical pentapeptide repeats ( × 15 and × 8 respectively) and expression of both were 163 required for maximal blocking. Bactericidal activity could be restored in sera by adsorbing 164 out these blocking antibodies with synthetic peptide containing the pentapeptide repeats (46). 165 In humans, α1,3 -Galactosyl antibodies (anti-Gal) are ubiquitously expressed making up 166 around 1% of total circulating IgG (47). Anti-Gal IgM, IgG and IgA were found not to bind 167 the LOS of N. meningitidis but to the pili of some strains (48). Total serum anti-Gal Ig was 168 found to inhibit complement-mediated killing of pili-producing N. meningitidis strains. When 169 purified, serum IgA1 and secretory IgA inhibited killing, whereas purified anti-Gal IgG did 170 not (48). Notably, anti-Gal has also been found to bind to LPS of enteric bacterial species 171 such as Klebsiella , Escherichia , Salmonella (49), and Serratia (50). 172 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 9 Disseminated Gonococcal Infection: Neisseria gonorrhoeae 173 Antibody has long been shown to be neccesary for complement-mediated killing of N. 174 gonorrhoeae. NHS contains bactericidal IgM antibody that can target and kill serum-sensitive 175 isolates of N. gonorrhoeae (51, 52). Strains that cause disseminated gonococcal infection are 176 overwhelmingly serum resistant to NHS (53), but have been shown to be killed by the 177 induction of specific antibody in serum from convalescent patients (54, 55). In some cases, 178 NHS was found to contain specific IgG antibody that could block the bactericidal effect of 179 both rabbit complement and convalescent sera (56-58). These inhibitory antibodies have been 180 found to be specific for either LOS or outer membrane proteins of N. gonorrhoeae . Antibody 181 to LOS, primarily IgM (51, 59), has been shown to be responsible for a major proportion of 182 bactericidal activity against N. gonorrhoeae (60, 61). In contrast, IgA specific for N. 183 gonorrhoeae LOS isolated from pooled NHS was shown to block the bactericidal killing of 184 IgG-mediated killing of the bacterium (60). 185 The best characterized gonnococcol antibody that inhibits complement-killing is specific for 186 an outer membrane protein; the reduction modifiable protein (Rmp, protein III). Antibody to 187 Rmp has been shown to increase bacterial burden in a mouse model as well as acting as a 188 biomarker for susceptibility to N. gonnorrhoea in humans (20, 22). Rmp is a 236 aa protein 189 that is physically associated with the porin Por and is highly conserved within the genus 190 Neisseria (62-64) Rmp is ubiquitously expressed in N. gonorrhoeae and demonstrates robust 191 immunogenicity (65). IgG specific for Rmp was shown to inhibit bactericidal killing by 192 protective antibodies, whether purified from human sera (65) or as a monoclonal antibody 193 (66), although only one of two monoclonal antibodies directed against Rmp promoted the 194 inhibitory effect (66). Bactericidal activity could be restored in non-killing serum from 195 convalescing patients by selectively adsorbing out the inhibitory anti-Rmp antibody. 196 Recently, antibodies with higher bactericidal activity against N. gonorrhoeae have been 197 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 10 induced by using a Rmp deletion strain (67). Rmp does exist in N. meningitidis (Class 4 outer 198 membrane protein) but the role of anti-Rmp antibodies in bactercidal killing of that species is 199 unclear. One study identified anti-class 4 antibodies that blocked killing of the strain in one 200 individual (68), however antibodies specific for class 4 induced by a menigococcal outer- 201 membrane vesicle vaccine did not inhibit complement-mediated killing (69). 202 Brucellosis: Brucella melitensis, suis and abortus 203 Brucellosis is the leading cause of zoonotic infection in humans worldwide and is caused by 204 multiple bacterial species in the genus Brucella Brucella are small gram-negative facultative 205 intracellular coccobacilli, with four species known to have moderate to significant 206 pathogenicity in humans: Brucella melitensis (sheep), Brucella suis (pigs), Brucella abortus 207 (cattle) and Brucella canis (dogs). High titers of specific antibody to Brucella species, likely 208 serum IgA, is known to inhibit agglutination reactions (53, 64, 70, 71), however a blocking 209 factor of complement-mediated killing was simultaneously being described In 1945, 210 Huddleson and colleagues showed that fresh plasma from cattle infected with B. abortus 211 could not kill the bacteria when given a source of complement. In addition, this plasma 212 inhibited the action of bactericidal plasma when mixed and was suggested to be antibody 213 (72). Similarly, guinea pigs infected with Brucella had serum that was bactericidal seven 214 days post-infection, but unable to kill the strain thirty days post-infection (73). The 215 bactericidal activity of fresh rabbit blood was also found to be inhibited when rabbits were 216 immunised extensively with B. abortus (74). Analysis of rabbit hyper-immune serum 217 indicated that IgA was primarily responsible for the bactericidal inhibition, with addition of 218 IgA blocking the bactericidal effect of IgM or IgG isolated from the sera (75), however the 219 role of other antibody isotypes could not be ruled out. 220 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 11 The process behind serum bactericidal inhibition in cattle was comprehensively studied by 221 Corbeil et al. where the authors investigated the differences in killing of smooth (O-antigen 222 expressing) and rough (O-antigen negative) strains of B. abortus after infection or 223 immunization with either strain. When specific antibody was present, killing of the rough 224 strain by serum was increased. In contrast, as specific antibody increased for the smooth B. 225 abortus , killing of that strain in serum significantly decreased (70). Antibodies involved in 226 the killing inhibition were identified as IgG1 and IgG2, with IgG2 driving significantly more 227 inhibition (70). Other studies have confirmed the blocking factor in bovine serum as IgG and 228 suggested that antibody under certain conditions, may actually promote establishment of 229 bovine brucellosis (76). 230 Pyelonephritis: Escherichia coli, Proteus mirabilis and morganii, Klebsiella aerogenes 231 Antibody that blocks complement-mediated killing has also been described in pyelonephritis 232 for a range of bacterial species including Escherichia coli, Proteus mirabilis and Klebsiella 233 sp. (77). In each case, the patient sera blocked the ability of NHS to kill their cognate strain in 234 a serum and strain specific manner. 235 In 1972, nine of 48 female patients with urinary tract infections were found to have a serum 236 factor that inhibited NHS-killing of their their cognate strain (78, 79). Of the nine patients, 237 eight were afflicted with pyelonephritis. Strains isolated from patients with this defect 238 included E.coli serotypes O2, O4, O7, O75, Proteus morganii, Proteus mirabilis and 239 Klebsiella sp. (78). This inhibitory factor was investigated in detail for the two patients 240 infected with Proteus sp. and confirmed to be due to IgG specific for LPS (80). Adsorbing 241 out all IgG, or antibody specific for LPS, was sufficient to restore bactericidal activity of the 242 serum (81). Finally, we recently found that 24% of 45 patients with E. coli urosepsis infection 243 had antibodies that inhibited serum-killing. These antibodies were IgG2 specific for LPS of 244 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 12 the E. coli strain. Isolates taken from patients with these serum-killing inhibitory antibodies 245 were more sensitive to NHS, suggesting the presence of these antibodies were required for 246 these strains to survive in the bloodstream (82). 247 Bacterial Lung Infections: Pseudomonas aeruginosa 248 Recurrent and chronic bacterial lung infections commonly affect those with bronchiectasis, 249 cystic fibrosis (CF) and chronic obstructive pulmonary disorder. P. aeruginosa is a major 250 cause of morbidity and mortality in these diseases, and once infection is established is 251 exceedingly difficult to eradicate. The first reported case of a blocking serum factor towards 252 Pseudomonas was Waisbren and Brown’s work (77). By investigating the sera of two 253 patients with chronic P. aeruginosa infection, the blocking factor was identified as IgG and 254 could be specifically adsorbed from the serum by using the cognate strain isolated from the 255 patient (83). In a study investigating forty-two patients with CF and P. aeruginosa infection, 256 six patients of the cohort had sera that was unable to kill their autologous strain, even though 257 it was sensitive to NHS killing (84). Additionally, these sera could kill heterologous strains of 258 P. aerugionsa from other patients, suggesting the presence of patient-specific blocking 259 antibodies in these sera (84). 260 The nature of blocking antibodies to P. aeruginosa was investigated in our study looking at 261 29 patients with bronchiectasis and chronic P. aeruginosa infection (85). Six of these patients 262 had sera that were unable to kill their cognate strain, despite the isolates being sensitive to 263 NHS killing. As little as 6% of patient sera added to NHS could completely inhibit 264 complement- mediated killing. ‘ I nhibitory sera’ had significantly higher binding of IgG2 to 265 their autologous strain and specifically removing these IgG2 from the sera restored 266 bactericidal activity. Conversely, reintroducing these purified IgG2 to NHS inhibited 267 complement-dependent killing. Furthermore, antibody specific for the O-antigen component 268 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 13 of LPS, but not lipid-A, inhibited complement-mediated killing. Thus, IgG2 specific for the 269 O-antigen was responsible for inhibiting complement-mediated killing. Importantly, patients 270 with ‘inhibitory antibodies’ had significantly worse lung fun ction than patients with normal 271 killing serum (23). 272 Salmonella infection: Salmonella enterica serovars Paratyphi B and Typhimurium 273 The first reported case of serum-killing inhibitory antibodies against Salmonella sp. was in 274 1953, when the presence of excess antibody to Salmonella R antigen (LPS core) was shown 275 to block serum bactericidal activity (86). The study by Waisbren and Brown also reported 276 two patients with inhibitory antibodies to their infecting Salmonella strain; a S. Paratyphi B 277 strain from a patient suffering peritonitis and a rectus abdominus muscle abscess, and a S. 278 Typhimurium strain from a patient suffering septiceamia and a splenic abscess (77). 279 The first detailed study of inhibitory antibody towards Salmonella was in 2010 when 280 MacLennan and colleagues investigated a cohort of HIV-infected African adults for their 281 serum’s ability to kill two S. Typhimurium strains, one resistant to NHS and one sensitive 282 (87). Within this cohort, 58% could not kill the sensitive strain, while 28% had lower 283 bactericidal activity than the NHS of the resistant strain. A high titer of anti- S. Typhimurium 284 IgG correlated with less bactericidal-killing by the patient sera. The non-killing sera could 285 inhibit the action of NHS. The blocking factor was identified as IgG specific for S. 286 Typhimurium LPS (87). Bactericidal activity could be restored in patient sera by either 287 adsorbing out the LPS-specific IgG or deleting O-antigen expression on the surface of the 288 Salmonella strain. In contrast, antibodies targeting outer membrane proteins were found to be 289 bactericidal (87). 290 In an analysis of 49 healthy adult serum samples, it was found that 48 had robust killing of a 291 S. Typhimurium strain with antibody speific for LPS identified as important for the 292 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 14 bactericidal activity (88). The single serum unable to kill the strain was found to contain 293 inhibitory antibodies, correlating with anti-LPS IgM. The serum also led to less complement 294 deposition on the bacterial strain. Thus, anti-LPS antibody was found to both promote and 295 inhibit complement-mediated killing (88). This result was supported when it was later shown 296 that any antibody specific for the LPS could inhibit complement-mediated killing, although 297 IgA and IgG2 had the strongest association (89). The anti-LPS IgG when sufficiently diluted 298 was found to be bactericidal, however when concentrated from NHS resulted in the blocking 299 of complement-mediated killing (89). 300 Individual cases: Actinobacillus pleuropneumoniae, Haemophilus influenzae 301 Actinobacillus pleuropneumoniae is a gram-negative encapsulated coccobacillus that causes 302 acute to chronic respiratory diseases in swine worldwide. Antibody that inhibited 303 complement mediated killing of the strain was identified in normal swine serum and hyper- 304 immune swine and guinea pig serum. The antibody was IgG specific for A. 305 pleuropneumoniae LPS (90). 306 Nontypable Haemophilus influenzae (NTHI) is a non-encapsulated gram-negative commensal 307 of the human nasopharynx. However, when present in the lower-respiratory tract it can 308 become an important pathogen associated with conditions including bronchitis, 309 bronchiectasis, pneumonia and chronic obstructive pulmonary disease (91). IgA when 310 purified from four out of five patients with NTHI infection inhibited the bactericidal activity 311 of NHS or their own serum to kill their cognate strain (92). Although the IgA is specific for 312 the NTHI strains, the precise target of the inhibitory IgA remains unknown. 313 Mechanisms underlying antibody-mediated serum resistance 314 Antibodies that inhibit the complement-mediated killing of bacteria can belong to various 315 isotypes: IgA, IgM, IgG, IgG1, IgG2 and IgG3 (23, 42, 46, 70, 89). Additionally, these 316 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 15 antibodies are specific for a variety of outer membrane structures such as LOS, LPS (core and 317 O-antigen), capsule, lipoproteins and outer membrane proteins (Table 2). Thus, there are 318 likely multiple mechanisms underpinning the antibody-mediated inhibition. However, many 319 similarities shared between the descriptions of antibody-mediated serum resistance that may 320 be informative. 321 In many of the cases reviewed here, the antibodies required for complement-resistance bind 322 either LPS or LOS. In the case of P. aeruginosa, the antibodies specific for the O-antigen but 323 not the Lipid-A/ core of the LPS were found to be responsible (23). This is reinforced by 324 studies where the inhibition is only found against smooth strains expressing O-antigen (70, 325 87). We and other groups have previously proposed two possible mechanisms of 326 complement-inhibition (23, 87). In both, high titers of antibody bind to the O-antigen of LPS, 327 and then exert their inhibitory effect either by 1) activating and depositing complement away 328 from the bacterial membrane or 2) by creating a physical antibody blockade that prevents 329 access of protective antibody and MAC to the cell surface. Evidence currently favours the 330 latter explanation, where multiple studies have found the inhibition of complement-mediated 331 killing to be dependent on high titers of the antibody. In N. meningitidis , if enough IgA was 332 present to form a ‘blockade’, no amount of IgM could restore bactericidal activity (93). This 333 antibody has been shown to bind and activate MAC formation (23, 87), however MAC 334 activation is likely not required for complement-resistance as one study determined that 335 removal of the Fc portion from IgA1 did not affect the inhibition (42). Finally, many isotypes 336 of antibody with disparate complement activation have been found to inhibit serum-killing. 337 Only one study has found IgM to inhibit serum-killing (89), with the majority of evidence 338 indicating that IgM towards LPS is bactericidal in serum (51, 54, 75, 93, 94). We hypothesise 339 that this may be due to the pentameric structure of IgM, which may interfere with the close 340 packing of the antibody required to form the protective barrier. IgM typically has lower 341 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 16 affinity for antigens than other isotypes of antibody. As inhibitory antibodies need to block 342 access of MAC to the membrane, we hypothesize that lower affinity antibodies may be 343 unable to block this insertion. Thus, we propose a mechanism of action where high titers of 344 antibody bind to an abundant antigen that is distal from the cell surface. These antibodies 345 completely cloak the bacteria, creating a physical barrier that prevents MAC insertion into the 346 membrane (Fig. 1A). Serum with low titers of antibody cannot cover the entire surface of the 347 bacteria, leading to bactericidal activity (Fig. 1B). Additionally, removal of the target epitope 348 restores complement-killing (Fig. 1C). This mechanism could be applied to capsule, LPS, 349 LOS or any numerous antigen distal from the cell surface. Differences in antibody titer or 350 abundance of antibody target may also explain differences in bactericidal effect. 351 Antibodies that protect bacteria from complement-mediated killing using this method have 352 previously been termed ‘ inhibitory ’ , ‘ blocking ’ or ‘paradoxical’ antibody, however all these 353 terms are broad or already have multiple uses in the field. We propose that these be termed 354 ‘cloaking antibodies’ which de scribes the mechanism of action; complete coverage that gives 355 protection, and is distinct from these other aforementioned terms. 356 Finally, proteins have also been identified as targets for antibodies that inhibit complement- 357 mediated killing; mainly Rmp from N. gonorrhoea and Lip/H.8 from N. meningitidis . The 358 mechanisms underlying the antibody inhibition are disparate. For Rmp, antibody binds and 359 promotes complement deposition, but does not lead to killing. It also can block the 360 bactericidal action of protective antibodies. It is thought that the presence of the anti-Rmp 361 antibody recruits complement proteins away from bactericidal sites of the N. gonorrhoea to 362 non-bactericidal sites (58) Whether this is due to steric hindrance, as with cloaking 363 antibodies, has yet to be determined. Anti- Lip/H.8 antibodies differ in that their presence 364 reduces complement activation (specifically C4), and although the antibodies could block the 365 action of some protective mAbs, it could not inhibit the bactericidal action of a mAb to anti- 366 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 17 PorA, possibly because PorA is a major and abundant porin of the N. meningitidis membrane 367 (46). Thus, blocking by Anti-Lip/H.8 antibodies is due to an overall reduction of complement 368 activation and can be overcome by protective antibodies. 369 Antibody-mediated serum resistance and disease enhancement 370 The causal link between antibody-mediated serum resistance and enhancement of infection is 371 clear in diseases that require serum resistance such as bacteraemia, urosepsis and meningitis. 372 The presence of the antibody allows otherwise serum-sensitive strains to survive in blood, 373 leading to higher burden of disease. Indeed, in E. coli urosepsis the isolates taken from 374 patients with high titers of cloaking antibodies were more sensitive to NHS killing, 375 suggesting that the presence of these antibodies were required for the infection to survive in 376 blood (82). In the case of facultative intracellular Brucellis , some authors have speculated 377 that protection from complement killing, while still promoting opsonophagocytosis may 378 facilitate higher infection of phagocytic cells between lysis/ infection cycles (76). 379 The link between cloaking antibody and disease severity in lung infections, such as those 380 shown in P. aeruginosa is less obvious, where complement-killing is not a main defence 381 against infection in the lung. Lung damage in chronic P. aeruginosa lung infections is usually 382 due to increased inflammation (95, 96). There are many ways antibodies may increase 383 inflammation including: increased C5a generation as MAC is activated (97), the 384 glycosylation state of the antibody (98) or differential Fc receptor activation (99). Antibody 385 to LPS has also been shown to inhibit alveolar phagocytosis of P. aeruginosa (see below). 386 Any one of these mechanisms may be responsible for the association with worse lung 387 function. 388 Finally, although cloaking antibodies are associated with worse disease in many bacterial 389 infections, it is important to note that antibodies to LPS, LOS or capsule are protective in 390 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 18 many of the diseases discussed above and are core components of current or developing 391 vaccines (100-104). These antibodies are protective through various mechanisms such as 392 preventing adhesion of the bacteria in the gastrointestinal tract (105). Despite this, vaccine 393 development efforts should also consider the possibility of these antibodies enhancing 394 bacterial infections, as it may also explain previous cases where LPS-based vaccinations have 395 led to worse clinical status in some patients (106, 107). Finally, it is worth remembering that 396 the serum of many mouse strains are unable to kill various gram-negative bacteria by 397 complement-mediated lysis (108, 109) and thus the role of serum resistance may not be 398 accounted for in many models. 399 VIRULENCE ENHANCEMENT BY ANTIBODY PROTEOLYSIS 400 To counteract host immune pathways, bacterial pathogens have evolved to express and 401 secrete proteases that target antibodies preventing normal immunoglobulin effector function 402 (110-112). Conceivably, this targeted cleavage of antibodies by bacterial proteases may 403 represent a distinct mechanism for exacerbating bacterial disease. In several examples, 404 pathogenic bacteria that produce IgA proteases are predominately associated with 405 colonisation of mucosal surfaces (113, 114). The capability to evade these IgA dominant 406 mucosal sites provides a fitness advantage for these pathogens. As an example, in a study by 407 Weiser et al. (17), the authors found specific antibody responses to the bacteria S. 408 pneumoniae (pneumococcus) can enhance infection. This study found that a secreted 409 bacterial protease cleaves IgA1 that has specificity to capsular polysaccharide (CPS). Rather 410 than the widely accepted function of IgA1 inhibiting adherence to pharyngeal epithelial cells, 411 presence of the antibody instead significantly augmented adherence in vitro . The authors 412 found that the IgA1 protease specifically cleaved antigen-bound CPS-specific IgA1. This 413 specific cleavage resulted in localised “unmasking” of the capsule allowing the not normally 414 readily accessible ChoP surface ligands, a choline containing molecule (phosphorylcholine), 415 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 19 to interact with the receptor for platelet-activating factor (rPAF) on cells in the infected host 416 consequently allowing bacterial cell adherence. The authors considered that such an 417 antibody-mediated mechanism would result in the ability of pneumococcus to persist on 418 mucosal surfaces of the host respiratory tract. 419 A mechanism instead exploiting the hinge region for proteolysis has also been observed for 420 immunoglobulin IgG. The extracellular secreted proteases glutamyl endopeptidase V8 421 (GluV8) of Staphylococcus aureus and Immunoglobulin-degrading enzyme (IdeS) of 422 Streptococcus pyogenes (115, 116), have been shown to cleave within the hinge region of 423 IgG molecules (of the CH2 linker) disabling binding to host Fc receptors. Several other 424 groups have reported that even a single cleavage in this region is able to abolish binding to Fc 425 receptors (117). Interestingly, cystatin C – a host cysteine protease inhibitor, is “hijacked” 426 and instead acts as cofactor for bacterial IdeS and accelerates IgG cleavage (118, 119). 427 Mechanistically, the literature is not clear but perhaps cleavage of the IgG immunoglobulin 428 serves two functions in precluding bacterial clearance by not only impeding the regulation of 429 a conducive immune response by Fc receptors binding but also masking epitopes to other 430 components of the host immune system. A similar tactic to evade host immune responses by 431 many pathogenic bacterial species is to express cell surface proteins (e.g. protein A of S. 432 aureus and Eib proteins of E. coli ) that bind antibodies outside the Fab region of IgG so that 433 these antibodies are incorrectly orientated and therefore functionally impaired from eliciting 434 an immune response (120, 121). Altogether, the presence of antibody in these scenarios 435 would instead actively enhance protection of the bacteria. 436 From these discussed studies, it becomes clearer that the exploitation by proteolysis of 437 immunoglobulin is a common strategy of pathogenic bacteria (Table 3). The evolution of 438 such different protease family classes (serine-, cysteine- and metallo- proteases) shows the 439 viability of such a strategy and how this may contribute in such distinct ways for a 440 on March 27, 2021 by guest http://iai.asm.org/ Downloaded from 20 mechanism of bacterial ADE infection. Furthermore, it is worth mentioning that some 441 protease