The Role of IgG Subclass in Antibody-Mediated Protection against Carbapenem-Resistant Klebsiella pneumoniae

Carbapenem-resistant Klebsiella pneumoniae is an urgent public health threat that causes life-threatening infections in immunocompromised hosts. Its resistance to nearly all antibiotics necessitates novel strategies to treat it, including the use of monoclonal antibodies. Monoclonal antibodies are emerging as important adjuncts to traditional pharmaceuticals, and studying how they protect against specific bacteria such as Klebsiella pneumoniae is crucial to their development as effective therapies. Antibody subclass is often overlooked but is a major factor in how an antibody interacts with other mediators of immunity. This paper is the first to examine how the subclass of anticapsular monoclonal antibodies can affect efficacy against CR-Kp. Additionally, this work sheds light on the viability of monoclonal antibody therapy in neutropenic patients, who are most vulnerable to CR-Kp infection.

sizes the importance of testing alternative therapies, including MAbs, against these pathogens (3). Much information regarding how antibody (Ab) structure influences interactions with pathogens remains to be discovered, and until recently the role of the Ab constant region and its different variants, or subclasses, had often been overlooked in therapeutic monoclonal Ab development. While four subclasses of IgG Abs exist in humans, the majority of MAbs used in the clinic are human IgG1 (hIgG 1 ), the most prevalent subclass (4). The subclass of an Ab (dictated by the number of disulfide bonds joining the heavy chains), its fragment crystallizable (Fc) region, and other aspects of the heavy chain, affect what immune receptors and adaptors the antibody binds. Subsequently, these interactions determine the amplitude and character of the immune response (5). Some subclasses interact with more immunostimulatory Fc receptors on professional phagocytes, increasing their activity, while others bind to immunosuppressing receptors that act to reduce collateral damage caused by excessive inflammation (6). Additionally, subclasses can be responsible for differences in antigen binding, even when Abs have identical variable regions (7,8). Understanding differences between IgG subclasses-how they bind, interact with the pathogen, and interact with other facets of immunity-is important to understanding which subclasses may provide a therapeutic benefit (8)(9)(10)(11)(12).
With the recent rise of multidrug-resistant Gram-negative bacteria, such as carbapenem-resistant Klebsiella pneumoniae (CR-Kp) (13), several laboratories have been focusing on developing antibodies against these pathogens (3,(14)(15)(16)(17). These bacteria frequently infect immunocompromised populations that lack robust innate and adaptive immune responses (18,19). Therefore, it is crucial to understand not only how different Ab subclasses act against these pathogens but also how they function in the context of immunocompromised states. We recently cloned murine IgG 3 (mIgG 3 ) monoclonal Abs that targeted the capsular polysaccharide (CPS) of wzi154 CR-Kp isolates, which fall within the clade 2 subfamily of the CR-Kp sequence type 258 (ST258) clonal group (14). Isolates of this conserved subgroup have been shown to be susceptible to Ab therapy through a variety of in vitro modalities such as killing by serum complement and action by neutrophils and macrophages, and such antibodies have been shown to be protective in vivo as well (14,15,20). We chose one of these, 17H12, to study the effects of switching IgG subclass on anti-Klebsiella Ab functionality. We report findings that the parent mIgG 3 was superior to the new murine IgG 1 (mIgG 1 ) variant in binding ability, initiation of complement-mediated bactericidal activity by serum, and activation of neutrophil-mediated killing at lower antibody concentrations. Conversely, the new mIgG 1 variant slightly outperformed the mIgG 3 parent in promoting macrophage-mediated phagocytosis of the bacteria. Finally, our comparison within a pulmonary mouse challenge model shows comparable overall efficacy of both subclasses in reducing bacterial organ burden at both lethal and nonlethal inocula in wild-type mice. Interestingly, efficacy of both antibodies was maintained in neutropenic mice except at the lethal inoculum.

RESULTS
Parent 17H12 mIgG 3 variant has superior binding of wzi154 capsular polysaccharide relative to that of the new subclass switch variant 17H12 mIgG 1 , despite identical variable regions. We first isolated a mIgG 1 variant of the mIgG 3 17H12 hybridoma line by using sib selection followed by fluorescence-activated cell sorting (FACS) and soft agar cloning, which we previously utilized (12). Although we sought to generate all three additional subclasses, only mIgG 1 and mIgG 2a variants were discovered in our initial screen, and only mIgG 1 variants could be enriched by downstream sib selection. The mIgG 1 hybridomas were verified to exclusively produce mIgG 1 , and the sequence of the variable region of the new clone was found to be identical to that of the mIgG 3 parent (the characteristics of which have been previously published [14]).
To investigate how subclass switching affected binding, we compared the affinity of the new mIgG 1 to that of its parent mIgG 3 against the wzi154 CPS originally used to generate the MAb (14). Analysis by enzyme-limited immunosorbent assay (ELISA) showed the mIgG 1 to have 4-fold less binding than its mIgG 3 counterpart, with 50% effective concentration (EC 50 ) values of 27.6 nM (95% confidence interval [CI], 18.8 to 40.6 nM) and 6.81 nM (95% CI, 3.00 to 13.2 nM), respectively (Fig. 1A).
Next, we compared the ability of each Ab to agglutinate CR-Kp clinical isolates, utilizing flow cytometry to measure relative clump sizes by forward scatter (21). We began testing the Abs with the previously studied CR-Kp wzi154 strain 39 (MMC39) (14), which we transformed with a novel green fluorescent protein (GFP)-expressing plasmid, pProbe-KtBl. Using this transformant, referred to here as MMC39-GFP, we noted that mIgG 3 promoted better agglutination than mIgG 1 ; mIgG 3 -opsonized bacteria demonstrated higher forward scatter than mIgG 1 -opsonized bacteria at the same concentration of antibody and also achieved maximum forward scatter at lower concentrations of antibody (see Fig. S1 in the supplemental material). As 30 g/ml provided the greatest disparity in agglutination between mIgG 1 and mIgG 3 , we chose this antibody concentration and then compared relative agglutination across a number of CR-Kp isolates (Table 1). These isolates include those collected from Montefiore Medical Center (MMC) and Stony Brook University (SBU), including those previously studied (14,22), as well as a previously studied isolate, 33576, from the mid-Atlantic states and its capsule-deficient variant (15) ( Table 1). These strains cover the three most prevalent wzi subgroups within the ST258 clone (22)(23)(24). Measuring the percentage of aggregates larger than a baseline formed by control bacteria untreated with antibody (21), we found that both Abs improved agglutination of nearly all wzi154 strains relative to that promoted by a control mIgG 1 but did not significantly promote agglutination of the capsule-deficient 33576 Δwzy strain or that of CR-Kp isolates of the wzi29 and wzi50 capsule types (Fig. 1B). Additionally, at this concentration the mIgG 3 parent caused a higher percentage above baseline of agglutination than that caused by mIgG 1 in 5 of 6 isolates. While the percent agglutination of strain 33576 did not significantly improve above baseline for either 17H12 mIgG 1 or mIgG 3 , this was likely due to high observed baseline aggregation of the strain. In contrast, histograms of gated 33576 (and MMC5) demonstrate clear shifts in forward scatter between the mIgG 3 , mIgG 1 , and control groups, and aggregated raw mean forward scatter data show significantly higher agglutination by either Ab of all wzi154 bacteria relative to that in both controls (Fig. S1).
As previous studies have shown the importance of the Fc region in Ab binding to antigen, we performed F(ab=) 2 digests of both Abs to determine whether this may hold true for 17H12. We determined by ELISA that digestion of 17H12 mIgG 3 led to a 2.7-fold loss of binding relative to whole mIgG 3 (Fig. 1C). In contrast, digestion of 17H12 mIgG 1 did not impact binding (Fig. S2).
Complement-dependent serum killing was exclusive to mIgG 3 . We next compared the ability of the two subclasses to mediate serum killing of CR-Kp. Using 20% fresh human serum, we determined that 17H12 mIgG 3 caused 90.1% and 92.7% reductions in CFU of MMC39 CR-Kp after 60 and 120 min, respectively. In contrast, the mIgG 1 caused 64.4% and 63.5% drops in CFU, respectively, similar to that caused by the control Ab ( Fig. 2A). In strain 33576, 20% human serum was found to be insufficient to reduce CFU under any condition (Fig. 2B), but bacterial replication was inhibited by both subclasses, while bacteria exposed to the control multiplied 6-fold over 120 min. Increasing the percentage of serum to 40% improved CFU reduction of the 33576 strain, but in an antibody-independent manner (see Fig. S3 in the supplemental material). As previously described, the 33576 Δwzy strain exhibits pronounced sensitivity to serum in the absence of capsule (15), and nearly complete killing occurred irrespective of treatment (Fig. S3). Variability in killing was observed depending on the human serum donor, but effects were consistent between experiments using serum from the same donor. Additionally, heat inactivation (HI) of serum abrogated all killing effects against MMC39 and instead allowed K. pneumoniae growth, which both Abs partially limited ( Fig. 2A).
We also specifically compared the relative amount of complement each antibody could fix. Using flow cytometry, we detected C3c and C5b-9 membrane attack complex deposition on MMC39, 33576, and 33576 Δwzy strains in the presence of either subclass ( Fig. 2C and D, Fig. S3). We observed the parent mIgG 3 to outperform the phosphatebuffered saline (PBS) control and the mIgG 1 in deposition of C3c onto MMC39, but not onto 33576, which exhibited high background C3c deposition (Fig. 2C). With both strains, however, C5b-9 deposition was found to be increased when bacteria were preopsonized with mIgG 3 (Fig. 2D). Controls confirmed that mIgG 3 capsule binding caused deposition of serum-based complement, with incubation of the bacteria in 0% ( Fig. 2C and D) or 20% HI serum resulting in no detectable deposition and the antibody failing to deposit additional complement onto the capsule-deficient 33576 Δwzy strain (Fig. S3).
Both subclasses improved macrophage-mediated phagocytosis, with mIgG 1 performing marginally better than mIgG 3 . We next compared the ability of the subclasses to contribute to cell-mediated action against CR-Kp. Monocytes and mac-  rophages are important in CR-Kp clearance (25), and we have previously demonstrated 17H12 mIgG 3 to enhance phagocytic uptake of numerous wzi154 CR-Kp strains (14). Therefore, we compared the ability of both variants to promote uptake using a CFU-based phagocytosis assay we previously performed (14,26). Our data show mIgG 1 to slightly improve J774A.1 macrophage phagocytosis of MMC39 relative to mIgG 3, a trend that was also suggested in the phagocytosis of strain 33576 (Fig. 3A). In contrast, the 33576 Δwzy strain was phagocytosed irrespective of treatment condition. The difference in phagocytosis between mIgG 1 and mIgG 3 was subtle and was not evident when phagocytosis of MMC39-GFP was observed under fluorescence microscopy ( Fig. 3B). Additionally, we found that uptake was not correlated with intracellular killing; after both J774A.1 macrophages and bone marrow-derived macrophages (BMDMs) had phagocytized the opsonized bacteria and external bacteria had been washed away, we observed by both CFU quantitation and by microscopy that the number of bacteria within these cells increased over time (data not shown). This observation suggests intracellular multiplication of CR-Kp after phagocytosis. Neutrophil killing of CR-Kp improved with lower concentrations of mIgG 3 than mIgG 1 . We next compared the ability of both antibodies to promote killing by neutrophils. Data in humans has shown that neutropenic patients may have reduced survival in cases of bacteremia caused by CR-Kp and other carbapenem-resistant Enterobacteriaceae (23). In mice, some studies have shown neutrophils to be important in CR-Kp clearance (15,27), while others have shown them to be less valuable (25). Using a tube-based incubation assay with human neutrophils, we observed both antibodies to promote neutrophil-dependent killing of MMC39 in the presence of 5% autologous serum at 40 g/ml (Fig. 3C). When we reduced the dose to 10 g/ml, however, mIgG 3 demonstrated improved efficacy, while mIgG 1 lost all efficacy relative to that of the control. Antibody-mediated killing by neutrophils was dependent on serum, as coincubation of neutrophils with HI serum failed to reduce bacterial CFU.
We then compared the ability of the two antibodies to promote neutrophil reactive oxygen species (ROS) production in response to CR-Kp. We used MMC39, MMC5, SBU32, and SBU34, and either 5% normal human serum (NHS) or 20% HI-fetal bovine serum (FBS) in reactions ( Fig. 3D; see also Fig. S4 in the supplemental material). These data indicate that both mIgG 1 and mIgG 3 promoted ROS production to comparable degrees. However, strains had differing ability to induce ROS production, with strains SBU32 and SBU34 having high constitutive production of ROS in the absence of either monoclonal Ab, whereas base ROS in MMC39 and MMC5 were nearly absent (Fig. 3D, Fig. 4). Furthermore, baseline ROS production in SBU32 and SBU34 occurred irrespective of whether NHS or HI-FBS was used, though NHS appeared to promote marginally higher production.
Both subclasses improved CR-Kp lung clearance in vivo, including within neutropenic mice. Finally, we compared the protective efficacy of the subclasses in vivo, using a pulmonary infection model we previously utilized in BALB/c mice (14). Because we observed these two MAbs to enhance neutrophil-mediated killing, we compared the relative abilities of the subclasses to control organ burden in both c57BL/6 wild-type mice and neutropenic mice, which were generated by Ly6G antibodymediated depletion. After depleting neutrophils or administering a control Ab, we infected mice with a sublethal dose of MMC39 preopsonized with either the mIgG 1 , mIgG 3 , a control mIgG 1 , or Tris-glycine buffer alone. As we found no significant differences between the mIgG control and the Tris-glycine vehicle, these groups were combined for analysis. We first observed higher CFU lung burden in neutropenic mice, in contrast to observations in previous studies (25,28). Additionally, we observed efficacy of both subclasses in reducing bacterial burden in the lung in both wild-type and neutropenic mice, with 17H12 mIgG 1 having a small but significant advantage over the mIgG 3 (Fig. 4A). Additionally, the mIgG 1 appeared to promote higher expression of the inflammatory cytokines gamma interferon (IFN-␥), tumor necrosis factor alpha (TNF-␣), interleukin-12 (IL-12), and IL-17 in neutropenic mice, suggesting higher im- Images were taken at ϫ40 magnification with an EVOS microscope using brightfield and GFP channels. (C) Killing of preopsonized MMC39 by human neutrophils after 60 min. Bars depict mean and SEM of three independent experiments. Within the results of the NHS-treated samples, differences in the variance of dose-matched treatment groups with and without neutrophils were assessed for significance by two-way repeated-measures ANOVA (variance between treatment groups, P Ͻ 0.001 for both 10 g and 40 g sets; variance comparing neutrophil status, P ϭ 0.135 and P ϭ 0.058, respectively) with results of Sidak's multiple-comparisons tests displayed in the graph. (D and E) Reactive oxygen species production by human neutrophils exposed to preopsonized MMC39, as measured by luminol luminescence, in the presence of NHS (D) or fetal bovine serum (FBS) (E). The left time lapse graphs are representative of five independent experiments. Right bar graphs show aggregate data of the area under the curve (AUC) and the maximum rate of change (max Δ) relative to those for PBS for all experiments. Differences in AUC and max Δ between control Ab, mIgG 1 , and mIgG 3 were assessed for significance by a Kruskal-Wallis test (P Ͻ 0.01 for NHS and FBS), with results of Dunn's test for multiple comparisons displayed in the graph. For all in-graph statistics, P values displayed in black are comparisons to the control IgG, whereas P values in red compare mIgG 1 with mIgG 3 . P values are indicated with ns (not significant) if P Ͼ 0.1, * if P Ͻ 0.05, ** if P Ͻ 0.01, and *** if P Ͻ 0.001. mune activation, while the mIgG 3 antibody generally exhibited reductions of these markers in neutropenic mice (Fig. 4B).
We proceeded to compare subclass efficacy at a lethal infectious dose, which has been previously observed to cause dissemination to the liver and spleen and death within 72 h. At this dose, both subclasses performed equally in nondepleted mice, reducing lung, liver, and spleen burden by at least 1 log (Fig. 4C). However, within neutropenic mice, mIgG 1 showed some loss of efficacy, reducing lung burden by only 0.52 log. Meanwhile mIgG 3 continued to reduce burden in all three organs by over 1 For all studies, overall differences in CFU and cytokines between treatment groups and between neutrophil status were assessed for significance by two-way ANOVA. Individual comparisons made between treatment groups of mice of the same neutrophil status (* symbols above), or comparisons made between wild-type or neutropenic mice given the same inoculum (# symbols below) were tested using Tukey's post hoc test with P values displayed in the graph. P values are replaced with ns (not significant) if P Ͼ 0.1, * if P Ͻ 0.05, ** if P Ͻ 0.01, or *** if P Ͻ 0.001. P values below plots compare CFU or cytokine levels of wild-type and neutropenic mice within the same treatment group. log, though results were more variable. Cytokine levels were also more variable as expected, although global increases in inflammatory cytokines such as granulocytemacrophage colony-stimulating factor (GM-CSF), IL-6, and IL-17 were observed compared to the nonlethal dose (Fig. S4). These cytokines appeared to drop in the presence of either subclass.
Overall, we observed protection of both antibodies in both immunocompetent and neutropenic mice, with a slight advantage of the mIgG 1 subclass in mediating small localized infection. At higher doses, both subclasses demonstrated equal efficacy in healthy mice, while only mIgG 3 subclass showed significant efficacy in neutropenic mice.

DISCUSSION
Though the pharmaceutical industry has focused primarily on development of hIgG 1 antibodies, there are increasing efforts to compare the efficacy of different antibody isotypes and subclasses in treating both cancer and infectious disease (29)(30)(31). Despite mIgG 2a and mIgG 2b being most similar to hIgG 1 , direct comparisons between mIgG 1 and mIgG 3 exclusively have provided important insights into antibody-mediated resolution of infection (12,32,33). Studies investigating anticapsular antibodies against Cryptococcus neoformans, for example, have suggested that mIgG 3 is poorly suited to protect against infection compared with mIgG 1 (34,35), whereas protection in mice from Bacillus anthracis spores was exclusively mediated by mIgG 3 (8). Subclass-specific antibody interactions with K. pneumoniae have not been previously studied in detail. Although humoral responses to a CR-Kp hexasaccharide vaccine were predominantly mIgG 1 (36), mlgG 3 , like hIgG 2 , is the primary humoral response to T-independent antigens such as polysaccharide capsules and is the predominant antibody subclass identified after vaccination with full-length capsule (14,37,38). This suggests that mIgG 3 has some evolutionary importance in protecting against encapsulated organisms.
While having identical variable regions, 17H12 mIgG 1 and 17H12 mIgG 3 differ in binding the CR-Kp capsular polysaccharide, inducing serum bactericidal activity, fixing complement, promoting phagocytosis, and promoting neutrophil-mediated bacterial reduction both in vitro and in vivo. The improved binding ability of mIgG 3 subclasses has been previously observed in antibodies against Burkholderia and group A Streptococcus capsules (33,39) and has been attributed to the ability of mIgG 3 to selfaggregate, creating opportunities for cooperative binding and increased avidity (40,41). Our binding and agglutination data extend this knowledge to anticapsular antibodies against CR-Kp, although we observed an intermediate level of binding after F(ab=) 2 digestion of mIgG 3 , whereas others have observed mIgG 3 F(ab=) 2 fragments to have no better binding than mIgG 1 fragments (33,39). This difference may be due to greater contribution of the mIgG 3 hinge region or disulfide bonds to K. pneumoniae CPS epitopes (40), and further digestion of the antibody into Fab fragments or replacement of heavy chain domains may shed more light on these interactions (8). Nonetheless, aggregation and cooperative binding appear to be important in defending against encapsulated organisms, since in addition to mIgG 3 and cold agglutinin IgM, the T-independent subclass hIgG 2 has also demonstrated the ability to self-aggregate (42). Such properties, however, make preparations of monoclonal antibodies exceedingly difficult to purify and store.
Additionally, agglutination of the 33576 strain by our antibodies suggest that clade 2 CPS functional epitopes could be conserved across geographic location. Our previous studies were restricted to wzi154 clinical isolates collected in the greater New York City area (14,22). Although further study of the wzi154 and other ST258 epitopes is warranted, such evidence adds more confidence to efforts to develop cross-protective antibodies and vaccines.
Several studies have shown the resistance of CR-Kp strains to serum and the efficacy of antibodies in overcoming this resistance (14, 15, 22, 43). Observing serum bactericidal activity is important when comparing antibody-mediated activity against CR-Kp.
Our results reiterate previous findings that mIgG 3 Abs promote greater serum killing, as well as C3c and C5b-9 deposition, than mIgG 1 Abs can. Fixation of complement by mIgG 3 has been shown to be mediated by its CH2 domain (44), and the poor ability of mIgG 1 to fix complement has also been observed (45,46). As previously found and also demonstrated in this study, the capsule is imperative to CR-Kp survival in blood (15). Thus, mIgG 3 may be more advantageous in limiting CR-Kp hematogenous dissemination through complement fixation.
The relative contributions of macrophages, monocytes, and neutrophils to CR-Kp clearance has been disputed. While evidence using cell-specific depletions in mice strongly suggested that lung clearance of CR-Kp is predominantly mediated by CCR2positive macrophages over neutrophils (25), studies examining the role of human and primate neutrophils in CR-Kp clearance have shown their efficacy in clearing bacteria in vitro (15,27,47). These contributions matter significantly to the field of anti-infective antibody therapy in the context of CR-Kp, as up to 86% of patients with CR-Kp bacteremia are neutropenic, and these patients have been found to face worse prognoses than patients without neutropenia (19,23). Our findings demonstrate that 17H12 mIgG 1 may promote improved phagocytosis of CR-Kp relative to that promoted by 17H12 mIgG 3 in murine J774 cells, as well as similar phagocytosis in bone marrowderived macrophages. This is interesting, as these antibodies are thought to act via different receptors on the macrophage surface (32,48). Nevertheless, phagocytosis of two CR-Kp strains did not correlate with killing of the bacteria, as CFU and visual evidence indicated that once inside, the CR-Kp was able to evade killing by the macrophage, and indeed to multiply within the cell. This phenomenon was also observed previously in non-CR K. pneumoniae strains, which were demonstrated to inhibit phagolysosome fusion (49). It is possible that coordination of macrophages with additional immune cells and cytokines in concert may be required for the full capability of antibody-mediated opsonophagocytosis to be realized (50)(51)(52)(53). Additionally, alveolar macrophages or inflammatory monocytes may have improved lysosomal capabilities relative to standard BMDMs (54), or macrophages may clear phagocytized bacteria self-destruction via autophagy or pyroptosis (50,51).
Our studies in human neutrophils showed that mIgG 3 promoted better clearance of CR-Kp by neutrophils at lower concentrations in the presence of serum. However, while antibody-mediated killing of CR-Kp by neutrophils depended on serum, the production of ROS upon stimulation with CR-Kp was not, as demonstrated with the production of ROS with heat-killed FBS-enriched media. Such findings suggest that ROS released by neutrophils in response to CR-Kp may be reactionary without being protective; further studies examining ROS responses in vivo, as well as studies examining other neutrophil protection mechanisms, such as lysosomal activity and neutrophil extracellular trap release, are warranted.
Our in vivo data provides several important findings. First, as previously stated, we observed CFU in the lungs of MMC39-infected mice to be higher in the neutrophildepleted mice in both sublethal and lethal challenges, suggesting that neutrophils are indeed important in protection against pulmonary infection by this wzi154 isolate. This runs counter to other studies that found no change in CFU in lungs of neutropenic mice (25,28). Our laboratory has previously discovered variability in the virulence of ST258 strains, including within wzi154 strains (22), and additional work has determined that neutrophils can clear some ST258 strains (47). Therefore, it is possible that immune responses to different CR-Kp strains may differ and thus be responsible for these differences.
Additionally, we observe potential differences in protection by the different subclasses. While both antibodies reduced bacterial burden in the lungs of mice, 17H12 mIgG 1 performed better, and the mIgG 1 was associated with higher levels of inflammatory cytokines in neutropenic mice than those in the control-treated or mIgG 3treated mice when given a sublethal dose. We suggest that at low inocula, monocytes and macrophages may be more important for infection control, and as mIgG 1 showed better opsonophagocytosis, it may function as the better subclass in these mice. As mentioned, bacterial control by macrophages could be augmented by other immune populations such as gamma delta T cells and innate type III lymphocytes, all of which may produce IL-17 to potentiate Klebsiella immunity (28,55,56). Increased levels of IL-17, IL-12, and other cytokines in the neutropenic mice may thus compensate for neutrophil responses through action by macrophages, monocytes, and other cells. IL-17, produced by resident lymphocyte populations, has been identified as indispensable in protection against K. pneumoniae pulmonary infection (28,(55)(56)(57)(58). However, these neutrophil-independent responses may be insufficient at higher inocula, as evidenced by the reduction of mIgG 1 efficacy in the neutropenic mice challenged with lethal infection. In contrast, drops in inflammatory cytokines in mIgG 3 -treated neutropenic mice may indicate that other components, such as complement, may be sufficient to control infection and require fewer compensatory distress cues for local control. Complement has been shown to be important in lung clearance of other pathogens early in infection (59,60). Furthermore, retention of efficacy by mIgG 3 in the high-inoculum neutropenic scenario may suggest a large role of complement in defending against more disseminated infection, when local control of infection by macrophages and other resident populations may be insufficient to compensate for the role of neutrophils. Studying subclass using complement depletion models may provide additional insights into the relative contribution of complement in antibodymediated protection.
Our study has several limitations. First, large heterogeneity in virulence exists between K. pneumoniae isolates, even within CR-Kp subsets (22), and our contrasting findings regarding neutrophil protection highlight the need to further study heterogeneity of the pathogen-immune response in numerous CR-Kp isolates (25). Furthermore, the functions of individual monoclonal antibodies, even those with the same subclass and similar target, can also be heterogeneous. Some anticapsular Streptococcus pneumoniae antibodies are better able to promote opsonophagocytosis, while others may directly interference with bacterial signaling and growth (61,62). Therefore, future studies of CR-Kp anticapsular antibodies should investigate several to better sample their potential. Finally, future studies of anti-CR-Kp antibodies must innovate the field of in vivo models used to study CR-Kp infection, as convenient and effective models that reproduce the chronic, persistent infection caused by CR-Kp in humans have been difficult to develop (27).
In conclusion, we find that the subclass differences of an anticapsular antibody can affect various facets of immune function against carbapenem-resistant Klebsiella pneumoniae but can exhibit similar efficacy in vivo. This information will promote future monoclonal antibody work on CR-Kp to provide effective therapies, and supports the potential of antibody 17H12 as a candidate to further study. Furthermore, we observe a role for neutrophils in antibody-mediated protection in vivo, encouraging further efforts to investigate the pathogen-host interactions of CR-Kp. between the ori and the existing kan r cassette. Insertion of the gene was confirmed both by sequencing and by digest with BssHII, which both the original pPROBE-Kt and the sh ble gene possessed. The plasmid was transformed into MMC39 by electroporation and plated onto Lennox (low-salt) LB with pH adjusted to 8.0 and containing 50 g/ml bleomycin (Zeocin; Thermo Fisher). Transformation was confirmed by growing positive colonies on 50 g/ml bleomycin and 50 g/ml kanamycin and observing GFP fluorescence of selected colonies under a microscope. To test for plasmid stability, we tested replicate plating of 100 unselected colonies onto selective plates, all of which grew. Additionally, nearly all bacteria screened after 10 serial exponential cultures of MMC39 in the absence of antibiotics were shown to be GFP-positive (GFP ϩ ) by microscopy. For all later experiments, MMC39-GFP isolates were streaked onto Miller LB agar supplemented with 50 g/ml of kanamycin and grown in kanamycin-supplemented Miller LB broth.
Subclass switch production and sequencing. The mIgG 1 switch variant of the 17H12 mIgG 3 murine hybridoma was generated as previously described (12). Briefly, the mIgG 3 parent hybridoma line was treated with endotoxin and IL-4, and spontaneous switch variants were identified through enzyme-linked immunosorbent spot (ELISpot). Sib selection was initially utilized, followed by two rounds of fluorescence-activated cell sorting (FACS) of cells stained with fluorescein isothiocyanate (FITC)-conjugated rat anti-mouse IgG 1 (catalog no. 1144-02; SouthernBiotech), and by rounds of soft-agar cloning in SeaPlaque agarose. Vials of frozen hybridoma clones were sent to GenScript for variable region exon sequencing and analysis through the IMGT/V Quest program.
Antibody purification. Antibodies were produced weekly over 6 months from respective hybridomas grown in CELLine (Wheaton) flasks fed with high-glucose Dulbecco's modified Eagle's medium (DMEM) plus 10% NCTC medium and 1ϫ penicillin-streptomycin and 1ϫ nonessential amino acids, supplemented with either 10% or 5% FBS in the inner and outer chambers, respectively. These antibodies were purified using Pierce protein G affinity chromatography per the manufacturer's instructions. Eluted antibody was neutralized in Tris-HCl (pH 8.0) and NaCl to final concentrations of 100 mM and 300 mM, respectively, then concentrated by centrifugal filtration (Amicon 30K), filter sterilized, snap-frozen in liquid nitrogen, and stored at Ϫ80°C until use. Concentration was determined by absorbance at 280 nM (extinction coefficient ϭ 1.4), which correlated with Bradford assay results. F(ab=) 2 generation. F(ab=) 2 fragments of 17H12 mIgG 3 and mIgG 1 were generated and purified using the Pierce F(ab=) 2 preparation kit and mouse IgG 1 Fab/F(ab=) 2 preparation kits following the manufacturer's instructions, except for digestion temperature and duration (10 min at ambient temperature for mIgG 3 and 36 h at 37°C for mIgG 1 ). Coomassie staining of SDS-PAGE in nonreducing conditions was utilized to ensure MAb/F(ab=) 2 purity after all purifications/digestions. Binding affinity. The EC 50 of the MAbs was calculated using ELISA as described previously (26). Briefly, polystyrene plates (Corning 3690) were coated with 0.5 mg/ml of ST258 clade 2 CPS (MMC34) in PBS, then blocked with 1% PBS-bovine serum albumin (BSA). The MAbs or F(ab=) 2 fragments were serially diluted starting at 1 M [assuming molecular weights (MW) of 150 kDa for full IgG and 110 kDa for F(ab=) 2 , respectively] and proceeding 2-fold. Antibody was detected using a horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG kappa secondary antibody (1:1,000, catalog no. PA1-86015; SouthernBiotech) and developed with 1-Step Turbo TMB (N,N,N=,N=-tetramethyl-1,3-butanediamine) ELISA substrate (Thermo Fisher) according to the manufacturer's instructions. Between steps, wells were washed four times with PBS-0.1% Tween 20. Experiments were repeated on two different days with two different antibody purification batches and digests to ensure reproducibility. Control antibodies were run in parallel as negative controls (catalog no. 0102-01 and 0105-01; SouthernBiotech).
Agglutination. Agglutination of bacteria by MAbs was detected by flow cytometry and confirmed by microscopy, similarly to previous studies (21,64). Washed cultures were diluted to approximately 1 ϫ 10 8 CFU/ml, and 40 l was added to 160 l of PBS with the appropriate antibody concentration and 0.5% BSA to give a final concentration of 2 ϫ 10 7 CFU/ml. The samples were mixed gently in roundbottomed flow cytometry tubes and incubated at 37°C in a shake incubator for 1 h, and later fixed by adding 200 l of 2% paraformaldehyde (PFA) and incubating for 10 min at room temperature (RT). Fixed samples were analyzed via FACSCalibur using forward and side scatter to determine clump sizes and also visualized on glass slides under phase-contrast microscopy (EVOS FL Auto, 40ϫ and 100ϫ objectives; Thermo Fisher). Voltages for flow cytometry were fixed throughout, but gating of bacteria was adjusted based on individual strains incubated in PBS alone to account for size differences between strains. In total, 50,000 events within the gate were counted per sample, representing ϳ75% of all recorded events. The percentage of positive agglutination events was calculated by measuring the percentage of gated cells whose forward scatter exceeded a value representing the largest 1% of events for bacteria treated with PBS alone. The control antibody utilized, 14G8, was an mIgG 1 against Staphylococcus aureus enterotoxin B (SEB) (65).
Serum resistance assays. Serum resistance/killing assays were modified from a previously described assay (22). Briefly, 250 l of a 1 ϫ 10 5 CFU/ml solution was added to 750 l of PBS containing 20% fresh or heat-inactivated (HI) human serum from a healthy donor. HI serum was generated from donor serum by a 30-min incubation in a 57°C water bath. Tubes were incubated at 37°C, rotating end over end. At 0, 60, and 120 min after mixing, 100 l was sampled from each tube, diluted, and plated onto LB agar for CFU quantitation. Percent survival was measured as a fraction of the CFU count at 0 min. Complement deposition assays. Flow cytometry was used to detect complement deposition of C3c and C5b-9, as previously described (14). Briefly, bacteria were diluted to 1 ϫ 10 8 CFU/ml in 1 ml PBS-BSA 1% or 20% fresh human serum (in PBS-BSA). PBS or 10 g/ml of antibody was then added, and bacteria were incubated either 20 min or 40 min at ambient temperature for C3c and C5b-9 deposition, respectively. Bacteria were washed, resuspended in PBS-BSA, and incubated with either FITC-conjugated sheep anti-human C3 (catalog no. AHP031F; Bio-Rad) at 1:500 or AF488-conjugated mouse anti-C5b-9 (ae11) (catalog no. 5120AF488; Novus Biologicals) at 1:150 or without antibody for 20 min at 4°C. After incubation, bacteria were washed and analyzed for fluorescence by FACSCalibur. Integrated geometric mean fluorescence was measured as the product of the percentage of gated events that passed a fluorescence threshold and the mean fluorescence of those events that passed the threshold.
Macrophage phagocytosis assays. BMDMs were differentiated from frozen bone marrow from 6 week-old c57BL/6 mice (Taconic) as previously described (66), except using pure macrophage colonystimulating factor (M-CSF) (10 ng/ml) rather than L929 medium as the M-CSF source for feedings on days 1 and 4. Differentiation of cells was confirmed by flow cytometry on cells stained with FITC-conjugated anti-F4/80 and BV510-conjugated CD11b (purity, Ͼ98% double positive). Macrophage phagocytosis as measured by CFU was performed similarly to previous protocols (22,26), Briefly, 1 ϫ 10 5 BMDM or J774A.1 cells were incubated overnight in wells of cell culture-treated 96-well plates. BMDMs were cultured in RPMI with HEPES and L-glutamine supplemented with 10% FBS and 1ϫ nonessential amino acids, while J774A.1 cells were cultured in DMEM supplemented with 10% FBS, 10% NCTC-109, and 1ϫ nonessential amino acids. The following day, 1 ϫ 10 7 /ml bacteria were opsonized for 20 min in respective cell culture media containing 40 g/ml of either mIgG 1 , mIgG 3 , or control mIgG 1 , and 100 l of this (multiplicity of infection [MOI] ϭ 10) was added to each well of the washed macrophage plates. After 30 min of incubation at 37°C in 5% CO 2 , cells were washed thrice and exposed to medium with 100 g/ml of polymyxin B for 20 min. Cells were washed again, and time 0 wells were immediately lysed twice with water and plated, while those at later time points remained in culture medium until needed. All conditions were performed in triplicate wells. The number of CFU calculated from LB plates was divided by the number of estimated cells plated to give the phagocytic index. Microscopy was performed using the MMC39-GFP strain and EVOS FL Auto (40ϫ objective; Thermo Fisher) using phase contrast and a GFP light cube.
Neutrophil killing assays. Neutrophil assays were adapted from a previous protocol (15). Briefly, 5 ϫ 10 5 bacteria opsonized for 30 min at RT in RPMI medium containing appropriate antibody concentrations were added to 5 ϫ 10 5 human neutrophils in RPMI medium containing a final concentration of 5% autologous fresh or HI serum. At 0, 15, 30, and 60 min, 100 l were sampled from reaction tubes, lysed in cold PBS plus 0.1% Triton X-100, diluted in PBS, and plated on LB agar for CFU quantitation. Percent survival was measured as a fraction of the CFU count at 0 min. Tubes containing sera but not neutrophils were run in parallel. Neutrophils were purified from whole blood using a MACSxpress whole blood neutrophil isolation kit (Miltenyi), treated once for 5 min with red blood cell lysis buffer, and resuspended in RPMI medium on ice until use. Flow cytometry performed on neutrophils purified from two experiments both showed a purity pf Ͼ99%.
Pulmonary infection experiments. We used c57BL/6 mice (Taconic) aged 7 to 9 weeks for all mouse experiments, and pulmonary infection was performed as previously (14,67). At 48 and 4 h prior to the procedure, mice were injected intraperitoneally with 225 g of rat anti-mouse Ly6G (1A8) or a control rat anti-mouse IgG 2a (2A3) (BioXcell). Neutrophil depletion was confirmed previously using flow cytometry of lung homogenates (Ly6G ϩ , Ly6C Ϫ , CD11b Ϫ ). Inocula were prepared by resuspending MMC39 in a Tris-glycine buffer containing 5 mg/ml of an ovalbumin control mIgG 1 (Crown Biosciences), 17H12 mIgG 1 , or 17H12 mIgG 3 to a final concentration of 6 ϫ 10 6 or 3 ϫ 10 7 CFU/ml. After 1 h of opsonization, 50 l of the inoculum was instilled into the surgically exposed trachea of a mouse under ketamine/xylazine using a bent 27-gauge needle. After 20 h, mice were euthanized, and lungs, liver, and spleen were collected and processed in NP-40 or PBS and diluted to enumerate CFU. Supernatants of lung homogenates used for cytokine analysis were stored at Ϫ80°C with 1ϫ Pierce proteinase inhibitor until testing using Bio-Plex Pro mouse cytokine Th17 panel A with additional GM-CSF and IL-12p70 singleplex sets on a Bio-Plex 200 Platform (Bio-Rad). Cytokine levels were normalized against total protein measured by Bradford assay.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. acapsular mutant. We also thank Bruna Seco and Peter Seeberger for their initial contributions to the project. Additionally, we thank Anne Savitt for her assistance in editing the manuscript.