IV BCG immunization drives higher and more durable plasma antigen-specific antibody titers
Following IV BCG immunization, there was a marked increase in Mtb whole-cell lysate reactive IgG and IgA titers compared to BCG administration by other routes (20). However, in this first study, the antibody responses elicited by the different BCG vaccination strategies to distinct Mtb antigen targets were not assessed. Thus, we sought to determine whether particular antigen-specific antibody populations are differentially induced by different BCG vaccination strategies. Antibody levels to a panel of Mtb antigens were compared using a custom, multiplexed Luminex assay (23). The antigen panel included: purified protein derivative (PPD) – a heterogenous collection of Mtb proteins (24), lipoarabinomannan (LAM) – a critical cell wall glycolipid (25), HspX – a stress induced intracellular protein (26), as well as PstS1 and Apa – both cell membrane associated glycoproteins linked to host cell invasion (27, 28). Of note, each of these antigens are expressed by both BCG and Mtb. Plasma samples collected pre-vaccination, week 8 post-vaccination, time of Mtb challenge (week 24), and post-infection (week 28) were analyzed, and fold change in antibody titer over pre-vaccination levels was calculated for each macaque at each timepoint.
Following immunization and prior to infection, antigen-specific IgG1 responses were detected in macaques across all vaccine arms. There were weak responses to all antigens in animals receiving standard ID BCG (Fig 1A). Conversely, those that received IV BCG vaccination displayed the largest increase in plasma IgG1 titers to nearly all tested antigens following vaccination. More specifically, PPD, LAM, PstS1, and Apa IgG1 titers in the IV BCG group were each significantly higher than those in the standard ID BCG group both at week 8 post-vaccination, and at the time of Mtb challenge (week 24 post-vaccination) (Fig 1A). The additional vaccination groups – high dose intradermal (IDhigh), aerosol (AE), and aerosol/intradermal (AE/ID) – trended towards higher IgG1 levels compared to the standard ID BCG group, though the differences were only significant for the IDhigh Apa-specific response (Fig 1A).
Antigen-specific IgA responses were also observed following vaccination across each of the experimental BCG vaccination groups, though the fold increases in IgA titer were not as prominent as those for IgG1. IV BCG vaccinated macaques elicited significantly higher IgA titers to LAM and PstS1 at week 8 post-vaccination and at the time of Mtb challenge compared to the standard ID BCG group, which did not generate a detectable increase in antigen-specific IgA titers to any of the antigens following vaccination (Fig 1B). Animals in the IDhigh group also elicited a significant increase in LAM IgA titers at week 8 post-vaccination (Fig 1B). However, minimal vaccine-induced plasma IgA responses to the additional antigens were detected in the other groups (Fig 1B).
Finally, antigen-specific IgM responses were detected in multiple experimental BCG vaccination groups, with IV and IDhigh vaccinated macaques mounting the strongest peripheral IgM responses to BCG vaccination. IV BCG vaccinated animals exhibited significantly higher LAM-, PstS1-, and Apa-specific IgM titers at week 8 following vaccination compared to the standard ID BCG group, while responses in the IDhigh group were not significantly different (Fig 1C). LAM IgM titers also trended higher at the time of challenge in the IV group, though the difference was no longer significant at this timepoint (Fig 1C).
Together, these data highlight peripheral differences in the antibody response to specific antigens induced by distinct BCG vaccination strategies. IV BCG immunized animals generated particularly high antigen-specific antibody levels in the plasma, with protein- and LAM-specific antibody responses persisting exclusively in IV immunized animals 24 weeks following vaccination, to the time of Mtb challenge. Further, standard ID BCG vaccination generated a weaker antigen-specific antibody response than the experimental vaccination regimens tested – each of which delivered a larger dose of BCG in addition to changing route – suggesting that altering both route and dose may result in enhanced peripheral humoral immune responses to BCG vaccination.
IV BCG vaccination uniquely elicits a robust lung-compartmentalized antibody response
We next aimed to profile the antigen-specific humoral immune response at the site of infection using bronchoalveolar lavage fluid (BAL) collected from each macaque pre-vaccination, week 4 post-vaccination, and week 16 post-vaccination – the final timepoint the BAL procedure was performed prior to Mtb challenge. IV BCG vaccination uniquely elicited a robust antibody response in the airways following vaccination (Fig 2A – C). Specifically, IV BCG vaccinated animals mounted IgG1, IgA, and IgM responses in the BAL that were significantly higher than the standard ID BCG group at week 4 across all mycobacterial antigens tested (Fig 2A – C). The magnitude of the responses was particularly striking, with over 100-fold increases in antibody levels observed for some IV BCG vaccinated macaques (Fig 2A). Most antibody responses elicited were transient, with only statistically significant levels of LAM-, PstS1-, and Apa-specific antibodies detected in the BAL 16 weeks following vaccination. A small number of macaques in the IDhigh, AE, and AE/ID groups additionally generated detectable antibody titers in the BAL following BCG vaccination (Fig 2A – C). However, these responses were limited to one or two macaques in each group, and were substantially lower in magnitude than responses generated in IV BCG immunized animals (Fig 2A – C). These data indicate that IV BCG vaccination alone induced strong, lung-compartmentalized, antigen-specific humoral immune responses. These antibodies contracted, but persisted at detectable levels in IV immunized animals for at least 4 months following immunization.
Antibodies from IV BCG vaccinated macaques mediate superior innate immune activation
Beyond their ability to bind and recognize pathogens or pathogen-infected cells, antibodies are able to deploy the anti-microbial activity of the innate immune system via Fc:Fc-receptor engagement to control a wide range of microbes (29). Further, Fcγ receptor (FcγR) signaling is necessary for the optimal survival and bacterial containment of Mtb in mice (30). Thus, we next measured the FcγR binding and functional capacity of plasma- and BAL-derived antibodies elicited in each BCG-vaccinated macaque to determine whether certain antibody FcγR binding profiles and/or antibody effector functions selectively tracked with distinct BCG vaccination strategies.
In the plasma, the IV BCG group displayed a trend towards higher levels of FcγR binding antibodies than the standard ID BCG group across nearly all antigens measured, including significantly increased PPD-, PstS1-, and Apa-specific FcγR2A and FcγR3A binding antibodies 8 weeks post-vaccination (Fig 3A). Further, although FcγR binding antibodies waned by the time of Mtb challenge across all vaccination groups, IV BCG vaccinated macaques maintained significantly higher levels of PPD- and PstS1-specific FcγR2A and FcγR3A binding antibodies close to the time of challenge, suggesting durable antibody functionality in this group (Fig 3A). To examine plasma antigen-specific antibody functionality, antibody-dependent phagocytosis by monocytes and neutrophils, as well as NK cell degranulation assays were performed. Each of these measurements were captured for LAM-specific antibodies, as each BCG vaccination regimen elicited detectable LAM-specific antibody titers in the plasma (Fig 1A – C). Antibodies from IV BCG vaccinated macaques induced the most potent antibody-dependent neutrophil phagocytosis, which was moderately, yet significantly higher than that observed in the standard ID BCG group at week 8 post-vaccination (Fig 3B). In contrast, limited differences were observed in LAM-specific antibody-dependent monocyte phagocytosis and antibody-dependent NK cell degranulation – a surrogate for antibody-dependent cellular cytotoxicity (ADCC) (31) – across the vaccine groups (Fig 3B).
In line with the elevated antibody levels observed in the BAL of IV BCG vaccinated macaques (Fig 2A – C), antibodies in the IV group demonstrated the highest levels of FcγR binding (Fig 3C). IV BCG immunized animals generated significantly higher levels of LAM-specific FcγR2A binding antibodies at week 4 post-vaccination (Fig 3C). In addition, antigen-specific FcγR3A binding levels in the BAL were particularly robust in the IV vaccinated group, with IV macaques displaying significantly higher levels of FcγR3A binding antibodies to PPD, LAM, PstS1, and Apa 4 weeks following vaccination (Fig 3C). LAM-, and PstS1-specific FcγR3A binding antibody levels remained significantly higher at week 16 post-vaccination (Fig 3C). Furthermore, BAL-derived antibodies in the IV group demonstrated superior LAM-specific functional activity. Specifically, BAL-derived antibodies from IV BCG vaccinated animals exhibited a trend towards stronger antibody-dependent monocyte phagocytosis activity (Fig 3D). More strikingly, the IV BCG group demonstrated significantly higher antibody-dependent neutrophil phagocytosis and NK cell degranulation activity week 4 vaccination, with little functionality observed in the other vaccination groups (Fig 3D). This activity returned to baseline in a majority of animals by week 16 post-vaccination (Fig 3D).
Previous data have linked an enrichment of FcγR3A binding and NK cell activating antibodies in the setting of LTBI, to enhanced intracellular Mtb killing in macrophages (32). Therefore, given the expansion of both of these humoral features particularly in the BAL of the IV immunized group 4 weeks post-vaccination, we next examined the anti-microbial activity of antibodies from each vaccination group in this context. Human monocyte-derived macrophages were infected with a live/dead reporter strain of Mtb (33), followed by the addition of pooled plasma or BAL from each BCG vaccination group. Plasma from the IV group did not drive significant Mtb restriction across either of the timepoints (Fig 3E). Conversely, the week 4 IV BCG BAL pool did drive moderate, yet significant intracellular Mtb restriction, whereas the IDhigh BCG BAL pool tended to enhance infection. These patterns were consistently observed across all tested macrophage donors (Fig 3F).
Taken together, these data highlight the induction of highly functional antibodies following IV BCG immunization in rhesus macaques. Further, the increases selectively observed in FcγR3A binding, NK cell degranulation, and intracellular Mtb killing in the BAL were particularly salient given recent associations reported between both FcγR3A binding, as well as NK cell activity, and improved Mtb control (32, 34).
Antigen-specific IgM titers in the plasma and BAL negatively correlate with Mtb burden
A spectrum of bacterial burden was observed in the lungs of rhesus macaques across the BCG vaccinated groups at the time of necropsy (20). Thus, despite IV immunization clearly affording optimal bacterial control following Mtb challenge, we next aimed to define whether any antibody features exhibited a robust relationship with lung Mtb burden.
In the plasma, 5 antibody measurements were significantly negatively associated with Mtb burden after multiple hypothesis testing correction (Fig 4A) (35). Surprisingly, each of the features identified were antigen-specific IgM titers at week 8 post-vaccination or at the time of challenge (Fig 4A and B), revealing an unexpected significant relationship between plasma antigen-specific IgM titers and improved outcome following Mtb challenge. In contrast, while higher antibody titers have historically been associated with elevated antigenic burden and enhanced Mtb disease, antibody levels and features were not identified that tracked positively with Mtb burden at either significance level (Fig 4A).
In the BAL, 18 antibody features were significantly negatively associated with Mtb burden after multiple hypothesis testing correction (Fig 4C) (35). Several antigen-specific IgG1, IgA, IgM, and FcγR binding measurements in the BAL at 4 or 16 weeks post-vaccination were negatively correlated with Mtb levels at the time of necropsy (Fig 4C and D). The majority of these features associated with reduced Mtb bacterial burden included antibody features present week 4 post-vaccination, the exception being LAM and PstS1 IgM titers at week 16 (Fig 4C and D). Again, none of the BAL antibody features measured had a significant positive correlation with Mtb burden at necropsy at either significance level (Fig 4C).
Collectively, the particularly low Mtb burden present in IV immunized animals indicate that the relationship observed between select humoral features and bacterial burden track with vaccination route, and thus may not represent independent correlates of protection. Nevertheless, these analyses point to the vaccine-induced humoral immune features which track most closely with improved microbial control in this vaccination cohort. Notably, IgM responses alone tracked with reduced Mtb burden close to the time of challenge across both compartments, potentially representing direct mechanistic correlates of immunity, or markers of a unique functional humoral immune response in these animals.
Antibody profiles accurately distinguish protected and susceptible BCG-vaccinated macaques
Given that many antibody titer and functional measurements were highly correlated, even across compartments, we next sought to determine whether a minimal set of antibody features could be defined that collectively tracked with Mtb control. Thus, macaques with a lung Mtb burden at necropsy below 1000 were categorized as protected (total n = 11; 9 IV BCG, 1 IDhigh, 1 AE/ID BCG), and those with an Mtb burden greater than or equal to 1000 were categorized as susceptible (total n = 37). Next, least absolute shrinkage and selection operator (LASSO) regularization was implemented on the standardized antibody data, removing variables unrelated to the outcome, as well as reducing the number of highly correlated features (36). Partial least squares discriminant analysis (PLS-DA) was then performed to visualize and quantify group separation (37, 38).
Robust separation was observed between protected and susceptible macaques on the basis of humoral profile (Fig 4E). The model distinguished protected from susceptible animals with a balanced cross-validation accuracy of 89.6% (Fig 4E). Remarkably, only 3 features were required to achieve this high level of predictive accuracy: BAL HspX-specific IgM at week 4, plasma LAM-specific IgG1 at week 8, and plasma LAM-specific IgM at the time of challenge. Each of these features contributed to separation along latent variable 1 (LV1) (Fig 4F). The selection of these three variables across distinct timepoints suggests that substantive humoral differences were present between protected and susceptible BCG-vaccinated macaques beginning in the lung in week 4, extending out to the time of challenge in the plasma. Further, this analysis demonstrates that protected and susceptible BCG-vaccinated macaques can be accurately resolved by simply using antibody titer measurements.
Protective vaccination via attenuated Mtb is associated with increased plasma IgM titers
While our analyses identified humoral features associated with reduced Mtb burden in BCG immunized animals, because vaccination route was so closely linked to protection in this cohort, the generalizability of these findings was unclear. Thus, we next queried whether similar humoral features were associated with Mtb control in an independent Mtb vaccination study in NHPs. Previous work demonstrated that AE vaccination with an attenuated Mtb strain (Mtb-ΔsigH) provided superior protection compared to AE BCG vaccination in rhesus macaques (22). Thus, antibody profiling was performed on the plasma of Mtb-ΔsigH or AE BCG vaccinated animals.
Using antibody titer measurements alone, Mtb-ΔsigH and BCG vaccination groups could be clearly separated using a principal component analysis (PCA) (Fig 5A). Analysis of the PCA loadings plot revealed that antigen-specific IgM responses primarily drove separation between the two groups, with antigen-specific IgM responses enriched among protected Mtb-ΔsigH vaccinated macaques (Fig 5B). Similarly, univariate analyses indicated that Mtb-ΔsigH vaccinated macaques elicited significantly higher LAM-specific IgM titers week 7 post-vaccination, as well as a trend towards increased Apa- and HspX-specific IgM titers (Fig 5C). In contrast, minimal differences in antigen-specific IgG1 and IgA titers were noted between the Mtb-ΔsigH and BCG groups (Fig S2). Finally, antibody responses to the Ebola virus negative control antigen were not detected in either group as expected regardless of isotype (Figs 5C and S2).
Thus, although the sample size from this cohort is small, increased plasma antigen-specific IgM titers tracked with reduced Mtb disease. A result similar to that observed in BCG immunized animals (Fig 4A), potentially hinting at a common association between antigen-specific IgM and vaccine-induced Mtb control.
Superior in vitro anti-microbial effect of LAM-specific IgM
Data from both the BCG route and the Mtb-ΔsigH immunization study pointed to an unexpected association of antigen-specific IgM titers with improved vaccine-induced Mtb control. However, whether elevated IgM levels represented a biomarker or contributed directly to anti-microbial control remained unclear. Given the emerging data pointing to an anti-microbial role for polyclonal IgG and monoclonal IgG and IgA antibodies against Mtb (32, 39–44), we next queried whether IgM also might harbor some anti-microbial capacity, using an engineered high-affinity LAM-specific antibody clone (A194) generated as an IgG1 and as an IgM (45).
In light of the previous observation that IgG1- and IgM-rich BAL from IV immunized rhesus macaques could drive intracellular Mtb killing in macrophages (Fig 3F), we first compared the anti-microbial activity of each isotype in a similar human monocyte-derived macrophage model. However, despite the anti-microbial signal observed in the BAL of IV BCG immunized animals, neither LAM-specific monoclonal antibody drove significant intracellular Mtb restriction in macrophages when added post-infection (Fig 6A).
While macrophages represent a primary cellular niche for Mtb in vivo during infection, we also probed the anti-microbial role for each LAM-specific antibody in a whole-blood model of infection – a system which queries the broader role of multiple immune cell types and components in microbial restriction. Specifically, fresh blood from healthy human donors was simultaneously infected with an Mtb luciferase reporter strain (46), and treated with each LAM-specific monoclonal antibody. Luminescence readings were then taken to obtain Mtb growth curves in the presence of each antibody treatment over the course of 120 hours. Remarkably, only the LAM-specific IgM antibody drove significant Mtb restriction in this system (Fig 6B). Further, the LAM-specific IgM antibody drove improved bacterial restriction compared to the IgG1 across nearly every donor tested (Fig 6B).
Ultimately, these data demonstrate that a high-affinity LAM-specific antibody clone drives improved Mtb restriction in whole-blood as an IgM, as compared an IgG1 variant, suggesting that in addition to representing an early marker of vaccine-induced Mtb control, Mtb-specific IgM antibodies have the potential to functionally contribute to immunologic control of Mtb.