Human antibodies targeting a Mycobacterium transporter protein mediate protection against tuberculos

Mycobacterium tuberculosis (Mtb) exposure drives antibody responses, but whether patients with active tuberculosis elicit protective antibodies, and against which antigens, is still unclear. Here we generate monoclonal antibodies from memory B cells of one patient to investigate the B cell responses during active infection. The antibodies, members of four distinct B cell clones, are directed against the Mtb phosphate transporter subunit PstS1. Antibodies p4-36 and p4-163 reduce Mycobacterium bovis-BCG and Mtb levels in an ex vivo human whole blood growth inhibition assay in an FcR-dependent manner; meanwhile, germline versions of p4-36 and p4-163 do not bind Mtb. Crystal structures of p4-36 and p4-170, complexed to PstS1, are determined at 2.1 Å and 2.4 Å resolution, respectively, to reveal two distinctive PstS1 epitopes. Lastly, a prophylactic p4-36 and p4-163 treatment in Mtb-infected Balb/c mice reduces bacterial lung burden by 50%. Our study shows that inhibitory anti-PstS1 B cell responses arise during active tuberculosis.

Exposure to Mycobacterium tuberculosis (Mtb) results in a spectrum of outcomes, one of which results in active tuberculosis (ATB) disease1. Both innate and adaptive arms of the immune response have been implicated in immunity to Mtb2,3, but the role of humoral immunity and more specifically, antibodies, remains controversial4,5. Several studies have suggested that antibodies may play a protective role in at least a proportion of otherwise healthy individuals who have a history of exposure to Mtb4,6,7,8, and antibody responses have been correlated with the protective efficacy of an experimental TB vaccine9. During active disease, higher antibody titers to the mycobacterial glycolipid lipoarabinomannan (LAM) have been correlated with decreased severity of infection10, and there is evidence of B cell dysfunction during active disease that resolves following treatment11. In addition, during an active infection, serum antibodies against the Mtb phosphate transporter PstS1 are detected12,13. However, functional characterization of monoclonal antibodies isolated from patients with ATB has not been carried out, and it is not clear whether such antibodies exhibit any antibacterial effects.

In the present study, we explore B cell responses during ATB, while focusing on the immune-dominant antigen PstS1. We isolate and analyze a total of 85 monoclonal antibodies (mAbs) from one patient with elevated anti-PstS1 responses and find two antibodies, p4-36 and p4-163, from distinct clones that demonstrate inhibitory activity against both BCG and Mtb. Structural analysis reveals that the two antibodies are directed against two different epitopes on PstS1 and do not compete with one another. When administered prior to infection of Balb/c mice with Mtb, both antibodies cause a 50% reduction in lung bacterial burden. Our data show that both p4-36 and p4-163 exhibit antibacterial activity and provide the first proof of concept that protective antibody responses can be generated during the course of active tuberculosis disease, a finding that may inform both therapeutic and preventive vaccine design.

Isolation of PstS1-specific mAbs

We recruited 26 in-patients with ATB (Supplementary Table 1). All patients received standard treatment, recovered, and were eventually discharged. We screened patient sera for anti-Mtb antibody responses prior to initiation of antibiotics using lysates from two pathogenic Mtb strains, H37Rv and CDC1551. Unlike healthy community controls, the majority of patients (23/26) showed serum reactivity by ELISA to at least one lysate (Fig. 1a, b). Testing of antibody titers against soluble lysate versus cell-wall components of nonpathogenic Mtb-H37Ra14 revealed responses against both components, but with higher reactivity against the Mtb surface (Fig. 1c). To identify antibody reactivity against specific Mtb proteins, we tested sera against five selected surface-exposed proteins (produced in-house, Supplementary Fig. 1) that had been described previously to elicit antibody responses in human infection (Fig. 1d–h)15,16. The strongest response was directed against PstS112 (Fig. 1d, i), which has been implicated in Mtb virulence13 and is one of three genes previously identified to contain amino acid sequence variants undergoing diversifying selection in conserved epitopes17,18. Therefore, we decided to characterize antibody responses to PstS1.

We focused on one patient, P.4, who had the strongest individual response to PstS1 and retained high anti-PstS1 responses over his treatment course (Fig. 2a). Hence, we sought to isolate mAbs directed against PstS1 and test their effect on experimental models of Mtb infection. We isolated peripheral blood mononuclear cells (PBMCs) from whole blood and sorted single IgG+, PstS1+ B cells, which comprised 0.5% of the total IgG+ B cell population (Fig. 2b)19. A total of 102 heavy and 90 light chains were amplified by single-cell Ig PCR;20 of these, 85 constituted natural heavy and light pairs (Supplementary Fig. 2). Consistent with previous reports of anti-Mtb B cell responses in humans7, the majority of the antibody sequences were not clonal. Based on VH DH JH, VL, and JL and >75% identity in CDRH321, a total of five clonal families were identified in 16 sequences (Fig. 2c). As previously described21, the VH sequences of the clonal heavy chains were more mutated when compared to the VHsequences that were not part of a clone (Fig. 2d), but both groups had a similar VH gene usage distribution (Supplementary Fig. 2b). We produced nine antibodies derived from four clones expressed in an IgG1 expression vector20 (Fig. 2e, f, expressed antibodies are in red). We were unable to amplify the light chains from Clone 5 variants and therefore antibodies from this clone were not produced. All tested mAbs, except for p4-31, reacted strongly both with PstS1 (with dissociation constants within the nanomolar range, Supplementary Fig. 3) and with Mtb lysates by ELISA (Fig. 2g–i). Among the nine antibodies we produced, mAbs p4-36 and p4-163 exhibited the strongest binding to bacterial lysates (Fig. 2h, i). These mAbs also showed binding to whole-bacteria H37Ra-mCherry compared with isotype control (Supplementary Fig. 4). We therefore decided to focus on these antibodies for the rest of the study.

First, since the anti-PstS1 mAbs were generated during natural infection occurring in patient P.4, we sought to investigate the impact of affinity maturation on their activity. Both p4-36 and p4-163 showed relatively low levels of somatic hypermutations, with p4-36 exhibiting 6 and 12 amino acid changes in heavy and light chains, respectively, and p4-163 exhibiting 10 and 8 amino acid changes in heavy and light chains, respectively. Nevertheless, we hypothesized that these mutations provide the mAbs with their ability to bind PstS1 and Mtb lysates. To test our hypothesis, we reverted the amino acid sequences of p4-36 and p4-163 back to their germline sequences (Supplementary Fig. 5a, b), similarly to what was previously carried out for other naturally elicited mAbs22, thus producing the unmutated predicted germline (gl) versions of each antibody: p4-36gl and p4-163gl. The binding of the germline variants was compared to the binding of the mature antibodies to PstS1, as well as H37Rv and CDC1551 Mtb lysates. Both mAbs showed a dramatic reduction in their binding when somatic hypermutations were reverted, with the exception of mAb p4-36gl that exhibited similar binding to PstS1 as the mature mAb (p4-36mt) by ELISA. However, SPR measurements showed a 4.7-fold reduction in affinity (Fig. 2j, Supplementary Fig. 5c–g). To further investigate whether mutations occurring in the heavy or the light chain contribute to antibody activity, we produced chimeric antibodies where the heavy chain is mature while the light chain is germline (HCmtLCgl), and vice versa (HCglLCmt). Testing those for binding showed that for p4-36, mutations in the light chain were more significant for binding than those in the heavy chain, as opposed to p4-163 where mutations in the heavy chain were more essential for binding compared to mutations in the light chain (Supplementary Fig. 5c–g). Overall, we conclude that both mAbs p4-36 and p4-163 are specific for PstS1 and acquired somatic hypermutations during natural infection that resulted in their improved binding to Mtb.

Activity ex vivo

Next, we sought to test the effects of the mAbs p4-36 and p4-163 on Mtb infection in culture. First, we tested the effect of our mAbs on bacterial entry. We infected phorbol 12-myristate 13-acetate (PMA) differentiated THP-1 macrophages23 with the attenuated H37Ra strain with and without PstS1-specific mAbs. The mAbs bound H37Ra within macrophages and did not prevent, but rather slightly increased, bacterial entry into the macrophages (Fig. 3a, upper panel, mAb binding is shown in the lower panel). We next asked whether the mAbs p4-36 and p4-163 could inhibit mycobacterial growth. For this, we used a whole-blood mycobacterial growth inhibition ex vivo assay (MGIA) where human whole blood cells from healthy donors are infected with pathogenic bacteria24. This system has been used previously to test polyclonal antibodies from healthy donors25 and allows testing of antibody activity in a more physiologically relevant in vitro system. Here, p4-36 and p4-163 were able to significantly restrict the growth of both Mycobacterium bovis-BCG (BCG) and pathogenic Mtb (Fig. 3b, c)24,26. This activity was not dose-dependent and was unique to mAbs p4-36 and p4-163, and was not observed with other anti-PstS1 mAbs, such as p4-31 and p4-141, showing that ELISA binding to Mtb lysates correlated well with their activity ex vivo.

We next sought to investigate the mechanism of mAb-induced CFU reduction over time. MAbs p4-36 and p4-163 did not inhibit infection, and immediately after infection, more antibody-bound bacteria were found in THP-1 cells (Fig. 3a). This suggests antibody-dependent opsonization of bacteria via Fc gamma receptors (FcγRs)6,8,24. To test this, we compared the antibody activity in MGIA by comparing the wild-type IgG1 antibodies with IgG1-N297A (NA), an aglycosylated Fc variant with no binding to FcγRs27. Treating with the NA variants abolished protection compared with wild-type IgG1 mAbs (Fig. 3d). To evaluate the contribution of FcγRs+ effector cells to the inhibitory activity of the antibodies p4-36 and p4-163, we repeated the MGIA experiments with selective FcγR blockers. Blocking the main FcγRs expressed by human macrophages, CD16 (FcγRIIIA) and CD32a, and CD32b (FcγRIIA and FcγRIIB) together eliminated the inhibitory activity of both p4-36 and p4-163 (Fig. 3e). By contrast, specific depletion of T cells or blocking MHC class II molecules did not affect mAb-induced CFU reduction (Fig. 3e). Altogether, our data imply that the Fc domain of the antibodies facilitates their antibacterial activity. We suggest that FcγRs can mediate bacterial opsonization into the cells, thus promoting the antibody inhibition of infection. Whether this Mtb inhibition is due to an active restriction of intracellular bacterial replication or due to induction of bacterial killing is yet to be determined.

Structure–function studies

To further understand the mechanism of the protective antibody binding to its target, we produced antigen-binding fragments (Fabs) of p4-163 and p4-36 and prepared Fab–PstS1 complexes for crystallization. We determined the structure of Fab p4-36 in complex with PstS1 at a resolution of 2.1 Å (Fig. 4a and Supplementary Table 2). Two PstS1–Fab p4-36 heterodimers were in the asymmetric unit of the crystal and had only minor differences in the constant domains of the bound Fabs (Supplementary Fig. 6). P4-36 binds a contiguous epitope located on an alpha-helix formed by residues 141–145 and preceding residues 136–137 and residues 139–140 (Fig. 4b and Supplementary Fig. 7a), within a small surface area of 583 Å2. The contacts between p4-36 and PstS1 are mostly Van der Waals and hydrogen bonds and contributed largely by complementarity-determining region 1 (CDR1) and CDR3 of the light chain (CDRL1 and L3) and CDR3 of the heavy chain (CDRH3) (Fig. 4b, Supplementary Fig. 8a, b and Supplementary Table 3). Ten hydrogen bonds are observed at the interface, and one salt bridge is formed between Asp36p4–36 CDRL1 and Lys136PstS1 (D36L[OD1]-K136PstS1[NZ]) (Fig.4b and Supplementary Table 3).

No crystal could be obtained with the complex of Fab p4-163 and PstS1. Thus, we switched to p4-170, which is a closely related variant of p4-163 and a member of the same B cell clone (Clone 1, Fig. 2e, f), and has 97.6% and 93.86% amino acid sequence identity to p4-163 heavy and light chains, respectively (Supplementary Fig. 5a, b). In addition, p4-170 exhibited similar inhibitory activity in MGIA to p4-163 (Supplementary Fig. 9). The Fab p4-170 and PstS1 complex could be crystallized, and the structure was determined at a resolution of 2.4 Å with one PstS1–Fab p4-170 heterodimer in the asymmetric unit (Fig. 4c and Supplementary Table 2). As revealed by the structure of the complexes, p4-170 binds to a different epitope from the one recognized by p4-36.

The epitope recognized by p4-170 is highly discontinuous and conformationally proximate to the Pi binding site in the middle of the molecule, between the two domains I and II of PstS1 (Fig. 4c). The epitope is a large surface of 977 Å2 and comprises three helices and four loops of PstS1 (Fig. 4d and Supplementary Fig. 7b). The interactions between p4-170 and PstS1 are mediated through salt bridges, hydrogen bonds, and Van der Waals contacts (Fig. 4d and Supplementary Table 4). Except for CDRH1, all the CDRs of p4-170 are involved in the interactions with residues 85, 88, and 92 of the PstS1 domain I and residues 125, 174, 177–178, 181–182, 190, 246–248, 267–268, 270–271, 275–276, and 279–281 of domain II (Fig. 4d, Supplementary Fig. 8c, d and Supplementary Table 4). The hydrogen bonds are formed with both the main-chain and side-chain atoms of PstS1. Four salt bridges are formed at the contact interface, including Glu88PstS1[OE1]–Arg55H[NE], GLu88PstS1[OE1]–Arg55H[NH1], Lys268PstS1[NZ]–Asp97L[OD1], and Asp279PstS1[OD2]–Arg54L[NH2] (Fig. 4c, d and Supplementary Table 4).

To confirm the key antibody: antigen contact residues identified by the structures, we generated a panel of point mutations in PstS1. Mutating residues at the p4-170:PstS1 interface, S246GPstS1, K268EPstS1, and D279APstS1 reduced the binding of all tested Clone 1 mAb variants, but not the binding of p4-36 (Fig. 4e). Amino acid substitutions in the alpha-helix that holds the epitope of p4-36 residues K136EPstS1, D139APstS1, and D140APstS1 reduced p4-36 binding. The binding of Clone 1 mAbs was also reduced, indicating that the alpha-helix bound by p4-36 might be essential for the folding of PstS1.

We next asked whether the binding of p4-36 or p4-170 could interfere with multimerization of the transporter. For this, we performed modeling of the binding of PstS1 to the transporter complex. According to our prediction, the bound antibodies show no blockage to the assembly of the PstA–B–C–S complex and do not block the binding of PstS1 to the PstA, B, C transporter complex (Fig. 5). We conclude that p4-36 and p4-170 do not function by inhibiting the transporter activities .

Activity in vivo

Finally, we tested the activity of the two mAbs p4-36 and p4-163 on Mtb infection in vivo. For this, we used wild-type Balb/c mice. We injected 0.5 or 1 mg mAb per mouse intraperitoneally five hours prior to aerosol infection with pathogenic Mtb. After 2 weeks, the mice were sacrificed and lung bacterial burden determined. Lung burdens were reduced in mice pretreated with mAbs p4-36 and p4-163, or both by approximately 0.5 log CFU (Fig. 6), which agrees with the MGIA results and verifying that the mAbs have anti-Mtb-inhibiting activity.

Here, we describe a patient, P.4, who had active tuberculosis and was selected from a cohort of infected patients due to his strong anti-Mtb serum responses to the known immunodominant antigen PstS1. While the activity of the two anti-PstS1 mAbs we isolated is modest (about 0.5 log), to our knowledge, this is the first description that naturally elicited PstS1 antibodies have anti-Mtb activity in infection.

PstS1 is a 38-kDa phosphate-binding periplasmatic protein, one of three subunits of the Mtb phosphate-specific transporter (Pst) complex12. PstS1 is an immunodominant marker for ATB28,29,30. PstS1 is also necessary for Mtb virulence; PstS1-deletion mutants were attenuated in a mouse-infection model13. Intriguingly, PstS1 was identified as one of only three Mtb genes that are subjected to evolutionary sequence diversification, suggesting that PstS1 variation plays a role in Mtb immune evasion17,18.

In the present study, we characterize two different sites on PstS1 that can be targeted by antibodies capable of inhibiting Mtb growth. P4-163 is a member of the largest B cell clone of nine mAbs that shared >93% sequence identity. The high-resolution structure of a clonal variant of p4-163, p4-170, revealed a large, sparse, and highly conformational epitope that is adjacent to the active site of PstS1. Intriguingly, seven of p4-170 contact residues (Lys268PstS1, Pro270PstS1, Ala271PstS1, Ile275PstS1, Ser276PstS1, Asp279PstS1, and GLy280PstS1) overlap with a highly conserved Mtb T-cell epitope 259-AAAGFASKTPANQAISMIDG-280, Immune Epitope Database number 3517,31. The mAb p4-36, a member of a different and smaller B cell clone, binds an alpha-helix structure within residues 136PstS1–145PstS1, with the epitope being 30 Å distant from the epitope bound by p4-170.

Donor P.4 had active tuberculosis; therefore, by definition, the mAbs we isolated did not prevent or eliminate Mtb infection in this patient. However, evidence from other infectious diseases shows that B cell responses develop in parallel and subsequent to high pathogen loads that stimulate B cell activation and maturation, causing neutralizing/protective antibodies to arrive ‘too little, too late', without a significant benefit to the individual that produced them22,32,33,34. Considering the low extent of affinity maturation, as reflected by the relatively low number of somatic hypermutations in p4-36 and p4-163, it is possible that these B cell clones only started their evolution and did not reach their full anti-Mtb potential. On the other hand, the fact that these antibodies ultimately arise provides evidence that B cells play an active role during ATB.

Despite the long-standing controversy, recent studies have demonstrated that in certain cases, anti-Mtb humoral responses can be protective7,24,35 or correlate with lack of active disease or even lack of infection4,6,35,36. Only one prior study profiled the human B cell responses against Mtb antigens on a monoclonal level7. This study attributed bacterial inhibition to IgA isotypes rather than IgG. However, in that study, ‘protection’ was defined as decreased viable Mtb within cells shortly after infection. Unfortunately, the study evaluated neither in vivo activity nor the precise molecular mode of antigen binding. While the Balb/c mouse model we use in this study does not recapitulate lung granulomas, which are the hallmark of active TB infection in humans, passive transfer of mAbs before pathogenic Mtb infection resulted in a significant reduction of bacterial lung loads in mice indicating a preventive mAb activity.

This is the first report of human anti-Mtb mAbs that have a modest, yet robust, activity in vivo, as well as resolving their structure along with their corresponding Mtb target. PstS1 DNA was previously reported to elicit protective T-cell immunity against Mtb in mice37. The fact that the germline version of p4-36, p4-36gl, binds recombinant PstS1 suggests that PstS1 could be used as an immunogen in humans to elicit B cell responses and p4-36-like antibodies in naive populations. Our study has implications for the development of new anti-Mtb therapies and prevention.