Cross-Neutralizing and Protective Human Antibody Specificities to Poxvirus Infections

Orthopoxvirus Infection in Humans Elicits a Complex B Cell Response Encoding Large Numbers of Clones Reactive with Antigens from Diverse Orthopoxvirus Species

We obtained peripheral blood mononuclear cells (PBMCs) from a donor who had recovered from a naturally occurring MPXV infection or from otherwise healthy subjects previously immunized with one of three different vaccine formulations (Table S1), IMVAMUNE (live attenuated modified vaccinia Ankara virus), Dryvax (a freeze-dried calf lymph produced vaccinia virus), or ACAM2000 (Vero cell culture produced vaccinia virus) (Verardi et al., 2012). To identify orthopoxvirus-specific B cell cultures, PBMCs were transformed with Epstein-Barr virus, and the supernatants from the resulting lymphoblastoid cell lines were screened by ELISA for binding to orthopoxvirus antigens. Hybridomas secreting human antigen-specific mAbs were generated from B cell lines secreting virus-specific antibodies, as previously described (Crowe, 2009). Because orthopoxviruses have so many protein components, we used complementary screening approaches to identify orthopoxvirus-specific mAbs. In one approach, we assessed hybridoma cell line supernatants for reactivity to each of 12 recombinant VACV protein antigens designated A21, A27, A28, A33, B5, D8, F9, J5, H2, H3, L1, and L5. The A33 and B5 proteins are surface antigens on the EV form of virus, while the remaining ten proteins are surface antigens on MV particles. In addition, we also screened for mAbs that bind to inactivated lysates of VACV-, CPXV-, or MPXV-infected cell monolayer cultures.

A total of 89 cloned hybridoma cell lines secreting human mAbs was isolated, including 44 lines from vaccinees and 45 from the donor with a history of MPXV infection (Table S1). The 89 mAbs were independent clones that displayed a high degree of sequence diversity, including a unique HCDR3 sequence for each mAb (Table S2). Thirty-two mAbs in the panel bound in ELISA to inactivated VACV-infected cell lysates only and thus their protein antigen specificity was uncertain initially. Binding of these mAbs was reassessed using VACV protein antigen microarrays, which revealed additional mAbs specific to D8, H3 A21, A25, H5, and I1 VACV proteins. Therefore, the mAb panel contained Abs to at least 12 antigens: D8, B5, A33, H3, L1, A27, I1, A25, F9, A28, A21, and H5 (Figure 1 A). The majority (62 of 89 [70%]) of purified mAbs reacted to one of six VACV antigens that were reported previously as major targets for neutralizing Abs in mice or humans (Moss, 2011), specifically A27, H3, D8, L1, B5, and A33. Sixteen percent (14 of 89) of mAbs in the panel reacted with VACV-infected cell lysate but not with a recombinant protein antigen, therefore they remain of unknown specificity (Figure 1A). MAbs that targeted the antigens VACV D8 and B5 were over-represented in the panel (35 of 75 mAbs) accounting for 47% of mAbs with known antigen specificity. Further analysis revealed several competition-binding groups among Ab specificities that bind to H3 or D8 antigens (Figure S1 ), indicating the presence of mAbs to several antigenic sites on these antigens.

We next assessed the cross-reactivity of individual VACV-reactive mAbs to CPXV, MPXV, or VARV by testing binding to CPXV-, MPXV-, or VARV-infected cell lysates or to recombinant VARV protein antigens that are orthologs of the identified VACV targets. A large fraction (45 of 73 [62%]) of mAbs that bound to VACV antigens in virus-infected cell lysate (Figure 1B) bound in a cross-reactive manner to the virus-infected lysates of all four Orthopoxvirus species tested, and the majority (70 of 73 [96%]) of mAbs cross-reacted with at least two orthopoxviruses (Figure 1C; Table S3). Remarkably, a large fraction (47 of 71 tested [66%]) of the mAbs with an established protein antigen specificity for VACV cross-reacted with the orthologous VARV antigens (Figure 1D; Table S3). The mAbs bound to recombinant antigens and/or infected cell lysates in a concentration-dependent manner (Figure S2 and data not shown), and the majority of them possessed half maximal effective concentration (EC₅₀) binding values of 1 μg/mL or lower, confirming their antigen-specific binding phenotype (Table S4). Therefore, the majority of mAbs in the panel exhibited binding patterns that suggested the potential to neutralize several Orthopoxvirus species that are infectious for humans.

The Majority of Human Neutralizing mAbs Recognized One of Six Antigens and Exhibited Cross-Neutralization for Several Orthopoxvirus Species

We next tested the mAbs in virus neutralization assays using MV or EV forms of VACV, CPXV, or MPXV. Neutralization potency of mAbs was assessed based on the half maximal inhibitory concentration (IC₅₀) and the maximum of neutralization effect (E_max) values. More than half (48 of 89 [54%]) of the mAbs possessed neutralizing activity (E_max ≥ 50%) at 100 μg/mL or lower concentration for at least one orthopoxvirus; 16 or 32 mAbs neutralized the EV or MV form of VACV, respectively (Figure 2 A). Of note, neutralizing activity for the majority of these Abs required complement (Table S5). Most (46 of 48 [98%]) of the neutralizing mAbs recognized one of six proteins, D8, L1, B5 A33, A27, or H3 (Figure 2B). Two remaining mAbs were from the subject with prior wild-type MPXV infection and recognized I1 or an undetermined MPXV antigen (Table S5).

A majority (38 of 48 [79%]) of neutralizing mAbs cross-neutralized at least two Orthopoxvirus species (mainly VACV and CPXV) and 12 of 48 (25%) mAbs neutralized three orthopoxviruses—VACV, CPXV, and MPXV (Figure 2C). Regardless of their antigen specificity, the neutralizing mAbs varied widely in their neutralization potency. IC₅₀ values of individual mAbs ranged from ∼0.02 to 100 μg/mL, and E_max values varied from 50% (the designated cut-off threshold to identify potent neutralizing clones) to 99.5% (Figure 2D; Table S5). Most of the neutralizing mAbs reduced plaque number only by ∼60%–80% at the highest tested concentration, regardless of antigen or form of virus targeted. The neutralizing activity of MV-targeted anti-D8, L1, A27, or H3, and EV-targeted anti-B5 mAbs were similar for VACV and CPXV, and two of six VACV and MPXV-neutralizing anti-A33 mAbs neutralized CPXV. None of the anti-D8 or anti-B5 mAbs neutralized MPXV, despite the ability of these mAbs to bind the corresponding MPXV ortholog protein (Tables S4 and S5). In contrast, the broadest cross-neutralizing activity (neutralization of VACV, CPXV, and MPXV), was detected in mAbs directed to A33, L1, A27, or H3 antigens (Figure 2D). Cross-neutralizing mAbs were isolated from most orthopoxvirus-immune subjects (Table S5). Together, our data indicate that mAbs induced by VACV immunization or MPXV infection that recognize any of six neutralizing determinants can inhibit several Orthopoxvirus species and also suggest that the broadest cross-neutralization is mediated predominantly by four Ab specificities.

Mixtures of Diverse mAb Specificities Possess Superior Cross-Neutralizing Activity for VACV, CPXV, MPXV, and VARV

We next designed two mixtures of mAbs, designated Mix6 and Mix4, containing diverse specificities with high neutralizing (low IC₅₀ and high E_max values) and cross-neutralizing activities to both MV and EV forms of virus (Figure 2D; Table S5). Mix6 contained single neutralizing mAbs directed to each of six antigens that targeted by neutralizing mAbs—MV proteins D8, A27, H3, and L1, and EV proteins B5 and A33. Mix4 was similar to Mix6, containing mAbs to A27, L1, B5, and A33, but lacked anti-D8 and H3 mAbs (Table S6) that were found to be non-protective in the lower respiratory tract infection model (detailed below). Both mixtures included four mAbs that exhibited similar binding for VACV proteins and the corresponding VARV protein orthologs (Figure S2). The neutralizing activity of Mix6 and Mix4 for VACV were higher than that of individual mAbs or VIGIV (Figure 3 A). Moreover, Mix6 and Mix4 cross-neutralized VACV, CPXV, MPXV, and VARV more potently than did VIGIV in EV and MV neutralization assays (Figures 3B and S3 ; VARV could only be tested in the MV assay, without complement). Therefore, neutralization and cross-neutralization are more efficiently achieved with mixtures of diverse mAbs specificities than with individual potently neutralizing mAbs.

Superior In Vivo Protection against VACV Infection Was Achieved by Administration of a Mixture of Human mAbs that Targeted Multiple Viral Antigens

We next evaluated the protective capacity of Mix6. Single-dose treatment with Mix6 one day before lethal intranasal (IN) challenge of C57BL/6 mice with VACV provided complete protection against weight loss and mortality (Figure 4 A). Mock-treated mice experienced severe illness and succumbed by day 7 post-inoculation (p.i.). The protection was associated with a profound (∼10⁶-fold) reduction of viral load in the lungs on day 7 p.i., when compared to the mock-treated group (Figure 4B). Notably, the level of protection provided by Mix6 was comparable to, if not higher than, that provided by prior immunization with a sub-lethal dose of VACV. In contrast, pre-treatment with VIGIV did not protect mice under the challenge conditions used. These mice were unable to control VACV replication in the lungs and succumbed by day 7 p.i., similarly to the mock-treated group (Figure 4B). These data indicate a high prophylactic potency of Mix6 for prevention of respiratory tract infection.

To further characterize the protective efficacy of Mix6, we tested it in a lethal model of systemic VACV dissemination using severe combined immunodeficiency (SCID) mice that lack adaptive immune responses but retain a functional complement system (Bosma and Carroll, 1991). Initially, we assessed the prophylactic effect of Mix6 given to mice by the intraperitoneal (i.p.) route 1 day prior to lethal i.p. virus challenge. Remarkably, single-dose pre-treatment with Mix6 provided sterilizing immunity in this model (Figure 4C). Mice pre-treated with a human mAb of irrelevant specificity succumbed to the disease by day 20 p.i., when the group of animals pre-treated with Mix6 was completely protected from death and any signs of disease. Clearance of human mAbs from animal blood rendered healthy mice susceptible to VACV re-infection (Figures 4C and 4D), demonstrating that the sterilizing immunity observed during primary VACV infection was mediated solely by the administered Mix6.

In summary, these findings demonstrate the high prophylactic potency of Mix6 for prevention of respiratory and systemically disseminated VACV infections.

Four Principal Antibody Specificities Participated in Protection against Respiratory VACV Challenge when Used in Mixture

We next determined the contribution of individual mAbs within Mix6 by assessing the protective capacity of single mAbs or their mixtures (Table S6). Both of the EV-targeted antibodies, anti-A33 and -B5, protected B6 mice from death and severe weight loss when administered alone or as a mixture (Mix6(ΔMV)) 1 day before IN VACV challenge. In contrast, none of the MV-targeted mAbs, or their mixture (Mix6(ΔEV)), protected mice in the same conditions (Figures 5 A and S4 ). A possible explanation for this result was that the VACV challenge conditions used in this model are quite stringent and likely do not allow detection of moderate levels of protection by some mAbs. The other possibility was that the selected mAb clones may bind to non-protective epitopes of their antigens. To investigate further, we assessed protection using a less severe upper airways infection mouse model (Figure S5 A). These conditions resulted in milder disease and less mortality. In addition, anti-D8 and H3 mAbs were tested as mixtures of five or three different epitope specificities that incorporated mAbs from different competition-binding groups for D8 or H3 antigens, respectively. These single antigen-specific mixtures thus recognized diverse epitopes in D8 or H3 antigen. In this less stringent challenge setting, anti-A27 mAbs prevented mortality and severe weight loss, showing these mAbs may contribute to the protective efficiency by Mix6. Anti-L1 mAbs and mixtures of anti-D8 or anti-H3 mAbs still were not protective (Figure S5A). Therefore, these monotherapy studies suggested three protective human mAbs specificities in this model—anti-B5, anti-A33, and anti-A27.

It was possible that some of the six Ab specificities contributed to protection in mixtures only in a cooperative manner that would not be detected by monotherapy studies. To detect such activity, we designed mixtures that were variants of the Mix6 that each lacked one mAb specificity (Table S6). Each of the Mix6 variant mixtures lacking one of the mAbs was protective, although mixtures lacking anti-L1, anti-A27, anti-A33, and anti-B5 were less efficient in protection against weight loss than Mix6 ( Figure 5B). Removal of the protective MV-targeted anti-A27 mAb from Mix6 did not affect the outcome of challenge substantially. However, exclusion of the MV-targeted anti-L1 mAb from Mix6 resulted in detectable weight loss upon infection, which was comparable to that seen when mice were pre-treated with Mix6 lacking either of the most potent EV-targeted mAbs (anti-A33 or anti-B5) identified in the monotherapy studies (Figures 5A and 5B). Moreover, Mix4 containing anti-L1, anti-A27, anti-B5, and anti-A33 mAbs conferred a level of protection equivalent to that of Mix6 (Figure 5C). Therefore, the mAbs in Mix4 appear to cooperate in achieving their protective effect.

One possible explanation for the diminished protection observed when a mAb mixture lacked a single MV- or EV-targeted mAb specificity was the decrease in total amount of mAb per treatment. Therefore, we next examined whether the lack of one mAb specificity in protective Mix4 could be compensated by using a mixture containing the same total amount of Ab by adding an equivalent amount of one of the retained mAb specificities targeting the same virion form. The monotherapy results suggested higher potency of anti-A33 and anti-A27 mAbs when compared to the EV- or MV-specific anti-B5 and anti-A27 mAbs (Figures 5A and S5A). Therefore, groups of mice were treated before VACV challenge with a variant of Mix4 that contained a 2-fold higher amount of the anti-A27 or anti-A33 mAb and lacked anti-L1 (designated as Mix4(ΔL1)), or anti-B5 (designated as Mix4(ΔB5)) mAb, respectively. An excess of anti-A27 or anti-A33 mAb did not restore the initial activity of Mix4 in absence of mAbs with anti-L1 or anti-B5 specificities, although the effect was minor under the challenge conditions used (Figure 5C). However, in more stringent challenge conditions, mice pre-treated with Mix4 exhibited significantly higher resistance to the disease and recovered faster compared to mice that received the Mix4 (ΔB5) containing a 2-fold higher amount of anti-A33 mAb (Figure S5B). This finding suggested Mix4 as a potent therapeutic mixture. Together, these results showed that four principal mAb specificities in Mix4 contributed to, and were required for, efficient protection against lethal respiratory tract VACV infection in the mouse model.

Therapeutic Effect of Mix6 when Given Up to 3 Days after Infection by the Respiratory Route

We next determined how long after respiratory infection Mix6 would exhibit a therapeutic effect, when treatment was delayed. For these studies, mice were immunized passively with Mix6 one day before or on the day of virus challenge, or 1, 2 or 3 days after virus challenge (Figure 6 ). As expected, the treatment was most efficient when administered before disease onset. Mice given Mix6 showed significant protection from weight loss if the treatment was given 1 day before, on the day of challenge or 1 day after infection. When the treatment was delayed until day 3, the time point when untreated animals developed disease due to profound virus burden in the lungs (data not shown), we observed protection from death, but only partial protection from weight loss (Figure 6). These data demonstrated that Mix6 mediated a therapeutic effect even when treatment was delayed, especially against lethality.

Diverse Human Ab Specificities Participate in Protection against Systemic VACV Infection

The experiments described above showed that a single dose of Mix6 given prior to systemic inoculation with a lethal dose of VACV conferred sterilizing protective immunity in SCID mice, which lack adaptive immunity (Figure 4C). Using this model, we next assessed the efficacy of monotherapy with individual mAbs, Mix6, or VIGIV that were given to mice 1 day after inoculation with a lethal dose of VACV (Figure 7 ). Similar to the respiratory challenge study, anti-H3 or anti-D8 mAbs were used as mixtures of several epitope specificities. Each of the Abs tested, including those identified as non-protective in respiratory tract infection, delayed morbidity and mortality in mice when compared to the animals in the mock-treated group. Moreover, delayed treatment with Mix6 conferred sterilizing immunity to inoculated mice, which all survived and lacked signs of illness for >155 days after VACV inoculation (Figures 7A and 7B). Together, these results demonstrated a high therapeutic potency of Mix6 and showed that diverse mAbs specificities may contribute to protection against systemically disseminated VACV infection.

Limitations, Caveats, and Open Questions

Using naturally occurring human mAbs isolated by hybridoma technology, this study revealed six principal cross-neutralizing human mAb specificities for VACV, CPXV, MPXV, and VARV. We next showed that these Ab specificities that are necessary and sufficient determinants of protection in murine challenge models. This work suggests that a mixture of these Abs could mediate cross-protective immunity to orthopoxviruses. As with most studies, there are several limitations of this work that we would like to point out.

The antibody discovery platform used likely allowed us to identify mAbs only from the most frequent classes of B cell memory clones that occur in human peripheral blood. Therefore, less frequent clones could be missing from our analysis.

• It remains unknown to what extent the B cell memory repertoire in the blood that we have studied corresponds to the antigen-reactive antibody protein repertoire in the serum that is secreted by long-lived plasma cells in the bone marrow. Future proteomics studies using emerging technologies might be able to address this question.

• Future development for use in humans of individual mAbs or mixtures described here against VACV, CPXV, and MPXV or VARV should include studies of larger animal models, such as non-human primates.

Key Resources Table

Contact for Reagent and Resource Sharing

Further information and requests for reagents may be directed to, and be fulfilled by the corresponding author: James E. Crowe, Jr. (james.crowe@vanderbilt.edu). Materials described in this paper are available for distribution under the Uniform Biological Material Transfer Agreement, a master agreement that was developed by the NIH to simplify transfers of biological research materials.

Experimental Model and Subject Details

Method Details

Quantification and Statistical Analysis

The descriptive statistics mean ± SEM or mean ± SD were determined for continuous variables as noted. Comparisons were performed using Wilcoxon rank sum test or the post hoc group comparisons in ANOVA; all tests were two-tailed and unpaired. Survival curves were estimated using the Kaplan Meier method and curves compared using the log rank test with subjects right censored, if they survived until the end of the study. ^∗ -p < 0.05; ^∗∗ - was used to reject a “null hypothesis.” ^∗ = p < 0.05; ^∗∗ = p < 0.01; ^∗∗∗ - = p < 0.001; ns – non-significant. Statistical analyses were performed using Prism v5.0 (GraphPad).