Screening mAbs against ebolavirus surface glycoprotein on SP-IRIS
To evaluate the performance of the mAbs against EBOV and SUDV GPs on SP-IRIS, we spotted antibodies on three SP-IRIS chips with 3 replicate spots for each antibody and performed virus detection experiments by incubating each chip with a different VSV pseudotype: rVSV-EBOV, rVSV-SUDV, and rVSV-LASV (for specificity test). Our antibody screening test included the ingredients of all three cocktails mentioned in the previous section, 13F6, 13C6, 6D8, 2G4, 4G7, 1H3, as well as additional antibodies: KZ52, 15H10, 3F10, 16F6, 5G10, 16H11, 19B3, and 19B4.
According to the binding data obtained from SP-IRIS for the rVSV-EBOV chip (Fig. 2a), the highest amount of virus binding was observed for 13F6, 13C6, 6D8 and 1H3 antibodies, 13F6 antibody giving the highest signal. 13F6 and 6D8 antibodies bind to the mucin-like domain (MLD), and 13C6 and 1H3 antibodies bind to the glycan cap (Fig. 3). Both of these regions are located in the outer region of the glycoprotein, therefore, our results can be explained by the easy accessibility of these domains. 2G4, 4G7 and KZ52 antibodies, all of which bind to the base domain of the GP, showed lower signals than the MLD and glycan cap binding mAbs. This may be due to the fact that epitopes of these mAbs are harder to access compared to more exposed MLD and glycan domains. Furthermore, these epitopes might be prevented from mAb binding due to the glycan coating over the GP.
KZ52 is a neutralizing mAb that was isolated from a patient who recovered from EBOV infection in the 1995 Zaire EBOV outbreak. This antibody was shown to bind a non-glycosylated region at the base domain of the GP. KZ52 blocks the cleavage of membrane-associated GP by host cell cathepsins, which is a critical step for virus entry into the cell.29 The cathepsin cleavage results in formation of a fusion-active GP that is capable of mediating membrane fusion. Despite showing similar GP binding affinity to 13C6 in ELISA and surface plasmon resonance (SPR) measurements performed with purified antigen in a previous work,30 the capture efficiency of KZ52 was much lower than that of 13C6 on SP-IRIS. The low binding level seen on SP-IRIS with this antibody is most likely due to the decreased accessibility of the base domain of membrane-associated GP combined with the limited flexibility of the surface attached antibodies. KZ52 antibody binds at a parallel angle to the base domain (Fig. 3) and this might pose a sterically hindered orientation for an antibody attached to the surface via multiple bonds.
Figure 2b shows the captured virus densities for the mAbs that showed either very little or no signal on SP-IRIS, as well as the negative antibody, 8G5, specific for wild-type VSV, on a smaller virus density scale to show the signal levels for these antibodies clearly. 15H10 and 3F10 produced very little signal on SP-IRIS whereas there was no significant detection with 16F6 antibody. The detection threshold which is calculated as the mean particle density captured on the 8G5 spots plus three standard deviations, is about 2,000 particles/mm2. According to data provided by USAMRIID (Supplementary Table S1), 3F10 and 16F6 mAbs showed a positive signal in antigen ELISA for EBOV glycoprotein. The fact that these mAbs worked well in an antigen ELISA but did not produce a strong signal for EBOV GP pseudotyped VSV on SP-IRIS platform can be explained by the differences in the accessibility of the specific epitope in the monomeric GP and the trimeric membrane-associated GP. Perhaps the epitopes for these mAbs are hindered from binding due to the organization of the monomers in the trimeric structure. Neither of these mAbs showed binding to the whole virus in flow cytometry (for Zaire species), which is also consistent with the very little signal or lack of signal on SP-IRIS (Supplementary Table S1).
15H10 mAb is a non-neutralizing antibody that binds to a site in the canonical heptad repeat 2 (HR2) region near the membrane-proximal external region (MPER) of the glycoprotein (Fig. 3).31,32 This antibody has a binding region that is closest to the viral membrane among all the antibodies tested, explaining the low binding level observed on SP-IRIS. Sudan specific antibodies (5G10, 16H11, 19B3, and 19B4) did not show any binding for the chip incubated with rVSV-EBOV (not shown in the graph), as expected, with particles densities well below the threshold.
rVSV-SUDV chip showed binding for 3F10, 5G10, and 16F6 antibodies, given in order of decreasing signal. (Supplementary Fig. S2). Other SUDV specific antibodies, 16H11, 19B3 and 19B4, did not bind to membrane-associated GP of SUDV, although they showed binding in ELISA (Supplementary Table S1). We reason that either the epitopes of these antibodies are not exposed in the GP associated with the viral membrane or the binding angles of the antibodies pose a constraint on the surface-immobilized antibodies. According to our results, 3F10 was the only mAb that recognized both EBOV and SUDV GP in a whole virus model. All three antibodies that were able to bind to rVSV-SUDV on SP-IRIS, 3F10, 5G10, and 16F6, can be used for SUDV detection in whole virus immunoassays, whereas 5G10 would be the most ideal capture antibody to differentiate between Zaire and Sudan species of ebolavirus. Our results demonstrate that SP-IRIS provides a sensitive and fast platform for screening mAbs for different viruses and also different species of a given virus to identify species-specific antibodies as well cross-reacting mAbs. We were able to test 16 different mAbs on a single chip, revealing their binding specificity and relative capture efficiencies. The third chip that was incubated with rVSV-LASV did not show any signal for any of the EBOV and SUDV-specific antibodies whereas there was significant binding for LASV GP specific 8.9F antibody. (Supplementary Fig. S3)
A model for EBOV GP epitope mapping based on SP-IRIS data Based on our results presented in Table 1, we can divide the Ebola virus antibodies (Zaire-specific) that we tested on SP-IRIS platform into three distinct groups (Fig. 3). Group 1 antibodies form the strong binding group and members of this group bind to either the MLD (13F6 and 6D8) or the glycan cap of the GP (13C6 and 1H3). Group 2 antibodies (4G7, 2G4, and KZ52), that show intermediate level of signal on SP-IRIS, bind to the G1/G2 interface (base domain) of GP. The final group, Group 3, is composed of mAbs that show very little or no binding on SP-IRIS. Among these, 15H10 binds to an antigenic site on GP2, HR2/MPER region, which is very close to viral membrane. 16F6 binds to GP1/GP2 area like Group 2 antibodies, however with a steeper angle and closer proximity towards the viral membrane. 3F10 mAb epitope is not known, however, based on low binding level observed on SP-IRIS, we can hypothesize that it binds to a region close to viral membrane, either to GP1/GP2 or MPER region. Overall, Group 3 antibodies, bind to less accessible areas of GP such as MPER or with angles that would be hard to achieve with surface immobilized antibodies due to steric hindrance and limited flexibility.
Table 1
mAbs against EBOV GP, their relative signal on SP-IRIS, binding domains on the GP, and neutralization activities
Anti-EBOV GP Ab
|
Signal on SP-IRIS
|
Binding Domain
|
Neutralization activity
|
13F6
|
++++
|
Mucin-like domain
|
Non-neutralizing
|
13C6
|
+++
|
Glycan cap
|
Non-neutralizing
|
6D8
|
+++
|
Mucin-like domain
|
Non-neutralizing
|
1H3
|
+++
|
Glycan cap
|
Non-neutralizing
|
4G7
|
++
|
GP1/GP2 (base)
|
Neutralizing
|
2G4
|
++
|
GP1/GP2 (base)
|
Neutralizing
|
KZ52
|
++
|
GP1/GP2 (base)
|
Neutralizing
|
15H10
|
+
|
GP2 (HR2/MPER)
|
Non-neutralizing
|
3F10
|
+
|
Unknown
|
Unknown
|
16F6
|
-
|
GP1/GP2 (base)
|
Neutralizing*
|
The degree of binding is denoted by +/- symbols: ++++ very high, +++ high, ++ medium, + little, - no binding. *Neutralization test for 16F6 was performed by USAMRIID.
Binding domain information also makes it possible to suggest a mechanism of action for a given antibody. For example, if an EBOV antibody is binding to MLD or the glycan cap, this may provide an evidence for the fact that it is a non-neutralizing antibody.23 Since the mucin-like and glycan domains are cleaved from the GP before host cell membrane fusion, these regions are not associated with neutralization process. Shedlock et al. showed that the cleavage of glycan cap occurred normally in the presence of mucin-like domain binding antibody, 13F6, showing that this antibody does not interfere with the cleavage process.29 Their work also demonstrated that 13C6 antibody delays the cleavage process, which might contribute to its protective activity. The fact that these antibodies provide protection in vivo can be result of a different mechanism such as antibody dependent cytotoxicity (ADCC). In ADCC, viral particles with GP bound antibodies are recognized by the Fc receptors on the killer cells of the immune system and destroyed. On the other hand, if an antibody binds to the base domain, it is most likely that it has neutralization activity. Neutralization activity can be realized through several different mechanism including prevention of membrane fusion, blocking the attachment to the host cell and aggregation of the viruses. Therefore, based on the model presented in the table in Fig. 3, it is possible to predict the binding domains of the newly identified antibodies as well as their mechanism of action using whole virus models and SP-IRIS platform.
Screening mAbs against Lassa virus on SP-IRIS
To evaluate the whole virus capture efficiency of 16 mAbs against Lassa virus glycoprotein, we spotted the following antibodies on two SP-IRIS chips: 8.9 F, 12.1 F, 10.4 B, 25.10 C, 19.7 E, 37.7 H, 13.4E, 2.4F, 37.2D, 36.1 F, 19.5 A, 4.1F, 18.5 C, 9.8 A, 37.2 G, and 24.6 C, with 8 replicate spots for each antibody as shown in Fig. 4a. We incubated both chips with 107 PFU/mL rVSV-LASV for 1 hour. The average virus densities for each LASV antibody obtained from SP-IRIS images are shown in Fig. 4b. Only 8 of the antibodies that showed significant binding on SP-IRIS are demonstrated in the bar graph. The antibodies that showed signal on SP-IRIS are 8.9 F, 12.1 F, 10.4 B, 25.10 C, 19.7 E, 37.7 H, 13.4 E, and 2.4F. The antibodies that did not show any signal on SP-IRIS are 37.2D, 36.1 F, 19.5 A, 4.1F, 18.5 C, 9.8 A, 37.2 G, 24.6 C. Since 8.9 F showed the highest amount of binding among these, we selected this antibody for the specificity test and limit-of-detection (LOD) experiment for the detection of rVSV-LASV.
To evaluate the specificity of the 8.9F mAb, we incubated two different SP-IRIS chips with different VSV psedotypes: rVSV-LASV and rVSV-EBOV. According to our specificity test results, 8.9F antibody binds to rVSV-LASV whereas it does not show a significant signal for rVSV-EBOV, demonstrating its ability to recognize LASV GP specifically (Fig. 5a). To evaluate the sensitivity of the rVSV-LASV detection in complex media using this antibody, we performed a dilution series experiment, where we incubated SP-IRIS chips with rVSV-LASV with 10-fold dilutions from a 107 PFU/mL stock, prepared in FBS, at the following concentrations: 103, 104, 105 and 106 PFU/mL. Figure 5b shows that the lowest concentration that was detected by SP-IRIS is 105 PFU/mL and LOD is between 104-105 PFU/mL. Given that the specificity experiment that used a 5 x 104 PFU/mL rVSV-LASV showed significant detection (Fig. 5a), we can conclude that the actual LOD is between 104 – 5 x 104 PFU/mL. Red dashed line in Fig. 5b shows the detection threshold obtained from 8.9F spots on a blank chip incubated with FBS alone. It is calculated as the mean of 6 spots plus three times the standard deviation.
Next, we wanted to see if there is a similar correlation between the level of SP-IRIS signal and the binding epitopes on the LASV GP as we observed with EBOV antibodies. We summarized the LASV antibodies tested on SP-IRIS in Table 2 that shows the SP-IRIS signal levels as well as the other information about the Lassa virus antibodies including antigen ELISA data, binding domains of the antibodies, and the neutralization activities, obtained from a previously published study.33 Fig. 6 shows a model of LASV GP structure adapted from the same study. LASV GP is composed of a glycoprotein complex (GPC) that has two subunits: GP1 and GP2. The 8.9F mAb that showed the highest capture efficiency on SP-IRIS, binds to GPC-C domain that involves both GP units. Among the 16 mAbs tested, 8.9F is the only antibody that binds to GPC-C domain. The fact that 8.9F does not recognize solubilized GP in an ELISA assay (Table 2) suggests that it is reactive against a conformational epitope formed in the membrane-associated GP. All of the other tested LASV antibodies showed signal on ELISA.
Table 2
The mAbs against LASV GP tested on SP-IRIS, observed SP-IRIS signal levels, signal on ELISA, binding domains on GP, and neutralization activities
Anti-LASV Ab
|
SP-IRIS
|
ELISA*
|
Binding Domain*
|
Neutralization activity*
|
8.9F
|
+++
|
Negative
|
GPC-C
|
Neutralizing
|
12.1F
|
++
|
Positive
|
GP1-A
|
Neutralizing
|
10.4B
|
++
|
Positive
|
GP1-A
|
Weak neutralizing
|
25.10C
|
++
|
Positive
|
GPC-A
|
Neutralizing
|
19.7E
|
+
|
Positive
|
GP1-A
|
Neutralizing
|
37.7H
|
+
|
Positive
|
GPC-B
|
Neutralizing
|
13.4E
|
+
|
Positive
|
GP2-L2
|
Non-neutralizing
|
2.4F
|
+
|
Positive
|
GP1-B
|
Non-neutralizing
|
37.2D
|
-
|
Positive
|
GPC-B
|
Neutralizing
|
36.1F
|
-
|
Positive
|
GPC-A
|
Neutralizing
|
4.1 F
|
-
|
Positive
|
GP2-L1
|
Non-neutralizing
|
18.5C
|
-
|
Positive
|
GPC-B
|
Weak neutralizing
|
9.8A
|
-
|
Positive
|
GPC-B
|
Weak neutralizing
|
37.2G
|
-
|
Positive
|
GPC-B
|
Weak neutralizing
|
24.6C
|
-
|
Positive
|
GP2-L3
|
Non-neutralizing
|
The degree of binding is denoted by the number of + symbols: +++ high, ++ medium, + little, - no binding. *Information about the antibodies obtained from [33]. 19.5A is not included in the table since there is no information on the binding domain or neutralization activity.
The table in Fig. 6 groups the LASV antibodies tested in this study according to the signal levels obtained from SP-IRIS and summarizes the correlation between the binding levels and associated epitopes. Group 1 represents the strong binding antibodies and has only one antibody, 8.9F. Group 2 is composed of antibodies that show intermediate signal level on SP-IRIS. The majority of the Group 2 antibodies are the ones that bind to GP1-A domain. This is expected since GP1 subunit includes the receptor binding site and also has less glycosylation, making it more accessible for antibody binding. Group 3 is formed by the antibodies that either showed little or no signal on SP-IRIS. Majority of the Group 3 antibodies bind either to GPC-B or GP2 linear region (GP2-L1 and GP2-L3). These areas most likely involve the areas of GP2 that are not easily accessible by the antibodies. These antibodies have either a weak-neutralizing activity or they don’t have neutralizing activity at all with the exception of 37.2D and 36.1F antibodies (Table 2). Although it is not possible to make a definite remark for a strong correlation between the SP-IRIS signal and neutralizing activity due to these exceptions, SP-IRIS can still provide a sensitive and high-throughput platform to screen diagnostic and therapeutic antibody candidates to find the antibodies with the best capture efficiency and highest potential to be effective against the LASV infection. Moreover, identifying more mAbs recognizing the receptor binding site on GP1 subunit which is less glycosylated would help improve the sensitivity of LASV detection. These antibodies can be easily determined on SP-IRIS since they would show high level of binding. According to our results, Group 1 and Group 2 antibodies would be the best candidates to choose as potential therapeutic antibodies against LASV infection.
Antibody competition assay for Ebola virus specific antibodies
As we mentioned earlier, ZMapp antibody cocktail has been shown to provide increased efficacy compared to the two previously evaluated MB-003 and ZMAb cocktails. ZMapp and ZMAb cocktails have 2G4 and 4G7 antibodies in common and they differ in the third component: ZMAb has 1H3, while ZMapp has 13C6. Therefore, one can conclude that 13C6 has some advantage compared to 1H3 in providing efficient protection. Both of these antibodies bind to the glycan cap and compete with each other. 13C6 binds at a perpendicular angel to the viral membrane whereas 1H3 binds at a less steep angle (Fig. 3).
To explain the superiority of 13C6 antibody over 1H3, we made the following hypothesis: 1H3 antibody hinders binding of either one or both of the base domain binding constituents of the ZMAb cocktail, or vice versa, i.e, binding of 2G4 and/or 4G7 antibodies inhibits the binding of 1H3, preventing it from realizing its function of triggering host immune response. Murin et al. reported competition assay results for EBOV antibodies using ForteBio Octet platform.23 Competition assay data presented in this work does not reveal any significant differences between 1H3 and 13C6 mAbs in terms of competition with the 2G4 and 4G7 antibodies. According to their data, binding of 4G7 decreases the binding of 1H3 and 13C6 to 41% and 42% of their binding without competition, respectively. Similarly, 2G4 decreases the binding of 1H3 and 13C6 to 48% and 44% of their uncompeted binding, respectively. When antibodies are introduced in the reverse order, both 13C6 and 1H3 affect binding of 2G4 and 4G7 similarly; 13C6 decreases binding of 2G4 and 4G7 to 62% and 51% of their uncompeted binding, respectively, and 1H3 antibody decreases their binding to 74% and 70% of their uncompeted binding, respectively. In fact, their data indicates that 13C6 interferes with 2G4 and 4G7 binding more than 1H3 does.
We developed a novel competition assay using SP-IRIS platform and pseudoviruses with membrane-associated GP to understand the interaction of the antibodies with the GP in its native state (Fig. 7). We spotted anti-EBOV GP antibodies (13F6, 13C6, 6D8, 1H3, 4G7, 2G4 and KZ52) on 8 SP-IRIS chips with 4 replicate spots per antibody. We first mixed 5 × 105 PFU/mL EBOV-pseudotyped VSV with each of the seven antibodies (at a final anti- body concentration of 1 µM) in separate microtubes and allowed the binding of the antibodies to the rVSV-EBOV particles for 20 minutes. Then, we incubated seven separate SP-IRIS chips with these mixtures for 1 h. We also incubated one SP-IRIS chip with the virus sample only as the reference chip to calculate the uncompeted binding level of the antibodies. We calculated the average virus densities for each antibody on the surface of each chip and calculated the percent binding for each antibody by comparing the binding on a given competition chip to the uncompeted binding from the reference chip. These percent binding values are given in Fig. 7.
Our results indicate that pre-decoration of virus particles with 4G7 and 2G4 antibodies reduces the binding of 1H3 significantly (to 23% and 41% of its binding alone, respectively). On the other hand, binding of 4G7 or 2G4 to the membrane-associated GP does not interfere with 13C6 antibody binding to the GP, with 73% and 92% competition binding percentages. Based on our results, we predict that binding of 2G4 and 4G7 antibodies causes a conformational change in the region where 1H3 binds, preventing the binding of 1H3, whereas they do not affect binding of 13C6. Therefore, it is possible that, when administered together, 2G4 and 4G7 antibodies will prevent binding of 1H3 to the glycan cap which will hamper the appropriate immune response and decrease the efficacy of treatment. Importantly, when antibody order is reversed, we observed that 13C6 enhances the binding of both 4G7 and 2G4 significantly, with 133% and 128% binding percentages, respectively, whereas 1H3 improves the binding of 2G4 only while reducing the 4G7 binding to 57%. When combined together, our results provide a possible explanation as to why 13C6 antibody in ZMapp cocktail is superior to 1H3 antibody that is in the ZMAb cocktail. 13C6 mAb enhances the binding of other two antibodies in the cocktail, 2G4 and 4G7, and its own binding is not affected significantly by them, rendering this cocktail more efficient than the 1H3, 2G4, and 4G7 mixture. These results suggest that competition between the antibodies in a cocktail has important implications for the efficacy of the treatment, and therefore should be carefully evaluated while designing mAb cocktails. Table 3 classifies the Ebola antibodies evaluated in the competition assay into different competition groups based on the percent binding levels on the competition chips. We suggest that mAbs in a given antibody cocktail that is being developed as an antiviral agent should be tested pairwise to see the effect of each mAb on the binding of the others. The mAbs that are from Group 3 and/or Group 4 would be ideal candidates for further evaluation in clinical trials.
Table 3
Classification of Ebola antibodies in the competition assay into competition groups based on their percent binding levels
Competition group
|
Binding percentage
|
Classification
|
Group 1 (black)
|
< 50%
|
Competing
|
Group 2 (dark gray)
|
50–75%
|
Intermediate competition
|
Group 3 (white)
|
75–110%
|
Non-competitive
|
Group 4 (light gray)
|
> 110%
|
Enhancing
|
The fact that our competition assay results are different than the ones obtained by Murin et al. (2014), can be explained by the major differences in the two competition assay designs. First of all, in the mentioned study, an ensemble-based method is used where surface immobilized soluble GP is reacted with the competing antibody pair successively. Ensemble-based techniques have limited sensitivities since they cannot resolve single binding events. On the other hand, SP-IRIS is a highly sensitive platform with single-particle visualization ability, making the competition assay more sensitive. Second, binding of the first antibody to the surface immobilized GP would decrease the binding of second antibody substantially, since GP is already sterically hindered by the surface attachment as well as the antibody 1 coverage. This may also explain why the enhancement affect is not observed in the cited work and why 13C6 binding is decreased substantially by 4G7 and 2G4 binding. Moreover, the soluble version of the glycoprotein would have different binding properties compared to the membrane-bound native glycoprotein used in our assay and, therefore, might not reveal the competition/enhancement events that are relevant from a therapy stand of point.