scFv were designed and cloned previously as described (38). In short, single constructs containing the variable heavy and light chain interspaced with a 15 or 18 amino acid glycine-serine linker of five HIV-directed bNAbs (CAP256.25, PGT121, 3BNC117, 8ANC195, and 10E8v4) were generated through overlapping PCR or ordered from GenScript (New Jersey, USA) (38). These scFv genes were then cloned into a CMV/R expression plasmid (AIDS Reagent Program, Division of AIDS, NIAID, NIH. For the lambda chain of PGT121, the pBR322 based lambda expression vector was used (AIDS Reagent Program, Division of AIDS, NIAID, NIH).

The constructs were grown in JM109 bacterial cells and extracted using a plasmid Maxiprep kit (Qiagen, Hilden Germany). Sequences were confirmed using the Applied Biosystems 3500xL Genetic Analyzer. Constructs were expressed as previously described (39). In short, HEK293F suspension cells at 1.5x10⁶ to 2x10⁶ cells/ml were cotransfected with linear Polyethylenimine hydrochloride (molecular weight, 40,000) at a 3:1 ratio with 1μg of plasmid per 1 ml of culture. Supernatants were harvested after 6 days.

scFv proteins were purified using Ni-Sepharose beads (GE Healthcare, Massachusetts USA), washed using a 30mM imidazole–phosphate-buffered saline (PBS) solution and eluted using 400mM imidazole in PBS. Glycerol was added to the elution at a final concentration of 5% to limit aggregation. Eluates were applied to Hiload Superdex 75 or Superdex 200 columns (GE Healthcare) equilibrated with PBS at pH 6.5 (5% glycerol with 0.02% sodium azide). The fractions corresponding to the size of the scFv were collected, pooled, and concentrated using Vivaspin concentrators or Vivapore static concentrators (GE Healthcare). The samples were dialyzed overnight at room temperature to remove sodium azide. Concentrations were measured on a NanoDrop device (Thermofisher, MA, USA), with extinction coefficients at 1% calculated using Expasy ProtParam (40) and characterized by SDS-PAGE. Molar weight was determined by using Expasy ProtParam (CAP256.25: 31.35kDa, PGT121: 28.89kDa, 3BNC117 28.47kDa, 10E8v4: 29.30kDa and 8ANC195 29.00kDa). scFv proteins were stored at -75°C.

IgG constructs were expressed in HEK293F cells as described previously (39). Supernatants were harvested after 6 days and purified using a protein A affinity column. Proteins were eluted using a 0.15M glycine buffer at pH 2.5 buffer into 1M Tris, pH 8, and were concentrated and dialyzed into PBS pH 6.5 containing 5% glycerol. Concentrations were measured on a Nanodrop using an Extinction Coefficient of 13.7 at a 1% solution. The molecular weight of the IgG was calculated using the Expasy ProtParam (40) of CAP256.25 (150.71kDa), 10E8v4 (147.29kDa), 3BNC117: (146.24kDa), PGT121: (146.2kDa), 8ANC195: (147.43kDa). IgG and proteins were stored at -75°C.

Plasmids containing HIV-1 envelope (gp160) genes cloned in the pcDNA™3.1D/V5-His-TOPO^® vector were co-transfected with pSG3^Δenv into HEK293T cells and cultured for 48-72hrs at 37°C. Supernatants were harvested and filtered through a 45μm filter and frozen at -80°C. Virus stocks were titrated on TZM-bl cells using a luciferase assay to determine a dilution yielding RLU at least 10-fold above the “cell only” background (40,000-100,000 RLU).

A panel of 43 viruses (41) plus BG505 N332 and CAP256_SU (CAP256.3mo.9C) (42, 43) representing HIV-1 clades A, B, and C was used to compare neutralization titers of IgG and scFv. Neutralization assays were performed in TZM-bl cells as described previously (44–46). Proteins were tested at 200μg/mL for the IgG (~146kDa) and 50μg/mL for the scFv (28-32 kDa). All assays were repeated at least twice. IC₅₀ and IC₈₀ of each antibody tested was calculated and geometric mean potency was calculated for both IgG and scFv using sensitive viruses only.

Combinations were tested by adding equimolar amounts of two or three scFv proteins in a neutralization assay as described above. A panel of 17 subtype A, B and C viruses were selected based on their sensitivity to at least 2 of the scFv in order to test neutralization of scFv combinations and confirm the Loewe Additivity and Bliss-Hill Independence models. Pre-dilutions containing scFv at 2μM and 10μM each were used to facilitate the assay set up. As a control, the single scFv were diluted to 10μM and 2μM as well and run alongside the combinations as a comparison. The highest concentration tested for the combinations was 30μg/mL or 1000nM. IC₅₀ and IC₈₀ of each experimental antibody combination was calculated, where each antibody is present at the concentration determined at IC₅₀ or IC_80.

The IC₅₀ and IC₈₀ values from experimental combinations were compared to the predicted IC₅₀ and IC₈₀ based on the Loewe Additive model and the Bliss-Hill model as previously described (18, 19).

The following formula is used to calculate Loewe Additivity.

Post analysis values were recalculated from nM into μg/mL based on the following formula.

Where (n) is the number of antibodies, Mw the molecular weight of the antibodies in the combination, and IC₅₀ the experimental IC₅₀ or theoretical IC₅₀ obtained. The same formula is used for IC₈₀ values replacing the IC₅₀ with IC₈₀ in the formula above. For resistant viruses, the model assumes the titer of the active scFv.

For the Bliss-Hill Independence model, the following formula was used to calculate the Hill function

Where c= bNAb concentration, k = IC₅₀, and

The combination neutralization curve is then calculated using the Bliss Independence model,

with f(A), f(B), etc. being the individual functions of the scFv antibodies. Combination molar IC₅₀ and IC₈₀ titers are calculated by setting f = 0.5 or 0.8 and assuming each scFv is present at the same molarity, and converted to µg/ml using the above formula.

Dual/triple coverage was calculated by considering a virus resistant if less than 2/3 antibodies in the combination were able to neutralize that virus at set concentrations.

We used the following formula for both models to determine which is more accurate in predicting the combination potencies for IC₅₀ and IC₈₀.

Synergy was predicted based on whether the experimental IC₅₀ and IC₈₀ were improved compared to the Loewe Additive model or the Bliss-Hill model. The formula below was used to characterize this effect, with the Bliss-Hill IC₅₀ replacing the Loewe IC₅₀ in the formula below.

Positive values indicate experimental titers lower (i.e. more potent) than predicted and imply synergy, while negative values indicate less potent experimental titers than predicted and imply antagonism. For a given combination of antibodies, the mean and the 95% Confidence interval were calculated using LogFold values for this combination against each virus in the panel.

All statistics were done using the GraphPad Prism 8 software. Fold differences of the experimental combinations were calculated with the most potent scFv in the mixture for both IC₅₀ and IC₈₀. A Wilcoxon signed-rank test was performed to test which model was more accurate at predicting the experimental titers for both IC₅₀ and IC₈₀. Pearson’s Correlation between experimental neutralization titers and titers based on the two models was calculated using the GraphPad software. A nonlinear fit model (log-log) was used to predict the slope between the experimental and predicted data, using a robust regression. A log-fold with a 95% confidence interval was used to determine if the fold difference is significantly different from 0, indicating either synergy or antagonism. Breadth-potency curves were drawn using a survival model in GraphPad. Significance was tested using the log-rank Mantel-Cox significance test.

To assess whether scFv of HIV bNAbs showed increased breadth and potency when used in combination, we tested five different scFv as part of dual and triple combinations. The scFv included those targeting the V2 (CAP256.25), the N332 supersite (PGT121), the CD4bs (3BNC117), and the MPER region (10E8v4), all of which were previously shown to retain significant activity compared to IgG (38). For this study, we added the antibody 8ANC195 that targets the gp120-gp41 interface so that all five major epitopes on the HIV trimer were covered (see single IC₅₀ and IC₈₀ data in nM Supplementary 1A and 1B and in µg/ml 1C and 1D respectively, (38)]. All scFv were expressed and purified by size exclusion columns and, with size and purity confirmed SDS-PAGE gels. scFv were stored in buffers containing 5% glycerol as determined previously to prevent aggregation (38).

Combinations of scFv were tested against a panel of 17 pseudoviruses from subtypes A, B, and C (n=3, 4 and 10, respectively). These were selected based on their sensitivity to at least three of the five scFv under investigation. Equimolar amounts (based on their molecular weights), of each scFv in combinations of two or three were tested, giving a total of 20 different combinations of antibodies (10 dual and 10 triple combinations), which were compared to single scFv neutralization titers. To standardize the output data, the concentration of the single scFv and the scFv combinations in μg/mL were calculated from the nM titers and the molecular weight of the scFvs ( Figure 1 , Supplementary Figure 2 ) (38). Combination titers IC₅₀ and IC₈₀ are reported as the concentration of each scFv in the mix in μg/mL or nM.

As expected, combinations of two scFv improved the coverage, with active scFv making up for the inactive scFv. For all dual and triple combinations, breadth reached 100% at IC₅₀ against this 17-virus panel ( Figure 1 ). The neutralization of individual viruses by two or three scFv was usually similar to the IC₅₀ of the most potent scFv in the combination ( Figure 1 ; fold differences are shown in the columns next to the IC₅₀s). Except for a few cases, the virus neutralization titers of the dual and triple combinations fell within 3-fold of the titer of the most potent single scFv in that combination. There was little to no loss of potency noted compared to the most potent scFv for the entire panel for all 20 scFv combinations tested (a total of 340 single test titers). A few instances of potential gain of potency compared to the most potent scFv were noted, with most of these linked to specific scFv combinations (bolded values in Figure 1 ).(See Supplementary Figure 2 for IC₅₀ titers in nM (A) and μg/mL (B) and for IC₈₀ values in nM (C) and μg/mL (D) respectively).

The combination of CAP256.25+10E8v4 scFv stood out, demonstrating an overall 4.9-fold improvement in geometric mean titer ( Figure 1B ). This was driven by five viruses of which two were subtype A, one was subtype B and two were subtype C indicating this was not subtype-specific. Five other dual scFv combinations showed improved potency over single scFv for single viruses but this was not linked to any specific combination of antibodies. The 3BNC117+10E8v4 combination showed an overall significant improvement in the geometric mean titer although this was not seen for individual viruses indicating complementarity of neutralization potency of the scFv rather than synergy.

For the triple combinations, five of the ten combinations showed >3-fold improved geometric means compared to the most potent single scFv with a 12-fold improvement for the CAP256.25+3BNC117+10E8v4 combination ( Figure 1C ). This was higher than for the dual combinations where only two out of ten showed an improvement in geometric mean potency and was driven by improvement in potency against single viruses (16 of the 170 virus-scFv combination pairings). This was due to the scFv (e.g. CAP256.25 and 3BNC117 scFv), in the combinations being potent against different viruses allowing for complementarity in neutralization potency and coverage. Triple combinations overall showed higher improvements in breadth compared to dual combinations or single scFv as a consequence of having more antibodies.

Some viruses appeared to be more sensitive to the effects of combined scFv. For example, a slight potency improvement was observed against the CAP61 and ZM249 viruses for the dual combinations, although for the latter loss of potency was also noted (fold difference <0.33 see Figure 1B ). Some enhancement in potency was also seen with these viruses plus Du156 and CAP88 for the triple combinations ( Figure 1C ).

Overall, we found either improved or similar titers for combinations compared to the most potent single scFv, particularly for triple combinations, indicating that combinations of scFv improve the coverage of a panel of viruses and the potency at which viruses are neutralized. We overlaid neutralization curves of the combinations with the single scFv used in the combination, to determine if there were distinct patterns associated with improved IC₅₀s or where we observed similar IC₅₀ for the combinations compared to the single scFv. For most dual and triple combinations (314/340), an additive effect was seen where the combination curve overlapped with the best scFv in the combination ( Figure 2A ). Here IC₅₀ did not show a potency improvement compared to the single scFv. A few combinations shifted the curve to the left compared to the curve of the most potent single scFv corresponding with fold potency improvements as seen in Figure 1 and indicating potential synergy (10/170 virus-dual combination pairings and 16/170 in the triple combinations showed improved IC₅₀s) ( Figure 2B ). In one case we observed decreased potency for the combination compared to the most potent scFv possibly indicating some antagonism ( Figure 2C ). With the addition of a third antibody this was partially negated ( Figure 2C right panel). Overall, most combinations did not show significant potency improvements or loss of potency, suggesting that they largely followed an additive model of potency.

To further explore whether combinations of scFv follow an additive model of interaction we compared the experimental results with predictions based on the single scFv titers using the Loewe Additive model and the Bliss-Hill models. As mentioned before, both these models assume no synergistic or antagonistic interactions between antibodies (18, 19). We calculated the combination scFv IC₅₀ and IC₈₀ titers based on these 2 models (Methods) and based on the geometric mean IC₅₀ and IC₈₀ of the repeats of single scFv obtained in the experiment. All scFv predicted data titers were compared to the IC₅₀ and IC₈₀ titers of the experimentally tested combinations.

There was a strong significant (p<0.0001) correlation between the experimental IC₅₀ and IC₈₀ and the predicted IC₅₀ and IC₈₀ of the dual combinations for both the Loewe and Bliss Hill model (Loewe, Pearson’s r=0.94 and r=0.79 respectively, Bliss-Hill, r=0.94 and 0.91 respectively) ( Figures 3A, B ). Similarly, for the triple combinations, there was a strong correlation between the predicted values and the experimental values (Loewe r=0.78 and r=0.79, Bliss-Hill: r=0.70 and r=0.95 respectively) ( Figures 3C, D ). For each combination, we also determined which model was better by comparing the mean absolute log difference between the experimental values and the predicted values for each individual combination. For the dual combinations, the Bliss Hill model was better at predicting two out of the 10 combinations (10E8v4 combined with CAP256.25 or PGT121), whereas no significant difference between the models was observed for any of the other combinations ( Figures 3E, F ). Similarly, for the triple combinations, most combinations did not show differences between the two models in the IC₅₀ with the Bliss-Hill model being better at predicting two combinations (8ANC195 + 10E8v4 with either CAP256.25 or PGT121) ( Figure 3G ). For the IC₈₀ of the triple combinations, 6/10 combinations were significantly better predicted by the Bliss-Hill model compared to the Loewe model ( Figure 3H ). This indicated that for most combinations both Loewe and Bliss-Hill could predict combination IC₅₀ titers well with a slight advantage for the Bliss-Hill model. However, the Bliss-Hill model is significantly better at predicting triple IC₈₀ titers compared to the Loewe model, with the latter underestimating the potency of the triple combinations at IC₈₀.

We next sought to determine if any of the combinations showed significant potency improvement or potency loss in either IC₅₀ or IC₈₀ by comparing the experimental data to the expected titers based on the Bliss Hill model. These analyses allowed us to determine if there is potential synergy (higher potency than predicted), as indicated by a log-fold difference >0.4 (~2.5 fold) and/or antagonism (lower potency than predicted) as indicated by a log-fold difference <-0.4 ( Figure 4 ). The dual combinations had a mean log-fold difference close to 0 (log-fold -0.4<CI<0.4) for both IC₅₀ and IC₈₀ indicating that there was neither antagonism nor synergy, although a few antibody combinations showed some improved potency (red dots, IC₅₀ n=3, 1.7% and IC₈₀: n=0, 0%) or decreased potency (blue dots, IC₅₀ and IC₈₀: n=6, 3.5%) for individual viruses these were relatively few and most showed less than 0.4 log fold difference when compared to model predictions ( Figures 4A, B ).

For the triple combinations, similar results were observed with 157/170 of the IC₅₀ and 154/170 of the IC₈₀ values showing less than a Log (2.5) difference. The mean on triple combinations was also close to 0 (log-fold -0.4<CI<0.4) for both IC₅₀ and IC₈₀. Similarly, only a few individual viral titers showed a potential loss of potency for the IC₅₀ (n=8, 4.7%) and IC₈₀ (n=11, 6.5%) or a potential gain of potency (IC₅₀ and IC₈₀: n=5, 3%) for triple combinations compared to the model ( Figures 4C, D ). None of the outliers in dual or triple combinations could be linked to specific viral signatures (data not shown). Notwithstanding these few outliers, we, did not observe any significant synergy nor antagonism for dual or triple combinations of scFv.

In order to assess their breadth-potency on a larger virus panel, we used the IC₅₀ and IC₈₀ data of single scFv to predict neutralization titers for combinations of two and three scFv, now including 8ANC195, on a 45-virus panel described in our previous study (38) (see Supplementary Figure 1 for single IC₅₀ in nM (A) and IC₈₀ in nM (B) or μg/mL (C, D) titers) (38). These theoretical scFv combinations were ranked based on their geometric mean potency using both IC₅₀ and IC₈₀ data for dual and triple combinations ( Figures 5A, B, E, F ). Combinations with the expected highest breadth and potency were plotted using breadth-potency curves ( Figures 5C, D, G, H ).

The best dual combinations, namely PGT121+10E8v4 (red), 3BNC117+10E8v4 (pink), and CAP256.25+10E8v4 (orange) were able to neutralize most viruses barring one at IC₅₀ and IC₈₀ ( Figures 5A, B ). The breadth-potency curves of the three 10E8v4-containing combinations reached 100% neutralization at the IC₅₀ and 98% neutralization at the IC₈₀ (colored breadth-potency plots versus grey, Figures 5C, D ) with most viruses neutralized between 0.1 and 1μg/mL at IC₈₀. The fourth combination of 10E8v4 with 8ANC195 was less effective due to the low potency of both antibodies and the limited breadth of 8ANC195, [see Supplementary Figure 3 (38)]. The other dual combinations showed reduced breadth (up to 30% at 50μg/mL) or reduced potency especially at IC₈₀ (grey dots and lines). Some combinations such as PGT121+3BNC117 had a potent geometric mean below 0.1μg/mL at IC₅₀, but there was a larger spread of neutralization titers and so breadth was lost at IC₈₀ ( Figures 5A, B ). (See Supplementary Figure 4 for IC₅₀ in nM (A) and μg/mL (B) and IC₈₀ in nM (C) and μg/mL (D) titers of dual, and triple combinations predicted by the Bliss-Hill Independence model).

The triple combinations showed higher levels of potency and complete neutralization with 8/10 combinations showing complete neutralization of the viral panel at IC₅₀, and 6/10 showing near-complete neutralization at IC₈₀ ( Figures 5E–H ). Moreover, 7/10 triple combinations had a geometric mean potency below 0.1μg/mL at IC₅₀ ( Figure 5E , green, dark green, and blue). Of these, three combinations all had a geometric mean IC₈₀ below 1μg/mL and neutralized up to 96% of viruses ( Figures 5F, H ). This indicated that triple combinations are able to reach nearly 100% neutralization at lower concentrations for both IC₅₀ and IC₈₀ (see Supplementary Figure 4 for IC₅₀ in nM (A) and μg/mL (B) and IC₈₀ in nM (C) and μg/mL (D) titers of dual, and triple combinations predicted by the Bliss-Hill Independence model).

The predicted breadth of each combination was next determined at four concentrations: 0.1, 1, 10, and 50μg/mL in addition to the geometric mean for the combinations tested (with a cutoff at 50μg/mL). Combining these scFv improved the coverage at all concentrations for both IC₅₀ and IC₈₀ ( Figure 6 ). Breadth for dual combinations ranged from 84% to 100% at IC₅₀ < 50µg/ml and 80% to 100% at IC₈₀ < 50µg/ml. For dual combinations, 100% breadth was reached for one combination at IC₅₀ < 1μg/mL, and for four combinations at IC₅₀ < 50μg/mL. For the IC₈₀, 98% breadth was obtained at 10μg/mL for the same combinations. Most combinations of three scFv had 100% breadth at IC₅₀ < 50 µg/ml, except for PGT121+3BNC117+8ANC195 and CAP256.25+PGT121+8ANC195, which had a breadth of 96% and 93% respectively. Breadth was especially improved for lower concentrations at IC₈₀. For example, breadth for single scFv ranged from 0 to 44% for IC₅₀ < 0.1µg/ml and 0 to 24% for IC₈₀ < 0.1µg/ml, whereas dual combinations ranged from 22 to 62% for IC₅₀ and 2 to 29% for IC_80. At the same cutoff, the breadth for combinations of 3 antibodies ranged from 47% to 82% for IC₅₀ and 16 to 40% for IC₈₀. Geometric mean potency for the panel also improved from 0.2-4.95 μg/mL for single scFv to 0.27-2.08μg/mL for dual combinations and 0.09-0.61μg/mL for triple combinations at IC₈₀. This was expected as the titers were calculated as containing the same amount of each scFv at that concentration.

Since PGT121, 3BNC117, and CAP256.25 all had good potencies as scFv, using these together with 10E8v4 resulted in improved potency of the combination. Moreover, scFv such as CAP256.25 and PGT121 and/or 3BNC117 have complementary profiles, with 3BNC117, for example, being especially potent against subtype B viruses and CAP256.25 being potent against subtype C viruses (see Supplementary Data 1 for the IC₅₀ and IC₈₀ of single scFv) (38). 10E8v4 is the broadest scFv and that resulted in the combinations containing 10E8v4 having the highest breadth. As a result, four of the five triple combinations with the best geometric mean titers included 10E8v4. Conversely, introducing less potent or less broad antibodies such as 8ANC195 resulted in less favorable antibody combinations, where 5 out of 6 combinations with the lowest potency and breadth contained 8ANC195, consistent with the relatively lower potency and breadth of this antibody. Of the top five, three combinations, CAP256.25+PGT121+10E8v4 and CAP256.25+3BNC117+10E8v4 and CAP256.25+8ANC195+10E8v4 neutralized all viruses below 1μg/mL and 98% of viruses at IC₈₀ below 10μg/mL. The best ranked PGT121+3BNC117+10E8v4 combination neutralized 96% of viruses below 1μg/mL and 98% of viruses below 10μg/mL at the IC₈₀ value. This analysis indicated a significant gain of potency and neutralization breadth at lower concentrations than single scFv at both IC₅₀ and IC₈₀. This indicated that combinations of scFv were able to increase the coverage at lower concentrations, which will be vital in clinical settings.

The AMP trial showed that viral isolates that were sensitive to VRC01 with an IC₈₀<1μg/mL were prevented from establishing infection in 75% of cases. As indicated in Figures 5 and 6 , scFv combinations can reach significant coverage by a single antibody at this potency with the best scFv combination potentially reaching a 96% protection level. However, other antibody classes may need higher levels of potency to provide similar protection as VRC01 and viral isolates which are neutralized close to this cut-off may allow for breakthrough. Dual coverage will be able to overcome these issues by neutralizing breakthrough viruses at higher potencies and providing a back-up in case antibody levels drop. Therefore, using this cut-off, we calculated the predicted dual coverage (i.e. by at least two scFv) of the combinations using the thresholds of single scFv IC₈₀ < 1µg/ml and < 10μg/mL. We found for dual combinations, the dual active coverage at 1μg/mL ranged from 20% with improvement at 10μg/mL to 51% for PGT121+10E8v4 ( Figures 7A, B ), This ranged from 0-20% at 1μg/mL and 9-62% for other combinations of two antibodies (See Supplementary Figure 5 ).

This was somewhat improved for triple combinations where we observed up to 58-80% dual active coverage at an IC₈₀ of 10μg/mL. The best 3-combinations,3BNC117+PGT121+10E8v4 showing the highest active coverage at 10μg/mL of 80%. Generally, triple combinations showed higher levels of active coverage by 2 antibodies and often included both 10E8v4 and 3BNC117. Similar to the dual combinations, low dual coverage was observed at 1μg/mL up to 38% (See Figure 7A , Supplementary Figure 5 . This could not be significantly improved for quadruple combinations which showed 44% and 87% dual coverage respectively at 1 and 10μg/mL. This loss of active coverage by multiple antibodies is also demonstrated in the breadth potency curves ( Figures 7C, D ). These curves clearly show that although at 10μg/mL most of the dual coverage is maintained there is a significant drop off at 1μg/mL. This was mainly due to antibodies such as 3BNC117, PGT121, and 10E8v4 showing only 31-49% breadth at 1μg/mL and the CAP256.25 scFv losing significant potency compared to the IgG (38). IgG combinations including this antibody could show slightly better active coverage by 2 antibodies, up to 53% at an IC₈₀ of 1μg/mL and up to 91% at 10μg/mL ( Supplementary Data 6A for predictions of IgG combinations titers in μg/mL and 6B for neutralization breadth at 4 concentrations and dual coverage by IgG) (38). The other IgG and scFv showed similar levels of potency with specifically the 10E8v4 scFv retaining all of its potency explaining why the IgG combinations did not show higher levels of dual active coverage at 1μg/mL ( Figure 7 , Supplementary Figure 6B ). This indicated that potency of single IgG and scFv needs to be improved to obtain higher levels of active coverage by 2 antibodies at lower antibody levels. The combination data demonstrate that the benefits of combinations of IgG targeting different HIV epitopes apply also to the smaller and more versatile scFv fragments.