Antibody Binding to Recombinant Adeno Associated Virus Monitored by Charge Detection Mass Spectrometry

Abstract

Recombinant adeno-associated virus (rAAV) is a leading gene therapy vector. However, neutralizing antibodies reduce its efficacy. Traditional methods used to investigate antibody binding provide limited information. Here, charge detection mass spectrometry (CD-MS) was used to investigate the binding of monoclonal antibody ADK8 to AAV serotype 8 (AAV8). CD-MS provides a label-free approach to antibody binding. Individual binding events can be monitored as each event is indicated by a shift of the antibody−antigen complex to a higher mass. Unlike other methods, the CD-MS approach reveals the distribution of antibodies bound on capsids, allowing AAV8 subpopulations with different affinities to be identified. The charge state generated by the electrospray of large ions is normally correlated with the structure, and the charge is expected to increase when an antibody binds to the capsid exterior. Surprisingly, binding of the first ADK8 to AAV8 causes a substantial decrease in the charge, suggesting that the first antibody binding event causes a significant structural change. The charge increases for subsequent binding events. Finally, high ADK8 concentrations cause agglutination, where ADK8 links AAV capsids to form dimers and higher order multimers.

Graphical Abstract

INTRODUCTION

With three FDA-approved treatments, recombinant adeno-associated virus (AAV) has emerged as one of the most promising gene delivery vectors.¹ Despite being nonpathogenic, AAV can still elicit an immune response.^2–4 Some potential patients have pre-existing neutralizing antibodies, the seroprevalence depending on age and region.^5,6 To avoid adverse reactions, the presence of neutralizing antibodies must be assessed prior to treatment. A large effort has been devoted to developing strategies for rAAV to avoid the body’s immune system.^7,8

Routine assays, such as total antibody⁹ and transduction inhibition,¹⁰ are often used to screen patients’ eligibility for treatment or to evaluate improvements in the ability of newly engineered capsids to evade neutralization. Enzyme linked immunosorbent assay (ELISA), a widely used total antibody assay, is labor intensive and time-consuming. Transduction inhibition assays measure the ability of any neutralizing factor, not just antibodies, to block the transduction of AAV capsids containing a special reporter genome that can be visualized (e.g., luciferase). Cryo-EM has been used to map the epitopes of various AAV serotypes.^11,12

The AAV capsid is heterogeneous. It contains 60 capsid proteins (Cp), but there are three Cp variants (VP1, VP2, and VP3) and 1891 possible combinations. The VP1/VP2/VP3 ratios are not fixed but have an ensemble average of around 1:1:10. The result is a capsid with an average mass of ~3.75 MDa with a distribution ~125 kDa wide. Such a heterogeneous mass distribution cannot be determined using conventional MS,¹³ but it can be measured by charge detection mass spectrometry (CD-MS).¹⁴ CD-MS has previously been used to examine AAV vector heterogeneity and genome integrity.^15–17 In this study, we have used CD-MS to investigate antibody binding to AAV capsids. Specifically, we monitored the binding of neutralizing antibody ADK8 to empty AAV serotype 8 (AAV8) capsids. A key advantage of CD-MS is that it does not require serotype specific labels; antibody binding is detected by a shift in the mass. Furthermore, using CD-MS, it is possible to determine the number of antibodies bound to each capsid, allowing the antibody binding reaction to be monitored at a level of detail not attainable by other methods.¹⁸ Using murine monoclonal antibody ADK8 (which binds intact AAV8 capsid) as a model, we find that binding a single antibody causes a substantial decrease in the charge of the AAV8 particle, suggesting that the binding of a single ADK8 molecule is sufficient to cause a significant structural change. We find that ADK8 binding is heterogeneous with a small component that resists antibody binding. Finally, high ADK8 concentrations cause agglutination, where ADK8 links AAV8 capsids to form dimers and higher order multimers.

EXPERIMENTAL METHODS

Samples.

Empty AAV8 was prepared by transfecting equal molar ratios of pRep2-Cap8 packaging plasmid and mini-adenovirus helper plasmid into HEK 293 cells (human embryonic kidney cells CRL-1573, ATCC) using PolyJet DNA In Vitro Transfection Reagent (SignaGen Laboratories). At 72 h after transfection, the vectors were harvested and purified by affinity chromatography. The vector capsid titer was measured with ELISA using ADK8 against AAV8 intact capsid antibody (Cat. No: 651160, Progen).

The AAV8 empty capsids (10¹³ particles/mL) were aliquoted and stored at −80 °C. When ready for use, the capsids were thawed at 4 °C. The thawed capsids were then exchanged into 200 mM ammonium acetate (Invitrogen, AM9071), which had been adjusted to a pH of 7.4, using Micro Bio-Spin columns. The capsids were then diluted 20× with the same 200 mM ammonium acetate solution.

ADK8 antibodies were received from Progen at a concentration of 5 mg/mL. The antibodies were exchanged into 20 mM ammonium acetate using Micro Bio-Spin columns, diluted to a concentration of 50 μg/mL with the same ammonium acetate solution, and aliquoted for future use. The aliquots were stored at −80 °C and thawed at 4 °C when they were ready for use.

AAV-Antibody Incubations.

7 μL aliquots of the diluted AAV solution were mixed with 3 μL aliquots of ADK8 solution. Prior to mixing, the ADK8 solution was diluted to give final ADK8 concentrations of 0.03 and 0.1 μM when mixed with the AAV solution. The AAV-antibody mixture was incubated for 1 h at room temperature in a Benchmark Roto-Therm Mini Plus at 30 rotations per minute. Following the 1 h incubation, the mixtures were electrosprayed at room temperature.

CD-MS Measurements.

CD-MS is a single particle technique where the m/z and charge are measured for individual ions.¹⁴ Ions are trapped in an electrostatic linear ion trap (ELIT) and oscillate back and forth through a conducting cylinder. The oscillating ion induces a charge on the cylinder that is detected by a charge sensitive amplifier. The resulting signal is digitized and analyzed using fast Fourier transforms. The fundamental frequency is used to calculate the m/z ratio, and FFT magnitudes are used to determine the charge. The mass of each ion is obtained from the product of its m/z and charge. Thousands of ions are measured, and the masses are binned into a histogram to give the mass spectrum.

RESULTS AND DISCUSSION

Figure 1a shows a typical CD-MS spectrum measured for the empty AAV8 capsid. The inset shows an expanded view of the main peak. The dashed red line in the inset is a Gaussian fit to the peak with a center of 3.76 MDa and a full width at half-maximum (fwhm) of 0.127 MDa. The peak center and width are consistent with previous CD-MS measurements for AAV8.^15,16 Figure 1b shows a charge versus mass heat map for empty AAV8 particles. This plot reveals correlations between mass and charge; in this case, the mass and charge distributions are both narrow. The inset represents an expanded view of the main feature. The charge distribution is centered on 153 e (elementary charges).

Figure 1.

CD-MS measurements of binding of the antibody ADK8 to AAV8. (a and b) Typical mass distribution and charge versus mass heat map for AAV8 prior to antibody exposure. The inset in part a shows an expanded view of the main peak. The dashed red line is a Gaussian fit, and the solid red line marks the peak center (3.76 MDa). In the heat map in part b, warmer colors indicate a higher signal density. The inset in part b shows an expanded view of the main feature, which is centered on a charge of 153 e. (c and d) Mass distribution and charge versus mass heat map measured after incubation with 0.03 μM ADK8. The inset in part d shows an expanded view of the main feature in the heat map showing that a low charge population has emerged. (e and f) Mass distribution and charge versus mass heat map measured after incubation with 0.1 μM ADK8. The red line in part e shows the center of the unbound AAV8 peak. The structures in part e indicate antibody binding (responsible for the peak at 6 MDa) and agglutination where the antibody cross-links capsids to form multimers (responsible for higher mass peaks). The spectra in a, c, and e are histograms generated using 20 kDa bins.

Figure 1c depicts a representative spectrum measured after incubation with 0.03 μM ADK8. The main peak has broadened to higher mass, and a small peak emerges in the mass distribution at around 8 MDa. An expanded view of the main peak is shown by the black line in Figure 2a. There is a series of poorly resolved peaks separated by around 150 kDa and a high mass tail that extends beyond 5.5 MDa. The poor resolution is mainly due to the heterogeneity of the AAV8 capsid. Because the sequence for ADK8 is not available, we measured its mass by CD-MS and obtained a value of 150.4 kDa (see Supporting Information). Thus, the separation between the peaks in Figure 2a is consistent with ADK8 binding. As an additional control, we incubated ADK8 with AAV9. ADK8 is reported to have no affinity for AAV9. CD-MS mass distributions measured for AAV9 with and without incubation with ADK8 were identical (see Supporting Information), confirming that ADK8 does not bind to AAV9, and showing that the ADK8 binding observed with AAV8 was specific. Thus, we are confident that the resolved peaks in Figure 2a, separated by around 150 kDa, are due to the stepwise addition of ADK8 to AAV8.

Figure 2.

Expanded view of mass distributions after AKD8 binding. (a) An expanded view of the mass distribution measured after incubation with 0.03 μM ADK8. The vertical gray lines show the peak centers for the AAV8 control and for AAV8 after sequential addition of ADK8 (assuming a mass of 150.4 kDa). (b) The spectrum in part a plotted using the same scale as in part c below. (c) Mass distribution measured after incubation with 0.1 μM ADK8. The gray vertical lines in parts b and c show the peak center for the AAV8 control. The dashed vertical line in part c shows the expected position of AAV8 with 13 ADK8 attached. The red lines show simulations that use a binomial distribution to account for the distribution of bound antibodies and use instrumental resolution and sample heterogeneity to give the peak widths of the individual adducts (see text).

The charge versus mass heat map corresponding to the mass distribution in Figure 1c is shown in Figure 1d, including an expanded view of the main features. The sharp distribution centered on 3.76 MDa and 153 e persists, representing AAV8 not bound to an antibody. However, when shifted to a slightly higher mass, there is a second distribution reflecting AAV8 with bound ADK8. A remarkable feature of this distribution is that it has a substantially lower charge than unbound AAV8. The addition of a single ADK8 causes the center of the charge distribution to decrease from 153 to 136 e, a drop of 11%.

Electrospray of large ions is expected to occur through the charge residue mechanism where the charge acquired by an ion reflects its size.^19,20 The charge can be used to deduce information about the structure of both large and small ions.^21,22 For AAV, empty and genome-containing capsids have similar charges (despite the mass difference) because the capsids are the same size.¹⁵ The charge increases for full AAV particles when they are heated, a result attributed to DNA extrusion.²³ In the present case, when antibodies bind to a virus capsid, the charge is expected to increase slightly because the antibody should bind to the outside of the capsid, increasing its size. However, we observe a decrease in charge of around 11% with the addition of the first antibody. Only the first ADK8 binding event causes the charge to decrease; i.e., the charge does not decrease further when additional ADK8 antibodies are bound. Each subsequent antibody addition causes the charge to increase by around 3 e. This is evident from the heat map in Figure 1d and further confirmed by studies at higher ADK8 concentration where more antibodies bind (see below). As noted above, an increase in charge with antibody binding is expected; therefore, it is only the first antibody binding event that is anomalous. The decrease in charge associated with binding the first antibody suggests that a substantial structural change is associated with that process.

Cryo-EM studies suggest that ADK8 Fabs bind to AAV8 around the protrusions on the 3-fold symmetry axis.¹¹ However, these studies were performed at saturation coverage (around 60 Fabs) and relatively low resolution, so they cannot really speak to the putative structural change that occurs when the first antibody binds. Mutagenesis studies show that the capsid protein peptide 586-LQQQNT-591 that lies in the binding footprint is critical for binding: when it was mutated, binding was inhibited.¹¹ This peptide is present in VP1, VP2, and VP3.

Figure 1e shows the mass distribution measured after the incubation of AAV8 with 0.1 μM ADK8. The corresponding charge versus mass heat map is shown in Figure 1f. The charge increases as more ADK8 antibodies bind, as explained above. In the mass distribution, there are now several broad peaks: the main one centered on 5.9 MDa, with smaller peaks centered on 11.7 and 17.2 MDa. The main peak centered on 5.9 MDa is attributed to the AAV8 capsid that has bound, on average, around 14 antibodies. The peaks at 11.7 MDa and 17.2 MDa in Figure 1f are probably due to dimers and trimers of AAV8. The expected masses of the dimer and trimer are 7.52 and 11.28 MDa, respectively. The measured masses are significantly larger; the extra mass can be attributed to antibody binding. On average, around 28 ADK8 are bound to the dimer, and around 40 are bound to the trimer. Recall that on average around 14 ADK8 are bound to the monomer so the number of ADK8 scales with the number of AAV8 in the multimers. There is no dimer present for the AAV8 sample without antibodies (Figure 1a). Thus, the antibody promotes aggregation of the AAV8. Antibodies could link AAV particles together if the two antigen binding sites on an antibody bind to different AAV particles. Agglutination, where cells, bacteria, viruses, or other particles with epitopes are clumped together by antibodies, is a well-known phenomenon.²⁴

Agglutination is barely detectable with an ADK8 concentration of 0.03 μM (see the small peak at around 8 MDa in Figure 1c). However, it becomes much more prevalent with an ADK8 concentration of 0.1 μM. Bearing in mind that each ion in the dimer peak contains two AAV8 capsids and each ion in the trimer peak contains three, around two-thirds of the AAV8 capsids are linked, with an ADK8 concentration of 0.1 μM. A dependence on the concentrations of both antibody and antigen is expected for agglutination.^25–27 The binding of the second antigen binding site to the same capsid may be frustrated at high antibody coverage because neighboring epitopes are already occupied, leaving one of the antigen binding sites dangling and free to bind to a different capsid.

If all ADK8 binding sites on AAV8 have the same affinity, then the distribution of bound ADK8 should follow a binomial distribution. The red line in Figure 2a shows the result of a simulation where the antibody distribution is binomial, and the peak widths for unbound AAV8 and the AAV8-ADK8 complexes are determined by the heterogeneity of the AAV8 capsid and the instrumental resolution (see Supporting Information for a more detailed description of the simulation). The parameters for the binomial distribution were adjusted to fit the two most prominent peaks in the spectrum. The positions of the poorly resolved features in the measured spectrum match the positions of the features in the simulation. However, the measured intensities for the higher mass peaks are much larger than the predicted intensities. This discrepancy may be caused by heterogeneity (where some particles have higher affinities), by positive cooperativity (where antibody binding causes the affinity to increase), or by both.

Figure 2c shows an expanded view of a portion of the mass distribution measured after incubation with 0.1 μM ADK8 (with the red line again showing a simulated binomial distribution for binding of ADK8 to AAV8). According to this simulation, it should still be possible to partially resolve peaks due to the addition of individual antibodies. There do appear to be poorly resolved features in the measured spectrum that align with the peak positions in the simulated spectrum. Figure 2b shows the mass distribution measured with an ADK8 concentration of 0.03 μM plotted on the same scale as the 0.1 μM distribution. At 0.03 μM, the average number of antibodies bound is around 2, while at 0.1 μM the average is around 13. The faster than linear increase suggests positive cooperativity. The measured distribution in Figure 2c is much broader than the calculated binomial distribution. There is a small peak at the mass of unbound AAV8, while at the other end, some capsids have bound over 20 ADK8 molecules.

CONCLUSIONS

CD-MS provides a label-free approach to monitoring antibody binding. Individual binding events can be detected, as each event is indicated by a shift in the mass of the antibody-antigen complex. The ability to monitor the distribution of bound antibodies is a feature of the CD-MS approach that makes it possible to determine whether there are subpopulations with different binding affinities. The distribution of ADK8 bound to AAV8 indicates a broad range of affinities, with one small component that resists binding a single antibody and another that binds over 20 antibodies under the same conditions.

For large ions, the charge generated by electrospray is expected to reflect the size of the ion. Thus, antibody binding to the exterior of a capsid is expected to result in a charge on the AAV8-ADK8 complex that is slightly larger than that on unbound AAV8. Surprisingly, we found that the addition of the first ADK8 to AAV8 causes the charge to decrease by 11%. Subsequent ADK8 additions caused the charge to increase, as expected. Thus, we hypothesize that the first ADK8 bound to AAV8 causes a structural change in the AAV8 capsid.

At high concentrations, ADK8 links AAV8 capsids together to form dimers and higher order multimers. Agglutination is a well-known phenomenon in antigen−antibody reactions, leading to the clumping of viruses, bacteria, or cells. It is the basis of many widely used assays including blood typing and antibody testing.^28–30 These assays consume milliliters of sample, while a CD-MS measurement can be performed with 10 μL. Thus, CD-MS may open new opportunities for agglutination when samples are limited.