Probing Antibody Binding to Canine Parvovirus with Charge Detection

1 hour ago - There are many techniques for monitoring and measuring the interactions between proteins and ligands. Most of these techniques are ...
0 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/JACS

Cite This: J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Probing Antibody Binding to Canine Parvovirus with Charge Detection Mass Spectrometry Carmen A. Dunbar,† Heather M. Callaway,‡ Colin R. Parrish,*,‡ and Martin F. Jarrold*,† †

Department of Chemistry, Indiana University, 800 E. Kirkwood Ave., Bloomington, Indiana 47405, United States Baker Institute for Animal Health, Department of Microbiology and Immunology, College of Veterinary Medicine, Cornell University, Ithaca, New York 14850, United States

‡

J. Am. Chem. Soc. Downloaded from pubs.acs.org by UNIV OF SOUTH DAKOTA on 11/07/18. For personal use only.

S Supporting Information *

ABSTRACT: There are many techniques for monitoring and measuring the interactions between proteins and ligands. Most of these techniques are ensemble methods that can provide association constants and in some cases stoichiometry. Here we use charge detection mass spectrometry (CDMS), a single particle technique, to probe the interactions of antigen binding fragments (Fabs) from a series of antibodies with the canine parvovirus (CPV) capsid. In addition to providing the average number of bound Fabs as a function of Fab concentration (i.e., the binding curve), CDMS measurements provide information about the distribution of bound Fabs. We show that the distribution of bound ligands is much better at distinguishing between different binding models than the binding curve. The binding of Fab E to CPV is a textbook example. A maximum of 60 Fabs bind and the results are consistent with a model where all sites have the same binding affinity. However, for Fabs B, F, and 14, the distributions can only be fit by a model where there are distinct virus subpopulations with different binding affinities. This behavior can be distinguished from a situation where all CPV particles are identical, and each particle has the same distribution of sites with different binding affinities. The different responses to viral heterogeneity can be traced to the Fab binding sites. A comparison of Fab binding to new and aged CPV capsids reveals that a post-translational modification at the binding site for Fab E (M569) probably reduces the binding affinity.

■

INTRODUCTION The interaction of viral capsids with host antibodies, generated after infection or vaccination, can result in neutralization of the virus and protection against infection. The specific interactions between viruses and host receptors or antibodies have been highly refined by evolution of both the virus and the host, leading to sophisticated interactions that balance the need for efficient viral infection and control of viral host ranges, while allowing viral success in the face of antibody challenge.1 Parvoviruses are small, single-stranded DNA viruses with genomes of about 5100 bases that are packaged into T = 1 icosahedral capsids.2,3 Parvovirus capsids are very stable and retain infectivity for days or weeks in the environment, but must also be able to release their genomes to begin replication after they bind to receptors and are endocytosed into the cell.4,5 This indicates that they must undergo a variety of controlled conformational changes during different stages of their life cycles. Canine parvovirus (CPV) is a host range variant of a virus related to feline panleukopenia virus (FPV) of cats and emerged in 1978 to cause a pandemic in dogs. CPV and FPV capsids bind to the transferrin receptor type-1 (TfR) to enter and infect cells, and the emergence of CPV in dogs resulted from changes in the CPV capsid that allowed it to bind the canine TfR and use it for infection.6,7 Antibodies © XXXX American Chemical Society

provide the majority of the protective immunity against CPV and FPV, in part because the virus circulates freely in the blood during infection and is exposed to antibodies.8 Parvovirus capsids are potent antigens and stimulate strong antibody responses after infection or vaccination.9 A number of studies have defined the antigenic structure of parvoviral capsids, or the related adeno-associated viral capsids, complexed with antibodies.10−12 Those studies show that antibodies bind to many of the exposed structures on the viral capsid surface. In studies of eight different mouse monoclonal antibody antigen binding fragments (Fabs) complexed with CPV or FPV capsids, it was seen that Fabs mostly saturated the available binding sites and their footprints covered ∼70% of the particle surface, but that they differed greatly in their efficiency of neutralization.10,13,14 Despite the importance of antibodies in mediating host immunity and in determining vaccine efficacy, it has proven difficult to quantitatively define the interactions of antibodies with individual capsids or to understand their dynamic properties. We lack a clear understanding of the dynamics and stoichiometry of antibody binding to individual viral Received: July 29, 2018

A

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society particles, which may be the key to efficient neutralization. Most techniques for monitoring and measuring the interactions between proteins and ligands are ensemble techniques such as isothermal titration calorimetry (ITC), surface plasmon resonance (SPR), or enzyme-linked immunosorbent assays (ELISA). While these methods provide measurements of equilibrium constants and in some cases stoichiometry, they do not reveal the interactions at the single particle level. Only a few single particle techniques can probe these types of interactions, but they require very specific sample preparation and few give stoichiometric information.15−20 Methods that can give stoichiometric information, such as Fö rster resonance energy transfer (FRET), require the presence of a fluorophore and cannot determine the distribution bound.20 In SPR based methods, a functionalized surface is used to probe protein−ligand interactions,21,22 including those between antibodies and viruses. Either the antibody or virion is bound to a functionalized surface, and binding of the ligand changes the intensity of the resulting beam reflected from the surface.23−25 SPR can determine the average number of ligands bound to viruses in aggregate but does not reveal the number of antibodies bound to individual particles or variability in that binding. ELISA is commonly used to measure the interaction of antibodies or antibody fragments with viruses,13 but it also does not reveal the binding levels to individual particles. ITC uses the heat loss or gain of the chemical reaction between substrate and ligand to obtain the binding constant, enthalpy of binding and number of binding sites,26 and again provides no information on the variability in the binding to individual particles. Charge detection mass spectrometry (CDMS) has been used to obtain masses of large particles such as viruses and multicomponent complexes.27−40 Here, we use CDMS to examine, for the first time, the details of the dynamics and stoichiometry of the interactions between a viral capsid and antigen binding fragments (Fabs) from a series of antibodies. The T = 1 parvovirus capsid displays 60 nominally identical copies of each epitope on its surface, and we test the Fabs of four antiviral antibodies that recognize defined structures on the capsid, as determined previously using cryo-EM.10 In this way, we are able to define and compare binding equilibria and reveal the occupancy of those antibodies on individual virus particles. The results show that each antibody has a unique profile of attachment to the viral capsid, despite showing similarly high affinities of binding when examined using ensemble methods.13

Table 1. Masses Measured by Conventional MS for the Antigen Binding Fragments Used in This Study antigen binding fragment Fab Fab Fab Fab

E B 14 F

measured mass (Da) 49 253 46 496 46 490 47 782

present in VP1. VP3 is derived from VP2 by proteolytic cleavage during maturation of full capsid.42−44 Empty capsids only contain VP1 and VP2 in a ratio of around 10:90.41,45 The sequence masses of VP1 and VP2 are 80 342 and 64 705 Da respectively,46 so the expected mass of an empty CPV particle is 3.987 MDa. The CMDS spectrum measured for empty CPV is shown in Figure 1. The measured spectrum (black points and line) has a peak centered on 4.063 MDa with a small high mass shoulder that extends beyond 4.5 MDa. The peak center (4.063 MDa) is 2% larger than the expected mass of a CPV capsid with a 10:90 ratio of VP1 and VP2. Counterions and adducts are known to lead to a measured mass up to 1% larger than the expected mass. The other 1% is attributed to the amount of VP1 being slightly larger than in the expected 10:90 VP1:VP2 ratio. The thin red line in Figure 1 shows the peak calculated for a binomial distribution with a 15:85 ratio of VP1:VP2. The peak was obtained by combining the binomial distribution for a 15:85 ratio (red histogram in Figure 1) with a Gaussian function representing the experimental resolution. Monitoring Fab Binding with CDMS: Binding Curves. Empty CPV capsids were incubated with one of the four fragment antibodies, and then the mass distributions of the resulting complexes were measured using CDMS to determine the number of bound Fabs. In these experiments, the concentration of the CPV capsid was held constant and the concentration of the Fab was increased until saturation was achieved. CDMS mass histograms measured for unreacted

■

RESULTS AND DISCUSSION Masses of Fabs and Empty CPV Capsid. The masses of the Fabs and the CPV capsid are required to interpret CDMS measurement of Fab binding to CPV. Sequence masses were not previously determined for the Fabs used in this work (14, B, E, and F), so the masses were measured by conventional LC-MS (Synapt G2, Waters). The results are given in Table 1. The CPV capsid is a T = 1 icosahedron containing 60 capsid proteins. Full (i.e., with DNA) and empty CPV particles are generated during infection, and those two capsid forms appear structurally equivalent when examined by X-ray crystallography.41 The studies reported here were performed with noninfectious empty particles. Three capsid proteins (VP1, VP2, and VP3) are present in the mature full particles. VP1 and VP2 are formed from different start codons and from alternatively spliced mRNAs, and the full sequence of VP2 is

Figure 1. CDMS spectrum measured for empty CPV particles. The black line and points show the measured spectrum plotted using 25 kDa bins. The red histogram shows the binomial distribution of masses expected for a capsid with a VP1/VP2 ratio of 15:85. The peak shown by the thin red line was obtained by combining the binomial distribution with a Gaussian function representing the experimental resolution. B

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society CPV and CPV with five Fab E concentrations ranging from 0.2 to 9.1 μM are shown in Figure 2. The peak center shifts to higher mass as the Fab E concentration increases. Addition of the Fab also causes the peak to broaden and then narrow. Peaks due to the additions of a specific number of Fabs are not resolved in Figure 2. However, the width of the measured mass distribution provides information about the distribution of Fab numbers bound at each concentration. To determine the number as well as the distribution of Fab E bound at each concentration, the measured peaks were fit with a series of Gaussians to account for the instrumental resolution and heterogeneity of the CPV capsid (see above). The width of the Gaussians was determined by fitting the width of the peak measured for the unreacted CPV. The Gaussians were stepped across the measured mass distributions by a mass equivalent to two Fabs. (We used a step of two Fabs because the peak for unreacted CPV has a full width at halfmaximum equivalent to around three Fabs.) The results of this fitting procedure are shown in Figure 3 for the five Fab E concentrations in Figure 2. The points in Figure 3 are the relative abundances of each Gaussian used to fit the measured peaks. These relative abundances were used to calculate the

Figure 3. Results of the fitting procedure used to analyze the measured mass distributions, accounting for instrumental resolution and the heterogeneity of the CPV capsid (see text). Representative results are shown for Fab E at the five concentrations shown in Figure 2. The points are the relative abundances of the Gaussians used to fit the measured mass distribution. These relative abundances were used to calculate the mean and distribution RMSDs that are plotted in Figures 4 and 5. The lines are Gaussians calculated using these quantities.

mean number of Fabs and the root-mean-square deviation (RMSD) of each mass distribution. The mean number bound was then plotted against Fab concentration to yield the binding curve for each Fab (Figures 4 and 5, A and B). The distribution RMSDs were also plotted against Fab concentration (Figures 4 and 5, C and D). According to cryo-EM studies, Fab E binding sites are roughly equidistant from the 2-, 3-, and 5-fold axes.14 Sixty Fab E are expected to bind per capsid at saturation.10 The Fab E binding curve is shown in Figure 4A. The points are the measurements and the line is a fit discussed below. Saturation occurs at around 60 Fabs. Fab B binds near the 3-fold axes of the capsid. There are expected to be 60 Fab B bound per capsid based on cryo-EM results.10 The Fab B binding curve is shown in Figure 4B. Saturation occurs at around 60 Fabs. Fab 14 binds close to the same site as Fab B, near the 3-fold axis. The Fab 14 binding curve is shown in Figure 5A. Saturation occurs with close to 60 Fab 14 bound. The Fab F binding site is roughly equidistant from the two, three and 5-fold axes. It is similar to and overlaps with the binding site for Fab E, The Fab F binding curve is shown in Figure 5B. In this case, saturation appears to occur with less than the expected 60 Fabs. Monitoring Fab Binding with CDMS: Distribution RMSDs. The plots of the distribution RMSDs against concentration for each Fab are shown in Figures 4 and 5, C and D. For Fab E, the distribution RMSD peaks during the portion of the binding curve where the number of Fabs bound is increasing rapidly, then decreases as the average number of Fabs bound approaches saturation. For Fabs B, 14, and F, the distribution RMSD increases rapidly as the number of Fabs bound increases, but the substantial narrowing of the distributions (seen for Fab E) is not observed for those Fabs. In those latter cases, the distribution RMSD remains relatively large as the Fab concentration is increased and saturation is approached. In the case of Fab 14 (Figure 5C), the distribution RMSD increases sharply at high Fab concentrations. This change is correlated with a small increase in the average number bound (Figure 5A). These observations

Figure 2. CDMS mass histograms measured for the reaction between CPV and Fab E. (A) Mass histogram of empty CPV with no Fab. The center of the mass distribution of unreacted CPV is indicated by the red dotted line at 4.06 MDa. B−F) are the CDMS histograms of CPV reacted with increasing concentration of Fab E: (B) 0.2 μM, (C) 0.5 μM, (D) 1.2 μM, (E) 3.4 μM, and (F) 9.1 μM. C

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

Figure 4. Binding curves and plots of distribution RMSDs for Fab E and Fab B reacting with CPV capsids. The binding curves (plots of the average number of Fabs bound versus initial Fab concentration) are shown in the upper panels. The lower panels show the distribution RMSDs plotted against Fab concentration. The results for Fab E are on the left of the figure, and those for Fab B are on the right. The points are the experimental measurements. The red lines show the prediction of the standard (Langmuir) model for ligand binding where all sites have the same intrinsic affinity. The blue lines show the predictions of a model where there are CPV capsid subpopulations with different intrinsic affinities. For the binding curve (upper plots) the blue line is beneath the red line.

Figure 5. Binding curves and plots of distribution RMSDs for Fab 14 and Fab F reacting with CPV capsids. The binding curves (plots of the average number of Fabs bound versus initial Fab concentration) are shown in the upper panels. The lower panels show the distribution RMSDs plotted against Fab concentration. The results for Fab 14 are on the left of the figure, and those for Fab F are on the right. The points are the experimental measurements. The red lines show the prediction of the standard (Langmuir) model for ligand binding where all sites have the same intrinsic affinity. The blue lines show the predictions of a model where there are CPV capsid subpopulations with different intrinsic affinities. For the binding curve (upper plots), the blue line is beneath the red line.

D

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

⟨m⟩ was calculated for each initial A concentration, and the results were fit to the measured binding curves using a leastsquares criterion. The fits are discussed below. This procedure yields optimized values of KA and n, the number of equivalent binding sites. For each initial A concentration, the optimized values of [VAm] provide the distributions of A that are bound to V. The RMSD of this distribution, σm, is given by

may indicate the onset of weaker, possibly nonspecific, binding of the Fab to the saturated capsid. Sources of Error. There are two main potential sources of error with the CDMS measurements described here. First, transfer of the virus−Fab complex through the electrospray interface into high vacuum could lead to collisional activation and when ligand binding is weak, the loss of some of the Fabs. The observation that the maximum number of bound Fabs is close to the expected number of epitopes on the T = 1 icosahedral capsid (60) for all Fabs suggests that loss of Fabs through collisional activation is not a significant source of error. Another potential source of error is the formation of nonspecific aggregates during electrospray. The large ions studied in this work are thought to be generated by the charge residue mechanism where the solvent evaporates from the electrospray droplet to leave behind a charged ion.47 If a droplet contains free Fabs in addition to a CPV(Fab)m complex, then the free Fabs and CPV(Fab)m complex could form a nonspecific aggregate as the droplet evaporates and the resulting CPV(Fab)m+xy+ ion would contain extra Fabs. An estimate of the average number of free Fabs present in a droplet can be obtained from the concentration and droplet size. The average size of the primary electrospray droplets can, in turn, be estimated from the electrospray conditions.48,49 For an estimated droplet size of 68 nm, the average number of free Fabs is 1.0 per droplet for a Fab concentration of 10 μM. Thus, nonspecific aggregation can only be responsible for the addition of a few extra Fabs at most. For Fab 14 the binding curve (Figure 5A) was measured for Fab concentrations well above the point where the number bound has leveled-off. If nonspecific aggregation was significant, the number bound would not level off. Potential errors in the CDMS mass measurements (which are relatively small) are discussed in the Materials and Methods Section below. Comparison with Predictions of the Standard Model for Ligand Binding. The absence of a lag phase in all the binding curves indicates that binding is noncooperative. The standard model for ligand binding to n identical and independent sites (i.e., noncooperative binding) describes a series of coupled equilibria: VA m − 1 + A V VA m

σm =

[VA m] n−m+1 KA = m [VA m − 1][A ]

(1)

Table 2. Parameters Determined from the Fit of the Standard Model to the Measured Binding Curvesa

(2)

where KA is the equilibrium constant for a single site and the other term is a statistical factor that results because there are n − m + 1 open sites for the forward reaction and any of the m adsorbed ligands can dissociate for the reverse reaction. Under the conditions used to perform the experiments, free A is not much larger than bound A, and so calculation of the equilibrium concentration [VAm] for given initial V and A concentrations leads to a series of coupled equations that were solved iteratively. The average number of A bound to V, ⟨m⟩, is then given by ⟨m⟩ =

∑ m[VA m] ∑ [VA m]

(4)

The calculated RMSDs for each initial A concentration will be compared to the values deduced from the CDMS measurements. The best fits of the standard model to the measured binding curves are shown by the red lines in Figures 4 and 5, A and B. In all cases, the standard model provides an excellent fit to the experimental results. Table 2 summarizes the parameters (n and KA) derived from the model. According to cryo-EM measurements and symmetry arguments, there are 60 equivalent sites for each Fab. For Fab E, the fit of the standard model optimized with 60 equivalent sites and for Fab 14 the best fit was obtained with 59. However, the fit assuming 60 equivalent sites for Fab 14 was only slightly worse (the RMSD was only 0.05% larger) and so the deviation from 60 is not significant. For Fab B, the best fit was obtained with 65 equivalent sites. The fit assuming 60 equivalent sites had an RMSD 50% larger than the fit with 65, so the deviation from 60 is significant here. For Fab F, the best fit was obtained with 56 equivalent sites and the deviation from 60 is significant again. The reduced number of equivalent sites for Fab F could be due to steric crowding reducing the affinity of the last few sites. While the binding curves are all well fit by the standard model, this is not true for the distribution RMSDs. The predictions of the standard model are shown by the red lines in Figures 4 and 5, C and D. In the case of Fab E, the standard model provides a good fit to the distribution RMSDs, except for the lower concentrations where the average number bound is increasing rapidly. On the other hand, for Fab B, 14, and F, the measured distribution RMSDs are much larger than predicted by the standard model.

where in this case V is the virus and A is the antibody. The equilibrium constant for the mth step is KA′ (m) =

∑ (m − ⟨m⟩)2 [VA m] ∑ [VA m]

Fab

n

KA

ΔG° (kJ/mol)

KD

binding site

Fab B Fab F

65 56

0.219 × 106 1.29 × 106

−30.5 −34.9

45.7 × 10−7 7.75 × 10−7

Fab E

60

2.38 × 106

−36.4

4.20 × 10−7

Fab 14

59

8.60 × 106

−39.6

1.16 × 10−7

3-fold axis equidistant from 2-, 3-, and 5-fold axes equidistant from 2-, 3-, and 5-fold axes 3-fold axis

a

n is the number of equivalent sites, and KA is the association constant for a single site. The association constant for the mth step incorporates a statistical factor (see eq 2). The free energy change for the association reaction is obtained from ΔG° = −RT ln KA. KD = 1/KA.

(3) E

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society Beyond the Standard Model: Capsids Are Identical but Binding Sites Are Not. The poor agreement between the predictions of the standard model and the distribution RMSDs for Fabs B, 14, and F led us to seek a modified approach that would provide a better match to the distribution RMSDs while simultaneously providing a good fit to the binding curves. In the standard model employed above, all binding sites are identical. This model is equivalent to the Langmuir adsorption isotherm used to model adsorption on surfaces. A reasonable next step would be to use a model where the sites are no longer equivalent. The Temkin adsorption isotherm was designed to account for adsorbate−adsorbate interactions.50,51 In this model, the heat of adsorption is coverage dependent. The equilibrium constant K’A(M) in eq 2 above can be rewritten as follows. n−m+1 n − m + 1 −ΔG / RT KA = e m m n − m + 1 ΔS / R −ΔH / RT e e = m

KA′ (m) =

(5)

The Tempkin adsorption model was designed to account for adsorbate−adsorbate interactions by making the heat of adsorption depend on the coverage. ΔHm = ΔH(1 − αθ )

(6)

where α is a constant and θ is the fractional coverage which is given by (m − 1)/n. Substituting into eq 5, n − m + 1 ΔS / R ΔH / RT(1 − αθ) e e m n − m + 1 ΔS / R −ΔH / RT αθ ΔH / RT e e e = m n−m+1 KA e βθ = m

KA′ (m) =

Figure 6. Comparison of the predictions of the Temkin model to the measured binding curve and distributions RMSDs. (A) Best fits of the Tempkin model to the binding curve for Fab B for a range of β values. The standard deviations of the fits for the different β values are given in the inset table. The lines in the plot are color coded to the β values in the table (purple −3.0, red −2.0, blue −1, green 0, and yellow +1). (B) Comparison of the measured and calculated distribution RMSDs. The lines are color coded as in (A).

(7)

where β is given by β=

αΔH RT

(8)

model where there are particle subpopulations with different binding affinities is a model with two subpopulations (high affinity and low affinity) with each subpopulation having n identical binding sites. In this case, we optimized the parameters of the model to simultaneously fit the binding curve and the distribution RMSDs, using a least-squares criterion. The best fits are shown as the blue lines in Figures 4 and 5. With this approach, it is possible to simultaneously obtain good fits to the binding curves and to the distribution RMSDs. In the case of the binding curves (Figures 4 and 5, A and B), the blue lines are not visible in most of the plots because they are completely under the red lines from the standard model. In the model described above, we assumed that there are two virus subpopulations with different binding affinities, and this approach provided a good fit to the experimental results. However, this does not necessarily mean that there are only two subpopulations, and more subpopulations may be revealed by higher resolution measurements. Table 2 shows a summary of the results from fitting the standard model to the binding curves. KA and n are shown for each Fab, along with the binding free energy determined from KA and KD (the reciprocal of KA). As discussed above, the number of available sites (n) is close to the expected value (60) in all cases. The KD values range from 0.1 to 4.5 μM. It

For β = 0, the Temkin model reduces to the Langmuir model. For β > 0, the binding energies increase with increasing coverage and for β < 0 the binding energies decrease with increasing coverage. Attempts to fit the Temkin model to the experimental data for Fab B are shown in Figure 6. Figure 6A shows fits to the binding curve. The inset table shows the standard deviations for the best fits as a function of β. The lines in the plot are color coded to the β values in the table (purple −3.0, red −2.0, blue −1, green 0, and yellow +1). The best fit is obtained with β = 0, and moving β away from zero in the positive or negative directions leads to poorer agreement between the predictions of the model and the experimental data. Figure 6B shows comparison between the model predictions and the distributions RMSDs. The lines in the plot are color coded to the β values in Figure 6A. The Temkin model does not significantly improve the agreement with the distribution RMSDs. Based on these results, we concluded that a model with heterogeneity in the binding sites (with all virus particles identical) could not simultaneously fit the binding curves and the distribution RMSDs. This led us to consider a model where there are particle subpopulations with different binding affinities. Beyond the Standard Model: Particle Subpopulations with Different Binding Affinities. The simplest F

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society should be noted that these are the KD values for a single site, and the equilibrium constant for the mth step is given by KD′ (m) =

m KD n−m+1

(9)

For n = 60, the statistical factor, m/(n − m + 1), ranges from 1/60 for the first step to 60 for the last. Comparison with Other Measurements. Most antibodies have KD values in the low micromolar (10−6 M) to nanomolar (10−7−10−9 M) range. The KD values reported in Table 2 fall into this range. The values found here indicate interactions at the weak end of the expected range, but this could be partly due to the fact antibodies have bivalent interactions versus monovalent for the Fabs.52 Another contributing factor may be that the CPV capsids are immature empty particles rather than mature virions. On the other hand, Nelson et al.13 have measured KD values for the binding of Fabs E, F, B, and 14 to CPV capsids using direct calibration ELISA.53 The measured KD values (1−6 nM) are much smaller than found here. The direct calibration ELISA method is a multistep measurement. The CPV capsids are immobilized in a well and exposed to a Fab solution. After a predetermined time the solution is removed and placed into another well with immobilized CPV. This process is repeated several times until all the Fab is adsorbed and then the absorbed Fabs are detected by means of a horse radish peroxidase-conjugated secondary antibody. Finally, the immobilized CPV concentration is determined by a Fab saturation analysis. Capsid Degradation and Effect on Fab Binding. Because the results described above showed evidence for capsid subpopulations with different binding affinities we were curious to see if the affinities changed with capsid age. To explore this question, we compared Fab binding to recently prepared CPV capsids and to several capsid samples that were purified more than 9 years ago, and that had been stored at 4 °C since that time. The mass distributions measured for the aged capsids are slightly broader than the recently prepared CPV capsids and shifted to slightly lower mass (by 18−51 kDa depending on the preparation). Supporting Information Figure S1 shows mass distributions measured for four aged CPV capsid samples. Based on a proteomics study of the posttranslational modifications (see below) the peak for the aged capsids should be shifted around 6 kDa higher in mass than the peak for the new capsids (assuming all aged samples have the same post translational modifications). Thus, the average masses of the aged capsids are around 24−57 kDa lower in mass than expected from the mass of the new capsids. This difference could be due to proteolysis or it could be due to the aged capsids containing slightly less VP1 and slightly more VP2 than the new capsids. If the mass shift is entirely due to changes in the VP1:VP2 ratio, then for the aged capsids this ratio ranged from around 12:88 to 9:91. These values are closer to the expected 10:90 ratio than the 15:85 ratio found for the newly prepared capsids (see above). The aged CPV capsid preparations were incubated with Fabs 14 and E, at relatively high Fab concentrations so that the capsids were expected to be close to saturation. After incubation with 19.2 μM Fab 14, the peak measured for the aged capsids was slightly lower in mass than the peak measured for the new capsids. However, the peak widths were similar. A representative example is shown in Figure 7A, where the red line shows the peak for new capsids and the black line shows the peak for aged capsids. The results for all four aged samples

Figure 7. Comparison of mass distributions measured after incubation of new and aged CPV capsids with Fab 14 (19.2 μM) and Fab E (5.7 μM). The black line shows the distributions measured for aged capsids, and the red line shows the results for new capids. Distributions for three additional aged samples are given in the Supporting Information.

are provided in Figure S2. The mass difference (which corresponds on average to around half a Fab) is small. Aging has little effect on the binding of Fab 14. On the other hand, incubation with 5.7 μM Fab E reveals a significant difference between aged and new capsids. A representative example is shown in Figure 7B and the results for all four aged samples are shown in Figure S3. In this case, the different aged samples show significantly different levels of Fab E binding; ranging from around 2 to 8 fewer Fabs than the new capsids. The peaks for the aged samples are also broader than the peak measured under identical conditions for the new capsids. The different levels of Fab binding for the aged capsids may indicate different levels of modification. The results described above indicate the binding of Fab E is influenced by capsid age while the binding of Fab 14 is not. Fab 14 binds near the 3-fold axis while Fab E binds nearly equidistant from the 2-, 3-, and 5-fold axes. To explore this difference further we used mass spectrometry to identify posttranslational modifications in the aged and new capsids. The capsids were digested using trypsin, desalted, and the masses of the resulting peptides determined using an Orbitrap mass spectrometer. We looked specifically for deamidated asparagine and oxidized methionine since the Fabs bind at the asparagine and methionine residues and modifications at these residues are expected to affect binding. We only considered post-translational modifications with relative abundances G

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society

variation between capsid subclasses or within the icosahedral structure. Our data suggest that the structural variation revealed by the antibody Fabs was localized to certain positions within the capsids, as the Fabs with different footprints were differently sensitive to the variation. Fab B and 14 both bind close to the 3-fold axes of the virus, while Fab E and F bind at overlapping sites roughly equidistant from the 2-, 3-, and 5-fold axes (see Table 2). Finally we note that there is evidence that Fab binding induces an allosteric change in the capsid that inhibits receptor binding.13,14 This raises the possibility that the structural variation revealed here could be induced in part by Fab binding. The newly prepared CPV capsids behave like a single homogeneous population when tested with Fab E, while the older capsids showed a consistent reduction in Fab E saturated binding suggesting that a portion of the aged capsids have sites with a reduced affinity toward Fab E. It appears to be unlikely that this is solely due to the post-translational modifications present in the aged capsids, because those were found to change in all positions in the capsids. This suggests that the heterogeneity results from other small structural differences within a portion of the Fab E footprint which is not present within the Fab14 footprint. Further analysis is required to identify the nature of the changes and how they affect antibody binding.

greater than 20%. Post-translational modifications that met these criteria are summarized in Table 3. There are more posttranslational modifications in the aged capsid than in the new. In particular, there is a relatively large number of oxidized methionine sites, which are often associated with protein degradation. There is an oxidized methionine at the Fab 14 binding site (M87). However, this modification is present in both the new and aged capsids. There are no significant differences in the post-translational modifications at the Fab 14 binding site, which is consistent with the relatively small difference in the binding affinities of the new and aged capsids. On the other hand, at the Fab E binding site M569 is 94% oxidized in the aged capsid, but not modified significantly in the new capsid. The M569 is within the footprint of the Fab E,14 so that the substantial difference in the binding affinities of the new and aged capsids found for Fab E could be due to the post-translational modification at that position.

■

CONCLUSIONS CDMS was used to probe the interactions between the canine parvovirus capsid and antigen binding fragments from a series of antibodies. The measurements provided information on the average number of Fabs bound to the CPV capsids and the distribution. A plot of the average number bound against the initial Fab concentration yields the binding curve. The binding curves could be fit using the standard model of ligand binding to obtain the association (or dissociation) constant and the number of binding sites. In the standard model, all binding sites are assumed to be identical. This model provided an excellent fit to the binding curves for all four Fabs examined in this work. However, the standard model could not account for the distribution of bound Fabs for three of the four Fabs examined. The measured distributions were much broader than predicted. Models where the binding sites are not identical (but all capsids are identical) could also not account for the distributions. However, a model with capsid subpopulations with different affinities could account for the distributions while also fitting the binding curves. These results demonstrate that the distribution of bound ligands is a much more sensitive probe of ligand binding than the binding curve alone. The nonequivalent binding of different Fabs is a novel finding, and difficult to explain from current structures, as those all use icosahedral averaging, and do not show any

■

Table 3. Post-Translational Modifications Found in New and Aged CPV Capsidsa % modified position

modification

new capsid

aged capsid

N25, N46, N47 N56, N64 M73 M87 M174, M183, M190 M319, M331 N466 M518 M569

1xDeamidated 1xDeamidated 1xOxidation 1xOxidation 3xOxidation 2xOxidation 1xDeamidated 1xOxidation 2xOxidation

100 94 99 100 0 0 74 0 0

100 97 90 92 80 95 0 95 94

MATERIALS AND METHODS

Preparation of CPV Capsids. Canine parvovirus (CPV) capsids were grown in Norden Laboratory Feline Kidney (NLFK) cells in 1:1 McCoy-Lebovitz 15 media with 5% fetal calf serum (FCS). The CPV stock was generated by transfecting NLFK cells with CPV infectious plasmid and passaging cell supernatants 3−4 times. The NLFK cells were seeded in 850 cm3 roller bottles (Corning) at 2 × 104 cells/cm3 and inoculated with 5 mL of CPV stock. After incubating at 37 °C for 4 days, 7.25 mL of 0.69 M Tris pH 8.7, 3.4% NP-40, and 0.19 mM EDTA was added to roller bottles. The cultures were frozen at −20 °C, thawed, and centrifuged at 12 700g for 30 min to remove cell debris. The resulting supernatant was incubated at 4 °C overnight with 4.25 mM polyethylene glycol 8000 (Fisher Scientific) and 0.5 M NaCl (Fisher Scientific), followed by centrifugation at 12 400g for 30 min in a Fiberlite F13 rotor (Thermo Scientific). The pellet was resuspended in 20−40 mL of 0.1% sarkosyl and 0.01 M Tris pH 7.5 and extracted with an equal volume of chloroform, prior to centrifugation at 11 800g for 15 min in a Fiberlite F21 rotor (Thermo Scientific). The aqueous fraction was layered over a 10 mL, 20% sucrose cushion and centrifuged at 30 000 rpm in a SW32Ti rotor (Beckman-Coulter) for 4 h at 12 °C. The pellet was resuspended in 1−2 mL 25 mM Tris, pH 7.4, layered over a 10− 40% sucrose gradient, and centrifuged at 30 000 rpm in a sw32Ti rotor for 4.5 h at 12 °C. Heavy (infectious, DNA-containing) and light (noninfectious, non-DNA containing) capsid bands were removed from the gradient. Empty capsids from the light band were used in these experiments. Some capsids were prepared as described in Agbandje et al.54 and stored in PBS with 0.05% sodium azide at 4 °C for over 9 years. Preparation of Antigen Binding Fragments (Fabs). Antibodies B, E, and F are rat IgG-2b, while antibody 14 is a mouse IgG-2a, and all were prepared and produced from hybridoma cells as described elsewhere55,56 Cells were cultured in DMEM (Dulbecco’s Minimal Essential Medium) with 10% FCS and nonessential amino acids in Lifecell gas permeable bags (Baxter Healthcare) at 37 °C, then cell supernatants were run through a HiTrap Protein G HP column (GE Healthcare) in 50 mM Tris and 150 mM NaCl pH 7.5, eluted with 0.1 M Citrate buffer pH 3.0, and immediately neutralized with 1 M Tris pH 9.0. Antibodies B and F were cleaved with papain and Fabs and purified as described in Nelson et al.13 Antibodies E and 14 were cleaved into Fabs by overnight incubation with immobilized

a

Modifications at the Fab 14 binding site are highlighted in bold, and modifications at the Fab E binding site are highlighted in italic (0 = not detected). H

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society papain (Fisher Scientific), according to the manufacturer’s protocol. The constant region (Fc) was separated from antibody 14 Fabs by incubating the cleaved protein with Protein A CL-4B Sepharose beads (Pharmacia Biotech) in 50 mM Tris pH 7.0. Antibody E constant regions were removed by running cleaved protein over a DEAESephadex A25 (Sigma-Aldrich) column in 0.01 M phosphate buffer pH 7.8 and collecting column flow-through. (Fab)2 and other incomplete cleavage products were removed via size exclusion chromatography over a Sephadex G100 column (Pharmacia) in PBS (phosphate buffered saline). Fabs were concentrated using an Amicon 10 kDa centrifugal filter (Millipore). Charge Detection Mass Spectrometry. Charge detection mass spectrometry is a single particle technique where the mass to charge ratio (m/z) and charge (z) are measured simultaneously and then multiplied to yield the mass of each ion. This process is repeated for thousands of ions, and the masses are binned to obtain a mass spectrum. The CDMS instrument, methods,57−59 and data analysis60,61 have been described in detail elsewhere. Briefly, ions are produced by a nanoelectrospray source (Advion Biosciences) and introduced into the instrument through a heated metal capillary. Ions pass through three differentially pumped regions containing an ion funnel, a hexapole, and a quadrupole. The nominal ion energy of 100 eV/z is set by the static potential applied to the hexapole. The ions are then focused into a dual hemispherical deflection energy analyzer (HDA). The HDA transmits ions within a narrow band of kinetic energies centered on 100 eV/z. The transmitted ions are focused into an electrostatic linear ion trap with a cylindrical charge detector tube at its center. The trap end-cap voltages are raised, and a trapping event occurs. After a user determined time (100 ms for these experiments), the end-caps are grounded and the trap is emptied. During a trapping event the ion oscillates back and forth through detector tube, inducing a periodic signal which is amplified, digitized and transferred to a computer for analysis. The digitized signals were analyzed by a Fortran program using fast Fourier transforms. Events where ions were not trapped for the whole trapping time were discarded. The uncertainty in the mass measured for each ion depends on the uncertainties in the m/z and charge determinations. Both are well characterized.60−62 In the case of the charge, the uncertainty is limited by electrical noise and depends on the square root of the trapping time. For a trapping time of 100 ms the uncertainty in the charge (RMSD) is around 1.3 e (elementary charges). The uncertainty in the m/z is limited mainly by the energy distribution of the ions. For the conditions employed here, the relative uncertainty (RMSD) in the m/ z is around 0.006. Combining the uncertainties in the charge and m/z leads to a relative uncertainty (RMSD) in the mass of around 0.009. Parvovirus Fab Reaction. Samples were received in PBS. Stock solutions were diluted in PBS, and reaction volumes were mixed and incubated at room temperature. The initial CPV capsid concentrations were 2.0 × 10−8 M for Fab E and Fab B, 4.0 × 10−8 M for Fab 14, and 7 × 10−9 M for Fab F. After a reaction time of 1 h, the mixtures were buffer exchanged using biospin size exclusion chromatography columns (Bio-Rad Laboratories) into 100 mM ammonium acetate and electrosprayed. Some of the reaction mixtures were monitored over a 24 h period and longer reaction times did not cause a change in the mass distributions, indicating that equilibrium had been established within 1 h. Measurements were performed with freshly prepared CPV capsids from four different preparations and with Fabs from two or three different preparations. The results obtained with the different preparations showed no significant systematic differences. The results from the different preparations were combined in Figures 4 and 5 and analyzed together. Post-Translational Modifications of Aged and New CPV Capsids. The aged and new CPV capsids were denatured using 8 M urea in 100 mM ammonium bicarbonate (Sigma-Aldrich). Denatured proteins were reduced with 4 mM tris(2-carboxyethyl)phosphine (Sigma-Aldrich) for 1 h at 56 °C, followed by alkylation using 8 mM iodoacetamide (Sigma-Aldrich) for 1 h at room temperature. Proteins were digested at 37 °C using trypsin (Promega) overnight. Peptides were desalted using Omix C18 micro pipet tips (Agilent) and

lyophilized to dryness. Desalted peptides were analyzed directly on an Orbitrap Fusion Lumos Mass Spectrometer (Thermo Scientific).

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/jacs.8b08050. Mass distributions for aged CPV capsids and binding of Fab 14 and Fab E to aged capsids (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*[email protected] *[email protected] ORCID

Martin F. Jarrold: 0000-0001-7084-176X Notes

The authors declare the following competing financial interest(s): The authors except M.F.J. declare no competing financial interest. M.F.J. is associated with a company that is developing charge detection mass spectrometry.

■

ACKNOWLEDGMENTS This material is based upon work supported by the National Science Foundation under Grant Number CHE-1531823 to M.F.J. and by NIH Grant R01 AI092571 to C.R.P. H.M.C. was supported by NSF-GRFP DGE-1650441.

■

REFERENCES

(1) Reading, S. A.; Dimmock, N. J. Neutralization of animal virus infectivity by antibody. Arch. Virol. 2007, 152, 1047−1059. (2) Halder, S.; Ng, R.; Agbandje-McKenna, M. Parvoviruses: Structure and infection. Future Virol. 2012, 7, 253−278. (3) Cotmore, S. F.; Tattersall, P. Parvoviruses: Small does not mean simple. Annu. Rev. Virol. 2014, 1, 517−537. (4) Kailasan, S.; Agbandje-McKenna, M.; Parrish, C. R. Parvovirus family conundrum: What makes a killer? Annu. Rev. Virol. 2015, 2, 425−450. (5) Cureton, D. K.; Harbison, C. E.; Cocucci, E.; Parrish, C. R.; Kirchhausen, T. Limited transferrin receptor clustering allows rapid diffusion of canine parvovirus into clathrin endocytic structures. J. Virol. 2012, 86, 5330−5340. (6) Palermo, L. M.; Hueffer, K.; Parrish, C. R. Residues in the apical domain of the feline and canine transferrin receptors control host specific binding and cell infection of canine and feline parvovirus. J. Virol. 2003, 77, 8915−8923. (7) Hueffer, K.; Parker, J. S.; Weichert, W. S.; Geisel, R. E.; Sgro, J. Y.; Parrish, C. R. The natural host range shift and subsequent evolution of canine parvovirus resulted from virus-specific binding to the canine transferrin receptor. J. Virol. 2003, 77, 1718−1726. (8) Parrish, C. R. Pathogenesis of feline panleukopenia virus and canine parvovirus. Bailliere Clin. Haematol. 1995, 8, 57−71. (9) Carmichael, L. E.; Joubert, J. C.; Pollock, R. V. A modified live canine parvovirus vaccine. II. Immune response. Cornell Vet. 1983, 73, 13−29. (10) Hafenstein, S.; Bowman, V. D.; Sun, T.; Nelson, C. D. S.; Palermo, L. M.; Chipman, P. R.; Battisti, A. J.; Parrish, C. R.; Rossmann, M. G. Structural comparison of different antibodies interaction with parvovirus capsids. J. Virol. 2009, 83, 5556−5566. (11) Tseng, Y. S.; Gurda, B. L.; Chipman, P.; McKenna, R.; Afione, S.; Chiorini, J. A.; Muzyczka, N.; Olson, N. H.; Baker, T. S.; Kleinschmidt, J.; Agbandje-McKenna, M. Adeno-associated virus serotype 1 (AAV1)- and AAV5-antibody complex structures reveal evolutionary commonalities in parvovirus antigenic reactivity. J. Virol. 2015, 89, 1794−1808.

I

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society (12) Tseng, Y. S.; Agbandje-McKenna, M. Mapping the AAV capsid host antibody response toward the development of second-generation gene delivery vectors. Front. Immunol. 2014, 5, 1−11. (13) Nelson, C. D. S; Palermo, L. M.; Hafenstein, S. L.; Parrish, C. R. Different mechanisms of antibody-mediated neutralization revealed using the Fab fragment of monoclonal antibodies. Virology 2007, 361, 283−293. (14) Organtini, L. J.; Lee, H.; Iketani, S.; Huang, K.; Ashley, R. E.; Makhov, A. M.; Conway, J. F.; Parrish, C. R.; Hafenstein, S. Nearatomic resolution structure of a highly neutralizing fab bound to canine parvovirus. J. Virol. 2016, 90, 9733−9742. (15) Clegg, R. M. Fluorescence resonance energy transfer. Curr. Opin. Biotechnol. 1995, 6, 103−110. (16) Thompson, N. L. In Techniques; Lakowicz, J. R., Ed.; Topics in Fluorescence Spectroscopy; Springer USA: New York, 1999; Vol. 1, pp 337−378. (17) Lagerkvist, A. C.; Foldes-Papp, Z.; Persson, M. A. A.; Rigler, R. Fluorescence correlation spectroscopy as a method for assessment of interactions between phage displaying antibodies and soluble antigen. Protein Sci. 2001, 10, 1522−1528. (18) Lieto, A. M.; Cush, R. C.; Thompson, N. L. Ligand-receptor kinetics measured by total internal reflection with fluorescence correlation spectroscopy. Biophys. J. 2003, 85, 3294−3302. (19) Vogel, S. S.; Thaler, C.; Koushik, S. V. Fanciful FRET. Sci. Signaling 2006, 2006, re2. (20) Ben-Johny, M.; Yue, D. N.; Yue, D. T. Detecting stoichiometry of macromolecular complexes in live cells using FRET. Nat. Commun. 2016, 7, 13709. (21) Green, R. J.; Frazier, R. A.; Shakesheff, K. M.; Davies, M. C.; Roberts, C. J.; Tendler, S. J. B. Surface plasmon resonance analysis of dynamic biological interactions with biomaterials. Biomaterials 2000, 21, 1823−1835. (22) Pattnaik, P. Surface plasmon resonance applications in understanding receptor-ligand interaction. Appl. Biochem. Biotechnol. 2005, 126, 79−92. (23) Dubs, M. C.; Altschuh, D.; Van Regenmortel, M. H. V. Interaction between viruses and monoclonal antibodies studied by surface plasmon resonance. Immunol. Lett. 1992, 31, 59−64. (24) Dimmock, N. J.; Hardy, S. A. Valency of antibody binding to virions and its determination by surface plasmon resonance. Rev. Med. Virol. 2004, 14, 123−135. (25) Gutiérrez-Aguirre, I.; Hodnik, V.; Glais, L.; Rupar, M.; Jacquot, E.; Anderluh, G.; Ravnikar, M. Surface plasmon resonance for monitoring the interaction of Potato virus Y with monoclonal antibodies. Anal. Biochem. 2014, 447, 74−81. (26) Damian, L. Isothermal titration calorimetry for studying protein-ligand interactions. Methods Mol. Biol. 2013, 1008, 103−118. (27) Fuerstenau, S. D.; Benner, W. H. Molecular weight determination of megadalton-DNA electrospray ions using charge detection mass spectrometry. Rapid Commun. Mass Spectrom. 1995, 9, 1528−1538. (28) Benner, W. H. A gated electrostatic ion trap to repetitiously measure the charge and m/z of large electrospray ions. Anal. Chem. 1997, 69, 4162−4168. (29) Mabbett, S. R.; Zilch, L. W.; Maze, J. T.; Smith, J. W.; Jarrold, M. F. Pulsed acceleration charge detection mass spectrometry: Application to weighing electrosprayed droplets. Anal. Chem. 2007, 79, 8431−8439. (30) Doussineau, T.; Kerleroux, M.; Dagany, X.; Clavier, C.; Barbaire, M.; Maurelli, J.; Antoine, R.; Dugourd, P. Charging megadalton poly(ethylene oxide)s by electrospray ionization. A charge detection mass spectrometry study. Rapid Commun. Mass Spectrom. 2011, 25, 617−623. (31) Doussineau, T.; Désert, A.; Lambert, O.; Taveau, J.-C.; Lansalot, M.; Dugourd, P.; Bourgeat-Lami, E.; Ravaine, S.; Duguet, E.; Antoine, R. Charge detection mass spectrometry for the characterization of mass and surface area of composite nanoparticles. J. Phys. Chem. C 2015, 119, 10844−10849.

(32) Doussineau, T.; Mathevon, C.; Altamura, L.; Vendrely, C.; Dugourd, P.; Forge, V.; Antoine, R. Mass determination of entire amyloid fibrils by using mass spectrometry. Angew. Chem., Int. Ed. 2016, 55, 2340−2344. (33) Elliott, A. G.; Harper, C. C.; Lin, H.-W.; Williams, E. R. Mass, mobility and MSn measurements of single ions using charge detection mass spectrometry. Analyst 2017, 142, 2760−2769. (34) Elliott, A. G.; Harper, C. C.; Lin, H.-W.; Susa, A. C.; Xia, Z.; Williams, E. R. Simultaneous measurements of mass and collisional cross-section of single ions with charge detection mass spectrometry. Anal. Chem. 2017, 89, 7701−7708. (35) Pierson, E. E.; Keifer, D. Z.; Selzer, L.; Lee, L. S.; Contino, N. C.; Wang, J. C.-Y.; Zlotnick, A.; Jarrold, M. F. Detection of late intermediates in virus capsid assembly by charge detection mass spectrometry. J. Am. Chem. Soc. 2014, 136, 3536−3541. (36) Lutomski, C. A.; Lyktey, N. A.; Zhao, Z.; Pierson, E. E.; Zlotnick, A.; Jarrold, M. F. HBV capsid completion occurs through error correction. J. Am. Chem. Soc. 2017, 139, 16932−16938. (37) Keifer, D. Z.; Motwani, T.; Teschke, C. M.; Jarrold, M. F. Measurement of the accurate mass of a 50 MDa infectious virus. Rapid Commun. Mass Spectrom. 2016, 30, 1957−1962. (38) Motwani, T.; Lokareddy, R. K.; Dunbar, C. A.; Cortines, J. R.; Jarrold, M. F.; Cingolani, G.; Teschke, C. M. A viral scaffolding protein triggers portal ring oligomerization and incorporation during procapsid assembly. Sci. Adv. 2017, 3, e1700423. (39) Pierson, E. E.; Keifer, D. Z.; Asokan, A.; Jarrold, M. F. Resolving adeno-associated viral particle diversity with charge detection mass spectrometry. Anal. Chem. 2016, 88, 6718−6725. (40) Chen, C.; Wang, J. C.-Y.; Pierson, E. E.; Keifer, D. Z.; Delaleau, M.; Gallucci, L.; Cazenave, C.; Kann, M.; Jarrold, M. F.; Zlotnick, A. Importin β can bind hepatitis B virus core protein and empty core-like particles and induce structural changes. PLoS Pathog. 2016, 12, e1005802. (41) Wu, H.; Rossmann, M. G. The canine parvovirus empty capsid structure. J. Mol. Biol. 1993, 233, 231−244. (42) Paradiso, P. R.; Rhode, S. L.; Singer, I. I. Canine parvovirus: A biochemical and ultrastructural characterization. J. Gen. Virol. 1982, 62, 113−125. (43) Tsao, J.; Chapman, M. S.; Agbandje, M.; Keller, W.; Smith, K.; Wu, H.; Luo, M.; Smith, T. J.; Rossmann, M. G.; Compans, R. W.; Parrish, C. R. The three-dimensional structure of canine parvovirus and its functional implications. Science 1991, 251, 1456−1464. (44) Xie, Q.; Chapman, M. S. Canine parvovirus capsid structure, analyzed at 2.9 Å resolution. J. Mol. Biol. 1996, 264, 497−520. (45) Cotmore, S. F.; Tattersall, P. The autonomously replicating parvoviruses of vertebrates. Adv. Virus Res. 1987, 33, 91−174. (46) Vihinen-Ranta, M.; Wang, D.; Weichert, W. S.; Parrish, C. R. The VP1 N-terminal sequence of canine parvovirus affects nuclear transport of capsids and efficient cell infection. J. Virol. 2002, 76, 1884−1891. (47) Fernandez de la Mora, J. Electrospray ionization of large multiply charged species proceeds via Dole’s charged residue mechanism. Anal. Chim. Acta 2000, 406, 93−104. (48) de Juan, L.; de la Mora, J. F. Charge and size distributions of electrospray drops. J. Colloid Interface Sci. 1997, 186, 280−293. (49) Hartman, R. P. A.; Brunner, D. J.; Camelot, D. M. A.; Marijnissen, J. C. M.; Scarlett, B. Jet Break-Up in Electrohydrodynamic atomization in the cone-jet mode. J. Aerosol Sci. 2000, 31, 65−95. (50) Brunauer, S.; Love, K. S.; Keenan, R. G. Adsorption of nitrogen and the mechanism of ammonia decomposition over iron catalysts. J. Am. Chem. Soc. 1942, 64, 751−758. (51) Masel, R. I. Principles of Adsorption and Reaction on Solid Surfaces; John Wiley & Sons: New York, 1996. (52) Yuan, W.; Parrish, C. R. Comparison of two single-chain antibodies that neutralize canine parvovirus: Analysis of an antibodycombining site and mechanisms of neutralization. Virology 2000, 269, 471−480. J

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Article

Journal of the American Chemical Society (53) Fuchs, H.; Orberger, G.; Tauber, R.; Geβner, R. Direct calibration Elisa: A rapid method for the simplified determination of association constants of unlabeled biological molecules. J. Immunol. Methods 1995, 188, 197−208. (54) Agbandje, M.; McKenna, R.; Rossmann, M. G.; Strassheim, M. L.; Parrish, C. R. Structure determination of feline panleukopenia virus empty particles. Proteins: Struct., Funct., Genet. 1993, 16, 155− 171. (55) Parrish, C. R.; Carmichael, L. E. Antigenic structure and variation of canine parvovirus type-2, feline panleukopenia virus, and mink enteritis virus. Virology 1983, 129, 401−414. (56) Parrish, C. R.; Carmichael, L. E.; Antczak, D. F. Antigenic relationships between canine parvovirus type 2, feline panleukopenia virus and mink enteritis virus using conventional antisera and monoclonal antibodies. Arch. Virol. 1982, 72, 267−278. (57) Contino, N. C.; Jarrold, M. F. Charge detection mass spectrometry for single ions with a limit of detection of 30 charges. Int. J. Mass Spectrom. 2013, 345, 153−159. (58) Contino, N. C.; Pierson, E. E.; Keifer, D. Z.; Jarrold, M. F. Charge detection mass spectrometry with resolved charge states. J. Am. Soc. Mass Spectrom. 2013, 24, 101−108. (59) Pierson, E. E.; Keifer, D. Z.; Contino, N. C.; Jarrold, M. F. Probing higher order multimers of pyruvate kinase with charge detection mass spectrometry. Int. J. Mass Spectrom. 2013, 337, 50−56. (60) Pierson, E. E.; Contino, N. C.; Keifer, D. Z.; Jarrold, M. F. Charge detection mass spectrometry for single ions with an uncertainty in the charge measurement of 0.65 e. J. Am. Soc. Mass Spectrom. 2015, 26, 1213−1220. (61) Keifer, D. Z.; Shinholt, D. L.; Jarrold, M. F. Charge detection mass spectrometry with almost perfect charge accuracy. Anal. Chem. 2015, 87, 10330−10337. (62) Hogan, J. A.; Jarrold, M. F. Optimized electrostatic linear ion trap for charge detection mass spectrometry. J. Am. Soc. Mass Spectrom. 2018, 29, 2086−2095.

K

DOI: 10.1021/jacs.8b08050 J. Am. Chem. Soc. XXXX, XXX, XXX−XXX