Two AdC68 constructs were manufactured within Pfizer. AdC68 material was formulated in A195 buffer (Xiang et al., 2002): 10 mM Tris, 75 mM NaCl, 5% w/v sucrose, 0.02% (w/v) 80 polysorbate 80, 1.0mM MgCl2, 0.1 mM EDTA, 0.5% (v/v) ethanol, 10 mM L-histidine, pH 7.4 (Mullins et al., 2021).

An accelerated stability study was performed in which adenovirus material was incubated at 26.5°C for up to 12 weeks. Throughout product development, these conditions were demonstrated to yield significant product degradation. A bulk AdC68 material was incubated at 26.5°C, and incremental time points were pulled from the bulk material and frozen at -80°C prior to analysis.

An adenoviral infectivity method was developed to monitor infectivity. For this assay, cultures of HEK293 cells (Cell biolabs, Catalog No. AD-100) were grown in a flask containing Dulbecco’s modified medium (Gibco, Catalog No. 11995) supplemented with 10% Heat Inactivated Fetal Bovine Serum (Gibco, Catalog No. 10082) and 1x Penicillin-Streptomycin (Gibco, Catalog No. 15140) in a 5% CO₂ atmosphere at 37°C. HEK293 cells were seeded into a 24-well flat bottom cell culture plate at 2.2E5 cells/well and then incubated in a 37°C, 5% CO₂ incubator. A dilution series of adenovirus sample was prepared in cell culture medium. The dilution series was then added to the wells of the cell assay plate about 1 h after cell seeding. The assay plate was then incubated for 2 days. The cells were then fixed with cold methanol at -20°C and then stained with a polyclonal antibody against adenovirus Hexon proteins, which is expressed upon infection of the cells with adenovirus sample. Positive stained cells in assay wells were counted under microscope fitted with a ×10 lens. Cells per view field were recorded and used to calculate the adenovirus sample infectivity, reported as infectious units (IFU)/mL as well as viral particles (vp)/IFU.

Select site-specific mutagenesis to the AdC68 construct was performed to investigate the impact of capsid deamidation. The mutation to the construct was performed by the company SGI-DNA. Mutations of N76, N414 and N76 & N414 on the hexon protein were generated, converting the DNA to encode aspartic acid from the original asparagine. Materials were manufactured and purified by Viraquest Inc., North Liberty, IA.

Multiple recombinant AdC68 constructs were designed during development with two selected to move forward into clinical trials (heretofore named AdC68 #1 and #2). In preparation for clinical manufacturing, infectivity and an AEX-HPLC purity assays (see Materials and Methods) were developed and qualified as part of the overall analytical control strategy. These tests were then utilized to evaluate samples from the 12-weeks accelerated stability study at 26.5°C to demonstrate these assays were stability indicating. Tables 1, 2 show that the infectivity of both developmental drug substance constructs decreased with a corresponding loss of purity and viral particle titer. Furthermore, there was a concomitant increase in the relative retention time, indicative of an acidic shift, of the viral particle peak. All changes observed in the data herein were significantly greater than the method precision results obtained during the method qualification (data not shown), indicating real changes in each attribute. A similar observation of an AEX-HPLC acidic shift for human Ad5 has been observed and investigated previously (Blanche et al., 2001).

The data from Table 1 can be readily visualized by inspecting overlaid chromatograms (Figure 1) from which the associated AdC68 #1 attribute data was calculated (Table 1). Sequential time points show a drop in the presumed AdC68 (“main”) peak area corresponding to the loss of viral particle titer. The increase in the early eluting “impurity” peak area, along with the reduction in the “main” peak area, corresponds to a decrease in the overall purity. The rightward trajectory of the main AdC68 (“main”) peak across the accelerated stability time course depicts the increase in retention time, which is a surrogate measurement of increased acidic surface charge. While cataloguing the “main” peak as the AdC68 product of interest for viral particle titer determinations (Table 1) was initially an anecdotal identification, additional characterization was used to confirm the “main” and “impurity” peak identities.

To characterize the “main” and “impurity” peak identities, 12-weeks degraded material was assessed. In Figure 2A, the unfractionated sample was shown to contain ample amounts of both “impurity” and “main” peaks in the analytical AEX-HPLC method. A modified, preparative AEX-HPLC method (see Materials and Methods) was then used to collect fractions of the “main” and “impurity” peaks. After several cycles of fractionation, like fractions were pooled together, concentrated and re-evaluated with the analytical AEX-HPLC method to ensure that the isolated peaks were in fact purified representations of “impurity” and “main” peaks (Figures 2B,C).

These fractions were then evaluated to determine their protein composition. Using a RP-HPLC/MS method similar to that described previously (Takahashi et al., 2006), the individual proteins within the samples were separated from each other for identification. The RP-HPLC/UV data demonstrated that the “impurity” peak contained primarily a single protein (Figure 2D), whereas the “main” peak contained many proteins and was representative of the chromatographic profile expected for adenovirus material (Figure 2E). The proteins from the “impurity” and “main” peaks were further examined by MS, which was directly coupled to the RP-HPLC separation. MS data confirmed that the “impurity” peak was comprised almost entirely of hexon protein, while the “main” peak contained all the proteins expected in a AdC68 viral particle, including hexon (Figures 2D,E) (Rux and Burnett, 2004; Russell, 2009; Nemerow et al., 2012; Ahi and Mittal, 2016). These observations were also consistent with an additional AdC68 construct (Supplementary Figure S1). The masses of hexon from both the main and impurity peaks are provided in Table 3. The observed masses of hexon for both the “main” and “impurity” peaks (104,844.8–104,847.7 Da) were abnormally high, especially when compared to hexon from an unstressed sample (104,842.1 Da), possibly indicative of deamidation of the protein.

RP-HPLC/MS is a denaturing analysis such that secondary, tertiary or quaternary structure of the proteins/complex are not retained. While the RP-HPLC/MS data demonstrated the presence of monomeric hexon, hexon is known to form noncovalent trimers on the adenovirus surface (Pichla-Gollon et al., 2007). To determine if the impurity, now identified as hexon, had such high-order structure, nSEC-HPLC/MS was implemented. The data clearly showed that the non-denatured, fraction collected impurity had a mass consistent with hexon trimer (Table 3). The higher-than-expected mass defect is likely explained by a combination of multiple deamidations, poor desolvation, and challenges with deconvolution at higher m/z.

With the identification of the major impurity concluded, the cause of the acidic shift was investigated by peptide mapping MS experiments. RP-HPLC-MS/MS data from the unfractionated sample time points in the accelerated stability study were collected and searched against various chemical modifications, including deamidation. While several modifications were observed across the entire adenoviral proteome, four hexon deamidation sites were evaluated extensively, including two that changed to a significant degree. Over the time-course, N76 and N414 reached almost 70 and 90% deamidation, respectively (Figure 3A). Conversely, many other sites of deamidation were observed but did not change significantly, including hexon N486 and N533 (Figure 3A).

The relative retention time shift from AEX-HPLC and the percent deamidation of N414 by RP-HPLC-MS/MS were plotted on the same graph and fitted to a single exponential equation to assess the relationship between deamidation and the relative retention time shift. For AdC68 #1, the generated curves and associated rate constants align excellently with each other (Figure 3B). Both curves showed good correlation with a single exponential equation indicating a single kinetic process and both rate constants were within error of matching. These observations were also consistent with an additional AdC68 construct (Supplementary Figure S2).

The pooled “impurity” and “main” peaks, along with the unfractionated sample from the 0 and 12-weeks timepoints, were also characterized by RP-HPLC-MS/MS. Specifically, the purpose of this study was to quantitate the deamidation sites of interest established in Figure 3A to ensure that deamidation events were not isolated to the “impurity” peak but also occurred on the intact adenoviral capsid prior to degradation. The data showed similar levels of N414 deamidation across the 12-weeks unfractionated sample and the “main” and “impurity” fractions (Figure 4). The data show somewhat higher levels of N76 deamidation in the “impurity” fraction compared to the “main” fraction (Figure 4). In both cases, the level of deamidation in the “main” fraction and the unfractionated sample at the 12-weeks accelerated timepoint were significantly higher than the 0-weeks timepoint. These observations were also consistent with an additional AdC68 construct (data not shown).

To conclude the deamidation inquiry, three hexon mutants were constructed: N414D and N76D single mutants and the N414D/N76D double mutant. These would mimic 100% deamidation at either or both sites for the asparagine residues that showed appreciable increases in deamidation upon accelerated stability. N414D was able to replicate and produce viable, infectious virus (data not shown). To assess the impact of the deamidation mutant on the relative retention time, the N414D mutant was tested by the analytical AEX-HPLC method, alongside the samples from the accelerated stability study (Figure 5). Upon analysis, the modified capsid eluted at a late relative retention time, consistent with the expectation that deamidation causes the AEX-HPLC retention shift in the non-mutant AdC68 accelerated stability samples (Figure 5). In contrast, the N76D single and N414D/N76D double mutant could not produce virus at any appreciable level. As such, the impact to infectivity and relative retention time shift could not be assessed.

As observed in Table 1 and Figure 1, viral particle concentration was reduced after accelerated stability conditions, as reflected by the drop in signal in the AEX-HPLC chromatogram. However, based on the magnitude of the particle concentration change vs. the magnitude of the decrease in infectivity, the change in particle concentration alone does not account for the decrease in infectivity. This is further demonstrated in the VP/IFU (Table 1) column, which adjusts the infectivity response to the particles dosed. The readout shows substantial reduction in per particle infectivity for the stressed samples.

Similarly, Table 1 and Figure 1 demonstrate a decrease in particle purity after accelerated stability conditions, as reflected by the increase in the “impurity” peak. As with particle concentration, the reduced purity alone would not account for the magnitude of reduced infectivity in the stressed samples. That said, the reduction in particle purity observed by AEX-HPLC does provide an avenue to investigate the mechanism of the viral particle concentration decrease, as the concentration decrease is likely linked to the increase of the impurity.

To understand the identity of the impurity, fractions were collected from the “impurity” and “main” peak observed by the analytical AEX-HPLC method after 12 weeks at 26.5°C. The profile from the unfractionated sample is shown in Figure 2A. A modified, preparative AEX-HPLC method was used for fraction collection to enable more concentrated fractions. Efficient enrichment of the “impurity” and “main” peaks were confirmed with the analytical AEX-HPLC method (Figures 2B,C). After successful enrichment, the “impurity” and “main” fractions were characterized by RP-HPLC/MS to identify differences between the fractions. Interestingly, the “impurity” fraction consisted primarily of a single protein (Figure 2D), which was identified as hexon by RP-HPLC/MS (Table 3). The “main” peak contained a profile indicative of adenovirus with many capsid proteins (Figure 2E), including hexon. RP-HPLC/MS was also used to confirm the identities of the capsid proteins from the “main” peak in an unstressed material, with the mass of hexon shown in Table 3. When compared with a representative unstressed and unfractionated sample, the mass of the “main” and “impurity” fractions from the sample subjected to 12 weeks at 26.5°C had higher observed masses by approximately 2–5 Daltons. This is possibly indicative of several asparagine deamidations, which is frequently observed in accelerated stability studies.

While the “impurity” was identified as hexon, we were interested in whether that “impurity” was monomeric or trimeric hexon, as it is noted that hexon exists as a trimer in the assembled adenovirus capsid (Pichla-Gollon et al., 2007). To assess this, the impurity peaks were evaluated by native SEC-MS, with a gradient that consisted of non-organic mobile phases and gentle MS source parameters directed at keeping non-covalent complexes assembled. The resulting spectra resulted in deconvolved masses between 314,545.2 and 314,559.7 Da, approximately three times the mass of the hexon in the RP-HPLC/MS data, indicating the impurity observed by AEX-HPLC is in fact hexon trimer (Table 3). A comparison of the observed to theoretical mass (314,526.0 Da) represents a large mass defect, but this defect can be explained by several sources. The theoretical mass does not consider hexon deamidation. As shown in the RP-HPLC/MS data, the observed hexon in the “impurity” and “main” peaks likely has several sites of deamidation present. If the monomeric hexon has on average 3 deamidations present on the capsid, the theoretical mass for the average trimer would be 9 Da higher than the theoretical mass listed in Table 3. Furthermore, deconvolution of spectra from native SEC-MS is challenging due to a limited number of charge states, higher m/z measurements, and poorer desolvation, often resulting in larger than expected masses. Given challenges in characterizing native spectra, and the fact that we know there is significant deamidation present on the hexon proteins in the 12-weeks accelerated stability sample, we feel that the observed mass is indeed reflective of hexon trimer.

As the change in purity and particle titer have been accounted for, the cause of the relative retention time shift was then explored. In previous work on human adenovirus (Blanche et al., 2001), a similar shift was observed and was isolated to the deamidation of four accessible hexon asparagine-glycine (NG) residue pairs. Such NG pairs are known hotspots for deamidation (Geiger and Clarke, 1987). The recombinant AdC68 constructs studied in the current manuscript had 3 such NG pairs in its hexon primary sequence: N414, N486 & N533, two of which align with the human adenovirus structure. Instead of focusing the search on just these residues, MS/MS data were evaluated in a data dependent manner to see if other sites of modifications were present and changing. Interestingly, N76 which has an asparagine-threonine (NT) pair, showed a deamidation propensity like N414, albeit at a slightly slower rate. The deamidation levels of N76, N414, N486 and N533 are plotted in Figure 3A. Overall, the high level of hexon deamidation aligns with the hypothesis of deamidation from the RP-HPLC/MS data at the intact capsid protein level. Furthermore, N414 deamidation was contrasted with the relative retention time, and the curves showed good correlation with a single exponential equation indicating a single kinetic process and similar rate constants (Figure 3B). N414 is at the external surface of the hexon trimer (Figure 6) and could be directly responsible for the acidic shift in the surface charge observed by AEX-HPLC.

To confirm this hypothesis, site-directed mutagenesis of N414 to D414 was performed, which mimics 100% deamidation at the N414 site. As expected, when the N414D mutant was run by AEX-HPLC, alongside the accelerated stability samples, the relative retention time was in line with the expected result based on the single exponential equation (Figure 5). While the relative retention time shift was explained by deamidation at N414, this mutation did not impact infectivity.

While deamidation at N76 is not unexpected based on the NT amino acid motif, structural analysis (Figure 6) shows that the N76 residue exists at the outer surface of the hexon trimer in a region that is buried on the viral particle surface (Reddy, 2017; Matteson et al., 2018). As such, we wanted to ensure that the deamidation was not observed only on the “impurity” fraction, which represents free hexon trimer. Therefore, we digested both the unfractionated 0-weeks and 12-weeks accelerated stability samples in addition to the “impurity” and “main” fractions and assessed the level of deamidation for both N76 and N414 (Figure 4). There is significant evidence that both N76 and N414 deamidation occurs on the “main” peak, which represents the assembled adenovirus capsid, as evidenced by the increase in deamidation compared to the 0-weeks sample. Furthermore, N414 was at roughly equivalent levels across all samples. While N76 deamidation was moderately more prominent in the “impurity” fraction compared to the “main” fraction, the abundance of deamidation in both fractions were still significantly larger than the 0-weeks sample. Thus, while moderate N76 deamidation may occur after particle degradation, there is still evidence that significant N76 deamidation occurs on assembled adenovirus capsid.

Site-directed mutagenesis of N76 to D76, as a mimic for the 100% deamidated residue was performed to investigate the impact of N76 deamidation on the relative retention time shift and infectivity. However, we were unable to form assembled particles after making this mutation. Collectively, evidence of deamidation at N76 at higher rates in the “impurity” peak and the inability to form capsids after site directed mutagenesis of N76 to D76, suggests that this modification may be deleterious to assembled capsids. Whatever the mechanism for its alteration, deamidation of N76 seems to be detrimental to the structural integrity of the viral capsid and might be a site of interest in future molecular designs to stabilize the product for long-term storage.