ABSTRACT
Human bocavirus 1 (HBoV1) is a parvovirus that gathers increasing attention due to its pleiotropic role as a pathogen and emerging vector for human gene therapy. Curiously, albeit a large variety of HBoV1 capsid variants has been isolated from human samples, only one has been studied as a gene transfer vector to date. Here, we analyzed a cohort of HBoV1-positive samples and managed to PCR amplify and sequence 29 distinct HBoV1 capsid variants. These differed from the originally reported HBoV1 reference strain in 32 nucleotides or four amino acids, including a frequent change of threonine to serine at position 590. Interestingly, this T590S mutation was associated with lower viral loads in infected patients. Analysis of the time course of infection in two patients for up to 15 weeks revealed a gradual accumulation of T590S, concurrent with drops in viral loads. Surprisingly, in a recombinant vector context, T590S was beneficial and significantly increased titers compared to that of T590 variants but had no major impact on their transduction ability or immunoreactivity. Additional targeted mutations in the HBoV1 capsid identified several residues that are critical for transduction, capsid assembly, or DNA packaging. Our new findings on the phylogeny, infectivity, and immunoreactivity of HBoV1 capsid variants improve our understanding of bocaviral biology and suggest strategies to enhance HBoV1 gene transfer vectors.
IMPORTANCE The family of Parvoviridae comprises a wide variety of members that exhibit a unique biology and that are concurrently highly interesting as a scaffold for the development of human gene therapy vectors. A most notable example is human bocavirus 1 (HBoV1), which we and others have recently harnessed to cross-package and deliver recombinant genomes derived from another parvovirus, the adeno-associated virus (AAV). Here, we expanded the repertoire of known HBoV1 variants by cloning 29 distinct HBoV1 capsid sequences from primary human samples and by analyzing their properties as AAV/HBoV1 gene transfer vectors. This led to our discovery of a mutational hot spot at HBoV1 capsid position 590 that accumulated in two patients during natural infection and that lowers viral loads but increases vector yields. Thereby, our study expands our current understanding of HBoV1 biology in infected human subjects and concomitantly provides avenues to improve AAV/HBoV1 gene transfer vectors.
INTRODUCTION
Parvoviruses are small nonenveloped viruses that package a single-stranded (ss)DNA genome of ∼5 to 6 kb. This genome contains two main open reading frames (ORFs) that comprise the nonstructural (ns) and the capsid genes (cap or vp). Bocaviruses (BoV), which belong to the autonomous parvoviruses, harbor an additional unique ORF that encodes the nucleophosphoprotein 1 (NP1). Intriguingly, a series of recent reports implies that ssDNA viruses, including parvoviruses, may evolve more rapidly than anticipated, evidenced by measurements of high nucleotide substitution rates (10−3 to 10−6 substitutions/site/year) that are comparable to the rate of RNA virus counterparts (1, 2). For example, high rates of 1 × 10−4 substitutions/site/year were inferred for some autonomous parvoviruses such as the carnivore parvoviruses (3, 4), human parvovirus B19 (5), and porcine parvovirus (6). Moreover, several studies have estimated similar rates of both structural and nonstructural parvovirus gene evolution. For example, in human bocavirus 1 (HBoV1), the np1 ORF shows the highest rate of mutations among the ns genes that correlates with significant changes in viral titer (7). This could be a result of the multiple roles of NP1 in viral replication (8) and capsid protein expression (9), in addition to its immunomodulatory function (10). Changes in the C-terminal part of the ns1 ORF were shown to directly influence the role of NP1 in viral genome replication (8), which underlines the importance of a tightly regulated coevolution of the nonstructural genes. Viral titers are also influenced by mutations in the structural vp gene, especially in the VP1u region that is vital for the infectivity of the virus (7, 11).
The parvoviral capsid is an important determinant of virus tropism, host range, and reaction to the immune system. It was shown that even small amino acid (aa) changes in the vp ORF can largely alter virus-cell interactions, including cell-type preference. For instance, adeno-associated virus type 1 (AAV1) and AAV6 share 99% aa identity but exhibit a distinct polarity bias in primary airway epithelia (pHAE) (12) as well as different abilities to transduce human and mouse hematopoietic stem cells (13). Even amino acid changes in the VP proteins within one AAV serotype can drastically alter viral features, as exemplified by the two subtypes of AAV3, AAV3a and AAV3b, which differ by only 6 aa but display distinct affinities to heparin (14, 15). Another example is the pair of rodent parvoviruses minute virus of mice (MVM)p and MVMi that share 97% sequence identity but differ in their in vitro (16) and in vivo (17) cell tropisms. Similarly, the canine parvovirus (CPV)-2 and CPV-2a strains differ only in four amino acids, which, however, lead to the extended feline tropism of CPV-2a (18).
This rich repertoire of parvoviruses with distinct properties has drawn enormous interest to their potential use as gene transfer vectors in cancer and gene therapy. In the recombinant genomes of these vectors, either all viral sequences or parts thereof are replaced by transgenes of interest. To package the recombinant genomes, the missing viral elements have to be supplied in trans during vector production (19–22). In particular, AAV has emerged as a promising viral vector following its extensive study for more than 5 decades. A major reason for its popularity is the feasibility to pseudotype recombinant AAV2 genomes with AAV capsids from other natural serotypes or synthetic variants, which allows for transgene delivery to different target organs (23). Another recent example of interesting parvoviral vectors are chimeric recombinant AAV (rAAV)/BoV vectors in which an rAAV genome is pseudotyped with the capsid proteins from primate BoV (22, 24). One of these BoV variants, HBoV1, has a unique tropism for the airways and has been utilized as gene delivery vehicle in vitro and in vivo (25). Curiously, despite the wealth of HBoV1 vp sequences that were isolated in many areas of the world, only one particular HBoV1 variant (GenBank GQ925675) has so far been used as viral vector.
Accordingly, we aimed to study whether and how naturally occurring variations in the HBoV1 vp sequence affect properties of the virus. To this end, we constructed a battery of new pseudotyped viral vectors from HBoV1 variants that were de novo isolated from patient samples or described in previous studies. Next, we packaged a Gaussia luciferase (Gluc)-encoding rAAV genome into each of the various HBoV1 capsids. This allowed us to characterize the effects of naturally occurring single-point mutations (“singletons”) on viral DNA packaging, transduction, and immunological reactivity with anti-HBoV1 antibodies. The results of our work reveal interesting and previously underappreciated aspects of HBoV1 biology and have important implications for the choice and use of rAAV/HBoV1 vectors in future gene therapy applications.
RESULTS
Analysis of capsid DNA and protein sequence diversity in naturally occurring HBoV1 variants.To analyze the natural sequence diversity of HBoV1 capsid genes and proteins, we collected a total of 64 samples from patients at the university hospitals in Heidelberg and Cologne (both in Germany) who had previously tested positive for HBoV1. These samples comprised tracheal secretions, aspirates, pharyngeal washes, and sputum as well as bronchioalveolar lavage samples collected from children and adults in the years 2014 to 2016. From these 64 samples, we were able to PCR amplify and sequence the entire HBoV1 capsid-coding region (2,016 bp) in 29 samples, i.e., in 45.3% of all samples (exemplified in Fig. 1A; full DNA sequences are shown in Data Set S1 in the supplemental material). Typically, failure to amplify or fully sequence the capsid gene correlated with low viral titers in the original sample, below 1 × 106 viral genomes (vg) per ml.
Isolation and sequencing of natural HBoV1 capsid variants. (A) Representative agarose gel showing the results of PCR amplification of HBoV1 capsid genes in patient samples. The expected size of the full-length PCR amplicon was 2,016 bp. POS, positive control (HBoV1 helper plasmid); NEG, negative control (H2O). Nucleotide (B) and protein (C) sequences of the HBoV1 capsid variants shown on the left in each panel. The second column depicts the number of nucleotide (B) or amino acid (C) differences compared to variant DQ000495 (shown at the top). The numbers above each column should be read vertically and indicate the position of each nucleotide (B) or amino acid (C) in the HBoV1 capsid gene or protein (VP1), respectively. A dot indicates no change. The four colors highlight the corresponding nucleotides and amino acids in the two panels. (D) Phylogenetic tree of the shown HBoV1 capsid variants derived by applying maximum likelihood methodology and 500 bootstrap repeats. Bootstrap repeats with a support of >60% are shown at the nodes. HBoV2 (GenBank NC012042), HBoV3 (NC012564), HBoV4 (NC012729), bovine parvovirus (NC001540), and canine minute virus (NC004442) were defined as the outgroup. Capsid variants that resulted from this work are marked with a dot.
Interestingly, alignment of these 29 capsid DNA sequences with the HBoV1 reference sequence that was first reported by Allander et al. in 2005 (26) (GenBank DQ000495; note that this is not the sequence that is utilized in current HBoV1 vectors) showed differences in 32 nucleotide positions. As summarized in Table S1 and Fig. 1B, the newly analyzed sequences carry between 4 and 19 mismatches with this reference sequence, corresponding to an average of 12.7 nucleotide differences per variant (368 variations in total, divided by 29 samples). Accordingly, their overall DNA sequence identity to DQ000495 is 99.1 to 99.8%, or 99.4% on average. Moreover, we noted that the mutations cluster in 14 of the 32 positions, namely, 441, 445, 714, 984, 1140, 1168, 1170, 1176, 1188, 1308, 1758, 1767, 1768, and 1785 (numbers are nucleotide positions in the HBoV1 vp1 capsid gene), where more than half of the 29 sequences differed from the reference.
At the protein level, these point mutations translated into substitutions at four positions in the 671-aa HBoV1 VP1 capsid protein compared to the DQ000495 reference (Fig. 1C). Each of the 29 sequences differed in one to three positions, namely, 68, 149, 474, or 590 (numbers are amino acid positions in VP1), corresponding to identities of 99.6 to 99.9% (99.8% on average). Hence, the majority of nucleotide exchanges were silent at the VP1 protein level. Again, we observed a clustering of the mutations, most notable at aa position 149 (red in Fig. 1B and C), where 100% of all analyzed sequences carry a threonine instead of the alanine reported by Allander and colleagues (26). The second striking difference is seen at position 590 (orange in Fig. 1B and C), where we detected a serine instead of a threonine in 15 of the 29 sequences. In addition, sample V1541706 carries asparagine instead of aspartate at position 68, and samples VK11443 and VK12783 have a replacement of serine with asparagine at position 474.
Next, we performed a phylogenetic analysis of all 29 DNA sequences together with 21 publicly available HBoV1 sequences from 14 countries and four continents. Additionally, we included the related viruses HBoV2, -3, and -4 (GenBank accessions NC012042, NC012564, and NC012729, respectively) as well as the nonprimate bocaviruses, canine minute virus (NC004442) and bovine parvovirus 1 (NC001540), which we collectively defined as the outgroup. This analysis confirmed that all 29 HBoV1 vp sequences cluster together with the 21 public sequences and are clearly distinct from the outgroup (bootstrap value of 99) (Fig. 1D).
Correlation of HBoV1 capsid sequence diversity and virus infectivity.In addition to the primary sequence, we analyzed the viral load in the original set of 64 HBoV1-positive patient samples that we had collected. By using quantitative (q)PCR, we succeeded at measuring the viral load for 39 samples, comprising the 29 for which we had previously obtained the full capsid sequence (see above) as well as 10 others where this information was lacking. The values obtained ranged from 1.82 × 103 to 6.13 × 1010 vg/ml, with a median of 3.91 × 106 vg/ml (Fig. 2A). Prior work had defined a cutoff of 1 × 106 vg/ml, above which, symptoms of HBoV1 infection manifested in outpatients or inpatients (27). Accordingly, we can classify 15 of the 39 samples as having a viral load below this cutoff, while the other 24 are above.
Correlation of HBoV1 capsid sequence and viral load in patient material. (A) Results of qPCR titration of viral loads in 39 different patient samples. Shown is the median value (3.91 × 106 vg/ml). (B) Correlation of aa at position 590 (threonine or serine) with viral load in 31 selected (see main text for criteria) samples. Shown are means plus range. ns, nonsignificant. (C) Sequencing results of selected regions of the HBoV1 capsid in material taken from patient A (sputum and tracheal secretions) at the four indicated time points (total collection period was approximately 15 weeks). Also shown are the corresponding viral loads. Letters and colors indicate the aa or nucleotide at each position (green, adenine; blue, cytosine; black, guanine; red, thymine). (D) Same as panel C, but data for patient B and material (tracheal secretions) collected over a period of 3 weeks. (E) Viral loads measured for patients A and B (vg/ml) and shown in panels (C) and (D) are plotted (y axis) against the measurement time points (days; x axis).
As described above, HBoV1 VP1 aa position 590 is a hot spot for a change from threonine in DQ000495 to serine. In line with this, we noted that roughly half of the samples whose viral loads we had determined carried a serine at this position. We thus correlated the occurrence of either threonine or serine with viral load for 31 of the 39 samples. This subset was selected based on the criteria that we were able to read >90% of the complete capsid sequence and that we were able to unanimously identify position 590 as threonine or serine. Remarkably, this analysis showed that HBoV1 variants carrying a serine (T590S, n = 17) have an ∼18-fold lower viral load than those with the originally reported threonine (T590, n = 14) (Fig. 2B). In detail, the average viral load for the T590S variant was 5.55 × 108 vg/ml, in contrast to 1.02 × 1010 vg/ml for variant T590.
Among the analyzed samples, several had been collected from the same patient(s) at various time points, which allowed us to study dynamic changes in the HBoV1 capsid sequence and measure alterations in the viral load during the course of an infection. The results are depicted in Fig. 2C for patient A (four time points over a period of 15 weeks) and Fig. 2D for patient B (five time points over a period of 3 weeks). Changes at the nucleotide level over time were observed at positions 804, 873, 1140, 1168, 1170, 1701, 1767, 1768, and 1785. Two examples that were identified in both patients and that are illustrated in Fig. 2C and D are a gradual change of thymine to cytosine at position 1140 and a change of guanine to adenine at position 1170. Intriguingly, while these two mutations were silent at the protein level, we found that the gradual replacement in both patients over time of the nucleotide sequence 5′-AA-3′ at position 1767/1768 by 5′-CT-3′ resulted in an exchange of threonine at aa position 590 to serine, i.e., the same mutation that we had already observed and highlighted before. Congruent with the data in Fig. 2B, we measured a consistent drop in viral loads over time that concurred with the accumulation of the T590S mutation. This is evidenced by a reduction of viral loads in patient A, from 3.11 × 109 vg/ml at the earliest time point to 7.46 × 103 vg/ml at the latest (week 15) (Fig. 2C and E). Likewise, the viral loads in patient B dropped from 1.57 × 1010 vg/ml to 3.91 × 106 vg/ml over the course of 3 weeks (Fig. 2D and E).
Dissection of the impact of changes in the vp ORF using recombinant HBoV1 vectors.To further study the impact of the observed natural or, introduced later in this work, synthetic point mutations in the HBoV1 capsid, we harnessed a streamlined system for production of recombinant HBoV1 gene transfer vectors that we have established recently (24). Its hallmark is the ability to package rAAV vector genomes encoding a reporter into HBoV1 capsids, by triple-transfecting HEK293T cells with three plasmids expressing all necessary factors, including the HBoV1 capsid proteins, and by purifying the resulting vector particles via iodixanol density gradient centrifugation. Specifically, here, we used an AAV vector genome expressing Gaussia luciferase (Gluc), which is a secreted protein that is easily detected in the supernatant of cultured cells. For the latter, we used primary human airway epithelia (pHAE) based on findings by others and us that these are highly susceptible to HBoV1 transduction (24, 28).
In total, we studied 18 HBoV1 capsid variants using this system, which comprised seven from our collection that had high viral loads between 3.4 × 109 and 6.1 × 1010 vg/ml and that exemplified the roughly equal distribution of either serine or threonine at position 590 (Fig. 3A and B; samples V1500812, V1602382, V1502611, V1512195, V1541706, V1613007, and VK11443). Compared to DQ000495, they differ in 5 to 18 nucleotides, or one to three amino acids. As two references, we included DQ000495 and GQ925675, i.e., the HBoV1 capsid variant that has been used in all recombinant HBoV1 vector preparations reported to date (22, 24, 25). Furthermore, we selected nine variants that were described by Principi et al. in 2015 and that differ from DQ000495 in one to six amino acids (27). As the original sequences were not available to us as molecular clones, we recapitulated only nonsynonymous substitutions that cause amino acid changes by successive overlap-extension PCR using DQ000495 as the template (Table S2 and S3). Of these nine capsids, seven were associated with high viral loads between 5 × 108 and 7.5 × 109 vg/ml in the study by Principi and colleagues (27). The remaining two variants from this study, named KPLGr1 and KPLGr2, represent two groups of patient-derived HBoV1 capsid sequences that were identical within each group and that were associated with low viral loads of 4.5 × 104 and 6.3 × 104 vg/ml, respectively. They were interesting, since KPLGr1 carries the T590S mutation, while KPLGr2 has T590, thus representing the distribution of these two residues and making these two variants useful as additional controls. Altogether, this set of 18 capsid variants was composed of six carrying T590 and 12 with T590S.
Selection and production of recombinant HBoV1 capsid variants. Nucleotide (A) and protein (B) sequences of the 18 HBoV1 capsid variants shown onthe left in each panel. The second column depicts the number of nucleotide (A) or amino acid (B) differences compared to variant DQ000495 (shown at the top). The numbers above each column should be read vertically and indicate the position of each nucleotide (A) or amino acid (B) in the HBoV1 capsid gene or protein (VP1), respectively. A dot indicates no change. The colors highlight the corresponding nucleotides and amino acids in the two panels. Panel B also shows the viral loads that were measured for each of the corresponding patient samples. Values marked with an asterisk were taken from Principi et al. 2015 (27) and rounded. (C) Vector titers (means ± SDs) were determined by qPCR for the shown 18 HBoV1 variants (each produced in nine 15-cm2 plates and purified). All titrations were performed twice except for those marked with an asterisk, which were performed once. The dotted line represents the average value of 1.1 × 1011 vg/ml. Colored bars represent randomly chosen candidates that are highlighted in the next section of this work. (D) Same results as in panel C but sorted by the presence of a threonine (n = 6) or serine (n = 12) at position 590. Shown are means ± SDs. ***, P < 0.001 (unpaired t test).
Titration of recombinant vectors based on HBoV1 capsid mutants.To produce vectors based on the aforementioned 18 HBoV1 capsid variants, we performed three independent runs using three 15-cm2 dishes each, following our triple-transfection and iodixanol purification scheme (24). Titration by quantitative (q)PCR revealed titers for the 54 (18 × 3) individual stocks between 6.51 × 108 to 6.61 × 1011 vg/ml (data not shown), with an average of 9.65 × 1010 vg/ml. The majority of capsid variants yielded titers in a range of 2 × 1010 to 2 × 1011 vg/ml, with the notable exception of KPLMI-30 and KPLMI-3503, whose average titers were 1.13 × 1010 and 2.15 × 109 vg/ml, respectively. In contrast, KPLGr1 and KPLGr2 that had been associated with low viral loads in patients before (27) consistently produced well in our hands.
We subsequently pooled all three independent preparations per capsid variant and further purified and concentrated them using Amicon Ultra-15 filter units. Final titers per nine 15-cm2 dishes per capsid variant were between 4.11 × 1010 and 2.4 × 1011 vg/ml (average of 1.1 × 1011 vg/ml) (Fig. 3C), again with the exception of KPLMI-30 and KPLMI-3503, which yielded 1.7 × 109 and 4.67 × 108 vg/ml, respectively. Titers for the 16 capsids that produced well were determined twice, since these stocks were used in later transduction experiments (see below). On average, concentration using the Amicon Ultra-15 filters resulted in a vector particle recovery of 42%.
Interestingly, we noted a highly significant difference between the HBoV1 capsids depending on the presence of a threonine or serine at position 590. While the average titer of the six T590 variants was 6.66 × 1010 vg/ml, it was 1.59 × 1011 vg/ml for all capsids with the T590S mutation, i.e., 2.39-fold higher (P = 0.0003) (Fig. 3D). Of note, for this calculation, we excluded the two low producers KPLMI-30 and KPLMI-3503, as they were obvious outliers. To facilitate the comparison of candidates and to highlight their differences throughout this work, five candidates were randomly chosen (including the two references DQ000495 and GQ925675) and colored.
Comparison of the transduction efficiency of all HBoV1 capsid variants in pHAE.To measure and compare the transduction efficiency of the 16 HBoV1 capsid variants that produced well, we used pHAE from five different donors (labeled 171469, 171476, 171834, 171905, and 171975). Each capsid was tested twice per donor (n = 10 for each variant) and compared to two negative controls (two wells of untransduced cells). Per transwell, we applied 3 × 108 vg, which corresponds to a multiplicity of infection (MOI) of 600 based on a count of roughly 5 × 105 cells per transwell. As the Gluc reporter that was encoded by all vectors is secreted from the cells, we were able to collect cell culture supernatants at three successive time points (day 6, 9 or 12 posttransduction) and thus analyze the kinetics of transduction.
Two general observations were, first, that the overall transduction efficiencies varied substantially between the donors, as best illustrated by the up to 10-fold differences in luciferase light units between donors 171476 and 171905 (Fig. 4A, also shown in Fig. 4D). Second, we noted donor-dependent relative differences in the performance of the various capsids. This is exemplified by variant KPLMI-499 (yellow bars) that gave robust luciferase expression in donors 171905 and 171975 but was less efficient in the other donors, 171469, 171476, and 171834 (Fig. 4A, also shown in Fig. 4D). For these two reasons, the raw data for all five donors are depicted individually in Fig. 4A and B.
Functional characterization of HBoV1 variants. (A) The indicated HBoV1 variants were tested for their transduction abilities in pHAE derived from five different patients (171834 to 171476). To this end, an rAAV-Gluc genome was packaged into each HBoV1 capsid, and pHAE were transduced apically at an MOI of 600. Gluc activity (means ± SDs) was measured in the medium 12 days posttransduction and plotted on the y axis as arbitrary light units (ALU). Colored bars represent candidates shown in Fig. 3C and are intended to facilitate comparison of candidates between pHAE derived from different patients. (B) Same results as in panel C but sorted by the presence of a threonine (n = 6, 2 transwells each) or serine (n = 12, 2 transwells each) at position 590. Shown are means ± SDs. ns, nonsignificant (multiple t test). (C) Flow cytometry analysis of untransduced pHAE derived from the indicated patient samples (n = 2 independent transwells per patient). Cells were stained for cell type-specific markers, i.e., β-tubulin IV (ciliated cells) and MUC5AC (goblet cells). Blue areas show the percentages of positive cells for the indicated markers. (D) Side-by-side comparison of selected HBoV1 variants highlighted in Fig. 3C and in panel A. Shown is their transduction ability in pHAE derived from different patients (indicated by the numbers below the x axis). Transduction efficiency was estimated from the Gluc reporter activity in the medium, which is plotted as ALU on the y axis (means ± SDs). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, nonsignificant (one-way ANOVA). (E) Transduction of pHAE with the shown HBoV1 variants at an MOI of 200. Gluc activity (means ± SDs) was measured in the medium 3, 9, and 12 days posttransduction and plotted on the y axis as ALU. NEG, negative control (untransduced cells).
Despite the donor variability, it was noteworthy that the two published variants, DQ000495 and GQ925675, were consistently among the top performers in all five donors. Furthermore, the HBoV1 reference sequence DQ000495 (green bars in Fig. 4A) that was originally reported in 2005 significantly outperformed GQ925675 (orange bars) in the pHAE derived from donor 171467 (Fig. 4A, also shown in Fig. 4D). This is intriguing considering that, to our best knowledge, the less-efficient GQ925675 forms the basis for all recombinant AAV/HBoV1 vectors that are currently in use by us and others. Next to these two historic controls, also, capsid variant V1512195 (blue bars, Fig. 4A and D) performed remarkably well in most cells, albeit we again noted some degree of donor dependency (e.g., in donor 171975, where it ranked second to last). In contrast, capsid variants V1613007, KPLMI-246, KPLMI-253, KPLMI-311, V1500812 (all in gray), and V1502611 (pink bars) were frequently among the least efficient of all capsid variants, also with a few donor-specific exceptions (e.g., the good performance of capsids V1500812 and V1502611 in donor 171469) (Fig. 4A). Besides, we measured a steady increase in luciferase transgene expression for all 16 capsids and all five donors from day 6 to 12 posttransduction (data not shown), congruent with prior notions of AAV/HBoV1 kinetics in this cell culture model (24).
Next, we compared transduction efficiencies at day 12 based on the presence of T590 (six variants) versus T590S (twelve variants) (Fig. 4B). We found no significant differences between the two groups in their transduction abilities but observed trends in donors 171834, 171469, and 171476, where variants with T590 had a slight advantage. Vice versa, the S590 variant tended to perform better in donors 171905 and 171975. Notably, as exemplified in Fig. 4C, we concurrently observed differences in the cellular compositions of the pHAE cultures that might have influenced the transduction efficiency and specificity of the HBoV1 variants. In particular, sample 171834 showed a much higher proportion of mucin-positive goblet cells versus tubulin-positive ciliated cells than samples 171905 and 171975, where ciliated cells predominated. An overview of the differential efficiency of selected variants in each of the five pHAE cultures is shown in Fig. 4D.
Finally, we separately tested the two HBoV1 variants that had yielded the lowest titers during vector production (Fig. 3C), KPLMI-30 and KPLMI-3503, in direct comparison to DQ000495 (Fig. 4E). For this, we used pHAE derived from two donors, one from the group of five that we had also used as described above (171476), while the other was used only in this experiment (171427). Moreover, owing to the limiting vector yields, we had to reduce the MOI from 600 to 200, corresponding to 1 × 108 vg per pHAE filter. Consistently, we found that these two HBoV1 variants mediated very low transduction that was 10- to 100-fold below that of the DQ000495 reference at all time points (Fig. 4E). Together, this implies that the multiple nucleotide and/or amino acid differences in these two capsids compared to the HBoV1 reference (Fig. 3A and B and Table S3) impact the abilities to both be produced as recombinant vector and mediate robust transgene expression, at least in pHAE.
Packaging and transduction efficiency of HBoV1 tyrosine mutants.An alternative approach to identify HBoV capsids with improved properties that complements screening of natural HBoV variants is rational engineering of the viral capsid. To this end, the modification of surface-exposed tyrosine residues in the viral capsid is especially promising, as these residues play important roles in assembly, ubiquitination, and degradation as well as in transcription and transduction of parvoviruses (29–35). Accordingly, we mutated six different tyrosine residues in the VP1 capsid protein to phenylalanine that we had originally predicted by structural modeling to be located on the HBoV1 capsid surface (Fig. 5A), namely, Y276, Y403, Y484, Y523, Y595, and Y657 (all VP1 amino acid numbering). To study the effects of each mutation on particle assembly and transduction, we packaged two transgenes, yfp (yellow fluorescent protein) or Gluc, into the different capsid variants. All mutants yielded largely comparable vector amounts (data not shown), implying that none of the studied tyrosine mutations affects vector production.
Synthetic tyrosine mutants of the HBoV1 capsid and their transduction properties. (A) Shown tyrosine residues (Y) were mutated to phenylalanines (F). Numbers indicate the position of the amino acid in VP1. Underlined nucleotides represent mutated residues that result in the corresponding amino acid change. (B) Gluc activity measured at the indicated time points. pHAE were transduced with the different tyrosine HBoV1 mutants from panel A at an MOI of 2 × 104 in the presence (+) or absence (−) of proteasome inhibitors. N, negative control (untransduced cells); wt, wild-type HBoV1 capsid (positive control). The dotted line represents the assay background. (C) HBoV1 capsid surface representation with surface-exposed tyrosine residues highlighted in green and the hydroxyl groups colored in red. In variants with an asterisk, the hydroxyl group is partially inaccessible. This image was generated using Chimera.
Next, to assess the ability of these mutants to transduce pHAE, they were added to the apical surface of transwells at an MOI of 2 × 104. This higher MOI (versus 600 before) was used, as we expected a lower infectivity for at least some of the tyrosine mutants and we wanted to be able to measure all transduction efficiencies in the same experiment and under identical conditions. Transductions were performed in the presence or absence of proteasome inhibitors to study whether the Y-to-F mutations had circumvented capsid ubiquitination and thereby alleviated HBoV1’s dependency on proteasome inhibition. Interestingly, we found that transduction of two of the six mutants was impaired at all time points (day 3 to 14) (Fig. 5B), namely, Y484 (∼6.8-fold reduction compared to that of wild-type HBoV1) and Y595 (∼4.5-fold reduction). In contrast, the other four mutants were indistinguishable from the wild-type control. Moreover, the Y484F mutation had actually increased the dependency of the cognate capsid on proteasome inhibitors, because this capsid was inert in their absence. In this respect, it differs from the Y595F mutant that showed a similarly reduced potency in the presence of proteasome inhibitors but that, unlike the Y484F variant, remained active also without these inhibitors (albeit at reduced efficiency, akin to all other variants).
Of note, after the completion of these experiments, the capsid structure of HBoV1 was determined by cryo-electron microscopy (36). The HBoV1 capsid structure revealed that of the six tyrosines that we studied here, only three are indeed located on the capsid surface, i.e., Y276, Y403, and Y595 (Fig. 5C). In addition, the hydroxyl group of Y276 is inaccessible for potential phosphorylation. In contrast, Y484, Y523, and Y657 are not exposed on the capsid surface. Most notably, mutation of Y484, which is located inside the capsid, had yielded the strongest phenotype, implying a mode of action that differs from the anticipated phosphorylation and ubiquitination (see Discussion).
Differential inhibition of HBoV1 capsid variants by human immunoglobulins.To determine the effect of anti-HBoV1 antibodies on the functionality of the different HBoV1 capsid variants, we first performed an enzyme immunoassay (EIA) (Fig. 6A). For this purpose, microtiter-plate wells were coated with the different capsid variants (antigens) at four different dilutions (only 8- and 16-fold dilutions are shown, as they resulted in a linear signal). Next, the reactivity of a human serum pool positive for HBoV1 antibodies was measured. We detected small differences typically within 1.3 optical density (OD) units between the variants in their ability to bind HBoV1-specific antibodies, except for GQ925675, which showed 2- to 4-fold reduced binding. Surprisingly, for reasons as-of-yet-unknown, capsid variant VK11443 that differs only in one nucleotide from GQ925675 and has an identical amino acid composition did not show the same reduction in antibody binding. Also notable are the two capsid variants KPLMI-30 and KPLMI-3503, which resulted in ODs of 3.10 and 0.10, respectively (at 8-fold dilution). Notably, the low viral titers of both KPLMI-30 and KPLMI-3503 (close to the background) limit the qPCR-based analysis, which requires encapsidated genomes to reliably estimate the amounts of viral capsid. Thus, the slightly higher OD value measured for KPLMI-30 than for the other variants (differences between 0.72 and 2.54 ODs) might result from the technical challenge in quantifying the proper amount of virus solution used in the EIA. The differential reactivity of these two variants with human antibodies is interesting in view of their strikingly reduced viral titers and transduction ability in pHAE (Fig. 3C and 4E, respectively). Based on the titer reduction, we speculated that these variants might have a defect either in assembly or in genome packaging. To resolve these possibilities, we studied viral VP expression by Western blotting (Fig. 6B). All variants expressed VP1, VP2, and VP3 proteins (without evidence for additional protein species) in the expected 1:1:10 stoichiometry, which shows that capsid protein expression is not limiting. Thus, the absence of signal in the EIA for KPLMI-3503 viral particles implies that this variant might have an assembly defect. In contrast, the high signal for KPLMI-30 in combination with the low measured viral titers hints at an impairment in genome packaging. Notably, we do not rule out the possibility that subtle differences in VP protein composition not detected by Western blotting may have contributed to the variations in transduction or packaging ability, a theory that could be studied with other more sensitive methods. Besides, we did not detect variations in the length of the encapsidated vector DNAs (data not shown), implying that genome integrity is not responsible for the differences in titer or transduction.
Analysis of the immunoreactivity of HBoV1 capsid variants. (A) EIA using pooled human sera positive for HBoV1. Iodixanol-purified viral stocks (adjusted to an average of 5 × 1010 vg/ml) of the indicated HBoV1 capsid variants were used to coat the wells of a microtiter plate (8- and 16-fold dilutions). EIA absorbance values (optical density [OD]) are plotted on the y axis. (B) Western blot analysis of the variants shown in panel A. HEK293T cells were transfected with the different HBoV1 plasmids to analyze the expression of the three capsid proteins, VP1, VP2, and VP3. (C) In vitro neutralization assay using commercially available human immunoglobulins (IVIg). Gluc-expressing vectors were preincubated with the indicated IVIg concentrations for 1 h at 37°C and then used to apically transduce pHAE at an MOI of 1 × 104 (5 × 109 vg per well). All assays were performed in duplicates. Shown are mean Gluc activity levels (light units) plus ranges measured at days 5 and 10 posttransduction.
Concerning the T590S variation, the EIA did not reveal obvious differences between the T and S groups in their binding to human anti-HBoV1 antibodies, which implies that aa position 590 does not confer a differential reactivity to HBoV1 capsid antibodies in human sera.
It is known for AAV vectors that antibody binding does not always result in neutralization of virus transduction (37). To test whether this applies to HBoV1 as well and to compare our different variants in a functional assay (transduction ability), we performed in vitro neutralization assays using commercially available pooled human immunoglobulins (IVIg). This mix of IgG antibodies from healthy individuals has previously been shown to potently reduce the activity of the standard GQ925675 vector, in line with its large seroprevalence in the human population (24).
Due to the limited availability of pHAE, we selected seven HBoV1 variants for the assay with equal T/S590 distributions, including the GQ925675 and DQ000495 reference strains. To this end, we mixed 5 × 109 vg per capsid variant (corresponding to an MOI of 1 × 104 vg per well) with six different IVIg concentrations and incubated these mixtures for 1 h at 37°C before adding them to pHAE from two different donors (Fig. 6C). These IVIg concentrations were shown in pilot studies to result in a reduction or complete inhibition of transduction with GQ925675 (data not shown). Comparison of Gluc expression from the different HBoV1 variants at day 5 and 10 showed no obvious differences between the two groups (T or S at position 590) in their transduction abilities in the presence or absence of IVIg, which supports the previous notion that this variation does not confer increased resistance to neutralizing antibodies. Still, contradicting this trend is the KPLGr2 variant, which differs only in aa position 590 from KPLGr1 but shows a higher resistance to IVIg, despite its stronger binding to antibodies in the EIA (Fig. 6A).
Analysis of evolutionary selection pressures on HBoV1.Intrigued by our finding that the natural T/S590 variation has a profound impact on viral titer, we asked whether this site was subject to positive selection pressure. To address this question, we used several methodologies comprising mixed-effects model of evolution (MEME), single likelihood ancestor (SLAC), fixed-effects likelihood (FEL), and fast unconstrained Bayesian approximation (FUBAR) (see Materials and Methods). Indeed, we found one site (using MEME) under “positive selection” (moderately significant with P = 0.51), namely, the above-mentioned aa 590 (Fig. 7A). However, a positive selection of T/S590 was not supported by the other methods used, which showed a neutral selection pressure at this position. Thus, albeit it is implied by the MEME results, we cannot firmly conclude that the observed substitution at this site has an impact on intraspecies transmission and adaptation of HBoV1.
Selection pressure acting on the intrahost level and comparison of the interhost genetic diversity of the VR-VIIIB. (A) Positive or negative selection pressure acting on the vp codons of the HBoV1 strains in this study. MEME, mixed-effects model of evolution; SLAC, single likelihood ancestor; FUBAR, fast unconstrained Bayesian approximation methods; FEL, fixed-effects likelihood. (B) Amino acid 590 variation among different primate bocaviruses.
Finally, we analyzed the conservation of the aa 590 residue in the VR-VIIIB by comparing the amino acid composition of this region to that of other primate BoVs (Fig. 7B). Notably, the gorilla bocavirus (GBoV) described by Kapoor et al. (38) is genetically most closely related to HBoV1 (at both the nucleotide and amino acid levels). Accordingly, the VR-VIIIB of HBoV1 is also most homologous to the one in GBoV, with both carrying a threonine at position 590, in contrast to an asparagine in HBoV2-4. Also interesting in this context is the profound ability of GBoV to transduce human airway epithelial cells that we have reported recently (24). Together, this may hint toward an interspecies transmission of HBoV1/GBoV.
DISCUSSION
The present work was fueled by a string of recent publications showcasing the great potential of recombinant gene transfer vectors derived by packaging of AAV vector genomes into HBoV1 capsids (22, 24, 25, 39). Most recently, these reports have inspired us to engineer similar vectors based on alternative BoV, i.e., HBoV2-4 and gorilla bocavirus (GBoV), leading to our discovery of their favorable properties for gene transfer into various primary human cells (24). At the same time, this work by others and us has revealed a series of gaps in our current understanding of fundamental BoV biology, whose resolution will not only benefit our knowledge of the bocaviral life cycle but also promises to foster the development of next-generation BoV gene therapy vectors.
In the first part of this work, we aimed to study the natural variation in HBoV1 isolates. Specifically, we focused our attention on the capsid (vp) ORF, which is the determinant of virus tropism and the subject of extensive research in viral vector development. Therefore, we amplified and sequenced 29 full-length vp sequences from patient samples collected in Heidelberg and Cologne. Despite the high degree of sequence conservation among the HBoV1 isolates, which is in line with previous reports (40), we detected an interesting hot spot for variation in VP1 at aa position 590 in around 50% of analyzed patient samples. This change results from a conversion of two nucleotides (5′-AA-3′) at positions 1767 to 1768 in vp1 to 5′-CT-3′. Interestingly, a previous study by Principi and coworkers, who analyzed samples from Milan (Italy), also showed this high prevalence of the T590S variation (27). The additional collection of several samples from the same individual over time allowed us to also follow the course of infection in two patients. Surprisingly, we found dynamic changes at the 590 aa position, starting with a clear abundance of threonine that was gradually replaced by a serine (see Fig. 2C and D). The emergence and persistence of the T590S switch could be explained by different events: (i) a de novo change of two adjacent nucleotides during virus evolution, which is, however, unexpected in view of the estimated mutation rate of primate bocaviruses (9 × 10−4 mutations/site/year) (41, 42), or (ii) as previously proposed by Martin and colleagues (43), a secondary infection with another strain, which might have a replication or immunological advantage and thus dominated over time. Another and rather rare scenario also reported in the above-mentioned study is the coinfection with two strains during the same primary event that have different kinetics or immunoreactivity. To conclusively identify the proper scenario, deep-sequencing analysis of samples from different time points has to be performed to detect even minor quantities of specific variants. Interestingly, as opposed to the study by Principi et al., the T590S change was concomitant with a decrease in overall virus load (27). However, these various study outcomes may have resulted from different time and end points in sample collection. At this point, it thus remains equally possible that the observed decrease at the endpoint of sample collection truly reflects a biological property of T590S or that a secondary infection has happened in these patients. Consequently, a firm conclusion regarding the persistence or spread of S590 versus T590 variants and their possible positive selection cannot be drawn until more samples have been collected and analyzed during symptomatic and asymptomatic periods.
In addition to the prominent T590S mutation, we detected additional amino acid changes resulting from single-nucleotide polymorphisms (SNPs): D to N (aa position 86), S to N (aa position 474), and a dominant A-to-T mutation (aa position 149) in all variants (compared to the DQ000495 reference strain). To experimentally unravel the roles of these naturally occurring SNPs or singletons in the vp ORF on the producibility and transduction ability of HBoV1, independent of the exact collection time points and sample types, we made use of a previously established recombinant vector system in which rAAV genomes are packaged into BoV capsids (22, 24). To this end, we selected variants from patient samples with approximately equal T/S distributions (Fig. 3A and B) and packaged an rAAV-Gluc vector into each capsid. The high amino acid identity of the HBoV1 variants in our study prompted us to include additional synthetic mutants into our screen that either (i) recapitulate several natural HBoV1 variants reported by Principi et al. (27) or (ii) were rationally designed, based on a previously published HBoV1 structure by Gurda et al. (44). During packaging of the rAAV-Gluc genomes into the different HBoV1 vp variants, we gained significantly higher titers with the ones harboring a serine at aa position 590 (see Fig. 3C and D), which was surprising as it is diametrically opposed to the decrease observed in the patient material. Hence, these findings allow us to conclude that the decrease in S590 viral load in patient material does not result from a reduced ability to produce viral progeny (Fig. 2B). Moreover, the consistently higher viral titers obtained using 590S variants indicate a direct effect of the variation on capsid assembly and/or genome packaging. Interesting in this context and supporting this hypothesis is that T/S590 lies in the capsid VR-VIIIB (also called “HI loop”) (Fig. 8A and B). This region belongs to the surface-exposed hypervariable regions and was linked to particle assembly and genome packaging in AAV (45), hinting at a similar function in the BoV context. In addition to its role in particle assembly, the VR-VIIIB is crucial for the externalization of the VP1u region during endosomal escape and thus contributes to particle infectivity (45). Accordingly, to test whether the vp variations studied in this work affected the transduction ability of HBoV1, pHAE grown on transwells were transduced from the apical side with equal numbers of viral particles. Transgene expression was followed over time by measuring the secreted reporter Gluc in the medium. Intriguingly, nearly all tested variants displayed high transduction abilities (except for KPLMI-3503, KPLMI-30, HBoV1 Y595F, and HBoV1 Y484F). Moreover, their activity was dependent on the transwell composition and thus varied among pHAE cultures, consistent with our prior observations with various primate BoV vectors (24). When we compared transduction abilities in the context of the T590S variation, we did not observe significant differences between the two groups. This is in line with a previous study using recombinant AAVs, which showed that swapping of the complete HI loop between serotypes with high sequence identity affects their producibility but not transduction ability (45). Hence, we concluded that the T/S variation affects the virus titers by playing a role in particle assembly and/or genome packaging but does not determine infectivity. In contrast, mutation of the tyrosine residue to phenylalanine at aa position 595, which flanks the VR-VIIIB in the HBoV1 capsid, resulted in a 4.5-fold reduction in virus infectivity. Importantly, this residue is highly conserved among primate BoVs except for HBoV4, where it can be naturally replaced by a phenylalanine (e.g., HBoV4 strain FJ973561).
Structural representation of the HBoV1 T590S variant (GQ925675) as well as that of the HBoV1 variants KPLMI-30 and KPLMI-3503. (A) Shown is the HBoV1 capsid surface with the 2-, 3-, and 5-fold symmetry axes as well as the location of the S590 residue (in green). (B) Ribbon diagrams of the VP3 monomer of the HBoV1 S590 variant. The S590 residue is represented by a green sphere and localizes to the VR-VIIIB (HI loop). Surface representation of the HBoV1 KPLMI-30 (C) or KPLMI-3503 (E) capsid. The aa residues 396 to 397 in KPLMI-30 are located on the capsid surface and are shown as green spheres, whereas the critical residues in KPLMI-3503 (534 to 536) reside inside the capsid and are not visible in the image in panel E. (D and F) Ribbon diagrams of the VP3 monomer of the HBoV1 variants in panels C and E, respectively. Residue changes compared to DQ000495 mentioned in panels C and E are shown as green spheres. Images in panels A, C, and E are radially colored from blue to white and red, representing capsid center to surface regions. These images were generated using Chimera. In panels B, D, and F, the capsid VRs from I to IX are indicated. These images were generated using PyMOL (http://www.pymol.org/).
An even stronger phenotype (6.8-fold reduction in infectivity) was obtained when Y484, located between VR-VI and VR-VII, was mutated to a phenylalanine. This residue lies inside the capsid and is also highly conserved among primate BoVs, indicating an important role in the BoV infection pathway, most likely perhaps for capsid assembly. Further analysis of Y484 in the wild-type HBoV1 capsid structure showed a potential hydrogen bond of the side chain’s hydroxyl group to the backbone (A435) of VR-V situated above this residue. Thus, an intriguing working hypothesis for future work is that the removal of the hydroxyl group by mutation to phenylalanine could alter the conformation of VR-V, which is a potential determinant of host tropism, and result in reduced virus infectivity or vector transduction efficiency.
The best performers in our transduction assays were DQ000495 and HBoV1 Y523F, which consistently mediated comparable or higher transduction than the standard BoV vector GQ925675 and thus represent promising candidates for a future application as viral vectors. In contrast to the favorable effect of S590, two of the reconstructed mutants (KPLMI-30 and KPLMI-3503) resulted in 64- and 239-fold lower average viral titers, respectively, despite the presence of the S590 residue. The EIA revealed the presence of assembled particles for KPLMI-30 but not KPLMI-3503, despite the presence of free VP proteins (compare Fig. 6A and B), which implies a defect in genome packaging, particle assembly, and/or antibody recognition. This hypothesis is supported by the localization of the different residues in both variants. In KPLMI-30, the two residues SA (at aa 396 to 397) are located on the capsid surface at the 5-fold canyon (Fig. 8C and D), i.e., a region surrounding the 5-fold axis channel, which serves as a portal for viral genome packaging (36, 46). In contrast, in KPLMI-3503, aa residues 534 to 536 (KPD) are inside the capsid between the HI loop and VR-VIII that is located on the sides of the 3-fold protrusions, i.e., the determinants of antibody recognition and infectivity in other parvoviruses (36) (Fig. 8E and F). Thus, we speculate that these residues might have led to structural changes that either negatively influenced capsid assembly or interfered with antibody binding. The latter would, however, not explain the 239-fold reduced viral titers. Consequently, at this point, the exact mechanisms underlying our observations remain unknown, and our assumptions require experimental validation. For instance, to unanimously determine whether aa 534 to 536 affect particle assembly or impact genome packaging, a different antibody recognizing a conformational epitope should be used in future experiments. Finally, it is interesting that the transduction ability of both mutants was severely compromised in pHAE (Fig. 4E), which seems to be at odds with the high titers (∼5 × 109 vg/ml) at which these mutants were detected in the original report (27). Here, it is important to mention again that only nucleotide changes that cause an amino acid change were transferred into our expression constructs. Thus, the effect of silent mutations that might have led to changes in different as-of-yet unknown ORFs within vp2 cannot be assessed in this setting.
In the last part of this work, we asked whether the T590S variation affects the serological reactivity of the HBoV1 capsid, which might have resulted in the emergence and/or persistence of this mutation in the clinical samples. This question is also important for the application of viral vectors in gene therapy, where preexisting neutralizing antibodies significantly lower the therapeutic benefit. To this end, we pursued two independent approaches: (i) EIA to assess the binding of antibodies in HBoV1-positive sera to the HBoV1 vp variants, and (ii) a functional assay in which the impact of a pool of human antibodies (IVIg) on the transduction abilities of the HBoV1 variants was studied.
Interestingly, we found only small differences between the variants in the EIA, except for the GQ925675 variant that showed 2- to 4-fold reduced binding to serum antibodies (Fig. 6A). This, however, did not directly correlate with the results from the transduction assays, where GQ925675 had no benefit compared to the other tested variants (Fig. 6C). This discrepancy was also observed in previous reports using AAV vectors and indicated that antibody binding is not always linked to particle functionality (37). The in vitro neutralization assay also did not reveal any correlation between the T/S590 variation and the susceptibility of the HBoV1 capsid to neutralization by pooled human IgG. One exception is the Gr2 variant, which was more resistant to neutralization than the other variants and which differed from Gr1 only in the presence of S590. This led us to the conclusion that T/S590 mostly affects the efficiency of virus production, yet it remains to be determined how this variation affects virus spread or latency.
The higher resistance of the KPLGr2 variant in this work is remarkable and would be advantageous for future applications of this variant as a viral vector. Thus, it would now be interesting to (i) validate the immune-escaping ability of KPLGr2 by testing its reactivity to different patient sera, and (ii) include other variants that have the same amino acid sequence as KPLGr2, namely, V1512195, V1502611, and V1500812, but that differ substantially in their nucleotide compositions. These variations have led to different transduction abilities in pHAE (Fig. 4A and D), which might be a result of alternative as-of-yet undiscovered ORFs that could also influence the immunoreactivity. Thus, it should be highly rewarding to additionally study the above-mentioned variants and the other BoV capsids reported in this work, as it will further enrich our knowledge of bocaparvovirus biology and support efforts to breed optimal viral vectors for human gene therapy.
MATERIALS AND METHODS
Plasmids and cloning procedures.The HBoV1 helper plasmid pCMVNS*Cap (GenBank GQ925675) was previously described (9, 39) and kindly provided by Ziying Yan. The DQ000495 vp sequence was ordered as a gene block from IDT (Leuven, Belgium). The gene block and HBoV1 vp sequences from clinical samples were amplified using primers 1 and 2 in Table S2 in the supplemental material (both with overhangs containing BsmBI restriction sites). The PCR product was cloned using a Golden Gate reaction into a previously described acceptor plasmid (pCMVNS*ΔVP-2×BsmBI) (24) lacking the vp sequence.
All other synthetic variants reported in this work were generated by introducing mutations using overlap extension (OE)-PCR as previously described (47). For each change, two PCRs were performed using overlapping primers (forward and reverse) containing the mutation(s) of interest (numbers 5 to 36 in Table S2) and two common external primers (3 and 4 in Table S2) with restriction sites (BstBI/EagI) that allow for the cloning of the end products into pCMVNS*Cap (see Tables S3 and S4 for an overview of the nucleotide changes introduced).
Phylogenetic analysis.Phylogenetic analysis of HBoV1 vp sequences was performed in MEGA7.0.26 (Pennsylvania State University, PA, USA). The evolutionary history was inferred using the maximum likelihood method. The percentages of replicate trees in which the associated taxa clustered together in the bootstrap test (500 replicates) are shown next to the branches. Only bootstrap values >70% are displayed.
Cell culture and patient material.HEK293T cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) with GlutaMAX (Thermo Fisher Scientific, Waltham, MA, USA), supplied with 10% fetal bovine serum (FBS) and 100 U/ml penicillin-streptomycin (both Merck/Sigma-Aldrich, Darmstadt, Germany). Polarized human airway epithelial cells were generated as previously described (24) from resected bronchial tissue and were obtained from Lung Biobank Heidelberg (member of the German Center for Lung Research, DZL) at the University Hospital Heidelberg, Germany. The cells were grown on ThinCerts (Greiner Bio-One, Frickenhausen, Germany) and differentiated at an air-liquid interface in PneumaCult ALI basal medium supplemented with PneumaCult ALI 10× supplement (both from StemCell, Vancouver, Canada).
Recombinant virus production.Pseudotyped BoV/AAV vectors were produced in HEK293T cells as previously described (24) using a triple plasmid transfection of (i) one of the BoV helper plasmids, (ii) a self-complementary (sc)AAV plasmid encoding Gluc (1.912-kb insert size), and (iii) pDGΔVP, a plasmid encoding rep from AAV2 and adenovirus helper genes (21). Cells were harvested 72 h posttransfection, and the crude cell lysate was processed for iodixanol gradient centrifugation as previously described (24, 48). The virus-containing 40% iodixanol fraction was pulled from the gradient, mixed with 15 ml phosphate-buffered saline (PBS), and applied to an Amicon Ultra-15 (Merck, Darmstadt, Germany) centrifugal filter unit (100,000 nominal molecular weight limit). Several centrifugation steps at 500 to 1,000 × g allowed for buffer exchange and concentration of viral preparations.
qPCR analysis of patient samples and recombinant virus titers.To determine viral titers, alkaline lysis was performed by mixing 10 μl of each virus stock with 10 μl Tris-EDTA (TE) buffer and 20 μl 2 M NaOH. The mixture was heated to 56°C for 30 min and then neutralized using 38 μl of 1 M HCl. Next, a 1:1,000 working solution was prepared and 5 μl was used in a TaqMan real-time PCR as previously described (48), using the 2× SensiMix II Probe kit (Bioline, Luckenwalde, Germany) and a probe binding in the promoter region (see Table S5 for probe/primer combinations).
To determine the viral load in patient samples, 2.5 μl of extracted DNA (QIAsymphony kit; Qiagen, Hilden, Germany) was directly mixed with 22.5 μl qPCR mix containing (i) 12.5 μl SensiMix SYBR No-Rox kit (Bioline), (ii) 0.25 μl of each forward and reverse primer (Table S5), and (iii) 9.5 μl H2O. The qPCR mixtures were measured in duplicates and run on a Rotor-Gene Q cycler (Qiagen) under the following conditions: initial activation (95°C, 10 s), followed by 40 cycles of (i) denaturation (95°C, 15 s), (ii) annealing (58°C, 20 s), and (iii) extension (72°C, 20 s). To ensure the detection of different HBoV1 strains, the forward and reverse primers were designed to bind in the relatively constant promoter region.
Gaussia luciferase assay.Gluc activity was determined in the cell culture medium as previously described (24). Briefly, 20 μl of the cell medium was incubated with 100 μl assay buffer supplied with coelenterazine (PJK, Kleinblittersdorf, Germany) at a final dilution of 11.7 μM. Gluc activity was detected in a GloMax96 microplate luminometer equipped with an automatic injector (Promega, Madison, WI, USA).
Enzyme immunoassay.HBoV1 variants selected from the transduction experiments were tested for their reactivity to human serum pools using an in-house IgG EIA. Therefore, 96-well microtiter plates were coated overnight at 4°C in duplicates with 4-, 8-, 16-, and 32-fold-diluted (in PBS) bocaviral stocks. Then, plates were coated with diluent LOY-X (Labsystems Diagnostics, Vantaa, Finland) three times for 10 min each. An HBoV1-IgG-positive serum pool (1:200) diluted in RED buffer (Kaivogen, Turku, Finland) was pipetted into each well and the plate was incubated at 37°C for 1 h. As control, a 1:200 dilution of an HBoV1-IgG-negative serum pool was used. After five washes with 0.05% Tween 20 in PBS, a horseradish peroxidase (HRP)-conjugated anti-human IgG (1:2,000; Dako/Agilent, Glostrup, Denmark) diluted in LOY-X was applied. For the detection of HRP-labeled antibodies, 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (Merck/Sigma-Aldrich) was added and incubated at room temperature for 20 min. Then, H2SO4 (0.5 M) was added to stop the reaction, and absorbances were measured at 450 nm using a MultiScan EX (Thermo Fischer Scientific).
Western blot analysis.Western blotting was performed as previously described (47). Briefly, 5 × 105 HEK293T cells were transfected with 2 μg of BoV helper plasmid. Three days posttransfection, the cells were harvested in 300 μl PBS, mixed with equal volumes of 2× SDS sample loading buffer, and boiled for 5 min at 95°C. The cell lysates were then centrifuged at 13,000 rpm, and 10 μl from each lysate was separated on 8% SDS-PAGE gels (Bio-Rad, Hercules, CA, USA). Next, the proteins were transferred to a nitrocellulose membrane (NeoLab, Heidelberg, Germany) using semidry transfer. For detection of the three capsid proteins (VP1, VP2, and VP3), an in-house-produced rabbit polyclonal anti-HBoV1 antibody was used at a 1:1,000 dilution. To produce the anti-HBoV1 antibody, rabbits were inoculated with HBoV1 VP3 virus-like particles (VLPs) that were produced in a baculovirus expression system. Serum was obtained on day 120 and tested for HBoV1 IgG by EIA. Immunization of rabbits was performed by GenScript (Piscataway Township, NJ, USA). For further details, see reference 49.
For detection of the primary antibody, an HRP-conjugated secondary donkey anti-rabbit antibody (NA934V; GE Healthcare, Chicago, IL, USA) was used (1:10,000). To visualize protein bands, the Lightning Plus-ECL reagent (PerkinElmer, Waltham, MA, USA) was added, and a chemiluminescence imager (Intas ChemoStar, Göttingen, Germany) was used to detect the signal.
Transduction of pHAE in the presence or absence of IVIgs.Primary HAE were incubated from the apical side with the different pseudotyped HBoV1 variants at an MOI of 1 × 104. To enhance transduction, two proteasome inhibitors were applied to the medium on the basolateral side, as previously described (22, 24), namely, 5 μM doxorubicin hydrochloride (Santa Cruz Biotechnology, Dallas, TX, USA) and 40 μM ALLN calpain inhibitor I (G-Biosciences, St. Louis, MO, USA). After 16 h, the virus was removed from the apical compartment and the medium replaced with fresh medium without inhibitors.
To perform transductions in the presence of IVIg (Kiovig; Baxalta, Bannockburn, IL, USA), the IVIg solution was diluted to working concentrations of 2, 10, 20, 40, 60, or 200 mg/dl. Next, equal volumes of virus and IVIg solution were mixed and placed at 37°C for 1 h. Positive (PBS, +HBoV1) or negative (+IVIg, −HBoV1) controls for transduction were also included. Transgene expression was measured at 5 or 10 days posttransduction.
Flow cytometry analysis.Characterization of pHAE cell composition using flow cytometry was performed as previously described (24). Briefly, to stain goblet and ciliated cells, the following primary antibodies were used: goblet cell marker MUC5A/C (ab3649; Abcam, Cambridge, UK) diluted 1:100 and ciliated cell marker β-tubulin IV (T7941; Merck/Sigma-Aldrich) diluted 1:100. Primary antibodies were incubated for 1 h at 4°C, followed by treatment with secondary anti-mouse antibody (AF-64 7 goat anti-mouse A21235; Thermo Fisher Scientific) for 30 min at room temperature. Cells were measured on a FACSVerse (BD Biosciences, Franklin Lakes, NJ, USA), and analysis was performed using Flowing Software (version 2.5.1; Turku Centre for Biotechnology, Turku, Finland). Only living cells were used for the analysis.
Statistical analysis.The statistical analysis was performed in PRISM Version 8.0 (GraphPad Software Inc., San Diego, CA). Two data sets in Fig. 2B or 3D were compared using an unpaired two-tailed t test analysis (for data in Fig. 2B, Welch’s correction was applied to account for highly significant differences in variances). The data sets in Fig. 4B were analyzed using a multiple t test. Statistical significance was determined using the Holm-Sidak method, with alpha at 0.05. Computations assume that all rows are samples from populations with the same scatter (standard deviation [SD]). In Fig. 4D, a one-way analysis of variance (ANOVA) with Dunnett’s test was used to compare each data set with the reference (DQ000495.1). Significance is denoted in the figures by asterisks above the SD or range bar (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; ns, nonsignificant).
Structural analysis.For the structural analysis, the model of the HBoV1 VP monomer (PDB ID 5URF) and the HBoV1 capsid 60-mer were downloaded from the VIPERdb online server (http://viperdb.scripps.edu) (50). Surface-exposed tyrosines were identified by generation of a gray surface representation of the capsid 60-mer followed by the coloration of tyrosine residues in Chimera (51). Similarly, the location of specific residues on the capsid surface such as serine 590 were identified. The depictions of the ribbon diagrams of the HBoV1 VP monomer were generated using the “smooth loop” option in PyMol (52).
Selective pressure analysis.Tests for negative or positive selections were conducted on the Datamonkey server (53). Methods used involve single likelihood ancestor (SLAC) (54), the fixed-effects likelihood (FEL) (54), the fast unconstrained Bayesian approximation (FUBAR) (55), and the mixed-effects model of evolution (MEME) (56). To reduce the probability of false-positive events, a P value threshold of 0.1 in SLAC, FEL, and MEME and a FUBAR posterior probability threshold of 0.9 were used to identify sites for selection.
Ethical approval.This study was carried out in accordance with the recommendations of the University Hospital Heidelberg with written informed consent from all subjects in accordance with the Declaration of Helsinki. All samples were received and maintained in an anonymized manner. The protocol was approved by the ethics commission at University Hospital Heidelberg under the protocol numbers S-270/2001 (collection of surgical material for lung research) and S-782/2018 (amplification of BoV sequences from patient material and application of derived BoV vectors in pHAE).
ACKNOWLEDGMENTS
J.F. and D.G. are supported by the Cystic Fibrosis Foundation (CFF; grant number GRIMM15XX0), the German Research Foundation (DFG; Cluster of Excellence Cell Networks, EXC81), and the Heidelberg Biosciences International Graduate School (HBIGS) at Heidelberg University. K.-P.L. is supported by an MD stipend from the German Center for Infection Research (DZIF, BMBF). D.G. received additional funding from the DZIF (TTU-HIV 04.803 and 04-815) and Collaborative Research Center SFB1129 (DFG, TP16; project number 240245660). M.A.S. and M. Meister received funding from the German Center for Lung Research (DZL, BMBF, 82DZL00402). M.A.-M. and M. Mietzsch were supported by NIH R01 GM082946. M.S.-V. and M.X. are funded by the Sigfried Jusélius Foundation and the Life and Health Medical Association, and M.X. received funding from the China Scholarship Council (CSC) and the Finnish-Norwegian Medical Foundation.
J.F., K.-P.L., and D.G. conceived the study. J.F. and K.-P.L. generated constructs and performed the majority of experiments. M.A.-M. and M. Mietzsch generated capsid structures and residue information. M.S.-V. and M.X. designed and performed EIAs. M.A.S. and M. Meister provided pHAE. O.S. and P.S. provided extracted DNA from patient material. J.F. and D.G. wrote the manuscript. All authors read the manuscript and approved the final version.
D.G. is a cofounder and shareholder of AaviGen GmbH. All other authors declare no competing financial interests.
FOOTNOTES
- Received 30 January 2020.
- Accepted 16 March 2020.
- Accepted manuscript posted online 25 March 2020.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.
