Abstract
Adeno-associated viruses (AAV) rely on helper viruses to transition from latency to lytic infection. Some recombinant AAV serotypes are secreted in a pre-lytic manner as extracellular vesicle (EV)-associated particles, although mechanisms underlying such are unknown. Here, we discover that the membrane-associated accessory protein (MAAP), expressed from a (+1) frameshifted open reading frame (ORF) in the AAV capsid (cap) gene, is a novel viral egress factor. MAAP contains a highly conserved, cationic amphipathic domain critical for AAV secretion. Wild type or recombinant AAV with a mutated MAAP start site (MAAPΔ) show markedly attenuated secretion and correspondingly, increased intracellular retention. Trans-complementation with recombinant MAAP restored extracellular secretion of multiple AAV/MAAPΔ serotypes. MAAP is sorted into recycling Rab11+ vesicles and strongly associates with EV markers upon fractionation. In addition to characterizing a novel viral egress factor, these studies highlight a prospective engineering platform to modulate secretion of AAV vectors or other EV-associated cargo.
Introduction
Adeno-associated virus (AAV) is a non-enveloped, single-stranded DNA virus belonging to the Dependoparvovirus genus within the Parvoviridae family1. Upon co-infection with a helper such as Adenovirus, Herpesvirus or Papillomavirus, AAV undergoes a transition from latent to lytic life cycle, hijacking the host cell machinery1. The AAV capsid consists of 60 capsid monomers of VP1, VP2, and VP3 at a ratio of 1:1:10 that packages a 4.7-kb single-stranded genome2. The AAV genome encodes replication (rep), capsid (cap), and assembly-activating protein (AAP) open reading frames (ORFs) flanked by inverted terminal repeats (ITRs), which are the sole requirements for genome packaging3–5. As such, the majority of the genome can be replaced by exogenous DNA sequences and packaged inside the AAV capsid to create a recombinant vector both in vitro and in vivo6,7. Some recombinant AAV serotypes appear to be secreted into cell culture media prior to lysis, albeit with variable efficiency8–10. Exactly how AAV exits the cell upon transitioning into this phase of replication remains unclear. Unlike other autonomous parvoviruses that undergo a lytic cycle, wild type AAV does not induce marked cytopathic effects (CPE) and therefore, cellular egress of virions is thought to be primarily driven by overexpression of Adenoviral or Herpesvirus proteins11,12.
Multiple studies have demonstrated that a significant fraction of recombinant AAV associate with extracellular vesicles (EVs) and are released into the supernatant fraction of the cell culture media13. It is well known that cells shed a variety of membrane-bound vesicles varying in size from 20 nm to 1 μm in diameter, which have been termed exosomes, microparticles or microvesicles (collectively referred to here as extracellular vesicles or EVs). Such EVs can package different macromolecules including proteins, nucleic acids and viruses, thereby making them an attractive therapeutic platform14. Despite being non-enveloped viruses, recombinant AAV capsids associated with EVs can enable efficient gene transfer to the retina, the nervous system, the inner ear15–17 and appear shielded from anti-AAV neutralizing antibodies17. However, the exact mechanism by which AAV associates with EVs remains to be determined.
Recent work has revealed a novel +1 frameshifted open reading frame (ORF) in the VP1 region of the AAV cap gene that mediates expression of the membrane-associated accessory protein (MAAP) (Fig. 1A), which was postulated to limit AAV production through competitive exclusion18. Here, we assign a novel function to MAAP in promoting EV mediated AAV egress from host cells. Specifically, we dissect the secondary structure elements in MAAP that contribute to this secretory function. Further, we demonstrate that MAAP transcomplementation can not only rescue but significantly alter the kinetics of viral secretion of multiple AAV serotypes. Further, independent of AAV infection, MAAP alone appears to possess a unique ability to induce hyper-secretion of homogenous EV fractions. Our findings highlight a viral egress mechanism evolved by AAV as well as highlight the potential to exploit this novel protein for loading therapeutic cargo including AAV vectors into EVs and AAV vector production in general.
Results
MAAP is a conserved AAV cap encoded protein with a predicted C-terminal membrane anchoring domain
Confirming that MAAP is a novel viral-encoded protein of unknown function, a pBLAST search of multiple AAV cap gene derived MAAP sequences on the National Center for Biotechnology Information (NCBI) website did not return any proteins with significant homology18. Amino acid sequence alignment of MAAPs derived from different AAV serotypes revealed conserved N- and C-terminal regions containing hydrophobic and basic amino acid residues interconnected by a threonine/serine (T/S) rich region (Fig. 1B). MAAP 3D structures generated using TrRosetta deep learning based modeling19 predicted the following, (i) a conserved N-terminal hydrophobic motif with both alpha helical and beta strand secondary structure elements; (ii) four T/S rich sequence clusters spanning 7-17 residues in length with last two being separated by a smaller alpha helical interspersed with basic residues and (iii) a C-terminal domain defined by another hydrophobic alpha helical motif merging into a cluster of arginine/lysine (R/K) residues (Fig. 1B, C). Importantly, secondary structure software analysis strongly predicts that the C-terminal domain constitutes a putative membrane binding, cationic amphipathic peptide (residues 96-114). It is noteworthy to mention that the secondary structure of MAAP is strikingly similar to the assembly activating protein (AAP), which is similarly encoded downstream from a (+1) frameshifted ORF in the cap gene. When combined with phylogenetic analysis using the neighbor-joining tree method, we observed that MAAPs from AAV serotypes 1,6,8,10,11 were tightly clustered, while other sequences, in particular, MAAP2, 5 and 9 showed significant divergence from other serotypes (Fig. 1D). We then transfected plasmids encoding recombinant MAAPs derived from the VP1 sequences of AAV serotypes 1, 2, 5, 8, and 9 and fused to a C-terminal green fluorescent protein (GFP) in vitro to assess their expression (Fig. 1E) and cellular localization. Fluorescence micrographs confirmed the propensity of MAAP to associate with cell surface membranes as well as subcellular organelles as observed by the punctate patterns throughout the cell (Fig. 1F). Taken together, these data confirm that MAAP is a novel AAV protein predicted to contain cationic amphipathic C-terminal domain potentially critical for membrane anchoring.
MAAP is essential for extracellular secretion of wild type and recombinant AAV particles
We then sought to determine whether MAAP played a role in the synthesis of (i) (pseudo)wild type AAV serotype 8 (i.e., wtAAV8 packaging AAV2 rep and AAV8 cap flanked by AAV2 inverted terminal repeats [ITRs])) (Fig. 2A) and (ii) recombinant AAV8 (i.e., rAAV8 packaging a chicken beta-actin promoter driven luciferase transgene flanked by AAV2 ITRs) (Fig. 2D). To ablate MAAP expression, we mutated the CTG start codon in the MAAP alternative open reading frame (ORF) without affecting the VP1 ORF in both wtAAV8 and rAAV8 plasmids. Culture media and cell pellets were harvested following co-transfection with an Adenovirus helper plasmid (and an additional ITR flanked luciferase encoding transgene cassette in case of rAAV) on days 3 and 5. Strikingly, quantitative PCR of viral genomes revealed a significantly higher (~1 log) amount of extracellular wtAAV8 particles in contrast to MAAPΔ particles recovered from media on day 3 (Fig. 2B). Further, we observed delayed secretion in case of MAAPΔ particles, which were equally apportioned between extracellular and cell lysate fractions on day 5. Although statistically significant, overall viral titers on day 5 were only minimally altered. Of the total virus produced, nearly 70% of wtAAV8 particles were secreted by day 3, while MAAPΔ particles recovered in the extracellular fraction comprised <10% of total (Fig. 2C).
A similar trend was observed in case of rAAV8 particles, with a 4-5 fold higher recovery from media over cell lysate and delayed secretion in case of MAAPΔ particles (Fig. 2E). Of the total virus produced, ~60% of rAAV8 particles were secreted by day 3 in contrast to <10% of MAAPΔ particles (Fig. 2F). Further evaluation of AAV capsid proteins – VP1,2 and 3 by western blot confirmed these results with undetectable to relatively lower levels in the extracellular fraction on days 3 and 5, respectively and correspondingly high(er) cellular retention in case of MAAPΔ particles (Fig. 2G,H). No differences were observed when comparing the transduction efficiency of rAAV8 and MAAPΔ particles in vitro (Fig. 2I). Furthermore, MAAPΔ recombinant virus showed similar VP1, VP2, and VP3 expression ratios and overall virus morphology compared to rAAV8 (Supplementary Fig. 1). Taken together, these results suggest that encoding MAAP from the alternative ORF in VP1 is essential for efficient cellular egress of AAV particles. Another interesting observation is the relatively low recovery of recombinant AAV serotype 9 (rAAV9) (~15%) and MAAPΔ particles (~5%) from media on day 3 (Supplementary Fig. 2). In contrast to rAAV8, rAAV9 particles appear to display delayed secretion as reported previously20, with only a modest difference in cellular egress efficiency compared to MAAPΔ particles. Thus, MAAP8 has a higher propensity to promote viral egress when compared to MAAP9. When taken together with the differences in sequence homology between AAV8 and AAV9, it is tempting to speculate that the cellular egress efficiency of different AAV capsids could be determined by their cognate MAAPs.
Helical and cationic domains are indispensable for MAAP expression and supporting AAV secretion
We next sought to determine the regions and/or domains of MAAP necessary to promote AAV viral egress. We first inserted a 3X-FLAG tag onto the C-terminus of MAAP8 under the control of its endogenous promoter and verified its expression and function (Supplementary Fig. 3). At the primary structure level, MAAP can be separated into multiple regions including N-terminus, Linker, C-terminus, hydrophobic regions 1 and 2, threonine/serine-rich regions 1-4, (T/S), and the basic region. Taking a structure function approach, we generated multiple MAAP8 deletion mutants to systemically dissect the regions of MAAP critical for AAV viral egress (Fig 3A). We discovered that deletions of the N or C-terminus severely diminished MAAP expression (Fig. 3B,C). MAAP deletions involving alpha helical domains or amphipathic domains rich in basic residues demonstrated the most deleterious effects on protein expression at steady state. Furthermore, of the MAAP8 deletion mutants that expressed, only T/S-rich region 2 and T/S-rich region 4 (MAAP8 Δ3 and MAAP8 Δ5 respectively) were found to be dispensable for MAAP function as measured through AAV secretion into the media (Fig. 3D-G). Taken together, these data identify critical secondary structure elements in MAAP that are likely essential for expression and function.
MAAP transcomplementation promotes extracellular secretion of diverse AAV serotypes
To determine whether MAAP expression regulates the secretion of other AAV serotypes, we mutated the CTG start codon in the MAAP ORF for rAAV serotypes 1, 2, 8, and 9, and compared viral titers in extracellular and cellular fractions at day 3 as described earlier. In parallel, we also evaluated whether MAAP transcomplementation could rescue the extracellular secretion of MAAPΔ rAAV particles. To achieve the latter, we expressed MAAP alone from the AAV helper plasmid containing rep and cap genes by mutating the start codons in the VP1,2,3 as well as AAP ORFs. Strikingly, viral titers associated with the cellular fraction were markedly increased for rAAV1, rAAV8 and rAAV9 (~4 to 7 fold), but not rAAV2 (Fig. 4A, C, E, G). In addition, overall recovered titers were increased moderately for the same serotypes (up to 2 fold). In corollary, we observed a striking impact of ablating or supplementing MAAP expression on extracellular vs cell-associated fractions of different AAV serotypes. Specifically, in case of rAAV1, this percentage was reversed from 60:40 to 20:80 upon MAAP ablation and restored to normal upon MAAP expression (Fig. 4B). A similar trend was observed for rAAV8 (~80:20 to 20:80 followed by restoration to normal upon supplementation) (Fig. 4F). Both rAAV2 and rAAV9 showed decreased secretion in general (~35:65) compared to rAAV1/8 with MAAP ablation further reducing extracellular viral titers to 15% and 20%, respectively (Fig. 4D, H). Another important observation is that MAAP8 trans-complementation not only fully rescued the extracellular secretion of rAAV1, rAAV2, and rAAV8 particles, but also doubled the recovery of rAAV9 MAAPΔ particles from media compared to rAAV9 particles (from 40% to 80%) (Fig. 4H). These results confirm the critical role played by MAAP in enabling extracellular secretion of AAV particles in a serotype-independent manner, albeit with different efficiencies. Further, our results demonstrate that trans-complementation of MAAP derived from AAV8 can not only rescue secretion of different AAV serotypes, but also potentially enhance the kinetics of secretion.
MAAP enables AAV egress by hijacking vesicular secretion pathways
To obtain further mechanistic insight into how MAAP enables AAV cellular egress, we first carried out quantitative confocal fluorescence microscopy of cells overexpressing hemagglutinin (HA) tagged MAAP8, which was confirmed separately by western blot analysis (Supplementary Fig. 4). MAAP8-HA colocalized significantly more with the exosomal biogenesis pathway marker Rab1121–23, than the late endo/lysosomal pathway marker Rab724 (Fig. 5A,B). Interestingly, MAAP9, which was associated earlier with slower secretion kinetics showed a similar co-localization to both Rab7 and Rab11 subcellular compartments (Supplementary Fig. 5). Overall, these results underscore the finding that MAAP is a key viral factor that mediates cellular egress of AAV particles through EV secretion pathways, an observation reported by others13,15–17,25,26. Further, our results also highlight potential differences in the structural attributes of different MAAPs that may explain the ability to exploit distinct secretory mechanisms that enables cellular egress of different AAV serotypes. Although the exact components of the vesicular secretion pathway exploited in MAAP-aided AAV egress still need to be elucidated, we carried out biochemical assays to explore potential MAAP-AAV capsid interactions. Notably, we did not find any evidence of direct interaction between MAAP and AAV capsid proteins or MAAP and AAP as determined by immunoprecipitation analysis (Supplementary Fig. 6A,B). However, it is important to note that MAAP8 fused to the BirA biotin ligase BioID227 was able to successfully biotinylate the AAV capsid proteins in vitro indicating that AAV and MAAP do share a proximal interaction within the cell (Supplementary Fig. 6C-F).
To further explore the biology of the MAAP-dependent AAV secretory process, we adopted a gradient centrifugation method to purify EVs from large volumes of cell culture supernatant28,29. Cell culture medium from suspension adapted HEK293 rAAV8/MAAPΔ producing cells (complemented in trans with MAAP8-HA) was processed by successive filtration and centrifugation steps to generate a crude EV pellet that was then separated on an iodixanol gradient (Fig. 5C). The gradient was split into 18 different fractions and probed for EV, MAAP and capsid specific antibodies. Using SDS-PAGE and immunoblot analysis, we found that MAAP associated with multiple EV containing fractions and also strongly associated with an EV/AAV containing fraction (Fig. 5D). To further understand this interaction between MAAP and extracellular vesicles, we purified EVs from the extracellular fractions of rAAV8 or rAAV8/MAAPΔ producing cells using the ExoQuick-TC kit30,31. In contrast to the rAAV8 control, we found negligible association of AAV particles associated with the purified EV fraction in the absence of MAAP as evidenced by western blot analysis of capsid proteins (Supplementary Fig. 7A). Strikingly, we also observed a markedly lower signal for CD8132,33, a marker for EVs in the rAAV8/MAAPΔ samples compared to rAAV8 controls (Supplementary Fig. 7A). These results further led us to explore the function of MAAP alone, unrelated to AAV biology. First, we observed that overexpression of MAAP-HA when compared to an HA tag only control yielded a significantly higher proportion of CD63+ EV fraction (Supplementary Fig. 7B). Second, when overexpressing MAAP8-GFP or a GFP only control, the former was significantly enriched in the purified EV fraction, directly corroborating the role of MAAP in promoting the secretion and loading of cargo into EVs (Supplementary Fig. 7C). Lastly, ultrastructural characterization of purified EV fractions using negative stain transmission electron microscopy (TEM) under different conditions (rAAV8, rAAV8/MAAPΔ, rAAV8/MAAPΔ + MAAP transcomplementation) revealed striking differences in the morphology, homogeneity and abundance of EVs (Supplementary Fig. 8). In particular, MAAP overexpression appears to promote a notable increase in spherical vesicles with a diameter ranging from ~20-50 nm. Interestingly, these vesicles appear to be relatively uniform and homogenous in their composition indicating MAAP may promote secretion of a specific type of extracellular vesicle.
Discussion
Dissecting new aspects of AAV biology is critical towards continued improvement of recombinant vectors for gene therapy applications. For instance, our lab and others previously dissected the functional attributes of the Assembly Activating Protein (AAP), which is encoded from an alternative ORF in the AAV cap gene4,5,34. These studies provided mechanistic insight into AAV capsid assembly in addition to highlighting capsid mutations that can adversely impact assembly and vector yield. Here, we provide structural and functional insights into the recently identified Membrane Associated Accesory Protein (MAAP), which is encoded from a different, alternative ORF within the AAV cap gene. We report that this highly conserved protein displays critical hydrophobic and cationic domains essential for membrane anchoring and promoting extracellular secretion of AAV particles. Although attenuation of MAAP expression does not affect viral yield, we demonstrate that extracellular secretion of AAV particles is drastically delayed with increased cellular retention of AAV particles. Our findings have immediate implications for recombinant AAV vector production, which is currently undergoing a revolutionary manufacturing scale-up phase7,35. Specifically, we demonstrate that MAAP transcomplementation can significantly alter the kinetics of AAV extracellular secretion potentially facilitating downstream purification only from cell culture media. Further, we note that mutation of the cap gene in engineered AAV capsids within the overlapping MAAP reading frame could impact vector yield and timing of harvest from cell lysate versus media fractions. Interestingly, we observed a significant titer increase on Day 3 with MAAP1Δ, MAAP8Δ and MAAP9Δ vectors. One possible explanation for the increased titer is that in cells transfected with MAAPΔ plasmids, empty rAAV capsids are not as readily secreted into the media and are retained in the nucleus, thereby increasing their availability for REP mediated genome packaging. In addition, we also observed that MAAP contributes to rapid secretion of what appears to be a homogenous EV fraction independent of AAV components. While this aspect needs to be explored further, it is exciting to note that MAAP might provide an orthogonal solution to loading therapeutic cargo into EVs.
From a structural perspective, MAAP is rich in alpha helical domains interspersed by T/S rich linker regions and basic residues. The most notable attribute of MAAP is a cationic amphipathic C-terminal domain, which is critical for membrane anchoring and extracellular secretion. Recent studies have revealed that PTGFRN and BASP1 can be utilized as scaffolds for loading various molecules onto the surface or into the lumen of EVs29. BASP1 is an inner membrane leaflet localized protein that utilizes a polybasic effector domain (PED) to bind to the EV membrane. Interestingly, the PED along with N-terminal myristoylation of BASP1 is essential for proper loading of carrier proteins to the inner membrane of EVs. Similarly, the matrix (MA) membrane proximal domain of HIV-1 Pr55GAG contains a highly basic region that is important for efficient membrane binding and proper targeting of MA to the plasma membrane36. MAAP also contains multiple basic regions that are essential for proper molecular function; however whether MAAP undergoes post-translational lipid modification remains to be determined. Nevertheless, our data suggests that similar to BASP1, MAAP can potentially be fused to protein cargo for loading into EVs.
Furthermore, our work has shed light on the interaction between MAAP, AAV and EVs. Using established MISEV criteria37, our work revealed that secreted AAV particles associate with multiple EV markers (CD81, CD63 and CD9), although a fraction of AAV particles appears to be free of any vesicular association. Notably, the EV fractions showed strong MAAP association further implicating this viral protein in vesicle mediated AAV cellular egress. Whether the free particles were previously associated with EVs that lysed or were secreted independently remains the subject of investigation. From a cell biology perspective, our data supports a model wherein MAAP likely exploits molecular interactions within the EV/exosomal pathway. Exactly how MAAP exploits this pathway and how AAV particles exploit these subcellular organelles prior to secretion warrants further investigation. It is plausible that MAAP may enable passive loading of AAV particles into EVs during secretion or alternatively, actively recruit AAV capsids into EVs through an unknown cellular bridging factor. These hypotheses are further corroborated by a consistent colocalization of AAV particles with EV markers and preliminary evidence from proximity ligation. Indeed studies exploring cellular egress of autonomous parvoviruses, such as Minute Virus of Mice (MVM) have shown viral particles to be sequestered into COPII-vesicles in the endoplasmic reticulum (ER) and transported via the Golgi compartment to the plasma membrane38. Remodeling of the cytoskeletal filaments through gelsolin mediated actin fiber degradation in infected cells have been shown essential for egress. Further, several biochemical factors such as sar1, sec24, rab1, the ERM family proteins, radixin and moesin have been implicated in this process of cellular exocytosis39. However, it should be noted that a virulence factor akin to MAAP in other parvoviruses has not been identified to date. In summary, we conclude that while dependoparvoviruses such as AAV rely on helper coinfection to trigger viral release through a lytic process, cellular egress likely accelerates and regulates viral dissemination and spread. In summary, our studies unequivocally implicate MAAP as a novel AAV egress factor.
Methods
Plasmid Constructs
MAAP DNA sequences from AAV serotypes 1,2,5,8 and 9 were synthesized and cloned into pcDNA3.1(+)-C-HA and pcDNA3.1(+)-C-eGFP expression vectors using HindIII and XbaI sites for MAAP 1,2,8,9 and EcoRV sites for MAAP5 (Genscript). All MAAP expression constructs were synthesized and cloned with an ATG start codon. The AAV8-Rep/Cap-VP* plasmid is a AAV2-Rep/AAV8-Cap plasmid with the start codons of VP1/2/3 and AAP mutated by site directed mutagenesis to prevent expression. The AAV8-Rep/Cap-MVP* additionally has a mutated MAAP start codon to prevent MAAP expression. The AAV8-MAAPΔ plasmid is a 2 AAV2-Rep/AAV8-Cap plasmid with the start codons of MAAP mutated by site directed mutagenesis to prevent expression. The AAV8-MAAP-3X-FLAG plasmid was generated by utilizing site-directed mutagenesis to incorporate a 3X-FLAG tag in frame onto the C-terminus of MAAP8 in the AAV8-Rep/Cap-VP* plasmid. AAV8-MAAP-3X-FLAG truncation mutants were generated through site directed mutagenesis. MAAP8 was cloned into the MCS-13X-Linker-BioID2-HA expression vector (Addgene #80899) using NheI and Age1 sites. All plasmid constructs were verified by DNA sequencing analysis.
Bioinformatics analysis and structural models
The amino acid sequences of 15 AAV serotypes were retrieved from GenBank. MAAP start and stop sites were defined as previously described18. Protein sequences were aligned using the ClustalW multiple-alignment tool40 and generated using Unipro UGENE software41. MAAP amino acid sequences from multiple AAV isolates were aligned using ClustalW, and phylogenetic trees were generated using the MEGAv7.0.21 software package42. The phylogeny was produced using the neighbor-joining algorithm, and amino acid distances were calculated using a Poisson correction43. Statistical testing was done by bootstrapping with 1,000 replicates to test the confidence of the phylogenetic analysis and to generate the original tree44. The percentage of replicate trees in which associated taxa clustered together in the bootstrap test is displayed next to the branches. Secondary structural elements were predicted using the JPred tool45. To predict membrane-binding, amphipathic α-helices, we used Amphipaseek (parameters: high specificity/low sensitivity)46. MAAP structural models were generated using the TrRosetta deep learning based modeling method19. Secondary structural depictions of these models were visualized using the PyMOL Molecular Graphics System (Schrödinger; https://www.pymol.org/2/).
Cellular assays, Immunoprecipitations and Western Blotting
For protein expression analysis, HEK293 cells seeded overnight in 6-well plates at a density of 3 × 105 cells per plate were transfected with a total of 2 μg DNA as indicated. HEK293 cell pellets overexpressing MAAP-HA, MAAP-GFP or expressing MAAP8-3X-FLAG were recovered 72 h post-transfection. Pellets were lysed in RIPA buffer with 1x Halt Protease Inhibitor (ThermoFisher) for 45 minutes at 4°C. Lysates were spun at max speed for 10 minutes at 4°C to remove cellular debris. 1X LDS sample buffer with 10mM DTT were added to cleared lysates and boiled for 2 min. Samples of cleared lysate were ran on Mini-Protean TGX 4-15% gels (Biorad), transferred onto PVDF with the Trans-Blot Turbo system (BioRad) and blocked in 5% milk/1x TBST. Blots were probed with mouse monoclonal anti-GFP antibody (1:1000 dilution, SC9996; Santa Cruz Biotechnology), rabbit polyclonal anti-HA SG77 antibody (1:1000 dilution, 71-5500; ThermoFisher Scientific), mouse monoclonal B1 hybridoma supernatant (1:50, 03-65158; ARP), mouse monoclonal anti-FLAG M2 antibody (1:1000 dilution, F180450UG; Sigma), mouse monoclonal anti-beta actin (1:1000 dilution, 8226; Abcam) as the primary antibody. Following three 1x TBST washes, samples were incubated with secondary antibodies conjugated to HRP at 1:20,000 in 5% milk/1x TBST for 1 hour (goat anti-mouse-HRP, 32430; ThermoFisher Scientific and goat anti-rabbit-HRP, 111-035-003; Jackson ImmunoResearch). Blots were developed using SuperSignal West Femto substrate (ThermoFisherScientific/Life Technologies) according to manufacturer instructions. For immunoprecipitation studies, HEK293 cells were transfected with pXR9 and MAAP9-pcDNA3.1(+)-C-HA for 72 hrs, then washed with 1X PBS and harvested in NP-40 with 1x Halt Protease Inhibitor (ThermoFisher) for 1 hr at 4°C. Lysates were spun at max speed for 20 minutes at 4°C to remove cellular debris. 10μL (2.5μg) of anti-HA SG77 antibody were added to 500 μL cleared lysate and incubated at 4°C for 3 hrs with nutation. Added 40 μL of pre-washed Protein G magnetic beads to pre-cleared lysate with antibody and carried out immunoprecipitations overnight on a nutator at 4°C. Bound protein was eluted in 10mM DTT and 1x LDS for 5 min at 95°C. Samples were then analyzed via SDS-PAGE (NuPAGE 4-12% Bis-Tris Gel) and transferred onto nitrocellulose membrane (ThermoScientific). Following blocking in 5% milk/1xTBST, samples were incubated with primary antibodies to either capsid (mouse monoclonal B1 hybridoma supernatant, 1:250, 65158; Progen), actin (mouse monoclonal anti-beta actin, 1:1000 dilution, 8226; Abcam), CD81 (mouse monoclonal anti CD81 M38, 1:1000, 10630D), CD63 (mouse monoclonal anti CD63 Ts63, 1:1000, 10628D), CD9 (mouse monoclonal anti CD9 Ts9, 1:1000, 10626D) or MAAP (mouse monoclonal anti-HA HA.C5 antibody, 1:1000 dilution, MA5-27543; ThermoFisher Scientific) overnight in 5% milk/1xTBST. Following three 1x TBST washes, samples were incubated with secondary anti-mouse antibody conjugated to HRP (goat anti-mouse-HRP, 32430; ThermoFisher Scientific) at 1:20,000 in 5% milk/1xTBST for 1 hour. The signal was the visualized via SuperSignal West Femto Maximum Sensitivity substrate (ThermoScientific) according to manufacturer instructions.
BioID2 expression and pulldown
For protein expression analysis, HEK293 cells seeded overnight in 6-well plates at a density of 3 × 105 cells per plate were transfected with a total of 2 μg of 13X-BioID2-HA or MAAP8-13X-BioID2-HA DNA. Cells were then supplemented with 50 μM biotin 24 hrs post transfection and allowed to label for 20 hrs. Cell pellets were recovered 20 hrs biotin supplementation and were lysed in RIPA buffer with 1x Halt Protease Inhibitor (ThermoFisher) for 45 minutes at 4°C. Lysates were spun at max speed for 10 minutes at 4°C to remove cellular debris. 1X LDS sample buffer with 10mM DTT were added to cleared lysates and boiled for 2 min. Samples of cleared lysate were ran on Mini-Protean TGX 4-15% gels (Biorad), transferred onto PVDF with the Trans-Blot Turbo system (BioRad) and blocked in 5% milk/1x TBST. Blots were probed with rabbit polyclonal anti-HA SG77 antibody (1:1000 dilution, 71-5500; ThermoFisher Scientific), mouse monoclonal anti-beta actin (1:1000 dilution, 8226; Abcam), and goat polyclonal anti-biotin (1:1000 dilution, 31852; ThermoFisher) as the primary antibody. Following three 1x TBST washes, samples were incubated with secondary antibodies conjugated to HRP at 1:20,000 in 5% milk/1x TBST for 1 hour (goat anti-mouse-HRP, 32430; ThermoFisher Scientific, goat anti-rabbit-HRP, 111-035-003; Jackson ImmunoResearch, mouse anti-goat-HRP, 31400, ThermoFisher).
Streptavidin pulldowns were performed as previously described47 with the following changes. HEK293 cells for each condition were grown on 2 x 15 cm2 plates until ~70% confluent and were transfected with 34 μg of plasmid DNA per plate. Each plate was transfected with 13X-BioID2 or MAAP8-13X-BioID2 along with pXX680, pTR-CBA-Luciferase, and AAV8-MAAPΔ. Media was supplemented with 50 μM biotin 48 hours post transfection and cells were harvested 20 hours post biotin supplementation. Biotinylated proteins were pulled down with Pierce High Capacity Streptavidin Agarose (20357; ThermoFisher) and were separated by SDS-PAGE and visualized using a SilverXpress silver stain kit (LC6100; ThermoFisher) according to the manufacturer’s instructions. Biotinylated proteins were also visualized by immunoblot and probed with rabbit polyclonal anti-HA SG77 antibody (1:1000 dilution, 71-5500; ThermoFisher Scientific), mouse monoclonal anti-beta actin (1:1000 dilution, 8226; Abcam), goat polyclonal anti-biotin (1:1000 dilution, 31852; ThermoFisher) and mouse monoclonal B1 hybridoma supernatant (1:50, 03-65158; ARP) as the primary antibody.
Recombinant and wild type AAV production, purification, and quantification
HEK293 (human embryonic kidney cells obtained from the University of North Carolina Vector Core) were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100U/ml penicillin, 100ug/ml streptomycin. Cells were maintained in 5% CO2 at 37°C. Recombinant AAV vectors were produced by transfecting HEK293 cells at ~75% confluence with polyethylenimine (PEI Max; Polysciences, 24765) using a triple plasmid transfection protocol with the AAV Rep-Cap plasmid, Adenoviral helper plasmid (pXX680), and single-stranded genomes encoding firefly luciferase driven by the chicken beta-actin promoter (ssCBA-Luc) or self-complementary green fluorescence protein driven by a hybrid chicken beta-actin promoter (scCBh-GFP), flanked by AAV2 inverted terminal repeat (ITR) sequences. Viral vectors were harvested from media and purified via iodixanol density gradient ultracentrifugation followed by phosphate buffered saline (PBS) buffer exchange. Titers of purified virus preparations were determined by quantitative PCR using a Roche Lightcycler 480 (Roche Applied Sciences, Pleasanton, CA) with primers amplifying the AAV2 ITR regions (forward, 5’-AACATGCTACGCAGAGAGGGAGTGG-3’; reverse, 5’-CATGAGACAAGGAACCCCTAGTGATGGAG-3’) (IDT Technologies, Ames IA).
Quantitative PCR analysis of AAV vector yield
HEK293 cells in six-well plates were transfected using PEI at ~75% confluence with Adenovirus helper plasmid (1μg), WT or MAAPΔ AAV-Rep/Cap plasmid (1μg), ITR-transgene plasmid (500ng), and AAV8-Rep/Cap-VP* or -MVP* (500ng). The AAV8-Rep/Cap-VP* plasmid is a AAV2-Rep/AAV8-Cap plasmid with the start codons of VP1/2/3 and AAP mutated to prevent expression. The AAV8-Rep/Cap-MVP* additionally has a mutated MAAP start codon. For 3-day experiments, media and cells were collected three days post-transfection. Cells were lysed by vortexing in a mild lysis buffer (10mM Tris-HCl, 10mM MgCl2, 2mM CaCl2, 0.5% Triton X-100 supplemented with DNAse, RNAse, and Halt Protease Inhibitor Cocktail) and incubated at 37°C for 1hr. Lysates were cleared by centrifuging at 21,000 rcf for two minutes. NaCl to 300mM was added to AAV2 lysates prior to centrifugation to prevent virus binding of the cell debris. Collected media and cleared lysates were assayed with qPCR for DNAse-resistant viral genomes as described above. For Day 3 and 5 experiments, media was collected and replaced on the first indicated day, and cells and media were harvested as described on the last day. For WT-like AAV8 transfections, cells were transfected with Adenovirus helper plasmid (1.9μg) and WT or MAAPΔ ITR-AAV2-Rep/AAV8-Cap-ITR plasmid (0.9μg), with media and cells collected on days 3 and 5 post-transfection as described above.
Confocal fluorescence microscopy
HEK293 cells were seeded on slide covers in 24-well plates at a density of 5e4 cells/well and allowed to adhere overnight. Cells were then co-transfected with Rab7-GFP, Rab11-GFP, MAAP8-pcDNA3.1(+)-C-HA and MAAP9-pcDNA3.1(+)-C-HA, and were incubated for 48 hrs at 37°C and 5% CO2. Cells were then fixed with 4% paraformaldehyde for 30 minutes and permeabilized with 0.1% Triton X-100 for 30 minutes. Following 1 hr of blocking with 5% Normal Goat Serum, cells were stained with rabbit polyclonal anti-HA SG77 antibody (1:100 dilution, 71-5500; ThermoFisher Scientific) primary for one hour, washed 3x with PBS, and then stained with fluorescent goat anti-rabbit Alexa Fluor 647 (1:400 dilution, ab150079; Abcam) secondary antibody and washed 1x with PBS. Cells were subject to 5 minutes staining with DAPI, and then mounted in Prolong Diamond (Invitrogen) and imaged using a Zeiss LSM 880 Airyscan confocal microscope. Co-localization analysis was performed by cropping the whole compartment (Rab11 or FIP3), or the whole cell, using Zeiss ZEN software with the Colocalization function. Threshold was automatically determined using the Costes method autothreshold determination. Pearson’s correlation coefficient was calculated for the analysis. Statistical analyses were carried out by the nonparametrical Mann–Whitney U test using Prism software (GraphPad).
Extracellular vesicle Isolation
For Figures 5C-D and Supplementary Figure 7D-H, a previously established EV isolation protocol was utilized29. Suspension adapted HEK293 (human embryonic kidney cells obtained from the University of North Carolina Vector Core) cells were grown in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100U/ml penicillin, 100ug/ml streptomycin and inoculated at a density of 5E+5 viable cells per milliliter. Cells were maintained in 5% CO2 at 37°C and grown overnight to a density of 1E+6 cells per milliliter for transfection. Cells were then grown for 3 days with viability measured daily. Cell density and viability were measured on a Countess II Automated Cell Counter (Invitrogen). Upon culture termination, cells were removed by centrifugation at 4,000 x g for 30 minutes at room temperature. Harvests were then concentrated using tangential flow filtration using a Minikros (100 kD) modified polyethersulfone (PES) hollow fiber filter (Repligen) on a KROSFLO KMPI system. Concentrated medium was then supplemented with 1mM MgCl2 and benzonase (20 U/mL, Sigma Aldrich) to digest extravesicular nucleic acids. Nuclease treated media was next transferred to 38-mL Ultra-Clear centrifuge tubes (Beckman Coulter) and centrifuged for 60 min at 133,000 x g at 4°C in a Sure Spin 630 36 mL rotor (Thermo Scientific). The media was discarded and the crude pellet was resuspended in a minimal volume of PBS. The resuspended pellet was then mixed with a 60% iodixanol solution (Optiprep, Sigma) and layered onto the bottom of a 38-mL Ultra-Clear centrifuge tube (Beckman Coulter). Lower-density solutions were prepared by diluting homogenization buffer (250 mM sucrose, 10 mM Tris-HCl, 1 mM EDTA, pH 7.4) to yield final iodixanol concentrations (vol/vol) of 45%, 30%, 23%, and 18%. Successive layers of 30% (9 mL), 23% (6 mL) and 18% (6 mL) iodixanol solutions were carefully pipetted on top. Resuspended pellet in PBS was added on top of the gradient. The gradient was ultracentrifuged in a Sure Spin 630 36 mL rotor (Thermo Scientific) for 16 h at 150,000 x g and 4°C to separate EVs from other cell culture supernatant contaminants. The gradient was then fractionated in 2 mL fractions and analyzed by biochemical analysis. For Supplementary Figure 7A-C and Supplementary Figure 8, EVs were isolated from tissue culture media using a commercial kit containing a polyethylene glycol (PEG) solution (ExoQuick-TC ULTRA kit, EQULTRA-20TC-1; System Biosciences) according to the manufacturer’s instructions.
Transmission electron microscopy
Isolated vesicular fractions or AAV viral particles (1×1010 vg) in 1x PBS samples were adsorbed onto 400 mesh, carbon coated grids (Electron Microscopy Sciences) for 2 min and briefly stained with 1% uranyl acetate (Electron Microscopy Sciences) diluted in 50% ethanol. After drying, grids were imaged with a Philips CM12 electron microscope operated at 80kV. Images were collected on an AMT camera.
Luciferase expression assays
A total of 1×104 HEK293 cells in 50μL DMEM + 10% FBS + penicillin streptomycin was then added to each well, and the plates were incubated in 5% CO2 at 37 °C for 24h. Cells were then transduced with AAV8-ssCBA-Luc vectors at a dose of 10,000 and 50,000 vg/cell. 48 hours post transduction, cells were then harvested and lysed with 25μL of 1×passive lysis buffer (Promega) for 30 min at room temperature. Luciferase activity was measured on a Victor 3 multi-label plate reader (PerkinElmer) immediately after the addition of 25μL of luciferin (Promega).
Statistical Analysis
Where appropriate, data are represented as mean or mean ± standard deviation. For data sets with two groups (Figure 2B and 2E; Supplemental Figure 1A), comparisons were made between all groups and significance was determined using a two-way ANOVA followed by a Sidak’s post-test. For data sets with at least three groups (Figure 3), comparisons were made between all groups and significance was determined by two-way ANOVA, with Tukey’s post-test. For analysis of confocal microscopy data (Figure 4A; Supplemental Figure 4A) significance was determined by a Mann-Whitney rank test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.
Author Contributions
ZE, PH, and AA designed all experiments and interpreted the data. ZE, PH and DO carried out all molecular biology, virus production and imaging studies. ZE, PH, HV and AA wrote the manuscript.
Competing Interests
AA and ZE have filed patent applications on the subject matter of this manuscript. AA is a founder at StrideBio, TorqueBio and WardenBio.
Acknowledgements
This study was funded by NIH grants awarded to A.A. (R01HL089221, UG3AR07336, R01GM127708; R01NS099371). We would like to acknowledge Ricardo Vancini and the Duke Research Electron Microscopy Service for their help in the preparation and imaging of TEM samples. We would also like to acknowledge Dana Elmore for her assistance with data analysis.