Adeno-Associated Virus and Hematopoietic Stem Cells: The Potential of Adeno-Associated Virus Hematopoietic Stem Cells in Genetic Medicines

Introduction

I was first introduced to adeno-associated virus (AAV) in the laboratory of James Rose at NIAID, NIH, during a postdoctoral fellowship. Intense discussions of AAV biology and its lifecycle were peppered with interesting anecdotes about the founding scientists who discovered AAV and elucidated the organization, expression, and replication, and the biology of this unique nonpathogenic, replication-deficient parvovirus. This is where I first encountered Barrie Carter, a neighbor at NIH, who played an instrumental role in deciphering every aspect of the AAV lifecycle from its genome organization, transcription, and replication to the role of the AAV-encoded proteins and the very first cloning of the AAV genome in a plasmid. ^{1 –10}

It was the molecular cloning of AAV ^{5,11 –13} (Fig. 1) that jumpstarted the path that led to the advent of the role of AAV in gene therapy. Indeed, the Carter group along with the Muzyzcka group, were the first to describe the use of recombinant AAV as a gene transfer vector. ^{12,14 –16} These stalwart scientists persisted in studying this innocuous virus that survived without killing its host or leaving a significant footprint. Studies on its compact genome that encodes three capsid proteins, four replication proteins, and accessory proteins, including assembly activating protein (AAP) and potentially protein X, revealed an efficient coding mechanism of overlapping reading frames and judicious use of a splice site to encode multiple proteins. ^{17 –19} Similarly, the efficient regulation of protein expression through the use of a start codon within an alternative splice site for VP1 and an ACG start codon for VP2 ^20,21 were just some of the lessons learned from studying AAV biology. Also fascinating was the unique single-stranded DNA genome of AAV bound by palindromic, high G:C inverted terminal repeats (ITRs) at either end, which folded back on themselves and served as primers for replication, a mechanism famously termed “boustrophedonic.” ²² With his foundational research on every aspect of AAV, Barrie Carter, along with a select few, laid the groundwork for our understanding of the life cycle, gene expression, and the tremendous potentials of AAV as we know it today. The first human trial of recombinant AAV vectors was initiated by Flotte and Carter in 1995, when adults with cystic fibrosis were treated with endobronchial instillation of AAV2, ¹⁰ which set the stage for the phase 2a and phase 2b trials of aerosolization. These studies marked the start of the AAV genetic medicine revolution of today. In his post-NIH years, Barrie continued to shepherd AAV gene therapy through biotechnology, commercial manufacturing, and beyond. With his groundbreaking work on the early studies that defined the AAV genome, the biology of AAV-encoded proteins—to the very first human clinical trial, Barrie Carter is one of the few scientists who played an instrumental role in the evolution of the field from the very beginning. Little did we know then that this replication defective, nonpathogenic virus would have such a major impact on novel therapies for incurable diseases and would revolutionize genetic medicine. Today, there are 205 AAV-based protocols for clinical trials registered at Clinicaltrials.gov. Regulatory agencies have recently approved several AAV “drugs” that are benefitting patients and extending lives, including, Glybera, ²³ Luxturna, ²⁴ and Zolgensma. ²⁵

OPEN IN VIEWER

Figure 1.

Time line of major events in AAV and HSC research. AAV, adeno-associated virus; HSC, hematopoietic stem cell.

AAV Antisense Vectors for the Inhibition of HIV Infection

The early 1990s, while the incidence of HIV infection was growing exponentially, efforts to design molecular and genetic strategies to inhibit HIV replication were being initiated. We decided to test whether we could induce intracellular resistance to HIV replication by transducing cells with an AAV2 vector encoding an antisense RNA to a region common to the HIV-1 TAR/3′UTR (untranslated region), which was present in all HIV-1 transcripts and genomic RNA. We showed that T cell lines transduced with the AAV antisense vector exhibited potent reduction in HIV-1 gene expression and infectious virus production, suggesting that this was a promising approach to gene therapy for HIV-1 infection ²⁶ (Fig. 1). However, it became obvious that simply transducing differentiated T cells alone would not be sufficient for long-term efficacy. Since macrophages and T cells served as natural targets of HIV-1 infection, AAV transduction of stem cell progenitors that differentiated into HIV targets would be necessary for long-term protection. This led to our work on exploring AAV transduction of hematopoietic stem cells (HSCs).

Early work on AAV gene therapy showed that it was a promising approach for gene delivery to somatic cells. ^{12,14 –16,27,28} Stem cells, on the other hand, which comprised the ideal target for genetic correction of diseases of the hematopoietic system, were also more challenging. Successful transduction of primitive HSCs capable of both self-renewal and multilineage differentiation held the promise of permanent cure for gene-based blood diseases. However, attempts to transduce the elusive HSC were thwarted by several significant obstacles: (1) true pluripotent stem cells are rare, (2) they are difficult to identify and even more difficult to purify, and (3) in vitro manipulation resulted in loss of the key stem cell properties of self-renewal and pluripotentiality.

AAV Vectors Transduce Non-Dividing Cells

Since primitive HSCs were known to exist in a quiescent state, it became important to determine whether AAV vectors were capable of transducing nondividing cells. We tested the ability of AAV vectors to transduce cells rendered nonproliferating either by treatment with the DNA synthesis inhibitors or contact inhibition and serum starvation. Similar levels of transgene expression in both actively dividing and nondividing cells indicated that AAV vectors were capable of transducing nondividing cells ²⁹ (Fig. 1). This was the first demonstration of AAV transduction of nondividing cells and indicated that the early steps of AAV vector transduction, including virus attachment, nuclear transport, uncoating, and second-strand synthesis, occurred successfully in nondividing cells. These results also suggested that AAV transduction of primitive quiescent HSCs may be feasible.

Gene Transfer Into CD34⁺ Cells

The vast majority of the early work on gene delivery to HSCs was being performed with retroviral vectors. CD34⁺ HSC could be primed to proliferate in vitro by treatment with cytokines, rendering them good targets for retroviral transduction since retroviral genomes underwent efficient reverse transcription and integration only in dividing cells. Thus, protocols were developed to transduce ex vivo isolated CD34⁺ HSC with retroviral vectors. Culture conditions optimized for retrovirus transduction often employed prestimulation of cells followed by culture at high cytokine concentrations, ensuring rapid cycling of CD34⁺ cells and resulting in efficient retrovirus transduction. ³⁰ However, it soon became clear that efficient retrovirus transduction of CD34⁺ cells in vitro did not correlate with long-term multilineage engraftment in vivo, and that in vitro culture at high cytokine concentrations resulted in the loss of self-renewal and multipotential properties of CD34⁺ cells.

AAV Transduction of CD34⁺ Cell In Vitro

It was also becoming increasingly clear that culture conditions optimized for retrovirus transduction of CD34⁺ cells did not result in efficient AAV transduction. Episomal AAV genomes were lost from rapidly dividing cells, making them poor targets for AAV transduction. It was necessary to identify culture conditions that would enable AAV transduction, while preserving viability and the multipotential nature of CD34⁺ cells. Using assays for reporter transgene expression, fluorescent in situ hybridization (FISH) and Southern analyses, we showed that very low cytokine concentrations were best for AAV transduction of CD34⁺ cells ^{31 –33} (Fig. 1). Higher cytokine concentrations that promoted rapid cell division resulted in poor AAV transduction of CD34⁺ HSC from human marrow and cord blood. We also showed that prestimulation of CD34⁺ cells was not required and was actually deleterious to AAV transduction. Comparisons of AAV transduction of rapidly and slowly dividing CD34⁺ subpopulations revealed that AAV vectors persisted longer in the more quiescent CD34⁺ subsets, while a loss of episomal AAV genomes was observed in rapidly dividing CD34⁺ cells ³⁴ (Fig. 1). CD34⁺ cells residing in the G0 phase of the cell cycle comprised the best targets for AAV transduction. Contemporaneously, several groups showed that AAV2 vectors transduced murine, nonhuman primate (NHP) and human hematopoietic stem and progenitor cells. ^{14,35 –41}

Using these culture conditions, we subsequently showed that AAV vectors transduced primitive CD34⁺ HSC capable of giving rise to differentiated hematopoietic colonies in long-term cultures up to 10 to 12 weeks after culture initiation. ³³ CD34⁺ cells showed chromosome-associated AAV signals by FISH. On evaluating AAV vector genomes in interphase nuclei using vector-specific fluorescent probes, we identified multiple episomal AAV vector genomes per nucleus. Interestingly, the episomal vector signals were found to be located at the periphery of the nuclei, colocalizing with heterochromatin. ³²

We showed that the more primitive marrow-derived CD34⁺CD38⁻ subpopulation of HSCs were the most quiescent subpopulation of CD34⁺ cells and primarily existed in the G0 phase of the cell cycle. ³⁴ This subset of primitive HSCs was enriched for hematopoietic progenitor activity and was transduced well by AAV vectors. Notably, we also showed that, although rapidly dividing CD34⁺ cells were transduced by AAV, the vector genomes were rapidly lost upon cell division. ³⁴

Importantly, these experiments explained the inability to successfully transduce CD34⁺ cells with AAV using culture conditions optimized for retrovirus vectors. When testing AAV transduction of CD34⁺ HSC, many groups used culture conditions optimized for retrovirus transduction. However these culture conditions promoted rapid cell division, which led to a loss of episomal AAV genomes. As a result of transducing rapidly cycling CD34⁺ cells, episomal AAV vector genomes were lost and stable transduction was not observed. The use of high concentrations of cytokines, while good for retrovirus transduction, resulted in poor AAV transduction of CD34⁺ cells. Somewhat counterintuitively, cytokine prestimulation of CD34⁺ cells is often still performed before AAV transduction, resulting in a loss of episomal AAV genomes from the rapidly dividing cells. However, AAV transduction of unstimulated, quiescent HSCs, ideally followed shortly by in vivo transplantation, resulted in successful long-term gene transfer.

In contrast to the readily detectable gene marking observed after AAV2 transduction, transgene expression appeared particularly poor in vitro in primitive quiescent CD34⁺CD38⁻ HSC at early time points after transduction. Several rate-limiting steps to efficient transgene expression have been demonstrated after AAV2 transduction, ^{40 –42} including donor variability in cell surface receptors for AAV, blocks to nuclear entry, uncoating and second-strand synthesis. Second-strand synthesis has been shown to be blocked when the cellular protein FKBP52 is bound to the D sequence of the AAV2 ITR. Pretreatment of cells with the receptor tyrosine kinase inhibitor tyrphostin A1 results in the phosphorylation and subsequent removal of FKBP52 from the D sequence, thereby allowing second-strand synthesis to proceed. ⁴³ We showed that second-strand synthesis is rate limiting for transgene expression in CD34⁺ HSC, and that tyrphostin treatment of cells leads to accelerated second-strand synthesis and augmented transgene expression. ³⁴

The Srivastava group showed that mutagenesis of surface exposed tyrosine residues to phenylalanine on AAV capsids reduced phosphorylation, ubiquitination, and proteosomal degradation of AAV virions, and resulted in increased trafficking to the nucleus and marked enhancement of transduction. ⁴⁴ In a series of collaborative studies, we showed that tyrosine modified AAV2 and AAV6 vectors showed significantly enhanced transduction in vitro and in vivo compared with wild-type AAV ^45,46 (Fig. 1). Transduction and persistence of AAV genomes in xenografted HSCs was evident long term in vivo. Similar results were observed in other tissues by other groups, ⁴⁷ suggesting that natural rate limitations to AAV transduction could be bypassed and would lead to further improvements in transduction.

Donor Variability

Our results indicated that most donors evaluated showed evidence of some level of AAV transduction. However, CD34⁺ cells from certain donors showed significantly lower levels of transduction than others. ^{31 –34} These results are in concordance with those of Ponnazhagan et al. ³⁹ who reported donor-to-donor variability of CD34 transduction by AAV vectors. Whether it is due to polymorphism of receptor expression or other factors remains unknown.

AAV Transduction of In Vivo Engrafting Self-Renewing Multipotential HSCs

Using a xenotransplant model, we showed that an AAV2 vector encoding the antisense RNA to HIV-1 transduced primitive highly purified CD34⁺CD38⁻ HSCs capable of engrafting in immune-deficient mice. ⁴⁸ Transplantation of AAV-transduced primitive quiescent CD34⁺CD38⁻ subset of HSCs led to long-term multilineage engraftment. AAV-transduced HSCs gave rise to cells of different hematopoietic lineages, including myeloid and lymphoid lineages, and CD34⁺ progenitors. The presence of AAV-marked cells of different hematopoietic lineages was detected to 6 months post-transplantation, confirming the multipotential nature of the AAV-transduced HSCs. Importantly, we showed that AAV-transduced HSCs were capable of serial engraftment, indicating their ability to self-renew. We additionally showed that primitive AAV-transduced HSCs harvested from the marrow of xenograft recipients retained the ability to give rise to multilineage hematopoietic colonies in vitro. These studies definitively showed that AAV2 vectors were capable of transducing primitive quiescent multipotential human HSCs capable of self-renewal and differentiation ⁴⁸ (Fig. 1).

AAV vectors are known to persist primarily in an episomal state following transduction. Chromosomal integration occurs at a very low frequency. This raises the question of how AAV vectors survive in HSCs. Since primitive HSCs are primarily quiescent and survive in a nondividing state for years, it is likely that AAV episomes persist for prolonged periods. It is of interest to note that stem cells are thought to possess an efficient DNA break repair system, ^49,50 and this may be responsible for the persistence of AAV in primitive quiescent stem cells. HSCs, in particular, have highly active DNA repair machinery, probably to guard against the accumulation of mutations and to ensure continued genomic integrity. ^{49 –51} Genes of DNA repair pathways have been found to be consistently overexpressed in stem cells. Whether these proteins or pathways play a role in AAV persistence in primitive HSCs is unknown.

Novel AAV Serotypes

Although most of the early work on AAV vectors was done with AAV2, in the early 2000s, evaluation of other AAV serotypes for gene transfer was underway. ⁵² Guangping Gao and James Wilson reported the identification and isolation of many naturally occurring novel AAV capsids from human and NHP tissues, including the human bone marrow. ^53,54

AAV6 Vectors Transduce CD34⁺ HSC Efficiently

In a collaborative study with the Srivastava group, we evaluated AAV6 vectors for transduction of CD34⁺ HSC. Our results showed that AAV6 transduced CD34⁺ HSC more efficiently than AAV2 in both short-term in vitro as well as long-term in vivo assays ⁴⁶ (Fig. 1). Transduction of CD34⁺ cells with AAV6 resulted in ∼20-fold greater genome copies per cell at 4–6 months post-transplantation compared with AAV2. Tyrosine mutants of AAV6 showed ∼40-fold higher transduction than AAV2. ⁴⁶ These studies indicated that different serotypes and the use of alternate capsids had the potential to further improve AAV transduction of CD34⁺ HSC. AAV6 is currently used widely for the delivery of donor DNA in conjunction with nuclease-based editing platforms, including CRISPR and zinc finger nucleases. ^{54 –57}

The Stem Cell-Derived AAVHSC

AAVHSCs mediate seamless, accurate genome editing in the absence of exogenous nucleases

Previous studies had demonstrated that AAV is capable of mediating genome editing without the requirement for double-stranded DNA breaks by exogenous nucleases. ^{62 –67} However, the reported editing frequencies were low, generally well below 0.5%. Upon testing the capacity of AAVHSC vectors to mediate genome editing, we observed that AAVs mapping to Clade F, including AAVHSC and AAV9, mediated efficient genome editing without the need for prior nuclease-mediated DNA breaks ⁶⁸ (Fig. 2).

OPEN IN VIEWER

Figure 2.

Events in the AAVHSC life cycle leading to genome editing: (1) binding of AAVHSC virions to cell surface receptor and co-receptors. (2) Endosomal entry of AAVHSCs. (3) Release of AAVHSCs from endosomes. (4) Entry of AAV through the nuclear pore complex. ^76,77 (5) Uncoating of AAVHSC virions and release of the single-stranded vector genome. (6) The single-stranded AAV vector genome with homology region and ITRs. (7) Recruitment of DNA repair proteins to the AAV editing genome. (8) Formation of nucleoprotein filament complex. (9) Homology search between editing vector genome and chromosomal DNA. ITRs, inverted terminal repeats.

Interestingly, we found that the lower efficiency of editing of AAV2, AAV6, and AAV8 was not due to their inability to transduce cells or deliver genomes to the nuclei, as AAV6 is known to transduce CD34⁺ and other cells efficiently. ⁴⁶ Since the genomes of all editing vectors were identical and based on AAV2, we concluded that capsid proteins play a role in determining editing efficiency. The role of AAV capsid proteins in modulation editing efficiency is under investigation. Figure 2 summarizes the potential sequence of events leading to AAV editing.

Notably, efficient in vitro editing was dependent upon high-quality, high-titered vector stocks and the use of high multiplicities of infection (MOIs), at or above 150,000 vector genomes/cells. ⁶⁸ Attempts to replicate AAVHSC editing at low MOIs with poor-quality AAV vector batches that cause toxicity have predictably failed. ^{67 –69} Rogers et al. ⁷⁰ and Dudek and Porteus ⁷¹ failed to observe Clade F AAV-mediated non-nuclease editing. However, their experiments were primarily performed at MOI: 5,000–10,000, due to toxicity of their vector stocks. These results are actually in concordance with Smith et al., ^68,69 where we showed that high MOIs were required to observe efficient editing. Vector preparations used for editing must be of sufficiently high quality such that no toxicity due to contaminants is observed at the high multiplicities required for editing. ⁶⁹

At the molecular level, editing was found to be seamless and highly accurate, with no evidence of on-target mutations. ⁶⁸ AAVHSC-based editing was guided by homology arms complementary to the genomic target site and operated exclusively by homologous recombination (HR), as evidenced by the absolute requirement for BRCA2, an essential mediator of HR. The complete absence of any evidence of NHEJ-based repair, including insertion/deletions and incorporation of AAV ITR, supported these conclusions. We showed editing at multiple genomic loci, including AAVS1 and Rosa26, and in a variety of cell types. AAVHSC-based editing was observed to be most efficient in primary cells and in vivo in mice. Remarkably, editing appeared to also occur efficiently in postmitotic cells, such as the adult liver in vivo, freshly isolated human liver sections, myocytes, and unstimulated primary CD34⁺ cells. The AAVHSC platform can be designed to mediate single-nucleotide changes in the genome, indicating its potential to correct point mutations. ⁶⁸

AAVHSC vectors mediated in vivo editing following intravenous delivery. We showed targeted insertion of a promoterless firefly luciferase open reading frame cassette into Intron 1 of the murine Rosa26 locus. ⁶⁸ Serial in vivo imaging of injected mice showed luciferase expression early after injection and was stable long term. Seamless editing was confirmed by sequencing and Southern analyses. As with other chromosomal loci, no ITR sequences were observed as being inserted into the targeted insertion site.

Overall, the AAVHSCs represent a single component editing platform that functions efficiently in vivo without the need for nuclease-mediated double-stranded DNA breaks and is being developed for therapeutic in vivo editing for inherited diseases, such as hemophilia and phenylketonuria (PKU). ⁷²

Clinical gene therapy of PKU with AAVHSC15

In 2019, Homology Medicines initiated a Phase1/2 open-label, randomized, concurrently controlled, dose escalation clinical gene therapy trial for PKU using an AAVHSC15 vector (pheNIX study, Clinicaltrials.gov Identifier NCT03952156). This trial, based upon Ahmed et al., ⁷³ is designed to test the safety and efficacy of a single intravenous injection of an AAVHSC15 vector encoding a functional copy of the phenylalanine hydroxylase (PAH) gene in adults with PKU. PKU is an autosomal recessive inborn error of metabolism due to an absence of functional PAH, a hepatic enzyme that converts phenylalanine to tyrosine. PAH deficiency leads to an inability to metabolize phenylalanine, resulting in developmental neurologic and metabolic problems. There is currently no adequate treatment for PKU that addresses the underlying genetic mutation. Management of PKU requires a highly restrictive diet, which leads to compliance failures.

Early results of the pheNIX gene therapy showed that intravenous injection of AAVHSC15 was well tolerated at both doses tested, which ranged from low to mid 1e13 vector genomes/kg (Homology Medicines, Inc. Webcast, December 17, 2019). No treatment-related adverse effects were observed at either dose, indicating safety of the in vivo use of AAVHSC15. One patient treated with the higher of the two doses showed a significant decline, by 48%, in serum phenylalanine levels (Fig. 3A) and a concurrent rise in serum tyrosine levels, by 85%, within 4 weeks after treatment initiation (Fig. 3B). The phenylalanine: tyrosine ratio declined by 72% by 4 weeks after injection (Fig. 3C). Thus, while the trial is still ongoing and data collection continues, early results support evidence of safety and efficacy of AAVHSC-PAH gene therapy for the treatment of PKU (Homology Medicines, Inc. Webcast, December 17, 2019. Data disclosed through December 2, 2019).

OPEN IN VIEWER

Figure 3.

Evaluation of AAVHSC15-PAH gene therapy in adults with PKU. Early results of pheNIX study, Clinicaltrials.gov Identifier NCT03952156. Data shown are from initiation of the trial until December 2, 2019. (A) Serum phenylalanine levels are shown. Each symbol represents a unique subject. The two patients shown in blue were treated at the low does. Open blue triangle, low dose, Patient 1. Open blue circle, low dose, Patient 2. Closed orange triangle, mid-dose, Patient 1. (B) Serum tyrosine levels. (C) Ratio of serum levels of phenylalanine to tyrosine. Low dose and mid dose represent low to mid 1e13 vector genomes/kg. Baseline and weekly measurements are shown. PAH, phenylalanine hydroxylase; PKU, phenylketonuria.

Preclinical Studies of AAVHSCs