Adenovirus vectors in hematopoietic stem cell genome editing

Abstract

Genome editing of hematopoietic stem cells (HSCs) represents a therapeutic option for a number of hematological genetic diseases, as HSCs have the potential for self-renewal and differentiation into all blood cell lineages. This review presents advances of genome editing in HSCs utilizing adenovirus vectors as delivery vehicles. We focus on capsid-modified, helper-dependent adenovirus vectors that are devoid of all viral genes and therefore exhibit an improved safety profile. We discuss HSC genome engineering for several inherited disorders and infectious diseases including hemoglobinopathies, Fanconi anemia, hemophilia, and HIV-1 infection by ex vivo and in vivo editing in transgenic mice, nonhuman primates, as well as in human CD34⁺ cells. Mechanisms of therapeutic gene transfer including episomal expression of designer nucleases and base editors, transposase-mediated random integration, and targeted homology-directed repair triggered integration into selected genomic safe harbor loci are also reviewed.

A major focus of gene therapy research has been genome editing of hematopoietic stem cells (HSCs), owing to that HSCs are the only cell type within the blood system possessing the ability of both self-renewal and multilineage differentiation. For congenital hematological disorders, the correction of a defect gene in HSCs implies that all progeny cells would carry the engineered traits and that a one-time treatment could provide a lifelong therapeutic and even curative remedy. In addition, efforts are also being dedicated to use HSC gene therapy for cancer and HIV-1 infection, based on the modification of host factors required for virus infection, directly targeting essential viral genes, or modulation of immune responses toward tumor destruction.

Efficient delivery of therapeutic transgene is a prerequisite for successful gene therapy. When gene therapy was conceptualized in the early 1970s, mammalian viruses were proposed as an effective vehicle to deliver a gene ‘drug’ [1,2]. In the ensuing nearly four decades, viral vectors have been intensively investigated and broadly used in gene transfer applications. A majority of the ongoing clinical trials are based on gene delivery using viral vectors including γ-retrovirus, lentivirus, adeno-associated virus, and adenovirus-derived vectors.

γ-retroviral vectors (γ-RVs), derived from murine leukemia virus, are one of the earliest vectors developed [3]. Based on ex vivo γ-RV transduction of CD34⁺ cells, several clinical studies for severe combined immunodeficiency disease (SCID) due to deficiency of the enzyme adenosine deaminase (ADA) and X-linked SCID have observed substantial overall improvements in clinical and immunological features in most of the patients [4-6]. However, as a result of insertional oncogenesis by γ-RV integration, leukemia was observed in five out of 20 X-SCID patients, which lead to one patient’s death [7,8]. The nature of vector insertion near oncogenes largely tempered further clinical use of early generations of γ-RV.

Lentiviral vectors (LVs) and recombinant adeno-associated viral vectors (rAAVs) are vigorously pursued in current HSC gene therapy. HIV-1-based LVs were pioneered in the second half of 1990s [9]. Compared to γ-RV, LV exhibited more efficient transduction of nondividing primitive HSCs and a safer integration profile [10]. Over recent years, LV-based HSC gene therapy has delivered promising results in patients with β-globin disorders [11,12]. However, the preferential insertion into active genes and local hotspots, and the observation of clonal expansion due to transactivation of HMGA2 in a treated β-thalassemia patient warrant concern over genotoxicity of early generation of LVs [13,14]. Additionally, current vesicular stomatitis virus G protein pseudotyped LVs are not suitable for in vivo systemic delivery because of neutralization by the human serum complement system [15]. In contrast, the largely episomal and nonpathogenic features of rAAVs have attracted much attention in the past two decades. Exceptional achievements using rAAVs have been made in both preclinical studies and human trials outside the HSC gene therapy field; for example, by in vivo delivery of rAAVs expressing Factor IX, more than fifty hemophilia B patients have been treated and most of the subjects have demonstrated efficient and relatively durable FIX activity [16]. For HSC genome editing, a recent study reported up to 90% homology-directed repair (HDR) by combining electroporation with CRISPR/Cas9 ribonucleoproteins (RNP) and transduction with a rAAV6 vector containing a donor transgene cassette [17]. Nevertheless, the < 5 kb packaging capacity, broadly detected preexisting immunity, and high cost for large-scale vector manufacturing present limitations to rAAV application.

The application of newer generation of adenoviral vectors (AdVs) may address the above limitations in HSC genome engineering and expand our treatment armamentarium for human diseases. In this review, we introduce the properties of a new generation of AdVs, that is, helper-dependent adenovirus vectors (HDAd) with HSC tropism, discuss various strategies for transient or stable transgene expression by HDAd, and then review recent advances in HSC gene editing for several representative blood disorders. Development of in vivo transduction is highlighted owing to its efficacy, technical simplicity, and potential for clinical application.

General concept of adenovirus vectors

Adenovirus was first isolated in the early 1950s [18], long before the discovery of other commonly adapted viral gene transfer vectors. When in vivo gene transfer was first attempted in the late 1980s for α1 antitrypsin deficiency, extensive knowledge had already been accumulated about first-generation AdVs (Group C serotypes 5 and 2 with E1 and/or E3 deletion). Their highly efficient transduction of a wide variety of dividing and quiescent cells, relative ease of preparation and purification, and well-characterized viral genome sequence and biology made AdV the vehicle of choice, leading to the first report of efficient in vivo gene transfer [19]. In the following decade, numerous studies have demonstrated the strong transduction efficiency of AdVs in multiple disease models. However, it was soon recognized that the first generation of AdVs is strongly immunogenic [20]. Local administration of AdVs elicits severe inflammatory responses and adaptive immune reactions, resulting in T cell-mediated clearance of transduced cells. In addition, adaptive antiviral immunity hindered repeated administration. Systemic delivery caused liver toxicity and triggered acute innate immune responses, which can be fatal at high vector doses [21]. On the other hand, the strong immunogenicity of AdVs can be harnessed as a beneficial property for oncolytic therapy [22,23].

Receptors and transduction

Virus tropism is largely dictated by a high-affinity attachment receptor. For human adenovirus, there are over 57 serotypes (Ad1–Ad57) with seven Groups (A–G). The serotype is defined historically by their characteristic neutralizing ability, and currently by DNA homology. Most adenovirus serotypes, including Group C Ad5, use the coxsackie adenovirus receptor (CAR) as a primary attachment receptor to target cells [24]. CAR is expressed in a broad range of tissues, rendering Ad5-based vectors a broad targeting spectrum. However, human HSCs express only low levels of CAR, making them refractory to infection by most adenoviruses [25].

Species B adenovirus tropism to human CD34⁺ cells

Group B viruses do not utilize CAR as a primary attachment receptor, and this influences their tissue tropism and subsequent disease pathogenesis. We and others have reported that group B adenoviruses are capable of using the complement regulatory protein CD46 as a primary attachment receptor [26,27]. CD46, also named membrane cofactor protein (MCP), is a type I transmembrane glycoprotein expressed on all nucleated human cells. It is described as a pathogen magnet [28], because CD46 also serves as a receptor for several other human pathogens, including laboratory strains of measles virus [29], human herpes virus 6 [30], pathogenic Neisseria [31], and Streptococcus pyogenes [32,33]. In humans, there are four major isoforms of CD46 (BC1, BC2, C1, and C2), depending on the alternative splicing of a region encoding an extracellular domain and the choice between one or two cytoplasmic tails, Cyt1 and Cyt2 [34]. CD46 expression is greatly up-regulated in malignant tumor cells [35-37]. Importantly, human HSCs express high level of CD46, rendering them highly susceptible to transduction by group B adenoviruses [25,38].

Three generations of adenovirus vectors

Adenoviral vectors are generated by replacing viral genes with foreign transgenes to be delivered to a wide variety of both dividing and nondividing cells, as well as mediating high levels of transgene expression. First-generation AdVs were constructed by deleting the early genes E1 and E3. However, at high multiplicities of infection, the E1 region becomes dispensable for replication. This may be related to the expression of specific host gene products with E1A-like activity, or to the proliferative state of cells. Leaky viral DNA replication is followed by low-level viral late protein production [39]. Accumulation of adenoviral cytotoxic late gene products following gene transfer causes direct cytopathic effects on the transduced cells and triggers host cellular immune response. These adverse effects can be aggravated by emergence of E1-containing (E1+), replication-competent adenovirus during the production of first-generation viruses in 293 cells.

These problems are diminished in second-generation adenovirus vectors, that is, vectors in which additional essential genes, E2 and E4, are deleted. However, it proved difficult to maintain cell lines that would efficiently transcomplement the deleted viral proteins.

Third generation of AdVs is devoid of all coding viral genes. Only inverted terminal repeats (ITRs) and the packaging signal are retained. The production of these vectors requires a helper virus to provide all structural proteins in trans; therefore, the vectors are also called HDAds or ‘gutless’ vectors. Cloning techniques such as Gibson assembly and recombineering have greatly simplified the construction of corresponding HDAd plasmids [40]. Furthermore, the properties of CRISPR/Cpf1 are favorable for engineering large DNA constructs including pHDAd genomes [41].

Helper-dependent adenovirus vectors

Helper-dependent adenovirus vector possess a number of properties that make them suitable tools for HSC genome editing.

Large packaging capacity

In theory, HDAds can accommodate transgene cassettes of up to 36 kb. For this reason, they are also called high-capacity AdV (HCAd) [42]. To facilitate vector purification and separation from the helper virus by cesium chloride ultracentrifugation, the insert size is usually kept within 28–32 kb. This payload is remarkably large, in sharp contrast to the 7 and 5 kb maximum carrying capacity of LVs and rAAVs, respectively [43]. The insert capacity of LVs and rAAVs limits their capability of incorporating large transgenes such as the most widely used CRISPR system (clustered regularly interspaced short palindromic repeats; ~ 5 kb with the premise of using minimally essential promoters and transcription elements), base editors (BEs; ~ 8 kb), and combinatorial systems (Fig. 1). With the utilization of HDAds, these cassettes as well as most of therapeutic genes can be readily accommodated. For example, we generated an HDAd vector that contained about 27kb of the human globin locus control region (LCR) [44]. The insert capacity is also relevant for using HDAd as donor vectors. It is thought that the efficiency of homologous recombination directly correlates with the length of the homology region. This has also been demonstrated in gene targeting studies with HDAd vectors [45-47].

Fig. 1.

Examples for genome engineering strategies using HDAd vectors. HDAd vectors do not contain any viral coding genes. Stuffer DNA sequences are usually included to make up the total genome size of 28–32 kb. The episomal nature of HDAds can be exploited for transient expression of nucleases (e.g., CRISPR/Cas9) and BEs (e.g., CBEs). Simultaneous editing of two target sites can be achieved by incorporating two guide RNAs (gRNA). An example of a SIN CRISPR vector is shown to demonstrate a strategy for controlling CRISPR duration. For integrating vectors using SB100× transposase, two vectors are co-delivered. One expresses the transgene (e.g., human γ-globin driven by β-globin LCR/promoter) flanked by two FRT-IRs. The other vector expresses SB100× transposase and flippase. For HDR-mediated targeted integration, one donor template is flanked by two homology regions neighboring targeted genomic sites. The second vector is used to express CRISPR for creating DSBs. Ad/AAV hybrid vectors mediate targeted integration into AAVS1 locus. Two vectors are used to express transgene flanked by AAV ITRs and AAV Rep78, respectively. Episomal expression can be combined with integrated expression. As an example, cloning of CRISPR outside of the two FRT-IRs leads to transient expression of the nuclease system while concomitantly allowing the γ-globin cassettes bracketed by two FRT-IRs for SB100×-mediated integration. 3′UTR, 3′ end UTR; APOBEC1, APOBEC1 cytidine deaminase; Cas9n, Cas9 nickase; EF1α, human elongation factor-1 α promoter; GOI, gene of interest; LCR, β-globin LCR; LHA, left HA; pA, polyA signal; RHA, right HA; U6, U6 promoter; UGI, uracil-DNA glycosylase inhibitor; ψ, packaging signal sequence.

Relatively low cost for vector manufacturing

A major obstacle for a widespread clinical application of gene therapy is the cost associated with clinical grade vector production. This is particularly the case for the production of rAAVs [48] and lentivirus vectors [49], which require, for each new vector preparation, large-scale plasmid transfection or more recently, transduction with baculovirus vectors that express corresponding viral proteins. In contrast, once a HDAd vector is generated and characterized, it can be used as a seed stock for further amplification.

The virus is produced at large scale in a 2-L spinner culture that routinely yields > 1 × 10¹³ viral particles per spinner (sufficient to treat a 10 kg patient) [50]. The relatively low costs for producing high yields of HDAd vectors are particularly relevant for in vivo gene therapy. Based on our experience, the costs for laboratory-grade HDAd vector are about $2000 per 1 × 10¹³ viral genomes. For intravenous injection, this would correspond to a cost of $10–20 per 25 g mouse. In comparison, for in vivo delivery of rAAV, 0.5–1 × 10¹³ viral genomes per adult mouse are routinely used, which would be a cost in the range of $500–$1000 per mouse (J. Chamberlain, personal communication). Ads including HDAds can be lyophilized and are stable at 4 °C [51].

Episomal expression

Adenoviral serotypes currently used as gene transfer vectors do not integrate into the host genome. Shortly after infection, the double-stranded viral DNA together with core proteins and the covalently linked terminal protein (TP) are translocated to the nucleus. TP binding confers stability of viral genomes, supporting long-term expression of the transgene, especially in nondividing cells. The episomal nature is a preferable feature for delivering designer nucleases such as CRISPR/Cas9 whose expression is only needed short-term to create the double-strand DNA breaks. We and others have produced HDAds expressing various nucleases [52-54]. In this context, it is noteworthy that the duration of Cas9 expression can be controlled by timed expression of anti-Cas9 inhibitors [55]. A recent report demonstrated a self-inactivating (SIN) CRISPR system by incorporating anti-Cas9 inhibitors in the same HDAd vector expressing CRISPR/Cas9 [56].

Integrated HDAd vector forms

Hematopoietic stem cell gene therapy requires that the gene modification is passed on to all progeny cells either through a permanent gene editing event or through gene addition after chromosomal integration. The latter can be achieved by various integration mechanisms, such as the Sleeping Beauty (SB) 100× transposase and HDR-mediated targeted integration. These approaches will be described in more detail in the following sections.

Absence of viral gene expression

Helper-dependent AdV do not contain any viral genes. Cytotoxicity associated with leaky viral gene expression seen with first-generation Ad vectors is therefore absent [57]. However, it should be mentioned that deletion of viral genes does not abrogate transcription-independent acute toxicity associated with virus uptake which will be discussed in the next section.

HDAd5/35++ with enhanced HSC tropism

Ad5 vectors use CAR for attachment to target cells. Group B adenoviruses utilize either desmoglein 2 (DSG2) [58] or CD46 [27] as primary receptors. While on HSCs CAR expression is only marginal, DSG2 and CD46 are uniformly expressed [59-62]. As consequence of this, HSCs can be readily transduced by members of the group B adenoviruses targeting either DSG2 [63] or CD46 [25], including Ad35, Ad11, Ad50, and Ad3. In order to be able to harness both, the stem cell transduction potential of these serotypes and the well-understood Ad5 vector system, fiber chimeric adenovirus vectors have been developed. These vectors are based on Ad5 and contain either just the fiber knob or the fiber knob and shaft of other serotypes. Examples include Ad5/3 [64], Ad5/35 [25], Ad5/11 [65], and Ad5/50 [66] fiber chimeric vectors. Studies done by us and others demonstrated that fiber chimeric Ad5/3 and Ad5/35 vectors can efficiently transduce HSCs of both humans and nonhuman primates (NHPs) [63,67]. Specifically, Ad5/35 vectors have been widely used for gene transfer into human HSCs [25,59,68-70] and human lymphocytes, the main target for HIV-1 [71-77]. Ad5/35 vectors also have been and are currently being employed in clinical trials to treat viral infections in immunocompromised individuals [76,78].

HDAd5/35++ denotes helper-dependent chimeric Ad5/35 vectors containing a mutant Ad35 fiber knob that was obtained by library screening. This mutant displayed over 25-fold higher affinity to CD46 and allowed for more efficient cell transduction at lower multiplicity of infection [79]. Of significance to innate acute toxicity is that upon systemic delivery, HDAd5/35++ showed much less liver sequestration and hepatocyte transduction than HDAd5 [38]. Although the definitive mechanism remains to be elucidated, this is probably due to the strong affinity of HDAd5/35++ to CD46 and a blockade of the interaction between Ad5 hexon and coagulation factor X (FX). The hexon–FX interaction is considered to mediate hepatocyte uptake of the virus [80-82].

Animal models for HDAd5/35++-mediated HSC gene therapy

Of all mammals, only NHPs express CD46 in a pattern similar to humans [83]. The only notable difference is that human erythrocytes lack CD46, while it is expressed on erythrocytes from NHPs. In previous studies, we showed that both baboons and macaques are adequate models for safety and efficacy studies with Ad35 fiber containing vectors or recombinant proteins [84-86].

In contrast to humans, where CD46 is expressed on all nucleated cells, CD46 is expressed only in the testis of mice. For further safety and efficacy studies, we therefore used human CD46 transgenic mice (strain MCP-8B) that express CD46 in a human-like pattern [87].

Another model is humanized mice. We transplanted human CD34 cells into sublethally irradiated, immunodeficient NOD.Cg-Prkdc^scid Il2rg^tm1Sug/JicTac (NOG) mice. Average engraftment rates at 6 weeks after transplantation were 25% based on huCD45 cells in the blood. Of note, all human cells express the HDAd5/35 receptor hCD46. Disadvantages of this model include an abnormal bone marrow structure and hematopoiesis [88].

Genome engineering strategies using HDAds

A variety of strategies have been adopted for genome engineering in HSCs. We will mainly discuss transient expression of nucleases and BEs, approaches to control transgene duration, as well as stimulation of transgene integration through transposases, homology-dependent repair (HDR), and hybrid vectors. A diagram is included to show these examples (Fig. 1).

Transient/episomal expression

Site-specific DNA breaks mediated by endonucleases

A major tool for genome engineering is engineered nucleases that generate double-stranded DNA breaks (DSBs) in preselected genomic sites (for a review, see [89]). The DSBs trigger endogenous cellular repair mechanism, that is, nonhomologous end joining (NHEJ), creating micro insertions or deletions (Indels) and, as a consequence, perturbing gene function. In the presence of a donor template, homology-dependent repair may occur, resulting in controlled modifications. There are now several different nuclease platforms available, with the most widely used ones being zinc finger nuclease (ZFN) [90], transcription activator-like effector nuclease (TALEN) [91], and CRISPR/Cas9 ([92,93]). By constructing a tandem array of various artificially engineered modules, ZFN and TALEN can target virtually any DNA sequence [94,95]. CRISPR/Cas9, derived from the CRISPR/Cas system of S. pyogenes, is composed of a human codon-optimized Cas9 protein and a single-guide RNA (sgRNA). It can be programmed to target any genomic locus followed by a 5′-protospacer adjacent motif sequence of NGG, with the specificity determined by the sgRNA containing a 20-nt guide sequence complementary to the genome locus of interest [92,93]. Nucleases are employed to knockout genes, correct frame shift mutations, delete or rearrange chromosomal regions, or knock-in wild-type or mutated cDNA into the endogenous or a heterologous site. Moreover, a variety of Cas9 derivatives and orthologs have been developed, for instance, CRISPRa [96], Cpf1 [97], saCas9 [98], and Cas13 [99] to name a few. Owning to the simplicity of manipulation and versatility, these robust tools allow widespread applications [100,101]. A CCR5-ZFN expressed by an Ad5/35 vector from Sangamo Therapeutics is currently tested clinically in HIV/AIDS patients [102]. In addition to our work with ZFNs, TALENs, and CRISPR/Cas9 expressed from HDAd5/35 vectors [52,53,55,103], many other research groups have demonstrated successful genomic editing using Ad vectors, including HDAd, expressing CRISPR/Cas9 [104-109]. Tsukamoto et al. reported the production of a panel of first generation of Ad vectors containing CRISPR/Cpf1 system [110,111].

High expression of endonucleases is considered to be harmful to the Ad producer cells and can lead to genomic rearrangements within the Ad genome or to low vector yields. A number of approaches have been developed to address this problem including the use of conditionally active promoters and downregulation of nuclease expression by shRNA [111]. We have exploited microRNA regulation to suppress ZFN and CRISPR/Cas9 expression in 293 and 116 cells [52]. In the context of HDAd production, we have also incorporated anti-CRISPR peptides into the helper vector [55,103]. Ng et al. expressed anti-CRISPR peptides in 116 cells [56].

Single nucleotide mutations by base editors

Base editors comprise a catalytically disabled nuclease, such as Cas9 nickase, fused to a nucleobase deaminase enzyme and, in some cases, a DNA glycosylase inhibitor. Currently, there are two categories, cytidine BEs (CBEs) and adenine BEs (ABEs) [112-114]. They are capable of installing precise nucleotide mutations at targeted genomic loci without creating DSBs. Nuclease-induced DSBs are found to be detrimental to host cells by causing unwanted large fragment deletion [115] and p53-dependent DNA damage responses [116,117]. This warrants concern particularly when editing fragile primitive stem cells. It is estimated that ~ 60% of pathogenic point mutations can be reversed by current BEs [118]. Current BE platforms appear to exert significant off-target activity at the RNA level [119,120]. With the rapid development of novel applications [121,122] and new BE variants [123,124], the amount of targetable mutations is projected to expand dramatically.

Adenoviral vectors are advantageous for delivery of BEs, with regard to the ~ 8 kb size for an all-in-one BE construct. We have successfully generated HDAd5/35++ vectors carrying CBE- or ABE-targeting critical motifs regulating γ-globin reactivation [125]. Naturally occurring loss-of-function mutations in the angiopoietin-like 3 (ANGPTL3) gene are associated with reduced blood triglycerides and low-density lipoprotein cholesterol. Chadwick et al. [126] obtained on average 35% on-target editing after intravenous injection of an Ad5 vector-bearing BE components targeting ANGPTL3 into mice. Another study performed by the same research group targeted the Pcsk9 gene using a similar in vivo method in mice [127]. Ad-BE-Pcsk9 installed 24% base conversion in the liver, leading to > 50% reduction of plasma PCSK9 protein level. By employing the latest versions of BEs with improved activity [128,129], rather than the suboptimal BE3 in their study, it is conceivable that higher editing efficacy is likely to be achieved. The efficacy of in vivo HSC base editing remains to be investigated.

Controlling transgene duration

Adenoviral vectors predominantly persist as episomal DNA [130]. Various mechanisms can be utilized to either shorten or extend the episomal transgene expression of Ad vectors [131-133]. More extensive E3 deletion in first generation of Ad vectors was associated with weaker immune responses and, consequently, contributed to extended transgene expression [134]. For pursuing prolonged episomal expression, HDAds are of significant advantage because of the absence of adaptive immunity against viral gene products [135]. It is known that the binding of TPs with ITRs is a major player in the stability of episomal DNA and its retention during mitosis. It is also found that the inclusion of retroviral LTRs in Ad vectors helped the virus persist [136]. Another interesting study presented by Gil et al. [137] showed that the introduction of self-replicating elements of Epstein-Barr virus into HDAd genome resulted in extended and nonintegrating expression.

With regard to delivery of nucleases and BEs, of more importance is to limit the duration of their expression in target cells. Extended presence of the editing machinery is considered to be detrimental to the cells because it can increase chances of off-target effects. This is particularly relevant for HSC gene editing. It is generally thought that once editing is completed, the editing machinery should be eliminated, the faster the better. This is called a ‘hit-and-run’ strategy [138,139]. As to controlling duration, electroporation of nuclease mRNA may be the best way. However, this approach is associated with significant cytotoxicity particularly for HSCs and, more importantly, not applicable for in vivo genome engineering. The naturally existing CRISPR inhibitors have shown potential to regulate the duration of CRISPR/Cas9 activity. CRISPR systems protect bacteria against invading bacteriophages. In response to this, phages have evolved counteracting peptides that bind to and inactivate Cas proteins as they search for foreign nucleic acid [140]. Among the reported peptides, AcrIIA2 and AcrIIA4 showed high anti-Cas9 potency [141]. We found that by timed delivery of a second HDAd5/35++ vector expressing AcrIIA2 and AcrIIA4, the Cas9 cleavage activity was not compromised in both primary cell lines and human CD34⁺ cells, demonstrating that a short-term expression is sufficient for CRISPR/Cas9 to perform editing. Importantly, the controlled Cas9 expression significantly preserved the engraftment capability of edited CD34+ cells in humanized NSG (NOD/Shi-scid/IL-2Rγ^nul1) mouse model [107]. Development of SIN and all-in-one CRISPR systems represents another promising strategy. This approach requires robust suppression of nuclease expression during vector production. By exploiting miRNA regulation to degrade Cas9 mRNA and anti-Cas9 peptides to inhibit Cas9 protein activity, Philip Ng’s group demonstrated the successful production of such HDAd vectors [56].

Integrated expression

Random integration mediated by transposase

In contrast to the intrinsic transient episomal expression, the incorporation of integrating mechanisms can lead to long-term transgene expression. For gene therapy of many diseases, the stable transgene expression is a prerequisite to confer sustainable treatment benefits. Transposon systems have emerged as promising tools for stable gene transfer. Ongoing clinical trials have employed them for therapeutic gene integration [142]. SB [143], piggyBac (PB) [144], and Tol2 [145] are among the most widely utilized transposon systems [146]. These systems usually require co-delivery of a donor transposon vector containing the gene of interests flanked by transposon DNA elements (i.e., inverted repeats/direct repeats, IR/DRs) and another vector expressing the transposase. Yant et al. [147] showed for the first time that systemic delivery of HDAd vectors carrying a SB system with a human factor IX transgene resulted in integration and persistent factor IX expression in mice. Since circular DNA is required for efficient transposition by the SB system, a Flp-FRT recombination system was also incorporated into the SB gene-containing vector [147]. Through molecular evolution, a new generation of hyperactive SB transposase (SB100×) exerted 100-fold enhancement in efficiency and strong activity in human CD34⁺ cells without affecting their multipotency [148]. Notably, 70% of the integrations occurred in intergenic regions. No preferential integration into genes was observed, in contrary to the integration pattern mediated by RVs and LVs. SB100× appeared to be superior to the PB system that exhibited a preference for integrations in regions surrounding transcriptional start sites and within long terminal repeat elements [149]. We have demonstrated the efficacy of SB100×-mediated integration into HSCs by HDAd5/35++ vectors in several disease models, including β-hemoglobinopathies, hemophilia A, and Fanconi anemia (FA) [150,151]. More studies have reported gene transfer by Ad vectors expressing other transposon systems, such as PB [152], T2TP [153], and bacteriophage-derived integrase PhiC31 [154]. Delivery of these transposons/integrases to HSCs by HDAd vectors remains to be fully explored.

Targeted integration into preselected sites

The semi-random integration pattern mediated by γ-RVs and LVs entails a risk of perturbing the expression of neighboring genes, including cancer-related genes. Theoretically, the random integration profile mediated by SB100× transposase and the lack of a preference for integration into active genes and promoters should be safer but concerns about genotoxicity remain. In addition, the integration into heterochromain associated sites may lead to gene silencing [44,155]. Therefore, a major effort in the field is aimed toward targeted transgene integration into preselected sites. A number of ‘safe harbors’ for targeted integration into the human genome have been suggested, for example, AAVS1 and CCR5 loci [156].

Targeted integration via homologous recombination without site-specific DNA breaks (DSBs) is inefficient [46]. It has been shown that DSBs stimulate HDR pathways and gene addition through homologous donor templates [157-161]. Hence, nucleases are commonly used to create DSBs, accompanied by the delivery of the donor cassette. However, the HDR pathway is still much less efficient that the NHEJ repair mechanism, particularly in quiescent stem cells [162,163]. Recently, with the remarkable advancement in genome editing techniques [17,164], the frequency of HDR has been improved dramatically. For ex vivo gene transfer into human CD34⁺ HSCs, zinc-finger nuclease mRNA and rAAV6-mediated donor template delivery resulted in ~ 15% targeted integration into the AAVS1 locus [165]. In another study that employed RNP electroporation to express CRISPR/Cas9 combined with rAAV6 transduction to deliver the donor template, the frequency of site-specific integration to correct the HBB sickle mutation was on average 25% [166]. These high-level HSC gene editing and HDR events are achieved through optimization of experimental parameters, such as modifications of sgRNA, time, and dose of AAV6 transduction. Of interest is to compare the performance of AdV with rAAV6 as the donor template vector. One study found that rAAV6 was more efficient than the first generation of Ad5/35 as a donor vehicle [167]. Recently, we demonstrated the utility of a HDAd5/35++ as a donor vector in mediating high frequency of HDR- and AAVS1-targeted integration [103]. In an in vivo HSC gene transfer study with AAVS1-transgenic mice, we have obtained efficient gene marking in peripheral blood cells with 80% of responsive rate. Of particular importance was that the gene expression level was significantly higher than that of SB100×-mediated random integration, which is most likely the result of integration into AAVS1, a locus with open chromatin [103]. Notably, HDAd5/35++ represents a potential good candidate for the delivery of large donor cassettes exceeding the capacity of rAAV6.

Other integrating hybrid vectors

In addition to the above-discussed Ad/transposase hybrid systems, other hybrid vectors have also been exploited for integrated expression of transgenes delivered by Ad vectors, such as AdV/γ-RV [168,169], AdV/foamy viral vectors [170,171], and Ad/AAV [172,173] (For a comprehensive review, see [174]). In essence, the integration mechanism relies on the incorporated viral elements from other viruses. As an example, we have previously reported on an HDAd5/35.AAV hybrid vector expressing a GFP reporter gene driven by the β-globin LCR. The LCR-GFP cassette with a length of 27kb was flanked by AAV ITRs. Integrated stable expression was observed in human CD34⁺ cells [44]. By co-infection with another HDAd5/35 vector expressing Rep78, the AAV gene responsible for AAV integration into the AAVS1 locus, targeted integration region was shown [175].

In vivo HSC transduction

Current HSC gene therapy involves harvesting hematopoietic stem/progenitor cells (HSPCs) from the patient, genetically engineering HSPCs ex vivo, and reinfusion back to the same patient following myeloablative conditioning. This procedure is not without risks. Commonly encountered challenges include insufficient collection of HSPCs by leukapheresis or bone marrow aspiration, genotoxicity and limitations of the viral vectors, loss of HSC multipotency during ex vivo manipulation, inadequate number of modified cells to confer successful engraftment, and morbidity and mortality associated with myeloablation. Moreover, the technical complexity and high cost of current ex vivo HSC gene therapy are barriers to a widespread application for common diseases. As an example, Strimvelis, an HSC gene therapy for ADA-SCID that has been approved for marketing in the European Union in 2016, comes at a price of close to $700 000 per treatment, and the regimen is only available at a single center in Europe, making access to it difficult even within the developed countries of Europe.

Our group recently developed an in vivo HSC transduction approach that offers the potential to address the aforementioned obstacles in the ex vivo procedure [38]. In our approach, the HSCs residing in the bone marrow niche were mobilized into the peripheral circulation through administration of granulocyte colonystimulating factor and the CXCR4 antagonist, Plerixafor/AMD3100. At the peak of mobilization (i.e., when HSPCs in the circulation reached the highest level), an integrating HDAd5/35++ vectors expressing a GFP reporter together with a SB100x/Flpe-expressing vector were injected intravenously. HSPCs transduced in peripheral blood homed back to bone marrow shortly after transduction. At week 12 post-transduction, 16.5% of total colony-forming cells stably expressed the reporter gene. This level of gene marking was maintained in HSPCs and lineage cells of secondary transplanted mice, confirming genetic modification of HSCs with multilineage, long-term repopulating capability [38]. However, it is noteworthy that the high level of gene marking in bone marrow HSCs was not reflected in peripheral blood mononuclear cells (PBMCs) of the primary mice. This was likely in part due to a lack of a proliferative stimulus for the modified HSCs. To overcome this drawback, we therefore combined the in vivo transduction with an in vivo selection mechanism by incorporating a mutant O⁶-methylguanine–DNA methyltransferase (mgmt^P140K) gene that confers resistance to O⁶-BG/BCNU (O⁶-Benzylguanine/Carmustine) [176-178]. After in vivo transduction and selection with three low doses of O⁶-BG/BCNU administered intraperitoneally with interval of 2 weeks, transgene marking in PBMCs was increased to > 50% [176]. Importantly, the approach has been tested by delivering several therapeutic genes, for instance, γ-globin, and with optimized procedures we routinely obtained more than 80% gene marking in HSCs and peripheral blood cells [53,103,179].

The in vivo HSC transduction approach has a number of preferable features. First, the ability of cost-efficient production of high yields of Ad vectors is desirable for in vivo transduction, which requires large amounts of vectors. Second, HDAd5/35++ vectors have a ~ 32 kb insert capacity, and importantly, high tropism to HSCs that uniformly express the HDAd5/35++ receptor, CD46, at high levels. Third, mobilized HSPCs are transduced with HDAd5/35++ vectors in the periphery, and home back to the bone marrow shortly after transduction. No ex vivo manipulation of HSCs and no myeloablation are needed. Fourth, the integrating HDAd5/35++ vectors are designed for persistent transgene expression. The SB100×-mediated integration shows a good safety profile without preferential insertion into exons. Notably, the selection mechanism can be combined with transient expression simply by inserting the genes outside of the two IR/DRs elements. Fifth, because of the fiber modification to HDAd5/35++ and dexamethasone administration prior to transduction, the procedure of systemic delivery is well tolerated by the mobilized animals reflected by low transduction of liver cells, no elevation of transaminases and pro-inflammatory cytokines. Ongoing studies in NHPs have confirmed the safety and efficacy of the in vivo transduction/selection approach.

A large fraction of humans has neutralizing serum antibodies directed against Ad5 capsid proteins, which will block in vivo transduction with Ad5/35 vectors, that is, vectors that contain Ad5 capsids proteins and chimeric Ad35 fibers. An alternative would be vectors derived from Ad35. (a) Ad35 is one of the rarest of the 57 known human serotypes, with a seroprevalence of < 7% [180-184]; (b) Ad35 is less immunogenic than Ad5 [185], which is, in part, due to attenuation of T-cell activation by the Ad35 fiber knob [186-188]; (c) after intravenous injection, there is only minimal transduction (only detectable by PCR) of tissues, including the liver, in human CD46 transgenic (hCD46tg) mice [189] and NHPs [190]. Work on HDAd35 vectors is ongoing in several laboratories including ours.

Figure 2 shows typical procedure for ex vivo and in vivo HSC gene transfer. Others have also reported in vivo HSC transduction approaches using nonviral [191] or viral vectors [192,193]. Readers are recommended to refer to our recent review for more examples [194].

Fig. 2.

Ex vivo VS in vivo HSC transduction. Ex vivo HSC gene transfer is started with harvesting white MNCs in blood by leukapheresis after mobilization. Alternatively, bone marrow cells can be collected by aspiration. HSPCs are usually enriched from total MNCs by selection for CD34⁺ cells. The HSPCs are then be transduced using viral vectors (usually SIN-LVs) carrying transgenes. Modified cells are cryopreserved for shipment or if multiple harvesting/transductions are needed for accumulating more cells. Subsequently, the modified HSPCs are thawed, recovered, and infused back to the same patient following myeloablative conditioning. For in vivo HSC transduction, viral vectors (e.g., HDAd5/35++ vectors) expressing transgenes are administered intravenously following mobilization of HSCs. HSCs egressed from bone marrow are transduced in blood. Shortly following transduction, the modified cells home back to bone marrow and persist long-term.

Progress in gene editing for hematological disorders

β-hemoglobinopathies

β-hemoglobinopathies are a group of genetic disorders with abnormal production of β-globin chain of hemoglobin—mainly including β-thalassemia and sickle cell disease (SCD). β-thalassemia is caused by diverse mutations resulting in various severity including thalassemia intermedia (β⁺) and major (β⁰). SCD results from a single nucleotide transversion (A–T) in the sixth codon of the β-globin gene HBB leading to Glu6-Val mutation. SCD is named based on the characteristic phenotype of red blood cells: a crescent or sickle shape caused by polymerization of the pathogenic β-globin (β^S) upon deoxygenation. Currently, treatments, including blood transfusion and hydroxyurea, are largely supportive. The only curative approach relies on allogenic HSC transplantation, which is not available to most patients due to lack of matching donors.

Gene therapy using autologous HSC transplantation has achieved encouraging progress. Current advancements in clinical trials mainly come from gene addition strategies. They are based on LV-mediated gene transfer of the wild-type human β-globin gene, the fetal γ-globin gene, or β-globin mutants with antisickling mutations. Specific vector examples that are under clinical trials include BB305 (β^A-T87Q, NCT02140554) [195], sG^bG (γ-globin, NCT02186418) [196,197], βAS3-FB (β^A-AS3, NCT02247843), GLOBE (β^A, NCT02453477) [198], and TNS9.3.55.A1 (β^A, NCT01639690) [199]. Among them, BB305 (trade name: Zynteglo) has received conditional approval from European Medicines Agency (EMA) [200]. Overall, therapeutic benefits have been observed in most of the reported patients after treatment, with non-β⁰/β⁰ patients showing more significant improvement [11-13,195,201,202]. Nevertheless, there are challenging aspects of LV design because of the requirement to achieve erythroid-specific and high-level globin expression while avoiding genotoxicity and not compromising viral titer. Although all these above vectors are delivered by SIN-LVs that should have an improved safety profile [203], the risk of insertional mutagenesis cannot be ruled out. A clonal dominance was observed in the first β-thalassemia patient who became transfusion independent after treatment with HPV569-transduced HSPCs (HPV569 is an early generation of BB305) [13]. This command caution and calls for longer follow-up to better evaluate the vector safety profile.

AdVs have become an efficient vehicle for HSC gene transfer since the development of vectors derived from group B adenoviruses [25]. We recently generated an integrating HDAd5/35++ vector system for γ-globin gene addition [204]. The γ-globin was driven by a 5 kb β-globin LCR/promoter. An mgmt^P140K gene under an elongation factor 1 α promoter (EF1α) was included for selection. The 12 kb transgene with regulatory elements was flanked by two FRT-IR/DRs sequences at each side. Upon co-transduction of lineage-negative cells from CD46 transgenic mice with another vector expressing a SB transposase and Flpe followed by transplantation into irradiated C57BL/6 mice (ex vivo transduction setting), γ-globin expression was detected in 80% of peripheral erythrocytes without selection. The percentage of γ-globin-positive erythrocytes increased to nearly 100% after selection. The γ-globin level was 10–20% of adult mouse globin. This ex vivo transduction efficacy was recapitulated in human CD34⁺ cells transplanted into NSG mice. Importantly, this vector system also worked efficiently in in vivo transduction settings. In a thalassemia CD46^+/+/Hbb^th-3 mouse model, we detected similar levels of γ-globin and a near complete phenotypic correction. Genome-wide integration analyses revealed a random pattern without a preference for genes [179], in agreement with previous reports [148,205]. To achieve higher levels of γ-globin expression, we established a targeted integration system with two vectors [103]. One contained a CRISPR/Cas9 specific for human AAVS1 locus. The other provided a donor template that carried the same 12 kb γ-globin transgene cassette flanked by two homology arms (HAs). The two HAs, amplified from human AAVS1 locus surrounding the CRISPR cutting sites, were further flanked by two CRISPR cutting sites to facilitate the release of donor template. In a human AAVS1 transgenic mice model, 80% responsive rate was obtained after transduction with the two vectors and in vivo selection. Of particular importance is that the γ-globin level reached on average 24% of adult mouse globin, significantly higher than levels achieved with SB100×-mediated random integration [103]. Integration site analyses revealed a predominant targeted integration into the intended AAVS1 locus. Another finding is that longer HAs appeared to increase the efficacy of targeted integration. It remains to be investigated whether further extension of the homology regions is beneficial. As an alternative target locus, one of the α-globin loci can be considered because an excess amount of α-globin production in β-thalassemia and SCD patients causes ineffective erythropoiesis and hemolysis [206].

Gene editing has also generated enormous enthusiasm for the treatment of hemoglobinopathies. The main objective is to reactivate the endogenous γ-globin that is silenced shortly after birth and capable of functioning as an alternative to β-globin, or to directly reverse the sickle mutation. Through genome-wide association study, BCL11A was found to be a master suppressor responsible for γ-globin dormancy [207,208]. Knockout of the BCL11A gene is not a viable option due to its indispensable roles in development. However, reports demonstrated that partially knockdown of BCL11A by targeting some critical motifs in its erythroid enhancer led to efficient de-repression of γ-globin while not affecting animal viability [209,210]. Strategies based on shRNA- and CRISPR-mediated down-modulation of BCL11A have entered clinical stage (NCT03282656 and NCT03655678). Using an HDAd5/35++ vector, significant γ-globin reactivation was observed by CRISPR/Cas9 targeting the critical GATAA motif in the BCL11A enhancer [55,211]. Alternative strategies include the disruption of binding sites of repressors within the HBG promoter region [212-214], the repair of the sickle mutation [215], and the forced modification of chromatin structure [216]. These strategies would be benefited considerably from recent advancement of gene editing efficiency. Latest reports have shown more than 90% disruption of the BCL11A GATAA motif in HSCs by electroporation of Cas9 RNPs [164]. When combining with rAAV6 transduction to deliver a donor template, over 20% HDR-mediated repair of the sickle mutation has been demonstrated [166]. As an alternative, Romero et al. [167] reported that RNP electroporation can be combined with Ad5/35 transduction for ex vivo gene editing of human HSCs to correct the sickle HBB mutation. Li et al. [217] demonstrated efficient repair of the sickle mutation in induced pluripotent cells using an HDAd5/35 vector. We constructed an HDAd5/35++ vector-expressing CRISPR/Cas9 specific for the BCL11A binding site in HBG promoter. Transduction of human CD34⁺ cells with this vector resulted in markedly increased production of γ-globin. In a β-YAC/CD46 transgenic mice model, in vivo transduction resulted in ~ 33% target site cleavage and ~ 15% γ-globin production over mouse β-major globin [53]. The degree of phenotypic correction of β-thalassemia or SCD is closely correlated with the level of γ-globin expression. We therefore combined the two mechanisms, that is, CRISPR/Cas9-triggered reactivation of endogenous γ-globin and SB100× transposase-mediated γ-globin gene addition, into one vector and obtained additive levels of γ-globin (~ 40% over mouse β-major globin) [218]. One additional advantage of this strategy is the shortened duration of CRISPR/Cas9 expression as a result of disruption of viral genome by transposition (Fig. 1). In addition to the CRISPR system, emerging base editing technologies are being pursued to precisely install single nucleotide mutations controlling fetal globin de-repression [125].

Fanconi anemia

Fanconi anemia is caused by mutations in any of the 22 identified FANC genes: FANCA–FANCJ, and FANCL–FANCW. FA pathway genes play critical roles in repair of damaged DNA—specifically, DNA interstrand cross-links. Thus, pathogenic mutations in FA genes result in chromosomal instability. FA patients are clinically manifested by congenital impairment, bone marrow failure, and cancer occurrence. Among these genes, FANCA mutations are responsible for over 60% of all cases of FA [219-221]. The same as β-thalassemia and SCD, allogenic transplantation with healthy donor HSCs is the only curative therapy.

Gene therapy for FA has been investigated for more than two decades. Early studies mainly used γ-RVs and led to no persistent therapeutic benefits [222,223], partly due to the inefficiency of γ-RVs to transduce primitive HSCs. With the advancement of LV vectorology, some important progress has been made. Juan Bueren’s research team from Spain recently published data from their clinical trial (NCT03157804) [224]. Four patients with FANCA mutations received transplantation with autologous CD34⁺ cells that were transduced ex vivo with a LV expressing a functional FANCA under a phosphoglycerate kinase (PGK) promoter. No conditioning was conducted prior to engraftment. Progressive repopulation of the bone marrow cells from gene-corrected HSCs was observed over a period of up to 30 months following treatment. Although the results are encouraging, continued surveillance will corroborate the integration safety profile and successful transduction of multipotent HSCs.

Besides LV-mediated gene transfer, other vectors are being investigated, such as foamy viral vectors [225], and measles virus envelope pseudotyped LVs [226]. HDAd5/35++ vectors also represent a promising strategy—particularly of relevance for in vivo HSC transduction. In addition to the above-described advantages of the HDAd5/35++ vector platform, several lines of evidence related to the nature and pathophysiology of FA further highlight the therapeutic potential of in vivo HSC gene transfer: (a) HSCs of FA patients are very fragile—making ex vivo manipulation more challenging; (b) generally, it is difficult to harvest large amounts of CD34⁺ cells from FA patients, and transplantation with insufficient number of gene-modified cells tends to result in reduced or even failure of engraftment; (c) due to proliferative advantage of gene-corrected HSCs over mutant HSCs, a few in vivo corrected primitive cells are likely to progressively reconstitute the blood system by overgrowing defect cells (supported by the evidence from natural gene therapy in monozygotic twins [227]). Our previous study demonstrated that in vivo transduction led to ~ 20% of gene-marked colony-forming unit (CFU)-forming cells in mice bone marrow [38]; therefore, selection may not be necessarily needed. Our HDAd5/35++ FancA gene therapy vector contained the human FancA gene under control of the PGK promoter as well as an EF1α-promoter-GFP gene expression cassette. Integration was mediated by SB100x. To generate a FancA model for in vivo HSC transduction with HDAd5/35++ vectors, we crossed CD46 transgenic mice with FancA^−/− mice. We found that GSCF/AMD3100 treatment of hCD46^+/−/FancA^−/− mice resulted in efficient HSC mobilization, that is, a > 10-fold increase in peripheral blood LSK cells 1 h after AMD3100, at which time animals were injected with the HDAd-FancA/GFP vector. Six weeks after HDAd injection, 120 mg/kg of the DNA damaging drug cyclophosphamide (CY) was given to determine the presence of any gene-modified FA cells, which should be more resistant to CY exposure than uncorrected FancA cells. Before CY (week 5) FancA mRNA was not detectable by qRT-PCR in PBMCs. One week after CY (week 7), the level of FancA mRNA was ~ 0.4% of the house-keeping gene mRPL10. The percentage of GFP-positive PBMCs before CY (week 5) was, on average, 0.25%. One week after CY, it increased to 0.8% and was then stable at a marking rate of ~ 15%. Bone marrow mononuclear cells (MNCs) harvested from mice 1 week after CY treatment, displayed a significantly greater resistance to the DNA cross-linker mitomycin C in progenitor CFU assays (> 3-fold more colonies than in control animals) indicating a partial phenotypic correction of FA. Two weeks after CY, hematological parameters were fully recovered and a second injection of CY (60 mg·kg⁻¹) was given. This preliminary study indicates that in vivo HSC gene therapy of FancA is possible without in vivo selection with O⁶BG/BCNU.

Hemophilia

Hemophilia is an inherited bleeding disorder resulting from a deficiency of blood coagulation factors. Genomic mutations in the X-linked factor VIII and IX coding genes (F8 and F9) cause hemophilia A and B, respectively. Severe forms of hemophilia are characterized by spontaneous bleeding episodes. Current standard treatment requires repeated infusions of recombinant FVIII and FIX factors that are very costly (~ $400 000 per year) and accompanied by complications.

Gene therapy for hemophilia—by providing a functional copy or by correcting mutations of the deficient F8 or F9 genes—may lead to sustained benefits. A number of features collectively make hemophilia an ideal target for gene therapy: the monogenic etiology, maintained clotting activity when synthesized by various cell types, low factor activity needed for clinical benefits (5% of normal activity), and simple approaches for activity measurement, etc. These features also have contributed to human trial achievements. After treatment, a majority of patients exhibited improvement in ongoing clinical trials, which are all based on liver-targeted rAAV vectors. Three of these trials sponsored by different companies have reached Phase III [two for hemophilia B (NCT03861273 and NCT03569891); one for hemophilia A (NCT03370913)] [16]. The results are undoubtedly encouraging. However, the widespread application of existing rAAV-based approach could face several obstacles: (a) mostly episomal nature of rAAV genomes affecting durability of efficacy, specifically in children with actively dividing hepatocytes (patients under 18 years old are excluded by most of current gene therapy trials); (b) the high cost of rAAV vector production, estimated to be > $1 M per patient; (c) the limited packaging capacity of rAAV which cannot accommodate large transcriptionally regulatory elements (viral titer drops precipitously with transgenes > 5 kb); and (d) the pre-existing immunity to AAV capsid leads to loss of efficacy and circumvents repeated administration (subjects with pre-existing antibodies to the used rAAV serotype are not eligible for current clinical trials) [228]. Hence, the search for alternative approaches could address these limitations and expand our treatment options.

Helper-dependent AdVs represent an effective delivery vehicle. For Ad5-based HDAd vectors, they retain extremely high tropism to hepatocytes while exhibiting much reduced immunogenicity. A series of studies have demonstrated that HDAd vectors mediated longterm therapeutic FVIII or FIX expression in transgenic mice, hemophilia dogs, and NHPs [229-234]. In addition, optimization of HDAd vector design, such as development of chimeric vectors with eliminated uptake by Kupffer cells [235], could further improve the transduction efficacy and safety profile. For integrated expression, SB transposase was observed to result in persistent hemostatic correction of canine hemophilia B with negligible toxicity [236]. Targeted integration of a corrective F9 gene into the ROSA26 safe harbor locus by AdV-delivered CRISPR/Cas9, and a donor template has led to phenotypic correction in a murine model with juvenile hemophilia B [106]. Efficient targeted integration of a F9 gene was also demonstrated by using a single HDAd vector simultaneously carrying CRISPR/Cas9 and a donor cassette [105]. Instead of liver-directed gene transfer, HSC-directed F8 gene transfer has been investigated and showed advantages in tackling pre-existing immunity [237]. We recently reported on an approach for HSC-derived, erythroid-specific F8 production based on HDAd5/35++ vectors [151]. It has been shown that the abundance and systemic distribution of erythroid cells can be harnessed for high-level production of therapeutic proteins with induced immune tolerance [238]. Using a vector carrying a GFP reporter under β-globin LCR/promoter together with the SB100× vector for targeted integration, we first showed that in vivo transduction led to sustained erythroid cell-restricted GFP expression, confirming the lineage specificity of LCR promoter. We then constructed a therapy vector by replacing GFP with ET3, a bioengineered human factor VIII [239]. After in vivo transduction, we detected stable supraphysiological levels (~ 300%) of ET3 in plasma. A phenotypic correction of bleeding was observed after in vivo HSC transduction of hCD46^+/+/F8^−/− hemophilia A mice in spite of high plasma anti-ET3 antibody titers. This suggests that ET3 levels were high enough to provide sufficient noninhibited ET3 systemically and/or locally (in blood clots) to control bleeding. The high-level ET3 production in erythroid cells did not affect erythropoiesis. Importantly, a phenotypic correction of bleeding was observed after tested in a hemophilia A mouse model. In addition to its relevance for hemophilia A gene therapy, our approach has implications for the therapy of other inherited or acquired diseases that require high levels of therapeutic proteins in the blood circulation.

HIV-1/AIDS

HIV infection is the etiological cause of AIDS. Since the first reported case of AIDS [240], more than 38 years have elapsed. It is estimated that there were 38 million people worldwide living with HIV-1/AIDS by the end of 2018 [241]. Although the introduction of effective antiretroviral therapy (ART) has resulted in massive reductions in HIV-1 incidence and substantially improved the life quality of HIV-1 patients, HIV-1 infection remains incurable to the general population, and no vaccine is currently available.

Gene therapy has been explored as a potential treatment option for HIV-1/AIDS. One of the strategies is to target the predominant coreceptor CCR5, which is required for HIV-1 entry into cells. It is inspired by the observation that some Caucasian individuals carrying an inactive mutant of CCR5 (CCR5Δ32) are resistant to HIV-1 infection [242], and ostensibly, this mutation does not affect their health status. Various approaches have been investigated, such as shRNA and nucleases. Perez et al. have demonstrated that an Ad5/35 vector-encoding ZFN.CCR5 efficiently knocked out CCR5 expression in up to 50% of human primary CD4⁺ T cells, the most important HIV-1 target in vivo. As expected, genetic disruption of CCR5 provided robust and heritable protection against HIV-1 infection [76]. In a clinical trial based on this approach, a survival advantage for CCR5-modified cells over unmodified cells and reduced viral DNA were found [243]. Using a similar vector, we delivered CRISPR/Cas9 to target CCR5 and achieved near complete inhibition of HIV-1 infection in edited primary CD4⁺ T cells [244]. Robert Doms’s group adopted two Ad5/35 vectors expressing ZFNs to simultaneously target CCR5 and CXCR4, the other HIV-1 coreceptor. This strategy conferred the cells with resistance to infection by both R5- and X4-tropic viruses, and ablation of the two receptors did not affect cell proliferation and engraftment of the modified CD4⁺ T cells [245]. All these studies were based on first-generation AdVs. It would be interesting to test the efficiency of HDAd5/35++ vectors.

In addition to differentiated T cells, HSC-directed CCR5 gene editing has also been investigated. It is largely inspired by the case of the ‘Berlin Patient’ [246], who has become free of viremia for more than 12 years since he received two rounds of HSC transplantations with donor cells homozygous for CCR5^Δ32. He is considered the first HIV-1 patient being functionally cured. Early this year, Gupta el al. reported a second case, the ‘London Patient’, who received a similar transplantation with CCR5^Δ32 HSPCs but using a milder conditioning regime [247]. This subject has been negative for HIV-1 traces for over 2 years. A research team from China has attempted to reproduce these two cases by creating CCR5-null CD34⁺ cells using CRISPR/Cas9 [248]. The edited cells were later transplanted into a patient with both HIV-1 infection and acute lymphoblastic leukemia. While the level of CCR5-negative T cells derived from edited CD34⁺ cells was modest (around 5%), the patient achieved remission for both HIV-1 infection and leukemia for over 19 months after the procedure, and the investigators did not observe gene editing-related side effects. It is likely that the remission was achieved by several effects in combination, such as myeloablation, donor cell-mediated immune clearance, and CCR5 gene editing. Despite of this progress, most of other allogenic transplantations attempting to re-create the case of Berlin patient did not succeed, partly due to graft versus host diseases (GvHD) and viral tropism shift as a result of CCR5 mutation [249]. This calls for further optimization of the approach with less invasive interventions.

Another strategy is to target the proviral DNA integrated into the host genome. Although ART could suppress viremia to a level below the detection limit by current approaches, viral rebound from latent viral reservoir eventually appeared in most patients after discontinuation of treatment. A general ideal is to make two DSBs by nucleases in proviral DNA (e.g., in the LTRs), leading to the removal of the viral genome [250]. Alternatively, targeting conserved and critical viral genes has been found to disable the provirus [251]. Nevertheless, this strategy faces profound challenges, including the diversity of viral quasi-species, the fast mutating nature of HIV-1, the transcriptionally inactive latency, the establishment of latent infection in long-lived cell subsets (such as resident macrophages in various tissues), and the lack of ideal animal models.

With these challenges, multipronged combinatorial strategies likely hold more promise. This, in some way, resembles the principle of ART that uses a cocktail of medications blocking different steps of HIV-1 replication. A number of combinations have been investigated, such as targeted integration of an antiviral inhibitor-encoding gene into CCR5 locus [252], simultaneous targeting the entry receptor and viral essential genes [253], and combining CRISPR/Cas9-mediated excision of provirus with ART [254]. The capacity of HDAd5/35++ vectors is advantageous with regard to accommodating multiple antiviral mechanisms, and at the same time incorporating a selection mechanism to transduced cells. Notably, in vivo HSC engineering is likely of significance regarding its capability to remodel the immune system with HIV-resistant cells while avoiding GvHD-related complications.

Conclusions and perspectives

Adenoviral vectors are the most widely used vector in clinical trials. By the end of 2018, there have been 569 out of 2930 gene therapy clinical trials that used AdVs, mainly in anticancer studies (Gene Therapy Clinical Trials Worldwide by The Journal of Gene Medicine). The development of HDAds substantially reduced the immunogenicity and improved the safety profile of AdVs as a gene delivery vehicle. Chimeric HDAd5/35++ vectors possess ~ 32 kb payload capacity that can accommodate most therapeutic genes and gene editing tools. Importantly, they can efficiently mediate gene transfer into HSCs with an episomal nature while formulable for integrated expression. The addition of AdVs into our toolkit has provided more gene transfer options that may be beyond the capability of other vectors.

Hematopoietic stem cell gene therapy could provide a curative treatment for a number of blood diseases. The conventional approach is based on ex vivo HSC gene transfer and has achieved encouraging results. Two of the six FDA- or EMA-approved gene therapy products are carried out by ex vivo HSC transduction—Strimvelis for ADA-SCID; and Zynteglo conditionally approved by EMA for β-thalassemia. However, the high cost and side effects limit the patient accessibility of ex vivo HSC gene therapy, particularly in resource limited regions where most hematological diseases prevail. In contrast, in vivo HSC transduction is highlighted by its relatively low cost and technical simplicity. It could be provided as an outpatient treatment. Its efficacy has been demonstrated in several murine disease models, including β-thalassemia and hemophilia A [151,179], and its safety is currently tested in NHPs. With more gene therapy products on the horizon (such as hemophilia), the application of in vivo HSC transduction could extrapolate genetic treatments to a larger patient population. Further improvements of in vivo HSC gene therapy with HDAd vectors on the road to clinical application include more effective mobilization protocols (particularly for some patients, such as SCD patients), complete elimination of innate responses upon intravenous injection, more advanced capsid modifications that circumvent pre-existing anti-Ad immunity, improved selection regimens [240,255], as well as new methods for purification of clinical grade HDAd vectors.

Abbreviations

ABEs

adenine BEs

Ad

adenovirus

ADA

adenosine deaminase

AdV

adenoviral vector

ANGPTL3

angiopoietin-like 3

ART

antiretroviral therapy

BEs

base editors

CAR

coxsackie adenovirus receptor

CBEs

cytidine BEs

CFU

colony-forming unit

CRISPR

clustered regularly interspaced short palindromic repeats

CY

cyclophosphamide

DRs

direct repeats

DSBs

double-stranded DNA breaks

DSG2

desmoglein 2

EF1α

elongation factor 1 α promoter

EMA

European Medicines Agency

FA

fanconi anemia

FIX

Factor IX

FVIII

Factor VIII

FX

factor X

GvHD

graft versus host diseases

HAs

homology arms

HDAd

helper-dependent adenovirus vectors

HDAd5/35++

helper-dependent chimeric Ad5/35 vectors containing a mutant Ad35 fiber knob

HDR

homology-directed repair

HSC

hematopoietic stem cell

HSPCs

hematopoietic stem/progenitor cells

IRs

inverted repeats

ITRs

inverted terminal repeats

LCR

locus control region

LV

lentiviral vector

MCP

membrane cofactor protein

mgmt

O⁶-methylguanine–DNA methyltransferase

NHEJ

nonhomologous end joining

NHPs

nonhuman primates

O⁶-BG

O⁶-Benzylguanine

PB

piggyBac

PBMCs

peripheral blood mononuclear cells

PGK

phosphoglycerate kinase

rAAVs

recombinant adeno-associated viral vectors

RNP

ribonucleoproteins

SB

sleeping beauty

SB100x

hyperactive SB transposase with100-fold enhanced activity

SCD

sickle cell disease

SCID

severe combined immunodeficiency disease

sgRNA

single-guide RNA

SIN

self-inactivating

TALEN

transcription activator-like effector nuclease

TP

terminal protein

ZFN

zinc finger nuclease

γ-RVs

γ-retroviral vectors