Genome editing and the next generation of antiviral therapy

Abstract

Engineered endonucleases such as homing endonucleases (HEs), zinc finger nucleases (ZFNs), Tal-effector nucleases (TALENS) and the RNA-guided engineered nucleases (RGENs or CRISPR/Cas9) can target specific DNA sequences for cleavage, and are proving to be valuable tools for gene editing. Recently engineered endonucleases have shown great promise as therapeutics for the treatment of genetic disease and infectious pathogens. In this review, we discuss recent efforts to use the HE, ZFN, TALEN and CRISPR/Cas9 gene-editing platforms as antiviral therapeutics. We also discuss the obstacles facing gene-editing antiviral therapeutics as they are tested in animal models of disease and transition towards human application.

Introduction

Over the last decade, the tools available for DNA manipulation have advanced significantly. A wide array of powerful gene-editing systems is now available, and researchers new to the field can easily use these tools with little instruction. Gene manipulation is being widely used in both basic research and the development of novel therapies, and it has become commonplace to knock genes in and or out of cells, or to manipulate the genomes of viruses, single cell organisms, and even complex vertebrates. Subsequently, the number of gene-editing publications in recent years has soared.

The workhorses of the gene-editing field are the targeted endonucleases. Targeted endonucleases are engineered enzymes that can introduce DNA double strand breaks (DSBs) with high specificity into desired target sequences. Several site-specific endonuclease platforms are now available, and their ability to precisely cleave a desired locus is being used to disrupt genes via mutagenic non-homologous end joining (NHEJ) or to direct homologous recombination mediated gene insertion/correction. Endonuclease platforms including meganucleases/homing endonucleases (HEs), zinc finger nucleases (ZFNs), Tal-effector nucleases (TALENS), and the RNA-guided engineered nucleases (RGENs) such as CRISPR/Cas9 can be used for gene editing, and they are discussed in greater detail elsewhere (Mussolino and Cathomen 2012; Rahman et al. 2011; Sander and Joung 2014; Stoddard 2011). Importantly, gene-editing tools like the targeted endonucleases are now being developed as therapeutics that can impact many types of disease at the genetic level.

Despite advances in modern medicine, a broad spectrum of infectious diseases pose a serious global health burden. Many pathogens currently have no vaccine, and cannot be treated with antimicrobial drugs. Other pathogens may be preventable through vaccination or treatable with drugs, but remain a serious health concern due to the chronic nature of the illness, or the lack of access to basic health care within the populations most at risk. Subsequently, much effort is required to develop new or alternative therapies that can be used to prevent or treat numerous serious diseases.

In this review, we introduce the recent development of gene-editing technologies as therapeutics to eliminate persistent/chronic viral infections. Hundreds of millions of people are affected worldwide by debilitating or life-threatening persistent viral infections (Boldogh et al. 1996; Kane and Golovkina 2010), and gene-editing technology provides a method to target therapeutics towards distinct viral targets. Since the first report of an endonuclease therapy that targeted the hepatitis B virus in 2010 (Cradick et al. 2010), more than 30 subsequent publications have described gene-editing therapeutics targeting the long-lived forms of viral nucleic acid that enable viral persistence. These therapies now offer the potential to functionally cure several chronic viral infections when used either alone, or in combination with other novel or existing antiviral therapies. Here we review efforts directed towards persistent viruses other than HIV, which will be discussed elsewhere in this special issue (Benjamin et al. 2016).

The current status of gene-editing antiviral therapies

Hepatitis B virus

Despite the development of an effective vaccine over 30 years ago, it is estimated that more than 240 million people globally are chronically infected with the hepatitis B virus (HBV), and more than 780,000 people are thought to die each year due to complications of chronic HBV infection (Lozano et al. 2012; WHO 2015a). A significant number of patients with chronic HBV develop life-threatening cirrhosis, and the high incidence of hepatocellular carcinoma caused by HBV makes it the second most common cause of infection-related cancer (de Martel et al. 2012). Although several drugs can be used to suppress HBV replication, they are not curative and can only slow down disease progression. HBV is a partially double-stranded DNA enveloped virus that causes life-long infections in approximately 5–10 % of infected individuals. Chronic HBV infections are maintained by the covalently closed circular DNA (cccDNA) HBV ‘minichromosome’, which accumulates in infected hepatocytes and acts as a permanent template for virus replication. Due to the life-long persistence of cccDNA in chronically infected patients, cccDNA elimination is a requirement for potential curative therapies.

The first study to propose that a targeted endonuclease could be used to cleave and disrupt viral genomic DNA was performed by Cradick et al. (2010). They were able to demonstrate that designer ZFNs targeting HBV genes could introduce mutations at the target site in a HBV-containing plasmid and inhibit production of pregenomic viral RNA in an in vitro model of replication. Since this first study, numerous other groups have employed similar strategies to target HBV. The first follow-up studies showed antiviral activity in vitro by ZFNs delivered using adeno-associated virus (AAV) vectors (Weber et al. 2014b), or by TALENs in vitro and in a mouse hydrodynamic injection model of HBV lacking cccDNA (Bloom et al. 2013; Chen et al. 2014). Importantly, the in vitro studies using TALENs showed for the first time that HBV cccDNA could be cleaved and disrupted directly.

More recently the CRISPR/Cas9 system has been utilized as a therapeutic to target HBV. CRISPR/Cas9 is able to inhibit HBV replication and target cccDNA in a number of in vitro models (Karimova et al. 2015; Kennedy et al. 2014a; Liu et al. 2015a; Ramanan et al. 2015; Seeger and Sohn 2014; Wang et al. 2015). It has also shown the ability to suppress HBV antigen expression from plasmids in a mouse hydrodynamic injection model of HBV that lacks cccDNA (Lin et al. 2014; Zhen et al. 2015). More recently it was demonstrated that a recombinant form of cccDNA called rcccDNA could be cleaved, disrupted, and/or cleared from mouse liver when Cas9 and guide RNAs were delivered using plasmids (Dong et al. 2015). In this new hydrodynamic mouse model rcccDNA enables viral persistence (Qi et al. 2014), making the work of Dong et al. the most comprehensive in vivo demonstration of an efficacious gene-editing HBV-directed therapeutic.

As endonuclease-based approaches to treat HBV are developed further, it will be important to assess their antiviral activity in more relevant in vivo models of HBV infection. HBV cannot fully replicate in mice, and several humanized mouse models with chimeric livers have been developed to study human liver diseases including HBV (Dandri and Lutgehetmann 2014). In the future the efficacy of HBV-specific endonucleases should be tested in such a model with fully replicating HBV. However, it is important to note that humanized liver models of HBV are limited by the lack of a functional immune system, which plays a significant role in the control of (or lack of) natural HBV infections (Ferrari 2015; Yoshio and Kanto 2016). Other aspects of treatment will also need to be addressed such as enzyme delivery, enzyme dosing, development of cleavage resistance, and the levels of intrahepatic cccDNA (Schiffer et al. 2013).

Epstein–Barr Virus

The Epstein–Barr Virus (EBV) is a large linear dsDNA gamma herpesvirus, to which more than 95 % of adults have been exposed (Kutok and Wang 2006; Longnecker et al. 2013). EBV causes infectious mononucleosis and is associated with a number of malignant tumors including Burkitt lymphoma, nasopharyngeal carcinoma, Hodgkin lymphoma and non-Hodgkin lymphoma. After infection, lytic replication of EBV first occurs in epithelial cells on mucosal surfaces, but the main targets of infection are B-lymphocytes where EBV establishes a lifelong latent infection. In latently infected B cells, the linear genome circularizes to form a long-lived episome that provides a template for viral persistence. There is currently no cure for EBV, so a gene-editing strategy that targets the latent viral episome for disruption would be an attractive therapeutic strategy to prevent reactivation from latency.

The CRISPR/Cas9 system has been used as a therapeutic to target EBV. Wang and Quake were able to target latent EBV genomes in a Burkitt Lymphoma cell line and demonstrated inhibition of viral replication and cellular proliferation (Wang and Quake 2014). Guide RNAs were targeted to EBNA1 to excise EBNA1 or inhibit gene regulation and replication, to EBV repeats to promote genome shredding, or to EBNA3C and LMP1 to inhibit cellular transformation. By combining a total of 7 sgRNAs they demonstrated a 65–85 % reduction in EBV viral loads in Raji cells. They also demonstrated restoration of apoptosis in treated cells and arrest of cellular proliferation. CRISPR/Cas9 treatment caused shredding of the viral genome, repressed functions supplied by targeted genes, and inhibited viral replication. In another study, Yuen et al. (2014) targeted the EBV BART region using CRISPR/Cas9. They were able to delete a segment of the BART promoter region and this led to a reduction in the miRNAs produced from BART, and also led to a reduction in the viral load. The miRNAs and protein products from the BART region are involved in cellular transformation and inhibition of apoptosis (Marquitz et al. 2011). Disruption of these viral genes can prevent re-activation from latency, cellular transformation and can potentially be curative. The encouraging data from these two studies suggests that the dsDNA EBV episome could be effectively targeted with an antiviral gene-editing therapeutic.

Herpes simplex virus

Herpes simplex viruses (HSV-1 and HSV-2) are neurotropic alpha herpesviruses that cause oral and anogenital ulceration, ocular disease, and in rare cases encephalitis. In different parts of the world an estimated 60–90 % of the population has been exposed to HSV-1 (Smith and Robinson 2002) and 10–80 % have chronic HSV-2 infections (Looker et al. 2015; Xu et al. 2002). Upon initial infection, HSV undergoes lytic replication in epithelial cells at mucosal surfaces. From there, HSV enters sensory nerve endings and travels via retrograde axonal transport to neuronal cell bodies in the trigeminal or dorsal root ganglia where it establishes life-long latency (Nicoll et al. 2012; Subak-Sharpe and Dargan 1998). In sensory neurons and autonomic ganglia, the HSV genome forms a long-lived circular viral episome that acts as the template for viral reactivation. HSV is able to sporadically reactivate from latency, which can lead to symptomatic disease at the peripheral site of infection. There is also frequent subclinical HSV shedding that is asymptomatic, but may be important in disease transmission (Mertz et al. 1992). There is currently no vaccine for HSV (Koelle and Corey 2003), but HSV can be treated with acyclovir, which inhibits lytic HSV replication, but does not cure latent HSV infections or prevent reactivation from latency. Therefore, a treatment for HSV that prevents reactivation from latency would be highly beneficial.

Curative therapies for HSV must eliminate or inactivate the HSV episome in sensory ganglia. Recently, three groups demonstrated that engineered meganucleases targeting the UL19 gene of HSV could disrupt the HSV genome and inhibit viral replication. Grosse et al. (2011) first demonstrated that meganuclease-mediated disruption of the HSV genome could inhibit lytic replication of HSV in vitro. Aubert et al. (2014) then showed that latent HSV genomes could be targeted for disruption by an HSV-specific meganuclease using a primary human fibroblast model of HSV latency. In this model of latency, disruption of the HSV genome was concomitant with reduction in viral titers following HSV reactivation. Finally, Elbadawy et al. used a HSV-specific meganuclease to introduce mutations into HSV genomes in explanted human corneas (Elbadawy et al. 2014). A 23.4 % reduction in HSV viral load was seen after treatment with HSV-specific meganuclease compared to controls. These data suggest that targeted disruption of latent HSV genomes in sensory ganglia could prevent reactivation from latency and offers hope for a sterilizing cure for HSV.

Human papillomavirus

The human papillomavirus (HPV) is a member of the extensive papillomavirus family of DNA viruses. Papilloma viruses infect basal cells of stratified epithelium within skin or mucous membranes, causing benign papillomas or pre-malignant lesions. HPV is the leading cause of infection-related cancer in humans (de Martel et al. 2012), and is directly responsible for more than 70 % of some prevalent cancers such as cervical cancer (WHO 2015c). An effective preventive vaccine that targets several high-risk types of HPV has been available for approximately 10 years, but vaccine use worldwide has been low and there remains a serious need for therapeutic interventions in patients who have already been exposed to HPV. After initial infection, the HPV genome forms a highly stable double-stranded DNA episome within an infected cell, and this episome enables HPV to establish a latent infection. As latent HPV infections can be present for years before viral reactivation occurs, it is important that new therapies efficiently target the viral episome for clearance or disruption.

As a virus that is responsible for a substantial disease burden with a well-characterized life cycle, HPV represents an attractive target for therapeutic gene editing. The HPV E6 and E7 viral proteins have long been known to be oncogenic, and therefore represent excellent targets for therapeutic endonucleases. The first attempts to target HPV with endonucleases used ZFNs that were designed to target the E6 or E7 genes of the high-risk HPV types 16 or 18 (Ding et al. 2014; Mino et al. 2013, 2014). These studies demonstrated that ZFNs could be used to disrupt HPV genomes in in vitro culture models and HPV-positive cervical cancer cell lines. Importantly, one of these studies also showed that ZFNs could be used to inhibit growth of HPV-positive tumor xenografts (Ding et al. 2014), demonstrating their potential uses as anti-cancer therapeutics. In another comprehensive study, it was demonstrated that TALENs could also be used as anti-HPV therapeutics (Hu et al. 2014a). TALENs targeting the E6 and E7 genes of HPV-16 and HPV-18 showed efficacy in several HPV in vitro models, and importantly were able to reverse the HPV-driven malignant phenotype of K14-HPV16 transgenic mice that are immune competent.

Recently the CRISPR/Cas9 system has also been used as an anti-HPV therapeutic, and reagents have been developed to target the E6 or E7 genes of the high-risk HPV types 6, 11, 16 or 18. CRISPR/Cas9 has shown anti-HPV efficacy in a number of in vitro cell culture models, and in an HPV-positive tumor xenograft model (Hu et al. 2014b; Kennedy et al. 2014b; Liu et al. 2015b; Yu et al. 2015; Zhen et al. 2014). Efficient restoration of the p53 pathway and up-regulation of pRb was shown, indicating that the functions of the HPV E6 and E7 viral proteins were being disrupted. Taken together this data suggests that the persistent HPV episome could be a viable target for antiviral endonuclease therapy.

Human T cell leukemia virus

Human T-cell leukemia virus (HTLV) is a delta retrovirus that can cause adult T-cell leukemia/lymphoma (ATLL) and is also linked with tropical spastic paresis (TSP), also known as HTLV-associated myelopathy (Gessain et al. 1985; Kondo et al. 1989; Osame et al. 1986; Poiesz et al. 1980). HTLV has also been associated with a variety of other inflammatory disorders including myositis, uveitis, and dermatitis. Upon infection of T lymphocytes, HTLV integrates its proviral genome into the host cell genome and establishes viral latency, only undergoing sporadic stochastic reactivation. To enable it to establish a life-long infection, the latent HTLV proviral genome does not express any viral proteins, and this helps prevent HTLV-infected cells from being cleared by the immune system. There are currently no treatment interventions for either acute or chronic HTLV infections.

The HTLV proviral genome is an attractive target for gene-editing enzymes, as disruption of the integrated HTLV viral genome could prevent reactivation from latency and potentially be curative. HTLV has been targeted for disruption by Tanaka et al, who developed ZFNs that target highly conserved regions of the HTLV long terminal repeats (Tanaka et al. 2013). They were able to demonstrate that ZFNs could introduce mutations in the LTR that inhibited LTR-driven HTLV gene expression. Furthermore, ZFN expression inhibited cellular proliferation in HTLV-1 transformed cell lines, and inhibited tumor growth in a xenograft model that used the HTLV positive adult T cell leukemia ED cell line. These results demonstrated the potential of endonucleases as therapeutics targeting integrated proviruses during retrovirus infections.

JC polyoma virus

JC polyoma virus (JC virus) is a double-stranded DNA virus of the family Polyomaviridae that was first isolated from an immunocompromised patient with progressive multifocal leukoencephalopathy (Padgett et al. 1971). JC virus is not known to cause disease pathology in immune-competent hosts, but is associated with disease in immunocompromised hosts including patients with cancer, HIV, inflammatory disorders, or receiving immunomodulatory therapies. The global prevalence of JC virus ranges between 20 and 60 % depending on the serological assay used (Egli et al. 2009; Gossai et al. 2016; Knowles 2006), and there are currently no curative therapies or vaccine for JC virus.

Gene-editing technologies offer a potential therapy for the eradication of persistent JC virus infections. Wollebo et al. (2015) demonstrated that JCV T-antigen expression could be inhibited using the CRISPR/Cas9 system and three guide RNAs that target segments of the JC virus genome encoding the T antigen. They were able to suppress T-antigen expression in SV40 or JC virus transformed cell lines and inhibit JC virus replication in permissive cell lines. The observations from this study are important, as some polyoma virus T-antigens are involved in cell transformation, and gene-editing therapeutics could therefore prevent the pathogenesis of polyoma virus associated tumors (Gupta et al. 2016).

Hepatitis C virus

Hepatitis C virus (HCV) is a positive-sense single-stranded RNA virus from the flavivirus family that is associated with hepatitis, liver fibrosis, cirrhosis and development of hepatocellular carcinoma. Almost 80 % of people who get infected with HCV go on to develop chronic infections, and worldwide an estimated 110–180 million people have chronic HCV (Gower et al. 2014; Messina et al. 2015; WHO 2015b). It is estimated that approximately 500,000 people die each year from complications that develop during chronic HCV infection. Until recently the mainstay of treatment for HCV was ribavirin and pegylated interferon alpha, but new direct acting antiviral (DAA) therapies were recently approved by the FDA that have enabled high cure rates in patients receiving therapy. Unfortunately, even after cure the chance of re-infection remains high for those with recurrent risk of exposure. Therefore, a therapy against HCV with long-lasting activity would be beneficial.

Unlike the other viruses described in this review, HCV is an RNA virus. Subsequently, the targeting of viral genomes in HCV infected cells with endonucleases would seem problematic, because of the DNA-specific nature of nucleic acid cleavage performed by most classes of endonuclease. However, a recent study targeting the highly conserved 5′UTR and 3′UTR regions of HCV with gene-editing therapeutics offered a viable treatment alternative. Price et al. (2015) investigated the impact of HCV UTR disruption on virus replication with the CRISPR/Cas9 system. Using a modified Cas9 from Franciscella novicida (FnCas9) they generated guide-RNAs that are complementary to the 5′UTR regions of the HCV genome. They showed up to a 40 % reduction in the translation of HCV proteins following treatment in Huh-7.5 human hepatocellular carcinoma cells. They also demonstrated that RNA targeting with CRISPR/Cas9 does not have the same requirements for a PAM sequence as DNA targeting, and that a nuclease-deficient FnCas9 with target guide RNAs could disrupt translation of viral products through binding to target RNA species. Although the antiviral effects seen were RNA cleavage independent and not strictly caused by gene editing, this study has opened up the possibility of using gene-editing reagents as antiviral therapeutics for RNA viruses. In the light of recent advances in DAA therapies that cure a high proportion of infected patients but not all, a gene-targeting therapy with efficacy against DAA-resistant HCV genotypes could provide a therapeutic pathway forward.

Hurdles to human application

New gene-editing reagents are developed almost daily, and the outlook for future therapeutic interventions is bright. However, a number of hurdles still remain before reagents that target persistent/chronic viral infections can be transitioned into the clinic.

Off-target endonuclease cleavage

One of the major obstacles to gene editing is the possibility that off-target cleavage can occur at a similar or identical site within the human genome. In theory off-target cleavage at an undesired location could introduce deleterious mutations in the genome, or lead to chromosomal translocation. Either of these events can cause cellular toxicity or transformation, although the outcome depends on whether an essential housekeeping gene or a gene involved in cell cycle regulation is altered. Due to the size of the human genome, probability suggests that an endonuclease target site of less than 17 base pairs has a 50 % or higher chance of being present within the human genome (Schiffer et al. 2012). Therefore, much effort has been put into analyzing the target-site cleavage specificity of each gene-editing platform when single or multiple nucleotide differences are present within homologous off-target sites in the host cell genome.

Some of the earliest efforts to assess the off-target cleavage activity of targeted endonucleases focused on HEs. Unlike other targeted endonuclease platforms, HEs use a single protein domain for target site binding and cleavage, and it was suggested that any off-target activity would be more limited and predictable because potential off-target sites need to be substrates for both binding and cleavage. Although off-target cleavage by HEs can been observed (Barzel et al. 2011), the levels seen depend on the particular HE and in general tend to be low. Subsequently, newer HE platforms have been developed that increase enzyme specificity, and subsequently the levels of HE-mediated off-target cleavage can be significantly minimized (Boissel et al. 2014; Wolfs et al. 2014).

Like TALENs and RGENs, ZFNs contain a DNA-binding domain, and a DNA-cleavage domain that is independent of the DNA target sequence. Therefore, the off-target activity of TALENs, RGENs, and ZFNs is defined by the specificity of their DNA-binding domains. For ZFNs this is complicated by the intricate context dependence of DNA binding for individual zinc finger domains when they are linked (Cornu et al. 2008; Sander et al. 2011). A number of studies have demonstrated that ZFNs can have significant off-target cleavage activity, although the extent of off-target cleavage is highly dependent on the ZFN being tested (Gabriel et al. 2011; Pattanayak et al. 2011). To limit the risk of off-target cleavage, modified ZFNs with altered FokI-derived cleavage domains have been developed (Miller et al. 2007; Szczepek et al. 2007). These ZFNs will only cleave DNA when the two ZFN subunits bind next to each other at their target site and an obligate FokI heterodimer forms. By using this approach the level of ZFN off-target activity can be significantly reduced, although some modified ZFNs can still be significantly cytotoxic (Weber et al. 2014b).

Like HEs and ZFNs, TALENs are able to cleave homologous off-target sites within the host cell genome (Guilinger et al. 2014). However, TALENs appear to have higher specificity for target-site binding and cleavage than ZFNs, as they are less likely to cleave off-target sites in the host cell genome and cause cytotoxicity (Mussolino et al. 2014). This improved specificity is likely because TAL-effector DNA-binding domains are less tolerant of local target sequence mismatches than ZFNs. It has also been suggested that TALENs have a favorable safety profile when compared to CRISPR/Cas9 (Frock et al. 2015), although cleavage at multiple off-target sites was seen, along with evidence of cleavage-induced chromosomal translocations.

Despite being the newest gene-editing tools available, RGENs have been studied extensively for their ability to mediate off-target cleavage. A number of studies have shown that the CRISPR/Cas9 system can cleave off-target sites, even when multiple target site mismatches are present (Fu et al. 2013; Kim et al. 2016; Tsai et al. 2015). Therefore, substantial work has gone toward limiting the off-target activity of the CRISPR/Cas9 system. Work from the Church and Zhang labs has shown that CRISPR/Cas9 systems that use larger PAM sequences have improved off-target cleavage profiles (Esvelt et al. 2013; Ran et al. 2015). Alternatively, the Joung lab has shown that guide RNAs with shorter sequences are less likely to cleave at off-target sites (Fu et al. 2014). More recently both the Joung and Zhang labs were able to engineer the spCas9 protein so that off-target cleavage was either significantly reduced or not detectable (Kleinstiver et al. 2016; Slaymaker et al. 2016). These latest advances in CRISPR/Cas9 reagents provide hope that off-target activity may not be a problem for gene-editing technologies in the future.

Endonuclease safety in vivo

Aside from toxicity caused by off-target cleavage, other safety risks must also be addressed before endonucleases are used in humans. All of the gene editing platforms described here and elsewhere (Mussolino and Cathomen 2012; Rahman et al. 2011; Sander and Joung 2014; Stoddard 2011) are at least partially derived from non-human organisms, so they represent potential immunogens if delivered to humans. Of major concern is the risk that expression of an engineered endonuclease within a target cell may make it a target for immune-mediated clearance in vivo.

To date very little has been done to investigate the potential in vivo toxicity or immunogenicity of engineered endonucleases. Nevertheless, the existing data suggests that targeted endonuclease expression in vivo may not be associated with significant toxicity. No report of in vivo toxicity was seen when HEs were expressed in transgenic mice (White et al. 2013), or delivered to mice by plasmids (Riu et al. 2005) or adenovirus vectors (Gouble et al. 2006). No ZFN-related in vivo toxicity or immunogenicity was seen in human clinical trials, where patients who received ZFN-treated T cells retained gene modified cells for at least 252 days post-transplant (Tebas et al. 2014), although T-cell expansion prior to infusion likely limited ongoing levels of ZFN expression. Similarly, no in vivo toxicity was observed when TALENs were used as therapeutics in mouse models of HBV and HPV (Bloom et al. 2013; Chen et al. 2014; Hu et al. 2014a). For CRISPR/Cas9 no associated toxicity was seen in CRISPR knock-in mice (Platt et al. 2014), or when saCas9 (Ran et al. 2015) or spCas9 (Swiech et al. 2015) were delivered to the liver and brain, respectively. Despite these observations, the effect of long-term exposure to targeted endonucleases remains unknown, and ‘safety switches’ like those being developed for cell-based therapies may be needed (Barese et al. 2015; Di Stasi et al. 2011). Conditional expression of gene-editing endonucleases from tissue-specific promoters, drug-inducible promoters or non-constitutive viral promoters may also improve treatment safety. Further studies are warranted to fully investigate the safety and immunogenicity of targeted endonucleases following long-term expression in vivo.

Delivery challenges

Possibly the greatest challenge facing endonuclease-mediated antiviral therapies is the need to deliver each virus-specific enzyme to all infected cells. For each persistent/chronic viral infection the hurdles to achieving this vary greatly, and have been discussed in detail elsewhere (Weber et al. 2014a). Essentially two main questions need to be addressed for each virus. Is there a therapeutic delivery system that can reach all of the target cell population? And can a given enzyme therapeutic be packaged into this delivery system?

The choice of delivery system used in all gene therapies is primarily dictated by the cell type or organ that must be transduced. Some viruses, such as HBV or HSV, are primarily found within a single cell type that is only found in a particular anatomic location, and this makes gene delivery easier. Other viruses, such as EBV and HTLV, target cell types that are distributed throughout the body, complicating delivery of therapeutics to all infected cells. A large number of non-viral and viral systems are available to achieve gene delivery to different virus-infected cell populations (reviewed in Pereyra and Herenu 2013), and for each target cell population more than one gene delivery system may enable efficient gene transfer. Each delivery system must have its efficacy, safety, and cost assessed on a per-virus basis.

For each gene-editing platform the requirements for delivery vary. HEs are derived from a single small gene so they can be efficiently delivered by most gene delivery systems. ZFNs and TALENs have two separate subunits that may be small (ZFNs) or large (TALENs), and the size constraints imposed by this restricts which gene delivery platform can be used. For example, ZFNs have been delivered by AAV vectors (Anguela et al. 2013; Li et al. 2011; Weber et al. 2014b), whereas TALENs have not. Additionally, the repeat domains present in ZFNs and TALENs make it difficult to incorporate separate subunits into a single vector because of the risk of recombination during production. CRISPR/Cas9 is limited by the need to deliver both a large Cas9 gene and a guide RNA, which makes their incorporation into some gene delivery vectors difficult. The discovery of smaller Cas9 genes has made this easier (Hou et al. 2013; Ran et al. 2015), but delivery of CRISPR/Cas9 to certain cell types may still be challenging. Ultimately each researcher must identify a gene delivery system that efficiently targets the correct target cells, and can accommodate their endonuclease of choice.

Treatment resistance

A large obstacle to any antiviral therapy is the potential to develop treatment resistance (Fig. 1). For gene-editing antiviral therapies, resistance could be caused by the acquisition of target site mutations that allow viral replication but prevent endonuclease binding or cleavage. Such mutations could be attained either during replication of the viral genome, or when a deleterious mutation is introduced following targeted gene editing. Alternatively, resistance could be gained when a defective viral gene target is restored to its wild-type state by a recombination event with a functional viral genome. Although de novo replication-derived resistance would be less likely to occur for the less error-prone DNA viruses compared to RNA viruses, any of these potential outcomes would be an unwanted consequence of a gene-editing antiviral therapy.

Fig. 1.

Generation of endonuclease-resistant virus. Upon endonuclease cleavage of an essential viral gene the DNA double strand break can be repaired perfectly (A), mutated to yield a replication incompetent virus (B), mutated to yield a treatment-resistant virus (C), or treated with a second endonuclease to prevent treatment resistance (D)

De Silva Feelixge et al. (2016) were the first to demonstrate that viral resistance can occur following endonuclease therapy. They showed that an insertion mutation introduced into the HIV provirus following ZFN therapy enabled virus replication and ZFN cleavage resistance. The mutant virus replicated at close to wild-type levels, and was infectious in activated primary T cells. However, resistance could be overcome when multiple regions of the viral genome were targeted for disruption with separate endonucleases. Wang et al. (2016a) further demonstrated that treatment resistant virus can emerge when HIV cultures are treated with CRISPR/Cas9 reagents that individually target different regions of the provirus. Resistance-yielding mutations were found in HIV target sites that were likely generated during the process of DSB repair after Cas9 cleavage. This CRISPR/Cas9 study was soon followed by another that also demonstrated resistance to cleavage at target sites present in HIV pol, env and LTR (Wang et al. 2016b). Although the high reverse transcriptase error rate of HIV makes the likelihood of de novo replication-derived resistance greater than for other DNA viruses, these three studies demonstrate that endonuclease-derived resistance could be a significant concern. Future studies should determine whether endonuclease resistance can occur for other viruses, and if so, whether combinations of endonucleases (De Silva Feelixge et al. 2016) or sgRNAs (Kennedy et al. 2015) can avoid the development of resistance.

Treatment cost

A major impediment to the widespread application of gene-editing antivirals is the potential cost to the populations most at risk of infection. Although in their relative infancy, gene therapeutics have been used widely in clinical trials, but the cost of reagent production has been extremely high, and the pricing of the recently approved gene therapy product Glybera suggests high costs will trickle down to patients (Morrison 2015). For many chronic diseases a one-off high-cost curative therapy might ultimately be cheaper than the cost of life-long noncurative treatment, but this is not true for all diseases. Subsequently, new models of gene therapy pricing are being discussed (Brennan and Wilson 2014; Touchot and Flume 2015).

For gene-editing reagents against viruses the large-scale production of plasmids, recombinant proteins or viral vectors that deliver antiviral therapeutics is also likely to be high. For example, Glybera, a therapeutic based on an AAV viral vector, was initially priced as high as $1.5 million. With some of the virus targets discussed here, mathematical modeling also suggests that multiple treatments will be likely, which would significantly increase treatment costs (Schiffer et al. 2013). In some cases, persistent viral infections can be adequately controlled with relatively inexpensive antiviral drugs, so this option would likely be preferred in many settings, particularly as cheap antiviral drugs become more readily available globally. Therefore, diligence will be required to ensure that gene-editing therapies will be affordable for populations in developing nations who are disproportionally affected by chronic and persistent infections.

Conclusions

Over the last 5 years advances in gene-editing technology have made it possible to target both DNA and RNA viruses with endonuclease-based antiviral therapeutics. This has opened the door for future therapies targeting chronic/persistent viral infections for which there are currently no or poor treatments. Although a fast-growing body of evidence suggests that endonucleases can be effective antiviral agents in vitro and in vivo, significant efforts are still needed to improve the efficacy, safety, and delivery of individual treatments. Nevertheless, the rapid advances in gene-editing technologies bring hope that this approach to treating persistent viral infections can 1 day lead to human application.