Prospects of Repeated Pulmonary Administration of Viral Vectors

Abstract

Pulmonary gene therapy may ultimately cure diseases such as cystic fibrosis, α1-antitrypsin deficiency, lung cancer, and pulmonary hypertension. Efficient expression of delivered genes in target cell types is essential for the achievement of this goal. To this end, readministration of viral vectors may be required: 1) to increase the percentage of transduced airway epithelial cells, 2) to direct gene transfer to individual lobes during different delivery sessions, or 3) to boost attenuated expression over time. Immune responses to viral proteins or viral encoded proteins are the greatest barrier to repeated vector administration.

Advances in understanding of innate and adaptive immune responses to viral vectors are informing the design and application of vectors for single and multiple dose regimens

This review focuses of vector re-administration to the mucosal surface of the lung. Here immune responses are generally viewed as undesirable. This contrasts with applications such as cancer therapies or vaccines, where immunostimulation is a goal. Pulmonary barriers to viral pathogens generally fall into two groups: physical and immunological. Physical barriers include mucociliary clearance and access to receptors (Figure 1). Immunological barriers are the innate and adaptive immune responses that have evolved to protect the host. Often, the pulmonary barriers to natural pathogens are just as effective at preventing gene transfer with therapeutic viral vectors. Innate immune responses to viruses or viral vectors are pre-existing or generated within minutes to hours following administration. In contrast, adaptive immunity requires days to weeks to develop. The innate immunity front-line in a healthy lung is patrolled by phagocytes (ie. alveolar macrophages) that quickly engulf vector particles and thereby prevent transduction of target cells. Other phagocytic cell types, such as neutrophils, dendritic cells and natural killer T cells, are recruited in response to viral vectors or pathogens (Figure 1).

Figure 1.

Schematic of potential barriers to single or repeated administration of viral vectors in the lung. PMN = polymorphonuclear leukocyte, ASL = airway surface liquid.

Phagocytic cells and epithelia express Toll-like receptors (TLRs) and RIG-I-like (retinoid-inducible gene 1-like) receptors (RLRs), which recognize nucleic acids or proteins derived from viral pathogens.¹ Examples include viral DNA, single-stranded RNA- and double-stranded RNA. Stimulation of TLRs initiates a signaling cascade consisting of at least two distinct pathways (Figure 2). The first is a MyD88-dependent activation of transcription factor NF-κB, which induces the secretion of proinflammatory chemokines and cytokines. The second pathway induces type I interferons (IFN). Subsequent induction of IFN-stimulated genes leads to the activation of antiviral factors, such as RNA-activated protein kinase. The RLRs such as RIG-I, Mda5, and LGP2 are expressed in the cytoplasm and recognize viral RNA.¹ The cytokine responses following delivery of a single dose of adenoviral (AdV) or adeno-associated virus (AAV) vector to the lung have been well studied. Many preclinical studies and clinical trials with adenoviral vectors document increases in systemic chemokines and pro-inflammatory cytokines including RANTES, IP-10, MIP-2, IFN-γ, TNF-α, IL-6, and IL-12 in humans and mice.

Figure 2.

Simplified schematic of innate immune system cell surface and endosomal pattern recognition receptors of viral nucleic acids and the activated signaling pathways relevant to gene transfer.

Cytokine responses following lentiviral vector (LV) delivery are less well studied. Following a single delivery of titer-matched lentivirus vector feline immunodeficiency virus (FIV), AdV5 or vehicle to the nasal epithelia of mice, we observed a transient, early increase in IL-6 and KC (human IL-8 ortholog).² These data suggest that vehicle impurities and/or simple physical irritation of the respiratory mucosa are sufficient to elicit transient cytokine release. Importantly, these cytokine responses did not preclude persistent transgene expression following lentiviral vector transduction.

Pre-existing immunity may prevent the success of a single administration of viral vector

Adenovirus

Most adult humans have preexisting immunity against many of the human AdV serotypes following exposure through natural infections. Pre-existing viral immunity is serotype-dependent and some AdV serotypes are less prevalent than others. Such immunity against a native serotype will significantly reduce the efficiency of gene transfer from a recombinant AdV of the same serotype.³ This observation raises the practical matter that surveillance for pre-existing immunity may need to be considered in the design of human gene transfer trials. For example, Roth and colleagues have found that patients with higher anti-AdV neutralizing antibody titers have reduced therapeutic responses to AdV-based cancer therapies.

Several groups have shown that adenoviral vectors will elicit sufficient humoral and cell-mediated immunity to prevent pulmonary re-administration. The cellular immune response, mediated through AdV-specific CD8+ T cells, eliminates the target cells expressing viral and transgene products. This causes a rapid loss of transgene expression in experimental animals inoculated with AdV.⁴

Adeno-associated virus

A number of AAV vector capsid serotypes have been delivered to the lung⁵ and in most cases, a second dose of AAV yields lower levels of pulmonary gene transfer because of the induction of neutralizing antibody responses. Several epidemiological studies report that AAV2 neutralizing antibodies are present in 20–67% of the human population (reviewed in ⁶). One theory implies that the number of antibodies required to neutralize the invading virus correlates to the surface area of the particle. Neutralization presumably occurs through physical interference of viral interactions with cell surface receptors.⁶

Lentivirus

In contrast to the encapsidated adenoviral and AAV vectors, the more genetically and structurally simple lentivirus particle contains fewer proteins. Therefore, the diversity of antigens presented by a pseudotyped lentivirus is typically more limited. While the consequences of repeat administration of lentivirus vectors are less well studied, some details are emerging. The responses to the replication competent viruses from which the vectors are derived have been described in some detail. For example, G-U-rich single-stranded RNA derived from HIV-1 can stimulate dendritic cells and macrophages via TLR-7 (mouse) or TLR-8 (human) and induce the secretion of IFN-α and other proinflammatory cytokines (reviewed in ¹). Furthermore, murine retroviruses activate B cells via TLR-4, initiating TLR-mediated production of proinflammatory cytokines and chemokines. Viruses and presumably viral vectors naturally activate TLRs and PKR and induce early innate responses. TLR signaling leads to DC activation, which in turn initiates CD4 and CD8 T cell differentiation necessary to generate adaptive cellular immune responses. In addition, one might anticipate that the vector particle itself or contaminants within the vector preparation may confer adjuvant activity. Clearly, AdV-, AAV-, or LV-induced immune responses direct the magnitude of T cell responses and ultimately shape the levels and persistence of vector expressed proteins.

Strategies to allow adenoviral vector re-administration have been assessed

The host responses to the re-administration of AdV are the most investigated; however, there is growing interest in anti-AAV and lentiviral vector adaptive immune responses. Robust pulmonary immune responses to AdV have been observed in rodents and non-human primates. In human trials with AdV, neutralizing antibodies were found in bronchoalveolar and nasal lavage and serum. Readministration of therapeutic doses of AdV to the airways in the absence of acute or chronic immunosuppression will require novel strategies.

Using alternate serotypes (serotype switching) is a strategy proposed to evade immune responses, allow repeated administration, and prolong expression from AdV. While Mastrangeli and co-workers demonstrated in animal studies that serotype switching between AdV 4, 5, and 30 confers some ability to readminister vector, this work has not been translated to the clinic. In addition, CTL responses to conserved vector proteins remain persistent problems limiting long term expression.

Covalent AdV capsid modification with polyethylene glycol (PEGylation) and other carriers has been demonstrated to mask vector epitopes and achieve protective activity against neutralizing antibodies in vitro, and to a lesser degree, in vivo. The covalent linkage of PEG to free lysine residues on the vector capsid confers this effect. Croyle and colleagues reported that the transduction efficiency of PEGylated AdV was not reduced in vivo. Eto et al. used a strategy that combined integrin-targeting RGD peptide and PEG to modify the AdV⁷ that enhanced gene transfer in both Coxsackie-adenovirus receptor (CAR)-positive and -negative cells.

Further, the formulation of AdV with lipids and other carriers may also facilitate AdV readministration. Formulation of AdV with dioleoylphosphatidylethanolamine (DOPE), carbamoyl cholesterol (DC-Chol), or dexamethasone-spermine delivered at day 0 allowed improved expression from a readministered AdV delivered at day 21 in contrast to no formulation.⁸ This approach reduced, but did not eliminate immune responses. The vector formulations were associated with a reduction in neutralizing antibodies, selected decreases in early cytokine responses, and decreased CD4 lymphocyte infiltrates in lung tissues. A posited mechanism is that the anti-inflammatory properties of the cationic lipid-dexamethasone-spermine formulation improve the profile for in vivo use. Additional studies in large animal models with more diverse immune systems are needed to appreciate feasibility of translating this finding into a human clinical application.

Helper-dependent (HD) or “gutted” adenovirus vectors were developed to eliminate the expression of viral proteins.⁹ It is important to note that the helper-dependent vectors retain all the AdV capsid proteins. In some settings, the expression from HD AdV is remarkably long-lived.¹⁰ Koehler et al ¹¹, demonstrated the dose-dependent feasibility of helper dependent (HD) AdV re-administration to the lungs of mice. Reporter gene (β-galactosidase) expression was ~10% reduced when the low dose (1.5 × 10¹⁰ particles) HD-AdV without a transgene was followed by HD-AdV with the β-galactosidase transgene. However, when HD-AdV was delivered at a higher dose (5 × 10¹⁰ particles) reporter gene expression was reduced about 80% compared to a single administration. Repeated administration of the higher vector dose was associated increased anti-AdV neutralizing antibodies in the BAL and serum compared to the lower dose protocol. In animal studies of systemic administration of PEGylated vectors, PEG treatment of HD AdV reduced the early innate immune responses to vector administration, in part due to reduced uptake of vector by macrophages and dendritic cells. These data indicate that while helper-dependent adenoviruses show promise for more prolonged expression in vivo as compared to first generation vectors, re-administration still poses significant challenges for long term therapies for chronic disease conditions.¹¹

In summary, for AdV to be an effective delivery vehicle for clinical applications requiring lasting expression, mucosal immunity to AdV must be monitored closely if vector is to be readministered. The success of gene therapy might be improved if AdV serotype switching is used to maximize the length of time between vector administrations. Further, this strategy can be used to avoid pre-existing immunity to serotypes common in human populations. The use of HD AdV and/or PEGylation may further increase the number of possible administrations. However, avoiding CTL responses to vector proteins remains a hurdle for successful re-administration, especially in the context of treatment strategies for chronic lung conditions.

Adeno-associated virus vectors can be re-administered successfully through use of novel serotypes, chemical formulation, or immunosuppression

Initial attempts at AAV readministration to the rabbit lung resulted in transduction that was not significantly different from that from single administration; in addition, repeated administration led to the appearance of neutralizing antibodies to the viral capsid. However, one additional application succeeded in transducing murine lung after treatment with anti-CD40 ligand and soluble CTL4-Ig fusion at time of initial AAV vector exposure, due to suppression of the humoral immune response. However, in non-human primates treatment with anti-CD40 ligand antibodies did not prevent virus-specific humoral responses upon second vector challenge. Additionally, repeat administration in the murine lung was achieved by utilizing alternate serotypes as well as chemical modification of the capsid of AAV2 to evade existing immunity.

Recent efforts to administer AAV repeatedly to the lung have shifted vector development from AAV2 capsids to novel serotypes such as AAV9 and those from nonhuman primates. To assess transduction efficiency of the novel serotype 9, Limberis and Wilson instilled mice intratracheally with AAV2/9 and AAV2/5 vectors expressing the hAAT human alpha-1 anti-trypsin cDNA and nuclear-targeted LacZ.⁵ They observed serum hAAT concentrations in AAV2/9-transduced mice to be roughly 74 times that of AAV2/5-transduced animals. AAV2/5 transduced both alveoli and conducting airways, whereas AAV2/9-mediated LacZ expression was seen primarily in alveoli. Limberis and Wilson reported relatively stable expression for nine months.⁵ They also noted that AAV2/9 could be readministered one month following initial exposure despite the presence of neutralizing antibodies. Mice previously exposed to AAV2/9 or 2/5 expressing hAAT were subjected to a second exposure of the same serotyped vector expressing LacZ. Initial neutralizing antibody titers were higher in serum from animals readministered with 2/5 than in those treated with 2/9. However, after six months neutralizing antibody levels to AAV2/9 exceeded those to 2/5. Vector transduction levels in the lung were reported to be roughly ten percent lower in animals readministered vector than those of naïve age-matched controls whereas there was no indication of decreased nasal epithelial transduction. Conversely, transduction was detected in neither lung nor nose in animals given a repeat dose of AAV2/5 serotype vectors. High levels of neutralizing antibodies, IgA, and IgG were detected in bronchoalveolar lavage fluid isolated from these mice in contrast to mice readministered with AAV2/9. The authors indicated that the antibody response to AAV2/5 vectors was largely systemic. Systemic responses to AAV2/9 were detectable but did not inhibit gene expression. They concluded that AAV2/9 was better able to evade mucosal immunity⁵ presumably by evading interaction with antigen-presenting cells as serotype-specific differences in activation of CD8 T cells to capsid were demonstrated.¹² However, this study did not address the time interval between vector administrations.

A report that investigated the interval between vector administrations was published recently. Sumner-Jones et al also determined the feasibility of administering AAV5 capsid/AAV5 genome vector (AAV5/5) following initial exposure to the same serotype.¹³ They delivered AAV5/5 vectors to the lungs and nasal cavities of mice in either a single dose or in two or three doses at 8 week intervals. They observed insignificant differences in expression from animals administered three doses compared with naïve mice. Conversely, animals administered a single dose of AAV5/5 showed high levels of expression. The decrease in expression from those receiving two and three doses correlated with increased neutralizing antibody titers. When the types of antibody responses were analyzed, the investigators found that the majority of neutralizing antibodies were raised against the AAV5/5 vector whereas little to no anti-transgene antibodies were detected. Additionally, significant repeat transduction was not observed when the interval between administrations was increased to 36 weeks. In contrast to previous reports, delivery of a soluble CTL4-Ig fusion did not enhance repeated transfer. In fact, there was no difference in expression in the lung between naïve animals and those readministered with AAV5/5CTL4Ig. Interestingly, anti-CTL4Ig neutralizing antibodies were detected in the sera in addition to anti-AAV5 antibodies. Notably, the investigators were able to achieve a successful repeat administration when the initial exposure was to the AAV2 capsid. These data represent the investigation of two delivery routes (intratracheal and intranasal) to one strain of laboratory mouse. The authors note that repeat administration via different routes of delivery and to different model strains might yield alternative results and should be investigated further.¹³

In addition to human serotypes, nonhuman primate AAV serotypes have been explored. AAVrh.10 serotyped vectors achieved high levels of human alpha 1 anti-trypsin (hAAT) in the presence of high level anti-AAV2 and AAV5 neutralizing antibody titers.¹³ Titer-matched doses of AAV5-, AAV2-, or AAVrh.10-hAAT were injected intrapleurally following exposure to AAV5 or AAV2 vector. AAVrh.10 vectors delivered comparable transgene levels in either immunized or naïve animals. This was in contrast to AAV2 vectors that failed to achieve detectable serum levels of hAAT when readministered to immunized mice. Similarly, AAV5 vectors achieved 980-fold lower levels of serum hAAT when readministered. Interestingly, mice first treated with AAV5 serotyped vector and then readministered AAVrh.10 vector showed serum hAAT levels increased by 300%. The authors conclude that the non-human primate AAVrh10 capsid pseudotype does not elicit an antibody response and could be used to retreat patients with existing immunity to AAV5 or AAV2. In addition, rhesus serotype AAV8 would also be an option as it achieved the second highest expression levels.¹⁴

Re-administration studies using lentiviral vectors are beginning to inform the field of their potential for repeated delivery

Lentiviral vectors are only beginning to make their debut in human clinical trials.¹⁵ Few studies to date have examined the consequences of repeated administration of retroviral- or lentiviral-based vectors in the airways or other tissues.

Using a luciferase reporter and bioluminescence imaging, we demonstrated that pseudotyping a feline immunodeficiency virus (FIV)- based lentiviral vector with the envelope glycoprotein from Autographa californica multicapsid nucleopolyhedrovirus (GP64-FIV) results in persistent in vivo gene expression following delivery of a single dose to mouse nasal epithelia. Further, we investigated the potential to repeatedly administer GP64-FIV to the respiratory tract and increase gene transfer in mice (Figure 3). Minimal innate and adaptive immune responses were observed following repeated topical delivery of a lentiviral vector to murine nasal epithelia. Interestingly, the vector failed to elicit immune responses that prevent re-administration even after 7 doses (1 dose/week). We demonstrated additive increases in transgene expression (both reporter genes and a therapeutic gene, erythropoietin) with repeat dosing. This increase in expression may represent both an increase in the percentage of cells expressing a transgene and an increase in the number of transgene copies/cell; however, no studies have yet specifically addressed this question.

Figure 3.

Gene transfer in murine nasal epithelia following 7 administrations (1 dose/7 consecutive days) of GP64 pseudotyped FIV expressing firefly luciferase. Pseudocolor overlay of bioluminescence signal on the gray scale photograph of the animal’s head.

Buckley et al. ¹⁶ recently investigated lentiviral reporter gene expression in mouse lung following a single fetal intra-amniotic administration, 3 administrations (1 fetal/2 neonatal), and 2 neonatal administrations. The fetal/neonatal re-administration achieved similar levels of epithelial cell transduction compared to a single fetal application but the number of transduced macrophages was increased. Furthermore, neonatal re-administration alone achieved a similar level of macrophage transduction but reduced levels of epithelial cell transduction, comparable to levels seen following a single neonatal administration.¹⁶ The authors speculated that this finding represents macrophages infiltrating after the initial dose of GP64 pseudotyped lentivirus and their subsequent transduction by the second and third doses.

It is possible that some envelopes are more immunogenic than others. However, it is unlikely that pre-existing immunity to the insect virus GP64 envelope glycoprotein will present a barrier in humans. We suspect that the GP64 glycoprotein, together with the vector associated proteins (ie. matrix, capsid, protease, integrase, and reverse transcriptase) are poorly antigenic when presented to the nasal mucosa. This phenomenon would presumably be the same in the conducting airways. In addition, the route of vector administration (pulmonary vs. systemic delivery) may significantly influence the subsequent development of immune responses. Naldini et al recently reported innate and adaptive immune responses to lentiviral vectors administered systemically and targeting hepatocytes in mouse models.¹⁷ Our results in the nasal airways suggest that mucosal application of a lentivirus vector is less immunostimulatory than systemic delivery.²

Prospects

Advances in immunology will likely inform the field as our understanding of the molecular underpinnings of innate immunity improves. One can envision that pharmacologic inhibitors of early innate immune responses might be developed to diminish deleterious responses or blunt adaptive immune responses. Capsid and envelope pseudotyping, directed evolution, and discovery of novel virus capsids or envelopes from other species may provide novel reagents.

To evade the neutralizing antibody response to AAV vectors, several solutions have been proposed – perhaps the most promising of which is modification of the AAV capsid. Several groups identified neutralizing epitopes on the AAV2 capsid through peptide scans, ELISA, and peptide competition experiments.⁶ Information on neutralizing epitopes coupled with work mapping receptor interaction domains resulted in discovery of capsid modifications which produce infectious vector that enhanced resistance to host neutralizing antibody responses.⁶

In addition to rational design, iterative rounds of library generation and selection –known as directed evolution – are effective in producing more infectious vectors in the lung. Recently, Excoffon and colleagues generated an AAV vector that was 100-fold more infectious than parental strains by subjecting the capsid genes from AAV2 and AAV5 to error-prone PCR and DNA shuffling. They subsequently selected the vectors on human airway epithelial cultures. The resulting variant was a chimera between AAV2 amino acids 1–128 and AAV5 amino acids 129–725 with an A581T point mutation.¹⁸ The mutation occurred in a region of AAV5 sialic acid binding which significantly increased binding and thus infectivity of the chimera. Furthermore, the newly created chimera carrying the cystic fibrosis transmembrane conductance regulator (CFTR) restored chloride current in CFTR-deficient human airway epithelia to that of wild-type levels. This result is an improvement over gene transfer experiments in the CFTR-deficient cells with AAV2 vector, which were unable to recover chloride current in these cells.

Brown and colleagues recently described a novel strategy to prevent off target expression of a LV delivered transgene in antigen presenting cells.¹⁹ A hematopoietic-specific microRNA, miR-142-3p, was incorporated downstream of the transgene, in this case, clotting factor IX. The result was long lasting transgene expression, no detectable factor IX antibodies, and phenotypic correction in a mouse model of hemophilia B. This study exemplifies the use of microRNAs to regulate vector tropism (reviewed in ²⁰) and may provide a valuable tool for viral vectors to evade the immune system and allow repeated administration to the lung.

In summary, repeated administration of viral vectors may be necessary for the stable expression of a therapeutic protein over the life of the affected individual should expression wane over time. Vector re-administration to the respiratory tract offers the possibility to increase the overall transduction efficiency. In addition, sequential vector administration to individual lung lobes may be advantageous for practical and safety reasons. However, before repeated administration of viral vector in feasible in a therapeutic setting, obstacles for each vector system must be addressed. Innate and adaptive immune responses to viral vectors will vary depending on the origin of the vector, the purity of the vector preparation, the envelope or capsid pseudotype, the nucleic acid content, the dose delivered, the route of delivery, and the number of administrations. Research in inbred mice has important implications for the translation of gene transfer technology into the development of human pulmonary disease therapies. It will be important in future studies to investigate these approaches in larger, more genetically diverse animal models.

Progress

• Advances in understanding of innate and adaptive immune responses to viral vectors are informing the design and application of vectors for single and multiple dose regimens.

• Pre-existing immunity may prevent the success of a single administration of a viral vector.

• Strategies to allow adenoviral vector re-administration have been assessed.

• Adeno-associated virus vectors can be re-administered successfully through use of novel serotypes, chemical formulations, or immunosuppression.

• Re-administration studies using lentiviral vectors are beginning to inform the field of their potential for repeated delivery.

Prospects

• Genome shuffling and directed evolution of the AAV capsid may help to evade immune responses.

• Mouse re-administration studies will be repeated in larger, more genetically diverse animal models.

• Novel vector formulations or transient immunosuppressive agents may facilitate vector re-administration.

• Incorporating microRNA targets into vectors may provide a novel strategy to prevent transgene expression in antigen presenting cells.