New Directions in Pulmonary Gene Therapy

Abstract

The lung has long been a target for gene therapy, yet efficient delivery and phenotypic disease correction has remained challenging. Although there have been significant advancements in gene therapies of other organs, including the development of several ex vivo therapies, in vivo therapeutics of the lung have been slower to transition to the clinic. Within the past few years, the field has witnessed an explosion in the development of new gene addition and gene editing strategies for the treatment of monogenic disorders. In this review, we will summarize current developments in gene therapy for cystic fibrosis, alpha-1 antitrypsin deficiency, and surfactant protein deficiencies. We will explore the different gene addition and gene editing strategies under investigation and review the challenges of delivery to the lung.

Introduction

Gene therapy holds promise for many disorders that may currently lack practical or effective treatments. Over the past decade, ex vivo gene therapy has advanced, with several treatments approved for clinical use.¹ In vivo somatic cell gene therapy for treatment of blindness caused by mutations in the RPE65 gene is approved by Food and Drug Administration (FDA) and clinical trials for hereditary hemophilias are yielding promising results.^1–3 However, several challenges and setbacks have prevented the advancements for treatment of lung disease. Clinical trials for cystic fibrosis (CF) began in the 1990s using gene addition strategies but did not achieve meaningful improvement in lung function. More than 20 years later, with 27 completed clinical trials, incremental yet substantial improvements in vector design, delivery, and efficacy have occurred.⁴

Continued progress in gene addition approaches has rekindled interest in therapies for lung diseases and other organ systems. In addition, the discovery and development of the clustered regularly interspaced short palindromic repeats (CRISPR)-associated system (Cas) for mammalian gene editing has revolutionized the field, providing new strategies to prevent or treat lung diseases. In this study, we review the advances in both gene addition and CRISPR/Cas-based gene editing approaches to effectively target the lung and treat CF and other genetic pulmonary diseases. We also look at the unique challenges to lung delivery and relevant exciting progress in pulmonary gene therapy.

Gene Therapy Strategies

Gene Addition

Initial approaches included the addition of a functional gene to affected cells to offset disease phenotypes.⁴ This is an excellent approach for recessive monogenic diseases resulting in loss-of-function phenotypes, such as CF, but would be ineffective for dominant negative, or gain-of-function diseases. Gene addition strategies commonly use constitutive promoters to express the gene of interest, but conditional or endogenous promoters are also used. Expression of a gene of interest can be transient or persistent depending on the therapeutic strategy, using genome integrating technologies or episomal maintenance of vector DNA. Examples of gene addition strategies will be discussed in more detail in later sections of this review.

CRISPR/Cas9 Gene Editing

Because of the recent explosion in gene editing technologies, we will briefly review the available tools. CRISPR arrays were first discovered in 1987 in Escherichia coli K12 and have since been observed in 90% of archaea and 50% of bacteria.^5,6 CRISPR/Cas systems act as a prokaryotic “adaptive immune system” to protect against infecting bacteriophages and other sources of exogenous DNA.^5,7,8 Although CRISPR/Cas systems are currently classified into six types (I–VI), each with its unique set of Cas nucleases, much of mammalian gene editing has been performed using Cas9 (type II) derived from Streptococcus pyogenes (spCas9).⁵

spCas9 is a multifunctional, multidomain protein comprising 1,368 amino acids.⁵ During the gene-editing process, Cas9 and guide RNA (gRNA), consisting of a target DNA-specific sequence (crRNA) complementary to the chromosomal sequence and a transactivating CRISPR RNA (tracrRNA) sequence that interacts with the Cas9, are introduced into the cell.⁵ The gRNA recognizes and binds to the target DNA near a protospacer adjacent motif (PAM) site. The PAM interaction (PI) domain of Cas9 recognizes and interacts with the PAM site.^5,9 Cas9 cleaves DNA 3 base pairs (bp) upstream of the PAM site by two distinct nuclease domains, an HNH-like endonuclease domain that cleaves the strand complementary to the guide sequence (target strand) and an RuvC-like domain that cleaves the nontarget DNA strand, causing a double-stranded break (DSB) in the genome.^5,9,10 This DSB triggers the cell's DNA repair mechanism, resulting in either (1) error-prone nonhomologous end joining (NHEJ) and resultant insertions or deletions (indels) that often lead to frameshift or nonsense mutations, or (2) homology-dependent repair (HDR) that requires a template in the form of single-stranded DNA (ssDNA) or double-stranded DNA (dsDNA) carrying the desired modification.⁵ HDR may also use the other host allele as a template, resulting in a nonediting outcome.^11–13 HDR is substantially less efficient than NHEJ, and it only occurs during the G2 and S phases of the replication cycle, whereas NHEJ can occur throughout the cell cycle.^5,12

Cas Variants

Although a majority of CRISPR gene editing studies use spCas9, several other Cas variants also show promising results. A second type II Cas protein, termed CRISPR from Prevotella and Francisella 1 (Cpf1) or Cas12a, was developed for mammalian gene editing.^5,14 Cas12a has four domains: RuvC-I and RuvC-II nuclease domains, a/beta domain, and zinc finger. Unlike Cas9, Cas12a lacks an HNH-nuclease domain, resulting in smaller molecular weight that offers advantages for some delivery approaches.⁵ Instead of creating a blunt-end DSB like spCas9, Cas12a causes staggered cuts, providing advantages for knock-in strategies.⁵ Cas12a recognizes a T-rich PAM site, rather than the G-rich PAM site recognized by Cas9.^5,15 This difference in PAM recognition can allow access to targeting of additional sites in the genome that were difficult to target with Cas9; however, Cas12a has been noted to have a low indel efficiency.¹⁵ Indel efficiency can be increased by modification of gRNA with addition of an 8-meric uridinylate-rich 3′ overhang.¹⁵ Cas12a also has a lower incidence of off-target edits than Cas9,¹⁶ possibly creating less safety concerns for clinical applications. Cas12a ribonucleoprotein (RNP) delivery to the airways by amphiphilic peptides achieved an average editing efficiency of 12% in the large airways and 10% in the small airways of ROSA^mT/mG mice.¹⁷ All cells in these mice contain a tdTomato cassette flanked by loxP sequences in the Rosa26 locus, and deletion of tdTomato by dual Cas12a targeting of these loxP sequences results in localized GFP fluorescence.¹⁷ This study shows the direct application of Cas12a RNPs in targeting the epithelia of the airways, a prominent area for pulmonary lung disease.

Additional protein engineering has improved the efficiency and specificity of editing with Cas12a and Cas9.^18,19 Using phage-assisted continuous evolution, in which evolution is driven by continuous selection of a desired phenotype linked to production of infectious phages containing the evolving gene,²⁰ spCas9 variants were developed to recognize non-NGG and non-G PAM sequences.^21,22 Directed evolution or protein engineering of spCas9 can improve target specificity and allow for targeting of sites previously inaccessible, creating better alternatives to native spCas9 gene editing in some instances. At present, evolved Cas9 variants that recognize non-NGG and non-G PAM sequences, xCas9(3.7) and SpCas9-NG, have only been assessed in cultured cells and in plants, so the in vivo efficacy and, ultimately, clinical applicability of these Cas variants remains to be determined.^23–26

A newer Cas variant developed for mammalian gene editing is derived from Deltaproteobacteria, termed CasX or Cas12e.^5,27 This is among the smallest Cas proteins, with <1,000 amino acids,²⁷ offering advantages for adeno-associated virus (AAV) packaging to provide space for other gene editing tools. Like Cas12a, Cas12e also creates a staggered DSB, but utilizes a longer PAM sequence of TTCN.²⁷ The indel efficiency and off-target incidence is yet to be determined, but its small size and nonpathogenic origin may offer advantages.

Cas variants have also been developed for epigenetic control of targeted genes. Enzymatically deactivated Cas9 (dCas9) that cannot cleave DNA yet retains the ability to bind target sequences is fused to either activator or repressor domains to upregulate or inhibit gene expression, respectively.^5,28–31 Termed CRISPRa when an activator domain is used and CRISPRi when a repressor domain is used, these Cas variants allow for reversible epigenetic control of specific genes. A CRISPRa strategy has been shown to effectively prevent muscle fibrosis and paralysis as well as improve hindlimb paralysis in a muscular dystrophy type 1A (MDC1A) mouse model through the upregulation of the LamaA1 gene.³²

Base Editing

Base editing provides a more efficient method to modify a single nucleotide than HDR.³³ Derived from a catalytically inactive spCas9 to a base editor, the Cas9 machinery binds to target DNA and causes displacement of a small segment of ssDNA into an R-loop structure without strand break creation.³³ Base editors then create single nucleotide changes in the R loop. There are two types of base editors: (1) cytidine base editors (CBEs), which can convert a C-G bp to a T-A bp and (2) adenine base editors (ABEs), which convert an A-T bp to a G-C bp. ABEs have a very high product purity (∼99.9%) and low frequency of indels (typically ≤0.1%) in cultured cells and mice,³⁴ compared with the CBE base pair conversion with 60–98% product purity.^33,35 The lower CBE product purity is thought to be associated with uracil N-glycosylase (UNG) expression, as UNG^−/− cells have increased base editing specificity.^33,35 Base editing has the potential to correct many monogenic disorders, with ABEs offering the potential to correct almost half of all pathogenic human mutations.³⁶ In CF, 15% of known single base pair mutations can potentially be corrected by CBEs and 46% using ABEs (CFTR2.org). One limitation for base editing is the presence of an applicable PAM site at the target site. ABE variants have recently been developed to use alternative PAM sequences, allowing access to genomic regions that were previously inaccessible.³⁷ Delivery of messenger RNA (mRNA) encoding CBE and gRNA successfully introduced base edits in mice and pig oocytes.^38,39 Chemically modified mRNA encoding ABE and gRNA can be used to edit otherwise hard-to-transfect cells.⁴⁰ Split intein AAV vectors offer another delivery system for base editors. In the short time since the development of base editing, this technique has been used to create novel mouse, rabbit, and sheep animal models, and to correct disease-related phenotypes in various animal models, displaying the broad potential for base editing in gene therapy research.^41–45

Prime Editing

Prime editing, a newer application of spCas9, allows modification of more than one nucleotide with greater efficiency than Cas9-HDR strategies. In this iteration of CRISPR/Cas, a modified spCas9 (nickase) creates a single-stranded break.⁴⁶ Fusion to a reverse transcriptase and use of an engineered guide RNA (prime editing guide RNA [pegRNA]) allows the nicked DNA to act as a primer for reverse transcription of the pegRNA template to DNA, resulting in the desired genetic alteration.⁴⁶ This strategy facilitates larger alterations than base editing, but with greater efficiency and less indel creation than conventional CRISPR/Cas9.

Prime editing could significantly expand the scope of gene therapy applications, with the ability to target as much as 89% of disease-causing mutations.⁴⁶ To date, this technology has only been tested in cultured cell lines, and its efficacy in vivo remains to be determined. Prime editors (PEs) are much larger than Cas9 protein, and effective delivery systems will need to be developed and optimized. In addition, the PE complementary DNA (cDNA) exceeds the packaging capacity of recombinant adeno-associated vectors, a common viral delivery vehicle. Finally, pegRNAs, which may be much longer than conventional gRNAs, could influence delivery efficiency when using a nonviral vector. Alternative or modified methods of PE delivery will likely need to be developed for efficient editing to occur in specific cell types and in vivo.

Barriers to Delivery: Unique Challenges of the Lung

The complex architecture of the lung allows the epithelial surfaces to be directly targeted for delivery of therapeutic cargoes. Topical administration through inhalation provides direct access to the epithelial cells lining the airways and airspaces. For vector or cell-based therapies, systemic delivery approaches through the vascular system are inefficient at targeting epithelial cells. Systemic approaches may be better suited for delivery to the vascular endothelial cell compartment. Therefore, a majority of pulmonary gene therapy approaches use aerosol administration. Although the lung allows for organ-specific delivery of topical/aerosol therapeutics, lung architecture and immune responses create formidable barriers for efficient delivery.

A mucus layer present in large and small airways comprises a mesh of gel-forming mucins. Mucins are large polymeric macromolecules containing negatively charged glycans.⁴⁷ Secreted mucins reside within airway surface liquid (ASL). ASL is derived from the secretions of submucosal glands and surface epithelia and is a rich source of secreted host defense peptides and proteins, antibodies, and the collectins surfactant proteins (SPs), SP-A and SP-D. Viral and nonviral vectors may be immobilized and inactivated in this layer through multivalent interactions with mucins, such as electrostatic and hydrophobic interactions and hydrogen bonding.^48,49 Steric obstruction may also contribute to vector immobilization. Vectors trapped in mucus may also be cleared through mucociliary clearance (MCC). Antibodies can move unimpeded within human mucus and ASL, but Fc regions can form low affinity interactions with the mucus.^49–51 Neutralizing antibodies against viral vectors, especially AAV and AdV, are present in airway secretions of some healthy and diseased individuals.^52–54

Antibody binding to viral vectors within ASL can cause their aggregation and clearance, reducing transduction. Repeat administration of AdV or AAV vectors results in the development of neutralizing anti-capsid antibodies, potentially reducing efficacy.^55,56 However, readministration of lentiviral vectors (LVs) in mouse models does not elicit development of systemic or local neutralizing antibodies.^57,58

Recent studies show that immune responses can be elicited by Cas9 variants.⁵⁹ S. pyogenes and Staphylococcus aureus are common human commensals that may also act as pathogens, and it is unsurprising that a subset of the human population may possess preexisting humoral and/or cell-mediated responses. Studies of AAV delivery of Cas9 in mouse models have shown immune responses can be elicited in response to Cas9 protein expression. Administration of AAV-delivered SaCas9 (AAV-CRISPR) to skeletal muscle leads to infiltration of CD4⁺ and CD8⁺ T cells, and the production of anti-Cas9 antibodies.⁶⁰ Systemically delivered AAV-CRISPR can also result in a humoral response in mice,⁶¹ the development of SaCas9-specific memory T cells, resulting in reduced efficacy upon repeat dosing,⁶² and CD8⁺ T cell-mediated killing of cells expressing SaCas9 in vivo.⁶³

Human sera can also contain preexisting antibodies toward Cas9 variants. Immunoglobulin G (IgG) antibodies targeting both SaCas9 and SpCas9 were found in 86% and 73% of donors, respectively, in cord blood from 22 donors.⁶⁴ In peripheral blood from 12 healthy donors, 67% of donors had anti-SaCas9 antibodies and 42% had anti-SpCas9 antibodies.⁶⁴ In a study with a large sampling size of 200 donors, 19% and 4.5% of donors were identified to potentially positive SaCas9 and SpCas9 antibodies, respectively, in a screening assay.⁶⁵ Of these donors, only 5% and 1.5% tested positive for anti-SaCas9 and anti-SpCas9 antibodies upon a more stringent confirmatory assay, suggesting antibodies targeting Cas9 may not be as prevalent in the general population.⁶⁵

Cell-mediated responses toward Cas9 proteins in humans has also been reported. Human peripheral blood mononuclear cells stimulated with SpCas9 or Cas12a lead to activation of T effector and T regulatory cells.⁶⁶ Although SpCas9-reactive T regulatory cells could mitigate the proliferation of effector T cells, enriched SpCas9-reactive effector T cells were shown to specifically target and lyse lymphoblastoid B cell lines that express SpCas9,⁶⁶ suggesting that the adaptive immune response may counteract therapeutic efforts of CRISPR. Further research into human immune response to SpCas9 and SaCas9 will be required to determine the effectiveness of these proteins in a clinical setting.

Efforts to suppress the host immune response to Cas9 proteins include transient immunosuppression and modification of Cas9. Immunodominant T cell epitopes of SpCas9 for HLA-A*02:01, the most common HLA type in European and North American Caucasians, have been identified, and mutation of MHC-binding anchor residues of these epitopes results in reduced T cell activation.⁶⁷ These modifications do not affect SpCas9 function or specificity, suggesting that such an approach could be used to suppress the host immune response to CRISPR-mediated gene therapy.⁶⁷ HLA allotype diversity may complicate the creation of a completely immunosilent Cas9 protein and may require allotype-specific modifications for more development of more effective treatments.

Pulmonary surfactant is present in alveolar lining liquid and comprises phospholipids, cholesterol, and SPs such as SP-A, B, C, and D. SP-A and SP-D play roles in host defense, as mice deficient in these proteins are more susceptible to bacterial and viral infections. SP-A and SP-D can decrease the transduction efficiency of nonviral vectors, causing aggregation and clearance.^68–70

Resident macrophages in the airways and alveoli also play critical roles in host defense and tissue remodeling. Particles in the size range of 250 nm to 3 μm are readily phagocytosed; however, particles smaller than 250 nm are taken up less efficiently.^49,71 As most viral and nonviral gene delivery vectors are below the 250 nm threshold, single particles are not easily targeted by macrophages, but vector aggregation may facilitate phagocytosis. Aggregation and engulfment of 20–110 nm silver nanoparticles was observed in macrophages isolated from bronchoalveolar lavage fluid in rats administered nanoparticles through aerosolization.⁷² Opsonization of vectors by SP-A and SP-D may also increase clearance by macrophages.⁴⁹

Vectors must also transverse the physical barriers of the differentiated epithelium. Tight junctions between adjacent cells prevent access to the basolateral surface, a site of many virus receptors.⁴⁹ To achieve durable gene editing of the large airways, progenitor basal cells will likely need to be targeted. Basal cells are a population of self-renewable multipotent cells responsible for the regeneration and homeostasis of the conducting airways. Their apical cytoplasm does not reach the lumen, limiting their transduction.⁷³ In addition, it has been observed that endocytosis from the apical cell surface of differentiated cells occurs at low efficiency,⁷⁴ which may prevent transduction by vectors that rely on this cell entry pathway.

Disease-related factors may also create barriers to epithelial cell transduction. In CF, loss of CFTR-dependent anion secretion results in reduced liquid secretion and an acidic ASL pH. This impairs antimicrobial ASL activity and reduces MCC.^75,76 Impaired MCC and reduced host defenses create a permissive environment for bacterial infections that leads to inflammation, airway obstruction, and airway remodeling. CF respiratory secretions contain elevated levels of endogenous DNA, actin filaments, and cellular debris owing to this inflammatory environment, and this may make secretions more viscous.⁷⁷ Bacterial biofilms can also form in the CF lung, adding an additional external barrier to epithelial cell access. Elevated oxidative stress can increase disulfide bond cross-links between mucin fibers, leading to smaller pore size of the mucus gel and more elastic mucus.^78,79 Studies have shown that CF mucus can impede both viral and nonviral vector entry to primary human airway epithelial (HAE) cultures.^49,80–84

Methods of Delivery: Viral and Nonviral Approaches

The delivery of cargoes for gene addition and gene editing components may be accomplished using the following formats: (1) delivery of DNA expression cassettes or mRNA encoding the required components, and (2) delivery of protein and DNA or RNA (in the context of CRISPR/Cas system: Cas protein, gRNA, and donor DNA, if required). The vector oftentimes dictates the type of cargo that is used. The delivery vehicle may be viral or nonviral.

Viral Delivery

Viral vectors are commonly used for pulmonary applications. Several viral vector systems are amenable to administration through inhalation and result in robust transduction of a variety of cell types in the lung.

Adeno-associated virus

AAV vectors have been used extensively in airway gene delivery in vivo and their safety record for human clinical trials is strong.^73,85–91 AAV transduction provokes little innate or adaptive immune responses upon the first treatment and shows minimal signs of cytotoxicity.⁴ The variety of AAV serotypes allows for transduction of different airway epithelial cell types. Clinical trials for AAV delivery for CF used the AAV2 serotype with an excellent safety prolife, but transduction efficiency of the airways was low.^55,92–94 In retrospect, one reason for this limited delivery efficiency could have been receptor availability. The AAV receptor (AAVR), encoded by gene KIAA0319L, is required for infection of AAV serotypes 1, 2, 3b, 5, 6, 8, and 9, and is exclusively localized to the basolateral and perinuclear compartments of HAE cells.⁹⁵ AAV2.5T, developed by directed evolution to transduce primary HAE cultures apically, resulted in 100-fold increase of apical transduction compared with the native AAV serotype 5.⁹⁶ AAV2.5T uses sialic acid as an attachment factor, and this interaction is required for internalization. Internalization of AAV2.5T is independent of the AAVR, as AAVR antibodies do not affect AAV2.5T transduction.^96,97

DNA shuffling of capsid subunits of various AAV serotypes can create recombinant AAV (rAAV) vectors with improved targeting and transduction efficiencies. This technique was used to select for AAV capsids that could efficiently transduce hepatocytes in the presence of pooled human antisera, resulting in the identification of an AAV2/AAV8/AAV9 chimera termed AAV-DJ.⁹⁸ DNA shuffling was used to identify AAV capsids with high tropism for lung endothelial cells and may be a powerful tool for discovery of novel AAV capsids for other lung compartments, such as airway epithelia or alveolar cells.⁹⁹ Display of peptide libraries in the AAV capsid can also be used in combination with capsid shuffling. Random AAV vector peptide display and capsid shuffling libraries have been used to screen capsids for cellular tropism both in vitro and in vivo.^98,100–103 DNA barcoding can be used in AAV capsid libraries to allow for tracking of capsids in a high-throughput setting.^100,101

The AAV capsid is an icosahedron consisting of capsid proteins VP1, VP2, and VP3 at a ratio ∼1:1:18.^104,105 Surface-exposed protrusions at the threefold axis of symmetry harbor five of the nine “variable loop” regions (VRs), which vary in amino acid composition between serotypes.^104,105 Differences between serotypes can primarily be mapped to this surface topology.^104,105 VR composition can affect tissue tropism and some VRs, such as VR-VIII of VP3, can withstand insertion of foreign sequences to alter tropism.^104,105 Selection of cell-type or tissue-specific targeting AAV capsids can be achieved by using peptide or designed ankyrin repeat protein (DARPin) insertion into the AAV capsid at these regions. Deverman et al. used a Cre recombination-dependent approach to select for capsids able to breach the blood–brain barrier.¹⁰⁶ Capsid libraries with a 7-mer peptide insertion were selected through administration to an astrocyte-specific Cre-expressing mouse model, resulting in isolation of capsid variants with a 40-fold increase in central nervous system transduction.¹⁰⁶ In a separate study, DARPins specific for Her2, CD4, and EpCAM were inserted into the AAV capsid sequence, resulting in specific targeting of tumors in a mouse model, CD4-positive cells in the spleens of humanized mice, and EpCAM-positive cells in whole human blood, respectively.¹⁰⁷

The airway mucus may impede efficient transduction of some AAV serotypes. AAV1, AAV2, and AAV5 can adhere to human airway mucus.^80,81 AAV6 transduction, however, is not reduced by endogenous mucus or mucus hypersecretion, whereas AAV1 transduction is reduced in primary HAE cultures or in a mouse model. A single amino acid change in the AAV6 capsid, K531E, increases susceptibility to mucus impairment, indicating glycan-specific interactions of serotypes.¹⁰⁸

Although AAV vectors can achieve high transduction efficiency in primary human airway epithelia, the packaging capacity is limited to 4.8 kb.¹⁰⁹ This may be a limitation for some gene addition and CRISPR/Cas approaches. The cystic fibrosis transmembrane conductance regulator (CFTR) cDNA sequence is ∼4.5 kb, which is too large to be packaged within an AAV expression cassette. Creation of a CFTR mini-gene through deletion of a portion CFTR regulatory (R) domain (CFTRΔR) and use of a minimal 83 bp synthetic promoter yielded an expression cassette that was effectively packaged into an AAV vector.^110,111 For gene editing, the AAV packaging size is too small to accommodate the full CRISPR/Cas cargo for spCas9. This obstacle can be overcome using a smaller Cas variant such as saCas9 or dual vectors, with one AAV containing the genetic sequence of Cas protein, and a second AAV encoding the gRNA and a marker gene or HDR template. A split Cas system, such as split-intein, can also be utilized, in which the N-terminal region of Cas9 is encoded in one AAV and the C-terminal region of Cas9 is encoded in a second AAV vector.^60,112,113 This strategy is also effective for delivery of base editors.^112,114,115 Dual AAV vectors were used to correct CFTR function in a CF pig model. One AAV carried an integrating CFTR expression cassette flanked by the piggyBac transposon and another AAV carried the piggyBac transposase, both administered through tracheal aerosolization.¹¹⁶ Although the transduction efficiency was not determined in this study, delivery was sufficient to correct the anion channel defect.¹¹⁶ Transduction efficiency of two AAV6 vectors has been shown to occur at similar levels as transduction of a single AAV6 vector in mice lungs, which is approached in 10% of airway epithelial cells.¹¹⁷ These studies highlight the feasibility of dual AAV vector strategies to effectively target the airways.

Although AAV vectors are very effective at transducing a variety of cell types in the lung, its small carrying capacity creates challenges for delivery of donor DNA for HDR. This application generally requires two vectors. AAV transduction can result in sustained expression of CRISPR components, as packaged DNA can persist episomally for the lifetime of the cell, or with modification, can rarely integrate into the host genome.

Another strategy to overcome genome size limitations is to create a parvovirus chimera in which a rAAV genome is packaged within the capsid of a different parvovirus with a larger packaging capacity. An early demonstration of this technique packaged rAAV genomes into a B19 capsid, resulting in effective targeting of erythroid cells.¹¹⁸ rAAV genomes have also been packaged into human bocavirus (HBoV) and gorilla bocavirus (GBoV) capsids to increase the packaging capacity by 1 kb.^119–121 These rAAV/HBoV or GBoV chimeras effectively transduced primary human airway epithelia cultures, lung organoids, and ferret lungs.^119–121 Although rAAV/HBOV chimeras using serotype HBoV1 can be effectively neutralized by human IgG, rAAV/HBoV4 and rAAV/GBoV were unaffected by such treatment, suggesting the potential use of these vectors in lung delivery.

A major limitation of parvovirus cross-genera pseudopackaging is production of high titer vector; however, recent progress in understanding HBoV1 viral replication has resulted in better production methods. In the prototypic production of rAAV/HBoV1 vector, transient transfection of HEK293 cells was performed with plasmids containing the (1) rAAV2 genome, (2) necessary adenovirus (Ad) helper genes (E2, E4Orf6, and VA RNA), (3) AAV2 Rep genes, and (4) HBoV1 capsids (VP1, VP2, and VP3) and nonstructural (NS) genes (NS1, NS2, NS3, NS4, and NP1).¹²¹ Typical yields were 10- to 20-fold less than that of rAAV2 vector.¹²¹

It was later discovered that NP1 is the only NS protein required for HBoV1 capsid expression and that it enables transcription of full-length cap mRNA that would otherwise be terminated at a polyadenylation (pA) site within the unique region of VP1.¹²² Silent mutation of the HBoV1 cap cDNA to omit pA sites allowed for the production of rAAV/HBov1 vector in the absence of any NS proteins, further enhancing the production yield.¹²³

Another production method of rAAV2/HBoV1 vector uses Sf9 insect cells and baculovirus expression vectors (BEVs).¹²⁴ Sf9 cells are infected with two BEVs: one harboring the rAAV2 genome, and the other carrying expression systems for AAV2 Rep78 and Rep52, as well as HBoV1 capsid proteins VP1, VP2, and VP3.¹²⁴ Addition of a third BEV that expresses HBoV1 NS protein NP1 enhances vector production yields similar to rAAV2 produced in Sf9 cells, although the proximal pA preventing HBoV1 cap expression in the absence of NP1 in mammalian cells do not appear to be effective in insect cells.¹²⁴ Overall, production of rAAV2/HBoV1 in Sf9 cells resulted in 10- to 100-fold more vector than rAAV2/HBoV1 produced in HEK293 cells using a transient transfection method of production.¹²⁴ Although the biological characteristics, such as transduction ability, of Sf9-produced rAAV are equivalent to HEK293-produced rAAV, rAAV2/HBoV1 produced in Sf9 cells had a five- to seven-fold reduced transduction of cultured HAE cells compared with rAAV2/HBoV1 produced in HEK293 cells.^124–128 This decreased transduction could be because of reduced genome packaging, as rAAV/HBoV1 produced in Sf9 cells had a high level of empty capsids (50–60% full particles) compared with vector produced in HEK293 cells (>95% full particles).¹²⁴ Variations in glycosylation of HBoV1 capsid proteins in the two cell lines could also affect transduction. Through further optimization of rAAV2/HBoV1 production methods, use of this vector in animal models and in clinical trials may become more feasible.

Analysis of variant HBoV1 isolates from human samples has identified various hot spots that can affect vector production yields and transduction of primary HAE cultures.¹²⁹ One capsid variant (T590S) increases AAV2/HBoV1 vector production without affecting transduction efficiency.¹²⁹ This amino acid variation is located in the HI loop of VP3, a hypervariable region that is linked to particle assembly and genome packaging in AAV.¹³⁰ Another isolated variant (Y523F) increased transduction of primary HAE cultures.¹²⁹ These capsid variants highlight the promising application of HBoV1 capsid modification to further enhance rAAV2/HBoV1 vectors for clinical applications.

Lentivirus

LVs are widely used, and several are FDA approved for use in clinical trials to treat diseases such as SCID, chronic granulomatous disease, adrenoleukodystrophy, and beta-thalassemia. LV therapies for these diseases largely involve ex vivo treatment of hematopoietic stem cells that are subsequently returned to the patient. Even with the success of ex vivo LV therapies, delivery to the lung remains challenging owing to the barriers previously discussed.

LVs are widely used in gene addition approaches to integrate expression cassettes into the host genome. They can also be used to deliver coding sequences for Cas proteins, gRNAs, and other machinery in a single vector. Multiple gRNAs can be encoded into a single lentivirus vector, each under its own cell type-specific promoter, to edit multiple genes in multiple cell types with one transduction.¹³¹ LVs can be pseudotyped with a variety of viral envelopes to alter tropism. They are commonly pseudotyped with the vesicular stomatitis virus (VSV) envelope, VSV-G, which efficiently transduces the basolateral surface of airway epithelia but has low levels of apical transduction because of receptor availability.^132–135 To enhance VSV-G pseudotyped LV transduction in vitro or in vivo, transient disruption of tight junctions through calcium chelators,¹³⁶ induced injury,¹³⁷ or treatment with mild detergents,¹³⁸ such as lysophosphatidylcholine (LPC) may be utilized. Multiple screens of different envelope pseudotypes for efficient apical transduction and virus production have been performed, with the envelope from Autographica californica multiple nucleopolyhedrovirus (AcMNPV) GP64 being suitable.^139–144 GP64-pseudotyped LVs can efficiently transduce the nasal passages and airways of mice and pigs.^132,141 Pseudotyping an SIV-based LV with modified Sendai virus (SeV) envelope hemagglutinin-neuraminidase (HN) and fusion (F) proteins also achieved robust apical transduction.^145,146 F/HN-SIV transduces the nasal epithelium of mice with 4.5% efficiency and apically transduces fully differentiated primary HAE cultures to generate functional CFTR chloride transmembrane conductance.¹⁴⁵

Whereas integration and persistent expression is advantageous for some gene addition and CRISPRa or CRISPRi approaches, in the case of gene editing, this may be undesirable. Anti-CRISPR proteins (Acrs), natural bona fide CRISPR-Cas antagonists, have been identified that inhibit Cas activity by preventing DNA binding, cleavage, crRNA loading, or complex formation.^147–149 Acrs can effectively inhibit Cas9 activity when delivered ex vivo to hematopoietic stem cells, decreasing Cas-associated cell toxicity, and improving engraftment in a mouse model.¹⁵⁰ Delivery of AcrIIC3 can inhibit Nme2Cas9 editing in mice, resulting in inhibition of editing in the heart and liver.¹⁵¹ Integration defective LVs with mutations in integrase may offer uses in this setting but have not been extensively studied to conclude whether integration is 100% deficient or only impaired.¹⁵² Small molecule inhibitors, doxycycline-dependent inactivation, truncated gRNAs, and synthetic oligonucleotides to silence Cas9 activity^153–157 have been developed to limit Cas9 persistence, and such approaches may be needed for editing applications.

Adenovirus

Adenovirus (Ad) vectors are used extensively as the many known serotypes can transduce a variety of cell types. Ad-CFTR vectors of serotypes 2 and 5 transduce primary HAE and were used to successfully transduce nasal turbinate epithelia in vivo, although transiently.⁴ Serotypes Ad-2 and Ad-5 utilize the coxsackievirus and adenovirus receptor (CAR) as a primary receptor, which is localized to the basolateral surface of human airway epithelia.¹⁵⁸ A chimeric vector Ad-5 carrying a heterologous fiber molecule from serotype 35 (Ad5F35) can transduce the apical surface of human airway epithelia and correct anion conductance in cells carrying the CFTR F508del mutation.¹⁵⁹ There are safety concerns regarding clinical use of Ad vectors, as several clinical trials for CF observed inflammatory responses. A nonhuman primate study also reported alveolar inflammation with high-dose administration of an Ad vector.^160,161 High systemic doses of Ad can elicit severe immune responses, as observed in the clinical trial in 1999 for ornithine transcarbamylase (OTC) deficiency, in which administration of 10¹³ viral particles administered through the femoral artery resulted in the death of a patient.^162,163 Owing to these safety concerns, Ad vectors are mostly used in the research setting, or are modified to decrease harmful immune responses. Studies of the innate immune responses to Ad vectors may achieve modifications that improve their utility.^164,165

Because of a relatively high seroprevalence of Ad-5 in the general population, alternative Ad vectors are being explored. Gorilla adenovirus isolates (GAd) can be converted to replication-deficient vectors that produce high-titer yields. GAd vectors have been used for vaccines against respiratory syncytial virus and Zika virus in preclinical studies to produce high-level antigen-specific antibodies.^166,167 GAd type 9, isolate 46 shows high lung endothelial cell tropism in a mouse model.¹⁶⁸ Like Ad-5, this GAd vector can be modified at the fiber knob to alter tissue targeting.¹⁶⁸ GAd vectors have a low seroprevalence and can undergo capsid engineering, possibly providing an alternative vector for gene delivery.

Helper-dependent adenovirus

Helper-dependent adenovirus (Hd-Ad) is attractive for delivery of large cargoes and CRISPR/Cas strategies that require the use of HDR donors. Hd-Ad has a large packaging capacity compared with AAV, with the ability to encode ∼36 kb.¹⁶⁹ This large packaging capacity can accommodate the genetic sequence of the Cas protein, gRNA, and donor DNA within a single vector. In Hd-Ad, all viral coding sequences are deleted to minimize viral protein production and produce a replication incompetent vector.¹⁶⁹ Hd-Ad vectors are less immunogenic than adenovirus vectors. A wide variety of serotypes may be used, but serotype 5 effectively transduces airway basal cells in vivo in mouse and pig models, as well as basal cells in vitro in primary HAE cultures.^4,170,171 Although it has been reported that Hd-Ad DNA can persist as episomal DNA, one study reported transient expression of Cas9 in a vector that encoded Cas9 upstream of donor DNA, with complete loss of Cas9 expression after 7 days.¹⁷² Although this phenomenon could be owing to dilution of the Hd-Ad DNA on serial passage, it is thought that the vector DNA is degraded upon donor DNA integration as a result of compromised vector DNA. Hd-Ad vectors may offer an alternative to transiently transduce cells for CRISPR/Cas-mediated gene therapy.

Nonviral Delivery

Nonviral vectors have been less widely used for lung delivery, owing to low transduction efficiency and inflammatory responses. Nonviral vectors continue to be pursued for delivery for the following potential advantages: (1) there is virtually no packaging limit, (2) many nonviral vectors are biodegradable and biocompatible, offering less cytotoxicity and adverse immune responses, (3) scale-up of production is relatively straightforward and economical, and (4) the transient expression of cargo may reduce off-target effects caused by constitutive expression of Cas proteins, host genome integration, and genotoxicity.

There is a resurgence of interest in nonviral delivery for pulmonary applications. The vectors are mainly composed of three configurations: (1) peptide-based, (2) lipid-based, and (3) polymer-based delivery. Other formulations of nonviral delivery systems, such as exosomal delivery, are also being explored. Although further discussion of this topic is outside the scope of this review, we refer the reader to the publications on inhaled nonviral delivery of CRISPR/Cas9 cargo¹⁷³ and nonviral strategies for treatment of CF.¹⁷⁴

Gene Therapy for Lung Disease

Cystic Fibrosis

CF is an autosomal recessive disease resulting from mutations in the CFTR gene.¹⁷⁵ CFTR encodes an anion channel that conducts chloride and bicarbonate across epithelial cells in the pancreas, liver, intestines, gall bladder, sweat gland, airways, and genital ducts.¹⁷⁶ Mutations disrupting CFTR function alter ASL volume and composition, impair host defenses, and decrease mucociliary transport, leading to bacterial infections, airway remodeling, and reduced lung function.^175,176 CF affects >70,000 individuals worldwide and is the most common life-threatening inherited disease among Caucasians.

To date, >2,000 variants have been identified in the CFTR gene, with at least 346 known to be disease causing.¹⁷⁶ Mutations are classified according to their impact on CFTR transcription and translation as follows: Class I mutations are nonsense or frameshift mutations that result in truncated or unstable mRNA that preclude the production of any functional protein. Class II mutations, including the most common allele F508del, result in protein misfolding or reduced trafficking to the plasma membrane. Class III mutations are characterized by dysfunctional channel gating. Class IV mutations produce CFTR protein with reduced conductance. Class V mutations are located within introns and cause aberrant mRNA splicing. Finally, class VI mutations produce functional, but relatively unstable CFTR protein at the cell surface. Of importance, a mutation can impair CFTR function in more than one way.

Since 2011, several small molecule CFTR correctors, which facilitate protein folding, and potentiators, which act to keep the CFTR channel open, have been developed as FDA-approved CF therapeutics.^177–180 To date, three correctors (lumacaftor, tezacaftor, and elexacaftor) have been approved for treatment of class II mutations. Tezacaftor and elexacaftor can be used in combination, as each small molecule targets its own specific site, to synergistically enhance protein folding and function.¹⁸⁰ Ivacaftor, at present the only FDA-approved potentiator, targets class III mutations, such as G551D, to increase channel conductance, and has been shown to be effective for some class IV and V mutations as well.¹⁸⁰ A triple therapy of tezacaftor, elexacaftor, and ivacaftor, termed Trikafta is currently approved for individuals 12 years and older.¹⁸¹ This treatment could benefit ∼90% of all individuals with CF within this age range, significantly advancing CF therapeutics.^180,181

Small molecule therapies are significant advancements but must be taken for the lifetime of the individual. Certain side effects, such as abnormal liver function, may also make these treatment options unsuitable for some patients.¹⁸⁰ Finally, small molecule therapies have little to no benefit to the ∼10% of individuals with more rare mutations. Gene therapy strategies have the potential to address these concerns and provide a mutation agnostic cure. Previous studies suggest that achieving 6–50% wild-type CFTR function can restore chloride transport in airway epithelia and could provide significant clinical benefit.^182–187 Several approaches, from gene addition, gene correction, and RNA correction, are currently being studied to restore CFTR function.

Gene addition

The first CF gene therapy studies involved the insertion of an additional copy of the CFTR gene, either integrated in the host genome or as an episome. Several viral and nonviral delivery strategies have been applied, all with varying degrees of persistence. Retroviral and LVs were used to deliver and integrate CFTR expression cassettes, resulting in restoration of CFTR function in HAE, mouse, pig, and ferret models.^4,188–191 AAV vectors, which persist episomally, are an effective means to deliver and restore CFTR function in vitro and in vivo in mouse, rabbit, and macaque models.^192–194 However, previous clinical trials utilizing AAV2 vectors did not achieve clinical significance.^{4,55,90,92–94,195} Inefficient gene transfer because of poor airway transduction by the AAV2 serotype, insufficient CFTR expression owing to use of the ITR promoter in the vector construct, and generation of therapy-inactivating host immune responses may have contributed to the disappointing outcomes of these previous clinical trials. Mild stabilization of lung disease was observed in the most recent nonviral delivery phase 2b clinical trial, in which repeated CFTR cDNA (pGM169) complexed with liposome complex GL67A was administered through nebulization.¹⁹⁶ Although the improvements in lung function were modest in the pGM169/GL67A-treated group, the investigators concluded that repeated administration of this nonviral therapy is safe and can change clinically relevant parameters of lung function, such as FEV₁.¹⁹⁶

Liposomes have been extensively studied for inhalable delivery of drugs, vitamins, and nucleic acids. The “gold standard” nonviral method of gene delivery to the lung is the cationic liposome GL67A. GL67A complexed with plasmid DNA (pDNA) carrying a CFTR expression cassette has shown encouraging results for pulmonary delivery in mice, sheep, and humans. The first clinical trial using GL67A as a vector delivery system resulted in ∼25% correction of the CF ion defect, although these results were only observed for 2 weeks. In addition, the treatment caused mild “flu-like” symptoms, most likely because of Toll-like receptor-9 (TLR-9) recognition of CpG content in the pDNA. Use of a third-generation, CpG-free plasmid maintained efficient delivery to mice lungs and reduced inflammatory responses to levels nearly indistinguishable to the vehicle control. In a second, phase 2b clinical trial, repeated administration of GL67A/pDNA complexes resulted in stabilization of lung function.

Hd-Ad systems effectively target the human airway epithelium and the airways of several animal models, providing a potential strategy for future clinical trials.^{4,170,197,198} Transposase-based methods, such as the piggyBac system, which can also be delivered through Ad or AAV vectors, achieved persistent CFTR expression in mice and phenotypic correction in a CF pig model.^4,116,199

rAAV/HBoV1 chimeras, in which an rAAV genome containing the full-length CFTR cDNA coding sequence driven by a CMV-β-actin promoter is packaged into a HBoV1 capsid, effectively transduced primary human polarized airway epithelial cultures from the apical surface and partially corrected CFTR-mediated chloride transport.¹²¹ These chimeras may offer an alternative vector delivery system if there are no blocking antibody barriers. AAV/HBoV1 vectors can also effectively transduce ferret lungs, offering an animal model in which this CFTR-carrying chimeric vector can be tested for phenotypic correction.¹²⁰

CRISPR/Cas9-mediated gene addition was used to insert a CFTR expression cassette into safe harbor loci in human and porcine cell lines.^172,200 In these studies, “all-in-one” HD-Ad vectors carrying expression cassettes for Cas9, a gRNA targeting either the AAVS1 or GGTA1 safe harbor locus in humans and pigs, respectively, and CFTR donor DNA with corresponding homology arms successfully integrated and expressed CFTR in a site-specific manner.^171,172,200 Transduced porcine IPEC-J2 cells exhibited an integration efficiency of 10%.²⁰⁰ Although efficiency was not calculated, transduced IB3–1/CF cells produced measurable CFTR chloride conductance with transient Cas9 expression, most likely owing to instability of vector DNA upon HDR of donor DNA.²⁰¹ This CRISPR/Cas gene addition strategy provides high levels of integration in cell lines without persistent expression of the Cas9 protein. Further studies in primary HAE cells and in vivo models are required to determine the feasibility of this approach in humans.

Coupling CFTR expression to another gene through the “GeneRide” approach is another strategy with therapeutic potential. This approach involves insertion of a promoterless therapeutic gene cDNA in-frame upstream of the stop codon of an endogenous gene, separating the endogenous gene and the therapeutic gene with a 2A ribosomal skipping site.²⁰² This results in mRNA production containing both endogenous and therapeutic mRNAs but two separate proteins. This technique allowed insertion of the human coagulation factor IX (hF9) into the hepatocyte-expressed albumin locus, ameliorating a hemophilia B phenotype in a mouse model.²⁰² The same albumin locus also used to insert a promoterless human uridine glucuronosyl transferase A1 (UGT1A1) cDNA, rescuing a stillborn phenotype in a Crigler–Najjar syndrome type I (CNSI) mouse model, although this did not normalize plasma bilirubin levels.²⁰³ Combining GeneRide with CRISPR/Cas9-mediated gene addition through a two AAV vector design: one carrying cDNA of saCas9 and gRNA targeting the albumin locus and the second vector carrying UGT1A1 cDNA flanked by albumin homology regions, resulted in a 26-fold increase in targeting efficiency, rescue of the CNSI mouse model stillbirth phenotype, and produced bilirubin levels comparable with wild-type littermates.²⁰⁴ The GeneRide approach shows potential for clinical applications for a variety of genetic disorders. Future inquiry into potential targeted loci, targeting efficiency, and off-targeting are needed to determine the therapeutic value of this approach for treatment of pulmonary diseases.

Gene editing

Cell penetrating peptides (CPPs), 5–30 amino acids in length that are cationic, amphipathic, or nonpolar can translocate the plasma membrane and facilitate cellular uptake.²⁰⁵ There are many CPPs that can transport different cargoes to a variety of cell types. Combinations of multiple CPPs may also be used to enhance delivery efficiency.²⁰⁵ In the context of CRISPR/Cas cargo delivery, CPPs can be linked to RNPs comprising Cas protein and gRNAs to mediate cellular uptake.²⁰⁵ Delivery efficiency can vary widely based on cargo type and the targeted cell type, and extensive optimization of CPPs is required. CPPs also tend to become trapped in endosomal compartments upon internalization.^206,207 Addition of an endosomal leakage domain (ELD) can be used to destabilize the endosomal membrane, allowing escape of Cas RNPs into the cytosol.^208,209 CPP-ELDs carrying Cas9 RNPs or Cas12a RNPs were shown to efficiently transduce primary HAE cultures, as well as the large and small airways of mice.¹⁷ These CPPs offer a promising delivery system for protein-based gene editing approaches for CF treatment. Recent evidence suggests that this Cas protein can effectively target primary human airway epithelia and murine lungs in vivo.¹⁷

Zinc finger nucleases (ZFN) are engineered nucleases that facilitate genome editing by binding a targeted DNA sequence through a DNA-binding domain and introducing a DSB through a DNA-cleaving domain. Bednarski et al. uses ZFNs targeting the 5′ end of exon 11 of the CFTR gene and introduced a donor construct containing exons 11–27 of CFTR to correct the F508del mutation in the CFBE41o− cell line, rescuing expression of CFTR mRNA and chloride conductance.²¹⁰ Although this approach is yet to be tested in vivo, the design of a super-exon could allow correction of all CFTR mutation downstream of exon 11.

With CRISPR/Cas technology, correction of disease-causing mutations has also been explored. Studies relying on HDR with donor CFTR DNA have corrected the F508del in human induced pluripotent stem cells (hiPSCs), intestinal organoids, and upper airway basal stem cell (UABC) organoids.^{171,211–213} Off-target activity as predicted in these studies were relatively low, occurring in 0–4% of alleles sequenced.^211–213 Correction efficiency in these studies, however, has been relatively low or has utilized selection of positive cells or colonies through expression of resistance markers. Gene editing of the CFPAC-1 homozygous F508del human cell line using a viral vector-, drug selection-, and reporter enrichment-free (VDR-free) approach resulted in just 1% of alleles corrected.²¹⁴ In vivo mouse studies have demonstrated 0.1% HDR-mediated CRISPR-Cas gene editing in the murine lung,²¹⁵ well below the threshold of CFTR phenotypic restoration, as well as an indel efficiency >1%.²¹⁵

An alternative strategy is transplantation of gene-corrected cells. Gene-corrected UABCs can be embedded onto pSIS membrane, FDA-approved porcine membranes used for sinonasal repair.²¹¹ This technology could be used to implant gene-corrected basal stem cells in the upper airways, areas that can become reservoirs for antibiotic-resistant bacteria.²¹¹ Although outside the scope of this review, ex vivo cell-based gene therapies are another approach under investigation for the treatment of lung diseases. Regardless of the approach, CRISPR/Cas-based HDR gene correction of CFTR mutations will need further progress to overcome the current low correction efficiency and the relatively high indel frequency.

Some mutations result in the creation of alternative splice sites within an intron of CFTR, resulting in creation of a pseudoexon or addition of sequence to a current exon.²¹⁶ These mutations are amenable to CRISPR/Cas-mediated NHEJ, by excision or disruption of the disease-causing mutation within the intron. Such an approach has been applied to 293T cell lines that contain either the c.1679 + 1634A>G, c.3140–26A>G, or C3718–2477C>T CFTR mutations.²¹⁷ Mutations c.1679 + 1634A>G and c.3140–26A>G create a splice acceptor site and disrupt 99% and 95% of CFTR transcripts, respectively.^218,219 CRISPR/Cas9-mediated NHEJ, using two gRNAs that target intronic sequences flanking the mutation, resulted in upward of 56% excision efficiency and up to 90% splicing rescue.²¹⁷ A third mutation, c.3718–2477C>T that creates a splice donor site,²²⁰ was also tested using this approach, achieving 55% excision efficiency and reduction of aberrant splicing to 17%.¹⁹³ This degree of editing is 10-fold higher than an HDR-mediated approach using the same cell type and CFTR locus.²²¹ This NHEJ strategy is effective for editing intronic CFTR splicing mutations but would not be feasible for mutations affecting canonical splice sites, as this would likely result in deletion of part of the protein coding sequence. NHEJ repair approaches could become a viable therapy for individuals with mutations that currently do not benefit from small molecule therapies.

Base editing has the potential to target many CF-causing mutations, through use of alternative Cas proteins with different PAM sequences, without the need of donor DNA. Of the mutations listed on the CFTR2 database, 11.7% can theoretically be targeted by CBEs and 31.8% by ABEs.²²² In an intestinal organoid model, Geurts et al. achieved 8.88% editing efficiency of the rare mutation R553X through adenine base editing using a canonical Cas9 fused to an ABE (cas9-ABE) and editing efficiencies of 1.43% when targeting mutations W1282X and R553X with a noncanonical Cas9 base editor (xCas9-ABE).²²² In this study intestinal organoids were dissociated into single cells and electroporated with plasmids encoding the Cas9-ABE and gRNA. When repeated using organoids derived from nasal brushings target editing of R553X achieved a similar efficiency of 8%,²²² highlighting a promising future for base editing in the treatment of CF lung disease.

Triplex-forming peptide nucleic acids (PNAs) are synthetic nucleotide analogs that can induce DNA repair through sequence-specific triplex formation at targeted genomic sites.²²³ Along with short donor DNA, PNAs have been used to edit the human β-globin and CCR5 genes.^224–227 Triplex-forming PNAs and donor DNA that target the F508del CFTR mutation delivered using polymer nanoparticles, achieved a modification frequency approaching 10% in human CFBE samples, and 5% in the mouse nasal epithelium with very little off-target effects (<0.0001%).²²⁸

Gene activation

Class V and other CF-causing mutations result in insufficient CFTR expression at the plasma membrane. Increasing production might restore chloride and bicarbonate conductance to near wild-type levels. Targeting a dCas9 activator (dCas9-VPR), in which a dCas9 is fused to the tripartite activator VP64-p65-Rta, with gRNAs that bind upstream of the CFTR promoter activated and increased expression of wild-type or F508del CFTR in cultured human nasal epithelial cells.²²⁹ Another epigenetic approach to increase CFTR expression is inhibiting BGas binding to CFTR. BGas is a long noncoding RNA (lncRNA) located in the antisense orientation of intron 11 of CFTR. BGas functions in concert with several proteins, including HMGA1, HMGB1, and WIBG, to alter chromatin structure and inhibit RNAP-II transcription of CFTR.²³⁰ Inhibition of BGas increases CFTR expression and chloride channel activity.²³⁰ Creation of a Gapmer, an antisense oligonucleotide with phosphorthioate bonds at the 5′ and 3′ ends, that targets BGas exon 1 (the exon that binds and suppresses CFTR transcription) increases CFTR expression sixfold, compared with a scrambled gapmer control.²²⁹ Both approaches further enhanced CFTR presentation to the cell surface with the addition of small molecule potentiators and correctors.²²⁹ Whether these documented increases in CFTR expression will provide therapeutic benefit remains to be seen. In addition, treatments using this strategy would not produce a cure but would need to be taken for the life of the individual. Despite these challenges, epigenetic modulation of CFTR could prove a useful therapeutic strategy in the future.

RNA editing

Like gene activation, RNA-directed editing can be used to provide a nonpermanent treatment for CF lung disease. Adenosine deaminases that act on RNAs (ADARs) are enzymes that catalyze site-directed mutagenesis of RNA by converting adenine to inosine, which is read as guanosine during translation.²³¹ This technology has been used to create a modified ADAR, termed λN-DD, in which a lambda phage N protein boxB that binds RNA is fused to the adenosine deaminase region of an ADAR.²³¹ This yields an ADAR that uses antisense RNA as a guide to bind CFTR mRNA. λN-DD has been shown to correct the W496X CFTR mutation in vitro with nearly 100% efficiency and within Xenopus oocytes to restore functional CFTR currents.²³¹

SP Deficiencies

Pulmonary surfactant is a mixture of phospholipids and SPs B and C that function to reduce surface tension at the air–liquid interface in the alveoli.²³² SPs are highly expressed in alveolar type 2 (AT2) cells.⁴⁹ Mutations in genes SFTPB, SFTPC, that encode SP-B and SP-C, and ATP-binding member 3 (ABCA3), which is involved in surfactant metabolism, cause SP deficiencies.²³³ SP deficiencies can cause loss-of-function phenotypes (SFTPB or ABCA3 mutations), owing to aberrant surfactant metabolism or SP function, or gain-of-function phenotypes (SFTPC mutations), because of cytosolic accumulation of abnormal SP in AT2 cells and disrupted lung development.²³³ SP deficiencies often present at birth with severe respiratory failure.²³³ Treatment options are mostly supportive or compassionate care, as donor availability for pediatric lung transplants is limited.

All known SFTPB mutations are autosomal–recessive, the most common mutation causing a frameshift in codon 121 (pPro133GInsTer95).²³³ An initial gene addition study utilized AdV delivery of human SFTPB to murine MLE12 cells. These mouse cells retain properties of AT2 cells but have very low SP-B levels and showed increased SP-B expression after Ad gene transfer.²³⁴ Delivery of the same AdV vector to cotton rats resulted in peak human SP-B 48–96 h postadministration, which subsequently diminished to undetectable levels.²³⁴ Administration of a SP-B expression plasmid by electroporation into a SP-B–deficient murine model, with expression controlled by a doxycycline inducible promoter, resulted in an eightfold increase in SP-B expression compared with a control plasmid treatment.²³⁵ SP-B plasmid treatment also resulted in a modest 2.5 to 5-fold increase in survival.²³⁵ Lung organoid models, termed alveolospheres, generated from hiPSCs, contain cells with canonical AT2 function and have been used to study the effectiveness of SFTPB gene editing.²³⁶ Lentivirus delivered wild-type (SFTPB) to alveolospheres derived from SFTPB(pPro133GInsTer95) hiPSCs resulted in successful transcription and translation of wild-type SP-B, as well as secretion of surfactant bioactive lipids.²³⁷ Electroporation of CRISPR/cas9 in the same model yielded similar results, including increased SFTPB mRNA, expression of mature SP-B protein, and generation of lamellar bodies.²³⁶ One study also showed effective SFTPB gene editing in a mouse model. Using a mouse with SP-B expression under a doxycycline-inducible promoter, nuclease-encoding chemically modified (nec) mRNA delivery, and donor DNA delivered through an AAV vector were used to insert a CAG promoter upstream of the SP-B start codon through HDR, resulting in prolonged survival of the mice.²⁰¹ These studies provide insight into the effectiveness of a gene editing strategy for SP-B deficiency.

The most common SFTPC mutation is the I73T missense mutation.²³⁸ Age of onset and disease severity are dependent on the specific SFTPC mutation and vary from severe respiratory failure at birth to idiopathic pulmonary fibrosis in adulthood.²³³ Knockout approaches may be a sufficient treatment option for this type of SP deficiency, as SFTPC null mice develop normal growth and lung function, whereas SFTPC(I73T) mice have an arrest of lung morphogenesis in late sacculation and are not viable.²³⁹ One recent study using CRISPR/Cas9-mediated knockout of SFTPC(I73T) in fetal mice shows the potential therapeutic applications of such an approach. An Ad vector encoding spCas9, EGFP, and gRNA was delivered to SFTPC(I73T) fetal mice through intra-amniotic injection, resulting in 64% reduction in AT2 cells that express HA-SFTPC(I73T) and a 22.8% increase in survival.²³⁸ Fetal lung editing through CRISPR/Cas9 has the potential to attenuate gain-of-function SP deficiency.²³⁸

Alpha-1 Antitrypsin Deficiency

Alpha-1 antitrypsin (AAT) deficiency (AATD) has a carrier frequency of 4% in the United States and is one of the most common genetic disorders in individuals with Northern European heritage.²⁴⁰ Mutations are named by their protease inhibitor (Pi) phenotype. The most common AAT mutation, E342K, results in the production of the PiZ form of AAT, also termed Z-AAT.²⁴⁰ Ninety percent of AATD patients are homozygous for PiZ allele. PiZ is prone to polymerization in the rough endoplasmic reticulum, impairing its secretion.²⁴⁰ PiZ results in two disease phenotypes: (1) a gain-of-function phenotype caused by aggregation of PiZ in hepatocytes that can lead to liver fibrosis, cirrhosis, and hepatocellular carcinoma, and (2) a loss-of-function phenotype with decreased secreted, functional AAT, leading to lung disease.²⁴⁰ PiZ leads to increased protease-mediated digestion of connective tissues in the lower airway and alveoli that can cause emphysema and chronic obstructive pulmonary disorder.²⁴⁰ About 5–10% of AATD patients will develop liver disease, and most will develop lung disease in adulthood.²⁴⁰ Current treatment options are limited for AATD. Patients who develop liver disease may require a liver transplant. Patients generally receive weekly infusions of AAT protein purified from healthy human donors.

AAT is a serine protease inhibitor (Serpin) encoded by the gene SERPINA1.²⁴⁰ AAT is a highly abundant protein, primarily produced by hepatocytes and secreted into the serum, but a small amount is also produced locally by bronchial epithelial cells and monocytes.²⁴⁰ Baseline human serum AAT levels range from 1,500 to 3,500 μg/mL (20–48 μM).²⁴¹ An AAT level of 11 μM in the serum or 1.2 μM in alveolar epithelial lining fluid (ELF) is considered to be protective, based on clinical observation of AATD patients.²⁴² One important function of AAT is to inactivate neutrophil elastase, an abundant protease released from neutrophil granules in response to various infectious or inflammatory stimuli.²⁴⁰

Gene addition strategies for AATD gene therapy involve delivery of the M-AAT coding sequence under the expression of a constitutive promoter. Early gene addition studies used retroviral vectors in murine or canine models of AATD, but these studies did not result in M-AAT expression at the therapeutic level.⁷⁵ Preclinical studies utilizing AAV vector serotypes AAV1, AAV2, AAV5, AAV6, AAV8, AAV9, and AAVrh.10 have been performed, with AAV1, AAV2, and AAVrh.10 serotypes moving onto clinical studies.^75,243 Both AAV1 and AAV2 vectors were administered through intramuscular injection and resulted in M-AAT expression below the therapeutic threshold. These trials likely failed because of the method of administration and the resultant targeted cells. Although, theoretically, any cell type can be modified to express M-AAT, as it is ultimately secreted into the serum, the method of administration can greatly affect the efficacy of the therapy. Using a murine model, intrapleural injection of an AAV5 vector resulted in higher human M-AAT levels than intramuscular delivery of the same dose.^75,244 This is most likely because of vector distribution after intrapleural administration, which allows for transduction of both the mesenchymal cells lining the pleura and systemic delivery through vector flow the visceral lymphatics, to systemic circulation, and then to hepatocytes. This would achieve both local M-AAT expression by mesenchymal cells that diffuses into the lung parenchyma and systemic M-AAT expression by hepatocytes that is secreted into the serum. A phase I/II clinical trial exploring intrapleural administration of an AAVrh.10 vector encoding M-AAT has recently concluded.²⁴⁵ The results of this trial have not yet been published, but murine model data suggest M-AAT expression may be higher than previous trials.

Several in vivo studies using a PiZ mouse model have shown therapeutic promise for CRISPR/Cas9-based gene therapy of AATD. One study took two parallel approaches using a dual AAV8 system, with one vector encoding saCas9, a smaller version of Cas9 derived from S. aureus, and the other vector encoding either gRNA or gRNA and donor DNA.²⁴⁶ The first approach used was to solve the gain-of-function phenotype of liver disease by creating indels through NHEJ. This resulted in 98% reduction in circulating PiZ and up to 86% reduction in PiZ aggregates within hepatocytes.²⁴⁶ The second approach was to correct loss-of-function phenotype of lung disease through HDR. This resulted in 5% correction efficiency. A third study used a similar dual AAV strategy, with one AAV9 vector that has broad cellular tropism encoding Cas9, and a second AAV8 with hepatocellular tropism carrying gRNA and a HDR template.²⁴⁷ Administration of these vectors resulted in an HDR frequency of 15–20% in newborn and adult PiZ transgenic mice and levels of secreted M-AAT of up to 71 μg/mL.²⁴⁷ A fourth study used a knock-in approach, targeting the ROSA26 safe harbor locus of C57B1/CJ mice.²⁴⁸ A dual Ad5 vector administration with one vector encoding human M-AAT donor DNA and the other vector encoding Cas9 and gRNA, achieved long-term expression of M-AAT lasting over 200 days and stable serum M-AAT level of up to 100 μg/mL.²⁴⁹

Future Considerations

Development of improved vectors for airway delivery and advancements in gene editing systems has brought remarkable progress. Major breakthroughs have occurred in CRISPR technology, including the advent of base editing and prime editing. Low indel and off-target frequencies with these approaches suggest that base editing and/or prime editing may be safer and more effective than traditional ZFN or CRISPR/Cas9 gene correction. Further study using in vivo models, optimization of correction efficiency, and adaptation to lung delivery will be required to assess therapeutic value.

As progress continues, open questions remain regarding the cell types targeted and the duration of treatment. Many airway epithelial cells that are accessible to vector delivery are terminally differentiated. As cells turn over, readministration may be required to continue to achieve therapeutic value. Investigation of the immune response and the potential for repeated dosing will be critical to future advancements.

Targeting of basal cells and other progenitor cell types could result in a lifelong cure using gene therapy; however, the effect of gene addition and gene editing on these cellular compartments is not fully understood. Determining transcriptomic changes in basal cells and other regional progenitor cell types after gene addition or editing and assessing their ability to differentiate and self-renew will provide important information on targeting these cell types for treatment of lung disease.

Author Disclosure

P.B.M. is on the scientific advisory board and receives support for sponsored research from Spirovant Sciences, Inc. A.V. has no competing financial interests.

Funding Information

This work was supported by the National Institutes of Health (UG3 HL-147366, P01 HL51670, P01 HL091842, P01 HL152960), the Cystic Fibrosis Foundation, The University of Iowa Center for Gene Therapy (DK54759), and the Roy J. Carver Chair in Pulmonary Research (P.B.M.).