Primary immunodeficiency diseases (PIDs) encompass a range of diseases due to mutations in genes that are critical for immunity. Haploinsufficiency and gain-of-function mutations are more complex than simple loss-of-function mutations; in addition to increased susceptibility to infections, immune dysregulations like autoimmunity and hyperinflammation are common presentations. Hematopoietic stem cell (HSC) gene therapy, using integrating vectors, provides potential cure of disease, but genome-wide transgene insertions and the lack of physiological endogenous gene regulation may yet present problems, and not applicable in PIDs where immune regulation is paramount. Targeted genome editing addresses these concerns; we discuss some approaches of CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats)/Cas system applicable for gene therapy in PIDs. Preclinical repair of gene mutations and insertion of complementary DNA restore endogenous gene regulation and they have shown very promising data for clinical application. However, ongoing studies to characterize off-target genotoxicity, careful donor designs to ensure physiological expression, and maneuvers to optimize engraftment potential are critical to ensure successful application of this next-gen targeted HSC gene therapy.
Primary immunodeficiency diseases (PIDs) encompass a range of disorders due to mutations in genes that are critical for immune function and result in absent, non-functioning or abnormally active immune cells. Over 300 genes have been associated with PIDs  and, with the availability of high-throughput next-generation sequencing, the list is rapidly growing longer. Better characterization of genetic defects provides the opportunity to design specific gene correction strategies for the treatment of PIDs, thus paving the way to precision medicine. In addition to delineating mutation types in PIDs, such as point mutations, chromosomal alterations (deletions, insertions, translocations) or copy number variation, mechanistic understanding of the genetic end products is also crucial as loss-of-function (LOF), gain-of-function (GOF), haploinsufficiency, or dominant negative phenotypes probably require different approaches tailored to each scenario when contemplating treatment strategies .
PIDs most commonly present with increased susceptibility to infections and evidence of immune dysregulation that may present as autoimmunity or hyperinflammation. Many recently defined PIDs include GOF mutations in genes that affect cellular signaling pathways or immune regulation. We will discuss the versatile applications of targeted genome editing using the CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats)/Cas system to both LOF PIDs and more challenging GOF PIDs. The clinical manifestations of the PID may influence the approaches for treatment of the disease. Previously well-described LOF mutations, such as those leading to severe combined immunodeficiency (SCID) due to IL2RG defects (X-linked SCID or SCID-X1) or adenosine deaminase deficiency (ADA–SCID), or chronic granulomatous disease (CGD), present early with severe infections. In contrast, some recently described PIDs with abnormalities in cellular immune signaling pathways may be diagnosed later in life with significant immune dysregulation presenting as autoimmunity [3,4]. Clinical management of PIDs with increased susceptibility to infections has included lifelong anti-microbial prophylaxis, immunoglobulin supplementation and, generally, consideration of hematopoietic stem cell (HSC) transplant (HSCT) from a human leukocyte antigen-matched allogeneic donor. For patients with defects in immune cell signaling, medical intervention can be tailored to enhance regulatory pathways or to inhibit inflammatory signals [5–8]. The design of genome-editing approaches to treat PID, therefore, requires consideration of the specific causative genetic mutation as well as the clinical presentation and condition of the patient.
The first successful allogeneic HSCT of a SCID-X1 patient over 50 years ago demonstrated that replacement with normal HSCs can cure this PID . Although potentially curative, allogeneic HSCT still suffers from the following problems: potential for graft versus host disease, graft rejection, toxicities related to chemotherapeutic or immunosuppressive agents and most critically, the availability of matched allogeneic donor grafts [10,11]. HSC gene therapy, however, can circumvent some of these problems by correction of autologous HSCs by insertion of a therapeutic gene into the chromosomal DNA of the HSC that engrafts and differentiates ideally with normal function. The insertion of the therapeutic gene relies on integrating vectors that deliver and insert one or more copies into the HSC. Using γ-retroviral vectors, initial HSC gene therapy trials demonstrated significant clinical benefits, especially where the insertion conferred a survival advantage to gene-corrected cells [12–14]. Unfortunately, lymphoid and/or myeloid insertional mutagenesis was observed in some PIDs including SCID-X1, CGD, and Wiskott–Aldrich syndrome (WAS) [15–17]. More recently, human lentivirus vectors have been used, with improved safety features and an integration pattern in the chromosomal DNA that does not favor promoters and enhancers, to reduce the risk of oncogenic activating integrations . Replication-deficient lentivectors have been examined in clinical trials for the treatment of PIDs, including SCID-X1, CGD, WAS, ADA–SCID with some positive results [19–22]. Lentivector gene therapy products for ADA–SCID, SCID-X1, and CGD are soon likely to become approved drugs [23,24]. The unequivocal clinical benefits following HSC gene therapy provide further incentive to design better strategies for specific gene modification of autologous HSCs for transplant to treat PIDs. Lentivector HSC gene therapy may yet suffer from two disadvantages (the potential long-term genotoxicity of the quasi-random genome-wide vector integrations and the lack of physiological regulation of gene expression) that may be eliminated by specifically targeted gene-editing techniques.
Targeted genetic modification may offer the greatest potential yet for definitively curing the widest spectrum of PIDs. Recent advances in nucleases have made possible precise genome editing by the use of programmable endonucleases including zinc finger nucleases, transcription activator-like effector nucleases, or megaTALs and more recently, the CRISPR/Cas9 system [25–31]. These genome engineering systems share an ability to recognize and target specific genomic DNA sequences and create a site-specific double-strand break (DSB) or a single-strand ‘nick’ in the DNA, thus stimulating the endogenous cellular repair machinery to bring about the genomic alteration. The main repair pathways are the error-prone non-homologous end joining (NHEJ) that results in insertions and/or deletions (indels) at the site of the break, and the homology-directed repair (HDR) that uses the sister chromatid or an exogenous DNA donor template to repair the break . In the context of PIDs, correction of genetic defects most frequently relies on HDR, although the NHEJ repair pathway can be harnessed to functionally inactivate a targeted locus (‘knock-out’) by introducing small indels.
We focus our discussion on the use of the CRISPR/Cas9 system as a representative of this class of gene-editing tools and on the genome-editing strategies that can be applied to potential cures of specific PIDs. Although gene-edited HSC may be the best definitive therapy, for some PIDs, for example in CD40L deficiency causing Hyper IgM syndrome, correction of primary immune cells, e.g. T-cells, may be sufficient for control of disease, either in isolation or to improve control of disease to reduce risks of definitive transplant, whether allogeneic or autologous gene-modified HSCs (Figure 1) [33,34].
Overview of the gene-editing therapy for primary PIDs.
CRISPR/Cas9 tools for gene editing
In addition to its versatility, the wide availability of CRISPR/Cas9 reagents has resulted in the rapid development of new tools, as summarized in Figure 2, and growing knowledge leading to creative solutions to solve the range of genetic mutations in PIDs. We present here some tools available to date and illustrate their use in specific PIDs.
CRISPR/Cas9 system for targeted genome editing.
The CRISPR/Cas9 system relies on the recognition of a specific genomic DNA target complementary to a short guide RNA sequence (gRNA, generally ∼20 nucleotides), a protospacer, adjacent to an appropriate protospacer adjacent motif (PAM). The formation of an RNA–DNA pairing induces a conformational change in the Cas9 nuclease that promotes its nuclease activity and the generation of a double-strand cleavage 3 bp upstream of the PAM sequence . The first described Cas9 nuclease, a class 2 type II CRISPR/Cas system derived from Streptococcus pyogenes (SpCas9), recognizes a 5′-NGG-3′ PAM sequence theoretically found every 8–12 bp in the genome [26,31,36]. Excellent works in the field have created and uncovered Cas9 variants requiring different PAM sequences [35,37] and nucleases from alternate bacterial species [Streptococcus thermophilus (PAM = 5′-NNAGAAW-3′ and 5′-NGGNG-3′ for CRISPR1 and 3, respectively, every 64 bp) or Neisseria meningitidis (PAM = 5′-NNNNGATT-3′ every 128 bp)] [25,38,39] have greatly extended the scope of targetable genomic sites.
The safety profile of the CRISPR/Cas system can be improved in many ways, most simply with the use of engineered SpCas9 High Fidelity variants [40,41]. Alternatively, studies using Cas9 nickases (Cas9n; D10A or H840A mutations) [26,42] or catalytically inactive Cas9 (fCas9) or dead Cas9 (dCas9) fused to a FokI nuclease domain [43–45] reported improved specificity by creating a DSB between the two binding sites. Cas9n can improve specificity by up to 1,500-fold compared with the wild-type (WT) Cas9 with fewer off-target cleavage events [46,47]. Also reported to possess greater specificities are Cas orthologs of the class 2 type V CRISPR/Cas system like Cpf1/Cas12a and Cas12b nucleases from other bacterial species [48–52] compared with S. pyogenes Cas9. Ingenious base editing approaches that convert mutants to WT sequence without DNA cleavage are thought to limit indels at target sites. Current strategy fuses Cas9 with a cytidine or adenosine deaminase to cause an irreversible conversion of C:G to T:A or A:T to G:C, respectively, without DNA cleavage [53–56]. The current limitations of this approach are the correction efficiency and a relatively limited scope of targetable mutations. Furthermore, more sensitive approaches revealed substantial off-target effects, with a 20-fold higher off-target frequency with cytosine base editors compared with adenosine base editors .
The next critical component of the CRISPR genome-editing system is the guide. Initially, the length of guide RNAs was around 100 nucleotides (spacer of 20 nucleotides complementary to the target genomic sequence) and they were generally produced by in vitro transcription. The use of single-guide RNA (sgRNA) with a shorter spacer of 17, 18, or 19 nucleotides improves specificity but may reduce the efficiency of genome-editing rates , and requires careful evaluation for each PID. Production of gRNA by in vitro transcription may be challenging for some groups and prohibitory for large-scale clinical grade manufacturing. Synthetic sgRNAs have potential advantages including chemical modifications that increase the stability of the sgRNA and genome-editing frequencies [59,60]. Recently, many have found Cas9 nuclease complexed with sgRNA to form a ribonucleoprotein (RNP) that supports earlier editing activity with less associated toxicity an attractive option (Figure 2a) [40,61,62].
Since the donor for the gene-editing system determines the end product, this component of the genome-editing system requires careful attention and will be discussed next in greater detail in the context of specific mutations and PIDs. Donors that have been used include single-stranded oligodeoxynucleotide (ssODN) [63,64], long ssDNA and linear or circular dsDNA [65–68], or complementary DNA (cDNA) within an engineered virus cassette such as integrase-defective lentivirus (IDLV; transient expression and weak integration capability)  or recombinant adeno-associated virus (rAAV), depending on the size of the donor (Figure 2). The rAAV as a donor delivery tool has low immunogenic potential and has high infectious capability even in non-dividing cells and low integration rates in the genome of the host cell [70–72]. They are generally capable of delivering up to 4.5–5 kb but by using two co-transduced AAV vectors, DNA cargos of up to 6.5 kb have been reported . For targeting human HSCs, rAAV serotype 6 was shown to have the greatest affinity, and a mutant AAV6 capsid provided further improvement [74–77]. In general, donors should have homology arms of 400 bp, codon optimization of the cDNA and modification of the PAM sequence to prevent re-cutting of the donor or gene-modified cells (Figure 2c) .
CRISPR/Cas9 approaches for HSC genome editing in PIDs
Two main considerations in determining a genome-editing strategy for PIDs are firstly, the mutation size (single or few nucleotides versus large deletions) and frequency (mutation hotspots versus disease-causing mutations distributed throughout the gene) that affect the choice of donor template, and secondly, the functional effect of the mutation. While LOF mutation can be boosted by correction of 1 or 2 alleles, diseases that result from heterozygote mutations are challenging due to the risks of unintended cuts on the normal allele that results in disease, or the residual effects of uncorrected mutant allele. Ideally, in haploinsufficiency mutations, one needs correction of the mutant allele without disrupting the other allele. However, in GOF mutations, the goal is to knock-out or repair the mutant allele while avoiding unintended cuts of the other allele. Great work in the design of broader choices for nucleases and with improved specificity may provide the necessary tools necessary for correction of these more challenging mutations [37,78,79].
LOF mutations that involve a single or few nucleotides are amenable to the use of a short oligonucleotide donor to repair the mutation resulting in a normal genetic sequence, i.e. ‘gene repair’ (Figure 2c). About two-thirds of human genetic diseases arise from single-base mutations. To correct such mutations in situ, the simplest approach is to use a short donor template, such as a ssODN [80–85] although linear or plasmid dsDNA donors have also been used [86,87]. Various studies have evaluated the effects of the length, sense, and symmetry of the donor ssODN as well as chemical modifications (such as phosphorothioate linkages both 5′ and 3′ ends) to induce higher HDR efficiency [60,81,86,88]. This strategy was employed to repair a relatively frequent mutation in the CYBB gene that encodes for gp91phox, a critical catalytic subunit of the phagocyte NADPH oxidase complex, which results in X-linked chronic granulomatous disease (X-CGD). Using a chemically modified 100 bp ssODN with symmetrical homology arms, delivered by electroporation with Cas9 mRNA and in vitro transcribed sgRNA, in vitro correction rates of 30–40% of HSC were achieved that, upon transplantation into NSG immunodeficient mice, resulted in 15–20% stable corrected cells at 5 months . Subsequent improvements to this system have included the use of chemically modified synthetic guides and non-targeting donor oligos that improved in vitro correction rates to 50–60% (unpublished). This level of correction is sufficient to reverse the disease phenotype of X-CGD as it relates to susceptibility of infections only. In contrast with some gene therapy approaches using exogenous promoters, the gene-edited corrected cells restored normal physiological regulation with normal amounts of protein expression in the appropriate mature differentiated phagocytes and functionally normal amounts of NADPH oxidase activity. Although consideration of individual mutations seems daunting at present, it is possible that with the rapid progress in producing synthetic guides and donor oligonucleotides, an individual mutation-specific repair system can be optimized for treatment within a few months for patients not needing emergency intervention. Optimization of this approach may provide proof-of-concept that gene-repaired HSCs can restore HSC compartment for gene therapy in other diseases, where a single LOF mutation may account for most or all patients, such as sickle cell disease  or p47phox-deficient CGD . Of note, the level of correction for reversal of disease phenotype for different LOF mutations likely varies and is important to define in the preclinical studies to provide the rationale for treatment. For example, diseases, such as SCID-X1 in which functional cells have survival advantage, require lower levels of (∼10%) correction to reverse disease phenotype .
While short donor oligos may be ideal for small mutations, repair of large deletions or insertions is not possible with this approach. Instead, an alternative strategy to address large mutations or simply address mutations located widely in the gene is to insert a functional cDNA into a targeted site, generally downstream of the endogenous promoter to restore physiological regulation (Figure 2c). Targeted integration (TI) near the start site could potentially address > 95% of the mutations. This ‘universal’ or ‘one-size-fits-all’ approach has been shown to achieve excellent rates of TI of the cDNA, or portions of genes at a targeted site (a ‘knock-in’) where the final genomic sequence is altered to provide a functional gene product under the control of the endogenous promoter. Although the donors can be delivered many ways, including IDLV [83,92], the best outcomes are observed following delivery by rAAV. Very encouraging preclinical data pave the way for imminent clinical application for treatment of SCID-X1 due to IL2RG gene defects [91–93] and CD40L for XHIM [34,94]. Other PIDs reporting encouraging results following targeted knock-ins include BTK for X-linked agammaglobulinemia (XLA)  and CYBB for X-CGD [76,96,97].
Recent advances in optimizing the CRISPR/Cas system have generally resulted in excellent TI rates over 50% in vitro. Critical factors to be taken into consideration to ensure successful clinical outcomes for gene-edited HSC gene therapy include an evaluation of the functional end product (protein expressed) as well as the engraftment capabilities following CRISPR/Cas9 gene editing. In addition to endogenous gene regulation, factors such as the Kozak consensus sequence, intronic regulatory elements, and polyadenylation length (poly(A)) should be taken into consideration when designing donors (Figure 2). To illustrate, we evaluated the ‘knock-in’ approach targeting the insertion of a partial cDNA (CYBB Exon 7–13) using our previously reported guide targeting CYBB E7 c.676 with the ssODN gene repair correction described above . Expression of the donor encoding CYBB Exons 7 to 13 is regulated by the endogenous CYBB promoter. We observed significantly reduced amounts of gp91phox protein per cell produced that was independent of TI rates (>50–70%). Substituting the poly(A) with a longer poly(A) significantly improved the amount of protein produced. Similarly, the addition of Woodchuck Hepatitis Virus Post Transcriptional Regulatory Element (WPRE) significantly increases expression of genes delivered by viral vectors [94,99,100]. Furthermore, there may be critical regulatory elements in introns lost with cDNAs that are devoid of all introns. This was illustrated by Sweeney et al. who showed that TI in CYBB E1 with a CYBB E1–13 cDNA failed to restore the gp91phox expression, while the expression was restored when the donor was inserted in Exon 2, suggesting the presence of critical regulatory elements in Intron 1 . This problem may be circumvented by the retention of the critical intronic sequences that may also allow the expression of isoforms under the physiological regulation of the promoter.
The engraftment capabilities of gene-edited HSCs need careful evaluation. We previously observed good engraftment rates of 30–40% following transplants of ∼1 million CRISPR/Cas9/ssODN gene-repaired HSCs per mouse (NOD SCIDyc-deficient strain) conditioned with 20 mg/kg busulfan. Our current experiences with AAV-mediated gene ‘knock-in’ approaches have observed lower engraftment rates in general (unpublished). Several groups have reported immune responses to viruses such as interferon-inducible antiviral factors or activation of p53-mediated DNA damage response induced by DSBs which may alter stemness of the HSCs and impair the ability to engraft [101–104]. Better understanding of the mechanisms involved may lead to discovery of agents that may counteract these responses, such as Cyclosporin H [105,106], or agents that enhance HSC self-renewal such as pyrimido-indole derivative UM171 and aryl hydrocarbon receptor antagonist StemRegenin 1 [77,107,108].
Since HDR occurs most efficiently during the S/G2 phases of the cell cycle when DNA is relaxed and more accessible , there are molecules targeting factors critical for DNA repair mechanisms that may also promote HDR. Synchronizing Cas9 expression with activation of the HDR machinery may also enhance templated repair [110–112] or by the use of inhibitors of key enzymes of the NHEJ pathway like SCR7, an inhibitor of the DNA ligase IV DNA-binding domain, or an inhibitor of 53BP1 (i53) [113,114]. A 50–70% increase in gene correction rates in vitro (both gene repair and targeted gene knock-in) was achieved with the use of the inhibitor of 53BP1 (unpublished). Long-term safety and efficacy studies of NSG transplants for gene-modified HSCs with the aid of i53 are ongoing.
GOF PIDs include recently characterized PIDs, such as phosphoinositide-3-kinase δ-syndrome (PIK3CD), or cytotoxic lymphocyte antigen-4 (CTLA-4) haploinsufficiency with pathological increased cellular signaling and hyperactivation, resulting in clinical disease [115–119]. In PIDs due to GOF mutations, transplants from allogeneic donors have shown that strategies that result in chimeric populations of gene-mutant cells and genetically normal cells will likely not reverse patient's clinical phenotype. Strategies to cause new mutant alleles may result in mutant phenotype, while TI into the normal allele will not alleviate the end product due to the mutant allele. There are challenging strategies that at least in theory may revert phenotype, such as targeted knock-out of the mutant allele and replacing the mutant allele with a TI of normal cDNA. Single nucleotide GOF mutations may also benefit from specific base editing that corrects the specific mutation without risks of DSB that potentially creates new mutants.
In conclusion, the CRISPR/Cas9 technology has transformed biomedical research and preclinical data strongly support the translation of specific genome editing for the treatment of a variety of primary immunodeficiencies. There are critical ongoing research efforts to develop more sensitive techniques to detect chromosomal alterations due to unintended effects of endonucleases. Thorough evaluation of engraftment capabilities of gene-edited HSCs particularly with AAV-delivered donors and detailed analysis of functional outcomes for each corrected gene defect is essential for effective correction of PIDs. Despite some concerns for off-target indels, the potential for circumventing genome-wide vector integration concerns, and to restore physiological control under endogenous promoters should greatly improve the safety of HSC gene therapy. Continued progress in gene-editing tools will hopefully provide broader range of choices for correcting currently challenging PIDs, especially those due to GOF mutations.
HSC gene therapy is clinically beneficial for some PIDs.
Targeted genome editing can avoid genome-wide integration and constitutive expression of transgene.
CRISPR/Cas approach to repair gene mutations or insertion of cDNA at endogenous gene locus can restore physiological regulation.
Preclinical CRISPR/Cas9 studies for some LOF mutation PID are ready for translation to the clinic.
Future direction — CRISPR/Cas9 is versatile and may provide innovative tools and strategies to address the range of genetic mutations in PIDs, such as haploinsufficiency, dominant negative or GOF mutations.
adenosine deaminase deficiency
chronic granulomatous disease
Clustered Regularly Interspaced Short Palindromic Repeats
Hematopoietic stem cell
non-homologous end joining
protospacer adjacent motif
Primary immunodeficiency diseases
recombinant adeno-associated virus
severe combined immunodeficiency
X-linked chronic granulomatous disease
This work is supported by the Intramural Research Program of the National Institutes of Allergy and Infectious Diseases and the National Institutes of Health.
We owe our gratitude to our patients who have volunteered and supported us in our research, our colleagues at the Dowling Apheresis Center, and the Center for Cell Engineering at the Clinical Center for support with research stem cell collections. We also thank Dr Kol Zarember for the helpful comments for the review.
The Authors declare that there are no competing interests associated with the manuscript.