CRISPR-Mediated Editing of the B Cell Receptor in Primary Human B Cells.

Vaccination approaches have generally focused on the antigen rather than the resultant antibodies generated, which differ greatly in quality and function between individuals. The ability to replace the variable regions of the native B cell receptor (BCR) heavy and light chain loci with defined recombined sequences of a preferred monoclonal antibody could enable curative adoptive cell transfer. We report CRISPR-mediated homologous recombination (HR) into the BCR of primary human B cells. Ribonucleoprotein delivery enabled editing at the model CXCR4 locus, as demonstrated by T7E1 assay, flow cytometry, and TIDE analysis. Insertion via HR was confirmed by sequencing, cross-boundary PCR, and restriction digest. Optimized conditions were used to achieve HR at the BCR variable heavy and light chains. Insertion was confirmed at the DNA level, and transgene expression from the native BCR promoters was observed. Reprogramming the specificity of antibodies in the genomes of B cells could have clinical importance.

Introduction

The CRISPR/Cas9 system enables rapid and precise genome engineering, and cell therapy is a particularly noteworthy application of this technology (Barrangou and Doudna, 2016). Ex vivo manipulation of cells is attractive because it obviates surmounting the formidable challenge of achieving efficient and cell-type-specific delivery in vivo; perhaps more importantly, ex vivo editing allows one to analyze and characterize edited cells before their adoptive transfer into patients (Barrangou and Doudna, 2016). Such quality control is very important, as in vivo editing by CRISPR/Cas9 may produce unexpected off-target mutations.

CRISPR-mediated ex vivo genome editing has been applied to correct the gene encoding hemoglobin in hematopoietic stem cells (HSCs) and/or progenitor cells, providing an innovative path to address β-hemoglobinopathies (Dever et al., 2016, Traxler et al., 2016). Genome engineering has also enabled deletion of CCR5 in hematopoietic stem/progenitor cells (HSPCs) (Holt et al., 2010) or CD4+ T cells (Perez et al., 2008), thereby protecting these cells from infection by HIV. Much effort has been made to edit HSCs and T cells, whereas far less attention has been given to the editing of B cells, despite the important role that they play in several immune processes, much of which is related to their ability to produce antibodies.

Monoclonal antibodies are the fastest growing class of therapeutic agents (Beck et al., 2010) and can be used to treat sundry pathologies, including autoimmune disease, cancer, and infectious disease. A main limitation associated with this therapeutic modality is the need for repeated administration—often for years or decades—which typically involves intravenous infusion at an ambulatory outpatient care center. Such logistics is very costly to the health care system and poses inconvenience to patients (Sylwestrzak et al., 2014) that may result in noncompliance. A second drawback of recombinant monoclonal antibodies is related to their production in cells of non-human origin (e.g., Chinese hamster ovary cells) or non-B-cell lineage (e.g., human embryonic kidney cells). The function of antibodies is strongly influenced by post-translational modifications (Li et al., 2015), which may differ between these cell lines and human B cells.

Harnessing the human antibody response is becoming increasingly feasible, as methodologies to isolate rare clones continue to improve (Wilson and Andrews, 2012, Sanjuan Nandin et al., 2017, Kwakkenbos et al., 2014, Franz et al., 2011). Primary human B cells have been transformed into stable cloned lines that secrete antibodies that neutralize respiratory syncytial virus in vivo (Kwakkenbos et al., 2010). The ability to induce the production of neutralizing antibodies to notable antigens by B cells in situ remains an unmet need, and repeated administration of recombinant products is not practical for several indications, particularly in the chronic therapeutic setting and for prophylaxis against infectious diseases.

The ability to replace the B cell receptor (BCR) heavy and light chains in an individual's B cells with sequences encoding a desired monoclonal antibody could lead to curative adoptive cell transfer. The antibody would be expressed dynamically and physiologically from its native enhancers and promoters in response to detection of antigen, resulting in the production of appropriate concentrations of antibody; such titrated dosing would be expected to ameliorate the undesirable side effects experienced by patients whose dose of recombinant product does not match their prevailing antigen concentration, which varies over time. In addition to defining specificity, this approach would generate autologous post-translational modifications. Such modifications can be optimized to program a preferred function (Lu et al., 2017), for example, by disrupting genes in other genomic loci that encode particular glycosyltransferases.

Although it has been shown that murine B cells (Cheong et al., 2016, Pogson et al., 2016, Chu et al., 2016) and primary human B cells (Hung et al., 2018, Wu et al., 2018) can be edited by CRISPR, homologous recombination (HR) at the BCR loci has been limited to hybridomas to date (Pogson et al., 2016). Herein, we sought to achieve HR at the BCR loci in primary human B cells for the first time. Such a demonstration would represent an important step toward achieving cellular humoral vaccines. Such cell therapies could replace repetitive administration of recombinant monoclonal antibodies, such as anti-tumor necrosis factor (TNF)-α, in the therapeutic setting. Knocking in sequences encoding broadly neutralizing antibodies against influenza or HIV (Walker and Burton, 2018) could also enable robust prophylaxis.

Results

Cas9 RNP Is Required to Edit Primary Human B Cells

First, we aimed to determine how Cas9 should be delivered to primary human B cells to enable genome engineering. Transfection via electroporation is a common approach for delivering exogenous biologics into primary immune cells. Cutting and editing can be achieved in primary human T cells following electroporation of mRNA encoding Cas9 (Eyquem et al., 2017) or Cas9/guide RNA (gRNA) ribonucleoproteins (RNPs) (Schumann et al., 2015). In primary human B cells, RNPs were required to achieve editing, as delivery of Cas9 encoded by DNA or mRNA in combination with gRNAs targeting the model CXCR4 locus did not produce any cutting, as evidenced by the T7E1 assay (Figure 1A, see Transparent Methods). This assay involves recognition and cleavage of mismatched DNA, which arises following non-homologous end joining.

Figure 1

Cas9 RNP Enables Efficient Genome Editing of Primary Human B Cells

(A) A T7E1 assay shows efficient cutting of the model CXCR4 locus following electroporation with Cas9 RNP, but not following electroporation with gRNA in conjunction with DNA or mRNA encoding Cas9. DNA was isolated 3 days post-transfection.

(B) TIDE analysis following electroporation of primary human B cells with DNA encoding Cas9 plus gRNA, mRNA encoding Cas9 plus gRNA, or Cas9 RNP targeting the model CXCR4 locus. DNA was isolated 3 days post-transfection.

(C) Flow cytometry analysis of cell surface CXCR4 expression confirms that target protein expression is greatly reduced following electroporation of Cas9 RNP. Cells were analyzed 6 days post-transfection. UT, untransfected.

These findings were validated by Tracking of Indels by DEcomposition (TIDE) analysis (Brinkman et al., 2014) of Sanger sequencing data (Figure 1B), which involves computational tracking of insertions and deletions by decomposition of the quantitative sequence trace data. It has been previously reported that use of RNP rather than DNA or mRNA increases cell viability substantially among human embryonic stem cells (Kim et al., 2014). RNPs also lead to substantially more specific genome modification than does DNA (Zuris et al., 2015). We next confirmed that disruption of this locus at the genomic level translated to decreased protein production. Flow cytometry revealed a 70% reduction in the levels of surface CXCR4 expression following RNP delivery (Figure 1C). Having demonstrated that gene disruption was possible, we sought to make use of the cell-intrinsic HR-directed DNA repair pathway to insert a DNA template of interest. During the previous cutting experiments, cells were cultured with interleukin-4, which was sufficient to maintain cell viability; however, to perform HR-mediated DNA repair, cells must proliferate as well. Thus the determination of suitable culturing conditions for primary human B cells was paramount to further investigation.

HR Is Achieved in Primary Human B Cells

Specifically, we had to identify a suitable activation protocol to perform HR on primary B cells. In vivo, activation of B cells can lead to differentiation into plasmablasts that perish within 1 week. In vitro, activated B cells were viable for more than 2 weeks, unless electroporation was performed in conjunction with the activation protocol, in which case viability was greatly decreased. We assessed various combinations of cytokines, Toll-like receptor agonists, and/or CD40 ligation (>30 conditions evaluated) and found that the CellXVivo Human B Cell Expansion Kit (R&D Systems) yielded the best activation and viability (data not shown). Locus-specific genome sequencing revealed that activation of B cells was required to achieve HR, as determined by CRISPR-GA analysis (Guell et al., 2014) of MiSeq data (Table 1). CRISPR-GA maps deep sequencing reads, estimates and locates insertions and deletions, and computes the allele replacement efficiency. Pre-activation of cells for 5 days led to more efficient editing than pre-activation for 3 days. Activation of cells pre- and post-transfection was required to achieve HR. We varied additional transfection conditions and determined that a Cas9:gRNA:HR template ratio of 100:580:100 pmol was optimal.

Table 1

Pre-activation of Primary Human B Cells Is Required to Achieve Homologous Recombination

	Pre-activation	Post-activation	HR Efficiency (%)
Untransfected controls	–	–	0.01
Donors	5 days	3 days	0.33
Donors	5 days		0.24
Donors	3 days	5 days	0.16
Donors	3 days		0.19
Donors		5 days	0.05

CRISPR-GA was used to analyze MiSeq data from at least two donors per group.

To analyze HR efficiency in a timely manner, we used a restriction digest assay that involved use of a single-stranded oligonucleotide containing a HindIII restriction site targeting CXCR4 (Schumann et al., 2015) (Figure 2A). This oligonucleotide was included in the transfection media along with an RNP targeting CXCR4, and editing was confirmed by TIDE analysis of Sanger sequencing data (Figure 2B) and by CRISPR-GA analysis of MiSeq data (Figure 2C). Although encouraging as a proof of concept, the MiSeq data indicated that the efficiency of editing was rather low, as was cell viability following activation and electroporation.

Figure 2

Site-Specific Introduction of a Restriction Site Is Achieved by Homologous Recombination in Primary Human B Cells

(A) A schematic of homologous recombination (HR) to insert a HindIII restriction site into the CXCR4 locus.

(B) TIDE analysis confirms editing in primary human B cells upon inclusion of a single-stranded DNA template.

(C) CRISPR-GA analysis of MiSeq data confirms insertion of the HindIII restriction site into the CXCR4 locus of primary human B cells, illustrating that HR is also possible. Cells were collected and analyzed 5 days post-transfection.

(D) Addition of the pan-caspase inhibitor Q-VD-OPH (“OPH”) improves cell viability, which was quantified by flow cytometry 5 days post-transfection. n = 3 independent experiments, mean ± SEM. *p < 0.05, calculated by unpaired Student's t test.

(E) A HindIII restriction digest shows that insertion is possible in activated primary human B cells and is improved by the addition of OPH. DNA was isolated 6 days post-transfection.

Caspase Inhibition Improves Cell Viability and HR Efficiency

To increase cell viability following electroporation, the pan-caspase inhibitor Q-VD-OPH (“OPH”) was added to the culture media, conferring a nearly 3-fold increase in the proportion of viable cells (Figure 2D). Addition of 10 μM OPH to the culture media improves cell viability following activation and electroporation, thereby leading to an increased amount of HR per cell input, as evidenced by HindIII restriction digest (Figure 2E).

HR Is Achieved at the BCR Heavy and Light Chain Loci

Although there are many loci of potential relevance for genome editing in primary human B cells, the BCR variable heavy and light chain loci are of particular interest, especially for adoptive cell therapy applications (Figure 3A). Top-performing gRNAs were identified by performing T7E1 assays after transfecting cells with RNPs targeting various sequences in the VH3-23, JH5, VK3-20, and JK4 loci (see Figures S1A–S1D). These loci were chosen because they are among the most frequently recombined loci (Prabakaran et al., 2012, Glanville et al., 2009, Boyd et al., 2010) and even co-recombine preferentially (Boyd et al., 2009, Xiao et al., 2013). Combinatorial cutting of both V and J loci was apparently achieved without loss of efficiency at either targeted locus (see Figures S1E and S1F). Although CRISPR is a tool that is known to enable multiplex editing in a given cell, these data do not rule out the remote possibility that the gRNAs cut individually in separate cells. We next established that a defined sequence could be inserted into a specific V region following a single cut or across V(D)/J following dual cutting (Figure 3B). Insertion was confirmed by cross-boundary PCR, which involved the use of a forward primer that is specific to the genome and a reverse primer that is specific to the inserted HR template (Figure 3C).

Figure 3

Defined Sequences Can Be Successfully Inserted into the BCR Heavy and Light Chain Loci of Primary Human B Cells

(A) A schematic of the conceptual framework of BCR reprogramming: an individual's B cells are recovered, modified ex vivo to express a monoclonal antibody of interest, and reinfused into the individual.

(B) A schematic of HR template insertion into or substitution across the BCR loci. A HindIII restriction site was introduced into the heavy and light chains, either by insertion into the V region or by replacement of a region spanning V/J.

(C) Site-specific insertion is confirmed by cross-boundary PCR. An amplicon is observed only if insertion into the defined genomic locus has occurred.

(D–F) Confirmation of insertion into the BCR loci by cross-boundary PCR. An amplicon is observed only if insertion has occurred. Cells were analyzed 5 days post-transfection. (D) Insertion efficiency into the heavy chain (VH3-23) locus occurs in a dose-dependent manner for increasing amounts of Nanobody HR template. (E) An HR template encoding the heavy variable and constant regions of adalimumab was successfully inserted into the heavy chain (VH3–23) locus of primary human B cells. (F) An HR template encoding the light variable and constant regions of adalimumab was successfully inserted into the light chain (VK3–20) locus of primary human B cells. Untransfected (UT) cells were used as unedited controls.

See also Figures S1 and S2.

Sequences Encoding Monoclonal Antibodies Can Be Inserted into Endogenous Loci

Replacement of the endogenously recombined BCR with a sequence encoding a defined monoclonal antibody of interest could allow for programmed lifelong immunity. Cutting at the V locus only—rather than at both V and J—allows one to reduce the number of gRNAs (and hence potential off-target editing) by 2-fold. It also allows one to insert a specific constant region with embedded functionality of choice (e.g., IgG1 versus IgG4; introduction of mutations into Fc to confer extended half-life, Zalevsky et al., 2010; or antibody-dependent cellular cytotoxicity versus complement-dependent cytotoxicity, Hessell et al., 2007) rather than rely on the class switch recombination that has occurred or will occur in the cell being edited. We thus focused on this approach, despite the fact that we confirmed that insertion across V(D)/J is possible (Figure 3C). Sequences encoding a nanobody (ozoralizumab) or a monoclonal antibody (adalimumab) that neutralizes TNF-α were inserted into the heavy and/or light (kappa) loci of primary human B cells isolated from donors, as confirmed by cross-boundary PCR (Figures 3D–3F, see Figures S2A–S2C).

Nanobody and Antibody Are Expressed from Endogenous Promoters

Next, we sought to demonstrate that it was possible to achieve intrinsic expression of an inserted sequence driven by the native BCR promoter(s). RT-PCR was performed to demonstrate expression from the endogenous heavy chain promoter at the mRNA level (see Figure S2D). Flow cytometry was used to assess expression from this locus of the FLAG epitope, which was fused to the C terminus of the nanobody (Figures 4A and 4B). Finally, it was observed that B cells could express both the heavy and light chains of a defined monoclonal antibody, wherein the variable chains were fused to defined constant regions and epitope tags (FLAG and hemagglutinin, respectively) (Figures 4C and 4D). Moving forward, marker-free co-selection might be used in tandem to enrich for edited cells (Agudelo et al., 2017).

Figure 4

Native Expression of Defined Sequences from the Heavy and Light Chain BCR Loci

(A) Insertion of an HR template encoding a nanobody yields expression from the heavy chain locus. A FLAG epitope tag was fused to the C terminus of ozoralizumab to enable detection by intracellular flow cytometry.

(B) Quantitation of the proportion of cells isolated from healthy donors expressing ozoralizumab by intracellular flow cytometry.

(C) Multiplexed insertion of the HR templates encoding the heavy chain (HC) and light chain (LC) of adalimumab yields expression of both chains, which were, respectively, fused to C-terminal FLAG and hemagglutinin (HA) epitope tags for detection by intracellular flow cytometry.

(D) Quantitation of the proportion of cells isolated from healthy donors expressing HC, LC, or both by intracellular flow cytometry. Cells were analyzed 6 days post-transfection. In the case of adalimumab insertion, only viable cells were analyzed.

TO, transfection only (no RNP or HDR template). n = 3 independent experiments, mean ± SEM. *p < 0.05, ***p < 0.005 calculated by unpaired Student's t test.

Discussion

To negate the need for repeated administration of recombinant monoclonal antibodies, several groups have strived to develop gene therapy-based approaches (Guijarro-Munoz et al., 2013). Ex vivo approaches involve transduction of allogeneic or autologous cells before adoptive cell transfer, typically as subcutaneous organoids. In vivo approaches include viral and non-viral transfection, typically via intramuscular injection. Viral vectors have been limited by safety concerns and clearance of cells expressing viral antigens; non-viral vectors have been limited by low transfection efficiency.

Vectored immunoprophylaxis represents a particularly interesting strategy to enable expression of broadly neutralizing antibodies to protect against, for example, influenza (Balazs et al., 2013) or HIV (Balazs et al., 2014), but the limitations associated with adeno-associated virus (AAV) vectors are considerable. These include the limited carrying capacity of the vector, modest expression levels of the transgene, a lack of regulation over such expression, and pre-existing and/or rapidly developing immunity to AAV (Brady et al., 2017). Human HSPCs have been engineered to express a broadly neutralizing anti-HIV antibody after in vitro maturation to B cells (Luo et al., 2009), although this was done using lentivirus-mediated gene therapy and therefore did not involve spatial addressing of the transgenes to the relevant loci. Although many cell types have been explored as in vivo bioreactors, the degree of success has been modest using orthogonal cells to produce monoclonal antibodies, which are naturally expressed by B cells. In non-B cells, transgene expression is typically driven by a constitutive promoter, which may not be physiologically appropriate.

It was recently shown that inserting a receptor of interest into its physiologically native locus can greatly improve function. Specifically, directing an HR template encoding a chimeric antigen receptor (CAR) to the T cell receptor (TCR) α constant locus leads to uniform CAR expression in primary human T cells isolated from peripheral blood (Eyquem et al., 2017). Moreover, the potency of these T cells is enhanced, as they confer vastly superior antitumor efficacy to CAR T cells that were generated by conventional retroviral methods or whose CAR was inserted into other loci (or regulated by non-native promoters). These data reveal that HR-mediated reconstitution of the native lymphocyte receptor locus yields optimal regulation of cell surface receptor expression. More recently, it was shown that the specificity of primary human T cells could be reprogrammed by replacing the endogenous TCR with a defined one that targets a known tumor antigen, NY-ESO-1 (Roth et al., 2018). Following recognition of the antigen, the TCR-engineered T cells produced effective antitumor responses both in vitro and in vivo, leading to substantial killing of cancer cells and delayed tumor progression.

Previous studies involving genome editing of B cells have been demonstrative, albeit distinct in scope from this work. The first report involved delivery of Cas9 and gRNA via retro- or lentivirus rather than RNP (Cheong et al., 2016). The authors demonstrated the ability to induce class switch recombination to a desired subclass by cutting in the constant regions, but did not perform HR or modify the variable regions that encode the antibody specificity. The work was performed in primary murine B cells, murine hybridomas, and human B cell lines. The first description of HR in B cells involved engineering of murine hybridomas to express a fusion construct encoding both a light chain and a variable heavy chain, resulting in full-length antibody expression (Pogson et al., 2016). These cells naturally proliferate, rendering HR readily feasible relative to resting B cells.

Editing of primary B cells has also been reported. Using a transgenic Cas9-expressing mouse, it was shown that primary murine B cells could be edited with high efficiency, thereby enabling a small-scale CRISPR-mediated screen to identify genes essential for B cell activation and plasma cell differentiation (Chu et al., 2016). The ability to achieve editing in primary human B cells, necessitating the introduction of Cas9 in addition to gRNA, was recently described (Hung et al., 2018). The method was applied to promote differentiation into plasma cells ex vivo by disrupting genes known to be involved in the regulation of development. The inclusion of either a single-stranded DNA oligonucleotide or AAV containing HR templates afforded HR. The resultant expression of BAFF from the CCR5 locus promoted the engraftment of such edited B cells into immunodeficient mice.

Herein, we sought to focus on the reprogramming of loci of particular interest within B cells, namely, the BCR heavy and light chain loci. Although previous efforts have focused on editing of these loci in murine hybridomas or on other loci in primary murine or human B cells, this work is the first to demonstrate the possibility of defining an antibody among a polyclonal population of cells that comprised heterogeneously recombined V and J segments. We demonstrate that PCR products are suitable HR templates, potentially obviating AAV constructs in primary human B cells. Future work will involve functional studies to confirm the neutralizing ability of the antibodies expressed from these cells, both in vitro and in vivo. Future efforts will also seek to improve editing efficiency. As mentioned above, selection of appropriate V or J segments influences the efficiency of editing, as some are preferentially recombined over others. The observed editing efficiency may also be influenced by cell viability, as cells whose BCR loci are cut but not corrected may not survive in the absence of tonic signaling (Yasuda et al., 2017).

In summary, we have demonstrated for the first time the ability to perform HR-mediated integration in the BCR variable loci of primary human B cells using the CRISPR/Cas9 system. Specifically, we have determined how to improve cell viability following activation and electroporation, and we have shown that a clinically used therapeutic monoclonal antibody can be inserted into and expressed from the endogenous BCR heavy and light chain loci. This method creates opportunities in the study of basic B cell biology as well as the development of cellular humoral vaccines, enabling one to consider vaccination from the perspective of the antibody rather than the antigen. Rather than accept that antibody repertoires are guided by unique germlines, V(D)/J recombination events, and somatic hypermutation in a given individual, it would be desirable to provide an optimal sequence to any person in need of a particular response. Memory B cells (Tangye and Tarlinton, 2009, Kurosaki et al., 2015) could be adoptively transferred (Li et al., 2011), leading to lifelong production of the antibody from long-lived plasma cells (Nutt et al., 2015). Although naturally a stochastic process, differentiation into plasma cells could be promoted by exposure to particular cytokines (Hasbold et al., 2004) or via transfection with appropriate transcription factors (Nutt et al., 2015). A safety switch such as inducible apoptosis, which has been demonstrated to be effective in patients (Di Stasi et al., 2011), may be included to prevent suboptimal outcomes. Clinical translation of genome-edited cellular therapies, whose function and safety can be controlled through synthetic biology (Fischbach et al., 2013), heralds an emerging era in the treatment of genetic, infectious, and acquired diseases (Cornu et al., 2017). Programming memory humoral immunity could have important clinical implications.

Limitations of the Study

The work described herein is proof of concept but would benefit greatly from additional studies. Although demonstration of genome editing at the endogenous BCR heavy and light chain variable regions—leading to expression from the native promoters—is exciting, future work that examines the function of the expressed antibodies will be elucidating. Although nanobody/antibody expression was suggested by preliminary proof-of-concept data, further validation is needed. Unfortunately, such experiments cannot be performed by the authors at this time. Future studies will include quantification of secreted antibody by ELISA with soluble target antigen. Because the antigen selected (TNF-α) has a natural receptor expressed on the surface of primary human B cells, certain studies were not possible. Investigation of antibodies that target orthogonal antigens (e.g., influenza or HIV) would enable assessment of target neutralization. Glycomic profiling will also be informative.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.