Disruption of HIV-1 co-receptors CCR5 and CXCR4 in primary human T cells and hematopoietic stem and progenitor cells using base editing

Abstract

Disruption of CCR5 or CXCR4, the main human immunodeficiency virus type 1 (HIV-1) co-receptors, has been shown to protect primary human CD4⁺ T cells from HIV-1 infection. Base editing can install targeted point mutations in cellular genomes, and can thus efficiently inactivate genes by introducing stop codons or eliminating start codons without double-stranded DNA break formation. Here, we applied base editors for individual and simultaneous disruption of both co-receptors in primary human CD4⁺ T cells. Using cytosine base editors we observed premature stop codon introduction in up to 89% of sequenced CCR5 or CXCR4 alleles. Using adenine base editors we eliminated the start codon in CCR5 in up to 95% of primary human CD4⁺ T cell and up to 88% of CD34⁺ hematopoietic stem and progenitor cell target alleles. Genome-wide specificity analysis revealed low numbers of off-target mutations that were introduced by base editing, located predominantly in intergenic or intronic regions. We show that our editing strategies prevent transduction with CCR5-tropic and CXCR4-tropic viral vectors in up to 79% and 88% of human CD4⁺ T cells, respectively. The engineered T cells maintained functionality and overall our results demonstrate the effectiveness of base-editing strategies for efficient and specific ablation of HIV co-receptors in clinically relevant cell types.

Graphical abstract

HIV co-receptor knockout using nucleases can protect T cells from HIV-1 infection, but can lead to chromosomal rearrangements from double-stranded DNA breaks. Knipping and colleagues use base editors to precisely ablate HIV-1 co-receptor expression. Highly efficient editing makes cells resistant to infection with HIV-1 pseudotyped viral vectors.

Introduction

Human immunodeficiency virus type 1 (HIV-1) causes chronic infection that can be managed by lifelong antiretroviral therapy (ART), but no curative options are currently available. T cells that express the CD4 receptor and one of the co-receptors CCR5 and CXCR4 are the main target cells for HIV-1 infection.^1,2 CCR5-tropic strains use the CCR5 receptor for cell entry, which is expressed by a variety of cells, such as T cells and macrophages, whereas CXCR4-tropic virus strains mainly infect T cells. Most HIV-1 infections are initially caused by CCR5-tropic strains, and CCR5^–/CXCR4^– dual tropic and CXCR4-tropic variants typically arise during the progression of infection.³ When untreated, the dominance of CXCR4-tropic strains leads to a rapid decimation of CD4⁺ T cell counts and ultimately ends in the fatal acquired immunodeficiency syndrome (AIDS).

A naturally occurring 32-base pair deletion (CCR5Δ32) in the CCR5 locus that prevents surface expression in homozygotes has been shown to confer resistance to CCR5-tropic HIV.⁴ This discovery led to the use of bone marrow from CCR5^Δ32/Δ32 donors to simultaneously treat leukemia and a concurrent HIV infection leading to the first functional cures of HIV-1 infections.5, 6, 7 Despite this important milestone in HIV treatment history, allogeneic stem cell transplantation is a high-risk procedure that is further limited by the availability of HLA-matched CCR5^Δ32/Δ32 donors, and is not a widely accessible option for the majority of patients.

Engineering an HIV-resistant immune system could be achieved by the disruption of CCR5 and CXCR4. The first strategy using zinc-finger nucleases (ZFNs) to knock out the CCR5 receptor⁸ was followed by similar gene editing strategies in hematopoietic stem and progenitor cells (HSPCs).⁹ Approaches for knockout of both receptors¹⁰ in T cells as well as studies using transcription activator-like effector nucleases^11,12 or the clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 system13, 14, 15 have also been detailed. Although gene editing tools are able to generate targeted DNA double-stranded breaks (DSBs) in almost any genomic locus of interest, their resolution depends on cell intrinsic repair mechanisms.¹⁶ Besides the desired outcome, DSBs can result in deleterious effects, such as large deletions or translocations, as well as activation of the p53 response.^17,18 Furthermore, DSB introduction at genomic sites other than the target site, called off-target activity, bears the risk of knocking out essential genes or inducing oncogenic transformation (e.g., inactivation of a tumor suppressor genes or translocations).^19,20

Unlike previous genome editing strategies, base editing uses programmable DNA-binding proteins to direct deaminases that modify targeted nucleotides in the genome, without requiring the introduction of DSBs.21, 22, 23, 24, 25 Cytosine base editors (CBEs)²¹ that convert C⋅G to T⋅A base pairs and adenine base editors (ABEs)²² that convert A⋅T to G⋅C base pairs have been employed and expanded to include numerous base editors with a diversity of editing windows, sequence context preferences, and PAM compatibilities using deaminases and DNA-binding domains with different properties.²⁴ Base editors are attractive tools for disease prevention applications by knocking out genes, such as CCR5 and CXCR4, because of their ability to generate precise editing outcomes with reduced risks of DSB-mediated byproducts or undesired cell-state changes induced by DSBs. The disruption of genes using base editors has previously been demonstrated by introducing premature stop codons^26,27 or eliminating start codons.²⁸ Here, we describe the use of CBE mRNA to install premature stop codons at the CCR5 and CXCR4 loci in primary human CD4⁺ T cells. We also delivered ABE mRNA to eliminate the CCR5 start codon in T cells and CD34⁺ HSPCs. The efficient disruption of co-receptor surface expression protected primary human CD4⁺ T cells from transduction with HIV-1 pseudotyped viral vectors and did not perturb T cell function. Our analysis of genome-wide gRNA-dependent off-target activity revealed off-target editing predominantly in intergenic and intronic genomic regions.

Results

Design of base editor target sites to disrupt HIV co-receptor expression

CBE and ABE (Figure S1A) have been shown to mediate edits with high precision, and low frequencies of indels and translocations relative to Cas9 nuclease-mediated disruption of genes.29, 30, 31, 32 CBEs can introduce premature stop codons by converting the codons coding for glutamine (CAA or CAG), arginine (CGA), or tryptophan (TGG) residues to TAA, TAG, or TGA.^26,27 Aiming to disrupt the essential HIV co-receptors CXCR4 and CCR5 by installing stop codons, we designed gRNAs in both CCR5 and CXCR4 for defining the optimal base editor candidate(s) (Figures 1 and S1A).

Figure 1.

Blocking HIV infection by introducing premature stop codons or eliminating start codons in HIV co-receptor genes using base editing

(A) Illustration of CXCR4 and CCR5 loci with locations of gRNA binding sites. Arrowheads indicate if the respective target base is located on the coding (right-pointing arrow) or non-coding (left-pointing arrow) strand. Boxes represent exons, colored portions represent protein-coding regions, and lines represent introns. (B) Sequences of candidate gRNAs for CXCR4 and CCR5 loci targeting and corresponding applied base editors. Underlined bases are the target bases, stop codons are highlighted in red, start codons are highlighted in green. When two bases are underlined, editing of both or either results in a stop codon. (C) Percentage of sequence reads with edited target base or (D) insertions and deletions (indels) at target sites analyzed by high-throughput sequencing on day 4 after electroporation. Analysis was done on day 4 with replicates from two to eight independent experiments in T cells from one to five independent donors. Data represented as mean ± SD. Indels, insertions and deletions.

Two gRNAs each in CXCR4 (gRNAs X4-1 and X4-2) and CCR5 (gRNAs R5-1 and R5-2) were designed to install premature stop codons with BE4max (Figures 1A and B), an optimized CBE³³ containing an APOBEC1 deaminase that requires the target base to be in the main editing window at positions 3 to 8 (counting the PAM as positions 21–23) as well as an NGG PAM downstream of the protospacer (Figure S1B). Although BE4max is a widely used base editor, these editing strategies were not ideal because the premature stop codon generated by R5-2 is proximal to the C terminus of CCR5, so a functional co-receptor may still be produced.³⁴ The R5-1 gRNA has only a single mismatch to the related CCR2 gene, which is an important chemokine receptor for monocytes, so this gRNA candidate could lead to detrimental off-target editing outcomes as has been observed previously.^8,35,36 Thus, we designed three additional gRNA candidates (gRNAs R5-3, R5-4, and R5-5) for use with BE4max containing SpCas9-NG³⁷ nickase (referred to as BE4max-NG), rather than SpCas9 nickase, and three gRNAs (gRNAs R5-6, R5-7, and R5-8) for use with evoCDA-BE4max (referred to as evoCDA, the SpCas9-NG variant was used in this study), which comprises an evolved CDA1 deaminase and has a wider editing window approximately at protospacer positions 1 to 13 (Figures 1A and S1B).³⁸ Notably, for gene knockout applications, bystander edits near the target nucleotide are generally not a concern, thus allowing the use of base editors with wide(r) activity windows.

ABEs have also been implemented for gene silencing by mutating the start codon from ATG to GTG or ACG.²⁸ Therefore, we used ABE8e, an ABE that contains a laboratory-evolved deoxyadenosine deaminase with substantially increased editing activity as compared with previous ABE variants,³¹ to eliminate the CCR5 initiating start codon. For this purpose, three gRNAs were designed, two for use with ABE8e SpCas9-NG³¹ (R5-9 and R5-10) and one for use with ABE8e Cas9-NRCH³⁹ (R5-11) (Figures 1A, 1B and S1). Because CXCR4 isoforms can use one of several potential start codons,^40,41 this ABE8e strategy was not readily applicable to CXCR4.

Base editors are highly efficient in primary human T cells with minimal insertions and deletions

Activated primary human T cells were electroporated with base editor mRNA and gRNAs and base-editing efficiency and indel frequency were determined by high-throughput sequencing (HTS). Editing efficiency varied based on the target site, but one or more gRNAs that yielded >80% of the desired edit after electroporation were found for each gene (Figure 1C). The frequency of indels after editing was generally lower when using BE4max and ABE8e, with slightly higher indel frequencies for evoCDA (Figure 1D).

For the CXCR4 locus both gRNAs showed efficient base editing, but X4-2 resulted in higher editing efficiency with 89.1% ± 2.6% of sequences containing the desired edit and lower indel frequency (0.6% ± 0.1%) compared with X4-1 (Figure 1C,D).

A comparison of CCR5 sites targeted with BE4max shows higher editing efficiencies (68.4% ± 8%) and lower indel frequencies (1.5% ± 0.2%) for R5-1 compared with R5-2 with only 37.1% ± 11.5% of sequences with the desired C⋅G to T⋅A conversion and 3.9% ± 0.1% of sequences with indels. Editing frequencies were generally lower for BE4max-NG as compared with all other base editors and ranged from 10.8% ± 7.4% (R5-5) to 36.5% ± 13.3% (R5-4). EvoCDA editing was efficient at the R5-6 and R5-7 target sites; however, the R5-6 gRNA appeared to hamper cell growth compared with R5-7 and so was not pursued further (Figure S2). Thus, R5-7 was the lead CBE editing candidate at the CCR5 locus with 75% ± 6% base editing and 5.1% ± 2% of sequences with indels (Figures 1C and 1D).

The ABE8e start codon editing strategy showed the highest editing frequencies of all compared sites with 95% ± 1.2% of sequences with A⋅T to G⋅C conversion for R5-9 using ABE8e-NG followed by 86.4% ± 4.9% for R5-11 with ABE8e-NRCH. All ABE8e target sites showed low indel frequencies between 0.6% ± 0.1% (R5-11) and 1.8% ± 0.6% (R5-9) (Figures 1C and 1D). These data show that both CXCR4 target sites (X4-1 and X4-2) as well as two CCR5 gRNAs for use with CBE (R5-1 and R5-7) and ABE (R5-9 and R5-11), respectively, yielded the highest editing efficiency. Therefore, we pursued these six gRNAs as the candidates for testing in the remainder of the study.

Base editors show low frequency of off-target editing and indels

To identify the most specific of these gRNA candidates, we experimentally identified off-target sites and conducted targeted HTS of the top off-target candidates. We employed “circularization for in vitro reporting of cleavage events” (CIRCLE-seq)⁴² as a platform for the identification of potential off-target sites due to its high sensitivity. Although CIRCLE-seq is based on in vitro cleavage of genomic DNA by Cas9 nuclease, and off-target identification pipelines have recently been developed specifically using base editors,^43,44 our comparison of the published data determined that CIRCLE-seq can identify base editor off-target sites with the greatest sensitivity and provide the largest list of potential sites to inform future clinical studies. We conducted CIRCLE-seq using the gRNAs X4-1, X4-2, R5-1, R5-7, R5-9, and R5-11, with the complete output from the off-target identification pipeline provided in Figure S3. From these, the 12 sites with the highest detected CIRCLE-seq read counts were PCR amplified for assessment by HTS after editing primary CD4⁺ T cells. Four of the top 12 predicted sites for the R5-1 gRNA did not have a substrate cytidine nucleotide in the protospacer sequence (OT3, OT6, OT8, and OT10), so three additional sites were sequenced (OT13, OT14, and OT15). One additional site was also sequenced for R5-11, because the 12th predicted off-target site was detected with the same number of reads as the 13th.

For X4-1 BE4max, off-target base editing was confirmed at one intronic site (OT5) with 22.4% ± 1.8% and <0.5% indels, with no editing detected at the other 11 examined sites. Similarly, X4-2 showed off-target editing at one intronic site (OT8) with an editing frequency of only ∼1.1% ± 0.2%, no detectable indels, and no editing detected at the other 11 examined sites (Figures 2 and S4). Base editing was confirmed at three R5-1 off-target sites with OT1 in a CCR2 exon, generating a stop codon with a frequency comparable with the target site. Indels were detected at OT3, which has no cytosine present in the protospacer (Figure S4). Base editing was detected at nine off-target sites for R5-7 evoCDA ranging from 0.3% ± 0.03% (OT5) to 39.9% ± 3.9% (OT8). Four of the confirmed off-targets are located in intergenic regions and the remaining five in introns distant from splice donor and acceptor sites. The off-target indel frequencies measured following evoCDA treatment were generally higher than for the other used base editors with the highest indel frequency at OT8 (6.6% ± 2.1%) (Figure S4).

Figure 2.

High-throughput sequencing analysis of genome-wide off-target positions identified by CIRCLE-seq

Potential gRNA-dependent off-target positions were identified using CIRCLE-seq and 12 to 15 of these target sequences were examined in base-edited T cells by high-throughput sequencing alongside the respective target site. (A) Percentage of sequences with C•G to T•A and A•T to G•C conversion at analyzed sites in base editor-treated and negative control samples are shown. The x axis defines the on- or off-target candidate site and the position of the putative base edit numbered in relation to the PAM for either cytosine (C) or adenine (A). For sequences with multiple bases amenable for editing (C or A, respectively) only the one with the highest editing efficiency is graphed. When no base is edited with >0.1%, the position that is closest to the target position is displayed. For base-editing efficiencies across the entire protospacer see Figure S4. (B) Sequence and genomic location of target sites and off-target sites with mean base-editing frequencies of >0.1% analyzed by high-throughput sequencing. Red letters represent mismatches to the target site. Sequencing analysis for each site was done on replicates from three independent experiments performed in CD4⁺ T cells from three unrelated donors (two donors for X4-2) and the mean and standard deviation are graphed. OT, off-target; ON, on-target site.

For R5-9 ABE8e-NG, off-target editing was detected at four intergenic and two intronic sites with editing frequencies of 0.2% ± 0.1% (OT11) to 10.9% ± 1.8% (OT10) (Figures 2 and S5). R5-11 ABE8e-NRCH off-target base editing was confirmed at two intergenic sites, but with only up to 1.4% ± 0.3% editing efficiency at OT13. Due to the largely overlapping R5-9 and R5-11 target site, 2 of the top 12 CIRCLE-seq sites coincided for R5-9 and R5-11, and one of these sites is a confirmed off-target site for both candidates (R5-9: OT10; R5-11: OT8). Congruent with extremely low indel frequency at the target sites, we did not detect indels at any of the analyzed off-target sites for R5-9 and R5-11 (Figure S4). Taken together, our specificity analysis observed variable off-target editing between gRNAs. One verified off-target of the R5-1 gRNA introduces a nonsense mutation, and so this candidate was excluded from further analysis. All other identified off-target events occurred in intergenic and intronic regions with no predicted detrimental implications.

Efficient introduction of stop codons and elimination of start codons reduces CCR5 and CXCR4 surface expression

After quantifying editing frequencies and off-target activity, we sought to examine the effect of our optimized editing strategy on cell surface receptor expression. Primary human CD4⁺ T cells were electroporated with the most efficient gRNA candidates along with base editor mRNA, then CXCR4 and CCR5 surface expression was analyzed after 4 and 11 days using flow cytometry (Figure 3A).

Figure 3.

Analysis of HIV co-receptor surface expression after base editing in primary CD4⁺ T cells

(A) Experimental timeline for editing and analysis of human CD4⁺ T cells. (B and C) Percentage of base-edited and untreated control T cells expressing CXCR4 (B) and CCR5 (C) on their surface on days 4 and 11 after electroporation. Shown are flow cytometry data from replicates from at least three separate experiments performed in CD4⁺ T cells from multiple, unrelated donors (three donors for X4-1, four donors for R59 and R511, six donors for R5-7, and seven donors for X4-2). Data in all graphs are represented as mean ± SD. See Figure S6 for representative flow cytometry data.

On day 4 after electroporation, X4-1 and X4-2 editing led to a reduction of surface CXCR4 expression by 75% and 86%, respectively (Figure 3B). Consistent with previous studies, we observed variability in CCR5 expression of unmodified control cells.45, 46, 47 Activation of T cells using CD3/CD28 beads and cytokines resulted in an elevated percentage of CCR5⁺ cells, particularly in unedited controls on day 11 after electroporation. Each of the tested gRNAs led to a stable reduction of CCR5 expression by 78%–89% on day 4 and 80%–94% on day 11 (Figures 3C and S6).

Tandem knockout of both co-receptors could protect CD4⁺ T cells from both CCR5^– and CXCR4-tropic HIV infection and thus would be of high therapeutic relevance. Toward achieving this goal, we co-delivered the most efficient CBE gRNAs at each locus (X4-2 and R5-7) along with BE4max or evoCDA (Figure 4A). The knockout efficiency for each gene was comparable with single gene knockout. The percentage of double-negative cells after editing with X4-2, R5-7 and evoCDA or BE4max increased from 1.5% ± 0.9% to 79.7% ± 7% (evoCDA) and 68.7% ± 7.9% (BE4max) on day 4 and from 2.9% ± 2.4% to 75% ± 8.3% (evoCDA) and 66.1% ± 9% (BE4max) on day 11 after electroporation (Figures 4B, 4C, and S7). The slight increase of unedited cells at day 11 is attributed to a modest accumulation of CCR5 surface expression in unedited cells that is observed with longer culture periods. Dual-edited T cells underwent further molecular and functional characterization. Quantitative RT-PCR revealed a concomitant loss of CXCR4 mRNA, while CCR5 mRNA was slightly elevated compared with controls (Figure 4D). These data suggest that the BE4max installed premature stop codon triggers nonsense-mediated decay while CCR5 editing may not. Regardless, functional receptors were lost from the cell surface as determined by flow cytometry. To assess whether editing impacted T cell function, control and dual-edited T cells were exposed to phorbol 12-myristate 13-acetate (PMA) and ionomycin and analyzed for cytokine production. We observed that both control and edited cells upregulated IL-2, IL-4, IFN-γ, and TNF-α in a similar fashion to one another (Figure 4E,F) showing that tandem edited T cells retain their functionality.

Figure 4.

Efficient simultaneous disruption of CCR5 and CXCR4 in primary CD4⁺ T cells

(A) Human CD4⁺ T cells were electroporated with both X4-2 and R5-7 gRNAs for multiplex editing using either evoCDA or BE4max base editor mRNA. Guide RNA sequences with target cytosines for stop codon introduction (underlined) are depicted and editing windows of the applied BE4max or evoCDA base editors are highlighted in light and dark blue, respectively. (B) Percentage of dual-edited CCR5^-/CXCR4^- CD4⁺ T cells on days 4 and 11 after treatment with X4-2, R5-7 gRNAs with evoCDA or BE4max. Data are represented as the mean ± SD of replicates from five independent experiments and donors. (C) Representative flow cytometry plots from experiments in (B) showing percentage of dual-edited CD4⁺ T cells (see Figure S7 for day 11 flow cytometry plots from the same experiment). (D) Quantitative RT-PCR of dual-edited T cells. CXCR4 and CCR5 transcript levels were measured in control and edited cells and normalized to the GAPDH house-keeping gene. Data are the mean and standard deviation of three separate donors analyzed in duplicate. (E and F) Cytokine analysis. Control (red shaded) or dual-edited (green shaded) cells were left unstimulated or stimulated with PMA/ionomycin and analyzed by FACS for the intracellular production of IL-2, IL-4, IFN-γ, and TNF-α. Representative histograms are shown in (E) and the mean fluorescence intensity ± SD from three individual donors is shown in (F).

Protection from infection with CCR5^– and CXCR4-tropic HIV-1 pseudotyped viral vectors

For an assessment of the ability of co-receptor knockout cells to protect against HIV infection, we generated green or red fluorescent protein (GFP, RFP)-expressing replication-incompetent lentivirus (LV) with CCR5^– or CXCR4-tropic HIV-1 envelopes. Primary human CD4⁺ T cells were treated with base editors and individual gRNA candidates or co-delivered X4-2 and R5-7. Subsequently, cells were transduced with the CCR5-tropic virus and/or the CXCR4-tropic virus and the infection efficiency of treated and untreated cells was assessed by flow cytometry of GFP and RFP expression, respectively (Figures S8 and S9).

The frequency of RFP-expressing unedited control cells was 50.6% ± 7.1%, and up to 88% protection from infection was achieved in edited cells. The frequency of RFP-expressing cells was reduced to 7.7% ± 1.8% using X4-2 treatment alone, 5.9% ± 2.2% when both X4-2 and R5-7 were co-delivered with evoCDA, and to 9.5% ± 2.3% when both X4-2 and R5-7 were co-delivered with BE4max (Figures 5B and 5C). CCR5-expression was also disrupted in multiplex editing samples (Figure 5B, right). Similarly, transduction with CCR5-tropic virus was significantly reduced in all edited samples by up to 86% as compared with unedited control samples, which showed 42.6% ± 7.8% of GFP⁺ cells (Figure 5D). The most efficient reduction in CCR5 expression and infection with CCR5-tropic virus was achieved using R5-9 resulting in 4.9% ± 1.7% of GFP⁺ cells. Co-delivery of R5-7 with X4-2 along with evoCDA led to comparable reduction of CCR5 surface expression and viral entry as R5-7 alone (10.9% ± 2.6% and 8.9% ± 3.7% of GFP⁺ cells respectively; Figures 5D and 5E). CXCR4 disruption was also efficient in multiplex editing samples (Figure 5D). These results show that our single or multiplex editing approaches are able to protect primary human CD4⁺ T cells from transduction with CCR5^– or CXCR4-tropic viral vectors. Collectively, these data demonstrate robust tandem receptor disruption leading to resistance to virus uptake in a manner that preserves T cell function.

Figure 5.

Base editing protects CD4⁺ T cells from transduction with CCR5-tropic or CXCR4-tropic HIV pseudotyped viral vectors

(A) Schema for editing, transduction, and analysis. (B) Tr712-LV-RFP is a replication-incompetent CXCR4-tropic viral vector that results in RFP expression upon transduction of CD4⁺CXCR4⁺ T cells. The first bar graph shows the percentage of single- or dual-edited CD4⁺ T cells expressing CXCR4 and RFP 2 days after Tr712-LV-RFP transduction, detected by flow cytometry. CCR5 expression of these samples is depicted in the top right bar graph, measured by flow cytometry. Data represented as mean ± SD of three or five independent experiments done in CD4⁺ T cells from different donors each. (C) Flow cytometry data representative for results graphed in (B) (see Figure S8 for gating strategy and additional flow cytometry data). (D) BaL-LV-GFP is a GFP-encoding replication-incompetent CCR5-tropic viral vector that requires cell surface expression of CD4 and CCR5 to enter cells. The graph at left shows the percentage of single-edited, dual-edited, and control CD4⁺ T cells expressing CCR5 and GFP 2 days after transduction with BaL-LV-GFP, detected by flow cytometry. CXCR4 expression of the same samples is depicted in the right bar graph, measured by flow cytometry. Results are represented as mean ± SD of three to five independent experiments performed in primary human CD4⁺ T cells from five different donors. (E) Representative flow cytometry data for the results graphed in (D) (see Figures S8 for gating strategy and additional flow cytometry data). p values were calculated using Student’s unpaired, two-sided t test to compare transduction efficiency of treated and control cells (n.s. p > 0.05, ∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, ∗∗∗∗p ≤ 0.0001). GFP, green fluorescent protein; RFP, red fluorescent protein.

CCR5 start codon editing leads to efficient CCR5 knockout in HSPCs

Due to the previously reported therapeutic benefit of CCR5 null HSPCs, which can be used as the basis for allogeneic or autologous bone marrow transplantation, we sought to determine whether our base-editing approach to knock out CCR5 is effective in human mobilized CD34⁺ HSPCs.^5,6,9,48 Because of the essential role of CXCR4 in stem cell maintenance,^49,50 a multiplex knockout approach in HSPCs is prohibitive. Therefore, we delivered the most efficient CCR5 gRNA candidates, R5-9 or R5-11, along with ABE8e-NG or ABE8e-NRCH mRNA, respectively, to mobilized CD34⁺ HSPCs (Figure 6A). Cells maintained high viability (Figure S10) and base-editing and indel frequencies were analyzed by HTS. The most efficient editing was achieved with R5-9 and ABE8e-NG, averaging 87.5% ± 7.6% with an indel frequency of 0.9% ± 0.02%. Editing with R5-11 and ABE8e-NRCH was less efficient, averaging 60.7% ± 2.7% with an equally low indel frequency of 0.9% ± 0.1% (Figure 6B).

Figure 6.

Base editing in mobilized human CD34⁺ hematopoietic stem and progenitor cells

(A) Experimental approach and timeline for base editing and analysis of HSPCs. (B) Percentage of sequences containing the respective target base edit or insertions and deletions (indels) at target sites analyzed by high-throughput sequencing. Analysis was done with replicates from three independent experiments in mobilized CD34⁺ cells from different donors each. Data are represented as mean ± SD. (C) Control and treated sample colony-forming unit (CFU) assay in methylcellulose. Six days after electroporation 100 cells were plated for each sample and colonies from erythroid (BFU/CFU-E) and granulocyte/macrophage (CFU-G/M/GM) progenitors were defined and counted 14 days after plating. Controls were cells that were electroporated with GFP mRNA and with no RNA (pulse control). Data are shown for three donors as the total number of colonies for either CFU-G/M/GM or BFU/CFU-E. (D) Colony sequencing. Individual colonies from CFU assays were obtained for DNA isolation and PCR to assess editing frequency via Sanger sequencing. Twenty-four total colonies were isolated for each lineage for each editor (two individual donors each). Data are plotted as total number of colonies that showed editing. Editing was defined as bi-allelic if the Sanger sequencing base call was 100%, mono-allelic if the distribution between edited and unedited was 50%. Mixed was defined as editing efficiency that was between 20% and 40%. (E) HSPC off-target analysis. Bulk electroporated HSPCs were analyzed for on-target and the CIRCLE-seq off-targets identified/screened in Figure 2 using the R5-9 (left panel) or R5-11 gRNA, respectively (right panel). For R5-9, long-term progenitors with a phenotype of CD34⁺CD45RA^–CD90⁺CD133⁺ were sorted to purity (Figure S11) and also screened for off-target editing. Data are from five donors for R5-9 and four from R5-11 (three controls for R5-11) are plotted as percent total reads with A > G editing ± SD.

We next used edited HSPCs in methylcellulose-based colony-forming unit assays (CFUs) and observed that the total number of colonies as well as the proportions of myeloid (CFU-M/G/GM) and erythroid (BFU-E, CFU-E) colonies was comparable in treated and untreated samples (Figure 6C). DNA was obtained from individual colonies and Sanger sequenced to assess editing. Similar to the bulk CD34 data in Figure 6B, we observed that editing with R5-9 was more efficient than R5-11 in single colonies of both the CFU-GM and BFU-E lineages (Figure 6D).

Next we analyzed on-target and off-target editing in the bulk and long-term repopulating (CD34⁺CD45RA^–CD90⁺CD133⁺) fraction⁵¹ of cells (Figure S11). Because R5-9 was the most efficient gRNA we focused on this population for sorting for long-term progenitors. Target site base editing was 80.8% ± 5.5% and CIRCLE-seq identified off-target sites observed in T cells also manifest in HSPC (OT1, OT9, and OT10), albeit at lower frequency (Figures 2 and 6E). The corresponding low indel frequency for both on- and off-target data (Figure S12) show that CCR5 start codon editing with our ABE strategy leads to highly efficient CCR5 targeting in primitive HSPCs that retain their lineage commitment potential.

Discussion

CCR5 receptor knockout in primary human T cells using designer nucleases has been demonstrated in multiple studies.^8,11,12,52 Eight completed and ongoing clinical trials employing ZFNs in autologous T cells (summarized in Ashmore-Harris and Fruwirth⁵³) reported engraftment and a survival advantage of modified cells, partly improved control of viral replication, and general safety of the treatment with only one treatment-related adverse event. Except for a potentially increased risk for West Nile virus infection, CCR5-deficiency has not been associated with significant health impairment.⁵⁴ As some HIV isolates are able to use CXCR4 instead of CCR5 or both as a co-receptor to infect cells, deficiency of both receptors would be ideal to protect T cells from HIV infection. Although the effects of CXCR4 deficiency on human T cells is not fully characterized, previous CXCR4 knockout studies reported no evident detrimental effects.^10,14,15 In addition, T cell-specific CXCR4-deficient knockout mice were shown to develop normal humoral and cellular immune responses.⁵⁵ Simultaneous knockout of both receptors has been shown in primary human T cells using gene editing (ZFN¹⁰ and CRISPR-Cas9 system^14,15). In this study, we aimed to circumvent the risks associated with DSB-mediated knockout by applying base editing to generate CCR5^– or/and CXCR4^– T cells. The low indel frequency caused by base editing results in a minimal p53 DNA damage response and reduces the risk of large deletions and translocations, especially for multiplex editing strategies.¹⁷ Using the CBEs BE4max and evoCDA, we observed precise and highly efficient introduction of stop codons at the CXCR4 and CCR5 loci in primary human CD4⁺ T cells. In agreement with previous work, we observed that overall editing frequencies were lower for BE4max-NG.³⁸ Adenine base editing using ABE8e to eliminate the CCR5 start codon was the most efficient approach and led to up to 95% ± 1.2% of sequences containing the desired A⋅T to G⋅C transition in primary human CD4⁺ T cells. The detected frequencies of target site indels, which are presumed to result from base excision repair and nicking of the non-edited strand, was <2% for most candidates.²⁴ EvoCDA caused elevated indel frequencies with up to 7.6% ± 3% of sequences with indels consistent with previous observations.³⁸

Using our lead gRNA candidates (X4-1, X4-2, R5-7, R5-9, and R5-11) we demonstrate successful and stable co-receptor knockout at the surface of primary human CD4⁺ T cells. The multiplex approach using X4-2 and R5-7 led to comparable knockout frequencies as when targeting only a single gene. Base editing to knock out CCR5 and CXCR4 resulted in protection from infection with lentiviral vectors with HIV envelopes at a similar frequency to the editing frequencies observed. Multiplex edited cells were also resistant to infection from both pseudotyped LVs, indicating that this efficient editing approach can generate T cells with resistance against infection. The dual-edited cells also showed the production of cytokines upon stimulation, demonstrating their functionality.

To assess whether any of the editing approaches tested in this study resulted in off-target editing, we used CIRCLE-seq to identify potential off-target sites and conducted HTS of the top predicted sites for each. Off-target editing was only observed in a single exon, and led to a premature truncation of CCR2, which shares homology with the target site in CCR5. The remaining off-target sites that were verified to be edited by the other five editing strategies did not appear to lead to any coding or regulatory change. These findings reinforce the value of conducting an assessment of gRNA-dependent off-target genome editing. RNA editing and gRNA-independent off-target editing56, 57, 58 have also been described in base-edited cells. These undesired outcomes may be reduced through the use of deaminase variants and/or by optimizing the duration and dose of treatment.^31,56,58 When delivering base editors via mRNA electroporation, as in this study, RNA expression is transient potentially diminishing off-target effects.

Because T cells have a finite lifespan, a truly curative treatment would require HIV-resistant HSPCs that give rise to resistant CD4⁺ T cells and other cell types that can be infected by HIV, such as CD4⁺ monocyte-derived macrophages. Because CXCR4 is critical for sustaining normal physiological function and homing of HSPCs,^49,50 it is not presently feasible in HSPCs to ablate this receptor. Potential options for addressing this are installation of “functional variants” that maintain CXCR4 homing functions while removing/minimizing HIV entry potential.⁵⁹ Therefore, we focused our efforts for CCR5-knockout with BEs in HSPCs, such as has been pursued using the CRISPR-Cas9 system¹³ and ZFN delivered as DNA,⁶⁰ adenoviral vector,⁶¹ or as mRNA,⁹ which are currently being evaluated in clinical trials (NCT02500849⁹ and NCT03164135⁶²). Mouse studies have shown no difference of engraftment potential between CCR5 ZFN mRNA-treated and untreated CD34⁺ cells.⁹ To mimic the homozygous CCR5Δ32 phenotype4, 5, 6 that mediates protection from CCR5-tropic HIV-1 infection by disruption of CCR5, we used our ABE strategy to eliminate the CCR5 start codon in CD34⁺ HSPCs. ABE8e base editing was highly efficient in CD34⁺ cells and the treated cells showed high viability with no skewing in colony formation as compared with untreated control cells, indicating that editing does not disrupt the hematopoietic potential of HSPCs. Furthermore, the feasibility of base editing in human peripheral-blood-mobilized CD34⁺ HSPCs has been demonstrated using CBE RNPs⁶³ and primary and secondary transplantation experiments revealed multilineage engraftment of edited human hematopoietic cells in mice, suggesting that this strategy can successfully edit long-term HSCs. Indeed, our fluorescence-activated cell sorting (FACS)-based sorting of long-term repopulating HSPCs showed highly efficient editing with minimal OT events.

Using healthy donor T cells we show highly efficient and specific HIV co-receptor CXCR4 and CCR5 disruption using base editing. Our approach would support autologous cell engineering given that HIV patient T cells have shown the ability to undergo large-scale expansion for adoptive cell therapy.^65,67 A key consideration is receptor knockout efficiency and fitness of the engineered cells. We observed robust and simultaneous inactivation of each receptor in a fashion that did not significantly impact cytokine production, cell viability, or expansion. Despite this, our strategy, as is the case for published data to date, does not mediate complete receptor ablation. Under adoptive therapy conditions this would result in unedited T cells able to be infected with HIV; however, the predominance of edited cells would be predicted to promote selection/expansion of the resistant (i.e., edited) cells.

Because engineered T cells do not result in elimination of the latent viral reservoir⁶⁷ HSPC strategies are attractive given that the only patients to be cured of HIV have received bone marrow transplantation with CCR5-deficient cells.^5,48,69 Therefore, we assessed DNA base editing in HSPCs and observed robust CCR5 ablation with a low incidence of off-target events. Collectively, our base-editing strategy in primary human T cells and HSPCs has significant translational application.

Materials and methods

Cell culture

Peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation using Lymphoprep and SepMate tubes (STEMCELL Technologies, Vancouver, Canada) from buffy coats obtained from Memorial Blood Centers in St. Paul, MN, without any identifying donor information. T cells were isolated from PBMCs using the EasySep Human T Cell Isolation Kit or the EasySep Human CD4⁺ T Cell Isolation Kit (both STEMCELL Technologies). Prior to electroporation, T cells were activated for 2 days with Dynabeads Human T-Expander CD3/CD28 beads (Thermo Fisher Scientific, Waltham, MA) and cultured in X-VIVO 15 Serum-free Hematopoietic Cell Medium (Lonza, Basel, Switzerland), supplemented with 5% AB human serum (Valley Biomedical, Winchester, VA), GlutaMAX (Gibco, Waltham, MA), N-acetyl-cysteine (Sigma Aldrich, St. Louis, MO), 50 U/mL penicillin, 50 μg/mL streptomycin (Gibco), 300 IU/mL IL-2, and 5 ng/mL recombinant human IL-7 and IL-15 (PeproTech, Cranbury, NJ). Cryopreserved CD34⁺ cells with greater than 90% purity, enriched from PBSC mobilized adult donors and selected on the Miltenyi CliniMACS system were obtained from the Core Center for Excellence in Hematology at the Fred Hutchinson Cancer Research Center without any identifying donor information. Cells were cultured in HSPC medium (SFEM II; STEMCELL Technologies) with 100 ng/mL each of stem cell factor (BioLegend, San Diego, CA), recombinant human thrombopoietin (BioLegend), recombinant human Fms-related tyrosine kinase 3 ligand (PeproTech), recombinant human IL-6 (PeproTech), 50 U/mL penicillin, and 50 μg/mL streptomycin (Gibco), 0.75 μM StemRegenin1 (STEMCELL Technologies) and 500 nM UM729 (STEMCELL Technologies). All cells were cultured at 5% CO₂ and 37°C.

Base editor mRNA production

BE4max mRNA was purchased from Aldevron (Fargo, ND) and included Cap1 addition and standard nucleotide triphosphates. All other base editors were produced as follows by TriLink BioTechnologies (San Diego, CA): mRNA was transcribed in vitro from PCR product using full substitution of uridine for an analog nucleotide: 5-methoxyuridine for CBEs mRNA or N1-methylpseudouridine for ABEs mRNA. mRNA was capped co-transcriptionally using CleanCap AG analog (TriLink BioTechnologies) resulting in a 5’ Cap 1 structure. In vitro transcription reaction was performed as described previously^64,66 and mRNAs were purified using RNeasy kit (QIAGEN). Mammalian-optimized UTR sequences (TriLink BioTechnologies) and a 120 base poly-A tail were included in the transcribed PCR product.

Pseudotyped virus production

The CCR5-tropic pseudotyped LV encoded for GFP to monitor infection and the envelope was derived from the HIV BaL isolate.⁶⁸ Virus particles were produced using a third-generation packaging system as described before.¹¹ The CXCR4-tropic pseudotyped virus contained a truncated 712 HIV (Tr712) envelope (pNL4-3^{BH10env 70} with a stop codon at position 713, lacking 144 amino acids⁷¹) and encoded RFP. Virus particles with Tr712 envelopes were produced by transient transfection of HEK293T cells. Transfection was performed by using the classical calcium phosphate method (solution A: 280 mM NaCl, 10 mM KCl, 1.5 mM Na₂HPO₄, 12 mM D-glucose, and 20 mM N-2-hydroxyethylpiperazine-N9-2-ethanesulfonic acid [HEPES] adjusted to pH 7.2 with NaOH; solution B: 2 mM CaCl₂). For display of the Tr712 envelope, 5 μg of the envelope plasmid was transfected together with 8.6 μg Gag-Pol packaging plasmid (8.91) and 8.6 μg of a plasmid encoding a self-inactivating LV-expressing RFP under the control of the spleen focus foamy virus promoter ([SIN]-HIVSFFVRFP). CXCR4^– and CCR5-tropic lentiviral particles were titrated on CD4⁺ Jurkat cells (Jurkat cells transduced with a CD4-encoding lentivirus and selected using puromycin) or PM1 cells (obtained through the NIH AIDS Reagent Program, Division of AIDS, NIAID, NIH⁷²), respectively, by adding serial dilutions of the viral supernatants. The quantification of infectious particles was calculated using RFP or GFP fluorescence determined by flow cytometry.

Electroporation and transduction

Two days after CD3/CD28 bead-activation, 3–3.5 × 10⁵ T cells were electroporated with 2 μg base editor mRNA and 30 pmol of synthetic gRNA (Synthego, Redwood City, CA; see Table S1 for gRNA sequences), 1 μg CleanCap EGFP mRNA (TriLink BioTechnologies) or without adding RNA (pulse control) using the Neon Transfection System (Thermo Fisher Scientific). Cells were resuspended in resuspension buffer T and electroporated in 10 μL Neon tips using the following electroporation parameters: 1,400 V, 10 ms, 3 pulses. Subsequently, cells were cultured in 24-well plates and T cell medium was added 2–3 days after electroporation. On day 4 after electroporation, cells were counted and viability was assessed using a Countess II FL Automated Cell Counter (Thermo Fisher Scientific). Cells were harvested for DNA isolation and flow cytometry on days 4 and 11 after electroporation. For transduction experiments, 96-well plates were coated with 3 μg/well of RetroNectin (recombinant human fibronectin fragment; TaKaRa Bio, Mountain View, CA) at 4°C overnight or at room temperature (RT) for 2 h. The RetroNectin was aspirated and the plate was incubated with 75 μL phosphate-buffered saline (PBS)/2% bovine serum albumin (Gibco) for 30 min at RT and washed with PBS once. The viral vector was diluted in 75 μL T cell medium (0.6–1 transduction particles per cell) and added to a well of the coated plate. T cell medium without virus was added to control wells and the plate was centrifuged for 2 h at 1,500 × g and 32°C. Cells were counted 5 days after base editor electroporation and 5 × 10⁴ cells were added to a well of the 96-well plate containing the viral vector. The plate was centrifuged for 5 min at 300 × g and cells were cultured for 2 days before flow cytometric analysis. CD34⁺ cells were thawed and cultured in HSPC medium for 2 days. A total of 1–2 × 10⁵ cells was nucleofected with 2 μg base editor mRNA and 50 pmol gRNA (Synthego) in Nucleocuvette Strips with the Amaxa 4D-Nucleofector (Lonza) using the P3 Primary Cell 4D-Nucleofector X Kit (Lonza) and program DZ100. Cells were cultured in 48-well plates and fresh medium was added 2 days after nucleofection. On day 4 after nucleofection cells were counted and viability was assessed using a Countess II FL Automated Cell Counter (Thermo Fisher Scientific) and cells were pelleted for DNA isolation.

Flow cytometry and cell sorting

Totals of 1 × 10⁵ to 2.5 × 10⁵ T cells were pelleted by centrifugation for 5 min at 300 × g and washed twice with FACS buffer (PBS, 2% fetal bovine serum, 0.2% sodium azide). Buffer was poured off and premixed antibodies were added to cells in remaining ∼150 μL buffer. To stain cells from transduction experiments, buffer was aspirated completely and antibody mix was prepared in 100 μL FACS buffer for resuspension of cell pellets. Cells were incubated in the dark for 25–35 min at RT. The following antibodies from BioLegend were used: PE/Cyanine7 or APC/Cyanine7 anti-human CXCR4 (clone 12G5, 5 μL), PE, or Alexa Fluor 647 anti-human CCR5 (clone HEK/1/85a, 5 μL), PerCP/Cyanine5.5-anti-human CD4 (clone OKT4, 2.5 μL) FITC anti-human CD8 (clone SK1, 5 μL), and Zombie violet fixable viability dye (1:500) for live/dead staining (BioLegend). Cells were washed twice and flow cytometry analysis was done on an LSRII instrument (BD Biosciences, San Jose, CA). UltraComp eBeads (Thermo Fisher Scientific) were used for compensation of anti-human CXCR4 and CCR5 antibodies.

For cytokine production analysis, control and dual-edited cells were left unstimulated or stimulated for 2 h with PMA/ionomycin cell stimulation cocktail (eBioscience, San Diego, CA). Monensin solution (BioLegend) protein transport inhibitor was added and live/dead staining was performed using Zombie violet fixable viability dye (BioLegend). The following antibodies (all from BioLegend) were used for intracellular cytokine staining: anti-human IL-4 clone MP4-25D2, anti-human IFNgamma clone 4S.B3, anti-human TNFalpha clone MAb11, and anti-human IL-2 clone MQ1-17H12.

For CD34 cell sorting, cells were electroporated as above with either R5-9 (ABE8e-NG editor) or R5-11 (ABE8e-NRCH) sgRNAs. 48 h after electroporation the cells were sorted by the University of Minnesota Flow Cytometry Core using a FACS LSR II for long-term hematopoietic progenitors with a phenotype of CD34⁺CD45RA^–CD90⁺CD133⁺ using the following antibodies from BioLegend: Alexa Fluor 488 anti-human CD34 clone 581, PE anti-human CD45RA clone HI100, PE/Cyanine7 anti-human CD133 clone S16016E, and APC anti-human CD90 (Thy1) clone 5E10.

Data were analyzed using the FlowJo software v.10.7.1 (TreeStar).

SpCas9-NG nuclease purification

SpCas9-NG nuclease was cloned into a pET42b plasmid with a 6×His tag, eliminating the normal glutathione-S-transferase fusion. BL21 Star DE3 chemically competent Escherichia coli cells (Invitrogen) were transformed with the plasmid and picked into 2×YT + 25 μg/mL kanamycin for overnight growth at 37°C. The next day, 1 L of pre-warmed 2×YT + 25 μg/mL kanamycin was inoculated at an optical density at 600 nm (OD₆₀₀) of 0.03 and shaken at 37°C for about 3 h until OD₆₀₀ reached 0.8. Culture was cold shocked in an ice-water slurry for 1 h, after which protein expression was induced by the addition of 1 mM IPTG. Culture was shaken at 16°C for 16 h to express protein. Cells were pelleted at 6,000 × g for 20 min and stored at −80°C. The next day, cells were resuspended in 30 mL cold lysis buffer (1 M NaCl, 100 mM Tris-HCl [pH 7.0], 5 mM TCEP, 20% glycerol, with three tablets of cOmplete, EDTA-free protease inhibitor cocktail [Millipore Sigma, Burlington, MA; 4693132001]). Cells were lysed by sonification at 4°C for a total treatment of 7.5 min, providing time to cool after every 3 s of treatment. Cell lysate was clarified for 20 min using a 20,000 × g centrifugation at 4°C. Supernatant was collected and added to 1.5 mL of Ni-NTA resin slurry (G Bioscience, 786-940, prewashed once with lysis buffer). Protein-bound resin was washed twice with 12 mL of lysis buffer in a gravity column. Protein was eluted in 3 mL of elution buffer (200 mM imidazole, 500 mM NaCl, 100 mM Tris-HCl [pH 7.0], 5 mM TCEP, 20% glycerol). Eluted protein was diluted in 40 mL of low-salt buffer (100 mM Tris-HCl [pH 7.0], 5 mM TCEP, 20% glycerol) just before loading into a 50-mL Akta Superloop for ion-exchange purification on the Akta Pure25 FPLC. Ion-exchange chromatography was conducted on a 5-mL GE Healthcare HiTrap SP HP pre-packed column. After washing the column with 15 mL low-salt buffer, the diluted protein was flowed through the column to bind. The column was washed in 15 mL of low-salt buffer before being subjected to an increasing gradient to a maximum of 80% high-salt buffer (1 M NaCl, 100 mM Tris-HCl [pH 7.0], 5 mM TCEP, 20% glycerol) over the course of 50 mL, at a flow rate of 5 mL per min. Fractions (1 mL) were collected during this ramp to high-salt buffer. Peaks were assessed by SDS-PAGE to identify fractions containing the desired protein, which were pooled and concentrated using an Amicon Ultra 15-mL centrifugal filter (100 kDa cutoff). SDS-PAGE stained with InstantBlue (Millipore Sigma; ISB1L) was used to visualize the purity after each step. Concentrated protein was quantified using a BCA assay (Thermo Fisher Scientific; 23227); the final concentration was 86.4 μM.

Cas9-NRCH nuclease purification

Cas9-NRCH nuclease was cloned into the protein expression plasmid pD881-SR (Atum, Newark, CA; FPB-27E-269). The expression plasmid was transformed into BL21 Star DE3 competent cells (Thermo Fisher Scientific; C601003). Colonies were picked for overnight growth in terrific broth (TB) + 25 μg/mL kanamycin at 37°C. The next day, 2 L of pre-warmed TB were inoculated with overnight culture at a starting OD₆₀₀ of 0.05. Cells were shaken at 37°C for about 2.5 h until the OD₆₀₀ was ∼1.5. Cultures were cold shocked in an ice-water slurry for 1 h, following which L-rhamnose was added to a final concentration of 0.8% to induce. Cultures were then incubated at 18°C with shaking for 24 h to express protein. Following induction, cells were pelleted and flash-frozen in liquid nitrogen and stored at −80°C. The next day, cells were resuspended in 30 mL cold lysis buffer (1 M NaCl, 100 mM Tris-HCl [pH 7.0], 5 mM TCEP, 10% glycerol, with five tablets of cOmplete, EDTA-free protease inhibitor cocktail tablets) (Millipore Sigma; 4693132001). Cells were passed three times through a homogenizer (Avestin EmulsiFlex-C3) at ∼18,000 psi to lyse. Protein was purified from the cell lysate using the same method as SpCas9-NG nuclease above. Concentrated protein was quantified using a BCA assay (Thermo Fisher Scientific; 23227); the final concentration was 305.5 μM.

CIRCLE-seq

CIRCLE-seq was performed and analyzed as described previously.^42,73 For gRNAs employed with BE4max (X4-1, X4-2, and R5-1), WT spCas9 protein was used (New England Biolabs, Ipswich, MA; M0386T), R5-7 and R5-9 were delivered in complex with SpCas9-NG protein and R5-11 was employed as RNP with SpCas9-NRCH. Before sequencing, the DNA was PCR amplified using PhusionU polymerase, and the products were quantified using a Qubit high-sensitivity kit after gel purification and sequenced on an Illumina MiSeq Sequencer (Illumina, San Diego, CA). Data were processed using the CIRCLE-seq analysis pipeline with parameters: “read_- threshold: 4; window_size: 3; mapq_threshold: 50; start_- threshold: 1; gap_threshold: 3; mismatch_threshold: 6; merged_analysis: True”. Parameters were modified for R5-9 and R5-11 data because target site reads were detected in raw reads but not displayed in the output table: “start_threshold: 7; gap_threshold: 10”. Because the Cas9 variant in ABE8e-NRCH has been shown to cleave target sites with little PAM restrictions, CIRCLE-seq analysis was done with an “NNN” PAM.³⁹ For each gRNA the top predicted sites with highest read counts were analyzed by HTS.

HTS and data analysis

Genomic DNA extraction was done using the PureLink Genomic DNA Mini Kit (Invitrogen) and elution in water. Genomic DNA was amplified using locus-specific primers (Integrated DNA Technologies, Coralville, IA) containing Illumina adapters (Tables S2 and S3). The PCR was performed using 100–200 ng genomic DNA, 1 μL each of forward and reverse primers and PhusionU Multiplex PCR Master Mix (Thermo Fisher Scientific) in a total volume of 20 μL with the following conditions: initial denaturation 30 s at 98°C; 30–31 cycles of denaturation at 98°C for 10 s, primer annealing at 64°C for 30 s, and elongation at 72°C for 15–30 s; final elongation at 72°C for 5 min. PCR products were analyzed on a 2% agarose gel supplemented with SYBR Safe DNA Gel Stain (Invitrogen, Carlsbad, CA) and product sizes were determined using MassRuler Low Range DNA Ladder (Thermo Fisher Scientific). To add Illumina sequencing adapters and unique barcodes to PCR products, a second PCR reaction was performed using 0.5 μL product from the first PCR, 1.25 μL each of forward and reverse Illumina barcoding primers and Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolabs) in a total volume of 25 μL with the following conditions: initial denaturation 1 min at 98°C; 30–31 cycles of denaturation at 98°C for 10 s, primer annealing at 60°C for 30 s, and elongation at 72°C for 30 s; final elongation at 72°C for 2 min. PCR products were analyzed as described above and purified using Agencourt AMPure XP PCR purification beads (Beckman Coulter, Brea, CA) at a beads/PCR product ratio of 0.75/1 eluting with water. PCR products were quantified using PicoGreen DNA quantification and an Agilent Bioanalyzer/TapeStation (Agilent Technologies, Santa Clara, CA) and the pooled library was validated using the Kapa Library Quantification Kit for Illumina platforms (Kapa Biosystems, Wilmington, MA). Libraries were sequenced on an Illumina MiSeq Sequencer using MiSeq Reagent v.3 (Illumina). Base editing and indel frequencies were analyzed using the CRISPResso tool v.2.0.29⁷⁴ in batch mode with no quality filtering with the following parameters: --quantification_window_size 17, --quantification_window_center -3, --ignore_substitutions. Base-editing percentages were calculated as the number of reads in the “Nucleotide_percentage_summary.xlsx” table that have the desired cytosine or adenine at the target position over the total aligned reads. For gRNA candidates with a CCA target codon, target site alleles showing C to T conversion of either or both cytosines (CT, TC, or TT) that all result in stop codons on the opposite strand were counted in the “Alleles_frequency_table_around_sgRNA.txt”. Frequencies of >0.1%, which is the limit of detection using Illumina MiSeq sequencing, were considered measurable. Indel percentages were calculated as the number of modified reads over the number of aligned reads from the “CRISPRessoBatch_quantification_of_editing_frequency.txt” table. Reference sequences were taken from University of California, Santa Cruz Assembly hg19 (February 2009).

Quantitative RT-PCR

Total RNA was isolated with the RNeasy MidiKit (QIAGEN, Hilden, Germany) and equal amounts of RNA were reverse transcribed with SuperScript IV VILO Master Mix (Thermo Fisher Scientific). Quantitative RT-PCR was performed using TaqMan gene expression assays and CCR5 (Hs99999149_s1), CXCR4 (Hs00607978_s1), and GAPDH (Hs02786624_g1) probes (Thermo Fisher Scientific). The assay was performed on the QuantStudio 3 Real Time PCR System (Thermo Fisher Scientific).

CFU assay

HSPCs were counted 6 days after nucleofection and plated at 100 cells per 35 mm well in triplicate in methylcellulose (MethoCult, STEMCELL Technologies). Cells were cultured for 14 days and colonies were enumerated and scored for morphology by an experienced blinded reviewer. The STEMvision instrument (STEMCELL Technologies) was used to document and capture images of the assays.

Statistical analysis

Unpaired two-sided t tests were conducted to compare the GFP/RFP fluorescence in transduced T cells using GraphPad Prism software v.8. No corrections were made for multiple comparisons.

Data availability

Data related to this article are provided in the supplemental information.

Supplemental information

Table S3

Circle-Seq output data of identified off target sites