All wild type TALEN pairs used were previously described. TRAC-targeting TALE nucleases were used in a proof-of-concept study to show feasibility of large scale production of off-the-shelf CAR T cells (Alzubi et al., 2021). TALE nucleases targeting CCR5 have been investigated in a comparative study benchmarking them against ZFNs (Mussolino et al., 2014). OH-TALEN encoding plasmids were produced in two steps by conventional cloning. First, the wild type FokI domains were excised using restriction enzymes PmeI and BamHI (NEB). Second, FokI domains encoding for OH-substitutions, ordered as gBlocks from IDT, were inserted using the NEBuilder^® HiFi DNA Assembly Master Mix following manufacturer instructions. All TALEN-encoding mRNAs were produced by in vitro transcription using the HiScribe™ T7 ARCA mRNA Kit with tailing (NEB) following manufacturer instructions. Target sequences of TALENs used in this study are shown in Supplementary Table S1.

PBMCs were isolated from leukocyte reduction system (LRS) chambers obtained from the Blood Donation Center (University of Freiburg, Medical Center) by density gradient centrifugation using Ficoll. Appropriate aliquots were resuspended in CryoStor™ CS10 (StemCell Technologies) for long-term storage in liquid nitrogen.

PBMCs/T cells were cultured at 37°C with 5% CO₂. Upon thawing, PBMCs were washed with PBS (300xg, 5min) and resuspended at a density of 2 × 10⁶ cells/ml in X-VIVO™ 15 (Lonza) medium supplemented with 200U/ml rhIL-2 (Immunotools) and seeded into 24-well plates (1 ml/well). Adherent cells were allowed to attach for 4 h. After 4 h, non-adherent cells were collected, counted, adjusted to a cell density of 1 × 10⁶ cells/ml with X-VIVO™ 15 (Lonza) medium supplemented with 200U/ml rhIL-2 (Immunotools) and re-seeded into 24-well plates (1ml/well). To each well, 5 µl of ImmunoCult™ Human CD3/CD28/CD2 T Cell Activator (StemCell Technologies) was added and the T cells activated for 72–96 h prior to gene editing.

Prior to gene editing, activation of T cells was assessed by staining for CD25. Downstream experiments were performed only when T cells were highly activated (>85% CD25-expression). For small scale transfer of TALEN mRNAs, 1 × 10⁶ cells were harvested by centrifugation (300xg, 5 min) and the supernatant removed. Cells were resuspended in 50 µl CliniMACS ^® Electroporation Buffer (Miltenyi Biotec). Just before electroporation, 7.5 µg of each left and right (or left/left, right/right) TALEN pairs were mixed with the resuspended T cells and electroporated using a CliniMACS ^® Prodigy with Electroporator unit (Miltenyi Biotec) with the previously described Setting 3 (Alzubi et al., 2021). Post electroporation, cells were recovered in 400 µl of pre-warmed X-VIVO™ 15 (Lonza) medium supplemented with 200 U/ml rhIL-2 (Immunotools) and seeded into two wells of U-shaped 96-well plates. Half of the cells were subjected to a transient low temperature shift to 32°C for 24 h before being shifted to 37°C. The other half was directly cultured at 37°C. Approximately half of the media was changed every 2 days and the cells split every 3–4 days.

All flow cytometry measurements were carried out using a BD Accuri C6 device. Cells were stained for 45–60 min at 4°C–8°C. T cell activation was assessed by staining with anti-human CD3-APC (Miltenyi Biotec, clone BW264/56) and anti-human CD25-PE (Miltenyi Biotec, clone 4E3). TRAC knock-out was assessed 6–7 days after gene editing by staining with anti-human TCRα/β-PE (Miltenyi Biotec, clone BW242/412) and anti-human CD3-APC (Miltenyi Biotec, clone BW264/56).

1 × 10⁶ cells were electroporated with 7.5 μg of left and right TALEN mRNAs. After 4 h of incubation at 37°C, 1 × 10⁵ cells were harvested (300xg, 5min) and washed with 500 μL FACS buffer (PBS supplemented with 5% FCS). Cells were subsequently treated with 100 μL BD Cytofix/Cytoperm™ (BD Biosciences). After 30 min incubation on ice, the cells were washed twice with 500 μL BD Perm/Wash Buffer (BD Biosciences). Permeabilized/fixed cells were stained with 50 μL rabbit anti-RVD antibody (Yang et al., 2022) (1:250) via incubation for 30 min on ice. After another wash step with 500 μL BD Perm/Wash Buffer, 100 μL of 1:500 diluted secondary goat anti-rabbit antibody (Life Technologies, clone A-11034) was added. Following an incubation for 30 min on ice, the cells were washed once with BD Perm/Wash Buffer and finally FACS buffer. Afterwards, the cells were analyzed using a BD Accuri C6 device.

Library preparations were basically performed as previously described (Turchiano et al., 2021) using samples that showed particularly high editing as determined by T7E1. In contrast to the original protocol, agarose gel-extraction using the QIAquick Gel extraction kit (QIAGEN) was performed on PCR fragments with a size of 200–500 bp originating from PCRII. This step was included to remove non-informative, short PCR fragments (<200 bp) prior to the barcoding step. NGS libraries were sequenced by Genewiz (part of Azenta Life Sciences) on Illumina HiSeq or NovaSeq with 2 × 150 bp read lengths. Only sites detected as significant hits in two technical replicates are depicted throughout this study. CAST-Seq and T-CAST analysis results for all TALENs targeting CCR5 and TRAC are provided in Supplementary Tables S4–S10. Oligonucleotides used for T-CAST are listed in Supplementary Tables S11.

Every single replicate, treated and untreated control, is processed independently from the alignment up to the cluster definition, as described in (39). Then, an overlap analysis is performed to unify the clusters from several replicates. Clusters overlapping or separated by less than 1,500 bp are merged and considered as a single translocation event [see (Turchiano et al., 2021) for details]. Based on the number of replicates, the user can define the minimum number of replicates where the site was found, and the minimum number of samples in which the site was significantly different from untreated control (i.e., the number of reads was significantly higher in treated vs. untreated based on Fisher’s exact test).

Barcode hopping: We introduced an additional filter to eliminate artifacts generated by barcode hopping events. Barcode hopping are identified by their low reads:hits ratio in comparison to real translocation events by the formula: log10 (reads:hits) distribution (<Q1—2.5*IQ).

Coverage: For the remaining sites, the read coverage is calculated in order to identify highly covered regions. Sites are divided into 100 bins of equal size. For each site, the coordinates of bin with the highest coverage across all replicates is used for downstream analysis instead of the whole site coordinates. This new feature restricts the alignment against the target sequence to a smaller, and highly covered region. This makes the alignment more specific and less prone to identification of false-positive OMTs/HMTs.

Alignment: A new TALEN-specific substitution matrix was implemented (Supplementary Tables S12) inspired by (18), and analysis restricted to four TALEN combinations: LF.LR, LF.RR, RF.RR, and RF.LR (L/RX, left/right; XF/R, forward/reverse). In order to determine the best combination, i.e., the one that is most likely cleaving an off-target site, different spacer lengths from 8 to 28 bp, are tested for each combination. Artificial sequences, representing binding sites of two TALEN arms separated by a spacer “N_k” of 8–28 nucleotides (k belong to 8:28) are tested. N can match any bases without cost, therefore the length of the spacer does not influence the alignment score by itself. An example sequence is shown in Supplementary Figure S2B. Alignment score is calculated using the pairwise Alignment function from Biostrings R package with a “local-global” alignment type. The different TALEN combinations and spacer lengths are first selected based on two criteria: a) The first (5′) aligned base is a T, b) the last (3′) aligned base is an A. Then we ordered them based on the alignment score and define the highest score as the most probable TALEN combination and spacer length for a given target site. The same approach was performed on randomly selected regions over the entire genome to determine the overall distribution of the alignment score on random sequences. p values of a given combination and spacer length are assessed based on the empirical cumulative distribution function. Sites with p values below 0.05 are considered as OMT. HMTs and NBSs were classified in the same way as described in (39).

TALEN target sites as well as putative off-target sites were amplified from 100 ng genomic DNA by standard PCR using Q5^® Hot Start High-Fidelity DNA Polymerase (NEB). PCR fragments were purified using either the QIAquick PCR Purification Kit or the QIAquick Gel Extraction Kit (both QIAGEN). Purified PCR products were pooled per sample and NGS libraries constructed using the NEBNext^® Ultra™ II DNA Library Prep Kit for Illumina^® (NEB). The resulting NGS libraries were quantified by ddPCR using the ddPCR™ Library Quantification Kit for Illumina TruSeq (Bio-Rad) and sequenced on Illumina HiSeq or NovaSeq platforms with 2 × 150 bp read length by NGS service provider Genewiz (part of Azenta Life Sciences). Reads were analyzed using the CRISPResso2 package (Clement et al., 2019) and the p values obtained from CRISPResso Compare. All primers used for Amplicon-NGS are listed in Supplementary Tables S11.

TALEN target sites were amplified from 100 ng of genomic DNA extracted using the NucleoSpin Tissue Mini Kit for DNA from cells and tissue (Macherey-Nagel) by standard PCR using Q5^® Hot Start High-Fidelity DNA Polymerase (NEB). Fragments were purified using the QIAquick PCR Purification Kit (QIAGEN) and subsequently denatured by incubation at 95°C for 5 min. Denatured DNA fragments were allowed to re-anneal through slow cooling of the samples to room temperature. Heteroduplex cleavage as surrogate readout for Indel formation/gene editing was visualized through enzymatic restriction of 100 ng of re-annealed sample with 7.5U of T7 endonuclease I (NEB) for 30 min at 37°C. T7E1 cleavage efficiency was determined by agarose gel electrophoresis and adjacent analysis of the gel images using ImageJ 1.47v.

In silico prediction of TALEN off-target sites was performed using the PROGNOS web tool (Cradick et al., 2014). For prediction, the individual RVDs were entered and up to six mismatches per TALEN half site allowed. Further, a distance of 10–25 bp between TALEN binding sites was allowed as well as formation of both hetero- and homodimers enabled. PROGNOS results are listed in Supplementary Tables S2 (CCR5) and S3 (TRAC).

CAST-Seq is a highly sensitive assay to nominate off-target (OT) sites by identifying gross chromosomal aberrations, such as large deletions and inversions at the on-target sites of designer nucleases. It was applied successfully to samples treated with various CRISPR-Cas9 nucleases (Turchiano et al., 2021; AlJanahi et al., 2022) but, to date, only to a single TALEN pair targeting the HBB locus (Turchiano et al., 2021). The fact that the bioinformatic nomination of an off-target site is fundamentally different for a dimeric TALEN pair as compared to a monomeric CRISPR-Cas nuclease, prompted us to revisit OT-activity of previously published TALEN targeting the CCR5 gene (Mussolino et al., 2014). To this end, TALEN-encoding mRNAs were produced and transferred into primary human T cells expanded from PBMCs. Amplicon NGS revealed that the CCR5-targeting TALEN caused small insertions and deletions (Indels) in up to 55% of alleles at the on-target site (Figure 1A). In addition, Indel formation at a frequency of 11.5% was observed at the previously described off-target site in CCR2 (Figure 1A) (Mussolino et al., 2014). Indeed, alignment of the TALEN target site in CCR5 with the off-target site in CCR2 situated 15 kb upstream, revealed a single mismatched binding by the right TALEN arm, while the left TALEN can bind to CCR2 without mismatches with an optimal spacer of 14 bp with reference to the right arm (Supplementary Figure S1A).

CAST-Seq analysis performed on these samples nominated in addition to CCR2 six high-scoring off-target mediated translocations (OMTs) harboring >20 CAST-seq hits. In addition to those seven off-target sites, three sites on chromosomes 4 (347 hits), chromosome 3 (66 hits), and chromosome 2 (41 hits) were identified (Figure 1B; Table 1). The current algorithm could neither classify them as OMTs nor as homology-mediated translocations (HMTs) but instead categorized them as natural break sites (NBS). Of note, according to the number of CAST-seq hits, translocations between the CCR5 on-target site and NBS#1 (347 hits) seems to be more frequent than translocations between CCR5 and OMT#1 (211 hits) (Table 1). We subsequently probed the predicted TALEN off-target site in OMT#1 for the formation of Indels by targeted amplicon NGS. To this end, we designed primers flanking the site at which two right TALEN arms was proposed to bind (4 or six mismatches, respectively) with an 18 bp spacer (Figure 1C), but did not find signs of TALEN-associated off-target activity (Figure 1D).

In order to better understand which chromosomal regions had actually translocated from chromosome 5 (OMT#1) to the on-target site, as well as how the deletion landscape at the on-target site looks, we produced plots showing the read coverage at these chromosomal regions (Figure 1E; Supplementary Figure S1B). The read coverage at the on-target site (Supplementary Figure S1B) revealed large deletions of several kilobases. It furthermore showed two distinct peaks, one at the CCR5 on-target site and a second peak at the known off-target site in CCR2, 15 kb upstream (Supplementary Figure S1B). This observation showcased that read coverage is a strong indicator for nominating an off-target site. In contrast, the predicted TALEN binding site in OMT#1 (Figure 1E green and purple dotted lines) were distant from the area of high coverage, as were the primers initially used to assess Indel formation at this site (Figure 1E arrows P#1 and P#2) We therefore designed primers P#3/P#4 flanking the highly covered region and repeated amplicon NGS, uncovering Indels in 6% of alleles (Figure 1F).

Coverage plots for OMT#2-#6 revealed that in only two cases (OMT#3 and OMT#6) the predicted TALEN off-target sites (green and purple dotted lines) were close to the region with the highest read coverage (Supplementary Figure S1C).

Incited by these results, we set out to improve the accuracy and fidelity of CAST-Seq through development of an optimized pipeline for TALENs called T-CAST. Taking advantage of the strong predictive value of read coverage at off-target sites, T-CAST splits up in a first step each translocated region into 100 bins of equal size. Next, the single bin, with the highest relative read coverage obtained from two (or more) CAST-Seq replicates is identified. A restricted region of ±100 bp from the highest bin was chosen, and the TALEN pair was aligned to this shorter region, using a 5′-T constraint, which is a prerequisite for TALEN-DNA engagement (Cornu et al., 2017). Moreover, to improve the predictive power of identifying the correct off-target site, we implemented a TALEN-specific substitution matrix, which accommodates for the fact that RVDs can bind to multiple DNA bases with varying affinity (Supplementary Table S12).

As shown in Figure 2A; Supplementary Figure S2A, TALEN binding site nomination is now restricted to the most highly covered regions identified by T-CAST. The implemented changes to the bioinformatics pipeline not only changed the identification of TALEN off-target sites, but also the classification for the observed chromosomal rearrangements. In addition to the known CCR2 OMT, 12 sites with more than 20 hits were classified as OMTs by T-CAST, on top of two NBS (Figure 2B; Table 2). While the translocation with chromosome 4 (347 hits, Table 1) was now classified as OMT, two sites previously classified as OMTs (OMT#2 and OMT#5) were now re-classified as NBSs (Table 2). Of note, the T-CAST predicted TALEN target site for OMT#2 (ANKRD55, listed as OMT#1 in Table 1) displayed seven mismatches in the first and eight mismatches in the second binding site of the TALEN left arm (Supplementary Figure S2B).

In order to validate the nominated off-target sites, we designed amplicon NGS primers flanking OMTs#1-9 and NBS#1/2 (Table 2). We obtained specific PCR products for all sites except OMT#7. NGS revealed significant Indel formation above background (0.1%) at all sites, except NBS#2 (Figure 2D), with Indel frequencies ranging from 0.12% (OMT#4) to more than 5% (OMT#1 and OMT#5). We challenged the T-CAST prediction for the most likely TALEN conformation to cause off-target activity by transferring mRNA encoding either only the left TALEN arm (L + L) or only the right TALEN arm (R + R) into activated T cells. In line with the T-CAST annotation, formation of Indels at off-target sites was observed in all samples treated with only left TALEN arm but not in cells exposed to the right TALEN arm only (Figure 2C). The exception is off-target activity at CCR2, for which both TALEN arms is necessary (L + R). Of note, the coverage plots in Supplementary Figure S1B uncovered all the events in the whole region between the CCR5 and CCR2, thus exposing CCR2 as off-target combined with large deletions. It is also worth mentioning that with the exception of CCR2 PROGNOS (Fine et al., 2014) did not predict any of these sites, even when allowing up to six mismatches per TALEN arm (Supplementary Table S2).

TALEN-mediated DNA cleavage is performed through dimerization of the two FokI domains upon binding in a tail-to-tail orientation of the two subunits. This can be brought about by two different TALEN arms but also through the formation of homodimers as described above (Figure 2C). In order to prevent homodimerization, charged residues within the two FokI dimer interface can be substituted by oppositely charged residues, exerting repellent electrostatic forces and preventing dimer formation of two identical TALEN subunits (Cade et al., 2012; Nakajima and Yaoita, 2013; Schwarze et al., 2021). These obligate-heterodimeric TALEN pairs reduce the number of possible TALEN conformation by 50% (Figure 3A).

Here, we validated two OH-TALEN scaffolds harboring either KKR-ELD or KVR-EAD substitutions in the FokI domain, either linked to the right or to the left TALEN arm. On-target activity of WT and OH-TALENs was assessed on genomic DNA isolated from primary T cells that were cultured constantly at 37°C or subjected to a transient temperature shift to 32°C for 24 h post-electroporation with TALEN-encoding mRNA (Figure 3B). Genotyping by T7E1 assay revealed significantly higher Indel formation for both OH-TALEN scaffolds, averaging 56% and 60% mutated alleles for KKR-ELD and KVR-EAD TALEN, respectively, in contrast to 36% mutated alleles upon transfer of WT TALEN under transient cold-shock (32°C) conditions. Under constant temperature at 37°C, no significant differences were observed between the activities of WT-TALEN and OH-TALEN, with 35%–47% cleavage in T7E1 assay (Figure 3B). Flow cytometric analysis upon intracellular staining 4 h after mRNA transfer revealed similar expression for all TALEN scaffolds (Figure 3C).

Primary T cells that were edited with CCR5-targeting OH-TALENs were subsequently subjected to T-CAST analysis. In contrast to WT-TALEN, no high-scoring chromosomal aberrations were observed in samples treated with KKR-ELD TALEN (Figure 3D) and only three structural variations with >20 T-CAST hits were identified in KVR-EAD TALEN treated cells (Figure 3E). Since no chromosomal translocations were detected in KKR-ELD samples, we decided to probe the two top-scoring NBSs (Table 3) for Indel formation. No Indels were detected at these sites in neither WT nor OH-TALEN treated samples (Figure 3F). Likewise, we analyzed all three nominated off-target sites identified in KVR-EAD TALEN treated samples as well as the two highest-scoring NBSs (Table 3) with respect to the presence of Indels by amplicon NGS. Indel formation above the limit of detection were only observed at OMT#1 and OMT#2 in KVR-EAD TALEN treated samples but hardly above background (Figure 3F). In addition, we tested for Indel formation at the six top-scoring OMTs identified previously for WT TALEN (Table 2) as well as the on-target site and the CCR2 off-target site. In line with the T7E1 results, the fraction of mutated alleles ranged from 56% (WT) to 68% and 72% for KKR-ELD and KVR-EAD, respectively (Figure 3G). Off-target mutagenesis in CCR2 was highest in KVR-EAD edited samples (20%) and lowest in KKR-ELD treated samples (6%). In line with off-target activity at CCR2, we detected a large 15 kb deletion between the CCR5 and CCR2 loci in all samples, irrespective of the scaffold (Supplementary Figure S3). In conclusion, both OH scaffolds greatly improved the specificity of the CCR5-targeted TALEN by abrogating activity at off-target sites that were bound by a single TALEN arm in a homodimeric tail-to-tail configuration.

The constant region of the T cell receptor α chain (TRAC) is an interesting target in CAR T cell immunotherapy. Disruption of TRAC results in the loss of expression of the entire T cell receptor complex (TCR), a major road block in allogeneic CAR T cell therapy (Lin et al., 2021). We have previously shown that a TALEN targeting TRAC can be used in large scale manufacturing of universal CAR T cells (Alzubi et al., 2021). Using T-CAST, we revisited the safety profile of the previously used WT-TALEN in comparison to OH-TALEN scaffolds. Genotyping by T7E1 assays revealed high on-target activity for all three TALEN scaffolds, each one achieving some 90% of T7E1 cleavage (Figure 4A). In line with our findings for CCR5-targeting TALENs, OH-TALEN targeting TRAC outperformed their WT counterpart when the primary T cells were subjected to a transient 32°C cold-shock post electroporation. These findings are mirrored by the phenotypic analysis of TCRα/β surface expression performed 1 week post-transfer of TALEN-encoding mRNA. Under transient cold-shock conditions, the fraction of TCRα/β-negative cells increased from 55% TCR-negative T cells upon transfer of WT-TALENs to 85% and 83% in T cells edited with KVR-EAD and KKR-ELD OH-TALENs respectively (Figure 4B). T cells cultured at 37°C revealed no significant differences with regard to knockout efficacies between the various scaffolds (Figures 4A, B). A representative flow cytometric analysis (Supplementary Figure S4) shows the two clearly separated populations of TCRα/β positive and TCRα/β negative cells, in agreement with monoallelic expression of the T cell receptor α chain.

Similar to our observations for TALEN targeting CCR5, the number of chromosomal rearrangements was highest in samples treated with the WT-TALEN scaffold (Table 4). T-CAST identified a total of five translocations with more than 20 T-CAST hits (Figure 4C). In KKR-ELD TALEN treated samples, no structural variation was identified (Figure 4D), while T cells edited with KVR-EAD TALEN displayed three chromosomal aberrations (Figure 4E). Of note, the highest-scoring rearrangement observed in the KVR-EAD samples, OMT#1 with 79 T-CAST hits, is in close proximity to the on-target site, thus likely representing a large deletion rather than an off-target site. A similar phenomenon was observed in KKR-ELD edited T cells (OMT#2, Table 4).

Validation of T-CAST results by targeted amplicon NGS confirmed high on-target activity for all TALEN scaffolds, with 76%–88% Indels at TRAC (Figure 4F). Assessment of off-targets OMT#1–4 identified for WT-TALEN confirmed off-target activity at OMT#1–3, with 6%–24% Indels at those sites in the WT-TALEN treated samples and absence of off-target mutagenesis in T cells edited with the OH-TALEN scaffolds. No Indels were observed at OMT#4 in any of the edited samples (Figure 4F), while OMT#5 could not be amplified. We furthermore probed OMTs associated with TALEN (OMT#1, 17 CAST-hits) and KVR-EAD TALEN (OMT#2, 32 CAST-hits). Despite the low number of T-CAST hits, a small but significant fraction of reads in both OH-TALEN edited T cell samples revealed off-target activity at this off-target site on chromosome 2 with 0.5%–0.7% of alleles with Indels. No Indels were detected at KVR-EAD OMT#2 (Figure 4F). Of note, similarly to the CCR5-targeting TALENs, there is no overlap between the PROGNOS-predicted sites for TRAC-targeting TALEN (Supplementary Table S3) and experimentally confirmed off-target sites identified by T-CAST.

In conclusion and in agreement with the results for the CCR5-targeting TALENs, both OH scaffolds greatly improved the specificity of the TRAC-targeted TALEN by abrogating off-target activity at sites that are cleaved by a homodimeric TALEN.