Toward eliminating HLA class I expression to generate universal cells from allogeneic donors

Ex vivo manipulation of autologous cell products that are then returned to the patient can restore cellular functions in individuals with incurable diseases.^1-5  However, this manufacturing of recipient-specific clinical-grade products is time-consuming and labor-intensive, as well as expensive, and the desired cells are often unavailable when required for many patients. Engraftment of donor-derived (allogeneic) cells to reconstitute cellular functions is advantageous compared with infusing patient-derived cells, as the ability to manufacture and validate therapeutic and fully functional cell preparations in advance improves safety, consistency, and availability. Survival of an allograft bearing disparate human leukocyte antigens (HLAs) in an immunocompetent recipient depends on avoiding or overcoming an immune response to the infused cells. Rejection is primarily mediated by host-derived T cells recognizing nonself major and/or minor histocompatibility antigens (mHAgs). Therefore, the most effective approach to sustaining allograft survival is to preclude mismatches between the donor and recipient HLA, as highlighted by the improved survival of HLA-matched grafts after allogeneic hematopoietic stem cell⁶  and solid organ transplantation.⁷  This led us to investigate whether an immune response could be avoided by eliminating expression of 1 or more mismatched HLAs on donor-derived cells.

Some viral proteins inhibit HLA folding and surface display, which allows infected cells to escape T-cell recognition,⁸  and enforced expression of these viral-derived transgenes can downregulate HLA expression.⁹  As an alternative, the Cre-LoxP system can be deployed to disrupt the β₂-microglobulin locus, and thus HLA class I expression, but this requires removal of antibiotic-resistant genes by Cre recombinase, which may introduce unwanted recombination events.¹⁰  We and others have previously attempted to downregulate HLA class I expression by introducing small interfering RNA targeting HLA heavy chains or β₂-microglobulin.^11-13  Although these posttranscriptional approaches reduce antigen levels, they require sustained transgene expression and, moreover, reduce but do not completely eliminate HLA expression. Given that an αβ T-cell receptor (TCR) response can be triggered by just a small number of cell-surface HLA molecules,¹⁴  we sought an alternative to achieve complete elimination of HLA. Here we show that transient expression of zinc finger nucleases (ZFNs)¹⁵  targeting the HLA-A locus can permanently and completely eliminate HLA-A expression from (1) a model cell line, (2) primary and genetically modified human T cells used in clinical trials, and (3) human embryonic stem cells (hESCs). These results highlight a path toward rapid human application, as circulating natural killer (NK) cells could be prevented from recognizing cells engineered to lose HLA expression.

Peripheral blood mononuclear cells (PBMCs) were obtained from healthy adult volunteer donors who had provided informed consent from Gulf Coast Regional Center (Houston, TX) in accordance with the Declaration of Helsinki, and who participated in research approved by the institutional review board of The University of Texas MD Anderson Cancer Center.

ZFNs containing 5 or 6 fingers were designed and assembled using an established archive of prevalidated 2-finger and 1-finger modules essentially as described.¹⁶  Briefly, the coding sequence of HLA-A was scanned for locations at which 2 such ZFNs (designated as ZFN-L and ZFN-R) could be targeted to sites that were separated by 5 base pairs and located on opposite DNA strands. The nucleotide targets for candidate ZFN pairs were then checked for divergence from other HLA coding sequences. Genes encoding the ZFN designs were assembled using a polymerase chain reaction (PCR)-based procedure and cloned into a plasmid.

HEK293 cells were maintained in Dulbecco’s modified Eagle medium (DMEM; Lonza, Basel, Switzerland) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Lonza) and 2 mmol/L l-glutamine (Invitrogen, Carlsbad, CA). Epstein-Barr-virus–transformed lymphoblastoid cell line (EBV-LCL), 721.221, EL-4, NALM-6, and Daudi cell lines were maintained in RPMI 1640 (Lonza) supplemented with 10% FBS and 2 mmol/L l-glutamine (designated as complete medium). The identity of these cell lines was confirmed by short tandem repeat DNA fingerprinting. CD8⁺ cytotoxic T-lymphocyte (CTL) clones specific for mHAgs were HLA-A*03:01-restricted 7A7 recognizing peptide RVWDLPGVLK encoded by PANE1 transcripts and HLA-A*02:01-restricted GAS2B3-5 recognizing CIPPDSLLFPA peptide from ORF +2/48 in C19ORF48. CTLs were thawed 1 day before the assay and maintained in RPMI 1640 supplemented with 10% human albumin serum, 2 mmol/L l-glutamine, 20 ng/mL interleukin 15 (IL-15; PeproTech, Rocky Hill, NJ), and 20 IU/mL IL-2 (Chiron, Emeryville, CA).

T cells were activated in vitro by stimulating PBMCs with OKT3 (eBioscience, San Diego, CA) preloaded onto artificial antigen presenting cells (aAPCs; clone 4: K562 cells genetically modified to stably coexpress CD19, CD64, CD86, CD137L, and a membrane-bound mutein of interleukin [IL-15] synchronously expressed with enhanced green fluorescence protein^17,18 ) at a ratio of 1:1 (T cells: γ-irradiated [100Gy] aAPC) in complete media supplemented with 50 IU/mL IL-2 (added every other day). OKT3-loaded aAPC were re-added every 14 days to sustain T-cell proliferation.

Our approach to manufacturing clinical-grade chimeric antigen receptor–positive (CAR⁺)T cells was adapted to generate CD19-specific T cells.¹⁹  DNA plasmids coding for Sleeping Beauty (SB) transposon CD19RCD28 and SB hyperactive transposase SB11 were simultaneously electro-transferred (Human T-Cell Nucleofector solution, program U-014), using a Nucleofector II device (Lonza), into T cells derived from PBMCs. A population of CAR⁺T cells was selectively numerically expanded by adding on the day of electroporation and re-adding every 14 days (at a 1:2, T cell:aAPC ratio) γ-irradiated (100 Gy) aAPC (clone 4 without OKT3 loading) in the presence of 50 IU/mL IL-2 (added every other day).

In vitro–transcribed mRNA species were prepared as previously described.²⁰  In brief, the DNA template plasmids coding for ZFN-L and ZFN-R (supplemental Figure 1) were linearized with XhoI. After in vitro transcription (RiboMAX Large Scale RNA Production System-T7; Promega, Madison, WI) and capping (ARCA cap analog; Ambion, Austin, TX) according to manufacturers’ instructions, poly-adenines were added using the poly A tailing kit (Ambion). The integrity of the mRNA species was validated on a denaturing 1% agarose gel with 3-(N-morpholino) propanesulphonic acid buffer, and concentration was determined by spectrophotometer (BioRad, Hercules, CA) at OD₂₆₀. The mRNA was vialed and stored at −80°C for one-time use.

For the modification of HEK293, expression vectors encoding HLA-A targeting ZFNs (supplemental Figure 1) were introduced by Nucleofection (Lonza) according to the manufacturer’s protocol and described elsewhere.²⁰  After electroporation, cells were immediately placed in prewarmed complete media and cultured at 37°C and 5% CO₂ for 4-6 hours, at which point 50 IU/mL IL-2 was added for further culture. In “cold shock” experiments,²¹  after overnight culture in a 37°C–5% CO₂ incubator, T cells were transferred to a 30°C–5% CO₂ incubator and cultured for 3 days before being returned to a 37°C–5% CO₂ incubator before analysis.

After washing cells with phosphate-buffered saline supplemented with 2% FBS and 2 mM EDTA, cells were labeled with phycoerythrin (PE) anti-HLA-A2 (BD Biosciences, San Jose, CA) at 4°C for 15 minutes, washed, and labeled with anti-PE microbeads (Miltenyi Biotec, Auburn, CA) for 10 minutes. After washing, labeled cells were passed through an LD Column (Miltenyi Biotec), and the flow-through fraction was collected and cultured. T cells were propagated on γ-irradiated OKT3-loaded aAPC, and CAR⁺T cells were propagated on CD19⁺aAPC (not OKT3-loaded) in complete media supplemented with 50 IU/mL IL-2 (added every other day).

The following antibodies were used: PE-conjugated antibodies; anti-HLA-A2 (clone BB7.2), anti-CD56 (clone B159), anti-HLA-DR (clone G46-6), mouse IgG2bκ, mouse IgG2aκ, FITC conjugated antibodies; anti-CD4 (clone RPA-T4), anti-CD8 (clone HIT8a), nonspecific mouse IgG1, PE and APC anti-CD3 (clone SK7), secondary reagent streptavidin-PE (all from BD Biosciences), biotin-conjugated anti-HLA-A3 (clone 4i153), APC anti-HLA-G (clone MEMG19), PE anti-HLA class I (clone W6/32; Abcam, Cambridge, MA), and PE anti-HLA-E (clone 3D12; Biolegend, San Diego, CA). The Alexa 488-conjugated anti-CD19RCD28 CAR antibody was generated in our laboratory. We added propidium iodide (Sigma-Aldrich) to exclude dead cells from analysis. Data were acquired on a FACS Calibur using CellQuest version 3.3 (BD Biosciences) and analyzed by FlowJo version 7.6.1 (Tree Star Inc., Ashland, OR).

The level of modification of the HLA-A gene sequence in ZFN-transfected cells was determined by the Surveyor nuclease assay.²²  In brief, genomic DNA from ZFN-modified cells underwent PCR with oligonucleotide primers designed to amplify the ZFN target regions within HLA-A2 and HLA-A3 genetic loci. After denaturing and re-annealing, Surveyor endonuclease (Transgenomic, Omaha, NE) was used to cut heteroduplex DNA products that were resolved by polyacrylamide gel electrophoresis. Percentage target modification was quantified by densitometry. The PCR primers used for the amplification of target loci were HLA-A3 forward, 5′-GGGGCCGGAGTATTGGGACCA-3′; HLA-A3 reverse, 5′-CCGTCGTAGGCGTCCTGCCG-3′; HLA-A2 forward, 5′-GGGTCCGGAGTATTGGGACGG-3′; and HLA-A2 reverse; 5′-TTGCCGTCGTAGGCGTACTGGTG-3′.

Target cells were labeled with 0.1 mCi ⁵¹Cr for 2 hours. After washing 3 times with ice-cold complete media, labeled cells were diluted and distributed at 10³ target cells in 100 μL per well in 96-well, v-bottomed plates. In the peptide titration assay, target cells were incubated with 10-fold serial dilutions of the peptides for 30 minutes at room temperature. CTLs were added at indicated effector-to-target ratios. After a 4- to 6-hour incubation at 37°C in 5% CO₂, 50 μL of cell-free supernatants were collected and counted on a TopCount device (Perkin Elmer, Shelton, CT). All assays were performed in triplicate. In some assays, parental HEK293 and HLA-A modified HEK293 clones were treated with 600 IU/mL interferon γ (IFN-γ; R&D systems, Minneapolis, MN) and 10 ng/mL tissue necrosis factor-α (TNF-α; R&D systems) for 48 hours before assay. The percentage specific lysis was calculated as follows: ([experimental cpm − spontaneous cpm]/[maximum cpm − spontaneous cpm]) × 100.

The hESC line WIBR3 (Whitehead Institute Center for Human Stem Cell Research, Cambridge, MA)²³  was maintained, as described previously,²⁴  on mitomycin C–inactivated mouse embryonic fibroblast feeder layers in hESC medium (DMEM/F12 [Invitrogen] supplemented with 15% FBS, 5% KnockOut Serum Replacement [Invitrogen], 1 mM glutamine [Invitrogen], 1% nonessential amino acids [Invitrogen], 0.1 mM β-mercaptoethanol [Sigma, St. Louis, MO], and 4 ng/mL FGF2 [R&D systems]). Targeted hESCs were differentiated into fibroblast-like cells as described previously.²⁵  Briefly, differentiation was induced by embryoid body (EB) formation in nonadherent suspension culture dishes (Corning, Corning, NY) in DMEM medium supplemented with 15% FBS for 5 days. EBs were subsequently plated onto adherent tissue culture dishes and passaged according to primary fibroblast protocols, using trypsin for at least 4 passages before the start of experiments.

hESCs were cultured in ρ-associated protein kinase inhibitor (Stemolecule; Stemgent, Cambridge, MA) 24 hours before electroporation. Cells were harvested using 0.05% trypsin/EDTA solution (Invitrogen) and resuspended in phosphate-buffered saline. Ten million cells were electroporated (Gene PulserXcell System, Bio-Rad: 250 V, 500 µF, 0.4 cm cuvettes) with 35 µg donor plasmid encoding the puromycin-resistant gene under the control of phosphoglycerate kinase promoter flanked by 5′ and 3′ groups homologous to the putative ZFN binding region of HLA-A24 and 7.5 µg of each ZFN-encoding plasmid, or 35 µg of donor plasmid and 10 µg of each ZFN-encoding mRNA. Cells were subsequently plated on DR4 mouse embryonic fibroblast feeder layers in hESC medium supplemented with ρ-associated protein kinase inhibitor for the first 24 hours. Puromycin selection (0.5 µg/mL) was initiated 72 hours after electroporation. Individual puromycin-resistant colonies were picked and expanded 10 to 14 days after electroporation. Correct targeting and gene disruption was verified by Southern blot analysis and sequencing of the genomic locus.

ZFNs were designed to cleave a predefined site within the genomic coding sequence of the endogenous human HLA-A genes (Figure 1A-B). Expression of these artificial nucleases in human cells should eliminate expression of HLA-A molecules via error-prone repair of introduced double-strand breaks leading to disruption of the reading frame. To evaluate the ability of these ZFNs to disrupt HLA-A expression, we initially used the HEK293 cell line, which coexpresses HLA-A*03:01 (HLA-A3) and HLA-A*02:01 (HLA-A2). After transfecting HEK293 cells with expression plasmids encoding the ZFNs, we used allele-specific PCR and the Surveyor nuclease assay²²  to quantify the level of gene modification at the anticipated ZFN target sites. These data demonstrated ∼10% modification of HLA-A3 and ∼6% modification of HLA-A2, using this ZFN pair (Figure 1C).

We used limiting dilution to obtain single-cell clones from the ZFN-modified HEK293 pool to assess the effect of disrupting HLA-A expression. Sequencing revealed clones that carried insertions or deletions (indels) within the expected ZFN-binding sites in HLA-A2, HLA-A3, or both alleles (supplemental Figure 2), which resulted in a frame shift leading to premature termination of translation. Because the steady-state level of HLA-A expression in HEK293 is low compared with hematopoietic cells, such as an EBV-LCL, we exposed the HEK293 cells to pro-inflammatory cytokines known to augment HLA levels.²⁶  The addition of IFN-γ and TNF-α increased expression of both HLA-A alleles in parental HEK293 cells (Figure 2A, top). In contrast, ZFN-treated single-cell–derived HEK293 clones harboring mutations in HLA-A2 and/or HLA-A3 did not express these proteins even after induction by IFN-γ and TNF-α (Figure 2A, bottom 3 panels). Next, we asked whether the loss of HLA-A expression on the ZFN-modified clones would preclude T-cell recognition by challenging with HLA-A3- and HLA-A2-restricted CTLs. As expected, an HLA-A3-restricted CTL clone 7A7 demonstrated robust specific lysis of the HLA-A3⁺ parental HEK293 line loaded with a serial dilution of the cognate peptide (Figure 2B, top). Clone 8.18 that had lost expression of HLA-A2 allele, but is wild-type at HLA-A3, was also lysed by this HLA-A3-restricted CTL. In contrast, when pulsed with the same peptide, clones 18.1 and 83, which had been edited to eliminate HLA-A3 expression, were not lysed by the HLA-A3-restricted CTL clone 7A7 (Figure 2B, top). We also evaluated the cytolytic activity of an HLA-A2-restricted CTL clone GAS2B3-5 and observed robust killing activity when presented with the parental HEK293 or the HLA-A2 wild-type clone 18.1, whereas the ZFN-modified HLA-A2^neg clone 8.18 and the HLA-A2/A3 double-knock out clone 83 were spared from lysis (Figure 2B, bottom). These data demonstrate that treatment with ZFNs completely eliminates HLA-A expression, resulting in protection from HLA-A restricted CTL-mediated killing, even under proinflammatory conditions known to upregulate endogenous HLA-A expression.

To extend our results to clinically relevant primary cells, we evaluated the activity of the HLA-A-specific ZFNs in human T cells. Because ZFNs require only short-term expression to achieve stable disruption of desired target genes, we transiently expressed ZFNs from an in vitro–transcribed mRNA. Electro-transfer of mRNA encoding the ZFNs into T cells from an HLA-A2 homozygous donor (HLA-A2 being the most common HLA-A allele in whites²⁹ ) rendered ∼19% of these T cells HLA-A2-negative (Figure 3A, top). We have previously demonstrated that transiently lowering the incubation temperature after transfection can increase ZFN activity.^20,21  Subjecting electroporated T cells to a transient hypothermia²¹  elevated the proportion of HLA-A2^neg cells to up to 57% in an mRNA dose-dependent manner (Figure 3A, bottom). Functional elimination of HLA-A2 was further confirmed by probing HLA-A2-transduced 721.221 cells with a pan class I antibody (supplemental Figure 3).

With a view to the clinical application of the HLA-targeted ZFNs, we evaluated the use of the “high-fidelity” obligate heterodimeric FokI domains EL:KK, which are designed to decrease potential off-target cleavage events by preventing homodimerization.³⁰  Use of mRNA encoding the EL:KK ZFN variants of ZFN-L and ZFN-R resulted in a marked increase in HLA-A^neg T cells, eliminating HLA-A expression in up to 52% of the T-cell population, despite limiting doses of mRNA (2.5 µg each ZFN; Figure 3B). A single round of HLA-A-positive T-cell depletion with antibody-coated paramagnetic beads readily increased the HLA-A2^neg T-cell fraction to more than 95% of the population without affecting CD4 or CD8 expression (Figure 4A). Analysis of this HLA-A2^neg population by the Surveyor nuclease assay (Figure 4B) and direct DNA sequencing (Figure 4C) revealed nearly 100% editing of the HLA-A2 alleles precisely within the region targeted by the ZFNs. Together, these data demonstrate that ZFN-driven genome editing can rapidly generate an HLA-A^neg T-cell population.

To demonstrate the potential utility of HLA editing, we next focused on a specific class of cells that could be broadly used in allogeneic settings after elimination of HLA expression: cytotoxic T cells genetically modified to express a universal CAR to redirect specificity toward tumor-associated antigens independent of HLA recognition.³¹  Indeed, we and others are currently infusing patient-specific CAR⁺T cells for the investigational treatment of CD19⁺ malignancies.^2,3,31-35  Recently, we have published results indicating that CAR⁺T cells retain redirected specificity for CD19 when ZFNs are used to eliminate endogenous αβTCR expression.²⁰  Such TCR-edited T cells demonstrate both improved potency and safety (amelioration of graft-versus-host disease) in vivo.³⁶  To further explore the potential of “off-the-shelf” T-cell therapies, we investigated whether ZFNs could eliminate HLA-A expression in CD19-specific CAR⁺T cells. PBMCs genetically modified by synchronous electro-transfer of DNA plasmids derived from the SB transposon/transposase system followed by selective propagation on CD19⁺ aAPC³⁷  resulted in expression of the CD19-specific CAR (designated CD19RCD28) in more than 90% of the T cells. These SB and aAPC platforms have been adapted for human application in 4 clinical trials (Clinicaltrials.gov trials NCT01497184, NCT01362452, NCT00968760, and NCT01653717). CAR⁺T cells were subsequently electroporated with in vitro–transcribed mRNA encoding the obligate heterodimeric variants of the HLA-A ZFNs. ZFN treatment successfully disrupted HLA-A2 expression in CAR⁺ T cells without selection, and this population was readily enriched to ∼99% HLA-A2^neg cells by negative selection for HLA-positive cells (Figure 5A). These cells were shown to maintain this new phenotype, as after 50 days of continuous coculture on CD19⁺ aAPC, ∼94% of the CAR⁺T cells remained HLA-A2^neg (not shown). Importantly, these HLA-A2^neg T cells evaded attack by HLA-A2-restricted CTL (Figure 5B) and maintained their antitumor activity, as evidenced by CAR-dependent lysis of CD19⁺ tumor targets (Figure 5C-D).

To broaden the application of allogeneic cells for therapeutic applications, including tissue regeneration, we sought to generate hESCs capable of evading T-cell recognition. By definition, all hESCs are allogeneic with respect to potential recipients and, on differentiation, will upregulate expression of HLA.³⁸  We genetically modified the HLA-A2⁺/A24⁺ hESC line WIBR3 with either mRNAs or DNA plasmids encoding ZFNs targeting HLA-A to evaluate the use of ZFNs for the generation of HLA-A^neghESC. To facilitate generation of HLA-A^neghESC, we codelivered a donor DNA plasmid encoding the puromycin resistance gene flanked by regions of homology surrounding the ZFN target site to mediate targeted integration by homology-directed repair (supplemental Figure 4). Puromycin-resistant clones were screened for modification of the HLA-A alleles by PCR sequencing of the ZFN target region and by Southern blot analysis of the targeted region, using probes located outside the donor homology groups. These clones, containing mutations in the ZFN target region in both HLA-A alleles, were differentiated into fibroblast-like cells and compared with a similarly differentiated unmodified parental hESC line. HLA expression was induced by treatment with IFN-γ and TNF-α and analyzed by flow cytometry. Although the parental cell line exhibited strong expression of both HLA-A alleles, all 3 knockout lines lacked cell surface expression of both HLA-A alleles (Figure 6). These data demonstrate the portability of the HLA-A knockout approach to hESCs, which may be a necessary step for persistence of these cells and their progeny posttransplantation.

A strategy to prevent rejection by the recipient of an allograft without the need to suppress host immunity or confine transplanted cells to privileged sites avoiding immune surveillance would significantly broaden the transplantation of clinical-grade allogeneic cells. However, to prevent donor-derived cells from being recognized by resident T cells, a graft would need to be devoid of expression of mismatched HLA, as antigen-specific TCRαβ can be activated by as few as 3 to 5 peptide/HLA complexes on the target cell.³⁹  Indeed, the potential for immune-mediated rejection is further compounded by the high precursor frequency of allo-reactive T cells (1 in 10³-10⁴).⁴⁰  We provide the proof of principle that ZFNs can target multiple HLA-A alleles for the complete and permanent elimination of HLA-A expression, resulting in cells that can evade T-cell recognition. We observed no apparent detrimental effects on cell growth or function of genome-edited T cells or hESCs as a result of ZFN treatment per se or because of the elimination of HLA-A expression itself. Indeed, T cells expressing a CAR that recognizes CD19, independent of HLA, have shown antitumor effects^2,3,33,34,41  and retained their antitumor activity when edited to be HLA-A-negative, suggesting that the ZFN and CAR technologies could be combined to generate off-the-shelf cellular immunotherapy options for patients with cancer.

While increasing the complexity associated with genetically editing multiple loci, the therapeutic strategy to generating off-the-shelf or universal T cells may be further enhanced if coupled with the elimination of other HLA class I as well as HLA class II molecules. This may be accomplished by simultaneous transfection of several pairs of ZFNs targeting a variety of genes.⁴²  Indeed, we validated this strategy by simultaneous modification of T cells with ZFNs to prevent expression of both HLA-A and TCR (supplemental Figure 5). This observation supports our future work to eliminate multiple target genes (including classical HLA molecules other than HLA-A) that may be achieved by a single electro-transfer event of mRNA species encoding multiple ZFNs.

The clinical application of infusing HLA^neg cells is appealing, as a single-donor–derived allograft can be preprepared for transfer across transplantation barriers into multiple recipients. This may be achieved by enriching for HLA^neg cells or taking advantage of in vivo selection after administration because of immune-mediated recognition and elimination of remaining disparate HLA⁺ unmodified cells. An additional potential clinical application of allogeneic cell transplantation is engraftment of cells derived from hESCs. These pluripotent stem cells are self-renewing, and yet can be differentiated in vitro into a range of cell and tissue types with the potential to support for cell replacement therapies. However, transplantation of hESCs, which by their definition are allogeneic with respect to the recipient, is complicated by the expression of disparate HLA on derived cells and tissues. We demonstrated a potential solution to this problem; namely, the use of ZFNs to permanently eliminate the expression of specific HLA molecules from the cell surface and thus prevent immune recognition of these epitopes. Even autologous induced pluripotent cells (iPSCs) may benefit from this approach, as T-cell mediated rejection of iPSC derivatives can result from the presentation of immunogenic antigens in the context of HLA.⁴³ 

Immune-mediated pressure on selecting for HLA expression is noted in some patients with idiopathic aplastic anemia who exhibit loss of heterozygosity of chromosome 6p, resulting in the loss of 1 HLA haplotype.⁴⁴  In these patients, HLA loss among populations of hematopoietic stem cells apparently leads to immune escape from pathogenic T cells recognizing auto-antigens and results in sustained and functional hematopoiesis. In addition, this loss of HLA did not appear to trigger recognition of hematopoietic stem cells exhibiting loss of heterozygosity by endogenous NK cells. Thus, absence of HLA expression does not necessarily result in NK-cell–mediated lysis.

A concern for loss of classical HLA class I expression is that this may render some allogeneic cells as targets for NK-cell–mediated cytotoxicity based on absence of ligands for KIR.⁴⁵  A potential solution to this issue may be expression of nonclassical HLA class I molecules such as HLA-E or HLA-G, which have been shown to protect cells from NK-cell–mediated lysis^46-49  and are much less polymorphic than classical HLA. Indeed, we have validated the approach of preventing NK-cell lysis by enforced expression of HLA-E and/or HLA-G on the HLA class I^null721.221 cell (supplemental Figure 6). This provides a solution to forestall elimination of administered HLA^neg allogeneic cells by recipient NK cells.

In summary, we demonstrate that transformed, primary, and pluripotent stem cells can be permanently modified by ZFNs to eliminate HLA-A expression. Our approach has the potential for rapid application in humans, as we employ mRNA electroporation (a process with growing use in GMP manufacturing of cellular products), resulting in the transient expression of ZFNs (which as a class have already been taken to the clinic). The ability to generate a priori cells from a single allogeneic donor to be infused across transplantation barriers into multiple recipients represents a significant step toward our goal of on-demand therapy with universal cells that can be predeployed at multiple sites and infused when needed, rather than when the cells are available.

We thank C. June (University of Pennsylvania) for help generating and providing aAPC clone 4; M. Fernández-Viña (Stanford University) for his assistance with HLA-typing and providing antibody; A. Multani (MD Anderson Cancer Center) for her assistance with karyotyping; L. Zhang, S. Hinkley, and A. Vincent (Sangamo Biosciences) for generation of the ZFN reagents; S. Abrahamson (Sangamo Biosciences) for editing the manuscript; and E. Lanphier (Sangamo Bioscience) for encouragement and support. The Cancer Center Support Grant core-supported facilities “Characterized Cell Line Core” undertook fingerprinting of cell lines.

This work was supported by grants from the Cancer Center (CA16672); R01 (CA124782, CA120956, CA141303; CA141303); R33 (CA116127); P01 (CA148600); Burroughs Wellcome Fund; Cancer Prevention and Research Institute of Texas; Caryn Papantonakis; CLL Global Research Foundation; DARPA (Defense Sciences Office); Department of Defense; Estate of Noelan L. Bibler; Gillson Longenbaugh Foundation; Harry T. Mangurian, Jr, Fund for Leukemia Immunotherapy; Institute of Personalized Cancer Therapy; Leukemia and Lymphoma Society; Lymphoma Research Foundation; The MD Anderson Cancer Center’s Sister Institution Network Fund; Miller Foundation; Mr Herb Simons; Mr and Mrs Joe H. Scales; Mr and Mrs Rick Calhoon; Mr Thomas Scott; National Foundation for Cancer Research; Paula Gavrel Asher Foundation; Pediatric Cancer Research Foundation; Robert J. Kleberg, Jr and Helen C. Kleberg Foundation; Uehara Memorial Foundation; and William Lawrence and Blanche Hughes Children's Foundation. This work was also supported by grants from the National Institutes of Health (R37, CA084198; and R01, CA087869 and HD045022 to R.J.).

Contribution: H.T., A.R., and F.S. designed, performed experiments, analyzed the data, and wrote the paper; C.Y., Y.Z., D.L.C., J.C.M., N.L., and Z.Z. performed experiments; E.H.W. and S.S.T. provided CTL clones; H.H. coordinated experiments; E.J.R., P.D.G., and M.C.H. designed experiments and wrote the paper; P.K., D.A.L., and R.E.C. analyzed the data; R.J. designed experiments and analyzed the data; and L.J.N.C. conceived the idea, coordinated the project, designed experiments, and wrote the paper.

Conflict-of-interest disclosure: A.R., Z.Y., N.L., Z.Z., J.C.M., M.C.H., E.J.R., P.D.G. are current (or former) employees of Sangamo BioSciences Inc. The remaining authors declare no competing financial interests.

Correspondence: Laurence J. N. Cooper, The University of Texas MD Anderson Cancer Center, Pediatrics, Unit 907, 1515 Holcombe Blvd, Houston, TX 77030; e-mail: ljncooper@mdanderson.org.