Research ArticleGENE EDITING

Gene editing to induce FOXP3 expression in human CD4+ T cells leads to a stable regulatory phenotype and function

See allHide authors and affiliations

Science Translational Medicine  03 Jun 2020:
Vol. 12, Issue 546, eaay6422
DOI: 10.1126/scitranslmed.aay6422

Enforced editing

Various autoimmune diseases could potentially be treated with regulatory T cells (Tregs), but there are many hurdles between this idea and clinical execution. Honaker et al. devised a gene-editing strategy to enforce expression of FOXP3, the master Treg transcription factor, in CD4+ T cells isolated from human peripheral blood, thereby overcoming limitations of Treg isolation and expansion. Resulting stable FOXP3 expression enabled a suppressive phenotype in vitro, and the edited cells were also functional in a xenogeneic graft-versus-host disease model and an experimental autoimmune encephalitis model. This approach has the potential to rapidly translate to clinical use.


Thymic regulatory T cells (tTregs) are potent inhibitors of autoreactive immune responses, and loss of tTreg function results in fatal autoimmune disease. Defects in tTreg number or function are also implicated in multiple autoimmune diseases, leading to growing interest in use of Treg as cell therapies to establish immune tolerance. Because tTregs are present at low numbers in circulating blood and may be challenging to purify and expand and also inherently defective in some subjects, we designed an alternative strategy to create autologous Treg-like cells from bulk CD4+ T cells. We used homology-directed repair (HDR)–based gene editing to enforce expression of FOXP3, the master transcription factor for tTreg. Targeted insertion of a robust enhancer/promoter proximal to the first coding exon bypassed epigenetic silencing, permitting stable and robust expression of endogenous FOXP3. HDR-edited T cells, edTregs, manifested a transcriptional program leading to sustained expression of canonical markers and suppressive activity of tTreg. Both human and murine edTregs mediated immunosuppression in vivo in models of inflammatory disease. Further, this engineering strategy permitted generation of antigen-specific edTreg with robust in vitro and in vivo functional activity. Last, edTreg could be enriched and expanded at scale using clinically relevant methods. Together, these findings suggest that edTreg production may permit broad future clinical application.


Autoimmune diseases such as type 1 diabetes mellitus (T1D), multiple sclerosis, and systemic lupus erythematosus (SLE) are chronic, life-threatening conditions that result from alterations in self-tolerance, leading to aberrant immune activity and end-organ pathology. Immune tolerance toward self-antigens is orchestrated at multiple levels of development and function. In the thymus, most developing T cells that express self-reactive T cell receptors (TCRs) are either deleted or converted into immunosuppressive CD4+ regulatory T cells (Tregs) (1, 2). The TCR repertoires of thymic Treg (tTreg) are thus skewed toward self-reactive specificities, crucial for their role in maintaining peripheral self-tolerance. tTregs act to suppress inflammation by both antigen-specific and antigen-independent mechanisms, including direct cell-cell contact and secretion of soluble cytokines that dampen cytotoxic effector T cell (Teff) function (3, 4). Multiple autoimmune conditions are characterized by dysfunctional or dysregulated Treg compartments (5, 6), and adoptive transfer of functional Treg has been explored in mouse models and early-phase clinical trials as a potential therapy for autoimmune disease (7, 8).

Therapeutic application of adoptively transferred Treg to treat autoimmune disease is limited, in part, by the scarcity of tTreg in the peripheral blood; tTreg with relevant TCR specificities would be even scarcer. Other challenges include expanding Treg to therapeutic numbers ex vivo while maintaining suppressive function and the ability of transferred cells to persist and proliferate after infusion. An alternative strategy is to convert the abundant circulating Teffs into functional Treg in vitro. Under defined culture conditions, Teff may take on properties of Treg, but while transfer of these cells has ameliorated disease in mouse models of autoimmunity, their adopted phenotype has proven unstable in vivo (9, 10). Stability of a Treg phenotype in converted cells is a critical concern in clinical applications for autoimmune disease, where the Teff repertoire may be skewed toward cells with self-reactive specificity and reversion might aggravate disease (11, 12). Treg-suppressive activity is driven by the transcription factor Forkhead box protein 3 (FOXP3) (13, 14). High FOXP3 expression in the absence of TCR activation is a defining marker of Treg and a functional requisite for Treg activity (1517). Inactivating FOXP3 mutations result in severe, multiorgan autoimmunity called immune dysregulation polyendocrinopathy enteropathy X-linked (IPEX) syndrome and is often fatal within months of birth (18, 19). Mice with mutations in Foxp3 also develop fatal autoimmunity and lack functional Treg (20).

Retroviral vectors delivering a FOXP3 complementary DNA (cDNA) expression cassette have been used to confer Treg-like properties to conventional T cells in humans and mice (2123). FOXP3-transduced CD4+ T cells express phenotypic markers and cytokine profile of Treg and suppress the expansion of activated Teff. The stability of this phenotype is dependent on the strength of the promoter and the maintenance of high expression of FOXP3, avoiding silencing of the transgene (21). However, the randomly integrating vector results in a heterogeneous cell product, with variable numbers of viral integration sites scattered nonspecifically across the genome. This heterogeneity may be undesirable in a clinical product, and there is a concomitant risk of genotoxicity, as well as vulnerability to silencing by local regulatory elements at sites of viral integration. Homology-directed repair (HDR)–mediated gene editing provides the ability to integrate a transgene at a precise genomic sequence, theoretically increasing the safety and stability of transgene expression.

Here, we use a targeted, high-efficiency gene-editing strategy to introduce a strong promoter precisely at the endogenous FOXP3 locus in primary human CD4+ T cells by HDR, enforcing stable expression of high quantities of FOXP3. The resulting gene-edited Treg-like cells (edTreg) exhibit a phenotype and cytokine profile that closely mirrors purified peripheral blood tTreg, and manifest robust immunosuppression in vitro and in vivo.


A gene-editing strategy to bypass epigenetic silencing in human peripheral blood CD4+ T cells

Transcription of FOXP3 in conventional T cells is epigenetically silenced via methylation events occurring within the Treg-specific demethylated region (TSDR) in intron 1 of FOXP3 (24). We hypothesized that insertion of a transcriptional promoter downstream of the TSDR, but upstream of the protein start codon, could drive constitutive FOXP3 expression in conventional T cells. We have previously developed methodologies to insert novel sequences into genetic loci of interest by HDR using engineered nucleases and adeno-associated virus (AAV) donor templates (2529). We therefore designed a gene-editing strategy to insert an MND (myeloproliferative sarcoma virus enhancer, negative control region deleted, dl587rev primer-binding site substituted) promoter into the FOXP3 locus to drive expression of endogenous FOXP3 in human T cells (Fig. 1A). For ease of detection of edited cells by flow cytometry, our initial editing strategy also introduced a coding sequence for green fluorescent protein (GFP) that upon successful HDR would become in frame with the N terminus of endogenous FOXP3 coding sequence. This mimics the sequence of Foxp3tm2Ayr mice (30), which have normal Treg development. Our gene-editing strategy should thus result in MND-regulated transcription of the endogenous FOXP3 gene, encoding an N-terminal GFP-FOXP3 fusion protein.

Fig. 1 Gene editing to impose stable FOXP3 expression in primary human CD4+ peripheral blood cells.

(A) Diagram of FOXP3 locus before (top) and after (bottom) successful gene editing using FOXP3 TALENs and the AAV FOXP3 [MND.GFP knock-in (ki)] donor template (middle). Exons are represented by blue boxes (on the basis of historical usage, exon 1 is used here to refer to first coding exon; the second exon within the FOXP3 gene); the position of the homology arms relative to FOXP3 is indicated by gray shading; green arrows show endogenous or introduced transcriptional start sites. The location of the TSDR is shown; most CD4+ T cells have a methylated TSDR that blocks transcription from the FOXP3 promoter. After editing, the MND promoter will drive expression of chimeric GFP-FOXP3 protein. (B) Timeline of key steps of gene editing and cell analysis. CD4+ T cells were activated with CD3/CD28 beads followed by gene editing (TALEN mRNA and AAV delivery) or controls (electroporated without TALEN mRNA and/or AAV). After further culture, the success of HDR was analyzed by flow cytometry. Edited cells were FACS-sorted and expanded for molecular, phenotypical, and functional analysis. (C) Representative flow plots (day 5; 2 days after editing) showing GFP and FOXP3 coexpression in cells edited with TALEN and AAV donor template versus mock-edited human T cells. (D) Scatter plot summarizing editing rates (%GFP+) as determined by flow cytometry 2 to 4 days after editing (n = 27 experiments using five different PBMC donors). Bar shows mean ± SD. (E) Schematic diagram (top) showing position of PCR primers used to detect HDR recombination events in gDNA and the expected amplicon sizes. The primers bind to FOXP3 outside of the AAV homology region. Representative agarose gel of PCR products from edited and control T cells is shown. MW, molecular weight.

We designed transcription activator-like effector nuclease (TALEN) pairs (sense and antisense) to induce DNA double-strand breaks near the FOXP3 start codon (fig. S1A). The donor DNA template was delivered using a recombinant AAV6 vector containing FOXP3 homology regions 5′ and 3′ of the TALEN-cleavage site flanking an MND promoter and GFP coding sequence (Fig. 1A). After purification and activation using CD3/CD28 beads, primary human CD4+ T cells were electroporated with forward and reverse FOXP3-specific TALEN mRNA pairs and subsequently cultured in medium containing the AAV6 donor template (Fig. 1B). To serve as a negative control, a portion of the activated CD4+ T cells were “mock”-edited by electroporating without TALEN mRNA and then culturing in medium without AAV. In genomic DNA (gDNA) extracted from CD4+ T cells treated with TALEN mRNAs only, we found an 80% frequency of DNA insertions or deletions (indels) at the target site, indicating efficient on-target cutting (fig. S1B). We did not detect any indels within any of the 15 highest-scoring possible off-target sites (table S1), suggesting target-specific cleavage using these TALENs. Control cells treated with the AAV donor template alone did not express detectable GFP at any time after transduction. Cells that received both FOXP3-specific TALEN mRNA and AAV expressed GFP, and GFP expression correlated linearly with that of intracellular FOXP3 (Fig. 1C), consistent with successful HDR driving expression of the FOXP3 gene with an N-terminal GFP fusion. On average, our editing protocol resulted in 37% GFP+ cells when cells were treated with both TALEN mRNA and the AAV donor template (Fig. 1D). To verify successful editing at the molecular level, GFP+ cells were purified by fluorescence-activated cell sorting (FACS) and gDNA amplified by polymerase chain reaction (PCR) using FOXP3-specific primers outside of both homology arms of AAV donor template (Fig. 1E). Only gDNA from cells treated with both TALEN and AAV gave the expected increase in amplicon size due to insertion of donor template (Fig. 1E). In the sample treated with TALEN and AAV, gDNA was isolated from FACS-purified GFP+ cells. Hence, most amplicons contain MND-GFP sequences. The two faint bands in the sample are nonspecific amplification products from the PCR. In addition, sequence analysis of GFP-sorted gene-edited cells confirmed the presence of the donor template integrated seamlessly at the FOXP3 locus (fig. S1C). These results validated our HDR approach as a method to induce high expression of the FOXP3 transcription factor by seamless, targeted gene editing in human CD4+ T cells.

FOXP3 editing promotes expression of proteins that discriminate thymic-derived Treg from Teff

The expression of FOXP3 is critical in establishing and maintaining the Treg transcriptome. We first compared the phenotype of gene-edited T cells expressing FOXP3 driven by the inserted MND promoter (referred to hereafter as edTreg) with that of tTreg or Teff isolated from the same peripheral blood sample. tTreg and Teff were purified by FACS based on the following cell surface marker staining: CD4+CD25++CD127 (tTreg) and CD4+CD25 (Teff). Example flow plots from before and after sorting are shown in fig. S2. Teff, tTreg, GFP+ edTreg, and mock-edited cells were expanded in parallel by culturing with CD3/CD28 beads followed by immunostaining for intracellular and surface markers associated with tTreg (Fig. 2A). Both tTreg and GFP+ edTreg expressed similarly high amounts of FOXP3 and CD25 (interleukin-2 receptor α or IL-2Rα) and reduced CD127 (IL-7Rα) (Fig. 2, B and C). Like tTreg, edTreg up-regulated expression of CTLA-4 and ICOS (Fig. 2, B and C). It is known that expression of FOXP3 and other Treg markers is transiently altered in Teff after activation (31); accordingly, both Teff and mock-edited cells showed moderate expression of Treg-associated markers (such as CD25 and FOXP3 up-regulation) in response to cell activation. After bead removal and subsequent cell culture, edTreg retained high expression of FOXP3, CD25, CTLA-4, and ICOS, whereas mock-edited cells diminished expression of Treg-associated markers. edTreg expressed intermediate quantities of the transcription factor Helios relative to tTreg or mock-edited T cells. At the same time in culture, we also compared expression of intracellular cytokines interferon-γ (IFN-γ), IL-2, tumor necrosis factor–α (TNF-α), and IL-4 after a 5-hour stimulation with phorbol 12-myristate 13-acetate (PMA) and ionomycin (Fig. 2D). Expression profiles of IFN-γ, IL-2, TNF-α, and IL-4 (Fig. 2D) were similarly reduced in tTreg and edTreg compared to mock-edited cells and Teffs. To assess whether CD25 up-regulation was correlated with increased downstream signaling, we measured the phosphorylation of signal transducer and activator of transcription 5 (STAT5) in response to varying IL-2 concentrations. edTreg had an increased sensitivity to low doses of IL-2 in comparison to mock-edited cells (fig. S3).

Fig. 2 Characterization of FOXP3-edited primary human CD4+ T cells.

Phenotype of FACS-purified tTreg (CD4+CD25++CD127) and Teff (CD4+CD25) T cells compared to mock-edited cells and edTregs. All flow plots shown were gated on live/lymphocyte/CD4+ and are representative findings. (A) Schematic diagram showing source of cells and timeline of culturing and manipulations for direct comparison of Teff, tTreg, edTreg, and mock-edited cell immunophenotypes from a single human PBMC sample. (B) Representative immunophenotype of edTreg and mock-edited cells compared with autologous tTreg and Teff. Flow plots show GFP and Treg-identifying markers versus FOXP3. (C) Histograms of markers shown in (B). (D) Profile of cytokines expressed in T cells after activation (PMA/ionomycin) and intracellular staining. Data in (B) to (D) are representative of multiple independent experiments including at least three independent human donors. (E) Hierarchical clustering of quantities of 30 “Treg gene signature” transcripts as determined by RNAseq using RNA obtained from activated and sorted Teff (CD4+CD25), tTreg (CD4+CD25++CD127), or edTreg. Each lane reflects data derived from the designated cell population generated from one of four independent human donors.

We next performed a global comparison of the transcriptomes of edTreg, tTreg, and Teff populations isolated from four different human donors using RNA sequencing (RNAseq). RNA was collected directly from FACS-purified cells after a 48-hour stimulation using CD3/CD28 beads and a subsequent 48-hour rest period (without bead stimulation; detailed experimental timeline is shown in fig. S4). Using rlog-transformed count values on a subset of genes showing the highest degree of global expression variation, we performed hierarchical clustering of the samples. We also looked at the deviation from the gene’s average expression across all samples and plotted this as a heat map, with samples showing the highest degree of similarity in gene expression being nearest to each other. Figure S5 shows the 100 most variable genes between each T cell subset. We observed the highest degree of clustering between the same T cell subset (Teff, tTreg, or edTreg). The next highest degree of similarity was observed between the tTreg and edTreg subsets. This hierarchy was maintained when we performed the clustering based on 30 transcripts previously shown to distinguish tTreg from Teff regardless of activation status (Fig. 2E) (32). In summary, edTregs differ from mock-edited T cells in that they exhibit phenotypic characteristics of tTreg, including expression of canonical surface and intracellular markers of tTreg, a tTreg-like cytokine production profile, greater sensitivity to IL-2, and a post-stimulation gene signature that is similar to sorted tTreg derived from the same donor.

In vitro immunosuppression by edTreg is dependent on functional FOXP3

A defining characteristic of tTreg is their ability to suppress the activation of autologous Teff. We first tested whether edTreg could suppress the proliferation of CD4+ T cells after polyclonal stimulation in vitro. We sorted edited CD4+ T cells based on GFP expression. GFP+ versus GFP sorted fractions were cocultured with fluorescently labeled autologous CD4+ Teff undergoing CD3/CD28 bead stimulation for 96 hours. Proliferation was read out using flow cytometry based on dilution of the fluorescent dye. Only the GFP+ cells suppressed the proliferation of Teff, showing that edTreg function is linked to the FOXP3+/GFP+ phenotype (Fig. 3A).

Fig. 3 edTregs generated using CD4+ T cells from healthy donors, but not from patients with IPEX, suppress the proliferation of activated autologous Teffs in vitro.

(A) Flow plots and histogram overlays of stimulated Teffs after 96-hour coculture with GFP+ or GFP fractions of edited CD4+ T cells. Division of Teff, labeled with CellTrace before coculture, results in dilution of CellTrace signal intensity per cell. The gated population in flow plots indicates dye-labeled Teffs. Histogram overlays are from the gated population shown in flow plots. (B) Flow plots and histogram overlays of CellTrace-labeled Teff cocultured for 96 hours with mock-edited or edTreg. All T cells used in each coculture study were autologous cells obtained from peripheral blood from healthy control or patients with IPEX. (C) Graph of geometric mean fluorescence intensity (MFI) of GFP-FOXP3 fusion protein in edTreg generated from healthy control (n = 1) and cells from patients with IPEX (n = 2).

Because our editing strategy up-regulates expression of the endogenous FOXP3 transcript by knocking in a strong upstream promoter, edTregs generated using input CD4+ T cells isolated from patients with IPEX, with loss-of-function mutations in FOXP3, are expected to be nonsuppressive. To directly demonstrate that edTreg immunosuppression is based on activation of endogenous expression of a functional FOXP3 protein, we compared the ability of editing to confer a suppressive phenotype between a healthy control donor and two patients with IPEX with FOXP3 mutations p.I363V (IPEX1) and p.A80Dfs*49 (IPEX2). Edited CD4+ T cells from the healthy subject gained suppressive function, whereas those from the patients with IPEX did not (Fig. 3B). Note that edTreg derived from CD4+ T cells isolated from IPEX subjects expressed similar amounts of GFP-FOXP3 fusion protein compared with edTreg derived from the healthy control donor (Fig. 3C). These results demonstrate that edTreg from healthy donors can suppress expansion of activated Teff in vitro, and this property is driven by increased expression of the functional endogenous FOXP3 in edited cells.

Comparison of edTreg with T cells expressing lentiviral-delivered FOXP3

Lentiviral (LV) transfer of a FOXP3 cDNA expression cassette into conventional T cells has been shown to confer a Treg-like phenotype and suppressive characteristics in vitro and in vivo (21, 33). As a comparison for our editing strategy, we generated an LV construct to deliver a cDNA encoding the same GFP-FOXP3 fusion protein made by our edTreg (Fig. 4A). The FOXP3 gene is located on the X chromosome, and because our experiments were conducted in CD4+ cells derived from male donors, there is only one FOXP3 allele per cell. The gene-editing and viral transduction procedures produced similar proportions of GFP+FOXP3+ cells (Fig. 4A). In contrast to the single copy of donor DNA per cell in edTreg, LV-treated cells (LV Treg) had an average of 3.0 LV integrations per GFP+ cell genome. Despite their copy number differences, the GFP mean fluorescence intensity (MFI) of GFP+ cells was consistently lower in LV Treg than in edTreg (Fig. 4B), perhaps indicating more efficient expression from a genomic versus cDNA context or LV integration in transcriptionally less permissive loci. FACS-purified LV Treg cells lost GFP expression more rapidly during 5 weeks in culture than edTreg (Fig. 4C; starting % GFP+ > 99% for both). When analyzed approximately 2 weeks after sorting, both methods of enforcing FOXP3 expression skewed the Teff toward tTreg phenotypes, including up-regulation of CD25, CTLA-4, and Helios and down-regulation of IL-2, TNF-α, and IFN-γ (Fig. 4D). Except for FOXP3, the percentage of cells expressing Treg markers, as well as the mean expression (as assessed by MFI) per cell, were similar between edTreg and LV Treg. We also demonstrated increased expression of the complement binding protein CD59 (based on flow cytometry staining; Fig. 4D), further validating our RNAseq data derived from tTreg and edTreg. Functionally, edTreg and LV Treg were equivalent in their ability to suppress polyclonal Teff proliferation in vitro (Fig. 4E).

Fig. 4 Comparison of FOXP3-edited and FOXP3 LV–transduced T cells.

(A) Diagram of LV construct: MND promoter will drive expression of a transcript encoding identical GFP-FOXP3 fusion protein as that of edTreg; transcript contains Woodchuck Hepatitis Virus Posttranscriptional Regulatory Element (WPRE) and poly(A) signals for efficient nuclear export and mRNA stability. Below are representative flow plots showing FOXP3 and GFP expression in mock-edited T cells, or sorted tTreg (CD4+CD25++CD127), edTreg, and LV Treg (CD4+GFP+), all after ≥14-day expansion in vitro with CD3/CD28 beads. (B) Mean viral copy number (±SD) of LV-transduced sorted cells (left; n = 6). Scatter plots (right) show the MFI of the GFP+ population for each sample (n = 5; P value from two-tailed Student’s t test). (C) Plot showing data points and simple linear regression of % GFP+ cells over time in culture after FACS purification; edTreg (n = 4) and LV Treg (n = 6); data from six experiments. Dashed lines indicate 95% confidence intervals; P value was obtained using an F test. (D) Bar graphs showing mean % of cells (top) and MFI (bottom) by flow cytometry staining for the proteins indicated. Viable singlets were further gated on CD4+GFP+ (LV Treg and edTreg), CD4+FOXP3+ (tTreg), or CD4+ (mock). In LV Treg and edTreg, MFI was calculated for the GFP-positive population only. Error bars show ±SD. An ordinary two-way ANOVA was performed, and P values were adjusted with Tukey’s multiple comparisons test. P values in black indicate comparison with mock-edited cells; those in red were comparison of groups indicated by dashed lines. (E) Percent suppression as a function of Treg or mock dilution (top). Histograms of proliferation dye at different ratios of Treg or mock to Teff (bottom). % suppression = [(% divided with no Treg − % divided with Treg)/% divided with no Treg] × 100.

Enrichment and expansion of edTreg using methods relevant for cell product manufacturing

Treg cells have a growth disadvantage in vitro and can be outcompeted by Teff (34). To obtain edTreg in numbers and purity sufficient for a therapeutic application, it will be necessary to expand edTreg in vitro and useful to enrich for successfully edited cells. After FACS-purifying GFP+ edTregs 2 days after editing (as in Fig. 2A), cells were cultured for 7 or 14 days. The edTreg expanded 23- to 48-fold in 7 or 14 days, respectively, yielding an average of 78 × 106 cells at 14 days from 1.5 × 106 sorted cells (Fig. 5A). As expected, greater expansion was achieved for mock-edited cells (158-fold). After 14 days of expansion, T cells from both mock-edited and edTreg groups exhibited a diverse TCR repertoire by Vβ spectratype analysis (fig. S6). We also assessed the methylation status of the TSDR CpGs within these cells. The TSDR affects transcription of the upstream FOXP3 promoter using epigenetic mechanisms. A TSDR with mostly nonmethylated CpGs is permissive for FOXP3 expression and is an epigenetic signature of tTreg, whereas CpGs in the TSDR of Teff are highly methylated, largely silencing FOXP3 expression. The MND promoter of edTreg is inserted 2.1 kb downstream of the TSDR, and its ability to drive FOXP3 expression in targeted cells could reflect demethylation of the TSDR or the ability to drive transcription despite TSDR methylation. Bisulfite conversion and subsequent colony sequencing of gDNA extracted from expanded edTreg and mock-edited cells revealed that nearly all CpGs within the TSDR of expanded edTreg were methylated (97%), as were those of mock-edited T cells (94%), whereas the TSDR CpGs of tTreg were minimally methylated (2.4%; Fig. 5B). These results imply that the MND promoter integrated downstream of TSDR bypasses the epigenetic silencing mediated by the TSDR at the endogenous upstream promoter. In addition, neither the presence of the MND promoter nor up-regulation of FOXP3 expression altered the TSDR methylation status.

Fig. 5 Characterization of edTreg after ex vivo clinical scale expansion or selection.

(A) Fold expansion of GFP+ edTreg and mock-edited CD4+ T cells at days 7 and 14 after an initial stimulation with CD3/CD28 beads. (B) Representative analysis of CpG methylation of the FOXP3 TSDR after expansion (top). Each row shows the methylation status (open circle = unmethylated; black circle = methylated; x = not determined) of the 12 TSDR CpGs for a single PCR clone. Bar graph (mean ± SD) summarizing % methylated TSDR CpGs from mock-edited cells and edTregs is at right (n = 3 experiments using different T cell donors). (C) AAV6 donor template designed to insert MND-LNGFR-2A upstream of FOXP3 coding sequence (top) and timeline of key steps of gene editing, cell enrichment, and analysis used in (D) to (G) (bottom). (D and E) Immunophenotype and cytokine production in mock-edited (with or without AAV delivery) versus LNGFR-enriched edTreg at day 12. (F and G) In vitro suppression assays were performed using autologous expanded tTreg, edTreg, and mock-edited cells (electroporated without TALEN mRNAs and without AAV transduction). (F) Representative histograms showing CTV fluorescence of labeled Teffs cultured alone (gray) overlaid with the histogram of CTV-labeled Teff in the coculture indicated (green). Numbers are the percentage of Teffs that have divided (bracketed). (G) Percent suppression of Teff proliferation is plotted versus the proportion of Tregs:Teffs in coculture; see Fig. 4E legend for % suppression calculation. Data shown in (D) and (E) are representative of three experiments. Data shown in (F) and (G) are representative of two experiments.

To allow enrichment of successfully edited cells, we next designed donor templates replacing the GFP coding sequence with that of a surface tag for use in cell selection. We tested two clinically relevant approaches for affinity selection: a truncated human low-affinity nerve growth factor receptor (LNGFRt; Fig. 5C) (35) and a truncated human epidermal growth factor receptor (EGFRt; fig. S7B) (36) that are expressed on the cell surface but lack intracellular signaling capabilities. Both epitope tags were in frame with a 2A ribosome-skip peptide (37) and the FOXP3 coding sequences. HDR resulted in surface expression of LNGFRt or EGFRt (linked at the C terminus to most of the 2A peptide) and endogenous FOXP3 (with one additional proline at the N terminus from the 2A peptide). After editing primary CD4+ T cells using the relevant AAV6 donor template, cell surface expression of each epitope was detected, allowing enrichment of edited cells to >80% purity using antibody affinity beads (fig. S7, A and B). LNGFRt-enriched (Fig. 5D) or EGFRt-enriched (fig. S7C, left) edTreg expressed FOXP3 at high MFIs and had tTreg-like phenotypes such as CD25++CD127−/low and reduced intracellular inflammatory cytokines (Fig. 5E and fig. S7C, right). Next, we tested the in vitro suppressive function of LNGFRt-enriched edTreg compared with mock-edited T cells or purified tTreg (Fig. 5, F and G). LNGFRt+ edTreg exhibited suppressive function equivalent to tTreg at all ratios of Treg:Teff. Thus, edTreg can be generated using a strategy that introduces a clinically relevant selection marker and can be expanded ex vivo to large numbers for potential clinical application. Enriched cells maintain a distinct immunophenotype that is independent of TSDR methylation status even after expansion ex vivo.

Adoptive transfer of autologous human edTreg suppress Teff-mediated GvHD in NSG mice

We next tested whether edTreg could suppress an immune response in vivo using a xenogeneic graft-versus-host disease (xeno-GvHD) model. In this model, injection of human Teffs into minimally irradiated immunodeficient nonobese diabetic (NOD)–scid–IL2rγNULL (NSG) mice provokes a polyclonal, major histocompatibility complex (MHC)–dependent immune response against multiple host tissues (38) and leads to symptoms of GvHD including hair loss, skin lesions, and weight loss. We first determined whether edTreg alone could become pathogenic in this model. We used AAV FOXP3[MND.GFPki] as the donor template in these experiments for ease of identifying edited cells in postmortem mouse tissues. After editing primary human CD4+ T cells, GFP+ cells were obtained by FACS and expanded for 7 days; mock-edited cells were not FACS-purified before expansion. Next, 8 × 106 edTreg or mock-edited T cells, or 4 × 106 CD4+ Teffs from the same human donor were injected into separate irradiated NSG mice and monitored for symptoms of GvHD several times weekly. All the mice treated with Teff developed overt symptoms of GvHD and were euthanized at predetermined humane endpoints as early as day 12 (fig. S8). GvHD was delayed in mice receiving mock-edited cells, with most reaching humane endpoints before 50 days after cell injection. In contrast, none of the mice receiving edTreg alone showed symptoms of GvHD and all survived through the 50-day experimental endpoint (fig. S8). Histological sections of tissues obtained from the Teff cohort showed extensive lymphocyte infiltration into multiple tissues including the lung, liver, intestine, skin, kidney, and pancreas (fig. S9) and higher histological scores relative to mice that received no T cells, confirming a diagnosis of GvHD.

To determine whether edTreg could suppress the development of GvHD in this model, 8 × 106 edTreg, mock-edited T cells, or tTreg (CD4+CD25++CD127 cells purified by FACS and expanded in culture) from the same peripheral blood sample were injected into NSG mice immediately after irradiation and allowed to engraft for 3 days before injection of 4 × 106 autologous CD4+ Teff (2:1 Treg:Teff ratio; Fig. 6A). A subset of animals from each cohort was sacrificed 14 days after injection to assess the effect of Treg on Teff expansion before the onset of overt GvHD; all remaining animals were monitored for up to 50 days, before being euthanized for tissue analyses. Animals engrafted with either edTreg or tTreg delayed severe weight loss (fig. S10A), showed lower disease score (fig. S10B), and had significantly (P = 0.0003) improved survival rates relative to recipients of mock-edited T cells (Fig. 6B).

Fig. 6 edTregs ameliorate the severity of xeno-GvHD in a humanized mouse model.

(A) Experimental timeline of murine xeno-GvHD studies; minimally irradiated NSG mice were assigned to one of the four cohorts shown, receiving freshly thawed human CD4+ T cells from PBMCs (Teff) with or without autologous edTreg, tTreg, or mock-edited cells (expanded 7 to 14 days in culture after FACS purification). (B) Kaplan-Meier survival curve of NSG mice monitored at least 50 days after Teff injection; mice were euthanized at predetermined humane endpoints. Colors and number of animals correspond to cohorts shown in (A); data are derived from five combined experiments; P values were derived using a log-rank test. (C) Flow plots of splenocytes from recipient mice on day 14 after Teff injection, showing the presence of human T cells (top); frequency (bottom left) and absolute number (bottom right) of human CD4+ T cells in specified tissues (n = 3 for each). (D) Representative flow plots showing GFP+ cells at day 14 (top); summary of % of GFP+CD4+ cells in indicated tissues at 14 or 50 days after Teff injection (bottom). (E) Phenotype of GFP+ cells in multiple tissues on days 14 and 66 (n = 3 for each), as measured by flow cytometry. (F) Flow plots showing intracellular IL-2 in GFP+ versus GFP T cells on day 50 and summary bar graph (n = 3). All flow plots in this figure are representative of repeated results; bar graphs show mean ± SEM; P values of statistically significant differences from two-way ANOVA are indicated in black type (indicating comparison with mock-edited group) or in red type (comparing groups indicated by dashed lines).

On day 14 after adoptive transfer of Teff, both the percentage and absolute number of human CD4+ T cells were decreased in the blood, spleen, liver, and lung of mice co-engrafted with edTreg or tTreg (Fig. 6C). edTregs (GFP+CD4+) were present in all examined tissues of edTreg recipient mice (Fig. 6D), ranging from a mean of 5% of CD4+ T cells in bone marrow (BM) to 28% in peripheral blood. In addition, GFPCD4+T cells (which were mostly Teff) in edTreg or tTreg recipients exhibited a more naïve (CD45ROCD62L+) phenotype and reduced expression of activation markers (CD38+ or CD40L+) compared with CD4+ T cells in mock-edited or Teff only groups (fig. S11A). Compared with mock-edited and Teff only groups, GFPCD4+ T cells in the peripheral blood of edTreg-engrafted cohorts reduced expression of intracellular proinflammatory cytokines (IFN-γ and IL-17A) and recipients of either edTreg or tTreg exhibited a lower T helper 1 (TH1)/TH2 ratio in the lung (fig. S11B). These data suggest that edTregs behave similarly to tTreg cells in this model, trafficking to multiple tissues and suppressing the activation of autologous CD4+ T cells.

GFP-expressing cells remained detectable in tissues (ranging from mean of 9% of CD4+ T cells in liver to 27% in lung) of animals engrafted with edTreg evaluated at the 50-day time point, demonstrating a stable phenotype despite a highly inflammatory environment (Fig. 6D). In most samples, GFP+ cells retained expression of Treg markers including increased FOXP3, CD25, and CTLA-4 and reduced CD127 (Fig. 6E). In some tissue samples (e.g., day 14 lung and day 50 blood and spleen), a proportion of GFP+ lymphocytes expressed CD127, consistent with a previous finding for in vivo activated Treg (39). FOXP3 was present at a high MFI in GFP+ cells in the spleen on day 50 as shown by intracellular staining; however, FOXP3 expression was difficult to quantify by our intracellular staining procedure in samples with low numbers of GFP+ cells in other organs. Importantly, and consistent with a regulatory phenotype, FOXP3+GFP+ cells on day 50 after transfer had low intracellular IL-2 compared with GFP cells (Fig. 6F). Together, these findings demonstrate sustained engraftment and functional regulatory activity of edTreg in a robust MHC class II–dependent in vivo inflammatory model.

To determine whether engineered CD4+ T cells with analogous Treg-like features and function could be developed using an alternative designer nuclease platform and a similar HDR-editing strategy, we developed a CRISPR-Cas9 reagent targeting FOXP3. We established a guide RNA (gRNA) targeting the first coding exon of human FOXP3 (upstream of the TALEN cleavage site) and an AAV HDR-donor construct with homology arms immediately upstream and downstream of Cas9-targeted sequences as shown schematically in fig. S12A. Cas9/gRNA cleaved the FOXP3 target at a high efficiency (88.7%; fig. S12B, left) with limited, in silico predicted, off-target site activity (fig. S12B, right, and table S2). Using the timeline described in fig. S12C, CRISPR-Cas9–mediated gene editing, LNGFR affinity enrichment, and cell expansion were performed to generate LNGFRt+ edTreg. Figure S12D shows an average HDR rate of 41% for generation of LNGFRt+ edTreg using independent donors. After affinity enrichment and subsequent expansion, LNGFRt+ edTreg exhibited high purity with ~80-fold expansion (fig. S12E). Comparable to TALEN-generated edTreg, Cas9-edited LNGFRt+ edTreg expressed Treg markers and reduced the production of inflammatory cytokines including IL-2, TNF-α, and IFN-γ (fig. S12, F and G). Cas9-edited LNGFRt+ edTreg also suppressed the xeno-GvHD immune response in NSG mice (fig. S12H). Because the edTreg features observed after HDR-based gene editing were equivalent between the TALEN and CRISPR-Cas9 nuclease platforms, we conclude that enforcement of MND-driven FOXP3 expression is responsible for the observed phenotypes.

Antigen-specific mouse edTregs limit priming of Teffs in a model of multiple sclerosis

For diseases where critical autoantigens are known (e.g., T1D or rheumatoid arthritis), antigen-specific Treg may be more potent for controlling disease (40, 41). To investigate edTreg function in an antigen-specific in vivo setting, we sought to edit T cells from myelin oligodendrocyte glycoprotein peptide fragment 35–55 (MOG)–specific TCR-transgenic mice [C57Bl/6-Tg(Tcra2D2,Tcrb2D2)1Kuch/J, abbreviated 2D2] (42). MOG challenge in combination with pertussis toxin disruption of the blood-brain barrier leads to experimental autoimmune encephalomyelitis (EAE), a mouse model of multiple sclerosis. Although tTregs expand and traffic to lesions in the central nervous system (CNS), MOG-induced EAE is not controlled by endogenous antigen-specific tTreg, possibly because of inflammatory Teff cytokines (43). We hypothesized that adoptive transfer of antigen-specific 2D2 edTreg could suppress Teff expansion in the periphery before these activated effectors migrate to the CNS. To test this hypothesis, we designed TALEN and AAV donor template reagents that would mimic as closely as possible the editing strategy used to generate the human edTreg (fig. S13A). After optimizing procedures for murine T cell stimulation and mRNA electroporation, we transfected murine T cells with mRNA encoding TALEN pairs specific for the first coding exon of mouse Foxp3. Seven to 9 days after transduction, ~80% of alleles contained indels based on colony sequencing of PCR-amplified gDNA (fig. S13B), indicating efficient target site cleavage. We next generated an AAV donor template that substituted mouse Foxp3 homology arms for the human sequences used in our previous HDR experiments; homology was proximal to, but not overlapping, the mouse TALEN binding sites. With slight modifications of the conditions used for human CD4+ T cell editing, including using AAV5 capsid for donor template transduction, we achieved ~25 to 30% editing rates (%GFP+) using 2D2 mCD4+ T cells isolated from spleen and lymph nodes (LNs) (Fig. 7A). GFP+ cells were phenotypically FOXP3+CD25+CTLA-4+ (Fig. 7B). CD4+ T cells were also isolated from 2D2neg littermates (C57Bl/6) and edited, resulting in edTreg with polyclonal TCR specificities for comparison with 2D2 edTreg. To assess in vivo function, we adoptively transferred 3 × 104 2D2 CD4+ Teffs in combination with 3 × 104 mock or edTreg, respectively, into lymphopenic Rag1−/− mice. Recipient mice were challenged with MOG35–55 peptide in adjuvant followed by pertussis toxin to disrupt the blood-brain barrier. Figure 7C shows the experimental timeline of cell transfer, immunization, and cell analysis. In this model, recipient animals developed symptoms of EAE (drooping tail progressing to tail then hindlimb paralysis) beginning at ~7 to 10 days after cell transfer. The immune response to EAE induction in Rag1−/− mice was robust, and the course of disease was not controlled by polyclonal or antigen-specific tTreg or edTreg (fig. S14), as predicted from previous studies (43).

Fig. 7 Generation of antigen-specific edTreg and in vivo modulation of antigen-specific Teff priming in a mouse model of multiple sclerosis.

(A) Flow plots showing GFP expression in MOG-specific mouse CD4+ T cells at day 2 after editing. The average percent of GFP+ cells across multiple experiments is shown at right (n = 10). (B) Flow plots of murine edTreg showing expression of relevant Treg markers. (C) Experimental timeline testing edTreg in murine EAE model. 2D2 Teffs were delivered with or without cotransferred edTreg generated from either 2D2 or C57Bl/6 mice into Rag1−/− recipient mice; all strains were on C57Bl/6 background. (D) Immunophenotype of T cells obtained from inguinal and axillary LNs in recipient mice at day 7 after cell transfer, as assessed by flow cytometry. Mean percentage (top) or total number (bottom) for total CD45+CD4+ cells and other indicated T cell subsets is shown. P values were obtained using either one-way (top left) or two-way ANOVA (top right and bottom) adjusted with Tukey’s multiple comparisons. (E) To label actively dividing cells, the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) was administered 30 min before sacrifice in selected animals. EdU incorporation in T cells was determined by intracellular labeling with an anti-EdU antibody and flow cytometry. Flow plots (top) are from T cells isolated from LNs 7 days after cell transfer; bar graphs summarize mean % of cells incorporating EdU in different cell subsets. Flow plots are representative of results from at least three independent experiments; bar graphs show mean ± SD; P values of statistically significant differences (one-way or two-way ANOVA with Tukey’s post hoc correction) are indicated above bars in black (indicating comparison with mock-edited group) or in red (comparing groups indicated by dashed lines).

Although neither tTreg nor edTreg controlled EAE progression in this model, we assessed the effect of these immunosuppressive cells on the responses of Teffs within the LNs. Recipient mice were euthanized at day 7 after transfer, and inguinal and axillary LNs were collected. The percentage of LN CD45+CD4+ T cells were twofold lower in recipients of antigen-specific 2D2 edTreg compared to recipients of mock-edited cells and 1.7-fold lower compared with recipients of polyclonal C57Bl/6 edTreg. The absolute number of CD4+CD45+ T cells was markedly reduced in both edTreg cohorts relative to the mock control, with 2D2 edTreg and C57Bl/6 edTreg having 49- and 18-fold fewer T cells, respectively. In all cohorts, most of the CD45+CD4+ cells were GFP, and fewer GFP T cells from mice receiving 2D2 edTreg expressed inflammatory cytokines IL-17A or IFN-γ (Fig. 7D). To determine whether the observed effect was due to reduced Teff proliferation, some animals were injected with the thymidine analog 5-ethynyl-2′-deoxyuridine (EdU) before sacrifice and its incorporation into gDNA was detected after a “click” reaction by flow cytometry (Fig. 7E). We found that 2D2 edTreg reduced the overall percentage of GFP cells that had incorporated EdU by 22 and 18% relative to groups treated with mock or polyclonal edTreg, respectively. EdU incorporation was present in GFP+ cells from 2D2 edTreg (~25%) and, to a lesser extent, from C57Bl/6 edTreg (~10%) recipient cohorts, consistent with the ability of tTreg to proliferate in vivo in response to self-antigen stimulation. These combined findings show that murine edTregs function in vivo to restrain the priming phase of pathogenic Teff and that antigen-specific edTregs exhibit greater activity and expansion in comparison with polyclonal edTreg.

Antigen-specific human T cells adopt a Treg phenotype after FOXP3 editing and are immunosuppressive in vitro

Our findings from the murine EAE model suggested that antigen-specific edTreg may have a greater efficacy in adoptive transfer applications than polyclonal edTreg. To determine whether our editing approach would be feasible with antigen-specific human T cells, CD4+ T cells from HLA DRB1*0401 donors were initially expanded for 7 days in the presence of the following pooled influenza and tetanus antigens: Clostridium tetani toxin (TT), influenza virus matrix protein 1 (MP), and influenza virus hemagglutinin (HA). The cultures were subsequently activated with CD3/CD28 beads before gene editing (Fig. 8A). After editing, cells were further expanded in the antigen cocktail for 4 to 7 days. At this time, the average editing rate (GFP+) was 28 ± 2.1% (Fig. 8B). Antigen-specific cells were purified by FACS after labeling with a mixture of phycoerythrin (PE)–conjugated influenza and tetanus MHC-II tetramers. These tetramer-positive edTreg (Tr+edTreg) recapitulated the immunophenotype of activated tTreg for canonical markers of Tregs: up-regulating expression of FOXP3, CD25, CTLA-4, LAG3, and Helios and suppressing IL-2 production (Fig. 8C and fig. S15A). Tr+edTregs were able to suppress polyclonal activated autologous CD4+Teff in vitro, unlike the Tr+Mock cells, indicating immunosuppressive function (fig. S15B).

Fig. 8 Generation of antigen-specific human edTregs and suppression of CD4+ Teff proliferative responses.

(A) Timeline of procedures used to generate antigen-specific edTreg. Peptide stimulation consisted of HA, MP, and TT peptides and autologous APCs (irradiated PBMCs). (B) Representative flow plots (left) and percentage (right) of GFP expression in tetramer-positive (Tr+; a mixture of MHC class II tetramers with influenza or tetanus peptides) human CD4+ T cells at 4 days after gene editing (n = 5). (C) Bar chart summarizing expression of Treg markers based on flow cytometry (mean ± SEM from at least three experiments); P values obtained by two-way ANOVA adjusted with Tukey’s multiple comparisons test; P values in black indicate comparison with mock-edited cells. (D) Diagram showing editing process and timing for in vitro suppression assays using antigen-specific (HA, MP, or TT) edTreg. Flow plots (below) show GFP+ editing rates of Tr+ cells at 3 days after editing and before flow sorting. After cell sorting and expansion, edTregs were cocultured with freshly thawed, autologous, EF670 proliferation dye–labeled CD4+ T cells and APCs along with either the pooled HA-MP-TT peptides or DMSO control. (E) Flow plots show proliferation of Teff cocultured with APCs and either DMSO or the peptide pool, with or without addition of antigen-specific edTreg. Teffs were gated as CD4+CTVEF670+. (F) Graph showing percent proliferation of Teff under each condition. Data shown in (D) to (F) are representative of two experiments.

We next developed methods to assess the specificity of immunosuppression mediated by antigen-specific edTregs. After edTregs were generated according to the schema shown in Fig. 8D and sorted, GFP+ edTregs were evaluated for antigen-specific immunosuppressive activity. Three days after editing, ~14% of cells were GFP+ and this population contained ~60% MP, HA, or TT Tr+ cells (Fig. 8D). FACS-sorted GFP+ cells were expanded for 10 days and subsequently cocultured with autologous Teff (CD4+CD25) and antigen-presenting cells [APCs; irradiated autologous peripheral blood mononuclear cells (PBMCs)], and either dimethyl sulfoxide (DMSO) (control stimulation) or a pool of MP, HA, and TT peptides (peptide stimulation). Strikingly, Teff proliferation, measured by dye dilution, was inhibited only in the presence of both antigen-specific edTreg and cognate peptides (Fig. 8, E and F), indicating the requirement for TCR engagement for edTreg activity.

It is known that tTregs suppress Teff of the same TCR specificity, as well as those with different TCR specificities that are in physical proximity, termed bystander suppression (44). We performed studies to directly test the capacity of human edTreg to suppress Teff with differing antigen specificities. To do so, we produced Teff with antigen specificities either identical to or different from the edTreg (see fig. S16 for experimental schematics) as follows. CD4+ T cells were stimulated with MP peptide for 10 days. MP Tr+ T cells were then sorted; some were retained for use as MP-specific Teff (MP-TCR Teff), and the remaining Tr+ T cells were activated by CD3/CD28 beads and edited to integrate MND-LNGFR-2A into FOXP3 as described in Fig. 5 and fig. S12A. After 7 days of expansion, MP Tr+ LNGFR+ and LNGFR cells were sorted to obtain MP Tr+ edTreg (MP edTreg) and MP Tr+ mock-edited cells (MP Mock), respectively (fig. S16, A and B). Mock-edited cells were used to control for the number of T cells per well, which by themselves suppress proliferation rates of the responder cells due to competition for growth factors such as IL-2. As shown in fig. S16C, MP edTreg, to a greater extent than MP mock-edited cells, suppressed MP-TCR-Teff proliferation in response to polyclonal stimulation (CD3/CD28 beads), indicating polyclonal immunosuppression. We next tested suppression in response to cognate peptides based on cytokine production of cocultured Teff, a more sensitive readout of immunosuppression where cell numbers were limited. Here, MP-peptide stimulated MP edTreg suppressed production of TNF-α, IL-17A, and IL-2 by MP-TCR Teff, whereas MP mock cells did not (fig. S16D). A similar strategy was used to test for bystander suppression of Teff proliferation. HA-specific Teffs (HA-TCR Teffs) were made by transduction of autologous CD4+ with a lentivirus delivering an HA-specific TCR expression cassette (fig. S16E). Although both MP edTreg and MP mock cells suppressed HA-TCR Teff proliferation in the setting of polyclonal stimulation, fewer MP-TCR Teffs proliferated in the presence of MP edTreg. MP edTreg and MP mock cells similarly suppressed HA-TCR Teff production of TNF-α, IFN-γ, and IL-2 with HA peptide stimulation, but MP edTreg showed further suppression of the cytokines in the presence of both the MP and HA peptides (fig. S16G). These results imply that antigen-specific edTregs require TCR engagement to inhibit production of inflammatory cytokines by Teffs and that edTregs exhibit bystander suppressive activity on Teffs with a different peptide specificity when edTregs encounter their cognate peptide. Together, these results demonstrate that CD4+ T cells from human peripheral blood can be enriched for target antigen specificity and modified by gene editing to impart tTreg-like phenotypic and suppressive properties that retain antigen specificity.


Evidence from early clinical trials and preclinical studies suggests that the adoptive transfer of Treg cells may represent a promising treatment modality for autoimmune disease. However, barriers to bringing Treg-based cell therapy to patients remain, including the difficulty of isolating and expanding the relatively small number of tTreg found in peripheral blood to large numbers ex vivo while maintaining the desired immunosuppressive phenotype; this challenge is magnified for antigen-specific Treg that might improve both safety and efficacy. In addition, tTreg in some patients may be intrinsically defective and/or may exhibit plasticity leading to proinflammatory function in vivo (45). Here, we demonstrate a strategy to generate stable Treg-like cells from abundant peripheral CD4+ T cells by inducing a defined genetic modification via HDR that drives high expression of FOXP3 from the endogenous locus. We found that gene-edited edTregs display a Treg-like phenotype even after long-term culture and that the adoptive transfer of edited cells protected mice in a severe xenogeneic GvHD model. Antigen-specific edTregs derived from murine Teff were able to suppress the proliferation of pathogenic effector cells in EAE, and we were further able to generate potent antigen-specific edTreg using tetramer-enriched T cells from human peripheral blood. We show that edTreg can be enriched and expanded at scale and, therefore, may have the potential for direct utility in the treatment of human autoimmune diseases.

The introduction of a strong promoter upstream of the first coding exon of the FOXP3 gene using TALEN and an AAV donor template resulted in stable expression of FOXP3 in an average of 37% of treated cells, with a range of 20 to 65%. A similar HDR-editing efficiency was achieved using the CRISPR-Cas9 nuclease platform. These HDR rates are consistent with our previously reported editing rates at other loci in activated T cells (e.g., CCR5, CD40L, and TRAC) (25, 27, 29). Using this editing strategy to generate either GFP-FOXP3 fusion protein or cis-linked LNGFRt surface expression as a marker for successfully gene-modified cells, edTregs are distinct from mock-edited T cells based on both high FOXP3 expression and other tTreg cell defining characteristics including up-regulated CD25, CTLA-4, and ICOS and the suppression of Teff proliferation and cytokine production in coculture. Other groups have previously shown that ectopic FOXP3 expression conferred some Treg-like phenotypic and suppressive properties to naïve Teff in vitro but that the consistency and stability of the phenotype are dependent on the maintenance of a high FOXP3 expression (21, 23, 46). Allan et al. (21) have reported that Teff transduced with an LV vector containing a FOXP3 cDNA driven by the EF1α promoter showed increased expression of CD25 and ICOS, but CTLA-4 expression was reduced compared to tTreg . Given that we detected high CTLA-4 in >90% of not only edTreg but also in T cells transduced with an LV vector containing a FOXP3 cDNA driven by the strong MND promoter, it is possible that up-regulation of CTLA-4 requires persistent, high expression of FOXP3.

In light of reports that Tregs can lose FOXP3 expression and become Teffs, or ex-Tregs (15, 47), the stability of FOXP3 expression and other effector proteins of immunosuppression is a critical concern for adoptive Treg cell therapies. Our adoptive transfer studies show that edTregs maintain suppressive function and tTreg-like phenotype in vivo for >50 days despite highly inflammatory conditions in the xenoGvHD model. Because MND-driven FOXP3 expression in edTreg is achieved in the setting of a fully methylated TSDR, we hypothesize that FOXP3 expression in edTreg will remain heritable and resistant to epigenetic events likely to negatively affect tTreg-derived cell products. Although there is evidence that FOXP3 expression must be consistent and at sufficient quantities to convert Teff to suppressive phenotypes (21), whether supraphysiologic FOXP3 expression has harmful effects is unknown, as is the impact of edTreg on infectious immune responses in vivo.

We compared our gene-editing method for generating edTreg to the induction of ectopic FOXP3 expression via a randomly integrating LV vector containing a FOXP3 cDNA. Although LV-treated cells in our experiments had, on average, three viral integrations, we observed less FOXP3 protein per cell in LV-treated cells versus edTreg by flow cytometry. This effect was independent of the TSDR, which was not altered by our gene-editing strategy and remained hypermethylated in edTreg. These findings suggest that expression of the endogenous gene is more efficient, perhaps because of cis-regulatory elements in the intronic regions, posttranscriptional effects of spliced versus unspliced mRNA, and/or a stabilizing effect of the untranslated region, features that are absent in cDNA-based LV. We observed a substantial loss of transgene expression by LV Treg compared to edTreg when expanded ex vivo that we hypothesize results from the selective advantage of cells having epigenetically silenced the LV provirus. An alternative approach for Treg therapy may include ex vivo conversion of peripheral Teff using CD3/CD28 stimulation in the presence of transforming growth factor–β (TGF-β) or all-trans retinoic acid to generate FOXP3+ cells, referred to as induced Treg (iTreg). Although iTreg can induce tolerance in murine models, FOXP3 expression is unstable in human iTreg and clinical trials, to date, have focused on studies of a distinct FOXP3neg IL-10–producing Tr1 subset [reviewed in (45, 48)].

Despite similar suppressive capacity in functional assays, we noted several characteristics where edTreg differed from FACS-purified tTreg. We found that expression of the Ikaros family transcription factor Helios, reported in 70 to 80% of Treg, was detectable in ~25% of activated edTreg (49). The MFI for Helios staining in edTreg was also lower than that of tTreg but higher than mock-edited cells. Studies of Helios-deficient Treg in Foxp3-Cre mouse models have led to conflicting results, and the role of Helios in Treg function remains incompletely defined (50, 51). Helios has been described as enforcing a Treg phenotype by binding to the FOXP3 promoter (52) and as facilitating epigenetic changes that suppress IL-2 production (53). These functions may not be required in edTreg, where we observe stable, high FOXP3 expression and similarly low IL-2 production versus tTreg after activation, apparently independent of Helios expression. It is notable that FOXP3 expression did not drive hypomethylation of the TSDR in light of evidence that FOXP3 in concert with Runx1 drives maintenance of active/demethylated TSDR (54). However, open chromatin may be required to initiate feed-forward maintenance of demethylated TSDR. Many factors have been reported to bind within the TSDR, and their roles and hierarchy in determining the epigenetic landscape of the locus remain to be determined. Nevertheless, that the TSDR remains methylated in edTreg suggests that the strength or location of the inserted MND promoter overcomes endogenous epigenetic regulation that might otherwise dampen stable, inheritable FOXP3 expression in these cells. A GFP-FOXP3 fusion protein similar to that generated by the editing strategy in some of our experiments has been described as a hypomorph in the NOD mouse, with selective defects in binding to transcriptional partners, and contradictory observations of differences in Treg cytokine production and function in various models (55, 56). Although we did not observe differences using GFP-FOXP3 fusion versus alternative constructs that do not produce a fusion protein, future clinically relevant products would clearly not use a GFP construct.

In the context of clinical application, antigen specificity is likely to be a key efficacy and safety feature for therapeutic suppressive cells. The engineering of antigen-specific edTreg described here is an additional important advance. Consistent with the distinct antigen-specific and antigen-independent suppressive functions of natural Treg, we found that mouse antigen-specific edTregs were better able to expand in vivo and suppress proliferation of a matched, mono-specific pathogenic Teff population compared with edTreg generated from polyclonal T cells. We also tested the capacity to perform gene editing after enrichment of circulating antigen-specific human CD4+ T cells using MHC class II tetramers. Gene editing of such cells resulted in Treg-like phenotypic and suppressive properties similar to those observed for bulk-edited CD4+ T cell populations, an important demonstration that our editing method is able to effectively endow memory CD4+ Teffs with these properties. Further, our studies using antigen-specific human edTreg demonstrate the capacity to exert both antigen-specific, direct and indirect (bystander) suppression of Teff. These findings support the concept that generation of a human edTreg product with a restricted, disease-relevant, antigen specificity may exert a broader suppressive impact in tissue-specific autoimmune disorders, similar to findings observed with antigen-specific murine tTreg (57, 58). As alternative strategies to confer antigen specificity, diverse engineering approaches could be used to combine the HDR-based FOXP3-editing approach described here with delivery of either a specific TCR (derived from tissue-specific pathogenic, or regulatory, CD4+ T cells) or a chimeric antigen receptor (CAR) to edTreg as a transgene. These latter approaches would have the advantage of eliminating the requirement for enrichment of rare cell populations during clinical manufacturing. In addition, a range of engineering strategies could be applied to achieve genetically enforced, coexpression of the targeted receptor and FOXP3.

As a proof of concept, the data presented here demonstrate that this HDR-editing approach creates FOXP3-expressing cells with stable Treg-like immunosuppressive properties. Although our limited analysis suggested no evidence for toxicity with either designer nuclease platform, direct clinical application will require extensive off-target analysis using a nonbiased approach performed with good laboratory practice (GLP) nuclease reagents in primary T cells from independent donors. Single-edit introduction of additional coding sequences into the FOXP3 locus could be used to impart features to improve edTreg manufacture or function. For example, clinically relevant surface markers (e.g., truncated LNGFR or EGFR) allowed enrichment of edTreg to ~90% purity using an affinity column. Although determining the engraftment capacity, long-term survival, and potential clinical benefit of edTreg will require carefully designed clinical trials, iterative improvements in the edTreg platform may permit broad clinical application. Such future application might include introduction of a FOXP3 cDNA for IPEX therapies, use of selection cassettes for in vitro or in vivo expansion, and/or co-delivery of TCRs or CARs recognizing common antigens in autoimmune disease or solid organ transplantation.


Study design

The objective of this study was to test whether stable, immunosuppressive Treg-like cells could be generated using a gene-editing approach to enforce FOXP3 expression in peripheral blood CD4+ T cells. Cell surface and intracellular markers of Treg cells have been well characterized in the literature and were assessed here after gene editing using flow cytometry and RNAseq analyses. These data were derived where possible using at least three biological replicates in independent experiments. Next, the ability of edited human T cells to suppress Teff immune responses was assessed both in vitro and in vivo. Investigators were not blinded to the treatment unless noted in the figure legend. Data points obtained from polyclonal in vitro immunosuppression assays represent the average of duplicate or triplicate samples within the same experiment. In vivo immune suppression was assessed using a xeno-GvHD model (for human edTreg) and a murine EAE induction model (for antigen-specific murine edTreg). In both models, endpoints and sample sizes were predetermined from pilot studies as well as published data. Note that humane endpoints for euthanizing animals were specified in advance. Sample sizes and number of biological and technical replications are provided in the figure legends. Primary data are reported in data file S1.

Primary human T cells

Human PBMCs, obtained by apheresis of healthy donors and viably frozen, were purchased from Fred Hutch’s Co-operative Center for Excellence in Hematology Cell Processing Core Facility. PBMCs from patients with IPEX were obtained by Seattle Children’s Hospital clinicians. All samples were obtained after informed consent, using protocols approved by the Institutional Review Boards of the Fred Hutch Cancer Research Center (FH985.03) or Seattle Children’s Research Institute (SCRI) (11738), respectively. CD4+ T cells were purified from PBMCs by negative selection using the EasySep Human CD4+ T Cell Enrichment Kit (STEMCELL Technologies) and then either frozen for later use or directly cultured for editing. Methods used for T cell culturing, editing, and LV transduction have been described previously (2527, 29, 59) and are summarized in Supplementary Materials and Methods.

Immunosuppression assays using polyclonal edTreg

CD4+ Teffs (“responders”) were isolated from fresh or frozen PBMCs, then immediately stained with CellTrace Violet (CTV; Thermo Fisher Scientific) according to the manufacturer’s instructions, and washed twice in phosphate-buffered saline (PBS). Cells were then plated in a round-bottom 96-well plate (1 × 104 to 5 × 104 cells per well) in T cell medium (including IL-2) and CD3/CD28 beads (1:10 to 1:32 bead:Teff). “Suppressor” cells (autologous edTregs or mock-edited T cells) were added to some wells, and cells were then cultured for 96 hours. All conditions were replicated in three wells, and replicate wells were combined before flow cytometry analyses. The addition of any T cell that competes for cytokines in the growth medium will reduce the apparent proliferation of the labeled responder cells. To prevent this from confounding our assay, in Fig. 3, the edTregs or mock-edited suppressor cells were rested in culture for 72 to 96 hours after CD3/CD28 bead stimulation. Rested cells were then resuspended in cytokine-free T cell medium for 24 hours before coculture. In Fig. 4E, autologous suppressors (mock, edTreg, tTreg, and LV Treg) were removed from CD3/CD28 beads and then irradiated at 30 Gy using a Mark I-30 cesium-137 irradiator immediately before coculture.

Mouse studies

NSG, C57Bl/6, 2D2 [C57Bl/6-Tg(Tcra2D2,Tcrb2D2)1Kuch/J], and Rag1−/− (B6.129S7-Rag1tm1Mom/J) mice were originally obtained from The Jackson Laboratory and housed in an Association for Assessment and Accreditation of Laboratory Animal Care–accredited specific pathogen–free facility at SCRI. Experimental mice were either purchased or bred in-house. All animal work was performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and was approved by the Institutional Animal Care and Use Committee of the SCRI.

Xenograft-induced GvHD

Human Tregs were purified by FACS based on GFP expression (edTreg) or CD25++CD127 expression (tTreg) and then expanded in parallel with mock-edited T cells in G-Rex 6 well or G-Rex 10 flasks (Wilson Wolf Manufacturing). Three days before delivery of the Teff xenograft, the indicated Tregs or control cells were delivered intraperitoneally (8 × 106/mouse) into irradiated (200 cGy) 8- to 10-week-old NSG mice. Antibiotic (Baytril) was added to drinking water (0.2 mg/ml; Bayer Healthcare) for 14 days after the irradiation. On experiment day 0, freshly isolated autologous human CD4+ Teffs were delivered intraperitoneally (4 × 106/mouse). Mice were weighed twice per week, and 7 of 11 experiments were scored for GvHD symptoms (60) by investigators who were blinded to which treatment each mouse received. Mice from each cohort were euthanized on day 14 or day 50 or at any time during the experiment if they lost more than 20% body weight. The presence of human CD4+ T cells and GFP-FOXP3+ cells was assessed by flow cytometry in multiple tissues, including peripheral blood, spleen, liver, lung, and BM. Some other organs including kidney, pancreas, skin, colon, and cecum were saved in formalin for histopathology.

EAE induction

In all experiments, 3.0 × 104 2D2 freshly isolated Teffs (total CD4+ fraction isolated from the spleen and LNs) were adoptively transferred into 7- to 13-week-old male or female Rag1−/− mice. In some experiments, 3.0 × 104 edTreg from 2D2 or C57Bl/6 mice (or mock-edited 2D2 T cells) were cotransferred along with Teff. Note that the edTregs or mock-edited cells were expanded in culture for 3 days after editing before transferring into mice. edTregs were FACS-sorted by GFP at day 2 after editing. One day after cell transfer (day 0), EAE was induced by subcutaneous injection of 100 μl of an emulsion containing 100 μg of MOG35–55 peptide (AnaSpec) in complete Freund’s adjuvant (Fisher Scientific). Mice also received 200 ng of Bordetella pertussis toxin (List Biological Laboratories) intraperitoneally on days 0 and 2. Mice were scored daily according to the guideline described by Hooke Laboratories ( In experiments shown in Fig. 7, mice from each cohort were euthanized 7 days after cell transfer. In some mice, 30 min before sacrifice, 1 mg of EdU (Thermo Fisher Scientific) in 200 μl of saline was delivered intraperitoneally. Lymphocytes from spleen and inguinal and axillary LNs were then collected for analysis by flow cytometry as described in the Supplementary Materials.

Antigen-specific human CD4+ T cells and edTregs

PBMCs were collected from adult healthy donors that were genotyped HLA DRB1*0401 as described (61) and viably frozen. Upon thaw, cells were seeded at 1 × 106 cells/ml in the presence of irradiated human PBMCs (30 Gy) and a mixture (20 μg/ml) of three peptides matching known immunogenic epitopes of common vaccines (GenScript): influenza virus MP 1, (97-116 VKLYRKLKREITFHGAKEIS), influenza virus HA (307-319 PVKQNTLKLAT), and TT (666-685 ETTGVVLLLEYIPEITLPVI). Seven to 11 days later, CD4+ T cells were isolated and stimulated with Dynabeads T-activator CD3/CD28 beads (1:1) for 72 hours. After stimulation, cells were edited as described. When edited cells were moved from 30° to 37°C, fresh immunogenic peptides were added to the medium along with autologous APCs (irradiated PBMCs). Seventy-two hours after editing, cells were stained with PE-labeled human leukocyte antigen (HLA) class II tetramers for each of the three peptides [made as described; (61)]. Tetramer+ (Tr+) GFP+ T cells were immunophenotyped by flow cytometry. In Fig. 8 (D to F), CD4+ T cells were isolated from PBMCs and stimulated by peptide and APCs for 9 to 10 days with IL-2. For MP/HA/TT-specific T cells, T cells were stimulated twice. Tetramer staining was performed on day 9 or day 10, and Tr+ cells were sorted if needed. Tr+ CD4+ T cells were activated by CD3/CD28 beads. Beads were removed after 2 to 3 days of activation; after overnight resting, cells were edited to integrate MND-GFP described in Fig. 1. Edited cells were sorted by GFP+ and expanded to set up suppression assays.

Immunosuppression assays using antigen-specific human edTreg

For antigen-specific suppression assay using primary CD4+ T cells as Teffs, CD4+CD25 cells freshly isolated from autologous PBMCs were incubated with APCs and either DMSO or peptide pool (MP, HA, and TT peptides) for 7 days with or without MP-, HA-, or TT-specific edTregs. CD4+CD25 cells and edTregs were labeled with Proliferation Dye eFluor 670 (EF670) and CTV, respectively, before coculture. Cells were harvested for surface staining, and dilution of EF670 was measured as antigen-specific Teff proliferation.

Statistical analysis

Statistical analysis of data obtained from in vitro and in vivo assays was performed using GraphPad Prism 7 and 8, with the exception of RNAseq data, which were analyzed as described for RNAseq. GraphPad analysis calculation of the mean, SEM, and SD, as well as comparisons of results using either Student’s t test, F test of linear regression, log-rank test, and one- or two-way analysis of variance (ANOVA) adjusted for multiple comparisons using with Tukey, Dunnett, or Sidak’s tests, as appropriate. Sample means were assumed to be normally distributed; no outliers were excluded. Details of the statistical tests used are indicated in each figure legend.


Materials and Methods

Fig. S1. FOXP3 TALENs catalyze efficient FOXP3 disruption and initiate seamless recombination of donor template.

Fig. S2. Purity of primary human tTregs and Teffs used in immunophenotyping assays.

Fig. S3. Human edTregs derived from CD4+ T cells exhibit increased sensitivity to IL-2.

Fig. S4. T cell sample preparation for RNAseq analysis.

Fig. S5. Transcriptome of Teff compared with tTreg and edTreg by RNAseq.

Fig. S6. The TCR repertoire after FOXP3 editing is highly polyclonal and similar to mock-edited T cells.

Fig. S7. Enrichment of edTreg using clinically relevant selection markers.

Fig. S8. edTregs, but not Teffs or mock-edited T cells, are nonpathogenic in a xeno-GvHD mouse model.

Fig. S9. Tissue histopathology studies in the xeno-GvHD mouse model.

Fig. S10. Body weight and GvHD scores of mice in the xeno-GvHD model.

Fig. S11. The phenotype and cytokine production of GFP cells in the xeno-GvHD model.

Fig. S12. edTregs generated using the CRISPR-Cas9 platform express Treg markers and promote immunosuppression in vivo.

Fig. S13. Design of TALENs targeting mouse Foxp3.

Fig. S14. EAE scores of mice with adoptive edTreg transfer and EAE induction.

Fig. S15. Antigen-specific human edTregs express Treg markers and are immunosuppressive in vitro.

Fig. S16. Antigen-specific edTregs exhibit antigen-specific and bystander suppression of Teff proliferation and cytokine generation.

Table S1. Rates of on- and off-target indels by human FOXP3 TALEN pairs.

Table S2. Top-ranked in silico predicted off-target indels of human FOXP3 Cas9 RNP.

Table S3. Oligonucleotide sequences.

Table S4. Flow cytometry antibodies.

Data file S1. Primary data.

References (6271)


Acknowledgments: We thank S. Khim for technical support, A. Timms for assistance with bioinformatic analysis of RNAseq, and J. Haddock for administrative support. We thank M. Jensen for providing biotinylated Erbitux and P. Linsley and F. Whitman for the HA-TCR lentivirus. Funding: This work was supported by the Helmsley Charitable Trust (to D.J.R. and J.H.B.), the SCRI Center for Immunity and Immunotherapies (CIIT) Program for Cell and Gene Therapy (PCGT), the Children’s Guild Association Endowed Chair in Pediatric Immunology (to D.J.R.), the Hansen Investigator in Pediatric Innovation Endowment (to D.J.R.), the Benaroya Family Gift Fund (to D.J.R.), and the American Diabetes Association (to A.M.S.). Author contributions: T.R.T., A.M.S., J.H.B., and D.J.R. conceived the approach and consulted on experimental design. Y.H., N.H., Y.X., L.F., D.H., Y.S., S.J.Y., C.L., T.T., E.M.D., and I.K. planned and performed the experiments. K.S. assisted in experimental design. K.S., Y.H., M.H., and D.J.R. wrote the manuscript. Competing interests: A patent application has been filed (D.J.R., A.M.S., I.K., T.T., K.S., and Y.H., “Expression of human Foxp3 in gene edited T cells,” U.S. patent application no PCT/US2019/029159). All other authors declare that they have no competing interests. Data and materials availability: All data associated with this study are available in the paper or the Supplementary Materials. RNAseq data are available at (doi:10.5061/dryad.02v6wwq08). AAV donor template plasmids and TALEN constructs are available from SCRI for academic use under a material transfer agreement with the Institute.

Stay Connected to Science Translational Medicine

Navigate This Article