Isolation of Pure Stable Human Treg Cells Based on Expression of GPA33

ABSTRACT

Adoptive cell therapy (ACT) with regulatory T (Treg) cells offers potential for treating immune‐mediated diseases. Ensuring the purity and stability of Treg cell products is critical for safe and effective therapies, particularly when targeting specific self‐antigens. The purest products are currently obtained using CD4⁺CD25⁺CD127^low/–CD45RA⁺ naïve (n)Treg cells. However, these still include cells lacking key transcription factors FOXP3 and Helios, able to produce inflammatory cytokines. Previously, we identified GPA33 as a surface marker for stable FOXP3⁺Helios⁺ human Treg cells and demonstrated that GPA33^high nTreg cell populations maintain higher purity during in vitro expansion compared with standard nTreg cells. However, the definition of the GPA33^high population among nTreg cells was arbitrary, and cell yields were low. Here, we show methods to unequivocally identify GPA33⁺ cells within the Treg cell fraction and generate Treg cell products that match nTreg cell populations in size and expansion capacity but exhibit superior FOXP3⁺Helios⁺ purity, lack effector cytokine production, and retain full suppressive function. Combining GPA33 with CD226 exclusion eliminates the need for CD127‐based gating. Post‐expansion, co‐expression of GPA33 with TIGIT reliably identifies lineage‐stable Treg cells. Thus, GPA33, alone or with CD226/TIGIT, is a robust marker for isolating Treg cells with enhanced therapeutic safety.

GPA33 identifies stable human regulatory T cells suitable for therapeutic use. GPA33⁺ Treg subsets expand efficiently, maintain high FOXP3⁺Helios⁺ expression, and lack production of IL‐2, IFNγ, and IL‐17A. GPA33, alone or combined with TIGIT, also enables robust post‐expansion isolation of pure, stable Treg cells for adoptive cell therapy.

1. Introduction

CD25⁺FOXP3⁺ regulatory T (Treg) cells are a specialized subset of CD4⁺ T cells with diverse immunosuppressive capabilities. As such, Treg cells play a pivotal role in the maintenance of immune tolerance [1, 2]. Adoptive cell therapy (ACT) using these cells is an attractive therapeutic alternative to the use of systemic immunosuppressants [1, 3, 4]. Promising results have been obtained in preclinical (murine) studies in which Treg cells were administered for the treatment of autoimmune disease, including type 1 diabetes (T1D), systemic lupus erythematosus (SLE), and experimental autoimmune encephalomyelitis (EAE), a model widely used for studying multiple sclerosis (MS) [5, 6, 7, 8, 9, 10]. Clinical trials in human patients have demonstrated the safety and tolerability of ACT using polyclonal, often autologous, Treg cells and provided some support for therapeutic efficacy [11, 12]. However, studies using mouse models suggest that there is significant room for improvement regarding therapeutic efficacy [13, 14, 15]. In particular, much is expected of therapies using Treg cells that express T cell receptors (TCR) or Chimeric Antigen Receptors (CAR) specific for (self‐)antigens from the tissues that must be protected, as Treg cell activity is dependent on activation of their antigen receptor [16, 17, 18, 19]. This approach does, however, put special demands on the purity of the Treg cell populations used, since contaminating conventional T (Tconv) cells expressing self‐reactive antigen receptors would pose the risk of auto‐aggression.

Phenotypically different cell populations have been used in clinical studies aiming to harness the regulatory power of Treg cells. In some cases, cells were used with a CD4⁺CD25^+/high phenotype, but this phenotype includes activated CD25⁺ Tconv cells [20, 21, 22]. Other studies used FOXP3⁺ “induced” Treg (iTreg) cells, generated in vitro by activating CD4⁺CD25^– Tconv cells in the presence of interleukin (IL)‐2 and TGF‐β. Such cells are also not the safest choice, as they lack the epigenetic Treg program that characterizes genuine Treg cells and can revert into Tconv cells [23, 24, 25, 26].

At present, most studies use Treg cells based on a CD4⁺CD25⁺CD127^low/– phenotype. Expression of CD127, the α‐chain of the IL‐7 receptor, is inversely correlated with the expression of hallmark Treg cell regulator FOXP3²⁷. Whilst reasonably useful for isolating Treg cells from healthy individuals, a caveat is that CD127 is downregulated after activation of conventional T cells [28, 29, 30]. This undermines its discriminatory capacity, especially when isolating Treg cells from patients with (auto)inflammatory disorders, where activated Tconv cells are often abundant [31, 32, 33, 34, 35]. A widely used method to improve the purity of therapeutic Treg cell populations relies on the use of the immunosuppressive drug rapamycin, which reportedly inhibits the proliferation of Tconv cells more than that of Treg cells [36, 37]. Indeed, a pure population of FOXP3⁺ cells is obtained when CD4⁺CD25⁺CD127^low/– cells from human blood are cultured (using antibodies to CD3 and CD28 in combination with recombinant IL‐2) in the presence of rapamycin [37, 38]. It has become clear, however, that the universal expression of FOXP3 in such populations is at least partially due to the ability of rapamycin to elicit expression of FOXP3 in Tconv cells [39, 40]. As these cells lose FOXP3 upon withdrawal of rapamycin and thereupon function as Tconv cells [39], therapeutic Treg cell populations obtained with the use of this drug may thus not be as safe as initially believed [41].

A further caveat with currently used purification protocols is that the CD4⁺CD25⁺CD127^low/– population arguably contains a mixture of Treg cells with variable degrees of commitment to the Treg cell lineage. Treg cells can develop from immature T cell precursors in the thymus (thymic (t)Treg cells), as well as from mature Tconv cells in the periphery (peripherally derived (p)Treg cells). The available evidence suggests that tTreg cells exhibit strong lineage fidelity, whereas pTreg cells can lose their suppressive capacity in an inflammatory environment due to unstable FOXP3 expression. Such cells can even acquire the ability to produce proinflammatory cytokines and cause pathology [42, 43, 44]. Therefore, tTreg cells likely represent the safer Treg cell type to use for ACT. Although tTreg cells can be distinguished from pTreg cells by the expression of Neuropilin‐1 (Nrp1) in mice, the expression of this surface molecule on Treg cells in humans is substantially different and cannot be used to identify human tTreg cells [45]. Co‐expression of the transcription factors FOXP3 and Helios reportedly marks lineage‐stable tTreg cells in mice [46] and presumably also in humans, although the latter is difficult to know with certainty [46, 47, 48, 49]. Although Helios^– human Treg cells can transiently express this factor upon activation [50], persistent expression of Helios seems the most reliable criterion available to identify stable Treg cells. In support of this, we previously showed that among (freshly isolated or in vitro expanded) CD4⁺CD25⁺CD127^low/– T cells, the ability to produce Tconv effector cytokines was found among FOXP3⁺Helios^– cells, but not among cells expressing both factors [26]. However, intracellular molecules cannot be used to isolate viable cells for therapy. Consequently, it is important to identify suitable surface markers for the isolation of lineage‐stable tTreg cells.

One such marker is the coinhibitory molecule T cell immunoreceptor with immunoglobulin and immunoreceptor tyrosine‐based inhibitory motifs (TIGIT). TIGIT is highly expressed by FOXP3⁺Helios⁺ Treg cells in human blood [51]. However, ex vivo isolated TIGIT⁺ Treg cells were refractory to expansion, necessary to obtain sufficient cells for ACT [51]. On the other hand, the costimulatory receptor CD226 (which competes with TIGIT for binding to CD155 expressed by antigen‐presenting cells) has potential as an exclusion criterion [52, 53]. Expression of CD226 was associated with a Tconv effector phenotype, while the absence of CD226 among CD4⁺CD25⁺ cells marked a population that mostly consisted of FOXP3⁺Helios⁺ Treg cells [54]. Finally, it has been proposed that CD45RA‐expressing “naïve” Treg (nTreg) cells within the CD4⁺CD25^high Treg cell compartment are the best population for adoptive Treg cell therapies, based on superior purity and stability aspects (with respect to FOXP3 expression and cytokine‐producing abilities) compared with CD4⁺CD25^high Treg cells [55, 56, 57]. Accordingly, (polyclonal) CD4⁺CD25⁺CD127^–CD45RA⁺ nTreg cells have been implemented in several ongoing clinical trials, including those for Crohn's disease and solid organ transplantation (NCT03185000, NCT04661254, NCT04817774) [56]. While an improvement over Treg cell populations used earlier, our own studies showed that the CD4⁺CD25⁺CD127^–CD45RA⁺ nTreg cell population often still contains a sizable Helios^– fraction of cells (up to 30%) that, importantly, possess the ability to produce inflammatory cytokines [26, 58]. Therefore, we believe that this population still does not possess the safety characteristics required for ACT with Treg cells expressing (self) tissue‐specific TCRs.

Using combined transcriptomic and proteomic analyses, we previously found that expression of the transmembrane receptor glycoprotein A33 (GPA33) reliably identifies FOXP3⁺Helios⁺ cells in the CD4⁺CD25⁺CD127^– population from human blood [26, 58]. We showed that FACS‐purified CD45RA⁺ nTreg cells with high expression of GPA33 uniformly and stably remained FOXP3⁺Helios⁺ during in vitro expansion for several weeks without the need to include rapamycin in the culture [26]. Moreover, GPA33^high nTreg cells refrained from producing Tconv effector cytokines after expansion in vitro, retained full CpG demethylation of the FOXP3 Treg‐specific demethylated region (TSDR), and were suppressive [26]. Although the purity of the population obtained made this an attractive selection method to prepare Treg products for ACT, the expression pattern of GPA33 within the nTreg cell population was insufficiently discriminatory for unequivocal discrimination between GPA33⁺ and GPA33^– cells. This necessitated selection for GPA33^high nTreg cells, which excluded many bona fide FOXP3⁺Helios⁺ cells for the sake of purity, leading to relatively low cell yields.

We here revisited our Treg cell isolation procedure to address these shortcomings. We show that isolation of GPA33⁺ (as opposed to GPA33^high) cells from the CD4⁺CD25⁺CD127^low/– Treg cell fraction yields populations of similar size as the CD45RA⁺ nTreg cell population, but with a much greater proportion of stable FOXP3⁺Helios⁺ cells. These GPA33⁺ Treg cells exhibit robust expansion capacities, harbor full suppressive capacity, and do not produce effector cytokines. We further find that the need for negative selection of CD127 can be circumvented by using GPA33 in combination with CD226. Finally, we showed that expression of GPA33 can be used to purify stable Treg cells after expansion in vitro, especially when combined with expression of TIGIT. All in all, we conclude that selection based on GPA33⁺ should be considered for the isolation of stable Treg cells for therapeutic purposes.

2. Materials and Methods

2.1. Sample Collection, Preparation, and Sorting

Human materials were obtained in compliance with the Declaration of Helsinki and the Dutch rules regarding the use of human materials from voluntary donors. Buffy coats were obtained from anonymized healthy donors after their written informed consent, as approved by the Ethical Advisory Council of Sanquin Blood Supply Foundation, Amsterdam, the Netherlands. Peripheral blood mononuclear cells (PBMC) were isolated from these buffy coats using standard Ficoll‐Paque density gradient centrifugation (GE Healthcare), after which anti‐CD4 microbeads (Miltenyi Biotec) were used for magnetic sorting of CD4⁺ cells according to the manufacturer's instructions. Cells were subsequently incubated for 15 min at room temperature with (a combination of) the following antibodies: anti‐CD4 PerCP/Cyanine5.5 (RPA‐T4), anti‐CD8 PerCP/Cyanine5.5 (SK1), anti‐CD25 BB515 (2A3), anti‐CD127 PE‐CF594 (A019D5), anti‐CD45RA BV785 (HI100), anti‐CD226 PE (DX11), and anti‐GPA33 AF647 (obtained from the Olivia Newton‐John Cancer Research Institute, Heidelberg, AU [59] and labeled with Alexa Fluor 647 Succinimidyl Ester (ThermoFisher Scientific)). Near‐IR (APC‐Cy7, Life Technologies) was used as a live/dead marker.

We FACS‐sorted for live, CD4⁺/CD8^–, CD25^–CD45RA⁺ (nTconv), CD25⁺CD127^low/–CD45RA⁺ (nTreg), CD25⁺CD127^low/–CD45RA⁺GPA33⁺ (GPA33⁺ nTreg), CD25⁺CD127^low/–GPA33⁺ (GPA33⁺ total Treg) and CD25⁺CD226^–GPA33⁺ (CD226^–GPA33⁺ Treg). In addition, CD25⁺GPA33⁺ cells were sorted. All cells were sorted using a FACS Aria III (BD Biosciences).

2.2. Cell Culture and Expansion

At the start of culture, cells were plated at a concentration of 2 × 10⁴ cells per well in 96‐well U‐bottom plates (Greiner Bio‐One). Cells were cultured in IMDM + 10% FCS + 1% L‐glutamine + 1% penicillin/streptomycin for up to 21 days in the presence of 0.1 µg/mL soluble anti‐CD3 mAb (M1654, clone 1XE, Pelicluster), 0.1 µg/mL anti‐CD28 mAb (clone CD28.2, eBioscience, ThermoFisher Scientific), and 300 IU/mL IL‐2 (proleukin, Novartis) at 37°C, 5% CO₂. Fresh medium with IL‐2 was added on day 4/5, day 11/12, and day 18/19 of culture. At days 7 and 14, cells were harvested, counted, split to the starting concentration of cells for further culture (and restimulated as described above) or used for other experiments/further analysis. At day 21, cells were counted and harvested for analysis.

2.3. Flow Cytometry

Cells were harvested, washed and labelled with (a combination of) the following antibodies: anti‐CD27 BV650 (O323), anti‐CD226 PE (DX11), anti‐TIGIT BV785 (A15153G), anti‐TIGIT BV421 (A1513G), anti‐GPA33 AF647 (as described above), FOXP3 PE‐Cy7 (236A/E7), Helios PerCP/eF710 (22F6), anti‐IL‐2 AF488 (MQ1‐17H12), anti‐IFNγ eFluor450 (4S.B3), anti‐IL‐17A BV605 (BL168) and anti‐IL‐10 PE (JES3‐9D7). Surface stainings were done in PBS containing 0.5% FCS for 15 min at room temperature. Dead cells were excluded from the analysis using Live/Dead Fixable Yellow Dead Cell Stain (BV510, Life Technologies, ThermoFisher Scientific). For intracellular stainings, cells were first fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience, ThermoFisher Scientific) according to the manufacturer's instructions before incubation with respective antibodies. Data were acquired on a FACSymphony A5 (BD Biosciences) using the FACSDiva software and analyzed using the Flowjo software (version 10.8.1). Statistical tests were performed using GraphPad Prism software (version 9.1.1). Results were considered statistically significant at p < 0.05. Flow cytometry was performed in adherence to “Guidelines for the use of flow cytometry and cell sorting in immunological studies (third edition)” [60].

2.4. Suppression Assay

Cryopreserved PBMCs were thawed at 37°C, washed with IMDM + 20% FCS, and labeled with CellTrace Violet (ThermoFisher Scientific). After 5 min of incubation with prewarmed IMDM + 10% FCS + 1% l‐glutamine + 1% penicillin/streptomycin, cells were washed twice with this medium. Labeled PBMCs were stimulated with anti‐CD3 mAb (PeliCluster; 0.01 µg/mL) and anti‐CD28 mAb (eBioscience, ThermoFisher Scientific) and cocultured with expanded (7 days) and washed Treg cell populations stimulated with anti‐CD3 mAb (PeliCluster; 0.01 µg/mL) and anti‐CD28 mAb (eBioscience, ThermoFisher Scientific; 0.01 µg/mL). Proliferation was analyzed by flow cytometry after 4 days of co‐culture at 37°C, 5% CO₂.

2.5. Cytokine Assay

At day 14, cells were harvested and replated in 96‐well U‐bottom plates (Greiner Bio‐One). Cells were either stimulated with PMA (10 ng/mL) and ionomycin (1 µM) or left unstimulated for 4 h in the presence of Brefeldin A (eBioscience, ThermoFisher Scientific). After 4 h, cells were stained and measured by flow cytometry.

3. Results

3.1. Selection for GPA33 Yields Populations of Similar Size and Expansion Capacity as nTreg Cells

We previously demonstrated that high expression of GPA33 among CD4⁺CD25⁺CD127^–CD45RA⁺ nTreg cells in human blood identifies a population with the attributes of tTreg cells [26]. Unfortunately, cell yields obtained using this method were much lower than those obtained by selecting for the total nTreg cell population, currently the state‐of‐the‐art in terms of purity and stability [55, 56, 61]. Therapeutic application in patients requires high numbers of cells. We, therefore, examined whether, by less restrictive gating on GPA33⁺ cells (instead of GPA33^high), larger populations can be obtained that still exhibit properties that compare favorably to those of the nTreg cell population.

As reference populations, we included in our analysis CD4⁺ naïve (n)Tconv (Fraction 1: CD45RA⁺CD25^–) and nTreg cells (Fraction 2: CD45RA⁺CD25⁺CD127^–), already used in clinical trials [56] (Figure 1A). Four fractions of CD4⁺ T cells (in human blood) expressing GPA33 were characterized. These included GPA33⁺ nTreg cells (Fraction 3: GPA33⁺CD25⁺CD127^–CD45RA⁺), GPA33⁺ total Treg cells (Fraction 4, GPA33⁺CD25⁺CD127^–), and GPA33⁺CD25⁺ T cells (Fraction 5). GPA33⁺ cell fractions were isolated by gating on the positive peak of flow cytometry histograms or biaxial flow cytometry plots displaying GPA33 expression from respective parent populations (Figure 1A–C). From the CD25⁺CD127^– gate, GPA33⁺ total Treg and/or GPA33⁺ nTreg cells can also be selected by plotting CD45RA against GPA33. This results in a clear population of GPA33⁺ cells, allowing for easy isolation of these cells (Figure 1B).

FIGURE 1.

Gating strategies for isolation of experimental GPA33⁺ cell populations using flow cytometry. PBMCs were isolated from buffy coats, after which CD4⁺ cells were magnetically sorted using microbeads. Cells were subsequently FACS‐sorted for the respective populations. (A) Gating strategy started with a selection for lymphocytes, viable single cells (Figure S1), followed by CD4⁺/CD8^– cells (top row, second from the left plot) out of which (clockwise starting from the right) CD25⁺CD226^–, total CD25⁺, CD25⁺CD127^– (total Treg), and CD25^–CD45RA⁺ (nTconv, Fraction 1 (Fr1)) cells were gated. From the first three populations, the GPA33⁺ fraction was gated using flow cytometry histograms by gating on the far‐right peak (resulting in Fraction 6 (Fr6), Fraction 5 (Fr5), and Fraction 4 (Fr4)). In addition, from the CD25⁺CD127^– gate, CD45RA⁺ (Fraction 2 (Fr2)) cells were gated, after which the GPA33⁺ (Fraction 3 (Fr3)) and GPA33^high (Fraction 7 (Fr7)) fractions were selected. (B) Gating of GPA33⁺ cells on the CD45RA against GPA33 (contour) plot within the CD25⁺CD127^– cell fraction. (C) Left: gating of GPA33⁺CD25⁺CD226^– (Fr6) cells on the CD226 against GPA33 (contour) plot within the CD25⁺ fraction. Right: representative flow cytometry histogram showing the expression of CD226 within the CD25⁺ population. (D) Left: representative flow cytometry histograms displaying the expression of CD127 in the indicated populations gated from the CD25⁺ gate as shown in (C). Right: bar graph showing the frequency of CD127⁺ cells in the indicated populations gated from the CD25⁺ gate as shown in (C) (n = 6 from six independent experiments). (E) Left: Representative CD226 versus GPA33 contour plot displaying gating on GPA33⁺CD226^– cells from the total CD4⁺ gate. Right: stacked bar graph showing the inclusion of CD25^– cells when gating on GPA33⁺CD226^– cells from the total CD4⁺ fraction (representative of n=6 from six independent experiments). (F) Left: Representative histograms displaying the expression of CD127 in CD25⁺ versus CD25^– cells in GPA33+CD226^– cells gated from the total CD4⁺ gate and right the quantification thereof (n = 6 from six independent experiments). Data are represented as mean ± SD; paired t‐test (F).

Apart from these populations, we characterized a CD4⁺CD25⁺ T cell population defined by expression of GPA33 and absence of CD226, a promising negative selection marker (Fraction 6: GPA33⁺CD25⁺CD226^–) [54]. This population can readily be obtained by first selecting CD25⁺CD226^– cells (among CD4⁺ T cells), followed by gating on GPA33⁺ cells (Figure 1A, top right plots). A clearly defined CD226^–GPA33⁺ population also appears when displaying these two markers directly from a CD4⁺CD25⁺ gate (Figure 1C, left). Among CD4⁺CD25⁺ cells, those that express CD226 also express CD127, identifying them as activated Tconv cells (Figure 1C,D). On the other hand, lack of CD226 expression excluded most, but not all, CD127⁺ Tconv cells (Figure 1C,D). Combined with GPA33 expression, however, the lack of CD226 identified a population of uniformly CD127^– cells (Figure 1C,D). Pregating on CD25⁺ cells is required for this, as applying this gating strategy from the total CD4⁺ gate results in the inclusion of a substantial fraction of CD25^– cells (Figure 1E) that are likely not Treg cells, given their significantly increased CD127 expression compared with the CD25⁺ portion (Figure 1F). Therefore, a combination of CD25, CD226, and GPA33 is sufficient to circumvent the need for CD127 gating, generally considered a critical marker for current Treg cell identification [27].

As we reported previously, the GPA33^high nTreg cell population (defined as cells with the top 20% highest expression of this molecule—Fraction 7, Figure 1A) was evidently much smaller than the total nTreg population (Fraction 2) [26]. However, selecting for the entire GPA33⁺ population among nTreg cells or from the total Treg cell (CD25⁺CD127^–) population yielded fractions (Fraction 3 and 4, respectively) of similar size to those of the total nTreg cell population (Fraction 2, Figure 2A) and significantly larger than the GPA33^high nTreg cell fraction (Fraction 7, Figure 2A). Populations of similar size were also obtained when gating on GPA33⁺CD25⁺CD226^– cells (Fraction 6, without selecting for the absence of CD127) (Figure 2A). The largest yields were obtained by gating on GPA33⁺CD25⁺ cells (Fraction 5), which were, on average, about 1.5‐fold larger than the total nTreg cell population (Fraction 2, Figure 2A).

FIGURE 2.

GPA33 isolation method yields Treg cells with high expansion capacity. The frequencies at day 0 and expansion of the respective FACS‐sorted populations were assessed. Cells were cultured for up to 21 days in the presence of soluble anti‐CD3, anti‐CD28, and IL‐2. (A) Bar graphs comparing the frequencies of total nTreg (Fr2, white), GPA33⁺ nTreg (Fr3, light gray), GPA33⁺ total Treg (Fr4, black), GPA33⁺CD25⁺ (Fr5, dotted), GPA33⁺CD25⁺CD226^– (Fr6, dark gray) and GPA33^high nTreg (Fr7, dashed) cells within the total live CD4⁺ fraction (n = 6 from six independent experiments). (B) Graph displaying the fold expansion of the concerned populations across the expansion period (at days 7 (n = 6), 14 (n = 6), and 21 (n = 4) (from at least four independent experiments). Note that fold expansion rates are relative to the previous timepoint. (C) Graph depicting the extrapolated cell counts on day 7, 14, and 21 of culture based on the respective fold expansion rates of the populations as shown in (B), except for GPA33^high nTreg cells (Fr7). Day 7 (n = 6), 14 (n = 6), and 21 (n = 4) from at least four independent experiments. Data are shown as mean ± SD; one‐way ANOVA (A).

The four GPA33⁺ populations, as well as nTreg cells, were expanded for up to 21 days with weekly restimulations using antibodies against CD3 and CD28 in the presence of recombinant IL‐2. Most robust expansion was seen after the first stimulation, and this rate diminished progressively with the second and third stimulations (Figure 2B). Importantly, all populations exhibited similar expansion capacity during culture, such that similarly high numbers of cells can be obtained from all GPA33⁺ populations as from the total nTreg cell fraction (Figure 2C). Given their similar population sizes and expansion capacities to those of the total nTreg (Fraction 2) cell population currently used for ACT, our results suggest that all GPA33⁺ populations tested should be able to generate sufficient cells for ACT.

3.2. Expanded GPA33⁺ Treg Populations Display High Purity and Are Unable to Produce Effector Cytokines

We previously showed that FOXP3⁺Helios⁺ Treg cells exhibit desirable properties for therapeutic use: while FOXP3^–Helios^– double negative (DN) and FOXP3⁺Helios^– single positive (FSP) cells could produce effector cytokines, this ability was absent from FOXP3⁺Helios⁺ double positive (DP) Treg cells, either freshly isolated or after expansion in vitro [26]. For this reason, we examined the expression profiles of these transcription factors in the various Treg cell populations (defined as described above) after expansion for up to 21 days of in vitro culture. GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), as well as GPA33⁺CD25⁺CD226^– (Fraction 6) cells had similarly high frequencies of FOXP3⁺Helios⁺ cells after two weeks of culture (Figure 3A,B), which, importantly, remained stable for the duration of the culture period up to three weeks (Figure 3C). In contrast, selecting for total nTreg (Fraction 2) cells or GPA33⁺CD25⁺ (Fraction 5) cells resulted in a markedly lower purity of FOXP3⁺Helios⁺ cells, which progressively decreased during culture (Figure 3B,C). The GPA33⁺CD25⁺ (Fraction 5) cell population yielded the highest frequencies of FOXP3^–Helios^– DN cells, in some donors approaching proportions found in Tconv (Fraction 1) cells (Figure 3B). This result is consistent with our earlier finding that a fraction of naïve Tconv cells expresses intermediate levels of GPA33 [26, 62]. Although expression of GPA33 is lost on Tconv cells after activation [62], this loss may not always be complete before CD25 appears on the surface of such cells.

FIGURE 3.

Expanded GPA33⁺ Treg populations display high purity, are unable to produce effector cytokines, and are suppressive. The respective FACS‐sorted populations were cultured for up to 21 days in the presence of soluble anti‐CD3, anti‐CD28, and IL‐2. Phenotypic analysis was performed throughout the culture period. (A) Representative flow cytometry contour plots showing expression of FOXP3 against Helios in the indicated populations at day 14 of expansion (n = 6 from six independent experiments). (B) Stacked bar graph displaying the frequencies of FOXP3⁺Helios⁺ DP (black), FOXP3⁺Helios^– FSP (light gray), FOXP3^–Helios^– DN (white) and FOXP3^–Helios⁺ HSP (dark gray) within the respective populations (n = 6 from six independent experiments) (C) Graph showing the frequency of FOXP3⁺Helios⁺ DP cells within the indicated populations after 7 (n = 6), 14 (n = 6) and 21 (n = 4) days of expansion (from at least four independent experiments). (D) After 14 days of expansion, cells were stimulated with PMA and ionomycin, after which flow cytometric intracellular cytokine staining was applied. Graphs depicting the percentage of IFN‐γ⁺, IL‐2⁺, and IL‐17A⁺ cells of the populations as indicated in (A) (n = 4 from four independent experiments). (E) After seven days of expansion, cells were co‐cultured with stimulated (anti‐CD3/anti‐CD28) responder cells for four days at different ratios as indicated. Graphs display the percentage of suppression (calculated as 100 – (percentage of proliferation in test condition/percentage of proliferation in responders only condition) × 100%) by total nTreg (Fr2, white), GPA33⁺ nTreg (Fr3, light gray), GPA33⁺ total Treg (Fr4, black), and GPA33⁺CD25⁺CD226^– (Fr6, dark gray) cells on CD4⁺ (above) and CD8⁺ (below) responder T cells (n = 2 from two independent experiments). Data are shown as mean ± SD; one‐way ANOVA (D).

Notably, we found total nTreg (Fraction 2) and GPA33⁺CD25⁺ (Fraction 5) cells to display high donor variability in their FOXP3^/Helios expression profiles (Figure 3B,C). In contrast, GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells consistently yielded expanded Treg products with high FOXP3⁺Helios⁺ content (Figure 3B,C).

To functionally investigate the purity of these cells, we examined the abilities of the different cell populations to produce Tconv effector cytokines after 14 days of in vitro culture, the typical expansion time of adult‐derived autologous Treg cells [12]. Similar to the GPA33^high nTreg (Fraction 7) cell population that we described earlier [26], we observed that the expanded GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cell populations all largely lack the ability to produce IL‐2, IFN‐γ, and IL‐17A (Figure 3D). These cytokines were, however, produced by cells derived from the total nTreg (Fraction 2) population, while the percentage of cytokine‐positive cells in the GPA33⁺CD25⁺ (Fraction 5) fraction approached that of Tconv cells, reinforcing the idea that this fraction of cells contains activated Tconv cells (Figure 3D). Overall, it appears that the use of GPA33 and CD25 alone results in a heterogeneous population with high donor‐to‐donor variability in terms of FOXP3⁺Helios⁺ purity and propensity for effector cytokine production. Therefore, this population was excluded from further analysis. Importantly, these data demonstrate that selecting for GPA33⁺ cells from a larger starting population than total nTreg cells, that is, from the total CD4⁺CD25⁺CD127^– Treg cell fraction, yields populations of cells with greater FOXP3⁺Helios⁺ purity and stability than total nTreg cells. Furthermore, by combining GPA33 with negative selection for CD226, a similarly pure and stable fraction can be obtained, obviating the need for CD127 in the isolation process.

3.3. GPA33⁺ nTreg (Fraction 3), GPA33⁺ Total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) Cells Are Fully Suppressive

To investigate the regulatory properties of GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells, we performed in vitro suppression assays. To this end, after seven days of expansion, the various expanded Treg cell populations were co‐cultured with CTV‐labelled PBMCs. After 4 days of co‐culture, cells were analysed. GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells clearly exhibited the capacity to suppress the proliferation of both CD4⁺ as well as CD8⁺ responder T cells in a cell concentration‐dependent manner, demonstrating that these stable fractions consist of fully functional Treg cells (Figure 3E). Moreover, their suppressive potential was comparable to that of total nTreg (Fraction 2) cells. All in all, these results show that GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells are bona fide Treg cells, also after in vitro expansion.

3.4. GPA33⁺ Treg Cells Maintain GPA33 Expression After in Vitro Culture

To obtain sufficient numbers for ACT, Treg cells must be expanded in vitro prior to transfusion. Although rigorous purification of the input population of cells before expansion is one way to go about it, it might be useful to further clean up populations after expansion if purity demands are very strict. Therefore, we investigated whether GPA33 is maintained during culture of the various Treg cell populations, so that it might be used to purge expanded populations from contaminating Tconv or unstable Treg cells. After 14 days of in vitro expansion, approximately 90% of GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells maintained expression of GPA33. Significantly fewer (on average ∼ 69%) GPA33⁺ cells were found in expanded populations stemming from the total nTreg (Fraction 2) cell population (Figure 4A,B). After 21 days of culture, a population of GPA33^– appears within the GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cell populations (Figure 4A), although 80%–90% of cells remained GPA33⁺. CD45RA⁺ Tconv (Fraction 1) cells rapidly lost GPA33 expression during culture (Figure 4A–C). GPA33 expression was maintained by GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells for the duration of the expansion period of 21 days, while GPA33 expression dropped in total nTreg (Fraction 2) cell populations after 14 days of in vitro expansion (Figure 4D). Initial selection for GPA33 expression ensured the isolation of Treg cells that near universally remained FOXP3⁺ (Figure 3). Nonetheless, GPA33 was not sufficient to always distinguish between FOXP3⁺ and FOXP3^– cells in expanded populations. Indeed, total nTreg cell populations sometimes contained cells that lacked expression of FOXP3, even though they were GPA33⁺ (Figure 4E,F). Thus, GPA33 expression on cells isolated freshly from blood helps robustly distinguish pure and stable cells from Tconv cells and Treg cells of lower purity and stability, but it does not by itself always accurately identify such cells after expansion.

FIGURE 4.

GPA33‐sorted Treg cells maintain GPA33 expression. The respective FACS‐sorted populations were cultured for up to 21 days in the presence of soluble anti‐CD3, anti‐CD28, and IL‐2. Phenotypic analysis was performed throughout the culture period. (A) Representative flow cytometry histograms showing the expression of GPA33 within the indicated populations at day 0, days 7, 14, and 21 of in vitro expansion (n = 5 ‐ 6). (B) Bar graphs depicting the quantification of the expression of GPA33 as shown in (A) at day 14 of culture within the total nTreg (Fr2, white), GPA33⁺ nTreg (Fr3, light gray), GPA33⁺ total Treg (Fr4, black) and GPA33⁺CD25⁺CD226^– (Fr6, dark gray) and Tconv (Fr1, dashed) cell populations (n = 5 from five independent experiments). (C) Bar graphs showing the frequency of GPA33⁺ cells within the indicated populations at day 14 of culture (n = 5 from five independent experiments). (D) Summary graph displaying the frequency of GPA33⁺ cells within the indicated populations across the expansion period (at days 7 (n = 6), 14 (n = 5), and 21 (n = 4) (from at least four independent experiments). (E) Representative flow cytometry contour plots showing expression of GPA33 against FOXP3 in the indicated populations at day 14 of culture. (F) Stacked bar graph displaying the quantification of GPA33^–FOXP3^– DN (white), GPA33^–FOXP3⁺ FSP (light gray), GPA33⁺FOXP3^– GSP (dark gray), and GPA33⁺FOXP3⁺ DP (black) cell frequencies within the respective populations as shown in (E) (n = 5 from five independent experiments). Data are shown as mean ± SD; one‐way ANOVA (B, C).

3.5. TIGIT Improves Identification of Stable Treg Cells After in Vitro Expansion

TIGIT is reportedly expressed in tTreg cells and is important for Treg cell effector function [26, 51, 63]. Strikingly, the majority of cells in two‐week expanded total nTreg (Fraction 2) cells lacked expression of this immunoregulatory molecule (Figure 5A,B). In contrast, most of the cells in the expanded populations, isolated based on expression of GPA33 (GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6)), were TIGIT⁺ (Figure 5A–C). The expression of this receptor in the GPA33⁺ populations remained stable for at least 21 days of in vitro expansion (Figure 5D). These high frequencies and levels of TIGIT expression further support the idea that the isolated GPA33⁺ populations contain high concentrations of tTreg cells [51].

FIGURE 5.

TIGIT improves the identification of stable Treg cells after in vitro expansion. The respective FACS‐sorted populations were cultured for up to 21 days in the presence of soluble anti‐CD3, anti‐CD28, and IL‐2. Phenotypic analysis was performed throughout the culture period. (A) Representative flow cytometry histograms showing the expression levels of TIGIT at day 14 of in vitro culture within the populations isolated on day 0 as indicated. (B) Quantification of the expression levels of TIGIT at day 14, as shown in (A) (n = 6 from six independent experiments). (C) Bar graphs depicting the frequency of TIGIT⁺ cells at day 14 of culture within the total nTreg (Fr2, white), GPA33⁺ nTreg (Fr3, light gray), GPA33⁺ total Treg (Fr4, black), GPA33⁺CD25⁺CD226^– (Fr6, dark gray) and Tconv (Fr1, dashed) cell populations isolated on day 0 (n = 6 from six independent experiments). (D) Summary graph showing the frequency of TIGIT⁺ cells at day 7, 14, and 21 of in vitro expansion of the populations isolated on day 0 as indicated (n = 4–6 from at least four independent experiments). (E) Representative contour plots displaying the expression pattern of GPA33 against TIGIT within the indicated populations at day 14 of in vitro culture. (F) Stacked bar graphs showing the distribution of GPA33^–TIGIT^– (black), GPA33⁺TIGIT^– (white), GPA33^–TIGIT⁺ (dark gray), and GPA33⁺TIGIT⁺ (light gray) cells after 14 days of in vitro culture of the populations isolated on day 0 (starting population) as indicated (n = 5 from five independent experiments). (G, H) Stacked bar graphs displaying the distribution of FOXP3^–Helios^– (DN, black), FOXP3⁺Helios^– (FSP, white), FOXP3^–Helios⁺ (HSP, dark gray) and FOXP3⁺Helios⁺ (DP, light gray) cells when gated on all, GPA33⁺, TIGIT⁺ or GPA33⁺TIGIT⁺ cells after 14 days of in vitro culture within the populations isolated on day 0 as indicated above each graph (n = 4–5 from at least four independent experiments). Data are shown as mean ± SD; one‐way ANOVA (B, C).

Given that both GPA33 and TIGIT seem to mark cells with the characteristics of tTreg cells [26, 51], we tested whether the combined expression of GPA33 and TIGIT may allow the identification of FOXP3⁺Helios⁺ tTreg cells after in vitro expansion. In line with this, we noted an enrichment for a TIGIT⁺GPA33⁺ phenotype in GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4) and GPA33⁺CD25⁺CD226^– (Fraction 6) cell populations when compared with the total nTreg (Fraction 2) cell population, while expanded Tconv (Fraction 1) cells exhibited very low frequencies of TIGIT⁺GPA33⁺ cells (Figure 5E,F). After 14 days of culture, GPA33⁺ alone failed to reliably identify FOXP3⁺Helios⁺ double‐positive (DP) cells in nTreg cell cultures (Figure 4E,G). Selection for expression of TIGIT did enrich for FOXP3⁺Helios⁺ DP cells in such cultures (Figure 5G). However, the best identification of FOXP3⁺Helios⁺ DP cells was obtained by selecting for the combination of TIGIT and GPA33 expression, mostly by eliminating the FOXP3^–Helios^– DN cell fraction (Figure 5G,H). The benefit of this post‐expansion selection was most prominent in expanded nTreg and total Treg cell cultures, given that the expanded GPA33⁺ Treg cells already gave rise to highly pure FOXP3⁺Helios⁺ populations.

We investigated whether other extracellular markers could enrich for FOXP3⁺Helios⁺ cells after 14 days of in vitro culture. These included CD27, reportedly involved in Treg cell stability and suppressive potency [64, 65, 66, 67]. Expanded GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) populations exhibited higher frequencies of GPA33⁺CD27 double positive cells than the expanded total nTreg (Fraction 2) cell population (Figure S2A (top), B (left)), but this was largely explained by the high frequencies of GPA33⁺ cells in these populations. Importantly, neither CD27 alone nor a combination of CD27 with GPA33 helped identify FOXP3⁺Helios⁺ DP cells in any of the expanded populations (Figure S3, top). We also tested whether expression of CD226 on expanded cells might be used to exclude FOXP3^– cells, given that CD226 has been used to distinguish Tconv from Treg cells [51, 54, 68]. However, despite the absence of CD226 on freshly isolated FOXP3⁺Helios⁺ DP cells (Figure 1), we found that most expanded Treg cells had acquired expression of this marker (Figure S2A (bottom), B (right)). Correspondingly, the lack of CD226 expression could not be used to identify FOXP3⁺Helios⁺ DP cells in any of the expanded populations (Figure S3, bottom).

In conclusion, expression of CD27 or CD226 cannot be used to clean up populations of Treg cells after expansion. Expression of TIGIT, on the other hand, does enhance the distinction between FOXP3⁺Helios⁺ DP cells from those with less desirable properties, but only after in vitro expansion. Moreover, the greatest purity of FOXP3⁺Helios⁺ DP cells post‐expansion is obtained through the combined use of GPA33 and TIGIT.

4. Discussion

It is widely believed that next‐generation Treg cell populations should be engineered or selected for reactivity to specific self‐antigens. This self‐reactivity increases the risk posed by contaminating Tconv or unstable Treg cells, necessitating the development of a method to identify and purify stable Treg cells. In this study, we further investigated the use of our previously reported surface marker GPA33 in the isolation of FOXP3⁺Helios⁺ Treg cells, which seem to possess the attributes of tTreg cells, believed to be the most stably committed to the Treg cell lineage. In particular, we investigated whether less stringent gating, that is, on GPA33⁺ as opposed to GPA33^high cells, in different populations yields a population of FOXP3⁺Helios⁺ cells that is large enough to be used for therapy. We demonstrate that selecting for GPA33‐positive cells in the total CD4⁺CD25⁺CD127^low/– Treg cell population (i.e., GPA33⁺ total Treg cells, Fraction 4) results in a cell fraction of similar size, but superior purity and stability, compared with total nTreg (Fraction 2) cells. We reveal that equally pure populations of FOXP3⁺Helios⁺ cells can be isolated by selecting for GPA33⁺ nTreg (Fraction 3) cells and cells with a GPA33⁺CD226^– surface phenotype among CD4⁺CD25⁺ cells (Fraction 6), with the latter circumventing the need for using CD127. Importantly, GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells had clear suppressive capabilities, comparable with those of total nTreg (Fraction 2) cells. We further note that although CD127 is not crucial, its replacement with CD226 is, as the combined use of GPA33 and CD25 alone results in a cell population with low FOXP3⁺Helios⁺ purity and the ability to produce effector cytokines. It seems likely that the major benefit of selecting against CD127 or CD226 expression is to exclude contaminating (activated) Tconv cells.

Interestingly, GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells exhibit highly similar phenotypic characteristics. These not only concern their intracellular expression profiles of FOXP3 and Helios, but also extracellular phenotypic (co‐)expression patterns of (GPA33 with) TIGIT, CD27, and CD226. These similarities suggest that while different gating strategies are applied to isolate these three populations, cells with overlapping (favorable) characteristics can be isolated when using GPA33. Given that these cells are most likely of thymic origin, they presumably carry TCRs directed toward self‐antigens rather than non‐self‐antigens [69, 70, 71]. Genetic modifications could enforce the therapeutic focus of these cells, for example, through the incorporation of receptors targeted toward relevant tissue‐specific antigens, depending on the disease condition [72, 73, 74, 75]. The capacity of GPA33⁺ nTreg (Fraction 3), GPA33⁺ total Treg (Fraction 4), and GPA33⁺CD25⁺CD226^– (Fraction 6) cells to efficiently expand in vitro whilst maintaining their suppressive capacity indeed makes them appealing candidate cells to generate safe Treg cell products directed at specific self‐antigens. In addition, expression of GPA33 and TIGIT can be used to further ensure the purity and safety of the product by allowing for depletion of potentially contaminating Tconv cells.

Our findings establish a refined framework for isolating human Treg cells at sufficient yield and stability to enable therapeutic exploration. Treg cell function is highly context‐dependent and shaped by the microenvironment, antigen availability, and cytokine milieu. Demonstrating that the populations identified here maintain lineage stability and suppressive function in vivo would therefore provide an important next step towards therapeutic application.

It should further be noted that cells in the present study were isolated from healthy individuals. Given that patients eligible for Treg ACT often suffer from chronic, systemic (auto)inflammation, further studies on how GPA33⁺ nTreg, GPA33⁺ total Treg, and GPA33⁺CD25⁺CD226^– cells behave in an inflammatory environment are warranted. Previously, we have shown that Treg cells derived from patients with T1D do express GPA33, both ex vivo and after expansion [26]. This suggests that GPA33 can be stably expressed by Treg cells in disease conditions where chronic inflammation occurs. We believe that the purity of the Treg cell populations obtained by the selection criteria characterized here makes GPA33‐based isolation procedures attractive for the preparation of safe Treg cell products for ACT.

Author Contributions

Florencia Morgana and Edith Slot designed and performed experiments. Florencia Morgana made figures and wrote the manuscript. Derk Amsen conceived of the project, supervised the work, and wrote the manuscript.

Ethics Statement

Human materials were obtained in compliance with the Declaration of Helsinki and the Dutch rules regarding the use of human materials from voluntary donors. Buffy coats were obtained from anonymized healthy donors after their written informed consent, as approved by the Ethical Advisory Council of Sanquin Blood Supply Foundation, Amsterdam, The Netherlands.

Conflicts of Interest

Derk Amsen owns a patent on the use of GPA33 for the purification of stable Tregs (WO 2018/178296). The remaining authors declare no conflicts of interest.

Supporting information

Supporting File 1: eji70107‐sup‐0001‐SuppMat.pdf.

Morgana F., Slot E., and Amsen D., “Isolation of Pure Stable Human Treg Cells Based on Expression of GPA33.” European Journal of Immunology 56, no. 2 (2026): e70107. 10.1002/eji.70107

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request (Derk Amsen, d.amsen@amsterdamumc.nl).