Carcinoembryonic antigen transgenic C57BL/6 mice were obtained from the Patterson Institute, Manchester, UK. The CEAtg mouse colony was bred by back-crossing with a colony of C57BL/6N mice (Charles River, Wilmington, MA, USA). Offspring mice were genotyped using the primer oligonucleotides 5′-CTGCAGCTGTCCAATGGC-3′ and 5′-CCTGGGACTGACCGGGAG-3′. C57BL/6 wild-type (wt) mice were used as controls. CEA-specific CAR transgenic (CEA-CARtg) C57BL/6 mice were generated by the laboratory of Prof. Abken (unpublished data). Briefly, embryonic stem cells were transfected with the Cre/loxP rosa26 vector containing a CD4 promoter-driven expression cassette coding for the fully murine SCA431scFv-mIgG-CD4(tm)-CD28–CD3ζ CAR. Blood T cells express the CEA-specific CAR on the cell surface. All experimental protocols were approved by the local animal welfare committee Agency for Nature, Environment and Consumer Protection of the State North Rhine-Westphalia (LANUV) and performed according to their guidelines.

Single cell suspensions from spleens and tracheobronchial lymph nodes (LNs) were obtained by meshing organs through a 70 µm cell strainer (BD, Franklin Lakes, NJ, USA), followed by lysis of erythrocytes and filtering through 30 µm cell strainer (Miltenyi Biotec, Bergisch Gladbach, Germany). Cells were stained with the anti-mouse monoclonal antibodies (mAb) CD4-FITC/clone GK1.5 (Southern Biotech, Birmingham, AL, USA), CD25-PerCP-Cy5.5/clone PC61 (BD), PE-conjugated anti-mouse CD25 mAb clone 7D4 (Miltenyi), FITC-labeled anti-mouse CD4 mAb clone GK1.5 (Southern Biotec), AF488-labeled anti-FoxP3/clone MF-14 (BioLegend, San Diego, CA, USA), RPE-conjugated anti-IgG₁ that binds to the CAR extracellular Fc domain (Southern Biotech), anti-latency-associated peptide (LAP)/clone TW7-16B4 (BioLegend), and “FoxP3 Staining Buffer Set” (Miltenyi Biotec). Data were acquired on FACS Canto II flow cytometer (BD) and analyzed using FlowJo software (FlowJo, LCC, Ashland, OR, USA).

T effector cells (Teffs) and Tregs were isolated from the murine spleens using the “Pan T Cell Isolation Kit II” and the “CD4⁺CD25⁺ Regulatory T Cell Isolation Kit” (Miltenyi Biotec), respectively, or using the “autoMACS” (Miltenyi Biotec). CAR Teffs were stained with PKH26 (Sigma-Aldrich, St. Louis, MO, USA) and stimulated for proliferation either by the immobilized agonistic mAb anti-CD3/clone 145-2C11 and anti-CD28/clone 37.51, or the anti-idiotypic CAR-activating mAb BW2064/36 directed against the IgG₁ spacer domain of the CAR (5 µg/ml coating concentration each). Teffs (10⁵ cells/well) were incubated with or without CAR Tregs (5 × 10⁴ cells/well) for 48 h. Intensity of PKH26 dye was recorded by flow cytometry using FACS Canto II and the proliferation rate of Teffs was calculated. For the LAP expression analysis, Treg cells were stimulated by incubation on plates coated with the anti-CD3 and anti-CD28 mAb, whereas anti-CEA CAR Tregs were incubated on plates coated with the BW2064/36 (anti-CAR mAb) for 48 h (5 × 10⁴ cells/well). Mock-coated plates (w/o) were used for control. LAP on the surface was detected using LAP-specific mAb using FACS Canto II. For IL-10 expression assay, CAR Tregs were cultured in triplicates in microtiter plates (PolySorp; Nalge Nunc, Rochester, NY, USA) precoated with the BW2064/36 mAb or an IgG₁ control mAb (Southern Biotech). After 72 h, IL-10 was measured in culture supernatants with the “mTh1/Th2/Th9/Th17/Th22/Treg Cytokine Panel 17-plex” (eBioscience/Thermo Fisher Scientific, Waltham, USA).

CAR Tregs and non-modified (wt) Tregs were retrovirally transduced to express Gaussia Luciferase (GLuc) and intravenously (i.v.) injected to ovalbumin (OVA)-sensitized CEAtg mice (24). 36 h later, Gluc-labeled cells were visualized by intraperitoneal (i.p.) injection of 100 µg benzyl-coelenterazine (PJK GmbH, Kleinblittersdorf, Germany). Lungs, spleen, stomach, and kidney were isolated and in the Petri dish recorded with 300 s exposure time using the Photon Imager (Biospace Lab, Nesles-la-Vallée, France). The threshold of bioluminescence signals was automatically determined using the Photo Vision software (Biospace Lab) and filtered against the background noise. Regions of interest were defined as regions above threshold and automatically gated by pre-defined program tools.

Ovalbumin (Grade V, Sigma-Aldrich) was used as model allergen (25). Lipopolysaccharide was removed from OVA by the Detoxi-Gel endotoxin removing gel and columns (Thermo Fisher Scientific). Clearance from endotoxin was confirmed by the Limulus amebocyte lysate test (Lonza, Basel, Switzerland). CEAtg mice were sensitized with two i.p. applications of 20 µg OVA in saline solution, adsorbed to 2 mg of aluminum hydroxide and magnesium hydroxide solution (alum; Inject alum adjuvant, Thermo Fisher Scientific), followed by intranasal (i.n.) exposures with 20 µg OVA at four consecutive days. Control mice were likewise treated with solutions without OVA. CAR Tregs for adoptive transfer were isolated from the spleens of the CEA-CARtg mouse or the C57BL/6 wt mouse using the “CD4⁺CD25⁺ Regulatory T cell Isolation Kit” (Miltenyi Biotec). Seven days after the first sensitization with allergen, CAR Tregs or wt Tregs were i.v. injected (1 × 10⁶ cells per mouse). One day after the last i.n. challenge, mice were sacrificed by asphyxiation with isoflurane (Baxter, Deerfield, IL, USA).

Airway hyper-reactivity was defined as an increase in dynamic lung resistance (R) in response to β-methacholine (MeCh; Sigma-Aldrich). One day after the last challenge with allergen, mice were deeply anesthetized with ketamine (Albrecht GmbH, Aulendorf, Germany) and xylazine (Rompun, Bayer Vital GmbH, Leverkusen, Germany), intubated, and mechanically ventilated via the flexiVent system (Scireq, Montreal, QC, Canada) using a tidal volume of 12 ml/kg at a frequency of 150 breaths/min. For baseline measurements mice were exposed to 0.9% (w/v) NaCl, aerosolized in the Aeroneb nebulizer (Inspiration Medical, Bochum, Germany). Subsequently, mice were provoked with increasing concentrations of MeCh (10, 20, and 30 mg/ml). After deep inflation of the lungs, single-frequency forced oscillations were performed; four peak values of R were analyzed for each MeCh concentration. AHR data were calculated as the change from baseline response.

Cryostat sections were stained with the biotin-conjugated anti-CEA mAb clone CB30 (Ancell, Bayport, MN, USA), streptavidin-horseradish peroxidase (BioLegend), and DAB chromogen substrate (Biozol, Eching, Germany). Sections were additionally stained with hematoxylin–eosin (H&E; Carl Roth, Karlsruhe, Germany) and analyzed using the Axiovert 400 M laser scan microscope (Carl Zeiss, Oberkochen, Germany). The left lung was fixed in 4% (w/v) formalin and embedded in paraffin. Tissue slices were stained with H&E (Merck, Darmstadt, Germany) or periodic acid-Schiff reagent (PAS; Sigma-Aldrich). Whole lung sections were scanned using the Keyence microscope BZ-9000 (Keyence, Osaka, Japan) at 100× magnification. Sum of H&E- or PAS-positive signal (pixels) per lung slice was quantified using in-house developed imaging software (25). For the immuno-histological detection of Tregs, lung and spleen cells were embedded in “Tissue-Tek O.C.T. Compound” (Sakura Finetek Europe B.V., Alphen aan den Rijn, Netherlands) and 5-µm cryostat sections were fixed in ice-cold acetone. Sections were stained for CD4 and FoxP3 expression with the fluorochrome-conjugated antibodies specific for mouse CD4-AF594/clone GKL1.5, FoxP3-AF488/clone 150D (BioLegend), and DAPI (4,6-diamidino-2-phenylindole dihydrochloride) for nuclear counterstain. For the detection of the CAR expression on the surface of Tregs, sections were incubated with the hybridoma-derived anti-idiotypic antibody BW2064/36 previously biotinylated by the Sulfo-NHS-LC-Biotin kit (Thermo Fisher Scientific). Following incubation with the primary antibody, antibody binding was visualized using AF-647 labeled streptavidin (BioLegend). Magnification was set to 60×. Slides were analyzed using the Olympus FV 1000 microscope (Olympus corporation, Tokio, Japan).

To obtain BALF, the right lung was flushed three times with 0.4 ml 2 mM EDTA in PBS. The total cell number was determined using the Cedex HiRes automated cell analyzer (Roche, Basel, Switzerland). To determine the cell types, cytospin slides were made with the CytoSpin 4 Cytocentrifuge (Thermo Fisher Scientific) and stained with May-Grünwald/Giemsa (Merck). Cell types were determined using standard microscopic criteria and counted in a blinded manner at 1000× magnification (Axiovert 40 CFL, Zeiss, Oberkochen, Germany).

Peripheral blood was clotted at room temperature for 20 min and the serum was collected. OVA-specific IgE, IgG₁, and IgG_2a were recorded by ELISA. Serum samples were incubated overnight in microtiter plates (Nalge Nunc) that were previously coated with 10 µg/ml OVA in bicarbonate buffer (pH 9.6). Specific binding was detected with respective goat anti-mouse horseradish peroxidase-conjugated anti-IgE antibodies (Southern Biotech), anti-IgG₁, or anti-IgG_2a antibodies (Bethyl Laboratories, Inc., Montgomery, TX, USA) and 3,3′,5,5′-tetramethylbenzidine as a colorimetric substrate. Optical density was determined at 450/630 nm using a microplate reader (Bio-Rad Laboratories, Inc., Hercules, CA, USA). OVA-specific IgE concentration in samples was determined using a standard calibration curve (AbD Serotec, Kidlington, UK). The titers of OVA-specific IgG₁ and IgG_2a were calculated by logarithmic regression as the reciprocal dilution of the sera. Cytokines in sera or cell culture supernatants were measured using the bead-based “Mouse 17-plex Bio-Plex multiplex system” and the Luminex xMAP device (Bio-Rad) or FlowCytomix (eBioscience/Thermo Fisher Scientific) and FACS Canto II.

The presence of CAR Tregs in tissues after adoptive transfer was additionally analyzed by RT-PCR. RNA was isolated from 4 to 5 million cells with “RNeasy Mini Kit” and “RNase-Free DNase Set” (Qiagen, Venlo, Netherlands). CAR cells were detected using OneStep RT-PCR Kit (Qiagen) and CAR specific primer oligonucleotides (5′-AAACAAACTGGAATGGATGGGCTACA-3′ and 5′-AACGTGGGATAACTACTCCACTGAT-3′).

Unpaired two-tailed Student’s t-test, one-way, or two-way ANOVA test with Bonferroni’s multiple comparison post-test (95% confidence interval) was performed using GraphPad Prism (San Diego, CA, USA). All data represent the mean ± SEM. Differences were considered statistically significant at p < 0.05 (*), p < 0.01 (**), and p < 0.001 (***).

The anti-CEA CAR was composed of the anti-CEA single chain variable fragment (scFv) for binding, the IgG₁ hinge-CH2CH3 as an extracellular spacer, the CD4 trans-membrane domain, and the intracellular CD28 and CD3ζ signaling domains for T cell activation upon binding to CEA (Figure 1A). CD4⁺CD25⁺ Tregs and CD4⁺CD25⁻ Teffs with anti-CEA CAR were isolated from the spleens of the anti-CEA CARtg mice. CAR Tregs stained positive for CD4, CD25, high level FoxP3 (Figure 1B), and the engineered CAR (not shown). To demonstrate the suppressive activity of CD4⁺CD25⁺ Tregs, cells were co-cultured with PKH26-labeled CD3⁺CD25^low CAR Teffs. CAR Tregs effectively repressed the proliferation of CD3⁺CD25^low CAR Teffs when stimulated through TCR/CD28 by the agonistic anti-CD3 and anti-CD28 mAb or through the CAR by the CAR-binding antibody BW2064/26 (Figure 1C). We conclude that the CD4⁺CD25⁺ CAR Treg cells constitute functionally active Tregs capable to suppress CAR Teff amplification. Treg cells, stimulated through the CAR, suppressed Teff cell proliferation more efficiently than after stimulation through CD3/CD28. This is likely due to differences in the CAR versus CD3/TCR-mediated T cell activation, i.e., the CD28ζ CAR provided much stronger CD28/CD3ζ signals than CD3/CD28 stimulation. Since the activation of Treg cells depends on the strength of the CD28 stimulation (19), stimulation through the CD28–CD3ζ CAR is likely superior to CD3/TCR stimulation. In accordance to our conclusion, CAR Treg cells showed higher activation levels than non-modified, wt Treg cells indicated by higher levels of LAP (TGF-β1) expression, as revealed by flow cytometry (Figure 1D). Furthermore, IL-10 production by CAR Tregs increased upon CAR stimulation through a specific antibody compared with the incubation with an isotype antibody of irrelevant specificity as control (Figure 1E). We concluded that antigen engagement by the CAR improved Treg cell activation in a specific fashion.

We used the immunocompetent CEAtg mice as recipients to evaluate in a clinically relevant model the immune-modulating effect of adoptively transferred CEA-specific CAR Tregs during induced allergic airway inflammation. The CEAtg mice express CEA under the control of the human CEA promoter on the luminal surface of the pulmonary (Figure 2A) and the gastrointestinal tract epithelia, closely mimicking the human situation. These mice were pre-sensitized by exposure to OVA; subsequently, one dose of non-preactivated, Gluc-labeled anti-CEA CAR Tregs or non-modified Tregs were applied by i.v. injection. CAR Tregs accumulated in the CEA⁺ lung, spleen, and stomach of OVA-treated mice 36 h after i.v. application while wt Tregs without CAR did far less (Figure 2B). Accumulation in CEA⁺ organs is specific since CAR Tregs did not substantially infiltrate the CEA⁻ kidneys.

We investigated whether adoptive transfer of CAR-redirected Tregs can suppress the clinical symptoms of experimental asthma in the mouse model. By systemic sensitization and local challenges with the model allergen OVA (Figure 3A), we induced the typical key features of asthma, such as AHR, eosinophilic airway inflammation, increased Th2 cytokine production, and elevated serum IgE levels. Control animals, inoculated with alum or NaCl solution without OVA, did not show these symptoms.

Adoptive transfer of CAR Tregs by one dose of i.v. injection to OVA-challenged CEAtg mice almost completely prevented the development of MeCh-induced AHR, whereas unmodified Tregs were less efficient as determined by invasive lung function measurements (Figure 3B). Histological examination of the lung sections revealed a substantial reduction of infiltrating inflammatory immune cells in the CAR Treg-treated mice compared with the non-treated asthmatic mice (Figure 3C). Treatment with Tregs without CAR did not substantially reduce airway inflammation demonstrating the superior therapeutic efficacy of redirecting Tregs by a CAR toward the affected organ.

Likewise, mucus production, one of the major hallmarks of asthma, was significantly suppressed upon CAR Treg cell therapy as revealed by PAS staining of the lung tissue (Figure 3D). In this respect, Tregs without CAR showed no substantial therapeutic effect.

Adoptive transfer of CAR Tregs reduced the number of eosinophils in the BALF by about threefold as compared with mice without Treg treatment while the number of macrophages was not altered (Figure 3E). Mice which received unmodified Tregs showed a less pronounced decrease in the number of total BALF cells and particularly of eosinophils.

Allergic asthma is characterized by a Th2-dominated immune response. To determine whether adoptive cell therapy with CAR Tregs modulates Th2 cytokine levels in OVA-challenged mice, we analyzed IL-5, IL-13, and IL-10 production, indicating progression of the disease. The superior suppression of asthma-like phenotype by CAR Treg treatment in relation to wt Treg injection was again demonstrated by significant reduction of antigen-specific IL-5 levels in cell culture supernatants of lung cells and splenocytes after in vitro restimulation with OVA (Figure 4).

We further examined IL-5 levels in mouse sera. CAR Treg-treated mice displayed significantly reduced IL-5 levels compared with untreated mice or mice treated with unmodified Tregs (Figure 5A). The effect was most prominent as early as 3 days after transfer of CAR Tregs, i.e., at day after second OVA inoculation.

OVA induced experimental asthma was accompanied by an increase of OVA-specific Ig levels in serum (Figure 5B). Adoptive CAR Treg transfer reduced the OVA-specific IgE levels, while non-modified Tregs did not. Essentially the same results were observed for IgG_2a, while IgG₁ levels were similarly reduced by Tregs with and without CAR (Figure 5B).

The pronounced effect of CAR Tregs compared with non-modified Tregs was due to CAR-mediated Treg cell activation. Transfer of CAR Tregs to OVA-treated wt C57BL/6 mice without CEA expression expectedly showed reduction in cell numbers in BALF compared to non-treated mice. However, the CAR Tregs were equally potent as the non-modified Tregs due to the lack of CAR-mediated Treg activation (Figure 5C). Accordingly, the levels of OVA-specific IgE were similarly reduced after Treg treatment with or without CAR (Figure 5D). We conclude that the suppressor activity of adoptively transferred Treg cells was substantially improved by anti-CEA CAR signaling through engagement of endogenous CEA.

To examine the presence of CAR Tregs in CEAtg mice after the last antigen challenge at day 12, lung and spleen cells were fluorescently stained with specific anti-CD4, anti-CAR, and anti-FoxP3 antibodies. At this time point, CD4⁺CAR⁺FoxP3⁺ cells were still present in the lung and spleen of the CAR Treg treated animals, whereas tissues from mice treated with non-modified Tregs did not stain positive for the CAR signal, as expected (Figure 6A). RT-PCR analysis confirmed the immunostaining data and furthermore revealed that the CAR Tregs also accumulated in the tracheobronchial LNs (Figure 6B).