Eight to twelve-week-old CAG-Epac1-camps and DEREG transgenic mice, as well as C57BL/6 wildtype mice, were used in this study. Generation of the CAG-Epac1-camps transgenic mice on FVB/N background was previously described (Calebiro et al., 2009). Briefly, in this transgenic mouse model, the cAMP sensor, Epac1-camps, is ubiquitously expressed under the control of cytomegalovirus enhancer/chicken β-actin (CAG) promoter in almost all tissues and cells (except erythrocytes and hair). Mice were kept under specific pathogen-free conditions in the central animal facility of the University Medical Center Hamburg-Eppendorf (UKE) and were handled according to international and national animal welfare guidelines with the organ isolation procedures approved by the Behörde für Justiz und Verbraucherschutz Hamburg (ORG1010, ORG946). Protocol for inducing experimental autoimmune encephalomyelitis and tissue sample collection was approved by the same authority (Nr 45/17).

Mouse (atrial NP, 1–28) and human natriuretic peptides (C-type NP, 1–53) were purchased from Bachem. BAY 60–7,550 (BAY) was ordered from Santa Cruz. 3-isobutyl-1-methylxanthine (IBMX) was obtained from AppliChem. All other chemicals were from Sigma-Aldrich.

Lymph nodes (superficial cervical, axillary, brachial, inguinal) and spleen were separately collected in ice-cold PBS. For immunoblotting single-cell suspensions were processed using the murine CD4⁺CD25⁺ Regulatory T cell Isolation Kit (Miltenyi Biotec) according to manufacturer’s instructions to obtain Treg and Tcon T cell subsets. For FRET measurements single-cell suspensions were processed using the murine CD4⁺ T cell Isolation Kit (Miltenyi Biotec) to enrich for CD4⁺ T cells via negative selection based on the manufacturer’s protocol and further labeled with CD25-Biotin antibody (Miltenyi Biotec) to separate CD4⁺CD25⁺ T cells from CD4⁺CD25⁻ T cells. T cell subsets were cultured 16–24 h in the presence or absence of Dynabeads Mouse T-Cell Activator CD3/CD28 (Thermo Fisher).

Murine T cells were centrifuged at 400x g for 4 min at 4°C, washed twice with ice-cold PBS, and lysed in RIPA buffer supplemented with Halt protease inhibitor cocktail (100x, Thermo Scientific). An equal number of cells was used continuously throughout the experiment. Samples were boiled at 95°C for 5 min. Protein lysates were loaded on 10% gel and SDS-PAGE was run followed by protein transfer onto nitrocellulose membrane (Amersham). Membranes were blocked in 3% or 5% non-fat milk prepared in TBS-T buffer, depending on the antibody, for 1 h at room temperature (RT). Subsequently, indicated primary antibodies were used in overnight incubation at 4°C: anti-PDE2A (1:750, Fabgennix), anti-PDE3B (1:2000, kindly provided by Dr. Sergei Rybalkin), anti-PDE4B (1:2500, Abcam) and anti-PDE4D (1:2500, Abcam). Anti-GAPDH (1:160000, Bio Trend) was used as a loading control in 30 min incubation at RT. After probing with the primary antibody, membranes were washed with TBS-T buffer three times and incubated for 1 h at RT with horseradish peroxidase (HRP)-conjugated secondary antibody (1:5000, goat anti-mouse or goat anti-rabbit, Biorad). To detect the signal, the SuperSignal West Pico PLUS Kit (Thermo Scientific) was used according to the manufacturer’s protocol. Densitometry analyses were performed on scanned blots in ImageJ software.

T cell subsets were obtained as previously described. T cells were plated to glass coverslips coated with poly-D-lysine (Sigma-Aldrich) for approximately 30–45 min before live cell imaging was performed. Coverslips with plated T cells were placed in the Autofluor cell chamber, non-adherent cells were removed with FRET buffer (144 mM NaCl, 5.4 mM KCl, 1 mM MgCl₂, 1 mM CaCl₂, 10 mM HEPES; pH = 7.3) and 400 µl of FRET buffer was subsequently added into the chamber. Data were obtained by monitoring FRET response with an inverted fluorescent microscope, Leica DMI 3000 B, equipped with an oil-immersion 63x/1.40 objective and a MicroManager 1.4. software. CoolLED, as a single wave emitting diode, was used as the fluorescent light source to excite cyan fluorescent protein (CFP) at 440 nm. Beam-splitter, DV2 Dual View (Photometrics), was used to split the emission light into CFP and yellow fluorescent protein (YFP) channel, simultaneously monitored on a CMOS (OptiMOS, QImaging) camera chip. Images were produced every 10 s. After reaching the stable baseline, different chemical compounds, diluted in FRET buffer, were added to the chamber in a volume of 400 µl. Raw data obtained by the aforementioned fluorescent microscope were analyzed in ImageJ and corrected offline with a spectral bleedthrough correction factor using Microsoft Excel.

For T cell activation assay, lymph nodes (superficial cervical, axillary, brachial, inguinal) and spleen from C57BL/6 wildtype mice were collected in ice-cold PBS. CD4⁺ T cells were isolated from single-cell suspensions using MojoSort CD4⁺ T cell isolation kit (BioLegend) according to manufacturer’s protocol. CD4⁺ T cells were seeded in an anti-CD3 (clone 145-2C11, BioLegend)-coated 96-well U-shape plate. Cells were supplemented with compounds (ANP and BAY) and soluble anti-CD28 (clone 37.51, BioLegend). Samples were incubated at 37°C and 5% CO₂ for 6 h. After incubation, cells were stained for T cell surface marker for 30 min at 4°C. After fixation and permeabilization with eBioscience™ Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher), Foxp3 was stained for 30 min at RT. The following antibodies were used: CD3ε (clone 145-2C11, BioLegend), CD4 (clone GK1.5, BioLegend), CD8a (clone 53–6.7, BioLegend), CD69 (clone H1.2F3, BioLegend), CD25 (clone PC61, BioLegend) and Foxp3 (clone FJK-16s, Thermo Fisher). Dead cells were excluded by adding Alexa Fluor 750 NHS Ester (Succinimidyl Ester) (Thermo Fisher). Nonspecific Fc receptor-mediated antibody binding was minimized by blocking with TruStain FcX anti-mouse CD16/32 (clone 93, BioLegend). Samples were acquired on the BD FACS LSR II analyzer (BD Bioscience) or FACSymphony A3 (BD Bioscience). Data were analyzed using FlowJo (version 10, BD Bioscience) (Supplementary Figure S1). ANP-mediated changes were indicated by calculating an activation index, denoted as the difference of CD25 or CD69 expression between compound-treated unstimulated and compound-treated anti-CD3/CD28 stimulated cells.

For induction of experimental autoimmune encephalomyelitis (EAE) 8–9 weeks old female B6-DEREG (C.B6-Tg(Foxp3-DTR/EGFP)23.2Spar/Mmjax) mice were immunized subcutaneously with 100 μg MOG_35-55 (peptides and elephants) in complete Freund’s adjuvant (BD Difco) containing 4 mg/ml Mycobacterium tuberculosis (BD Difco). 250 ng pertussis toxin (Merck Millipore) were injected intraperitoneally in 100 μl PBS on the day of immunization and 48 h later. Animals were scored daily from d7 for clinical signs of disease. At the peak of disease (d14/d15), immunized mice were sacrificed. Age and gender-matched naïve mice were sacrificed as healthy controls. T cells from acute EAE mice and healthy controls were isolated from the peripheral lymphatic organs (axillary, brachial, superficial cervical, inguinal lymph nodes and spleen). T cells were isolated using the MojoSort CD4⁺ T cell isolation kit (BioLegend) according to manufacturer’s instructions. Immune cells were identified by staining with following antibodies: CD4 (clone 145-2C11), TCRb (clone RM4-5), CD8a (clone 53–6.7), CD19 (clone 6D5) and CD11b (clone M1/70) (all BioLegend). Dead cells were excluded by adding Alexa Fluor 750 NHS Ester (Succinimidyl Ester) (Thermo Fisher). Nonspecific Fc receptor-mediated antibody binding was minimized by blocking with TruStain FcX anti-mouse CD16/32 (clone 93, BioLegend). Tcon and Treg were sorted with the FACS Aria III cell sorter (BD Bioscience). RNA was isolated using RNeasy Mini Kit (Qiagen) and cDNA was synthesized with RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Fisher). For quantitative real-time PCR, following TaqMan Gene Expression Assays were used: Mm01277042_m1 (mouse Tbp) and Mm01136644_m1 (mouse Pde2a) (both Thermo Fisher). Samples were analyzed with QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher) and by using the ΔCt method for relative quantification.

Data shown in bar graphs are presented as mean ± standard error of the mean (SEM). In immunoblot experiments, N represents the number of mice from which individual samples were prepared. For FRET measurements, number of mice (N) and number of single measured cells (n) are indicated below the graphs. Statistical analysis was performed using GraphPad software. Normal distribution was tested using Kolmogorov-Smirnov test with Dallal-Wilkinson-Lille approximation to calculate p value. Differences were tested using one-way ANOVA followed by Sidak’s or Dunett’s multiple comparison tests, Student’s t-test or Mann-Whitney U test, as appropriate, and indicated in the figure legends. Significant differences are indicated as *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Previous work that elucidated the impact of different PDEs in T cell subsets showed that the Pde3b gene, which encodes for the PDE isoform PDE3B, a cGMP-inhibited phosphodiesterase, is downregulated in Treg by strong Foxp3-mediated transcriptional repression (Zheng et al., 2007). However, the expression and regulation of the cGMP-activated phosphodiesterase family PDE2A in T cells have not been systematically investigated. To address the role of PDE2A and to confirm earlier findings on PDE3B in distinct T cell subsets, we analyzed the expression of both PDEs at the protein level. MACS-purified Tcon and Treg from mouse splenocytes and lymph nodes were cultured for 24 h in the absence or presence of anti-CD3/CD28 Dynabeads to induce T cell activation, followed by immunoblot analysis. We confirmed a significantly higher expression of PDE3B in both non-activated and activated Tcon as compared to Treg which barely express this PDE subfamily (Figure 1A). However, T cell activation did not have any effect on PDE3B expression. By contrast, stimulation of TCR engagement in Tcon led to a marked increase in PDE2A protein expression. Notably, this effect was detected exclusively in Tcon, whereas activation of Treg did not induce PDE2A upregulation (Figure 1B). Since the exposure of mouse thymocytes (Michie et al., 1996) and human T cells to the adenylyl cyclase activator forskolin or even TCR ligation was reported to upregulate PDE4 activity (Erdogan and Houslay, 1997) or protein levels of short PDE4B and PDE4D isoforms upon >24 h exposure (Peter et al., 2007), respectively, we tested whether TCR stimulation also changes PDE4 protein expression in mouse Tcon and Treg. Interestingly, under our experimental conditions we could not detect activation-dependent changes in PDE4B and PDE4D content (Figures 1C,D), although their expression in Tcon was constitutively higher than in Treg which is compatible with previously published data (Vang et al., 2013).

To further assess the cAMP hydrolyzing activity of PDE2A and PDE3B, we established real-time live-cell imaging of cAMP in T cells using the highly sensitive cytosolic FRET biosensor Epac1-camps (Supplementary Figure S2) (Börner et al., 2011; Sprenger et al., 2012). We isolated T cell subsets from splenocytes of transgenic mice ubiquitously expressing this biosensor (Calebiro et al., 2009) and cultured them overnight to analyze both activated and non-activated T cells. During FRET experiments, T cells were firstly stimulated with adenosine (ADO) to evoke intracellular cAMP production, followed by inhibition of specific PDEs using the selective PDE2A inhibitor BAY 60–7550 (BAY) or the selective PDE3 inhibitor cilostamide (CILO). The maximal cAMP response of the biosensor was subsequently induced by the non-selective PDE inhibitor 3-isobutyl-1-methylxanthine (IBMX). PDE2A inhibition led to significantly stronger cAMP response in activated as compared to non-activated Tcon (Figures 2A,B,E). Again, this effect was not observed for Treg (Figure 2E and Supplementary Figure S3), compatible with exclusive upregulation of PDE2A upon Tcon activation (Figure 1B). FRET responses to the PDE3 inhibitor CILO were higher in Tcon compared to Treg (Figures 2C,D,F and Supplementary Figure S3) as expected from the expression pattern of PDE3B (Figure 1A). However, no significant differences in PDE3 responses were observed in naive T cells compared to activated T cells in both Tcon and Treg (Figure 2F). Both BAY (in vitro IC₅₀ of ∼4 nM) and CILO (in vitro IC₅₀ of ∼70 nM) are selective for PDE2 and PDE3 (>100-fold selectivity vs other PDEs), respectively, in the concentration used (Sudo et al., 2000; Masood et al., 2009; Perera et al., 2015). To confirm their specificity, we have repeated these experiments using another pair of selective inhibitors PF-05180999 for PDE2A (Helal et al., 2018) and cilostazol for PDE3 (Schrör, 2002), which led to comparable results (Supplementary Figure S4). Thus, consistent with upregulation of PDE2A upon Tcon activation, FRET measurements also revealed higher functional activity of PDE2A in activated Tcon.

Since the activity of PDE2A and PDE3 is modulated by cGMP, we next studied how stimulation of cGMP production in T cells affects cAMP responses. To understand this cGMP/cAMP cross-talk which is best studied in other cells types upon co-stimulation of β-AR and NP receptors (Götz et al., 2014; Perera et al., 2015), we first stimulated T cells with the unselective β-AR agonist isoproterenol (ISO) to increase cAMP production. Next, we applied saturating concentrations of the natriuretic peptides ANP or CNP on top of ISO to increase cGMP generation and explore how it modulates intracellular cAMP levels in T cells. To fully saturate the FRET sensor at the end of the experiment, we applied the adenylyl cyclase activator forskolin (FSK) together with IBMX. Real-time measurements revealed that both ANP and CNP augmented ISO-stimulated cAMP responses in non-activated Tcon (Figures 3A,E,F). This was abolished by pretreatment of T cells with CILO, suggesting that cGMP inhibits the PDE3-mediated cAMP degradation (Figures 3B,E,F). In sharp contrast, activated Tcon responded to ANP and CNP with a decrease in cAMP levels, and this effect was sensitive to the PDE2A inhibitor BAY (Figure 3C,D–F, Supplementary Figure S3). This suggests that cGMP stimulates PDE2A activity, which dominates during T cell activation, due to the induction of PDE2A expression. However, no significant difference could be observed when looking at ANP or CNP effects in Treg, with or without TCR activation (Figures 3E,F), confirming that the modulation of cGMP/cAMP cross-talk depends on the T cell activation state and PDE2A induction exclusively in Tcon.

Finally, we set out to test the functional relevance of our findings and investigated the effects of ANP on T cell activation. Previously, it has been shown that increased cAMP levels inhibit TCR-induced T cell activation (Ruppelt et al., 2007; Brudvik and Taskén, 2012). Since PDE2A as a cGMP-activated PDE is upregulated during Tcon activation, we hypothesized that a rise of intracellular cGMP induced by ANP, via its receptor GC-A, could lower intracellular cAMP levels via PDE2A activation, further promoting T cell activation. After short-term engagement of the TCR expression of the early activation markers CD25 and CD69 on purified Tcon was significantly higher upon ANP treatment (Figures 4A,B) compared to vehicle treated cells. Co-treatment with BAY abolished the effect of ANP, confirming that GC-A/cGMP signaling acts via PDE2A to promote T cell activation (Figures 4A,B), while BAY alone had no significant effect (Supplementary Figure S5).

Finally, to test whether upregulation of PDE2A can also be detected in in vivo-activated conventional T cells, we analyzed Pde2a mRNA expression in FACS-purified Tcon that were isolated from draining lymph nodes and spleen from control mice or mice at the peak of EAE disease severity. Indeed, in this model of multiple sclerosis we could detect a 2-fold induction of Pde2a (Supplementary Figure S6) in Tcon, suggesting that this mechanism might play a role in T cell activation after autoantigen immunization.