IL-2 activity was determined by STAT5 activation (phospho-STAT5 (pSTAT) signal in a flow cytometry-based assay. Mouse splenocytes were treated with ACK buffer (155mM Ammonium Chloride +10mM Potassium Bicarbonate +0.25mM EDTA in H₂0, pH7.4), to lyse RBC and plated in complete RPMI media (RPMI1640 containing 10% FBS, buffered with HEPES GlutaMax, Penicillin-streptomycin, beta-mercaptoethanol, sodium pyruvate, and non-essential amino acid). Following a brief equilibration at 37°C, pre-diluted human WT IL-2, H16R, V91K D20A M104V (3x), D20W (all Fc-fused human IL-2 were produced at Amgen), or recombinant human IL-2 (R&D, cat# 10453-IL) was added to the wells and incubated for 20 min at 37°C. Cells were pelleted at 4°C and fixed in 1x Foxp3 fix/perm buffer and stained for Foxp3 using a kit (eBioscience cat # 00-5523). After Foxp3 staining, cells were fixed in Perm Buffer III of Phosflow kit (BD #558050) and stained for phospho STAT5, followed by extensive washing in FACS staining buffer (BD #554657), and stained for cell surface markers per standard protocol. pSTAT5 median fluorescence intensity (MFI) values from the Treg-gated population (Foxp3⁺ CD4⁺ CD3⁺ T cells) were used to generate IL-2 dose-response curves, and non-linear regression analysis was performed to generate EC50 values GraphPad Prism (v9.02; GraphPad Software, Inc., San Diego, CA).

Characterization of the in vivo Treg response and bulk RNA-Seq analysis were performed in 8-12 week-old C57BL/6 female mice. IL-2 was dosed at indicated doses by subcutaneous injection. When specified, isotype control (LTF-2, BioXcell #BE0090) or MHC II blocking (M5/114, BioXcell #BE0108) antibody was administered by intraperitoneal injection at 500 μg per dose on day -1 and day 0. Animals were sacrificed on day 4 or 5 for analysis.

All experimental studies were conducted under protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Amgen. The Amgen IACUC protocol number for IL-2 administration of mice is 2022-1526 and that for psoriasis model is 2022-1531. Animals were housed at Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International-accredited facilities (at Amgen) in ventilated micro-isolated housing on corncob bedding. Animals had ad libitum access to sterile pelleted food and reverse osmosis-purified water and were maintained on a 12:12 h light:dark cycle with access to environmental enrichment opportunities.

Male WT C57L/6 (7-8 weeks old) were purchased from Jackson Laboratory (CA, USA). Twenty-five micrograms of Fc-IL-2 proteins were dosed subcutaneously over the shoulders. Blood specimens were collected at various times post-injection (0.083h, 1h, 6h, 24h, 48h, 96h, 168h, 240h, 336h and 504h), incubated at ambient temperature for approximately 20 min or until fully clotted and then centrifuged to separate out the serum. All serum specimens were stored at -70°C ( ± 10°C) until use in analytical assays. PK parameters were estimated from mean serum concentration-nominal time data by noncompartmental analysis using Excel addins PKSolver (17). Serum concentration-time profiles were generated using GraphPad Prism (v9.02; GraphPad Software, Inc., San Diego, CA).

In vivo pSTAT5 response in mouse splenocytes was measured using the Phospho (Tyr694)/total STAT5a,b assay whole cell lysate kit (Meso Scale Discovery, #15163D) per manufacturer’s instructions and read on a Sector Imager 6000 (Meso Scale Discovery).

Single cell suspensions of skin and splenocytes were surface stained with antibodies, washed with PBS, then Live/Dead stained, fixed and intracellularly stained as previously described (16). Skin samples were passed through 40μm filter tubes final time before data acquisition. Flow cytometry was performed on BD Symphony A3 Cytometer (BD, San Jose, CA, USA).

Four different IL-2 treatment conditions (PBS, WT IL-2, H16R, and 3x mutein) within 2 antibody treatment groups (isotype and MHC II blocking antibodies) were evaluated with 4 mice in each group, resulting in a total of 32 Treg samples. CD25- Tcons from two mice from each group were included for comparison but were omitted from subsequent deep analysis. Treg and Tcon cells were purified from total splenocytes using a commercial mouse Treg isolation kit (Miltenyi # 130-091-041) following the manufacturer’s protocol. Individual mice were processed as separate samples. Tcons were collected as the flow-through fraction in the CD25 positive selection step. The bound fraction was eluted and collected as the Treg cells. Small aliquots were stained for CD3/CD4/CD25/Foxp3 to check for purity and indicated that most samples showed greater than 90% Foxp3⁺ staining post purification.

Sorted CD25- Tcon or Treg cell RNA was extracted by using Qiagen RNeasy mini kit (Qiagen #74106). The concentration and quality of the isolated RNA samples were assessed and determined on ThermoFisher NanoDrop 8000 spectrophotometer and Agilent 4200 TapeStation system, respectively, and had RNA integrity numbers at least 9.1. 200ng or 500ng RNA was used to prepare cDNA library by using a protocol modified from the Illumina TruSeq Stranded mRNA kit (Illumina #20020595). Briefly, after two rounds of polyA⁺ RNA selection and the step of fragmentation and priming at 94°C for 4 minutes, the fragmented RNA was reverse-transcribed to cDNA using SuperScript™ III (ThermoFisher Scientific Invitrogen™ #18080093) and RNase-out (ThermoFisher Scientific- Invitrogen™ #10777019) at 50°C for 1 hour. Sequentially, the products of the 1^st strand cDNA synthesis reaction were converted to double stranded cDNAs and subjected to end-repair, A-tailing, and adapter ligation by following up the commercial instruction. The constructed libraries were amplified and barcoded using the PCR program: a denaturation step at 98°C for 30 seconds, 15 cycles of 95 °C for 10 seconds, 60 °C for 30 second, 72°C for 30 seconds, and a 72 °C extension cycle for 5 minutes followed by a hold step at 4 °C. The amplified products were subjected to a clean-up step using AMPure XP bean (Beckman Coulter #63881) at the 0.85x ratio of beads to the PCR reaction volumes. After quantifying the clean-up products, the final barcoded libraries were sequenced to a minimum depth of 30 million paired-end reads at the 150nt read length on an Illumina Nextseq2000.

RNA-Seq sequencing reads were aligned using OSA aligner (18) embedded in v10.0 of the Omicsoft Array Suite (Qiagen) to mouse genome version B38 based on gene model GENCODE v19. Gene level expression were quantified based on Omicsoft implementation of RSEM (19), and represented as normalized fragments per kilobase per million reads (FPKM). 2D Principal component analysis (PCA) plots were generated based on transformed, variance stabilized counts (using the variance-stabilizing transformation (VST) function in DESeq2) of all expressed genes for all samples or Treg samples to explore the distribution of samples across cell type and treatment conditions. Genes that were expressed at very low levels (counts per million mapped reads > 0.5 in less than 4 samples), genes that did not code for protein were filtered out before downstream differential expression analysis by DESeq2 (v1.30.1) with default parameters (20). Six pairwise comparisons were performed (WT vs PBS, H16R vs PBS, 3X vs PBS in Isotype and MHC II blockade conditions respectively). All differentially expressed genes (DEGs, FDR<0.1) were Z-score normalized versus PBS treatment, and clustered (hierarchical clustering, default parameters) and visualized in a heatmap using ComplexHeatmap (v.2.6.2) (21). Boxplots of example gene expression across conditions were generated by QIAGEN OmicSoft Studio (V11.7, Qiagen). UpSet plots were generated with UpSetR (v1.4.0) R package, to visualize the intersections of significantly up/down-regulated genes (FDR<0.1) in WT vs PBS, H16R vs PBS and 3X vs PBS in the isotype control antibody-treated group.

We employed WGCNA (v1.71) R package to construct the weighted gene co-expression network (22, 23). Briefly, 19,662 genes with a total read count of more than fifty in all 32 Treg samples were kept for the analysis. The R function pickSoftThreshold was used to calculate the soft thresholding power and the optimal power was set as 5 to ensure scale-free topology fit index is above 0.9. A hierarchical clustering dendrogram of the topology overlap matrix (TOM) matrix was constructed with the parameter minModuleSize as 20 to identify gene co-expression modules. The module eigengenes (MEs) were defined as the first principal component expression of genes within a given module and calculated using the moduleEigengenes function in the R WGCNA package. It can be considered as a representative of the gene expression profiles within a module, which was further correlated with various sample traits by Pearson’s correlation.

DEGs from different comparison group and WGCNA module genes were further analyzed by Reactome pathways (2016 version) using clusterProfiler (v3.18.1) (24) and ReactomePA (v1.34.0) R packages (25) or canonical pathways Ingenuity pathway analysis (IPA v01-20-04; Qiagen).

Lu et al. murine IL-2 treatment single cell RNA-Seq processed data were generated in house and previously published (26). Processed scRNA-Seq count matrix and metadata by Miragaia et al. (27) was obtained from https://figshare.com/projects/Treg_scRNA-seq/38864. Processed count data with cell classification from Reynolds et al. were downloaded from ArrayExpress (E-MTAB-8142). Processed count data were integrated with meta data provided by author, further filtered, and normalized to focus on Tregs with Seurat (v4.0.6) (28) R package. Gene expression across different Treg subgroups were visualized by DotPlot function by Seurat.

Tregs were isolated from total splenocytes using Treg purification kit according to manufacturer’s instructions (Miltenyi #130-091-041). CD8 T cells were isolated from naïve C57BL/6 spleens with EasySep purification kit according to manufacturer’s instructions (Stem Cell Technologies #19853). 1e5 CD8 T cells were co-cultured in 0.2ml complete RPMI media with 2.5e4 T cell activating Dynabeads (Gibco #11456D) and Treg at indicated ratios overnight at 37°C. Cells were washed, stained, acquired on BD Symphony cytometer and analyzed with Flowjo 10.5.3 (Treestar). Geometric mean fluorescence intensity (gMFI) of CD69 in the CD8 T-gated population was used as activation readout.

Quantification of Fc-IL2 proteins in mouse serum was performed using electro-chemiluminescent immunoassays on the Sector S 600 plate reader (Meso Scale Discovery) with biotinylated anti-human IL-2 antibody (R&D Systems, BAF202) as the capture reagent and ruthenylated mouse anti-human IgG Fc (Amgen clone 1.35.1 Mab) as the detection reagent. The interpolation of analyte serum concentrations from standard curves was performed in Watson LIMS™ Non-GLP Prod.

Six to eight week-old female BALB/c mice were purchased from Charles River Laboratory and maintained on a 12-hour light/dark cycle under specific pathogen-free conditions with access to food and water ad libitum. Protocols were approved by Institutional Animal Care and Use Committee of Amgen. Mice were administrated subcutaneously with 25μg IL-2 muteins or PBS a day before (day -1) imiquimod application. A topical dose of 62.5mg of Imiquimod cream (5%), containing 3.125mg active compound, (Aldara; AMA HEALTH CARE INCDBA BURTS PHARMACY) was applied daily on the shaved back skin and right ears for 5 consecutive days. All mice were weighed daily, and the percent body weight was calculated as the percent of the initial body weight. The ear thickness was measured daily with a caliper. Blood samples were collected on d2 (d3 post IL-2 muteins injection). Harvested tissues were processed for histopathological analysis or flow cytometry analysis.

To explore the effects of attenuated IL-2 signaling in vivo, we chose two human IL-2 muteins that exhibited attenuated activity compared to WT IL-2 in in vitro phospho-STAT5 (Tyr694 pSTAT5) assay. These IL-2 molecules were fused to effector function-less human IgG1 Fc, thus exhibiting reduced binding to Fc-gamma receptor and increased in vivo half-life. Both muteins demonstrated significantly differentiated levels of attenuation compared to WT IL-2. H16R showed approximately a 20-fold shift in EC50 compared to WT IL-2 in human Tregs (1.168 nM vs. 0.057 nM). D20A V91K M104V (hereon referred to as 3x mutein) exhibited approximately 84-fold reduction in activity by EC50 (4.763 nM vs. 0.057 nM) and significantly reduced Emax compared to both WT IL-2 and H16R and is thus considered a far-attenuated mutein. D20W served as a negative control because it binds to IL2RA weakly but does not induce significant pSTAT5 response in vitro (16).

Using the STAT5 activation assay, we compared the rank order of activity of these human IL-2 molecules in human and mouse Tregs. As shown in Figure 1A, the EC50 values for human WT IL-2, the muteins, and recombinant IL-2 showed a strong correlation (R² = 0.9860) between human and mouse Tregs (Table 1). Additionally, the Emax values of the IL-2 muteins showed reduced maximal activities compare to WT IL-2 in mouse Tregs (Figure 1B), as previously observed in human Tregs. Of note, the activity of 3x mutein was less than 30% of WT IL-2 by Emax. D20W did not induce significant activity in mouse Tregs, similar to what was observed in human Tregs.

To characterize the effect of WT and attenuated IL-2 in vivo, mice were given a single dose of 1, 5, or 25 μg of WT IL-2, H16R, or 3x mutein. The inactive D20W mutein was given only at the 25 μg dose. On day 4, spleens were harvested and analyzed by flow cytometry. As shown in Figure 1C, the percentages of CD25⁺ Foxp3⁺ CD4 T cells representing Tregs increased in mice treated with IL-2, particularly with H16R or 3x IL-2 mutein. Both the percentages of Tregs within the CD4 T cell compartment and total Treg numbers increased in a dose-dependent manner in response to WT IL-2, H16R, and 3x muteins (Figure 1D). D20W did not induce significant response. Interestingly, the maximal Treg response generated by WT IL-2 and the muteins did not correspond to the relative rank order of activity in the in vitro STAT5 activation assays. Both attenuated muteins robustly increased Treg number and representation to a significantly greater extent compared to that induced by WT IL-2. In contrast to Tregs, other cell types including Tcon (Foxp3^- CD4 T) and NK cells showed responses that were commensurate with the in vitro activity of these molecules. WT IL-2 significantly increased the total Tcon and NK cell numbers while H16R and 3x did not (Figure 1D). The combination of the heightened Treg and reduced non-Treg responses confirmed that attenuated IL-2 muteins increased Treg selectivity in vivo. Additionally, and perhaps more surprisingly, we observed that the attenuated muteins demonstrated a disproportionately robust activity on Tregs in vivo, incongruent with the level of activity measured in vitro. The fact that the 3x mutein, which is a far-attenuated mutein, enhanced Treg expansion as well as WT IL-2 was unexpected.

It has been proposed that attenuated IL-2 selectively enhances Treg response due to the prolonged half-life and reduced binding to non-Tregs in vivo (29). We performed a mouse pharmacokinetic (PK) study and measured the serum Fc-IL-2 concentrations at various time points after dosing. We observed that the kinetics of WT and H16R were similar, as their concentrations dropped to undetectable levels beyond 96 hours (Figure 1E). In contrast, the serum levels of 3x mutein remained sustained and was detectable up to 240 hours. Compared to the reported half-life of Proleukin [0.74h, (30)], the half-lives of the Fc-fused WT, H16R, and 3x mutein were much longer, at 12.17, 8.43, and 29.3 hours respectively (Table 2). Additional PK parameters of Fc-IL-2 were generally consistent with the notion that IL-2 with reduced affinity to the receptor remained in circulation longer, although this effect was much more pronounced for 3x mutein compared to H16R.

We evaluated the in vivo pharmacodynamic (PD) activity of IL-2 by measuring the pSTAT5 signal and Treg expansion in splenocytes. Our data showed that the peak of the pSTAT5 signal was seen between 2 and 4 hours for all three IL-2 molecules, with WT IL-2 reaching the highest peak signal intensity compared to H16R or 3x mutein (Figures 1F, S1). Interestingly, 3x mutein uniquely maintained elevated pSTAT5 signals through much later time points, extending to 16 hours and beyond. The percentages of Tregs also increased over time, with delayed kinetics compared to the pSTAT5 response. The increase was seen as early as 16 hours post IL-2 dosing and further increased at 72 hours, at which point the two attenuated muteins showed greater Treg expansion compared to WT IL-2. These data showed a correlation between 3x mutein’s ability to generate a sustained IL-2 signal and Treg response in vivo. However, the PK and PD profiles of WT IL-2 and H16R were similar; thus it is unclear whether the superior activity of H16R in Treg expansion can be explained solely by its PK and PD profile.

Treg responses to IL-2 can be influenced by environmental signals beyond IL-2 availability. These include signals generated by the TCR (8), costimulatory receptors [e.g., CD28 (31), ICOS (32)], tumor necrosis factor receptor (TNFR) superfamily members [e.g., 4-1BB (33), OX40 (34), GITR (35)], cytokines [e.g., IL-33 (36), TGF-β (37)], and hormones such as leptin (38). We hypothesized that weak IL-2 muteins synergized with one or more of these signals in vivo to induce robust Treg response, rather than acting solely through STAT5 activation immediately downstream of the IL2R complex. Since TCR signal is an important driver of Treg responses, we focused on evaluating the contribution of TCR signal to the IL2-dependent Treg response. To this end, mice were pretreated with isotype control or MHC II-blocking antibody, followed by administration of a single dose of IL-2 (Figure 2A). IL-2 treatment increased the representation and the number of Tregs among CD4 T cells when the mice were pre-treated with the isotype control antibody, similar to the previous studies without any antibody treatment. In contrast, the Treg response was significantly diminished when mice were pretreated with MHC II blocking antibody (Figure 2B). By cell number, Treg expansion was greatest with the slightly attenuated mutein H16R, at nearly 8-fold increase over the PBS treated control (Figure 3C). The far-attenuated 3x mutein induced a similar Treg expansion response as WT IL-2. MHC II blockade dramatically reduced this response, indicating that the TCR signal is required for IL-2-driven Treg expansion in vivo. We also evaluated Treg activation response to IL-2 by assessing the expression of several known activation markers including Ki67, Gitr and Cd25. Both H16R and 3x muteins robustly induced expression of Treg activation markers, and the far-attenuated 3x mutein exerted an even stronger effect than WT IL-2 (Figure 2D). Furthermore, and in contrast to the Treg cell number, blocking the TCR signal had little to no effect. Tcons did not upregulate Ki67 or Gitr in vivo in response to any of the IL-2 molecules and the MHC II blockade had no impact (Figure S2), indicating that IL-2 selectively activated Tregs. These data clearly demonstrated the distinct requirement of TCR signal in IL2-dependent Treg expansion versus activation and suggested that the attenuated muteins could activate Tregs even when the TCR signal was compromised.

We sought to further understand how attenuated IL-2 muteins generated strong Treg activation responses in vivo by performing transcriptomic analyses. Applying the same study design to treat animals with isotype control or MHC II blocking antibody followed by IL-2 (Figure 2A), purified Tregs on day 4 were analyzed by bulk RNA-seq analysis. We also included Tcons (CD25- CD4 T cells) from two of the four mice from each treatment group as comparison. Principal component analysis (PCA) revealed a clear separation between the Treg and Tcon samples, confirming successful cell purification and expected distinct transcriptional programs of Tregs and Tcons (Figure S3). PCA of Treg samples revealed that the maximum variation was driven by treatment conditions; all IL-2 treatment conditions were clearly separated from PBS control, with the two IL-2 muteins further clustering apart from the WT IL-2 treatment. We observed further data separation based on antibody treatment (isotype versus MHC II blocking), particularly amongst samples treated with IL-2 (WT and the two muteins). The PBS-treated samples showed no separation based on antibody treatment, suggesting that blocking TCR signal by itself had little impact in the absence of IL-2 treatment within the short timeframe (4 days) of this study. Taken together, these data suggest a prominent change in gene expression following treatment with IL-2, with impact from the TCR blockade, particularly in the IL2-treated samples.

We analyzed the transcriptional response induced by WT IL-2, H16R and 3x mutein in the isotype antibody-treated group and performed differential gene expression analysis comparing the changes observed under IL-2 treatment versus baseline (PBS treatment). We set low cutoff thresholds (FDR<0.1, no fold change cutoff) to broadly capture all changes that may be of significance in the biological context. Comparison of the differentially expressed genes (DEGs) across different treatment groups was visualized by the UpSet plot (Figure 3A), which showed that IL-2 treatment led to both increased and decreased gene expression. At an FDR < 0.1, we observed a total of 10356 differentially expressed genes (DEGs) between any IL-2 treatment versus PBS. The majority (52% or 5342 genes) of the DEGs overlap across all treatments, with 48% (2559 out of 5347 genes) in downregulated and 54% (2764 out of 5098 genes) in upregulated DEGs. However, we also observed significant changes linked only to the IL-2 mutein treatments, indicating a potentially unique transcriptional response associated with attenuated IL-2 muteins.

To map out the gene response patterns across treatments, we subjected the set of DEGs identified in Figure 3A to unbiased hierarchical clustering. We identified a total 9 different clusters as indicated in a heatmap in Figure 3B. Clusters 1 and 2 captured genes that were largely downregulated by all IL-2 treatments, with the attenuated muteins inducing a greater degree of downregulation compared to WT IL-2. Clusters 5 - 9 captured genes that were largely upregulated across all IL-2 treatments. Clusters 8 and 9 showed a pattern that correlated with the in vitro activity of the three IL-2 molecules, i.e., ranking in the order of WT>H16R>3x by pSTAT5 response, while clusters 6 and 7 showed an opposite trend. Clusters 3 - 5 captured small numbers of genes that appeared to show conflicting trends amongst the different IL-2 species.

Surveying for a few exemplary genes that are commonly identified as IL-2 response genes in Tregs showed that they were scattered across different clusters (Figure S4 and Supplementary data file S1). We found many genes canonically associated with Treg function including Lrrc32 (Garp), Il10, Ctla4, and Amphiregulin (Areg) in cluster 9, and these genes showed a stronger response to WT and H16R than to 3x. Il2Rra (Cd25) was found in cluster 6 along with many genes involved in cell cycle regulation (e.g., Ccnb1, Ccnb2, Ccnd2), and these genes were induced to a greater extent by H16R and 3x mutein than WT IL-2. The fact that cell cycle regulators were expressed highly by the attenuated muteins was consistent with the greater expansion of Tregs associated with the muteins. Others such as Ki67, Il1rl1 (St2), Foxp3, Socs2, and Stat5b, that are known to contribute to IL-2 and/or Treg response pathways were found to be elevated by all three IL-2 species in cluster 7. Thus, attenuated IL-2 muteins can generate a canonical IL-2 activation response in Tregs, although with some differences compared to the WT IL-2 response.

To better understand the differences between WT IL-2 and mutein-generated responses in Tregs, we performed enriched pathway analysis based on Reactome 2016 (25), which suggested that individual clusters were associated with multiple pathways with overlapping biological implications (Figure S4). IL-2 treatment prominently increased expression of genes (Figure S4) that broadly regulate immune responses including some key Treg-associated genes as mentioned above (cluster 9), genes in the chemokine and cell adhesion receptor pathway (cluster 8, Ccr8, Icam1, Itgb8), and metabolic pathways such as pyruvate dehydrogenase complex (cluster 5, Pdk1, Pdha1, Atp5a1, Vdac1), consistent with the known role of IL-2 in Treg activation, migration, and metabolic reprogramming. The attenuated IL-2 muteins enhanced cell cycle (cluster 6) and axon guidance/Rho GTPase pathway (cluster 7, Cdc42, Pak2, Nck1, Nck2) responses, consistent with the robust Treg responses observed in mice treated with H16R or 3x mutein. Pathways involved in cholesterol biosynthesis (cluster 3, Mvk, Dhcr24, Sqle), TNFRSF-mediated non-canonical NF-kB pathway (cluster 2, Nfkb2, Relb, Traf3), and ribosomal RNA (rRNA) modification pathway (cluster 1, Ddx21, Nol11, Wdr36, Wdr75) were repressed in response to IL-2. Subtle differences between WT IL-2 and attenuated muteins were captured in cluster 4 which included several interferon signaling pathway genes (e.g., Irf7, Isg15, Ifit1), with greater degree of repression associated with the muteins, and an opposite trend in cluster 3 (the cholesterol biosynthesis pathway) where WT IL-2 led to a greater degree of repression.

Shared gene responses amongst the different IL-2 molecules are summarized in a dot plot in Figure 3C. Genes regulated by all three IL-2 molecules (2574 genes) are associated with RHO GTPases, phagocytosis, programmed cell death, and TCR signaling pathways, which point to fundamental effects of IL-2 on Tregs. Altered genes shared by WT and H16R (310 genes) are enriched for interferon signaling and cholesterol biosynthesis, while WT and 3x mutein shared only a small number of genes (87 genes) associated with pyruvate metabolism. Genes uniquely altered by both muteins (824 genes) are enriched for TGF-beta and NTRK signaling, providing additional insights into how IL-2 muteins may enhance Treg function, considering the role of TGF-beta signaling in promoting Treg function and peripheral Treg cell differentiation. This strengthens the notion that IL-2 (or specifically, the attenuated muteins) enforces Treg function and phenotype by feed-forward mechanisms. Furthermore, 3x mutein uniquely regulated genes (437 genes) highlighting additional cell cycle-related pathways, whereas H16R uniquely regulated genes (311 genes) involved in mRNA processing, DNA repair and snRNP assembly. These results suggest that the attenuated IL-2 muteins modulated shared and unique biological processes in Tregs compared to WT IL-2.

The inclusion of MHC II blockade treatment showed that certain aspects of the Treg response to IL-2 were co-regulated by the TCR signaling while others were not. To investigate the functional gene regulatory network underlying the IL2-induced transcriptional responses and the crosstalk with TCR signaling, we performed weighted correlation network analysis (WGCNA) (22, 23) of gene expression and identified 11 co-expression modules of genes (Figures 4A, B). The relationship amongst the modules and the number of genes in each module are presented in a dendrogram representation, which highlighted that the turquoise module, containing the largest number of genes, separated out as a distinct module.

We performed correlation analysis to understand how individual modules related to various treatment conditions and summarized the result in a heatmap (Figure 4C). For example, columns 1-5 show how the modules associate with MHC II blockade vs. isotype treatment across different IL-2 treatment conditions. The green module showed a statistically significant negative correlation with MHC II blockade, and thus it represented a gene module that showed dependence on the TCR signal regardless of the IL-2 status.

Similar analysis was performed across columns 6-11 to characterize IL2-dependent gene responses, and the turquoise module emerged as a prominent module that showed consistently positive correlation with the IL-2 treatment compared with the PBS controls. The turquoise module showed sensitivity to TCR blockade with WT IL-2 treatment, while the attenuated muteins elicited exaggerated responses that were not diminished by TCR blockade (Figures 4C, S5). This pattern suggested that WT IL-2 and TCR signals synergistically induced this module but that attenuated muteins could stimulate response even when TCR signal was limited. In contrast, the blue module was suppressed by IL-2 treatment, with greater suppression by muteins compared to WT. TCR blockade prevented the WT IL2- but not the attenuated mutein-mediated response.

The red and purple modules also showed a trend of negative correlation with MHC class II blockade, with the attenuated muteins generating stronger effect compared to WT IL-2. These trends suggested that the attenuated muteins induced a distinct signal from WT IL-2, with an ability to amplify signals where TCR signal is limited or reduced.

To explore biological processes represented by these gene modules, we conducted pathway enrichment analysis using Ingenuity Pathway Analysis (IPA), focusing primarily on the green module as the Treg gene module regulated by the TCR signal, and the turquoise module as that regulated by the combination of the IL-2 and TCR signals, with amplified induction by the attenuated muteins.

As shown in Figure 4D, the genes in the turquoise module were significantly enriched in pathways related to cellular metabolism, including mitochondrial dysfunction (e.g., Casp3, Gpx7, Prdx5, Sod2) and oxidative phosphorylation (e.g., Cyb5a, Ndufa11, Uqcrc1). T cell signaling-related pathways were also highlighted as the CTLA4 signaling in cytotoxic T (e.g., Akt1, Cd3d, Lck, Zap70), SLE in T cell signaling pathway (e.g., Foxp3, Icos, Rhog, Stat3), regulation of IL-2 expression in activated and anergic T lymphocytes (Calm1, Nfkb1, Smad3, Tgfb1), T cell exhaustion signaling pathway (e.g., Batf, Foxo1, Il10rb, Pdcd1, Prdm1, Tigit) and PI3K/AKT signaling pathway (Il1rl1, IL2ra, Itgb7). These gene profiles and associated pathways suggest that IL-2 and TCR signals synergize to enhance Treg’s metabolic fitness and increase their function, and the attenuated IL-2 muteins further amplified these effects with reduced requirement for TCR signaling. The genes in the green module were significantly enriched in the TNFRSF-related signaling pathways as suggested by death receptor signaling (e.g., Nfkbid, Rel), RANK signaling in osteoclasts (e.g., Map2k6, Traf6), NRF2-mediated oxidative stress response (e.g., Atf4, Sqstm1) and p38 MAPK signaling pathways. Since the green module genes are expressed in a TCR-dependent manner, this suggested that the TCR signal is essential for engaging these pathways in Tregs.

To illuminate biological pathways that are relevant to Treg states and function that underlie homeostatic and disease conditions, we curated a list of genes representing the pathways in the turquoise and green modules, which are summarized in a heatmap (Figure 5). These genes showed expression patterns that were consistent with the module response pattern as indicated by the module-trait analysis and were used as reference genes to further query pathway involvement in Treg responses in vivo.

We used a previously published single cell RNA seq (scRNA-Seq) dataset generated from mouse Tregs responding to a mouse IL-2 mutein (26) to confirm the relevance of our reference genes, which were generated from mouse Tregs responding to human IL-2 muteins. As shown in Figure 6A, these genes were largely enriched in the Treg population that had been annotated as “activated” by IL-2 compared with the “resting” Treg population. The transitional population annotated as “Primed/activated” showed intermediate expression. This analysis demonstrated that mouse Tregs responded to human IL-2 in vivo similarly as they did to murine IL-2, and the gene responses captured by the turquoise and green modules reflected the Treg activation state.

To deconvolute which Treg subsets responded to IL-2, we used a comprehensive mouse Treg dataset from a study published by Teichmann and colleagues (27), which had defined functional Treg subsets including Treg_stress, Treg_STAT1, Treg_NLT-like, Treg_effector (eTreg), and Treg_suppressive, which were proposed to represent the various stages of Treg adaptation during homeostatic circulation based on kinetic trajectory analysis. Additionally, it had mapped subpopulations of Tregs in lymphoid tissues (spleen, brachial lymph node (bLN) and mesenteric LN (mLN), collectively referred to as lymphoid tissue, or LT) and the mucosal barrier tissues (skin and colon, referred to as non-lymphoid tissue, or NLT). Querying this dataset for the expression of the turquoise and green module genes showed that these genes were preferentially enriched in Treg subsets that had been annotated as suppressive, NLT, NLT-like, and to a lesser extent, eTreg subpopulations (Figure 6B). Based on these results, we propose that IL-2 promotes Tregs to adapt aspects of the tissue-resident NLT phenotype and to take on functional traits of effector and suppressive Treg cells. Additionally, transcriptional signatures accompanying these phenotypic changes can be captured in lymphoid tissues like the spleen.

The turquoise and green module gene signatures were prominently enriched in Tregs from NLTs, such as the skin and the colon (Figure 6C). Among the turquoise module pathways, the SLE T cell signaling pathway, regulation of IL-2 expression, and PI3K/AKT signaling pathways (mostly represented by cytokine receptor and adhesion receptor genes) were highlighted, while the green module showed a prominent association with the death receptor signaling. Interestingly, the CTLA4 signaling in cytotoxic T pathway genes, including those proximally involved in TCR receptor signaling (e.g, Cd3d, Cd3g, Lck), showed an opposite trend and these genes appeared to be expressed at slightly higher levels in LT than in NLT Tregs. The mitochondrial dysfunction and oxidative phosphorylation pathway genes also tended to be expressed at higher levels in the LT Tregs. These results indicate alternative IL-2 response profiles between LT and NLT Tregs, which consequently may reflect differences in their sensitivity to TCR signaling, tissue tropism, and/or metabolic response to their environment.

One hypothesis around inflammatory diseases is that reduced number of Tregs and/or activation limit the suppression of inflammatory response in tissues. To evaluate Treg activation state in human diseases, we characterized the turquoise and green module Treg gene signatures in human inflammatory disease datasets. Because these gene modules showed a compelling association with skin-resident Tregs, we focused our analysis on two human skin inflammatory diseases, atopic dermatitis and psoriasis, using the dataset published by Teichmann, Watt, Haniffa and colleagues (40). We queried the turquoise and green module reference genes to determine whether Tregs displayed activation gene signatures in healthy versus inflamed tissues (Figure 6D). In this analysis, IL2-induced Treg activation genes were modestly represented across healthy, non-lesional or lesional tissues, with some that differed between healthy vs non/lesional skin. For example, the CTLA4 signaling in cytotoxic T pathway genes in the turquoise module were expressed robustly in healthy skin but appeared reduced in both lesional and non-lesional skin in AD and psoriasis. The genes in the green module were not expressed in Tregs in healthy skin, while a few (e.g., Lmna, Rel, Atf4, Sqstm1, Dusp10) appeared to be elevated in skin Tregs in disease patients. Overall, the inflamed skin Tregs did not express high levels of IL-2 activation signature genes, suggesting that Treg activation is limited at disease sites.

Functionally activated effector Tregs (eTregs) are distinguished from naïve central Tregs (cTregs) by multiple markers, some of which are directly involved in Treg’s ability to suppress inflammation and mediate tissue repair. As a readout of Treg functionality following IL-2 treatment, we assessed transcript levels for genes that are considered eTreg markers. As shown in Figure 7A, eTreg marker genes including Icos, Ctla4, Klrg1, and Il10 were robustly upregulated in response to WT IL-2, and this response was largely abolished by the MHC II blocking treatment. Thus, WT IL-2 enhances eTreg phenotype in a TCR-dependent manner. The attenuated muteins H16R and 3x muteins also upregulated the eTreg marker genes. Interestingly, some genes induced by the attenuated muteins were not abrogated by the MHC II blockade (e.g., Icos, Ctla4, and Il10) and for some, such as Il10 and Lag3 (not shown), the response was even elevated. These results suggested that the attenuated IL-2 muteins could enhance the eTreg phenotype even when the TCR signal was compromised.

To determine whether IL-2’s positive impact on eTreg gene signatures consequently led to enhanced Treg function, we performed an in vitro suppression assay. In a co-culture assay where purified CD8 T cell activation was assessed by CD69 expression, Tregs from mice treated with WT IL-2 showed a greater degree of suppression compared to Tregs from PBS-treated mice, suggesting that WT IL-2 enhanced Treg suppressive function in vivo (Figure 7B). Interestingly, MHC II blockade had little impact on the ability of IL-2 to enhance Treg function in this assay, suggesting that blocking the TCR signal in vivo did not impair Treg suppressive function in the short term up to 4 days. The two attenuated muteins also enhanced Treg function, although the level of suppression was intermediate between the WT IL2- and PBS-treated Tregs.

Molecular pathways implicated in the regulation of eTreg differentiation include the NF-kB pathway, triggered by engagement of one or more of the TNFRSFs expressed on Tregs (35), and the TCR pathway (8, 41), triggered when Tregs encounter the cognate antigen. Furthermore, several Foxp3 target genes including Irf4, Rela, and Prdm1 (encoding Blimp-1) have been shown to confer specialized eTreg phenotype and function (42, 43). We interrogated whether IL-2 regulated the expression of Irf4, Rela, and Prdm1 in Tregs in vivo and whether TCR signal was involved. The Irf4 and Rela expression did not increase in response to WT IL-2 or the attenuated muteins, while Prdm1 expression did, although the overall expression level was low (Figure 7C). These results suggest that IL-2 induces eTreg marker gene expression without directly increasing the expression levels of Irf4, Rela, and Prdm1.

We had observed that the turquoise module included genes that are known to drive TCR signaling. Thus, we considered the possibility that IL-2 enhanced TCR signaling in a positive feedback loop, as a mechanism of enhancing the eTreg gene signature downstream of TCR. In support of this hypothesis, Zap70, Trac (encoding TCR alpha), and Slp76 expression increased in response to IL-2, and interestingly, the attenuated muteins were able to enhance their expression even when TCR signaling was blocked (Figure 7D). In fact, Zap70 expression in response to the muteins was further amplified when TCR signaling was blocked. Additionally, several genes known to be involved in the calcium signaling pathway downstream of TCR, including Stim1 (Figure 7D), Orai1, and Nfatc3 (data not shown), were elevated, indicating that attenuated IL-2 signal may also enhance the calcium signaling pathway in Tregs.

Based on these results, we hypothesized that the IL-2 crosstalk with the TCR pathway directly enhanced Treg sensitivity to antigenic signal and adaptation of the eTreg phenotype. We tested this hypothesis by evaluating the expression of Nur77, which is a TCR-dependent gene known to be upregulated in circulating Tregs (9). Intracellular staining of Nur77 showed that Tregs from mice treated with 3x mutein indeed showed a strikingly elevated level of Nur77 (Figure 7E).

Given that the IL-2 response gene signature was particularly enriched in the skin, we decided to evaluate the effects of IL-2 molecules in a mouse model of psoriasis. Human Fc-conjugated IL-2 molecules lose in vivo exposure as early as after the first dose (Figure S6), potentially due to anti-drug antibody (ADA) response; thus, we opted to evaluate the impact of IL-2 with a single dose in a short-term psoriasis model induced by application of the imiquimod (IMQ) cream.

We administered a prophylactic dose of IL-2 on day -1 and induced disease by daily application of the IMQ cream for 5 days (Figure 8A). On day 2 (3 days after IL-2 treatment), we detected a robust increase in the percentage of Tregs in blood following treatment with WT IL-2 or H16R, while 3x mutein showed a trend of increase (Figure 8B). The attenuated muteins induced negligible NK and Tcon responses, but WT IL-2 induced significant expansion of NK cells. These results confirmed that attenuated IL-2 muteins induced Treg expansion in vivo with increased Treg selectivity even in an inflammatory setting. Furthermore, Tregs in mice treated with IL-2 expressed higher levels of activation and eTreg markers, including Gitr, Icos, and Klrg1, validating that IL-2 induces eTreg marker gene expression at the protein level (Figure 8C).

With a single dose administration of IL-2, we did not observe disease improvement measured by body weight and skin thickness (Figures S7A, B). In this model, the skin pathology is characterized by immune cell infiltration, with gamma delta (γδ) T cells specifically implicated as the main driver of disease (44). On day 5 at the end of the study, the percentage of γδ T cells was strikingly elevated, while the percentages of Tregs and Tcons were reduced, in the skin of diseased (i.e., IMQ-treated) compared to naïve mice (Figure 8D). Mice treated with WT IL-2 maintained these same disease-associated changes, while H16R and 3x mutein treatment suppressed the accumulation of γδ T and NK cells in the skin (Figure 8E). Evaluation of Ki67 showed that the skin γδ T and NK cells proliferated similarly in all diseased animals regardless of the IL-2 treatment (Figure S7C). These results suggest that IL-2 promotes expansion of circulating Tregs, enhances the eTreg phenotype and suppresses immune cell (including γδ T and NK cells) trafficking to the inflamed tissue. In our study, the two attenuated muteins demonstrated improved Treg selectivity and ability to suppress pathogenic cell infiltration compared to WT IL-2 in a disease setting.