Research ArticleCancer

Interleukin-2 signals converge in a lymphoid–dendritic cell pathway that promotes anticancer immunity

See allHide authors and affiliations

Science Translational Medicine  16 Sep 2020:
Vol. 12, Issue 561, eaba5464
DOI: 10.1126/scitranslmed.aba5464

Hidden virtues of interleukin-2

Interleukin-2 (IL-2) is an immunostimulatory cytokine that is being tested as an anticancer therapy due to its ability to activate T cells and natural killer cells. Unexpectedly, Raeber et al. discovered that IL-2 can also activate dendritic cells, an entirely different type of cells involved in antitumor immunity. The authors analyzed data from a clinical trial in human patients being treated with recombinant IL-2. They also pursued more detailed studies in mouse models to identify the mechanisms responsible for dendritic cell activation, helping to explain how IL-2 sensitizes tumors to immune destruction.


Tumor-infiltrating dendritic cells (DCs) correlate with effective anticancer immunity and improved responsiveness to anti–PD-1 checkpoint immunotherapy. However, the drivers of DC expansion and intratumoral accumulation are ill-defined. We found that interleukin-2 (IL-2) stimulated DC formation through innate and adaptive lymphoid cells in mice and humans, and this increase in DCs improved anticancer immunity. Administration of IL-2 to humans within a clinical trial and of IL-2 receptor (IL-2R)–biased IL-2 to mice resulted in pronounced expansion of type 1 DCs, including migratory and cross-presenting subsets, and type 2 DCs, although neither DC precursors nor mature DCs had functional IL-2Rs. In mechanistic studies, IL-2 signals stimulated innate lymphoid cells, natural killer cells, and T cells to synthesize the cytokines FLT3L, CSF-2, and TNF. These cytokines redundantly caused DC expansion and activation, which resulted in improved antigen processing and correlated with favorable anticancer responses in mice and patients. Thus, IL-2 immunotherapy–mediated stimulation of DCs contributes to anticancer immunity by rendering tumors more immunogenic.


Dendritic cells (DCs) are a subgroup of professional antigen-presenting cells considered indispensable in orchestrating T cell responses to intracellular pathogens and tumors (13). Human blood DCs have traditionally been subdivided into conventional DCs (cDCs) and plasmacytoid DCs (pDCs); however, results from single-cell RNA and protein analyses identified additional subpopulations (4, 5). In mouse lymphoid organs, DCs are correspondingly divided into cDCs and pDCs. Transcriptionally, cDCs can be further differentiated into type 1 cDCs (cDC1) that are controlled by interferon regulatory factor 8 (IRF8) and basic leucine zipper transcription factor activating transcription factor–like 3 (BATF3) and type 2 cDCs (cDC2) that are controlled by IRF4 (2, 3). The transcription factor E2-2 regulates pDCs (3). Phenotypically, cDC1s are characterized by the presence of CD141 (also known as BDCA-3) and DNGR-1 (also termed CLEC9A) in humans and CD8α and CD103 (also known as integrin αE), DNGR-1, and the chemokine receptor XCR1 in mice. Conversely, cDC2s are marked by CD1c (also referred to as BDCA-1) in humans and CD4 and CD11b in mice (6).

DC subsets in nonlymphoid tissues, including the tumor microenvironment (TME), vary considerably in terms of phenotypic and functional properties (7). In cancer, rare tumor-infiltrating cDCs attract T cells to the TME, where the cDCs stimulate CD8+ T cells by presenting tumor antigens and producing interleukin-12 (IL-12) (814). However, the upstream molecular and cellular factors favoring the on-demand generation and expansion of cDCs in antitumor responses are ill-defined. Two studies have implicated natural killer (NK) cells in facilitating DC infiltration of tumors, which correlated with prolonged survival in humans (15, 16).

NK cells are lymphoid cells, and their survival and homeostasis depend on signals mediated through the common gamma chain cytokine receptor (γc, also termed CD132), encoded by Il2rg. Members of the CD132 cytokine family comprise IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 (17, 18). IL-2 signals either through an intermediate-affinity dimeric IL-2 receptor (IL-2R), composed of IL-2Rβ (CD122) and CD132, or a trimeric IL-2R additionally including IL-2Rα (CD25). The dimeric receptor is found mainly on memory CD8+ T and NK cells, whereas the trimeric receptor is predominantly found on regulatory T (Treg) cells at the steady state and is transiently up-regulated on recently activated effector T cells (17, 19). In addition to its effects on T cells and NK cells, IL-2 can also stimulate innate lymphoid cells (ILCs), particularly type 2 ILCs (ILC2), NKT cells, and activated B cells, as well as certain nonimmune cells (1921). However, IL-2 is not known to affect DC homeostasis in vivo.

Thus, it was entirely unexpected to observe a prominent increase in several DC subsets in both mice and humans during our studies on IL-2 immunotherapy. We here describe a pathway driven by IL-2 and mediated by innate and adaptive lymphoid cells. This pathway stimulates DCs and promotes their expansion, processes that contribute to improved antitumor immune responses, correlating with prolonged survival in mice and humans.


IL-2 immunotherapy expands cDCs in mice and humans

DCs are characterized by the absence of lineage (Lin) markers, have intermediate (int) or high (hi) CD11c, and can further be subdivided into CD11cint B220hi pDCs and CD11chi major histocompatibility complex class II (MHC-II)hi cDCs, including CD11blow XCR1+ CD8α+ DNGR-1 (CLEC9A)+ cDC1s and CD11bhi XCR1 cDC2s (Fig. 1A and fig. S1A). A short course of three injections of recombinant human IL-2 (IL-2; teceleukin) increased total counts of cDCs in spleens of adult wild-type mice (Fig. 1B). This expansion was due to active proliferation of cDCs, as evidenced by increased incorporation of the thymidine analog bromodeoxyuridine (BrdU) into cDCs (Fig. 1C). To assess whether this IL-2 effect was caused by binding of IL-2 to CD25hi or CD122hi cells, we used CD25-biased IL-2/anti-IL-2 (5344) antibody complexes (IL-2/5344) and CD122-biased IL-2/anti-IL-2 (NARA1) antibody complexes (IL-2/NARA1; referred to as IL-2cx exclusively hereafter) (2224). Both CD25- and CD122-biased IL-2/anti-IL-2 antibody complexes stimulated qualitatively and quantitatively comparable expansion and proliferation of splenic DCs as unbiased IL-2 (Fig. 1, B and C, and fig. S1, B and C), whereas treatment of Il2rg−/–Rag2−/− mice with IL-2cx did not expand DCs (fig. S1D), thus excluding non–IL-2–mediated stimulation by interactions of anti–IL-2 antibody with Fc receptors on DCs. Moreover, IL-2cx not only mediated expansion of splenic cDCs but also increased total cDC, cDC1, and cDC2 counts in mouse lymph nodes (fig. S1E).

Fig. 1 IL-2 immunotherapy expands cDCs in mice and humans.

(A) Representative gating strategy for mouse splenic DC subsets including plasmacytoid DCs (pDC) and conventional type 1 and type 2 DCs (cDC1 and cDC2). (B) Total splenic cDC counts after treatment with IL-2 or IL-2 complexes (IL-2cx) in wild-type (WT) mice. Data are presented as means ± SEM (n = 7 to 9 mice per group from N = 3 independent experiments). (C) Proliferation of splenic cDCs from mice receiving the indicated treatments was measured by bromodeoxyuridine (BrdU) incorporation over 3 days. Data are presented as means ± SEM (n = 7 mice per group from N = 3 independent experiments). (D and E) Quantification of total numbers (D) and proliferation (E) of splenic cDC subsets from mice receiving the indicated treatments. Data are presented as means ± SEM [n = 19 (D) and 16 (E) mice per group from N = 8 (D) and 6 (E) independent experiments]. (F) Study design of the investigator-initiated clinical trial testing 1.5 million international units (MIU) aldesleukin (recombinant human IL-2) injected subcutaneously daily on five consecutive days with blood draws before and after aldesleukin injection (top). Corresponding gating strategy of human cDC subsets identifying CD141+ cDC1 and CD1c+ cDC2 (bottom). (G) Percentages of Ki67+ proliferating cDC1 (n = 8) and cDC2 (n = 10) in peripheral blood from patients before and after aldesleukin treatment. (H) Biological processes gene ontology (GO) terms enriched in RNA-seq data of IL-2cx–treated WT mice compared to untreated mice (n = 4 mice per group from one experiment). The numbers in parentheses refer to the count of significantly up-regulated genes within each GO term. (I) Top 75 up-regulated (top) and top 75 down-regulated (bottom) genes in IL-2cx–treated WT mice compared to untreated mice. P values calculated with Kruskall-Wallis test with Dunn’s multiple comparison test (B), Mann-Whitney test (D and E), or paired t test (G). APC, allophycocyanin; PE, phycoerythrin.

A detailed analysis of DC subsets showed that administration of IL-2cx expanded counts of all lymphoid-resident cDC subsets, including CD11blow XCR1+ cDC1s and CD11bhi XCR1 cDC2s, whereas pDCs remained unchanged compared to untreated mice (Fig. 1D and fig. S1F). These increased cell counts were accompanied by proliferation of all cDC subsets and pDCs, as evidenced by BrdU positivity (Fig. 1E and fig. S1G).

We also analyzed human CD11c+ MHC-II (HLA-DR)+ DCs within an investigator-initiated clinical trial (termed Charact-IL-2; NCT 03312335) using recombinant human IL-2 (aldesleukin) immunotherapy (Fig. 1F). We observed an increase in the proliferation of CD141+ cDC1s and CD1c+ cDC2s (Fig. 1G). Different clinical trials testing aldesleukin reported the expected proliferation of CD4+ and CD8+ T cells and NK cells upon aldesleukin immunotherapy (25, 26). However, by comparing Ki67+ DCs on day 0 (before) and day 5 (that is 1 day after the last injection) of a 5-day course of daily aldesleukin, we found an increase of proliferating cDC1s and cDC2s (Fig. 1G).

Mature cDCs, identified as CD11chi MHC-IIhi cells that are negative for colony-stimulating factor 1 receptor (also termed CD115), were purified from spleens of control versus IL-2cx–treated mice and submitted to RNA sequencing (RNA-seq). Gene ontology (GO) analysis of the transcriptome of mature cDCs showed overrepresentation of genes associated with cell proliferation and activation of the immune system, including the GO terms cell cycle, cell division, DNA replication, chromosome segregation, nucleosome assembly, protein localization to kinetochore, DNA recombination, chromosome condensation, kinetochore organization, and microtubule depolymerization (Fig. 1H). Other terms, such as response to virus, response to interferon-β, and response to interferon-α, indicated that cDCs were exposed to an inflammatory environment (Fig. 1H and fig. S1H). Analyzing the top 75 up- and 75 down-regulated genes in more detail, we found that the genes in cDCs from the IL-2cx–treated mice were enriched in GO term genes associated with the complement system (Cfb, C3, C1qc, and C4b), cytotoxicity (Gzmb and Gzma), and the integrin Itgae, which encodes CD103 and is associated with migratory cross-presenting DCs (Fig. 1I). Contrarily, Cd207, which encodes Langerin, and certain chemokine receptors and ligands (Ccl22, Cxcl12, and Ccr2) were down-regulated. CCL22 was suggested as important in DCs for interaction with CCR4+ Treg cells, thus implicating CCL22 in the control of excessive T cell responses (27).

Together, these results established pronounced stimulation of mouse and human cDCs after IL-2 immunotherapy in vivo. This stimulation of DCs resulted in proliferation of both cDC1s and cDC2s.

IL-2 causes proliferation of mature cDCs and differentiation of DC precursors

The observed proliferation and expansion of cDCs could be mediated by proliferation of mature cDCs or differentiation of hematopoietic DC precursors or both. To address this issue, we conducted a cell cycle analysis of mature splenic cDCs using propidium iodide staining. In line with published results stating 5% of lymphoid-resident cDCs are cycling (28), we detected 3.2 and 1.9% of splenic cDCs in S and G2/M phase, respectively (Fig. 2A). Proliferating cDCs from mice treated with IL-2cx increased to 10.1% in S and 6% in G2/M phase (Fig. 2A).

Fig. 2 IL-2 causes proliferation of mature cDCs and differentiation of DC precursors.

(A) Cell cycle analysis of splenic cDCs from IL-2cx–treated and untreated wild-type mice with the Watson model for cell cycle analysis for each cell cycle state including gap 0 (G0), gap 1 (G1), synthesis (S), gap 2 (G2), and mitosis (M). Data are presented as means ± SEM (n = 7 mice per group from N = 3 independent experiments). (B) Representative pseudocolor plots of bone marrow cells from untreated mice showing gating strategy for identifying monocyte/DC progenitor (MDP) cells, common dendritic cell progenitor (CDP) cells, precursor (pre)–cDC1, and pre-cDC2. (C) Total counts of precursor populations in bone marrow cells in mice receiving indicated treatments. Data are presented as means ± SEM (n = 11 mice per group from N = 3 independent experiments). (D) Total cell counts of splenic cDC1 and cDC2 from matched mice described in (C). Data are presented as means ± SEM (n = 11 mice per group from N = 3 independent experiments). (E) Activated caspase-3 in cDCs from untreated and IL-2cx–treated mice. Data are presented as means ± SEM (n = 4 to 5 mice per group from N = 2 independent experiments). (F) Survival of mature cDCs 48 hours after adoptive transfer of cDCs from both untreated and IL-2cx–treated donors into untreated recipient mice by intrasplenic injection. Data are presented as means ± SEM (n = 10 mice per group from N = 4 independent experiments). (G) Survival of mature cDCs 48 hours after adoptive transfer by intrasplenic injection of cDCs from untreated donor into either untreated or IL-2cx–pretreated recipient mice. Data are presented as means ± SEM (n = 8 to 9 mice per group from N = 2 independent experiments). P values were calculated using multiple t tests (A and D), Mann-Whitney test (C), or unpaired t test (E to G).

To evaluate cDC differentiation from bone marrow precursors, we hierarchically assessed the differentiation steps of DCs, starting from monocyte/DC progenitors (MDPs), followed by common DC progenitors (CDPs) and the clonogenic progenitors precursor (pre)–cDC1 and pre-cDC2 (Fig. 2B) (2, 6, 2933). IL-2 immunotherapy in mice resulted in a decrease of MDPs, CDPs, and pre-cDC2 cells in the bone marrow (Fig. 2C), with simultaneous increase of splenic cDC1 and cDC2 (Fig. 2D). Bone marrow pre-cDC1 cells did not change, possibly due to compensatory proliferation (Fig. 2C). Because CD135 is used to identify DC precursors, the observed decrease in bone marrow counts could be partly affected by down-regulation of CD135 upon engagement with its ligand Feline McDonough Sarcoma (FMS)–like tyrosine kinase 3 ligand (FLT3L).

Investigating the possibility that IL-2 treatment affects survival of cDCs, we measured intracellular caspase-3, which induces apoptosis downstream of the intrinsic pathway regulated by the B cell leukemia 2 (BCL-2) protein family and the extrinsic pathway mediated by death receptors activated at the cell surface (34). However, we did not observe any significant difference in caspase-3–positive splenic cDCs between untreated and IL-2cx–treated mice (P = 0.6794; Fig. 2E). Staining for live cells versus dead cells with Annexin V and 7-AAD confirmed these results (fig. S2A). In addition, in the intrinsic apoptosis pathway, neither the amount of BCL-2, which is the inhibitor of the pro-apoptotic BCL-2–associated X (BAX) protein (35), nor the amount of BAX was different between cDCs from untreated and IL-2cx–treated mice (fig. S2, B and C). These results suggested no survival benefit after IL-2 treatment, yet they were measured indirectly by apoptosis-regulating proteins. Direct measurement of survival by adoptive transfer of cDCs from untreated and IL-2cx–treated mice into untreated recipients revealed a small, but significant, survival advantage of the IL-2cx–treated cDCs (P = 0.0015; Fig. 2F). Because of the increased differentiation of DC progenitors, this effect could be attributed to the inclusion of “younger” cDC subsets in the transferred cDCs from IL-2cx–treated mice. To exclude this possibility, we reversed the experimental setup and transferred cDCs from untreated mice to either untreated or IL-2cx–pretreated mice, observing a small but insignificant increase in the IL-2cx–pretreated mice (P = 0.3148; Fig. 2G).

Together, these results suggested that IL-2 treatment facilitates proliferation of mature cDCs and accelerated differentiation of DC progenitors to mature DCs. To a lesser extent, IL-2 treatment might enhance DC survival.

Mature DCs and their precursors lack functional IL-2Rs

Given the profound stimulatory effects of IL-2 on cDCs, we investigated IL-2R abundance on cDCs and their precursors at the mRNA and protein levels. Quantitative polymerase chain reaction (qPCR) of Il2r transcripts from untreated and IL-2cx–treated cDCs, CD4+ Foxp3 conventional T (Tcon), and CD4+ Foxp3+ Treg cells revealed that cDCs contained Il2rg mRNA, but they lacked Il2ra or Il2rb mRNA (Fig. 3A). In comparison, Tcon and Treg cells expressed mRNA for all three IL-2R subunits (Fig. 3A). We also confirmed these results at the protein level by surface staining for IL-2R subunits on mature mouse cDCs and Treg cells in untreated and IL-2cx–treated mice (Fig. 3B). As previously published (36, 37), cDCs up-regulated CD25, but not CD122, upon stimulation with two different Toll-like receptor (TLR) ligands (fig. S3A). We investigated IL-2R abundance on DC precursors. MDP and CDP lacked detectable amounts of CD132; pre-cDC1 had a slightly detectable amount of CD132 and pre-cDC2 had clearly detectable CD132 (Fig. 3C). MDP, CDP, pre-cDC1, and pre-cDC2 all lacked detectable surface CD122 and CD25, even after IL-2cx treatment (Fig. 3C and fig. S3B).

Fig. 3 Mature DCs and their precursors lack functional IL-2Rs.

(A) Relative mRNA expression of Il2ra (encoding CD25), Il2rb (CD122), and Il2rg (CD132) in cDCs from untreated (black dots) and IL-2cx–treated (red dots) mice compared to the expression in conventional T (Tcon) and regulatory T (Treg) cells from the same mice (n = 2 mice per group and treatment condition). Control conditions included splenocytes isolated from untreated wild-type (WT), Il2ra−/−, Il2rb−/−, and Il2rg−/− mice (n = 1 mouse per group). Data are presented as means ± SEM (N = 1 independent experiment with 4 technical replicates per mouse). (B) IL-2R subunit abundance on the surface of mouse splenic cDCs (left) and Treg cells (right) by flow cytometry. Displayed are representative histograms from n = 9 mice and N = 3 independent experiments showing the intensity of IL-2R subunit staining in untreated (blue lines) and IL-2cx–treated (red lines) mice and the fluorescence minus one (FMO) control (gray shaded areas). (C) IL-2R subunit abundance on the surface of DC precursors in bone marrow. MDP, monocyte/DC progenitor; CDP, common DC progenitor; pre-cDC1, precursor cDC1; pre-cDC2, precursor cDC2. Shown are representative histograms from n = 9 mice and N = 3 independent experiments showing the intensity of IL-2R subunit staining (red lines) and FMO control (gray shaded areas). (D) Phosphorylation of signal transducer and activator of transcription 5 (STAT5) in mouse cDCs (red lines) and Treg cells (black lines) upon incubation with titrated concentrations of IL-2 (top) or mouse CSF2 (mCSF2; bottom). Data are presented as means ± SD (n = 2) from one of three independent experiments. (E) IL-2R subunit abundance on human CD11c+ HLA-DR+ cDCs compared to the abundance on CD4+ T cells. Shown are representative histograms from n = 6 individual donors and N = 3 independent experiments of the respective intensity of IL-2R subunit staining (red lines) and FMO control (gray shaded areas). (F) Evaluation of cytokine signaling by measurement of phosphorylated STAT5 (pSTAT5) in human cDCs and CD4+ T cells after incubation with either IL-2 (red lines; 1000 IU/ml) or human CSF2 (hCSF2; blue lines; 200 ng/ml). Shown are representative histograms from n = 6 individual donors and N = 3 independent experiments of the respective intensity of pSTAT5 staining and FMO control (gray shaded areas). (G) Lethally irradiated Il2rg−/− mice reconstituted with a 1:1-mix of CD45.1+ WT and CD45.2+ Il2rg−/− bone marrow cells receiving indicated treatments 2 months after reconstitution. Data are presented as means ± SEM (n = 5 to 9 mice per group from N = 3 independent experiments). BM, bone marrow. P values were calculated using unpaired t test.

To directly assess whether mature cDCs had functional IL-2Rs, we analyzed downstream intracellular signaling by signal transducer and activator of transcription 5 (STAT5), which becomes phosphorylated (pSTAT5) upon activation of a functional IL-2R by IL-2. As a positive control for STAT5 signaling, we also tested the cDCs response to CSF2 [also termed granulocyte-macrophage CSF (GM-CSF)], which also activates STAT5. Confirming the lack of functional IL-2Rs, even the highest concentrations of IL-2 tested failed to induce pSTAT5 in cDCs, whereas Treg cells readily responded to the same range of IL-2 concentrations in a dose-dependent manner (Fig. 3D). Conversely, pSTAT5 increased in cDCs incubated with mouse CSF2, whereas Treg cells remained unaffected by this stimulation (Fig. 3D). The same results applied to cDCs from IL-2cx–treated mice (fig. S3C). Furthermore, Lin CD135+ DC precursors, including MDPs, CDPs, pre-cDC1, and pre-cDC2, from untreated and IL-2cx–treated mice did not show increased pSTAT5 upon stimulation with a high dose (1000 IU) of IL-2, whereas incubation with mouse CSF2 increased pSTAT5 (fig. S3D). Thus, DC precursors do not have functional IL-2Rs.

Similar to their mouse counterparts, human cDCs were positive for CD132 but lacked detectable CD25 and CD122 at the cell surface, unlike CD4+ T cells that served as controls (Fig. 3E). Accordingly, stimulation of human cDCs with a high IL-2 concentration did not increase pSTAT5, whereas the same treatment induced pSTAT5 in human CD4+ T cells, and human CSF2 increased pSTAT5 in human cDCs (Fig. 3F).

Supporting these findings, IL-2cx treatment of bone marrow chimeric mice carrying a 1:1-mix of CD45.1-congenic wild-type and CD45.2-congenic Il2rg−/− DCs resulted in equal expansion of both wild-type and Il2rg−/− cDCs (Fig. 3G). Collectively, these data support a model whereby IL-2 immunotherapy indirectly activates cDCs and their precursors through the induction of intermediary factors.

IL-2 immunotherapy causes production of several DC mitogens

Searching for candidates that stimulate cDCs and their precursors, we decided to assess FLT3L and CSF2, because these cytokines contribute to the development and survival of DCs during steady state and inflammation, respectively (38). Untreated Flt3l−/− mice had reduced cDC counts compared to wild-type mice, amounting to ~1 × 106 per spleen in wild type (Fig. 1B) versus ~2.5 × 104 per spleen in Flt3l−/− (Fig. 4A), thus confirming the importance of FLT3L for cDC homeostasis. However, similar to wild-type mice (Fig. 1B), IL-2 treatment caused a fourfold expansion of cDCs in Flt3l−/− animals (Fig. 4A). Likewise, use of IL-2cx in Csf2−/− mice, which contained reduced steady-state numbers of cDCs compared to those in wild type, resulted in an increase of splenic cDC counts (Fig. 4B). We predicted that FLT3L and CSF2 compensated for the absence of each other, but unexpectedly, Flt3l−/− Csf2−/− double-knockout mice also showed expansion and proliferation of cDCs and corresponding subsets upon IL-2 treatment (Fig. 4, C and D, and fig. S4A).

Fig. 4 IL-2 immunotherapy induces production of several DC mitogens.

(A to C) Total splenic cDC counts from Flt3l−/− (n = 9 to 11 mice from N = 4 independent experiments) (A), Csf2−/− (n = 4 to 6 mice from N = 2 independent experiments) (B), or Flt3l−/− Csf2−/− (n = 20 mice per group from N = 5 independent experiments) (C) mice treated with IL-2cx. Data are presented as means ± SEM. (D) Proliferation of cDCs in Flt3l−/− Csf2−/− mice shown by BrdU incorporation over 3 days. Data are presented as means ± SEM (n = 14 mice from N = 3 independent experiments). Gate indicates region quantified. (E and F) Total splenic cDC counts in Flt3l−/− Csf2−/− (n = 5 to 6 mice per group from N = 3 independent experiments) (E) or WT (n = 9 mice per group from N = 3 independent experiments) (F) mice receiving indicated treatments. Data are presented as means ± SEM. (G) Expansion of WT and Tnfrsf1a/b−/− cDCs after IL-2cx treatment of mice previously irradiated lethally and reconstituted with a 1:1 mix of bone marrow cells from the indicated donor mice. Data are presented as means ± SEM (n = 12 mice per group from N = 4 independent experiments). (H) Measurement of mouse FLT3L, CSF2, and TNF in serum using enzyme-linked immunosorbent assay (ELISA). Data are presented as means ± SEM (n = 6 to 11 mice per group from N = 3 independent experiments). (I) Measurements of human FLT3L and TNF using proximity extension assay and CSF2 using ELISA in sera of patients before and after receiving aldesleukin as described in Fig. 1F (n = 10 to 12 individual patients). P values were calculated using Mann-Whitney test (A), unpaired t tests (B, C, G, and H), repeated measure one-way ANOVA with Holm-Sidak correction for multiple comparison (E), mixed-effects model with Holm-Sidak correction for multiple comparison (F), or paired t test (I). n.d., not detectable.

In previous experiments, we observed that IL-2cx treatment increased tumor necrosis factor (TNF) in a T cell–dependent manner (39). We thus hypothesized that TNF stimulates cDC expansion in Flt3l−/− Csf2−/− mice. Blocking TNF with the soluble TNF receptor 2-Fc-IgG1 fusion protein etanercept reduced cDC expansion to close to background amounts in IL-2cx–treated animals (Fig. 4E). Etanercept treatment of wild-type mice also reduced IL-2cx–mediated expansion and proliferation of cDCs, suggesting a FLT3L- and CSF2-independent effect of TNF on DC homeostasis (Fig. 4F and fig. S4B). Moreover, IL-2cx treatment of bone marrow chimeric mice containing a 1:1-mix of wild-type and Tnfrsf1a/b−/− bone marrow showed reduced expansion of Tnfrsf1a/b−/− cDCs, thus indicating a direct effect of TNF on cDCs (Fig. 4G).

Measurement of serum cytokines in mice showed marked increase in FLT3L and TNF upon IL-2 immunotherapy, whereas CSF2 was not detectable (Fig. 4H). The latter might be due to local action of CSF2 rather than systemic secretion (38). We found similar results for these cytokines in human serum before and after aldesleukin treatment (Fig. 4I). In summary, these results indicated that IL-2 mediates cDC expansion through secretion of the cytokines FLT3L, CSF2, and TNF, all of which directly stimulate cDCs (referred to hereafter as DC-active cytokines).

IL-2–stimulated innate and adaptive lymphoid cells produce DC-active cytokines

Our results indicated that cDCs were expanded by cytokines secreted by IL-2–responsive cells. Because both hematopoietic cells and nonhematopoietic cells have IL-2Rs (23), we investigated whether cDC-stimulating cells are of hematopoietic origin. To this end, we generated bone marrow chimeric mice in which wild-type or Il2rg−/− recipients were lethally irradiated (950 RAD) and reconstituted with wild-type or Il2rg−/− bone marrow (Fig. 5A, left). Only mice reconstituted with wild-type bone marrow, but not those with Il2rg−/− bone marrow, showed expansion of splenic cDCs after IL-2 immunotherapy, indicating that IL-2–responsive cells are of hematopoietic origin (Fig. 5A).

Fig. 5 IL-2–stimulated innate and adaptive lymphoid cells produce DC-active cytokines.

(A) Mice with chimeric bone marrow (BM) were generated by adoptive transfer (AT) of either Il2rg−/−, CD45.1 wild-type (WT), or CD45.2 WT lineage-depleted BM cells into lethally irradiated Il2rg−/− or CD45.1 hosts. Two months later, mice were treated as indicated. Data are presented as means ± SEM (n = 4 to 11 mice per group from N = 2 to 4 independent experiments). (B) Ki67+ proliferating splenic innate lymphoid cells (ILCs, left) and total ILC counts (right) upon IL-2cx treatment. ILCs were defined as Lin (CD3, CD5, CD8, CD11b, CD11c, CD19, B220, Ter119) CD127+ CD90+; with ILC1 defined as TBET+, ILC2 as GATA3+, and ILC3 as RORγt+ (fig. S5E). Data are presented as means ± SEM (n = 8 to 14 mice per group from N = 3 to 5 independent experiments). (C) Expansion of splenic cDCs in Tcrbd−/− mice after treatment with IL-2cx. Data are presented as means ± SEM (n = 10 to 13 mice per group from N = 4 independent experiments). (D) Expansion of splenic cDCs upon IL-2cx treatment in Rag−/− mice additionally depleted of NK cells, ILCs, or both. NK cells were depleted by injection of an anti-NK1.1 mAb; ILCs were depleted by injection of an anti-CD90.2 mAb. Data are presented as means ± SEM (n = 6 to 9 mice per group from N = 3 to 4 independent experiments). (E to G) Diagram of experimental setup: ILC precursors or control common lymphoid progenitor (CLP) cells were purified with fluorescence-activated cell sorting (FACS) before adoptive transfer into sublethally irradiated Il2rg−/− or Il2rg−/− Rag2−/− recipient mice (E). Two months later, immune cell populations reconstituted from ILC precursors (AT ILC, top) and CLP (AT CLP, bottom) were analyzed to verify successful engraftment (F). Mice reconstituted with ILC precursors were subsequently treated as indicated, and splenic cDCs were quantified (G). Data are presented as means ± SEM (n = 6 to 7 mice per group from N = 3 independent experiments). (H) Flow cytometry–based RNA assay showing expression of Flt3l, Csf2, and Tnf mRNA geometric mean fluorescence intensity (gMFI) in CD4+ regulatory T (Treg), CD4+ conventional T (Tcon), CD8+ T, natural killer (NK) cells, ILC1, ILC2, and ILC3. Data are presented as means ± SEM (n = 6 mice per group from N = 2 independent experiments). (I and J) Counts of indicated immune cell subsets producing CSF2 (upper graphs) or TNF (lower graphs) in mice treated with IL-2cx or left untreated (n = 8 to 10 mice per group from N = 2 to 3 independent experiments) (I) and ILCs (n = 6 mice per group from N = 3 independent experiments) (J). Data are presented as mean values ± SEM. Statistical significance was tested with Mann-Whitney test (A), multiple t tests (B, H, and I), unpaired t test (C, G, and J), or one-way ANOVA with Holm-Sidak correction for multiple comparison (D).

To evaluate potential target cells, we searched the ImmGen database (40) to identify mouse immune cells with functional IL-2Rs, comprising at least those with both CD122 and CD132. By examining Il2ra, Ilr2b, and Il2rg expression, our search revealed that, in addition to T, NK, and NKT cells, ILCs had detectable amounts of each of these transcripts (fig. S5A). As previously published (2224, 39, 41, 42), we confirmed in vivo expansion and proliferation of CD4+ T, CD8+ T, and NK cells upon IL-2cx treatment (fig. S5, B to D). Furthermore, all ILC subsets, including type 1 ILCs (ILC1), type 2 (ILC2), and type 3 (ILC3), showed proliferation and expansion upon IL-2 immunotherapy (Fig. 5B and fig. S5E).

To assess the contribution of each IL-2–responsive cell subset to cDC homeostasis and IL-2–mediated cDC expansion, we used different knockout and antibody-depletion strategies. Although Tcrbd−/− mice, lacking both αβ and γδ T cells, showed reduced counts of splenic cDCs compared to wild type, amounting to ~1 × 106 per spleen in wild type (Fig. 1B) versus ~3 × 105 per spleen in Tcrbd−/− (Fig. 5C), IL-2cx treatment stimulated expansion of cDCs in the Tcrbd−/− animals (Fig. 5C). We used Rag1−/− mice, lacking T and B cells, and depleted the mice of NK cells, ILCs, or both by injection of anti-NK1.1 or anti-Thy1.2 antibodies or both antibodies. Both cell types contributed to IL-2–mediated cDC expansion (Fig. 5D). The incomplete inhibition of cDC expansion in Rag1−/− mice depleted of NK cells and ILCs is likely due to tissue residency of ILCs, which are incompletely amenable to antibody-mediated depletion (43).

To further study the contribution of ILCs to IL-2–induced cDC proliferation, we reconstituted sublethally irradiated (450 RAD) Il2rg−/− and Il2rg−/− Rag2−/− mice with ILC2 precursor (ILC2p) cells (Fig. 5E and fig. S5F) (4345). Two months later, ILC2p cells had developed to all ILC subsets but not into B, T, and NK cells (Fig. 5F, top). Although sorted ILC2p cells were used for the reconstitution of Il2rg−/− Rag2−/− mice, we detected not only ILC2s but also RORγt+ ILC3s and RORγt GATA3 ILC1s (Fig. 5F, top). Control mice reconstituted with common lymphoid progenitor (CLP) cells harbored all ILC subsets, as well as B, T, and NK cells (Fig. 5F, bottom). Subsequent treatment with IL-2cx induced the expansion of splenic cDCs in ILC2p-reconstituted mice (Fig. 5G).

To investigate whether lymphoid cells produced cDC-active cytokines, we measured intracellular RNA transcripts of FLT3L, CSF2, and TNF with flow cytometry, which showed that T, NK cells, and ILCs produced Flt3l at steady state (Fig. 5H). IL-2 treatment of mice further stimulated production in CD4+ Treg, CD8+ T, and NK cells, but not in CD4+ Tcon cells and ILCs (Fig. 5H). Under steady-state conditions, low amounts of Csf2 transcripts were detected, and IL-2cx treatment of mice induced expression in NK cells, ILC1, and ILC2, but not in T cells and ILC3 (Fig. 5H). IL-2 treatment of mice increased intracellular Tnf in all subsets (Fig. 5H).

Intracellular assessment of CSF2 and TNF upon in vitro stimulation confirmed expansion of CSF2+ and TNF+ cells in all investigated lymphocyte subsets and ILCs, except for TNF+ CD4+ Tcon cells, which did not expand (Fig. 5, I and J). Together, these results demonstrated that IL-2 stimulates innate and adaptive lymphoid cells to produce DC-active cytokines, which then stimulate the expansion of cDCs.

IL-2 immunotherapy–activated cDCs facilitate antitumor responses in mouse and human

We investigated the effects of IL-2 treatment on the function of cDCs. Treatment of mice with IL-2cx induced up-regulation of CD40, CD80, CD86, and MHC-I, but not MHC-II, on cDCs (Fig. 6A). These changes are indicative of mature cDCs with increased potential of cross-presentation and costimulation for T cell activation (10, 46, 47). RNA-seq data confirmed that cDCs from IL-2–treated animals showed up-regulation of genes associated with antigen processing and presentation (fig. S6A).

Fig. 6 IL-2 immunotherapy–activated cDCs facilitate antitumor responses in mouse and human.

(A) Abundance of CD40, CD80, CD86, MHC-I, and MHC-II on splenic cDCs of untreated and IL-2cx–treated mice displayed as representative histograms (left) and fold change of gMFI normalized to untreated (right). Data are presented as means ± SEM (n = 9 mice per group from N = 4 independent experiments). (B) Measurement of antigen uptake and processing by cDCs isolated from untreated and IL-2cx–treated mice after 6, 24, 48, and 72 hours. Data are presented as means ± SEM (n = 9 mice per group from N = 5 independent experiments). (C) Diagram of experimental setup (left): Ovalbumin peptide–loaded, purified cDCs were injected into the footpad of naïve mice; after 24 hours, CFSE-labeled OT-1 cells were adoptively transferred. Two to 3 days later, popliteal lymph nodes were removed and OT-1 cell proliferation was quantified (right). Data are presented as means ± SEM (n = 11 to 12 mice per group from N = 5 independent experiments). (D) Tumor growth kinetics (left) and quantification of tumor-infiltrating cDCs (day 11) (middle and right) in B16-F10 melanoma–bearing mice treated with IL-2cx or anti–PD-1 or left untreated. Treatment was initiated when tumors were visible and palpable and applied every other day (see also fig. S6B, top). Data are presented as means ± SEM (n = 9 mice per group from N = 3 independent experiments). For tumor volume data, P value for IL-2cx compared to untreated is indicated. (E) Correlation of cDC1 and cDC2 with tumor volume of tumor-bearing mice on day 11 from same samples as in (D). (F) Diagram of experimental paradigm (left). Right: Quantification of tumor-infiltrating cDCs in BrafCA PtenloxP Tyr::CreERT2 mice harboring 4-hydroxytamoxifen–induced tumors. Treatment was started once tumors were visible and palpable. Data are presented as means ± SEM (n = 12 to 14 mice per group from N = 2 to 3 independent experiments). (G and H) Tumor growth kinetics in Cd11c-DTR (n = 9 to 10 mice per group from three independent experiments) (G) and Zbtb46-DTR (n = 9 to 10 mice per group from three independent experiments) (H) BM chimeric mice harboring B16-F10 melanoma. Treatment with IL-2cx and anti–PD-1 was initiated 4 days after tumor cell injection when tumors were visible and palpable and applied three times weekly for the duration of the experiment. To deplete DCs, diphtheria toxin (DT) was applied three times weekly starting 1 day before tumor inoculation and for the duration of the experiment (see also fig. S6B, bottom). Data are presented as means ± SEM. (I) Survival analysis of patients with cutaneous melanoma according to IL-2 signature and BATF3 expression (cutoff at 50%). (J to L) Correlation of BATF3 with IL-2 signature (J), with CD4 and CD8B expression, with NK cell signature (K), and with FLT3LG-CSF2-TNF expression signature (L). Data were retrieved from TCGA (n = 473 individual patients). Statistical significance was tested with Mann-Whitney test (A), two-way ANOVA with Holm-Sidak correction for multiple comparison (B and D, left), unpaired t test (C), one-way ANOVA with Holm-Sidak correction for multiple comparison (D, middle graph cDC1), Kruskall-Wallis test with Dunn’s multiple comparison test (D, right graph cDC2, and F), Pearson correlation coefficients (E and J to L), Mixed-effects model with Holm-Sidak correction for multiple comparison (G to H), and log-rank test (I).

To measure antigen uptake and processing, we used DQ-ovalbumin in which fluorescence is an indirect measure of lysosomal degradation of ovalbumin in cDCs. Compared with cDCs from untreated mice, cDCs from IL-2cx–treated mice showed greater antigen uptake and processing after 24, 48, and 72 hours of in vitro culture (Fig. 6B).

On the basis of these data, we hypothesized that cDCs from IL-2–treated mice have greater capacity to activate T cells than those from untreated mice. We isolated cDCs from untreated or IL-2cx–treated mice, incubated these cells with a 24–amino acid ovalbumin peptide (DEVSGLEQLESIINFEKLAAAAAK) with improved cross-presentation properties (48), and adoptively transferred these cDCs into naïve mice by footpad injection. After 24 hours, carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled antigen-specific OT-1 CD8+ T cells were adoptively transferred into the mice, and the proliferation of these T cells was assessed 2 to 3 days later (Fig. 6C). The results confirmed our hypothesis: cDCs from IL-2cx–treated animals induced significantly more expansion of OT-1 CD8+ T cells (P = 0.0206; Fig. 6C).

To investigate whether IL-2 immunotherapy expanded tumor-infiltrating cDCs, we quantified cDCs in B16-F10 melanoma–bearing mice receiving IL-2cx (fig. S6B). Tumor-infiltrating CD103+ cDC1s and cDC2s were increased on day 11 after tumor implantation, and tumor growth was delayed in IL-2cx–treated compared to growth in untreated mice (Fig. 6D). On the contrary, anti-programmed cell death 1 (PD-1) antibody treatment neither increased tumor-infiltrating CD103+ cDC1 or cDC2 nor delayed tumor growth compared to untreated animals (Fig. 6D and fig. S6C). The failure of anti–PD-1 monotherapy to reduce growth of B16-F10 melanoma is consistent with observations of other groups (49, 50). Further analyzing these data, we observed a negative linear correlation between tumor-infiltrating cDC1 and tumor volume (Fig. 6E). We did not observe a significant correlation for tumor-infiltrating cDC2 (P = 0.0872; Fig. 6E). We observed increased numbers of tumor-infiltrating cDC1 and cDC2 in the inducible melanoma model BrafCA PtenloxP Tyr::CreERT2 (Fig. 6F).

To clarify the role of cDCs in orchestrating IL-2–mediated antitumor responses, we evaluated B16-F10 melanoma growth in two different DC depletion models: the Cd11c-DTR model, which depletes all CD11c+ cells upon administration of diphtheria toxin (DT), and the Zbtb46-DTR model, which specifically depletes cDCs (3). In bone marrow chimeras generated using these transgenic mice, IL-2cx–mediated tumor control was lost when DCs were continuously depleted beginning 1 day before tumor implantation (Fig. 6, G and H, and fig. S6, D and E). DC-depleted mice not receiving immunotherapy did not show accelerated tumor growth compared with mice that were not DC depleted, likely due to the aggressive growth of B16-F10 melanoma. Similar to the other B16-F10 results (Fig. 6D), anti–PD-1 treatment did not delay tumor growth.

Analyzing The Cancer Genome Atlas (TCGA) data of human skin cutaneous melanoma revealed prolonged survival of patients with tumors with a high IL-2 signature, consisting of IL2 and 10 IL-2–induced genes (50% threshold; Fig. 6I) (51). We observed a similar trend for patients with BATF3high tumors (Fig. 6I). The hazard ratio for tumors with high expression of the IL-2 signature was even lower than that for BATF3high tumors (hazard ratio 0.57 and 0.68, respectively), suggesting that the IL-2 signature has a better predictive value. Further data mining uncovered a positive relationship between the IL-2 signature and BATF3 (Fig. 6J). Other correlations were found between BATF3 and CD4, CD8B, and an NK cell signature (Fig. 6K). BATF3 also correlated with the CSF2-FLT3L-TNF cytokine signature and the individual cytokines CSF2, FLT3L, and TNF (Fig. 6L). More detailed analysis revealed a positive correlation between IL2 and BATF3 as well as between each IL2R subunit transcript and BATF3 (fig. S6F).

Together, these data showed that IL-2 immunotherapy promotes activation and expansion of tumor-infiltrating cDCs, which enhances antigen presentation and CD8+ T cell activation. These effects correlated with improved antitumor responses in mice and humans.


Previous seminal studies established a crucial role of intratumoral CD103+ cDCs in stimulating CD8+ T cells for efficient antitumor responses (811). These cells depend on FLT3L, CSF2, BATF3, and IRF8. However, these cells are extremely rare in the TME. Our work identified an IL-2–lymphoid cell–cDC pathway that robustly expands and stimulates this BATF3+ IRF8+ CD103+ cDC1 subset. This pathway depended on production of FLT3L, CSF2, and TNF by several ILC subsets, NK cells, and T cells. These DC-active cytokines induced differentiation of DC precursors, as well as expansion and activation of mature cDCs. Thus, our findings extend work showing that NK cells attract DCs into the TME by secreting XCL1 and CCL5 (15) and serve as the main producers of FLT3L in mouse melanoma (16). Moreover, we showed that the IL-2–lymphoid cell–cDC pathway stimulates and expands cDC2, which have been implicated in priming antitumor CD4+ T cell responses (46). Although FLT3L and CSF2 play critical roles in cDC expansion (28, 38), TNF induces cDC maturation that is critical for efficient antiviral responses (52). Our findings suggested a role for TNF in cDC differentiation in vivo, supporting previous studies showing such an effect in vitro where TNF skews differentiation of precursors from monocytes to DCs by inhibiting the IL-6/CSF1 pathway (53).

Some previous studies suggested that DCs have functional IL-2R and that IL-2 has a direct effect on DC development in vitro (54, 55); other studies provided somewhat contradictory data based on the absence of CD122 on DCs (5658). Mouse (59) and human (60) DCs can have detectable CD25, especially upon stimulation with TLR ligands (36, 37) or after culture with CSF2 (56). Functionally, up-regulation of CD25 on activated DCs could facilitate trans-presentation of IL-2 to T cells early during T cell activation (61, 62). Concerning CD122, most studies suggest that directly isolated or CSF2-cultured mouse and human DCs do not have detectable CD122 (5658). Nevertheless, two carefully conducted in vitro studies suggested an inhibitory role of IL-2 on FLT3L-mediated DC differentiation from mouse bone marrow (54, 55). The discrepancy between the observed inhibitory effect of IL-2 and our study might be explained by the different conditions, cytokines, and other growth factors in vitro versus in vivo. Here, neither directly isolated DCs nor DC precursors had CD122, even after in vivo IL-2 or TLR ligand treatment. Furthermore, we could not detect activation of the IL-2 signaling pathway after incubation of DCs and DC precursors with high concentrations of IL-2, which, to our knowledge, has not been shown in previous studies claiming functional IL-2Rs on DCs. Thus, DCs and DC precursors do not have functional IL-2Rs. Last, Il2rg−/− cDCs expanded equally to Il2rg+/+ cDCs in bone marrow chimeric mice upon IL-2 treatment, strongly supporting our model of indirect DC expansion through secondary secretion of DC-active cytokines upon IL-2 treatment in vivo.

Our work thus uncovers an aspect of IL-2–mediated antitumor responses beyond the direct effects on IL-2–responsive effector cells. One could speculate that IL-2 treatment might turn poorly immune infiltrated, that is, “cold,” tumors into highly infiltrated, “hot,” tumors, which are responsive to immune checkpoint inhibitors. In line with this suggestion, IL-2 immunotherapy in melanoma-bearing mice increases intratumoral TNF and CCL5 production and PD-1 ligand (PD-L1) abundance on cancer cells, thus rendering the tumor hot (39). Moreover, in patients, treatment with the PEGylated IL-2 molecule bempegaldesleukin (also known as NKTR-214) increased PD-L1 on previously PD-L1–negative urothelial cancer biopsies after 3 weeks of treatment, indicating the conversion of a cold to a hot TME (63).

Furthermore, our data could explain why IL-2 immunotherapy has, in some patients, resulted in remarkably long-term antitumor responses lasting 20 years and more (64). This finding could indicate the priming and programming of long-lived memory CD8+ and CD4+ T cells during IL-2 immunotherapy by direct IL-2 signals (65) and restimulation of CD8+ and CD4+ T cells by IL-2–mediated expansion of intratumoral cDC1 and cDC2. On the basis of this suggestion, we hypothesize that the herein-described adaptive cDC poiesis could affect the generation and development of long-lived memory CD8+ T cells. These suggestions might also apply to bacterial and viral infections. Thus, early cDCs presenting high loads of antigens could prime T cells to preferentially adopt a short-lived effector fate, whereas a second wave of cDCs due to the IL-2–lymphoid cell–cDC pathway could result in cDCs carrying lower concentrations of antigenic peptides, thereby inducing memory T cells.

Future studies should address the following limitations of our study. Our patient tumor results are derived from the TCGA database and reflect correlations between different gene signatures. It would be interesting to extend our investigations to melanoma patient samples from tumors and secondary lymphoid tissues, ideally in a paired manner before and after high-dose aldesleukin treatment. However, these samples are very difficult to obtain as aldesleukin has rarely been used for cancer immunotherapy to date. Furthermore, our investigations have focused on mouse and human melanoma. Assessment of other cancer types after IL-2 immunotherapy could reveal a similar pathway that results in DC expansion. Another topic concerns depletion of ILCs. Using monoclonal antibodies (mAbs) targeting NK cells and ILCs only partially abrogated IL-2cx–mediated DC expansion. This could be due to incomplete depletion of tissue-resident ILCs, which should be assessed in future studies using transgenic mouse models allowing specific and complete depletion of selective ILC subsets.

In summary, our work further sheds light on an IL-2 immunotherapy–mediated IL-2–lymphoid cell–cDC pathway (fig. S7). Our findings support the investigation of potential combinatorial approaches in which IL-2 treatment could render poorly immunogenic cancers amenable to treatment with immune checkpoint inhibitors.


Study design

The aim of this study was to assess the mechanism of IL-2–mediated DC expansion in mice and humans. To this end, we used different mouse models, blood samples collected from IL-2–treated patients, and publicly available datasets. Sample size was chosen empirically based on results of previous studies. Reporting of mouse studies followed the Animal Research: Reporting of In Vivo Experiments guidelines. In general, experiments aimed to include in total 10 mice per group and were repeated three times with precise numbers for each individual experiment provided in the figure legends. Mice were randomly assigned to different treatment groups and stratified according to gender and age. Investigators were not blinded. No data outliers were excluded. Primary data are reported in data file S1.

Clinical trial and human samples

Human samples were collected within the clinical trial “Open-label, Monocentric, Phase II, Investigator-initiated Clinical Trial on Unbiased Characterization of Immunological Parameters in Interleukin-2-treated Systemic Lupus Erythematosus” (Charact-IL-2, ClinicalTrials identifier: NCT03312335) and the “Fundamental research project for phenotypical and functional characterization of different leukocyte subsets in healthy and diseased individuals” (PFCL-1, BASEC no. 2016-01440). Both projects have been reviewed and approved by the competent Swiss authorities and have been carried out in accordance with principles enunciated in the current version of the Declaration of Helsinki, the guidelines of Good Clinical Practice, and Swiss legal requirements. Before enrollment into the clinical trial or sample collection, written informed consent was obtained. Human blood was collected into EDTA vacutainer tubes (BD Biosciences) followed by Ficoll-Paque PLUS (GE Healthcare) gradient centrifugation for peripheral blood mononuclear cell (PBMC) isolation. Isolated PBMCs were frozen in fetal calf serum (FCS; Gibco) containing 10% dimethyl sulfoxide (Sigma-Aldrich) and stored for less than 1 year in liquid nitrogen before analysis. Serum was isolated from blood collected with clot activator vacutainer tubes (BD Biosciences) and stored for less than 18 months at −80°C before analysis.

For evaluation of IL-2–mediated expansion of cDCs and lymphocytes, blood from patients with systemic lupus erythematosus was collected prior and after a 5-day course of daily 1.5 million international units (MIU) of aldesleukin (Proleukin, Novartis Pharma), according to the study protocol.


C57Bl/6J mice were purchased from Charles River Laboratories. Tcrbd−/− (JAX stock no. 002122), Rag1−/− (JAX stock no. 002216), Csf2−/−, Flt3l−/−, CD45.1+ (JAX stock no. 002014), Il2ra−/− (JAX stock no. 002462), Il2rb−/− (JAX stock no. 002816), Il2rg−/− (JAX stock no. 003174), Rag2−/− Il2rg−/− (Taconic stock no. 4111), OT-1 (JAX stock no. 003831), Cd11c-DTR (JAX stock no. 004509), Zbtb46-DTR (JAX stock no. 019506), Tnfrsf1a/b−/− (JAX stock no. 003243), Foxp3-GFP-DTR (JAX stock no. 016958), and BrafCA PtenloxP Tyr::CreERT2 (JAX stock no. 013590) were obtained from the Jackson laboratory (JAX) or from the Swiss Immunology Mouse Repository. Csf2−/− (JAX stock no. 026812) and Flt3l−/− (MMRRRC stock no. 37395-JAX) mice were provided by M. Manz (University Hospital Zurich) and A. Rolink (University of Basel), respectively, and crossed to generate Csf2−/− Flt3l−/− mice. Female and male mice were used for experiments at 2 to 5 months of age. The experiments were conducted in a randomized fashion by unblinded investigators. Mice were held in a specific pathogen–free facility at the University Hospital Zurich according to institutional guidelines. Animal experiments received prior approval by the veterinary office of the Canton of Zurich (license numbers 142/2017 and 246/2016) and were conducted in accordance with Swiss Federal and Cantonal Law.

In vivo treatments

Recombinant human IL-2 (teceleukin, Roche) was obtained from the National Cancer Institute of the National Institutes of Health. IL-2/NARA1 (IL-2cx) and IL-2/5344 antibody complexes were prepared by mixing 15,000 IU IL-2 and 15-μg anti-IL-2 mAbs per injection, as previously described (2224). IL-2cx, IL-2/5344 antibody complexes, or 200,000 IU IL-2 were injected daily for three consecutive days unless stated otherwise. Where mentioned, depletion of specific immune cells was performed using 250-μg anti-NK1.1 mAb (clone PK136, Leinco Technologies) and 500-μg anti-Thy1.2 mAb (clone 30H12, BioXcell). Depleting mAbs were administered intraperitoneally daily throughout the experiment starting 1 day before treatment initiation with IL-2 complexes. For in vivo activation of DCs, 100-μg polyinosinic–polycytidylic acid (Sigma-Aldrich) and 100-μg Pam3Csk4 (InvivoGen) were injected daily for three consecutive days.

To determine immune cell proliferation, BrdU (B5002, Sigma-Aldrich) was given to mice either by daily intraperitoneal injection (1 mg per injection) or in drinking water (0.8 mg/ml), as established (66). BrdU-incorporated cells were labeled using the FITC BrdU Flow Kit (BD Biosciences) according to the manufacturer’s instructions. Where indicated, 500 μg of etanercept (Enbrel, Pfizer) was injected concomitantly with IL-2cx for three consecutive days.

Flow cytometry

Single-cell suspensions of lymph nodes, spleens, and bone marrow were prepared according to standard protocols. For all experiments evaluating DCs, organs were digested in phosphate-buffered saline (PBS) containing collagenase D (1 mg/ml; Roche) for 15 min in the incubator at 37°C. Tumors were cut into small pieces and incubated in 10 ml of dissociation buffer [RPMI 1640, 5% FCS, deoxyribonuclease I (10 μg/ml; Sigma-Aldrich) and collagenase type I (200 U/ml; Thermo Fisher Scientific)] for 60 min at 37°C, shaking with 25 rpm. Cell suspensions were then passed through a 70-μm cell strainer. In experiments with intradermally implanted tumors, after one wash, a Percoll (40 and 70%; GE Healthcare) gradient centrifugation was performed. All cell suspensions were stained for flow cytometry analysis using flow cytometry buffer (PBS with 2% FCS and 2 mM EDTA) and fluorochrome-conjugated antibodies for at least 20 min at 4°C. A list of used antibodies is provided in table S1. Intracellular staining with Ki67, FOXP3, GATA3, RORγt, TBET, BAX, BCL-2, CSF2, and TNF was performed following the manufacturer’s instructions using the eBioscience Foxp3/Transcription Factor Staining Buffer Set (Thermo Fisher Scientific). Caspase-3 was measured with the Active Caspase-3 Apoptosis Kit (BD Biosciences) according to the manufacturer’s instructions. Annexin V and 7-AAD live and dead cell staining was conducted according to the instructions of the Annexin V Apoptosis Detection Kit with 7-AAD (BioLegend).

For STAT5 phosphorylation analysis, freshly isolated mouse splenocytes or human PBMCs were incubated for 25 min at 37°C as indicated with either mouse CSF2 (Peprotech), human CSF2 (Peprotech), or IL-2 (teceleukin, Roche). Cells were directly fixed with paraformaldehyde (final concentration 1.5%; Sigma-Aldrich) followed by permeabilization with methanol (Cantonal Pharmacy Zurich) and staining with fluorochrome-conjugated antibodies in flow cytometry buffer.

Cell cycle analysis was conducted with FxCycle PI/RNase Staining Solution (Thermo Fisher Scientific) after ethanol fixation (final concentration 70%; Cantonal Pharmacy Zurich) according to the manufacturer’s instruction.

Intracellular RNA was measured with PrimeFlow RNA assay (Thermo Fisher Scientific) with target probes against Actb (Ref 60384), Flt3l (Ref 64644), Csf2 (Ref 64983), and Tnf (Ref 60209) according to the manufacturer’s instructions.

Samples were acquired with a BD FACS Canto II, BD LSR II, or BD FACSymphony flow cytometer (BD Biosciences) and analyzed using FlowJo software (BD Biosciences). Where indicated, fluorescence-activated cell sorting (FACS) was performed using a BD FACS Aria III 5L (BD Biosciences).

Generation of bone marrow chimeras

Bone marrow chimeric mice were generated by adoptive transfer of 1 to 2 × 106 lineage (Lin)–depleted bone marrow cells from donor mice into lethally (950 RAD) irradiated host mice. Lineage depletion was done by magnetic negative selection (MoJoSort Streptavidin Nanobeads, BioLegend) with biotinylated antibodies targeting CD3, CD19, NK1.1, and Ter119 (all from BioLegend). For ILC reconstitution experiments, ILCp (CD127+ α4β7+ CD25+ CD135 CD117) and, where indicated, CLPs (CD127+ α4β7 CD135+) were FACS-sorted from Lin (CD3, CD4, CD5, CD8, CD11b, CD11c, CD19, CD335, B220, Gr-1, Ter119, NK1.1, and F4/80) bone marrow, followed by adoptive transfer into sublethally (450 RAD) irradiated Il2rg−/− or Il2rg−/− Rag2−/− mice. All mice were rested for at least 6 to 8 weeks before treatment with IL-2cx.

Cell cultures

For DQ-ovalbumin assays, splenocytes were isolated as described above followed by incubation in DQ-ovalbumin (Life Technologies) at a concentration of 50 μg/ml in full RPMI 1640 medium [RPMI 1640 (Gibco) containing 10% FCS (Gibco), 4 mM l-glutamine (Thermo Fisher Scientific), and penicillin-streptomycin (Thermo Fisher Scientific)] for the time indicated.

For intracellular TNF or CSF2 measurements in lymphocytes, splenocytes of wild-type mice were stimulated for 4 or 8 hours in full RPMI 1640 medium containing either phorbol 12-myristate 13-acetate (PMA; 50 ng/ml; Sigma-Aldrich), ionomycin (0.5 μg/ml; Sigma-Aldrich), and brefeldin A (BFA; 10 μg/ml; Sigma-Aldrich) for TNF evaluation or PMA (100 ng/ml), ionomycin (1 μg/ml), and 2 μM monensin (BioLegend) for CSF2 evaluation. For intracellular TNF and CSF2 measurements in ILCs, splenocytes of Tcrbd−/− mice were stimulated for 6 hours in full RPMI 1640 medium containing PMA (100 ng/ml), ionomycin (1 μg/ml; Sigma-Aldrich), and BFA (10 μg/ml; Sigma-Aldrich).

Adoptive DC transfer by intrasplenic injection

For studies assessing the persistence of mature DCs, cell trace violet–labeled (labeling for 7.5 min at 37°C; Thermo Fisher Scientific), magnetically purified CD11c+ (MoJoSort Mouse CD11c Nanobeads, BioLegend) splenocytes isolated from untreated or IL-2cx–pretreated CD45.1+ or CD45.2+ mice were adoptively transferred by subcapsular injection into the spleen of either untreated (Fig. 2F) or IL-2cx–pretreated recipients (Fig. 2G). IL-2cx–treated donors received IL-2cx on three consecutive days before isolation of splenocytes, whereas IL-2cx–treated recipients received IL-2cx on three consecutive days with the last injection coinciding with the day of adoptive DC transfer.

OT-1 proliferation

CD45.1 mice were treated with IL-2cx for three consecutive days or left untreated followed by isolation of splenocytes and magnetic purification of CD11c+ cells with (MojoSort Mouse CD11c Nanobeads, BioLegend). Isolated cells were incubated in RPMI 1640 (Gibco) containing 10% FCS (Gibco), 4 mM l-glutamine (Thermo Fisher Scientific), penicillin-streptomycin (Thermo Fisher Scientific), and ovalbumin peptide (50 μg/ml) (DEVSGLEQLESIINFEKLAAAAAK, obtained from the peptide synthesis facility of Leiden University Medical Center) for 2 hours before injection into the foot pad of CD45.2 recipient mice (48). Purified and CFSE (2 to 3 × 106; Thermo Fisher Scientific)–labeled OT-1 T cells were adoptively transferred to recipients 24 hours later, and proliferation was measured 2 to 3 days after adoptive transfer.

Tumor experiments

The mouse B16-F10 (CRL-6475, American Type Culture Collection) cell line was cultured in Advanced Dulbecco’s modified Eagle’s medium (Gibco) containing 10% FCS (Gibco), 4 mM l-glutamine (Thermo Fisher Scientific), and penicillin-streptomycin (Thermo Fisher Scientific). Recipient mice were intradermally engrafted with 1 × 106 B16-F10 cells. Treatment was started when tumor nodules were visible and palpable on day 4 after tumor inoculation by using IL-2cx or 100-μg anti–PD-1 (clone RMP1-14, Leinco Technologies) injected every other day (fig. S6B, top) (24). Mice were euthanized on day 11 for isolation of tumor-infiltrating lymphocytes. For evaluating tumor growth under DC-deficient conditions, lethally irradiated C57Bl/6J host mice were reconstituted with Lin-depleted Cd11c-DTR or Zbtb46-DTR bone marrow, as described above, followed by treatment with 0.5-μg DT (Sigma-Aldrich) three times weekly starting 1 day before tumor inoculation. IL-2cx and anti–PD-1 treatment was continuously administered three times weekly starting day 4 after tumor inoculation (fig. S6B, bottom). Tumor volume was calculated according to the following formula: V = 2/3 × π × ((a + b)/4)3, a (mm) was the length and b (mm) was the width of the tumor.

For inducible tumors, BrafCA PtenloxP Tyr::CreERT2 mice were topically applied on the skin with 2 μl of 25 mg/ml 4-hydroxytamoxifen (≥70% Z isomer, Sigma-Aldrich) dissolved in dimethyl sulfoxide (Sigma-Aldrich) on three consecutive days. Approximately 21 to 28 days later when tumor nodules have formed, mice were treated with IL-2cx or anti–PD-1 every other day, as described above (Fig. 6F).

Cytokine measurements

Serum samples from mice were collected and stored at −80°C until cytokine testing. Serum cytokine concentrations were determined using Mouse GM-CSF enzyme-linked immunosorbent assay (ELISA) MAX Standard (BioLegend), Mouse TNF-α Quantikine HS ELISA kit (R&D Systems), and Mouse Flt-3 L ELISA Kit (Thermo Fisher Scientific). Measurements were performed in accordance to the manufacturer’s instructions.

Human serum samples before and after aldesleukin treatment were collected and stored as described before. GM-CSF was measured using Human GM-CSF ELISA MAX Standard (BioLegend). Human FLT3L and TNF were measured with proximity extension assay using the Olink Inflammation Panel.

Gene expression correlation analyses

For mouse Il2r expression analysis, RNA-seq and microarray data were retrieved from the ImmGen database (40) and plotted as relative expression in indicated immune cell subsets. The RNA-seq dataset for skin cutaneous melanoma was downloaded with Firebrowse (The Broad Institute) from TCGA (67). Gene expression was log2-transformed (log2[gene expression + 1]) and different genes of interest were correlated against each other. For gene signatures, the mean expression of all signature genes was calculated and plotted against BATF3 expression. The NK cell signature was defined by the genes NCR1, NCR3, KLRB1, CD160, and PRF1 (15). The IL-2 signature was defined by the IL-2-target genes LTA, ARHGDIA, PKD2, LTB, KLF6, PDIA3, GZMB, PSMB1, S100A4, and FLT3LG (51) and IL2. Significance of correlative gene expression was calculated by using Pearson’s correlation coefficient. Survival analysis was done with gene expression profiling interactive analysis 2 (68).

Quantitative polymerase chain reaction

RNA was isolated with the RNeasy Plus Micro and Mini Kits (Qiagen) from 250,000 to 1,000,000 FACS-sorted cDCs (CD3 CD19 NK1.1 CD45RB MHC-II+ CD11c+), CD4+ Tcon (CD3+ CD4+ Foxp3), and CD4+ Treg cells (CD3+ CD4+ Foxp3+) from Foxp3-GFP-DTR mice. Control RNA was isolated from untreated bulk wild-type, Il2ra−/−, Il2rb−/−, and Il2rg−/− splenocytes. After cDNA reverse transcription (High Capacity cDNA Reverse Transcription Kit, Applied Biosystems), qPCR analysis was performed using SYBR green (KAPA SYBR FAST, Roche) on a 7900HT Fast Real-Time PCR System (Applied Biosystems), and ΔΔCt values were calculated. The following primer sequences were used: Hprt, GAT CAG TCA ACG GGG GAC AT (forward) and ATC CAA CAA AGT CTG GCC TGT (reverse); Il2ra, AAC CAT AGT ACC CAG TTG TCG G (forward) and TCC TAA GCA ACG CAT ATA GAC CA (reverse); Il2rb, TGG AGC CTG TCC CTC TAC G (forward) and TCC ACA TGC AAG AGA CAT TGG (reverse); Il2rg, GGT TAC TGA ATA CCA AGG GAA CTT T (forward) and TGG CAG AAC CGT TCA CTG TA (reverse).

RNA sequencing

Forty thousand splenic mouse cDCs from untreated and IL-2cx–treated wild-type mice were separated by FACS in RLT Plus lysis buffer (Qiagen) containing 1% 2-mercaptoethanol (Sigma-Aldrich). Subsequently, RNA was isolated using the RNeasy Plus Micro Kit (Qiagen). RNA-seq and data processing were done by the Functional Genomics Center Zurich, as briefly described here. The RNA extracted from sorted cells was quantified for quality and concentration using the TapeStation RNA high sensitivity kit (Agilent). The estimated RNA integrity numbers were in a range of 8.4 to 9.8 showing a good RNA quality; the RNA concentrations of 1.1 to 10 ng/μl recommended the use of a low input library preparation method. Accordingly, the SMARTer Stranded Total RNA Seq Kit v2 (Takara Bio) was used to prepare cDNA by universal priming (with 3-min fragmentation) and to deplete ribosomal cDNA with ZapR v2 and R Probes v2. The libraries were quantified by Tapestation D1000 (Agilent) measurements, and they showed fragment sizes around 320 bp. The libraries were sequenced on a HiSeq 4000 platform using 125 cycles single-read targeting ~40 M reads per sample. Adapters and low-quality tails were trimmed from reads before read alignment. STAR aligner (v2.5.4b) (69) was used to align the RNA-seq dataset to Ensembl genome build GRCh38.p10 (Release 91). Gene expression counts were calculated with feature counts from Bioconductor package Rsubread (v1.32.1) (70). A gene was considered as expressed if, in at least one group of the comparison, it had more than 10 counts in more than half of the samples. Differentially expressed genes were detected using Bioconductor package EdgeR (v3.20.6) (71). Gene set enrichment analysis was done with GO analyzer for RNA-seq and other length-biased data (goseq, v1.30.0) (72). For the heatmap display, the top 75 up-regulated and top 75 down-regulated genes were log2-transformed (log2[normalized gene count + 1]) for calculation of Z-scores.


Specific software used for analysis is mentioned where applied. Two cartoons were created using

Statistical analyses

Detailed information on sample numbers and statistical tests used are described in the figure legends. Calculation was performed with the GraphPad Prism 8 software package. In general, differences between two groups were tested with t tests for normally distributed data or Mann-Whitney test for not normally distributed data. Multiple comparison was done with either one-way analysis of variance (ANOVA) followed by Holm-Sidak’s multiple comparison test for normally distributed data or Kruskall-Wallis test followed by Dunn’s multiple comparison test for not normally distributed data. Time-dependent comparisons were tested with two-way ANOVA or mixed-effects model, followed by Holm-Sidak’s multiple comparison test. Normality testing was done using d’Agostino-Pearson test. Correlation analysis was done with Pearson correlation coefficient and survival analysis with Logrank test. For all statistical analyses, significance was accepted at the 95% confidence level (P < 0.05). Exact P values are provided; n.s. indicates not significant.


Fig. S1. Different IL-2 formulations cause cDC1 and cDC2 expansion.

Fig. S2. Apoptosis-mediating proteins in cDCs do not change after IL-2cx treatment.

Fig. S3. IL-2R expression in stimulated cDCs and IL-2cx-treated cDCs and DC precursors.

Fig. S4. Expansion of cDC subsets in Flt3l−/− Csf2−/− mice and cDC proliferation in etanercept-treated wild-type mice.

Fig. S5. IL-2R expression on different leukocytes and expansion of CD4+ T, CD8+ T, and NK cells by IL-2 complexes.

Fig. S6. DCs show improved antigen processing and facilitate antitumor responses upon IL-2cx treatment.

Fig. S7. Model of IL-2–mediated DC expansion and tumor infiltration.

Table S1. Antibodies and fluorescent dyes used for flow cytometry.

Data file S1. Primary data.


Acknowledgments: We thank L. Bürgi for help with experiments, U. Steiner, E. Käser, and C. Defila for support with the clinical trial, Y. Zurbuchen and C. Cervia for support in maintaining the mouse colony during the SARS-CoV-2 pandemic, M. Manz (University Hospital Zurich) and A. Rolink (University of Basel) for providing mice, and the members of the Boyman laboratory for discussion and critical reading of the manuscript. Editorial services were provided by N. R. Gough (BioSerendipity, LLC, Elkridge, MD). Funding: This work was supported by the Swiss National Science Foundation (310030-172978 to O.B.), the Hochspezialisierte Medizin Schwerpunkt Immunologie (HSM-2-Immunologie; to O.B.), the Clinical Research Priority Program CYTIMM-Z of the University of Zurich (to O.B.), Swiss Cancer Research grant (KFS-4136-02-2017 to O.B.), Swiss Cancer MD-PhD fellowship (MD-PhD-3557-06-2015 to M.E.R.), Fritz Rohrer-Fonds (to M.E.R. and O.B.), Olga Mayenfisch Stiftung (to M.E.R.), EMDO Stiftung (to M.E.R. and O.B.), and the Young Talents in Clinical Research Fellowship by the Swiss Academy of Medical Sciences and Bangerter Foundation (YTCR 32/18 to M.E.R.). Author contributions: M.E.R. designed and performed experiments, conducted the clinical trial, analyzed the data, and wrote the manuscript. R.A.R. designed and performed experiments and gave scientific input. D.S. and U.K. performed experiments. O.B. designed and analyzed experiments, supervised the study and the clinical trial, and wrote the manuscript with input from all the authors. Competing interests: O.B. is a founder and shareholder of Anaveon AG and holds patents on improved IL-2 formulations (patents WO2017122130A1, WO2014100014A1, WO2016005950A1, and WO2017121758A1). M.E.R., U.K., and O.B. are listed as inventors on a patent application on the use of IL-2 formulations for DC immunotherapy. The other authors declare that they have no competing interests. Data and materials availability: The raw dataset generated by RNA-seq has been deposited in the ArrayExpress database at EMBL-EBI ( under accession number E-MTAB-8991. All other data associated with this study are present in the paper or the Supplementary Materials. Materials are available upon request from the corresponding author.

Stay Connected to Science Translational Medicine

Navigate This Article