Effects of interleukin-2 in immunostimulation and immunosuppression

Clues supporting the existence of IL-2 emerged in 1976. That year, after successfully expanding T cells in vitro, Morgan et al. (1976) reported the presence of a T cell growth factor in the conditioned media of phytohemagglutinin-stimulated blood lymphocytes. In the early 1980s, human IL-2 was firmly identified as a variably glycosylated ∼15.5-kD protein (Robb and Smith, 1981), thereafter purified (Smith et al., 1983), and finally cloned (Taniguchi et al., 1983). At the same period, the IL-2 receptor (IL-2R) was discovered (Kuribayashi et al., 1981; Robb et al., 1981), thus solving the first type I cytokine/receptor complex.

By permitting a prolonged culture of T cells, the discovery of IL-2, initially called “T cell growth factor,” facilitated molecular and cellular investigations that precipitated, for example, the characterization of the TCR and its function (Allison et al., 1982;Haskins et al., 1983), or the identification of the first human retrovirus: human T cell leukemia virus (HTLV-1; Poiesz et al., 1980). Initial studies performed in vitro concluded to a critical role of IL-2 in the development of effector T lymphocytes. Moreover, experimental investigations conducted in a chicken model of autoimmune thyroiditis revealed a pro-autoimmune effect of IL-2 and IL-2R–expressing T lymphocytes (Krömer et al., 1985), an observation that was mechanistically explained by the capacity of IL-2 to reverse anergy of self-reactive T cells in mice (Gonzalo et al., 1993) and simultaneously validated by clinical studies in humans showing that cancer patients treated with high-dose (HD) IL-2 frequently developed autoimmune thyroiditis (Krouse et al., 1995).

However, in vivo studies conducted in the 1990s in mouse strains lacking IL-2 or IL-2R subunits led to a revision of the concept that the IL-2/IL-2R system would be solely involved in immunostimulatory circuities. Indeed, rather than harboring an immunodeficiency, these animals demonstrated lymphadenopathy, uncontrolled proliferation of peripheral activated T cells, and signs of autoimmunity (Sadlack et al., 1993;Suzuki et al., 1995; Willerford et al., 1995). Such observations unveiled the existence of immunosuppressive mechanisms critically relying on IL-2 and later attributed to regulatory CD4⁺ T cells (Tregs; Sakaguchi et al., 1995; Malek et al., 2000, 2002).

The immunomodulatory effects of IL-2, mainly on effector and regulatory T lymphocytes, have been exploited for treating various pathologies, though with limited clinical benefits so far. In this line, a recombinant human IL-2 called aldesleukin (brand name: Proleukin) was approved for the treatment of kidney cancer and melanoma as early as 1992 and 1998, respectively (Alva et al., 2016).

After introducing some fundamental aspects of IL-2 biology, the present review will summarize current strategies to introduce IL-2 into the immunotherapeutic armamentarium.

IL-2 is mainly produced by CD4⁺ T lymphocytes (naive, memory, and T helper [Th] 1) following antigenic stimulation, by type 2 and 3 innate lymphoid cells in the small intestine, and to a lesser extent by activated CD8⁺ T cells, B cells, and by other innate immune entities such as natural killer (NK) and NKT lymphocytes, dendritic cells (DCs), monocytes, or mast cells (Malek, 2008; Wojciechowski et al., 2009; Hershko et al., 2011; Zelante et al., 2012; Zhou et al., 2019). In naive T lymphocytes, the engagement of the TCR and co-stimulatory molecules (e.g., CD28) within an immunological synapse activates activator protein 1 (AP-1), NFκB, and NFAT (Fig. 1). In cooperation with constitutive factors, these transcription factors promote the expression of the IL2 gene (Serfling et al., 1995). IL2 transcription occurs within 30 min after stimulation but is transient, declining to background levels within 24–48 h. Additionally, post-transcriptional regulatory mechanisms further restrict the availability of IL-2 mRNAs, the levels of which usually peak at 4–8 h after stimulation (Jain et al., 1995). The turnover of IL-2 mRNAs is mostly controlled by proteins interacting with an AU-rich cis element (ARE) in their 3′-untranslated region. Among these trans-acting factors figure nuclear factor 90 (NF90) and tristetraprolin. NF90 is activated by protein kinase (PK) B (best known as AKT) upon CD28 co-stimulation, or by PKC upon restimulation with PMA, and then exported from the nucleus to the cytosol. There, NF90 binds to ARE and stabilizes IL-2 mRNAs, thus allowing their translation (Pei et al., 2008; Zhu et al., 2010). In contrast, tristetraprolin is expressed in T lymphocytes following activation and plays a critical role in the rapid decay of IL-2 mRNAs, as its interaction with ARE promotes its degradation (Ogilvie et al., 2005; Yang et al., 2015). Altogether, these transcriptional and post-transcriptional mechanisms control the magnitude and duration of IL-2 production by activated T cells.

Once secreted, IL-2 is consumed in an autocrine/paracrine manner by neighboring cells that harbor its receptor, IL-2R (Fig. 2). The latter consists of a hetero-complex of up to three subunits: α, β, and γ, also known as CD25, CD122, and CD132, respectively. Although each receptor subunit may independently bind IL-2 with a weak affinity (Kd: ∼10⁻⁸–10⁻⁷ M), only the intermediate-affinity βγ dimeric (Kd: ∼10⁻⁹ M) and the strong-affinity αβγ trimeric IL-2R (Kd: ∼10⁻¹¹ M) mediate intracellular signal transduction (Flynn and Hartley, 2017). The γ subunit is ubiquitously expressed on most hematopoietic cells. It is also referred to as the “common” chain (labeled “γ_c”), as it is shared with the receptors for IL-4, -7, -9, -15, and -21. The β subunit appears constitutively expressed at various levels on T and NK(T) lymphocytes. In contrast, the α chain/CD25 is detected on early thymocytes, absent on naive/memory T cells (particularly in mice, but detectable on a minor fraction of human naive/memory T cells), transiently exposed on activated/effector T lymphocytes, and preferentially/more stably expressed on Tregs (Kmieciak et al., 2009;Flynn and Hartley, 2017). Within the myeloid compartment, monocytes display the intermediate-affinity βγ receptor, whereas DC subtypes can present the three subunits of IL-2R (Bosco et al., 2000; Herr et al., 2014; Kitashima et al., 2018). Interestingly, DCs may also supply the α chain in trans, thus supporting high-affinity binding of IL-2 to naive T lymphocytes that undergo antigen priming (Wuest et al., 2011). Additionally, some nonhematopoietic cell types may harbor either (i) dimeric βγ receptors, as this applies to intestinal epithelial cells, dermal fibroblasts, or fibroblast-like synoviocytes; or (ii) the high-affinity αβγ IL-2R, as reported for endothelial cells, proximal tubular epithelial cells, or gingival fibroblasts. As an aside, IL-2R expression has also been documented for malignant cells such as melanoma and cervical tumor cells (Valle-Mendiola et al., 2016).

Binding of IL-2 to IL-2R activates JAK1/3. In turn, these kinases ignite the phosphoinositide-3-kinases (PI3Ks)/phosphatidylinositol 3,4,5-trisphosphate (PIP3)/AKT/mechanistic target of rapamycin (mTOR)/p70^S6K, and Ras/Raf/mitogen-activated protein kinase kinases 1 and 2 (MAP2K1/2, also known as MEK1/2)/extracellular signal-regulated kinases 1 and 2 (ERK1/2) signaling cascades and phosphorylate STAT5. Activation of mTOR, p70^S6K, AKT, and ERK1/2 modulates the activity and de novo expression of multiple downstream regulators involved in protein synthesis, autophagy, cell metabolism, survival, proliferation, and differentiation. Concomitantly, activated STAT5 determines the fate of the cell by transactivating numerous target genes (Fig. 2). While these three signaling cascades are stimulated in effector T cells, it is important to note that the STAT5 pathway is predominantly triggered in Tregs (Cheng et al., 2011). In T lymphocytes, depending on their subtype, STAT5-stimulated genes can encode the following: various cytokine receptors including IL-2Rα and IL-2Rβ (positive feedback), IL-4Rα or IL-12Rβ; diverse proteins involved in cell proliferation and survival (e.g., PIM1, MYC, cyclins, and B cell lymphoma (BCL) 2); some effector molecules such as granzyme B, CD178 (best known as Fas-ligand), and some cytokines like IFN-γ, TNF-α, or IL-4; and some regulators of immune cell functions such as the suppressor of cytokine signaling 1 and 2 (SOCS1/2), and the transcription factor forkhead box P3 (FOXP3). IL-2-activated STAT5 also represses the expression of particular effector genes such as IL17A or BCL6 (Kovanen et al., 2005; Grange et al., 2013; Knosp et al., 2013; Kanai et al., 2014; Ross and Cantrell, 2018). Phenotypic changes consecutive to IL-2 stimulation are described in the following paragraph.

IL-2 is a pleiotropic cytokine with immunostimulatory or immunoinhibitory activity depending on the target cell. Its effects on nonhematopoietic and innate immune cells remain poorly deciphered (Valle-Mendiola et al., 2016). Above all, IL-2 stands out as a well-established regulator of T cell development and homeostasis.

CD4⁺FOXP3⁺CD25⁺ Tregs are responsible for maintaining immunological self-tolerance and for down-regulating inflammatory and adaptive immune responses. As opposed to conventional T cells (Tconv cells), Tregs do not produce IL-2 but critically rely on its presence for their differentiation from immature single-positive CD4⁺ T cells in the thymus (referred to as natural/thymic Tregs) or from naive CD4⁺ T cells in the periphery (referred to as peripheral Tregs), when they acquire FOXP3, and hence for their expansion, survival, and immunoregulatory functions (including the secretion of the immunosuppressive cytokines IL-10 and TGF-β) in the periphery (Kündig et al., 1993;Sadlack et al., 1993; Suzuki et al., 1995; Willerford et al., 1995; Malek et al., 2002; D’Cruz and Klein, 2005; Fontenot et al., 2005; Burchill et al., 2007; Zheng and Rudensky, 2007; Malek, 2008; Tang et al., 2008; Campbell and Koch, 2011; Goldstein et al., 2013; Lin et al., 2013; Chinen et al., 2016; Ross and Cantrell, 2018).

Additionally, IL-2 regulates lineage commitment of CD4⁺ Th cell subsets. Th1 lymphocytes promote cellular immune responses against intracellular microbes and cancer cells. Th1 cell differentiation is triggered by IL-2–mediated expression of IL-12Rβ in naive CD4⁺ T cells. Then, together with IL-12 produced by APCs, IL-2 coordinates the expression of the transcription factor T-Box21 (TBX21, best known as T-bet), which orchestrates Th1 differentiation and the production of type 1 cytokines like IFN-γ (Reem and Yeh, 1984; Shi et al., 2008; Liao et al., 2011). Th2 lymphocytes regulate humoral immunity to extracellular parasites and bacteria, as well as allergic inflammation. In the presence of IL-4, IL-2 sensitizes cells to the Th2 program by up-regulating IL-4Rα expression and by inducing epigenetic changes that boost IL-4 secretion (Ben-Sasson et al., 1990; Zhu et al., 2003;Cote-Sierra et al., 2004; Le Gros et al., 2008; Liao et al., 2008). Th17 cells coordinate the adaptive defense against extracellular pathogens, mediate tissue inflammation, and participate in many autoimmune pathologies. IL-2 inhibits the differentiation of Th17 cells through various mechanisms including repression of retinoic acid receptor–related orphan nuclear receptor γt, IL-6Rα, and IL-17A. These three factors are required for the development and the proinflammatory activity of Th17 cells (Laurence et al., 2007;Liao et al., 2011; Yang et al., 2011). Accordingly, it has been reported that IL-2 consumption by Tregs facilitates Th17 differentiation in Candida albicans infection (Pandiyan et al., 2011). Similarly, IL-2 inhibits the differentiation of T follicular helper (Tfh) CD4⁺ T cells, which control B cell responses and germinal center formation. Thus, IL-2 activates the expression of positive regulatory domain zinc finger protein 1 (PRDM1, best known as B lymphocyte–induced maturation protein 1 [BLIMP-1]) which in turn trans-represses gene expression of BCL6, a transcriptional regulator required for Tfh development (Johnston et al., 2009; Ballesteros-Tato et al., 2012). Interestingly, in an influenza infection model, high levels of IL-2 prevented the development of T follicular regulatory (Tfr) cells in the course of the infection. However, once the virus was eliminated and the response resolved, some Tregs down-regulated IL-2Rα/CD25 and up-regulated BCL6 before differentiating into Tfr cells. These Tfr cells migrated to B cell follicles to prevent clonal expansion of self-reactive B lymphocytes (Botta et al., 2017).

In naive CD8⁺ T lymphocytes, IL-2 mediates the acquisition of the effector cytotoxic phenotype following antigen encounter by promoting the secretion of IFN-γ, TNF-α/β, granzyme B, and perforin. The intensity of the IL-2/IL-2R signaling activity shifts the fate of CD8⁺ T cells toward a short-lived effector or a long-lived memory phenotype. De facto, high levels of IL-2 stimulate the synthesis of BLIMP-1, which controls the effector transcriptional program and inhibits the expression of central memory markers such as BCL6, IL-7Rα/CD127, and CD62L. On the contrary, low levels of IL-2 impair the synthesis of effector molecules, while authorizing the (re)expression of BCL6, IL-7Rα/CD127, and CD62L, thus driving activated CD8⁺ T cells toward the memory compartment (Manjunath et al., 2001; Williams et al., 2006; Kalia et al., 2010; Pipkin et al., 2010;Ross and Cantrell, 2018; Spolski et al., 2018). Similarly, low doses of IL-2 drive a memory or Tfh-like phenotype in CD4⁺ T lymphocytes (Boyman and Sprent, 2012). Importantly, BLIMP-1 induction in effector cells ultimately leads to a reduced production of IL-2, accompanied by a contraction of the effector alcove and the appearance of an exhausted state (Gong and Malek, 2007; Beltra et al., 2016; Zhu et al., 2017). Interestingly, this exhausted phenotype, characterized by surface exposure of programmed cell death protein 1 (PD-1) or up-regulation of IL-2Rβ, is reversible. As evoked in the next section, this peculiarity may be exploited in immunostimulatory treatments by combining IL-2 and anti-PD1.

The exquisite sensitivity of both effector and regulatory T lymphocytes to IL-2 has designated this cytokine as a potential ally for treating immune-related diseases, spurring interest in both its immunosuppressive and immunostimulatory effects. Fine-tuning of the balance between these two functionally contrasted T cell subsets is at the heart of IL-2–based immunotherapies. Immunoinhibitory IL-2 treatments aim at selectively expanding Tregs over effector T cells, whereas immunostimulatory IL-2 interventions should proceed inversely. To achieve these opposite goals, fundamental distinctions of the IL-2/IL-2R system within the targeted populations are being exploited. Thus, Tregs constitutively display the IL-2Rαβγ trimer, while activated effector T cells only transiently express IL-2Rα.

When contemplating the addiction of Tregs to IL-2 and their preponderant role in maintaining immune tolerance and suppressing inflammation (Sakaguchi et al., 1995), it is a posteriori no surprise that inactivation of the IL2,IL2RA, IL2RB, STAT5, or FOXP3 genes in mice led to Treg cell depletion/dysfunction and to multi-organ autoimmune and inflammatory syndromes (Sadlack et al., 1993; Suzuki et al., 1995; Willerford et al., 1995; Malek et al., 2002; Fontenot et al., 2003;Snow et al., 2003; Burchill et al., 2007; Yao et al., 2007). Importantly, similar clinical manifestations have been detected in patients affected by mutations in IL2RA or FOXP3, leading to the immunodysregulation polyendocrinopathy enteropathy X-linked (IPEX) syndrome (Bennett et al., 2001;Wildin et al., 2001; Moraes-Vasconcelos et al., 2008;Goudy et al., 2013). Based on these observations, reestablishing Treg cell functions thanks to the administration of recombinant human IL-2 has been considered for the treatment of autoimmune, inflammatory, or graft-versus-host (GvHD) diseases (Krömer et al., 1985;Andreu-Sánchez et al., 1991; Koreth et al., 2011;Saadoun et al., 2011; Hartemann et al., 2013; Matsuoka et al., 2013; Castela et al., 2014; He et al., 2016).

For immunosuppression, the common strategy relies on the administration of low-dose (LD) IL-2 (as opposed to HD IL-2, which has been clinically approved for cancer therapy), based on the rationale that limited concentration of IL-2 restricts its capture to Tregs (rather than Tconv cells), as Tregs constitutively express high amounts of the high-affinity IL-2Rαβγ (Klatzmann and Abbas, 2015). The proof of concept of this approach was attained when LD IL-2 reverted and prevented experimental type 1 diabetes (T1D) in mice (Tang et al., 2008; Webster et al., 2009; Yu et al., 2009; Grinberg-Bleyer et al., 2010; Baeyens et al., 2013; Pérol and Piaggio, 2016). LD IL-2 also achieved disease control in other indications, but only when delivered chronically, like in Alzheimer’s disease (Dansokho et al., 2016), or when combined with other drugs such as glucocorticoids in GvHD (Pérol and Piaggio, 2016) andTrypanosoma cruzi infection (González et al., 2015) or rapamycin in skin transplantation (Pérol and Piaggio, 2016). In the clinic, LD IL-2 has been evaluated for the treatment of T1D (Long et al., 2012; Hartemann et al., 2013; Todd et al., 2016), chronic GvHD (Koreth et al., 2011,2016; Matsuoka et al., 2013; Kennedy-Nasser et al., 2014), systemic lupus erythematosus (SLE; He et al., 2016; Humrich and Riemekasten, 2016), alopecia aereata (Castela et al., 2014), and vasculitis associated with chronic hepatitis C virus infection (Saadoun et al., 2011). Overall, LD IL-2 was well tolerated, not only due to the dosing (<3 × 10⁶ IU/d) but also due to a preference for s.c., rather than systemic, administration. In all these pathological conditions, LD IL-2 led to an increase in circulating Tregs, and some clinical responses were observed. Nevertheless, to increase clinical efficacy, improvements are needed as IL-2 has a short half-life, requires repeated administration, and can unwantedly activate NK cells (Ye et al., 2018).

The short half-life of IL-2 (10–85 min in serum) is intrinsic to its small molecular weight that falls beneath the glomerular filtration cutoff estimated at 30–50 kD (Konrad et al., 1990; Ruggiero et al., 2010). Hence, to maintain efficient IL-2 bioavailability, administration must be repeated at close intervals. Alternatively, prodrug versions of recombinant human IL-2 decorated with releasable polyethylene glycol (PEG) chains have been synthesized and allowed: (i) a sustained release of the cytokine; (ii) a prolonged stimulation of the IL-2R signaling pathway; and (iii) a biased binding of IL-2 either to IL-2Rα/CD25 or IL-2Rβ/CD122, depending on the sites of PEGylation (Charych et al., 2017). For instance, a PEGylated IL-2 named NKTR-358 demonstrates a lower affinity for IL-2Rβ than for IL-2Rα and preferentially activates Tregs over Tconv cells. NKTR-358 restored Treg function in murine and simian models of SLE and cutaneous hypersensitivity, respectively, encouraging its clinical evaluation in patients (Cully, 2017; Table 1).

IL-2/anti–IL-2 antibody immunocomplexes (referred to as “IL2Cxs”) are also designed to increase the half-life of IL-2 (Létourneau et al., 2010). Moreover, depending on the fine specificity of the antibody, IL2Cxs can redirect IL-2 toward IL-2Rαβγ⁺ Tregs (pro–Treg cell function) or IL-2Rβγ⁺ NK and naive/memory CD8⁺ T cells (pro-effector; Boyman et al., 2006). For instance, the mAb JES6-1 recognizes an epitope on murine IL-2 that contacts IL-2Rβ and γ. Interestingly, JES6-1 sterically blocks the interaction of IL-2 with IL-2Rβ and allosterically reduces IL-2 affinity for IL-2Rα. Experimentally, IL-2/JES6-1 selectively binds to and expands Tregs as they present sufficient IL-2Rα to displace the mAb (Spangler et al., 2015). Pro-Treg IL2Cxs have been successfully used in mouse models of experimental autoimmune encephalomyelitis (Webster et al., 2009), diabetes (Pérol and Piaggio, 2016), allergy (Smaldini et al., 2018), atherosclerosis (Dinh et al., 2012), and solid organ transplantation (Pérol and Piaggio, 2016). When complexed to recombinant human IL-2, the first clinical grade pro-Treg anti–IL-2, named F5111.2, demonstrated effectiveness in inducing T1D remission in diabetic mice, and in diminishing the severity of xeno-GvHD and experimental autoimmune encephalomyelitis (Trotta et al., 2018).

An alternative approach consisting in fusing IL-2 to the α chain of IL-2R has recently been evaluated in rodents (Ward et al., 2018). This IL-2–CD25 fusion protein exhibited an increased half-life and selectively expanded Tregs in vivo. In nonobese diabetic mice, administration of IL-2–CD25 enhanced Tregs in the endocrine pancreas and diminished the occurrence of diabetes. These results encourage the clinical evaluation of such a new class of IL-2 derivatives for treating autoimmunity and other pathologies that result from an exacerbated immune response (Ward et al., 2018).

Elucidation of the quaternary structure of IL-2 assembled to IL-2Rαβγ (Wang et al., 2005) has facilitated the engineering of a series of IL-2 “muteins” with variable affinity to IL-2Rα (Rao et al., 2003, 2005; Carmenate et al., 2013), IL-2Rβ (Liu et al., 2009; Levin et al., 2012; Mitra et al., 2015; Peterson et al., 2018), and IL-2Rγ (Liu et al., 2009;Mitra et al., 2015; Carmenate et al., 2018). These IL-2 analogues elicit graded (agonistic, mixed, or antagonistic) and differential signaling outputs downstream of IL-2Rαβγ and IL-2Rαβ, ultimately affecting the Treg/effector ratio. For instance, the IL-2 mutein H9-RETR (nine amino acid substitutions) was engineered to bind IL-2Rβ with a rather high affinity and block its heterodimerization with IL-2Rγ. H9-RETR antagonized signal transduction by native IL-2 and prevented ex vivo proliferation of pre-activated human CD8⁺ T cells, as well as the cytolytic activity of NK cells (Mitra et al., 2015).

Like their WT counterpart, IL-2 muteins suffer from limited bioavailability in vivo. As a remedy, they are frequently fused with an mAb or a crystallizable fragment (Fc) of an antibody, thus generating so called “immunocytokines.” Lately, a human IL-2 mutein harboring an N88D substitution responsible for a reduced affinity for IL-2Rβγ was fused to a nontargeted effector-function-silent human IgG1. In macaques, this long-lived IgG-(IL-2N88D)₂ fusion protein sustained preferential amplification of Tregs (Peterson et al., 2018). In a GvHD murine model, the stabilized antagonist H9-RETR-Fc4 remarkably extended survival (Mitra et al., 2015). Preclinically, several IL-2–Fc molecules demonstrated therapeutic ability to induce transplantation tolerance (Zheng et al., 2003; Millington et al., 2012; Jindal et al., 2015; Mitra et al., 2015) or to prevent autoimmune disorders (Zheng et al., 1999; Bell et al., 2015). An Fc–IL-2 mutein called AMG-592 (Amgen) was designed to have greater half-life than native IL-2, as well as an increased affinity for IL-2Rα. In human peripheral blood mononuclear cell cultures, AMG-592 preferentially expanded Tregs over effector T cells and lowered the production of pro-inflammatory cytokines in comparison to native IL-2. In an ongoing first-in-human trial, AMG-592 was well tolerated without severe adverse events and increased the Treg/Tconv cell ratio (Tchao et al., 2017). AMG is now being evaluated in clinical trials for the treatment of rheumatoid arthritis, SLE, and GvHD (Table 1). Another Fc–IL-2 mutein with increased affinity for IL-2Rα, named DEL-106 (Delinia-Celgene-BMS), is under development (Cully, 2017).

Mathematical prediction models, validated by experimentations, demonstrated that Tregs locally outcompete Tconv cells in consuming IL-2, as long as its autocrine/paracrine level does not reach the T cell activation threshold. Beyond this point, activated T cells engage a positive feedback loop that will temporarily up-regulate the high-affinity IL-2R, support their proliferation, and secure their effector program (Busse et al., 2010).

In line with these theoretical considerations, i.v. infusions of HD IL-2 (6–7.2 × 10⁵ IU/kg/dose, 12–15 doses/d) have been conceived for the treatment of malignant and infectious diseases. In Western countries, HD IL-2 received approval for the care of metastatic renal cell carcinoma and melanoma in the 1990s (Alva et al., 2016). Seminal publications reported an overall response rate (ORR) of 14–23% (including 5–8% of complete responders [CRs]) in patients with kidney cancer (Fyfe et al., 1995, 1996) and a 16% ORR (6% CRs) in melanoma patients (Atkins et al., 1999). In the field of infectious disease, HD IL-2 has essentially been evaluated in AIDS as an adjuvant to antiretroviral therapy. A meta-analysis compiling data from 25 completed trials revealed an increase in the CD4⁺ T cell count in the presence of IL-2. However, IL-2 supplementation did not reduce mortality or the risk of opportunistic infections, and even tended to increase the rate of severe adverse events, thus discouraging further investigations in HIV-positive patients (Onwumeh et al., 2017).

The mitigated success of HD IL-2–based cancer therapy can be attributed to the following: (i) its short bioavailability; (ii) an undesired expansion of Tregs that dampens antitumor immunity; and (iii) its dose-dependent toxicity. Indeed, elevated systemic (endocrine) levels of cytokines are nonphysiological (because cytokines are by definition paracrine factors) and hence accompanied by deleterious effects. Thus, systemic injection of IL-2 causes vascular leak syndrome due to bystander damage of IL-2Rαβγ⁺ endothelial cells and due to the unwarranted release of pro-inflammatory cytokines from T and NK cells (Epstein et al., 2003; Boyman and Arenas-Ramirez, 2019).

On one hand, the deleterious side effects of HD IL-2 must be attenuated to enhance its therapeutic index. Lower dosages and alternative delivery routes, such as s.c., i.m., or intralesional injections, have been tested (Tang and Harding, 2019). Only intratumoral administration outperformed standard HD i.v. IL-2, reaching up to 62% CR in melanoma patients with skin and soft-tissue metastases (Konrad et al., 1990; Palmer et al., 1993; Ravaud et al., 2002; Radny et al., 2003; Yang et al., 2003; Geertsen et al., 2004).

On the other hand, the therapeutic index of HD IL-2 may be improved by the selective stimulation of effector cells instead of Tregs. Multiple groups are introducing IL-2 in combinatorial regimens alongside immunomodulatory regimens, with the hope of improving efficacy and eventually reducing IL-2 dosage. Thus, ≥45 trials are active, recruiting patients in diverse oncological indications (http://www.clinicaltrials.gov) to evaluate HD IL-2 in combination with surgery, chemotherapy, radiotherapy, adoptive T cell therapy, anti–PD-1 or anti–CTLA-4, tumor-targeting mAbs, or cancer vaccines.

In parallel, attempts are ongoing to modify the pharmacological properties of IL-2. For example, pro-effector IL2Cxs in which an anti–IL-2 antibody sterically impedes the interaction between IL-2 and IL-2Rα, but allosterically enhances IL-2 affinity for IL-2Rβ, are being developed (Spangler et al., 2015). Antibodies with such characteristics include S4B6 (anti-mouse IL-2), as well as Mab602 and NARA-1 (anti-human IL-2; Boyman and Arenas-Ramirez, 2019). Preclinical studies in melanoma-bearing mice revealed superiority of S4B6 IL2Cxs over free IL-2 because: (i) the IL-2 interaction with IL-2Rα⁺ endothelial cells was disrupted and vascular leak syndrome prevented; and (ii) naive IL-2Rβγ⁺ CD8⁺ T and NK cells were preferentially amplified over IL-2Rαβγ⁺ Tregs. Encouragingly, IL2Cxs showed greater antitumor activity and some synergy with immune checkpoint inhibitors (Krieg et al., 2010; Létourneau et al., 2010; Arenas-Ramirez et al., 2016; Caudana et al., 2019).

A PEGylated IL-2 named NKTR-254 demonstrated a much more reduced affinity for IL-2Rα than for IL-2Rβ. A Phase 1/2 clinical trial has recently been completed in patients with late stage solid tumors (NCT02869295). Preliminary results indicate a favorable safety profile, a 10–30% shrinkage of the tumor burden in 23% of the patients, and an increase of tumor-infiltrating CD8⁺ T and NK cells with minimal impact on Tregs (Marin-Acevedo et al., 2018). Additional Phase 1–3 trials are now enrolling patients (Table 2).

Based on in vitro evolution assays and protein crystallography, Levin et al. (2012) designed the IL-2 mutein H9 (the IL-2 agonist that sourced H9-RETR). H9 harbors five mutations affecting core residues (i.e., L80F, R80D, L85V, I86V, and I92F). These substitutions induce a natural conformational switch responsible for a much stronger interaction with IL-2Rβ and an optimal downstream signaling activation not requiring IL-2Rα (Levin et al., 2012). The so called “superkine” H9 favored the expansion of CD8⁺ T and NK cells, demonstrated lower toxicity, and improved antitumor activity in several murine cancer models. Interestingly, IL-2 superkines demonstrated similar therapeutic activity as immunocomplexed WT IL-2 (Levin et al., 2012; Tang and Harding, 2019). In the clinic, the IL-2 mutein BAY50-4798 (Bayer) failed to demonstrate sufficient efficacy in advanced melanoma and renal cancer (ORR < 5%), leading to its discontinuation (Margolin et al., 2007). Retrospectively, dual introduction of an IL-2Rα bias plus the N88R mutation (which disfavors IL-2/IL-2Rβ interaction; Peterson et al., 2018) likely conferred an undesired pro-Treg cell activity to BAY50-4798.

Immunocytokines, which are hybrid proteins composed of WT or mutant IL-2 fused to Fc domains or tumor-targeting mAbs, are being developed. Signs of efficacy were observed in preclinical melanoma models treated with IL-2–anti-ganglioside GD2 (Becker et al., 1996a,b). The humanized version of such a construct, called EMD273063 (Merck), demonstrated a weak efficacy in metastatic melanoma (ORR = 7.1%; Albertini et al., 2012). Additional immunocytokines already completed clinical evaluations such as an anti-EpCAM-IL-2 (EMD273066/tucotuzumab celmoleukin), an anti-single/double-stranded DNA-(D20T)IL-2 mutein (EMD521873), and a fibronectin-targeted L19 diabody-IL-2 (Darleukin), with no or only marginal responses (Johannsen et al., 2010;Eigentler et al., 2011; Albertini et al., 2012; Connor et al., 2013; Gillessen et al., 2013; Laurent et al., 2013; van den Heuvel et al., 2015). However, as a possible exception to this rule, intralesional therapy of metastatic melanoma with a combination of recombinant IL-2 protein with the human mAb fragment L19 (L19–IL-2) together with another similar protein in which IL-2 is replaced by TNF-α (L19-TNF-α) showed a 53.8% ORR and induced abscopal effects (Danielli et al., 2015). An alternative strategy consists in fusing IL-2 with IL-2Rα to sterically prevent interactions with the high-affinity receptor. Such a compound, ALKS4230, is being tested in two Phase 1/2 trials in combination with the anti-PD1 antibody pembrolizumab (Table 2).

Another procedure applied to chimeric antigen receptor T cell technology is exemplified by the ortho–IL-2/ortho–IL-2Rβ system. It consists of a mutated IL-2Rβ, ortho–IL-2Rβ, which no longer binds native IL-2 but its mutant form, ortho–IL-2. The expression of ortho–IL-2Rβ on chimeric antigen receptor T cells confers selective proliferation following ortho–IL-2 administration, avoiding bystander activation of other IL-2R⁺ cells like Tregs (Sockolosky et al., 2018).

Finally, Garcia’s team generated a computationally designed IL-2/15 hybrid molecule of 100 amino acids that differs from the natural IL-2 not only in sequence but also in topology, ultimately preserving its ligation to IL-2Rβγ but not to IL-2Rα. This so called “neoleukin” Neo-2/15 showed greater antitumor activity and tolerance than WT IL-2 in preclinical experiments (Silva et al., 2019).

Layer after layer, we keep uncovering the complexity of IL-2 biology and incrementing IL-2 therapy. Following decades of deceptions, we may finally meet broader success. In this line, standard therapy with nonmodified recombinant IL-2 protein still has room for improvement, as illustrated by its increased therapeutic index following an intralesional route. Regarding the innovative approaches embodied by IL-2 muteins such as superkines or antagonists, immunocomplexes, and immunocytokines, drawbacks have emerged. They include the intrinsic immunogenicity and poor bioavailability of muteins, the potential disassembly of immunocomplexes in vivo, or the unexpected depletion of T cell subsets targeted by IL-2 mutein-Fc immunocytokines (Vazquez-Lombardi et al., 2017). Additional rounds of preclinical research and clinical investigations will likely be necessary to unchain the full therapeutic potential of agents targeting the IL-2/IL-2R system. Most importantly, immunomodulatory combination therapies, particularly with immune checkpoint inhibitors, have proven preclinical efficiency. In the clinic, administration of HD IL-2 appears essential for the efficacy of cancer treatments relying on the adoptive transfer of ex vivo expanded autologous T cells (e.g., tumor-infiltrating lymphocytes, T cells with an engineered TCR, or a chimeric antigen receptor) with remarkable objective response rates witnessed in melanoma (34–56%), lymphoma (80–100%), and leukemia (67–100%; Rosenberg and Restifo, 2015; Boyiadzis et al., 2018). In this setting, IL-2–based therapeutics could emancipate as adjuvant agents that will enhance the efficacy of established immunotherapeutics.

G. Kroemer is supported by the Ligue contre le Cancer (équipe labelisée); Agence National de la Recherche—Projets blancs; Agence National de la Recherche under the frame of E-Rare-2, the ERA-Net for Research on Rare Diseases; Association pour la Recherche sur le Cancer; Association “Le Cancer du Sein, Parlons-en!,” Cancéropôle Ile-de-France; Chancellerie des Universités de Paris (Legs Poix), Fondation pour la Recherche Médicale; a donation by Elior; European Research Area Network on Cardiovascular Diseases Project on Metabolic Therapy for Managing Diastolic Heart Failure; Gustave Roussy Odyssea, the European Union Horizon 2020 Project Oncobiome; Fondation Carrefour; High-end Foreign Expert Program in China (GDW20171100085), Institut National du Cancer; Institut National de la Santé et de la Recherche Médicale Hétérogénéité Tumorale et Ecosystème; Institut Universitaire de France; LeDucq Foundation; the LabEx Immuno-Oncology (ANR-18-IDEX-0001); the Recherche Hospitalo-Universitaire en santé Torino Lumière; the Seerave Foundation; the Site de Recherche Intégrée sur le Cancer (SIRIC) Stratified Oncology Cell DNA Repair and Tumor Immune Elimination; and the Site de Recherche Intégrée sur le Cancer Cancer Research and Personalized Medicine.

G. Kroemer has been holding research contracts with Bayer Healthcare, Genentech, Glaxo Smyth Kline, Institut Mérieux, Lytix Pharma, PharmaMar, Sotio, and Vasculox. G. Kroemer is the chair of the Scientific Advisory Board of Medicenna and a member of the Board of Directors of the Bristol Myers Squibb Foundation France. G. Kroemer is a scientific co-founder of everImmune and Samsara Therapeutics. The other authors declare no competing financial interests.

Author contributions: J.G. Pol and P. Caudana wrote the manuscript. J.G. Pol and J. Paillet made the tables and figures and wrote the legends. E. Piaggio and G. Kroemer designed the plan of the review and edited the manuscript.