A first point to clarify is to identify which form of circulating IL-15 is present in the serum of HD. Indeed, it has been postulated that all circulating IL-15 found in the serum of human and mice exists as heterodimeric complex associated with the sIL-15Rα, thus excluding the presence of circulating cytokine in its monomeric form.30 This assertion, in our opinion, is only partially plausible in mice in which sera there is a constant detection of sIL-15Rα that binding to circulating IL-15 may generate, at least in part, a fraction of the sIL-15/IL-15Rα complex.20 30 31 In addition, a recent paper shows that, in murine sera, the free monomeric sIL-15 form is more abundant than the sIL-15/IL-15Rα complex, this latter representing only 14% of the circulating cytokine.5 By contrast, in mice stimulated with lipopolysaccharide or polyinosinic-polycytidylic acid (pI:C), sIL-15 was mostly associated to IL-15Rα, as a sIL-15/IL-15Rα complex.26 These data strongly suggest that in ‘resting mice’ only a minor fraction of IL-15 is represented by the soluble complex, which may be necessary to feed some lymphoid subsets. On the contrary, in inflammatory and infectious settings most of circulating IL-15 is converted to the soluble complex isoform thus acting as a rapid alert response to pathological situations.

In human HD sera, the monomeric IL-15 serum concentration ranges between 1 and 6 pg/mL32–35 and its role is to foster specific lymphoid subsets.10–13 Indeed, in pathological conditions, such as lymphodepleted patients, serum level of monomeric sIL-15 rapidly raises to supraphysiological concentrations due to the absence of the natural cellular sinks. The cut-off point, defining elevated supraphysiological concentrations, is 20 pg/mL.35 These values can be detected in several inflammatory diseases29 32–35 and in arthritis patients, where they are associated to more severe disease.32

Concerning the detection of sIL-15/IL-15Rα in human sera, the few papers that have investigated this point showed that sIL-15/IL-15Rα has been detected at concentrations ranging between 43 and 231 pg/mL in the plasma of eight out of eight metastatic melanoma patients undergoing lymphodepletion30 which, per se, could amplify the phenomenon. By contrast, Bouchaud et al have studied 37 HD showing that their serum levels of the sIL-15/IL-15Rα were barely detectable.29

In figure 2A, we have investigated both the presence and the level of the sIL-15/IL-15Rα complex in plasmas of 35 HD: about one in three of the plasmas resulted positive with values ranging from 0.3 ng/mL up to 5 ng/mL. Plasmatic absence of the soluble complex may simply reflect a constant consumption by PB natural cellular sinks for this cytokine. Alternatively, the complex is not secreted in order to avoid the excessive stimulation of immune cells in comparison with that induced by monomeric sIL-15.14–18 On the other hand, plasmatic detection in one third of the ‘HD’ could be due to benign unapparent inflammatory conditions at the time of the blood withdrawal. Indeed, in animal models, the transient secretion of the sIL-15/IL-15Rα complex is elicited by inflammatory stimuli and viral infections thus acting as response to specific situations requiring rapid and powerful activation of cytolytic subsets.36 Interestingly, two of the enrolled HD, subsequently resulted to be affected by chronic renal diseases.

In humans, differently from mice,29 30 sIL-15Rα is not detectable or it is present at very low levels in the sera of HD but it is present in patients affected by different diseases.37 40 For instance, T-cell large granular lymphocyte leukemia (LGL) is characterized by increased serum levels of sIL-15Rα, composed by the natural shed form of sIL-15Rα as well as the alternatively spliced form of sIL-15Rα suggesting different cellular sources. In addition, IL-15Rα expression is upregulated in patients' PB mononuclear cells as well. Upregulation of IL-15Rα could protect IL-15 from proteolysis, decrease the IL-15 response threshold of the cells, and sIL-15Rα complexed with monomeric circulating IL-15 (sIL-15) may generate the soluble complex. Both events may participate to progression of the disease.37 Moreover, LGL cells express tp-IL-15,42 that via signaling in cis and in trans 14 18 22, could facilitate leukemic cells expansion.42

IL-15 can act as growth and viability factor for malignant T cells in lymphoma patients. Skin lesions and PB T-cells of cutaneous T-cell lymphoma patients show over-expression of both IL-15 mRNA and protein.43 44 In these patients, survival of malignant CD4⁺ T-cells is dependent, in the early stages, on IL-15 supplied from the microenvironment, but during disease progression, malignant cells, through autocrine IL-15 production, may autonomously support their growth and survival.43 In human T-cell lymphocytic virus leukemia, viral Tax protein induces the production of both IL-15 and IL-15Rα and this IL-15 autocrine loops leads to leukemia progression.44 45 In multiple myeloma, malignant plasma cells display functional IL-15 receptor with increased expression of the IL-15Rα chain and autocrine production of IL-15, these phenomena favor survival and proliferation of malignant cells independently from their microenvironment.46 All these data are summarized in (online supplemental table 1).

Increased levels of sIL-15Rα in the serum of patients with head and neck cancer and high intratumoral IL-15 concentration, are highly correlated with poor clinical outcome.40 47 Clinical studies report a correlation between high intratumoral IL-15 concentrations and poor clinical outcome also in patients with lung cancer.48

Concerning human melanomas, we had previously shown that IL-15 produced by primary melanoma cell cultures, was involved in tumor escape mechanisms and in the activation of pro-inflammatory signals.8 49 In analogy, injection of small concentrations of human recombinant IL-15 (rhIL-15) in nude mice caused the development of much more aggressive tumors on subcutaneous implantations of human melanoma cells.50 The above-mentioned data,8 40 47–50 together with the detection of high levels of sIL-15/IL-15Rα in the serum of metastatic melanoma patients,30 strongly suggests that the generation of intratumoral and/or circulating sIL-15/IL-15Rα complexes may contribute to the development of a tumor microenvironment favorable to tumor progression and immune escape.

However, there are exceptions to the above described behavior since, in some solid tumors, the IL-15/IL-15R complex may also play an antineoplastic role. For instance, in mice implanted with various tumor cell lines, the sIL-15/IL-15Rα complexes are abundant in the interstitial fluid of freshly implanted tumors and their expression anticipates the infiltration of tumor infiltrating lymphocytes that play a major role in tumor regression. By contrast, the relative levels of sIL-15/IL-15Rα complexes are low in mice with advanced tumors and are associated with tumor progression. However, in advanced tumors the sIL-15/IL-15Rα complexes can be rescued by intratumoral activation of stimulator of interferon genes leading to tumor regression.51 These are intriguing results, even if they have been obtained in mice implanting a limited number of cell lines. It could be interesting to evaluate these parameters studying the development of chemically or virus induced tumors, looking for the density of stromal cells producing the sIL-15/IL-15Rα complexes, together with the quali-quantitative characteristics of infiltrating immunocompetent cells. Concerning human tumors, triple negative breast cancer cells overexpress IL-15Rα, and the concomitant expression of IL-15 in IL-15Rα-expressing cells, stimulates an autocrine signaling cascade in the absence of IL-2Rβ- and IL-2Rγ_c chains, promoting cell proliferation, migration and blocking apoptosis. Nevertheless, in this subset of high-grade breast tumors, coexpression of IL-15 and IL-15Rα is associated with better survival outcomes. This is likely, due to the fact that IL-15 and IL-15Rα coexpressed on cancer cells could paracrinely activate infiltrating mononuclear cells potentially leading to an anti-tumor immune response.52

Intrarenal IL-15, produced by tubular and cortical cells, plays a major role in the preservation of epithelial homeostasis (for review see53). By contrast, renal clear carcinoma cells (RCCs) do not secrete IL-15,27 but express tmb-IL-1554 55 and do not express neither in vitro nor in vivo the γ_c chain and JAK3, its signaling partner.56 Thus, in RCCs, both recombinant IL-15 (rhIL-15) and tmb-IL-IL5 transduce an unbalanced signal promoting the epithelial mesenchymal transition, thus behaving as protumorigenic factors.54 56 By contrast, renal cancer stem cells express the IL-15Rα/IL-2Rβ/γ_c heterotrimeric receptor. This receptor, when stimulated by rhIL-15, promotes the generation of polarized, non-tumorigenic and IL-15 secreting epithelial cells that behave as ‘normal cells’ being capable of autocrine IL-15 production.53 57 Thus, IL-15 may display at the same time pro-tumor and antitumor effects depending on the subset of renal cancer cell involved. It is tempting to speculate that the induction of γ_c chain obtained in vitro in RCC cells using ERK1/2 inhibitors,53 in vivo could induce a modified microenvironment more suitable for IL-15 therapeutic use.

In prostate cancer (PC), IL-15 overexpression is associated with recurrence-free survival suggesting that IL-15 could represent a potential prognostic biomarker.58 Subsequently, it has been shown that IL-15 decreases PC cells migration and invasion in vitro and in vivo. In addition, the presence of IL-15 in the tumor microenvironment decreases proliferation and blood vessel formation, suggesting that IL-15 may display antitumor effects inducing inflammation and shaping the stroma.59 Immunostaining of 70 colon cancer biopsies and implantation of human colon cancer cells in nude mice show that IL-15 is produced by metastatic colon carcinoma cells and can induce hyperplasia in the mucosa adjacent to colon cancer, thus contributing to angiogenesis and tumor progression.60 In addition, IL-15 promotes in vitro the proliferation, motility, and invasiveness of human colon cancer cells as well as increase their resistance to apoptosis.61 By contrast, in human patients, it has been recently shown that loss of IL-15 expression in about 30% of metastatic patients correlates with lower T-cell density, decreased T-cell proliferation, high risk of relapse and decreased survival.62

The above-mentioned results (summarized in online supplemental table 2), show how the role of intratumoral IL-15 is complex, ambivalent and dependent on several, not fully elucidated, factors such as the type of IL-15 produced, the IL-15-Rα chain isoforms involved in sIL-15/IL-15Rα complexes, the presence of functional IL-15 receptor on tumor cells as well as their response to stromal and endogenous IL-15. Thus, it is difficult to foresee if circulating and intratumoral IL-15 will behave as friend or foe, at present we suspect that on the long term the role of foe will emerge.

Given the target cell specificity, IL-15 looks more promising than other cytokines in enhancing the efficacy of effector cells against tumors and viral infections.63–65 Indeed, rhIL-15 or the sIL-15/IL-15Rα used either alone or in combination with other treatments, have shown efficacy in several experimental tumor models66–69 (data summarized in table 1). However, these results should be revisited taking into account which species has been employed as source of IL-15 in the experiments. An underestimated point concerns the type of soluble complex employed in the different mouse models. Indeed, human IL-15/IL-15Rα heterodimers are highly effective without causing side effects16 17 70–75 By contrast, mouse sIL-15/IL-15Rα or human sIL-15/IL-15Rα in humanized mice are associated with important detrimental effects or generation of inefficient NK cells, highlighting the existence of negative feedbacks counteracting excessive NK stimulation.76–78 The above-mentioned discrepancies indicate that mouse cells are more sensitive to murine IL-15 than to human IL-15 and suggest the existence of species-related differences in the activity of IL-15 in mouse models. This must be taken into account when comparing the effects of IL-15 across species.79

Another point to consider is the target specificity of the different IL-15 isoforms. In this context, using a new series of transgenic mice delivering IL-15 as sIL-15 or complexed to IL-15Rα, distinct effects have been identified on NK or CD8⁺ T-cells.80 Thus, mature NK cells are rescued exclusively by monomeric IL-15 (sIL-15) that also mediates a powerful antimetastatic response in a model of mouse melanoma.80 These data underline the specific role of NK cells in the control of metastatic spread.80–82 In this model, even if the levels of IL-1580 are clearly supra physiological, it seems that, at least in mice, each IL-15 form could have specific targets.

Finally, we should also consider the IL-15-dosing regimens.

A first study employing transgenic mouse engineered to constitutively overexpress sIL-15 (mean±SEM, 186.7±41.8 pg/mL) exhibited early expansions of NK and CD8⁺ T-cells. Later, these mice developed fatal lymphocytic leukemia with a T-NK phenotype.83 These data suggest that in vivo prolonged exposure to supra physiological concentrations of IL-15 may be highly detrimental. More recently, two other studies investigated the impact of transient and prolonged in vivo stimulation by the sIL-15/IL-15Rα complex on NK and CD8⁺ T cells. Elpek and coll. have shown that 2 weeks stimulation of CD8⁺ T cells with sIL-15/IL-15Rα induced an activated phenotype, robust T-cell receptor stimulation and effector function. By contrast, this regimen caused an accumulation of mature NK cells with considerably impaired activation, cytotoxicity and proliferative activity.78 More recently, Frutoso et al, have shown in vivo that NK cells respond to a single stimulation cycle with sIL-15 or the sIL-15/IL-15Rα complexe, while they become hyporesponsive and fail to expand after a second cycle of stimulation. This could be due to the contemporary expansion of CD44⁺CD8⁺ T cells. These data highlights as an excessive stimulation of NK cells by sIL-15 and sIL-15/IL-15Rα complexes may induce NK hyporesponsiveness.84 Indeed, prolonged NK cell activation using ionomycin as well as repeated exposure to Toll-like receptors (TLRs) agonists and85 repeated injections of interferon (IFN)-α86 induced decreased cytolytic function and cytokine production.87 Both TLRs and IFN-α stimulation increase production of IL-15, IL-15Rα and/or of the soluble complex36 85 86 88 89 that are potentially responsible of NK cells exhaustion. These data would suggest that regimens based on prolonged stimulation with IL-15 and IL-15 superagonist should be avoided.

rhIL-15 is currently under investigation in several Phase I clinical trials (NCT02395822, NCT01385423, NCT03388632, NCT01572493 NCT01021059 NCT01369888. NCT01875601) in which the cytokine is administered employing different dosing regimens, including intravenous bolus and subcutaneous and continuous infusion of monomeric sIL-15 both in adult and pediatric solid tumors.90–94 There is also a number of clinical trials (NCT01727076, NCT01885897, NCT02523469), based on the utilization of an IL-15N72D/IL-15Rα-Fc super-agonist complex, termed N-803 (formerly Alt-803, from NantKwest, Culver City, California, USA) supposed to maximize expansion and priming of immune cells (NK and CD8⁺ T-cells).95 96 However, in the latter setting a drawback rises from the rapid consumption of the injected IL-15 agonists due to the rapid and huge expansion of the target immune cells.97 The above-mentioned clinical trials are summarized in table 2.

In addition, the repeated stimulation with rhIL-15 in patients with cancer may result detrimental. Indeed, a recent study has shown that one rhIL-15 infusion led to an increase in NK cell expansion and proliferation in the PB of patients, whereas a second infusion of rhIL-15 resulted in a decrease in NK cell responses.98 Moreover, two recent studies also show that repeated cycles of N-803 injection may lead to reduced biological responsiveness in NK cells of SIV⁺ macaque99 and, in a phase 1b trial, in patients with metastatic non-small cell lung cancer.100

These data open up two questions: NK hyporesponsiveness observed in cancer patients receiving repeated cycles of IL-15 or soluble complex is induced by the contemporary expansion of an inhibitory T cell subset as reported in mice84 or may be directly induced by IL-15 and/or the super-agonist complex? If the second hypothesis holds true, ex vivo protocols for the expansion of NK cells intended to be reinjected in patients with cancer as a complementary immunologic treatment should be revisited. Felices and coll. have investigated this point and their data indicate that continuous treatment of NK cells, over a 9-day period, with rhIL-15 at 20 ng/mL resulted in decreased viability, cell cycle arrest gene expression pattern, decreased function both in vitro and in vivo, and reduced tumor control.101 These data are truly alarming, since the IL-15 concentration employed (20 ng/mL) thought being supra-physiologic, is in the range of those employed by dozens of laboratories working on NK cells. These data are difficult to explain, but it is possible that this behavior occurs commonly at different degrees and may be underestimated in the routinely practice. An alternative explanation could reside in the proneness of recombinant IL-15, suspended in classical buffers, to rapidly undergo self-aggregation with loss of biological activity.102 Thus IL-15 suspended in classical buffers maintained for even few days at 4°C, may lose biological activity in an unpredictable way generating incertitude concerning the real concentration employed that could result partially inefficient in stimulating NK cells. In this respect, the use of an anti-cytokine aggregation buffer solves the above mentioned problems (unpublished results).

The presence of high level of the sIL-15/IL-15Rα complex in lymphodepleted melanoma metastatic patients could be essentially due to the lymphodepletion that would cause the loss of natural sinks leading to temporary increase of the plasmatic sIL-15/IL-15Rα complex. In order to solve this question, we analyzed the plasma from 35 non-lymphodepleted metastatic melanoma patients. We observed that one third of them resulted highly positive before anti-BRAF treatment and maintained this positivity after therapy showing long-lasting supra-physiological plasmatic levels of sIL-15/IL-15Rα only in a subset of patients (figure 2A). Since NK cells are major players in the control of the metastatic evolution,82 we wondered whether the high levels of the sIL-15/IL-15Rα complex, detected in the plasma of metastatic melanoma patients undergoing30 or not lymphodepletion, would display or not biological activity on NK cells. In the latter instance, high level of a soluble complex lacking biological activity29 39 could act as a decoy cytokine unable to provide homeostatic signals to their cellular sinks present in the PB.10–13 By contrast, constant high plasmatic levels of a soluble complex at high biological activity could induce suboptimal NK activation of: (1) PB-NK cells for instance in metastatic patients, or (2) NK cells pre-expanded ex vivo and destined to NK cell-based anti-tumor therapies.96 97 The latter issue could be tentatively solved by the in vivo silencing of the IL-15Rα-IC3 isoform responsible for the assembly of the soluble complex at high biological activity.41 In order to investigate the latter point we developed an in vitro protocol for mimicking this situation. For this purpose, NK cells were expanded during 12 days with IL-2, shortly starved and then refed for 8 days either with the hyperagonist IL-15clpx15 or with IL-2 (starved-Ctrl). At the same time, NK cells were continuously maintained in IL-2 without starvation and served as control (Ctrl). NK cells fed with Hyperagonist IL-15clpx exhibited a statistically significant reduction of cytotoxic effects on K-562 target cells as compared with Ctrl and starved-Ctrl NK cells (figure 2B), suggesting that Hyperagonist IL-15clpx could really cause direct functional NK impairment.

Finding a single agent-based therapy effective in causing complete tumor regression and long-lasting anti-tumor responses represents a real challenge owing to the multiple immunological barriers/escape mechanisms present in tumors. While having some efficacy in inducing tumor regression as a monotherapy, IL-15 isoforms show a greater potential when used in combination with other immunotherapies. Indeed, very efficient anti-tumor effect was detected in the preclinical pulmonary metastasis model using CT26 colon carcinoma cells, when N-803 was combined with the checkpoint inhibitors anti-CTLA-4 and anti-PD-L1. This treatment markedly expanded subsets of NK and memory CD8⁺ T cells. The improved NK cells function played a fundamental role in reducing tumor metastases and mice survival.103 A marked antitumor effect was obtained in combination therapies using sIL-15 or IL-15/IL-15Rα with checkpoint blocking antibodies targeting either PD-L1 or CTLA-4. In addition, combination of IL-15 superagonists (hyperagonist RLI and N-803) and blocking antibodies against PD-L1 and CTLA-4 resulted in remarkable improvement in antitumor responses leading to increased survival and reduced tumor growth in several preclinical models (for review see ref. 94). Moreover, in MHC class I-deficient melanoma, refractory to CD8⁺ T cell-mediated immunity, combination of IL-15 with TIGIT blockade could rescue antitumor NK cell-mediated cytotoxic activity and inhibit tumor metastasis in two mouse melanoma models.104 Recently, the group of Huntington identified the cytokine-inducible SH2-containing protein (CIS) as a critical negative regulator of IL-15 signaling in NK cells. Targeting CIS improved NK cell ex vivo proliferation and anti-tumor functions without any apparent secondary or off-target effects. Thus, the inhibition of CIS constitutes a novel method of unleashing the NK cell antitumor response with important perpesctives in clinical immunotherapy.105 Finally, an important synergistic antitumor activity was shown after administration of agonistic anti-CD40 Abs in combination with IL-15 in preclinical models of colon, prostatic and pancreatic cancers.67 106 107 These novel preclinical data support the scientific basis for clinical trials employing this combination immunotherapy strategy for the treatment of patients with cancer. The promising results of pre-clinical models led to the development of human protocols. For instance, the combination of ex vivo IL-15-activated NK cells with anti-NKG2A blocking mAbs revealed an effective strategy for the treatment of multiple myeloma patients.108

Overall, these data strongly support the notion that IL-15-based harnessing of NK cells, in combination with checkpoint inhibitors, may effectively unleash NK cells thus representing an effective therapeutic strategy, particularly in HLA-class I tumors that are ‘invisible’ to CD8⁺ T lymphocytes. This concept has relevance not only in adult malignancies, but also in the setting of several refractory or recurrent high-risk pediatric tumors.94