New, Low–Molecular Weight Chemical Compounds Inhibiting Biological Activity of Interleukin 15

Chronic overproduction of IL–15 contributes to the pathogenesis of numerous inflammatory and autoimmune disorders. Experimental methods used to reduce the cytokine activity show promise as potential therapeutic approaches to modify IL–15 signaling and alleviate the development and progression of IL–15–related diseases. We previously demonstrated that an efficient reduction of IL–15 activity can be obtained by selective blocking of the specific, high affinity subunit alpha of the IL–15 receptor (IL–15Rα) with small–molecule inhibitors. In this study, we determined the structure–activity relationship of currently known IL–15Rα inhibitors in order to define the critical structural features required for their activity. To validate our predictions, we designed, analyzed in silico, and assessed in vitro function of 16 new potential IL–15Rα inhibitors. All newly synthesized molecules were benzoic acid derivatives with favorable ADME properties and they efficiently reduced IL–15 dependent peripheral blood mononuclear cells (PBMCs) proliferation, as well as TNF–α and IL–17 secretion. The rational design of IL–15 inhibitors may propel the identification of potential lead molecules for the development of safe and effective therapeutic agents.


Introduction
Interleukin 15 (IL-15) is a member of the common receptor gamma chain (γ c ) family, which also includes IL-2, IL-4, IL-7, IL-9, and IL-21. IL-15 is widely expressed and acts pleiotropically on many immune and non-immune cell types affecting their development and function [1]. To deliver its signal, IL-15 uses a heterotrimeric receptor composed of a specific, high affinity subunit alpha of the IL-15 receptor (IL-15Rα), IL-2/IL-15 specific receptor beta (IL-2/IL-15Rβ, CD122) and a common cytokine receptor gamma (γ c , CD123). Depending on the arrangement of the receptor units, the IL-15 signal can be delivered either in cis or trans mode. In cis-presentation, IL-15Rα and IL-2Rβ/γ c are expressed on the surface of the same cell while in trans-mode, which dominates in vivo, IL-15Rα presents IL-15 to neighboring cells bearing IL-15Rβ/γ c .
Errors leading to IL-15 overproduction directly contribute to the pathogenesis of numerous inflammatory and autoimmune disorders, e.g., rheumatoid arthritis, psoriasis, multiple sclerosis, celiac disease, type 1 diabetes, and alopecia areata. Increased concentrations of IL-15 are also linked to T-cell leukemias and graft rejection. Targeting IL-15 has been shown to be a valuable therapeutic method in many preclinical studies and clinical trials, however, to date none of the potential IL-15 inhibitors have been approved for clinical use. The experimental approaches used to inhibit cytokine activity include soluble IL-15Rα [2][3][4], antibodies inhibiting IL-2/IL-15Rβ [5] or IL-15 [6,7], a peptide inhibitor of γ c [8] or a modified IL-15 molecule with competitive antagonist activity [9]. We have previously demonstrated that selective blockage of IL-15Rα with small molecule inhibitors also leads to an efficient reduction of IL-15 activity. The efficacy of this approach correlates well with the significance of IL-15Rα in controlling IL-15 action. The accumulated evidence shows that IL-15 acts predominantly not as a monomer, but in complex with IL-15Rα. IL-15Rα is required as an intracellular chaperone for IL-15 secretion, complexed with IL-15 it prolongs the cytokine half-life, it is also necessary for optimal signaling through the IL-2Rβ/γ c dimeric receptor [10].
In the current study, we attempted to define the critical structural features of IL-15Rα inhibitors required for their activity against IL-15Rα. We based our research on previously identified, first potential IL-15Rα inhibitors selected on the basis of in silico modelling, pharmacophores screening, and prediction of ADME properties [11]. Furthermore, to validate our estimates, we designed, synthesized, and subjected to in vitro analysis a group of 16 new compounds presented in this paper.
All tested molecules reduced IL-15 dependent responses with significantly higher efficacy compared to currently known IL-15Rα inhibitors. The data obtained show that the applied approach allows the design of small molecular inhibitors of IL-15 Rα and could provide the results necessary to progress toward the successful development of a therapeutic agent.

Structural Characterization of Active and Inactive IL-15 Rα Inhibitors from the Group of Benzoic Acid Derivatives
Among the previously described molecules identified through in silico modeling and pharmacophores screening as putative small molecule IL-15Rα inhibitors, a vast majority, i.e., 14 out of 16 molecules, are benzoic acid derivatives [11] (Table 1). However, only half of them, that is seven molecules, exhibit inhibitory activity against IL-15, as demonstrated in biological in vitro tests (Table 1A). In our attempts to identify structural features necessary for these compounds activity, we have revealed that most of the active molecules (R11-R16) were benzoic acids derivatives that contained either acyclic-succinate or maleate or cycliccyclohexyl or norbornane substituents bearing free carboxylic group. More specifically, R11 and R13 were aminomethylbenzoic acid derivatives, R16 was a derivative of p-aminobenzoic acid, while R12 and R15 were derivatives of 3,5-diaminobenzoic acid.
The very high prevalence of benzoic acid derivatives in the group of the IL-15Rα inhibitors is not surprising. The X-ray structure of the IL-15R/IL-15 complex reveals that the interaction between interleukin and its receptors is stabilized mainly by two salt bridges between two arginine moieties of IL-15Rα and two glutamic acid residues of IL-15 (Arg26-Glu53 and Arg35-Glu46), located approximately 8-9 Å from each other (Figure 1a) [12].
During our previous in silico search for potential IL-15Rα inhibitors we specifically targeted to disrupt these interactions; therefore, most of the identified compounds had either two chemical groups with a formal +1 charge with high affinity to IL-15 or two moieties bearing a formal -1 charge with the high affinity to IL-15Rα [11]. The high prevalence of compounds with these moieties is likely due to the fact that carboxylic groups are the most common negatively charged moieties in all organic systems and the presence of benzoic acid produces relatively rigid molecules able to target both arginine residues of IL-15Rα which are relatively far apart. It is also worth noting that in the structure of the entire IL-15/IL-15R quaternary complex the targeted region of IL-15Rα has the most basic character, as there are in total four basic residues (Lys17, Arg24, Arg26 and Arg35), therefore binding of the newly designed inhibitors to other parts of the IL-15/IL-15R complex, such as the β or γ subunits, is unlikely [13].  During our previous in silico search for potential IL-15Rα inhibitors we specifically targeted to disrupt these interactions; therefore, most of the identified compounds had either two chemical groups with a formal +1 charge with high affinity to IL-15 or two moieties bearing a formal -1 charge with the high affinity to IL-15Rα [11]. The high prevalence of compounds with these moieties is likely due to the fact that carboxylic groups are the most common negatively charged moieties in all organic systems and the presence of benzoic acid produces relatively rigid molecules able to target both arginine residues of IL-15Rα which are relatively far apart. It is also worth noting that in the structure of the entire IL-15/IL-15R quaternary complex the targeted region of IL-15Rα has the most basic character, as there are in total four basic residues (Lys17, Arg24, Arg26 and Arg35), therefore binding of the newly designed inhibitors to other parts of the IL-15/IL-15R complex, such as the β or γ subunits, is unlikely [13].

Design and In Silico Analysis of New Potential IL-15Rα Inhibitors
Advances in virtual screening approaches and a better understanding of the 3D structure and molecular determinants of the specific high-affinity interactions between IL-15 and its heterotrimeric receptor complex have led to the discovery of several novel small molecules that effectively reduce IL-15 activity either by interfering with IL-15/IL-2Rβ or γc [14]. However, despite the appealing advantages of inhibiting IL-15 by selectively blocking IL-15Rα, there has been little progress in developing small molecule inhibitors of this receptor subunit. In the attempt to design highly active IL-15Rα inhibitors, we based our approach on recognition of features characteristic of currently known active and inactive IL-15Rα inhibitors. As shown in Table 1, putative IL-15Rα inhibitors (R11-R16) are the benzoic acid derivatives, which contain an amide bond in the side chain due to acylation of the amino group with dicarboxylic acids, such as succinic (R15, R16), maleic (R12), 1,2-cyclohexanedicarboxylic (R11) and 2,3-norbornanedicarboxylic (R13) acids. We aimed to generate non-chiral compounds to simplify analytical procedures and avoid potential issues with their synthesis and purification. Structures of the newly synthesized compounds are presented in Figure 2.

Design and In Silico Analysis of New Potential IL-15Rα Inhibitors
Advances in virtual screening approaches and a better understanding of the 3D structure and molecular determinants of the specific high-affinity interactions between IL-15 and its heterotrimeric receptor complex have led to the discovery of several novel small molecules that effectively reduce IL-15 activity either by interfering with IL-15/IL-2Rβ or γ c [14]. However, despite the appealing advantages of inhibiting IL-15 by selectively blocking IL-15Rα, there has been little progress in developing small molecule inhibitors of this receptor subunit. In the attempt to design highly active IL-15Rα inhibitors, we based our approach on recognition of features characteristic of currently known active and inactive IL-15Rα inhibitors. As shown in Table 1, putative IL-15Rα inhibitors (R11-R16) are the benzoic acid derivatives, which contain an amide bond in the side chain due to acylation of the amino group with dicarboxylic acids, such as succinic (R15, R16), maleic (R12), 1,2-cyclohexanedicarboxylic (R11) and 2,3-norbornanedicarboxylic (R13) acids. We aimed to generate non-chiral compounds to simplify analytical procedures and avoid potential issues with their synthesis and purification. Structures of the newly synthesized compounds are presented in Figure 2. The free energy of binding (ΔGbind) values for the new molecules were within the range of −11.9 and −14.6 kcal/mol which indicated a very strong binding required for drug-like compounds ( Table 2).   The free energy of binding (∆G bind ) values for the new molecules were within the range of −11.9 and −14.6 kcal/mol which indicated a very strong binding required for drug-like compounds ( Table 2). The highest ∆G bind value of -14.60 kcal/mol and the lowest binding constant (K i ) of 19.8 pM were observed for 7a ( Table 2).
The structure of the IL-15/IL-15Rα complex (PDB code: 2Z3Q) is mainly stabilized with two salt bridges (Arg26:IL-15Rα with Glu53:IL-15 and Arg35:IL-15Rα with Glu46:IL-15) and two strong hydrogen bonds: Arg24:IL-15Rα with the Glu53 backbone oxygen atom (IL-15) and the Arg35 backbone oxygen atom (IL-15Rα) with Tyr26:IL-15. Thus, to interfere with the formation of the IL-15/IL-15Rα complex the potential antagonist should target Arg24, Arg26, and/or Arg35 residues of the receptor ( Figure 3). A similar strategy was used previously by us to design a set of ligands with high affinity to IL-15Rα and disrupting IL-15/IL-15Rα binding [11]. As shown in Figures 1 and 3, carboxylic groups of 7a strongly interact with Arg24/Arg26/Arg35 residues of IL-15Rα and may destabilize IL-15/IL-15Rα complex by hindering interactions between Arg26 from IL-15Rα and E53 from IL-15. The binding poses of the remaining new compounds are presented in the Supplementary Information (Figures S1-S16). Interestingly, we predict that all of the studied compounds bind to at least two of the three crucial Arg residues (Arg24, Arg26, or Arg35). Additionally, in some of them, the carboxylic group of ligands makes a strong salt bridge to Lys17, located also close to the IL-15/IL-15Rα interface ( Figure 1). Based on these results and the relatively high Gibbs free binding energy values obtained for all investigated compounds we can expect that in each case we should observe at least a moderate disruption of the IL-15/IL-15Rα interactions. Table 3 presents the main results of the computational ADME prediction. As shown, all new molecules exhibited drug-like properties. Only minor problems were indicated for 7a, 7e, and 7h: no primary metabolites were indicated for these molecules placing them outside of the range of 95% of drugs. Furthermore, the low values of predicted Caco-2 cell permeability identified for all of the molecules gave rise to one Jorgensen's rule of three violation and indicated their possible low oral availability.
all of the studied compounds bind to at least two of the three crucial Arg residues (Arg24, Arg26, or Arg35). Additionally, in some of them, the carboxylic group of ligands makes a strong salt bridge to Lys17, located also close to the IL-15/IL-15Rα interface (Figure 1). Based on these results and the relatively high Gibbs free binding energy values obtained for all investigated compounds we can expect that in each case we should observe at least a moderate disruption of the IL-15/IL-15Rα interactions.  Table 3 presents the main results of the computational ADME prediction. As shown, all new molecules exhibited drug-like properties. Only minor problems were indicated for 7a, 7e, and 7h: no primary metabolites were indicated for these molecules placing them outside of the range of 95% of drugs. Furthermore, the low values of predicted Caco-2 cell permeability identified for all of the molecules gave rise to one Jorgensen's rule of three violation and indicated their possible low oral availability.    Interestingly, six of the new putative IL-15Rα inhibitors were previously described in the scientific literature . Compounds 6a, 6b, 6c, and 7a, 7b, 7c were used to develop the quantitative structure-activity relationship model of acetylcholinesterase inhibitors as potential drugs for Alzheimer's disease [16]. Compounds 6a, 6b, and 6c were also tested as potential antibacterial and antifungal agents [17,18], while 7a, 7b, and 7c were shown to express strong activity as human carbonic anhydrase isoenzymes I and II (hCA I and hCA II) inhibitors [19]. Based on virtual screening, 6b was identified as an inhibitor of the oncogenic fusion protein RUNX1/ETO [20], a fusion protein comprising leukemiainitiating transcription factor that interferes with the RUNX1 function [21], and a weak inhibitor of the SH2 domain of the tyrosine kinase P56 LCK [22].

The Effect of Novel Benzoic Acid Derivatives on PBMC Viability, IL-15-Dependent PBMC Proliferation and TNF-α and IL-17 Release
All cell types constituting peripheral blood mononuclear cells (PBMC), i.e., B cells (~15%), T cells (~70%), monocytes (~5%), and natural killer (NK) cells (~10%), respond to IL-15. Thus, we used PBMC to assess the biological effectiveness of novel benzoic acid derivatives as potential IL-15 inhibitors. To exclude non-specific effects linked to cell death, all of the synthesized compounds were first screened for their cytotoxicity. No cytotoxic effect was observed for evaluated molecules up to 5 mM (data not shown). In all further experiments, the new compounds were used at 5 µM concentration.
PBMCs do not proliferate spontaneously in vitro, but they divide and release TNF-α and IL-17 in response to IL-15 stimulation [4]. As shown in Figure 4, all the tested compounds significantly inhibited IL-15 depended cell proliferation and strongly inhibited IL-15-dependent PBMC release of TNF-α and IL-17, respectively. The new molecules exerted their biological activity at 5 µM concentration which is significantly lower as compared to the previously reported benzoic acid derivatives, for which the active concentrations were in the range of 50-200 µM (Table 1A) [11]. Importantly, the observed inhibitory effect was superior also to cefazolin, a derivative of 7-aminocephalosporanic acid and a firstgeneration cephalosporin antibiotic cefazolin which was identified as a small-molecule inhibitor of IL-15Rα and showed promising potential in human proof-of-concept study as a repurposing drug candidate for psoriasis therapy [23]. In vitro, cefazolin reduced IL-15dependent TNF-α and IL-17 synthesis at 50 µM and 300 µM concentrations, respectively.

Computational Methods
The crystal structure (PDB code: 2Z3Q [12]) of the IL-15/IL-15Rα has been prepared by removing the IL-15 cytokine and water molecules (if present) and adding hydrogen atoms with AutoDockTools 4 [24]. Atomic interaction energy grids were calculated using

Computational Methods
The crystal structure (PDB code: 2Z3Q [12]) of the IL-15/IL-15Rα has been prepared by removing the IL-15 cytokine and water molecules (if present) and adding hydrogen atoms with AutoDockTools 4 [24]. Atomic interaction energy grids were calculated using probes corresponding to each atomic type found in the ligand, at 0.375 Å grid resolution. We have used four different boxes: a 126 Å cubic box centered on the protein and including the entire receptor, and three smaller boxes (50 × 80 × 40 Å, 40 × 70 × 30 Å and 30 × 60 × 30 Å) centered on the protein-protein binding region. In the docking part, we have used Autodock 4.2 [24] with the Genetic Lamarckian Algorithm and standard options, including 100 dockings per compound and 5,000,000 energy evaluations per docking [25]. In all docking experiments, each ligand has been treated in a fully flexible manner with the Gasteiger partial charges added by AutoDockTools 4. In the first phase of docking, the protein has been treated as a completely rigid model, also with Gasteiger partial charges. In the second phase of docking, after finding the likely docking region the protein has been treated as a rigid model, but with four residues (Lys17, Arg24, Arg26, and Arg35) described in a fully flexible manner. For each study's compound, we have estimated the free energy of binding as well as the inhibition constant K i (at 298.15 K) using the approach implemented in Autodock, which uses semiempirical force field to evaluate the sum of differences in energies between unbound and bound states of the ligand. All 2D figures of binding sites were prepared using Discovery Studio Visualizer (Dassault Systemes, 2015). ADME properties were computationally modeled using QikProp ver. 4.6 software (Schrodinger Inc., New York, NY, USA) using default settings.

Compound Synthesis
Fine chemicals and solvents were purchased from commercially available vendors and were used without further purification. Methyl 4-chloro-4-oxobutanoate (4) and methyl 4-chloro-4-oxo-2(Z)-butenoate (5) were obtained according to the literature procedures [26,27]. TLC analyses were performed on plates precoated with silica gel (Merck 60 F 254 , 0.25 mm). HPLC analyses were performed on a Waters system equipped with a LiChrosphere ® 100 RP-8 HPLC column and PDA 2996 detector (190-400 nm, 1.2 nm) using ACN:0.1% TFA in H 2 O (50/50 v/v) as a mobile phase. Samples were dissolved in the mobile phase at a concentration of 0.1 mg/mL (7c and 7f) or 0.25 mg/mL (the remaining compounds). Analyses were performed at the wavelengths given for each separately described compound. Melting points were determined throughout DSC measurements (DSC822e cell) with an IntraCooler (Mettler Toledo GmbH, Gießen, Germany) in the nitrogen atmosphere.
Samples (5-10 mg used as received) in a standard (40 µL) aluminum pan were heated from 25 to 300 • C (10 • C/min). IR spectra were recorded in a KBr pellet containing ca 1 mg of the tested compound and ca 200 mg of KBr using Nicolet iS10 spectrometer in the range of 4000-400 cm −1 (resolution 4 cm −1 ). NMR spectra were recorded on a Varian VNMRS-600 spectrometer in 25 • C for DMSO-d 6 solutions using TMS as the internal standard in the following ranges (δ ppm): 0-14 ( 1 H), 20-180 ( 13 C) and -146 to -124 ( 19 F); chemical shifts (δ) are quoted in ppm and coupling constants (J) in Hz. Mass spectrometry (MS and HRMS) was carried out using AutoSpec Premier (Waters) spectrometer with EI ionization. Absorbance measurements in colorimetric in vitro tests were performed using SPECTROstar Nano spectrometer at the wavelength suggested by tests manufacturers.

Synthesis of Succinic Derivatives
Synthesis of compounds 6a-6f and 6h was performed according to Scheme 1A. Succinic derivatives 6b-f and 6h were obtained according to the same protocol, using appropriate aminobenzoic acid (1b-f, h) and succinic anhydride (2) 11.33 (s, 1H) NH, 8.48

Statistical Analysis
Statistical significance was assessed by ANOVA with the Dunnett post hoc test. p values below 0.05 were considered statistically significant. Statistical analyses were performed using GraphPad Prism 9.3.1 software (GraphPad Software). Data were presented as the mean ± SEM from at least three independent experiments.

Conclusions
Abnormal expression or dysregulated IL-15 signaling plays a key role in the pathogenic development of numerous autoimmune and inflammatory diseases. First positive proofof-concept for the clinical effects of topical cefazolin, first-generation cephalosporin which was identified as a small molecule IL-15Rα inhibitor and used in the treatment of psoriasis confirms this approach [23]. In the present study, we determined the structure activity relationship of benzoic acid derivatives representing the predominant subset of the currently known IL-15Rα inhibitors and identified the structural requirements necessary for the anti-IL-15 activity. This allowed us to design a series of 16