Substituted arylsulphonamides as inhibitors of perforin-mediated lysis

The structure-activity relationships for a series of arylsulphonamide-based inhibitors of the pore-forming protein perforin have been explored. Perforin is a key component of the human immune response, however inappropriate activity has also been implicated in certain auto-immune and therapy-induced conditions such as allograft rejection and graft versus host disease. Since perforin is expressed exclusively by cells of the immune system, inhibition of this protein would be a highly selective strategy for the immunosuppressive treatment of these disorders. Compounds from this series were demonstrated to be potent inhibitors of the lytic action of both isolated recombinant perforin and perforin secreted by natural killer cells in vitro. Several potent and soluble examples were assessed for in vivo pharmacokinetic properties and found to be suitable for progression to an in vivo model of transplant rejection.


Introduction
Perforin is a 67 kDa, calcium-dependent glycoprotein expressed by only the natural killer (NK) cells and cytotoxic T lymphocytes (CTLs) of the mammalian immune system [1,2]. These "killer lymphocytes" utilise the pore-forming ability of perforin as a critical component of the granule exocytosis pathway; the principal mechanism used by NK and CTL cells for tumour immunosurveillance and as a defence against viral infection and intracellular pathogens [3]. Identification of a target cell by an effector cell results in the formation of an immune synapse whereupon CTL (or NK) secretory granules polarise to the site of contact. These granules contain both perforin and a group of pro-apoptotic serine proteases known as granzymes, and upon fusion with the CTL plasma membrane, release their luminal contents into the synapse [2,4]. Perforin performs a key role in this process because entry of the granzymes required for cell death into the target cell cytosol is solely dependent on its presence [1,5].
Although perforin is synthesized and secreted into the immune synapse as a monomer, it rapidly binds to the target cell membrane through its calcium-dependent C2 domain [6,7] and oligomerises into large transmembrane pores composed of approximately 24 perforin monomers. This process was elucidated using a combination of the perforin X-ray crystal structure and cryoelectron microscopy to reconstruct an entire perforin pore [8]. Electron microscopy, X-ray crystallography and functional studies have also shown that the process involves electrostatic interactions which include a salt bridge formed between R213 on the 'front' surface of one monomer interacting with E343 on the 'back' surface of the adjacent monomer [9]. Similarly, mutational studies reveal that D191, which is immediately adjacent to R213, makes interactions that are key to oligomerisation and that substitution with a bulky hydrophobic residue (D191V) abrogates this process [9].
Until recently, the precise mechanism of granzyme entry into the target cell was debated, but it is beyond any doubt that the pore-forming activity of perforin is indispensable. In essence, secreted perforin forms large (18 nm diameter) transmembrane pores on the surface of the target cell, through which the granzyme monomers (4 nm diameter) diffuse into the cytosol [10,11]. Once internalised the granzymes cleave key substrates to initiate rapid apoptotic death [5,10e13]. Unlike the granzymes, which are encoded by many genes and are, therefore, subject to considerable redundancy of function, the gene encoding perforin (PRF1) is present as a single copy in all mammals. Gene-targeting studies in mice [1] and naturally occurring disease-causing mutations in humans [14,15] confirm that perforin deficiency cannot be compensated by any other protein. This makes perforin an ideal target for therapeutic intervention.
While perforin is a key component of the immune response, inappropriate activity has also been implicated in a number of human immunopathologies and therapy-induced conditions. These include cerebral malaria, insulin-dependent diabetes, juvenile idiopathic arthritis and postviral myocarditis [16e18], as well as therapy-induced conditions such as allograft rejection and graft versus host disease [19e21]. Our current goal is to seek small molecule inhibitors of perforin as potential immunosuppressive agents for the treatment of autoimmune diseases and other conditions characterised by dysfunction of this pathway. This should be a highly selective strategy since perforin is expressed exclusively by CTL and NK cells, in contrast to approaches using conventional immunosuppression treatments which indiscriminately depress immune function [22e24].
Based on an initial hit from a mass screen [25], we have previously designed and optimised inhibitors of perforin that can; (i) block recombinant purified perforin, (ii) block perforin delivered by intact NK cells and, (iii) withstand incubation in serum (e.g. 1; Fig. 1

) [26e30].
While these compounds appeared highly promising, replacement of the 2-thioxoimidazolidinone moiety that contained a potential Michael acceptor and showed variable toxicity toward perforin-producing NK cells proved problematic. This issue was only overcome when we amalgamated our own finding that an aryl sulphonamide could act as a bioisosteric replacement [30] with a strategy implemented by GSK workers, where a thiazolidinedione was replaced with a pyridyl-linked benzenesulphonamide to give 2 [31]. This approach resulted in a new series of benzenesulphonamide-based perforin inhibitors, exemplified by 3, which were potent, soluble and essentially non-toxic toward NK cells [32]. In the following report we extend our study to explore whether it is possible to further modulate activity and physicochemical properties through variation of the sulphonamide linker, linker position, and substitution on the central pyridine ring and terminal benzene (Fig. 2).

Chemistry
The majority of the target compounds were constructed from right to left starting with our previously published key iodide 75 [32] (Scheme 1). Suzuki reaction of 75 with a variety of commercially available aminopyridine boronates under standard conditions gave the required amine intermediates 76e79 which were subsequently reacted with a range of substituted aryl sulphonyl chlorides. The 5-amino-3-pyridine derivative 78 [32] in particular was employed in the preparation of all compounds in Table 3. One exception was where the central pyridine ring was replaced with a benzene; in this case the Suzuki step was carried out with 2methyl-5-nitrobenzeneboronic acid, the nitro compound (80) hydrogenated to give the amine (81), which was then reacted with either 2,4-difluorobenzenesulphonyl chloride or 2pyridinesulphonyl chloride to afford 13 and 15 respectively. Finally, amido-linked compound 8 was prepared by reaction of 78 with 2,4-difluorobenzoic acid chloride.
In a smaller number of cases, mostly those examples with substitution on the central pyridine ring, the target compounds were effectively synthesized from two halves; the fully elaborated left-hand side benzenesulphonamide subunit (e.g 82e85) and key iodide 75 as the right-hand side (Scheme 2). The intermediate bromides 82e85 and 91e93 were prepared from a variety of commercially available aryl sulphonyl chlorides and substituted 3aminopyridines (or anilines) under standard conditions. In the case of 85, the sulphonamide NH was methylated with NaH and MeI in DMF to give 86, and for 91e93 where protection of this NH was required for the subsequent coupling to be successful, the alkylation was carried out with (chloromethoxy)ethane to give 94e96. All bromides were converted to the corresponding boronates 87e90 and 97e99 under palladium-catalysed conditions using bis(pinacolato)diboron and KOAc in DMSO, then finally reacted in a Suzuki step with iodide 75 to introduce the thiophene-N-methylisoindolinone subunit (6, 10e12, 17, 18, 20). Where required (for 17,18,20), deprotection was carried out under acidic conditions. A limited number of "reverse" sulphonamides were also prepared (Scheme 3). In the case of target compound 7, intermediate bromide 100 was prepared from 2,4-difluoroaniline and 5bromopyridine-3-sulphonyl chloride. Protection of the sulphonamide was not required and conversion to the boronate 101 and subsequent Suzuki coupling with 75 to give 7 proceeded smoothly. Likewise, bromide 102 was prepared from 2-aminopyridine 3bromo-4-methylbenzenesulfonyl chloride, converted to the boronate 107 and therein coupled with 75 to afford the final product 16. For bromides 103 and 104, protection of the sulphonamide NH (105, 106) was required for the sequential borylation (108, 109) and Suzuki reactions to proceed in good yield, and afforded targets 14 and 19 respectively in good yield.

Inhibition of recombinant perforin-mediated lysis
In our initial report describing the discovery of benzenesulphonamide-based inhibitors of perforin [32], analogue design focussed on exploration of the thiophene and N-methylisoindolinone subunits which comprise the right-hand side of the molecule. For the current study we sought to further optimise potency and physicochemical characteristics through manipulation of the central pyridine and sulphonamide linker, as well as employing a wide range of substituents on the left-hand benzene (or aryl) ring. Table 1 shows a group of compounds that explore the effect of changing the position of the sulphonamide link to the pyridine, the location of the pyridine nitrogen, and modification/replacement of the sulphonamide itself. By moving the sulphonamide link from the 5-position (3) to the 4-position (4), activity is abolished. If the sulphonamide is retained in the preferred 5-position and the pyridine nitrogen moved to the 4-position (5), activity is lost 4-fold compared to the lead, 3 (IC 50 s ¼ 4.70 and 1.17 mM respectively). The requirement for a free NH in the sulphonamide was probed through methylation of 3 to give N-methyl compound 6. It appears likely that this acidic hydrogen is required for interaction with the target protein since a 14-fold drop in activity to IC 50 ¼ 16.1 mM was observed. A "reverse" sulphonamide (7) is however still acceptable, with less than a 2fold reduction in activity (IC 50 ¼ 2.24 mM). Finally, replacement of the sulphonamide with a carboxamide (8) results in complete loss of potency, further supporting our hypothesis that the sulphonamide NH is required in the linker for optimal activity.
The effect of variation about the central pyridine ring was investigated next (Table 2). Direct replacement of the pyridine (3) with a benzene (9) resulted in a loss of activity from IC 50 ¼ 1.17 mM to 5.74 mM. Introduction of a 2-fluoro-(10) or 2-chloro-substituent Scheme 1. Reagents and conditions: (i) Boronate, Pd(dppf)Cl 2 , EtOH/toluene, 2 M Na 2 CO 3 , reflux; (ii) H 2 , 60 psi, 1:1:1 EtOH/EtOAc/THF, RT, 3 h; (iii) 2,4-Difluorobenzenesulfonyl chloride or pyridine-2-sulfonyl chloride, pyridine, 0e45 C; (iv) 2,4-Difluorobenzenoic acid chloride, pyridine, 0e45 C, 16 h. (11) to the pyridine gave similar activity to 3 (IC 50 s ¼ 1.03 and 1.99 mM respectively), while the electron-donating 2-methoxysubstituent (12) gave an analogue which was somewhat poorer (IC 50 ¼ 3.56 mM). Benzene-linked compounds (13e19) were then explored to determine if substitution on the benzene ring could improve activity. Introduction of a single methyl group on 9 to give 13 resulted in complete loss of potency, however at this point we reasoned that the increased lipophilicity and reduced solubility associated with the presence of two benzene rings (9 and 13) could be improved if we effectively moved the central pyridine of 3 to the other side of the sulphonamide link (2-pyridylsulphonamide compounds 14e20). This enabled us to explore a small number of compounds containing substitution on the central benzene ring, without further detriment to the overall physicochemical properties. Although far from a comprehensive list, the introduction of methyl (15; IC 50 ¼ 15.4 mM), fluoro (17; >20 mM), trifluoromethoxy (18; >20 mM) and methoxy (19; 9.68 mM) groups proved unsuccessful. The effect of reversing the sulphonamide orientation is consistent with   Table 3.
Clearly substitution on the benzene ring is required, with perhaps because all these halogens are less electron-withdrawing than fluorine, a concept that we explore in more detail below. A different trend was observed when the electron-donating substituents methoxy (37e40) and trifluoromethoxy (41e43) were employed; here the meta-position was favoured over ortho-and para-. The two meta-substituted examples 38 (2.56 mM) and 42 (1.42 mM) showed significantly better activity in comparison to other positional isomers. We then surveyed a variety of electronwithdrawing substituents by preparing compounds 44e59. As a group, these broadly paralleled the fluorine-substituted compounds discussed above, where the ortho-or para-positional isomers showed superior potency over the corresponding metaexamples. More specifically, for the trifluoromethyl-(44e47), ester-(51e54) and sulphonyl-(56, 57) substituted compounds, the ortho isomer was best (IC 50   benzene ring and through to the sulphonamide, enhancing interactions with the protein and resulting in improved activity. Hybrid compounds 60e65 were also prepared to investigate whether the effects of individual substituents could be combined. The resulting activities were neither synergistic nor additive, bringing no further gain to the overall potency. A set of four compounds (66e69) with a heterocycle (pyridine or thiophene) linked to the sulphonamide were also prepared. The preference for the heteroatom to be located directly next to the sulphonamide bond was clear with the 2-pyridyl and 2-thiophenyl compounds 66 and 68 (both IC 50 s ¼ 1.07 mM) far superior to the corresponding 3linked isomers 67 and 69 (15.13 and 12.51 mM respectively).
Finally, a set of compounds containing a variety of substituted heterocycles were prepared (70e74), but with the exception of the 4-oxazole 70 (IC 50 ¼ 3.05 mM), none showed much promise.

Advanced assessment of selected compounds
Having shown that a range of benzenesulphonamides block lysis by recombinant perforin, a subset of promising examples was identified to test for inhibitory effect on the lytic action of whole NK cells. Compounds were selected based on potency, and included several for which the Jurkat IC 50 s were >20 mM, to further validate our use of this higher throughput screen as our primary assay. The inhibitors were co-incubated with KHYG1 human NK cells in medium for 30 min at room temperature, 51 Cr-labelled target cells were added, and the resulting level of chromium release used to determine residual lytic activity and thus degree of inhibition. The use of whole NK cells to deliver perforin provides a more realistic model of conditions in vivo compared to isolated recombinant protein which acts indiscriminately. Recognition of aþ target cell, formation of a synaptic cleft, and release of the granular contents into the cavity between effector and target are all required elements for lysis to occur. Confirmation that the observed level of inhibition is due to blocking the activity of perforin rather than non-specific killing of the effector cell was also sought by measuring the viability of the NK cells 24 h later. Our lead compound for the current work and most potent compound from our previous study [32], 2,4-difluorobenzene 3, is included as a reference point (Table 4). One notable omission from this table is the potent 4-carboxylic acid-substituted compound 55 as this was toxic  to the NK cells and therefore the degree of inhibition was unable to be determined.
All of the compounds with Jurkat IC 50 s < 10 mM showed excellent suppression of NK-cell mediated killing of labelled target cells (68e96% at 10 mM), however this activity did not correlate exactly with their potency against isolated recombinant protein. This finding may reflect the varying ability of the respective inhibitors to access perforin located in a synaptic cleft within a complex biological milieu. Two examples with halogen on the central pyridine ring (10,11) and the 2-nitrobenzene compound 58 were particularly effective in blocking NK cell action (93, 95 and 96% inhibition respectively). Results from the set of compounds with IC 50 s > 20 mM broadly validated our use of this assay as a primary screen, with four of six examples showing no potency in either assay (16,17,47,57). Compound 13 demonstrated 50% inhibition at 10 mM, perhaps not surprisingly given that the only structural change from 3 was the replacement of a bridging pyridine with a benzene ring. The one unexpected outlier was the 3-cyano compound 49 which, although it had poor activity against isolated recombinant perforin, had excellent activity against perforin produced by whole NK cells. The NK cells also retained excellent viability across all examples, consistent with our previous findings for this series [32] and in contrast to earlier reported classes [26e28]. Preliminary physicochemical data was collected on the same (active) subset of compounds in order to assess their potential for progression to in vivo pharmacokinetic (PK) studies (Table 5). Following conversion to the corresponding sodium salts the solubility varied widely, with the 2,4,6-trifluorobenzene (26) and 4cyanobenzene (50) analogues being highly soluble, while the presence of 2-fluoropyridine (10), 2-nitrobenzene or the more lipophilic trifluoromethylbenzene group (44, 45) had a negative impact on solubility. All examples tested showed good stability in aqueous solution over 24 h, however results were more varied in the presence of human, rat and mouse microsomes. While 10, 11, and 58 showed acceptable stability (>70% parent after 30 min) across all three species, the remaining compounds (3, 26, 45, 50, 59 and especially 44) showed moderate to poor stability with human microsomes. This data in combination with poor solubility resulted in the elimination of 44 and 45 from consideration for the in vivo PK studies reported in section 2.4 below.

In vivo pharmacokinetics
The in vivo PK parameters were measured for seven compounds selected on the basis of the in vitro assessment described above.
Plasma pharmacokinetics were determined in male CD-1 mice for compounds 3, 10, 11, 26, 50, 58 and 59 (Table 6). Blood samples were collected at 5e8 time-points after dosing the compounds at 10 mg/kg in a solution of 20% hydroxypropyl-b-cyclodextrin by intraperitoneal (IP) injection. For analysis of the samples, chromatographic conditions were optimised by HPLC for each compound of interest and an internal standard. A liquid chromatography with tandem mass spectrometry (LC-MS/MS) method was then developed and validated for quantitation of each analyte.
Compound 3 possessed the best half-life (12 h) but showed much lower C max and AUC than other examples. While 10, 11 and 26 had shorter half-lives (4.5e9.5 h) they also showed the highest C max (236, 149 and 124 mmol/L) and exposures (AUC ¼ 2885, 2364 and 1019 mmol/L*h respectively). 4-Cyanobenzene (50) and both nitrobenzene isomers (58, 59), had even shorter half-lives, however the maximum concentration and overall exposure reached was still superior to the compound with the overall longest half-life (3).

Conclusions
Between our previous report [32] and the current study, we have explored in detail how changes made on a benzenesulphonamide-based template affects perforin inhibitory activity. Analysis of the resulting SAR shows that although a range of variations were explored throughout the molecule, the isoindolinone, thiophene, pyridine and sulphonamide link of 3 are probably close to optimal, while there is some tolerance for substitution on the central pyridine ring and the terminal benzene  ring. A smaller panel of compounds was selected based on the following criteria; in vitro potency against isolated recombinant perforin and whole human NK cells, lack of toxicity against NK cells, solubility and stability (aqueous and microsomal). For this group the in vivo PK parameters were assessed to select potential candidates for evaluation in a mouse model of transplant rejection. We sought acceptable half-life (potentially impacts dosing frequency), C max and AUC to ensure sufficient exposure to maximise our chances of achieving efficacy. Other key parameters taken into consideration for selection of the final in vivo candidates were in vitro potency and solubility. While the in vivo mouse efficacy studies are carried out using IP dosing, ultimately this product would be administered to human patients intravenously, meaning that sufficient solubility in a formulation close to physiological pH is crucial. Given the above criteria, one example from the panel that appears to possess a suitable balance of properties is the 2-chloro pyridine 11. This compound shows excellent potency (without toxicity) in vitro, good solubility, stability across all three species of microsomes and acceptable PK properties. Compounds 58 and 3 are also worth consideration; 58 because it is the most potent compound of the set, and 3 based on a combination of potency, solubility and half-life. The next step will be to conduct studies to determine if these compounds are capable of preventing transplant rejection in a mouse model.

Chemistry
Elemental analyses were performed by the Microchemical Laboratory, University of Otago, Dunedin, NZ; values are indicated by the symbols of the elements and were within ±0.4% of the theoretical values. Melting points were determined using an Electrothermal Model 9200 and are as read. Several compounds were examined as sharp-melting solvates, on which elemental analyses were determined. NMR spectra were measured on a Bruker Advance 400 MHz spectrometer and referenced to Me 4 Si. Mass spectra were recorded either on a Varian VG 7070 spectrometer at nominal 5000 resolution or a Finnigan MAT 900Q spectrometer. All final compound purities were determined to be >95% by HPLC on an Alltech Alltima C18 column (3.2 Â 150 mm, 5 mm) eluting with 5e80% MeCN/45 mM NH 4 HCO 3 .
Iodide 75 [32] (500 mg, 1.41 mmol) and 5-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)pyridin-2-amine (465 mg, 2.11 mmol) were dissolved in a mixture of toluene (8 mL) and EtOH (4 mL). A solution of 2 M Na 2 CO 3 (2 mL) and Pd(dppf)Cl 2 .CH 2 Cl 2 (57 mg, 0.07 mmol) were added and the entire mixture heated at reflux under N 2 for 2 h. Upon cooling, the desired product precipitated from the reaction mixture, was isolated by filtration, and washed with H 2 O, CH 2 Cl 2 and MeOH. No further purification was required, giving 76 as a green-yellow solid (326 mg, 72%); mp (MeOH/ CH 2 Cl 2 ) 252e254 C. 1   To a solution of 76 (120 mg, 0.37 mmol) in dry pyridine (12 mL) under N 2 at RT, was added 2,4-difluorobenzenesulphonyl chloride (159 mg, 0.74 mmol) in CH 2 Cl 2 (1.5 mL) dropwise over 5 min. The mixture was stirred at 45 C under N 2 for 16 , and the solvent then removed under reduced pressure. The reaction was quenched with a little water and the resulting solid collected by filtration and washed with water and Et 2 O. Purification was carried out by trituration with hot CH 2 Cl 2 /MeOH solution to give 4 as a pale brown solid (65%); mp (CH 2 Cl 2 /MeOH) 269e272 C. 1  In some cases a bis-sulphonamide was also formed; here a second step was introduced where the crude product was treated with a 1:1 mixture of 1,4-dioxane and 2 M NaOH. The monosulphonamide resulting from subsequent acidification of the reaction mixture was isolated by filtration, washed well with water, and dried. Purification was carried out by flash column chromatography (MeOH/CH 2 Cl 2 gradient).

Inhibition of perforin-mediated lysis of Jurkat cells
The ability of the compounds to inhibit the lysis of labelled nucleated (Jurkat T lymphoma) cells in the presence of 0.1% BSA was measured by the release of 51 Cr. Jurkat target cells were labelled by incubation in medium with 100 mCi 51 Cr for 1 h. The cells were then washed three times to remove unincorporated isotope and re-suspended at 1 Â 10 5 cells per mL in RPMI buffer supplemented with 0.1% BSA. Each test compound was pre-incubated to concentrations of 20, 10, 5, 2.5 and 1.25 mM with recombinant perforin for 30 min with DMSO as a negative control. 51 Cr labelled Jurkat cells were then added and cells were incubated at 37 C for 4 h. The supernatant was collected and assessed for its radioactive content on a gamma counter (Wallac Wizard 1470 automatic gamma counter). Each data point was performed in triplicate and an IC 50 was calculated from the range of concentrations described above. where x is the point the inhibitor intersects with the DMSO on a curve. In the example shown ( Fig. 3) "x" is the x-intercept corresponding to the point on the DMSO control curve that yields the same level of 51 Cr release as the test compound at an E/T ratio of 16:1.

Toxicity to KHYG1 NK cells
The toxicity assay was carried out in exactly the same manner as the killing assay above, but instead of adding the labelled K562 target cells, 100 mL of RPMI 0.1% BSA was added. Cells were incubated for 4 h at 37 C and then washed Â3 in RPMI þ 0.1% BSA. Cells were then re-suspended in 200 mL of complete medium and incubated for 18e24 h at 37 C. Trypan blue was added to each well. Viable (clear) cells and total (clear þ blue) cells were counted, and the percentage of viable cells was calculated compared to DMSO treated cell control (% viability).