The general synthetic route used to prepare the target compounds 7-12 is outlined in Scheme (1). Synthesis started with the conversion of sulindac, sulindac sulfide or sulindac sulfone (1-3) to the methyl esters, which were converted to the corresponding aldehydes (4-6) by treating with DIBALH. Finally, reductive amination of the aldehydes with various substituted amines afforded compounds 7-12 in moderate to good yields [33].

Scheme (1). Synthetic pathways to analogs 7-12. Reagents and conditions: (a) MeOH, SOCl2 (b) DIBAL-H, Toluene (c) NaBH4, MeOH, rt (d) Na(OAc)3BH, 1,2-DCE, rt.

Sulindac sulfide ethane amine 13 was prepared from sulindac sulfide (1) following the four-step sequence shown in Scheme (2). The sequence involved the reduction of the acid to the corresponding alcohol, which was then converted to an iodide by treating with tetrabutylammonium iodide, pyridine and trifluoromethanesulfonic anhydride [34]. The iodide was then converted to the corresponding azide using sodium azide followed by a Staudinger type reduction [35] to give sulindac sulfide ethane amine 13. Amine 13 was further transformed to amide 14 using standard peptide coupling conditions.

Scheme (2). Synthetic pathways to analogs 13-14. Reagents and conditions: (a) BH3/THF, 0 oC-rt (b) TBAI, Pyridine, CH2Cl2, Tf2O, -78 oC-rt (c) NaN3, MeCN, reflux (d) PPh3, rt (e) 2-Furoic acid, HBTU, Et3N, MeCN, rt.

Scheme (3). Synthetic pathways to analogs 14-59. Reagents and conditions: (a) Oxalyl chloride, DMF, CH2Cl2, rt (b) trimethylsilylazide, CCl4, rt-50 oC (c) AcOH/HCl (d) RCO2H, HATU, DIEA, MeCN (e) RSO2Cl, NMI, Pyridine (f) RCHO, Na(OAc)3BH, 1,2-DCE, rt

Methane amine analogs (16-17) of different sulindac derivatives were synthesized from their corresponding acids (1, 15) by treating with oxalyl chloride followed by trimethylsilyl azide and acetic acid [34]. These amines were further modified in three ways: 1) conversion to amides by reacting with various acids (18-33); 2) conversion to sulfonamides by reacting with sulfonyl chlorides and pyridine (34-47) and; 3) reductive amination of 16 or 17 with various aldehydes and Na(OAc)₃BH to afford bis amino substituted sulindac methane amines (48-61) Scheme (3) [33].

Using quantitative high-throughput screens (qHTS) all synthesized compounds were screened against three cancer cell lines: HT29 colorectal carcinoma, PC3 prostate and MDA-MB-231 breast cancer cell lines according to reported methods – see Supplementary Material, Appendix A. Screening results are summarized in (Tables 1-6).

Table (1) lists the anticancer activity of compounds 7-12 in colon, prostate and breast cancer cell lines. Compound 10, the amine analog of SSA, was found to be active, but slightly less potent than SSA. Compound 9 with a 4-methylthiobenzylidene ring at the C-1 position and 1-pipyridinylethylaminoethyl group at the C-3 position had activity very similar to 10. Among benzylaminoethyl analogs of sulindac sulfide, sulfoxide and sulfone (7, 11 and 12), the sulindac sulfone analog 12 displayed significant anticancer activity. The furan-2-ylmethylaminoethyl analog of sulindac sulfide (8) was active against HT29, PC3 and MDA-MB-231 cells, but was slightly less potent than SSA.

Sulindac sulfide ethane amine 13 Table (2) was prepared to study the role of substituents at the amino ethane group. Compound 13 was modestly less active than its 1-pipyridinylethyl and N,N-dimethylaminoethyl counterparts 9 and 10, but more active than its benzyl derivative 7 (Table 1). No significant differences were observed in the case of 13 and its furan-2-ylmethyl analog 8. These results demonstrate that appropriate amino ethane analogs show significant activity relative to sulindac and its metabolites as well as other analogs. Conversion of 13 to its reverse amide analog 14 slightly reduced potency by two fold, and the reverse amide was also modestly less potent than reduced analog 8.

Two analogs 16 and 17 Table (3) were prepared to study the effect of substituting an amino methyl linker for the amino ethyl linker in 13. Both analogs were less active than 13 against all three cancer cell lines, while their potencies were comparable.

Compounds 16 and 17 were further modified to reverse amide analogs 18-33 Table (4), sulfonamide derivatives 34-47 Table (5) and disubstituted amino derivatives 48-61 Table (6) to explore the role of different functional groups in anticancer activity. Compounds 18-33, which contain an amide moiety at the C-3 position, displayed significant anticancer activity in all the three cell lines except their furan-2-yl analogs 19 and 27. Importantly, 3,4,5-trimethoxybenzylidene-3-indenylmethyl-2-phenylacetamide (26) was the most effective and potent inhibitor of cancer cell growth among all the compounds tested, with CC₅₀ of <98.0 nM in colon and breast cancer cells and 520.0 nM in prostate cancer cells. Among all the sulfonamide derivatives 34-47, 4-thiomethylbenzylidine analogs 34-40 were more active than their 3,4,5-trimethoxybenzylidine analogs 41-47. Conversion of methaneamine linker to a series of disubstituted methaneamine compounds reduced potency, except for the pyridyl analogs 49 and 56.

Table 4. Screening data for compounds 18-33.

Cmpd	R, R1, R2 & R3	CC50 (µM)
		HT29	PC3	MDA-MB-231
18	R1 & R3 = H, R2 = SCH3, R = benzyl	3.56±1.25	18.86±9.71	4.06±1.97
19	R1 & R3 = H, R2 = SCH3, R = furan-2-yl	33.25±3.43	39.79±6.81	41.16±7.60
20	R1 & R3 = H, R2 = SCH3, R = >	3.64±0.21	4.81±0.99	7.24±0.56
21	R1 & R3 = H, R2 = SCH3, R = aminomethyl	5.49±0.21	7.27±0.31	7.30±0.62
22	R1 & R3 = H, R2 = SCH3, R =	5.98±0.24	6.03±0.41	7.43±0.87
23	R1 & R3 = H, R2 = SCH3, R =	4.39±0.29	6.21±0.49	6.70±0.71
24	R1 & R3 = H, R2 = SCH3, R =	4.23±0.16	5.16±0.50	6.81±0.63
25	R1 & R3 = H, R2 = SCH3, R =	3.15±0.17	6.66±0.42	4.35±0.19
26	R1, R2 & R3 = OCH3, R = benzyl	<0.098±0	0.52±0	<0.098±0.04
27	R1, R2 & R3 = OCH3, R = furan-2-yl	>50.00	>50.00	>50.00
28	R1, R2 & R3 = OCH3, R = methyl	6.44±1.40	23.26±2.61	5.77±1.55
29	R1, R2 & R3 = OCH3, R = aminomethyl	8.51±1.70	17.92±2.08	9.92±2.13
30	R1, R2 & R3 = OCH3, R =	11.05±1.35	26.53±6.01	11.94±2.27
31	R1, R2 & R3 = OCH3, R =	5.21±0.39	11.95±1.47	8.87±1.38
32	R1, R2 & R3 = OCH3, R =	4.87±0.11	8.11±0.41	7.30±0.73
33	R1, R2 & R3 = OCH3, R =	5.67±1.67	11.94±1.56	8.20±0.99

Most analogs in the presented series are lipophilic while also containing a positively charged amine. Lipophilic basic amines can accumulate in cell membranes if their lipophilic atom groups form favorable non-polar contacts with lipid chains while their charged amine is interacting with fatty acid head groups. Such association with lipids can lead to cytotoxicity through membrane destabilization. Therefore, we evaluated a number of representative compounds against BJ cells, a normal human foreskin fibroblast cell line, where cytotoxicity was determined at 10 µM drug concentration, as described previously [36]. Results of cytotoxicity evaluations for selected compounds are shown in Table S-1 (Supplementary materials: Appendix A). BJ cytotoxicity evaluation suggests that these compounds are not cytotoxic. In particular, compound 26 possessing the highest potency against all three cancer cell lines shows EC₅₀ > 22.73 µM in the BJ cytotoxicity assay. Further, Table S1 includes screening results against a panel of additional cancer cell lines performed at St. Jude Children’s Research Hospital for selected compounds (as described in Supplementary materials: Appendix A). Compound 26 stands out as having inhibitory activity against five acute lymphoblastic leukemia cell lines used on this panel (Table S1), and thus this compound may be considered a viable lead for further development. Compounds 20, 23, 24, 26, 32 and 56 were active against multiple leukemia cell lines showing a wide range of efficacy while also inhibiting the three (HT29, PC3, MDA-MB-231) cancer cell lines with CC₅₀ < 10 µM potency.

The binding mode of sulindac, a known COX-2 inhibitor was predicted based on the X-ray structure of COX-2 (PDB code 4COX) using Induced Fit docking as implemented in Schrödinger software Fig. (2). The carboxylate group of sulindac forms a hydrogen bond with Y355 and highly favorable, salt bridging/hydrogen bonding interactions with R120, while the conformation of the R120 side chain is restrained by a salt bridge to E524. These favorable contacts are also present in the X-ray structure of COX-2 (PDB code 4COX) through the carboxylate group of co-crystallized indomethacin. Docking places the carboxylate of sulindac in overlapping position with that of indomethacin co-crystallized in COX-2.

The carboxylate functionality of sulindac was replaced by various basic amine containing groups in the presented amine derivative series, none of which are suitable for forming favorable polar interactions with Arg120. In addition to the loss of salt bridging and hydrogen bonding interactions, basic amine groups close to the guanidinium moiety of R120 would not be favorable. This is illustrated by superimposing a representative analog (10) onto sulindac in Fig. (2). Compound 10 could not be docked favorably into COX-2, even while treating residues in the binding site region as flexible. Compound 10 has a very similar structure to SSA (sulindac sulfide amide). The difference between their structures is that an amide group in SSA is replaced by a more basic methylene-amine group in compound 10 Fig. (1), Scheme (1). Superposition of compound 10 onto sulindac places this group in approximately overlapping space with the carboxylate of sulindac and near R120 Fig. (2). Interestingly, SSA inhibits human colon tumor cell lines while showing no activity against COX-1 and COX-2 [30]. The lack of COX-2 activity of SSA supports our analogous, model based prediction that compound 10 and its analogs containing basic amine substituents are unlikely to inhibit COX-2. Similar reasoning was incorporated in SSA analog designs previously as in our approach to design out COX-2 inhibitory activity through replacement of the carboxylate of sulindac with a variety of amine containing substituents [15].

Fig. (2). Sulindac (orange carbons) docked into the COX-2 active site and the superimposed compound 10 (green colored carbons).

The docked pose of sulindac suggests further vectors to explore for the design of new analogs inactive against COX-2. In the docked pose of sulindac the benzylidene moiety is accommodated in a deep-lying pocket, surrounded by largely non-polar residues as shown in Fig. (2): F381, L384, W387, F518, M522, V523, L352, as well as (residues not shown for clarity) G526, A527, Y385. Tight packing of residues around the benzene ring especially in the region of C3, C5 include the side chains of Y385, W387 and backbone atoms of M522, V523, which restrict pocket volume available for 3-, 5- substituents. The 4-methylsulfinyl substituent is fitted tightly in a small cavity forming a hydrogen bonding interaction with W387. Limited cavity space in this region suggests that 3-, 4-, 5-substitutions of the benzene ring may lead to steric hindrance, interfering with the binding of sulindac analogs to COX-2. Therefore benzene substitutions have been also explored in the presented analog series.

Favorable permeability and metabolic stability are key properties relevant to oral bioavailability that should be considered early on during lead optimization. In order to assess physicochemical properties of our most potent series (7-33) we correlated logD and molecular weight (MW) of the compounds and selected the analogs that map to an optimal ‘Golden Triangle’ region as shown in Fig. (3), as proposed based on results of previous studies [37, 38]. For example, Johnson and co-workers [37] used in vitro permeability data available in Caco-2 cells for 16,227 compounds and in vitro metabolic clearance data derived from human liver microsome (HLM) stability for 47,018 compounds for a correlation analysis with computed physicochemical properties [37]. Among physicochemical properties, logD was found to positively correlate with permeability, except at high logD values. Metabolic clearance was negatively correlated with logD and molecular weight (MW). Permeability and HLM stability data have been combined in order to deduce the optimal ranges for logD and MW which describe compounds that possess both, good permeability and metabolic stability properties. The optimal ranges map approximately to a triangular area (‘Golden Triangle’) in a plot of MW versus logD (with a baseline at MW 200 between logD -2 to 5 and apex at MW 450 between logD 1 – 2). Similar results have been obtained based on the logD and Caco-2 permeability data analysis of 9,571 structurally diverse compounds [38]. Therefore, we chose to correlate MW and logD for the compounds in our series and identify those that map to or near the optimal ranges. Fig. (3) illustrates several active analogs that map to the “goldilocks area” (13, 16, 17) or near the borderline of this optimal zone (21, 29, 30), which may be associated with desirable in vivo permeability and metabolic clearance properties based on the mentioned published findings. Table S2 (Supplementary Materials: Appendix A) lists computed physicochemical properties for the series 7-33.

Fig. (3). Active analogs mapping onto or near the optimal ‘Golden Triangle’ zone (yellow) are shown on a molecular weight (MW) versus LogD correlation plot.

Screening against HT29 colorectal carcinoma, PC3 prostate and MDA-MB-231 breast cancer cell lines are described. Results of additional cancer cell line screens and cytotoxicity data is summarized in Table S1. Computed physicochemical properties of compounds 7-33 are listed in Table S2.

General experimental methods, synthetic procedures and analytical data are provided.