Di-, tri- and tetra-5'-O-phosphorothioadenosyl substituted polyols as inhibitors of Fhit: Importance of the alpha-beta bridging oxygen and beta phosphorus replacement.

BACKGROUND: The human FHIT gene is inactivated early in the development of many human cancers and loss of Fhit in mouse predisposes to cancer while reintroduction of FHIT suppresses tumor formation via induction of apoptosis. Fhit protein, a diadenosine polyphosphate hydrolase, does not require hydrolase activity to function in tumor suppression and may signal for apoptosis as an enzyme-substrate complex. Thus, high affinity nonhydrolyzable substrate analogs may either promote or antagonize Fhit function, depending on their features, in Fhit + cells. Previously synthesized analogs with phosphorothioadenosyl substitutions and "supercharged" branches do not bind better than natural substrates and thus have limited potential as cellular probes.
RESULTS: Here we link adenosine 5'-O-phosphates and phosphorothioates to short-chain polyols to generate a series of substrate analogs. We obtain structure-activity data in the form of in vitro Fhit inhibition for four types of analog substitutions and describe two compounds, inhibitory constants for which are 65 and 75-fold lower than natural substrates.
CONCLUSIONS: The best Fhit inhibitors obtained to date separate two or more 5'-O-phosphoromonothioadenosyl moieties with as many bond lengths as in AppppA, maintain oxygen at the location of the alpha-beta bridging oxygen, and replace carbon for the beta phosphorus.

Background

Loss of Fhit protein is among the earliest known events in the development of a variety of the most common and lethal human malignancies [1]. Loss of Fhit leads to cells that are deficient in programmed cell death and that form tumors in mice while Fhit reexpression in Fhit-cancer cells reduces tumorigenicity and restores programmed cell death [2-6]. Infection of Fhit +/- mice with viruses that re-express Fhit reduce cancer occurrence [7], apparently by killing pre-neoplastic cells that have lost the wild-type Fhit allele. Human Fhit protein, a member of the Fhit branch of the histidine triad superfamily of nucleotide-binding proteins, binds and hydrolyzes diadenosine polyphosphates such as ApppA and AppppA (1) into AMP plus ADP and ATP, respectively [8-10]. His96, which is responsible for covalent catalysis and more than 4× 106-fold of rate enhancement in ApppA hydrolysis [9,11-13], is nonetheless dispensable for ApppA-binding and tumor suppression, suggesting that Fhit function in tumor suppression depends on formation of an E-S complex [2,12]. If Fhit-substrate complexes promote tumor suppression by stimulating a pro-apoptotic effector, then Fhit inhibitors that resemble natural substrates may promote Fhit function. Similarly, Fhit inhibitors with normative features may antagonize Fhit function. Either class of compounds may be important in dissecting Fhit cell biology and regulating apoptosis.

Making use of a synthesis strategy to link adenosine 5'-O-phosphates and phosphorothioates to short-chain polyols [14], we evaluate four inhibitor parameters and obtain compounds, inhibitory constants for which are as much as 70-fold lower than natural substrates. Key features of the best candidate agonist compound 6b and candidate antagonist compound 12b are conservation of the length of the polyphosphate replacement, use of 5'-O-phosphorothioadenosyl residues, no replacement for oxygen at the location of the α-β bridging oxygen, and replacement of carbon for the β phosphorus. Compound 12b contains additional negatively charged substituents that may facilitate Fhit inhibition [15] while rendering the compound antagonistic to Fhit function in the cell. Covalent structures of AppppA (1) and compounds 6b and 12b are provided in Figure 1.

Figure 1

Structures of Fhit substrate 1 and Fhit inhibitors 6b and 12b.

Results and Discussion

A series of compounds 2–6 were synthesized and evaluated for Fhit inhibition that link two 5'-O-phosphorothioadenosyl or AMP groups with five reagents ranging in size from ethylene glycol to meso-erythritol. Sodium salts of each compound were titrated into assays of purified Fhit [16] with 1.8 μM fluorescent substrate, Appp-S-(4-4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacine-3-yl)methylaminoacetyl (ApppBODIPY) [10], and competitive Ki values were obtained by calculating the inhibitor concentration-dependence in reduction of kcat/Km(apparent) for substrate hydrolysis [10]. As shown in Table 1, inhibitors 2b and 3b with two-carbon diol linkers were either only as inhibitory as the natural substrate 1 or substantially less so. Inhibitor 4b with a four-carbon linker was as ineffective as 3b. Inhibitors 5a and 6a that substitute a central CH2-PO2-CH2 or CH2-CH(OH)-CH2 for the PO2-O-PO2 of 1 were as inhibitory as 1 is a good substrate. The phosphorothioate analogs, 5b and 6b, bound approximately 10 and 75 times better to Fhit than 1. Thus, conservation of bond-lengths between adenylate moieties of 1, 5 and 6 is conducive to binding Fhit.

Table 1

Fhit inhibition by AppppA analogs varied in interadenylate linker and phosphorothiolation.

		X = O	X = S
Comp.	R1	Ki (nM)	Ki (nM)
		a	b
1		2600	-
2		-	2600
3		-	86000
4		-	86000
5		1500	220
6		4700	35

Because phosphorothioate analogs of 5 and 6 were better inhibitors than the corresponding phosphates, phosphorodithioate analog 6c was prepared together with five additional compounds as phosphates and phosphorothioates. Phosphorodithioate 6c was a less effective inhibitor (Ki = 8600 nM, not shown in Tables) than corresponding phosphate 6a and phosphorothioate 6b and, as demonstrated in Tables 1 and 2, the phosphorothioate congener of every compound made as a phosphate and a phosphorothioate had a lower Ki value. The contribution of particular P-chiral phosphorothioate stereoisomers to inhibition has not been examined. When diadenosine 5',5"'-(P1, P2-methylene-P3-thio)-P1, P3-triphosphate[17] was crystallized with wild-type and mutant Fhit, the α-phosphorothioate group was found in the mutant but not wild-type active site, suggesting that α-phosphorothioate inhibitors may be slow substrates [12]. Similarly, Frey and co-workers found that Fhit slowly cleaves both Rp and Sp stereoisomers of γ-(m-nitrobenzyl)-adenosine 5'-O-(l-thiotriphosphate) with modest and similar beneficial effects on Km as compared to the corresponding phosphate [13].

Table 2

Fhit inhibition by AppppA analogs varied in phosphorothiolation and three modifications of glycerol.

				X = O	X = S
Comp	R1	R2	R3	Ki (nm)	Ki (nM)
				a	b
6	O	OH	H	4700	35
7	O	CH3	H	-	76000
8	O	NH3	H	-	3000
9	O	H	H	-	230
10	NH	H	H	-	3300
11	S	H	H	-	5200
12	O	-CH2-O-P-(X,O)-OAdo	-CH2-O-P-(X,O2)	1500	40
13	O	-CH2-O-P-(X,O)-OAdo	-CH2-O-P-(X,O)-OAdo	3200	65
14	O	-CH2O-P-(X,O2)	-CH2-O-P-(X,O2)	900	700
15	O	-O-P-(X,O)-OAdo	H	1500	78
16	NH	-O-P-(X,O)-OAdo	H	-	2700
17	O	-O-P-(X,O2)	H	420	110

Given the inhibitor activity of 6b, the contribution of the oxygen in the position of the α-β bridging oxygen of 1 was examined. As can be seen in Table 2, comparison of 9b with 10b and 11b and of 15b with 16b indicated that imido or sulfur substitutions for oxygen reduce enzyme inhibition by an order of magnitude. Derivatives of 6b with one or two additional functionalities on the central carbon were characterized. Analysis of 9b indicated that loss of the polar hydroxyl group reduces inhibitory activity and analysis of 7b and 8b indicated that nearly isosteric groups that are nonisoelectronic are not tolerated. The magnitude of these effects was surprising and may be related to altered conformations upon manganese coordination. Earlier, Blackburn and co-workers made "supercharged" methane-trisphosphonic acid AppppA analogs containing a central carbon from which three adenylate or phosphate moieties are bonded [15]. In work presented here, tripodal inhibitors 15a and 17a showed, respectively, micromolar and submicromolar efficacies while the phosphorothioate counterparts 15b and 17b showed 100 nM efficacy.

Derivatives of pentaerythritol, tetrapodal compounds 12,13 and 14, were also evaluated for Fhit inhibition. While compound 14b, containing two phosphorothioylated branches without adenosine, was barely a submicromolar inhibitor, compounds 12b and 13b, which contain respectively three and four CH2-phosphorothioadenosyl groups bonded to the central carbon were 40 nM and 65 nM inhibitors. Thus, while the simplest compound 6b was initially rendered less inhibitory by modification because its central hydroxyl was important for inhibition, addition of one or two CH2-phosphorothioadenosyl groups restored inhibition. We presume that compounds 6b, 12b and 13b with 2, 3 or 4 identical CH2-phosphorothioadenosyl groups can present a manganese bound 6b-like complex to Fhit in a similar manner. On the basis that Fhit binds diadenosine polyphosphates with one AMP group buried and the other adenosine solvent-exposed in a specific conformation [12], effective tri and tetrapodal inhibitors such as 12b, 13b and HC(ADP)3[15] may exist in solution predominantly in a conformation that resembles Fhit-bound AppppA.

On the basis of observations [12,13] discussed above, one would expect AppppA with two phosphorothioadenosyl substituents to be a slow substrate with a relatively low Km. However, carbon in place of the β phosphorus (for example in 6b) would make the leaving group the alkoxide – OCH2-CH(OH)-CH2O-phosphorothioadenosyl rather than α-thio ATP. Presumably, because normal Fhit products are mononucleotides with low pKa values, the enzyme does not have a group to protonate an alkoxide leaving group and thus substitution of carbon for the β phosphorus turns a substrate into an inhibitor.

Conclusions

Tumor suppression by Fhit is not destroyed by mutation of the nucleophilic His96 to Asn, a mutation that specifically reduces kcat[2,12]. Thus, evidence suggests that the proapoptotic function of Fhit depends on formation of an E-S complex. If the substrate-dependent signaling model is correct, then compound 6b, as a nearly isosteric AppppA analog with a low Ki, may promote Fhit signaling in Fhit+ cells. Compound 12b, which one would expect to bind Fhit with bulky phosphorothioate and phosphorothioadenosyl groups interfering with putative effector binding, may consequently prove to be antagonistic to Fhit function. Ongoing in vitro studies aim to evaluate the stereochemistry of phosphorothioyl binding to Fhit while in vivo studies test the effects of these compounds on programmed cell death. Phosphorothioate analogs 6b and 12b that preserve the α-β bridging oxygen but substitute carbon for the β phosphorus, achieving 65 to 75-fold binding advantages over AppppA, are expected to be critical for cell biological characterization.

Materials and Methods

Synthesis and Characterization

The synthesis and physicochemical characteristics of compounds 5, 6a and b, 12 through 15, and 17 have been described [14]. Compounds 2–5, 8b, 9b were obtained in the reaction of two equivalents of 5'-O-(2-thiono-1,3,2-oxathiaphospholane)-N6,N6,O2',O3'-tetrabenzoyladenosine (18) with one equivalent of ethylene glycol (for 2b); O,O-dimethyl-D,L-tartrate (for 3b); meso-erythritol (for 4b); O-methyl bis(hydroxymethyl)phosphinate (for 5); 2-amino-1,3-propanediol (for 8b); 1,3-propanediol (for 9b). To prepare compound 6c, two equivalents of 5'-O-(2-thiono-1,3,2-dithiaphospholane)-N6,N6,O2',O3'-tetrabenzoyladenosine (19) were condensed with one equivalent of glycerol. Compounds 7b, 10b, 11b, 16b were obtained in the reaction of N6,N6,O2',O3'-tetrabenzoyladenosine with 2-thiono-1,3,2-oxathiaphospholane derivatives of 2-methyl-1,3-propanediol (for 7b); 1,3-diaminopropane (for 10 b); 1,3-propanedithiol (for 11b) and l,3-diamino-2-hydroxypropane (for 16b). Compounds 12 through 14 were obtained in the reaction of tetra-oxathiaphosphothioylated erythritol (20) with two equivalents of -N6,N6,O2',O3'-tetrabenzoyladenosine (21) for 12; three equivalents of 21 for 13; four equivalents of 21 for 14. Compound 15 was obtained in the reaction of tri-oxathiaphosphothioylated glycerol (22) with three equivalents of compound 21[14].

Synthesis of 5'-O-(2-thiono-1,3,2-oxathiaphospholane)-N6,N6,O2',O3'-tetrabenzoyladenosine (18)

N6,N6,O2',O3'-tetrabenzoyladenosine (1 mmol) was reacted with 2-chloro-1,3,2-oxathiaphospholane (1.1 mmol) in pyridine solution (3 ml) in the presence of elemental sulfur (5 mmol). After stirring for 12 h at room temperature, the solvent was removed under reduced pressure and the crude product was purified by silica gel column chromatography using chloroform:hexane (8:2) as an eluent to provide 18 in 72% yield [31P NMR 105.2 ppm (d), FAB-MS (M-1) m/z 820].

Synthesis of 5'-O-(2-thiono-1,3,2-dithiaphospholane)-N6,N6,O2',O3'-tetrabenzoyladenosine (19)

N6,N6,O2',O3'-tetrabenzoyladenosine (1 mmol) was reacted with 2-chloro-1,3,2-dithiaphospholane [18] in pyridine solution (3 ml) in the presence of elemental sulfur (5 mmol). After stirring for 12 h at room temperature, the solvent was removed under reduced pressure and the crude product was purified by silica gel column chromatography using chloroform:hexane (8:2) as an eluent to give 19 in 68% yield [31P NMR 115.2 ppm, FAB-MS (M-1) m/z 836].

Synthesis of tetra-oxathiaphosphothioylated erythritol (20) or tri-oxathiaphosphothioylated glycerol (21)

2-chloro-1,3,2-oxathiaphospholane (8 mmol) was added to a suspension of elemental sulfur (10 mmol) in pyridine (10 ml). A solution of glycerol (1.5 mmol) or erythritol (1.5 mmol) in pyridine (1 ml) was then introduced to the reaction mixture and stirred for 12 h at room temperature. Crude products were purified by silica gel column chromatography using chloroform:hexane (7:3) to give 22 in 77% yield [31P NMR 105.6 ppm (m), FAB-MS (M-1) m/z 505] or 21 in 60% yield [31P NMR 105.1 ppm (m), FAB-MS (M-1) m/z 687], respectively.

Condensation of oxathiaphospholane derivatives with polyols

A mixture of 1,8-diazabicyclo(5,4,0) undec-7-ene with the corresponding polyol (one equivalent of 1,8-diazabicyclo(5,4,0) undec-7-ene per -OH function) in 1 ml acetonitrile solution was added to the solution of one molar equivalent of oxathiaphospholane derivative (18,19,20 or 21) dissolved in dry acetonitrile (5 ml). The reaction mixtures were stirred at room temperature for 4 hours and then solvent was removed under reduced pressure. Purified compounds were obtained by Sephadex A-25 ion-exchange chromatography using a linear gradient of ammonium bicarbonate buffer (pH 7.5) as eluent, and their physicochemical characteristics are given in Table 3.

Table 3

Physicochemical characteristics of compounds 2–17.

Comp.	31P NMR in D2O	MS-MALDI	Yield
No	H3PO4 chemical shifts [ppm]	(M-1) m/z	[%]
2	58.95; 57.64	752	24
3	58.82; 57.69	840	53
4	57.14; 56.92	812	50
5*	58.46; 58.24; 30.28	815	54
6a*	2.3	750	35
6b*	56.83	782	42
6c	115.86; 115.159	814	25
7b	57.64; 57.33	780	21
8b	59.13; 58.94	781	10
9b	57.32; 58.15	766	34
10b	60.65; 60.31	764	55
11b	71.85; 71.17	798	40
12*	56.94; 46.82	1267	20
13*	56.28	1516	35
14*	57.12; 46.71	1018	21
15*	57.56	1127	60
16b	60.17; 59.13; 58.49; 57.41	1125	15
17*	56.61; 45.95	877	37

*Compounds described in ref. [14].

Enzyme inhibition assays

Inhibitors, at approximately 4,2, 1, 0.5 and 0.25 times Ki value, were added to assays of Fhit with ApppBODIPY [10]. Ki values were obtained by calculating the inhibitor concentration-dependence in reduction of kcat/Km (apparent) as earlier described [10].