Synthesis, Stability, and Biological Studies of Fluorinated Analogues of Thromboxane A2.

Platelet activation results in the generation of thromboxane A2 (TxA2), which promotes thrombus formation by further amplifying platelet function, as well as causing vasoconstriction. Due to its role in thrombus formation and cardiovascular disease, its production is the target of antiplatelet drugs such as aspirin. However, the study of TxA2-stimulated cellular function has been limited by its instability (t 1/2 = 32 s, pH = 7.4). Although more stable analogues such as U46619 and difluorinated 10,10-F2-TxA2 have been prepared, we targeted a closer mimic to TxA2 itself, monofluorinated 10-F-TxA2, since the number of fluorine atoms can affect function. Key steps in the synthesis of F-TxA2 included α-fluorination of a lactone bearing a β-alkoxy group, and a novel synthesis of the strained acetal. F-TxA2 was found to be 105 more stable than TxA2, and surprisingly was only slightly less stable than F2-TxA2. Preliminary biological studies showed that F-TxA2 has similar potency as TxA2 toward inducing platelet aggregation but was superior to F2-TxA2 in activating integrin αIIbβ3.

Introduction

Thromboxane A2 (TxA2) is produced enzymatically from arachidonic acid through the action of several enzymes including cyclooxygenase (COX) and thromboxane synthase in response to tissue injury, promoting hemostasis, vasoconstriction, and wound healing.[1−3] However, these necessary features for survival can also cause death to those susceptible to or suffering from cardiovascular disease (CVD).[4−9] Current first-line therapy involves the use of nonsteroidal anti-inflammatory drugs (NSAIDs) which block >95% of COX1 activity and therefore TxA2 production.[10] However, the treatment suffers from side effects associated with shutting down the whole prostanoid cascade and with resistance in some patient groups.[11]

The study of TxA2 has been limited by its high instability (t1/2 = 32 s, pH = 7.4)[1] and so a number of more stable analogues have been prepared in which one or both oxygens of the strained acetal have been replaced by carbon,[12] sulfur,[13] or a less strained bicyclic structure (e.g., U46619, Figure ).[14,15] A different strategy is to retain the strained acetal but reduce the rate of hydrolysis by incorporating either bromine[16] or, more importantly, fluorine[17−20] atoms at the C-10 position (Figure ). Although the synthesis[21,22] of monofluorinated F-TxA21 has been attempted,[23] only the difluoro analogue 2 has succumbed to total synthesis,[17] which showed similar potency in platelet aggregation to the parent compound.[20] The stability of 2 has only been investigated using a model compound (3), which, as expected, showed much higher stability than TxA2 (t1/2 > 30 days, pH = 7.4).[20] We were interested in targeting F-TxA21 since the number of fluorine atoms can have a significant impact on function. For example, in a comparative study of the CHF- and CF2- phosphonate analogues of sn-glycerol-3-phosphate, O’Hagan found that the monofluorinated was better than the difluorinated substrate for the dehydrogenase enzyme.[24−26] We now report the first synthesis of F-TxA21 and compare its stability and biological activity with that of F2-TxA22.

Figure 1

Thromboxane A2 and its analogues.

Results and Discussion

Our retrosynthetic analysis of F-TxA2 is shown in Scheme . We envisioned forming the strained acetal by an intramolecular cyclization and introducing the upper side chain by a Wittig reaction on the corresponding fluorinated lactol. Lactone 6 could be obtained by fluorination of the enolate of lactone 7, which itself could be synthesized by Baeyer–Villiger oxidation of ketone 8. Ketone 8 could then be obtained from conjugate addition of the lower side chain 9 to our key enal intermediate 10 followed by ozonolysis. At the outset, the main challenges presented in the synthesis were formation of the strained acetal and fluorination of the enolate bearing a potential leaving group at the β-position.

Scheme 1

Retrosynthesis of Fluorinated Thromboxanes from Bicyclic Enal

Synthesis of Fluorinated Thromboxane A2

Our synthesis began from PMB-acetal 12, available in 3 steps in high er using our established proline-catalyzed aldol dimerization of succinaldehyde (Scheme ).[27,28] Initially, we elected to carry through the major β-isomer of the acetal to simplify analysis. Conjugate addition of the mixed vinyl cuprate 13 followed by trapping with TMSCl and ozonolysis[27] gave ketone 14 which was converted into the key lactone intermediate 15 through a Baeyer–Villiger oxidation[29,30] (64% yield, over 3 steps).

Scheme 2

Synthesis of Key Lactone Precursor and Completion of the Synthesis of the Monofluorinated Thromboxane A2

With a scalable synthesis of lactone 15 in hand, we embarked on the fluorination reaction. Lactones bearing siloxy and benzyloxy groups in the β-position are particularly prone to elimination upon deprotonation and have to be trapped by reactive electrophiles at low temperature.[31−35] Initial investigation showed that NFSI was a sufficiently reactive electrophile, and after optimization we found that the reaction proceeded with good selectivity (10:1 dr) and yield (51%) using 1.2 equiv KHMDS and 2.5 equiv NFSI in Et2O. Following PMB deprotection with DDQ, we explored the Wittig reaction with (4-carboxybutyl)triphenyl-phosphonium bromide (18), but this invariably led to intractable mixtures. We suspected that the lactone was interfering in this step and so converted lactone 16 into siloxyacetal. This time, following PMB deprotection, Wittig reaction using phosphonium salt 18 with t-BuOK surprisingly gave the corresponding epoxide in 67% yield.[36] To avoid epoxide formation, we screened alternative conditions and found that using LiHMDS with a ratio of hemiacetal (17):Wittig salt:LiHMDS of 1:4:8 at 0 °C gave the corresponding alkene in 82% yield as a separable 5:1 mixture of Z/E isomers after esterification with TMSCHN2. Selective desilylation of the TIPS group with TBAF/AcOH gave the required lactol 19 in 98% yield.[37−41]

To complete the synthesis of 10-F-TxA2, we required a method for the construction of the strained acetal. Owing to its known sensitivity and the low yields previously obtained for the construction of this motif, we decided to explore this key step on model substrate 23. This was prepared from d-arabinal-derived glycal 22 by fluorination with selectfluor (Scheme ).[42−44] Two methods for making the strained acetal had been reported previously, Still’s Mitsunobu reaction[45,46] and Fried’s displacement of the mesylate,[17−20] but neither was successful on hemiacetal 23 as shown in Scheme . These synthetic hurdles required us to find a new method to make the strained acetal. Shoda reported that treatment of unprotected glycopyranoses with 2-chloro-1,3-dimethylimidazolinium chloride (DMC) gave the corresponding 1,6-anhydro sugars directly.[47,48] This reagent was tested on hemiacetal 23, but although we did not obtain the desired acetal 24 directly, we did isolate chloride 26 with complete chemoselectivity. The fortuitous formation of the (unstable) chloride presented another opportunity, since glycosyl chlorides can be activated by silver salts to promote their displacement.[49] Indeed, treatment with Ag2O promoted cyclization giving the acetal 24 in 40% yield, providing a novel solution to the synthesis of strained acetals.

Scheme 3

Formation of Strained Acetal on Model Hemi-acetal 23

Moving onto the real target, brief optimization of the chlorination/cyclization steps was again required but optimum conditions were quickly established. Treatment of hemiacetal 19 with 6 equiv of each of the chlorination reagent, DIPEA, and Ag2O gave the desired acetal 21 in 52% yield (Scheme ). Finally, hydrolysis of 21 with 1.0 N NaOH in 50% 1,4-dioxane/water followed by deprotection with TBAF furnished F-TxA21 in 78% yield.

We also tried to prepare the other diastereoisomer 10α-F-TxA2 from the minor diastereomer formed in the fluorination of lactone 15. While we were able to carry this diastereoisomer through to the corresponding diol (hydroxy hemiacetal, diastereomer of 19), attempts to prepare the chloride and the subsequent cyclization were thwarted by competing elimination and hydrolysis.

By adapting this strategy, we were able to prepare F2-TxA2 (see Supporting Information), so that its stability and biological activity could also be assessed. With both fluorinated TxA2 analogues in hand, we were then able to compare their stabilities with the parent TxA2 and study their biological activity.

Stability Studies of Fluorinated Thromboxane A2 and Model Compounds

The hydrolytic stability of TxA2 at pH 7.4 (37 °C) was measured and found to have a t1/2 of 32 s.[1] Fried measured the stability of his F2-TxA2 model compound 3, which is similar in structure to F2-TxA2, at pH 1.27 (22 °C) to have a t1/2 of 86 min. While this 108 difference in rate constant is interesting to note, the difference in pH and temperature of these measurements renders a direct comparison of stability, and an assessment of the effect of fluorine, very difficult. Hence, we sought to compare the stability of TxA2 with its fluorinated analogues by measuring the kinetics of hydrolysis under the same conditions. Using 19F NMR to monitor the decay of the acetal moiety, we determined pseudo first-order rate constants for the hydrolysis of 1 and 2 (Table ) under buffered conditions. At pH 7.4, we found that F-TxA2 (1) has a half-life of 20 days, which is 105 more stable than TxA2. Interestingly, F2-TxA2 (2) was only 1 order of magnitude more stable at pH 7.4 with a half-life of over 40 weeks. We then measured hydrolysis rates of 1 and 2 at lower pHs (Table ), where, as expected, decreasing the pH decreased the stability. The rate of hydrolysis we measured for F2-TxA2 (2) at pH 1.25 (t1/2 = 64 min) was in good agreement with that of Fried’s model compound 3 at pH 1.27 (t1/2 = 86 min).[20,50]

Table 1

Kinetics of Hydrolysis.a

compound	pH	k1′ (s–1)	t1/2
F-TxA2 (1)	7.40	3.93 × 10–7	20 days
	2.42	8.93 × 10–5	2.2 h
F2-TxA2 (2)	7.40	2.5 × 10–8	46 weeks
	2.42b	1.01 × 10–6	190 h
	1.80	5.46 × 10–5	3.5 h
	1.25	1.80 × 10–4	64 min

Hydrolyses of 1 and 2 were measured under buffered conditions (50 mM), using 19F NMR to monitor the decay of the ketal. k1′ = pseudo first-order rate constants. t1/2 = half-life.

Average of two runs.

The marginal increase in stability of 2 compared to 1 at pH 7.4 was unexpected, as the increase in stability caused by inductive effects of the electronegative fluorine atoms is usually additive.[51,52] Thus, we speculated that there might be a strong stereoelectronic effect governing the stability of the strained acetal. Unfortunately, we were not able to prepare 10α-F-TxA2 to test this, so we compared the stability of the two diastereoisomers of model compound 24 (3α-24 with 3β-24, Scheme ). Indeed, we measured a very substantial difference in hydrolysis rate between the isomers: 3β-24 was ca. 200× more stable than 3α-24. The greater lability of 3α vs 3β-24 presumably originates from having a better σ-donor (C–H vs C–F bond) aligned to the incipient oxocarbenium ion, as supported by DFT calculations on a model substrate (Scheme ; see Supporting Information for further discussion). Our inability to make 10α-F-TxA2 could therefore be due to its greater instability. Furthermore, as 3a-24 exhibited a half-life of just 15 h at pH 7.4, it is likely that 10α-F-TxA2 would not have been suitable for biological studies (see Supporting Information for full details). These studies therefore reveal that the stability derived from the stereoelectronic effect of an antiperiplanar fluorine is very significant compared to a syn-periplanar fluorine and provides a rationale for the nonadditive inductive effect of fluorine atoms on acetal hydrolysis.

Scheme 4

Investigations into Hydrolysis of Model Compound 24

Biological Studies

To evaluate the biological activity of the fluorinated thromboxanes 1 and 2, concentration–response experiments were performed on human platelets, and platelet aggregation was recorded by light transmission aggregometer. The stable PGH2 analogue U46619 has been used widely as a standard of comparison for evaluating TxA2-like activity and so was included in this study.[53−56] Concentration–response curves were fitted (Figure ) and EC50 values were calculated (Table ). The data show that F-TxA2 has similar activity as U46619 in inducing platelet aggregation but is almost 3-fold less potent than F2-TxA2. While F2-TxA2 was more potent, the EMax was significantly lower than U46619 and F-TxA2, suggesting partial agonism at TxA2 receptors. As platelet amplification pathways such as ADP release and integrin αIIbβ3 outside-in signaling can potentially mask a weaker agonist response in aggregation experiments, we were also interested to study a more direct functional readout of platelet activation: integrin αIIbβ3 activation. Interestingly, we found that, in contrast to F-TxA2 and U46619, F2-TxA2 induced only weak integrin αIIbβ3 activation (Figure ). Since both F-TxA2 and U46619 have similar activity in both aggregation and integrin αIIbβ3 activation experiments and U46619 has comparative activity to TxA2,[23] our data strongly indicates that F-TxA2 is a closer mimic to TxA2 than F2-TxA2. Further biological and pharmacological studies are ongoing.

Figure 2

TxA2-like properties of mono- and difluorinated TxA2 analogues on platelet aggregation. Aggregation of human platelet-rich-plasma induced by U46619, F-TxA2, and F2-TxA2 (average ± SEM, n = 3).

Table 2

Concentration of TxA2 Analogues Which Produces 50% Of Maximal Aggregation

compound	pEC50a ± SEM, n = 3	EC50b (μM), n = 3	EMaxc (%) ± SEM, n = 3
U46619	5.85 ± 0.06	1.4	84.1 ± 2.7
F-TxA2 (1)	5.80 ± 0.18	1.6	77 ± 13.1
F2-TxA2 (2)	6.30 ± 0.12	0.5	69.3 ± 4.5

pEC50, the negative logarithm of EC50.

EC50, the concentration of agonist that produces 50% of maximum response.

EMax, the maximum aggregation in platelet-rich plasma.

Figure 3

TxA2-like properties of mono and difluorinated TxA2 analogues on platelet integrin αIIbβ3 activation. Washed platelets were stimulated with U46619, F-TxA2, and F2-TxA2 in the presence of 1 μM ADP for 15 min and integrin αIIbβ3 activation was determined using FITC-PAC1 by FACS analysis. Data is expressed as a percentage of the maximal α-thrombin (0.5 U/mL) response (average ± SEM, n = 5).

Conclusions

In summary, we have developed novel syntheses of chemically stable fluorinated thromboxanes, utilizing our key enal intermediate, which is readily available in high ee. The total synthesis of the F-TxA2 and F2-TxA2 were completed in 17 and 18 steps, respectively, from 2,5-dimethoxytetrahydrofuran. The scalable route enabled >100 mg of advanced material (e.g., 21) to be prepared for chemical and biological screening. In addition to overcoming some unexpected challenges associated with incorporating and carrying fluorine through a synthesis, we have also developed a new method for constructing the highly strained acetal. As expected, F-TxA2 does indeed possess markedly greater stability than TxA2, enabling it to be further studied in biological assays. Preliminary biological studies showed that F-TxA2 is the closest mimic to date of TxA2 having similar potency toward inducing platelet aggregation, and is considerably superior to F2-TxA2 in activating integrin αIIbβ3.