Studies Toward the Synthesis of Caramboxin Analogues

Intrigued by the recent discovery of caramboxin by Brazilian researchers, we present the results from our studies toward the racemic synthesis of caramboxin analogs through the orthocarboxylation of 3,5-dimethoxy benzyl derivatives. Three different approaches were tested, and the route involving a Vilsmeier-Haack formylation followed by a Lindgren oxidation provide a potential intermediate for the synthesis of several caramboxin analogs.


Introduction
Originally from Asia, star fruit or carambola (Averrhoa carambola) is a star-shaped fruit popularly consumed and used as traditional medicine in tropical countries around the globe.The toxic effect of this fruit, which mechanism was unclear, involves not only neurotoxicity, but also nephrotoxicity even for people with normal renal function. 1Recently, a review of publications from 2000 to 2014 related to the toxicity of carambola noted 27 deaths from 110 patients.In addition, the most commonly reported symptoms after consuming carambola are hiccups, vomiting and confusion. 2Although several studies have suggested that oxalate (C 2 O 4 2- ) is the responsible for toxicity, in 2013, Brazilian researchers 3 isolated a neurotoxin, caramboxin (1) (Figure 1), that can inhibit the GABAergic system (related to the central nervous system).
Compound 1 contains a phenylalanine skeleton, and the absolute configuration of the carbon α to the amino acid has only been inferred by comparison with the [α] D signal of L-phenylalanine. 3So far, only the 2D chemical structure of 1 was confirmed by a computational study using DFT (density functional theory) calculations. 5In this case, the theoretical nuclear magnetic resonance (NMR) chemical shifts are in accordance with the experimental NMR measured in dimethyl sulfoxide (DMSO-d 6 ).
To the best of our knowledge, there are no reports dedicated to the total synthesis of 1.In 2012, Quintiliano and Silva 4 reported a 10-step synthesis of the tetrahydroisoquinolinic derivative 2 starting from dimedone (Figure 1).Years later, an unsuccessful attempt to convert 2 to 1 was only reported in the PhD thesis of the same author. 6n view of the recent and the important discovery of caramboxin, studies toward the synthesis of derivatives of 1 could also be of great importance.The development of synthetic routes to obtain the core of caramboxin could contribute to a possible total synthesis of 1.
Structural analysis of 1 reveals an intriguing carboxyl group at the ortho position of the phenylalanine moiety, which we consider a significant synthetic challenge.Although a synthesis of (DL)-o-carboxy 13 C-phenylalanine starting from o-bromotoluene was reported, 7 the required benzylic bromination of toluene could be difficult if an activated aromatic analog is applied.Thus, we investigated three synthetic pathways for the direct o-carboxylation of the aromatic ring to synthesize 3-methoxylated lactam 4 (Scheme 1).Hydrolysis of 4 could lead to the 3-methoxylated derivative of caramboxin (3).
The first pathway involves the Bischler-Napieralski (BN) cyclization of N-Boc-protected ester 7 (Scheme 1, route A).Kim and co-workers 8 developed an in situ Friedel-Craftstype intramolecular cyclization of N-Boc carbamates via isocyanate intermediate 5 using triflic anhydride (Tf 2 O) and

Studies Toward the Synthesis of Caramboxin Analogues
Ronaldo E. Oliveira Filho, a Vanessa M. Higa a and Álvaro T. Omori * ,a 4-(dimethylamino)pyridine (DMAP).In this case, however, N-Boc carbamates containing an α-ester group (compound 6) were not tested.Isocyanate intermediate 5 could also be generated in situ through a Curtius rearrangement of monohydrolyzed malonate 7 (Scheme 1, route B).This second pathway arose from the work reported by Judd et al. 9 These authors reported a one-pot procedure to obtain dihydroisoquinolin-1-ones from activated dihydrocinnamic acids through a modified Curtius rearrangement in the presence of BF 3 •OEt 2 .However, although a protocol to transform monoester malonic acids into N-Boc carbamates has been reported, 10 the direct conversion of compound 7 into isocyanate 5 in the presence of a monoester is still unknown.
Finally, the third route was proposed to prioritize the C-H functionalization step (Scheme 1, route C).Compound 4 could be obtained through an intramolecular alkylation of N-amide malonate diester 8 followed by a decarboxylation step.Selective oxidation of aldehyde 9 could lead to the acid precursor of amide 8. Interestingly, compound 9 had previously been synthesized by Danishefsky and co-workers 11 through the Vilsmeier-Haack formylation of 10.

Bischler-Napieralski pathway
The Bischler-Napieralski approach had begun with commercially available 3,5-dimethoxybenzoic acid 11 (Scheme 2).Reduction of 11 with LiAlH 4 , followed by benzylic bromination assisted by PBr 3 in dioxane provided bromide 12 in almost quantitative yield over two steps. 12To insert the stable enolate fragment, nucleophilic substitution of 12 with the carbanion formed from previously prepared N-Boc malonate 14 with Cs 2 CO 3 gave desired alkylated product 15 in 80% isolated yield.Another alkylation protocol using microwave irradiation at high temperature was applied; 13 however, in this case, we observed that the Boc group from 14 is heat sensitive, and a decrease in the isolated yield was observed.
With carbamate 15 in hand, two sets of cyclization conditions using Tf 2 O in 2-chloropyridine (2-ClPy) or DMAP, according to Banwell's protocol, were tested. 14ven in the presence of the two methoxy groups, in both cases, NMR analysis showed no evidence of cyclization product in the aromatic region.The major product obtained in both cases was the free amine 16 with 85 and 40% yield using DMAP and 2-ClPy, respectively.Spyropoulos and Kokotos 15 proposed the formation of an imino triflate intermediate when Tf 2 O and N-Boc protected amino acids are mixed.We believe the formation of 16 might have occurred probably by the formation of the isocyanate followed by hydrolysis or by simple deprotection of the Boc through traces of trifluoromethanesulfonic acid. 16lthough the cyclization of N-Boc amides with Tf 2 O has been reported, 17 no examples containing diesters groups was found.Identical conditions with hydrocinnamic acid derivatives afforded the desired lactam.However, no reports using malonic acid monoesters were found. 18n order to verify the influence of the Boc group, the same reaction was conducted using an NHAc group (17).In this case, the desired BN product (18) was obtained in 20% isolated yield.Thus, probably the lability of the Boc group and the purity of the triflic anhydride are compromising the success of the cyclization.

Curtius rearrangement pathway
Since the BN pathway did not provide the desired cyclized product, we focused our efforts on the Curtius approach (Scheme 3).Thus, starting with the same bromide, 12, the alkylation with diethyl malonate under microwave irradiation 13 followed by monohydrolysis with an equimolar amount of KOH provided monoester acid 7 in good overall yield.Deprotonation of the diethyl malonate with NaH gave the same diester 19, in only 48% isolated yield.
Different bases for the Curtius rearrangement using diphenylphosphoryl azide (DPPA) with compound 7 were tested. 9Among them, only triethylamine (TEA) provided corresponding isocyanate 5 in 59% isolated yield.However, the intramolecular S E Ar was not observed.Another attempt using a greater amount of BF 3 at higher temperature (90 °C) only afforded a trace amount of recovered 5. Analogous to the BN approach, the failure of the reaction can be attributed to the presence of the ester.

Vilsmeier-Haack pathway
In view of the difficulty of the ortho-carboxylation of functionalized aromatics, likely due to chemoselectivity issues, we decided to prioritize the formylation in the beginning of the route through the Vilsmeier-Haack (VH) reaction of a single substrate.According to the literature, the VH reaction of benzylic alcohol 10 was achieved by Danishefsky and co-workers. 11Following the same protocol, using freshly distilled POCl 3 , we obtained highly functionalized chloride 9 in 90% isolated yield (Scheme 4).Due to the presence of the labile benzylic chloride, the oxidation of the benzaldehyde to the corresponding benzoic acid was carefully studied.The oxidation of the aldehyde is crucial for the synthesis of the amide.The results of the tested oxidation protocols are summarized in Table 1.
Depending on the oxidation protocol tested, compounds 20, 21 and 22 were obtained in different ratio.Initially, the protocols for benzaldehyde oxidation using Oxone® 19 and H 2 O 2 /AgNO 3 20 did not provide any polar compounds by thin layer chromatography (TLC) analysis.Additionally, oxidation using KMnO 4 revealed only traces of compound 20 or a 1:1 mixture of 20 and lactone 21 (entry 4). 21indgren oxidation is one of the mildest protocols to oxidize benzaldehydes to the corresponding benzoic acids. 22his reaction uses sodium chlorite as the oxidant and is operationally simple.However, due to the formation of hypochlorite in situ, chlorination of the activated aromatic ring can be observed.The results of the Lindgren reaction, shown in Table 1 (entries 5 to 12), suggests that the reaction is dependent on the temperature, the reaction time and the careful addition of the chlorite.Extending the reaction time (entry 9) and adding the sodium chlorite in one portion led exclusively to chlorinated acid 22. Very slow addition of sodium chlorite at low temperature provided mixtures of 20 and 21 (entries 5 to 8).Reactions at higher temperatures (entries 11 and 12) gave lactone 21, 23 indicating that acid 20 is quite sensitive.Danishefsky and co-workers 11 reported a 7:1 ratio of 20 and 22, and no formation of lactone 21 was observed.In our case, a higher proportion of the acid was obtained (19:1); however, the isolated yield was only 16% (entry 7).Another protocol using acetone as the solvent gave similar results to those achieved in the THF/H 2 O system (entry 10). 24Even in under buffer conditions (entries 11 and 12), lactone 21 was also obtained.
In terms of isolated yield, we could not reproduce the Danishefsky protocol.Purification of 20 in the presence of lactone 21 by chromatographic methods (silica gel, alumina, preparative TLC, and preparative high performance liquid chromatography (HPLC)) and by other methods (acid-base extraction and recrystallization) failed in our hands.In all purification attempts, we observed the lactonization of 20.
The position of the chlorine atom on the aromatic ring in 22 was determined by two-dimensional NMR through analysis of the HMBC (heteronuclear multiple-bond correlation) spectra.Long distance heteronuclear coupling constants ( n J C-H , n ≥ 2) 25 of compound 22 are shown in Table 2.
Next, the conversion of the acid to the amide was studied (Scheme 5).The diethyl aminomalonate hydrochloride (13) was chosen to guarantee the following intramolecular alkylation step.Thus, impure acid 20 and dichloride 22 were reacted with HBTU (2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate) in the presence of triethylamine in dichloromethane (DCM) at room temperature. 26Desired amide 8 was obtained in low yield (7%) and the major compound was lactone 21, which was obtained in 66% isolated yield.As expected, 20 was found to be unstable in the presence of triethylamine.However, when dichlorinated acid 22 was submitted in the same conditions that were used for 20, we could obtain amide 23 in 68% yield.The presence of the chlorine atom on the aromatic ring in 23 makes the molecule less sensitive to lactonization.Based on the low yields with sensitive acid 20, we decided to continue the pathway using amide 23 for the next steps.
We considered the intramolecular cyclization of 23 to be the crucial step for this route (Scheme 6).Due to the successful use of Cs 2 CO 3 in the BN pathway, we decided to subject dichloride 23 to similar conditions.A catalytic amount of KI was added to increase the reactivity of the benzylic chloride portion. 27In this case, a mixture of lactam 24 and iodide 25 were obtained in low yields.However, compound 25 was recovered and reacted again with Cs 2 CO 3 to furnish 24.The low yields in the first cyclization protocol can be attributed to the reaction being conducted at room temperature.A better conversion was observed when a higher temperature and microwave (MW) irradiation were used to recycle 25 to 24.The overall isolated yield of 24 was 27% over two steps.Lactam 24 possesses the majority of the functional groups present in caramboxin.Attempts to decarboxylate one of the esters and open the lactam ring in a one-pot fashion were carried out (Scheme 7). 28In the presence of 6 mol L -1 HCl at 130 °C, several byproducts were obtained.Liquid chromatography-mass spectrometry (LC-MS) analysis did not provide any evidence of a possible hydrolysis product.
The hydrolysis of amides is usually difficult.Thus, to avoid the d-lactam opening, a portion of remaining acid 22 was esterified by CH 3 I (Scheme 8).Corresponding methyl ester 27 was then submitted to the benzylic alkylation.In this case, we used protected diethyl acetamidomalonate 28, and corresponding product 29 was obtained in a higher yield (60%) than what was achieved with the intramolecular version (23 to 24).The hydrolysis of amide triester 29 was partially successful.In fact, the decarboxylation of only one of the malonate esters occurred, affording acetamide benzyl methyl ester 30 in quantitative yield.Unfortunately, further acid hydrolysis of 30 with longer reaction times and at higher temperatures gave complex mixtures of products.
As previously mentioned, on the basis of the high sensitivity of acid 20, we anticipated that the Vilsmeier-Haack product, stable aldehyde 9, could be converted to a less reactive benzonitrile analogue, which could later be converted to the corresponding acid by hydrolysis.Thus, the treatment of 9 with NaN 3 in POCl 3 generated nitrile 31 in reasonable yield (67%) (Scheme 9). 29Alkylation of 31 using 28 was accomplished under similar conditions to those mentioned before affording the malonate 32.At this point, the hydrolysis of the latter compound was more carefully investigated.
Since the last hydrolyses were carried out in acidic media, we decided to test the reactions under basic conditions with two distinct protocols.With the first set of conditions using 2 mol L -1 NaOH and an ultrasonic bath at 80 °C, 30 we obtained only the corresponding dicarboxylate disodium salt 33.On the other hand, using KOH as the base in refluxing ethanol, 31 we obtained only decarboxylation derivative 34.Interestingly, in the last attempt, when compound 33 was subjected to acidic hydrolysis (2 mol L -1 HCl) for an extended period, we could obtain imino tetrahydroisoquinolinone 35 in low isolated yield (12%).In this case, the expected hydrolysis of the acetamido group and the malonate decarboxylation occurred; however, the nitrile group was attacked by the free amino group, and a similar reaction was reported before by Hamley and co-workers. 32

Conclusions
We presented three different approaches for the ortho-carboxylation of 3,5-dimethoxy benzyl derivatives toward the preparation of caramboxin analogs.We observed that the insertion of the carboxyl group early on in the synthesis helps avoid chemoselective issues.The Vilsmeier-Haack formylation was chosen due to the concomitant halogenation of the benzylic alcohol, which facilitates the malonate alkylation.For the synthesis of caramboxin, a study regarding the regioselectivity of the VH will be necessary.We also conclude that the "protection" of the carboxylate by esterification or by functional group interconversion seems to be more attractive than lactamization by an intramolecular alkylation.Lastly, the challenges presented by the final steps, mainly the hydrolysis of the amide and the esters, require more detailed study.
70 µm) from Fluka Analytical.The gas chromatography-mass spectrometry (GC-MS) analysis was made in one ion trap, Varian 4000 from Federal University of ABC.The LC-MS analysis was made in quadrupole Agilent 6130 Infinity coupled to an Agilent 1260 HPLC system, from Federal University of ABC.The HRMS analysis was made in micro-TOF (time of flight) Bruker Daltonics from São Paulo University.The 1 H and 13 C NMR were made on Varian (500 MHz) from Federal University of ABC and Varian AIII or Bruker DPX-300 (300 MHz) from São Paulo University.The solvents used were deuterated chloroform (CDCl 3 ) and deuterated dimethyl sulfoxide (DMSO-d 6 ).The melting point analyses were made on Büchi B-540 or EZ-Melt SRS-Stanford Research Systems from Federal University of ABC.The purification on HPLC was made in Waters coupling with UV-Vis detector model 2489, using a semipreparative column Phenomenex C18.The Microwave™ synthesis system were made on CEM Focused, model Discover, from Federal University of ABC.