The term "fleximers" was introduced in 2001 by the Seley research group,1 referring to synthetic analogs of nucleosides in which the purine base has been "splitted" into two linked heterocycles – imidazole and pyrimidine. Initially intended as model compounds for studying the binding sites of enzymes and characterizing the interactions between proteins and nucleic acids, compounds of this type have subsequently shown remarkable antiviral and antitumor activity. In the light of recent events, there is a considerable interest in the possible applications of such compounds against coronavirus infections.2 The category of fleximers can be extended also to several other synthetic nucleosides, the heterocyclic base of which contains two (or more) linked planar rings. All of them can be viewed as isosteric analogs of natural nucleosides that possess considerable pharmacological potential. The methods for synthesis, description of properties, and the biological activity of fleximer nucleosides are thoroughly considered in the current review article.

Routes of synthesis leading to fleximer nucleosides

The first fleximer nucleosides – isosteric analogs of adenosine 1 and guanosine 2 – were obtained from 4,5-dibromo-1-(2,3,5-tri-О-benzyl-β-D-ribofuranosyl)-1Н-imidazole (3) through a rather involved multistep synthesis. The key step in the synthesis of the "splitted" structure of the heterocyclic base was the assembly of a fused tricyclic system with a central thiophene ring. This tricyclic system was subjected to Raney nickel desulfurization, providing the desired fleximer1 (Scheme 1). Similarly to any other synthesis of nucleoside analogs, there was the issue of which synthetic step would be most convenient for the formation of the glycosidic bond. The attempts to perform glycosylation of a previously assembled tricyclic base 4 did not quite succeed – instead of the expected mixture of glycosides 5 and 6, only the nontarget isomer 6 was obtained in a moderate yield3 (Scheme 2). As a result of this, the strategy was changed to assembling the ring system on the basis of the nucleosidic precursor 3.

scheme 1

Scheme 1

scheme 2

Scheme 2

The structures of the first synthesized fleximers 1 and 2 (Scheme 1) contained a bond between the С-5 carbon atom of imidazole ring and the С-6 carbon atom of pyrimidine ring (distal fleximers). The same research group later also obtained proximal fleximers with a conserved purine moiety, but with a different position of bond between the rings: between the С-4 carbon atom of imidazole and the С-5 carbon atom of the pyrimidine ring. The synthesis of these fleximers was achieved using a new route based on the palladium-catalyzed cross coupling, which was subsequently applied to the synthesis of nearly all described fleximer nucleoside analogs4 (Scheme 3).

scheme 3

Scheme 3

4-Iodo-1-(2,3,5-tri-О-benzyl-β-D-ribofuranosyl)-1Н-imidazole (7) was used in a Suzuki reaction with boronic acid 8 in the presence of a palladium catalyst and the obtained fleximer precursor 9 was directly converted into xanthosine isostere 10 and into the substituted diamino derivative 11 (Scheme 3). Compound 11, on the other hand, was used for the synthesis of fleximer guanosine analog 12 and isoguanosine analog 134 (Scheme 4). Similar methods provided access also to the proximal and distal fleximer 2'-deoxynucleosides,5 acyclic fleximer nucleoside analogs,2, 6 as well as the fleximer analogs of neplanocin А, a carbocyclic nucleoside antibiotic.7

scheme 4

Scheme 4

The distal "flexible guanine" isomer 14 was obtained from imidazole organozinc derivative 15 (Scheme 5), while the same authors did not succeed in deprotecting the proximal isomer.8 In more detail, synthetic methods for the nucleosides with "classical" fleximeric nucleobases have been described in a monograph that was published in 2018.9

scheme 5

Scheme 5

Several fleximer nucleosides having significantly simpler structures were synthesized for modeling the process of genetic code recognition. In that case, the classic synthetic sequence was applied to the preparation of nucleoside analogs: the synthesis of compounds 16a,b was followed by their glycosylation (Scheme 6). Fleximers 17a,b were then converted through known procedures into the amino derivatives (Х = NH2) and functionalized at the amino group. The yields of compounds 17a,b were in the range from 10 to 37%, probably due to the formation of isomeric products at the glycosylation stage.10

scheme 6

Scheme 6

Another class of fleximer nucleosides was created with the aim of searching for new antiviral and antitumor drugs, therefore pharmacologically active isosteres were chosen instead of natural purine nucleosides as a structural model. The synthesis of a series of "inverted" carbocyclic fleximers 18a–c, in which a pyrimidine ring was linked to the pseudo-carbohydrate moiety, gave 68–89% yields in the key step. The bond between the pyrimidine ring and the second heterocycle was formed by a condensation reaction of 5-bromouracil derivative 19 with organostannanes according to the Stille reaction11 (Scheme 7). The same approach was also used for obtaining the derivatives with 2,3-dihydroxycyclopent-4-en-1-yl and 2,3-dihydroxycyclopentan- 1-yl carbocycles.12

scheme 7

Scheme 7

In many cases, click reactions were used for the fleximer synthesis, in particular Huisgen's cycloaddition of azides to alkynes. Distal fleximers 20 were obtained upon microwave irradiation in the presence of a ruthenium catalyst,13 while proximal nucleosides 21 were prepared by a copper-catalyzed Huisgen reaction14 (Scheme 8). The yields of the protected nucleosides ranged from 75 to >99%.

scheme 8

Scheme 8

Similar methods provided access to bistriazole fleximers, related to ribavirin (1-(β-D-ribofuranosyl)-1H-1,2,4- triazole-3-carboxamide) – a known antiviral drug.15 Starting from protected structural analogs of ribavirin, containing an azido group at position 3 of the heterocyclic base, a total of 11 fleximer ribonucleosides 22 were synthesized, which had various alkyl and aryl substituents in the 1,2,3-triazole ring. Remarkably, the isomer with an azide group at position 5 did not participate in cycloaddition reaction under the same conditions15a (Scheme 9).

scheme 9

Scheme 9

Similarly, 10 acyclic bitriazole acyclic nucleosides 2315b were also synthesized (Scheme 9), while a series of 9 acyclic ribavirin fleximer analogs 24 containing a С–С bond between the triazole rings were obtained by a reaction of the protected acyclic nucleoside 25 with alkyl and aryl azides15c (Scheme 10).

scheme 10

Scheme 10

Analogs of ribavirin that contain aryl substituents at position 5 of 1,2,4-triazole ring can be viewed as a simplified variant of fleximer nucleosides. A series of 10 such ribosides were obtained in 72–91% yields from protected 5-bromoribavirin and arylboronic acids using a Suzuki reaction, that proceeded under microwave irradiation in PhMe,16 and 13 acyclic analogs too.17

The structure of triazolyloxadiazole ribonucleosides 26 seems to be closer to the initial concept of fleximers as purine nucleobase isosteres. They were obtained by cyclization of amidoxime 27 in the presence of tetrabutylammonium fluoride (TBAF)18 or O-(benzotriazol-1-yl)- N,N,N',N'-tetramethyluronium tetrafluoroborate (TBTU)19 (Scheme 11).

scheme 11

Scheme 11

Compounds 26 and 2'-deoxyribonucleosides containing the same bases were obtained also by a different route. Starting from 1H-1,2,4-triazole-3-carboxamide (28), the substituted triazolyloxadiazoles 29 were obtained in three steps in the overall yield of 68–71% (Scheme 12). The attempts at direct glycosylation of these bases according to the traditional procedure were not successful: the oxadiazole ring was destroyed. At the same time, these compounds appeared to be excellent substrates for recombinant purine nucleoside phosphorylases (PNP, ЕС 2.4.2.1), enzymes applied to the synthesis of various unnatural nucleosides by a transglycosylation reaction, essentially representing a nucleobase transfer.18 Although the substrate specificity of these enzymes was not very high,20 the participation of bases 29 in the reaction indirectly confirmed the postulated isosterism of fleximer bases and purines.

scheme 12

Scheme 12

It has been recently also shown that enzymatic transglycosylation enabled the preparation of "classic" fleximers. The catalyst in this process could be not only PNP from E. сoli, but also nucleoside-2'-deoxyribosyltransferase (LlNDT, EC 2.4.2.6) from L. leichmannii, allowing to synthesize 2'-deoxyribonucleosides 30 and 31. Due to the substrate specificity of these enzymes, the reaction proceeded mostly at N-1 position of the imidazole ring. Despite the fact that both the proximal and distal isomers of nucleobases participated in this reaction, only the proximal compounds 31 and 32 were synthetically accessible21 (Scheme 13).

scheme 13

Scheme 13

The distal isomers may be obtained from the enzymatic reaction products of tricyclic base 3321a by desulfurization on Raney nickel or Suzuki reaction of nucleoside 34 with arylboronic acids21b (Scheme 14).

scheme 14

Scheme 14

The structural features of fleximers

The fleximer nucleoside analogs were initially designed on the basis of the assumption that mutual rotation of the rings in the modified nucleobase should increase the accessibility of the structure to interactions within enzyme active sites.3 These compounds were presented as "molecular chameleons" that could adapt to their environment and therefore, presumably, be more active in the roles of enzyme inhibitors or agonists.4 However, a question was immediately raised about the actual degree of flexibility in these fleximers. The theoretically free rotation of the rings relative to each other must in reality be limited by intramolecular and intermolecular interactions. The comparative conformational analysis of fleximers 1 and 2 and the natural purine nucleosides, relying on the data of one-dimensional differential NOE spectra and ab initio calculations showed that the pool of N conformations (north conformations) in the case of fleximers was considerably larger than for natural purines.22 This could, for example, explain the unexpected earlier observation: fleximer guanosine analog 2 inhibited S-adenosyl- L-homocysteine hydrolase (EC 3.3.1.1) – an enzyme that is specific for adenosine, while fleximer 1 (an adenosine analog) did not interact with this enzyme.23 Both ribosides and acyclic fleximers24 predominantly existed in their anti conformations in solutions (compounds 35–38, Fig. 1). Computational structure modeling of the "classic" imidazole-pyrimidine fleximers showed that the energy barrier for rotation around the С–С bond linking the heterocycles was equal to about 40 kJ/mol. Besides that, the calculated bond energies for fleximer–pyrimidine pairs showed that the hydrogen bond properties in these modified nucleobases imitated the properties of the respective natural purines. The interaction with carbohydrate moiety often led to preferred nonplanar conformation of the two rings: the rings were oriented at 25–40° relative to each other, while the energy barrier was decreased.25

Figure 1.
figure 1

The conformations of fleximer ribosides 35–38.

Full size image

X-ray structural studies of bitriazole fleximers 22 (Scheme 9) showed that the triazole rings in the bitriazole base were coplanar.15a The nitrogen-containing heterocycles were also coplanar in the nucleobase moieties of compounds 35–38, which were obtained from the substituted compounds 21 (Scheme 8), as their conformation was stabilized by the additional hydrogen bonds between the rings (Fig. 1).

The biological activity of fleximers

The fleximers intended for studying enzyme–substrate interactions showed a broad range of biological activity, as should be expected. Similarly to other nucleoside analogs, they exhibited antiviral and antitumor properties associated with their isosteric similarity to the natural nucleobases. The justification for such activity is considered in detail in several review articles.26

The interest of researchers has been most attracted by the varied antiviral activity of fleximers. Acyclic analogs 39a,b of the "classic" fleximers (Fig. 2) are active against the coronaviruses HCoV-NL63 and MERS,2, 6 the filoviruses, in particular various strains of the Ebola virus,27 while the ribavirin-type fleximers have shown activity against the hepatitis С virus (compound 26 (R = Ph, Scheme 11) had half maximum inhibitory concentration (IC50) of 8.8 μg/ml, half maximum cytotoxic concentration (CC50) of 380 μg/ml), as well against the herpes and influenza A viruses,18 and the tobacco mosaic virus.15b

Figure 2.
figure 2

Acyclic analogs of "classic" fleximers – compounds 39a,b showing antiviral activity.

Full size image

A detailed study by Seley-Radtke and coworkers revealed the relationships between the structure of functionalized pyrimidine ring and the activity of imidazole-pyrimidine fleximers against the Ebola virus, as well as mentioned some unpublished data on antiflavivirus properties of the "classic" fleximers.24 The correlation between the biological properties and structural features of "flexible" nucleobases was discussed in a monograph published in 2014.28

Acyclic bitriazole nucleosides 24 (Scheme 10) exhibited antiproliferative activity against the drug-resistant pancreatic cancer cell line MIA PaCa-2, which is associated with caspase-dependent apoptosis. These compounds were also characterized with respect to their immunomodulatory properties. Interestingly, the most active among acyclic bitriazole nucleosides 24 was the protected intermediate 40 containing (pyren-1-yl)methyl substituent at the 1,2,3-triazole ring (Fig. 3), and not the nucleoside derived from it.15c Antiproliferative activity against the cell line MIA PaCa-2 was also exhibited by close structural analogs of bitriazole fleximers, ribonucleosides containing an arylamino group at the 1,2,4- triazole ring.29

Figure 3.
figure 3

Acyclic fleximer nucleoside analog 40, carbocyclic fleximer nucleosides 42a–c, 43a–c, and the fleximer nucleosides 41, 44.

Full size image

There is a scarcity of available information in the literature on the biological targets and established mechanisms of action for fleximer nucleosides. As already pointed out, the distal fleximer guanosine analog 2 (Scheme 1) inhibited the S-adenosyl-L-homocysteine hydrolase,23 while its triphosphate was an excellent substrate for human fucose-1-phosphate guanylyltransferase (EC 2.7.7.30).30 Compound 40 activated caspase-3 and caspase-7 (EC 3.4.22.56 and EC 3.4.22.60, respectively), causing apoptosis of cancer cells, while simultaneously acting as agonist of the TLR7 receptor (tolllike receptor 7) that initiates immune response mechanisms.15c Compound 41 inhibited the cytosolic 5'-nucleotidase II (EC 3.1.3.5),14b while carbocyclic nucleosides 42a–c and 43a–c (Fig. 3) inhibited adenosine deaminase (EC 3.5.4.4).12

Proximal nucleotides 44 (Fig. 3) were studied in the role of substrates for DNA polymerases. Commercially available DNA polymerases were used: endonuclease-free Klenow fragment KF (exo-), Deep Vent (exo-), and terminal deoxynucleotidyl transferase. Triphosphates 44 were found to act as substrates for all of these enzymes, with adenine as the preferred base pair during DNA chain elongation. The derivative with 3-aminophenyl substituent was integrated in the chain opposite to adenine by the KF polymerase (exo-) only 37 times less effectively than the native thymidine triphosphate. The chain elongation terminated after the inclusion of a single arylimidazole nucleotide.31 This observation suggests that a possible mechanism for the antiviral action of compounds 39a,b (Fig. 2) relies on the termination of viral nucleic acid replication.

A quite significant number of interesting compounds in which planar rings of the nucleobases are separated by more than one bond are described in a review article32 and a series of subsequent publications.33 Compounds that are structurally related to fleximers have already entered pharmaceutical development. Thus, pevonedistat (MLN4924)34 is currently in phase I of clinical trials against acute myeloid leukemia, while regadenoson (CVT-3146, Lexiscan) has been approved by the U.S. Food and Drug Agency (FDA) as agonist of adenosine receptor А2А for the treatment of cardiovascular diseases. While a question may be raised whether such compounds should be categorized as fleximers, as the majority of them are not isosteres of purine or pyrimidine nucleosides, it should be recognized that the conformational flexibility of the heterocyclic moiety may affect their biological activity.

In the conclusion, it should be noted that even though the majority of key synthetic work devoted to fleximer nucleosides has been published many years ago, the most promising studies from the aspect of biological properties and medicinal chemistry applications are quite recent. This gives a reason to hope that the research involving fleximer nucleosides may have a great future.