From ribavirin to NAD analogues and back to ribavirin in search for anticancer agents

Introduction

In 1972, ribavirin (virazole) was synthesized by Robins and coworkers [1] and was found to be a broad-spectrum antiviral agent active against both DNA and RNA viruses [2]. Although its mechanism of action is not fully understood, even after more than 40 years of studies and more than 13 000 research articles,¹ ribavirin is still used in the clinic today and has recently played an important role in therapies against hepatitis C virus infection (in combination with other drugs and interferons). Ribavirin is a ribonucleoside, structurally similar to the natural 5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside (AICA riboside) (Figure 1). When phosphorylated in the cell to the monophosphate level, it mimics inosine monophosphate (IMP) and inhibits IMP dehydrogenase (IMPDH), a key enzyme in de novo synthesis of purine ribonucleotides. As a triphosphate, it is structurally similar to the ‘cap’ of mRNA and interferes with the translation processes. The ribavirin’s nucleobase could also be considered as a fragment of adenine or guanine (depending on its conformation around the carboxamido bond), which is able to hydrogen bond with both uracil and cytosine. This leads to mutations often responsible for an ‘error catastrophe’ [3]. All the above activities severely influence replication of RNA viruses [4]. How ribavirin works against DNA viruses remains a mystery, especially because the 2′-deoxy-ribavirin does not show any biological activity.

Figure 1

Structures of ribavirin, AICA riboside, and tiazofurin.

In the subsequent years, numerous analogues of ribavirin have been synthesized as potential antiviral agents, among them tiazofurin [5] (Figure 1). Tiazofurin (in contrast to ribavirin) is a C-nucleoside, in which the natural glycosyl carbon-nitrogen (C-N) bond is replaced by carbon-carbon (C-C) linkage.

Contrary to expectations, this heterocyclic nucleoside did not show any antiviral activity. After some time on the shelf, tiazofurin was rediscovered as a potent inhibitor of proliferation of several cancer cell lines, including acute and chronic myelogenous leukemia (AML and CML, respectively) [6]. It was soon established that the anticancer activity of tiazofurin was related to its inhibition of IMPDH [7–9]. Tiazofurin was phosphorylated to the 5′-monophosphate but did not inhibit IMPDH as a substrate analogue. In contrast to ribavirin, it was instead coupled in the cell with ATP (by NMN adenylyl transferase) to form a nicotinamide adenine dinucleotide (NAD) analogue – tiazofurin adenine dinucleotide, called TAD [8, 10, 11] (Figure 2).

Figure 2

Structures of NAD and TAD.

TAD mimics NAD but cannot participate in hydride transfer and selectively inhibits IMPDH. Two isoforms of the human enzyme are known [12, 13] (IMPDH1 and IMPDH2), and TAD inhibits both of them with equal potency (K_i=110 nm). Tiazofurin entered several clinical trials as a potential anticancer agent [14–19] and was approved by the Food and Drug Administration (FDA) as an orphan drug for patients in blast crisis of CML [20]. However, its broad application is limited because of the development of resistance, mainly due to the expression of phosphodiesterase called TAD-ase, which cleaves the TAD pyrophosphate bond to the inactive tiazofurin and adenosine monophosphates [21, 22].

In the 1980s, at the Memorial Sloan-Kettering Cancer Center, classes of C-nucleosides (related to ψ-isocytidine) were developed as potential anticancer agents [23–25]. It was anticipated that some of these compounds might mimic well the natural nicotinamide riboside and possibly show a potent anticancer activity. In addition, if these C-nucleosides were synthesized as their corresponding NAD (TAD)-like analogues, their therapeutic potential might be significantly improved. Indeed, if they are resistant to the action of phosphodiesterases, enter the cell, cannot participate in hydride transfer, and selectively inhibit IMPDH, they might be of high therapeutic potential. Herein is a review describing our early enthusiasm, some first-generation inhibitors, and recently developed derivatives as well as offering some comments about the future of the field.

Early enthusiasm

The attachment of nicotinamide ring to the sugar moiety through a C-C glycosyl bond instead of the C-N bond leads to its C-nucleoside isosters (Figure 3), which more closely resemble the parent nicotinamide riboside than tiazofurin [26, 27]. Contrary to expectations, we found that these nucleosides did not show any anticancer activity. Possibly, they were not converted in the cell into their corresponding NAD analogues, or if converted, these dinucleotide analogues did not inhibit IMPDH.

Figure 3

Structures of nicotinamide riboside and its C-nucleoside isosters, as well as NAD analogues – C-NAD and C-PAD.

When NAD analogues were prepared and submitted to biological evaluation against cellular dehydrogenases, C-NAD (but not C-PAD) was found to inhibit alcohol dehydrogenase (K_i=1 nm) about ~20 000-fold more potently than other cellular dehydrogenases, including IMPDH [28]. Meanwhile, the synthetic ribavirin adenine dinucleotide (1) was poor inhibitor of IMPDH (K_i=235 μm) and AICA analogue 2 was inactive [29] (Figure 4).

Figure 4

Structures of ribavirin and AICA adenine dinucleotides 1 and 2.

Interestingly, when the benzamide riboside was prepared [30], it proved to be as potent as tiazofurin in anticancer activity and was found to be metabolized in the cell to the corresponding NAD analogue [31] – benzamide adenine dinucleotide (BAD) [32] (Figure 5).

Figure 5

Structures of benzamide riboside and BAD.

As mentioned above, tiazofurin and benzamide riboside are effectively converted in the cell into their corresponding NAD analogues, which is rather a unique metabolic activation. Grifantini and coworkers [33–35] prepared numerous analogues of tiazofurin, which followed this pathway as well, and in recent reviews, tiazofurin and some of its analogues containing heterocyclic sugar analogues were reported [36, 37]. All the above NAD analogues cannot be considered as potential drugs because, as pyrophosphates, they are not able to cross the cell membrane and do not enter the cell. Even when they were synthesized in the cell, they were metabolically unstable being cleaved by cellular phosphodiesterases. However, when their pyrophosphate oxygen is replaced by a carbon atom, these compounds become bis(phosphonate)s, known to enter cells and be resistant to phosphodiesterase cleavage. Thus, some nucleosides that were inactive at the nucleoside level, when chemically converted into NAD analogues, might show interesting biological activities. Consequently, synthetic bis(phosphonate) NAD analogues independent of metabolic activation and resistant to the action of cellular phosphodiesterases were expected to show improved therapeutic potential.

Mycophenolic adenine dinucleotides (MAD) analogues

Mycophenolic acid (MPA), a potent and selective inhibitor of IMPDH, is approved as an immunosuppressant [38]. Numerous attempts to use MPA as an anticancer agent failed, mainly due to its unfavorable metabolic profile. In the human body, MPA is extensively glucuronidated and removed in urine (a dose of 2–3 g a day is needed to maintain the therapeutic level of the drug) [39]. MPA binds to IMPDH at the nicotinamide mononucleotide subdomain (N subsite) of the cofactor (NAD)-binding pocket. The adenosine subdomain (A subsite) is empty. The attachment of adenosine to the MPA moiety through bis(phosphonate) linker afforded the corresponding MAD derivatives (Figure 6). These compounds were able to enter cells, showed nanomolar inhibition of the enzyme, and, most importantly, were resistant to glucuronidation [40]. Thus, they severely inhibit proliferation of several cancer cell lines with IC₅₀s in low micromolar or even sub-micromolar range.

Figure 6

Structures of MPA and bis(phosphonate) MAD analogues 4 and 5.

In general, the NAD(P)-binding domain is modular in nature [41–43]. In addition to N and A subsites, it also contains the pyrophosphate-binding subsite (P subsite or P-groove). Each subsite plays an important role in securing proper and tight binding of NAD-like inhibitors; however, each has its own requirements.

The N subsite is the most restricted, and we used tiazofurin, benzamide riboside, or MPA as the replacement for NMN. These compounds played a role of an anchor, which holds well only at the N subdomain of the cofactor-binding pocket of human IMPDH (Figure 7).

Figure 7

The modular nature of the cofactor-binding domain of NAD-dependent IMPDH (PDB entry 1NFb).

The A-subsite region of the NAD-binding domain of NAD-dependent enzymes is the most diversified. Even isoforms of IMPDH show differences. In IMPDH2, the adenine ring of NAD is sandwiched between the side chains of His 253 and Phe 282. Thr 45 is in close contact with the C2 and N3 of the adenine ring. In IMPDH1, these amino acids are replaced by Arg 253, Tyr 282, and Ile 45 [44]. Modification of the adenine system at C2 was therefore an attractive idea. Indeed, the substitution of TAD (K_i=110 nm) with an ethyl group afforded a 2-ethyl analogue 3 (K_i=1 nm), which proved to be 100-fold more potent against IMPDH1 than TAD (K_i=110 nm). Insertion of other substituents (such as iodo, phenyl, amino, aminophenyl, ethylene) at the C2 of adenine afforded generally more potent compounds, but no significant selectivity against IMPDH1 or IMPDH2 was observed.

The corresponding bis(phosphonate) MAD analogues substituted at the adenine moiety were then synthesized. The 2-ethyl analogue 4 was a little less potent against IMPDH1 (K_i=16 nm) and IMPDH2 (K_i=38 nm) but a new 4-pyridyl derivative 5 showed sub-nanomolar inhibition of the type 1 isoform (K_i=0.5 nm) with K_i=14 nm against IMPDH2. As expected, these compounds inhibited the proliferation of K562 cells of CML (4, IC₅₀=1 μm, and 5, IC₅₀=0.4 μm). They were as potent as the tyrosine kinase inhibitor gleevec (imanitib, IC₅₀=0.6 μm), heralded as a ‘magic bullet’ against CML. It is likely that these two MAD analogues 4 and 5, due to a different mechanism of action, might show synergistic effect with imanitib (or with second-generation drugs, i.e. dasatinib and nilotinib) [45].

Mycophenolic adenosine analogues without dinucleotide connection

The P-groove subdomain is a large, shallow cavity where interactions of the enzyme with the phosphate groups of the cofactor analogue are weak, maintained mainly by water molecules (represented by red stars in Figure 7). It is likely, therefore, that modification of the natural pyrophosphate linkage or even replacement of its phosphorus atoms may be well tolerated. In fact, we synthesized a variety of NAD analogues in which the mycophenolic and adenosine moieties were connected by linkers of a different length and chemical character. Bis(sulfonamide) or triazole analogues (Figure 8) illustrate well such limitless possibilities.

Figure 8

Structures of bis(sulfonamide) and triazole mycophenolic adenosine analogues.

No matter how drastic the modification – all non-dinucleotide analogues were found to be potent nanomolar inhibitors of IMPDH (range=34–167 nm for the isoform type 2). Unexpectedly, however, when the bis(phosphonate) group was replaced and the negative charge removed, all of the neutral analogues lost their ability to inhibit proliferation of cancer cell lines. It is possible that these neutral nicotinamide adenine analogues are not able to enter the cell.

Numerous carboxylic acids were used for esterification of poorly absorbed nucleosides to make prodrugs that helped transfer polar nucleosides across the cell membrane. A well-known example is isobutyric acid ester of 2′-deoxy-2′fluoro-2′-C-methylcytidine (RG7128) [46], a predecessor of sofosubuvir (sovaldi), the blockbuster anti-hepatitis C drug (Figure 9).

Figure 9

Structures of RG7128 (isobutyric prodrug of PSI 6130) and phosphoroamidate prodrug Sovaldi.

We reversed this idea and used a nucleoside (adenosine) to facilitate the transport of MPA into the cell. We synthesized mycophenolic adenosine ester 6 and found it, as expected, to be a potent inhibitor of IMPDH1 (K_i=163 nm), IMPDH2 (K_i=102 nm), as well as sub-micromolar inhibitor (IC₅₀=0.5 μm) of K562 cell proliferation. This adenosine prodrug, in contrast to morpholinoethyl ester of MPA (MMF, FDA approved; Figure 10) was found to be resistant to glucuronidation.

Figure 10

Structures of MMF and mycophenolic adenosine ester 6 and mycophenolic adenosine amide 7.

In addition to its inhibitory activity against IMPDH, ester 6 is expected to serve as a depot form of MPA due to the action of cellular esterases. As cellular esterases work in the stereoselective manner, we also synthesized a mycophenolic adenosine ester 6a containing l-adenosine. This isomer was found to be only 2-fold less potent (IMPDH2 K_i=0.23 μm) than its natural counterpart and approximately a 10-fold weaker inhibitor of the type 1 isoform (IMPDH1 K_i=2.5 μm). The stability of l- and d-isomers in human plasma was dramatically different. Surprisingly, unnatural l-isomer 6a was hydrolyzed by esterases much faster (half-life t_1/2=~1 h) than natural d-isomer 6 (t_1/2=~24 h). Next we prepared the amide derivative 7, which showed even more potent inhibition of IMPDH2 (K_i=84 nm) and, as expected, was neither glucuronidated nor hydrolyzed. However, it showed no activity in vitro against K562 cell proliferation.

To improve the metabolic stability of MPA-adenosine esters, compounds with more stable linkers could be prepared. Insertion of an amino acid between MPA moiety and adenosine would lead to new analogues with improved metabolic profile [47]. Earlier, we and others [48, 49] attached methyl esters of a number of amino acids to MPA, among them l- and d-valine methyl esters. Both compounds were found to be moderate inhibitors of IMPDH2 (K_i~1 μm). The hydrolysis of the methyl ester group and the attachment of 5′-amino adenosine afforded two new mycophenolic-valine-adenosine diamides 8 and 8a, which showed a dramatic improvement of inhibitory activity against IMPDH (Figure 11).

Figure 11

Mycophenolic-(d)- and (l)-valine-adenosine diamides 8 and 8a.

The compound 8 containing d-valine inhibited IMPDH1 (K_i=3 nm) 30-fold more potently than IMPDH2 (K_i=88 nm). This example illustrates well the ability of the P-groove for accommodation of much larger structural elements then the natural pyrophosphate group. Actually, compounds 8 and 8a, as well as most of other mycophenolic-amino acid-adenosine diamides (not discussed herein), showed potent enzymatic activity but no activity at the cellular level. Thus, our search for metabolically suitable linkers continues.

Back to the future

Recently, ribavirin has been reported as a promising drug against cancer, including breast, colon, lung, and AML [50–53]. The compound entered numerous clinical studies against variety of cancers and is still being evaluated in the clinic [54]. It was clearly established that the antitumor activity of ribavirin is related to the inhibition of an oncogene, the eukaryotic translation initiation factor eIF4E [55–57]. Interestingly, ribavirin has also anticancer potential because its 5′-monophosphate acts as a mimic of IMP and inhibits IMPDH [58, 59] often overexpressed in cancer cells [60, 61]. Synergistic inhibition of cancer cell proliferation by ribavirin and tiazofurin has been reported [62]. Ribavirin 5′-triphosphate is structurally similar to the m⁷G ‘cap’ structure (Figure 12), and its cap-related activity affecting translation of RNA has been studied and reported (also in relation to cancer) [63–65].

Figure 12

The m⁷G cap and ribavirin ‘cap’ structures.

Finally, it was found that ribavirin inhibits histone methyl transferase, and it was declared as ‘a tri-targeted antitumor repositioned drug’ [66]. Consequently, it is possible that ribavirin effectiveness in patients with AML is because of the combination of its multitarget activity.

It is interesting to note that in the recent phase II clinical trial, the problem of resistance to ribavirin emerged. It was found that it is caused by an inducible glucuronidation, possibly at the amido group of the drug [67]. It has now been proposed to use ribavirin in combination with inhibitors of glucuronyl transferases.

In conclusion, both ribavirin and NAD analogues (distant relatives of ribavirin) are still under investigation as potential anticancer drugs. It remains to be seen if these compounds would emerge in the near future as agents potent and safe enough to be introduced to the clinic.