Class 1a drugs are nucleobase, nucleoside or nucleotide analogues that are erroneously incorporated in place of canonical purine or pyrimidine bases into RNA, usually having similar effect on DNA. For a summary of the canonical nucleobases, nucleoside and nucleotides see Figure 2. These compounds are subjected to processing by cellular metabolic enzymes to generate multiple active compounds that impinge nucleotide biology, earning them the broad name of antimetabolites. As well as being incorporated into RNA or DNA, these compounds can inhibit metabolic enzymes to prevent synthesis of new nucleic acid strands by depleting nucleotide pools.

5-flurouracil (5-FU), first synthesised as an anti-cancer compound by C. Heidelberger in 1957 (Duschinsky et al., 1957), is a classic example in which the nature of the drug is being re-explored in relation to RNA alterations and RNA damage. 5-FU is an uracil analogue with fluorine in place of a hydrogen at ring position 5. This gives 5-FU the ability to disrupt metabolism modifications that occur at position 5, and also alter the Hoogsteen edge when incorporated into nucleic acids. Throughout the late 1950s to early 2000s, research boomed in the study of 5-FU’s MOA, revealing profound RNA damaging effects of 5-FU in strains of E. coli, viruses and various forms of cancer (Horowitz and Chargaff, 1959; Andoh and Chargaff, 1965; Graham and Kirk, 1965; Kadowaki et al., 1965; Wilkinson and Crumley, 1977; Spiegelman et al., 1980a; Spiegelman et al., 1980b; Ghoshal and Jacob, 1997). Subsequently, the concept that 5-FU targets RNA was surpassed by the rise in interest of the role 5-FU plays in limiting DNA synthesis, which is engrained as the current dogma. In truth, compounds such as 5-FU have multiple MOAs, which we still do not fully understand (Figure 3A). This is despite 5-FU being given to around 2 million patients worldwide each year to treat diseases such as colorectal, breast and pancreatic cancer. The dogmatic MOA is that 5-FU inhibits thymidylate synthase (TS); an enzyme essential for the conversion of deoxyuridine monophosphate to deoxythymidine monophosphate (dTMP) via methylation of uridine at position 5. This reduces the pool of deoxythymidine triphosphate (dTTP) for use by DNA polymerases, stopping the synthesis of new DNA strands. At the same time 5-FU leads to incorporation of fluorodeoxyuridine triphosphate (FdUTP) into DNA hence replacing dTTP. It was shown that FdUTP in DNA is recognised as an error, stalling DNA synthesis and promoting cancer cell death (reviewed in Miura et al., 2010). Although this is the widely accepted view of 5-FU, the 5-FU metabolite fluorouridine triphosphate (FUTP) is in fact incorporated 3,000-15,000 fold more into RNA than FdUTP is into DNA, and knockdown of DNA repair proteins in colon cancer cell lines, fails to alter cancer cell sensitivity to 5-FU (Pettersen et al., 2011). Furthermore, a recent study showed that resistance to 5-FU is driven by functional changes in RNA binding proteins and not by changes in TS activity (Liang et al., 2022). Consistent with this, 5-FU blockade of TS activity has also been challenged; supplementing 5-FU treated cancer cells with dTMP, which is metabolically downstream of TS, did not significantly reverse 5-FU cytotoxicity (Pettersen et al., 2011), suggesting that 5-FU is not dependent on blocking TS activity. An intriguing recent study showed that this TS blockade can in fact act as a molecular trap for the RNA precursor metabolite FdUMP (Kurasaka et al., 2022). TS in 5-FU resistant colon cancer cells trap more FdUMP than cells naïve to treatment, resulting in reduced cell death due to reduced active 5-FU metabolites. How this relates to 5-FU incorporation into RNA is of great interest.

In support of RNA damage as the determinant of the response to 5FU, mice administered uridine to compete with 5-FU metabolites for RNA incorporation 2 h after 5-FU dosing showed a reduction of intestinal epithelial cell apoptosis compared to 5-FU alone (Pritchard et al., 1997). Furthermore, the level of RNA incorporation of 5-FU correlates with cytotoxicity in human breast and colon cancer cell lines (Kufe and Major, 1981; Glazer and Lloyd, 1982). In agreement, measuring bulk incorporation of 5-FU metabolites into RNA identified that the presence of RNA damage correlates with drug efficacy in a panel of responsive and non-responsive model cell lines (Chen et al., 2023). In this study, it was also found that in cells from tumours that respond to 5-FU in the clinic it was the active incorporation of 5-FU metabolites into RNA that determined drug response, not any effect on overall RNA or DNA synthesis. Overall, these studies point to a considerable role for RNA damage in the cytotoxicity of 5-FU, but also highlights heterogeneity in the molecular effects of the drug. Understanding all of 5-FU’s molecular impacts, and how these vary between cell types, will be the first step towards directing 5-FU towards specific mechanisms and cell types. Previous research has shown that pairing other anti-cancer drugs or biological compounds with 5-FU can indeed impact how 5-FU is utilised in cells. In the colon cancer cell line HCT-8 pre-treatment with thymidine sensitised the cells to 5-FU and reduced viability in comparison to single treatments (Greenhalgh and Parish, 1990). Although the study used 5-FU concentrations above clinically achievable levels it is important to note that pre-treatment with thymidine increased the quantity of 5-FU incorporated into RNA by 1.5-fold in preference to DNA. Furthermore 5-FU increased overall nucleolar surface area and cell size, likely indicating an accumulation of pre-ribosomes containing 5-FU (Greenhalgh and Parish, 1990).

How pervasive is 5-FU incorporation into RNA? With new technological advances in the way we isolate and detect RNA, 5-FU-damaged RNA species such as mRNAs, rRNAs, microRNAs, tRNAs can readily be identified. Therizols et al. (2022), adapted a liquid chromatography method coupled to high resolution mass spectrometry and quantified the number of 5-FU containing ribosomes across multiple colorectal cancer cell lines as well as breast cancer and pancreatic cancer cell lines. The group found between 7 and 14 5-FU molecules were present in cytoplasmic ribosomes, confirming that these fluorinated ribosomes can escape quality control processes. These results were reflected in colorectal cancer xenografts derived from mice and tissue from patients treated with 5-FU. Interestingly, fluorinated ribosomes had altered activity and favoured the translation of specific mRNAs. Isolation of fluorinated ribosomes by sucrose gradient revealed increased translation of the cancer promoting mRNA insulin growth factor-1 receptor (IGF-1R) compared to non-fluorinated ribosomes (Therizols et al., 2022). Due to the multifaceted nature of 5-FU it is difficult to attribute these changes to alteration solely in rRNA, but the link to 5-FU damaged RNA in turn providing a resistance mechanism is of incredible interest. Of note the relative abundance of 5FUTP to canonical UTP after a 24 h incubation with 5-FU in HCT116 was quantified in a metabolomic analysis at ∼1:300 (Ser et al., 2016). Therizols et al. (2022) performed the exact same experiment (time, drug concentration, cell line) and observed a relative incorporation of 5FUTP:UTP into rRNA of ∼1:200. The convergence of this data is startling and indicates the extent of 5-FU damage to RNA is likely governed largely by drug exposure. Consistent with high levels of 5-FU incorporation in ribosomes, a genome wide screen in yeast found that 5-FU induced haploinsufficiency in only 7 genes, all of which are related to the processing of the ribosome subunits (Lum et al., 2004). This included four of the ten components of the RNA degrading exosome. These data imply that 5FU perturbs exosome activity in degrading unwanted RNA species, and furthermore implies this is the major pathway linked to cell fitness, at least in this yeast model. The biophysical changes in nucleic acids that result from 5-FU incorporation are not fully understood. Due to presumably random incorporation of 5-FU into RNA species it is difficult to study specific effects. Structurally, 5-FU does not disrupt Watson-Crick base pairing, instead sitting on the Hoogsteen edge. It has been suggested that Hoogsteen base pairing is more stable between 5-fluorouridine:adenosine than the canonical uridine:adenosine, but whether this has functional consequences is unknown (Abdulnur, 1976).

As well as rRNA, 5-FU incorporation has functional consequences in other RNAs. 5-FU stops the formation of pseudouridine, which is essential for the activity of U2 snRNA in splicing. 5-FU was found to inhibit U2 pseudouridylation and thus stop the splicing of pre-mRNAs in a Xenopus model where endogenous U2 snRNA was depleted (Zhao and Yu, 2006). Using both in vitro transcribed 5-FU RNAs and in vivo synthesised 5-FU RNA this work demonstrated how 5-FU metabolites have profound and specific effects on essential RNA biology. Similarly, 5-FU inhibits the post-transcriptional methylation of uridine at position 5 of most tRNAs, which is an essential modification for their activity. Using 5-FU as a chemical crosslinking probe in a novel FICC-seq methodology, Carter et al. defined the mammalian methyl-transferase responsible for uridine methylation due to the pervasive incorporation of 5-FU within tRNAs (Carter et al., 2019). Expanding this approach to identify more binding partners of 5-FU-damaged RNA should allow the molecular response to be defined in detail. As a final remarkable example of 5-FU incorporation having functional consequences, 5-FU in the mRNA synthesised from mutant TP53 alleles was able to promote readthrough of the mutation, restore p53 protein expression and induce apoptosis (Palomar-Siles et al., 2022). More studies need to be conducted to elucidate other RNA targets of 5-FU and how this can influence multiple biological factors such as cancer progression, resistance, known side effects and most importantly patient outcome.

Azacytidine is an antimetabolite developed in the late 1960s and is well documented for its effectiveness against leukaemia (Von Hoff et al., 1976; Veselý and Čihák, 1977). It is approved for use in patients with myelodysplastic syndrome (MDS) and acute myeloid leukaemia (AML). Azacytidine is a cytidine nucleoside analogue with a nitrogen substituted for carbon at position 5. This substitution leaves Watson-Crick base pairing unaffected and has been reported to have little effect on Hoogsteen pairing (Oellerich et al., 2019). Azacytidine has a similar MOA to 5-FU in terms of RNA and DNA incorporation (Figure 3B). Early studies showed that administration of azacytidine to leukemic mice depleted endogenous 2′deoxy-cytidine uptake into cells and blocked incorporation into DNA, but did not affect cytidine uptake into RNA. In addition, a heavy labelled radioactive azadeoxycytidine confirmed efficient integration into DNA (Von Hoff et al., 1976). In vitro investigations demonstrated cells arrested in G1 phase when treated with azacytidine did not synthesise DNA however if cells had already entered the S phase at the time of treatment, continued to synthesise DNA but failed to undergo mitosis (Tobey, 1972), suggesting azacytidine’s cytotoxic effects can be delayed and halts cell cycle progression. A well-documented MOA of azacytidine is the inhibition of cytidine methylation in de novo DNA strands. Since methylation in cancer can lead to selective gene expression, e.g., hypermethylation of tumour suppressor genes such as retinoblastoma (RB1) (Greger et al., 1989) and cell cycle checkpoint genes p16INK4A and p14(ARF) (Baur et al., 1999) and hence their repression, or the hypomethylation of mismatch repair (MMR) genes such as human-mut-L-homologue (hMLH1) (Fleisher et al., 2001), reactivation of cancer suppressive genes by azacytidine is a mechanism of interest (Christman, 2002). Although reports demonstrate the DNA incorporation and damaging effects by azacytidine, it is known that the drug is not limited to incorporating into DNA. In fact, 80%–90% of azacytidine is incorporated into RNA, with only 10%–20% incorporated into DNA (Li et al., 1970). Like its effects on DNA, mounting evidence confirms azacytidine inhibition of RNA methylation. The RNA methyltransferase DNMT2 enzyme methylates tRNA^Asp at cytidine 38 (C38). tRNA^Asp charging and delivery of aspartate to translating ribosomes is dependent on C38 methylation by DNMT2 (Shanmugam et al., 2015). Treating the colorectal cancer cell line HCT116 and the MCF7 breast cancer cell line with azacytidine significantly reduced the number of C38 methylated residues in tRNA^Asp (Schaefer et al., 2009). Azacytidine also reduced the viability of myeloid leukemic cell lines and was associated with reduced methylation of C38 and C39 in tRNA^Asp. Therefore, the cytotoxic effects of azacytidine may stem from failure to load tRNAs with amino acids, which would cause mistranslation and ribosome pausing. The MOA of azacytidine is broad and has been shown that when incorporated into newly synthesised RNA strands triggers endoplasmic reticulum stress response and cell death (Pawlak et al., 2022). Despite the success achieved with azacytidine treatment of leukemic cancers, in which the 12-month overall survival rate is 75% in MDS, 60% AML and 75% chronic myelomonocytic leukaemia (CMML), more than 60% of patients do not respond to the drug and over 80% do not experience complete disease remission (Helbig et al., 2019). It is interesting to speculate on how patient response is linked to RNA damage. Understanding the specifics of what dictates the levels of and response to azacytidine damaged RNA may be central to improving patient outcomes.

Other nucleobase/nucleoside analogues that cause Class 1a RNA damage are the thiopurines; azathioprine, 6-mercaptopurine and 6-thioguanine. These are commonly used to treat AML but have also proved effective as immunosuppressants and are prescribed to patients with chronic bowel diseases, psoriasis, and rheumatoid arthritis. The known MOA of thiopurines is their misincorporation in place of canonical guanosine bases during DNA synthesis while thiodeoxyguanosine triphosphate (TdGTP), also leads to the depletion of purine pools by inhibiting metabolic enzymes (Figure 3C). Thioguanines incorporated into DNA are subsequently heavily methylated to form S6-methylthioguanine in nascent DNA strands. Erroneous pairing of S6-methylthioguanine to thymidine is recognised as an error by the MMR system, further contributing to cytotoxicity (Swann et al., 1996). The azathioprine metabolite 6-thioguanosine can also inhibit GTP-dependent proteins such as the GTPase Rac1 (Tiede et al., 2003), highlighting the multiple mechanisms of thiopurine compounds (Coulthard et al., 2018).

Thiopurines have been known to incorporate into RNA for decades with 6-thioguanosine seen pervasively in rRNA and mRNA after exposure in cultured cells (MELVIN et al., 1978). Furthermore, high levels of 6-thioguanosine in RNA significantly reduced overall protein synthesis and protein half-life, which was attributed to an erroneous decoding of damaged RNA. Similar effects were seen on RNA and protein synthesis in vivo (Kwan et al., 1973). Consistent with deleterious effects of 6-thioguanosine upon incorporation into RNA, leukemic cells treated with 6-thioguanine exhibited greater adenosine-to-inosine RNA editing via induction of the editing enzyme ADAR2 (You et al., 2023). Furthermore, this work showed that 6-thioguanosine incorporation was biased towards shorter RNAs than longer RNAs, giving an intriguing insight into the unknown RNA damage biases of class 1a compounds. In terms of relative effects on DNA and RNA, 6-mercaptopurine metabolites were shown to be incorporated into DNA around 4 times more prevalently than RNA in a mouse lymphoma cell line (Tidd and Paterson, 1974).

Although thiopurines have proved highly successful in the treatment of malignant disease including patients undergoing organ transplant there is cause for concern of drug-induced secondary cancer, particularly after long term use (Karran et al., 2003; Karran, 2007). Further investigation is needed into the outcomes of thiopurines incorporation into RNA. Understanding the consequences of thiopurine damaged RNA may significantly change how the drugs are used and increase their safety in long term use.

The cytidine nucleoside analogue gemcitabine is used for the treatment of pancreatic tumours, as well in some cases of ovarian and bladder cancer. Structurally it differs from cytidine due to the inclusion of two fluorines in place of the 2′hydroxyl group and its adjacent hydrogen atom in the ribose sugar. Thus, unlike the other drugs in this class listed above, the base of gemcitabine is identical to its orthologue, with the difference in the sugar. Gemcitabine undergoes processing to impinge on multiple aspects of pyrimidine metabolism (Bergman et al., 2002) (Figure 3D). Deamination of gemcitabine and its metabolites generates uridine analogues containing the same difluororibose modification that inhibit thymidylate synthase much like 5-FU metabolites. However, gemcitabine is also metabolised into phosphorylated forms that can be incorporated into DNA and RNA. The relative incorporation into RNA or DNA was seen to be comparable for some cell lines, but divergent in others, with either a preference for RNA or DNA noted (Ruiz van Haperen et al., 1993). The cell line with the highest incorporation of gemcitabine into RNA was the most sensitive to drug treatment. Gemcitabine causes termination of DNA synthesis after the next nucleotide has been incorporated, through a well described mechanism that is difficult for cells to resolve (reviewed by Plunkett et al., 1995). In contrast, gemcitabine appears to have a less of an inhibitory effect on RNA synthesis, which persists after exposure, giving a strong time dependent increase in gemcitabine incorporation into RNA (Ruiz van Haperen et al., 1993).

Gemcitabine can be incorporated into RNA (and indeed DNA) as 2′,2′-difluorinated cytidine, but also as 2′,2′-difluorinated uridine, due to the action if cytidine deaminase (Bergman et al., 2002). Indeed, exposing cells to 2′,2′-difluorinated uridine resulted in incorporation into RNA and DNA, indicating that it can be phosphorylated and is an acceptable substrate for RNA and DNA polymerases (Veltkamp et al., 2008). Derissen et al. (2018) analysed the pharmacokinetics of the cytidine and uridine metabolites derived from gemcitabine in patient plasma and peripheral blood cells. They found higher levels of the cytidine analogues than the uridine analogues, but observed triphosphate forms of both nucleotides at levels where nucleotide incorporation occurs in vitro. Importantly, this work only quantified free nucleotides, with the gemcitabine metabolites incorporated into RNA and DNA removed by precipitation with perchloric acid prior to analysis. Determining the extent of RNA damage by these antimetabolites is now a priority to shed new light on their pharmacokinetics and target engagement.

Class 1b agents are also incorporated into cellular RNA, but in doing so terminate the transcription process, resulting in truncated RNAs. This termination occurs due to either i) the lack of a 3’ hydroxyl group for conjugation of the subsequent nucleotide or ii) the presence of other moieties that block the addition of the next nucleotide. They have primarily been designed and applied as antivirals. Here we focus on targeting the two types of retroviruses, whose RNA genomes are either reverse transcribed into DNA or simply duplicated directly into new RNA copies. Both activities rely on viral RNA dependent DNA/RNA polymerase enzymes, with these enzymes the target of the nucleoside antivirals. Compound design has aimed to specifically inhibit the synthesis of these DNA or RNA genomes by the viral DNA- or RNA-dependent RNA polymerases (Ami and Ohrui, 2021). But, given the structural similarity of the modified nucleotide antiviral metabolites to the canonical nucleotides there are significant off-target effects on the host cell polymerases.

Nucleoside antivirals were rapidly approved to combat the human immunodeficiency virus (HIV) pandemic due to their ability to halt reverse transcription of the HIV RNA genome (Feng, 2018). Combinations of these compounds remain the standard of care for people infected with HIV, who require constant suppression of reverse transcription to reduce further genomic incorporation of the viral genome. The first nucleotide antiviral approved to treat HIV was zidovudine (also called azidothymidine) which has an azide group at the 3′ position, causing termination of reverse transcription (Figure 4). Over 20 similar compounds with changes at the 3’ (as well as 1′, 2′ and 4′) positions have also been approved for treatment of HIV, with these and other compounds also approved for viruses such as hepatitis C virus (HCV) and SARS-CoV2 (Ami and Ohrui, 2021). Different modifications at each position change the efficacy of these compounds against specific viruses, while also altering the side effects associated with them (Figure 4). This is a clear indication that the molecular targets behind the side effects are the host (human) nucleotide polymerases, which will utilise the different antivirals to differing extents (Moyle, 2000). Fortunately, the human polymerases appear to be more selective than the viral polymerases, generating a therapeutic window for viral targeting, with the toxicity profile of these compounds reduced by increasing divergence from the canonical nucleotides (Ami and Ohrui, 2021) (Figure 4). An excellent example of this in action is the antiviral stavudine, which is thymidine analogue lacking both the 2′ and 3′ hydroxyl groups—i.e., the minimal divergence from the canonical nucleoside possible (Figure 4). Stavudine is effective as an anti-HIV agent, but patients suffer from severe side effects, leading to its removal from the market in 2020.

The toxicities arising from these antivirals range from peripheral neuropathy, cardiomyopathy, lactic acidosis and cytopenia (Feng, 2018). Many of these side effects stem from changes in post-mitotic cell function, leading to an investigation of the mitochondrial DNA polymerase, rather than the nuclear DNA polymerases, as the molecular mechanism behind the side effects. Mitochondrial toxicity was first attributed to the inhibition of the mitochondrial DNA polymerase, but in some cases toxicities were observed in the absence of changes in mitochondrial DNA levels (Moyle, 2000). This led to an interest in the mitochondrial RNA polymerase (POLRMT) as an alternative source of toxicity. Arnold et al. (2012) used recombinant POLRMT and purified eukaryotic RNA polymerase II to screen for their ability to utilise 12 nucleoside antiviral analogues that cause termination of HCV RNA genome synthesis. The authors found that all the analogues were viable substrates for POLRMT but were less effectively used in their RNA polymerase II assays. Of note, the compounds which retain their 3′ hydroxyl groups did not cause chain termination by the host cell RNA polymerases. Likewise, a cell-based screen of compounds that had failed in trials against HCV due to toxicity found that they selectively inhibit POLRMT over host cell DNA polymerases, with little effect on the nuclear host RNA polymerases (Feng et al., 2016). Expanding this further, additional nucleoside antivirals were found to be substrates for POLRMT in vitro, with some resulting in chain termination despite possessing a 3′ hydroxyl while others did not terminate elongation (Jin et al., 2017; Lu et al., 2018). These methodologies open the possibility to define the substrate repertoire of nucleosides that POLRMT can use, which should now be applied to the obligate chain terminating antivirals lacking 3′ hydroxyls. Remdesivir, a prodrug that is metabolised to 2′cyano-modified adenosine analogue, is the only approved antiviral for patients with SARS-CoV2 infection and has shown remarkable tolerance. Consistent with this, molecular profiling of how remdesivir interacts with host cell polymerases showed very little impact (Xu et al., 2021) (Figure 4). Going forward, performing in silico simulations of POLRMT binding to antiviral nucleosides is an exciting development (Freedman et al., 2018), which may also be possible for the recently solved structures of the human nuclear RNA polymerases (Ramsay et al., 2020; Aibara et al., 2021; Misiaszek et al., 2021).

The success of antiviral nucleosides is an excellent demonstration of structure-function relationships between the analogues and the polymerases that require their canonical analogues for function. It remains to be seen if these compounds designed to be antivirals may be effective at targeting genetic human diseases due to their ability to target human polymerases. Resistance to nucleoside antivirals occur rapidly through mutations in the viral polymerases (Seley-Radtke and Yates, 2018), and it will be interesting to see if human polymerases are also able to become resistant under the same selection pressure.

The simplest alkylating agents covalently conjugate methyl groups to specific reactive positions on the bases of DNA and RNA. Most of the analyses in this area have focused on the relative damage of DNA base positions and their consequences. This can guide assumptions about how RNA damage may mirror that of DNA but highlights the need to study this in more detail. The methylating agent temozolomide is approved for treatment of malignant glioma; it is a prodrug that undergoes non-enzymatic activation to form a highly reactive methyldiazonium ion (Denny et al., 1994) (Figure 5). This ion methylates a range of positions on DNA, and presumably RNA, with a preference for N7 of guanosine (70%), N3 of adenosine (9%) and O6 of guanosine (5%) (reviewed in Barciszewska et al., 2015). The relative preferences in RNA are not known but are likely different to those in DNA due to the higher solvent accessibility of RNA’s bases. Consistent with this, endogenous RNA in human cells in culture receives around 20-fold more oxidative damage per base than DNA (Hofer et al., 2005). Furthermore, the methylating tool compound methyl methanesulfonate, often used to mimic temozolomide damage in laboratory experiments, results in over 10-fold more RNA damage than DNA damage in E. coli (Vågbø et al., 2013) (Figure 5). Intriguingly, in this study an enrichment for RNA over DNA methylation was seen at the N1 position of adenosine, potentially due to this position participating in base pairing in helical DNA while being more readily solvent exposed in RNA. The overwhelmingly higher incidence of RNA damage over DNA damage by these monovalent methylating agents lead to the interesting dual hypothesis that RNA acts both as a sponge for damage to protect DNA and also as an early warning system of potential genotoxicity (Vågbø and Slupphaug, 2020). Understanding what responds to this early warning is of great interest, and if inhibited could enhance the effects of these RNA/DNA damaging agents in the future.

The weaponised mustard gas that lead to the analysis of similar compounds for anti-cancer therapy was a bifunctional sulphur mustard called bis-chloroethyl sulphide (Singh et al., 2018). Using similar chemicals to treat patients brought success with the nitrogen mustards, with compounds such as chlorambucil used to treat leukemic patients. Subsequent improvements of this same bifunctional mustard moiety resulted in the widely used compound cyclophosphamide (Figure 6). Cyclophosphamide is used to treat several neoplasms, including lymphomas, myelomas, breast cancer, ovarian cancers, and neuroblastoma. Due to cyclophosphamide’s immunosuppressive properties, it is also used for treatment of autoimmune diseases such as lupus and multiple sclerosis (Ogino and Tadi, 2022). As a bifunctional alkylating agent, cyclophosphamide can form cross links between nucleic acid strands which induce apoptotic events. Surprisingly, the DNA adducts of the major active component of cyclophosphamide, phosphoramide mustard, primarily form monofunctional adducts on guanosines or the phosphate backbone with a maximum of only 12% bifunctional crosslinks observed (Povirk and Shuker, 1994) (Figure 6). This low incidence of crosslinking is consistently seen across all similar bifunctional alkylating compounds.

Cyclophosphamide metabolic activation occurs in the liver with the first event being hydroxylation by the enzyme cytochrome P450 to non-toxic products 4-hydroxycyclophosphamide followed by aldophosphamide. Aldophosphamide is further processed by phosphodiesterases to acrolein and phosphamide mustard. Both products are capable of DNA alkylation and inducing cell death (reviewed by Voelcker, 2020) (Figure 6). Studies in mid to late 20th century evidence the ability of cyclophosphamide to also exert cytotoxic effects through RNA damage. Neo-natal mice treated with cyclophosphamide demonstrated a significant decrease in the incorporation of ¹⁴C uridine into newly synthesised RNA strands. Cyclophosphamide treatment reduced RNA synthesis in the livers and brains of these mice by 64% and 36% respectively. (Short and Gibson, 1973). In addition, studies using E. coli strains showed that the cyclophosphamide metabolite 4-hydroxyperoxycyclophosphamide (4-OOH-CP) could be used to determine the relative effects of cyclophosphamide on RNA and DNA. RNA was predominantly targeted by 4-OOH-CP, in a dose-dependent manner (Frank et al., 1981). Furthermore 4-OOH-CP targeted specific RNA species and prevented charging of tRNA^Leu with leucine amino acids. These RNA effects resulted in cyclophosphamide stalling protein synthesis (Frank et al., 1981), most likely through interrupting RNA synthesis and regulatory processes such as amino acid loading of tRNAs. Given that this work is 50 years old, these mechanisms should be revisited using newer technologies to elucidate other RNA targets that could be linked to cytotoxicity. The toxic by-product of cyclophosphamide metabolism, acrolein, can also directly damage DNA and RNA (Figure 6). Acrolein is released in the formation of aldophosphamide and is known to have mutagenic effects (Singh et al., 2014). Patients taking cyclophosphamide for prolonged times, for example, to treat vasculitis disease, have a higher risk of developing lymphoma or bladder cancer (Shang et al., 2015). Using A549 cancer cells a study showed the RNAse and DNase sensitive accumulation of acrolein within RNA-rich nucleoli (Wang et al., 2016). Functionally, this perturbed rRNA synthesis, as shown by reduced nascent RNA synthesis within nucleoli and a decrease in pre-rRNA 45S transcript levels. Consistent with defective ribosome synthesis, acrolein reduced global protein synthesis and the abundance of translating polysomes, while also causing structural changes to the nucleolus. How much the acrolein by-product contributes to cyclophosphamide cytotoxicity in cancer patients is an area that requires further investigation. Since acrolein does have devastating side effects, cyclophosphamide is paired with the acrolein neutralising medicine sodium-2-mercaptoethane-sulfonate (MESNA). MESNA however also induces adverse effects (Links and Lewis, 1999), therefore quality of life for the patient may be heavily reduced. The multimodal effects of cyclophosphamide’s metabolic products on nucleic acid biology offer a fascinating opportunity to study the relative roles of RNA and DNA damage, with huge scope of using this knowledge to improve patient treatment options.

Although not strictly alkylating agents, platinum-based compounds have very similar MOAs by bifunctionally linking to nucleic acids, usually via guanosine N7 but also via adenosine N7. The bifunctional nature of these agents is almost always engaged, usually through intrastrand linkage of eitherdirectly proximal bases. These compounds were discovered by chance in 1965 when Barnett Rosenberg found that products derived from platinum electrodes inhibited bacterial cell division (Weber, 2015). Cisplatin was the first platinum based drug and is frequently used to treat breast, bladder, ovarian and testicular cancers, but also proved effective against head and neck carcinomas, lymphomas and lung cancer (Eastman, 1987; Weber, 2015). The limitations of cisplatin are the multitude of adverse reactions, which brought about a 2nd generation of platinum drug called carboplatin. Unfortunately, both platinum drugs cause neurotoxicity and in addition patients become resistant to therapy after long term use. In effort to limit neurotoxicity and overcome resistance, the 3rd generation platinum drug oxaliplatin was developed. Oxaliplatin is commonly used to treat colorectal cancer in combination with 5-FU (Giacchetti et al., 2000). The MOA of platinum compounds is similar to nitrogen mustards; they form intra-strand and inter-strand cross links of DNA strands. Cisplatin and oxaliplatin form cross-links between guanine bases creating bulky adducts that alter DNA and RNA structure (Eastman, 1987; Chaney et al., 2005). What differs between the drugs are the DNA repair proteins that recognise cisplatin or oxaliplatin induced adducts (Wang and Lippard, 2005). Ultimately platinum drugs are lethal to cancer cells through removal of bulky adducts by MMR and damage recognition proteins. Breaks in DNA strands occur because of these repair processes, with the accumulation of DNA breaks leading to programmed cell death.

The current view that platinum drugs exert cytotoxic effects solely through DNA damage has recently been challenged. A mammalian RNA interference strategy revealed that oxaliplatin’s MOA differs significantly from those of cisplatin and carboplatin (Bruno et al., 2017). Short hairpin RNAs were designed to target genes involved in regulating cell cycle checkpoint and cell death and transfected cells were exposed to platinum drugs, resulting in a gene signature for each compound. Subsequent analysis revealed that oxaliplatin clustered with transcription and translation inhibitors such as actinomycin D, puromycin and rapamycin, while cisplatin and carboplatin showed similar activity to compounds that induce DNA damage repair. The DT40 avian cell line often used in gene manipulation studies due to its high targeted integration efficiency, was used to target DNA damage repair proteins. Forty DT40 cell lines each missing a specific repair protein revealed treatment with cisplatin enhanced cell sensitivity, specifically knock out of homologous recombination repair genes such as BRCA2. In contrast, BRCA2 knockout cells were significantly less sensitive to oxaliplatin (Bruno et al., 2017). This implies that oxaliplatin cytotoxicity is not predominantly occurring through platinum-based DNA lesions. In fact, high-performance liquid chromatography (HPLC) of DNA isolated from the ovarian cancer cell line A2780, confirmed that out of the expected 60% of platinum G-G adducts, 33% was retrieved in cisplatin treated cells while only 7% of adducts could be retrieved in cells dosed with oxaliplatin. Cisplatin is more reactive than oxaliplatin, producing a higher amount of G-G and A-G adducts, ultimately demonstrating that cisplatin binds DNA faster than oxaliplatin. Oxaliplatin adducts do occur, but at a much slower rate and lower frequency (Saris et al., 1996). In support of these findings, assessment of γH2AX expression; a common marker used to confirm the presence of DNA damage repair events, showed oxaliplatin treated cells lacked γH2AX foci compared to cisplatin, indicative of failure to initiate a DNA damage response (Bruno et al., 2017). Furthermore, the same work showed that oxaliplatin did not result in DNA damage measured by the comet assay, whereas cisplatin produced archetype comet tails indicative of damage to genomic DNA. In truth this will not be black and white, both cisplatin and oxaliplatin will cause damage to both DNA and RNA. For example, there is strong evidence that cisplatin damages RNA within ribosomes and arrests elongation along mRNAs in in vitro translation experiments (Heminger et al., 1997; Melnikov et al., 2016). However, the consensus now is that there is a preference for cisplatin to damage DNA and for oxaliplatin to damage RNA (Figure 7).

Evidence also supports a role for oxaliplatin in inducing ribosomal stress, an effect likely linked to damage to rRNA. Ribosomal stress results in activation of the tumour suppressor protein p53, through the elevation of non-ribosome associated ribosomal proteins. Instead, these ribosomal proteins bind and sequester the E3 ubiquitin-protein ligase HDM2 (human double mutant minute 2 homolog) thereby relieving the HDM2 block on p53 and permitting p53 activation (Knight et al., 2016). Following this mechanism, silencing of ribosome proteins such as RPL11 conferred resistance to oxaliplatin, while no effect was seen with RPL11 knockdown in cisplatin treated cells (Bruno et al., 2017). To further support the role of oxaliplatin cytotoxicity mediated through the disruption of ribosome biogenesis, studies have explored redistribution of nucleolar markers and disruption of nucleolar structure in response to drug induced nucleolar stress. The nucleolus is structured into liquid subphases; the fibrillar centre (FC), dense fibrillar centre (DFC) and granular component (GC) (reviewed by Boisvert et al., 2007). Each phase is essential for the processing of rRNA transcripts and assembly of ribosome subunits. Transcription of rDNA occurs in the FC, while pre-rRNA is processed and modified to 18S, 5.8S and 28S rRNA transcripts in the DFC. Within the GC the pre-RNA transcripts mature, the 5.8S is combined with the 28S pre-RNA. Mature rRNAs form ribosome subunits that can then be exported to the cytoplasm (Boisvert et al., 2007). Nucleolar structures can be identified by the abundance of proteins located at each liquid subphase; nucleophosmin (NPM1) and SURF6 are distributed in the GC and fibrillarin (FBL) is abundant in the DFC. NPM1 is of particular interest as it is a chaperone protein that translocates from the nucleolus to the nucleoplasm and has a broad role in the cell including in ribosome biogenesis, DNA repair mechanisms to apoptosis (Box et al., 2016). NPM1 redistribution was demonstrated in A549 cells treated with oxaliplatin, using concentrations below that do not induce excessive cell death (Sutton et al., 2019). No change in NPM1 positioning in the nucleolus was evident in carboplatin treated cells while a small change was seen with cisplatin although only at concentrations with significant cell death. This re-localisation of NPM1 coincides with reduced rRNA transcription. After pulse labelling A549 with radiolabelled ³²P followed by chasing with oxaliplatin, extensive NPM1 re-localisation was observed just 3 h after drug treatment, simultaneously to reductions in the 47S and 28S rRNA transcripts (Sutton and DeRose, 2021). This recent work has given cause to reanalyse oxaliplatin as an RNA-trophic drug. In agreement, a subsequent independent study also found oxaliplatin to result in disintergation of nucleolar liquid subphases, as a consequnce disrupting the rRNA assembly line (Schmidt et al., 2022). Although, as mentioned above, the molecular preferences of the platinum compounds are by no means black and white. For example, cisplatin is able to damage rRNA in yeast, a method developed based on the ability of platinum adducts to inhibit reverse transcription (Plakos and DeRose, 2017). Specific sites are seen to be enriched for platination in a dose responsive manner, which correspond to the solvent exposed sections of rRNA, inviting the speculation that the more RNA-rich human ribosome may present more target sites for platination (Petrov et al., 2014).