1Graduate Field of Biochemistry, Molecular and Cellular Biology, Cornell University, Ithaca NY 14853

1Graduate Field of Biochemistry, Molecular and Cellular Biology, Cornell University, Ithaca NY 14853

2Division of Nutritional Sciences, Cornell University, Ithaca NY 14853.

2Division of Nutritional Sciences, Cornell University, Ithaca NY 14853.

Thymidylate (dTMP) biosynthesis plays an essential and exclusive function in DNA synthesis and proper cell division, and therefore has been an attractive therapeutic target. Folate analogues, known as antifolates, and nucleotide analogs that inhibit the enzymatic action of the de novo thymidylate biosynthesis pathway and are commonly used in cancer treatment. In this review, we examine the mechanisms by which the antifolate 5-fluorouracil, as well as other dTMP synthesis inhibitors, function in cancer treatment in light of emerging evidence that dTMP synthesis occurs in the nucleus. Nuclear localization of the de novo dTMP synthesis pathway requires modification of the pathway enzymes by the small ubiquitin-like modifier (SUMO) protein. SUMOylation is required for nuclear localization of the de novo dTMP biosynthesis pathway, and disruption in the SUMO pathway inhibits cell proliferation in several cancer models. We summarize evidence that the nuclear localization of the dTMP biosynthesis pathway is a critical factor in the efficacy of antifolate-based therapies that target dTMP synthesis.

Folate functions as a family of enzyme cofactors that chemically activate and transfer one-carbon units for fundamental cellular processes including DNA replication and repair, mitochondrial protein synthesis and amino acid interconversions and catabolism 1. The single carbons carried by tetrahydrofolate (THF) are generated from the catabolism of the one-carbon donors that include serine, glycine, sarcosine, dimethylglycine and histidine. Folate-activated single carbons are required for de novo purine synthesis, de novo thymidylate (dTMP) biosynthesis, for the remethylation of homocysteine to methionine and for the synthesis of fmettRNAMet 1. Collectively, these pathways are known as folate-mediated one-carbon metabolism (FOCM), which is a network of interconnected folate-dependent metabolic pathways that are compartmentalized in the mitochondria, cytosol and the nucleus.

In mitochondria, the inner membrane folate transporter, SLC25A32, is essential for mitochondrial folate accretion 2, and the mitochondrial and cytoplasmic folate pools don't freely exchange 3. In mitochondria, one-carbon donors are catabolized to generate formaldehyde in the form of 5,10-methylene-tetrahydrofolate, which is subsequently oxidized to 10-formyl-tetrahydrofolate. Formate is then liberated from 10-formyl-tetrahydrofolate in an ATP-generating reaction, and translocates to the cytosol and nucleus to support FOCM in those compartments 4. Mitochondria also contain an anabolic pathway to generate dTMP for mitochondrial DNA synthesis 5, and mitochondrial FOCM is also required for fmettRNAMet synthesis, which is involved in the initiation of mitochondrial protein synthesis.

In the cytosol, serine and mitochondrial-derived formate are the two primary sources of folate-activated one-carbons for folate-dependent biosynthetic reactions. Formate accounts for 70-90% of total one-carbon units for cytoplasmic FOCM in MCF-7 cells 6,7, whereas serine catabolism in the cytoplasm accounts for 10-30% of total one-carbon units. Serine is a direct source of single carbons in the cytoplasm through the activity of serine hydroxymethyltransferase 1 (SHMT1) which catalyzes the transfer of the one-carbon unit from serine to tetrahydrofolate to form glycine and 5,10-methylene-tetrahydrofolate (Figure 1).

Folate carries one-carbon units required for de novo thymidylate synthesis, de novo purine synthesis, and methylation reactions. THF, tetrahydrofolate; DHF, dihydrofolate; dUMP, deoxyuracil monophosphate; dTMP, deoxythymidine monophosphate; SAM, S-adenosylmethionine; SAH, S-adenosylhomocysteine; MTHFD1, methylenetetrahydrofolate dehydrogenase; (S), synthase; (C), cyclohydrolase; (D), dehydrogenase; SHMT1, serine hydroxymethyltransferase; TYMS, thymidylate synthase; DHFR, dihydrofolate reductase; MTHFR, methylenetetrahydrofolate reductase.

Mitochondrially-derived formate is also a source of 5,10-methylene-tetrahydrofolate. It is condensed with tetrahydrofolate to form 10-formyl-tetrahydrofolate (Figure 1) in an ATP-dependent reaction catalyzed by the synthase (S) activity of the trifunctional enzyme methylenetetrahydrofolate dehydrogenase (MTHFD1). 10-formyl-tetrahydrofolate can serve as a cofactor that supplies the 2 and 8 carbon of the purine ring for de novo purine synthesis. Alternatively, MTHFD1 can further process 10-formyl-tetrahydrofolate to 5,10-methenyl-tetrahydofolate through its cyclohydrolase (C) activity, and then to 5,10-methylene-tetrahydrofolate through the NADPH-requiring dehydrogenase (D) activity of MTHFD1. 5,10-methylene-tetrahydrofolate exists at a branch-point in the FOCM network. It is the substrate for 5,10-methylene-tetrahydrofolate reductase (MTHFR). The NADPH-dependent synthesis of 5-methyl-tetrahydrofolate by MTHFR is essentially irreversible in vivo, hence 5-methyl-tetrahydrofolate is committed to serving as a cofactor for homocysteine remethylation to methionine, catalyzed by the vitamin B12-dependent enzyme methionine synthase (MTR). Methionine is a precursor of S-adenosylmethionine (SAM), which is a required cofactor for over 100 cellular methylation reactions including DNA, histone, protein, and phospholipid methylation and for neurotransmitter synthesis 1. 5,10-methylene-tetrahydrofolate is also the source of one-carbons for de novo dTMP biosynthesis (Figure 1).

The de novo dTMP synthesis pathway catalyzes conversion of deoxyuridylate (dUMP) to dTMP, and involves four enzymes. SHMT1 and MTHFD1 each independently generate 5,10-methylene-tetrahydrofolate from tetrahydrofolate using serine and formate as one-carbon sources, respectively, as described above. Thymidylate synthase (TYMS) utilizes 5,10-methylene-tetrahydrofolate as a cofactor in the conversion of dUMP to dTMP (Figure 1), oxidizing tetrahydrofolate to dihydrofolate. Dihydrofolate reductase (DHFR) catalyzes the NADPH-dependent reduction of dihydrofolate to tetrahydrofolate (Figure 1).

In addition to folate-dependent de novo synthesis, thymidylate synthesis can occur through a salvage pathway. Salvage synthesis occurs through the action of thymidine kinase (TK), which catalyzes the ATP-dependent phosphorylation of the nucleoside thymidine to the nucleotide dTMP. TK1 is found in the cytoplasm and nucleus whereas TK2 is found within the mitochondria 8. Both TK1 expression and nuclear localization increase during S-phase of the cell cycle and in response to DNA damage 9, but TK1 activity is low in quiescent cells. TK2 is not cell cycle regulated, rather it is constitutively expressed and contributes to the mitochondrial dTMP pool to support mitochondrial DNA synthesis, which is required even in post-mitotic tissues 10. In animal cells, salvage pathway synthesis generally does not provide sufficient levels of dTMP to meet DNA synthesis requirements.

Impaired de novo dTMP synthesis creates imbalances in dNTP pools, which are mutagenic 11, stall replication forks, and induce cell cycle arrest and apoptosis12. Impaired de novo dTMP synthesis also increases rates of uracil misincorporation into DNA13,14. DNA polymerases can incorporate either dTTP or dUTP during DNA replication when there is an “A” base on the template strand, and will increase dUTP incorporation into DNA when dTTP becomes limiting. Both rates of cell division and of uracil accumulation into DNA may be affected by the dUTP/dTTP ratio15.

The main consequence of uracil in DNA is DNA single- and double-strand breaks generated by DNA repair enzymes. Mammalian cells possess redundant pathways to both limit and repair uracil misincorporation. Deoxyuridine triphosphatase (dUTPase) catabolizes dUTP making it unavailable for DNA synthesis15. Once dUTP is incorporated into DNA, one of several uracil DNA glycosylases initiate base excision repair. Uracil DNA glycosylase (UNG2 is the nuclear isozyme and UNG1 is the mitochondrial isozyme) is considered to be the primary glycosylase that targets uracil in DNA. However, other glycosylases contribute to uracil excision including Single-Strand-Selective Monofunctional Uracil-DNA Glycosylase 1 (SMUG1). In base excision repair, DNA glycosylases cleaves uracil from the sugar phosphate backbone, leaving an abasic (AP) site. This site is then cleaved by AP endonuclease (APE1) creating a single-strand break. Single-strand breaks can either be processed by short- or long-patch repair, removing a single or several nucleotides, respectively. DNA polymerase ß fills these sites 16,17. Hence, excision of uracil directly leads to the formation single- and double-strand breaks that can be lethal 18. Excessive repair of double-strand breaks can lead to activation of p53-mediated apoptosis 19.

The impact of altered expression of the enzymes that constitute the de novo dTMP pathway on the functioning of de novo dTMP biosynthesis and DNA stability has been studied in genetic knockout mouse models, and in studies of gene variants and inborn errors of metabolism (IEMs).

Mice homozygous for a single amino acid change at position 106 in TYMS, generated through N-ethyl-N-nitrosourea (ENU) mutagenesis, display an early embryonic lethal phenotype, demonstrating that TYMS is essential. Interestingly, mice heterozygous for the mutation are viable but compensate for the mutation by increasing TYMS protein synthesis such that heterozygous mutants express more TYMS in the liver than wild-type littermates 20.

TYMS expression is highly regulated at the transcriptional and translational levels. TYMS transcription is regulated by activating factors that bind to 28-bp tandem repeats in the promoter. The 28-bp tandem repeats in the 5'UTR most commonly exist as either two repeats (2R) or three repeats (3R), though some populations have been reported to have four to nine repeats 21,22. These repeat sequences contain enhancer box (E-box) sequences, which bind numerous upstream stimulating factors to increase TYMS transcription 21. TYMS expression is higher in individuals with the 3R allele compared to the 2R allele, and this affects both response to chemotherapeutic agents and risk for several types of cancer 21. For instance, those carrying the 2R allele may have increased risk for colorectal cancer and head and neck squamous cell cancer, though risk may be different among different populations 21. Recent investigations have assessed the relationship between TYMS polymorphisms and cancer risk on treatment response, but findings are not consistent23-27.

TYMS mRNA levels increase as much as 20-fold in S-phase for DNA synthesis 28, driven by the E2F transcription factor 29. E2F also drives expression of other proteins required for cell cycle progression, DNA damage response, and stress response 30. E2F is regulated by the retinoblastoma (Rb) tumor suppressor protein. In quiescent cells, hypophosphorylated Rb binds E2F and represses transcription. However, in proliferating cells Rb becomes hyperphosphorylated which causes dissociation of Rb from E2F, allowing for E2F-activated transcription. E2F hyperphosphorylation also contributes to stimulating the G1 to S transition in some cell types 30-32. The TYMS protein binds to its own mRNA, which serves to suppress its own translation. Interestingly, TYMS protein can only bind its mRNA when its active site is unoccupied by folate or nucleotide substrates. This regulatory response can modulate response to 5-FU and possibly other antifolate compounds by modulating rates of TYMS translation 33.

There are no reports characterizing a DHFR-deficient mouse, although given its role in regenerating tetrahydrofolate cofactors from dihydrofolate, it is assumed to be essential for life. Patients with IEMs in DHFR usually present with megaloblastic anemia as a result of inadequate RBC precursor synthesis due to insufficient dTMP synthesis to support DNA replication, resulting in impaired cell division 34,35. These patients also exhibit neurological and psychomotor impairments due to low folate levels within the cerebrospinal fluid. Both hematological and neurological symptoms improve clinical symptoms in most cases treated with high doses of reduced folate such as folinic acid 34,35.

Transcription of DHFR, like TYMS, in S-phase is driven by the E2F transcription factor which (for DHFR) works in concert with the Sp1 transcription factor 32. DHFR protein also binds its own mRNA when its active site is unoccupied and thereby autoregulates its translation. One of the first successful and still widely used antifolates, methotrexate (MTX), targets DHFR 36. Binding of MTX to DHFR inhibits DHFR activity, and also disrupts the interaction between DHFR and its own mRNA, relieving the repressed DHFR translation and leading to increased DHFR protein levels 32,37,38. Many cells develop resistance to MTX due to de-repression of DHFR translation by preventing DHFR from binding to its own mRNA 36. A 19-bp deletion polymorphism in intron 1 of DHFR affects both DHFR transcription and translation, and has been linked to increased risk for several folate-associated pathologies, but not all studies agree and/or risk varies by population (reviewed extensively in Abali et al 32).

The Shmt1 knockout mouse is viable and fertile, and its primary phenotype is decreased rates of de novo dTMP synthesis and increased incorporation of uracil in nuclear DNA 39,40. The Shmt1 mouse is the only FOCM enzyme knockout mouse model to develop folate-responsive NTDs 41. Shmt1 heterozygosity on the Apcmin/+ background increased tumor number and tumor load compared to Shmt1+/+ mice on the Apcmin/+ background. A common polymorphism in the SHMT1 gene, C1420T (rs 1979277), exacerbated the effect of the MTHFR polymorphism that increased the risk of cardiovascular disease in the Nurses’ Health Study 42 and also contributed to reduced risk of ALL in individuals who are homozygous for the TYMS 3R allele 43.

SHMT1 is translated through both a cap-dependent mechanism and through an internal-ribosome entry sequence (IRES) (cap-independent) mechanism 44-46. IRES-mediated translation of SHMT1 is mediated by the mRNA binding protein eIF2 and is responsive to several factors including heavy-chain ferritin binding and UV-induced DNA damage 44,46. Interestingly, although the SHMT1 IRES is part of the 5'UTR of the mRNA, cooperation between ITAFs in both the 5'UTR and the 3'UTR work in concert to facilitate SHMT1 translation 45.

Mthfd1gt/gt embryos exhibit an early embryonic lethal phenotype; Mthfd1gt/+ embryos are viable and do not develop NTDs 47. Recently, several IEMs in MTHFD1 in humans have been identified using exome-capture sequencing, and these mutations result in clinical phenotypes including severe combined immunodeficiency (SCID) 48,49. One proband inherited two mutations resulting in decreased MTHFD1 protein levels and severely impaired cyclohydrolase/dehydroganse (C/D) enzyme activity. Cultured fibroblasts from the proband exhibited elevated homocysteine, impaired de novo dTMP synthesis and DNA damage, while de novo purine biosynthesis was unaffected 50. The common MTHFD1 G1958A (R653Q) polymorphism (rs2236225) results in decreased 10-formyltetrahydrofolate synthetase activity 51. The rs2236225 polymorphism is associated with increased risk for neural tube defects and fetal loss as well as breast and gastric cancers 52-55.

Mitochondria also synthesize dTMP from both salvage (via TK2) and de novo (via SHMT2, DHFRL1, and TYMS) pathways. IEMs resulting from either compromised mitochondrial DNA (mtDNA) synthesis or nucleotide precursor synthesis lead to mitochondrial depletion syndromes (MDS). IEMs in genes including TK2, TYMP, RRM2B, POLG, and DGUOK are some examples of genes associated with MDS, and this topic has been recently reviewed elsewhere 56. Chinese Hamster Ovary (CHO) cells lacking SHMT2 exhibit increased uracil in mtDNA and impaired mitochondrial de novo dTMP synthesis 5.

Shortly after the discovery of folate as an essential factor for mammalian DNA synthesis, folate antagonists were developed as anticancer agents to impair DNA synthesis in transformed cells. Pharmaceuticals that target enzymes in the folate-dependent dTMP biosynthesis pathway were developed including aminopterin, methotrexate, and nucleoside analogs 57. 5-fluorouracil (5-FU) is the most widely used of the therapeutic agents that target TYMS and is used in the treatment of breast, colorectal, epithelial, esophageal and pancreatic cancers 58. The antifolates Pemetrexed and Raltitrexed also inhibits TYMS and are used to treat lung and colorectal cancer 57.

The canonical mechanism of 5-FU-mediated cell death is TYMS inhibition. Thymidine phosphorylase (TP) converts 5-FU into 5-fluorodeoxyuridine (5-FdUrd), and TK1 phosphorylates 5-FdUrd, generating 5-fluorodeoxyuracil monophosphate (FdUMP) (Figure 2). The 5-FU metabolite, FdUMP, acts as a dUMP substrate analogue for TYMS. FdUMP, TYMS, and 5,10-methylene-tetrahydrofolate form a stable ternary complex 59,60 resulting in covalent bond formation between FdUMP and TYMS. The inactivation of TYMS activity causes dUMP accumulation and depleted dTTP pools, resulting in “thymineless cell death” 61. Impaired dTMP biosynthesis results in accelerated rates of genomic uracil incorporation 39,40 and DNA repair 16,17. The resulting futile cycles of uracil misincorporation and excision result in an accumulation of DNA strand breaks, as described earlier.

5-FU can be incorporated into DNA, RNA, or bind to TYMS, increasing uracil in DNA. It is unknown whether FdUMP binding impairs TYMS nuclear localization. It is also unknown how TYMS inhibition and impaired formation of the complex of de novo dTMP synthesis enzymes each contribute to uracil incorporation in DNA. Noncoding RNA refers to snRNA and rRNA incorporation, as described in section 3.3. 5-FU, 5-fluorouracil; 5-FdUrd, 5-fluorodeoxyuridine; FdUMP, 5-fluorodeoxyuracil monophosphate; FdUDP, 5-fluorodeoxyuracil diphosphate; FdUTP, 5-fluorodeoxyuracil triphosphate; FUMP, 5-fluorouracil monophosphate; FUDP, 5-fluorouracil diphosphate; FUTP, 5-fluorouracil triphosphate; OPRT, orotate phosphoribosyltransferase; UMPK, uridine monophosphate kinase; UDPK, uridine diphosphate kinase; TK1, thymidine kinase 1; TMPK, thymidylate kinase; NDPK, Nucleoside diphosphate kinase; TYMS, thymidylate synthase.

In addition to TYMS inhibition, FdUTP can be incorporated into DNA. Thymidylate kinase (TMPK) converts FdUMP to FdUDP, and nucleoside diphosphate kinase (NDPK) phosphorylates FdUDP to FdUTP (Figure 2). There is evidence to suggest that FdUTP incorporation contributes significantly to 5-FU-mediated toxicity. In mammalian cells, 5-FU treatment leads to 5-FdU incorporation into DNA at levels higher than uracil incorporation in DNA. However in yeast exposed to 5FU, uracil levels in DNA exceed genomic 5-FdUTP levels 62. Inhibition of UNG expression sensitizes the ovarian cancer cell line OVCAR-5 to 5-FdU, but not to Raltitrexed. This suggests that 5-FdU in DNA has a significant toxic effect 63 compared to dU incorporation alone, as would be expected from Raltitrexed treatment.

Current evidence suggests that cytotoxicity of 5-FdUMP present in the genome is a consequence of DNA repair and the resulting lesions. In vitro assays with DNA polymerase δ indicate that synthetic oligonucleotides containing 5-FdU do not decrease rates of DNA synthesis or stall DNA replication itself 63. 5-FdUTP that is incorporated into DNA is recognized and excised by base excision repair machinery by the same mechanisms that remove genomic uracil 64. In the absence of UNG, 5-FdUMP in DNA is removed by another DNA glycosylase, thymine DNA glycosylase (TDG). TDG-mediated excision results in persistent strand breaks and delayed S-phase progression. Unlike other uracil DNA glycosylases, TDG is post-translationally modified by the Small Ubiquitin-like Modifier (SUMO), and SUMOylation is required for TDG dissociation from AP sites and subsequent APE1 loading 65. This requirement for SUMO modification is hypothesized to delay onset of base excision repair, leading to increased half-life of these intermediate sites which can lead to replication fork stalling and collapse. Additionally, MEFs derived from Tdg−/− mice are resistant to 5-FU 66. Hence repair of repair of 5FdU in DNA by TDG plays a meaningful role in 5-FU mediated toxicity.

5-FU incorporation into RNA may or may not make meaningful contributions to its cytotoxicity. 5-FU is converted to 5-fluorouracil triphosphate (FUTP), through the collective actions of orotate phosphoribosyltransferase (ORPT), uridine monophosphate kinase (UMPK), and uridine diphosphate kinase (UDPK) (Figure 2). Alternatively, 5-fluorouridine (5-FUrd) can be administered and converted to FUMP via uridine kinase (UK), and subsequently converted to FUTP via UMPK and UDPK. FUTP can be incorporated into newly synthesized RNA. Following exposure to 5-FU, its incorporation into RNA is reported to be 3,000 – 15,000 times higher than its incorporation into DNA; furthermore, 5-FU cytoxicity is partially rescued by ribonucleosides, but not dexoyribonucleosides in HeLa cells and SW480 colon cancer cells 67. HeLa cells exposed to 5-FUrd, but not 5-FdUrd, exhibit stress granule formation. Stress granules are cytosolic aggregations of RNA and protein formed under stress, which house stalled translation complexes 68. Stress granules are hypothesized to be a defense to prevent improper translation, and can undergo re-initiation of translation or degradation. This effect is also seen with other RNA incorporating drugs 68.

The cytotoxicity of 5-FU in RNA appears to be caused by impaired RNA processing, as opposed to direct translational inhibition as 5-FU substitution in mRNA does not affect the rate of in vitro translation 69. 5-FU incorporation into rRNA in vitro prevents its degradation by exoribonuclease Rrp6, suggesting turnover of rRNA is impaired in yeast cells 70. 5-FU incorporation into mRNA also alters splicing in vitro 71, as 5-FU treated HeLa cells exhibited impaired splicing of pre-mRNA. Injection of 5-FU synthesized mRNA did not exhibit impaired splicing 72, rather incorporation of 5-FU into snRNA impairs the function of the splicing machinery. However, investigations of 5-FU tumor resistance have not revealed a role for 5-FU incorporation into RNA as having a primary effect in 5-FU efficacy as a chemotherapeutic agent.

Cancer cells can become resistant to 5-FU by increasing rates of 5-FU catabolism or by increasing rates of de novo dTMP biosynthesis. Dihydropyrimidine dehydrogenase (DPD) in the liver catabolizes 5-FU into dihydrofluorouracil (DHFU), metabolizing 80% of the 5-FU 73 (Figure 2). Mutations in the DPYD gene, which encodes DPD, result in impaired liver DPD activity, a condition known as DPD deficiency. DPD deficiency is linked to increased cytotoxicity in 5-FU treated patients 74. Conversely, the efficient degradation of 5-FU by DPD limits the usefulness of the drug. The oral 5-FU prodrug Capecitabine circumvents this degradation.

TYMS levels can increase in response to 5-FU treatment 58, and increased TYMS gene expression is associated with poor response to 5-FU treatment 75-78. Elevated levels of DPD, TYMS, and thymidine phosphorylase (TP) are correlated with increased 5-FU chemoresistance and decreased survival Bai, et al. 79. However, Li et al found there was no correlation observed between mRNA levels of TP, DPD, or TYMS and breast carcinoma tumor size, node status, or histological grade, although elevated TYMS mRNA levels were found to be correlated with shorter survival. The mRNA expression of these genes had high variation, which may have obscured significant correlations80. Increased TYMS and TP levels were correlated with lymph node recurrence in advanced gastric cancer 81, and consistent with observations that injected TP overexpressing cells in BALB/c mice are sensitized to 5-FU 82. As extensively reviewed in Bronckaers et al, TP promotes tumor growth and metastasis, but is also necessary to activate 5-FU prodrugs 73.

In humans the addition of leucovorin to 5-FU treatment has been shown to be more effective than 5-FU alone 83. Leucovorin, which is a natural form of folate, 5-formyl-tetrahydrofolate, can be converted to 5,10-methylene-tetrahydrofolate in the cell, supporting the formation of the TYMS-5-FdUMP-5,10-methylene-tetrahydrofolate ternary complex 59,60. The addition of leucovorin has been shown to sensitize cancer cells to 5-FU 84. Leucovorin supplementation has been shown to increase ternary complex formation in MCF-7 breast cancer cells and NCI H630 colon cancer cells 85. These findings indicate that elevated cellular folate levels increase 5-FU efficacy. The FOLFOX regimen, which is formulated to include oxaliplatin, (a platinum-based agent which forms intra- and inter-strand crosslinks in DNA), leucovorin and 5-FU, has been shown to further increase 5-FU efficacy in the treatment of metastatic colorectal cancer 86.

Until recently, it was assumed that dTMP synthesis occurred in the cytoplasm, with dTMP translocating to the mitochondria or nucleus for DNA replication. The enzymes that constitute the de novo dTMP synthesis pathway were originally localized to the cytosol in both mammalian cells and yeast. However, in the 1980s, there were reports demonstrating that TYMS could also be localized to the nucleus. Thereafter, co-sedimentation and co-fractionation experiments in mammalian cells indicated that some, but not all, of the enzymes involved in deoxynucleotide biosynthesis were found to co-localize in a multiprotein complex with the DNA replication machinery, and this complex was referred to as the “replitase”87-89. Interestingly, the nuclear localization of TYMS was not observed in yeast, rather S. cerevesiae TYMS was found to localize to the nuclear periphery 90. Hence, it was not clear whether the entire de novo biosynthesis pathway was present in the nucleus, or if dTMP biosynthesis occurred in the nucleus.

Radiolabeling experiments using isolated nuclei from S-phase blocked cells demonstrated that nuclei are capable of de novo thymidylate biosynthesis using either serine (via SHMT) or formate (via MTHFD1) as one-carbon donors 7,91. This raises the possibility that dTMP synthesis occurs exclusively in the nucleus only during S phase. Formation of a multienzyme complex and the associated protein-protein interactions are essential for the functioning of the de novo dTMP synthesis pathway, as sonicated nuclei have greatly reduced de novo dTMP biosynthesis activity92. Furthermore, the de novo dTMP biosynthesis complex was found to localize to sites of DNA synthesis and associate with nuclear lamin proteins and other components of the DNA replication machinery including PCNA 92. The co-localization of the de novo dTMP synthesis pathway to cellular compartments associated with DNA synthesis is not unique to the nucleus, as the de novo dTMP synthesis pathway also has been found to function in purified mitochondria, involving the enzymes TYMS, DHFR2, and SHMT2 5.

The mammalian enzymes that constitute the de novo dTMP biosynthesis pathway including DHFR, SHMT1, TYMS, and MTHFD1 undergo SUMOylation at the G1/S boundary and in response to UV-induced DNA damage leading to SUMO-dependent nuclear import 46,93,94. The multienzyme complex responsible for nuclear de novo dTMP biosynthesis is essential to prevent high levels of uracil incorporation into DNA in mice 40. Previous studies have demonstrated that SHMT1 expression can determine rates of de novo thymidylate biosynthesis in cultured cells 95 and expression of catalytically-inactive SHMT1 mutants also stimulate de novo dTMP biosynthesis 92. Yet, formate, through the activity of MTHFD1, is the primary source folate-activated one-carbons for de novo dTMP synthesis 6. These experimental findings were explained by the observation that SHMT1 acts as a scaffold protein that is essential for the co-localization of MTHFD1, TYMS and DHFR, and that SHMT1 nuclear localization is necessary for de novo dTMP complex formation 44,50,92. SHMT1 binds tightly to lamin A and lamin B proteins, and serves to tether TYMS, DHFR and MTHFD1 to the nuclear lamina 50,92. Interestingly, overexpression of the SHMT1 cDNA in mice results in high levels of SHMT1 protein in liver, but impaired nuclear localization of SHMT1, leading to depressed rates of de novo dTMP biosynthesis in isolated nuclei, and up to 10-fold increased levels of uracil in mouse liver nuclear DNA40. These studies demonstrate the importance of nuclear localization of the de novo dTMP synthesis complex in preventing uracil accumulation in DNA and maintaining genome stability.

Formation of a nuclear de novo dTMP synthesis complex appears to be unique to mammalian cells, and the evolution of nuclear dTMP synthesis may provide several advantages in limiting genome instability. In response to DNA damage, yeast increase dNTP concentrations 6-8 fold 96 and E. coli increase dNTP concentration 1.8-3.7 fold 97, whereas mammalian cells do not show increased dNTP concentrations following DNA damage 11,98. In both S. cerevesiae and E. coli, the increase in dNTP levels results in higher rates of mutagenesis 96,97. Thus, dTMP synthesis localized to sites of DNA synthesis may lower mutation rates. Nuclear dTMP synthesis at sites of DNA synthesis allows for selective concentration of this deoxyribonucleotide at its only site of utilization, thereby limiting unnecessary synthesis and its incorporation into RNA in the cytosol. Provision of dTTP directly at sites of DNA synthesis may also permit the regulated and selective incorporation of dTTP or dUTP into DNA, although there is no evidence currently that uracil incorporation into the genome is regulated or selectively incorporated.

SUMOylation of proteins is known to affect protein-protein interactions, cellular localization, and substrate stability 99. SUMOylation is necessary for proper functioning of many proteins involved in critical cellular processes such as DNA replication and repair 100-102, transcriptional regulation 103-105, maintenance of chromosomal structure 106, nuclear morphology, and nuclear import. For example, the E3 SUMO ligase PIASy stimulates SUMOylation of both p53 and Rb, inducing premature senescence 107, indicating SUMOylation is used as a regulator of gene transcription by the cell. Disruption of SUMOylation machinery results in DNA replication and repair defects, suggesting SUMOylation of these components is critical. Loss of the SUMO-conjugating protein ligase Ubc9 leads to chromosome condensation and segregation defects in mouse embryos, and Ubc9-deficient blastocysts have defects in nuclear and subnuclear architecture. Loss of Ubc9 leads to a collapse in the RanGDP-GTP gradient, which is critical for nuclear import and export 108. Cancer models show upregulation of the various SUMO components. Ubc9 is upregulated in HPV-positive cervical lesions 109, and in malignant and peritumoral squamous cell carcinoma tissues 110. As reviewed in Mattoscio et. al., upregulation of several SUMO proteases, which remove SUMO modifications (SENP1, SENP3, SENP5), has been shown to be critical for some cancer cell proliferation 111-114, whereas SENP2 is downregulated in hepatocellular carcinoma and bladder cancer 115,116.

The SUMO machinery has been suggested to be an attractive target for the development of anticancer therapies, as disruption of SUMO components has been shown to affect cancer cell proliferation. Reduced Ubc9 expression suppresses growth of KRAS mutant colorectal cancer cells 117, and Spectinomycin B1, an inhibitor of Ubc9, inhibits proliferation of MCF-7 breast cancer cells 118. Reduced expression of the SUMO-activating enzyme subunit 2 (SAE2) and Ubc9 impair cell proliferation in cancer cells as does reduced expression of SAE2 alone in U2OS osteosarcoma cells. HCT116 colorectal carcinoma cells with reduced SAE2 expression exhibit elevated rates of apoptosis and senescence, with xenografts exhibiting lower rates of tumor growth in mice 119. Disruption of deSUMOylating enzymes also impairs cancer cell proliferation. Reduced expression of SENP1 leads to growth arrest and apoptosis in multiple myeloma cells 120 and inhibits cell proliferation in prostate cancer cells 121. Reduced SENP5 expression inhibits cell proliferation and causes defects in nuclear morphology in HeLa cells 122.

Advances in the fundamental understanding of the nuclear compartmentation of de novo dTMP biosynthesis prompts a reevaluation of the mechanisms of 5-FU cytotoxicity. There is evidence that nuclear compartmentation of the de novo dTMP synthesis pathway modifies 5-FU efficacy. A higher ratio of nuclear:cytoplasmic TYMS was shown to be significantly correlated with decreased survival in colorectal cancer123. Additionally, increased nuclear TYMS expression is associated with poor response to 5-FU in colorectal carcinoma 124. Other antifolates may function to impair nuclear dTMP synthesis, as MTX has been shown to almost completely block nuclear translocation of DHFR in acute myeloid leukemia cells 125. Thus, nuclear localization of the de novo dTMP biosynthesis pathway may be a critical predictor of 5-FU mediated toxicity. Interestingly, a common SHMT1 variant, SHMT1 C1420T (rs 1979277), results in an amino acid substitution (Leu474Phe) located at the interface where Ubc9 and SHMT1 interact when SHMT1 undergoes SUMOylation 93. This SHMT1 polymorphism impairs the post-translational SUMO modification of SHMT1 and its nuclear localization 94, and hence would be expected to decrease total dTMP synthesis capacity. SHMT1 C1420T was associated with better response and longer progression-free survival in patients with metastatic colorectal cancer treated with bevacizumab+ FOLFIRI (5-FU, leucovorin, irinotecan) 126. Barcelos et. al. also found that SHMT1 C1420T was associated longer progression-free survival with FOLFIRI treatment in metastatic colorectal cancer 127. The discovery of SUMO-dependent nuclear de novo dTMP synthesis in mammalian cells raises the possibility of novel strategies to target de novo dTMP biosynthesis, with a focus on SHMT1 SUMOylation as it is essential for formation of the metabolic complex responsible for nuclear dTMP synthesis.

To improve efficacy of antifolate based therapies, subcellular localization and the factors which modify localization of target proteins must be taken into account. Considering that nuclear localization of the de novo dTMP biosynthesis complex is critical for proper dTMP synthesis, further studies are needed to determine if pharmaceuticals that target enzymes in dTMP synthesis affect nuclear localization of TYMS (as shown with DHFR and MTX), and if nuclear localization of the pathway enzymes influences drug efficacy and cancer survival. At a more fundamental level, further studies are needed to determine if antifolate binding to their respective enzymes induce conformational changes that could affect SUMOylation and nuclear import. Because of SHMT1's critical role in nuclear localization of the de novo dTMP biosynthesis complex, impairing SHMT1 nuclear import or lamin-binding capacity may improve antifolate based therapies. Considering that SUMOylation plays a critical role in nuclear localization of the de novo dTMP biosynthesis complex, and in DNA replication and repair in general, targeted SUMO-based inhibitors could be used to augment antifolate treatments.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.