Human UMP-CMP Kinase 2, a Novel Nucleoside Monophosphate Kinase Localized in Mitochondria*

Enzyme deficiency in the salvage pathway of deoxyribonucleotide synthesis in mitochondria can cause mtDNA depletion syndromes. We have identified a human mitochondrial UMP-CMP kinase (UMP-CMPK, cytidylate kinase; EC 2.7.4.14), designated as UMP-CMP kinase 2 (UMP-CMPK2). The C-terminal domain of this 449-amino acid protein contains all consensus motifs of a nucleoside monophosphate kinase. Phylogenetic analysis showed that UMP-CMPK2 belonged to a novel nucleoside monophosphate kinase family, which was closer to thymidylate kinase than to cytosolic UMP-CMP kinase. Subcellular localization with green fluorescent protein fusion proteins illustrated that UMP-CMPK2 was localized in the mitochondria of HeLa cells and that the mitochondrial targeting signal was included in the N-terminal 22 amino acids. The enzyme was able to phosphorylate dUMP, dCMP, CMP, and UMP with ATP as phosphate donor, but the kinetic properties were different compared with the cytosolic UMP-CMPK. Its efficacy to convert dUMP was highest, followed by dCMP, whereas CMP and UMP were the poorest substrates. It also phosphorylated the monophosphate forms of the nucleoside analogs ddC, dFdC, araC, BVDU, and FdUrd, which suggests that UMP-CMPK2 may be involved in mtDNA depletion caused by long term treatment with ddC or other pyrimidine analogs. UMP-CMPK2 mRNA expression was exclusively detected in chronic myelogenous leukemia K-562 and lymphoblastic leukemia MOLT-4 among eight studied cancer cell lines. Particular high expression in leukemia cells, dominant expression in bone marrow, and tight correlation with macrophage activation and inflammatory response suggest that UMP-CMPK2 may have other functions in addition to the supply of substrates for mtDNA synthesis.

Nucleotide synthesis is a basic biological process for cell proliferation and almost all other physiological activities in the cell. Two pathways have been reported for nucleotide synthesis: the de novo pathway and the salvage pathway (1). In the de novo pathway, the synthesis of nucleotides starts from small molecules whereas in the salvage pathway, free nucleosides are directly used to synthesize ribonucleotides and deoxyribonucleotides.
To figure out the complete enzymatic steps of the salvage pathway for deoxyribonucleotide synthesis in mitochondria, we cloned and characterized the human homolog (hypothetical protein LOC129607, NCBI accession NP_997198) of a mouse gene, which was designated as thymidylate kinase family LPSinducible member (TYKi) because of a putative thymidylate kinase domain (29). Previous studies showed that the expression of murine TYKi was induced or up-regulated by LPS and several cytokines (TNF␣, IFN␥, IL-1␤, IFN␣) and was downregulated by TGF␤ (29 -35), whereas there are no reports about its protein properties so far. Here, we cloned the full-length coding sequence of the human cDNA and found that the gene product was localized in the mitochondria of HeLa cells. We expressed this protein in insect cells and purified it to homogeneity. The recombinant protein phosphorylated CMP, UMP, * This work was supported in part by grants from the Swedish Cancer Society and the Swedish Research Council. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains a supplemental dCMP, dUMP, and monophosphates of the pyrimidine nucleoside analogs ddC, dFdC, araC, BVDU, and FdUrd. Based on its substrate recognition and intracellular location we designated this novel enzyme as mitochondrial UMP-CMP kinase and named it UMP-CMPK2. The relative phosphorylation efficacy of the natural substrates were dUMP Ͼ dCMP Ͼ CMP Ͼ UMP, which is different from the properties of cytosolic UMP-CMPK.
The characteristics and phylogenetic analysis demonstrated that this protein comes from a novel family of UMP-CMP kinases.
Phylogenetic Analysis of UMP-CMPK2-Multiple sequence alignments were accomplished with Kalign program on the EBI server and edited with the GeneDoc v2.6 program. The rooted tree was constructed with PhyML and was plotted with iTOL v1.01.
Subcellular Localization of UMP-CMPK2-Two plasmids were constructed for protein expression in mammalian cells. The first plasmid was based on pEGFP-N1 vector (Clontech). Two oligos 5Ј-GAA TTC ATG GCC TTC GCC CGC CGG CTC CTG CGC GGG CCA CTG TCG GGG CCG CTG CTC GGG CGG CGC GGG GAT CCA-3Ј and 5Ј-GGA TCC CCG CGC CGC CCG AGC AGC GGC CCC GAC AGT GGC CCG CGC AGG AGC CGG CGG GCG AAG GCC ATG AAT TCA-3Ј were used to form a dsDNA fragment encoding the first 22 amino acids of UMP-CMPK2. This dsDNA was ligated into the EcoRI-BamHI site of pEGFP-N1 to create an expression plasmid pN22-GFP.
The second plasmid was based on pcDNA-DEST47 destination vector (Invitrogen). Using pENTR-UCMPK2-E448G as template, with forward primer 5Ј-CAC CAT GGC CTT CGC CCG CCG GCT C-3Ј and reverse primer 5Ј-GTG AGG ATC CGG TCC ACT AAA ACT ATT CTG GAT TAG G-3Ј in which the stop codon was deleted for C-terminal fusion purpose, full-length coding region of UMP-CMPK2 cDNA was amplified by PCR under the same condition of the first PCR except the following parameters were used: 1ϫ (94°C, 3 min), 30ϫ (94°C, 15 s; 58°C, 30 s; 68°C, 2 min) and 1ϫ (68°C, 10 min). PCR products were cloned into pENTR TM /S.D./D-TOPO to create an entry clone. The mammalian expression plasmid pUCMPK2-E448G-GFP was constituted by recombination of this entry clone and pcDNA-DEST47 (Invitrogen) according to the manufacturer's instructions.
The HeLa cell line (American Type Culture Collection) was transfected with pN22-GFP and pUCMPK2-E448G-GFP according to the method described (19). The cells were stained with MitoTracker Red (Molecular Probes) 48 h after transfection. Cell fluorescence was imaged with a Nikon Eclipse E600 microscope equipped with a SPORT RT digital camera.
Expression and Purification of Recombinant UMP-CMPK2-The cDNA sequence without putative mitochondrial targeting signal was amplified by PCR with forward primer 5Ј-CAC CGG ATC CAT GGT CTG CGC TGG GGC CAT GG-3Ј (with BamH I site), reverse primer 5Ј-GAA TTC CTA GTG ATG GTG ATG GTG ATG CGG TTC ACT AAA ACT ATT CTG-3Ј (with EcoRI site and six histidine codons), Platinum Pfx DNA polymerase (Invitrogen) and 1ϫ PCRx Enhancer solution. The cycling parameters are 1ϫ (94°C, 2 min), 30ϫ (94°C, 30 s; 58°C, 30 s; 68°C, 2 min) and 1ϫ (68°C, 10 min). PCR products were cloned into pENTR TM /SD/D-TOPO to create an entry clone. The UMP-CMPK2⌬21-6His coding sequence from this entry clone was inserted into the BamHI-EcoRI site of the pBacPAK8 transfer vector (Clontech) to get a transfer construct. Recombinant virus was constructed by cotransfecting Spodoptera frugiperda (Sf9) cells with the transfer construct and BacPAK6 viral DNA according to the manufacturer's manual.
Protein expression in Sf9 cells and the affinity chromatography were modified from reported protocol (36). Sf9 cells were harvested 72 h after infection with 10 p.f.u. of recombinant virus. The cells were lysed, and the extract was cleared by centrifugation at 40,000 rpm ϫ 30 min, 4°C, using a Beckman 45TI rotor. The recombinant proteins were purified with TALON Metal Affinity Resin (Clontech), desalted on PD-10 column (Amersham Biosciences) and followed by anion-exchange chromatography with a MonoQ HR5/5 (Amersham Biosciences) column pre-equilibrated in low salt Buffer IEX (20 mM Tris-HCl, pH 8.5, 50 mM NaCl, 10% glycerol, 1 mM DTT, and 1ϫ protease inhibitors). After elution in a linear gradient (10 ml) of Buffer IEX (0.05-1 M NaCl), the recombinant proteins were concentrated with Centricon 10 devices (Millipore) and applied to a Superose 6 10/300 GL column (Amersham Biosciences) pre-equilibrated in Buffer GF (20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 10% glycerol, and 1 mM DTT). The proteins were eluted at a flow rate of 0.30 ml/min with 1.5 bed volumes of Buffer GF. The calibration of column was carried out with Bio-Rad gel filtration standard.
Isoelectric Point Determination-The pI of recombinant protein was determined on a PhastSystem with IEF PhastGel IEF 3-9 media (Amersham Biosciences) according to the manufacturer's instruction.
Northern Blotting-The human cancer cell line Northern blot including the mRNA from 8 different cancer cell lines was purchased from Clontech Inc. A 644-bp cDNA encoding the C-terminal domain was used as probe. Hybridization was carried out according to the manufacturer's instructions.

RESULTS
Cloning of the Full-length cDNA and Primary Structure Analysis of the UMP-CMPK2 Protein-The human UMP-CMP kinase 2 is currently recorded as hypothetical protein LOC129607 in Entrez Gene data base of NCBI. The genomic sequence is localized at 2p25.2. It is regarded as a thymidylate kinase family LPS-inducible member because its homology with mouse TYKi protein, which was LPS-inducible in macrophages and was supposed to have thymidylate kinase activity based on sequence similarity with thymidylate kinase (29). Although many ESTs already have been cloned, no mRNA or cDNA that covers the full-length coding region has been characterized.
To clone the full-length cDNA, we selected the ATCC cDNA clone IMAGE3063188 (NCBI accession BC089425) as our PCR template. By adding the coding sequence of 8 missing amino acids in the forward primer, we PCR amplified the full-length cDNA of UMP-CMPK2. Because this cDNA was derived from BC089425, which contains a point mutation at the codon of the 448th amino acid (GGA (codon of glycine) in BC089425 and GAA (codon of glutamic acid) in genomic DNA and all other ESTs), we designated the protein coded by this cDNA sequence as UMP-CMPK2-E448G. For protein expression, this point mutation was corrected by using the genomic sequence in reverse primers for PCR amplification. An interesting feature of the cDNA sequence is that the first half of the coding sequence has high GC content. From the translation start site, the 650-bp cDNA at the 5Ј-end has 76.3% GC content, whereas the 3Ј-end cDNA sequence only has about 52.6% GC.
UMP-CMPK2 has 449 amino acids, which is longer than UMP-CMPK (228 amino acids). Primary structure analysis with Motif Scan and multiple sequence alignment with other thymidylate kinases showed that the protein sequence had a thymidylate kinase domain (Fig. 1B). The C-terminal domain contains all the consensus motifs: P-loop (ATP/GTP binding motif A), LID domain, adenine-base binding loop, catalytic site, and potential substrate binding site (Fig. 1B).
Phylogenetic analysis was accomplished among selected species, which evolutionarily cover a long distance range from prokaryotes to primates (Fig. 1A). UMP-CMPK2, UMP-CMPK, thymidylate kinase (TMPK), cytidylate kinase (CMPK), and uridylate kinase (UMPK) from some prokaryotes were selected for comparison. On the phylogenetic tree, it is clear that the distance between the UMP-CMPK2 family and the TMPK fam-ily was much shorter than the distance between UMP-CMPK2 and UMP-CMPK. The predicted UMP-CMPK2 of dogs has quite low homology with other UMP-CMPK2 members (Fig. 1,  A and B). Subcellular Localization of UMP-CMPK2 in HeLa Cells-Original analysis with PSORT program showed that UMP-CMPK2 had high probability of being a mitochondrial protein with a putative cleavage site of the mitochondrial targeting signal after the 20th or the 32nd amino acid. To confirm this prediction, a dsDNA fragment encoding the first 22 amino acids of UMP-CMPK2 was synthesized and cloned into the pEGFP-N1 vector to construct N22-GFP fusion gene. After transfection with the pN22-GFP plasmid, HeLa cells were cultured for ϳ48 h and then were stained with MitoTracker Red. Results showed that most of the green fluorescence of GFP distributed in a typical fibrillar pattern as mitochondria usually do. Overlay of the fluorescence signal of GFP and MitoTracker Red demonstrated the fibrillar mitochondrial localization in yellow color (Fig. 2, top panel).
The full-length UMP-CMPK2-E448G cDNA was fused to the GFP gene and was transiently expressed in HeLa cells as a UMP-CMPK2-E448G-GFP fusion protein. As shown in Fig. 2, bottom panel, the green signal of UMP-CMPK2-E448G-GFP was also overlaid with the red signal of MitoTracker Red, which confirmed the mitochondrial localization of UMP-CMPK2 in HeLa cells.
Protein Expression and Purification of Recombinant UMP-CMPK2-At first, we thought that this protein would have TMPK activity based on BLAST results and interpretation of the mouse homolog (29). Initially, UMP-CMPK2 was expressed as His-tagged and GST-tagged recombinant proteins in Escherichia coli, respectively. We cleaved off the tags with protease and purified the proteins to high homogeneity (Ͼ95%). All enzymatic activity assays performed under conditions for TMPK activity only showed UMP-CMP kinase activity. No phosphorylation of dTMP was detected (data not shown). Considering a report that human TMPK activity was undetectable in transformed yeast or from proteins of several E. coli expression systems (38), we overexpressed His-tagged recombinant UMP-CMPK2 in insect cells and obtained pure proteins for enzymatic assay.
First we designed a new set of primers for PCR amplification. The forward primer started at the 22nd amino acid to remove the cleavable putative mitochondrial targeting signal. One ATG codon was added before the 22nd amino acid to start translation. The reverse primer was located around the position of stop codon, but the stop codon was replaced by a histidine codon followed by 5 more to encode the His 6 tag. The E448G point mutation was also corrected to glutamic acid (Glu, E) by using the genomic sequence. The PCR products were used to construct recombinant Baculovirus carrying UMP-CMPK2⌬21-6His fusion gene.
The Sf9 cells were infected with recombinant virus and harvested 3 days later. To characterize the properties of the enzyme, UMP-CMPK2⌬21-6His protein was first purified with affinity chromatography. The protein was quite pure according to the SDS-PAGE data (Fig. 3B, lane 2). To reach a higher purity we performed two more purification steps: anion-exchange and gel filtration chromatography. The final products had very high purity (Ͼ95%) as shown by SDS-PAGE (Fig. 3B) and gel filtration diagram (data not shown).
The pI value of the recombinant protein was 6.05 as determined by isoelectric focusing method along with pI markers of pH 4.55-8.65 (Fig. 3A). The protein did not bind the MonoQ matrix stably at pH 8.0, especially when large amounts of pro-teins were loaded. Instead we used pH 8.5 for anion-exchange chromatography. Because the pH value in mitochondria is about 8.0 (39, 40), we kept pH 8.0 for gel filtration chromatography and the following activity assays. Gel filtration diagram showed two peaks of UMP-CMPK2. Referring to the molecular size calibration curve from gel filtration molecular size standards, the major peak was monomer, whereas the minor peak had molecular weight of ϳ89 kDa indicating dimer form of the recombinant protein.
Substrate Specificity of UMP-CMPK2-Substrate specificity of recombinant protein was studied with thin layer chromatography (TLC) assay by using the [␥-32 P]ATP as phosphate donor. As shown in Fig. 4A, the recombinant UMP-CMPK2 could phosphorylate CMP, UMP, dCMP, and dUMP.
We also investigated the phosphorylation of dCMP and dUMP analogs using a two-step enzymatic method. For dCyd analogs, the human deoxycytidine kinase (dCK) was added into each reaction to catalyze the first phosphorylation. The results showed that ddC-MP, dFdC-MP, and araC-MP can be phosphorylated by UMP-CMPK2 (Fig. 4B). For dUrd analogs, we used human TK2 for the coupled phosphorylation. BVDU-MP and FdU-MP were proved to be substrates of UMP-CMPK2 in this assay (Fig. 4C).
Kinetic Properties of UMP-CMPK2-The Michaelis-Menten kinetic properties of the enzyme were determined with four natural substrates using reversed-phase HPLC ( Table 1). The production of dUDP was calculated referring to the chromatography data of UDP because there was no pure dUDP standard available. UMP-CMPK2 showed preference for dUMP and dCMP compared with CMP and UMP. The dUMP was the preferred substrate and the V max /K m value was about 35-fold higher than that for dCMP, and about 1600-fold higher than that for the poorest substrate UMP. The differences in V max were about 11-fold ranging from 0.19 mol/mg/min (UMP) to 1.77 mol/mg/min (dCMP). The V max of CMP and dCMP was similar, whereas V max of dUMP was 2.5-fold of the maximum rate for UMP. The differences in K m were 63-fold ranging from 0.10 mM (dUMP) to 6.30 mM (UMP). There was less than 3-fold difference between the K m of CMP and dCMP ( Table 1).
Expression of UMP-CMPK2 in Cancer Cells-A human Northern blot with poly(A) mRNA of eight different cancer cell lines was used. The mRNA expression was exclusively detected in chronic myelogenous leukemia K-562 and lymphoblastic leukemia MOLT-4 (Fig. 5). The former cell line had higher expression level than the latter one correlated to the actin mRNA abundance. Based on the RNA molecular standards, the estimated size of UMP-CMPK2 mRNA is about 3.3 kb, which is close to the size of NM_207315.2 (3009 bp) reported in NCBI.
No expression was detected in other six cell lines.   A, determination of the pI value of recombinant protein by isoelectric focusing. The gel was Coomassie-stained, and the distances from the cathode to each protein pI marker were used to draw the pH gradient profile calibration curve. B, SDS-PAGE of proteins from different purification steps. Lane 1, total soluble proteins of Sf9 cells expressing UMP-CMPK2⌬21-6His, 10 g. Lane 2, proteins after affinity chromatography, 3 g. Lane 3, proteins after anion-exchange chromatography, 3 g. Lane 4, proteins after gel filtration chromatography, 3 g. No heat denaturation before loading.

DISCUSSION
In mammalian cells, UMP-CMP kinase and thymidylate kinase are the only known pyrimidine nucleoside monophosphate kinases that have been identified. UMP-CMPK phosphorylates CMP, UMP, dCMP, and dUMP, whereas TMPK phosphorylates dTMP and dUMP (41). The majority of studies on mammalian nucleoside monophosphate kinases have been focused on cytosolic enzymes. Here, for the first time, we report a human mitochondrial UMP-CMP kinase, which was designated as UMP-CMP kinase 2.
The overlap of GFP signal and MitoTracker Red fluorescence indicated that UMP-CMPK2 was localized in the mitochondria of HeLa cells. The mitochondrial targeting signal was included in the first 22 amino acids, and the sequence around serine forms a sequence pattern similar to reported R-10 conserved sequence motif of mitochondrial targeting signal RX(F/I/L)SX 6 (42). The exact cleavage site needs to be determined by experimental methods.
The amino acid sequence of UMP-CMPK2 has high homology with TMPK. It contains putative motifs that are conserved FIGURE 4. Screening of nucleoside monophosphate and nucleoside analog specificity of mitochondrial UMP-CMP kinase 2. Coupled assays were used to phosphorylate the nucleoside analogs. A, 1 mM natural nucleoside monophosphates as substrates. 0.1 g UMP-CMP kinase 2 was used in each reaction; 1 h of incubation at 37°C. B, human deoxycytidine kinase (dCK) was used to phosphorylate cytidine analogs to get monophosphate substrates for UMP-CMPK2⌬21-6His. 1.0 g of each enzyme per reaction was used; 2 h of incubation at 37°C. C, human mitochondrial TK2 was used to phosphorylate uridine analogs to get monophosphate substrates for UMP-CMPK2⌬21-6His. 0.5 g of each enzyme per reaction was used; 2 h of incubation at 37°C.  in TMPK, such as P-loop, LID domain, and catalytic site, as reported in mouse TYKi by Lee et al. (29). However, we were not able to detect any TMPK activity from highly purified recombinant UMP-CMPK2 expressed in insect cells or in E. coli. Because the mitochondrial targeting signal was removed in all expression systems used and all the protein preparations had similar UMP-CMP kinase activities, the N-terminal sequence is not likely to alter the enzyme activity. We conclude that the expressed recombinant protein is not a thymidylate kinase but instead a mitochondrial UMP-CMP kinase belonging to a novel protein family. This family is close to TMPK, but the distance to cytosolic UMP-CMPK is relatively far away, which supports the high phosphorylation efficacy of UMP-CMPK2 on dUMP. Kinetic properties of UMP-CMPK2 were significantly different from those of cytosolic UMP-CMPK. Human cytosolic UMP-CMPK was reported to preferentially phosphorylate ribonucleotides (43,44). In contrast, mitochondrial UMP-CMPK2 had higher efficacy for deoxyribonucleotides than for ribonucleotides. The best substrate was dUMP followed by dCMP, whereas the relative efficacy with CMP was lower than for dCMP. As the poorest substrate, UMP showed very low efficacy with the lowest V max and the highest K m . The K m values for CMP and UMP of the recombinant UMP-CMPK2 were higher than those of UMP-CMPK. The K m of dCMP is between the two reported values of UMP-CMPK. However, the K m for dUMP was lower than that of UMP-CMPK (43). All the specific activities for CMP and UMP were lower than those data of UMP-CMPK, but the V max /K m of dUMP from UMP-CMPK2 is higher (ϳ1200-to 2400-fold) than the value of UMP-CMPK reported by Van Rompay et al. (43,44). Referring to the regulation of UMP-CMPK by ATP and magnesium reported by Hsu et al. (45), we used 2 mM ATP and 6 mM Mg 2ϩ in the kinetic assays for dCMP. Our observations showed that the kinase activity for dCMP of recombinant UMP-CMPK2 was significantly inhibited at high concentration of ATP (data not shown), which indicated that the activity of UMP-CMPK2 might be regulated by ATP concentration. In summary, current data indicate that mitochondrial UMP-CMP kinase has low affinity to pyrimidine ribonucleotides and it may mainly participate in dUTP (or just dUDP) and dCTP synthesis in mitochondria.
Deoxycytidine analogs and deoxyuridine analogs are important drugs widely used as antiviral and anticancer agents. For instance, ddC, BVDU, and BVaraU are used as antiviral agents; dFdC, araC, and FdUrd are used in anticancer therapy (46 -51). Using coupled enzymatic assays, recombinant UMP-CMPK2 could phosphorylate the monophosphate forms of all above analogs except BVaraU. Under long term treatment, several deoxynucleoside analogs, such as ddC, can cause mtDNA depletion after incorporation into mtDNA by DNA polymerase ␥ (52). Our data propose the possibility that UMP-CMPK2 may be involved in the activation of ddC and other pyrimidine analogs in mitochondria. Further studies on UMP-CMPK2 and other enzymes of the same pathway, will contribute to increased knowledge that may reduce mitochondrial toxicity caused by these compounds.
The expression of UMP-CMPK2 was detected in two leukemia cell lines, and several sources have demonstrated a special expression pattern of this gene. The expression profile from UniGene data base (ID: Hs.7155) shows the highest expression level in bone marrow, whereas at all other tissues have relative rare transcripts. Many tissues, including tissues rich in mitochondria (such as muscle and heart), show no detectable transcripts of UMP-CMPK2. Among all investigated tumors, leukemia cells show the most abundant expression whereas other cells have very low levels or undetectable transcripts of this gene. Another expression profile from Genomics Institute of the Novartis Research Foundation also shows similar pattern. All these data suggest that UMP-CMPK2 may not be crucial for general cell activities, such as mtDNA synthesis, but play a role during hematopoiesis or lymphocytopoiesis.
Previous studies reported that only one UMP-CMP kinase was present in calf thymus, rat liver, or Yoshida sarcoma cells (41,53), and there was no UMP-CMPK activity detectable in purified mitochondrial extracts from HeLa S3 cells (44) probably because of a very low expression level of UMP-CMPK2 in these tissues as shown by the expression profiles mentioned above. In addition, the antibodies developed for cytosolic UMP-CMPK may not recognize UMP-CMPK2 because of the relative long distance between them in evolution.
The studies of the mouse TYKi gene suggested a correlation between this protein and the immune response. LPS is a bacterial endotoxin which is a powerful activator of macrophages that in turn can recognize and destroy invading microorganisms and tumor cells and orchestrate the process of inflammation. Murine TYKi was reported to be significantly induced by LPS in macrophages, and this induction could be modulated by myeloid differentiation protein-88 (29,30). The number and size of mitochondria in macrophages were greatly increased after LPS induction (54). In other tissues, TYKi was also up-regulated by several inflammatory response-related cytokines (such as TNF␣, IFN␥, IL-1␤, and IFN␣) and infections of parasites (31,33,55). All these expression profiles suggest that UMP-CMPK2 is actively involved in macrophage activation and the inflammatory response. The human UMP-CMPK2 consists of 449 amino acids, which is about twice the size of UMP-CMPK. It can be separated into two domains according to sequence properties: the N-terminal domain and the C-terminal domain. All motifs participating in phosphoryl transfer are predicted to be located at the C-terminal domain, whereas there are no data available suggesting a function of the N-terminal domain. An interesting composition is that 9 of 13 cysteines crowd in the first 189 amino acid residues, which suggests that this region may have a tight or complex tertiary structure, oxidation/reduction activity or other special features. More than 60% of leucines, alanines, prolines, and glycines are located in the N-terminal domain. No leucine zipper pattern has been found and in contrast to the leucine-rich property, there are no isoleucines in this domain. Thus, the leucine rich domain may contribute with special properties, like protein-protein interactions, a possibility that will be further investigated. The two distinct domains of UMP-CMPK2: the N-terminal domain with unknown function and the C-terminal domain with the nucleoside monophosphate kinase function, suggest that UMP-CMPK2 may be a bifunctional protein with other biological functions in addition to its UMP-CMP kinase activity.
UMP-CMP kinase 2 should be responsible for phosphorylation of dCMP and dUMP in mitochondria (Fig. 6). Considering the pyrimidine metabolism pathway in cytosol, mitochondria may have their own thymidylate synthase. It is also possible that dCMP deaminase and dUTP diphosphatase activities exist in mitochondria. Although the newly discovered mitochondrial pyrimidine nucleotide carrier (PCN1) has high specificity for dTTP (56), it is not likely to provide sufficient dTTP for mtDNA synthesis in resting cells. The accumulated dTMP from phosphorylation of thymidine by TK2 (2)(3)(4)(5), and from the import of cytosolic dTMP by a highly selective and active dTMP transport system (57), must be phosphorylated to dTDP and then dTTP through the salvage pathway. A thymidylate kinase should exist to catalyze this phosphorylation in mitochondria.
In summary, UMP-CMP kinase 2 is the first pyrimidine nucleoside monophosphate kinase that has been identified in human mitochondria. The expression profiles of UMP-CMP kinase 2 suggest that the dUDP metabolism pathway in mitochondria may play a unique role. Our current studies revealed its different characteristics compared with cytosolic UMP-CMP kinase. Further studies of this enzyme will contribute to further understanding the salvage pathway for nucleotide synthesis in mitochondria, to elucidate the actual role of the dUDP metabolism pathway and to reduce the mtDNA damage caused by nucleoside analogs.