Mammalian Pyrimidine Biosynthesis: Fresh Insights into an Ancient Pathway*

Pyrimidine nucleotides play a critical role in cellular metabolism serving as activated precursors of RNA and DNA, CDP-diacylglycerol phosphoglyceride for the assembly of cell mem-branes and UDP-sugars for protein glycosylation and glycogen synthesis

Pyrimidine nucleotides play a critical role in cellular metabolism serving as activated precursors of RNA and DNA, CDPdiacylglycerol phosphoglyceride for the assembly of cell membranes and UDP-sugars for protein glycosylation and glycogen synthesis (1)(2)(3). In addition, uridine nucleotides act via extracellular receptors to regulate a variety of physiological processes (4). There are two routes to the synthesis of pyrimidines; nucleotides can be recycled by the salvage pathways or synthesized de novo from small metabolites (Fig. 1). Most cells have several specialized passive and active transporters (5,6) that allow the reutilization of preformed pyrimidine nucleosides and bases.
The relative contribution of the de novo and salvage pathways depends on cell type and developmental stage. In general, the activity of the de novo pathway is low in resting or fully differentiated cells where the need for pyrimidines is largely satisfied by the salvage pathways (7). In contrast, de novo pyrimidine biosynthesis (Fig. 1) is indispensable in proliferating cells in order to meet the increased demand for nucleic acid precursors and other cellular components. Consequently, the activity of the de novo pathway is subject to elaborate growth state-dependent control mechanisms. Pyrimidine biosynthesis is invariably up-regulated in tumors and neoplastic cells (8), and the pathway has been linked to the etiology or treatment of several other disorders including AIDS (9), diabetes (10), and various autoimmune diseases (11) such as rheumatoid arthritis. This review focuses on the structure and regulation of the pyrimidine biosynthetic complexes and the interplay of the diverse control mechanisms operative in mammalian cells.

The Multifunctional Protein CAD 1
Since the pioneering discoveries of Jones, Hoogenraad and others in 1971 (1), the physical association of the first three enzymes of the de novo pyrimidine pathway, carbamoyl-phosphate synthetase (CPSase), aspartate transcarbamylase (ATCase), and dihydroorotase (DHOase) has been documented in many animal cells. In 1977, Stark and his associates (12) made the remarkable discovery that all three activities are carried on a single polypeptide. The 243-kDa CAD polypeptide associates to form hexamers and higher oligomers (13) so that the mass of the complex exceeds 1.4 MDa or about one-half the size of the ribosome. The domain structure (Fig. 2) has been mapped, and the function of each domain has been assigned (14).
Carbamoyl-phosphate Synthetase-CPSase catalyzes the synthesis of carbamoyl phosphate from glutamine, bicarbonate, and two ATP molecules. Carbamoyl phosphate biosynthesis ( Fig. 1, circled 1) is a complex process involving four partial reactions (15,16) catalyzed by the GLN, CPS.A, and CPS.B domains of the molecule (Fig. 2). The 40-kDa GLN domain is an amidotransferase that generates ammonia by glutamine hydrolysis. The synthetase domain consists of two homologous 60-kDa halves, CPS.A and CPS.B, thought to have arisen by an ancestral gene duplication and fusion (17). These subdomains are functionally equivalent (18) but assume specialized functions when fused together in the intact molecule (19,20). CPS.A catalyzes the ATP-dependent activation of bicarbonate forming carboxyphosphate, which then reacts with ammonia to form carbamate. Carbamoyl phosphate is formed in a second ATP-dependent phosphorylation on CPS.B. All of these intermediates are labile but are effectively sequestered within the complex, passing between the active sites via narrow tunnels that snake through the interior of the molecule (21). CPSase catalyzes the rate-limiting step in de novo pyrimidine biosynthesis and, as discussed below, controls the flux through the pathway (1).
Aspartate Transcarbamylase-The reaction of carbamoyl phosphate and aspartate to form carbamoyl aspartate is catalyzed by ATCase (Fig. 1, circled 2). A single CPSase in Escherichia coli supplies carbamoyl phosphate for both pyrimidine and arginine biosynthesis. Thus, the bacterial ATCase catalyzes the first committed step in the pyrimidine pathway and is allosterically regulated (22). In contrast, there are two CPSases in mammalian cells, CPSII, the CAD activity committed to pyrimidine biosynthesis, and CPSI (23), an ammonia-dependent enzyme that initiates urea biosynthesis in the mitochondria. Because CAD CPSase catalyzes the first step in the pathway, ATCase is unregulated and there is no counterpart of the regulatory chain found in E. coli ATCase. The isolated CAD ATCase domain, from proteolytic digests (24) or expressed in E. coli (25), is a homotrimer of 34-kDa subunits. Kinetic and modeling studies showed that the mammalian domain shares a common catalytic mechanism, oligomeric structure, and tertia- ry-fold with the E. coli ATCase catalytic subunit (24 -27), including a composite active site comprised of residues from adjacent subunits. Qiu and Davidson (27) showed that substitutions at the trimer interface disrupt the CAD oligomeric structure suggesting that the trimeric ATCase interactions are a crucial organizing element in the hexamer.
Dihydroorotase-The 46-kDa CAD DHOase domain, which catalyzes the reversible condensation of carbamoyl aspartate to dihydroorotate ( Fig. 1, circled 3), is a zinc metalloenzyme. Whereas the active site of E. coli DHOase has two zinc ions and a carboxylysine that bridges the metal centers (28), the mammalian DHOase domain probably belongs to a different subgroup of the amidohydrolase superfamily (29). An extensive phylogenetic analysis (30) classified DHOases into two major classes thought to have arisen by an ancestral gene duplication. Type I DHOases are the most ancient and include domains of multifunctional proteins, such as CAD, subunits of multienzyme complexes, and monofunctional enzymes. The type II enzymes, such as E. coli DHOase, are a more recent evolutionary development, are smaller (38 kDa), and have undergone appreciable changes in sequence. The isolated CAD DHOase domain (31)(32)(33)(34) has only one zinc atom and is larger than its bacterial counterpart consistent with its assignment as a type I DHOase.

Dihydroorotate Dehydrogenase
Mammalian DHOdhase is a 43-kDa flavoprotein (FMN) localized in the mitochondria that oxidizes dihydroorotate to orotate (Fig. 1, circled 4). The electrons are transferred directly to the respiratory chain via ubiquinone (1). Biochemical and microscopic studies (1,35) showed that the enzyme is an integral membrane protein localized in the inner mitochondrial membrane with the active site facing the inner membrane space (Fig. 3). Mitochondrial import (36) is governed by an uncleaved bipartite sequence at the amino end of the polypeptide (Figs. 2 and 3) that consists of a mitochondrial targeting sequence (MT) and a membrane stop-transfer sequence (MA) that anchors the protein to the inner membrane.
Structural studies showed that truncated human DHOdhase (37), lacking the bipartite sequence, consists of two domains (Fig. 2). The active site is located within the large catalytic domain (CAT), whereas the small domain (QT) forms a tunnel that leads directly to the bound FMN and provides access to ubiquinone. Orotate is completely buried on the distal side of FMN so it is unlikely to enter via the same tunnel. Instead, a flexible loop moves out of position to accommodate substrate binding on the distal side of FMN. Antiproliferative agents such as leflunomide bind within the tunnel blocking access of ubiquinone to the active site.

The Multifunctional Protein UMP Synthase
UMP synthase is a bifunctional protein that catalyzes the last two steps of de novo pyrimidine biosynthesis ( Fig. 1) (1). The mammalian protein (Fig. 2) consists of a 24-kDa orotate phosphoribosyltransferase (OPRTase) that catalyzes the transfer of PRPP to orotate forming OMP (Fig. 1, circled 5) and a 28-kDa orotidine-5Ј-phosphate decarboxylase (ODCase) that decarboxylates OMP (Fig. 1, circled 6) forming UMP (38). Sedimentation analysis of mouse UMP synthase (39) identified a monomeric and two dimeric species, 5.1 S and 5.6 S. The monomer lacks ODCase activity, and the 5.1 S dimer is only partially active. The formation of the fully active 5.6 S dimer is induced by the binding of OMP or nucleotide analogs.

CTP Synthase
CTPSase (Fig. 1, circled 9) catalyzes the ATP-dependent transfer of the amide nitrogen of glutamine to the C-4 position of UTP to form CTP (40). Although there are no structural studies, sequence analysis revealed that CTPSase consists of two domains (Fig. 2): an amidotransferase domain (GLN) that hydrolyzes glutamine and an amidator domain (AMD) that catalyzes the ATP-dependent phosphorylation of UTP and its subsequent reaction with ammonia to form CTP. This reaction is the rate-limiting step in the formation of cytosine nucleotides and as such represents another important control locus in pyrimidine biosynthesis.

Intracellular Location of the Pyrimidine Biosynthetic Enzymes
Biochemical and microscopic studies showed that CAD is primarily cytosolic (35,41) with a smaller fraction in the nucleus. In the cytosolic compartment, CAD and UMP synthase are localized around and outside the mitochondria, and CAD appears to be associated with the cytoskeleton. Mitochondria are known to be anchored to the cytoskeletal network, so an interesting possibility is that CAD binds to and translocates along the filament to the mitochondria where DHOdhase is located (Fig. 4, path A). The physical association of CAD with the mitochondria is an attractive idea because under physiological conditions, the equilibrium strongly favors the formation of carbamoyl aspartate over dihydroorotate (42). Docking CAD near the mitochondria may allow a more efficient capture of dihydroorotate by DHOdhase and prevent the accumulation of carbamoyl aspartate in the cell.
The role of CAD in the nucleus was initially not given much credence (41). However, Angeletti and Engler (43) subsequently found that CAD is associated with the nuclear matrix in adenovirus infected cells, where it anchors the pTP protein at sites of active replication. Moreover, the ura2 protein, a CAD homolog in Saccharomyces cerevisiae consisting of CPSase, ATCase, and an inactive DHOase domain, was found within the nucleus (44).
More recently (45), a plasmid encoding a CAD-green fluorescent protein fusion protein has been transfected into BHK cells making it possible to monitor the movement of CAD in live cells. These studies showed that CAD is localized in the cytosol in resting cells or during the G 1 phase of the cell cycle but that a substantial fraction, 30%, is translocated into the nucleus during the S phase when the demand for pyrimidine nucleotides reaches a peak.
The localization of CAD in the nucleus near the site of RNA and DNA synthesis might seem to be an efficient arrangement, but the exclusive localization of DHOdhase within the mitochondria undermines this argument. Dihydroorotate synthesized in the nucleus would have to exit and diffuse into the mitochondria where it is oxidized to orotate, and orotate would then reenter the cytoplasm (Fig. 4, path B). This circuitous route would seem to negate the putative functional efficiency of locating CAD in the nucleus. Thus, the role of nuclear CAD remains uncertain, although it is possible that nucleocytoplasmic translocation plays a role in the regulation of the pathway or that CAD has a moonlighting function, perhaps in DNA replication or cell division, that is unrelated to pyrimidine biosynthesis.

Regulation of Pyrimidine Biosynthesis
The intracellular nucleotide pools are controlled within narrow limits in normal, resting cells but expand 3-to 4-fold in tumor cells (3). Even larger, 8-fold, increases occur in mitogenstimulated lymphocytes (7). The increased demand for nucleotides is satisfied in large part by up-regulation of de novo pyrimidine biosynthesis as a result of increased intracellular enzyme levels and metabolic control mechanisms.
Gene Expression-CAD gene expression is controlled at both the transcriptional and posttranscriptional level and is upregulated when resting cells enter the proliferative phase (46 -48). Myc binding to an upstream E box was found to be responsible for the increase in CAD gene transcription that occurs at the G 1 /S boundary as cells traverse the cycle (49,50). At the protein level, CAD is rapidly degraded at the onset of apoptosis by caspase cleavage (51). Less is known about the transcriptional regulation of the genes encoding the other pyrimidine biosynthetic enzymes, but there are many studies that show that the intracellular concentrations of these enzymes, and in some instances the mRNA, are appreciably higher in tumors and other rapidly growing cells (8).
The rate of cell growth probably sets the level of CAD and other enzymes and establishes the basal rate of de novo pyrim-idine biosynthesis, but the rapid changes in flux through the pathway requires precise metabolic control exerted by allosteric effectors and the activity of signaling cascades.
Metabolic Control-The CPSase activity of CAD is the major locus of control of de novo pyrimidine biosynthesis (1). The enzyme is subject to feedback inhibition by the end product UTP and is allosterically activated by PRPP. PRPP, a UMP synthase substrate (Fig. 1), is a feedforward activator and helps coordinate pyrimidine and purine production because it is also a substrate in the first step of purine biosynthesis. The allosteric effectors bind to a regulatory subdomain (Fig. 2, B3) at the carboxyl end of CPS.B in CAD (52) and other CPSases (53,54). Several studies (1) suggest that allosteric regulation of CAD governs the rate of pyrimidine biosynthesis in vivo. For example, expansion and depletion of the UTP and PRPP pools in cultured cells, tissue slices, and whole animals produce dramatic effects on the activity of the pyrimidine biosynthetic pathway that correlate with the effects of these ligands on the CAD CPSase activity. A nice illustration is provided by a recent analysis (55) showing that the suppressor of black mutation in Drosophila disables UTP inhibition of CAD allowing the pyrimidine pools to expand, thus providing the ␤-alanine precursors needed for normal cuticle pigmentation.
Carrey (56) made the important discovery that cAMP-dependent protein kinase A (PKA) phosphorylates purified CAD at two sites (Fig. 2). Modification of Ser-1406 in the regulatory subdomain (Fig. 2) abolishes UTP inhibition (56) and appreciably decreases the affinity of the enzyme for PRPP (57). The loss of feedback inhibition would be expected to stimulate pyrimidine biosynthesis and allow unchecked expansion of the nucleotide pools needed for cell proliferation; however, this effect would be partially offset by the reduction of PRPP activation.
Further insight into the growth state-dependent regulation of the pathway was provided by the discovery (58) that epidermal growth factor stimulation promotes the phosphorylation, via the MAP kinase (Erk1/2) cascade, of a single residue, Thr-456, in the CPS.A (A1) domain of CAD (Fig. 2). This modification has no effect on any CAD activity but converts UTP from an inhibitor to a modest activator and appreciably stimulates PRPP activation. Both changes in the allosteric transitions would be expected to increase the flux through the pyrimidine pathway. Moreover, a specific Thr residue (Fig. 2, Thr1037) in the B1 subdomain was found to be autophosphorylated both in vivo and in vitro (59), resulting in a selective increase in CPSase activity and modulation of the allosteric transitions.
CTP synthase activity regulates the relative size of the UTP and CTP pools and also helps coordinate the production of pyrimidine and purine nucleotides. The enzyme is allosterically activated by GTP and is inhibited by the end product CTP (Fig.  1). The yeast enzyme is phosphorylated by both PKA and protein kinase C (60). The phosphorylated enzyme is a dimer, but the binding of the substrates, UTP and ATP, induces the formation of the active tetramer. In contrast, the dephosphorylated enzyme is an inactive dimer that does not undergo nucleotide-induced tetramerization. Less is known about the mammalian enzyme. Human CTPSase is regulated by GTP and CTP (61), but whether or not it is also controlled by phosphorylation is unknown.
Growth State-dependent Regulation of the de Novo Pathway-A low basal level of pyrimidine biosynthesis is needed to sustain resting cells. The activity of the pathway increases 8-fold when BHK cells enter the exponential growth phase and then drops precipitously to basal levels as the culture becomes confluent (62). The transition to exponential growth is associated with a large increase in MAP kinase activity and the phosphorylation of the CAD MAP kinase site (Thr-456). As a result, UTP inhibition of CAD is abolished, and PRPP activation increases 21-fold, changes in the allosteric transitions that can account for the stimulation of pyrimidine biosynthesis. As the cultures approach confluence and growth ceases, Thr-456 is dephosphorylated, and there is a concomitant increase in PKA phosphorylation of CAD. The response to PRPP rapidly decreases, and the activity of the pyrimidine biosynthetic pathway is down-regulated. The sequential changes in CAD phosphorylation state coincide with the up-regulation of the pathway as the cells approach S phase and are reversed at the S/G 2 boundary as pyrimidine biosynthesis is down-regulated (45). The lack of down-regulation of pyrimidine biosynthesis in tumorigenic breast cancer cells, MCF7, has been attributed to the elevated MAP kinase activity that leads to persistent phosphorylation of the CAD MAP kinase site and a concomitant blockage of PKA phosphorylation of CAD (63).
These observations are consistent with a model (Fig. 5) in which the rate of pyrimidine biosynthesis is constrained in resting cells by UTP inhibition. As the cells enter the cycle, the MAP kinase mediated phosphorylation of CAD serves as a molecular switch uncoupling feedback inhibition and allowing the nucleotide pools to expand. CAD activity is now primarily controlled by PRPP, which maximally activates during S phase as a result of the sequential MAP kinase and PKA-mediated phosphorylations.

Conclusions
The steps involved in pyrimidine biosynthesis occur nearly universally in all organisms; however, compared with the monofunctional bacterial proteins, the eukaryotic enzymes have a more complex structural organization and a more sophisticated mode of control. The pathway is subject to diverse regulatory mechanisms including allosteric inhibition and activation, phosphorylation, and perhaps changes in intracellular location. Deciphering the operation and interplay of these controls in the cell remains a fascinating challenge.