Reversible Compartmentalization of de Novo Purine Biosynthetic Complexes in Living Cells

See allHide authors and affiliations

Science  04 Apr 2008:
Vol. 320, Issue 5872, pp. 103-106
DOI: 10.1126/science.1152241


Purines are synthesized de novo in 10 chemical steps that are catalyzed by six enzymes in eukaryotes. Studies in vitro have provided little evidence of anticipated protein-protein interactions that would enable substrate channeling and regulation of the metabolic flux. We applied fluorescence microscopy to HeLa cells and discovered that all six enzymes colocalize to form clusters in the cellular cytoplasm. The association and dissociation of these enzyme clusters can be regulated dynamically, by either changing the purine levels of or adding exogenous agents to the culture media. Collectively, the data provide strong evidence for the formation of a multi-enzyme complex, the “purinosome,” to carry out de novo purine biosynthesis in cells.

Purines are not only essential building blocks of DNA and RNA, but, as nucleotide derivatives, they also participate in a multitude of pathways in both prokaryotes and eukaryotes (1, 2). Biosynthetically, adenosine and guanosine nucleotides are derived from inosine monophosphate (IMP), which is synthesized from phosphoribosyl pyrophosphate (PRPP) in both the de novo and salvage biosynthetic pathways (Fig. 1). The salvage pathway catalyzes the one-step conversion of hypoxanthine to IMP by hypoxanthine phosphoribosyl transferase (HPRT), whereas the de novo pathway consists of 10 chemical reactions that transform PRPP to IMP. In higher eukaryotes (such as humans), the de novo pathway uses six enzymes, including three multifunctional enzymes: a trifunctional protein, TrifGART, that has glycinamide ribonucleotide (GAR) synthetase (GARS) (step 2), GAR transformylase (GAR Tfase) (step 3), and aminoimidazole ribonucleotide synthetase (AIRS) (step 5) activities; a bifunctional enzyme, PAICS, that has carboxyaminoimidazole ribonucleotide synthase (CAIRS) (step 6) and succinylaminoimidazolecarboxamide ribonucleotide synthetase (SAICARS) (step 7) activities; and a bifunctional enzyme, ATIC, that has aminoimidazolecarboxamide ribonucleotide transformylase (AICAR Tfase) (step 9) and IMP cyclohydrolase (IMPCH) (step 10) activities. In contrast, prokaryotes, such as Escherichia coli, use only monofunctional enzymes throughout this pathway, except for the bifunctional ATIC.

Fig. 1.

The de novo purine biosynthetic and salvage pathways. The de novo pathway transforms PRPP to IMP in 10 steps. Steps 2, 3, and 5 are catalyzed by a trifunctional enzyme, TrifGART; steps 6 and 7 are catalyzed by a bifunctional enzyme, PAICS; and Steps 9 and 10 are catalyzed by a bifunctional enzyme, ATIC.

Although studies of the individual enzymes in vitro have revealed much about their respective mechanisms of action, a number of in vitro attempts using kinetic analysis and/or binding measurements to demonstrate protein-protein interactions have not been fruitful, with few exceptions (36). In addition, there is scant evidence from in vivo cellular studies for the hypothesis that these enzymes act in concert within a multienzyme complex (7). In this study, we used fluorescence microscopy to investigate whether functional multienzyme complexes involved in de novo purine biosynthesis form in living mammalian cells under purine-rich and purine-depleted conditions, which affect the rate of metabolic flux (810).

We selected for study two human (h) enzymes involved with de novo purine biosynthesis as initial candidates for involvement in a multienzyme complex in vivo: the nonsequential hTrifGART protein, which catalyzes steps 2, 3, and 5, and formylglycinamidine ribonucleotide synthase (hFGAMS), which catalyzes step 4. These proteins were fused to either a green fluorescent protein (GFP) or an orange fluorescent protein (OFP) and were transiently expressed in HeLa cells that had been cultured in either purine-rich [minimum essential medium and 10% fetal bovine serum (FBS)] or purine-depleted (RPMI 1640 and dialyzed 5% FBS) media (9, 11). When expressed individually in cells grown in purine-rich media, both the hTrifGART (Fig. 2A) and hFGAMS (Fig. 2C) proteins exhibited diffuse cytoplasmic distributions, as previously found in 293T fibroblast cells and with their prokaryotic counterpart, E. coli cells (7). However, when these constructs were expressed in purine-depleted cells, we observed cytoplasmic clustering, estimated at ∼27% of the hTrifGART-GFP (Fig. 2B) and ∼77% of the hFGAMS-GFP (Fig. 2D) proteins. To determine whether the two enzymes colocalized into these clusters in low-purine conditions, we transiently coexpressed hTrifGART-GFP and hFGAMS-OFP in HeLacells (Fig. 2, G, H, and I); ∼60% of the transfected cells exhibited colocalization within the clusters (Fig. 2I).

Fig. 2.

Cellular localization of enzymes participating in the de novo purine biosynthetic pathway. (A to D) hTrifGART-GFP [(A) and (B)] and hFGAMS-GFP [(C) and (D)] transiently expressed in HeLa cells in purinerich [(A) and (C)] and purine-depleted [(B) and (D)] media. (E) Immunofluorescence of endogenous hTrifGART protein expressed in HeLa cells cultured under low-purine conditions. (F) Merged image by immunofluorescence of hTrifGART-GFP (green) and endogenous hTrifGART (red) expressed in HeLa cells grown in purine-depleted media. (G to I) hTrifGART-GFP [(G) and green in (I)] co-clusters with hFGAMS-OFP [(H) and red in (I)] in HeLa cells under purine-depleted conditions. (J to L) The control hC1THF-GFP [(J) and green in (L)], which does not participate directly in de novo purine biosynthesis, does not form clusters or colocalize with hFGAMS-OFP [(K) and red in (L)] in HeLa cells grown in purine-depleted media. Scale bar, 10 μm.

Z-scan imaging with confocal laser-scanning microscopy confirmed that the hTrifGART-GFP and the hFGAMS-OFP proteins were co-clustered in the cytoplasm (fig. S1). We also observed co-clustering between hFGAMS-GFP and hTrifGART-OFP after we reversed the fusion of the fluorescent probes in HeLa cells that were maintained in purine-depleted media (fig. S2). Additionally, similar cellular localization experiments were carried out with hTrifGART-GFP and hFGAMS-GFP in two additional human cell lines, HTB-125 and HTB-126; the results suggested that the purine-dependent clustering appears to occur in other human cell lines [supporting online material (SOM) Text]. These findings, when combined with the imaging results described for HeLa cells, suggest that the formation of this multienzyme complex involved in the de novo purine biosynthetic pathway may be functionally relevant in mammalian cells, particularly those challenged with minimal media.

To verify that the clustering we observed was not due to an artifact of expressing the fluorescent fusion constructs but was in fact present with endogenously expressed protein, we conducted immunofluorescence imaging of endogenous hTrifGART protein in HeLa cells grown in purine-depleted media; clustering was clearly observed (Fig. 2E), similar to what we found for the hTrifGART-GFP protein under the same conditions (Fig. 2B). We also identified some clustering of endogenous hTrifGART in the presence of purine-rich media (fig. S3), which we attribute either to the sensitivity of immunofluorescence in detecting protein expression or to the loss of cytosolic materials during fixation. We further demonstrated that endogenous hTrifGART co-clustered with hTrifGART-GFP in HeLa cells that were maintained in purine-depleted media (Fig. 2F), which supports our initial observations with the GFP fusion constructs. Taken together, these results suggest that multienzyme complexes form to satisfy the cellular demand for de novo purine biosynthesis.

We expanded our studies to include all six human enzymes of the pathway (Fig. 1): PRPP amidotransferase (hPPAT) (step 1), hTrifGART, hFGAMS, hPAICS, adenylosuccinate lyase (hASL) (step 8), and hATIC. As a control, we also included a tetrahydrofolate (H4F)–utilizing trifunctional enzyme, hC1THF, that possesses 5,10-methylene-H4F dehydrogenase, 5,10-methenyl-H4F cyclohydrolase, and 10-formyl-H4F synthetase activities. hC1THF is not directly involved in the de novo purine pathway but is responsible for synthesizing a key cofactor, 10-formyl-H4F, for both hTrifGART and hATIC activities (5, 12). We found that all six enzymes that directly participate in the de novo purine biosynthetic pathway co-clustered with hFGAMS-OFP in the cytoplasm of cells maintained in purine-depleted media (Fig. 2, G to I, and Fig. 3); however, hC1THF-GFP was diffuse throughout the cytoplasm, and thus did not co-cluster with hFGAMS-OFP (Fig. 2, J to L).

Fig. 3.

Colocalization of GFP fusion proteins with hFGAMS-OFP in HeLa cells grown in purine-depleted media. hPPAT-GFP (A), hPAICS-GFP (D), hASL-GFP (G), and GFP-hATIC (J) were co-transfected with hFGAMS-OFP (B, E, H, and K, respectively). (C, F, I, and L) Green channels correspond to the GFP-fusion constructs, and red channels correspond to hFGAMS-OFP in the merged images. Scale bar, 10 μm.

When the individual fluorescent protein constructs were transiently expressed in purine-depleted HeLa cells, the frequency of clustering was generally very low: ∼5 to 8% for hPPAT-GFP, hPAICS-GFP, hASL-GFP, and GFP-hATIC, but ∼27% for hTrifGART-GFP and ∼77% for hFGAMS-GFP. However, when co-transfected with hFGAMS-OFP under these conditions, the clustering frequency increased substantially to ∼22 to 32% for hPPAT-GFP, hPAICS-GFP, hASL-GFP, and GFP-hATIC, and ∼60% for hTrifGART-GFP. The percentage of colocalization of GFP-fusion proteins with hFGAMS-OFP was calculated by dividing the number of pixels that showed both green and red colors by the total number of pixels that showed either green or red colors (11, 13). Colocalization analysis collectively revealed that 78 ± 6% of GFP-fusion proteins co-clustered with hFGAMS-OFP. However, the analysis for hC1THF-GFP and hFGAMS-OFP exhibited 21 ± 5% colocalization, which suggests that these proteins do not colocalize in clusters: The representative image (Fig. 2L) clearly shows hFGAMS-OFP clusters on a diffuse background of hC1THF-GFP. Scattergram analysis was carried out as an alternative means of examining colocalization, and it showed the compartmentalization of all six enzymes participating in the de novo purine biosynthesis (table S1 and fig. S4). The varying percentage of cells exhibiting clustering under low-purine conditions may be attributed to cells not being synchronized with respect to the cell cycle. The G1 and S phases require more purines to support RNA and DNA synthesis than the other phases of the cell cycle (14); thus, we speculate that the population of the cells exhibiting clustering may represent cells in the G1 or S phases. This rationale is further supported by the previously measured ∼five- to sixfold increase in the level of mRNAs for PPAT and SAICARS in the G1 and S phases over those in the G0 phase of synchronized rat 3Y1 fibroblasts (15).

To assess the reversibility of the multienzyme complexes, we imaged the cellular distribution of hFGAMS-GFP while purine-depleted medium was exchanged for purine-rich medium, and observed that the hFGAMS-GFP clusters dispersed within ∼0.5 to 1.5 hours of HeLa cells being incubated with purine-rich medium (Fig. 4, A and B). When purine-depleted medium was subsequently restored, clusters began to re-form in ∼1 to 2 hours (Fig. 4, B and C). Additionally, we performed fluorescence recovery after photobleaching on hFGAMS-GFP in HeLa cells that had been incubated in purine-depleted medium. We clearly observed mobility of the clusters through the rapid fluorescence recovery as a function of time (fig. S5). Collectively, the predominance of clustering appears to be linked to insufficient levels of purines, which suggests that there is a cellular sensor of the external purine conditions and that the mobile clusters represent functional complexes.

Fig. 4.

Reversible formation of clusters by hFGAMS-GFP. (A) hFGAMS-GFP clusters when HeLa cells are exposed to purine-depleted medium. (B) Clusters disperse within ∼0.5 to 1.5 hours upon incubation with purine-rich medium. (C) The clusters begin to re-form within ∼1 to 2 hours after purine-depleted medium is restored. Scale bar, 10 μm.

Furthermore, we imaged the effects of introducing either an exogenous purine source, hypoxanthine, or a well-known antagonist of hFGAMS, azaserine, to inhibit the metabolic flux in the de novo pathway of HeLa cells grown under purine-depleted conditions (8). Hypoxanthine, when added to fibroblast cultures in concentrations as low as 1 to 10 μM, significantly inhibits (>50%) de novo purine biosynthesis through its effect on adenosine and guanosine nucleotides in the cellular pools (16, 17), and the increased levels of adenosine and guanosine nucleotides modulate activities of PPAT as well as PRPP synthetase (8). In addition, cells grown in purine-depleted media are deficient in adenosine 5′-triphosphate, with both the restoration of cellular growth rates and the timing of the G1-to-S transition dependent on both the de novo and salvage pathways (14). We propose that the initial increased clustering is a response to hypoxanthine-induced inhibition of the de novo pathway and that the retention of the clusters reflects demand for purines in cellular growth (fig. S6, A, B, and solid squares in E). When we incubated azaserine with purine-depleted cells, we anticipated that hFGAMS clusters would at least be maintained because of the greater demand for purine biosynthesis. An azaserine block increases the rate of the early steps of de novo purine biosynthesis in fibroblasts that lack HPRT activity (17), which is consistent with our observed maintenance of the clusters (fig. S6, C, D, and open circles in E). Collectively, the association and dissociation of the multienzyme complex reflected changes in cellular purine levels that were imposed by the addition of external reagents that regulate purine metabolic flux.

From our studies, the various roles that the multienzyme complex plays in cells are not yet clear. These functional complexes may produce efficient substrate channels that link the 10 catalytic active sites. To date, extensive kinetic analyses of E. coli PPAT and GARS (6), Acetobacter aceti purE and purK (bacterial counterparts of CAIRS) (3), and recent crystal structures of octameric hPAICS (18) have revealed putative substrate channels for individual steps in the de novo pathway. Additionally, clustering of the 10 active sites may provide an efficient means of globally regulating purine flux under varying environmental conditions. However, we have not yet identified a specific subcellular compartment with which these multienzyme complexes associate, in contrast to, for example, the actin cytoskeletal association observed for aldolase (19). However, our cross-linking experiments with formaldehyde on living cells grown in purine-depleted media, and subsequent Western blot analysis, revealed that endogenous hTrifGART proteins produced a cross-linked adduct of molecular weight greater than 500 kDa (fig. S7), which suggests that hTrifGART is indeed compartmentalized in close proximity with other proteins, whose identities are under investigation. Moreover, recent proteomic studies suggest that hTrifGART and hFGAMS are probably phosphorylated at binding motifs recognized by 14-3-3 proteins that play important roles in many biological processes, including cell-cycle regulation and signal transduction (20), which raises the intriguing possibility that phosphorylation is involved in regulation of the complex.

Although the association of metabolic enzymes has been claimed for enzymes involved in the glycolytic pathway (21, 22), to our knowledge this complex, the glycosome, has not been identified in living mammalian cells. By analogy, the present clusters observed in the de novo purine biosynthetic pathway may constitute a “purinosome.” The formation of the purinosome appears to be dynamically regulated by stimulation of de novo purine biosynthesis in response to changes in purine levels. The purinosome may be a general phenomenon in all cell types during specific stages of the cell cycle, along with posttranslation modifications. Because of the relevance of de novo purine biosynthesis to human diseases, the purinosome may represent a new pharmacological opportunity for therapeutic intervention.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S9

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article