Developmental function of Nm23/awd: a mediator of endocytosis
- 117 Downloads
The metastasis suppressor gene Nm23 is highly conserved from yeast to human, implicating a critical developmental function. Studies in cultured mammalian cells have identified several potential functions, but many have not been directly verified in vivo. Here, we summarize the studies on the Drosophila homolog of the Nm23 gene, named abnormalwingdiscs (awd), which shares 78% amino acid identity with the human Nm23-H1 and H2 isoforms. These studies confirmed that awd gene encodes a nucleoside diphosphate kinase, and provided strong evidence of a role for awd in regulating cell differentiation and motility via regulation of growth factor receptor signaling. The latter function is mainly mediated by control of endocytosis. This review provides a historical account of the discovery and subsequent analyses of the awd gene. We will also discuss the possible molecular function of the Awd protein that underlies the endocytic function.
KeywordsNm23 awd Prune Killer-of-prune Drosophila
Nm23 was first isolated as a cDNA clone that was down-regulated in highly metastatic derivatives of murine melanoma cell line K-1735 . The gene is highly conserved in the eukaryotes from yeast to mammals. In vertebrates, it consists of a gene family of eight.
Nm23 has been shown to inhibit metastasis, but not primary tumor growth, in xenografts of cells from human breast cancer, murine melanoma, rat colon cancer, and human oral squamous cancer . However, in clinical cancer samples, this correlation is much weaker. While in initial breast cancer studies, reduced Nm23 expression levels were consistent with increased metastasis , in later studies, no clear correlation could be discerned when comparing benign tumors and metastatic cancers [4, 5, 6, 7, 8]. It has become clear that Nm23 plays a critical role in normal tissue functions, and both up- and down-regulation of this gene can disrupt growth and differentiation.
The function of Nm23
Further complicating the issue of Nm23’s role in tumor progression is the multiple enzymatic functions assigned to Nm23: (a) Nucleoside diphosphate kinase (NDPK) that transfers the terminal phosphate group from ATP to a non-ATP nucleoside diphosphate (such as GDP) through the formation of an intermediate histidine-phosphate linkage at histidine 118 (residue 119 in Drosophila) . (b) DNA binding and DNA nuclease involved in transcriptional regulation [9, 10]. (c) DNase activity that is activated by granzyme A in caspase-independent apoptosis . (d) Histidine-dependent protein kinase [12, 13, 14]. This enzymatic activity is similar to that of NDPK but toward protein substrates . Some tantalizing kinase targets have been identified such as aldolase C and kinase suppressor of Ras (Ksr), but their in vivo relevance requires further investigation . It is therefore critical that the apparently complex Nm23 functions be examined in physiologically relevant model systems. The Drosophila model to date has provided the most extensive set of studies, in part because Drosophila genome encodes only one major Nm23 homolog that accounts for >98% of the NDPK activity in the larval extracts . Interestingly, many of these Drosophila studies originated from completely unrelated circumstances.
The case of Killer-of-prune
The discovery of awd
In a totally different line of work in the 1970s and 1980s, Allen Shearn was interested in the development of imaginal discs. Imaginal discs are small epithelial structures that are set aside during embryogenesis for future development of adult tissues, including all appendages and the eye–antenna complex. The imaginal discs have been a favorite system of Drosophila geneticists for studying developmental control of cell growth and differentiation. The Shearn laboratory conducted a genetic screen for late larval/pupal lethality, in the belief that among these mutations would be those regulating imaginal disc development. Indeed, 20 alleles, representing 12 genes, from this screen showed disc defects [20, 21]. One of these alleles encodes a single 0.8-kb transcript, mapped to the distal tip of the chromosome arm 3R, and was named abnormalwingdiscs (awd) . The mutant imaginal discs, including wing, leg, and eye-antenna, could not differentiate properly, and showed extensive cell death. In a more complicated organ transplant experiment, they demonstrated the role of awd in ovary development. In Drosophila as in mammals, gonad development begins with the specification of germ cells at the very early stage of embryogenesis . During gastrulation in both flies and mammals, the primordial germ cells move into the embryonic proper and coalesce with somatic gonadal mesoderm to form the primitive gonads. Dearolf et al. first transplanted the premature ovaries containing both somatic cells and germ cells from mutant larvae into wild-type surrogate females of corresponding age. None of the transplanted mutant ovaries developed fully. On the other hand, homozygous mutant germ cell precursors isolated from early awd mutant embryos transplanted into wild-type embryos (before gastrulation) gave rise to normal ovaries and fertile gametes . Since in the germ cell transplant, the surrogate wild-type females provided the somatic components of the future ovaries, the observation indicates that the ovarian phenotype in awd mutants is the result of defects in the somatic components (follicle cells) of the ovaries. This will become apparent later (see below).
In a case of accidental but beneficial scientific convergence, the isolation and cloning of awd happened at about the same time Patricia Steeg identified Nm23 as the first metastasis suppressor gene , and it was quickly realized that awd is the Drosophila homolog of Nm23, which shares 78% amino acid identity with the Nm23-H1 and H2 isoforms .
Enzymatic activity of the Awd protein
The early studies from the Shearn laboratory showed that awd mutant larval brain exhibited mitotic defects correlated with defective microtubule polymerization . In a display of remarkable insight, the Shearn group noted that Nickerson and Wells had reported the purification of a 17-kDa protein, the same size as the Awd protein, from bovine brain microtubules that exhibited nucleoside diphosphate kinase activity (NDPK) . Remarkably, antibody raised against the bovine NDPK cross-reacted with Awd. They went on to show that larval extracts from awd mutant larvae contained dramatically reduced NDPK activity . In a concurrent but independent study, the NDPK enzyme from Dictyostelium discoideum was isolated and the amino acid sequence showed 60% identity with Awd . All these suggested that the Nm23/Awd family of proteins is a NDPK enzyme. This notion soon proved to be true .
In a subsequent series of elegant studies, the Shearn laboratory showed that recombinant Awd protein contained NDPK activity in vitro , and that mutation at the presumptive active site histidine residue (H119A) abolished this NDPK activity and failed to rescue awd mutant phenotype when introduced as a transgene . Interestingly, they also found that human Nm23-H2, but not H1, could completely rescue the awd lethal phenotype when expressed from the awd gene promoter, and neither could rescue the female sterile phenotype . The implication is that, as suggested by the authors, the NDPK activity is necessary but not sufficient to account for all the awd functions. Although this notion could be explained as a failure of the human proteins to interact with all the native Drosophila targets, it may also suggest that activities other than NDPK are physiologically relevant.
The identity of K-pn
In another interesting twist, soon after they isolated the awd alleles, the Shearn laboratory set out to isolate deletion mutants of awd to serve as complete null alleles. This led them to revisit the K-pn mutant. Back in the 1960s, Lifschytz and Falk reasoned that the way to resolve the K-pn function was to isolate revertants (Supplementary Table 1) of the K-pn conditional lethal phenotype. They took advantage of Sturtevant’s original demonstration: complete lethality of male offspring from pn mother and K-pn father. They mutagenized the K-pn males with X-ray, which produced chromosomal deletions, mated them with pn females, and recovered rare male progenies . These male escapers represented the K-pn revertants. It is important to note, from the hindsight, that this mutagenesis worked because K-pn is a dominant allele; that is, one copy of (heterozygous) K-pn can cause lethality in pn mutant (Fig. 1). Therefore, loss-of-function mutation in one K-pn allele alone could result in viable male offspring. This also confirmed that K-pn is a gain-of-function or neomorphic mutant (Supplementary Table 1), since its phenotype could be ameliorated by loss-of-function mutations. However, surprisingly, all attempts to generate homozygous stocks of these revertants failed, suggesting that all of the revertants were homozygous lethal . In addition, these revertants could not complement each other for their lethality (Supplementary Table 1). This is peculiar since non-complementation usually indicates that the alleles involved are different mutations of the same gene. Since loss-of-function K-pn mutation itself was expected in the revertants, could this mean all revertants they isolated by chromosomal deletion (X-ray irradiation) were new mutations in the K-pn gene? This seemed unlikely at the time and so these authors suggested that the revertants generated by irradiation—usually large deletion mutants—uncover a few very closely linked genes including K-pn (so they could not complement each others; Supplementary Table 1), and that these genes are homozygous lethal. Shearn reasoned that since K-pn and awd are closely linked at the tip of the chromosome arm 3R—note that he had no reason to suspect that awd and K-pn are the same gene—he could generate at least a few new awd deletion alleles, in addition to the expected K-pn deletion alleles, by repeating the mutagenesis scheme of Lifschytz and Falk. To their surprise, all the K-pn revertants generated failed to complement the existing awd alleles . More importantly, awd cDNA cloned from the K-pn mutant could cause lethality in pn mutants when expressed as a transgene . Subsequent cloning confirmed that K-pn in fact is an allele of awd, which carries a proline to serine substitution at residue 97 [26, 27]. This allele was renamed awdK-pn (Supplementary Table 1). Interestingly, proline 97 is a highly conserved residue among NDPKs [26, 27, 31]. The NDPK from K-pn is not altered in its catalytic activity, but the peptide chain is more flexible and exhibited reduced stability to heat treatment or denaturing agents . The increased flexibility is consistent with the neomorphic nature of the mutant, as it is possible that the mutant protein can interact with additional proteins. However, the Steeg laboratory showed that the K-pn mutation (P96S in Nm23-H1) retained the NDPK activity but lost the histidine-dependent protein kinase function , and that the mutant could not inhibit in vitro motility of cancer cells . The relative importance of NDPK and protein kinase functions of Nm23/Awd obviously requires further studies, and the enigmatic K-pn mutant may yet hold the key to this question.
The function of Pn
Unfortunately, the functional basis of the K-pn phenotype has remained a mystery. What does a fruit fly eye color mutant have to do with a metastasis suppressor gene? The pn gene was cloned and sequenced soon after the identity of K-pn was revealed [34, 35]. Since the NDPK function is to generate non-ATP nucleoside triphosphates, notably GTP, the authors’ suggestion that Pn protein may be a GTPase-activating protein (GAP) generated plenty of excitement, as this function would link Nm23/Awd to the all-important GTPases such as the Ras family of proteins or G-proteins. However, subsequent analysis did not support a GAP function for Pn. Further sequence analyses suggested that Pn belongs to a large family of phosphodiesterase  and that human Nm23-H1 physically interacts with human Pn . It was also shown that human Pn protein contains a cyclic nucleotide phosphodiesterase activity that promotes cancer cell motility . At the time of Sturtevant, there were only three independent alleles of pn (alleles 1, 2 and 3), all of which showed lethal interaction with K-pn. All are insertions that resulted in peptide truncations of varying lengths . These mutants do not produce proteins detectable by a polyclonal antibody raised against full-length protein. Most, if not all, of the pn mutants isolated later show lethality when combined with K-pn. Whether they show residual enzymatic activities is not known. Curiously, some of these mutants resulted in short peptide truncation at the C-terminus. These observations indicate that the C-terminus is critical for function or protein stability.
Another important link among Nm23/Awd, Pn, and the nucleoside diphosphate–triphosphate conversion function is the original Pn function. pn mutants affect Drosophila eye color. The Drosophila red eye pigment is a derivative of pteridines (drosopterin). The first and rate-limiting step in pteridine synthesis is catalyzed by the enzyme GTP cyclohydrolase (GTPCH), which converts substrate GTP to dihydroneopterine triphosphate . It has been shown that in pn mutants, the pteridine level is reduced and the activity levels of GTPCH become variable . Since Nm23/Awd as a NDPK is presumed to supply nucleoside triphosphate, thus driving the forward reaction catalyzed by GTPCH, the defects in pn mutants are consistent with the genetic data that implicate Pn as an antagonist of Nm23/Awd. The antagonistic relationship between Nm23 and Pn was later demonstrated in human cells [37, 38].
Sturtevant himself offered a hypothesis in explaining the K-pn/pn interaction : “One possible hypothesis is that pn leads to the production (or accumulation) of some substance that is absent, or present only in small amounts, in pn+ flies, and that K-pn converts this substance to a different and toxic substance. Such a hypothesis at least suggests experiments to be carried out, and offers some hope of relating the phenomena to the reactions concerned in the synthesis of pigments.” This insight suggested a link to GTP metabolism and may still serve as a guide for Nm23 researchers today. No such substance has been detected so far. However, indirect evidence may hint at the validity of the hypothesis. In a series of protein analyses, Timmons and Shearn showed that the phosphorylation (activation) pattern of K-pn protein is not altered compared with the wild-type Awd protein . This observation, combined with the finding that K-pn mutant changed the substrate-binding cleft of the NDPK enzyme, suggested to the authors that K-pn may catalyze an ectopic substrate [27, 41]. More recently, the Shearn laboratory repeated the genetic screen for K-pn revertants and isolated a rare non-awd allele [42, 43]. Interestingly, this allele encodes a protein exhibiting glutathione S-transferase (GST) activity. GST is usually a detoxification enzyme, but in some cases, addition of glutathione to certain substances can lead to cytotoxicity . Although direct biochemical evidence is lacking, the authors suggested that the GST activity is needed in the pn-K-pn flies to convert the hypothetical ectopic substrate into the toxic substance. Mutation of the GST function therefore rescued the pn-K-pn lethality.
awd in neurotransmission
In another case of serendipity, a team of neurobiologists in Arizona and India were interested in using a genetic approach to solve the endocytic pathway that regulates neurotransmission at the synaptic junctions. One ingenious idea they came up with was to use a temperature-sensitive allele of the Drosophila dynamin gene, shibire (shi). At non-permissive temperature (29°C), the flies become paralyzed (shibire is the Japanese term for numbness) due to defective uptake of neurotransmitters at the neuromuscular junctions [38, 45]. In order to identify other players in the vesicle recycling pathway, the authors conducted a genetic screen for enhancers of the shi phenotypes. That is, they looked for second site mutations that would make shi flies pass out at a lower temperature (25°C). They isolated three such alleles and to their (and everybody else’s) surprise, all three alleles were mutations of the awd gene . The authors went on to show that the existing awd mutations could indeed cause the same enhancer-of-shi phenotype, and the His119 mutant did not confer such activity. This important finding suggested a highly specific and almost exclusive functional interaction between dynamin and awd.
awd and epithelial tubule morphogenesis
Our laboratory has been interested in the developmental function of the von Hippel-Lindau tumor suppressor gene (VHL), the human germline mutations of which predispose carriers to kidney cancer, a blood vessel tumor called hemanigioblastoma, and a number of other, less frequent tumors . We had shown that VHL loss-of-function in Drosophila caused disruption of an epithelial tubule organ termed trachea . We subsequently found that Drosophila VHL protein physically interacts with Awd in a yeast two-hybrid screen (unpublished data). This prompted us to examine the possible trachea phenotype in awd mutants. This was an attractive idea since the trachea is an epithelial tubule organ that has been used to model the mammalian vascular system as well as other epithelial ductal tissues such as mammary gland and kidney tubules. The implications for human epithelial tumors and perhaps angiogenesis would be highly interesting.
Interestingly, in collaboration with Janis O’Donnell, we showed that the neurotransmitter dopamine could promote endocytosis in the tracheal cells . Dopamine synthetic pathway enzyme tyrosine hydroxylase (TH) requires the cofactor tetrahydrobiopterin, which also relies on the activity of GTPCH for biosynthesis. Therefore, the cellular level of dopamine is also dependent on the function of GTPCH (encoded by the Punch gene in Drosophila) . We demonstrated that Punch mutant exhibited the same tracheal phenotype as in awd and shi, with over-accumulation of FGFR on the tracheal cell surface. Furthermore, awd and Punch double mutants resulted in exacerbated phenotype . Thus, awd and Punch may function in parallel to promote endocytosis of FGFR, or they can work in the same GTP-GTPCH enzymatic pathway that leads to elevated endocytosis. The latter, indirect role for awd is not favored in light of the strong and specific genetic interaction between awd and shi/dynamin . In addition, we showed that in Punch mutants dynamin is down-regulated and diffused in the tracheal cells , while we did not observe such change in dynamin expression pattern in awd mutants (unpublished observation).
Thus, these studies reaffirmed the endocytic function of Awd and suggested a plausible mechanism for the Nm23 antimetastasis function. That is, Nm23/Awd may be involved in internalization and turnover of growth factor receptors. In the absence of Nm23/Awd, the cells are over-sensitized to chemotactic signals, resulting in elevated metastatic potential. It would be interesting to examine whether the pn–awd functional relationship can be manifested in this epithelial model system.
awd and epithelial cell invasion
Given our finding that Awd expression is lost in border cells prior to initiation of cell migration , we tested if forced expression of Awd in the border cells could dominantly interfere with developmentally programmed invasion of these cells. Indeed, re-expression of Awd in border cells reduced the level of Pvr and ameliorated downstream signaling and blocked migration  (see Fig. 3b for an example). Over-expression of a dominant-negative shi/dynamin transgene could result in over-accumulation of Pvr and spinning (without forward moving) of border cells, and this phenotype could be rescued by co-expression of awd . Interestingly, Awd also regulates the turnover of Domeless, the Drosophila cytokine receptor that is also required for border cell migration, but has no effect on the level of an ectopically expressed human transmembrane protein CD8 . Therefore, the presumed endocytic function of Awd is not promiscuous. The exact specificity of this function is currently unknown. It is worth noting that Awd protein is detected only in the somatic follicle cells and not in the germ cells , which is consistent with the early Shearn finding that the ovarian function of awd is exerted in the somatic component of the ovary, not in the germ cells .
The molecular function of Awd protein
The genetic studies in Drosophila clearly implicated a role of Awd in endocytosis. The most straightforward explanation for this function is that Nm23/Awd is a GTP supplier for dynamin GTPase. However, a guanine nucleotide exchange factor (GEF)-like function is not likely since physical interaction between Nm23/Awd and dynamin has not been demonstrated in physiological conditions. It is therefore possible that Nm23/Awd affects dynamin-mediated endocytosis through an indirect mechanism. Further genetic experiments should help resolve this issue.
There are indeed other ways Awd can regulate endocytosis. For example, David Deitcher  proposed that the NDPK activity of Nm23 can provide CTP to the enzyme cytidine diphosphate diacylglycerol synthase (CDS) that catalyzes a rate-limiting step in the phosphatidylinositol (4,5)-bisphosphate [PtdIns(4,5)P2] synthetic pathway . PtdIns(4,5)P2 has been implicated as a second messenger that regulates clathrin-mediated endocytosis of receptors [61, 62]. Dynamin is known to bind PtdIns(4,5)P2 through its pleckstrin homology domain  and PtdIns(4,5)P2 is also required in subsequent steps in endocytosis . On the other hand, since other small GTPases such as Rab5 and Arf6 are also regulators of endocytosis, a GTP supplier function of Nm23/Awd may also apply in such capacity. It is interesting to note that among the myriad of interacting proteins of Nm23 in mammalian cells, many are related directly or indirectly with the GTPases, such as Arf6 , TIAM1 (a guanine exchange factor for Rac) , Lbc (a guanine exchange factor for Rho) , and Rad . Whether these GTPase-related functions hold true requires further in vivo investigation.
The developmental studies showed that substitution of the active-site histidine residue that is critical for the NDPK activity could not stall border cell migration . This is consistent with previous finding that this residue is required for rescuing the enhancer of shi phenotype . Curiously, this residue is not required for suppressing the in vitro motility (assayed by Boyden chamber) of the metastatic breast cancer cells [32, 33]. However, the histidine substitutions employed in the two systems are different (phenylalanine in human versus alanine in fly). It is therefore difficult at this time to draw direct comparison. On the other hand, human mutants that affect the histidine-dependent protein kinase activity failed to suppress motility of the cancer cells [32, 33]. So far, very few Nm23 protein kinase targets have been identified , but it would be highly interesting to investigate whether any of the endocytic components would be a phosphorylation target of Nm23/Awd.
The story of Drosophilaawd illustrated the value of an accessible genetic system that can bring seemingly unrelated interests together in solving a complex biological problem. Much of the progress in awd research appears serendipitous, but the underlying theme is that the careful use of the genetic system can lead to unexpected and valuable new insights, which may be otherwise difficult to attain using biased approaches. We have shown an important endocytic function for awd. This should lead to further investigation, in both mammalian and fly systems, into possible Nm23/Awd functions in regulating pathways that are known to involve endocytosis and vesicle recycling, such as epithelial cell adhesion and integrin turnover. More recent results, both published and unpublished, have also indicated an interaction between Nm23/Awd and the tumor suppressor protein VHL. For example, Drosophilaawd and vhl mutants show similar tracheal phenotypes . Human VHL can stabilize Nnm23-H1 in the endocytic pathway that regulates FGFR signaling strength . We also found that Awd and Drosophila VHL protein can physically interact (Dammai and Hsu, unpublished observations). Such interaction is likely independent of the known E3 ubiquitin ligase activity of VHL, and thus independent of the hypoxia-inducible-factor function , because VHL stabilizes Nm23, not destabilizing it. Nonetheless, such novel interactions warrant further investigations. Although so far in Drosophila the main Awd activity appears to be an involvement in endocytosis, which can manifest into various physiological functions, Nm23 homologs in other developmental systems have suggested other functions such as regulation of cyclin dependent kinase inhibitors (reviewed in Bilitou et al., this issue). Also, mouse Nm23 has been shown to localize to primary cilia . This is highly interesting considering the early observations that Nm23/Awd is associated with microtubules [17, 24]. Such activity may link Nm23/Awd function to cilia-dependent signaling pathways such as sonic hedgehog, Wnt, and TGFβ . The established Drosophila assay system—imaginal discs, trachea, and the border cells—should continue to provide a valuable tool for testing the Nm23/Awd functions in vivo, including those of the various disease-related Nm23 mutants.
The work is supported by grants from the National Institutes of Health to T.H. (RO1GM57843) and V.D. (RO1CA128002).
- 4.Sirotkovic-Skerlev M, Krizanac S, Kapitanovic S, Husnjak K, Unusic J, Pavelic K (2005) Expression of c-myc, erbB-2, p53 and nm23-H1 gene product in benign and malignant breast lesions: coexpression and correlation with clinicopathologic parameters. Exp Mol Pathol 79:42–50. doi:10.1016/j.yexmp.2005.02.004 PubMedCrossRefGoogle Scholar
- 5.Galani E, Sgouros J, Petropoulou C, Janinis J, Aravantinos G, Dionysiou-Asteriou D, Skarlos D, Gonos E (2002) Correlation of MDR-1, nm23-H1 and H Sema E gene expression with histopathological findings and clinical outcome in ovarian and breast cancer patients. Anticancer Res 22:2275–2280PubMedGoogle Scholar
- 10.Ma D, Xing Z, Liu B, Pedigo NG, Zimmer SG, Bai Z, Postel EH, Kaetzel DM (2002) NM23-H1 and NM23-H2 repress transcriptional activities of nuclease-hypersensitive elements in the platelet-derived growth factor-A promoter. J Biol Chem 277:1560–1567. doi:10.1074/jbc.M108359200 PubMedCrossRefGoogle Scholar
- 25.Wallet V, Mutzel R, Troll H, Barzu O, Wurster B, Veron M, Lacombe ML (1990) Dictyostelium nucleoside diphosphate kinase highly homologous to Nm23 and Awd proteins involved in mammalian tumor metastasis and Drosophila development. J Natl Cancer Inst 82:1199–1202. doi:10.1093/jnci/82.14.1199 PubMedCrossRefGoogle Scholar
- 27.Timmons L, Xu J, Hersperger G, Deng XF, Shearn A (1995) Point mutations in awdKpn which revert the prune/Killer of prune lethal interaction affect conserved residues that are involved in nucleoside diphosphate kinase substrate binding and catalysis. J Biol Chem 270:23021–23030. doi:10.1074/jbc.270.39.23021 PubMedCrossRefGoogle Scholar
- 32.Freije JM, Blay P, MacDonald NJ, Manrow RE, Steeg PS (1997) Site-directed mutation of Nm23-H1. Mutations lacking motility suppressive capacity upon transfection are deficient in histidine-dependent protein phosphotransferase pathways in vitro. J Biol Chem 272:5525–5532. doi:10.1074/jbc.272.9.5525 PubMedCrossRefGoogle Scholar
- 33.MacDonald NJ, Freije JM, Stracke ML, Manrow RE, Steeg PS (1996) Site-directed mutagenesis of nm23-H1. Mutation of proline 96 or serine 120 abrogates its motility inhibitory activity upon transfection into human breast carcinoma cells. J Biol Chem 271:25107–25116. doi:10.1074/jbc.271.41.25107 PubMedCrossRefGoogle Scholar