NEMO specifically recognizes K63-linked poly-ubiquitin chains through a new bipartite ubiquitin-binding domain.

An important property of NEMO, the core element of the IKK complex involved in NF-kappaB activation, resides in its ability to specifically recognize poly-ubiquitin chains. A small domain called NOA/UBAN has been suggested to be responsible for this property. We recently demonstrated that the C-terminal Zinc Finger (ZF) of NEMO is also able to bind ubiquitin. We show here by ZF swapping and mutagenesis that this represents its only function. While neither NOA nor ZF shows any preference for K63-linked chains, we demonstrate that together they form a bipartite high-affinity K63-specific ubiquitin-binding domain. A similar domain can be found in two other proteins, Optineurin and ABIN2, and can be freely exchanged with that of NEMO without interfering with its activity. This suggests that the main function of the C-terminal half of NEMO is to specifically bind K63-linked poly-ubiquitin chains. We also demonstrate that the recently described binding of NEMO to linear poly-ubiquitin chains is dependent on the NOA alone and does not require the presence of the ZF.

A highly sensitive cell-based luciferase assay for high-throughput automated screening of SARS-CoV-2 nsp5/3CLpro inhibitors.

Effective drugs against SARS-CoV-2 are urgently needed to treat severe cases of infection and for prophylactic use. The main viral protease (nsp5 or 3CLpro) represents an attractive and possibly broad-spectrum target for drug development as it is essential to the virus life cycle and highly conserved among betacoronaviruses. Sensitive and efficient high-throughput screening methods are key for drug discovery. Here we report the development of a gain-of-signal, highly sensitive cell-based luciferase assay to monitor SARS-CoV-2 nsp5 activity and show that it is suitable for the screening of compounds in a 384-well format. A benefit of miniaturisation and automation is that screening can be performed in parallel on a wild-type and a catalytically inactive nsp5, which improves the selectivity of the assay. We performed molecular docking-based screening on a set of 14,468 compounds from an in-house chemical database, selected 359 candidate nsp5 inhibitors and tested them experimentally. We identified two molecules which show anti-nsp5 activity, both in our cell-based assay and in vitro on purified nsp5 protein, and inhibit SARS-CoV-2 replication in A549-ACE2 cells with EC50 values in the 4-8 µM range. The here described high-throughput-compatible assay will allow the screening of large-scale compound libraries for SARS-CoV-2 nsp5 inhibitors. Moreover, we provide evidence that this assay can be adapted to other coronaviruses and viruses which rely on a viral protease.

A highly sensitive cell-based luciferase assay for high-throughput automated screening of SARS-CoV-2 nsp5/3CLpro inhibitors.

Effective drugs against SARS-CoV-2 are urgently needed to treat severe cases of infection and for prophylactic use. The main viral protease (nsp5 or 3CLpro) represents an attractive and possibly broad-spectrum target for drug development as it is essential to the virus life cycle and highly conserved among betacoronaviruses. Sensitive and efficient high-throughput screening methods are key for drug discovery. Here we report the development of a gain-of-signal, highly sensitive cell-based luciferase assay to monitor SARS-CoV-2 nsp5 activity and show that it is suitable for high-throughput screening of compounds in a 384-well format. A benefit of miniaturisation and automation is that screening can be performed in parallel on a wild-type and a catalytically inactive nsp5, which improves the selectivity of the assay. We performed molecular docking-based screening on a set of 14,468 compounds from an in-house chemical database, selected 359 candidate nsp5 inhibitors and tested them experimentally. We identified four molecules, including the broad-spectrum antiviral merimepodib/VX-497, which show anti-nsp5 activity and inhibit SARS-CoV-2 replication in A549-ACE2 cells with IC 50 values in the 4-21 µM range. The here described assay will allow the screening of large-scale compound libraries for SARS-CoV-2 nsp5 inhibitors. Moreover, we provide evidence that this assay can be adapted to other coronaviruses and viruses which rely on a viral protease.

Inhibition of NEMO, the regulatory subunit of the IKK complex, induces apoptosis in high-risk myelodysplastic syndrome and acute myeloid leukemia.

In high-risk myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML), blasts constitutively activate the antiapoptotic transcription factor nuclear factor-kappaB (NF-kappaB). Here, we show that this NF-kappaB activation relies on the constitutive activation of the IkappaB kinase (IKK) complex, which is formed by the IKKalpha, IKKbeta and IKKgamma/NF-kappaB essential modulator (NEMO) subunits. A cell-permeable peptide that mimics the leucine zipper subdomain of IKKgamma, thus preventing its oligomerization, inhibited the constitutive NF-kappaB activation and induced apoptotic cell death in a panel of human MDS and AML cell lines (P39, MOLM13, THP1 and MV4-11). Small interfering RNA-mediated knockdown of the p65 NF-kappaB subunit or the three IKK subunits including IKKgamma/NEMO also induced apoptotic cell death in P39 cells. Cell death induced by the IKKgamma/NEMO-antagonistic peptide involved the caspase-independent loss of the mitochondrial transmembrane potential as well as signs of outer mitochondrial membrane permeabilization with the consequent release of cytochrome c, apoptosis-inducing factor and endonuclease G. Primary bone marrow CD34(+) cells from high-risk MDS and AML patients also succumbed to the IKKgamma/NEMO-antagonistic peptide, but not to a mutated control peptide. Altogether, these data indicate that malignant cells in high-risk MDS and AML cells critically depend on IKKgamma/NEMO to survive. Moreover, our data delineate a novel procedure for their therapeutic removal, through inhibition of IKKgamma/NEMO oligomerization.

The tRNA-dependent activation of arginine by arginyl-tRNA synthetase requires inter-domain communication.

The tRNA-dependent amino acid activation catalyzed by mammalian arginyl-tRNA synthetase has been characterized. A conditional lethal mutant of Chinese hamster ovary cells that exhibits reduced arginyl-tRNA synthetase activity (Arg-1), and two of its derived revertants (Arg-1R4 and Arg-1R5) were analyzed at the structural and functional levels. A single nucleotide change, resulting in a Cys to Tyr substitution at position 599 of arginyl-tRNA synthetase, is responsible for the defective phenotype of the thermosensitive and arginine hyper-auxotroph Arg-1 cell line. The two revertants have a single additional mutation resulting in a Met222 to Ile change for Arg-1R4 or a Tyr506 to Ser change for Arg-1R5. The corresponding mutant enzymes were expressed in yeast and purified. The Cys599 to Tyr mutation affects both the thermal stability of arginyl-tRNA synthetase and the kinetic parameters for arginine in the ATP-PP(i) exchange and tRNA aminoacylation reactions. This mutation is located underneath the floor of the Rossmann fold catalytic domain characteristic of class 1 aminoacyl-tRNA synthetases, near the end of a long helix belonging to the alpha-helix bundle C-terminal domain distinctive of class 1a synthetases. For the Met222 to Ile revertant, there is very little effect of the mutation on the interaction of arginyl-tRNA synthetase with either of its substrates. However, this mutation increases the thermal stability of arginyl-tRNA synthetase, thereby leading to reversion of the thermosensitive phenotype by increasing the steady-state level of the enzyme in vivo. In contrast, for the Arg-1R5 cell line, reversion of the phenotype is due to an increased catalytic efficiency of the C599Y/Y506S double mutant as compared to the initial C599Y enzyme. In light of the location of the mutations in the 3D structure of the enzyme modeled using the crystal structure of the closely related yeast arginyl-tRNA synthetase, the kinetic analysis of these mutants suggests that the obligatory tRNA-induced activation of the catalytic site of arginyl-tRNA synthetase involves interdomain signal transduction via the long helices that build the tRNA-binding domain of the enzyme and link the site of interaction of the anticodon domain of tRNA to the floor of the active site.

Genomic organization of the rat aspartyl-tRNA synthetase gene family: a single active gene and several retropseudogenes.

The genomic organization of the gene encoding rat aspartyl-tRNA synthetase (AspRS), a class II aminoacyl-tRNA synthetase (aaRS), was determined. A single active gene and several pseudogenes were isolated from a rat genomic DNA library and characterized. The active DRS1 gene encoding the rat AspRS spans approximately 60 kb and is divided into 16 exons. Exons 8-16, encoding the nt-binding domain of the synthetase, are clustered in the 3'-region of the gene, whereas exons 3, 4, and 5, encoding the anticodon-binding domain are separated by large introns (up to 15 kb) containing LINE sequences. One of the pseudogenes, psi DRS1, has a nt sequence 93% identical to that of the complete cDNA sequence of rat AspRS but several stop codons interrupt the coding sequence, thus identifying psi DRS1 to an inactive processed pseudogene. Two repetitive elements from the LINE family are inserted into psi DRS1. Calculation of nt substitution rates suggests that psi DRS1 sequences arose approximately 27 Myr ago. The other pseudogene, psi DRS2, should be more ancient. Taken together, these results clearly demonstrate that the AspRS gene family is composed of only one active gene. The availability of the gene structure of AspRS could help to clarify molecular evolution of class II aaRS.

The in vitro DNA binding properties of NDP kinase are related to its oligomeric state.

Genetic and biochemical evidences suggest that the enzymatic activity of NDP kinase is necessary but not sufficient for its biological function. While the human NDPK-B binds specifically single-strand polypyrimidines sequences, the hexameric enzyme from Dictyostelium does not. We demonstrated by electrophoretic mobility shift assay and filter binding assay that a dimeric mutant from Dictyostelium binds to an oligodesoxynucleotide while the wild-type does not. These data suggest that the differences in the DNA binding properties of several eucaryotic NDP kinases might be correlated to the differences in the stability of their hexameric structure.

Natural variation can significantly alter the sensitivity of influenza A (H5N1) viruses to oseltamivir.

Geographic spread of highly pathogenic avian H5N1 influenza viruses may give rise to an influenza pandemic. During the first months of a pandemic, control measures would rely mainly on antiviral drugs, such as the neuraminidase (NA) inhibitors oseltamivir and zanamivir. In this study, we compare the sensitivities to oseltamivir of the NAs of several highly pathogenic H5N1 viruses isolated in Asia from 1997 to 2005. The corresponding 50% inhibitory concentrations were determined using a standard in vitro NA inhibition assay. The K(m) for the substrate and the affinity for the inhibitor (K(i)) of NA were determined for a 1997 and a 2005 virus, using an NA inhibition assay on cells transiently expressing the viral enzyme. Our data show that the sensitivities of the NAs of H5N1 viruses isolated in 2004 and 2005 to oseltamivir are about 10-fold higher than those of earlier H5N1 viruses or currently circulating H1N1 viruses. Three-dimensional modeling of the N1 protein predicted that Glu248Gly and Tyr252His changes could account for increased sensitivity. Our data indicate that genetic variation in the absence of any drug-selective pressure may result in significant variations in sensitivity to anti-NA drugs. Although the clinical relevance of a 10-fold increase in the sensitivity of NA to oseltamivir needs to be investigated further, the possibility that sensitivity to anti-NA drugs could increase (or possibly decrease) significantly, even in the absence of treatment, underscores the need for continuous evaluation of the impact of genetic drift on this parameter, especially for influenza viruses with pandemic potential.

Aspartyl-tRNA synthetase from rat: in vitro functional analysis of its assembly into the multisynthetase complex.

In mammalian cells, nine aminoacyl-tRNA synthetases, including aspartyl-tRNA synthetase, are associated within a multienzyme complex. Rat aspartyl-tRNA synthetase has a N-terminal polypeptide extension of about 40 amino acid residues which can be removed without impairing its catalytic activity. Earlier, in vivo studies showed that enzymes deprive of this N-terminal segment behave in vivo as free entities. We designed an experimental in vitro approach, based on the exchange of the complexed endogenous enzyme by free recombinant species, to assess the contribution of that domain in the association of aspartyl-tRNA synthetase to the complex. A phosphorylation site was introduced at the N-terminus of rat aspartyl-tRNA synthetase. The enzyme served as a reporter protein to evaluate the dissociation constants of native and N-terminal-truncated species towards the complex. Our data show that a moderate but significant drop in affinity is inferred by the removal of the N-terminal domain. The results suggest that this domain binds to another component of the complex, but might primarily serve a targeting function absolutely required in vivo for the assembly within the multienzyme structure.

The binding mode of human nucleoside diphosphate kinase B to single-strand DNA.

In this paper, we studied the interaction of the human isoform B of nucleoside diphosphate kinase (NDP kinase B) with the nuclease hypersensitive element (NHE) present in the promoter element of the c-myc oncogene. The DNA-binding properties of NDP kinase B and other NDP kinases are compared and the nucleotide requirement for binding are discussed. Using quantitative methods, we identified the DNA-binding sites on the protein and we proposed a structural model for a complex of one hexameric NDP kinase B with an oligonucleotide.

Single strand DNA specificity analysis of human nucleoside diphosphate kinase B.

Nucleoside diphosphate kinases (NDP kinases) form a family of oligomeric enzymes present in all organisms. Eukaryotic NDP kinases are hexamers composed of identical subunits (approximately 17 kDa). A distinctive property of human NDPK-B encoded by the gene nm23-H2 is its ability to stimulate the gene transcription. This property is independent of its catalytic activity and is possibly related to the role of this protein in cellular events including differentiation and tumor metastasis. In this paper, we report the first characterization of human NDPK-B.DNA complex formation using a filter-binding assay and fluorescence spectroscopy. We analyzed the binding of several oligonucleotides mimicking the promoter region of the c-myc oncogene including variants in sequence, structure, and length of both strands. We show that NDPK-B binds to single-stranded oligonucleotides in a nonsequence specific manner, but that it exhibits a poor binding activity to double-stranded oligonucleotides. This indicates that the specificity of recognition to DNA is a function of the structural conformation of DNA rather than of its specific sequence. Moreover, competition experiments performed with all nucleotides provide evidence for the contribution of the six active sites in the DNA.protein complex formation. We propose a mechanism through which human NDPK-B could stimulate transcription of c-myc or possibly other genes involved in cellular differentiation.

Expression of rat aspartyl-tRNA synthetase in Saccharomyces cerevisiae. Role of the NH2-terminal polypeptide extension on enzyme activity and stability.

Cytoplasmic aspartyl-tRNA synthetase from mammals is one of the components of a multienzyme complex comprising nine synthetase activities. The presence of an amino-terminal extension composed of about 40 residues is a characteristic of the eukaryotic enzyme. We report here the expression in the yeast Saccharomyces cerevisiae of a native form of rat aspartyl-tRNA synthetase and of two truncated derivatives lacking 20 or 36 amino acid residues from their amino-terminal polypeptide extension. The three recombinant enzyme species were purified to homogeneity. They behave as alpha2 dimers and display catalytic parameters in the tRNA aminoacylation reaction identical to those determined for the native, complex-associated form of aspartyl-tRNA synthetase isolated from rat liver. Because the dimer dissociation constant of rat AspRS is much higher than that of its bacterial and yeast counterparts, we could establish a direct correlation between dissociation of the dimer and inactivation of the enzyme. Our results clearly show that the monomer is devoid of amino acid activation and tRNA aminoacylation activities, indicating that dimerization is essential to confer an active conformation on the catalytic site. The two NH2-terminal truncated derivatives were fully active, but proved to be more unstable than the recombinant native enzyme, suggesting that the polypeptide extension fulfills structural rather than catalytic requirements.

Switching the amino acid specificity of an aminoacyl-tRNA synthetase.

The accuracy of protein synthesis essentially rests on aminoacyl-tRNA synthetases that ensure the correct attachment of an amino acid to the cognate tRNA molecule. The selection of the amino acid substrate involves a recognition stage generally followed by a proofreading reaction. Therefore, to change the amino acid specificity of a synthetase in the aminoacylation reaction, it is necessary to alleviate the molecular barriers which contribute its editing function. In an attempt to accommodate a noncognate amino acid into the active site of a synthetase, we chose a pair of closely related enzymes. The current hypothesis designates glutaminyl-tRNA synthetase (GlnRS) as a late component of the protein synthesis machinery, emerging in the eukaryotic lineage by duplication of the gene for glutamyl-tRNA synthetase (GluRS). By introducing GluRS-specific features into the Rossmann dinucleotide-binding domain of human GlnRS, we constructed a mutant GlnRS which preferentially aminoacylates tRNA with glutamate instead of glutamine. Our data suggest that not only the transition state for aminoacyl-AMP formation but also the proofreading site of GlnRS are affected by that mutation.

The p43 component of the mammalian multi-synthetase complex is likely to be the precursor of the endothelial monocyte-activating polypeptide II cytokine.

p43 is one of the three auxiliary components invariably associated with nine aminoacyl-tRNA synthetases as a multienzyme complex ubiquitous to all eukaryotic cells from flies to humans. The cDNA encoding the hamster protein was isolated by using mixed oligonucleotides deduced from peptide sequences. The 359-amino acid protein is the hamster homologue of the recently reported murine and human EMAP II cytokine implicated in a variety of inflammatory disorders. The sequence of several proEMAP II proteins suggests that the p43 component of the complex is the precursor of the active mature cytokine after cleavage at a conserved Asp residue. The COOH-terminal moiety of p43 is also homologous to polypeptide domains found in bacterial methionyl- or phenylalanyl-tRNA synthetases and in the yeast Arc1p/G4p1 protein that associates with yeast methionyl-tRNA synthetase. Our results implicate the COOH-terminal moiety of p43 as a RNA binding domain. In the native state, as a component of the multisynthetase complex, p43 may be required for tRNA channeling and, after proteolytic processing occurring in tumor cells, would acquire inflammatory properties possibly related to apoptosis. The release of a truncated p43 from the complex could be involved in mediation of the signaling of tumor cells and stimulation of an acute inflammatory response.

Polyanion-induced alpha-helical structure of a synthetic 23-residue peptide representing the lysine-rich segment of the N-terminal extension of yeast cytoplasmic aspartyl-tRNA synthetase.

Conformational studies were performed on the synthetic tricosapeptide N-acetyl-SKKALKKLQKEQEKQRKKEERAL-amide, representing the highly basic segment (residues 30-52) of the N-terminal extension of yeast cytoplasmic aspartyl-tRNA synthetase. Circular dichroism experiments show that, in aqueous solution at neutral pH, the peptide adopts a random conformation. The effects of pH, temperature, addition of trifluoroethanol (TFE), and titration with polyanions on the conformation of the peptide were studied. In TFE or in the presence of an equimolar concentration of (phosphate)18, the peptide adopts a 100% alpha-helical conformation. A partially alpha-helical conformation is induced by (phosphate)4 or d(pT)8 (respectively 40% and 35% helical content). Raising the pH in aqueous solution promotes 75% alpha-helicity, with a transition pK of 9.9 reflecting deprotonation of lysine residues. On the basis of these results, nuclear magnetic resonance studies were carried out in TFE as well as in aqueous solution in the presence of (phosphate)18, to determine the structure of the molecule. Complete 1H resonance assignments were obtained by conventional two-dimensional NMR techniques. A total of 138 interproton constraints derived from NOESY experiments were used to calculate the three-dimensional structure by a two-stage distance geometry/simulated annealing procedure. The two deduced structures were highly similar and show that nine cationic residues are segregated on one face of a helical structure, providing an ideal polycationic interface for binding to polyanionic surfaces.

Peptidic determinants and structural model of human NDP kinase B (Nm23-H2) bound to single-stranded DNA.

Isoform B of human NDP kinase (NDPK-B) was previously identified as a transcription factor stimulating in vitro and ex vivo the transcription of the c-myc oncogene, which involves this enzyme in carcinogenesis. We have studied the enzymatic properties of NDPK-B in the presence of several single-stranded oligonucleotides. We show that the oligonucleotides are competitive inhibitors of the catalytic activity, indicating that the active site acts as a binding template for the anchorage of the oligonucleotide. Furthermore, the presence of a guanine at the 3'-end of several different aptamers increases its affinity 10-fold. To define the surface of the protein contacting the DNA within the nucleoprotein complex, we used single nanosecond laser pulses as the cross-linking reagent and MALDI-TOF mass spectrometry to identify cross-linked peptides purified from proteolytic digests of the cross-linked complex. Using 11-mer and 30-mer single-stranded oligonucleotides, the same three different nucleopeptides were identified after irradiation of the complexes, indicating a common binding mode for these two aptamers. Taken together, these results allowed us to propose a structural model of NDPK-B bound to single-stranded DNA.

Functional replacement of hamster lysyl-tRNA synthetase by the yeast enzyme requires cognate amino acid sequences for proper tRNA recognition.

We cloned the cDNA encoding a 597-aa hamster lysyl-tRNA synthetase. This enzyme is a close homologue of the 591-aa Saccharomyces cerevisiae enzyme, with the noticeable exception of their 60-aa N-terminal regions, which differ significantly. Several particular features of this polypeptide fragment from the hamster lysyl-tRNA synthetase suggest that it is implicated in the assembly of that enzyme within the multisynthetase complex. However, we show that this protein domain is dispensable in vivo to sustain growth of CHO cells. The cross-species complementation was investigated in the lysine system. The mammalian enzyme functionally replaces a null-allele of the yeast KRS1 gene. Conversely, the yeast enzyme cannot rescue Lys-101 cells, a CHO cell line with a temperature-sensitive lysyl-tRNA synthetase. The yeast and mammalian enzymes, overexpressed in yeast, were purified to homogeneity. The hamster lysyl-tRNA synthetase efficiently aminoacylates both mammalian and yeast tRNA(Lys), whereas the yeast enzyme aminoacylates mammalian tRNA(Lys) with a catalytic efficiency 20-fold lower, as compared to its cognate tRNA. The 152-aa C-terminus extremity of the hamster enzyme provides the yeast enzyme with the capacity to complement Lys-101 cells. This hybrid protein is fairly stable and aminoacylates both yeast and mammalian tRNA(Lys) with similar catalytic efficiencies. Because this C-terminal polypeptide fragment is likely to make contacts with the acceptor stem of tRNA(Lys), we conclude that it should carry the protein determinants conferring specific recognition of the cognate tRNA acceptor stem and therefore contributes an essential role in the operational RNA code for amino acids.

Coupling between catalysis and oligomeric structure in nucleoside diphosphate kinase.

A dimeric Dictyostelium nucleoside diphosphate kinase has been stabilized by the double mutation P100S-N150stop which targets residues involved in the trimer interface (Karlsson, A., Mesnildrey, S., Xu, Y., Moréra, S., Janin, J., and Veron, M. (1996) J. Biol. Chem. 271, 19928-19934). The reassociation of this dimeric form into a hexamer similar to the wild-type enzyme is induced by the presence of a nucleotide substrate. Equilibrium sedimentation and gel filtration experiments, as well as enzymatic activity measurements, show that reactivation of the enzyme closely parallels its reassociation. A phosphorylatable intermediate with low activity participates in the association pathway while the dimeric form is shown totally devoid of enzymatic activity. Our results support the hypothesis that different oligomeric species of nucleoside diphosphate kinase are involved in different cellular processes where the enzymatic activity is not required.

The mechanism of phosphorylation of anti-HIV D4T by nucleoside diphosphate kinase.

The last step in the intracellular activation of antiviral nucleoside analogs is the addition of the third phosphate by nucleoside diphosphate (NDP) kinase resulting in the synthesis of the viral reverse transcriptase substrates. We have previously shown that dideoxynucleotide analogs and 3'-deoxy-3'-azidothymidine (AZT) as di- or triphosphate are poor substrates for NDP kinase. By use of protein fluorescence, we monitor the phosphotransfer between the enzyme and the nucleotide analog. Here, we have studied the reactivity of D4T (2',3'-dideoxy-2',3'-didehydrothymidine; stavudine) as di- (DP) or triphosphate (TP) at the pre-steady state. The catalytic efficiency of D4T-DP or -TP is increased by a factor of 10 compared with AZT-DP or -TP, respectively. We use an inactive mutant of NDP kinase to monitor the binding of a TP derivative, and show that the affinity for D4T-TP is in the same range as for the natural substrate deoxythymidine triphosphate, but is 30 times higher than for AZT-TP. Our results indicate that D4T should be efficiently phosphorylated after intracellular maturation of a prodrug into D4T-monophosphate.

Structural and catalytic properties and homology modelling of the human nucleoside diphosphate kinase C, product of the DRnm23 gene.

The human DRnm23 gene was identified by differential screening of a cDNA library obtained from chronic myeloid leukaemia-blast crisis primary cells. The over-expression of this gene inhibits differentiation and induces the apoptosis of myeloid precursor cell lines. We overproduced in bacteria a truncated form of the encoded protein lacking the first 17 N-terminal amino acids. This truncated protein was called nucleoside diphosphate (NDP) kinase CDelta. NDP kinase CDelta had similar kinetic properties to the major human NDP kinases A and B, but was significantly more stable to denaturation by urea and heat. Analysis of denaturation by urea, using size exclusion chromatography, indicated unfolding without the dissociation of subunits, whereas renaturation occurred via a folded monomer. The stability of the protein depended primarily on subunit interactions. Homology modelling of the structure of NDP kinase CDelta, based on the crystal structure of NDP kinase B, indicated that NDP kinase CDelta had several additional stabilizing interactions. The overall structure of the two enzymes appears to be identical because NDP kinase CDelta readily formed mixed hexamers with NDP kinase A. It is possible that mixed hexamers can be observed in vivo.