Discovery of small molecule CD40-TRAF6 inhibitors.

The CD154-CD40 receptor complex plays a pivotal role in several inflammatory pathways. Attempts to inhibit the formation of this complex have resulted in systemic side effects. Downstream inhibition of the CD40 signaling pathway therefore seems a better way to ameliorate inflammatory disease. To relay a signal, the CD40 receptor recruits adapter proteins called tumor necrosis factor receptor-associated factors (TRAFs). CD40-TRAF6 interactions are known to play an essential role in several inflammatory diseases. We used in silico, in vitro, and in vivo experiments to identify and characterize compounds that block CD40-TRAF6 interactions. We present in detail our drug docking and optimization pipeline and show how we used it to find lead compounds that reduce inflammation in models of peritonitis and sepsis. These compounds appear to be good leads for drug development, given the observed absence of side effects and their demonstrated efficacy for peritonitis and sepsis in mouse models.

X-linked Charcot-Marie-Tooth disease, Arts syndrome, and prelingual non-syndromic deafness form a disease continuum: evidence from a family with a novel PRPS1 mutation.

BACKGROUND: X-linked Charcot-Marie-Tooth disease type 5 (CMTX5), Arts syndrome, and non-syndromic sensorineural deafness (DFN2) are allelic syndromes, caused by reduced activity of phosphoribosylpyrophosphate synthetase 1 (PRS-I) due to loss-of-function mutations in PRPS1. As only few families have been described, knowledge about the relation between these syndromes, the phenotypic spectrum in patients and female carriers, and the relation to underlying PRS-I activity is limited. METHODS: We investigated a family with a novel PRPS1 mutation (c.830A > C, p.Gln277Pro) by extensive phenotyping, MRI, and genetic and enzymatic tests. RESULTS: The male index subject presented with an overlap of CMTX5 and Arts syndrome features, whereas his sister presented with prelingual DFN2. Both showed mild parietal and cerebellar atrophy on MRI. Enzymatically, PRS-I activity was undetectable in the index subject, reduced in his less affected sister, and normal in his unaffected mother. CONCLUSIONS: Our findings demonstrate that CMTX5, Arts syndrome and DFN2 are phenotypic clusters on an intrafamilial continuum, including overlapping phenotypes even within individuals. The respective phenotypic presentation seems to be determined by the exact PRPS1 mutation and the residual enzyme activity, the latter being largely influenced by the degree of skewed X-inactivation. Finally, our findings show that brain atrophy might be more common in PRPS1-disorders than previously thought.

Blocking CD40-TRAF6 signaling is a therapeutic target in obesity-associated insulin resistance.

The immune system plays an instrumental role in obesity and insulin resistance. Here, we unravel the role of the costimulatory molecule CD40 and its signaling intermediates, TNF receptor-associated factors (TRAFs), in diet-induced obesity (DIO). Although not exhibiting increased weight gain, male CD40(-/-) mice in DIO displayed worsened insulin resistance, compared with wild-type mice. This worsening was associated with excessive inflammation of adipose tissue (AT), characterized by increased accumulation of CD8(+) T cells and M1 macrophages, and enhanced hepatosteatosis. Mice with deficient CD40-TRAF2/3/5 signaling in MHCII(+) cells exhibited a similar phenotype in DIO as CD40(-/-) mice. In contrast, mice with deficient CD40-TRAF6 signaling in MHCII(+) cells displayed no insulin resistance and showed a reduction in both AT inflammation and hepatosteatosis in DIO. To prove the therapeutic potential of inhibition of CD40-TRAF6 in obesity, DIO mice were treated with a small-molecule inhibitor that we designed to specifically block CD40-TRAF6 interactions; this compound improved insulin sensitivity, reduced AT inflammation, and decreased hepatosteatosis. Our study reveals that the CD40-TRAF2/3/5 signaling pathway in MHCII(+) cells protects against AT inflammation and metabolic complications associated with obesity whereas CD40-TRAF6 interactions in MHCII(+) cells aggravate these complications. Inhibition of CD40-TRAF6 signaling by our compound may provide a therapeutic option in obesity-associated insulin resistance.

Application of a cell-based protease assay for testing inhibitors of picornavirus 3C proteases.

Proteolytical cleavage of the picornaviral polyprotein is essential for viral replication. Therefore, viral proteases are attractive targets for anti-viral therapy. Most assays available for testing proteolytical activity of proteases are performed in vitro, using heterologously expressed proteases and peptide substrates. To deal with the disadvantages associated with in vitro assays, we modified a cell-based protease assay for picornavirus proteases. The assay is based on the induction of expression of a firefly luciferase reporter by a chimeric transcription factor in which the viral protease and cleavage sites are inserted between the GAL4 binding domain and the VP16 activation domain. Firefly luciferase expression is dependent on cleavage of the transcription factor by the viral protease. This biosafe assay enables testing the effect of compounds on protease activity in cells while circumventing the need for infection. We designed the assay for 3C proteases (3C(pro)) of various enteroviruses as well as of viruses of several other picornavirus genera, and show that the assay is amenable for use in a high-throughput setting. Furthermore, we show that the spectrum of activity of 3C(pro) inhibitor AG7088 (rupintrivir) not only encompasses enterovirus 3C(pro) but also 3C(pro) of foot-and-mouth disease virus (FMDV), an aphthovirus. In contrary, AG7404 (compound 1), an analogue of AG7088, had no effect on FMDV 3C(pro) activity, for which we provide a structural explanation.

A 3-base pair deletion, c.9711_9713del, in DMD results in intellectual disability without muscular dystrophy.

We have identified a deletion of 3 base pairs in the dystrophin gene (DMD), c.9711_9713del, in a family with nonspecific X-linked intellectual disability (ID) by sequencing of the exons of 86 known X-linked ID genes. This in-frame deletion results in the deletion of a single-amino-acid residue, Leu3238, in the brain-specific isoform Dp71 of dystrophin. Linkage analysis supported causality as the mutation was present in the 7.6 cM linkage interval on Xp22.11-Xp21.1 with a maximum positive LOD score of 2.41 (MRX85 locus). Molecular modeling predicts that the p.(Leu3238del) deletion results in the destabilization of the C-terminal domain of dystrophin and hence reduces the ability to interact with ß-dystroglycan. Correspondingly, Dp71 protein levels in lymphoblastoid cells from the index patient are 6.7-fold lower than those in control cell lines (P=0.08). Subsequent determination of the creatine kinase levels in blood of the index patient showed a mild but significant elevation in serum creatine kinase, which is in line with impaired dystrophin function. In conclusion, we have identified the first DMD mutation in Dp71 that results in ID without muscular dystrophy.

A functional variant in the CFI gene confers a high risk of age-related macular degeneration.

Up to half of the heritability of age-related macular degeneration (AMD) is explained by common variants. Here, we report the identification of a rare, highly penetrant missense mutation in CFI encoding a p.Gly119Arg substitution that confers high risk of AMD (P = 3.79 × 10⁻6; odds ratio (OR) = 22.20, 95% confidence interval (CI) = 2.98-164.49). Plasma and sera from cases carrying the p.Gly119Arg substitution mediated the degradation of C3b, both in the fluid phase and on the cell surface, to a lesser extent than those from controls. Recombinant protein studies showed that the Gly119Arg mutant protein is both expressed and secreted at lower levels than wild-type protein. Consistent with these findings, human CFI mRNA encoding Arg119 had reduced activity compared to wild-type mRNA encoding Gly119 in regulating vessel thickness and branching in the zebrafish retina. Taken together, these findings demonstrate that rare, highly penetrant mutations contribute to the genetic burden of AMD.

A complex V ATP5A1 defect causes fatal neonatal mitochondrial encephalopathy.

Whole exome sequencing is a powerful tool to detect novel pathogenic mutations in patients with suspected mitochondrial disease. However, the interpretation of novel genetic variants is not always straightforward. Here, we present two siblings with a severe neonatal encephalopathy caused by complex V deficiency. The aim of this study was to uncover the underlying genetic defect using the combination of enzymatic testing and whole exome sequence analysis, and to provide evidence for causality by functional follow-up. Measurement of the oxygen consumption rate and enzyme analysis in fibroblasts were performed. Immunoblotting techniques were applied to study complex V assembly. The coding regions of the genome were analysed. Three-dimensional modelling was applied. Exome sequencing of the two siblings with complex V deficiency revealed a heterozygous mutation in the ATP5A1 gene, coding for complex V subunit α. The father carried the variant heterozygously. At the messenger RNA level, only the mutated allele was expressed in the patients, whereas the father expressed both the wild-type and the mutant allele. Gene expression data indicate that the maternal allele is not expressed, which is supported by the observation that the ATP5A1 expression levels in the patients and their mother are reduced to ∼50%. Complementation with wild-type ATP5A1 restored complex V in the patient fibroblasts, confirming pathogenicity of the defect. At the protein level, the mutation results in a disturbed interaction of the α-subunit with the ß-subunit of complex V, which interferes with the stability of the complex. This study demonstrates the important value of functional studies in the diagnostic work-up of mitochondrial patients, in order to guide genetic variant prioritization, and to validate gene defects.

CING: an integrated residue-based structure validation program suite.

We present a suite of programs, named CING for Common Interface for NMR Structure Generation that provides for a residue-based, integrated validation of the structural NMR ensemble in conjunction with the experimental restraints and other input data. External validation programs and new internal validation routines compare the NMR-derived models with empirical data, measured chemical shifts, distance- and dihedral restraints and the results are visualized in a dynamic Web 2.0 report. A red-orange-green score is used for residues and restraints to direct the user to those critiques that warrant further investigation. Overall green scores below ~20 % accompanied by red scores over ~50 % are strongly indicative of poorly modelled structures. The publically accessible, secure iCing webserver ( https://nmr.le.ac.uk ) allows individual users to upload the NMR data and run a CING validation analysis.

Enzyme-specific activation versus leaving group ability.

Enzyme-specific activation and the substrate mimetics strategy are effective ways to circumvent the limited substrate recognition often encountered in protease-catalyzed peptide synthesis. A key structural element in both approaches is the guanidinophenyl (OGp) ester, which enables important interactions for affinity and recognition by the enzyme--at least, this is usually the explanation given for its successful application. In this study we show that leaving group ability is of equal or even greater importance. To this end we used both experimental and computational methods: 1) synthesis of close analogues of OGp, and their evaluation in a dipeptide synthesis assay with trypsin, 2) molecular docking studies to provide insights into the binding mode, and 3) ab initio calculations to evaluate their electronic properties.

Evolution and diversification of the organellar release factor family.

Translation termination is accomplished by proteins of the Class I release factor family (RF) that recognize stop codons and catalyze the ribosomal release of the newly synthesized peptide. Bacteria have two canonical RFs: RF1 recognizes UAA and UAG, RF2 recognizes UAA and UGA. Despite that these two release factor proteins are sufficient for de facto translation termination, the eukaryotic organellar RF protein family, which has evolved from bacterial release factors, has expanded considerably, comprising multiple subfamilies, most of which have not been functionally characterized or formally classified. Here, we integrate multiple sources of information to analyze the remarkable differentiation of the RF family among organelles. We document the origin, phylogenetic distribution and sequence structure features of the mitochondrial and plastidial release factors: mtRF1a, mtRF1, mtRF2a, mtRF2b, mtRF2c, ICT1, C12orf65, pRF1, and pRF2, and review published relevant experimental data. The canonical release factors (mtRF1a, mtRF2a, pRF1, and pRF2) and ICT1 are derived from bacterial ancestors, whereas the others have resulted from gene duplications of another release factor. These new RF family members have all lost one or more specific motifs relevant for bona fide release factor function but are mostly targeted to the same organelle as their ancestor. We also characterize the subset of canonical release factor proteins that bear nonclassical PxT/SPF tripeptide motifs and provide a molecular-model-based rationale for their retained ability to recognize stop codons. Finally, we analyze the coevolution of canonical RFs with the organellar genetic code. Although the RF presence in an organelle and its stop codon usage tend to coevolve, we find three taxa that encode an RF2 without using UGA stop codons, and one reverse scenario, where mamiellales green algae use UGA stop codons in their mitochondria without having a mitochondrial type RF2. For the latter, we put forward a "stop-codon reinvention" hypothesis that involves the retargeting of the plastid release factor to the mitochondrion.

In silico identification of potential cholestasis-inducing agents via modeling of Na(+)-dependent taurocholate cotransporting polypeptide substrate specificity.

Na(+)-dependent taurocholate cotransporting polypeptide (NTCP, SLC10A1) is the main transporter facilitating the hepatic uptake of bile acids from the circulation. Consequently, the interaction of xenobiotics, including therapeutic drugs, with the bile acid binding pocket of NTCP could lead to impairment of hepatic bile acid uptake. We pursued a 3D-pharmacophore approach to model the NTCP substrate and inhibitor specificity and investigated whether it is possible to identify compounds with intrinsic NTCP inhibitory properties. Based on known endogenous NTCP substrates, a 3D-pharmacophore model was built, which was subsequently used to screen two virtual libraries together containing the structures of 10 million compounds. Studies with Chinese hamster ovary cells overexpressing human NTCP, human hepatocytes, ex vivo perfused rat livers, and bile duct-cannulated rats were conducted to validate the activity of the virtual screening hits. Modeling yielded a 3D-pharmacophore, consisting of two hydrogen bond acceptors and three hydrophobic features. Six out of 10 structurally diverse compounds selected in the first virtual screening procedure significantly inhibited taurocholate uptake in the NTCP overexpressing cells. For the most potent inhibitor identified, an anthraquinone derivative, this finding was confirmed in human hepatocytes and perfused rat livers. Subsequent structure and activity relationship studies with analogs of this derivative indicated that an appropriate distance between hydrogen bond acceptor features and presence of one or two negative charges appear critical for a successful NTCP interaction. In conclusion, pharmacophore modeling was successfully used to identify compounds that inhibit NTCP. Our approach represents an important first step toward the in silico flagging of potential cholestasis-inducing molecules.

Papain-specific activating esters in aqueous dipeptide synthesis.

Enzymatic peptide synthesis has the potential to be a viable alternative for chemical peptide synthesis. Because of the increasing commercial interest in peptides, new and improved enzymatic synthesis methods are desirable. In recently developed enzymatic strategies such as substrate mimetic approaches and enzyme-specific activation, use of the guanidinophenyl ester (OGp) group has been shown to suffer from some drawbacks. OGp esters are sensitive to spontaneous chemical hydrolysis and the group is expensive to synthesize and therefore not suitable for large-scale applications. On the basis of earlier computational studies, we hypothesized that OGp might be replaceable by simpler ester groups to make the enzyme-specific activation approach to peptide bond formation more accessible. To this end, a set of potential activating esters (Z-Gly-Act) was designed, synthesized, and evaluated. Both the benzyl (OBn) and the dimethylaminophenyl (ODmap) esters gave promising results. For these esters, the scope of a model dipeptide synthesis reaction under aqueous conditions was investigated by varying the amino acid donor. The results were compared with those obtained from a previous study of Z-X(AA) -OGp esters. Computational docking analysis of the set of esters was performed in order to provide insight into the differences in the reactivities of all the potential activating esters. Finally, selected ODmap- and OBn-activated amino acids were applied in the synthesis of two biologically active dipeptides on preparative scales.

Structure based hypothesis of a mitochondrial ribosome rescue mechanism.

BACKGROUND: mtRF1 is a vertebrate mitochondrial protein with an unknown function that arose from a duplication of the mitochondrial release factor mtRF1a. To elucidate the function of mtRF1, we determined the positions that are conserved among mtRF1 sequences but that are different in their mtRF1a paralogs. We subsequently modeled the 3D structure of mtRF1a and mtRF1 bound to the ribosome, highlighting the structural implications of these differences to derive a hypothesis for the function of mtRF1. RESULTS: Our model predicts, in agreement with the experimental data, that the 3D structure of mtRF1a allows it to recognize the stop codons UAA and UAG in the A-site of the ribosome. In contrast, we show that mtRF1 likely can only bind the ribosome when the A-site is devoid of mRNA. Furthermore, while mtRF1a will adopt its catalytic conformation, in which it functions as a peptidyl-tRNA hydrolase in the ribosome, only upon binding of a stop codon in the A-site, mtRF1 appears specifically adapted to assume this extended, peptidyl-tRNA hydrolyzing conformation in the absence of mRNA in the A-site. CONCLUSIONS: We predict that mtRF1 specifically recognizes ribosomes with an empty A-site and is able to function as a peptidyl-tRNA hydrolase in those situations. Stalled ribosomes with empty A-sites that still contain a tRNA bound to a peptide chain can result from the translation of truncated, stop-codon less mRNAs. We hypothesize that mtRF1 recycles such stalled ribosomes, performing a function that is analogous to that of tmRNA in bacteria.

Flexible protein-ligand docking using the Fleksy protocol.

Considering protein plasticity is important in accurately predicting the three-dimensional geometry of protein-ligand complexes. Here, we present the first public release of our flexible docking tool Fleksy, which is able to consider both ligand and protein flexibility in the docking process. We describe the workflow and different features of the software and present its performance on two cross-docking benchmark datasets.

Phosphoribosylpyrophosphate synthetase superactivity and recurrent infections is caused by a p.Val142Leu mutation in PRS-I.

We identified a novel missense mutation, c.424G>C (p.Val142Leu) in PRPS1 in a patient with uric acid overproduction without gout but with developmental delay, hypotonia, hearing loss, and recurrent respiratory infections. The uric acid overproduction accompanying this combination of symptoms suggests that the patient presented with phosphoribosylpyrophosphate (PRPP) synthetase superactivity, but recurrent infections have not been associated with superactivity until now. However, recurrent infections are a prominent feature of patients with Arts syndrome, which is caused by PRPS1 loss-of-function mutations, indicating that the patient reported here has an intermediate phenotype. Molecular modeling predicts that the p.Val142Leu change affects both allosteric sites that are involved in inhibition of PRPS1 and the ATP-binding site, which suggests that this substitution can result both in a gain-of-function and loss-of-function of PRPP synthetase. This finding is in line with the normal PRPP synthetase activity in fibroblasts and the absence of activity in erythrocytes of the present patient. We postulate that the overall effect of the p.Val142Leu change on protein activity is determined by the cell type, being a gain-of-function in proliferating cells and a loss-of-function in postmitotic cells. Our results show that missense mutations in PRPS1 can cause a continuous spectrum of features ranging from progressive non-syndromic postlingual hearing impairment to uric acid overproduction, neuropathy, and recurrent infections depending on the functional sites that are affected.

Gaussian interaction profile kernels for predicting drug-target interaction.

MOTIVATION: The in silico prediction of potential interactions between drugs and target proteins is of core importance for the identification of new drugs or novel targets for existing drugs. However, only a tiny portion of all drug-target pairs in current datasets are experimentally validated interactions. This motivates the need for developing computational methods that predict true interaction pairs with high accuracy. RESULTS: We show that a simple machine learning method that uses the drug-target network as the only source of information is capable of predicting true interaction pairs with high accuracy. Specifically, we introduce interaction profiles of drugs (and of targets) in a network, which are binary vectors specifying the presence or absence of interaction with every target (drug) in that network. We define a kernel on these profiles, called the Gaussian Interaction Profile (GIP) kernel, and use a simple classifier, (kernel) Regularized Least Squares (RLS), for prediction drug-target interactions. We test comparatively the effectiveness of RLS with the GIP kernel on four drug-target interaction networks used in previous studies. The proposed algorithm achieves area under the precision-recall curve (AUPR) up to 92.7, significantly improving over results of state-of-the-art methods. Moreover, we show that using also kernels based on chemical and genomic information further increases accuracy, with a neat improvement on small datasets. These results substantiate the relevance of the network topology (in the form of interaction profiles) as source of information for predicting drug-target interactions. AVAILABILITY: Software and Supplementary Material are available at http://cs.ru.nl/~tvanlaarhoven/drugtarget2011/. CONTACT: tvanlaarhoven@cs.ru.nl; elenam@cs.ru.nl. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Papain-catalyzed peptide bond formation: enzyme-specific activation with guanidinophenyl esters.

The substrate mimetics approach is a versatile method for small-scale enzymatic peptide-bond synthesis in aqueous systems. The protease-recognized amino acid side chain is incorporated in an ester leaving group, the substrate mimetic. This shift of the specific moiety enables the acceptance of amino acids and peptide sequences that are normally not recognized by the enzyme. The guanidinophenyl group (OGp), a known substrate mimetic for the serine proteases trypsin and chymotrypsin, has now been applied for the first time in combination with papain, a cheap and commercially available cysteine protease. To provide insight in the binding mode of various Z-X(AA)-OGp esters, computational docking studies were performed. The results strongly point at enzyme-specific activation of the OGp esters in papain through a novel mode of action, rather than their functioning as mimetics. Furthermore, the scope of a model dipeptide synthesis was investigated with respect to both the amino acid donor and the nucleophile. Molecular dynamics simulations were carried out to prioritize 22 natural and unnatural amino acid donors for synthesis. Experimental results correlate well with the predicted ranking and show that nearly all amino acids are accepted by papain.

Snooker: a structure-based pharmacophore generation tool applied to class A GPCRs.

G-protein coupled receptors (GPCRs) are important drug targets for various diseases and of major interest to pharmaceutical companies. The function of individual members of this protein family can be modulated by the binding of small molecules at the extracellular side of the structurally conserved transmembrane (TM) domain. Here, we present Snooker, a structure-based approach to generate pharmacophore hypotheses for compounds binding to this extracellular side of the TM domain. Snooker does not require knowledge of ligands, is therefore suitable for apo-proteins, and can be applied to all receptors of the GPCR protein family. The method comprises the construction of a homology model of the TM domains and prioritization of residues on the probability of being ligand binding. Subsequently, protein properties are converted to ligand space, and pharmacophore features are generated at positions where protein ligand interactions are likely. Using this semiautomated knowledge-driven bioinformatics approach we have created pharmacophore hypotheses for 15 different GPCRs from several different subfamilies. For the beta-2-adrenergic receptor we show that ligand poses predicted by Snooker pharmacophore hypotheses reproduce literature supported binding modes for ∼75% of compounds fulfilling pharmacophore constraints. All 15 pharmacophore hypotheses represent interactions with essential residues for ligand binding as observed in mutagenesis experiments and compound selections based on these hypotheses are shown to be target specific. For 8 out of 15 targets enrichment factors above 10-fold are observed in the top 0.5% ranked compounds in a virtual screen. Additionally, prospectively predicted ligand binding poses in the human dopamine D3 receptor based on Snooker pharmacophores were ranked among the best models in the community wide GPCR dock 2010.

NDUFB7 and NDUFA8 are located at the intermembrane surface of complex I.

Complex I (NADH:ubiquinone oxidoreductase) is the first and largest protein complex of the oxidative phosphorylation. Crystal structures have elucidated the positions of most subunits of bacterial evolutionary origin in the complex, but the positions of the eukaryotic subunits are unknown. Based on the analysis of sequence conservation we propose intra-molecular disulfide bridges and the inter-membrane space localization of three Cx(9)C-containing subunits in human: NDUFS5, NDUFB7 and NDUFA8. We experimentally confirm the localization of the latter two, while our data are consistent with disulfide bridges in NDUFA8. We propose these subunits stabilize the membrane domain of complex I.

PRPS1 mutations: four distinct syndromes and potential treatment.

Phosphoribosylpyrophosphate synthetases (PRSs) catalyze the first step of nucleotide synthesis. Nucleotides are central to cell function, being the building blocks of nucleic acids and serving as cofactors in cellular signaling and metabolism. With this in mind, it is remarkable that mutations in phosphoribosylpyrophosphate synthetase 1 (PRPS1), which is the most ubiquitously expressed gene of the three PRS genes, are compatible with life. Mutations described thus far in PRPS1 are all missense mutations that result in PRS-I superactivity or in variable levels of decreased activity, resulting in X-linked Charcot-Marie-Tooth disease-5 (CMTX5), Arts syndrome, and X-linked nonsyndromic sensorineural deafness (DFN2). Patients with PRS-I superactivity primarily present with uric acid overproduction, mental retardation, ataxia, hypotonia, and hearing impairment. Postlingual progressive hearing loss is found as an isolated feature in DFN2 patients. Patients with CMTX5 and Arts syndrome have peripheral neuropathy, including hearing impairment and optic atrophy. However, patients with Arts syndrome are more severely affected because they also have central neuropathy and an impaired immune system. The neurological phenotype in all four PRPS1-related disorders seems to result primarily from reduced levels of GTP and possibly other purine nucleotides including ATP, suggesting that these disorders belong to the same disease spectrum. Preliminary results of S-adenosylmethionine (SAM) supplementation in two Arts syndrome patients show improvement of their condition, indicating that SAM supplementation in the diet could alleviate some of the symptoms of patients with PRPS1 spectrum diseases by replenishing purine nucleotides (J.C., unpublished data).