PTMD 2.0: an updated database of disease-associated post-translational modifications

Abstract

Various post-translational modifications (PTMs) participate in nearly all aspects of biological processes by regulating protein functions, and aberrant states of PTMs are frequently associated with human diseases. Here, we present a comprehensive database of PTMs associated with diseases (PTMD 2.0), including 342 624 PTM–disease associations (PDAs) in 15 105 proteins for 93 types of PTMs and 2083 diseases. Based on the distinct PTM states in diseases, we classified all PDAs into six categories: upregulation (U) or downregulation (D) of PTM levels, absence (A) or presence (P) of PTMs, and creation (C) or disruption (N) of PTM sites. We provided detailed annotations for each PDA and carefully annotated disease-associated proteins by integrating the knowledge from 101 additional resources that covered 13 aspects, including disease-associated information, variation and mutation, protein–protein interaction, protein functional annotation, DNA and RNA element, protein structure, chemical–target relationship, mRNA expression, protein expression/proteomics, subcellular localization, biological pathway annotation, functional domain annotation and physicochemical property. With a data volume of ∼8 GB, we anticipate that PTMD 2.0 will serve as a fundamental resource for further analysing the relationships between PTMs and diseases. The online service of PTMD 2.0 is freely available at https://ptmd.biocuckoo.cn/.

Graphical Abstract

Graphical Abstract.

Introduction

Protein post-translational modification (PTM), a reversible covalent modification after protein translation, is one of the most important regulatory mechanisms orchestrating a broad spectrum of biological processes (1–3). Various PTMs regulate protein structural and functional dynamics by altering residue hydrophobicity and protein state and stability (4,5). Over 680 types of PTMs have been discovered to date (http://www.uniprot.org/docs/ptmlist.txt), among which phosphorylation and ubiquitination have been the most studied due to their high abundance in cells (3,6). In 1992, Edomond H. Fischer and Edwin G. Krebs were awarded the Nobel Prize in Physiology or Medicine for their discovery of reversible protein phosphorylation as a biological regulatory mechanism (7), which was subsequently strongly implicated in orchestrating signal transduction, cytoskeleton rearrangement and cell cycle progression, among other critical cellular processes (1,8). In 2004, Irwin Rose, Aaron Ciechanover and Avram Hershko were awarded the Nobel Prize for identifying ubiquitination as a transferable signal for the degradation of proteins by the proteasome (9). Additionally, other modifications, such as acetylation, regulate transcription and cellular metabolism through the modification of histones and non-histone proteins (10,11).

With the rapid progress in the study of PTMs, accumulating evidence has revealed that abnormal PTM statuses are frequently implicated in a variety of human diseases, such as neurodegenerative diseases, cancers and diabetes (12–14). For example, site-specific phosphorylation of microtubule-associated protein tau (MAPT) inhibits amyloid-β toxicity, which provides insights into postsynaptic processes in Alzheimer's disease (AD) and highlights the pathogenic role of MAPT/tau phosphorylation in neuronal toxicity (15). Also, S-phase kinase-associated protein 2 (SKP2), an oncoprotein and cell cycle regulator, has been reported to promote the phosphorylation, ubiquitination and degradation of the tumour suppressor programmed cell death protein 4 (PDCD4) in breast cancer cells (16). Moreover, genetic mutations in a serine/threonine protein kinase gene, leucine-rich repeat kinase 2 (LRRK2), are considered the most common cause of familial Parkinson's disease (PD) (17). It has been reported that autophosphorylation of Ser1292 indicates LRRK2 kinase activity and contributes to the cellular effect of PD mutations (18). These studies indicate strong associations between PTMs and a wide range of diseases. Therefore, a comprehensive and integrated data resource of PTM–disease associations (PDAs) with detailed annotations would offer a fundamental data resource for further elucidating the molecular mechanisms underlying human disease pathology.

In 2018, we developed PTMD 1.0, a database of PTMs that are associated with diseases, containing 1950 known PDAs in 749 proteins for 23 PTM types and 275 disease types from the literature (19). In recent years, an increasing number of studies have focused on analysing the relationships between PTMs and diseases. We continuously maintained and updated the database, and PTMD 2.0 contained 342 624 PDAs in 15 105 proteins for 93 types of PTMs and 2083 diseases (Supplementary Table S1). First, we manually collected 6740 high-quality and experimentally identified PDAs from the literature, with detailed descriptions of PTM–disease relationships taken from the primary references (Supplementary Table S2). Also, we integrated and processed 335 884 potential PDAs from three public databases, including ActiveDriverDB (20,21), BioMuta (22) and PhosphoSitePlus (23), which map and characterize PTM sites that overlap with disease-associated mutations. Furthermore, we carefully annotated disease-associated proteins by integrating the knowledge from 101 additional resources that covered 13 aspects, including disease-associated information, variation and mutation, protein–protein interaction, protein functional annotation, DNA and RNA element, protein structure, chemical–target relationships, mRNA expression, protein expression/proteomics, subcellular localization, biological pathway annotation, functional domain annotation and physicochemical properties (Supplementary Table S3). Compared with PTMD 1.0, the number of reported PDAs increased 3.5-fold. Additionally, PTMD 2.0 contained approximately 8 GB of data, representing a 40-fold increase in volume compared with that of PTMD 1.0 (∼0.2 GB). We anticipate that the updated PTMD 2.0 will serve as a fundamental resource for further analysing the relations between PTMs and diseases.

Construction and content

Data collection, curation and integration

In this update, we focused primarily on expanding new PDA events and introducing new PTM types from the literature and PTM databases. First, we update the existing PTM types in PDAs. We conducted a comprehensive search on PubMed via multiple keywords and manually curated the abstracts or full texts of each paper published after 1 January 2018. These keywords included but were not limited to ‘phosphorylation’, ‘ubiquitination’, ‘acetylation’, ‘glycosylation’, ‘methylation’, ‘nitration’, ‘S-palmitoylation’, ‘S-nitrosylation’ and ‘SUMOylation’. In addition to updating existing modification types, we also incorporated several new PTM types into our database. By searching PubMed with relevant keywords such as ‘benzoylation’, ‘sulfation’, ‘succinylation’, ‘neddylation’, ‘ampylation’, ‘isonicotinylation’, ‘umpylation’, ‘glycation’, ‘lactylation’ and other modifications, we curated experimentally identified PDAs from the literature.

To gain further insight into the relationships between PTMs and diseases, we classified all collected PDAs into six classes according to the status of PTM events in diseases (Supplementary Figure S1). (i) Upregulation (U): the PTM level is increased in diseases. For example, the expression of mHtt increases the phosphorylation of p53 at Ser46, causing it to interact with Pin1 and dissociate from iASPP, which induces the expression of apoptotic target genes (24). (ii) Downregulation (D): the PTM level is downregulated in diseases. In leukaemia, PRL2 downregulates the phosphorylation of CBL at tyrosine 371, reducing FLT3 ubiquitination and degradation and thereby enhancing FLT3 signalling (25). (iii) Presence (P): the presence of a PTM event is associated with disease progression. For example, phosphorylation of EZH2 activates STAT3 signalling via STAT3 methylation and promotes the tumorigenicity of glioblastoma stem-like cells (26). (iv) Absence (A): the absence of a PTM event is involved in disease progression. Loss of acetylation at Lys16 and trimethylation at Lys20 of histone H4 are common hallmarks of human cancer (27). (v) Creation (C): a mutation event (single amino acid or indel mutation) creates one or multiple PTM sites or increases the protein PTM level in disease. Many mutations causing human genetic disease may create new N-glycosylation sites, the pathogenic effects of which result from the addition of N-linked carbohydrates (28). (vi) Disruption (N): a mutation event that disrupts one or multiple PTM sites or reduces protein PTM levels in diseases. A typical example of this type of disease-associated PTM is that the T313M mutation of PINK1 disrupts its MARK2-mediated phosphorylation and is implicated in PD (29). Therefore, the former two types (U and D) emphasized the changes in PTM levels in diseased tissues compared with normal tissues. In contrast, types P and A highlighted the associations of the presence or absence of a PTM event with disease progression. The last two types, C and N, focus on the impact of mutation events on PTMs in diseases.

Next, we integrated these newly reported PDAs with data from PTMD 1.0 and three compelling public databases, including ActiveDriverDB (20,21), BioMuta (22) and PhosphoSitePlus (23). ActiveDriverDB is a web-based proteo-genomics database that annotates and interprets protein-coding genetic variations through 552 068 experimentally identified PTM sites to assess their functional impact on human diseases, including cancer (20). BioMuta is an annotated database that maps single-nucleotide variations (SNVs) at the genome, protein and gene levels, and its latest version, BioMuta v3.0, includes 4 684 236 non-synonymous SNVs across 18 269 genes, primarily linked to 77 cancer types as classified by Disease Ontology (22,30). PhosphoSitePlus is a manually curated, open resource for studying protein PTMs in humans and mice, featuring over 330 000 non-redundant PTM sites, with data primarily from mass spectrometry. In its ‘PTMVar’ dataset, there have been over 25 000 PTM sites affected by missense mutations and associated with various diseases (23). The newly reported PDAs were collected from the literature, and then combined with existing known PDAs in PTMD 1.0 and PhosphoSitePlus (23). Additionally, we mapped and characterized PTMs that are potentially affected by genetic variants linked to diseases from the three databases. This comprehensive integration allowed us to enhance the database significantly. In total, we manually collected 6740 experimentally identified PDAs from the literature and further integrated 335 884 potential PDAs from the three databases (Supplementary Table S1). Compared with PTMD 1.0, PTMD 2.0 achieved a 3.5-fold increase in the reported PDAs and additionally supplements potential PDAs that integrate PTM sites affected by disease mutations. This substantial expansion underscores our commitment to providing a more comprehensive resource for understanding the intricate relationships between PTMs and diseases.

Comprehensive annotation of disease-associated proteins and PDA sites

PTMD 2.0 was developed as a protein-centred database to facilitate comprehensive research. Each protein entry in the database is assigned a unique PTMD ID, which serves as the primary accession identifier, e.g. PTMD00006, for the cellular tumour antigen p53 in Homo sapiens. For each protein entry, basic annotations, such as UniProt/RefSeq/Ensembl accession numbers, protein names/synonyms, gene names/synonyms, gene IDs, NCBI Taxa IDs, functional descriptions, protein sequences, nucleotide sequences, Gene Ontology terms and relevant keywords, were integrated from UniProt (31–34). A representative 3D protein structure was selected from the Protein Data Bank (PDB) (35) and AlphaFold Protein Structure Database (AlphaFold DB) (36), if available, for visualization of known PDA sites. Each PDA site includes information on the modification position, residue, type of modification, regulation type and associated disease. For each known PDA event, a brief description of the regulatory role in disease was provided, as well as PMIDs of primary references to ensure traceability and credibility. For potential PDAs integrated from other databases, the following information was additionally provided: the sequence position of the disease-associated mutation affecting the modification site, the original amino acid altered by the mutation, the variant amino acid resulting from the mutation, a 15-amino-acid flanking peptide around the mutation, the mutation’s position relative to the PTM, the impact on the modification and PolyPhen results, which predict the potential impact of amino acid substitution on protein stability and function (37), if available.

We further enhanced the database by compiling and integrating knowledge from 101 additional public databases, providing thorough annotations of disease-associated proteins and PDA sites (Supplementary Table S3). These public resources cover 13 aspects: (i) variation and mutation; (ii) disease-associated information; (iii) protein–protein interactions; (iv) protein function; (v) DNA and RNA elements; (vi) protein structure; (vii) chemical–target relationships; (viii) mRNA expression; (ix) protein expression/proteomics; (x) subcellular localization; (xi) biological pathways; (xii) domain annotation; and (xiii) physicochemical properties (Figure 1). Details on processing the annotations of each resource are described in detail in the Supplementary Methods. All annotations can be downloaded from PTMD 2.0 (https://ptmd.biocuckoo.cn/download.php).

Figure 1.

The procedure for constructing PTMD 2.0. First, we manually collected experimentally identified disease-associated proteins and PDA sites from PubMed. Also, we integrated and processed data from four public databases, including PTMD 1.0 (19), ActiveDriverDB (20,21), BioMuta (22) and PhosphoSitePlus (23) (Supplementary Table S1). Furthermore, we annotated the disease-associated proteins via the knowledge from 101 additional databases that cover 13 aspects: (i) variation and mutation; (ii) disease-associated information; (iii) protein–protein interaction; (iv) protein function; (v) DNA and RNA element; (vi) chemical–target relationships; (vii) protein structure; (viii) mRNA expression; (ix) physicochemical properties; (x) protein expression/proteomics; (xi) subcellular localization; (xii) biological pathway; and (xiii) domain annotation (Supplementary Table S3). Ac, acetylation; Gly, glycosylation; Kla, lysine lactylation; Me, methylation; P, phosphorylation; SUMO, SUMOylation; Ub, ubiquitination.

Data statistics of PTMD 2.0

To explore the relationships between known PTMs and diseases, we conducted a statistical analysis of reported PDA events (Supplementary Table S2). Previously, in PTMD 1.0, we grouped 23 PTM types into nine supertypes, including acetylation, glycosylation, methylation, nitration, S-palmitoylation, phosphorylation, S-nitrosylation, SUMOylation and ubiquitination. For example, serine phosphorylation, threonine phosphorylation and tyrosine phosphorylation, along with dephosphorylation and autophosphorylation, were all classified under the supertype of phosphorylation, while monoubiquitination, polyubiquitination, deubiquitination and autoubiquitination were classified into the supertype of ubiquitination. With the expanding modification types, we added 70 new modification types and grouped all PTM types into 49 supertypes. The numbers of modification events are shown for each PTM type (Figure 2A). According to our results, the top three supertypes of PTMs with the greatest number of disease associations are phosphorylation (4181 PDAs, 68.09%), ubiquitination (761 PDAs, 12.39%) and acetylation (625 PDAs, 9.27%), whereas other PTM supertypes have relatively fewer disease associations. We also grouped 2083 types of diseases into 34 supertypes based on tissue location (Figure 2B). Among all the diseases, the supertype of nervous system disease has the largest number of associations with PTMs (13.8%), which indicates potentially important roles of PTMs in neurodegenerative diseases. Many abnormal PTM events occur extensively in numerous neurodegenerative diseases, such as AD and PD, and are involved in neuronal dysfunction and cell death (18,38,39). These significant associations suggest that targeting abnormal PTM events may offer effective therapeutic strategies. Additionally, breast cancer ranks second in terms of the number of PDAs. PTM events may influence breast cancer by regulating protein stability and cell cycle progression (40,41). Among the six types of PDAs, 61.14% of PTMs were upregulated and 21.23% were downregulated in diseases (Figure 2C). These findings suggest that alterations in PTM levels may significantly contribute to disease development, demonstrating great potential as prognostic markers and therapeutic targets for disease treatment. Moreover, 12.28% of disease-associated PTM events are characterized by the presence of PTMs, whereas only 2.79% are characterized by their absence. These findings indicate that PTM events preferentially occur during disease progression, whereas the development of only a small proportion of diseases is associated with the loss of PTMs. Additionally, the number of disease associations within each type of PDA varied remarkably among the 34 supertypes of disease (Figure 2D).

Figure 2.

Data statistics and analysis using known PDAs in PTMD.2.0. (A) The 93 PTM types were classified into 49 supertypes. The numbers of known PDA sites are shown for each PTM type. (B) The percentage of different supertypes of diseases in PTMD 2.0. In total, 2083 types of diseases are classified into 34 supertypes based on tissue information. (C) The number of known disease associations within each type of disease-associated PTM. (D) Distribution of the six types of known PDAs across disease supertypes. The dot size indicates the number of PDAs per disease. ‘U’ and ‘D’ indicate upregulated and downregulated PTM levels in diseases, respectively. ‘P’ and ‘A’ denote the presence and absence of a PTM event associated with disease progression, respectively. ‘C’ represents mutations creating or increasing PTM sites or levels, whereas ‘N’ represents mutations disrupting or reducing PTM sites or levels. PTM, post-translational modification; PDA, PTM–disease association.

Usage

The online service of PTMD 2.0 was developed in an easy-to-use manner. Here, we selected human p53 as an example to describe the use of PTMD 2.0. On the browse page, we implemented three types of browsing options, including ‘Browse by PDAs’, ‘Browse by Diseases’ and ‘Browses by PTMs’ (Figure 3). In the option of ‘Browse by PDAs’, users can select the regulation type ‘Upregulation’ and then choose the relevant PTM type ‘phosphorylation’, which will generate a tabular list of all PDAs that are upregulated and involve phosphorylation (Figure 3A). For the option of ‘Browse by Diseases’, the 34 supertypes of diseases were intuitively visualized as icons around a human body (Figure 3B). Users can click these icons to explore specific diseases related to each tissue. By clicking on ‘Ovary’, users can view a table of specific types of kidney diseases. Additionally, clicking the ‘ovarian cancer/carcinoma’ link will display all PDA substrates associated with ovarian cancer disease. For the option ‘Browse by PTMs’, users can click on the link ‘phosphorylation’ to select from various phosphorylation types, including tyrosine phosphorylation, threonine phosphorylation, serine phosphorylation, phosphorylation, dephosphorylation and autophosphorylation, to find the associated PDA substrates (Figure 3C). By clicking on the PTMD ID ‘PTMD00006’, users can access the protein page for p53 (Figure 3D). An ‘enlarging’ button was implemented below the structure window, allowing the 3D structure of human p53 to be viewed in a larger window. In addition to the basic information and details on PDA sites, additional annotations could be accessed either by clicking the ‘Annotation’ button on the left bar or by clicking the ‘Integrated Annotation’ icon adjacent to the PTMD ID (Figure 3D). Each type of additional annotation can be presented by clicking on its corresponding name. For example, users could click on ‘dbSNP’ under the variation and mutation section to view all single nucleotide polymorphisms identified in the TP53 gene from the single nucleotide polymorphism database (dbSNP) (34) (Figure 3E).

Figure 3.

The browsing options of PTMD 2.0: (A) Browse by PDAs; (B) Browse by Diseases; and (C) Browse by PTMs. (D) The tabular list and the protein page of human p53. In addition to the basic information and details on the PDA sites, additional annotations can be accessed. (E) Annotation page of p53. For example, dbSNP single nucleotide polymorphisms identified in the TP53 gene are shown (67).

Additionally, multiple search options were implemented. On the search page (http://ptmd.biocuckoo.cn/advance.php), users can choose from four search options, including ‘Simple Search’, ‘Advanced Search’, ‘Batch Search’ and ‘BLAST Search’, to query data in PTMD 2.0 (Supplementary Figure S2). The ‘Simple Search’ option was also presented at the home page, and users could click on ‘Example’ and then ‘Submit’ to search for human p53 and its homologues in other species.

Discussion

Since the discovery of the first phosphorylated protein, casein, in 1883, the identification and characterization of protein PTMs has been intensively studied for over 135 years (42). Extensive experimental studies have demonstrated that abnormal PTM status is often involved in disease development and progression (12,43), but the total number of aberrant PTM events associated with human disease remains unclear. A comprehensive and integrated data resource detailing disease-associated PTMs would significantly help with a deeper understanding and exploration of the relationship between PTMs and disease and elucidate the mechanisms underlying various diseases. For this purpose, we developed PTMD 1.0, which contains 1950 PDAs in 749 proteins for 23 PTM types and 275 diseases. In this study, we developed PTMD 2.0, which contains 342 624 PDAs in 15 105 proteins for 93 types of PTMs and 2083 diseases. Compared with PTMD 1.0, the updated PTMD 2.0 database provided a significantly expanded range of modification types and diseases, with substantially enriched annotations. The current dataset includes significantly more information, incorporating reported PDAs from the literature and potential PDAs from other databases, reflecting the considerable expansion of our data collection efforts. The volume of experimentally verified data sourced from the literature has expanded 3.5-fold compared with the previous release, whereas the total number of disease-associated proteins has increased 20-fold. Furthermore, the number of modification types has increased 4-fold, and the number of disease types has increased 7.5-fold. Moreover, PTMD 2.0 contains approximately 8 GB of data, representing a 40-fold increase in volume from PTMD 1.0 (∼0.2 GB).

The information on PDAs curated in our database allows users to query and explore the relationships between PTMs and diseases, enabling researchers to investigate these PTM events in various contexts to understand their specific implications in disease progression and treatment strategies. For example, the tumour suppressor p53 has been widely studied because of the complexity of its function and regulation in carcinogenesis (44). Many important modifications of p53, which play significant functional roles, have been confirmed (45–47). Ser46 phosphorylation of p53, triggered by severe or persistent stress, shifts the p53 response from cell cycle arrest to apoptosis (48,49). According to information from our database, Ser46 phosphorylation increases with mutant huntingtin (mHtt) expression in Huntington disease (HD) patients (24), leading to Pin1-mediated conformational changes and neuronal apoptosis, exacerbating HD pathology. Moreover, Ser46 phosphorylation decreases in colorectal, breast and hepatocellular cancers (50–52). Furthermore, studies have shown that Ser15 phosphorylation of p53 induces dissociation from MDM2, stabilizing the p53 protein (53). According to our database, Ser15 phosphorylation enhances p53 stability and function, increasing its ability to inhibit tumour cell proliferation and induce apoptosis. By preventing the formation of the p53/MDM2 complex, Ser15 phosphorylation prevents p53 degradation via the ubiquitin–proteasome pathway, increasing p53 stability and levels within lymph node carcinoma of the prostate prostate cancer cells, making it a critical factor in the antitumour process (54).

In addition to the basic information and details on the PDA sites of each protein entry, additional annotations from 101 additional resources were integrated to cover 13 aspects. For example, from the Catalogue Of Somatic Mutations In Cancer (COSMIC), which contains cancer genes and mutations (55), there were 1129 records of missense mutations that change known PTM sites of the human p53 protein (Figure 4). Among the 29 protein sites altered by COSMIC mutations, 12 were reported to be associated with diseases. The disease-associated information revealed that 4040 relevant disease terms were associated with the TP53 gene according to Diseases 2.0 (56). Human p53 acts as a scaffold protein of liquid–liquid phase separation in the nucleus (57–59), interacts with 870 other proteins and is post-transcriptionally targeted by 130 miRNAs according to the annotations in miRTarBase, a data resource for maintaining experimentally validated microRNA–target interactions (60). With 58 resolved 3D structures in the PDB (35), the p53 protein can be targeted by 859 chemicals, such as fluorouracil (61), trientine (62) and wortmannin (63). According to the annotations of Reactome, a well-curated pathway database (64), the protein p53 is involved in 112 biological pathways.

Figure 4.

Overview of integrated annotations for human p53. A brief summary of all 101 data resources used in this study is shown in Supplementary Table S3. The details on processing each of the resources are presented in the Supplementary Methods.

In the future, we will regularly maintain and update the database if new disease-associated PTM events become available. Also, we plan to include additional PTM types and related diseases by expanding the scope of public resources for further integration. Of note, one of the limitations in PTMD 2.0 is the variability in the quality of annotations for both PDAs and proteins, with some being potentially incorrect or inconclusive. For example, although Ser46 phosphorylation of human p53 has been well documented to be frequently downregulated in cancer, the corresponding experiments were mainly performed in various cancer cell lines (50–52). Additional evidence in animal models and/or cancer patients will be included, if available, in the near future, and such information will strengthen the confidence of PTMD data. For protein functional annotations, one of the integrated databases is The Animal Transcription Factor DataBase (AnimalTFDB) 3.0, which contains both known and predicted transcription factors and and cofactors in animals (65). Thus, for proteins in PTMD 2.0 annotated as potential transcription factors, the transcriptional activity should be experimentally validated prior to further analysis. Besides directly using the online service, PTMD 2.0 also provided an application programming interface (API) for accessing the data, with the help of the National Genomics Data Center (66). For example, users can retrieve all PDAs of p53, using the following API link: https://ngdc.cncb.ac.cn/search/api/specific?db=ptmd&q=PTMD00006. In our future plan, we will provide more API options to increase the database’s accessibility, allowing researchers to integrate the database into their workflows and tools more seamlessly. We believe that PTMD 2.0 will serve not only as a fundamental resource for analysing PDAs but also as a catalyst for new discoveries and advancements in understanding the molecular mechanisms underlying diseases.

Data availability

PTMD 2.0 is freely accessible at https://ptmd.biocuckoo.cn/. More details are available under the ‘Download’ and ‘User guide’ pages on the website.

Supplementary data

Supplementary Data are available at NAR Online.

Funding

Funding for open access charge: Natural Science Foundation of China [32341020, 32341021 and 31930021]; National Key R&D Program of China [2022YFC2704304 and 2021YFF0702000]; Hubei Innovation Group Project [2021CFA005]; Interdisciplinary Research Program of HUST [2023JCYJ010 and 2024JCYJ013]; Hubei Province Postdoctoral Outstanding Talent Tracking Support Program; Research Core Facilities for Life Science (HUST).

Conflict of interest statement. None declared.