Cross-species cluster co-conservation: a new method for generating protein interaction networks
- Anis Karimpour-Fard†1,
- Corrella S Detweiler†2,
- Kimberly D Erickson2,
- Lawrence Hunter1 and
- Ryan T Gill3Email author
© Karimpour-Fard et al.; licensee BioMed Central Ltd. 2007
Received: 5 July 2007
Accepted: 5 September 2007
Published: 05 September 2007
Co-conservation (phylogenetic profiles) is a well-established method for predicting functional relationships between proteins. Several publicly available databases use this method and additional clustering strategies to develop networks of protein interactions (cluster co-conservation (CCC)). CCC has previously been limited to interactions within a single target species. We have extended CCC to develop protein interaction networks based on co-conservation between protein pairs across multiple species, cross-species cluster co-conservation.
The exponential increase in sequence information has widened the gap between the number of predicted and experimentally characterized proteins. At present, about 400 microbial genomes are fully sequenced. The prediction of protein function from sequence is a critical issue in genome annotation efforts. Currently, the best established method for function prediction is based on sequence similarity to proteins of known function. Unfortunately, homoogy-based prediction is of limited use due to the large number of homologous protein families with no known function for any member. An alternative method for predicting protein function is the phylogenetic profiles approach, also known as the co-conservation (CC) method first introduced by Pellegrini et al. . Co-conservation predicts interactions between pairs of proteins by determining whether both proteins are consistently present or absent across diverse genomes [2–8]. CC methods have been shown to be more powerful than sequence similarity alone at predicting protein function.
Even though all CC methods rely on the premise that functionally related proteins are gained or lost together over the course of evolution, several different strategies for performing CC studies have been reported. For example, Date et al.  used real BLASTP best hit E-values normalized across 11 bins instead of binary classification for conservation, while Zheng and coworkers  constructed phylogenetic profiles using presence/absence of neighboring gene pairs. Alternatively, Pagel et al.  constructed phylogenetic profiles between domains, instead of genes, and then created domain interaction maps. Barker et al.  applied maximum likelihood statistical modeling for predicting functional gene linkages based on phylogenetic profiling. Their method detected independent instances of protein pair correlated gain or loss on phylogenetic trees, reducing the high rates of false positives observed in conventional across-species methods that do not explicitly incorporate a phylogeny .
Cross-species clustered co-conservation
CS-CCC identifies interactions that could not be identified by CCC
CS-CCC reveals how proteins that function in distinct but coordinated processes may have evolved
Figure 4 shows a cluster containing proteins that function in distinct but inter-dependent processes. For instance, in P. aerginosa, flagella, chemotaxis machinery, and type IV pili are important for bacterial biofilm formation [13, 14] and are co-conserved. Type IV pili mediate twitching motility, which is important for subsequent spreading of the bacteria over the surface and the formation of microcolonies within a developing biofilm . Twitching motility proteins PilJ and PilK are co-conserved within this cluster and are highly interconnected with flagella and chemotaxis proteins. Flagellar motility appears to be required for approaching surfaces, and 17 flagellar proteins are co-conserved (Figure 4c). Chemotaxis is required for the bacteria to swim towards nutrients associated with a surface. P. aerginosa has two chemotaxis signaling systems, and proteins representing both are in the biofilm cluster (CheR1, CheR2, CheA, CheW, PA0173, PA0178; PctA, PctB, PctC). These data suggest that chemotaxis, flagella, and pili proteins may be co-conserved because they all contribute to biofilm formation. Moreover, the inclusion of P. aerginosa in the CS-CCC analysis brought pili proteins into the biofilm cluster, suggesting that in some bacteria, all of these processes co-evolved. Thus, CS-CCC can identify co-conserved networks of proteins that function in biochemically distinct pathways but that contribute to complex biological phenomenon.
RpoN connects RpoN-regulated proteins with flagella and with type III secretion system proteins
Homology between co-conserved flagellar and T3SS genes
S. typhimurium LT2
E. coli 0157
P. aerginosa (PAO1)
CS-CCC can be used to assign function to unstudied proteins
Genes that function in biofilm formation
Small clusters can contain proteins that function in the same processes
Examination of small protein clusters revealed that most pairs or triplets contain proteins that function in the same processes. To further test this observation, we experimentally examined the triplet containing YcgB, YeaH, and YeaG, which cluster together across different bacteria (Figure 5b). Because independent data indicate that yeaH, but not yeaG, contributes to antimicrobial peptide resistance in S. typhimurium , we determined whether strains lacking ycgB have a similar phenotype. Strains lacking ycgB were indeed sensitive to antimicrobial peptides (unpublished data). Thus, CS-CCC analyses revealed previously unknown protein interactions that provided sufficient justification to test a specific biological hypothesis suggested by these interactions.
When proteins are not identified as co-conserved using CS-CCC
In this study, we have shown that CS-CCC of proteins provides important information. Both the presence and the absence of clustered co-conservation for any given protein are informative. There are at least two reasons why proteins that function together are not co-conserved in a species: first, a protein is found only in certain organisms or a protein function is performed by different proteins in different organisms; and second, a result is a false negative.
A protein is found only in certain organisms: T3SS effectors
Effector proteins are secreted by T3SS machinery and function to alter host cell physiology . A bacterial species can have many effectors but they generally do share apparent sequence homology, either within or between bacteria . We examined 49 known SPI2 and SPI1 effectors in S. typhimurium LT2 and 40 known effectors in P. aeruginosa and found that none of these proteins are co-conserved. In contrast, some of the known translocon T3SS proteins, which form the secretion apparatus, are highly co-conserved (Figure 4c). Thus, while CS-CCC offers insights into the function of proteins that are co-conserved, our results show that some of the non co-conserved proteins, such as effectors, are organism specific.
A result is a false negative: flagella and RpoN
Our analysis of false negatives reveals that the CS-CCC method produces some false negatives. For instance, there is no co-conservation between RpoN and flagella in E. coli 0157, S. flexneri and P. aeruginosa (Figure 4c). However, it has been experimentally shown in P. aeruginosa that many flagellar genes, such as flhA and flhB, are regulated by RpoN . In addition, an RpoN consensus sequence is located in the intergenic region between flhB and flhA . These data suggest that the absence of co-clustering of RpoN with flagellar proteins in P. aeruginosa is a false negative result. Thus, when proteins are not co-conserved, it cannot be concluded that they are functionally unrelated. This result further underlines the value of developing and comparing interaction networks from multiple genomes when attempting to infer function.
There are also some situations in which a result is both a false negative and the protein in question is found only in certain organisms. The bacterial flagellum is a complex molecular system with multiple components required for functional motility. It extends from the cytoplasm to the cell exterior. Not only are flagella organelles of locomotion, but they also play important roles in attachment and biofilm formation. There are common themes in flagellar protein control and assembly, but there also appears to be variation among organisms. Some of the flagellar proteins are not co-conserved in any of the bacteria of our study, such as, three ring proteins (FlgH, FlgI, and FliF), and some of the axle-like proteins FliE, FlgB, FlgF, FlgL, and FliD. FliE has been shown to physically interact with FlgB . The stator motor proteins MotA and MotB are also not co-conserved. Thus, CS-CCC analysis of the flagellar cluster yields both false negative results and is also a consequence of species-specific proteins. This also illustrates that determining why proteins are not co-conserved can be difficult, without additional information.
Large volumes of data make computational methods feasible, exciting, and preferable to gene-by-gene homology searches. We have shown that use of CS-CCC expands protein interaction networks to include proteins with distinct functions that are involved in coherent biological processes, offers insight into the function of uncharacterized proteins, reveals unique information about each genome examined, and gives insight into the process of evolution.
Protein co-conservation can be a result of many factors, including vertical inheritance or functional selection. Thus, we have examined patterns of CCC within and across several bacteria using CS-CCC. Our analysis showed that this computational approach provides us with more information than the traditional homology approaches or CCC. Homology approaches to protein function are based on similarity to other proteins with known functions and are limited by the fact that many proteins have unknown functions. While homology-based methods can be effective for predicting the functions of remote homologs, these methods perform poorly as the evolutionary distance between homologous proteins increases. Even a sophisticated homology-based method fails to successfully assign functions to most of the proteins for a particular organism. CCC, on the other hand, is not strictly based on homology but is limited by its ability to analyze only a single species at a time. In contrast, CS-CCC examines each cluster across multiple species and reveals interactions that both homology-based methods and CCC fail to identify. Use of CS-CCC allows researchers to extend the protein interaction network to better understand pathways involving multiple proteins with multiple functions. Therefore, the CS-CCC method is a significant advance and will be useful for researches in many different fields of biology.
Prediction by CS-CCC provided us with global views of six complete bacterial genomes. Identification by CS-CCC of proteins that cluster together enabled more accurate predictions of the biological roles that proteins with previously unstudied functions may play. For instance, proteins that function in distinct but coordinated processes can be co-conserved across species even though not all processes occur in all bacteria (Figure 4c). In addition, in large, highly interconnected clusters in which most of the proteins have unknown functions, it is likely that they all function together in a common phenomenon. The GGDEF/EAL cluster is an example of this, as many of the previously unknown proteins in this cluster play roles in biofilm formation (Figure 5a). Even small protein clusters identified by CS-CCC are likely to consist of proteins that function in the same process, as shown by COG/TIGR analysis and experimentally (Figure 5b). These analyses provide evidence that the CS-CCC method is a reliable predictor of functional relationships.
For any given method, there are advantages and disadvantages. The number of false positives and false negatives is a key measurement of accuracy. In our case, the number of false negatives is not possible to estimate without performing many additional laboratory experiments. However, our evaluation of CS-CCC showed that the number of false positives was low. Since this method was evaluated based on our selected bacteria, there may be some bias toward overestimation of accuracy when applied to other organisms, and this remains to be tested. In addition, we have shown that our results can be sensitive to the number of bacteria included in our analysis. Finally, there may be some aspects of the bacteria we chose that are not representative of other bacteria, further reducing the generality of these results. Thus, while the report here represents a compelling demonstration of the value of performing CCC across multiple species, future efforts should be focused on developing better understanding of which and how many organisms to include in CS-CCC studies.
Materials and methods
Bacteria used to create CS-CCC graphs
We chose to focus on the Gamma subgroup of proteobacteria because members of this subgroup are among the best characterized, including whole genome sequences and curated datasets of protein functions and interactions. The genomes of five closely related Gamma Gram-negative and one low G+C bacteria (B. subtilis) were used to evaluate the CCC method. Substantial experimental data exist for all six bacteria. The gammaproteobacteria included E. coli (K12 and O157-O157:H7 EDL933), S. flexneri (2a str. 2457T), S. typhimurium (LT2), and P. aeruginosa (PAO1). E. coli (K12) is the most intensively studied Gram-negative bacteria and is the closest studied relative of P. aeruginosa, and S. typhimurimum LT2. E. coli (O157-O157:H7 EDL933) is a clinical isolate from raw hamburger meat implicated in hemorrhagic colitis outbreak, and S. typhimurium LT2 causes enteritis in humans. P. aeruginosa is an opportunistic pathogen and is the major cause of morbidity and mortality in patients with cystic fibrosis; P. aeruginosa PAO1 was isolated from a wound . P. aeruginosa is a versatile Gram-negative bacterium that also thrives in soil, marshes and coastal marine habitats, and on plant tissues . E. coli K12 diverged 4.5 million years ago (MYA) from O157, an estimated 100 MYA from Salmonella, 200 MYA from Pseudomonas, and 1,200 MYA from Bacillus. Thus, we examined a combination of pathogenic and non-pathogenic organisms that range from closely to distantly related.
Construction of CS-CCC graphs
Comparison of genomes examined in this study
No. of annotated genes
No. (%) of co-conserved genes
No. of co-conserved protein pairs
E. coli (K12)
E. coli (O157-O157:H7 EDL933)
Shigella flexneri 2a str. 2457T
Salmonella typhimurium LT2 + pSLT plasmid
4,857,432 + 93,939
4,425 + 102
P. aeruginosa (PAO1)
Finally, for comparison of each cluster across different species (CS-CCC), we used BFS to build a network (source network) for a set of target proteins from the source genome. We then built networks for each additional organism that contained proteins with the same name as at least one of the proteins from the source networks. This process identifies proteins and protein interactions that are consistently identified across multiple species (colored gray in Figure 1c) or that are unique to individual species (colored red in Figure 1c). This same method can be used to further parse such networks to identify combined, common and unique networks present for specific proteins across a collection of organisms (Figure 1c). In this way, CS-CCC builds on information generated by CCC (Figure 1b) to provide more accurate and genome-specific protein function assignment. We used protein name to map links across conserved species (thus, links are not explicitly based on orthology) [37–39]. Like all methods, the use of protein names has both advantages and disadvantages. Here, protein name was chosen in order to validate that CS-CCC provides new and biologically informative data not accessible by CCC alone. For this purpose, we chose to validate this method using named proteins where functional information was available. While this is appropriate for method validation, the disadvantage is that there are problems with annotation due in part to a lack of standardization, which would limit the number of proteins for which this analysis can be reliably performed. In light of this limitation, we considered using reciprocal homology as an alternative to protein name. We found that this introduces unacceptable levels of cross-talk, much of which is likely noise. Addressing this limitation is an important area for continued effort.
Data are available upon request.
Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 is a figure that shows the reliability of predicted protein interaction pairs using TIGR role categories at three different confidence levels.
Clusters of Orthologous Groups
cross-species clustered co-conservation
methyl-accepting chemotaxis protein
million years ago
The Institute for Genomic Research
type III secretion systems.
We thank Norman Pace for excellent discussions, Daniel Barker and Sonia M Leach for reading the manuscript and helpful comments. We also thank Kevin B Cohen for helpful comments. This study was supported by NSF grant BES0228584, and NIH grants K25_AI064338, R01-AI-072492A, and R01-LM-008111.
- Pellegrini M, Marcotte EM, Thompson MJ, Eisenberg D, Yeates TO: Assigning protein functions by comparative genome analysis: protein phylogenetic profiles. Proc Natl Acad Sci USA. 1999, 96: 4285-4288. 10.1073/pnas.96.8.4285.PubMedPubMed CentralView ArticleGoogle Scholar
- Marcotte EM, Pellegrini M, Ng HL, Rice DW, Yeates TO, Eisenberg D: Detecting protein function and protein-protein interactions from genome sequences. Science. 1999, 285: 751-753. 10.1126/science.285.5428.751.PubMedView ArticleGoogle Scholar
- Strong M, Mallick P, Pellegrini M, Thompson MJ, Eisenberg D: Inference of protein function and protein linkages in Mycobacterium tuberculosis based on prokaryotic genome organization: a combined computational approach. Genome Biol. 2003, 4: R59-10.1186/gb-2003-4-9-r59.PubMedPubMed CentralView ArticleGoogle Scholar
- Huynen M, Snel B, Lathe W, Bork P: Predicting protein function by genomic context: quantitative evaluation and qualitative inferences. Genome Res. 2000, 10: 1204-1210. 10.1101/gr.10.8.1204.PubMedPubMed CentralView ArticleGoogle Scholar
- Mellor JC, Yanai I, Clodfelter KH, Mintseris J, DeLisi C: Predictome: a database of putative functional links between proteins. Nucleic Acids Res. 2002, 30: 306-309. 10.1093/nar/30.1.306.PubMedPubMed CentralView ArticleGoogle Scholar
- Bowers PM, Pellegrini M, Thompson MJ, Fierro J, Yeates TO, Eisenberg D: Prolinks: a database of protein functional linkages derived from coevolution. Genome Biol. 2004, 5: R35-10.1186/gb-2004-5-5-r35.PubMedPubMed CentralView ArticleGoogle Scholar
- Date SV, Marcotte EM: Protein function prediction using the Protein Link EXplorer (PLEX). Bioinformatics. 2005, 21: 2558-2559. 10.1093/bioinformatics/bti313.PubMedView ArticleGoogle Scholar
- von Mering C, Huynen M, Jaeggi D, Schmidt S, Bork P, Snel B: STRING: a database of predicted functional associations between proteins. Nucleic Acids Res. 2003, 31: 258-261. 10.1093/nar/gkg034.PubMedPubMed CentralView ArticleGoogle Scholar
- Zheng Y, Roberts RJ, Kasif S: Genomic functional annotation using co-evolution profiles of gene clusters. Genome Biol. 2002, 3: RESEARCH0060-10.1186/gb-2002-3-11-research0060.PubMedPubMed CentralView ArticleGoogle Scholar
- Pagel P, Wong P, Frishman D: A domain interaction map based on phylogenetic profiling. J Mol Biol. 2004, 344: 1331-1346. 10.1016/j.jmb.2004.10.019.PubMedView ArticleGoogle Scholar
- Barker D, Pagel M: Predicting functional gene links from phylogenetic-statistical analyses of whole genomes. PLoS Comput Biol. 2005, 1: e3-10.1371/journal.pcbi.0010003.PubMedPubMed CentralView ArticleGoogle Scholar
- Mytelka DS, Chamberlin MJ: Escherichia coli fliAZY operon. J Bacteriol. 1996, 178: 24-34.PubMedPubMed CentralGoogle Scholar
- Korber DR, Lawrence JR, Caldwell DE: Effect of motility on surface colonization and reproductive success of Pseudomonas fluorescens in dual-dilution continuous culture and batch culture systems. Appl Environ Microbiol. 1994, 60: 1421-1429.PubMedPubMed CentralGoogle Scholar
- Pratt LA, Kolter R: Genetic analysis of Escherichia coli biofilm formation: roles of flagella, motility, chemotaxis and type I pili. Mol Microbiol. 1998, 30: 285-293. 10.1046/j.1365-2958.1998.01061.x.PubMedView ArticleGoogle Scholar
- Manson MD, Armitage JP, Hoch JA, Macnab RM: Bacterial locomotion and signal transduction. J Bacteriol. 1998, 180: 1009-1022.PubMedPubMed CentralGoogle Scholar
- Galan JE, Collmer A: Type III secretion machines: bacterial devices for protein delivery into host cells. Science. 1999, 284: 1322-1328. 10.1126/science.284.5418.1322.PubMedView ArticleGoogle Scholar
- Mecsas JJ, Strauss EJ: Molecular mechanisms of bacterial virulence: type III secretion and pathogenicity islands. Emerg Infect Dis. 1996, 2: 270-288.PubMedPubMed CentralView ArticleGoogle Scholar
- Fleiszig SM, Arora SK, Van R, Ramphal R: FlhA, a component of the flagellum assembly apparatus of Pseudomonas aeruginosa, plays a role in internalization by corneal epithelial cells. Infect Immun. 2001, 69: 4931-4937. 10.1128/IAI.69.8.4931-4937.2001.PubMedPubMed CentralView ArticleGoogle Scholar
- Simm R, Morr M, Kader A, Nimtz M, Romling U: GGDEF and EAL domains inversely regulate cyclic di-GMP levels and transition from sessility to motility. Mol Microbiol. 2004, 53: 1123-1134. 10.1111/j.1365-2958.2004.04206.x.PubMedView ArticleGoogle Scholar
- Spiers AJ, Bohannon J, Gehrig SM, Rainey PB: Biofilm formation at the air-liquid interface by the Pseudomonas fluorescens SBW25 wrinkly spreader requires an acetylated form of cellulose. Mol Microbiol. 2003, 50: 15-27. 10.1046/j.1365-2958.2003.03670.x.PubMedView ArticleGoogle Scholar
- Choy WK, Zhou L, Syn CK, Zhang LH, Swarup S: MorA defines a new class of regulators affecting flagellar development and biofilm formation in diverse Pseudomonas species. J Bacteriol. 2004, 186: 7221-7228. 10.1128/JB.186.21.7221-7228.2004.PubMedPubMed CentralView ArticleGoogle Scholar
- Huang B, Whitchurch CB, Mattick JS: FimX, a multidomain protein connecting environmental signals to twitching motility in Pseudomonas aeruginosa. J Bacteriol. 2003, 185: 7068-7076. 10.1128/JB.185.24.7068-7076.2003.PubMedPubMed CentralView ArticleGoogle Scholar
- Karp PD, Riley M, Saier M, Paulsen IT, Collado-Vides J, Paley SM, Pellegrini-Toole A, Bonavides C, Gama-Castro S: The EcoCyc Database. Nucleic Acids Res. 2002, 30: 56-58. 10.1093/nar/30.1.56.PubMedPubMed CentralView ArticleGoogle Scholar
- Georgellis D, Kwon O, De Wulf P, Lin EC: Signal decay through a reverse phosphorelay in the Arc two-component signal transduction system. J Biol Chem. 1998, 273: 32864-32869. 10.1074/jbc.273.49.32864.PubMedView ArticleGoogle Scholar
- Tenorio E, Saeki T, Fujita K, Kitakawa M, Baba T, Mori H, Isono K: Systematic characterization of Escherichia coli genes/ORFs affecting biofilm formation. FEMS Microbiol Lett. 2003, 225: 107-114. 10.1016/S0378-1097(03)00507-X.PubMedView ArticleGoogle Scholar
- Ren D, Bedzyk LA, Ye RW, Thomas SM, Wood TK: Differential gene expression shows natural brominated furanones interfere with the autoinducer-2 bacterial signaling system of Escherichia coli. Biotechnol Bioeng. 2004, 88: 630-642. 10.1002/bit.20259.PubMedView ArticleGoogle Scholar
- Lynch MD, Warnecke T, Gill RT: SCALEs: multiscale analysis of library enrichment. Nat Methods. 2007, 4: 87-93. 10.1038/nmeth946.PubMedView ArticleGoogle Scholar
- Erickson KD, Detweiler CS: The Rcs phosphorelay system is specific to enteric pathogens/commensals and activates ydeI, a gene important for persistent Salmonella infection of mice. Mol Microbiol. 2006, 62: 883-894. 10.1111/j.1365-2958.2006.05420.x.PubMedPubMed CentralView ArticleGoogle Scholar
- Gophna U, Ron EZ, Graur D: Bacterial type III secretion systems are ancient and evolved by multiple horizontal-transfer events. Gene. 2003, 312: 151-163. 10.1016/S0378-1119(03)00612-7.PubMedView ArticleGoogle Scholar
- Hueck CJ: Type III protein secretion systems in bacterial pathogens of animals and plants. Microbiol Mol Biol Rev. 1998, 62: 379-433.PubMedPubMed CentralGoogle Scholar
- Saijo-Hamano Y, Uchida N, Namba K, Oosawa K: In vitro characterization of FlgB, FlgC, FlgF, FlgG, and FliE, flagellar basal body proteins of Salmonella. J Mol Biol. 2004, 339: 423-435. 10.1016/j.jmb.2004.03.070.PubMedView ArticleGoogle Scholar
- Stover CK, Pham XQ, Erwin AL, Mizoguchi SD, Warrener P, Hickey MJ, Brinkman FS, Hufnagle WO, Kowalik DJ, Lagrou M, et al: Complete genome sequence of Pseudomonas aeruginosa PA01, an opportunistic pathogen. Nature. 2000, 406: 959-964. 10.1038/35023079.PubMedView ArticleGoogle Scholar
- NCBI Genbank Protein Annotation. [http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi]
- Winsor GL, Lo R, Sui SJ, Ung KS, Huang S, Cheng D, Ching WK, Hancock RE, Brinkman FS: Pseudomonas aeruginosa Genome Database and PseudoCAP: facilitating community-based, continually updated, genome annotation. Nucleic Acids Res. 2005, D338-343. 33 Database
- Institut Pasteur. [http://genolist.pasteur.fr/SubtiList/]
- Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N, Schwikowski B, Ideker T: Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003, 13: 2498-2504. 10.1101/gr.1239303.PubMedPubMed CentralView ArticleGoogle Scholar
- Snel B, van Noort V, Huynen MA: Gene co-regulation is highly conserved in the evolution of eukaryotes and prokaryotes. Nucleic Acids Res. 2004, 32: 4725-4731. 10.1093/nar/gkh815.PubMedPubMed CentralView ArticleGoogle Scholar
- Penkett CJ, Morris JA, Wood V, Bahler J: YOGY: a web-based, integrated database to retrieve protein orthologs and associated Gene Ontology terms. Nucleic Acids Res. 2006, W330-334. 10.1093/nar/gkl311. 34 Web Server
- Fitch WM: Homology a personal view on some of the problems. Trends Genet. 2000, 16: 227-231. 10.1016/S0168-9525(00)02005-9.PubMedView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.