Horizontal gene transfer and diverse functional constrains within a common replication-partitioning system in Alphaproteobacteria: the repABC operon
© Castillo-Ramírez et al; licensee BioMed Central Ltd. 2009
Received: 9 May 2009
Accepted: 18 November 2009
Published: 18 November 2009
The repABC plasmid family, which is extensively present within Alphaproteobacteria, and some secondary chromosomes of the Rhizobiales have the particular feature that all the elements involved in replication and partitioning reside within one transcriptional unit, the repABC operon. Given the functional interactions among the elements of the repABC operon, and the fact that they all reside in the same operon, a common evolutionary history would be expected if the entire operon had been horizontally transferred. Here, we tested whether there is a common evolutionary history within the repABC operon. We further examined different incompatibility groups in terms of their differentiation and degree of adaptation to their host.
We did not find a single evolutionary history within the repABC operon. Each protein had a particular phylogeny, horizontal gene transfer events of the individual genes within the operon were detected, and different functional constraints were found within and between the Rep proteins. When different repABC operons coexisted in the same genome, they were well differentiated from one another. Finally, we found different levels of adaptation to the host genome within and between repABC operons coexisting in the same species.
Horizontal gene transfer with conservation of the repABC operon structure provides a highly dynamic operon in which each member of this operon has its own evolutionary dynamics. In addition, it seems that different incompatibility groups present in the same species have different degrees of adaptation to their host genomes, in proportion to the amount of time the incompatibility group has coexisted with the host genome.
The repABC plasmids are a typical genome component of many Alphaproteobacteria species. In fact, more than 20 Alphaproteobacteria species have at least one repABC plasmid (see refs [1, 2] for recent reviews), these repABC plasmids may be the commonest plasmids in Alphaproteobacteria species. In some species these repABC plasmids constitute a significant amount of the bacterial genome; such is the case of Rhizobium leguminosarum 3841, in which repABC plasmids account for 35% of the genome . This plasmid family includes several incompatibility groups, meaning that more than one type of repABC plasmid can reside in the same bacterial species [1, 2]. For instance, Rhizobium etli CFN42 has 6 plasmids, all of them repABC plasmids . In contrast to other low copy-number plasmids, in which the elements involved in plasmid replication and segregation are located on different loci (each one under its own regulatory circuit), the repABC plasmids contain all the elements required for replication and partition within the repABC operon. In general, this transcriptional unit comprises three protein-encoding genes (repA, repB, and repC) and a gene encoding a small antisense RNA (ctRNA) , which is located within the repB-repC intergenic region. The proteins encoded in the repABC operon have an intricate relationship, with RepA and RepB interacting both with themselves and with each other. These proteins, in conjunction with the centromere-like sequence, parS, function as the plasmid's segregation machinery [1, 2, 6]. On one hand, RepA is a transcriptional repressor of the operon, while RepB acts as its co-repressor by contacting the operator sequence. The third protein-encoding gene of the operon, repC, is essential for plasmid replication; it encodes the initiator protein, RepC, which exerts its function by binding the origin of replication located within its own coding sequence [1, 2, 6]. Taking these observations into account, it is reasonable to hypothesize that the repABC operon is under concerted evolutionary pressures aimed at maintaining its functionality and avoiding incompatibility with other repABC operons. Remarkably, this operon is not only the replication system of repABC plasmids, but of some secondary chromosomes of some Rhizobiales species. For instance, the second chromosomes of Agrobacterium vitis S4 and Agrobacterium tumefaciens C58 have a repABC origin of replication .
At the structural level, the various repABC operons are only superficially homogeneous; they are highly diverse in DNA sequence, and some possess specific structural elements shared only by few members of the family. These distinctive elements fall into three types: (a) the number and class of regulatory elements involved in operon transcription; (b) the number and position of centromere-like sequences (parS sequences); and (c) the presence of peptide-encoding minigenes . Several Alphaproteobacteria genomes possess repAB genes that are not in close association with the ctRNA or repC sequences. However, it has been shown that replication of some Alphaproteobacteria plasmids depends only on RepC and a ctRNA, without the involvement of the repAB genes. This suggests that fusion of different modules could participate in the generation of new repABC plasmids, indicating that the different elements may have experienced different evolutionary histories.
Plasmid stability requires an exquisite balance among all of the interacting molecules involved in plasmid replication and segregation. Perturbation of this balance, for example by the introduction of any replication or segregation element in excess, could lead to plasmid incompatibility. It has been shown that repABC plasmids contain at least four elements involved in plasmid incompatibility: the RepA and RepB proteins, the small antisense RNA, and the parS sequences [6, 8–10]. Phylogenies made with RepA, RepB, and RepC proteins have shown that different replicons residing in the same bacterial strain tend to belong to different clades . Other study found that phylogenetic analyses of repABC gene lineages had a lack of evolutionary congruence with the species tree . These observations suggest that divergent evolution followed by episodes of horizontal transfer have played a central role in originating new incompatibility groups. We might therefore expect that incompatibility groups residing in the same genome would be different enough so as to not interfere with each other.
In this study, we analyzed three aspects of repABC operons. First, because it is known that repABC operon has been horizontally transferred, through phylogenetic analyses, we examined horizontal gene transfer of entire operon versus horizontal transfer of individual genes within this operon. This is a key point, since a previous study has shown that some bacterial operons present horizontal gene transfer events that affect not the entire operons but single genes within the operons . Second, we determined the degree of differentiation among repABC operons from different plasmids residing in the same strain (which implies different incompatibility groups). Third, we established the degree of evolutionary adaptedness among different repABC operons coexisting in a single species. In principle, because all the elements of the partition and replication systems are contained in the same operon and the encoded proteins interact, these elements might be expected to present almost the same history. Contrary to this, we found significantly different histories for the various elements of the repABC operon. Moreover, we detected different selective constraints among the elements composing the operon, and even within individual components. As expected, when different incompatibility groups coexisted in a species, these groups were clearly differentiated from one another. Finally, we found different levels of adaptation to the host genome within and between repABC operons coexisting in the same species.
The collection of homologous repABC operons
To date, at least 81 repABC operons have been recognized across the class Alphaproteobacteria . Because we wanted to utilize only homologous groups with the same domain structure, we established strict criteria for defining homologous rep genes and operons (see Methods). As a result of this, we analyzed only 49 operons herein (see Additional file 1). Twenty-one genomes had at least one repABC operon, and most of the operons were located on plasmids. A few genomes, such as those from genera Brucella and Agrobacterium, had repABC operons located on replicons that are considered secondary chromosomes (see Additional file 1). Two Rhizobium species, R. etli CFN42 and R. leguminosarum 3841, had the highest number of repABC operons, with seven operons each. All plasmids from these species had a single operon, with the exceptions of plasmid p42f from R. etli CFN42 and plasmid pRL11 from R. leguminosarum 3841, which each had two operons per plasmid. We also found six faulty operons that were missing one of the three protein-encoding genes; five out of six were composed of repA and repB genes, while the remaining one consisted of repA and repC. In four of six cases, the faulty operons coexisted with complete operons. In many species only one gene was present; by far the most widely distributed gene was repA, followed by repC (see Additional file 1).
There is no a single history for the repABC operon
Different levels of functional restriction within and between Rep proteins
Estimates of the best amino acid models for the individual Bayesian phylogenies
Shape parameter α
P. Invariant sites
11.987 (11.128 12.906)
0.933 (0.776 1.102)
0.065 (0.022 0.107)
WAG (PP 1)
19.952 (18.617 21.323)
1.721 (1.487 1.983)
0.0698 (0.041 0.103)
WAG (PP 1)
17.678 (16.49 18.922)
1.122 (0.993 1.265)
0.068 (0.040 0.098)
JTT (PP 1)
Well differentiated incompatibility groups
Average between-locus distance for the different loci
P. Invariant sites
Shape parameter α
Codon Adaptation Index as a measure of evolutionary adaptedness
The repABC operon is not only important because it is the replication-partition system of repABC plasmids, a common component of Alphaproteobacteria species, but because it is also the replication-partition system of some secondary chromosomes in Alphaproteobacteria species. Our present analyses functioned at two levels: within the repABC operon and between repABC operons in those cases where several repABC operons coexisted in the same genome. We did not find a single history within the repABC operon; clearly, each protein had its own phylogeny. This is somewhat surprising, since repA, repB, and repC form an operon, and it would seem that they should have similar histories if the entire operon had been horizontally transferred. Instead, even RepA and RepB, which compose the partition system and physically interact, had different phylogenies. This contrast with a recent work in which relaxase sequences were used as tools for classification of conjugative systems. In that study it was found that relaxases and the IV coupling proteins (T4CP), which map next to each other and belong to a minimal gene set that allows plasmid to be conjugally transmitted, evolve congruently for long periods of time . Thus, it seems that compared with some elements of the transfer machinery the repABC replication-partition system is highly diverse.
Quite notably, every single gene of this operon presented evidence of horizontal gene transfer. In situ gene displacement is a likely process behind this, since the structure of the repABC operon is completely conserved. We think in situ gene displacement could have occurred through homologous recombination, as we found homologous recombination events across the 3 rep genes. Although in situ gene displacement appears unlikely, there is evidence that shows that this process is not that scarce. Omelchenko et al found that within the bacterial operons they had analyzed in situ gene displacement was a frequent event . A striking difference between in situ gene displacement and other types of horizontal gene transfer events is that the former leaves intact the operon structure, so that, the operon is completely functional.
The proteins differed not only at the topological level, but also at the level of functional restriction. RepA and RepC, which belong to different systems, were under similar levels of functional restriction, suggesting that key elements of the partitioning and replication systems are under similar functional restrictions. In contrast, RepB had a very different level of functional restriction. We also found different levels of functional restrictions within proteins. For example, the ATPase domain of RepA (Figure 3, family MipZ), which forms a complex with the chromosome partitioning protein and is indispensable for partitioning, presented the lowest substitution rates. As well the only recombination event presented in repA did not affect the ATPase domain but a relatively unconserved part of the gene. Therefore, it seems that the different proteins, and even the different parts of the proteins themselves, are under different functional and/or structural constraints. Of the three genes studied, repA was the most conserved and might have the highest expression level. This is not unexpected, as RepA is known to have several functions, and its expression is required in both the presence and absence of partition, suggesting the need for high-level translation in order to maintain sufficient RepA levels. In contrast, repC, which is a replication initiator protein, had the lowest CAI values, perhaps due to the higher levels of homologous recombination in this gene (see below). Horizontal gene transfer could be very important in allowing the variability of this operon. Indeed, if horizontal gene transfer had not affected the genes within the operon, these genes would have to have a single evolutionary history. Instead, we found that the reverse was true. The proteins encoded in those genes not only presented different phylogenies, they also had different functional restrictions, even within the proteins themselves, and the CAI values differed among the genes. Given the presence of differences at several levels, it is very logical to think that horizontal gene transfer has unconnected the various portions of the operon, allowing each part to have a particular evolutionary history. In this way, genes with very different functional restrictions could be located next to each other, as seen for repB and repA.
The existence of multiple repABC operons located on different replicons in the same genome implies the presence of different incompatibility groups. We herein showed that when multiple repABC operons coexisted in the same genome, they were well differentiated from one another. We did not find evidence of homologous recombination in these cases; this is not unexpected, since homologous recombination would homogenize the sequences, meaning that the different groups would no longer be compatible with each other. The intergenic region, which encodes a small antisense RNA (a very important determinant for incompatibility), was highly conserved and found to be under high functional restriction, yet it did not have any invariant sites. Although this sequence has changed only minimally due to functional restrictions, it has still accumulated sufficient changes to allow the coexistence of the different incompatibility groups. In agreement with our within-operon analysis, repA and repC were highly conserved, with repC being the most highly conserved between operons (it had the smallest average distance). As mentioned above, repC also had the most homologous recombination events. This suggests that homologous recombination might be reducing the divergence of repC, potentially also explaining the low CAI values for this gene (homologous recombination would be erasing any improvement in the CAI values). In a report on the genome sequence of R. leguminosarum, Young and coworkers suggested that a recent recombination event had taken place, and divergence of RepC was not critical for plasmid compatibility . Here, one of the recombination events detected in repC involved the sequence from pRL8, which is a plasmid of R. leguminosarum 3841.
Different repABC operons had distinct levels of adaptation to their host genome, with no two repABC operons presenting the same CAI values. We think that amelioration might be playing a role in the adaptation of repABC operons to their hosts. Plasmids p42a and p42d were suggested to be newly acquired plasmids based on their lower GC values, poor conservation, and poor functional connectivity with the rest of the genome . These two plasmids had the worst CAI values, implying that they are not well adapted to their host's genome. In contrast, the operon from p42f, which appeared to be the oldest plasmid harbored within R. etli CFN42, had the highest CAI values, suggesting that this operon is highly adapted. These findings indicate that the longer a repABC operon coexists with its host genome, the more adapted the operon becomes. This may result in more effective replication and partitioning processes. As well plasmids, which had the most adapted operons, presented essential genes as well; for instance plasmids pRL11, pRL12, and pRL10, which all have essential genes , had the operons with higher CAI values than the rest of plasmid of R. leguminosarum 3841.
In summary, we herein report finding different histories and functional constraints within the repABC operon. In addition, when multiple repABC operons were present in the same genome, they had different levels of adaptedness to the host genome, and this seems to be related to the length of time each operon had been associated with the host genome. Finally, horizontal gene transfer with conservation of the operon structure provides a highly dynamic operon in which each member could have its own evolutionary dynamics.
Detection of homologous genes and operons
We first identified the homologous of the RepA, RepB, and RepC proteins across the known Alphaproteobacteria genomes (see Additional file 6). The RepA, RepB, and RepC proteins from symbiotic plasmids of R. etli CFN42 and S. meliloti 1021 were used as seeds, and were queried against the proteomes encoded by the other genomes (Additional file 6), using BLAST  with an E-value cutoff of 1.0e-12. We retained all cases where a seed protein had a hit in any other proteome and the proteins aligned along at least 70% of their lengths. We then selected for DNA sequences wherein repA was next to repB, and repB was next to repC (by definition, the only gene between repA and repC was repB), this was taken as a complete operon. The homologous protein groups contained only proteins whose genes formed complete operons. For each homologous protein group, we constructed an alignment with MUSCLE , and used this alignment to infer a phylogeny (see below). To generate the DNA alignments of repA, repB, and repC, we used their protein alignments as references, and performed nucleotide alignment using the "tranalign" program from The European Molecular Biology Open Software Suite (EMBOSS) . The recombination analysis was carried out on these DNA alignments.
Other sets of DNA alignments were created for each of the operons contained in R. etli CFN42 and R. leguminosarum 3841. The intergenic region between repB and repC was also considered. We then used jModelTest  to carry out statistical selection of the best-fit models of nucleotide substitution for every DNA alignment. Finally, maximum likelihood distance matrices were inferred using the model specifications from jModelTest; this was done with PUZZLE .
Phylogenies were created using MrBayes v3.1.2 , allowing the MCMC sampler to explore all of the fixed-rated amino acid models included in MrBayes. The number of rate categories for gamma distributions was set to four, with a proportion of sites allowed to be invariable. We performed two runs with four chains each, for 5,000,000 generations. Trees were sampled every 1000 generations, 20% of all generations were removed as burn-in, and a consensus tree was taken. We also estimated the best amino acid models, including the amino acid matrices with the highest posterior probability, estimates of the proportion of invariable sites, and estimates of the gamma shape parameter.
A strict consensus tree was created from all three Bayesian phylogenies, using CONSENSE .
We established the similarities of the phylogenies using the Robinson and Fould distance (RFd), as calculated with TREEDIST .
We used confidence sets to assess whether the differences in topology between the individual Bayesian phylogenies exceeded those expected to occur by chance. We used expected likelihood weighting , which provides a simple and intuitive method for making multiple comparisons of models and constructing the corresponding confidence sets. This test has the benefit of being less conservative than the Shimodaira-Hasegawa test . The topologies tested included those from the RepA, RepB, and RepC phylogenies. PUZZLE  was used to carry out this test for each protein alignment.
Although methods that use the substitution patterns or incompatibilities among sites seem be the most powerful strategy for identifying the presence of recombination events, no single method seems to perform optimally under all different scenarios . Thus, the best strategy is often to use a combination of methods. Here, we used the RDP3 program , which implements a number of methods for identifying recombination events, including GENECONV , RDP , MaxChi , Chimera , SisCan , and Bootscanning . We identified a recombination event as valid when at least three of the six methods indicated positive findings.
Functional regions and among-site rate variation in Rep proteins
We identified the various protein domains by applying the Pfam-A component of Pfam . For this analysis, the RepA, RepB, and RepC proteins of symbiotic plasmid p42d from R. etli CFN42 were queried against Pfam-A. For every position of each protein alignment, a substitution rate was assigned using a discrete-gamma distribution. The discrete-gamma distribution used five rate classes and was implemented through PUZZLE.
Codon Adaptation Index as measure of evolutionary adaptedness
This analysis was done only for the repA, repB, and repC genes located on operons found within species R. etli CFN42 and R. leguminosarum 3841. We used the utility "cusp" from EMBOSS to calculate a codon usage table for the genes encoding the ribosomal proteins in each species. Using these tables as a reference, we applied the "cai" program of the EMBOSS suite to calculate Codon Adaptation Indices for the repA, repB, and repC genes.
We thank Luis Lozano for his comments on the manuscript. SC-R acknowledges a PhD fellowship from CONACyT, and thanks the Computational Unit of the Instituto de Biotecnología, particularly Jerome Verleyen, and the "Macroproyecto de Tecnologías de la Información y la Computación de la UNAM," for the use of their computer facilities. This work was partially supported by PAPIIT Grant IN205808. SC-R would like to extend warm thanks to Valeria Lobos for all the non-academic support, which is the one that most counts.
- Cevallos MA, Cervantes-Rivera R, Gutierrez-Rios RM: The repABC plasmid family. Plasmid. 2008, 60 (1): 19-37. 10.1016/j.plasmid.2008.03.001.View ArticlePubMed
- Pappas KM: Cell-cell signaling and the Agrobacterium tumefaciens Ti plasmid copy number fluctuations. Plasmid. 2008, 60 (2): 89-107. 10.1016/j.plasmid.2008.05.003.View ArticlePubMed
- Young JP, et al: The genome of Rhizobium leguminosarum has recognizable core and accessory components. Genome Biol. 2006, 7 (4): R34-10.1186/gb-2006-7-4-r34.PubMed CentralView ArticlePubMed
- Gonzalez V, et al: The partitioned Rhizobium etli genome: genetic and metabolic redundancy in seven interacting replicons. Proc Natl Acad Sci USA. 2006, 103 (10): 3834-9. 10.1073/pnas.0508502103.PubMed CentralView ArticlePubMed
- Kumar CC, Novick RP: Plasmid pT181 replication is regulated by two countertranscripts. Proc Natl Acad Sci USA. 1985, 82 (3): 638-42. 10.1073/pnas.82.3.638.PubMed CentralView ArticlePubMed
- MacLellan SR, et al: The expression of a novel antisense gene mediates incompatibility within the large repABC family of alpha-proteobacterial plasmids. Mol Microbiol. 2005, 55 (2): 611-23. 10.1111/j.1365-2958.2004.04412.x.View ArticlePubMed
- Slater SC, et al: Genome sequences of three agrobacterium biovars help elucidate the evolution of multichromosome genomes in bacteria. J Bacteriol. 2009, 191 (8): 2501-11. 10.1128/JB.01779-08.PubMed CentralView ArticlePubMed
- Chai Y, Winans SC: RepB protein of an Agrobacterium tumefaciens Ti plasmid binds to two adjacent sites between repA and repB for plasmid partitioning and autorepression. Mol Microbiol. 2005, 58 (4): 1114-29. 10.1111/j.1365-2958.2005.04886.x.View ArticlePubMed
- Ramirez-Romero MA, et al: Structural elements required for replication and incompatibility of the Rhizobium etli symbiotic plasmid. J Bacteriol. 2000, 182 (11): 3117-24. 10.1128/JB.182.11.3117-3124.2000.PubMed CentralView ArticlePubMed
- Venkova-Canova T, et al: Two discrete elements are required for the replication of a repABC plasmid: an antisense RNA and a stem-loop structure. Mol Microbiol. 2004, 54 (5): 1431-44. 10.1111/j.1365-2958.2004.04366.x.View ArticlePubMed
- Cevallos MA, et al: Rhizobium etli CFN42 contains at least three plasmids of the repABC family: a structural and evolutionary analysis. Plasmid. 2002, 48 (2): 104-16. 10.1016/S0147-619X(02)00119-1.View ArticlePubMed
- Omelchenko MV, et al: Evolution of mosaic operons by horizontal gene transfer and gene displacement in situ. Genome Biol. 2003, 4 (9): R55-10.1186/gb-2003-4-9-r55.PubMed CentralView ArticlePubMed
- Strimmer K, Rambaut A: Inferring confidence sets of possibly misspecified gene trees. Proc Biol Sci. 2002, 269 (1487): 137-42. 10.1098/rspb.2001.1862.PubMed CentralView ArticlePubMed
- Yang Z: Among-site rate variation and its impact on phylogenetic analyses. Trends Ecol Evol. 1996, 11 (9): 6-10.1016/0169-5347(96)10041-0.View Article
- Finn RD: The Pfam protein families database. Nucleic Acids Res. 2008, D281-8. 36 Database
- Sharp PM, Li WH: The codon Adaptation Index--a measure of directional synonymous codon usage bias, and its potential applications. Nucleic Acids Res. 1987, 15 (3): 1281-95. 10.1093/nar/15.3.1281.PubMed CentralView ArticlePubMed
- Garcillan-Barcia MP, Francia MV, de la Cruz F: The diversity of conjugative relaxases and its application in plasmid classification. FEMS Microbiol Rev. 2009, 33 (3): 657-87. 10.1111/j.1574-6976.2009.00168.x.View ArticlePubMed
- Altschul SF, et al: Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 1997, 25 (17): 3389-402. 10.1093/nar/25.17.3389.PubMed CentralView ArticlePubMed
- Edgar RC: MUSCLE: a multiple sequence alignment method with reduced time and space complexity. BMC Bioinformatics. 2004, 5: 113-10.1186/1471-2105-5-113.PubMed CentralView ArticlePubMed
- Rice P, Longden I, Bleasby A: EMBOSS: the European Molecular Biology Open Software Suite. Trends Genet. 2000, 16 (6): 276-7. 10.1016/S0168-9525(00)02024-2.View ArticlePubMed
- Posada D: jModelTest: phylogenetic model averaging. Mol Biol Evol. 2008, 25 (7): 1253-6. 10.1093/molbev/msn083.View ArticlePubMed
- Schmidt HA, et al: TREE-PUZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics. 2002, 18 (3): 502-4. 10.1093/bioinformatics/18.3.502.View ArticlePubMed
- Ronquist F, Huelsenbeck JP: MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics. 2003, 19 (12): 1572-4. 10.1093/bioinformatics/btg180.View ArticlePubMed
- Felsenstein J: PHYLIP (Phylogeny Inference Package) version 3.6. 2005, Department of Genome Sciences, University of Washington, Seattle
- Posada D, Crandall KA, Holmes EC: Recombination in evolutionary genomics. Annu Rev Genet. 2002, 36: 75-97. 10.1146/annurev.genet.36.040202.111115.View ArticlePubMed
- Martin DP, Williamson C, Posada D: RDP2: recombination detection and analysis from sequence alignments. Bioinformatics. 2005, 21 (2): 260-2. 10.1093/bioinformatics/bth490.View ArticlePubMed
- Padidam M, Sawyer S, Fauquet CM: Possible emergence of new geminiviruses by frequent recombination. Virology. 1999, 265 (2): 218-25. 10.1006/viro.1999.0056.View ArticlePubMed
- Smith JM: Analyzing the mosaic structure of genes. J Mol Evol. 1992, 34 (2): 126-9.PubMed
- Gibbs MJ, Armstrong JS, Gibbs AJ: Sister-scanning: a Monte Carlo procedure for assessing signals in recombinant sequences. Bioinformatics. 2000, 16 (7): 573-82. 10.1093/bioinformatics/16.7.573.View ArticlePubMed
- Salminen MO, et al: Identification of breakpoints in intergenotypic recombinants of HIV type 1 by bootscanning. AIDS Res Hum Retroviruses. 1995, 11 (11): 1423-5. 10.1089/aid.1995.11.1423.View ArticlePubMed
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.