https://orcid.org/0000-0002-1704-2587
Figures
Abstract
The process of gene birth is of major interest with current excitement concerning de novo gene formation. We report a new and different mechanism of de novo gene birth based on the finding and the characteristics of a short non-coding sequence situated between two protein genes, termed a spacer sequence. This non-coding sequence is present in genomes of Mus musculus, the house mouse and Philippine tarsier, a primitive ancestral primate. The ancestral sequence is highly conserved during primate evolution with certain base pairs totally invariant from mouse to humans. By following the birth of the sequence of human lincRNA BCRP3 (BCR activator of RhoGEF and GTPase 3 pseudogene) during primate evolution, we find diverse genes, long non-coding RNA and protein genes (and sequences that do not appear to encode a gene) that all stem from the 3’ end of the spacer, and all begin with a similar sequence. During primate evolution, part of the BCRP3 sequence initially formed in the Old World Monkeys and developed into different primate genes before evolving into the BCRP3 gene in humans. The gene developmental process consists of the initiation of DNA synthesis at spacer 3’ ends, addition of a complex of tandem transposable elements and the addition of a segment of another gene. The findings support the concept of the spacer sequence as a starting site for DNA synthesis that leads to formation of different genes with the addition of other sequences. These data suggest a new process of de novo gene birth.
Citation: Delihas N (2022) An ancestral genomic sequence that serves as a nucleation site for de novo gene birth. PLoS ONE 17(5): e0267864. https://doi.org/10.1371/journal.pone.0267864
Editor: Ruslan Kalendar, University of Helsinki: Helsingin Yliopisto, FINLAND
Received: January 24, 2022; Accepted: April 17, 2022; Published: May 12, 2022
Copyright: © 2022 Nicholas Delihas. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting information files.
Funding: The author received no specific funding for this work.
Competing interests: The author has declared that no competing interests exist.
Introduction
Protein genes are created by varied processes that include gene duplication [1–5], retrogenes [6] and de novo formation [6–12]. With respect to the latter, Knowles and McLysaght [8] first reported that several human protein-coding genes arose by a de novo mechanism, and Wu et al [9] identified 60 protein-coding genes that are also born by a de novo process. De novo gene birth has been defined as “the process by which new genes evolve from DNA sequences that were ancestrally non-genic” [10]. Less has been reported on origins of long intergenic noncoding RNA (lincRNA) genes. However, some examples are lincRNA genes created from pseudogenized protein genes [13] and lincRNA family genes formed by gene duplication [14]. In addition, the formation of a new human lincRNA gene by transcriptional readthrough has been reported. The work of Rubino et al [15] shows that by use of the transcriptional apparatus of an existing gene and transcriptional readthrough to a small intergenic sequence that represents a functional unit, a new gene is created. This new lincRNA gene is thought to participate in regulation of the immune system. This study has similarities to the work of Shiao et al [16] concerning de novo acquired 3’ UTRs that may play important functions of retrogenes, and that of Stewart and Rogers [17] in terms of the recruitment of non-coding sequences with chromosomal rearrangements and the resultant formation of new protein genes. Thus far, the creation of lincRNA genes appears similar to that of protein genes.
In this paper, a different process of de novo gene birth is presented that involves an ancestral non-coding genomic sequence that does not form part of new genes but is intimately associated with new gene birth. It was previously thought that the lincRNA FAM247 family gene sequence may serve as a nucleation site for new gene formation [18]. By nucleation site we mean a genomic region that serves as a foundation for new gene sequence growth. Presented here is a non-coding DNA sequence, termed a spacer sequence that is attached to the gamma-glutamyltransferase 1 (GGT1) gene. It is present in the rodent house mouse (Mus musculus) and ancestral prosimian primitive primates such as Philippine tarsier (Carlito syrichta) and is evolutionarily conserved in the genomes of all higher primates. It consists of less than 4000 bp, and can contain small sections of the 5’ end of lincRNA FAM247 gene family sequence. We consider this spacer sequence to be a nucleation site for new and multiple gene formations. This is based on finding different genes and incomplete sequences that have no gene annotations in primate genomes, where all sequences/genes stem from the 3’ ends of spacers. Incomplete sequence growth that has been seen in some primate genomes appears to represent a trial and error process in gene formation. The chimpanzee genome provides an example of the combination of the duplication of the spacer sequence and de novo gene formation at the duplicated genomic locus, which is partly analogous to chromosomal rearrangements and the resultant generation of de novo genes described by Rogers and Stewart [17]. Eight experimentally and/or computationally determined genes have been detected that stem from spacer sequences during primate evolution and all sequences start with the elongation of the FAM247 sequence. Discussed also is the possible involvement of the higher order structure of the FAM247 5’ end sequence with DNA elongation during sequence growth.
To help support the concept of the spacer sequence as a nucleation site for de novo sequence/gene development, presented is the evolutionary formation of the human long non-coding RNA gene, the lincRNA gene BCRP3 (the BCR pseudogene 3) and propose a model for the formation of the BCRP3 sequence in the Rhesus monkey. This model helps delineate a generalized model of new gene formation by the initiation and elongation of DNA sequence at the spacer sequence 3’ end and followed by addition of other chromosomal sequences to form new genes. Together with the analysis of the development of the BCRP3 sequence in other primates, this supports the concept of the spacer sequence as a nucleation site for de novo sequence/gene development.
Aside from non-coding RNA genes, it is possible that the ancient GGT5 protein gene, whose sequence also begins with an FAM247 sequence with the protein gene linked to a spacer sequence, is found in non-mammalian ancestors and may also have formed via spacer sequence initiation. The zebrafish genome may be an ancestral example where GGT5 was born, but the spacer sequence significantly diverged during evolution, which makes further assessment of spacer involvement in GGT5 formation difficult.
Results
Properties of spacer sequences
Computational alignment and search programs were used to analyze genomes of primates and other species. The primitive early primate, Philippine tarsier genome was found to display a small genomic spacer sequence (2872 bp) situated between the protein genes GGT1 and GGT5 (Fig 1). The spacer sequence between GGT1 and GGT5 in the house mouse Mus musculus genome is also shown. A large expansion at this genomic region occurred during primate evolution as the Rhesus monkey sequence between genes GGT1 and GGT5 shows an increase in size to 216,200 bp; this sequence expansion is on chr10 and the sequence is also found inverted with chromosomal rearrangements (Fig 1). The chimpanzee genome continued this genomic expansion with an increase to 343,330 bp; the human genome maintained most of this sequence but decreased by ~10%. The spacer sequence between GGT1 and GGT5 of the primitive primate Philippine tarsier is found in the higher primates linked to the GGT1 gene after genomic expansion. The expanded genomic regions contain duplicated sequences that have provided for the formation of a number of new genes and family of genes. Of significance, the GGT1-spacer 3’ ends are found to be chromosomal sites where diverse genes and/or sequences originate in various primate species, the spacer 3’ end serving as the starting point for growth of new sequences and/or genes.
The house mouse and Philippine tarsier (member of ancestral primates) are in the top two drawings. The lengths of genomic regions between genes GGT1 and GGT5 in the higher primates are shown below. The chromosomal region is inverted in Rhesus and other primates. Chromosomal locations are also shown above the schematics. The approximate evolutionary time is on the right. Genomes of these species were analyzed from the NCBI data base (https://www.ncbi.nlm.nih.gov).
Fig 2a depicts several diverse genes in genomic regions that follow the 3’ ends of the GGT1-spacer sequences, and these genes are present in different species. Additionally, in humans, gene duplication of the GGT-spacer motif gives rise to both GGT-related family genes and the FAM247 lincRNA family genes (Fig 2b). In terms of mechanism of initiation and growth of newly formed sequences from spacer 3’ ends, we do not know the source of the FAM247 template or the primer for DNA synthesis, or even if there is a template involved in new DNA synthesis.
a. Diverse genes (GGT5, LOC107973089, LOC112206779, LOC107000612, BCRP3 and FAM247) present in different species that stem from GGT1-spacer sequence 3’ ends. The FAM247 sequence (highlighted in green) is partially or totally present in all genes and/or sequences linked to spacers. The yellow highlighted regions represent Alu/LINE TE tandem repeat arrays. The tan areas contain sequences of the BCR gene. b. Diagrammatic representation of the spacer sequences that lead to gene sequences, The green highlighted regions represent FAM247 sequences present in different genes that start close to the 3’ ends of spacers. GGT-related genes are highlighted in re and BCR-related in tan. There are FAM247 5’ end sequences present in some spacers, e.g., mouse and tarsier but not depicted in the diagram. b. The GGT-spacer-FAM247 gene family sequences present in different segmental duplications in human chr22 [14]. There appears to be no transcript expression from the FAM247 sequence associated with the GGTLC3-spacer, although the sequence has 99.4% identity with lincRNA gene FAM2347A and contains the entire FAM247A sequence [15]. GGT family genes GGT2, GGTLC3, GGT3P, and GGTLC5P are protein or pseudogenes that developed at duplicated loci. 25bp and 354 bp of FAM247 are not in the GGT5 sequences of the mouse and Tarsier, respectively, and 34 and 36 base pairs of the 5’ end of FAM247 are not in the BCRP3 sequence GGT5 genes of Rhesus and humans, respectively.
The 3’ ends of the mouse and tarsier spacer sequences are defined by the start of the GGT5 gene (Fig 1). Because of sequence expansion in the higher primates, the GGT5 gene locations cannot be used to define the ends of spacers. However, we have used the presence of other genes, or the FAM247 sequence where there are no gene annotations to define the 3’ end of spacers in the higher primates. The 5’ half sequence of FAM247 is present in all genes/sequences that stem from spacer 3’ ends; the term FAM247 is used throughout the manuscript to denote the lincRNA FAM247A gene sequence or part of it. The presence of the 5’ end FAM247 sequence is helpful in estimating the 3’ ends of the spacers from species where there are no gene annotations immediately following the spacer but where the FAM247 sequence is present, e.g., in the baboon, gibbon and orangutan genomes. Thus, we can only roughly estimate the 3’ ends of the spacers in most primates. With genes BCRP3 (in humans) and GGT5 (in tarsier) that are attached to spacers, small sections of the FAM247 5’ ends are within the spacers, e.g., positions 1–32 of FAM247 are within the spacer and the FAM247 sequence within the BCRP3 gene starts with FAM247 position 33.
Table 1 shows a high conservation of the overall bp sequence of spacer sequences between the primates, and a 56% sequence identity between the mouse and human spacers. In addition, the spacer 3’ regions display blocks of totally invariant sequences amongst the primates and the mouse (Fig 3, highlighted in light blue,). The functions of these conserved sequences are not known, but because of their invariance over evolutionary time, they may function to initiation gene sequence from the spacer 3’ end.
Alignment of five spacer sequences from mouse, tarsier, Rhesus, chimpanzee and humans. Only section of the spacer 3’ terminal ends, the start of the mouse GGT5 and the start of FAM247 sequences are shown. Spacer terminal ends: Mouse, at 3161 bp; tarsier, 2872 bp; Rhesus, 2933 bp; chimpanzee, 3893 bp; human, 3920 bp. Light blue highlighted, conserved sequence blocks that are conserved in all species analyzed,. Green highlighted, the start of FAM247 sequence. Red highlighted, start of GGT5 gene sequence in the mouse genome. In the higher primates, since GGT5 is distal to GGT1, the start of the GGT5 sequence can not be used to define the 3’ ends of the spacers and the FAM247 sequence has been used. The lengths of spacer 3’ ends vary between species. However, the spacer end of the chimpanzee is shown, i.e., position 3893 bp of the chimpanzee sequence from GGT1.end-FAM247.start.chimp.ref, which ends before the FAM247 sequence begins. In humans, the BCRP3 gene contains the FAM247 sequence but starting with position 33 of the FAM247, with positions 1–32 bp of FAM247 present in the human spacer, therefore we have defined the start of the BCRP3 gene sequence as the human spacer 3’ end. The alignment of the complete sequences used is in S2 Fig.
There are spacer sequences between GGT1 and GGT5 in the zebrafish and opossum genomes, but these have significantly diverged in base pair sequence and do not display conserved sequence blocks; they have not been included in the comparisons in Table 1 or in Fig 3.
Additional sequence conservation is within the spacer 5’ end region, and it is shown in the alignment of the 5’ end spacer gene sequences from all species considered (S1a Fig).This alignment also has an added sequence, the NCBI sequence termed: GGT1 RefSeq, Homo sapiens gamma-glutamyltransferase 1 (GGT1) NG_008111.1 (website: Homo sapiens gamma-glutamyltransferase 1 (GGT1), RefSeqGene on chromosome 22). Note that the GGT1 RefSeq contains the entire GGT1 gene sequence but also includes regions beyond the 5’ and 3’ ends of the gene, for example, the (GGT1), RefSeqGene extends 2010 bp beyond the GGT1 gene 3’ end. The sequence alignment between spacer sequences from different species, with the extended 2010 bp sequence included, show a similarity in sequence from position 1–1451 bp of the (GGT1) RefSeqGene with sequences of the 5’ ends of the spacers, particularly with sequences from the Rhesus to human. Of significance, the distantly related prosimian primitive primate gray mouse lemur spacer shows a particularly high identity (84%) with part of the human (GGT1) RefSeqGene 3’ end sequence (S1b Fig); thus, a segment of the 5’ region of the spacer sequence shows a high evolutionary conservation that spans ~55 million years. The 2010 bp sequence, which follows the GGT1 gene 3’ end, makes up a large portion of the spacer sequence in humans.
GGT5
Of the three experimentally determined genes that are linked to spacer sequences, i.e., GGT5, BCRP3 and the FAM247A-D gene family, GGT5 is the most difficult to analyze in terms of mechanism of formation. The gene is present in the genome of zebrafish, an ancestral vertebrate species that predate the rodents and primates. However, the gene bp and protein aa sequences have significantly diverged over evolutionary time. Although there are small blocks of aa acid sequences such as 280PPPPAGGA287 in the zebrafish GGT5b aa sequence that are totally conserved in all species analyzed, i.e., zebrafish, opossum, mouse and all primates, the overall zebrafish GGT5b gene aa sequence shows an identity of only 48% relative to the human sequence, and showing a poor aa sequence similarity with other GGT5 genes. However, there is a continuum of decline of aa identity relative to the human gene aa sequence during evolution that shows a continuous sequence drift for this gene (S3 Fig). The aa sequence blocks of 100% aa identity, such as the one shown above, may be related to important functional roles of these invariant segments from the GGT5 protein. Included in S3 Fig is the GGT5 aa sequence of the opossum (Monodelphis domestica, gray short-tailed opossum), which is approximately 175 MYA in evolutionary age and thus the evolutionary appearance of the opossum predates the appearance of rodents. Addition of the opossum aa sequence aids in the assessment of the evolutionary changes in GGT5 aa sequence and pattern of change and supports the continuum of evolutionary changes observed. From the GGT5 aa sequences that have been analyzed, the data suggest that the GGT5 genes from zebrafish to humans are evolutionarily related.
Evidence was presented to show that GGT5 exon1 consists entirely of the FAM247 sequence in humans, primates and the mouse, but the FAM247 presence in zebrafish was uncertain [18]. Here we show evolutionary changes of GGT5 exon1 aa sequences, with the opossum exon1 aa sequence included (Fig 4); this helps show the trend in loss of conserved aa found with evolutionary time, but also supports the evolutionary conservation of certain aa positions, which are found to be highly biased in terms of the presence in different regions of the peptide chain (Fig 4). There is a substantial loss of conserved aa residues in the first two thirds of the sequence, but a stability at the carboxyl terminal end of the exon1 sequence where a majority number of aa residues do not change from primates, rodents, opossum and zebrafish (Fig 4). Thus, although the overall percent identity of GGT5 exon1 aa sequences from zebrafish and opossum compared to that of humans is poor, the invariant aa positions of exon 1 and their highly biased locations in the peptide chain suggest an FAM247-type sequence also forms exon1 of zebrafish and opossum GGT5 genes. Development of the zebrafish GGT5 gene’s 5’ end sequence may have begun with the FAM247 sequence, but how the GGT5 sequence was extended and matured to its full sequence during its birth, either in zebrafish or another early ancestor, is not known.
Top. The percent identities between the human exon 1 and other species. Bottom. Amino acid sequences alignment showing tinvariant aa residues with *. Aligned by Clustal2.1 (https://www.ebi.ac.uk/Tools/msa/clustalo/).
BCRP3: Formation of the BCRP3 sequence in Rhesus
In terms of gene expression, the human BCRP3 gene produces one transcript that is expressed primarily in the testes (NCBI https://www.ncbi.nlm.nih.gov/gene/?term=Homo+sapiens+BCRP3) [19]. Structurally, BCRP3 consists of approximately the 5’ half of FAM247A gene sequence, an Alu/LINE TE tandem repeat array, and a copy of a segment of the BCR (the BCR activator of RhoGEF and GTPase) gene sequence [18] (Fig 5a). The BCRP3 gene offers an interesting picture of how a gene sequence was created and evolved in primates over evolutionary time. Using sequence blast searches, the earliest detection of the BCRP3 sequence is in the Old World monkeys, the Rhesus monkey and baboon. In Rhesus, the BRCP3 sequence is found in chr10, linked to the GGT1-spacer at its 3’ end, however the BCRP3 sequence has differences; primarily, it is shorter compared to the human BCRP3 (S4 Fig). The Rhesus BCRP3 sequence contains the 5’ half of the FAM247 sequence (with 88% identity compared to the human BCRP3), significant differences in the Alu/LINE TE tandem repeat array (Table 2) and a sequence segment of the BCR gene sequence with a significantly shorter BCR component compared to the human BCRP3 gene (Fig 5a).
a. A schematic of the composition of the BCRP3 sequence in Rhesus compared to that of the human BCRP3 gene. b. The proposed formation of the BCRP3 sequence in Rhesus, with three steps that involve initiation of synthesis and sequence growth, followed by addition of a complex of TE motifs and ending with addition of a segment of a gene from another part of the genome.
There are no genes annotated at the start of the BCRP3 sequence, but there is a gene stemming from the 3’ end of the Rhesus BCRP3 sequence, the computationally derived protein gene LOC107000612 annotated as breakpoint cluster region protein-like (Fig 5a); it comprises 2889 bp and is homologous to the human BCRP3 3’ end segment with 90% identity. Thus, a putative gene stems from the Rhesus BCRP3 sequence but the major portion of the BCRP3 sequence has no annotations: the Rhesus protein gene LOC107000612 greatly differs from the human lincRNA BCRP3 gene.
A model for the formation of the BCRP3 sequence in Rhesus is described below. Fig 5b graphically shows the proposed model of BCRP3 formation.
Model of the formation of the BCRP3 sequence in Rhesus monkey
- Initiation and growth of the BCRP3 sequence begins at the 3’ end of the spacer with the elongation of the FAM247 sequence up to FAM247 position 5955 bp (Fig 5b, section 1). The spacer may provide signals to initiate synthesis of the FAM247 sequence.
- An array of contiguous Alu/LINE tandem repeats, other TEs, and AT simple repeats are added to the FAM247 sequence (Fig 5b, section 2). Table 2 shows the TE tandem repeats. A search for a copy of a similar Alu/LINE TE tandem array in other parts of the Rhesus genome was negative. This is in contrast to the human Alu/LINE TE tandem array in the BCRP3 gene where an almost identical TE Alu/LINE array is present in the human IGL locus [18]. How the TE tandem repeats were added to the growing sequence in Rhesus is not known. However, there are significant differences with TE insertions and simple repeats between the Rhesus and human TE tandem arrays (Table 2). Also, the tandem repeat in human BCRP3:
AluSg- AluSx1- AluSg- L1MEg- AluSx1- L1MEg- AluSg4 consists of a nearly perfect tandem repeat array with no extraneous base pairs between repeating TEs, which is not the case for the Rhesus array (S5 Fig). This suggests a de novo formation of TE arrays with each species. - The Rhesus BCR (BCR activator of RhoGEF and GTPase) is a large gene of 133735 bp. A small section (6682 bp) of the 3’ end of BCR is copied and transferred to the growing Rhesus BCRP3 sequence (Fig 5b, section 3). The segment of the BCR gene present in the Rhesus monkey is homologous to the human BCRP3 gene sequence (with 82% identity) but its length is shorter than that in the human BCRP3 gene (Fig 5a). A copy of part the Rhesus BCR gene sequence may have been transferred to the growing Rhesus BCRP3 sequence linked to the Rhesus GGT1-spacer. There is also a partial copy of the Rhesus BCR sequence present in the Rhesus IGL locus, but it is unlikely the source of the BCR sequence in the Rhesus BCRP3 as the BCR fragment in the IGL locus is not long enough.
BCRP3: The process of formation of the BCRP3 sequence in the baboon, gibbon and orangutan
The baboon is classified as part of the Old World monkeys and is related to the Rhesus monkey, but diverged ~2 MYA [20]. It has partially developed the BCRP3 sequence at its genomic GGT1-spacer locus but did not progress as far as the Rhesus in sequence development. It has the FAM247 sequence up to FAM247 position 5955 bp at 92% identity with the human BCRP3 gene (compared with the Rhesus BCRP3 sequence at 88%) and has the repeat Alu/LINE TE array (Table 2). The tandem repeats of the baboon are more similar to the human repeats than to those of the Rhesus, but missing in the baboon TE tandem array are an Alu, and MADE1 that are present in the human array at the 3’ end (Table 2). Significantly however, the baboon BCRP3 sequence differs from that of the Rhesus in that it does not have a copy of the BCR gene segment (S6 Fig) but in terms of similarity of the partial sequence, it is closer to the human BCRP3. In addition, there are no genes annotated at the locus where the homologous partial BCRP3 sequence resides in the baboon genome. Thus, there is no apparent explanation for synthesis of the partial BCRP3 other than a failed attempt to synthesize a more complete BCRP3 type sequence or produce a sequence that can encoded a gene. However, the closer identity of the baboon partial sequence with the human BCRP3, compared to the Rhesus sequence identity with the human BCRP3 suggests the Rhesus developed the BCRP3 sequence to form the LOC107000612 protein gene.
Fig 6 summarizes the variety of sequences that stem from the 3’ ends spacer sequences in different species of the superfamily Hominoidea. The gibbons (Nomascus leucogenys (northern white-cheeked gibbon) are part of the family Hylobatidae, a branch of the superfamily Hominoidea (that consists of the human-like apes and humans) but are the lesser apes or small apes. Their evolutionary appearance is ~17MYA. Of major interest, at the GGT1-spacer locus, only part of the FAM247 sequence has formed up to FAM247 position 4467bp, which is shorter than the Rhesus FAM247 sequence at 5955 bp, but it displays a high identity with the human FAM247 sequence (95%). In addition, at this chromosomal locus, the gibbon sequence does not have a Alu/LINE TE tandem array and does not have a copy of the BCR segment of the BCR gene. There are no annotated genes that stem from the partial FAM247 sequence. Thus, it appears to have initiated a partial human FAM247 gene sequence with a high identity with the human FAM247 at the gibbon GGT1-spacer locus, but was unsuccessful in completion of a full FAM247 sequence, the presumed end result.
a. BCRP3 sequence present at loci that contain a duplication of the GGT-spacer motif showing diverse genes stemming from the BCRP3 sequence. b. The presence of the FAM247 sequence at different GGT-spacer loci in gibbon, chimpanzee and human genomes.
The gibbon genome, however, has formed an almost complete BCRP3 sequence, but at another chromosomal locus, a GGT-spacer duplication locus that has the GGT2 gene-spacer sequence (not the GGT1-spacer) (Fig 6a). Although it has a base pair identity of 95% compared to the human BCRP3, there are several gaps in the sequence and one large additional sequence (3307bp) present in the gibbon BCRP3 sequence that is not present in the human BCRP3 gene (S7 Fig). There are two genes annotated within part of the BCRP3 sequence, the gibbon LOC115835989 breakpoint cluster region protein-like and LOC115835847, the putative POM121-like protein 1 (Fig 6a). It appears the gibbon formed a sequence close to that of the human BCRP3 but may have used this sequence to form two genes of its own.
The orangutans are also part of the superfamily Hominoidea and are classified with the great apes. The orangutan appeared evolutionarily about ~9 MYA. Similar to the baboon, the orangutan has formed only a part of the BCRP3 sequence at its GGT1-spacer sequence locus and appears to have “regressed” in capacity to mature the BCRP3 sequence compared to the Rhesus. The sequence includes the FAM247 sequence and the tandem TE repeat array, but does not have the BCR sequence that forms the 3’ end region of the Rhesus BCRP3 sequence (Fig 6a) (S8 Fig). In addition, the partial sequence formed by the orangutan has several small sequence repeats that may represent polymerase stuttering. It also has no putative genes that are annotated within the FAM247-Alu/line TE tandem repeat sequence. Thus, the orangutan, which is evolutionarily more advanced than the Rhesus monkey has not formed the BCRP3 sequence comparable to that of the Rhesus. Similar to the baboon, the orangutan may have come to a “dead-end” in producing a more extended or complete BCRP3 sequence.
The Alu/LINE TE repeat region of the orangutan does have major differences in the middle of the sequence compared to the human Alu/line TE tandem repeat. There are insertions of three copies of an SVA_A retrotransposon and it is missing two LiMEg elements (S9 Fig). SVA insertions are known to affect function [21,22]. The three SVA_A retrotransposon insertions in the orangutan sequence may be related to an inability to form a more complete BCRP3 sequence, however the baboon, which also contains no BCR sequence, does not have retrotransposon insertions in its BCRP3 sequence; thus it is unlikely the retrotransposons are the cause of the partial sequence in the orangutan.
BCRP3: Formation of the BCRP3 sequence in the chimpanzee
Similar to the gibbon, the chimpanzee genome also took an unexpected pathway with respect to BCRP3 sequence development where BCRP3 synthesis occurred at a different chromosomal locus from the GGT1 site of synteny, at a locus that represents a duplication of the GGT-spacer motif (Fig 6a). This locus encodes a glutathione hydrolase light chain 2-like protein gene (LOC100610580) and it is 130,795 bp removed from the GGT1 chromosomal locus of synteny in the chimpanzee genome. There is a full sized BCRP3 sequence formed by the chimpanzee at this duplication site with a high identity, 98% compared with the human BCRP3 sequence (S10 Fig). The chimpanzee has the identical Alu/LINE TE array in its BCRP3 sequence as the human BCRP3 gene except for differences in several subfamilies of Alus, and repeat sequences that are present in the human BCRP3 gene and not in the chimpanzee (Table 2). The major overall differences between the chimpanzee and human BCRP3 sequences are an AluY insertion in the human sequence and an AluSx insertion in the chimpanzee; these Alus are outside of the region containing the repeat Alu/LINE TEs. Thus, the chimpanzee, together with the gibbon, formed the complete the BCRP3 sequence as opposed to the baboon or orangutan, but the chimpanzee formed a sequence that is closer to the human BCRP3 than that of the gibbon.
The duplication locus in the chimpanzee, which has the glutathione hydrolase light chain 2-like protein (LOC100610580) and the BCRP3 sequence, shows a computationally predicted gene annotated as LOC11220671, a breakpoint cluster region protein-like pseudogene (Fig 6a). This gene is 5323 bp in length and has a sequence that is homologous to positions 11710 bp-17033 bp of the human BCRP3 gene; it thus has only about one quarter of the BCRP3 gene sequence. The functions of both of the putative pseudogene LOC112206717 and the human pseudogene BCRP3 gene are unknown.
In humans, an inactive glutathione hydrolase 2 protein (LOC102724197) has been annotated in an Unlocalized Scaffold region of chr22 (NT_187386.1). The NCBI transcript table shows 14 transcripts associated with LOC102724197. The inactive glutathione hydrolase 2 protein gene does have a linked spacer sequence but interestingly, it also has the complete FAM247 sequence instead of the BCRP3 related sequence that is found linked to chimpanzee LOC100610580, the glutathione hydrolase light chain 2-like protein (Fig 6a and 6b). An FAM247 sequence may have formed de novo at this scaffold region of human chr22; alternatively, the FAM247 sequence at this locus may have originated by gene duplication.
FAM247: Formation of complete sequence in chimpanzee and the gene in human
Fig 6b shows a schematic of the FAM247 sequence that is present at the chimpanzee GGT1 locus. This sequence has a 97% identity with the human FAM247A lincRNA gene sequence and contains the entire length of the FAM247A gene sequence (S11 Fig). The chimpanzee GGT1 linked spacer may have served as a nucleation site to initiate FAM247 synthesis, but unlike the synthesis of the BCRP3 sequence, there was continued synthesis of FAM247 until the complete FAM247 sequence was formed. There may be signal(s) directing the addition of TE tandem repeats to an FAM247 growing sequence, but in the absence of such signal(s), there may be continued FAM247 sequence growth. However, why there is a very partial FAM247 sequence in the baboon genome is not understood. Part of the chimpanzee FAM247 sequence is annotated as two genes, LOC107973089, and LOC112206779, and both are termed uncharacterized protein genes. Thus, the computationally derived protein genes in the chimpanzee differ from the human lincRNA genes where both types of genes stem from the same DNA sequence, or part of it.
Human genome chr22 has at least four copies of the FAM247 sequence and three genes that represent the FAM247 A,C, and D lincRNA family (Fig 2b) [15]. The evolutionary relationship of the chimpanzee FAM247 sequence to the human FAM247 long non-coding RNA gene family is unclear. It is not known if there was de novo synthesis of the FAM247 gene in humans and subsequent duplication of the linked sequence by segmental duplications (see Babcock et al for chr22 duplications [23]), or that the sequence was inherited from the chimpanzee and the FAM247 sequence developed in humans into the FAM247 gene with minor mutations and an association with a transcriptional apparatus [24]. With respect to the GGT protein family genes and human chr222 segmental duplications, the GGT1 gene sequence appears to have been modified to form various members of the GGT family at different chromosomal loci that consist of duplications of the GGT-spacer sequence (Fig 6b) [14].
Discussion
In this manuscript we describe the evolutionary development of the BCRP3 sequence in primates, and this developmental process suggests de novo gene birth that arises from an intergenic spacer sequence. This non-coding DNA spacer sequence, which is situated between genes GGT1 and GGT5 in genomes of ancestral prosimian primitive primates remained attached to the GGT1 gene after the large primate genomic expansions and has been conserved during primate evolution. Examples are provided to show that varied sequences and diverse genes stem from the GGT1-spacer 3’ end, or from a duplicated spacer sequence; there are also incomplete sequences that stem from spacer 3’ ends that have no gene annotations. Differences found in primate BCRP3 sequence development, e.g., in FAM247 lengths, different arrays of TE tandem repeats, and varied BCR sequences suggest the start of new sequence growth ab initio (i.e., from the beginning) at each primate spacer site. Functions of invariant base-pairs in the spacer sequence are unknown, however they may be significant in initiation of DNA synthesis.
To further support the concept of the spacer as a gene forming site, finding negative controls would be useful, e.g., finding homologous spacers in the human genome that are not linked to a gene or to the FAM247 sequence. It is likely that a segment of the 5’ end of the FAM247 sequence needs to be attached, or be part of the spacer for initiation and elongation of gene synthesis. Also, locating homologous but mutated spacer sequences that have mutations in one or more of the invariant bp positions, and also do not have a linked gene at the 3’ end of the spacer would support an association of invariant base pairs with gene/sequence formation.
FAM247, whose 5’ side makes up the entire human GGT5 exon 1 sequence, appears to also be present in the zebrafish GGT5 exon 1, based on conserved amino acid analyses. Other data show that the 3’ end of the human FAM247, a sequence, which forms exon 11 and the 3’ UTR of the ubiquitin specific peptidase 18 (USP18) gene transcript [15] is present in zebrafish [18]. Thus, sections of the FAM247 sequence have been present in an early ancestor approximately 300 million years ago. The primate GGT5 gene appears to be descendant from an early ancestor, such as zebrafish, and GGT5 may have initially been born from a spacer sequence starting with an FAM247 type sequence in zebrafish or another ancestor. However, in terms of how the GGT5 gene sequence was elongated and completed, this is difficult to determine with current data.
As the FAM247 sequence formed parts of genes and functional elements during evolution, it would be unusual if this sequence was an isolated example. There should be other sequences that formed parts of multiple, diverse genes and/or functional elements in different life forms during evolution, as well as the presence of other spacer-type sequences.
A model shows how the long non-coding RNA gene, BCRP3 is born in the Rhesus monkey. It consists of initiation of FAM247 sequence growth by the spacer, the elongation of the FAM247 sequence, followed by addition of a complex of tandem transposable elements and ending with the transfer of a copy of the BCR gene segment to the newly formed sequence. The baboon, which together with the Rhesus monkey is part of the Old World monkeys, and the orangutan that is a part of the hominoids (great apes) appear to have both come to a dead end in BCRP3 development and did not progress to the extent of the Rhesus in BCRP3 sequence maturation. Both primate species show a limited BCRP3 sequence with no tandem TEs or other sequences attached. In addition, the partly formed BCRP3 sequences in the baboon and orangutan genomes have no known or predict genes stemming from the partial sequences. The gibbon only formed a partial FAM247 sequence at its GGT1-spacer locus and with no annotated genes predicted to be encoded within the sequence. These examples suggest a trial and error process in BCRP3 and FAM247 sequence maturation for these species.
The final formation of the BCRP3 gene in humans suggests a long-term evolutionary process involving gene development. Guerzoni and McLysaght [11] previously described the de novo formation of primate protein genes over evolutionary time; thus, the process of long-term gene development during evolution may have occured with both protein and non-coding RNA genes.
Interestingly, the chimpanzee developed both the complete BCRP3 and FAM247 sequences and with a high identity of both sequences with the human gene sequences, but these sequences were formed at different chromosomal loci from those found in humans or the sites of synteny. This leaves the unanswered question of how the FAM247 gene sequence was formed in humans, i.e., by inheritance of the FAM247 the sequence from the chimpanzee followed by translocation of the sequence, or by de novo formation of the FAM247 sequence at the human locus having a GGT-spacer sequence and the RNA transcriptional apparatus to form an FAM247 RNA transcript.
The TE ALU/LINE repeats of the BCRP3 sequences have similarities to chromosomal satellite sequences, e.g., HSAT1, an element that was originally found on the Y chromosome but is also present but abundantly found on chr22 [25–27]. How the tandem TE repeats that are present in BCRP3 originated in each primate species is not known. However, McGurk and Barbash [28] have pointed out that formation of tandem arrays may begin as TE dimer insertions followed by expansion to a tandem array. Also of interest are models for the birth of genomic satellite DNA repeats [29], which may pertain to the Alu/LINE TE tandem array seen here.
We do not know the function of the Alu/LINE TE tandem arrays. With centromere and pericentromeric satellites, some play a role in heterochromatin formation in Drosophila and mammals [30]. The BCRP3 gene is situated in a pericentromeric region of human chr 22, which may be relevant. Other and diverse roles of satellites have been outlined [31].
Lastly, the mechanism of initiation of DNA synthesis, the DNA template for FAM247 synthesis, or if there is a template involved is a “black box”. In terms of non-template DNA synthesis, Liang et al [32] studied DNA synthesis with a thermophilic restriction-endonuclease-DNA polymerase and described synthesis of DNA without a template or primer; a role in the development of genes during early evolution was hypothesized. In addition, it was shown that a hyperthermophilic archebacterial DNA polymerase can elongate palindromic and imperfect palindrome tandem repetitive DNA [33]. The FAM247 5’ end sequence begins with a small imperfect palindrome; the sequence then continues to approximately 2000 bp with a number of repetitive base pairs, and then is followed by TEs (S12 Fig). Thus, the higher order structure of the 5’ side sequence of the of FAM247 may be of relevance. With the BCRP3 gene, which contains part of the FAM247 sequence, the imperfect palindrome lies within the spacer sequence as the 5’ side of the FAM247 sequence within BCRP3 starts at bp position 33 bp of FAM247 and positions 1–32 bp are within the spacer and not the gene. The tarsier GGT5-linked spacer contains 357 bp of the FAM247 5’ end sequence and has several perfect palindromes and one imperfect palindrome. Can this suggest template free elongation of FAM247 synthesis via higher order 2nd structure participation? Experimental studies are needed, and the significance the of FAM247 5’ end imperfect palindrome needs to be assessed.
Methods
Primate species genomes
Genomic sequences of species listed were accessed using Home gene NCBI: (https://www.ncbi.nlm.nih.gov/gene) and BLAST Local Alignment (https://blast.ncbi.nlm.nih.gov/BlastAlign.cgi)
Species
Humans, Homo sapiens (NCBI:txid9606)
Chimpanzee, Pan troglodytes (NCBI:txid9596)
Orangutan, Pongo abelii (:Sumatran orangutan) (NCBI:txid9601)
Baboon, Papio anubis (olive baboon) (NCBI:txid9554)
Rhesus, Macaca mulatta (Rhesus monkey) (NCBI:txid9544)
Tarsier, Carlito syrichta (Philippine tarsier) (NCBI:txid1868482)
Lemur, Microcebus murinus (gray mouse lemur) NCBI:txid30608)
Opossum, Monodelphis domestica (gray short-tailed opossum) (NCBI:txid13616)
Mouse, Mus musculus (house mouse) NCBI:txid10090)
Zebrafish, Danio rerio (zebrafish) (NCBI:txid7955)
Gene source
The NCBI/NLM data base was the source of the chromosomal locations of genes, gene annotations and gene sequences of primate and other species, Website: home gene NCBI, https://www.ncbi.nlm.nih.gov/gene
Nucleotide and amino acid sequence alignment programs
The EMBL-EBI sequence analysis program, Clustal Omega Multiple Sequence Alignment (https://www.ebi.ac.uk/Tools/msa/clustalo/) [34] was primarily used for alignments of nucleotide and amino acid sequences as well as determining identities between sequences. The identities represent only aligned sequences and do not including gaps sequences. It should be pointed out that the percent identities can vary in comparisons of homologs with lower similarities.
Pairwise Sequence Alignment, EMBOSS Stretcher (https://www.ebi.ac.uk/Tools/psa/emboss_stretcher/) and EMBOSS Needle [34] were employed for aligning two sequences.
Transposable elements and simple repeat analyses
RepeatMasker (http://www.repeatmasker.org/cgi-bin/WEBRepeatMasker) was employed to determine the TE Alu/LINE repeat sequences. Both search engines rm -BLAST and AB-BLAST were used. Minor differences between results from both search engines did not affect the results or conclusions. An additional related resource is the Dfam data base [35], the data base for repetitive DNA families. It should also be pointed out that there are can ambiguities in annotation of TE subfamilies [36], this was not a problem in comparing TE patterns from different species. Dr. Jessica Storer, Institute for Systems Biology, provided TE and repeat sequence data present in the BCRP3 sequence using an updated RepeatMasker program.
RNA expression
The expression of BCRP3 expression from normal tissues from website: www.ncbi.nlm.nih.gov/gene/, human tissue-specific expression from the New Transcript table subfamilies [19].
Supporting information
S1 Fig.
a. Sequence alignment of the (GGT1), RefSeqGene with the GGT1-spacer sequences from mouse and primate species. b. Alignment of sequence from the Microcebus murinus (gray mouse) lemur with part of the 3’ end sequence the (GGT1), RefSeqGene sequence.
https://doi.org/10.1371/journal.pone.0267864.s001
(PDF)
S2 Fig. Alignment of complete sequences from spacers.
https://doi.org/10.1371/journal.pone.0267864.s002
(PDF)
S3 Fig. Amino Acid sequence alignment of GGT5 from various species.
https://doi.org/10.1371/journal.pone.0267864.s003
(PDF)
S4 Fig. The alignment of the BCRP3 sequence present in the Rhesus locus that contains the sequence between LOC106996293 and GGT1 and human BCRP3 gene.
https://doi.org/10.1371/journal.pone.0267864.s004
(PDF)
S5 Fig. Human BCRP3 gene Alu/LINE TE tandem repeat sequence compared with the Rhesus tandem repeats.
https://doi.org/10.1371/journal.pone.0267864.s005
(PDF)
S6 Fig. Alignment of the BCRP3 sequence from the baboon with the BCRP3 gene and Rhesus BCRP3 sequences.
https://doi.org/10.1371/journal.pone.0267864.s006
(PDF)
S7 Fig. Alignment of the gibbon sequence between GGT2 and GGT1 with the human BCRP3 sequence.
https://doi.org/10.1371/journal.pone.0267864.s007
(PDF)
S8 Fig. Nucleotide sequence alignment of the orangutan sequence that contains the BCRP3 sequence, with the human BCRP3 gene sequence.
https://doi.org/10.1371/journal.pone.0267864.s008
(PDF)
S9 Fig. The orangutan tandem TE repeat array showing three SVA_A insertions.
Data kindly provided by Dr. Jessica Storer.
https://doi.org/10.1371/journal.pone.0267864.s009
(PDF)
S10 Fig. Alignment of the chimpanzee sequence between genes LOC112206721-LOC112206738 (containing the GGT1-spacer duplication locus) with the human BCRP3.
https://doi.org/10.1371/journal.pone.0267864.s010
(PDF)
S11 Fig. Alignment of the chimpanzee sequence between GGT1 and LOC749026 with the FAM247A sequence in humans.
https://doi.org/10.1371/journal.pone.0267864.s011
(PDF)
S12 Fig. RepeatMasker analysis of the FAM247 sequence and Secondary structure predictions of FAM247 imperfect palindrome.
https://doi.org/10.1371/journal.pone.0267864.s012
(PDF)
Acknowledgments
I am grateful to Dr. Jessica Storer, Institute for Systems Biology, Seattle, Washington for providing updated RepeatMasker data for transposable element analyses.
References
- 1.
Ohno S. Evolution by gene duplication. Springer-Verlag.1970. ISBN 0-04-575015-7.
- 2. Ohno S. Gene duplication and the uniqueness of vertebrate genomes circa 1970–1999. Semin Cell Dev Biol. 1999;10(5):517–522. pmid:10597635.
- 3. Korbel JO, Kim PM, Chen X, Urban AE, Weissman S, Snyder M, et al. The current excitement about copy-number variation: how it relates to gene duplications and protein families. Curr Opin Struct Biol. 2008;18(3):366–374. pmid:18511261.
- 4. Conant GC, Wolfe KH. Turning a hobby into a job: how duplicated genes find new functions. Nat Rev Genet. 2008;9(12):938–950. pmid:19015656.
- 5. Larson RT, Dacks JB, Barlow LD. Recent gene duplications dominate evolutionary dynamics of adaptor protein complex subunits in embryophytes. Traffic. 2019; 20(12):961–973. pmid:31518038
- 6. Long M, Betrán E, Thornton K, Wang W. The origin of new genes: glimpses from the young and old. Nat Rev Genet. 2003;4(11):865–875. pmid:14634634.
- 7. Levine MT, Jones CD, Kern AD, Lindfors HA, Begun DJ. Novel genes derived from noncoding DNA in Drosophila melanogaster are frequently X-linked and exhibit testis-biased expression. Proc Natl Acad Sci U S A. 2006; 103(26):9935–9939. pmid:16777968.
- 8. Knowles DG, McLysaght A Recent de novo origin of human protein-coding genes. Genome Res. 2009;19(10):1752–1759. Epub 2009 Sep 2. pmid:19726446
- 9. Wu DD, Irwin DM, Zhang YP. De novo origin of human protein-coding genes. PLoS Genet. 2011;7(11):e1002379. Epub 2011 Nov 10. pmid:22102831.
- 10. Van Oss SB, Carvunis AR. De novo gene birth. PLoS Genet. 2019 May 23;15(5):e1008160. pmid:31120894
- 11. Guerzoni D, McLysaght A. De Novo Genes Arise at a Slow but Steady Rate along the Primate Lineage and Have Been Subject to Incomplete Lineage Sorting. Genome Biol Evol. 2016;8(4):1222–32. pmid:27056411.
- 12. Luis Villanueva-Cañas J, Ruiz-Orera J, Agea MI, Gallo M, Andreu D, Albà MM. New Genes and Functional Innovation in Mammals. Genome Biol Evol. 2017;9(7):1886–1900. pmid:28854603.
- 13. Liu WH, Tsai ZT, Tsai HK. Comparative genomic analyses highlight the contribution of pseudogenized protein-coding genes to human lincRNAs. BMC Genomics. 2017;18(1):786. pmid:29037146.
- 14. Delihas N. Formation of human long intergenic non-coding RNA genes, pseudogenes, and protein genes: Ancestral sequences are key players. PLoS One. 2020 Mar; 15(3):e0230236. eCollection 2020. pmid:32214344.
- 15. Rubino E, Cruciani M, Tchitchek N, Le Tortorec A, Rolland AD, Veli Ö et al. Human Ubiquitin-Specific Peptidase 18 Is Regulated by microRNAs via the 3’Untranslated Region, A Sequence Duplicated in Long Intergenic Non-coding RNA Genes Residing in chr22q11.21. Front Genet. 2021 Feb 3;11:627007. eCollection 2020. pmid:33633774.
- 16. Shiao MS, Khil P, Camerini-Otero RD, Shiroishi T, Moriwaki K, Yu HT. et al. Origins of new male germ-line functions from X-derived autosomal retrogenes in the mouse. Mol Biol Evol. 2007;24(10):2242–53. pmid:17646254.
- 17. Stewart NB, Rogers RL Chromosomal rearrangements as a source of new gene formation in Drosophila yakuba. PLoS Genet. 2019;15(9):e1008314. eCollection 2019 Sep. pmid:31545792.
- 18. Delihas N. Genesis of Non-Coding RNA Genes in Human Chromosome 22-A Sequence Connection with Protein Genes Separated by Evolutionary Time. Noncoding RNA. 2020;6(3):36. : pmid:32899105.
- 19. Fagerberg L, Hallström BM, Oksvold P, Kampf C, Djureinovic D, Odeberg J. et al. Analysis of the human tissue-specific expression by genome-wide integration of transcriptomics and antibody-based proteomics. Mol Cell Proteomics. 2014. 13, 397–406. pmid:24309898
- 20. Zinner D. Keller C. Nyahongo JW, Butynski TM, de Jong YA, Pozzi L. Distribution of mitochondrial clades and morphotypes of baboons Papio spp. (Primates: Cercopithecidae) in eastern Africa. J. East African Nat. Hist. 2015. 104, 143–168.
- 21. Savage AL, Bubb VJ, Breen G, Quinn JP. Characterisation of the potential function of SVA retrotransposons to modulate gene expression patterns. BMC Evol Biol. 2013;13:101. pmid:23692647.
- 22. Vogt J, Bengesser K, Claes KB, Wimmer K, Mautner VF, van Minkelen R, et al. SVA retrotransposon insertion-associated deletion represents a novel mutational mechanism underlying large genomic copy number changes with non-recurrent breakpoints. Genome Biol. 2014;15(6):R80. pmid:24958239.
- 23. Babcock M, Pavlicek A, Spiteri E, Kashork CD, Ioshikhes I, Shaffer LG, et al. Shuffling of genes within low-copy repeats on 22q11 (LCR22) by Alu-mediated recombination events during evolution. Genome Res. 2003;13(12):2519–32. pmid:14656960.
- 24. Wang Y, Liu F, Wang W. Dynamic mechanism for the transcription apparatus orchestrating reliable responses to activators. Sci Rep. 2012;2:422. pmid:22639730.
- 25. Marques-Bonet T.; Ryder O.A.; Eichler E.E. Sequencing primate genomes: What have we learned? Annu. Rev. Genomics Hum. Genet. 2009, 10, 355–386). pmid:19630567
- 26. Kato T, Kurahashi H, Emanuel BS. Chromosomal translocations and palindromic AT-rich repeats. Curr Opin Genet Dev. 2012; 22(3):221±8. pmid:22402448.
- 27. Delihas N. A family of long intergenic non-coding RNA genes in human chromosomal region 22q11.2 carry a DNA translocation breakpoint/AT-rich sequence. PLoS One. 2018; 13(4):e0195702. pmid:29668722.
- 28. McGurk MP, Barbash DA. Double insertion of transposable elements provides a substrate for the evolution of satellite DNA. Genome Res. 2018; 28(5):714–725. pmid:29588362.
- 29. Ahmad SF, Singchat W, Jehangir M, Suntronpong A, Panthum T, Malaivijitnond S, et al. Dark Matter of Primate Genomes: Satellite DNA Repeats and Their Evolutionary Dynamics. Cells. 2020; 9(12):2714. pmid:33352976.
- 30. Shatskikh AS, Kotov AA, Adashev VE, Bazylev SS, Olenina LV. Functional Significance of Satellite DNAs: Insights From Drosophila. Front Cell Dev Biol. 2020;8:312. pmid:32432114.
- 31. Thakur J, Packiaraj J, Henikoff S. Sequence, Chromatin and Evolution of Satellite DNA. Int J Mol Sci. 2021; 22(9):4309. pmid:33919233.
- 32. Liang X, Jensen K, Frank-Kamenetskii MD. Very efficient template/primer-independent DNA synthesis by thermophilic DNA polymerase in the presence of a thermophilic restriction endonuclease. Biochemistry. 2004;43(42):13459–66. pmid:15491153].
- 33. Ogata N. Elongation of palindromic repetitive DNA by DNA polymerase from hyperthermophilic archaea: a mechanism of DNA elongation and diversification. Biochimie. 2007;89(5):702–12. pmid:17321663.
- 34. Madeira F, Park YM, Lee J, Buso N, Gur T, Madhusoodanan N, et al. The EMBL-EBI search and sequence analysis tools APIs in 2019. Nucleic Acids Res. 2019;47(W1):W636–W641. pmid:30976793].
- 35. Storer J, Hubley R, Rosen J, Wheeler TJ, Smit AF. The Dfam community resource of transposable element families, sequence models, and genome annotations. Mob DNA. 2021;12(1):2. pmid:33436076).
- 36. Carey KM, Patterson G, Wheeler TJ. Transposable element subfamily annotation has a reproducibility problem. Mob DNA. 2021;12(1):4. pmid:33485368.