Transposon-derived and satellite-derived repetitive sequences play distinct functional roles in Mammalian intron size expansion.

BACKGROUND: Repetitive sequences (RSs) are redundant, complex at times, and often lineage-specific, representing significant "building" materials for genes and genomes. According to their origins, sequence characteristics, and ways of propagation, repetitive sequences are divided into transposable elements (TEs) and satellite sequences (SSs) as well as related subfamilies and subgroups hierarchically. The combined changes attributable to the repetitive sequences alter gene and genome architectures, such as the expansion of exonic, intronic, and intergenic sequences, and most of them propagate in a seemingly random fashion and contribute very significantly to the entire mutation spectrum of mammalian genomes. PRINCIPAL
FINDINGS: Our analysis is focused on evolutional features of TEs and SSs in the intronic sequence of twelve selected mammalian genomes. We divided them into four groups-primates, large mammals, rodents, and primary mammals-and used four non-mammalian vertebrate species as the out-group. After classifying intron size variation in an intron-centric way based on RS-dominance (TE-dominant or SS-dominant intron expansions), we observed several distinct profiles in intron length and positioning in different vertebrate lineages, such as retrotransposon-dominance in mammals and DNA transposon-dominance in the lower vertebrates, amphibians and fishes. The RS patterns of mouse and rat genes are most striking, which are not only distinct from those of other mammals but also different from that of the third rodent species analyzed in this study-guinea pig. Looking into the biological functions of relevant genes, we observed a two-dimensional divergence; in particular, genes that possess SS-dominant and/or RS-free introns are enriched in tissue-specific development and transcription regulation in all mammalian lineages. In addition, we found that the tendency of transposons in increasing intron size is much stronger than that of satellites, and the combined effect of both RSs is greater than either one of them alone in a simple arithmetic sum among the mammals and the opposite is found among the four non-mammalian vertebrates.
CONCLUSIONS: TE- and SS-derived RSs represent major mutational forces shaping the size and composition of vertebrate genes and genomes, and through natural selection they either fine-tune or facilitate changes in size expansion, position variation, and duplication, and thus in functions and evolutionary paths for better survival and fitness. When analyzed globally, not only are such changes significantly diversified but also comprehensible in lineages and biological implications.

Introduction

Repetitive sequence (RS) elements are characterized as multi-copied sequences in two broadly defined classes: satellite sequences (SSs), including both micro-satellites and mini-satellites, and transposable elements (TEs) that are characterized based on sequence identity and structure, biogenesis, insertion site preference, and degree of redundancies.1,2 The RSs are evolutionarily active and show significant influences on the structures of genes and genomes, and are thus highly relevant to biological functions.3,4 It has been reported that TE-free regions are negatively selected for certain regulatory elements throughout vertebrate genomes, although the conservation of the sequence contents is often variable.5,6 Furthermore, TEs have different distributions among exonic, intronic, and intergenic regions.7 Indeed, a small number of TE classes are still active, generating population differentiation,8 and the compositional dynamics of genomic sequences exhibits step-by-step evolutionary changes as a consequence of competitions between host genomes and parasitic sequences.3 In addition, TE transposition often serves as a driving force for the conversion of introns into exons or gaining novel introns as well as alternatively spliced transcripts.9–11 Therefore, new sequence integration and the balance of exons and introns in number, length, and ordinal position of a gene provide basic materials for species evolution.12

Different subfamilies of TEs have seemingly diverse influences on genes and genomes by changing sequence length to variable extents. Specifically, due to the distinction between “copy-and-paste” of retrotransposons and “cut-and-paste” mostly used by DNA transposons, the former should be a primary player in the event of genome size increase.2 Introns are considered as the major “warehouse” of TEs11,13 and certain families of TEs are observed to correlate with functional genes, such as between mammalian interspersed repeats (MIRs) and immune genes.13 Exploiting the relationship between sequence composition and polymorphism, we noticed that minimal introns (introns in a minimal size range) have unique features distinct from larger introns and demonstrated how these smaller introns escape from TE-driven insertions and also largely free from SS-driven intron expansion.14–16 As many vertebrate genomes have now been sequenced, we are able to address more questions on TE- and SS-driven intron expansions in different vertebrate lineages. In particular, we would like to understand how intron expansion relates to gene functions among the three subgroups of mammals—primates, large mammals, and rodents—and what are the roles of mutation and natural selection played in the course of genome evolution.

Results

Intron size increase often involves lineage-specific changes in RS contents in the context of genes

To investigate the relationship between intron size and repeat insertion in a comparable fashion, we divided introns into ten size intervals for the convenience of in-depth analysis since in general introns tend to cluster at certain size ranges (Fig. 1). According to the relationships among shape-variable curves from the three repeat types, retrotransposons, DNA transposons, and satellites, we found that RSs of the twelve mammals fell into two basic patterns. The first pattern is SS-rich, including three rodent species and two primitive mammals, and its repeat abundance ranks as retrotransposon > satellite > DNA transposon. The second pattern, including the rest of the seven mammals, has a repeat content order of retrotransposon > DNA transposon ≥ satellite (the subequal sign is true only for macaque). In addition, we observed an up-convex curvature of retrotransposon distribution and an up-concave curvature of DNA transposon and satellite distributions with the exception that the curves of satellite distribution in mouse and rat are near-linear, indicating that SSs play a relatively dominant role in their intron size expansion. As to the difference between the non-mammal vertebrates and the mammals, we found that DNA transposons have higher abundance but decreasing slope with intron size increase than the other two patterns in both zebrafish and frog. However, this phenomenon disappears and changes into lower abundance and an increasing slope with intron size increase in anole and chicken. The abundance of retrotransposons is lower than those of satellites in zebrafish, frog, and anole, and the abundance of retrotransposons is higher than that of satellites but the mode of slope remains the same in chicken and the mode of slope changes into descending in all twelve mammals.

Figure 1

Percentage of introns with retrotransposons, DNA transposons, and satellites.

Notes: The fractions of introns with repeats are displayed over intron length intervals. The ten intervals of intron lengths are defined as: 1, (50–150); 2, (151–300); 3, (301–600); 4, (601–1000); 5, (1001–1400); 6, (1401–2000); 7, (2001–3000); 8, (3001–5000); 9, (5001–10000); and 10, (10001+).

We subsequently tried to find the major TE families that influence intron size in each vertebrate species or lineages by calculating the fraction of introns possessing a particular RS class (Table 1). First, SINEs are supreme in overall abundance among all TEs in mammals. In the primates, Alu and MIR are most abundant. In the two small rodents, mouse and rat, B1, B4, and B2 are most abundant, whereas in guinea pig, the larger rodent of the group, B1 and B4 are most abundant. Second, for the four most abundant TE families in each species, the four large mammals, cow, panda, horse, and elephant, share MIR, L1, and L2, as well as other species-specific TEs that include BovA for cow, tRNA-Lys for panda, and SINE:SINEs that are specific for horse and elephant. MIR is abundant in all twelve mammals; opossum and platypus rank as the top two but the three rodents appear behind all the rest mammals. Third, the three lower vertebrates, chicken, anole, and frog, have CR1, Sauria, and Harbinger as the most abundant TEs, respectively. Zebrafish appears to have the most diverse DNA transposons and they are all quite abundant: DNA:DNA, hAT, hAT-Charlie, TcMar-Tc1, En-Spm, hAT-Ac, and Harbinger. Fourth, concerning satellite sequence classes, we found that all SSs are prevalent in the sixteen vertebrates but mouse, rat, zebrafish, and opossum are more SS-rich among all.

Table 1

Percentage of introns with classified into repetitive families.

Class/family	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16
DNA:Chapaev-Chap3	–	–	–	–	–	–	–	–	–	–	–	–	–	13%	–	–
DNA:DNA	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	35%
DNA:En-Spm	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	14%
DNA:Harbinger	–	–	–	–	–	–	–	–	–	–	–	–	–	–	21%	11%
DNA:MER1_type	–	21%	19%	16%	–	21%	–	12%	11%	12%	–	–	–	–	–	–
DNA:T2	–	–	–	–	–	–	–	–	–	–	–	–	–	–	16%	–
DNA:TcMar-Tc1	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	15%
DNA:TcMar-Tigger	10%	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
DNA:hAT	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	23%
DNA:hAT-Ac	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	11%
DNA:hAT-Charlie	21%	–	–	–	19%	–	20%	–	–	–	17%	–	–	–	12%	20%
LINE:CR1	–	–	–	–	–	–	–	–	–	–	23%	–	19%	–	–	–
LINE:L1	27%	27%	27%	27%	27%	27%	26%	23%	20%	20%	30%	–	–	–	–	–
LINE:L2	27%	27%	25%	21%	25%	29%	26%	–	–	13%	36%	54%	–	–	–	–
LINE:Penelope	–	–	–	–	–	–	–	–	–	–	–	–	–	12%	–	–
LINE:RTE	–	–	–	13%	–	–	–	–	–	–	13%	–	–	–	–	–
LINE:RTE-BovB	–	–	–	–	–	–	18%	–	–	–	–	–	–	–	–	–
LTR:ERV1	–	–	–	–	–	–	–	–	–	–	18%	–	–	–	–	–
LTR:ERVL-MaLR	14%	–	–	–	–	–	15%	–	–	–	–	–	–	–	–	–
LTR:MaLR	–	13%	13%	–	–	11%	–	19%	17%	13%	–	–	–	–	–	–
SINE:Alu	49%	49%	51%	–	–	–	–	–	–	–	–	–	–	–	–	–
SINE:B1	–	–	–	–	–	–	–	41%	35%	37%	–	–	–	–	–	–
SINE:B2	–	–	–	–	–	–	–	31%	30%	–	–	–	–	–	–	–
SINE:B4	–	–	–	–	–	–	–	32%	30%	26%	–	–	–	–	–	–
SINE:BovA	–	–	–	36%	–	–	–	–	–	–	–	–	–	–	–	–
SINE:ID	–	–	–	–	–	–	–	12%	22%	–	–	–	–	–	–	–
SINE:MIR	35%	36%	36%	29%	33%	37%	35%	15%	13%	21%	62%	57%	–	–	–	–
SINE:SINE	–	–	–	20%	–	28%	45%	–	–	–	25%	–	–	–	–	–
SINE:Sauria	–	–	–	–	–	–	–	–	–	–	–	–	–	15%	–	–
SINE:V	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	12%
SINE:tRNA-Glu	–	–	–	16%	–	–	–	–	–	–	–	–	–	–	–	–
SINE:tRNA-Lys	–	–	–	–	34%	–	–	–	–	–	–	–	–	–	–	–
Satellite:Satellite	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	16%
Simple repeat:Simple repeat	27%	26%	27%	20%	24%	17%	18%	44%	40%	26%	31%	20%	13%	30%	14%	34%

Notes: The percentages are fractions of introns with the selected repeats over all introns in the listed species and only those greater than 10% are showed in the table. The species codes are: 1, human; 2, orangutan; 3, macaque; 4, cow; 5, panda; 6, horse; 7, elephant; 8, mouse; 9, rat; 10, guinea pig; 11, opossum; 12, platypus; 13, chicken; 14, anole; 15, frog; and 16, zebrafish.

We further identified abundant TE families in each species and have several significant observations (Fig. 2). First, there are near-linear distributions of MIR in introns with a length range of 150 bp–10,000 bp and rapid accumulations of introns over 10,000 bp in the primate and large mammal lineages. In contrast, there is a drastic slowing-down in the rodents, particularly mouse and rat. Aside from this, slowing gains of MIR are also seen in the two primitive mammals. Second, the trends of L1 and L2 insertions over intron sizes are also interesting; the two curves intersect in the large mammals and primates but do not in opossum, where we observe L1 < L2 before and L1 > L2 after the intersections. Third, the distribution of primate-specific Alu repeats has an up-convex curvature, an indication of early saturation and preferred insertions in relatively small introns as compared to LINEs and other SINEs. The rodent-specific B1, in contrast, has a near-linear distribution and is more prevalent than B2 and B4. SINE:ID, unique to mouse and rat, seems more active in rat than in mouse. Fourth, distinctly different from what in other mammals, L2 in platypus behaves similarly to its MIR.

Figure 2

Percentage of introns with selected repeat families.

Note: The intron length intervals are defined in the same way as what in Figure 1.

RS-centric intron expansion involves both size and position effects

To look into distinctive effects of TEs and SSs on intron size and position parameters, we divided introns into four basic classes: TS (both RSs), T (TEs), S (SSs), and N (neither TE nor SS). We focused on three essential intron features: fraction, length, and relative position in a gene. We made the following observations (Fig. 3). First, when plotting the percentage of introns in the four classes, we found that the pattern is rather heterogeneous, ie, the primates, the large mammals, and platypus are grouped together in a pattern of T > N > TS > S, showing a transposon-dominant pattern, so is opossum that has a pattern of T > TS > N > S. Second, mouse and rat form their own group, as it is noticed that both have more satellite sequences than other mammals: TS > N > T > S. Third, aside from the dominant TS-free group or N, guinea pig (N > T > TS > S), frog (N > T > TS > S), and chicken (N > T > TS > S) all have more transposons in their introns than satellites. Fourth, anole and zebrafish have a pattern of N > TS > T > S, in a similar path as compared to mouse and rat regardless of N. If we pick a single most abundant RS-containing intron group, TS, T, S, and N, for a species, the fractions are 39.6%, 52.7%, 12.8%, and 72% in mouse, platypus, anole, and chicken, respectively.

Figure 3

Percentage of the numbers of the four intron classes.

Note: TS, T, S, and N stand for introns with TE and SS, TE only, SS only, and without any of the two basic types, respectively.

We also investigated the size relevance of introns according to two simple size intervals: ≤1000 bp and >1000 bp. Obviously, the absolute majority of introns in N are small, ≤1000 bp, as opposed to the fact that the greater majority of introns in TS and T are larger, >1000 bp. When examining the median length, we found that intron length increase is correlated with the complexity of RS insertions: TS > T > S > N (Fig. 4). We also observed that the TS intron group tends to be near the 5′-end of genes as opposed to the N intron group that tends to be near the 3′-end of the genes in primates, large mammals, rodents, opossum, and frog, as well as that the TS intron group tends to be near the 5′-end of the genes in platypus, chicken, and anole (Fig. 5). The extremely biased distributions are seen in mouse, where the transposon-rich introns tend to be near the 3′-end, and in zebrafish, where all four intron groups show no significant bias.

Figure 4

Length comparison of the four intron classes.

Note: The asterisks indicate significant differences between neighbouring data groups based on Wilcoxon rank sum test and cut-off <0.05.

Figure 5

Position index comparisons for the four intron classes.

Note: The asterisks indicate significant differences between neighbouring data groups based on Wilcoxon rank sum test and cut-off <0.05.

We further examined both length and position effects for four selected transposons: LTR, LINE, SINE, and DNA. Their intron length medians rank as LTR > DNA > LINE > SINE in the primates, the large mammals, and opossum (Fig. 6). In the three rodents, mouse and rat form a unique league themselves with a length order of DNA > LINE > LTR > SINE, but guinea pig stands alone with a similar pattern to other non-rodent mammals: LTR > DNA > LINE > SINE. In addition, the platypus introns with LTR or DNA transposons tend to be larger in size, in comparison with those of LINE- or SINE-containing introns. In contrast, the chicken introns with LINE tend to be smaller, when compared to those with SINE, DNA or LTR. There are other independent patterns such as LTR > SINE > LINE > DNA and LTR > LINE > SINE > DNA in frog and zebrafish, respectively. An exception is unique to anole, where the order becomes LINE > SINE > DNA when LTR is absent. The most likely reason is the lack of well-classified LTR consensus in the RepeatMasker default library due to high diversity of transposable elements in anole, especially when compared to mammals.17 In the primates, the large mammals, and guinea pig, the median position index ranks as LTR < DNA < LINE < 0, and the introns with SINEs in cow, panda, horse, human, and guinea pig have a slight bias toward 5′-end (data not shown). In both mouse and rat, the introns with DNA transposons have the most 5′-end biases and those with SINEs have the least 5′-end biases. In the two primitive mammals, opossum and platypus, their LTRs and DNA transposons tend to be inserted into introns near the 5′-end. The chicken introns harbouring LTRs or DNA transposons have a stronger bias toward insertions at the 5′-end than those with LINE. The order of the median intron position index for anole is LINE < SINE < DNA < 0. The positional preference for the frog introns is the proximity of 5′-end but that of DNA transposon- containing introns is the weakest. In zebrafish, introns with LINE, SINE or LTR have a stronger 5′-end preference, and those with LTR have the least bias.

Figure 6

Length comparisons of the four TE-containing intron classes.

Note: The asterisks indicate significant differences between neighbouring data groups based on Wilcoxon rank sum test and cut-off <0.05.

Intronic RS-abundance and RS-specificity define characteristic gene functions in different mammalian lineages

We first classified genes in a similar way to what we did for introns: (1) TS, genes have both transposons and satellites in their introns; (2) T, genes have only transposons in their introns; (3) S, genes have only satellites in their introns; (4) N, genes have neither transposons nor satellites in their introns. In general, we observed an order of TS > N > T > S in chicken and anole, but a different order of TS > T > N > S in the rest vertebrates. When compared the same RS classes from different species, the most abundant four classes for TS, T, S, and N are 83.1% in mouse, 33% in horse, 8.32% in chicken, and 28.4% in chicken, respectively (Fig. 7). Furthermore, we considered functional categorization of the four gene classes in the four mammalian lineages: mammals, primates, large mammals, and rodents. We found diverse development- and transcription-related functions in S and/or N genes, including “embryonic skeletal system development” and “transcription regulator activity” in mammals (Table 2), “negative regulation of neuron differentiation” and “gene expression” in primates (Table 3), “midbrain development” and “regulation of transcription” in large mammals (Table 4), and “inner ear morphogenesis” and “regulation of gene expression” in the rodents (Table 5). There are also lineage-specific and tissue-specific profiles for the expression of these genes. For instance, “hormone activity” of N genes is shared by all the major groups of mammals and “pheromone binding” of S genes is unique to the rodents. There are also genes with immunological functions identified in the primate S (eg, “positive regulation of chronic inflammatory response to antigenic stimulus”) and N genes (eg, “MHC class I receptor activity”), in S genes of the large mammals (eg, “antigen processing and presentation”), and in N genes of the rodents (eg, “inflammatory response”). In addition, some TS genes are related to fundamental structures and metabolic functions, including “cytoskeleton” and “protein homodimerization activity” in the mammals, “extracellular matrix structural constituent” and “regulation of cell shape” in the primates, “ATP biosynthetic process” in the large mammals, and “acyltransferase activity”, “protein ubiquitination”, and “phosphoinositide binding” in the rodents. There are also rodent TS genes involved in the nervous system and being response to external stimulus or environment. As to T genes, mitochondrial structure related functions are found in both the primates and the large mammals.

Figure 7

Percentage of genes in four classes.

Note: TS, T, S, and N denote genes with TE and SS, TE only, SS only, and with none of the two repeat types, respectively.

Table 2

Mammal-specific GO term enrichment of the four gene classes.

Class	GO code	GO name	1	2	3	4	5	6	7	8	9	10	11	12
TS	GO:0016324	Apical plasma membrane	*	–	–	–	–	–	*	*	*	–	–	–
TS	GO:0005516	Calmodulin binding	*	–	–	*	–	–	–	*	*	–	–	–
TS	GO:0006812	Cation transport	*	–	–	*	–	–	–	*	*	–	–	–
TS	GO:0005856	Cytoskeleton	*	–	–	*	–	–	–	*	*	–	–	–
TS	GO:0005829	Cytosol	*	–	–	–	–	–	*	*	*	–	–	–
TS	GO:0005783	Endoplasmic reticulum	–	–	*	*	–	–	–	*	*	–	–	–
TS	GO:0005887	Integral to plasma membrane	*	–	–	–	*	–	*	*	*	–	*	–
TS	GO:0023034	Intracellular signaling pathway	*	–	–	*	–	–	–	*	*	–	–	–
TS	GO:0005216	Ion channel activity	*	–	*	*	–	–	–	*	*	–	*	–
TS	GO:0008237	Metallopeptidase activity	*	–	–	–	–	–	*	*	*	–	–	–
TS	GO:0042803	Protein homodimerization activity	*	–	–	*	–	–	–	*	*	–	–	–
T	GO:0005576	Extracellular region	–	*	*	*	*	*	*	–	*	–	–	–
S	GO:0030326	Embryonic limb morphogenesis	*	–	–	*	–	–	–	*	*	–	*	–
S	GO:0009954	Proximal/distal pattern formation	*	–	–	*	–	–	–	*	*	*	*	–
S	GO:0030528	Transcription regulator activity	*	–	*	*	*	*	*	*	*	–	*	–
N	GO:0009952	Anterior/posterior pattern formation	*	*	*	*	*	*	*	*	*	*	–	–
N	GO:0048706	Embryonic skeletal system development	–	*	*	–	–	*	*	–	–	*	–	–
N	GO:0048704	Embryonic skeletal system morphogenesis	–	–	*	*	–	*	*	–	*	–	*	–
N	GO:0005576	Extracellular region	*	*	*	–	*	*	*	*	*	*	–	–
N	GO:0005179	Hormone activity	*	*	*	–	–	*	*	–	–	*	–	–
N	GO:0030528	Transcription regulator activity	*	–	*	*	*	*	*	*	*	–	*	–

Notes: The species codes are the same as what listed in Table 1. The asterisks indicate enrichment of GO terms.

Table 3

Primate-specific GO term enrichment of the four gene classes.

Class	GO code	GO name	Human	Orangutan	Macaque
TS	GO:0005201	Extracellular matrix structural constituent	*	–	–
TS	GO:0031965	Nuclear membrane	*	–	–
TS	GO:0008360	Regulation of cell shape	*	–	–
T	GO:0019882	Antigen processing and presentation	*	–	–
T	GO:0019886	Antigen processing and presentation of exogenous peptide antigen via MHC class II	*	–	–
T	GO:0002504	Antigen processing and presentation of peptide or polysaccharide antigen via MHC class II	*	–	–
T	GO:0004004	ATP-dependent RNA helicase activity	*	–	–
T	GO:0005125	Cytokine activity	–	*	–
T	GO:0022625	Cytosolic large ribosomal subunit	*	–	–
T	GO:0010008	Endosome membrane	*	–	–
T	GO:0004308	Exo-alpha-sialidase activity	*	–	–
T	GO:0031640	Killing of cells of another organism	*	–	–
T	GO:0005765	Lysosomal membrane	*	–	–
T	GO:0042613	MHC class II protein complex	*	–	–
T	GO:0032395	MHC class II receptor activity	*	–	–
T	GO:0005763	Mitochondrial small ribosomal subunit	*	–	–
T	GO:0000398	Nuclear mRNA splicing, via spliceosome	*	–	–
T	GO:0005730	Nucleolus	*	–	–
T	GO:0019887	Protein kinase regulator activity	*	–	–
T	GO:0003723	RNA binding	*	–	–
T	GO:0008380	RNA splicing	*	–	–
T	GO:0019843	rRNA binding	*	–	–
T	GO:0005681	Spliceosomal complex	*	–	–
T	GO:0006414	Translational elongation	*	–	–
T	GO:0017070	U6 snRNA binding	*	–	–
S	GO:0004869	Cysteine-type endopeptidase inhibitor activity	–	*	–
S	GO:0044424	Intracellular part	–	–	*
S	GO:0045665	Negative regulation of neuron differentiation	*	–	–
S	GO:0009887	Organ morphogenesis	*	–	–
S	GO:0002876	Positive regulation of chronic inflammatory response to antigenic stimulus	*	–	–
S	GO:0002925	Positive regulation of humoral immune response mediated by circulating immunoglobulin	*	–	–
S	GO:0010843	Promoter binding	*	–	–
S	GO:0007519	Skeletal muscle tissue development	–	*	–
S	GO:0005164	Tumor necrosis factor receptor binding	*	–	–
N	GO:0002474	Antigen processing and presentation of peptide antigen via MHC class I	*	–	–
N	GO:0007267	Cell-cell signaling	*	–	–
N	GO:0009987	Cellular process	*	–	–
N	GO:0010467	Gene expression	*	–	–
N	GO:0008201	Heparin binding	*	–	–
N	GO:0042309	Homoiothermy	–	–	*
N	GO:0050825	Ice binding	–	–	*
N	GO:0048535	Lymph node development	*	–	–
N	GO:0032393	MHC class I receptor activity	*	–	–
N	GO:0000122	Negative regulation of transcription from RNA polymerase II promoter	–	–	*
N	GO:0048663	Neuron fate commitment	*	–	*
N	GO:0005184	Neuropeptide hormone activity	*	–	–
N	GO:0004522	Pancreatic ribonuclease activity	*	–	–
N	GO:0010552	Positive regulation of gene-specific transcription from RNA polymerase II promoter	–	–	*
N	GO:0045084	Positive regulation of interleukin-12 biosynthetic process	*	–	–
N	GO:0045944	Positive regulation of transcription from RNA polymerase II promoter	–	–	*
N	GO:0050826	Response to freezing	–	–	*
N	GO:0016471	Vacuolar proton-transporting V-type ATPase complex	*	–	–

Note: The asterisks indicate significant enrichment of GO terms.

Table 4

Large-mammal-specific GO term enrichment of the four gene classes.

Class	GO code	GO name	Cow	Panda	Horse	Elephant
TS	GO:0006754	ATP biosynthetic process	–	–	–	*
TS	GO:0015662	ATPase activity, coupled to transmembrane movement of ions, phosphorylative mechanism	*	–	–	–
TS	GO:0006821	Chloride transport	–	–	–	*
TS	GO:0007214	Gamma-aminobutyric acid signaling pathway	–	–	–	*
TS	GO:0051536	Iron-sulfur cluster binding	–	–	*	–
TS	GO:0016459	Myosin complex	–	–	–	*
TS	GO:0004725	Protein tyrosine phosphatase activity	–	–	–	*
TS	GO:0005097	Rab GTPase activator activity	*	–	–	*
TS	GO:0032313	Regulation of Rab GTPase activity	–	–	–	*
TS	GO:0048010	Vascular endothelial growth factor receptor signaling pathway	–	–	–	*
T	GO:0022900	Electron transport chain	*	–	–	–
T	GO:0007186	G-protein coupled receptor protein signaling pathway	*	–	–	–
T	GO:0016021	Integral to membrane	*	–	–	–
T	GO:0005743	Mitochondrial inner membrane	*	–	–	*
T	GO:0005747	Mitochondrial respiratory chain complex I	–	*	–	–
T	GO:0005515	Protein binding	–	–	–	*
T	GO:0070469	Respiratory chain	*	–	–	–
S	GO:0019882	Antigen processing and presentation	–	*	–	–
S	GO:0042742	Defense response to bacterium	*	–	–	–
S	GO:0030901	Midbrain development	*	–	–	–
S	GO:0048663	Neuron fate commitment	*	–	–	–
S	GO:0045449	Regulation of transcription	–	*	–	–
N	GO:0022627	Cytosolic small ribosomal subunit	–	–	*	–
N	GO:0016021	Integral to membrane	*	–	–	–
N	GO:0045449	Regulation of transcription	–	*	–	–

Note: The asterisks indicate significant enrichment of GO terms.

Table 5

Rodent-specific GO term enrichment of the four gene classes.

Class	GO code	GO name	Mouse	Rat	Guinea pig
TS	GO:0015629	Actin cytoskeleton	*	–	–
TS	GO:0008415	Acyltransferase activity	–	*	–
TS	GO:0045177	Apical part of cell	*	–	–
TS	GO:0006915	Apoptosis	*	–	–
TS	GO:0030424	Axon	*	*	–
TS	GO:0008013	Beta-catenin binding	–	*	–
TS	GO:0005975	Carbohydrate metabolic process	*	–	–
TS	GO:0007049	Cell cycle	*	*	–
TS	GO:0051301	Cell division	*	–	–
TS	GO:0042995	Cell projection	*	–	–
TS	GO:0009986	Cell surface	*	*	–
TS	GO:0016568	Chromatin modification	*	–	–
TS	GO:0000777	Condensed chromosome kinetochore	*	–	–
TS	GO:0016023	Cytoplasmic membrane-bounded vesicle	–	*	–
TS	GO:0031410	Cytoplasmic vesicle	*	*	–
TS	GO:0030425	Dendrite	–	*	–
TS	GO:0006281	DNA repair	*	*	–
TS	GO:0009055	Electron carrier activity	*	–	–
TS	GO:0005768	Endosome	*	*	–
TS	GO:0009897	External side of plasma membrane	–	*	–
TS	GO:0031012	Extracellular matrix	–	*	–
TS	GO:0005925	Focal adhesion	–	*	–
TS	GO:0005525	GTP binding	*	–	–
TS	GO:0005096	GTPase activator activity	*	*	–
TS	GO:0004386	Helicase activity	*	–	–
TS	GO:0042802	Identical protein binding	*	*	–
TS	GO:0030027	Lamellipodium	*	–	–
TS	GO:0016042	Lipid catabolic process	*	–	–
TS	GO:0042470	Melanosome	–	*	–
TS	GO:0008168	Methyltransferase activity	*	–	–
TS	GO:0005874	Microtubule	*	–	–
TS	GO:0008017	Microtubule binding	*	–	–
TS	GO:0005739	Mitochondrion	*	*	–
TS	GO:0007067	Mitosis	*	–	–
TS	GO:0006397	mRNA processing	–	*	–
TS	GO:0043066	Negative regulation of apoptosis	–	*	–
TS	GO:0043025	Neuronal cell body	–	*	–
TS	GO:0005634	Nucleus	*	*	–
TS	GO:0030165	PDZ domain binding	–	*	–
TS	GO:0048471	Perinuclear region of cytoplasm	*	*	–
TS	GO:0005777	Peroxisome	*	*	–
TS	GO:0035091	Phosphoinositide binding	*	–	–
TS	GO:0043065	Positive regulation of apoptosis	–	*	–
TS	GO:0043123	Positive regulation of I-kappaB kinase/NF-kappaB cascade	*	–	–
TS	GO:0014069	Postsynaptic density	–	*	–
TS	GO:0006813	Potassium ion transport	*	–	–
TS	GO:0042734	Presynaptic membrane	–	*	–
TS	GO:0043234	Protein complex	*	*	–
TS	GO:0032403	Protein complex binding	*	*	–
TS	GO:0019904	Protein domain specific binding	–	*	–
TS	GO:0046982	Protein heterodimerization activity	–	*	–
TS	GO:0019901	Protein kinase binding	–	*	–
TS	GO:0008104	Protein localization	–	*	–
TS	GO:0008565	Protein transporter activity	*	–	–
TS	GO:0016567	Protein ubiquitination	*	*	–
TS	GO:0045449	Regulation of transcription	*	–	–
TS	GO:0006974	Response to DNA damage stimulus	–	*	–
TS	GO:0042493	Response to drug	–	*	–
TS	GO:0001666	Response to hypoxia	–	*	–
TS	GO:0007584	Response to nutrient	–	*	–
TS	GO:0014070	Response to organic cyclic substance	–	*	–
TS	GO:0004871	Signal transducer activity	*	*	–
TS	GO:0005625	Soluble fraction	*	*	–
TS	GO:0015293	Symporter activity	–	*	–
TS	GO:0019717	Synaptosome	*	*	–
TS	GO:0005802	Trans-Golgi network	*	–	–
TS	GO:0006511	Ubiquitin-dependent protein catabolic process	*	–	–
TS	GO:0004842	Ubiquitin-protein ligase activity	–	*	–
S	GO:0009653	Anatomical structure morphogenesis	–	*	–
S	GO:0001658	Branching involved in ureteric bud morphogenesis	*	–	–
S	GO:0045165	Cell fate commitment	*	*	–
S	GO:0042733	Embryonic digit morphogenesis	*	–	–
S	GO:0060441	Epithelial tube branching involved in lung morphogenesis	*	–	–
S	GO:0042472	Inner ear morphogenesis	*	–	–
S	GO:0003676	Nucleic acid binding	–	*	–
S	GO:0048709	Oligodendrocyte differentiation	*	–	–
S	GO:0001569	Patterning of blood vessels	*	–	–
S	GO:0005550	Pheromone binding	*	–	–
S	GO:0008284	Positive regulation of cell proliferation	*	–	–
S	GO:0010552	Positive regulation of gene-specific transcription from RNA polymerase II promoter	*	–	–
S	GO:0045666	Positive regulation of neuron differentiation	*	*	–
S	GO:0010468	Regulation of gene expression	*	*	–
S	GO:0048536	Spleen development	–	*	–
S	GO:0030878	Thyroid gland development	*	–	–
S	GO:0016564	Transcription repressor activity	*	–	–
N	GO:0006935	Chemotaxis	*	–	–
N	GO:0001533	Cornified envelope	*	–	–
N	GO:0006952	Defense response	*	–	–
N	GO:0042742	Defense response to bacterium	*	–	–
N	GO:0005615	Extracellular space	*	*	–
N	GO:0006954	Inflammatory response	*	*	–
N	GO:0007389	Pattern specification process	–	*	–
N	GO:0004252	Serine-type endopeptidase activity	–	*	–

Note: The asterisks stand for significant enrichment of GO terms.

The insertion profiles of TEs and SSs are diverse among the vertebrate genomes

We evaluated the expansion strength of TEs and SSs in introns based on the ratio of the repeat length over the corresponding RS-free length (Table 6). We found that zebrafish has the strongest expansion strength among TS, T, and S genes, whereas chicken has the weakest strength in TS and S genes and anole has the weakest strength in T genes. In the mammals, opossum has the strongest strength in TS and S genes but T genes have the most strength in platypus. A striking observation is the fact that the strength of TS genes is greater than the sum of both T and S genes in the mammals, and we saw the opposite phenomenon in the non-mammalian vertebrates (Table 6).

Table 6

Comparisons of incremental ratio of TEs and SSs.

Species	TS	T	S	T + S
Human	0.833	0.612	0.080	0.693
Orangutan	0.736	0.574	0.067	0.641
Macaque	0.700	0.569	0.063	0.632
Cow	0.661	0.435	0.065	0.500
Panda	0.512	0.384	0.050	0.434
Horse	0.545	0.426	0.059	0.486
Elephant	0.660	0.484	0.094	0.578
Mouse	0.500	0.346	0.071	0.418
Rat	0.458	0.329	0.078	0.406
Guinea pig	0.302	0.244	0.056	0.301
Opossum	0.867	0.556	0.103	0.660
Platypus	0.749	0.623	0.091	0.714
Chicken	0.090	0.183	0.023	0.205
Anole	0.116	0.126	0.035	0.161
Frog	0.330	0.339	0.081	0.420
Zebrafish	1.202	1.207	0.263	1.471

Note: Incremental ratio is defined as X/(1 − X), where X equals to the median length percentage of repeats in introns.

When integrating the content of intronic repeats in individual genes based on orthology (unique homologous gene in each species), we discovered different topological structures (Fig. 8). The shared clusters between the two trees are the human-orangutan and the mouse-rat clades, the distant relationship to chicken, and the approximation of zebrafish to placental mammals as compared to the other three non-mammalian vertebrates. With regard to TEs, the primates and the large mammals are remarkably distinct from the rest species and are closer to the mouse-rat clade as compared to guinea pig. With regard to SSs, opossum is clustered with the primates as well as the rodents and the four large mammals rather than the other primitive mammal, platypus.

Figure 8

Topological trees constructed based on TE (A) and SS (B).

Note: A detailed procedure is described in Methods.

Discussion

Other than whole genome duplication, the complexity of vertebrate genomes builds upon many unique sequence and functional features but one of them is genome expansion that compounds with the expansion of gene and intron sizes. There are three essential ways to increase genome sizes.18,19 The first is to increase the number of genes through genome and gene duplications. The second and also the foremost important mechanism is gene size expansion through intron size and number increases.20 The final way is the expansion of intergenic sequences and auxiliary chromosomal structures. With regard to the diversity of RSs and insertion/expansion mechanisms, we classified intron expansion into two categories: TE-driven and SS-driven,2,21 and speculated that they may play distinct roles in the intron size expansion of mammalian genomes. First, the profiles of TE insertions can be classified at levels of species and lineages, such as primates, large mammals, and rodents, and we did observe similar modes within lineages and distinctions among lineages. However, exceptions do exist as the rodents are not always cohesive—guinea pig behaves differently from mouse and rat concerning many RS counts. Second, we would like to emphasize the effect of RS expansion event rather than copy number counts, and we hope to see a clear and direct picture that correlates intron size variation with RS insertion.

In general, both TEs and SSs are reported to be non-randomly distributed among eukaryotic genomes.1,21–23 On one hand, there is strong negative selection to protect essential sequences in genomes for the transmission of basic genetic information in a relative shorter evolutionary time scale, such as protein-coding sequences or exons. On the other hand, RSs are indispensable as the prime power and raw materials for genomes to evolve for better fitness, to generate complexity and diversity, and to promote speciation and population dynamics.2,24 Therefore, RSs have strong influences on gene expression and regulation indirectly through variations in intron length and content.10,13 One mechanism shared by all the studied vertebrates is that both TE and SS insertions increase intron size but the strength of the former is much greater than that of the latter. In fact, after eliminating RS insertions in all introns, we observed that the tendency of length increase in the four intron classes remains the same. In other words, the large introns remain large in size even without RS insertions in all four intron classes and so do small introns. However, the introns of anole and chicken genomes are exceptional, where the intron size definitions may shift or not be clearly distinguishable between large and small when RS insertions are removed from the intron sequences (data not shown). We observed a non-random and unbalanced expansion mechanism of intron size evolution: larger introns tend to grow faster than smaller ones when introns are enlarged to a certain size or over a specific threshold. Furthermore, we investigated relationship and mechanism of TE- or SS-driven intron expansions. Satellites can increase intron size at an early or primitive stage as they change intron size in a relatively limited scale, but transposons are capable of increasing intron size in a larger (such as LINEs) and more massive (such as LTRs in multiple insertions) scale and thus have stronger influence on intron size expansion. Most importantly, we observed a synergy between TE-driven and SS-driven insertions, providing a greater degree of intron expansion

To understand the possible roles of RS families on gene and intron size expansions, we paid special attention on intron length and positioning within a transcript and on functional enrichment in the context of TE- vs. SS dichotomy among species and lineages. For instance, we found that TS-containing introns have a 5′-end bias in all vertebrates but zebrafish and that the RS-free (or the N class) introns have a 3′-end bias in all mammals but platypus. We have recently identified distinct functional profiles of genes at different evolving rates in primates, large mammals, and rodents,25 and in this study we used a similar classification scheme to investigate protein-coding genes with RS-driven intron expansion. For instance, DNA transposon-containing introns tend to be smaller in fraction, larger in size, and biased toward 5′-end enrichment in mouse and rat. We also pointed out that genes with TE-free introns are enriched in both development and transcription and genes with SS-containing introns are mostly immunity-related in primates and large mammals.13 We also extracted function categories in nervous systems for mammalian genes possessing SS-containing introns since microsatellite alternations may lead to neurological disorders.26 Previous studies proposed that microsatellites are unevenly positioned within different regions of protein-coding genes such as UTRs, exons, and introns, and they may play functional roles in regulating gene expression, splicing, mRNA export, and response to external environment.27 Most SSs that we studied are microsatellites, and we demonstrated that there are functional biases in SS-insertions, such as promoter-related regulatory genes as one of the major categories. In addition, SSs preferentially reside in heterochromatins at or near centromeres and telomeres, where transcriptional activities are rarely discovered. However, if detected, the genes are usually development-related and involved in epigenetic regulation and DNA methylation; the latter two lead to the alteration of chromatin state and may in turn regulate the expression of SS-containing noncoding RNAs.28,29 We concluded that combined or independent effects of species/lineage-specific TEs and SSs may play an important role in functional differentiations of intron-containing protein-coding genes. At present, the sequence-similarity-based RS library is mostly composed of known TEs, especially the collection of mammal-specific sequences. As increasing number of completed high-quality non-mammalian vertebrate genomes are being sequenced, together with the help of de novo identification technologies,30,31 there should be more novel species-specific TEs discovered, adding stronger validation power to the current study.

It is vital for us to track down the precise timing of intron evolution and expansion, such as in a context of lineages, especially the number of introns per gene and the length variation of introns.32 Spliceosomal introns are the great majority in vertebrate genomes, albeit opposing hypotheses on the origin of introns, “intron-early” and “intron-late”, which argue that introns of this particular type is either more ancient or late comers.33 Further analyses on genomes based on taxonomy suggested that intron loss is the dominant phenomenon with position- and phase-specificity in modern mammals and perhaps large amount of intron gains occurred at the early stage of animal evolution,34–36 and recent study has found several cases of intron gains happened in the ancestor of placental mammals in transposon-derived domestication-related genes.37 Moreover, gene length is correlated with gene expression levels and breaths and is affected by RS insertions, such as L1 and MIR.38 Housekeeping genes are often highly-expressed and harbor smaller introns to reduce the processing cost of transcription, including time and energy. In contrast, tissue-specific genes are often lowly-expressed and harbor larger introns, requiring more effective and complex regulatory elements.38,39 Our data, based on a RS-centric stratification approach, showed that intron expansion is strongly influenced by not only RS types but also insertion timing, and the latter is manifested as species-specific propagation of distinct RSs. A comparative study concerning the five teleost genomes indicated that zebrafish experienced an ancient large-scale RS-induced intron expansion, and RS profiles of such expansion is rather distinct from the other four fishes with relatively lower insertion frequency.40 Based on these observations, we suspect that the RS content diversity that we observed among vertebrate introns or genes may not be straightforward to characterize with regard to precise timing as the samples we used are still in a limited scope. Insertions of both TEs and SSs should avoid making damages to key regulatory sequences, such as the splice sites, the branch point, the polypyrimidine tract, and other uncharacterized functional elements, and have potential co-evolving patterns with neighbouring sequences;41 and in particular, TEs (eg, SINEs) facilitate the splicing of larger introns via the formation of secondary structure in mammals.42 TE- and SS-derived RSs are forced to cluster or locate in intronic regions and seldom occur in core regulatory regions that are constantly under strong positive or negative selections.

Methods

We obtained RepeatMasker repetitive elements and Ensembl gene structure annotation data from UCSC Genome Database FTP server (ftp://hgdownload.cse.ucsc.edu/), including those from human, orangutan, macaque, cow, panda, horse, elephant, mouse, rat, guinea pig, opossum, platypus, chicken, anole, frog, and zebrafish (Table 7). We excluded genes that do not encode proteins or have very short introns (<50 bp) from our analysis. For each gene, we only keep the longest primary transcript and/or that has the largest number of exons. Concerning the possible overlapping regions in different repeat families or sub-families, we only counted once when a sequence is used multiple times and otherwise indicated. We also collected the gene-transcript-protein relationship, protein sequences, and Gene Ontology (GO) annotations from Ensembl web or FTP sites (http://www.ensembl.org, ftp://ftp.ensembl.org), and used Fisher Exact Test to find the enriched GO terms and adopted the Bonferroni corrections with a cut-off of 0.1 to reduce false positive rate. To compare the major phylogenic groups in mammals, we regarded the four non-mammalian vertebrates as out-group and considered four divisions (some are obviously lineages and others are not): mammal-specific (occurring only in 12 mammals), primate-specific (occurring only in human, orangutan and/or macaque), non-primate large-mammal-specific (occurring only in cow, panda, horse and/or elephant) and rodent-specific (occurring only in mouse, rat and/or guinea pig). We defined normalized position index as (2*IO-IN-1)/IN, where IO stands for intron order in a gene along the transcription direction and IN is total intron number in a gene. In general, we classified repeat elements into two types of transposons or TE (LTR, LINE, SINE and DNA transposon, in which the former three classes are retrotransposon) and satellites or SS (satellite and microsatellite repeats). We prepared orthologous groups using the inflation parameter = 2 in popular MCL algorithm (http://micans.org/mcl/) to cluster gene families after a protein-based all-to-all-blast with a cut-off of 1e-5.43 And then we only selected the groups containing 16 genes and each gene can be assigned a species for phylogenetic analyses. Finally, we used the fraction of number and length of introns in a unit of gene to evaluate the contents of transposons and satellites for 357 orthologous genes, which form a high-dimensional vector for each species. Furthermore, we used the modified cosine of vector included angle to measure the distance of compared species vectors,44 and adopted a way similar to classical UPGMA (Unweighted Pair Group Method with Arithmetic Mean) clustering technology.45 In brief, we began with the twelve initial species and combined the nearest two neighbor species into one cluster and considered the center of the two points in the space as the new vector of the new node and then repeated the process until all nodes came into one cluster. We employed TreeView program to visualize the result of the tree-like structure.46

Table 7

Species names and the numbers of introns used in this study.

Short name	Full name	Version	Number of introns
Human	Homo sapiens,	hg19	191,918
Orangutan	Pongo pygmaeus abelii	ponAbe2	108,083
Macaque	Macaca mulatta	rheMac2	135,376
Cow	Bos taurus	bosTau4	155,350
Panda	Ailuropoda melanoleuca	ailMel1	136,011
Horse	Equus caballus	equCab2	128,897
Elephant	Loxodonta africana	loxAfr3	127,667
Mouse	Mus musculus	mm9	183,175
Rat	Rattus norvegicus	rn4	154,905
Guinea pig	Cavia porcellus	cavPor3	119,495
Opossum	Monodelphis domestica	monDom5	137,533
Platypus	Ornithorhynchus anatinus	ornAna1	101,406
Chicken	Gallus gallus	galGal3	128,491
Anole	Anolis carolinensis	anoCar2	122,041
Frog	Xenopus tropicalis	xenTro3	136,091
Zebrafish	Danio rerio	danRer7	207,279