Short interspersed elements (SINEs) are a major source of canine genomic diversity

Abstract

SINEs are retrotransposons that have enjoyed remarkable reproductive success during the course of mammalian evolution, and have played a major role in shaping mammalian genomes. Previously, an analysis of survey-sequence data from an individual dog (a poodle) indicated that canine genomes harbor a high frequency of alleles that differ only by the absence or presence of a SINEC_Cf repeat. Comparison of this survey-sequence data with a draft genome sequence of a distinct dog (a boxer) has confirmed this prediction, and revealed the chromosomal coordinates for >10,000 loci that are bimorphic for SINEC_Cf insertions. Analysis of SINE insertion sites from the genomes of nine additional dogs indicates that 3%-5% are absent from either the poodle or boxer genome sequences—suggesting that an additional 10,000 bimorphic loci could be readily identified in the general dog population. We describe a methodology that can be used to identify these loci, and could be adapted to exploit these bimorphic loci for genotyping purposes. Approximately half of all annotated canine genes contain SINEC_Cf repeats, and these elements are occasionally transcribed. When transcribed in the antisense orientation, they provide splice acceptor sites that can result in incorporation of novel exons. The high frequency of bimorphic SINE insertions in the dog population is predicted to provide numerous examples of allele-specific transcription patterns that will be valuable for the study of differential gene expression among multiple dog breeds.

Short interspersed elements (SINEs) are retrotransposons that have accumulated to very high copy numbers in many mammalian genomes. For example, at least 300 Mb (10%) of the human genome is composed of a single family of SINEs, known as Alus (Schmid 1996; Lander et al. 2001; Venter et al. 2001). SINEs accumulate by a “copy and paste” mechanism. Following transcription by RNA polymerase III, the transcripts can be reverse-transcribed and integrated into the genome at distinct locations (Eickbush 1992; Ohshima et al. 1996). There are no known mechanisms for specific removal of inserted SINEs.

SINEs must consume resources of their host for replication, expression, and amplification. In addition, novel transposition events can cause severe disruption of their host's cellular activities (see below). However, it is unclear whether SINEs are primarily intracellular parasites of defenseless host genomes, or if they are symbionts that are tolerated because of their occasional positive influences on genome evolution (Brosius and Gould 1992; Makalowski 2000). They have certainly been implicated in the dynamics of genome evolution, whereby new functional elements appear, and old ones become extinct. First, unequal homologous recombination between Alu elements has clearly contributed to human genomic diversity (Deininger and Batzer 1999). This process appears to underlie the diversification of specific genes (e.g., tropoelastin) (Szabo et al. 1999), or of large genomic regions that encompass multiple genes (e.g., segmental duplications) (Bailey et al. 2003). Second, during retrotransposition of a donor element, transcription past its normal cleavage site can lead to the transduction of 3′-sequences that flank the donor element. Previously this phenomenon has been described only for long interspersed elements (LINEs) (Goodier et al. 2000; Pickeral et al. 2000), although we also see evidence for SINE-mediated transduction of 3′-sequences in the dog genome (see Results and Discussion). Third, transcription of eukaryotic retrotransposons can interfere with expression of neighboring genes (Han et al. 2004). Transcripts of SINEs have also been reported to stimulate protein translation in a response to cellular stress (Schmid 1998; Rubin et al. 2002). Fourth, the insertion of SINEs within genes can have significant effects on mRNA splicing and protein expression. Approximately 75% of human genes contain Alus. There is now abundant evidence that retrotransposition of these elements into exons, or close to mRNA splicing signals, can have dramatic effects on the expression of cellular protein activities (Muratani et al. 1991; Wallace et al. 1991; Vidaud et al. 1993; Janicic et al. 1995; Halling et al. 1999; Mustajoki et al. 1999; Sukarova et al. 2001; Claverie-Martin et al. 2003; Ganguly et al. 2003). It is likely that they can also produce more subtle effects. Indeed, it has been estimated that at least 5% of alternatively spliced exons in the human transcriptome are derived from Alus (Sorek et al. 2002), and processes by which intronic Alus can become “exonized” have been described (Vervoort et al. 1998; Lev-Maor et al. 2003).

Although the human genome contains more than 1 million Alu elements, the vast majority were inserted prior to divergence of the human and ape lineages, and are therefore fixed in the genomes of current primate populations. However, among the ∼5000 young Alus that have integrated during the past 4-6 million years, ∼1200 elements have inserted so recently that they are bimorphic with respect to the presence or absence of insertion in different human genomes (Batzer and Deininger 2002). A genomic locus is considered bimorphic for a SINE insertion if it has two alleles in the general population that are distinguished by the absence or presence of a SINE insertion. A recent analysis of the dog genome, based on survey-sequencing, concluded that recent amplification of canine SINEs has led to a much higher frequency of bimorphic SINE insertions (Kirkness et al. 2003). However, in the absence of a draft genome sequence, the genomic context of these sequence variations was generally unclear. Here, we have extended the initial observations by identifying genomic coordinates and context for more than 10,000 bimorphic SINE insertions. We consider the potential phenotypic consequences of this genomic variability, and the potential utility of SINEs as abundant, evenly distributed polymorphisms that can help us better understand the ancestral relationships between the diverse dog populations of today.

Results and Discussion

Approaches to identify loci that are bimorphic for SINE insertions

For this study, we compared an assembly of survey-sequence data from a poodle genome (derived from 1.5× coverage) (Kirkness et al. 2003), and a draft sequence of a boxer genome (derived from 7.5× coverage) (http://www.genome.gov/12511476). These have been termed CanSS and CanFam1, respectively. The focus of the comparison was a family of SINE elements, termed SINEC_Cf (RepBase release 7.11). The SINEC_Cf repeats comprise a major subfamily of canine-specific SINEs that are likely derived from a tRNA, and contain internal control elements for transcription by RNA polymerase III (Minnick et al. 1992; Bentolila et al. 1999). A SINEC_Cf repeat (∼200 bp) can be distinguished from related SINEs by a two-base insertion (RG) at position 91. Homologous SINEs have been described in a variety of carnivore species (Vassetzky and Kramerov 2002). Previously, we reported that CanSS contains ∼233,000 fragments of SINEC_Cf repeats, with a combined length of 33.8 Mb (Kirkness et al. 2003). As expected for a more complete assembly, RepeatMasker analysis of CanFam1 (http://genome.ucsc.edu/cgi-bin/hgTables) identified fewer SINEC_Cf fragments (∼170,000), but a similar combined length (29.3 Mb). Both analyses indicate that SINEC_Cf sequences represent ∼15% of the total length of all SINEs in the dog genome.

We surveyed a large selection of SINEC_Cf sequences for their distribution between the two sequenced genomes. Using an approach described previously (Kirkness et al. 2003), segments of CanSS that contain full-length SINEC_Cf repeats with flanking sequences were masked for common repeats, and aligned with CanFam1. Loci were considered as potentially bimorphic for SINEC_Cf insertions when both flanks of a CanSS SINE were contiguous on CanFam1 (i.e., the SINE is absent from CanFam1), and the match was unique. The requirement for a unique match is a more stringent criterion than was applied previously, and, as a consequence, heterozygous SINE insertions in the boxer genome were not scored. Examples of such heterozygosity were implied when the flanks of a CanSS SINE could be aligned to two regions of CanFam1 (one on a defined chromosome, and one on the 91 Mb of sequence that has not been assigned to a specific chromosome). However, the requirement for a unique alignment ensured that SINE insertions within any recently duplicated regions of the dog genome were not mistakenly considered as bimorphic insertions. Alignment of 50,500 CanSS segments with CanFam1 revealed 3987 loci (7.9%) that are predicted to be bimorphic for SINEC_Cf insertions (Table 1; Supplemental Table S1). Using the same approach, 92,580 SINE-containing segments of CanFam1 were aligned with CanSS to identify 6575 loci (7.1%) where the SINEC_Cf repeat is absent from the homologous region of CanSS (Table 1; Supplemental Table S1). Again, it should be noted that this is a minimum estimate, as it discounts the 7% of SINEC_Cf insertions that are predicted to be heterozygous in the sequenced poodle genome (Kirkness et al. 2003). In combination, these analyses revealed 10,562 loci for which there is strong evidence of bimorphism for SINE insertions between the two sequenced genomes.

Table 1.

Genomic coordinates of bimorphic SINE insertions

Dog breed	Sequence ID	Start coordinate	End coordinate	Poodle sequence ID	Expected insertion	Boxer chromosome	Coordinate	Orientation
Boxer	chr01	3320493	3320616	AACN010667811	557	chr01	3320493	+
Poodle	CE563274	497	374			chr01	3903727	—
Boxer	chr01	4033961	4034084	AACN010457626	424	chr01	4033961	+
Boxer	chr01	4263197	4263074	AACN010384528	799	chr01	4263197	—
Boxer	chr01	4674003	4673880	AACN010148521	1453	chr01	4674003	—
Poodle	CE666043	360	237			chr01	7738326	+
Poodle	AACN010508306	188	311			chr01	8008980	—
Poodle	AACN010763349	556	433			chr01	8380849	—
English shepherd	TI389899990	168	291			chr01	8480134	+
Boxer	chr01	8688840	8688963	CE514608	246	chr01	8688840	+
Boxer	chr01	8883852	8883975	AACN010090863	287	chr01	8883852	+
Poodle	AACN010699373	517	640			chr01	9062009	+
Boxer	chr01	9483194	9483317	CE826798	292	chr01	9483194	+
Boxer	chr01	9813098	9813221	AACN010538407	224	chr01	9813098	+
Poodle	CE366308	330	207			chr01	10844127	+
Boxer	chr01	11362455	11362578	AACN010063024	1426	chr01	11362455	+
Boxer	chr01	11748564	11748441	AACN010413170	664	chr01	11748564	—
Boxer	chr01	11931616	11931493	AACN010029257	3387	chr01	11931616	—
Poodle	AACN010775895	173	296			chr01	12079479	+
Boxer	chr01	12104226	12104103	AACN010165887	1149	chr01	12104226	—
Boxer	chr01	12141439	12141562	AACN010415886	845	chr01	12141439	+
Poodle	CE792917	540	417			chr01	12168215	+
Boxer	chr01	12285426	12285549	AACN010039890	320	chr01	12285426	+
Boxer	chr01	12591205	12591328	CE574347	179	chr01	12591205	+
Poodle	AACN010707012	641	764			chr01	12615665	+
Poodle	AACN010435963	443	566			chr01	13122073	—
Boxer	chr01	13391958	13391835	AACN010271007	572	chr01	13391958	—
Poodle	AACN010604073	291	168			chr01	13465651	+
Poodle	CE571271	203	326			chr01	14175158	—
Boxer	chr01	14417857	14417980	AACN010113778	158	chr01	14417857	+
Poodle	AACN010184489	1793	1670			chr01	14507734	—
Poodle	AACN010749955	477	600			chr01	14517253	+
Boxer	chr01	14564268	14564391	AACN010271540	329	chr01	14564268	+
Boxer	chr01	14929470	14929593	AACN010717696	548	chr01	14929470	+
Boxer	chr01	15150723	15150600	AACN010916377	44	chr01	15150723	—
Boxer	chr01	15151989	15152112	AACN010425836	1089	chr01	15151989	+
Poodle	CE412993	68	191			chr01	15183132	—
Poodle	CE056395	581	458			chr01	15200679	—
Poodle	AACN010373145	322	199			chr01	15623445	+
Poodle	AACN011003395	69	192			chr01	15849317	+
Poodle	AACN010870015	437	314			chr01	15885701	+
Boxer	chr01	16017574	16017697	AACN010108721	892	chr01	16017574	+
Poodle	AACN010615570	227	104			chr01	16441116	+
Boxer	chr01	16771014	16771137	AACN010816174	568	chr01	16771014	+
Boxer	chr01	16948663	16948786	AACN010808689	337	chr01	16948663	+
Boxer	chr01	17165450	17165327	AACN010826518	634	chr01	17165450	—
Bedlington terrier	TI356819018	473	350			chr01	17632332	—
Poodle	AACN010370372	423	546			chr01	21284948	—

Each SINE is defined by a dog breed, a sequence ID, and the start and end coordinates of the core SINE consensus sequence within the sequence ID. For boxer, the sequence IDs and coordinates refer to CanFam1 (http://genome.ucsc.edu). For poodle, they refer to GenBank accessions (http://www.ncbi.nih.gov/Genbank/); for other breeds, they refer to sequences that are available from the NCBI Trace Archive (http://www.ncbi.nlm.nih.gov/Traces/). For boxer SINEs that are absent from the poodle genome, the table lists a homologous poodle sequence ID, and the coordinate where the SINE insertion would be expected. For all SINEs, the coordinates of the insertion and the orientation of the SINE are mapped relative to CanFam1. The complete table of 11,265 SINEs is provided as Supplemental Table S1.

It was of interest to determine how many additional bimorphic loci may exist in the general dog population that are not revealed by the large collections of sequence data from two specific dogs. In order to address this question, we used ∼1 million random genomic sequence reads that have been generated from nine dogs of different breeds, four wolves, and a coyote (http://www.genome.gov/12511476). Reads that contained SINEC_Cf repeats, and sufficient flanking sequence, were processed and aligned with both CanSS and CanFam1 as described above. For the nine domesticated dogs, analysis of ∼1000 SINEC_Cf repeats per dog indicated that 7.0%-10.9% were absent from CanSS, 6.7%-11.5% were absent from CanFam1, and 2.8%-5.0% were absent from both genomes (Table 2). The corresponding values for novel SINE insertions in the genomes of wild canids overlapped the range for domesticated dogs (4.1%-4.9%; Chinese and Spanish wolves), or were substantially larger (7.9%-9.2%; Californian coyote, Alaskan wolf, and Indian wolf). However, the sample sizes for bimorphic SINE insertions were relatively small for each of the wild canids (15-36), and the differences between them cannot yet be considered as significant.

Table 2.

Survey of bimorphic SINE insertions from random genomic sequence of multiple dogs

Breed	Sequence reads	SINEC_Cf with unique flanks	SINEC_Cf absent from CanSS (%)	SINEC_Cf absent from CanFam1 (%)	SINEC_Cf absent from both (%)
Beagle	99,648	990	69 (7.0)	66 (6.7)	28 (2.8)
Labrador Retriever	99,744	942	78 (8.3)	69 (7.3)	28 (3.0)
German Shepherd	100,743	987	72 (7.3)	75 (7.6)	30 (3.0)
Italian Greyhound	98,208	889	79 (8.9)	72 (8.1)	33 (3.7)
English Shepherd	99,648	954	88 (9.2)	81 (8.5)	38 (4.0)
Bedlington Terrier	102,240	969	106 (10.9)	92 (9.5)	40 (4.1)
Portugese Water Dog	97,728	1013	95 (9.4)	94 (9.3)	42 (4.1)
Alaskan Malamute	100,704	932	96 (10.3)	87 (9.3)	45 (4.8)
Rottweiler	102,143	1040	106 (10.2)	120 (11.5)	52 (5.0)
Chinese Gray Wolf	23,423	169	16 (9.5)	17 (10.1)	7 (4.1)
Spanish Gray Wolf	22,176	185	18 (9.7)	15 (8.1)	9 (4.9)
Californian Coyote	23,790	240	29 (12.1)	28 (11.7)	19 (7.9)
Alaskan Gray Wolf	21,696	220	34 (15.4)	36 (16.4)	20 (9.1)
Indian Gray Wolf	22,560	227	32 (14.1)	29 (12.8)	21 (9.2)

Whole-genome shotgun reads from the NCBI Trace Archive were derived from nine dogs of different breeds, four wolves, and a coyote. Reads that contained SINEC_Cf sequences with nonrepetitive flanks were searched against CanSS and CanFam1, as described in the Methods section.

The preceding analysis indicates that many thousands of SINEC_Cf insertions within canine genomes are not represented in either CanSS or CanFam1. However, the analysis also demonstrates that random genomic sequencing is an inefficient means to identify these novel loci. Only ∼1% of the sequence reads contained sufficient SINE and flanking sequences to permit comparison with the two reference genomes. Consequently, we have developed a methodology that specifically targets SINEC_Cf repeats and flanking sequence for amplification and sequencing. The methodology is centered on a simple inverse-PCR that exploits both the high level of sequence conservation between SINEC_Cf repeats, and the sequence variation between SINEC_Cf repeats and related canine SINEs (Fig. 1). The consensus sequence of SINEC_Cf repeats contains a single, conserved CATG sequence that can be cleaved by the frequently cutting restriction enzyme, NlaIII. After self-ligation of digested genomic fragments, sequences upstream of SINEC_Cf repeats are amplified selectively with primers corresponding to segments of the SINEC_Cf repeat that differ from related SINEs. Cloning of the PCR products yields libraries of SINEC_Cf flanking sequences that can be sequenced readily. Limited sequencing of seven trial libraries demonstrated that >88% of inserts contain SINEC_Cf repeats (with >100 bp of flanking sequence). Importantly, after alignment of these flanking sequences with CanSS and CanFam1, 2.7%-6.8% identified the loci of novel SINEC_Cf insertions. In order to validate these data, 29 of the loci were amplified from the genomic DNA of 8-15 dogs (three to five breeds). Of these, 26 were confirmed as sites of variable SINEC_Cf content, and three were monomorphic for the absence of a SINE in all of the tested dogs (Fig. 2). When scaled up, this simple assay should identify several thousands of novel bimorphic SINEC_Cf insertions, and complement subtractive hybridization approaches that were recently reported to identify bimorphic Alu insertions in human populations (Mamedov et al. 2005).

Figure 1.

Construction of libraries that are enriched for SINEC_Cf elements and flanking sequence. (A) Genomic DNA is cleaved with the frequently cutting restriction enzyme, NlaIII. (B) The cleaved fragments are self-ligated. (C) The circularized products are subjected to PCR using SINEC_Cf-specific primers. (D) The linear products are size-selected and cloned in a plasmid vector. (E) Inserts are sequenced with a vector-specific primer.

Figure 2.

Validation of putative bimorphic SINE insertions. Six representative examples (a-f) of PCR products from loci that were predicted to contain bimorphic SINE insertions after analysis of SINEC_Cf libraries (see Fig. 1). Primers were designed to the flanks of the predicted insertion site, and templates were genomic DNA from a combination of mixed-breed dogs (lane 2), three Bernese Mountain dogs (lanes 3-5), three pugs (lanes 6-8), and three dachshunds (lanes 9-11). Product lengths were derived from DNA markers (lanes M), and were consistent with the presence (+) or absence (–) of a SINE insertion.

The current collections of canine genomic sequence data have therefore revealed >11,000 loci that are bimorphic for SINEC_Cf insertions (Supplemental Table S1). An additional 86,000 loci that contain a SINEC_Cf repeat on at least one allele of CanFam1 and CanSS can also be examined for variability in a wider selection of dogs. Furthermore, the inverse-PCR approach described above permits novel SINEC_Cf insertions to be identified with ∼90-fold higher efficiency than from random genomic sequence reads. Together, these approaches should permit the identification of at least 10,000 additional bimorphic loci in the general dog population. This genomic variability will be a very useful resource for the study of ancestral relationships between different canine lineages. Specifically, bimorphic SINE elements offer two advantages over other types of common polymorphisms. First, the presence of a SINE element represents identity by descent, since the probability that two different young SINE repeats would integrate independently at the same chromosomal location is small. Although there have been a few reports of parallel insertion events (Kass et al. 2000; Cantrell et al. 2001), these are considered to be very rare, at least for primate Alus (Roy-Engel et al. 2002; Salem et al. 2005). Second, the ancestral state of each SINE insertion polymorphism is known to be the absence of the SINE element, and this can be used to root trees of population relationships. In contrast, other types of genetic polymorphism, such as VNTRs and SNPs, can be identical by state if they have arisen from independent parallel mutations at different times and have not been inherited from a common ancestor. Bimorphic Alu-insertion polymorphisms have been used to study human origins, ancestral relationships, and demography (for review, see Batzer and Deininger 2002; Watkins et al. 2003). In some respects, dog breeds resemble geographically isolated human populations, but with a higher degree of isolation, and narrower bottlenecks. In addition, the complex genomic structure of modern dog populations presents specific challenges, owing to the recent origin of most dog breeds (<300 yr), and their derivation from multiple ancestral types (Parker et al. 2004).

Previously, evolutionary studies of canine lineages have focused mainly on variations of mitochondrial DNA or VNTRs (Vila et al. 1997; Savolainen et al. 2002; Koskinen 2003; Parker et al. 2004). These approaches indicate that modern dog breeds were first domesticated from wolves, possibly in East Asia, and that many dog breeds that share morphologies, behaviors, and geographical origins can be segregated by genotype. However, in common with most types of marker, variations of mitochondrial DNA and VNTRs have limitations for evolutionary analyses (Ellegren 2000; Sigurgardottir et al. 2000). Bimorphic SINE insertions offer the advantages of identity-by-descent, and easy typing methodologies, that make these abundant variations a valuable additional resource for identifying the ancestral relationships between different dog breeds, and between domesticated dogs and wild canids.

It is also relevant to note that recent observations of extensive linkage disequilibrium in the dog indicate that association studies to find genes that contribute to diseases and traits could be conducted using as few as 30,000 evenly distributed genomic markers (Sutter et al. 2004). It is conceivable that bimorphic SINEs could provide many of these markers if the throughput for SINE-typing could be scaled up to that currently used for SNPs. One approach for high-throughput typing could use the total amplification of SINEC_Cf flanks (Fig. 1D). The products would be labeled and hybridized to microarrays of oligonucleotides that represent known SINE flanks. By this means, genomes could be scored for the presence or absence of many thousands of SINEs in a single hybridization assay.

Recent LINE activity in the canine genome

Active long interspersed elements (LINEs) are autonomous retrotransposons that likely provide the enzymic activities that are required by SINEs for their propagation (Jurka 1997; Dewannieux et al. 2003). The high frequency of bimorphic SINE insertions in the dog genome may therefore be indicative of highly active LINEs. Identification of active LINEs can pose a major problem for genome assemblies that are based on the whole-genome shotgun approach. Owing to their abundance, similarity, and the fact that they cannot be spanned by individual sequence reads, LINEs are often imprecisely assembled as collapsed contigs. For example, the draft mouse genome contained only 12 full-length LINEs with intact open reading frames (ORFs), although at least 3000 were predicted to exist (Waterston et al. 2002). Similarly, our analysis of CanFam1 revealed only four LINEs with two intact ORFs among the 3226 LINEs that are long enough to be functional (>4.5 kb). The vast majority of these elements contain frameshift mutations or in-frame stop codons. However, at present, we cannot readily distinguish assembly errors from genuine mutations that disrupt the ORFs. Consequently, the number of canine LINEs that are potentially active cannot be estimated reliably.

Another approach to compare recent LINE activity in the dog and human genomes is an analysis of 3′-truncated LINEs. Recent differences in LINE activity should be reflected by detectable differences between the numbers of recent LINE insertions. Owing to the fact that most LINE insertions are 3′-truncations, this analysis is not dependent on precise assembly of full-length elements. We considered the 3′-terminal 500 bases of the youngest known dog and human LINEs (L1_Y_Cf, and L1HS respectively; RepBase Update 9.1). These were aligned with the dog and human genomes, and alignments that spanned at least 98% of the query sequence were categorized by percentage nucleotide identity. For L1_Y_Cf, there were 2700 genomic segments that shared at least 98% identity. For L1HS, the value was 792. That is, among the youngest elements, there are approximately threefold more L1_Y_Cf - like elements in the dog genome, than L1HS-like elements in the human genome. This is consistent with a recent higher level of LINE activity in the dog lineage.

Evidence of recent LINE activity, in the form of bimorphic LINE insertions, is a difficult problem to represent in genome assemblies because the shorter allele (lacking the LINE) is preferentially selected for the final assembly. This is exemplified in the CanFam1 assembly. A BAC clone (GenBank accession no. AC147784.3) from the same dog that was sequenced for CanFam1, contains a full-length L1_Y_Cf (bases 83,981-90,278) with two uninterrupted ORFs. However, although the BAC sequence is represented in CanFam1 (chr29, bases 36,992,169-37,155,933), the LINE is absent. Analysis of the raw sequence reads (from the Trace Archive) that cover the insertion site reveals two alleles in which the element is either absent (e.g., accession nos. 294,160,392 and 285,880,943) or present (e.g., accession nos. 290,601,686 and 237,812,340). This is clearly an example of a bimorphic LINE insertion. We have begun to examine this phenomenon in more detail by using the 3′-ends of LINEs (plus unique flanking sequence) from the poodle genome sequence. When searched against CanFam1, alignments that span the flanking sequence, but not the LINE sequence, indicate potential bimorphic insertions of LINEs. Our preliminary analysis suggests that there are more than a hundred of such candidate regions that can now be tested by amplification of the regions from multiple dog genomes.

The analysis of LINEs in the dog genome has revealed examples of intact elements that are potentially active, and evidence for a higher level of recent activity than in the human genome. It is therefore possible that increased LINE activity has contributed to the relatively high frequency of bimorphic SINE insertions in the canine genome.

Evidence for SINE-mediated transduction of 3′-flanking sequences

When transcription of a retrotransposon fails to terminate at the end of the element, the additional downstream genomic sequence that is transcribed can be mobilized to a new genomic location along with the element. This mechanism of transduction is thought to occur during retrotransposition of LINE-1 elements when the normal polyadenylation signal is bypassed in favor of a second, downstream signal (Goodier et al. 2000; Pickeral et al. 2000). However, transduction of 3′-flanking sequences by SINEs has not been described previously. Analysis of flanking sequences and target site duplications for SINEC_Cf elements in the dog genome revealed several examples of short genomic segments (60-120 bp) that appear to have been transduced during retrotransposition of SINEC_Cfs (Supplemental Table S2). These examples include a short segment of Chromosome 8 that is replicated downstream of a SINEC_Cf at eight other genomic locations. They also include a bimorphic insertion, where a SINE appears to have transposed a short segment from Chromosome 13 to Chromosome 1 of the boxer genome, although the latter locus lacks both elements in the sequenced poodle genome. The genomic variability that results from bimorphic insertions of SINEs may therefore extend to additional flanking sequences for some active SINEC_Cfs. Although the mechanism for these putative transduction events remains to be explored, the sequences are consistent with transcription of active SINEC_Cf repeats through their 3′-flanking sequences, with polyadenylation at downstream cleavage sites.

Distribution of bimorphic SINE insertions across the dog genome

The 92,580 SINEC_Cf repeats from CanFam1, and the 11,265 SINEC_Cf repeats that are bimorphic between CanFam1, CanSS, and the Trace Archive reads, are distributed among all chromosomes with frequencies of 29.7-43.8 per Mb and 2.8-6.0 per Mb, respectively (Table 3). The local GC content for 1 kb upstream (median = 38.2%) and 1 kb downstream (38.0%) of SINEC_Cf insertion sites is only slightly lower than the genome average for 1-kb nonoverlapping windows of CanFam1 (39.5%). In addition, there is no indication that SINEC_Cf repeats are preferentially located within genes. The Ensembl annotation of CanFam1 (http://www.ensembl.org/Canis_familiaris; release 27.1.1) identifies 18,201 genes. These span 34% of the CanFam1 sequence, and contain 36% of the CanFam1 SINEC_Cf repeats (and 33% of the bimorphic SINEC_Cf repeats) that were identified in this study. Across the whole genome, 48% of annotated genes contain at least one CanFam1 SINEC_Cf (mean; 3.8 per gene), and 14% contain at least one bimorphic SINEC_Cf (mean; 1.5 per gene).

Table 3.

Distribution of SINEC_Cf repeats and bimorphic SINEC_Cf insertions (bSINEC_Cf) among dog chromosomes and annotated genes

		SINEC_Cf		bSINEC_Cf		Annotated genes
Chromosome	Length (Mb)	No.	per Mb	No.	per Mb	No.	% with SINEC_Cf	% with bSINEC_Cf
1	124.898	4520	36.2	516	4.1	1058	45.4	11.5
2	87.725	3125	35.6	373	4.3	672	52.7	14.9
3	95.080	3116	32.8	398	4.2	457	55.6	16.2
4	91.326	3183	34.9	383	4.2	521	49.3	14.2
5	92.065	3103	33.7	360	3.9	916	43.7	9.9
6	79.106	2938	37.1	332	4.2	805	48.1	12.3
7	83.037	3054	36.8	338	4.1	586	58.0	15.4
8	77.375	2931	37.9	452	5.8	567	51.1	16.6
9	53.643	2347	43.8	312	5.8	846	44.8	13.4
10	72.717	2720	37.4	341	4.7	600	53.5	16.0
11	75.770	2856	37.7	316	4.2	480	49.2	11.9
12	75.456	3048	40.4	399	5.3	539	46.6	16.0
13	66.160	2124	32.1	257	3.9	348	48.6	12.6
14	63.550	2261	35.6	269	4.2	360	52.8	15.8
15	67.238	2626	39.1	264	3.9	461	52.5	12.1
16	60.308	2066	34.3	251	4.2	388	47.9	14.9
17	66.886	2521	37.7	326	4.9	545	49.2	13.4
18	66.174	2208	33.4	278	4.2	851	33.5	7.3
19	56.914	1978	34.8	272	4.8	176	58.0	18.2
20	61.173	2381	38.9	282	4.6	884	39.6	9.3
21	53.028	1896	35.8	268	5.1	408	44.9	15.4
22	64.236	2350	36.6	324	5.0	205	52.7	20.5
23	55.586	2000	36.0	273	4.9	290	64.5	20.7
24	50.732	1810	35.7	267	5.3	487	45.2	14.2
25	54.437	1922	35.3	271	5.0	379	49.3	17.2
26	41.010	1668	40.7	244	5.9	425	47.5	15.8
27	49.087	2089	42.6	267	5.4	471	48.2	16.3
28	42.450	1479	34.8	166	3.9	310	57.1	14.2
29	44.860	1528	34.1	180	4.0	185	57.8	14.6
30	43.165	1708	39.6	199	4.6	361	59.8	19.1
31	41.216	1350	32.8	204	4.9	173	48.6	19.7
32	41.834	1834	43.8	252	6.0	204	60.3	27.0
33	34.471	1328	38.5	198	5.7	205	65.4	25.9
34	45.233	1389	30.7	216	4.8	211	48.8	19.4
35	29.507	875	29.7	146	4.9	168	44.0	18.5
36	33.961	1172	34.5	146	4.3	163	58.3	18.4
37	33.897	1118	33.0	164	4.8	207	56.0	16.4
38	26.492	863	32.6	146	5.5	193	44.0	14.5
X	126.913	5429	42.8	356	2.8	735	44.6	8.2
Un	91.063	3666	40.3	259	2.8	361	21.6	5.0

Insertion of SINEs close to exons has been shown to cause aberrant splicing of transcripts in both human (Wallace et al. 1991; Ganguly et al. 2003) and dog (Lin et al. 1999). Insertion of SINEs within exons has also been reported to disrupt gene expression and cause diseases in both human and dog. For example, an Alu insertion within exon 11 of CLCN5 is associated with Dent's disease in human (Claverie-Martin et al. 2003), while a SINEC_Cf insertion within exon 2 of the PLPTA gene is associated with centronuclear myopathy in Labrador retrievers (Pele et al. 2005). The Ensembl annotation of CanFam1 lists 163 exon junctions that are within 100 bases of a SINEC_Cf repeat. However, none of the 92,580 SINEC_Cf repeats from CanFam1 are annotated as residing within an exon. This is likely to be misleading for at least two reasons. First, there are examples of introns within annotated genes that consist entirely of SINEC_Cf sequence (Ensembl genes, ENSCAFG00000000578, ENSCAFG00000000879, ENSCAFG00000001129, ENSCAFG00000004064, ENSCAFG00000008336, ENSCAFG00000017490, ENSCAFG00000017848). If these are, indeed, transcribed genes, the SINEC_Cf insertion would be expected to affect gene expression. However, it is possible that at least some of these examples are processed pseudogenes that have no functional significance whether or not they have acquired SINE insertions. The second reason for an absence of SINEC_Cf sequences in annotated exons relates to our limited knowledge of canine gene structures. Relative to human and mouse, the dog is not well represented in GenBank by cDNA sequences. For example, dbEST (http://www.ncbi.nlm.nih.gov/dbEST; release 100104) contains 6.0 million expressed sequence tags (ESTs) from human, 4.2 million from mouse, but only ∼155,000 from dog. Consequently, the gene annotation of CanFam1 relies more heavily on sequence comparisons with genes from other species that have been validated by ESTs and full-length cDNAs. Such comparisons would not be expected to reveal dog-specific repeats within predicted transcripts.

Transcription of SINEC_Cf elements

Although current annotation of CanFam1 fails to identify SINEC_Cf repeats within predicted exons, an analysis of dog ESTs identified 120 examples of cDNA clones that contain complete or partial SINEC_Cf sequences (Table 4). Approximately half of these are located in the “sense” orientation at the 3′-ends of cDNAs. This common location may be an experimental artifact, caused by oligo(dT) priming of cDNA synthesis at internal, A-rich regions of the primary transcript that are provided by the SINE. However, it is also possible that at least some of these examples arise from the use of the known polyadenylation signals within most tRNA-related SINEs, including SINEC_Cf (Borodulina and Kramerov 2001). Depending on the context of these signals, they could cause premature cleavage and polyadenylation of Pol II-derived dog mRNAs. The cDNAs that terminate with a SINEC_Cf sequence include four examples in which the SINEC_Cf repeats are absent from CanFam1, and are therefore likely to represent bimorphic insertions (Table 5). Notably, three of these four examples are located within annotated genes that are transcribed in the same orientation as the SINEC_Cf repeat that terminates the cDNA. It will be of interest to know if such SINE insertions cause a significant level of premature polyadenylation for the transcripts of the genes in which they have inserted. This should be relatively easy to determine using tissues from dogs that are heterozygous for the SINE insertions, as these tissues should also express normal transcripts from one allele.

Table 4.

Dog ESTs that contain SINEC_Cf sequences

EST ID	Start	End	Orientation	Type	EST ID	Start	End	Orientation	Type	EST ID	Start	End	Orientation	Type
BI395282	100	292	(+)	1	CK999028	9	191	(–)	1	CF411877	509	701	(+)	2
BI396013	247	428	(+)	1	CN000189	1	200	(–)	1	CK997835	373	553	(–)	2
BI817000	122	307	(+)	1	CN000259	1	185	(–)	1	CN000198	540	735	(+)	2
BM536771	134	316	(–)	1	CN000862	1	198	(–)	1	CN001701	354	549	(–)	2
BM538738	29	206	(–)	1	CN003480	9	177	(–)	1	CN003009	132	323	(+)	2
BM540896	1	178	(–)	1	CN003481	52	232	(+)	1	CN003100	198	370	(–)	2
BQ091505	75	251	(+)	1	CN004223	4	186	(–)	1	CN003272	372	553	(+)	2
BQ091808	34	230	(–)	1	CO615070	1	177	(–)	1	CN003734	467	653	(+)	2
BQ233944	258	445	(+)	1	CO618415	1	170	(–)	1	CN003833	466	649	(+)	2
BQ233951	80	248	(+)	1	CO622584	1	192	(–)	1	CN004818	224	406	(–)	2
BQ234215	272	443	(+)	1	CO622645	1	184	(–)	1	CN005854	337	520	(+)	2
BQ234327	32	200	(+)	1	CO624201	14	199	(–)	1	CO620822	82	253	(–)	2
BQ234548	80	248	(+)	1	CO631931	1	181	(–)	1	CO630670	115	296	(+)	2
BQ235019	404	574	(+)	1	CO633032	16	212	(–)	1	CO688822	181	367	(+)	2
BQ235154	401	585	(+)	1	CO633878	19	187	(–)	1	CO708098	261	444	(+)	2
BQ235636	335	521	(+)	1	CO634532	1	173	(–)	1	BM536886	187	337	(–)	3
BQ235876	129	321	(+)	1	CO635086	1	173	(–)	1	BM539513	338	459	(–)	3
BQ788275	291	460	(+)	1	CO677843	1	171	(–)	1	BM540124	47	198	(–)	3
BQ788475	291	465	(+)	1	Z97735	3	192	(–)	1	BQ234114	244	395	(–)	3
BU744447	60	271	(–)	1	Z97747	1	180	(–)	1	BU749657	397	532	(–)	3
BU744716	70	232	(–)	1	Z97810	30	213	(–)	1	BU751410	37	167	(–)	3
BU744952	61	227	(–)	1	BI389234	124	312	(–)	2	CK996159	35	156	(–)	3
BU745150	87	273	(–)	1	BM538367	123	312	(–)	2	CK996423	275	436	(+)	3
BU749508	66	248	(–)	1	BM541044	246	428	(+)	2	CN002508	157	318	(–)	3
BU751148	61	243	(–)	1	BM735692	178	358	(–)	2	CO629118	58	205	(+)	3
BU751489	114	297	(–)	1	BQ234408	316	500	(+)	2	CO668671	394	545	(+)	3
CF406750	65	260	(–)	1	BQ235081	316	500	(+)	2	CO675560	265	414	(+)	3
CF407503	69	250	(–)	1	BQ235579	204	335	(–)	2	CO675838	117	270	(+)	3
CF408296	58	245	(–)	1	BU750849	192	398	(+)	2	CO684930	118	267	(+)	3
CF408332	59	246	(–)	1	BU751135	148	345	(+)	2	BF228953	25	193	(+)	4
CF409934	107	310	(–)	1	BU751296	557	726	(–)	2	BF228988	18	194	(+)	4
CF410986	145	327	(–)	1	BU751297	324	492	(+)	2	BQ234445	5	187	(+)	4
CF411174	218	411	(–)	1	CF406935	280	463	(+)	2	BQ234448	10	146	(+)	4
CF413125	130	324	(–)	1	CF406941	595	755	(–)	2	BQ235047	9	143	(+)	4
CK996008	4	199	(–)	1	CF408437	286	452	(+)	2	BQ235092	10	150	(+)	4
CK996233	93	277	(–)	1	CF409696	259	439	(–)	2	BQ235470	9	144	(+)	4
CK997330	1	181	(–)	1	CF409992	367	556	(+)	2	BQ235629	9	191	(+)	4
CK997507	3	174	(–)	1	CF410165	274	451	(–)	2	BQ235644	9	169	(+)	4
CK997864	1	172	(–)	1	CF410277	296	489	(–)	2	BQ290082	9	136	(+)	4
CK998336	23	191	(–)	1	CF411091	230	412	(+)	2	CK995704	1	162	(+)	4

Each EST is represented by its GenBank accession number, the start and end coordinates of the SINEC_Cf sequence within the EST, the orientation of the SINEC_Cf sequence (+/^–), and the type of insertion (1–4). Four types of insertion were distinguished. A complete SINEC_Cf sequence was located at either the 3′-end of the cDNA (type 1) or within the cDNA sequence (type 2). Partial SINEC_Cf sequences that have arisen from the use of splice acceptor sites within the element were termed type 3. Finally, there were ESTs that consist entirely of SINEC_Cf sequence (type 4).

Table 5.

Examples of dog cDNAs that terminate with a bimorphic SINEC_Cf (+) sequence

	SINEC_Cf(+)	GGGATCC..167..AAAAA	Ensembl gene prediction
1	BI817000(+)	1-101-CCATTAGAGGATATAAAGAAGGGATCC..174..AAAAA-307	ENSCAFG00000016640 (+)
	Cfa6	18637451-18637551-CCATTAGAGGATATAAAGAAgattgct
2	BQ234215(+)	9-251-CAGAGGGTTAAAATGCTTCAGGGATCC..160..AATAA-443	ENSCAFG00000010152 (+)
	Cfa21	49587244-49587487-CAGAGGGTTAAAATGCTTCAaattgtc
3	CK997864(–)	1-372-TCAGCTGGTAGGAATGAATTGGGATCC..156..TAAAA-564	–
	Cfa31	20121516-20121868-TCAGCTGGTAGGAATGAATatggatat
4	BQ233944(+)	10-237-GTGTTAGAACTCCATCAACAGGGATCC..176..AAAAA-445	ENSCAFG00000013709 (+)
	Cfa17	66653098-66653326-GTGTTAGAACTCCATCAACAttttccc

The termini of the consensus SINEC_Cf (+) sequence is included for comparison. For examples 1–4, the GenBank accession numbers (and orientations) refer to dog ESTs [e.g., BI817000 (+)] that can be aligned with specific chromosomal regions of CanFam1. The nucleotide identity between the EST and genomic sequences is terminated by a SINEC_Cf sequence at the 3′-end of the cDNA sequence. These are all examples of transcribed SINEC_Cf insertions that are bimorphic in the dog population. Three of the four examples map within introns of putative genes in the (+) orientation (Ensembl annotation).

Among the ESTs of Table 4, there are also examples of dog cDNAs that have acquired additional exons (relative to their human orthologs) owing to the splicing of transcribed SINEC_Cf elements. An example is illustrated by the canine Tipin gene (Fig. 3). Relative to human, the dog Tipin transcript has acquired an additional exon owing to the insertion of a SINEC_Cf repeat downstream of the first exon. When transcribed in the (–) orientation, a SINEC_Cf repeat provides characteristic sequence motifs that permit it to act as a 3′-splice acceptor site, resulting in activation of a cryptic 5′-splice site downstream of the element. A similar mechanism has been described for exonization of an Alu within the RPE gene in primates (Krull et al. 2005). In the case of canine Tipin, the novel exon is predicted to be untranslated although its effect on Tipin protein expression is currently unknown. Table 6 provides details of 10 distinct cDNAs where SINEC_Cf repeats are spliced into transcripts by using precisely the same splice acceptor site. For most of these, alignment of the cDNA sequences with CanFam1 confirms the predicted splicing pattern. However, there are also two examples (9 and 10) in which the SINEC_Cf repeat is absent from CanFam1. These likely represent additional examples of bimorphic SINEC_Cf insertions that are transcribed. Four of the cDNAs appear to represent dog orthologs of known human genes, and the SINE insertion disrupts the homologous open reading frame for three of these (see legend to Table 6).

Figure 3.

Splicing of a SINEC_Cf sequence into the canine mRNA for Tipin. (A) The 5′-end of the canine Tipin gene differs from that of human by inclusion of an additional untranslated exon that is derived from a SINEC_Cf. (B) Alignment of sequences representing the 5′-ends of Tipin mRNA from human (GenBank accession no. BP248314) and dog (BM540124) with genomic sequences from human (NCBI build 34) and dog (CanFam1). Nucleotide identity between the human and dog genomic sequences is shaded, exons are boxed, and intronic sequences are in lower case. Between exons 1 and 2, the dog genome contains a SINEC_Cf sequence that is flanked by a characteristic 15-bp duplication of the insertion site (white lettering on black). This element provides a 3′-splice acceptor site, and causes activation of a cryptic 5′-splice site downstream of the element, resulting in the incorporation of the SINEC_Cf sequence within the dog Tipin mRNA.

Table 6.

Examples of dog cDNAs that have incorporated a SINEC_Cf (–) sequence via a splice acceptor site within the element

	SINEC_Cf(–)	TTTTTATTTATTTATGATAGTCACACAGAG.. 141.. GGATCCC
1	BM540124(+)	38-GCCCCGGAGG TCACACACAC.. 134.. GGATCCC AGGTCATCAC-208
	Cfa30	33601222-GCCCCGGAGGgtgagagctg... tttatgatagTCACACACAC.. 136.. GGATCCC AGGTCATCAC-33600139
2	CN002508(+)	154-CACTGATGAG TCACACAGAG.. 138.. GGATCCC CATTTCATTT-328
	Cfa2	68067110-CACTGATGAGgtatgaattg... tttatgatagTCACACAGAG.. 138.. GGATCCC CATTTCATTT-68064412
3	CO675838(–)	306-GCACACTCTG TCACACAGAG.. 138.. GGATCCC AAGCCAAGCA-478
	Cfa6	11279927-GCACACTCTGgtcagttcca... tttatgatagTCACACAGAG.. 141.. GGATCCC AAGCCAAGCA-11279642
4	BM536886(+)	177-CAGCCCACAG ACATAGAGAG.. 134.. GGATCCC CTCATCTGAC-347
	Cfa9	26197334-CAGCCCACAGgtaaagtatt... ttcatgatagACATAGAGAG.. 134.. GGATCCC CTCATCTGAC-26192495
5	CO668671(–)	109-CATCCCAGAG GCNCNCAGTG.. 136.. GGATCCC CTACATTGCT-281
	Cfa14	47081174-CATCCCAGAGgtaagagtat... tttatgatagGCACACAGTG.. 138.. GGATCCC CTACATTGCT-47047028
6	BU749657(+)	322-GCAATCACGG AGAGAGAGAG.. 114.. GGATCCC CGGGTTAAGG-472
	Cfa15	52463523-GCAATCACGGgtaaggtctt... tttattatagGCACACAGTG.. 114.. GGATCCC CGGGTTAAGG-52469700
7	BQ234114(+)	233-CCCACCACAG TCACAGAGAG.. 136.. GGATCCC AGATCCATCT-405
	Cfa26	24551559-CCCACCACAGgtataaacaa... tcagtcacagTCACAGAGAG.. 136.. GGATCCC AGATCCATCT-24550169
8	CO684930(–)	328-ATCAAAGCTG TCACACAGAG.. 132.. GGATCCC CATGTGTGTG-496
	Cfa27	34794751-ATCAAAGCTGgtgagataca... tctatgatagTCACACAGAG.. 138.. GGATCCC CATGTGTGTG-34795736
9	CO675560(–)	183-TCTTCAAGCT TCACAGAGAG.. 134.. GGATCCC CCGTTAGTTG-353
	Cfa22	49073614-TCTTCAAGCTgtgagtgcgg... TTAGTTG-49088654
10	CO629118(–)	320-ACCTCTAAAG TCNCAGAGAG.. 134.. GGATCCC TATGCTCACT-493
	Cfa35	28868224-ACCTCTAAAGgtcagtcaca... TATGCTCACT-28867050

The location of the splice site (↓) follows a canonical “AG” dinucleotide, and is illustrated on the consensus SINEC_Cf (–) sequence. For examples 1–8, the GenBank accession numbers (and orientations) refer to dog ESTs [e.g., BM540124 (+)] that can be aligned with specific chromosomal regions of the dog genome sequence. The alignments of EST and genomic sequences include exons (upper case), introns (lower case), and coordinates of the aligned EST and genomic sequences. All introns are flanked by canonical “gt” and “ag” dinucleotides. For examples 9 and 10, sequences that flank the SINEC_Cf of the EST can be aligned with genomic sequence, but the SINEC_Cf sequence is absent. These are examples of bimorphic SINEC_Cf insertions that have become incorporated within cDNAs as alternative exons. For three of the dog ESTs (CN002508, BM536886, CO668671), the SINE insertion disrupts the open reading frame of the homologous human cDNA (phosphorylase kinase, β subunit, NM_000293; TRIM37 mRNA, NM_015294; phosphodiesterase 1C mRNA, NM_005020).

The recent expansion of SINEs in the dog genome, reflected by a high frequency of bimorphic SINE insertions, provides a unique opportunity to explore the influence of SINEs on the evolution of a mammalian genome. For many thousands of genes, an individual dog will carry two alleles that differ by their content of SINEs. It is therefore possible to assess the impact of SINEs on gene expression patterns within individuals (or even within individual cells) rather than requiring a comparison between multiple individuals or between multiple species. The high frequency of bimorphic SINE insertions in the dog is predicted to provide numerous examples of allele-specific splicing patterns that can be studied further by correlating their potential functional effects with their distribution between dog breeds. Consequently, it is likely that canine bimorphic SINE insertions will provide us with evidence of how insertion elements can mold a mammalian genome, as well as the means to identify genetic relationships between the diverse lineages of current canine populations.

Methods

Loci that are bimorphic for SINEC_Cf insertions between CanSS and CanFam1

The unmasked assembly of a draft boxer genome sequence (CanFam1) was downloaded from the UCSC Genome Bioinformatics site (http://hgdownload.cse.ucsc.edu/downloads.html#dog). The unmasked assembly of a survey-sequenced poodle genome sequence (CanSS) has been described previously (Kirkness et al. 2003). A sequence consisting of bases 1-124 of the SINEC_Cf consensus (RepBase release 7.11) was searched against CanSS using NCBI BLAST (version 2.2.4). The output was filtered for alignments that included at least bases 5-120 of the SINEC_Cf consensus, had no gaps, and had fewer than 11 mismatches. Aligned segments of the survey sequence, together with 50 bases of 5′-sequence, and 175 bases of 3′-sequence, were extracted from the contigs, and masked for canine SINEs and low-complexity sequences using RepeatMasker (version 07/02). The output was filtered for sequence fragments that retained at least 30 consecutive unmasked bases on both flanks of the masked SINEC_Cf sequence. These sequences were then searched against CanFam1 using NCBI BLAST (-W 15, -v 5, -b 5, -F F). In order for a SINEC_Cf to be scored as potentially bimorphic, it was necessary for the flanks of the query to align with only a single fragment of CanFam1, and for these flanks to be contiguous on the homologous CanFam1 fragment. The same approach was used to identify SINEC_Cf repeats in CanFam1 that are absent from CanSS.

Loci that are bimorphic for SINEC_Cf insertions from multiple breeds of dog

Approximately 1 million whole-genome shotgun reads that were derived from nine dogs of different breeds, four wolves, and a coyote were downloaded from the NCBI Trace Archive (ftp://ftp.ncbi.nih.gov/pub/TraceDB). The correlation between “center-_project” IDs (S229-S245) and specific dog breeds was provided by Kerstin Lindblad-Toh (The Broad Institute). Selection of SINEC_Cf repeats and flanking sequences was performed as for CanSS and CanFam1 segments (see above), except that they were restricted to bases 25-675 of each shotgun read (in order to avoid low-quality bases). The filtered sequences were then searched against both CanSS and CanFam1, and scored as described above.

Evidence for SINE-mediate transduction of 3′-flanking sequences

For each of the 92,580 SINEC_Cf-containing segments of CanFam1, 100 bases that flank the 3′-end of the element were masked (RepeatMasker) and searched against the complete collection of 143,080 SINEC_Cf-containing segments from CanFam1 and CanSS using NCBI BLAST. Distinct segments that shared nucleotide identity for >50 consecutive bases were subject to further manual alignment, and annotated for SINEC_Cf sequences, target-site duplications, and transposed 3′-sequences. Potential transduction events were indicated when a target-site duplication (plus 3′-flanking sequence) of one element was contained within the target site duplications of another.

Characterization of SINEC_Cf sequences within dog ESTs

Approximately 155,000 dog ESTs in dbEST (http://www.ncbi.nlm.nih.gov/dbEST/; release 100104) were searched with bases 1-124 of the SINEC_Cf consensus sequence as described above for CanSS and CanFam1. Those ESTs that contained SINEC_Cf sequences were downloaded from dbEST and aligned to CanFam1 using BLAT (http://www.genome.ucsc.edu/cgi-bin/hgBlat) and NCBI BLAST.

Assay for identification of novel loci that are bimorphic for SINEC_Cf insertions

Dog genomic DNA (100 ng; Novagen) was digested with NlaIII in 20 μL, heat-inactivated (65°C, 20 min), and ligated overnight at 20°C in 500 μL with 10,000 U of DNA ligase (New England Biolabs). After phenol-extraction and ethanol-precipitation, ∼20 ng were amplified in 50 μL with 0.5 U of Platinum Taq DNA Polymerase, 0.2 mM each dNTP, 1× buffer, 1.5 mM MgCl₂, and 1.2 μM of the following primers: 5′-GGTATCAACGCAGAGTGGCC GCCTCGGCCCTGGGCCAAAGGCAGG, 5′-GGTATCAACG CAGAGTGGCCGCCT, 5′-ATTCTAGAGGCCATTACGGCCTC GAATCCCACRTCRGGCTCCYRG, 5′-ATTCTAGAGGCCATTAC GGCCTCG.

The PCR-amplification was 30 cycles of 95°C (45 sec), 60°C (1 min), and 72°C (2 min). Products of >300 bp were purified from agarose gels using the QIAquick Gel Extraction system (Qiagen), and cloned using the TOPO TA Cloning system (Invitrogen). After transformation, plasmid templates were prepared from white colonies and sequenced using the M13F primer. Seven independent libraries were constructed, and 772 clones were sequenced. High-quality sequence data were obtained for 81%-92% of the clones from each library.