Tm1: A Mutator/Foldback Transposable Element Family in Root-Knot Nematodes

Three closely related parthenogenetic species of root-knot nematodes, collectively termed the Meloidogyne incognita-group, are economically significant pathogens of diverse crop species. Remarkably, these asexual root-knot nematodes are capable of acquiring heritable changes in virulence even though they lack sexual reproduction and meiotic recombination. Characterization of a near isogenic pair of M. javanica strains differing in response to tomato with the nematode resistance gene Mi-1 showed that the virulent strain carried a deletion spanning a gene called Cg-1. Herein, we present evidence that the Cg-1 gene lies within a member of a novel transposable element family (Tm1; Transposon in Meloidogyne-1). This element family is defined by composite terminal inverted repeats of variable lengths similar to those of Foldback (FB) transposable elements and by 9 bp target site duplications. In M. incognita, Tm1 elements can be classified into three general groups: 1) histone-hairpin motif elements; 2) MITE-like elements; 3) elements encoding a putative transposase. The predicted transposase shows highest similarity to gene products encoded by aphids and mosquitoes and resembles those of the Phantom subclass of the Mutator transposon superfamily. Interestingly, the meiotic, sexually-reproducing root-knot nematode species M. hapla has Tm1 elements with similar inverted repeat termini, but lacks elements with histone hairpin motifs and contains no elements encoding an intact transposase. These Tm1 elements may have impacts on root-knot nematode genomes and contribute to genetic diversity of the asexual species.


Introduction
Root-knot nematodes (RKNs), comprising the genus Meloidogyne, are obligate parasites causing damage to plant roots [1] and weakening plants by acting as carbon sinks [2,3]. Worldwide, RKNs represent a serious economic problem, causing a 5-12% reduction in crop productivity [4]. Species within the genus Meloidogyne differ widely in their modes of reproduction. The three species generally regarded as causing the most damage globally are part of what has been called the M. incognita (Mi)-group: M. incognita, M. javanica, and M. arenaria. All three are parthenogenetic species in which the eggs are products of mitosis and develop without fertilization [5,6,7]. These asexual species are closely related and are thought to have originated from interspecific hybridization followed by the loss of meiosis [8,9,10]. Despite their lack of meiosis, this group of RKNs is evolutionarily successful, capable of parasitizing thousands of plant species and acquiring heritable virulent phenotypes [7]. The mechanisms by which these asexual RKN species acquire the ability to parasitize additional host taxa are unknown, but genomic change mediated by transposable elements and repetitive sequences has been proposed [11,12].
Transposable elements are potent mutagens and can shape genomes in many ways, ranging from large-scale chromosomal rearrangements to subtle alterations in gene regulation [13,14]. Two general mechanistic classes of transposable elements exist [15]: Class I-retrotransposons, which replicate and insert in new genomic locations through an RNA intermediate; and Class II-DNA transposons, which transpose through cut-and-paste mechanisms without an RNA intermediate. Typical autonomous Class II elements encode a transposase gene flanked by terminal inverted repeats (TIRs). The transposase protein recognizes the TIRs and facilitates mobilization of the element to a new genomic location. Non-autonomous elements do not encode a functional transposase, but can be mobilized in trans by transposase proteins produced by an autonomous element. Most transposable elements are also flanked by target site duplications (TSDs) produced as a byproduct of the transposon integration process [16].
A widely studied example of the genetic variability within the apomictic RKNs is the ability to acquire virulence by escaping recognition by tomato plants with the root-knot nematode resistance gene, Mi-1. Mi-1 confers resistance that is broadly effective against the related Mi-group asexual root-knot nematode species [17,18] and, surprisingly, some isolates of potato aphid and whitefly [19,20]. However, Mi-1 does not confer resistance to the RKN species M. hapla, a facultative parthenogen in which gametes are products of meiosis, but in the absence of males, sister nuclei re-fuse to produce parthenogenetic diploid progeny [21,22]. Populations of Mi-1-resistance breaking asexual nematodes have been found both in the field and after greenhouse selection [23,24,25,26,27,28]. Gleason et al. [24] sought to identify genetic differences between an avirulent isolate of M. javanica (VW4) and a greenhouse-derived virulent derivative, VW5. Results of both gDNA and cDNA AFLP analyses were consistent with these two nematode strains being isogenic with the exception of the deletion spanning a putative gene named Cg-1 in VW5. The known 727 bp of the Cg-1 locus was found to express 456 nt-long mature transcript without obvious coding potential, capable of producing peptides of only 3-32 amino acids, and containing a histonehairpin motif near its 39 end. Except for its discovery in Cg-1, this motif has been identified only in the 39 UTR of metazoan, but not plant, histone genes where it is required for processing of histone mRNAs and export of these transcripts from the nucleus [29,30]. Southern analyses and PCR demonstrated that Cg-1 is a member of a gene family with 8 copies in the M. javanica genome, and transcript analysis indicated that several of these sequences are expressed. Soaking preinfective juveniles of VW4 in dsRNA corresponding to parts of the Cg-1 sequence resulted in a virulent phenotype capable of parasitizing Mi-1 tomato, supporting a role for Cg-1 in Mi-1-mediated recognition by the resistant host. However, because of the apparent lack of coding potential, the molecular function of Cg-1 and its role in mediating response to tomato with Mi-1 has remained elusive.
Recently, the genome sequences of two root-knot nematodes species, M. incognita and M. hapla, became available and provide a resource for further investigation of the Cg-1 locus [10,31]. Herein we present a detailed analysis of the Cg-1 locus, and demonstrate that the previously described Cg-1 gene is within a novel class II transposable element with structurally complex terminal-inverted repeats similar to those of Foldback (FB) transposable elements. We have named this element Tm1 (for Transposon in Meloidogyne-1). We analyze and contrast the Tm1 transposon families of M. incognita and M. hapla. While Tm1 elements display evidence of transposition activity during the radiation of RKN species, our analyses suggest Tm1 elements are currently inactive and a decaying transposon family in M. incognita, M. javanica, and M. hapla. We propose that despite the lack of obvious recent transposition activity, Tm1 elements may be part of the molecular toolkit asexual root-knot nematodes use to acquire genetic diversity.

Results
The Cg-1 locus contains a transposon-like structure We used the available 727 bp of the Meloidogyne javanica VW4 Cg-1 locus (GenBank EU214531.1) [24] to query the published genome sequence of M. incognita, a RKN species closely related to M. javanica. The highest-scoring BLASTN match (E = 6.70e 248 ) was on M. incognita contig 1763 (GenBank CABB01001763.1). We noted that sequences 59 of the M. javanica Cg-1 transcript are present on contig 1763 in an inverted-repeat orientation flanked by 9 bp direct repeats, suggestive of a transposable element ( Figure 1A). The sequences between these inverted repeats are similar to Cg-1 with the exception of the presence of a second, unrelated 936 bp inverted-repeat element flanked by 7 bp target site duplications being inserted within the region of homology ( Figure 1A). In other words, the sequence on M. incognita contig 1763 appears to contain two nested transposable elements: a 936 bp element residing within another novel transposable element sharing sequence identity to M. javanica Cg-1 ( Figure 1A). We designed two PCR primers to test whether M. javanica Cg-1 was also flanked by inverted repeats. The 39 end of primer SG1 binds a 5 bp sequence (CAATGA) present in Cg-1 but absent in other members of the gene family [24]. A second primer, SG6, binds within the presumptive inverted repeat. Using SG1 and SG6, we amplified a 629 bp fragment from genomic DNA of M. javanica strain VW4, but not VW5 (data not shown), confirming specific amplification of the Cg-1 locus and supporting the presence of inverted repeats flanking sequences coding for the Cg-1 transcript.
To investigate the M. javanica Cg-1 locus further, we screened a genomic DNA library containing ,3 kb VW4 HindIII fragments, a size previously demonstrated to harbor Cg-1 [24], using primers SG1 and SG6. We obtained a 3154 bp clone containing Cg-1 (GenBank EU214531.2; Figure 1A), and sequence analysis confirmed that the Cg-1 sequence is flanked by inverted repeats. Sequences outside the inverted repeat element of M. javanica Cg-1 diverge completely from those outside the element on M. incognita contig 1763, further suggesting that these inverted-repeat sequences correspond to the boundaries of transposable elements. However, the inverted repeats flanking M. javanica Cg-1 do not have target site duplications and the 59 TIR appears to be truncated ( Figure 1A). The element at Cg-1 is 82.7% identical to that on M. incognita contig 1763 excluding the nested 936-bp insertion.
In an attempt to identify additional sequence flanking Cg-1, we screened a BAC library of M. javanica VW4 by PCR using primers SG1 and SG6. Despite several attempts, we were unable to identify a clone corresponding exactly to Cg-1 in the BAC library. However, by lowering the annealing temperature of our PCR reaction (Materials and Methods), we obtained a 26 kb BAC clone (clone 18M17, GenBank HQ122410.1). The sequence of this clone identified an element similar to that found at Cg-1 ( Figure 1A). This element was 1080 bp with terminal inverted repeats flanked by 9 bp target site duplications. The sequence is 84.9% identical to Cg-1, but sequences outside the element on BAC 18M17 are different from those found flanking elements at Cg-1 and Mi contig 1763, again suggestive of a transposable element.
The three putative transposable elements shown in Figure 1 are similar in sequence and display no obvious protein-coding potential. Between their inverted repeats, each contains a histone hairpin motif ( Figure 1A) [24,29]. Analysis of the TIRs of each element reveals a composite structure ( Figure 1B). The TIRs of the elements on contig 1763 and BAC 18M17 have asymmetric 7 bp motifs at their outer ends, internal to the 9 bp target site duplications, here called Motif A1 (59-CGGTTAA-39) and Motif A2 (59-CCTACCC-39). Internal to Motif A1 and A2 is a 3 bp sequence, Motif B (59-GGA-39). The largest portion of the TIR is composed of a variable number of 13 to 15 bp (most often 14 bp) tandem repeats with the consensus 59-CGATTCAGTATCCGC-39 (Motif C). Due to differing numbers of repeats of Motif C, the TIRs at each end of the same element, as well as between elements, differ in size. The internal ends of the TIRs (closer to the element center) contain a conserved 12 bp purine-rich sequence stretch (Motif D; 59-GGGAAAAGGGGA-39), followed by one additional unit of Motif C.
To determine whether the nematode genomes carried paralogous loci lacking an element and the associated target site duplication (e.g., empty sites) corresponding to the insertion site of the element on BAC 18M17, we surveyed several M. javanica and M. incognita isolates by amplifying genomic DNA with primers 17-F and 17-R, which flank the putative transposable element ( Figure 1A). We obtained ,0.4 kb amplicons from all tested RKN isolates ( Figure 1C), approximately 1 kb shorter than the 1493 bp amplicon produced from BAC clone 18M17 and corresponding to the expected length for an empty site. Sequencing of the 0.4-kb amplicon from VW4 (GenBank HQ122411.2) confirmed that these fragments correspond to a paralogous locus without the inverted repeat element and target site duplications. A second PCR test using a primer within the inverted-repeat element (SG2) and the flanking 17-R primer ( Figure 1A) demonstrates that the tested M. javanica isolates, but not the M. incognita isolates, contain a paralog with an inverted repeat element similar to that on BAC 18M17 ( Figure 1C).
Taken together, the identification of 3 similar elements in distinct genomic contexts, two of which are flanked target site duplications, and the identification of paralogous loci with and without an element support the hypothesis these sequences are transposable elements. We have named this transposon family, members of which have novel TIRs with internally redundant motifs, Tm1 (Transposon in Meloidogyne-1). Elements closely resembling those found at Cg-1, Mi contig 1763, and on BAC 18M17 and containing a histone-hairpin motif will be referred to as Tm1-HH.

Identification of a Tm1 element encoding a 496 amino acid putative transposase
The three Tm1-HH elements discussed previously lack potential to code for a transposase protein and are thus nonautonomous elements. Reasoning that an autonomous Tm1 element is likely to share the same terminal inverted repeats, we amplified nematode DNA using only the SG6 primer, which binds within the TIRs ( Figure 1B). We obtained predominant amplicons of approximately 0.9 kb and 2.3 kb in all tested M. javanica and M. incognita isolates, but no amplicons from the more distantly related M. hapla VW9 (Figure 2A). The 0.9 kb amplicons are the size expected from Tm1-HH elements such as those found at Cg-1 and on BAC 18M17 ( Figure 1A). Three of the 2.3-kb amplicons (one each from M. javanica VW4, M. javanica Yolo-1, and M. incognita LB-2) were cloned and sequenced (GenBank accessions HM470231.1, HM470232.1, and HM470230.1, respectively). These three clones were very similar in sequence, with the two M. javanica amplicons sharing 99% identity and the M. javanica VW4 and M. incognita LB-2 amplicons sharing 97% identity. Because the binding site of SG6 was within the TIRs ( Figure 1B), the cloned amplicons from VW4, Yolo-1, and LB-2 lack the full TIR. Therefore, we used BLASTN to query the M. incognita genome and identified a highly similar element on contig 274 (GenBank CABB01000274.1) (E value = 0). This element is 2593 bp long, and is flanked by 9 bp target site duplications ( Figure 2B). The TIRs are structurally similar to those of the Tm1 elements described above and carry the same set of motifs ( Figure 2B and Table S1).
A query of the Pfam protein database [32] with the DNA sequences of the 2.5 kb Tm1 element on contig 274 element revealed coding potential for a MULE domain transposase [33,34]. Thus, we refer to the 2.5 kb Tm1 element as Tm1-A (Tm1-Autonomous). However, no long contiguous ORF was found suggesting the presence of introns. To identify transcripts encoded by Tm1-A elements, we used primers within the predicted coding region (M8-1F and M8-3R; Figure 2B) to amplify partial Tm1-A transcripts from cDNA of M. javanica Yolo-1 (GenBank HQ122409.1; Figure 2B) and M. incognita LB-2 (not shown). We were not successful in obtaining full-length cDNA, so the 59 and 39 ends of the transcript were deduced by comparing the Yolo-1 Tm1-A cDNA sequence to the genomic Yolo-1 Tm1-A sequence. The nearest in-frame start codon (AUG) and stop codon (UGA) flanking the cloned cDNA were found; a spliced leader 1 (SL1) acceptor site (UUUUCAG) [35] was found 59 of the predicted start codon, and several potential polyadenylation signal sequences (composed of AUUAAA, AAUAAU, GAUAAA, and two repeats of AAAAAA) [35] were identified 39 of the stop codon. Taken together, the Yolo-1 Tm1-A element is predicted to express a transcript comprised of 7 exons ( Figure 2B) and encoding a protein of 496 amino acids ( Figure 2C).
The highest BLASTP hits of the predicted full-length 496 amino acid protein encoded by the M. javanica Yolo-1 Tm1-A element (GenBank ADM16638.1) correspond to predicted proteins deduced from the genome sequences of aphids and mosquitoes (Table 1). Pfam v. 24.0 [32] revealed that these predicted proteins have a domain structure similar to transposases encoded by Phantom elements, a recently described subclass of Mutator Class II transposable elements that is widely distributed in eukaryotic taxa including many animals [36] ( Figure 2C). Residues 5-66 of the putative Yolo-1 Tm1-A transposase have similarity (E = 1.3e 27 ) to a FLYWCH zinc-finger domain (Pfam PF04500), a protein interaction domain in some isoforms of Drosophila Mod(mdg4) proteins [37]. Pfam also identified Yolo-1 Tm1-A residues 197-294 as a MULE domain (E = 9e 213 ) (Pfam PF10551) [33,34]. MULE domains are found in transposases of the Mutator family originally identified in plants [38] and the prokaryotic IS256 family [39].

Tm1 elements are low copy and do not display evidence of recent transposition
To determine the copy number of Tm1-A elements, a Southern blot of HindIII-digested genomic DNA from several isolates representing three RKN species was probed with an internal portion of the Tm1-A element ( Figure 2B). All surveyed M. javanica isolates have a single ,11 kb hybridizing HindIII fragment, while all surveyed M. incognita isolates have a single ,8 kb hybridizing fragment ( Figure 3A). No HindIII fragments hybridizing to the Tm1-A probe were detected in M. hapla VW9, consistent with PCR results (above) and bioinformatic analyses (discussed below). The same blot was stripped and re-probed with a fragment of beta actin (GenBank AF532605; Materials and Methods). The beta actin hybridization patterns were also similar between isolates of the same species, although M. javanica Yolo-1 lacked hybridizing fragments of approximately 2.5 and 1.25 kb when compared to other M. javanica isolates. A Southern blot of M. javanica VW4 genomic DNA cut with BamHI, XbaI, or XhoI and probed with Tm1-A also revealed only a single hybridizing fragment (13 kb, 12 kb, 11 kb, respectively; data not shown). These results are consistent with the presence of a single copy of Tm1-A in the genome of tested M. javanica and M. incognita isolates, and the absence of a conserved copy in M. hapla VW9. Furthermore, the similar size of the HindIII fragments suggests this element is likely in the same genomic location within the isolates of each species.
To determine whether Tm1-HH elements like that at Cg-1 are in distinct genomic positions between isolates, a portion of the element at Cg-1 ( Figure 1A) [24] was used to probe a second Southern blot of HindIII-digested genomic DNA ( Figure 3B). With the exception of the previously noted absence of a 3.1 kb hybridizing fragment in M. javanica VW5 representing the deletion of Cg-1 [24], the same pattern of approximately nine strongly hybridizing bands was seen in all M. javanica isolates (VW4, Yolo-1, and Yolo-4; Figure 3B). This is in agreement with previous findings obtained by cloning and sequencing PCR products that M. javanica has 9 copies of sequences with 87 to 100% identity to Cg-1 [24]. In contrast, only weakly hybridizing fragments are observed in the M. incognita isolates suggesting that any copies present are sufficiently diverged and incapable of hybridizing with the probe at the stringent conditions used. This is consistent with our query of the M. incognita genome, which found the Tm1-HH element on contig 1763 to be the most similar sequence (72.4% identity over the length of the probe). The similar hybridization patterns present in the tested M. javanica isolates suggest that like Tm1-A, non-autonomous Tm1-HH elements have not been mobilized since these isolates diverged.  We sought to identify additional Tm1 elements in the published genomes of M. incognita and M. hapla using a bioinformatics approach. In short, we used a BLASTN-based algorithm to search for two or more matches on the same contiguous sequence with similarity to a 38 bp region within the Tm1 TIRs in an inverted orientation (Materials and Methods). This method identified 39 elements in M. incognita and 22 elements in M. hapla, ranging in length from 0.17 to 2.5 kb (Tables S1 and S2; summarized in Table 2).
Tm1 elements of M. incognita could be divided into three categories based upon features of sequences between the TIRs (Table S1). Only 1 transposase-encoding element, on contig 274 as previously noted, was identified. Eight Tm1 elements, including the aforementioned element on contig 1763, included sequence that matched or resembled the histone hairpin of the Cg-1 element. The region with similarity to the Cg-1 element varied, but generally spanned the ,24 nt histone hairpin and 300-400 nt 39 of this sequence, in some cases extending to the TIR. We refer to these elements as the histone-hairpin class (Tm1-HH). Tm1-HH elements are about 900 bp except for two, including the previously described element on contig 1763, which contain insertions (Table  S1). Upstream of the histone hairpin region, elements in this group vary in their similarity to Cg-1, with the element of contig 1763   (Table S2). The Tm1-ML element class is thus the only one clearly shared between M. incognita and M. hapla, though a neighbor-joining consensus tree generated from an alignment of Tm1-ML elements demonstrates the M. hapla elements are more closely related to each other than the majority of M. incognita elements ( Figure S1). Two elements in M. hapla (on contigs 402 and 53) contain internal sequences that did not fit within the Tm1-A, -D, -HH, or -ML classes (noted as ''other'' in Table S2).
None of the non-autonomous Tm1 elements have significant coding potential, and BLASTX fails to detect similarity to known or predicted proteins in the GenBank nr (non-redundant) database (all E values .1). Similarly, conceptual translations of predicted genes and ORFs residing within Tm1 elements produce only short peptides not significantly similar to known or predicted proteins (all E values .1; data not shown).

Tm1 TIRs have both conserved and variable motifs
The TIRs of Tm1 elements annotated in the M. incognita and M. hapla genomes resemble the TIRs of Tm1 elements described previously ( Figure 1A, 2B), but have some differences in the composition and sequence of motifs ( Figure 4A While Tm1-ML elements lack the GA-rich Motif D, this motif is generally present and highly conserved in other Tm1 elements (Tables S1 and S2). The sequence consensus of Motif D differs between M. incognita (59-GKGGAAAAGGGGA -39; where K = G or T) and M. hapla (59-GCCAAAAGGGGA -39). In both species, the 59 end of Motif D is less conserved than is the 39 end ( Figure 4B). The SG6 primer binding site spans part of the variable portions of Motif D ( Figure 1B), accounting for our inability to detect all Tm1 elements in M. javanica, M. incognita, and M. hapla via PCR.
In addition to differences in motif composition, the TIRs of each element differ in sequence. BLAST2 was used to identify local regions of similarity between the two TIRs of each element, allowing automatic exclusion of asymmetric regions such as Motif A1, A2, and asymmetric numbers of Motif C units. Subsequently, the BLAST2 identified regions were aligned for comparison. Of the 61 elements analyzed, only 5 elements (Tables S1 and S2) display 100% sequence identity between the BLAST2-defined portions of their TIRs. On average, the two TIRs of a single element are 93.15% identical due to SNPs and small indels.

Tm1 elements are commonly flanked by 9 bp target site duplications
Analysis of complete Tm1 elements having both terminal motifs A1 and A2 revealed that in most cases (45 of 53), elements are flanked by 8-10 bp (though most commonly 9 bp) target site duplications (Tables S3 and S4). Comparison of TSD sequences did not reveal a clear consensus sequence, suggesting Tm1 elements do not have target sequence preferences at their integration site.

Discussion
A deletion in the M. javanica genome encompassing a candidate effector gene led us to the discovery of a novel transposable element family that we have named Tm1. These elements are defined by the shared novel structure and sequence of their TIRs, which contain 4 sequence domains and varying numbers of an internally redundant ,14 bp sequence motif and, thus, resemble Foldback (FB) elements, a heterogeneous group of transposons with composite TIRs of multiple domains and tandem repeats [43,44,45,46]. Tm1 elements contain at their boundaries asymmetric 7 bp terminal motifs, generally Motif A1 at one end of the element and Motif A2 at the other, that are outside of the inverted repeat. The presence of short, non-inverted sequences at the ends of elements is unusual, but has been observed in other Mutator-like elements, which are thus sometimes said to have sub-terminal inverted repeats (sub-TIRs) [36,47]. Additional conserved motifs characterizing Tm1 elements include a 3 bp Motif B and a GArich motif D, which is generally followed by a unit of Motif C.
Based on bioinformatic analysis of available genomic sequence, M. incognita Tm1 elements can be placed in three general categories: Tm1-A, Tm1-HH, and Tm1-ML. The isolates of M. incognita and M. javanica that we examined carry a single Tm1-A element encoding a protein resembling a Mutator-like (MULE) transposase. The presence of 9 bp TSDs flanking most Tm1 transposons is consistent with the size commonly generated by MULE domain transposases [47,48,49] and other members of the IS256/MULE transposon family [50,51,52,53]. The domain structure of the putative transposase places Tm1-A and the Tm1 family within the Phantom subclass of Mutator elements with highest similarity to sequences from insects [36]. The predicted transposases of Phantom elements are characterized by an N-terminal, FLYWCH zinc-finger DNA binding domain and a MULE transposase domain. It is unclear whether the Tm1-A transposase can bind DNA as it lacks the final histidine residue to form a complete C2H2 zinc-finger motif within the FLYWCH domain. However, it is possible Tm1-A can localize to DNA in a manner similar to the Drosophila FLYWCH domain isoforms of Mod(mdg4) proteins, which do not bind DNA directly but modulate gene expression through interactions with DNA binding proteins [37,54,55]. Further biochemical analysis of the Tm1-A-encoded protein is needed to clarify any function.
Recently some Foldback elements have been grouped with Mutator transposons into a superfamily because of similarity between their encoded transposases [56]. Our finding of a Phantom-like transposase within FB-like TIRs supports this grouping. However, it is clear that despite their structural similarity, FB-like elements are not necessarily phylogenetically related as different classes of transposases have been shown to reside within their TIRs and mediate their transposition. For example, the FB transposon Galileo in Drosophila harbors a P element transposase [57] whereas the Arabidopsis FB element FARE contains a MULE domain transposase [46]. In addition, while some Phantom transposases are flanked by FB-like TIRs or sub-TIRs, others are flanked by structurally simpler TIRs more common in the Mutator superfamily [36]. Such observations have led others [36,57] to propose that FB-like transposon families evolved independently from structurally simple transposons through the gradual expansion of TIR length. A similar case of TIR expansion toward a FB-like structure appears to have occurred during the evolution of the RKN Meloidogyne chitwoodi Tc1/Mariner-like element Mcmar, which has 355 bp TIRs composed of direct repeats and palindromes unlike most Tc1/ Mariner elements, which have simple 26-30 bp TIRs [58].
We hypothesize the non-autonomous Tm1 elements are derived from an ancestral autonomous Tm1-A element, in a manner consistent with models of transposon family evolution [59]. For example, Pack-MULES are derivatives of plant Mutator transposons that have acquired fragments of host genes [60,61]. Tm-1-HH elements, characterized by the presence of a highly conserved histone hairpin motif, may represent a similar acquisition of host sequences by a Tm1 element. Prior to the discovery of Cg-1 [24], histone-hairpin motif had been found only in the 39 UTR of transcripts of replication-dependent histone genes in metazoans. The highly-conserved 24-nt hairpin and sequences 39 serve as binding sites for an elaborate processing machinery that produces a specific endonucleolytic cleavage of the mRNA just 39 of the hairpin. As a result of this cleavage, these histone mRNAs are not polyadenylated [29]. Interestingly, the region of the Tm-1-HH element that is most conserved among this group is 39 of the histone hairpin structure, that is, the 39 end of the transcript that is predicted to be cleaved off by hairpin processing machinery. Further studies will be required to determine if the histone hairpin region is under selection and if RNAs produced by Tm1-HH elements have a function.
Transposable elements can be acquired through horizontal gene transfer (HGT; for reviews see [62,63,64]). However, despite the similarity of the Tm1-A transposase to insect Phantom transposases (Table 1), Tm1-A lacks sufficient DNA sequence conservation with these elements to demonstrate HGT as has been done for other transposons [65,66,67,68]. A more parsimonious assumption is that the Tm1 elements described here evolved from a functional Tm1-Alike element present in the ancestral Meloidogyne species rather than a recent horizontal gene transfer event from insects. Though firm conclusions cannot be drawn from current data, our observations are consistent with a model in which the Tm1-HH and Tm1-D classes arose as distinct deletion/insertion derivatives of Tm1-A in the M. incognita group and M. hapla lineages, respectively. Additionally, though the Tm1-ML elements are structurally similar between M. incognita and M. hapla, most show evidence of having evolved independently as deletion derivatives of larger Tm1 elements in each lineage: the majority of (20 of 30) M. incognita Tm1-ML elements have a M. incognita-type domain A1 (CGGTTAA; Figure 4) and the majority of (9 of 12) M. hapla Tm1-ML elements have a M. hapla-type domain A1 (CGGATAA; Figure 4). Additionally, the M. hapla Tm1-ML elements form a well-supported clade distinct from most M. incognita Tm1-ML elements ( Figure S1).
Our data is consistent with Tm1 elements having been mobile during the evolution of root-knot nematode species. The presence of an additional 8 Tm1-HH elements similar in sequence to that at Cg-1 in M. javanica but not in M. incognita ( Figure 3B) suggests that this particular element has expanded in the M. javanica lineage or lost in the M. incognita lineage. The presence of an insertion in the site flanked by primers 17-R and 17-F in M. javanica, but an empty site only in M. incognita suggests that this may be a relatively recent transposon insertion site (Fig. 1C). However, our Southern analyses also suggest neither Tm1-A or Tm1-HH elements have experienced recent transposition activity. Each isolate of M. javanica and M. incognita has a species-specific, rather than isolatespecific, hybridization pattern with the Tm1 probes ( Figure 3). In fact, our Southern analysis indicates that the beta actin genes of Meloidogyne are more polymorphic than Tm1 elements ( Figure 3A). It is important to note that most nematode isolates tested in this study (with the exception of M. incognita 557R, which was derived from a field collection in North Carolina) were collected from different locations in the California Central Valley and a wider sampling may be necessary to reveal rare Tm1 transposition events.
Consistent with the molecular observations suggesting recent Tm1 element inactivity, bioinformatically-detected Tm1 elements in the M. incognita and M. hapla genomes display sequence features consistent with senescence and deterioration. As Class II transposon families age, the number of autonomous elements decreases while the number of non-autonomous members increase and diverge in sequence [59,69]. Moreover, as transposable elements are commonly deleterious, asexual lineages are predicted to contain largely inactive and decaying transposable elements [70]. Consistent with these hypotheses, tested M. javanica and M. incognita isolates have retained only a single putatively autonomous Tm1-A element while M. hapla encodes no elements capable of encoding a transposase, and the majority of the Tm1 family in both M. incognita and M. hapla is composed of MITE-like elements. Furthermore, no two elements are identical in sequence, suggesting they have accumulated mutations, including insertions, since the expansion of the Tm1 family. Additionally, some Tm1 elements are missing TIR motifs, differ in copy number of Motif C repeats, or lack TSDs. One Tm1 element in M. hapla (contig 1532) has Motif A2 at the end of both TIRs and lacks a TSD, and may have arisen as a crossover event between two Tm1 elements. Moreover, the Tm1-HH element at Cg-1 appears to be degenerate and lacks terminal Motif A2 ( Figure 1A), possibly due to an intra or inter-element recombination event ( Figure 1A). Together these observations support that Tm1 elements represent a currently inactive and senescent transposon family in root-knot nematodes.
The role of the loss of the Tm1-HH element at the Cg-1 locus in the acquisition of ability to bypass resistance in tomato mediated by the Mi-1 gene remains unclear. Current models based on similar plant resistance genes predict that the Mi-1 protein either directly or indirectly recognizes the presence of a nematode product (elicitor) to trigger defense responses [71]. Previous results [24] suggest the virulent M. javanica VW5 strain is missing the elicitor that is present in its progenitor strain VW4 and support the importance of the Cg-1 transcript for Mi-1-mediated resistance. However, several Tm1-HH elements similar to that at Cg-1 are present in both strains VW4 and VW5, and DNA blots showed that Cg-1 is present in the virulent M. javanica field isolates Yolo-1 and Yolo-4 ( Figure 3 and [24]). It may be that the Tm1-HH element carrying Cg-1 regulates expression of a cis-linked nematode gene that is responsible for triggering resistance. Transposable elements contain enhancers and promoters, often in their TIRs, and thus can act as regulators of cis-linked genes (reviewed in [13]). RNAi silencing of Cg-1 may reduce or eliminate expression of the cis-linked gene by, for example, modifying chromatin structure in the region. In the case of VW5, the deletion of Cg-1 extends at least 2 kb 59 of the Tm1-HH element (S. Gross and V. Williamson, unpublished observations) and, thus, the cislinked elicitor gene may be deleted. However, these ideas remain untested as the extent of the deletion has not been determined and a cis-linked effector has not yet been identified.
Given the presence of multiple Tm1-HH elements in our virulent and avirulent isolates, the truncated nature of the Tm1-HH element at Cg-1, data suggesting the deletion extends beyond the Tm1 element at Cg-1, and the apparent inactivity of extant Tm1 elements, it is unlikely that canonical cut-and-paste transposition of Tm1 caused deletion of Cg-1 in VW5. However, other rearrangements involving the Tm1-HH element at Cg-1 may have occurred, such as ectopic recombination and chromosome breakage [72,73]. In fact, internally repetitive TIRs of Foldback elements appear to be hotspots for ectopic recombination, as has been shown in studies of chromosome rearrangements in natural populations of Drosophila [74,75]. In asexual lineages, which cannot acquire beneficial alleles that may arise in other members of the population, recombination between repetitive sequences leading to genome alterations may prove advantageous. Similar to Barbara McClintock's ''Genome Shock'' theory [76], asexual RKN parasites may use genomic change to their advantage when faced with the need to surmount plant resistance genes or perish. Such a mechanism has been proposed for some pathogenic bacteria, where effector molecules are clustered in ''pathogenicity islands'' flanked by transposable and repetitive elements and readily altered to avoid host defense responses [77,78]. Perhaps because its asexual reproduction mode benefits from use of repetitive sequences for adaptation, the genome of M. incognita has much more repetitive DNA than that of the sexual species M. hapla [10,11]. Whether Tm1-mediated genomic alterations produced the deletion of an effector molecule that allowed this nematode strain to bypass recognition by the tomato resistance gene product remains to be determined. However, such activity is consistent with hypotheses that transposons and repetitive elements contribute to creating genetic diversity in asexual organisms such as the root-knot nematodes M. javanica and M. incognita [11,12], and may represent an example of transposable element exaptation [13,79].

Molecular analyses
To prepare DNA, 200-300 ml of packed frozen eggs were pulverized in a mortar and pestle with liquid nitrogen. An equal volume of homogenization buffer (200 mM NaCl, 200 mM Tris pH 7.5, 20 mM EDTA, 2% SDS, 0.04 M 2-mercaptoethanol) was added to the mortar and grinding continued until eggs were completely homogenized. The homogenate was incubated at 40uC for 30 minutes with 0.2 mg/ml proteinase K. Homogenate was extracted using an equal volume of a 1:1 mixture of phenol:chloroform, then repeated with chloroform only. DNA was precipitated with 2 volumes ethanol and 0.1 volumes 3 M sodium acetate, washed in 70% ethanol, and resuspended in TE (10 mM Tris pH 8.0, 1 mM EDTA). Phenol:chloroform extraction and ethanol precipitation were repeated, and the DNA was resuspended in TE, treated with RNAse A, and quantified using a fluorometer. Prior to all PCR reactions, DNA was diluted to 5 ng/ml in TE.
Restriction digests of either 5 or 10 mg of genomic DNA were performed according to the restriction enzyme manufacturer's directions (New England Biolabs). DNA was separated in 1.0% LE agarose and 16 TBE and blotted to Hybond-N nylon membranes (GE Healthcare Life Sciences, Piscataway, NJ) following standard protocols [85]. Tm1-A probe templates were synthesized from a M8-1F and M8-6R PCR amplicon generated from a plasmid containing a 2.3 kb SG6 amplicon from VW4. Template for Tm1-HH probe was generated by amplification with primers a5 and a6 from a plasmid containing a cloned VW4 a5-a6 amplicon [24]. Beta actin probe was amplified from M. javanica VW4 genomic DNA template using primers Mj bActin-F and Mj bActin-R (95uC-30 s; 58uC-30 sec, 72uC-1 min). Probe templates were isolated from 1% TBE gels using a QIAquick Gel Extraction Kit (Qiagen, Valencia, CA). Randomly-primed probes were labeled with [a 32 P]-dATP (Perkin Elmer, Waltham, MA) according to published protocols [86]. Blots were hybridized in aqueous buffer at 60uC following standard protocols [85].
A plasmid library of ,3 kb HindIII fragments of M. javanica VW4 genomic DNA was constructed as follows: 50 mg genomic DNA was digested with HindIII and separated on a 0.8% LE agarose gel in 16 TBE. Four size fractions spanning the range of 2-4 kb were isolated from the gel, and DNA was extracted from these fractions using a QiaQuick Gel Extraction Kit (Qiagen, Valencia CA). The four fractions of gel-extracted DNA were amplified using primers SG1 and SG6 to identify the fraction containing the highest amount of Cg-1 PCR template. The HindIII fragments were cloned into Litmus38i (New England Biolabs) that had been previously treated with alkaline phosphatase. Approximately 10,000 independent clones were screened via PCR using pooling methods described previously [87].
A BAC library of M. javanica VW4 DNA was prepared with high molecular weight DNA isolated from J2s by Lucigen Corporation (Middleton, WI). A total of 12,672 BAC clones with an average insert size of 50 kb was screened via PCR with SG1 and SG6 primers using PCR conditions as above but with the annealing temperature lowered from 58uC to 54uC. The identified BAC clone (BAC 18M17) was isolated and purified according to the instructions provided by Lucigen. Sequencing and sequence assembly of BAC 18M17 was completed by the UC Davis College of Agricultural and Environmental Science Genomics Facility.

Genbank deposits
The following sequences were deposited at GenBank: M.
Predicted protein sequences were analyzed using Pfam v. 24.0 [32]. Nuclear localization signals were identified using PSORT II [40]. GeneMark-ES [93,94] was used to predict putative genes, using the C. elegans ES-3.0 model for gene prediction parameters. Figure S1 Unrooted neighbor-joining consensus tree demonstrating the relationship between M. hapla (red) and M. incognita (blue) Tm1-ML elements. Branch bootstrap support is noted on branches. Tree was generated using Geneious Aligner and Geneious Tree Builder (Drummond et al. 2008). Branch lengths were transformed to uniform length for clarity.

(TIF)
Table S1 Tm1 elements of Meloidogyne incognita. Target site duplication (TSD) length is indicated if both terminal motifs A1 and A2 are present; n.a. (not analyzed) indicates one or both of these motifs is missing; n.f. (not found) indicates that neither motif A1 or A2 was found. GenBank accession numbers, location on contig, TSD sequences, and sequences of 7 bp terminal motifs are provided in Table S3. When possible, elements are oriented with the Left TIR beginning with Motif A1 and the Right TIR terminating with Motif A2. Letters in parenthesis denote Tm1 elements on the same contig. Notes: a Identity between TIRs excluding motif A1 or A2. Tm1 elements with nested, uncharacterized transposable elements are noted as follows: b Contains 937 bp insertion flanked by 9 bp TSD,