The polychromatic Helitron landscape of the maize genome

Abstract

Maize Helitron transposons are intriguing because of their notable ability to capture gene fragments and move them around the genome. To document more extensively their variability and their contribution to the remarkable genome structure variation of present-day maize, we have analyzed their composition, copy number, timing of insertion, and chromosomal distribution. First, we searched 2.4 Gb of sequences generated by the Maize Genome Sequencing Project with our HelitronFinder program. We identified 2,791 putative nonautonomous Helitrons and manually curated a subset of 272. The predicted Helitrons measure 11.9 kb on average and carry from zero to nine gene fragments, captured from 376 different genes. Although the diversity of Helitron gene fragments in maize is greater than in other species, more than one-third of annotated Helitrons carry fragments derived from just one of two genes. Most members in these two subfamilies inserted in the genome less than one million years ago. Second, we conducted a BLASTN search of the maize sequence database with queries from two previously described agenic Helitrons not detected by HelitronFinder. Two large subfamilies of Helitrons or Helitron-related transposons were identified. One subfamily, termed Cornucopious, consists of thousands of copies of an ≈1.0-kb agenic Helitron that may be the most abundant transposon in maize. The second subfamily consists of >150 copies of a transposon-like sequence, termed Heltir, that has terminal inverted repeats resembling Helitron 3′ termini. Nonautonomous Helitrons make up at least 2% of the maize genome and most of those tested show +/− polymorphisms among modern inbred lines.

Helitrons are a class of eukaryotic transposable elements (TE) discovered only recently from a computational analysis of the fully sequenced genomes of Arabidopsis, rice, and C. elegans (1). These transposons vary greatly in abundance, even among closely related species. In fruit flies, for instance, they can comprise from 1 to 5% of the genome (2). Although they appear not to be present in humans (3), they can make up as much as 3% of the genome in other mammals (4). Among well studied plants, Helitron content has been estimated to be >2% in Arabidopsis thaliana (1), <0.01% in maize (5), and to range from 0.03 to 1.02% in different rice species (6). However, these values are most likely underestimates because Helitrons are hard to detect computationally given their lack of classical transposon structural features, such as terminal inverted repeats (TIRs) and target site duplications (TSDs).

The remarkable variation in genome structure among maize inbred lines has been attributed to recent LTR retrotransposon insertions and to the differential presence of genes or gene fragments (7, 8). Two Helitrons, HelA and HelB, carrying fragments from four different genes, accounted for all of the genic differences distinguishing the bz1 locus haplotypes of inbred lines McC and B73 (9). From a comparison of allelic BAC contigs, Morgante et al. (10) calculated that ≈10,000 gene fragments were inserted differentially in the genomes of the inbreds B73 and Mo17. Eight of nine insertions analyzed appeared to be nonautonomous Helitron-like transposons containing gene fragments, leading the authors to conclude that Helitrons were responsible for most of the genic sequence diversity between those two maize inbreds. Wang and Dooner (11) compared vertically the bz1 genomic regions of eight different inbred lines and land races that shared only 25% to 84% of their sequences and established that Helitrons could account for as much as 8% of the total sequence variation among haplotypes.

Most known maize Helitrons are relatively large (2–30 kb) nonautonomous elements containing gene fragments or pseudogenes (9, 10, 12, 13), although smaller elements (<1 kb) without gene fragments have also been identified (11). Some elements have fragments of a gene encoding a “RepHel” protein with replication initiator and DNA helicase domains (2). These elements are postulated to be derived by deletion from putative autonomous elements encoding an intact RepHel protein that presumably mobilizes Helitrons by rolling circle (RC) transposition (1, 10). However, evidence that this protein is involved in Helitron transposition is strictly circumstantial. In fact, it was recently reported that Helitrons can excise and leave footprints, an outcome not expected from RC transposition (14), so Helitrons may exhibit both copy-and-paste and cut-and-paste modes of transposition.

Whether Helitrons can capture an intact gene remains an open question. The presence within a Helitron of an almost complete maize gene has been reported for a cytidine deaminase and a P450 monooxygenase gene (15, 16), but there is no evidence that either is functional. Several Helitrons have been reported to produce chimeric transcripts that fuse exons from different genes (8–10, 12). Thus, Helitrons resemble the rice Pack-MULE TEs described by Jiang et al. (17) in producing chimeric transcripts from multiple gene fragments and their main importance may lie in their potential to create new plant genes through the rearrangement and fusion of fragments from multiple genomic loci.

Hollister and Gaut (18) investigated a class of abundant Arabidopsis Helitron elements named Basho and found that these elements had undergone rapid expansion and relatively few gene capture events. Sweredoski et al. (19) identified 23 Helitrons in maize and 552 Helitrons in rice by BLASTN searches of the nr database and concluded that there were clear differences between maize and rice with respect to gene capture. However, only a handful of maize Helitrons have been discovered by sequence alignments. Many more putative Helitrons were discovered recently by application of the HelitronFinder program to the GenBank nr database and several were confirmed empirically (20). Here, we have analyzed the Helitron content of the almost complete B73 inbred genome sequence generated by the Washington University in St. Louis Genome Sequencing Center. We find that the Helitrons of maize are more variable than those of any species analyzed to date. To gain a better understanding of their contribution to the highly polymorphic genome of present-day maize, we have analyzed their gene content, copy number, chromosomal distribution, and timing of insertion.

Results

Computational Identification and Analysis of Nonautonomous Helitrons in B73.

We used HelitronFinder (20), a program designed to detect Hel1 elements (21), to search 2.4 Gb of genome sequences from the maize inbred B73 (Release 2a.50, October, 2008). The program identifies putative Helitrons on the basis of conserved sequences at the Hel1 5′ and 3′ ends and assigns them to high, medium, or low probability categories, depending on the degree of sequence conservation. Among 16,205 BACs, we identified 2,791 putative nonautonomous Helitrons in the likeliest category. The maximum, minimum, median, and average sizes of the predicted Helitrons are 50 kb, 126 bp, 7,342 bp, and 11.9 kb, respectively. The 2,791 nonautonomous Helitrons make up 33.4 Mb or ≈1.39% of the sequenced B73 genome.

Alignment of the 5′ terminal 20 bp and 3′ terminal 30 bp of 21 Helitrons from different subfamilies (see below) revealed a high degree of conservation (Fig. 1). Analysis of the 3′ end, including the canonical palindromic sequence presumably involved in hairpin formation and transposition (1), confirmed that all of the identified Helitrons belong to the Hel1 group (20). Over 70% of the identified palindromes started within 33 bp from the 3′ end.

Fig. 1.

Alignment of the 5′ terminal 20 bp (Left) and 3′ terminal 30 bp (Right) of the Hel1 family. One representative of each of 21 Hel1 subfamilies was chosen for the alignment. Asterisks above the figure mark sites with 100% consensus among all sequences.

Verification of Putative Helitrons.

We selected nine predicted B73 Helitrons for verification by PCR in a panel of 13 inbreds. Presence of a PCR band of the size expected from a vacant site in at least one inbred line and its absence in B73 constitutes evidence of a +/− polymorphism in the panel (20). Primers for two predicted Helitrons gave nonspecific amplification and those for a third one failed to amplify fragments in any inbred, so these three were not considered further. The PCR results for the remaining six are summarized in Table 1. The polymorphic distribution of vacant and occupied sites for all six tested Helitrons establishes that they are not fixed residents of the maize genome and validates their prediction by HelitronFinder.

Table 1.

Molecular verification of predicted Helitrons

Inbred line	Predicted Helitrons
	Silco_101	Silico_102	Silico_103	Silico_106	Silico_107	Silico_110
4Co63	Vacant	Vacant	Occupied	Vacant	Vacant	Occupied
A188	Vacant	Occupied	Vacant	Occupied	Vacant	Occupied
BSSS53	Vacant	Vacant	Occupied	Vacant	Vacant	Occupied
B73	Occupied	Occupied	Occupied	Occupied	Occupied	Occupied
CML139	Vacant	Vacant	Occupied	Vacant	Vacant	Vacant
H99	Vacant	Occupied	Occupied	Occupied	Vacant	Occupied
I137TN	Vacant	Occupied	Occupied	Occupied	Vacant	Occupied
Ki3	Vacant	Vacant	Occupied	Vacant	Vacant	Occupied
M14	Vacant	Occupied	Occupied	Vacant	Vacant	Occupied
Mo17	Vacant	Vacant	Occupied	Occupied	Vacant	Occupied
McC	Vacant	Vacant	Occupied	Occupied	Vacant	Vacant
W22	Vacant	Vacant	Occupied	Occupied	Vacant	Vacant
W23	Occupied	Vacant	Occupied	Vacant	Vacant	Occupied

Vacant (no Helitron): PCR product obtained. Occupied (Helitron present): no PCR product. B73 is the positive control and H2O served as negative control.

Fig. 2 shows the results of Helitron vacant site amplification for Silico_102: Inbred lines 4Co63, BSSS53, CML139, Mo17, McC, W22, and W23 produce a strong PCR band, showing that they lack Silico_102 at this genomic location and are, thus, polymorphic with B73. For each predicted Helitron, the vacant site nature of the PCR-amplified band was confirmed by sequencing. The distribution pattern for all six Helitrons tested differs among the 13 tested inbred lines, implying recent Helitron movements in maize.

Fig. 2.

PCR verification in a panel of 13 inbred lines for Silico_102. Lanes from 1 to 15 are: DNA marker, 4Co63, A188, BSSS53, B73, CML139, H99, I137TN, Ki3, M14, Mo17, McC, W22, W23, and H₂O, respectively. The arrow points to the Helitron vacant site band in inbreds that lack Silico_102 at this genomic location. The diffuse larger fluorescent band in some lanes corresponds to genomic DNA.

An alternative approach to validate the predicted Helitrons is to compare duplicated regions of the maize genome. HelitronFinder predicts that there is a Helitron in chromosome 1 BAC AC190706. The AC190706 sequences 500 bp upstream and 400 bp downstream of the predicted Helitron are 100% identical with those of the homeologous region (22, 23) in chromosome 4 BAC AC212409 (Fig. S1).

Gene Fragments Captured by Helitrons.

To identify Helitrons that had captured gene fragments, we performed a BLASTN search (24) against transcribed monocot sequences in the TIGR Database (25). Approximately 94% of the predicted Helitrons were found to carry fragments from transcribed genes. We further curated 272 of our 2,791 predicted Helitrons. Only 27 (10%) lacked gene fragments. Of these, the longest was 4,071 bp long, but 20 were smaller than 1 kb. Thus, most agenic Helitrons detected by HelitronFinder are smaller than 1 kb.

We carefully characterized a subset of 158 of the curated Helitrons in terms of percentage sequence identity among their 5′ and 3′ ends, copy number, size, and gene fragment content. The results obtained after removing duplicate entries from overlapping BAC sequences are summarized in Table 2. Elements sharing ≈90% end sequence identity grouped into subfamilies, which, in general, shared all or some gene fragments (Table S1). The size of the elements ranged from 0.2 kb to 39.2 kb, the number of gene fragments, from 0 to 8, and the copy number of elements within a subfamily, from 1 to 79.

Table 2.

Characteristics of maize Helitron subfamilies

Helitron subfamilies	Copy number	5′ end identity, %	3′ end identity, %	Size, kb	No. of gene fragments
Hel1_100a-e	5	80–100	70–100	6.2–11.6	2–7
Hel1_101a-c	3	100	94–98	5.8–6.2	4–5
Hel1_102a	1	NA	NA	7.9	6
Hel1_103a	1	NA	NA	5.3	2
Hel1_104a	1	NA	NA	0.9	0
Hel1_105a-a4	79	61–100	78–100	1.4–10.4	1
Hel1_106a-z	26	57–100	86–100	3.8–39.2	1
Hel1_107a-b	2	100	100	8	4
Hel1_108a-b	2	100	100	3.6–9.9	3
Hel1_109a	1	NA	NA	39.4	6
Hel1_110a-c	3	100	100	8.8	3
Hel1_111a	1	NA	NA	15.1	2
Hel1_112a	1	NA	NA	6.6	1
Hel1_113a-b	2	100	100	2.1	0
Hel1_114a-b	2	85	88	8.2–11.0	2–4
Hel1_115a-b	2	100	100	2.7–23.5	1–2
Hel1_116a-b	2	100	98	13.1–13.6	7
Hel1_117a	1	NA	NA	4.5	1
Hel1_118a	1	NA	NA	16.4	5
Hel1_119a	1	NA	NA	2.1	1
Hel1_120a	1	NA	NA	7.9	5
Hel1_121a-d	4	100	100	2.6	1
Hel1_122a	1	NA	NA	10.9	7
Hel1_123a-b	2	61	90	0.5	2–3
Hel1_124a	1	NA	NA	17.6	7
Hel1_125a-h	8	95–100	82–94	0.9–7.2	0
Hel1_126a	1	NA	NA	14	8
Hel1_127a	1	NA	NA	12.3	3
Hel1_128a-b	2	95	85	0.2	0

NA, not applicable

Two subfamilies, Hel1_105 and Hel1_106, were particularly numerous. The 79 members of the Hel1_105 subfamily (Table S2A) contain sequences homologous to a rice gene encoding a phosphatase-like protein (Os05g0358500 NP_001055303). Most of the Helitrons in this subfamily carry a segment of the gene consisting of ≈180 bp from exon 2 and ≈240 bp from exon 3, but a few either lack the fragment from exon 3 or carry an additional 40-bp fragment from exon 4. The size of most Helitrons in this subfamily ranges from 1.5 to 2.5 kb, the sole exception being Hel1_105Y3, which includes an 8-kb retrotransposon. A BLASTN search of the B73 genome sequence with the 1,617-bp Hel1_105a element uncovered 1,505 additional members of the Hel1_105 subfamily, making this the largest subfamily of gene-fragment-bearing Helitrons in B73 (Table S3A). HelitronFinder detected most of these elements, but assigned them a medium probability value because of polymorphisms in the 5′ and 3′ end sequences.

The second largest subfamily is Hel1_106, composed of Helitrons that have captured fragments of a CesA1 cellulose synthase gene. Twenty-six members were identified among the manually curated Helitrons (Table S2B) and 58 additional members from a BLASTN search of the B73 sequence database using Hel1_106a as query (Table S3B). The average size of these elements is ≈4 kb. Most carry a 1.12-kb fragment of CesA1 (NP_001104954.1 GI: 162460417) that extends from the middle of exon 1 to almost the 3′ end of intron 2. Some contain shorter fragments from the same CesA region. The much larger 39-kb Hel1_106s element contains, in addition, a retrotransposon insertion. In all of these cases, the introns separating adjacent exons have also been captured by Helitrons, allowing us to predict that the CesA1 gene fragment was probably captured from a genomic copy in chromosome 8 (AC198758).

We manually annotated 100 additional Helitrons that had captured gene fragments (Table S4: Hel1_200a to Hel1_311a). We did not determine the copy number of this group because we were mainly interested in the diversity of maize genes that can contribute fragments to Helitrons. Some Helitrons appear to have two 5′ ends and only one 3′ end or vice versa (e.g., Hel1_201a). Although these apparently double-headed or double-tailed Helitrons may move as a unit, we only considered the internal Helitron component with single 5′ and 3′ ends in our analysis of gene content. These 100 Helitrons carried a variable number of gene fragments, ranging from one to nine. The size of the four Helitrons that captured nine gene fragments (Hel1_303a, Hel1_302a, Hel1_299a, and Hel1_284a) ranged from 13 to 34 kb. Hel1_303a carried fragments from nine genes, some of which are linked in the genome. The genes for flavonol synthase, TPR-like protein, and PDIL 1–5 genes are all on chromosome 9 at physical map locations of 22.55, 29.84, and 22.28 Mb, respectively. Similarly, four of the nine gene fragments in Hel1_299a come from genes in chromosome 3 that map at 31.10, 136.42, 31.56, and 41.70 Mb, respectively. However, each of these Helitrons also contained gene fragments from different chromosomes. The fragments captured by these 100 manually annotated Helitrons come from >200 different genes. Thus, the diversity of Helitron-embedded gene fragments is much greater in maize than in rice or Arabidopsis.

Our 272 curated Helitrons captured gene fragments from 376 different genes (Table S5). This indicates that maize Helitron sequences contain a wide variety of gene fragments. Genes encoding retrotransposon gag-pol proteins or fragments thereof are also abundant in Helitrons. Copies of elements with at least 70% sequence identity usually retain common gene fragments. The increase in length of related Helitron copies is often the result of retrotransposon insertions. Some possibly full-length genes may have been captured by Helitrons, such as a putative casein kinase I by Hel1_309a. The majority of these are putative genes and additional evidence is required to verify that they are functional.

Helitron Copy Numbers.

Among manually annotated Helitrons, smaller ones tend to have the highest number of copies (e.g., Hel1_105: Table 2). Outside of the two largest copy-number subfamilies, Hel1_105 and Hel1_106, there are 14 subfamilies with one copy; 8 with two copies; 2 with three copies; and 1 each with four, five, and eight. Most single-copy Helitrons measure several kilobases and only one of them is smaller than 1 kb.

The agenic Helitrons Hel1-4 and Hel1-5 discovered by Wang and Dooner (11) from a vertical comparison of bz1 haplotypes were scored by HelitronFinder as being less probable than the 2,791 previously discussed Helitrons and were omitted from the above analysis. To obtain an estimate of the copy number and distribution of these Helitrons in B73, we conducted BLASTN searches against the current release of the maize genome sequence database (3b.50: Feb. 2009) using each of these sequences as queries. Hel1-5 turned out to be not only the most numerous Helitron, but probably the most numerous transposon in maize. It is present in so many copies in the corn genome (≈3,000: Table S6) that we have renamed it Cornucopious. The Cornucopious subfamily consists of a variety of elements with different degrees of sequence identity to the original 994-bp Hel1-5 element (11).

The 469-bp agenic Hel1-4 element discovered in the bz1 haplotype of A188 is not found intact in B73. Hel1-4-homologous sequences are present in several copies in B73, but they are split. The 5′ terminal 336 bp are present in 43 copies that cover >70% of the sequence at >90% identity. The 3′ terminal 104 bp are present in 158 copies that cover >66% of the sequence at >90% identity. Interestingly, these 158 copies are part of a larger (≈390-bp) sequence flanked by perfect 37-bp TIRs that are homologous to the 3′ end of Hel1-4 and that we have designated Heltir for Helitron-like sequence with TIRs (Fig. 3). Heltirs have two Helitron 3′ ends, are almost invariably flanked by an A and a T (aCTAGGT..ACCTAGt), as would be expected from the canonical site of insertion of Helitrons, and are inserted in TA-rich regions. An example of a Heltir from chromosome 10 is given in Fig. S2.

Fig. 3.

Comparison of Hel1 and Heltir end sequences. The 5′ terminal 20 bp and 3′ terminal 30 bp of Hel1 and Heltir elements are shown. The TIRs of the Heltir element are marked by an arrow and their sequence identity to the 3′ end of Hel1 elements is indicated by vertical strokes. The palindromic sequences at the 3′ end of Hel1 and both ends of Heltirs are underlined and the AA at the center of the Hel1 palindrome is in bold.

Besides these two short elements, there are probably other agenic elements that were given a low probability score by HelitronFinder and could be present in multiple copies in the maize genome. The sequence variability of Helitrons, even at their relatively conserved termini, makes it difficult to estimate the true percentage of the maize genome that is made up by this transposon superfamily.

Helitron Divergence Time.

The divergence time of Helitrons cannot be estimated as readily as that of LTR retrotransposons because Helitrons lack duplicate sequences that become homogenized upon transposition. Hollister and Gaut (18) used terminal branch length (TBLs) as a proxy for estimating the time of Helitron insertions. We applied their method to estimate the time of insertion of elements in the two most abundant gene-capturing subfamilies, Hel1_105 and Hel1_106. The majority of Helitrons in these two subfamilies inserted recently in the maize genome: 71% of Hel1_105 elements and 69% of Hel1_106 elements (Table S7) are <1 million years old and >96% of elements in both subfamilies are <5 million years old. This dating indicates that most recognizable Helitrons in the genome sequences of present day maize have transposed recently.

Dynamic Distribution of Helitrons.

Most Helitrons appear to be proliferating actively and are able to capture genes or gene fragments from multiple locations in the genome. The BAC sequence AC183508 (Table 3) contains an interesting nested Helitron, Hel1_260a/Hel1_261a (Table S4). Although nested elements are common among LTR retrotransposons, this was the only Helitron nest detected in B73. However, some of the predicted Helitrons have multiple ends, either two or three 5′ ends or two 3′ ends (Table 3). Hel1_103a has three 5′ ends and 17 others have two 5′ ends, but all of them have one 3′ end. Nine Helitrons have two 3′ ends, but a single 5′ end. This lack of symmetry could be due to failure of the HelitronFinder program to detect a less conserved 5′ or 3′ end or, alternatively, to a real truncation of the elements. We recently found a case of two Helitrons adjacent to, rather than nested within, each other in the r1 region from chromosome 10 (20) and here we have found an additional case in sequence AC187094 from chromosome 7 (Table S3A and Table S4).

Table 3.

Helitrons with multiple ends

Helitron	Accession no.	Start 5′_1*	Start 5′_2*	Start 5′_3*	End 3′_1*	End 3′_2*
Hel1_103a	AC190799.1	77100	76942	73315	71740
Hel1_105 h2	AC200479.3	72973			71230	58668
Hel1_105a3	AC202878.2	99277	99572		101195
Hel1_109a	AC149034.3	54590	51684		15101
Hel1_201a	AC191691.3	170683	167108		160573
Hel1_203a	AC183928.4	71437	62566		58077
Hel1_206a	AC184164.2	130023	131805		138112
Hel1_225a	AC185470.2	152279	161454		162631
Hel1_231a	AC186166.3	95281	93840		92310
Hel1_233a	AC186223.3	35296			38049	58820
Hel1_237a	AC186815.3	50586	54373		59681
Hel1_239a	AC185211.3	90120	92586		94472
Hel1_240a	AC186329.3	32604			42102	55571
Hel1_244a	AC177874.2	192724	183714		176844
Hel1_245a	AC185499.2	161144			154517	154206
Hel1_248a	AC190964.2	168988			163728	160378
Hel1_257a	AC190498.1	24663	17072		15643
Hel1_260a/261a	AC183508.3	112810	87052		78952	78616
Hel1_274a	AC187845.3	116665	117678		120258
Hel1_276a	AC177926.2	139016			132784	120710
Hel1_285a	AC177904.2	155390			123713	105891
Hel1_289a	AC193322.3	41857			57972	58095
Hel1_292a	AC190626.3	142543	144299		149859
Hel1_297a	AC186577.3	26879	27883		30462
Hel1_298a	AC186577.3	39172	42750		50201
Hel1_299a	AC186189.3	90063	86847		75250
Hel1_301a	AC191802.3	26106			27813	68593
Hel1_306a	AC191705.3	128700	128105		127390

*Location of the 5′ end first base or the 3′ end last base in the BAC sequence listed under the accession no. column.

To examine whether Helitrons were distributed randomly among the 10 maize chromosomes, we tested by χ² whether the distributions of the two most abundant subfamilies, Cornucopious and Hel1_105, deviated significantly from those expected based on chromosome lengths. Both of them did (Fig. S3). The calculated χ² values are 28.98 and 24.95, respectively, and the probability of obtaining deviations from expectation as large as or larger than those observed based on the null hypothesis is <0.01 in both cases, suggesting that the distribution of Hel1 type Helitrons is nonrandom in maize. Interestingly, however, a χ² test of heterogeneity indicates that the two observed distributions do not differ significantly from each other (χ² = 16.86, 9 df, P > 0.05)

Discussion

Nonautonomous Helitrons make up at least 2% of the maize genome, a much higher percentage than that reported in ref. 5. We estimate that there are ≈4,000 gene-fragment-bearing Hel1 Helitrons in B73 and that the average Helitron outside of the Hel1_105 subfamily contains 1.75 genes. Thus, our number is in agreement with the ≈10,000 +/− gene sequence polymorphisms between B73 and Mo17 detected by Morgante et al. (10) and interpreted by them as being caused most likely by nonautonomous Helitron-like transposons. We are currently modifying our HelitronFinder program with the 2,791 predicted Helitrons as our training dataset in an attempt to make it less stringent, but still accurately predictive. A preliminary survey of the B73 sequence database with this less stringent version of HelitronFinder yields 5,000 additional predicted Helitrons. Furthermore, known short agenic Helitrons, which were not placed in the highest probability category of predicted Helitrons by our original HelitronFinder program, can be very numerous. A BLAST search with the agenic 0.9-kb Hel1-5 element first identified in the bz1 haplotype of inbred I137TN (11) revealed >3,000 copies in the B73 genome, making it the most abundant corn transposon and prompting us to rename it Cornucopious. Given the large number of copies of this element, we did not attempt to resolve duplicate entries in the htgs maize sequence database. Nevertheless, even assuming a 30% overlapping coverage in the sequence of the B73 genome (www.maizesequence.org/faq.html), the Cornucopious Helitron subfamily would still be represented by >2,000 members. It is clear from the above discussion that the percentage of Helitrons in the maize genome could be much higher than we report here.

The chromosomal distribution of the two largest subfamilies of Hel1 Helitrons, Cornucopious and Hel1_105, was found not to be proportional to chromosome length in either case. If Helitrons did transpose by RC replication and formed tandem arrays, as reported in Myotis (4), some chromosomes might accumulate a disproportionate number of a particular Helitron relative to other chromosomes. However, we did not observe Helitron concatemers and the elements in the two examples of adjacent Helitrons that we detected in the B73 genome differed from each other. Although the calculated χ² values are significant, the differences between the observed and expected distributions (Fig. S3) are not glaringly obvious and, in fact, the two observed distributions do not differ from each other. A plausible explanation of this result is that the maize genome is not yet finished and sequences from some chromosomes, such as 4 and 10, may be over-represented in the current sequence release (www.maizesequence.org/index.html).

Not all maize Helitrons were found to be inserted between an A and a T. We identified 18 Helitrons (6.6%) that were inserted into CT, GT or TT dinucleotides, instead of the characteristic AT dinucleotide. Elements inserted into a TT target site have been reported before (10). This finding has prompted us to remove the strict requirement for a specific nucleotide adjacent to the Helitron 5′ end from our HelitronFinder program to increase its ability to find maize Hel1 elements.

Analysis of the gene content of Helitrons revealed that maize Helitrons have the capacity to capture a wide variety of gene fragments. These fragments showed different degrees of similarity with homologous genes found in maize or rice. The fragments captured by our set of 272 curated Hel1 Helitrons range in size from 28 bp to 7.6 kb and come from 376 different genes. In contrast, the gene fragments found in Arabidopsis thaliana Basho elements measure from 30 to 350 bp and derive from only five genes (18). Thus, maize Helitrons possess a greater, possibly species-specific, ability to capture gene fragments. Several of the annotated Helitrons may contain full-length genes. However, because many of those genes are hypothetical, putative or unknown and the corresponding full-length cDNAs have not been found, additional evidence is required to confirm that the proteins they encode are expressed and functional. Some elements captured gag-pol gene fragments, creating mimics of the retrotransposon insertions into Helitrons found by Morgante et al. (10) and by us here (e.g., Hel1_105y3). Other TE nests, such as Helitrons into LTR retrotransposons (15, 16) and DNA transposons into Helitrons (10) were also observed in this study.

Although Helitron-borne maize gene fragments were identified by their homology to rice genes, some are also homologous to previously annotated maize genes. In a few cases, the host gene contributing a specific fragment could be identified on the basis of high (>95%) sequence identity. Helitrons bearing fragments from multiple genes were analyzed for possible linkage of the donor genes. Some Helitrons, like Hel1_303a and Hel1_299a, appeared to have captured fragments from relatively closely linked donor genes, but adjacent gene fragments often came from different chromosomes. For example, Hel1_299a carries adjacent fragments from a chromosome 3 gene encoding a carbohydrate transporter and a chromosome 5 gene encoding an unknown protein. As noted by Morgante et al. (10), gene fragments tended to have the same transcriptional orientation. The overall ratio of 5′-3′ vs. 3′-5′ orientation of fragmented genes relative to the ends of the capturing Helitron was 4:1 (427:102) in 156 annotated maize elements.

The original model of gene capture by Helitrons (26) invoked a malfunction or deletion of the 3′ end RC terminator and the de novo formation of a terminator-like signal located downstream. Although the probability of random DNA functioning as a de novo terminator has been questioned (2), our study shows that the 3′ termini of maize Helitrons are more variable, their sequence identity ranging from 11% to 100%, whereas the 5′ termini are more conserved, their identity ranging from 57% to 100%. Our dataset includes several complex Helitrons with more than one 5′ or 3′ terminus. Their occurrence could be explained by deletion of either the 5′ or 3′ internal end from nested Helitrons or from closely linked Helitrons, as in the model of gene capture via two Helitrons proposed by Kapitonov and Jurka (2). Therefore, both possibilities regarding the occurrence of multiple 3′ end fragments could be valid. However, one aspect of gene capture not readily explained by either model is the strong asymmetry in the orientation of Helitron-captured gene fragments discussed above.

The agenic ≈0.4-kb Hel1-4 Helitron (11) is unusual in that related TIR-containing sequences, which we have named Heltirs, are found in the B73 genome in a high number of copies (≥158). Their structure and repetitive nature suggests that they are transposons. Their TIRs are perfect, measure 37 bp, are highly similar to the conserved Hel1-4 3′ end, and are flanked by an A and a T, observations that suggest an unexpected relationship between Helitrons and TIR-flanked transposons (Fig. 3). If a TIR-binding, conventional transposase mobilizes Heltirs in maize, the same enzyme could be responsible for the recently reported maize Helitron excision footprints (14).

“Young” Helitrons, with an estimated time of insertion <1 million years, comprise 69% to 72% of the two largest subfamilies of gene-fragment-bearing Helitrons, Hel1_105 and Hel1_106. “Old” Helitrons, with an estimated time of insertion >5 million years, comprise only 2.6 to 7.7% of the total. Our analysis reveals that an expansion of elements with a size between 1 and 2 kb occurred <1 million years ago. We found that small elements tend to have more copies, indicating that they transpose more easily. However, although a large fraction of small Helitrons inserted <1 mya, older members of the same subfamily are also observed, indicating that small Helitrons can persist in the genome for a long time. Thus, maize Helitrons represent a group of transposons that have actively reshaped the genome of present day maize. No statistically significant correlation between the size and age of Helitrons was observed. In contrast, Hollister and Gaut (18) noted that larger Helitron elements are less likely to persist in the Arabidopsis genome and suggested that this may be due to selection against the deleterious effects of interHelitron ectopic recombination. Whether maize Helitrons can participate in such ectopic recombination events today is questionable. In one study designed to detect meiotic recombination between two pairs of Helitrons at allelic locations, no recombination events were detected and the elements were found to be densely methylated (27). The identification of additional new elements and the expanded curation of our dataset to include all predicted Helitrons will allow us to draw a clearer picture of the evolutionary dynamics of maize Helitrons.

With the maize genome sequencing and annotation on the way, it will be interesting to investigate further the relationship between Helitron age and size, the possibility of linkage among captured fragments, the distance of Helitrons to donor genes, and the gene density of donor chromosome segments. The sequencing of additional maize lines will reveal whether other Helitron transposons, different from the highly abundant Cornucopious of B73, have amplified to a high degree. If so, the differential escape from transposition suppression by the host genome of a specific transposon within a family would be reminiscent of the dramatic amplification of an mPing MITE described by Wessler and associates in four different rice strains (28).

Materials and Methods

Prediction of Helitrons.

We identified Helitrons computationally with the HelitronFinder program (20). We then validated a sample of the predicted Helitrons by both molecular and in silico methods. The predicted Helitrons were used to retrain HelitronFinder so as to produce more generalized predictions. We annotated a portion of the predicted output Helitrons by blasting them against the htgs GenBank databases. Because of redundancy in the B73 GenBank entries, we individually checked the chromosomal clone location of curated Helitrons to eliminate potential duplicates.

Identification of Gene Fragments.

We downloaded the TIGR Plant Transcript Assemblies (25) into our server. All 2,791 predicted Helitrons were compared by BLASTN (24) against the sequence databases of monocot ESTs, full length and partial cDNAs. We then curated a subset of 272 Helitrons to identify captured gene fragments. The GeneSeqer tool from MaizeGDB (http://maizegdb.org) was used to infer potential exon/intron structure in premRNA. The predicted protein sequence was compared by BLASTP against the nr/protein National Center for Biotechnology Information database. The predicted gene sequences were blasted against the Helitron sequences to identify the embedded gene fragments. We further conducted BLASTX searches against the nr protein sequences in the National Center for Biotechnology Information database to identify protein-coding regions with an e value cutoff of e⁻¹⁰.

Estimation of the Time of Divergence of Helitrons.

We adopted the terminal branch lengths method (18) to estimate the time of Helitron insertions, using a substitution rate of 1.3 × 10⁻⁸ per site per year for intergenic regions (29). We aligned the 3′ terminal 50 nucleotides of the Hel1_105 and Hel1_106 Helitrons by ClustalX (30) and carried out a neighbor-joining phylogenetic analysis for each subfamily using the MEGA 4 (31) software package with 1,000 bootstrap replicates.

Molecular Validation of Predicted Helitrons.

The inbreds used in the Helitron validation panel were: 4Co63, A188, BSSS53, B73, H99, M14, Mo17, W22, and W23, North American breeding lines; CML139, from CIMMYT; I137TN, from South Africa, and Ki3, from Thailand (32). McC carries the Bz-McC haplotype in a W22 background (9). Primer design, cloning, and sequencing were as described in ref. 20.

Note Added in Proof.

The maize genome sequence data used in the present analysis are reported in: Schnable PS, et al. (2009) the B73 maize genome: Complexity, diversity and dynamics. Science, 10.1126/science.1178534.