In parallel to the application of new analytical methods, the avian phylogenetic tree has undergone substantial changes in the past decades. But even today many branchings remain highly controversial, although it is widely accepted that modern birds fall into two major clades, (1) Palaeognathae, a clade comprising rheas, kiwis, ostrich, emus, cassowaries, and tinamous and (2) Neognathae, comprising Galloanseres (fowl and waterfowl) and Neoaves (all other taxa) 12. Within Galloanseres, Galliformes (gamebirds) are traditionally classified into five families: Megapodiidae (megapodes, brush turkey and allies), Cracidae (currassows, guans, and chachalacas), Odontophoridae (New world quails), Numididae (Guinea fowl), and Phasianidae (pheasants, peacocks, partridges, and allies) 3456.

While virtually all studies identify megapodes and cracids as successive sister taxa of the remaining Galliformes, the branching orders of Odontophoridae, Numididae, and Phasianidae, with its presumed subfamilies Tetraoninae (grouses) and Meleagridinae (turkeys), are less clear. Especially, the interrelationships between Numididae, Odontophoridae, and Phasianidae are considered a "major puzzle" of galliform phylogeny 7. Also debated are the exact affinities of Tetraoninae and Meleagridinae. In traditional classifications, these two taxa are separated from the other Phasianidae, whereas molecular sequence analyses support a position of Tetraoninae and Meleagridinae deeply within a clade including the other Phasianidae 891011. Recently, Kaiser et al. 12 investigated phylogenetically informative retropositions in Galliformes and found significant support for two clades, (I) a monophyletic Phasianidae including Meleagridinae and Tetraoninae and (II) a clade comprising Meleagridinae, Tetraoninae, Phasianus, and Tragopan. The relative positions of Numididae and Odontophoridae and the topology of the phasianid tree differed from other sequence-based studies depending on the genes investigated and the analytical methods used 91011131415161718.

Discrepancies in phylogenetic reconstructions based on various paleontological, morphological, behavioral, and molecular methodologies are often due to the presence of homoplastic characters 19. Markers that are less likely to be confounded by problems of homoplasy include rare genomic changes (RGC) such as random insertions and deletions (indels) and retroposed elements 20. Indels are frequently used for phylogenetic reconstructions 1921222324 and the presence/absence patterns of retroposed elements have proven invaluable for reconstructing virtually ambiguity-free phylogenetic trees 2526272829. Presence/absence data resemble virtually homoplasy-free multistate characters with an extremely large possible number of unique character states. Steel and Penny 30 suggest that for this kind of data, maximum parsimony converge to a maximum likelihood estimator. The clear "presence" of a retroposed element at orthologous positions in related taxa indicates a derived condition acquired via a common ancestor, while its "absence" in more distant taxa represents the plesiomorphic condition prior to integration. Retroposed elements contain several features that, on their own, are very unlikely to occur twice independently at orthologous genomic positions. These include defined subtypes of retroposed elements, diagnostic mutations, and characteristic truncations of the consensus retroelements. Although presence/absence patterns are virtually homoplasy-free, there does exist, as for any other marker system, a low probability of lineage sorting 20 and a slight chance of exact excision of retroposed elements with perfect direct repeats 31. These caveats aside, a statistical framework was developed to evaluate presence/absence data 32, and presence/absence patterns of retroposed elements have now been successfully used to reconstruct, for example, the placental mammalian tree at the superordinal level 27, the monophyly of Cetartiodactyla 33 and Pegasoferae 34, the position of Primates within Supraprimates 29, and internal primate relationships 353637.

In the chicken genome, retroposed elements of the chicken repeat 1 (CR1) family of Long INterspersed Elements (LINEs), with more than 200,000 copies, constitute 80% of all interspersed repeats and 3.1% of the entire genome 3839, while the second largest fraction of retroposed elements, the Long Terminal Repeat elements (LTRs) of endogenous retroviruses, with 12,000 copies, constitute only 4.7% of all interspersed repeats 3839. As CR1 elements do not show target site duplication (direct repeats) 4041, excisions, such as that described by van de Lagemaat 31, cannot occur, making them the most suitable retroposed elements in bird genomes for phylogenetic purposes 42.

While full-length CR1 elements are 4.5 kb long and contain two open reading frames 41, most CR1 sequences are truncated copies of their autonomous full-length master genes 3943. CR1 subtypes are characterized by diagnostic mutations that occurred in their specific master genes. Different full-length copies of master genes remained transcriptionally active over long, overlapping, periods of time and distributed corresponding retroelements in specific waves of activity as has been described in penguins 44. To efficiently select phylogenetically informative CR1 elements from the chicken genome, it would be helpful to know which CR1 elements were active at which evolutionary time points. As CR1 elements, like most other retroposed elements, integrate almost randomly into the genome, they also frequently insert into other CR1 copies. But, at a given point in time only the active CR1 subtypes can insert into copies of their own or other CR1 subtypes (Figure 1). This provides information about which 'host' CR1 subtypes were already integrated at this particular time point. If the reverse case, in which the 'host' subtype inserted into an active CR1 subtype, cannot be found in the entire genome, one can assume that the 'host' subtype was probably already inactive at this particular time point. Thus, an analysis of the patterns of nested CR1 elements (transpositions in transpositions that we call the TinT method, Figure 1) provides a relative timetable of active CR1 elements. A first step towards a genome wide characterization of the activity ranges of CR1 elements is to search for the distribution patterns of nested retroposons. Churakov et al. 45 recently applied a novel method based on the single-case patterns of nested retroposons to characterize the historical appearance of various armadillo-specific SINE subfamilies. Similarly, Ichiyanagi and Okada used this method to determine the full lengths of SINEs in zebrafish 46 and Pace and Feschotte to investigate DNA transposon activity in the human genome 47.

To further investigate and reconstruct the phylogenetic tree of Galliformes, we developed a multidimensional, computer applicable model for computing the frequencies of TinT genome-wide. Using this model, we describe the waves of activity or fixation patterns of various CR1 subtypes. We then used this information to directly select specific subtypes of retroposed elements that were active on the galliform evolutionary lineage leading to the chicken (Gallus gallus). These element subtypes were used to experimentally extract phylogenetic informative orthologous sequences from representative loci of all galliform families. As LTR elements also insert into each other, but were not frequent enough to apply the TinT method, we investigated their random insertions in the chicken genome for potential phylogenetic use. Furthermore, other phylogenetic signals (indels) observed during the genomic alignments of our presence/absence loci provided support for additional clades. With this multifaceted approach we reconstructed the major aspects of galliform evolution.

To obtain a relative temporal order of CR1 element activity for phylogenetic use, we explored the patterns of nested retroposed CR1 subtypes (Figure 1) from the chicken genome. From a genome wide collection of annotated retroposed elements we extracted all 1,978 cases of nested CR1 elements. The resulting matrix (additional data file 1) was used to calculate a multidimensional model (additional data file 1) giving the maximum probability of activity for each of the 22 CR1 subtypes on a relative timescale (TinT, see additional data file 1). The model makes the following assumptions: (i) For each CR1 subtype there was one limited period of activity. (ii) There was no known target site preference for the CR1 subtypes, thus each individual copy could have inserted at any random position in the genome. This could have been either an anonymous sequence or another CR1 copy. (iii) The number of copies of any given CR1 subtype in the genome reflects the duration of its activity. (iv) The temporal fixation rate of each CR1 subtype can be described by a normal distribution as is shown by the divergences of its single copies from their consensus sequences (additional data file 1). (v) The probabilities of fixation among the individual CR1 copies during their specific activity periods were relatively equal (equal promoter activity, equal affinity of reverse transcriptase to the mRNA, and an equal availability of reverse transcriptase). Based on these assumptions we developed a function describing the behavior of the fixation of each CR1 subtype on a relative timescale. Using the maximum likelihood approach we calculated the maxima of probability of fixation for each CR1 subtype.

The results show that the various CR1 subtypes differ not only in the median time points on the relative timescale during which they were actively fixed in the genome, but also in the lengths of these time windows, revealing three major peaks of concentrated CR1 activity/fixation (Figure 2A, additional data file 1). The oldest peak is dominated by CR1 subtypes Y4, D, X2, E, C4, Y, F2, D2, and X elements, the middle peak by CR1 subtypes Y2, C3, G, F2, X1, D2, H, Y4, E, and C, and the youngest peak by CR1 subtypes H2, F0, B2, F2, D2, and C2.

To verify the relative times of TinT activities we calculated the average divergencies of all CR1 subtypes from their consensus sequences. Assuming random accumulation of mutations, the degree of divergency should then be age-related. There was a significant correlation between the relative time scales of the cumulative TinT and the CR1 divergencies (R = -0,6489, P << 0.01) (additional data file 1).

Armed with the relative ages of chicken CR1 subtypes deduced from the cumulative TinT, we selected a representative set of diverse CR1 elements, whose activity periods spanned the entire time frame of galliform evolution, to use as experimental probes of phylogenetic branch points of the other galliform species. Intronic sequences of the chicken genome (17,300 introns; maximal length of 1 kb) were screened for embedded CR1 elements and LTR elements (RepeatMasker) (300 CR1 and 45 LTR elements). These were inspected by eye in the genome browser (UCSC) 48 and the most conserved loci (120 cases), representing the broad activity range of CR1 subtypes as compiled by the TinT method, were chosen to generate conserved PCR primers for amplification of orthologous loci in representative galliforms (see Methods for species sampling). Of these, 19 loci were successfully amplified in all important taxa, revealing a total of 25 phylogenetically informative retroposed elements. CR1 elements present in these amplified loci, along with those presented by Kaiser et al. 12 show that at least the CR1 subtypes E, Y2, X2, Y4, F2, D2, H2, C, C2, B2, H, and G were active during galliform evolution. Moreover, the results of the cumulative TinT analysis are clearly in line with the exemplarily shown activities of certain CR1 elements identified by their presence/absence patterns in various species (Figure 2B). The elements represented in the older cumulative TinT peak we found as well to be active during the first divergences in galliform evolution, while the elements of the second peak were active during times of younger divergences. Thus the phylogenetic markers present an actual calibration of the TinT relative timescale. In further investigations we focused on screening for intronic CR1 elements with TinT-selected activity patterns in a galliform-wide amplification.

As our analysis of chicken sequences did not furnish elements that retroposed after the divergence of the lineage leading to turkeys from the one leading to the chicken, to provide phylogenetic information to solve the potential sister group relationships on this lineage we scanned the available turkey genomic trace sequences (about 6 million) for insertions of relatively young repeats (CR1-C2 and CR1-B2). Cases with elements absent in chicken orthologous genomic loci were selected in the genome browser (UCSC) 48 and we generated conserved PCR primers for eight loci.

Altogether, we extracted a total of 23 orthologous loci containing 25 phylogenetically informative insertions of CR1 and LTR elements. The presence/absence patterns of these elements support ten clades within Galloanseres and Galliformes (Figure 3, additional data file 2): (1) One element (CR1-E) was found in species of Galliformes and Anseriformes but not in the outgroup zebra finch (Taeniopygia guttata; Passeriformes), thus supporting the clade Galloanseres. (2) Two elements (CR1-Y2 and CR1-X2) were found at orthologous genomic positions in all galliforms except Megapodiidae (megapodes). Although the pattern of the three possible branching orders – with the prior hypothesis at the first position – is not significant ([200]; p = 0.111) according to Waddell et al. 32, this is the first unambiguous retroposon evidence to support the local tree topology supported by morphological 16 and mitochondrial sequence data 1015. (3) We found eight independent retropositions (2× CR1-F2, 3× CR1-D2, 2× CR1-X2, 1× CR1-Y4) shared by Numididae, Odontophoridae, and Phasianidae that were not present in Anseriformes, Megapodiidae and Cracidae. Assuming a prior hypothesis based on morphological 16, nuclear 9, and mitochondrial sequence data 1015, these markers, together with two insertions found by Kaiser et al. 12, provide statistically significant support ([10 0 0], p < 0.0001) for this branch of galliform evolution. (4) Odontophoridae and Phasianidae share one orthologous CR1-H2 insertion that is missing in Anseriformes, Megapodiidae, Cracidae, and Numididae. While morphological 16 and nuclear sequence data 9, as well as the combined data set of Crowe et al. 11 support a sister group relationship between these two taxa, mitochondrial analyses 1015 support a topology in which Numididae are the sister taxon of Phasianidae. Our results provide the first, unambiguous retroposon evidence supporting the morphological and nuclear data. (5) We found five independent retroposon insertions (2× CR1-B2, 1× CR1-C2, 1× CR1-C and one GGLTR4a) in all Phasianidae species examined that were absent in all other galliform families. The monophyly of Phasianidae was also significantly supported by four additional independent CR1 insertions discovered recently by Kaiser et al. 12. Thus the monophyly of Phasianidae (including Tetraoninae and Meleagridinae) is significantly supported by nine independent retroposition insertions ([9 0 0]; p < 0,0001). (6) One CR1-B2 and one CR1-C element were found in all investigated phasianids except Rollulus rouloul. Thus, together with the one additional insertion detected by Kaiser et al. 12, there is now significant support ([3 0 0]; p = 0.037) for a hitherto unexpected clade. (7) All investigated phasianids except Rollulus and Gallus were found to share a CR1-C retroposition. An additional independent CR1 insertion supporting the same topology ([2 0 0]; p = 0.11) was found by Kaiser et al. 12. (8) We found two insertions of CR1-B2 elements in all investigated phasianids except the Rollulus, Gallus, and Pavo species. These findings lend additional support to the seven independent insertions found by Kaiser et al. 12 in similar species. The pattern of [9 0 0] is highly significant (p < 0.0001). (9) Two LTR (GGERVL18LTR) insertions were present in orthologous positions in Gallus gallus and Gallus lafayetii that were absent in all other analyzed birds. (10) One CR1-D2 insertion was found in Crax alector and Crax fasciolata, which was absent in all other investigated galliforms.

Within the loci containing the retroposed elements, we also found support for additional clades by the presence of 95 random intronic indels (see Figure 3). One indel was found only in the Meleagridinae and Tetraoninae species, a grouping that was also recently indicated by one CR1-insertion 12. For example, twelve indels were exclusive to Tetrao and Tympanuchus, grouping these two species of the subfamily Tetraoninae together, three independent indels were specific for intronic sequences of Perdix and Chrysolophus, seven were unique to the two Chrysolophus species, seven were unique to a clade comprising Pavo and Afropavo, in agreement with Kimball et al. 49 and four were unique to Pavo muticus and Pavo cristatus. Twenty-one independent indels group together the two odontophorid genera Callipepla and Colinus, and eighteen unite Crax alector and Crax fasciolata. Together with all the retropositions presented in points 4, 5, 6, 7 and 8 above (Fig. 3), these data clearly support a sister group relationship between Afropavo and Pavo, rejecting an earlier morphology-based hypothesis of a clade comprising Afropavo and Numididae 50. Although low complexity RGCs have a higher probability of being homoplastic 23 than do the insertion patterns of retroposons, we did not find any indels contradicting the topology of the tree supported by retroposed elements.

We provide the first retroposon evidence that Odontophoridae are the sister taxon of Phasianidae, which is also supported by at least one morphological feature, the presence of a well-developed intermetacarpal process on the carpometacarpus (see however, Stegmann (1978) concerning the possibility of a secondary loss of this process in Numididae 51). Our study further provides clear evidence against a recently hypothesized sister group relationship between Perdix and Meleagridinae (Crowe et al. 2006; note that the morphological data set used in this analysis contains several incorrect character scorings 52).

To the best of our knowledge, it has not yet been pointed out that the sister group relationship between New World turkeys and Palaearctic grouse, which is also supported by analyses of sequence data 1114, indicates a New World origin of grouse. Because the clade (Meleagridinae + Tetraoninae) is nested within taxa with predominantly Asian distributions, the stem species of this clade probably reached the New World from Asia.

A striking observation from the TinT data is that the maximal frequencies of individual CR1 subtype fixation rates fall in close temporal proximity to one another and tend to be concentrated in distinct temporal waves, as is visible from the cumulative curve (Figure 2). Interestingly, two of these peaks of CR1 fixation rates coincide with the two most highly supported branches, indicating long internal branches and/or high retroposition fixation rates. At least five parameters might affect the fixation rate of retroposons in a population in a given branch of the phylogenetic tree: (I) the promoter activity of the master gene, (II) the overall availability of enzymatic retroposition machinery and (III) its affinity to the master RNA, (IV) the population size, and (V) the branch length. In small populations the fixation of a single retroposition event is much more likely than in large populations 53. As the cumulative curve (Figure 2) reflects the timeframes of maximum fixation rates of several independent CR1 subtypes, it is unlikely that individual peaks are the result of promoter activity. The peaks might reflect times of over expression of the total retroposition machinery or might be due to severe population bottlenecks in the ancestral chicken lineage.

The GenBank accession numbers for the sequences discussed in this paper are [EU054465–EU054818].

In summary, we present a method to calculate the relative times of maximum fixation frequencies for retroposons. We applied the TinT method to obtain a temporal order of the activity of different CR1 subtypes. The results of the TinT method enabled us to preselect potential phylogenetically informative presence/absence loci to test hypotheses for specific internal branches of a phylogenetic tree. With this preselection strategy, we found the first retroposition evidence supporting successive sister taxon relationships between Megapodiidae, Cracidae, Numididae, and the remainder of galliform birds. Highly significant support is presented for the first time for the monophyly of the phasianoid clade comprising Numididae, Odontophoridae and Phasianidae (Figure 3, point 3). One marker suggests that Numididae are the sister taxon of a clade comprising Odontophoridae and Phasianidae. Five independent retroposon insertions presented in this study, along with four previous ones, offer overwhelming support for the monophyly of all investigated phasianid species (Figure 3, point 5). We present the first significant support for a sister group relationship between Rollulus and all other investigated phasianids (Figure 3, point 6), and additional retroposon evidence for the monophyly of a clade containing Pavo, Afropavo, Coturnix, Chrysolophus, Perdix, Tragopan, Tympanuchus, Tetrao and Meleagris (Figure 3, point 7) and for a clade comprising Chrysolophus, Perdix, Tragopan, Tympanuchus, Tetrao, and Meleagris (Figure 3, point 8). Complementary information from random indels indicate the existence of a clade including Tympanuchus, Tetrao, and Meleagris another clade comprising Tympanuchus and Tetrao, one comprising Chrysolophus and Perdix to the exclusion of Tragopan, Tympanuchus, Tetrao, Meleagris, and the other investigated galliform taxa, and finally a clade comprising Pavo and Afropavo to the exclusion of all other investigated taxa.

The mathematical TinT model, applied to birds, is currently in the process of being tested in several mammalian groups. We believe that it will prove to be a significant tool for all genomic projects characterizing the activity periods of retroposed elements.

JOK, GC, and JS conceived and designed the experiments. JOK, AM, and GC performed the experiments, collected data, or did experiments for the study. JOK, AM, GC, AK, and JS analyzed the data. JB and JS contributed reagents/materials/analysis tools. JOK, GM, and JS wrote the paper.

All authors have read and approved the final paper.