Introduction

The Harpy Eagle, Harpia harpyja (Aves: Accipitridae; Linnaeus, 1758), a critical apex predator, is one of the largest extant eagles on Earth. It occurs widely in the Neotropics from Central to South America, with half of its distribution in the Brazilian Amazon1. The Harpy Eagle occurs in low population density throughout its range and is characterized by a sizable territory and a patchy distribution2. Despite showing reversed sexual dimorphism in body size, they have a positive female-biased sex ratio at birth3. As an apex predator, it feeds mainly on medium-sized and tree-dwelling mammals, especially sloths and small monkeys4,5. The Harpy Eagle is one of two raptor species with documented cases of persecution (i.e., direct killing), which, in conjunction with habitat decrease, disturbance, and reduction of food sources6, increases its extinction risk.

The IUCN classification of the Harpy Eagle as vulnerable is reflected in the low genetic diversity documented for the species1, with signs of recent bottlenecks reducing heterozygosity, even though the species retains high genetic diversity in mitochondrial markers throughout its distribution7. Individuals from the Brazilian southern Amazon and Atlantic Forest populations show high gene flow with low population structure, while the northern Amazon population shows greater differentiation1. However, estimates of population dynamics through time and effective population size (Ne), two key features for assessment of genetic health of any species and development of high-impact conservation strategies, still need to be improved for the Harpy Eagle7.

Previous studies showed that the Harpy Eagle has many large-scale rearrangements in its genome architecture when compared to the more distantly related Galloanserae (i.e., chicken, geese, ducks, quails, and pheasants)8,9. Transposable elements (TEs) have been associated with large-scale structural variants, including chromosomal rearrangements, transcriptional regulation, and methylation10. Within Aves, the chicken and zebra finch genomes are the best annotated and have the most in-depth descriptions of TEs11, even though robust descriptions of the TE landscape for a diversity of non-model bird species have been recently done12,13,14,15,16. However, as studies of TEs are heavily impacted by genome assembly contiguity and quality, TE landscapes are still restricted to a few model species for which chromosome-level genomes produced using long-read data are available11. In general, genome-wide patterns indicate a reduced activity of TEs across birds, with a strikingly low presence of TEs (lower than 15%, except Piciformes; see17) possibly limiting genome size expansion and rearrangements10,11,12. Consequently, birds have a remarkably stable genome size and large-scale synteny in macrochromosomes12,18. Despite the overall trend of genome architecture conservation, genome assemblies of diverse Aves groups are adding evidence of extensive variation in some lineages16. Studies of comparative genomics benefit from high-quality genomes to elucidate structural variants and large- and micro-scale synteny.

Here, we report on a high-quality, chromosome-level genome assembly of the Harpy Eagle and bring evidence to a long-standing enigma in the apparent genome architecture (bauplan) reshuffling within Accipitridae. We provide structural coding gene annotation, identify transposable elements, and compare the genome architecture of the Harpy Eagle to species with chromosome-level genome assemblies available for other Accipitridae, the Cathartidae (New World vultures and condors), and the basal Galloanserae. Our findings support a major reshuffling event at the base of the Accipitridae-Pandionidae split or earlier. We also used the genome reported herein to reconstruct the Harpy Eagle effective population size through time, contextualizing the current genetic diversity documented for the species. Given this apex predator's ecological importance, cultural relevance, and conservation needs, the reference genome for the Harpy Eagle is a valuable asset for future studies on this species and will heavily contribute to population and conservation studies.

Results

Genome sequencing and assembly

We generated the near-complete chromosome-level genome assembly for the Harpy Eagle (named bHarHar1.0) (Fig. 1; Table 1). We obtained 37.68 Gb of reads and a coverage of 31.41X for PacBio reads, 707.27X for Bionano optical mapping, and 157.04X for Hi-C. Coverages of HiFi-CCS reads and Hi-C libraries were considered for all reads prior to filtering and trimming. We estimated the genome size at 1.23 Gb and heterozygosity at 0.306% through the k-mer analysis from raw read data using Genomescope (Fig. 1a). The repetitive content was initially estimated from sequencing reads at 9.4% of the total genome length (Fig. 1a). The final curated assembly was 9% larger at 1.35 Gb (1,351,447,071 bases), distributed in 322 scaffolds (664 contigs), with N50contigs = 16.8 Mb and N50scaffolds = 58.1 Mb. Over 96% of the assembled sequence was assigned to 30 pseudo-chromosomes (28 autosomes plus Z and W sex chromosomes). The assembled chromosome set is consistent with a previously reported karyotype of 2n = 588,19. Macrochromosome length varied from 18.49 Mb to 109.61 Mb, the 5 microchromosomes varied from 0.13 Mb to 1.1 Mb, while the sex chromosomes had lengths of 117.05 Mb (Z chromosome) and 37.4 Mb (W chromosome) (Fig. 1b). We observed a negative correlation between chromosome length and interchromosomal chromatin contacts, with more contacts occurring between the 15 smallest macrochromosomes, both Z and W sex chromosomes, and all five microchromosomes (Fig. 1c). The read k-mer completeness was estimated at 97.86%, and the consensus quality value was equal to 60.9562 (0.8 errors/Mb). BUSCO estimates showed a complete genome retrieving 99.7% of single copy genes using the Aves reference database (aves_odb10) and only 0.4% of duplicated sequences remaining (Fig. 1d). The complete circularized mitogenome was recovered from the final assembly and had all 37 expected coding genes distributed in 17.74 Kb. Hereafter, we renamed the scaffold 439, identified as the mitochondrial genome, as chrM (NW_026293368.1 in NCBI’s deposited assembly). The nuclear genome is deposited under assembly accession number GCF_026419915.1 in NCBI.

Figure 1
figure 1

Genome assembly statistics. (a) GenomeScope k-mer profile; (b) Length distribution for each of the assembled pseudo-chromosomes; (c) Hi-C interaction heatmaps for the curated bHarHar1.0 assembly. Chromosomes are displayed in size order from top to bottom and from left to right. Arrows indicate the chromosome W; (d) Snail plot for the genome assembly of Harpia harpyja, bHarHar1.0 with summary statistics. The plot shows the N50 metrics, base pair composition, and BUSCO gene completeness, generated with Blobtoolkit v2.6.4 (https://github.com/blobtoolkit/blobtoolkit)65.

Full size image
Table 1 Summary statistics for the Harpy Eagle (Harpia harpyja) genome assembly (bHarHar1.0).
Full size table

Transposable elements annotation

We built a custom intact TE library for H. harpyja, the first for the Accipitriformes order to date. The library of intact elements included 327 LTRs, 1,703 LINEs/SINEs, and 4843 DNA elements and miniature inverted-repeat transposable elements (MITEs) (7228 intact transposable elements in total) (Fig. 2A). The main class of intact elements found was the CACTA DNA transposon, with 2852 elements, followed by the CR1 LINE retrotransposons (1397 elements) and the Mutator DNA transposon class (1102 elements). These intact elements represent the TEs potentially still active in the H. harpyja genome, directly impacting the evolution of the genomic architecture of the species.

Figure 2
figure 2

Overall annotation of repeat elements considering (a) intact elements and (b) annotated insertions found on the Harpy Eagle genome; (c) repeat landscape transposable elements in the Harpy Eagle genome in terms of K2P divergence; (d) relative repeat landscape in terms of divergence time. TE annotation was done with EDTA v2.0.1 and summary output was plotted with LibreOffice’s spreadsheet editor v24.2.5 (https://libreoffice.com.br/).

Full size image

This library was subsequently used to identify ancient genome-wide transpositions (Fig. 2a). We found that 17.26% of the total genome was masked as repetitive sequences. Transposable elements represent the most significant portion of non-redundant masked sequences (16.33%), with 480,259 independent TE insertion events. Among the annotated elements, the Unknown class represents 36.5% of the masked bases (Fig. 2a). A single element (TE_00000121, Supplementary Table S1) accounted for 25.3% of all masked bases. We searched this TE sequence for TE-related motifs in DFAM and protein motifs in Uniprot and only found disordered domains. The second most prevalent TE was the CR1 retroelement, with 16.9% of insertions. The third most enriched class (10.8%) is the CACTA transposon class (Fig. 2b).

The Kimura two-parameter (K2P) distance between the intact elements and each identified insertion showed seven activity peaks of transposable elements in the Harpy Eagle genome (Fig. 2c). The oldest peak is composed mainly of the CR1 retrotransposon, predominating in the most ancient transposition events within 26% K2P or older (Fig. 2c) at over 98 million years ago (MYA; Fig. 2d). The CR1 insertions represented more than 50% of all TE insertions until ~ 63 MYA when the insertion events of DNA transposons (CACTA elements) and LTR Gypsy started to take place. CACTA elements comprise most TE insertions ~ 54 MYA (43%). Noteworthy, we detected the spread of an unknown element (TE_00000018) also at ~ 54 MYA, reaching a peak of 47.9% of all insertions at 48 MYA, simultaneous to the increase of Mutator DNA transposon insertion events (10.9% of all insertions at the same time). This was followed by a period (41–33 MYA) of relative stability for all TE classes, except for solo-LTRs (peak ~ 33–37 MYA with mean of 24.67% of all insertions) and the Helitron rolling-circle elements. After this period, we could observe a small increase in CR1 elements, peaking at ~ 31 MYA, but that was followed by silencing, reducing its relative content to ~ 1.8% of total insertions ~ 20 MYA. Solo-LTRs peak again at ~ 22–24 MYA. Concomitant to this, started the most pronounced spread, in relative and absolute terms, of the unknown element described above (TE_00000121, Supplementary Table S1), hereafter named Harpia harpyja 1 TE (HarHar1), which covered 87.1% of all insertions by ~ 17 MYA. This TE had no hits to the DFAM curated TE hidden Markov models (Ficedula albicollis and Uraeginthus cyanocephalus data) and no hits to UniProtKb and SwissProt protein databases. After translating the HarHar1 sequence in the six frames, we also searched for protein domains in the Harpia harpyja 1 TE, but none were detected. The second most recent event of elevated transposition activity showed the proliferation of the Retrovirus-like and Gypsy LTRs ~ 4–11 MYA, summing 73.9–58.3% of all insertions. Curiously, the prevailing elements in the most recent insertion events (nearly intact elements, transposition events in the last 2 million years) are the DNA transposons CACTA (42.4%) and Mutator (15.3%) and the Helitron rolling-circle DNA TE (12.6%). Retrotransposons and the group of unknown elements compose only a minor fraction of nearly intact TEs.

Coding gene annotation

The NCBI eukaryotic annotation pipeline annotated 16,803 unique protein-coding loci (which span 44,808 mRNAs) and 174 pseudo-genes; and 4251 lncRNAs, 41 snRNAs, 268 snoRNAs, and 367 tRNAs among non-coding RNAs. Most of the annotated coding loci (16,363) were also identified as known protein coding loci deposited in public sequence databases (only 440 were uncharacterized proteins), of which 99.8% (16,341 loci) were located on the thus far named chromosome scaffolds. When considering the distribution of coding loci across the genome, microchromosomes had higher gene density than macrochromosomes (Fig. 3a) despite the linear scaling of gene number with chromosome length (Fig. 3b).

Figure 3
figure 3

Gene density per chromosome. (a) Genes per kilobase for each assembled chromosome; (b) Number of coding loci per chromosome length.

Full size image

We compared the annotated Harpy Eagle coding sequences with those of the Eurasian Goshawk (Accipiter gentilis), the Golden Eagle (Aquila chrysaetos), the California Condor (Gymnogyps californianus), and a new de novo assembly of chicken (Gallus gallus). We assigned orthology based on reciprocal best hits in all-vs-all pairwise comparisons. We found 15,636 orthogroups (OGs), of which 12,838 had genes from all species considered (12,646 OGs had exactly 1:1 relationships and are thus the most suitable gene set for phylogenetic inference). The Harpy Eagle coding genes were distributed in 14,987 OGs, 16 species-specific (total 121 coding genes). Moreover, 189 Harpy Eagle genome coding genes were not assigned to any orthogroup. Thus, we have identified 310 coding genes specific to this lineage.

Comparative genomics

We found strong synteny between the genomes of the Harpy Eagle and the other two species of Accipitridae (the more closely related Eurasian Goshawk, A. gentilis, and the more distantly related Golden Eagle, A. chrysaetos; Fig. 4; Supplementary Figure S1), preserving the same collinear gene blocks. Despite the overall trend of macrosynteny conservation within Accipitridae, assuming the G. gallus genome architecture as the ancestral bauplan, we identified 60 major rearrangements (Fig. 5; Supplementary Table S2) comparing the five species considered here (all rearrangements were manually checked with cross-species HiC mapping, using bHarHar1 as reference; Supplementary Figure S2). Of these, 15 small intrachromosomal translocations and one fission (all occurring on the ancestral linkage group equivalent to the chicken chromosome 4) are evident when comparing the G. gallus and G. californianus genomes. These rearrangements persisted in the same configuration in the other Accipitriformes analyzed (see Supplementary Table S2 for a detailed description of all rearrangements identified) and might be characteristic of the Accipitriformes genome bauplan.

Figure 4
figure 4

Patterns of chromosome evolution within Accipitriformes and Galloanserae. Pairwise macrosynteny based on 20 gene collinear blocks as reciprocal best hits orthologs. Numbered bars indicate putative chromossomes for each analyzed species. Links indicate the boundaries of syntenic gene blocks identified by MCScanX. Color codes are based on the Gallus gallus chromosome organization. Data was plotted using SynVisio web application (https://synvisio.github.io/)83 and edited manually for better readability. Arrows denote inverted sequence orientation in alignment. Chromosomes not shown in this figure did not show any syntenic gene block.

Full size image
Figure 5
figure 5

Chromossomal rearrangement counts within Accipitriformes in comparison with the Gallus gallus ancestral bauplan. Counts were computed based on the synteny analysis shown in Fig. 4. Any rearrangement was considered exclusive when found only in a single species. Abbreviations: GG Gallus gallus, GC Gymnogyps californianus, AC Aquila chrysaetos, AG Accipiter gentilis, and HH Harpia harpyja.

Full size image

Within Accipitriformes, the synteny alignment of the three Accipitridae species and the California Condor (Cathartidae) revealed 23 chromosome fissions occurring in the largest autosomes (equivalent to gg1 to gg5), with 19 tandem fusions of microchromosomes with macrochromosome fragments, and 3 intrachromosomal translocations (Fig. 5). Thus, most chromosomal rearrangements detected (n = 45) were found exclusively within the Accipitridae family. H. harpyja shows four exclusive end-to-end (tandem) chromosome translocations, originating chromosomes 1, 4, 6, and Z, by fusing gene blocks that are apart in the other four species considered here (e. g., the orthologous sequence to the autosomes ag23, ac20, gc13, and gg12, from A. gentilis, A. chrysaetos, G. californianus, and G. gallus, respectively, are fused to the sex chromosome Z in the Harpy Eagle). In turn, the A. gentilis genome shows seven exclusive chromosome breaks relative to the other Accipitridae (Fig. 4; Supplementary Table S2). Aquila chrysaetos has the most conserved genome architecture amongst the Accipitridae sampled, when compared to the ancestral genome bauplan of G. gallus and G. californianus, with only one exclusive tandem translocation, originating the chromosome ac5 (Fig. 4).

Past effective population size dynamics

We reconstructed the past demographic history of the Harpy Eagle (Fig. 6). The PSMC analysis revealed an overall declining trend in the effective population size over the past 1 million years, with a steeper reduction in the last 20 thousand years, encompassing the Last Glacial Maximum (LGM, 18–23 thousand years ago—KYA), and continuing until the onset of the Holocene. This effective population decline went from ~ 8000 individuals (~ 0.5–1 million years ago—MYA) to about ~ 4,000 individuals (~ 200 KYA). This was followed by periods of increase and decrease in Ne around a mean (from 200 to 20 KYA) of 4000 individuals again until ~ 20 KYA. Nevertheless, the more recent Ne intervals in the PSMC for the Harpy Eagle are especially concerning since they reach Ne < 1000 and represent the lowest effective population size throughout its evolutionary history (Fig. 5). We observed only minor Ne variations over warming (~ 147–122 KYA) and cooling (~ 123–65 KYA) periods of climatic change20. Notwithstanding, starting ~ 20 KYA, the continuous downturn in genetic diversity of the Harpy Eagle spanned the Last Glacial Maximum and the Holocene, two sequential periods characterized by opposing cooling and warming phases, respectively, and which did not impact the continuous and steep Ne reduction observed during this period.

Figure 6
figure 6

Past effective population size (Ne) estimates of the Harpy Eagle. The darker line depicts the Ne through time, while lighter colors represent bootstrap estimates over 100 replicates. Gray arrows denote the inflection points of effective population size. Shaded areas denote the LGM, and the warming and the cooling period discussed by Germain et al20.

Full size image

Discussion

Here, we have assembled a reference genome for the Harpy Eagle and characterized its genome in the context of other members of the same order (Accipitriformes) for which chromosomal level genomes are available and a more distantly related basal avian lineage (Galloanserae). We inferred the TE landscape from a custom repeat library for the Harpy Eagle genome, the first such library for the order Accipitriformes, and described overall genomic features. We also reconstructed the past demography of the species to infer changes in its effective population size. Below, we discuss how the genomic features combined with its evolutionary history might have contributed to the evolution of the Harpy Eagle and other Accipitridae.

Harpy eagles have a similar genome organization pattern observed in other birds and non-avian reptiles, with macro- and microchromosomes, a greater concentration of genes on the latter, and higher microchromosome recombination rates21. Indeed, microchromosomes have been considered genome building blocks due to their frequent rearrangements across species21. Previous studies have shown a high degree of genome synteny between Galloanserae (galliform and anseriform birds) and Neoaves (most other birds, including passerines)17,22. Birds are known for having highly syntenic genomes across the entire group23 and even share large syntenic blocks with turtles24. However, our results show an interesting pattern of genomic rearrangement within Accipitriformes. Following the elementary algebraic operations in chromosome rearrangements proposed by Simakov et al.25, we show that only reversible macrosyntenic changes (tandem or end-to-end translocations and fissions) occurred within the Accipitridae and between the more distantly related Cathartidae and Galloanserae. These chromosomal rearrangements occur when whole chromosomes are fused end-to-end or break on intergenic sequence and thus preserve the overall gene order of the syntenic blocks comprising each chromosome. In contrast, most rearrangements between Accipitridae versus Cathartidae plus Galloanserae are macrochromosome fissions combined with microchromosome fusions, followed in some cases by small synteny-breaking rearrangements within gene space, which are irreversible changes in genomic architecture.

While birds are known for having an overall stable chromosome order18, we observed that several reshuffling events happened throughout the evolutionary history of Accipitriformes, followed by restabilization of genome architecture in its most diverse and widespread lineage, the Accipitridae. Previous studies based on chromosome painting had already identified major syntenic disruptions within Accipitriformes, demonstrating a similar genomic bauplan amongst Accipitridae and Pandionidae (ospreys), but one that differed from the basal-most Cathartidae, which, in turn, conserved the same overall syntenic association present in the basal Galloanserae within Aves9,26,27. However, in the only study to analyze these chromosomal reshuffling events in Accipitriformes from a phylogenetic perspective, species of Falconidae (falcons) were included in the analyses and were assumed as closely related to Accipitridae (eagles, hawks, harriers, kites, and Old-World vultures) and Pandionidae (ospreys) to the exclusion of Cathartidae (New-World vultures9). The conclusions drawn from this study, however, need to be re-evaluated given that whole-genome data has confirmed the paraphyly of diurnal birds of prey, with Cathartidae, Sagittariidae (secretary birds), Pandionidae, and Accipitridae grouping together in one clade (Accipitriformes) which, in fact, is sister to nocturnal birds of prey (owls, Strigiformes) and therefore totally separated from the Falconidae26. Unfortunately, no chromosome painting data or chromosomal level genomes are available for Sagittariidae (which includes a single species, the Secretarybird Sagittarius serpentarius). Still, these two combined datasets support the evolution of a more conserved ancestral genome architecture until at least the base of the Accipitriformes, given the basal divergence of Cathartidae from all other Accipitriformes26, contrasting strongly to a derived “reshuffled” bauplan shared at least by Pandionidae and Accipitridae9,26,27. Chromosome-level genomes or chromosome painting data for Sagittariidae should elucidate whether this unique lineage shares a genomic bauplan that is either more similar to that of Cathartidae and Galloanserae or to those found in Pandionidae and Accipitridae9,26,27.

Another organizational characteristic is that repeat content in birds is generally low when compared with other animals (lower than 15%17), except for Piciformes (woodpeckers and toucans), which have > 20% repeat content12 and sparrow songbirds (Passerellidae) with > 30% repeat content16. For the Harpy Eagle, we found more than twice the repeat content (16.33%) reported for most songbirds (7.8%) and parrots (9.8%, 28), a closer figure to Piciformes. Over 36% of all TE insertions in the Harpy Eagle genome are from unknown elements. Two unknown elements comprise over 88% of all insertions identified (Supplementary Table S1). The most prevalent of these two unknown elements is solely responsible for the peak of activity at ~ 17 MYA (mean K2P = 8.7%). CR1 nonLTR retrotransposons are the second most prevalent superfamily but are concentrated in the ancient transposition events, fading out to become virtually absent in the last 3 MYA in the Harpy Eagle. As shown by Kapusta and Suh11, LTR retrotransposons are especially active in Accipitriformes, but we found more Gypsies (17.8 Mb masked) than ERVs (12.3 Mb) in the Harpy Eagle genome.

We found evidence of significant changes in the TE content of the Harpy Eagle genome over the past 98 MYA. Major shifts in TE content are coincident or precede phylogenetic diversification events in the Accipitridae family. The most prevalent element in avian species, the LINE CR1 elements29,30,31 are replaced by a concomitant expansion of Gypsy LTR TEs and CACTA DNA transposons after the Cretaceous–Paleogene extinction event ~ 65 MYA, and the split between Cathartidae and Accipitridae families. Bursts of solo-LTRs precede major expansions within the Accipitridae family. The occurrence of solo LTRs results from ectopic recombination events between very similar LTRs, usually associated with recent expansions of these elements. The split of the H. harpyja from the other Harpiinae at ca. 16.4 MYA32 is also coincident with an unknown TE burst ~ 13 to 22 million years ago (Fig. 2C and D). Taken together, these TE expansions may have played a role in diversification events within this group. Bursts of TE activity have been associated with speciation events33,34, as chromosome rearrangements can lead to postzygotic (reproductive) isolation. Widespread transposition events, often provoked due to recent invasion through horizontal transfer or reactivation of previously silenced ancient elements, can alter drastically the genome bauplan in a population, especially those with low effective population size, since natural selection is relatively less effective than genetic drift in small populations. Studies using high-quality genomes and repeat landscapes have supported the idea that increased TE activity might correlate with speciation timing across species33. In mammals, ERVs have been associated as mediators of genomic plasticity by facilitating recombination and DNA rearrangements35. In birds, CR1 LINEs were associated with speciation events in songbirds33. Here, we show that major changes in the relative TE content of different classes, namely Gypsy retrotransposons and solo-LTRs, DNA transposons CACTA and an unknown element occurred concomitantly to the reshuffling in the ancestral genomic bauplan to all Accipitridae. Following the concept of “ecology of the genome”36, similar to the deer mouse37, Harpy eagles have a recent expansion of an unknown TE (HarHar1) that caused a shift in its TE landscape and drove LINEs (common in other avian species) to lower relative abundances in the genome. The causes and consequences of this shift in abundance and how competition for insertion sites between LINEs, LTRs, and unknown TEs37 shaped the genome architecture of Accipitridae, remains to be elucidated following the availability of more high-quality genomes for this and related families.

The past demographic history of the Harpy Eagle provides evidence of a very concerning scenario. Between ~ 300 and 20 KYA, Ne estimates fluctuated relatively little in the Harpy Eagle. However, this trend was followed by a steep decline starting at approximately ~ 20 KYA to less than 10% of the Ne initially calculated. Despite the lack of reliability in Ne estimates younger than ~ 10 KYA, characteristic of PSMC analysis, the contemporary population size is estimated to be declining fast with rapid habitat loss (as discussed in Kaizer et al.5). The overall declining trend since ~ 20 KYA is thus supported by independent evidence. Given the species’ deep cultural significance to Amerindian societies of the Amazon region38 and hunting reports, the Ne decrease since the LGM may have multiple causes. Many species had associated decreasing population sizes due to past climate change compounded by anthropogenic impact, possibly from the first waves of human migration to South America39,40,41. Recent evidence for earlier presence of human populations in the American continent pushes back the occupation to even before the LGM, ~ 30 KYA42,43,44,45 (as an open debate, see also Surovell et al.46). Thus, considering recent analysis on the impact of human activity in the extinction of megafauna, mammals, and birds47,48,49 (but see also Stewart et al.50,51), we can assume the decline in population size of the Harpy Eagle overlapped the human occupation of South America.

The recent fires in the Brazilian Amazon, Pantanal, and Cerrado biomes, compounded with extensive deforestation and climate change, pose dire threats to the survival of one of the world’s largest extant eagles. The drastic reduction in population size under stressful environmental conditions diminishes the power of selection to purge and control transposable elements, old and new, and expose the Harpy genome to a reshuffling of syntenic gene blocks, higher rates of gene regulation-disrupting transposition events, and ectopic recombination among terminal repeats from TEs52. This, in combination with other factors such as low genetic diversity and higher rates of inbreeding, may contribute to a lower genetic health of the species. In summary, the high-quality genome presented here offers a robust foundation for informed conservation strategies of the Harpy Eagle. By providing a detailed genetic blueprint, this genome enables the development of targeted genetic management strategies for both in situ and ex situ conservation programs, ensuring the maintenance of genetic diversity of wild and captive populations. The development of molecular markers derived from the reference genome can facilitate the monitoring of wild and captive populations, enabling researchers to track genetic diversity, gene flow, and potential signs of inbreeding depression in real-time. Additionally, using the reference genome presented herein to support the re-sequencing of natural populations will allow for regional inferences of past demographic trends that might identify historical strongholds for the species. Finally, by mapping the TE landscape in the Harpy Eagle genome, it will be possible to monitor its influence on deleterious mutations in different natural and captive populations, guiding conservation programs. This comprehensive approach, integrating genomic insights with field monitoring, is essential for the long-term conservation of the Harpy Eagle and other threatened species.

Materials and methods

Sampling and sequencing

A blood sample was taken from a rescued female specimen shot illegally by hunters in a forest clearing near Parintins, Amazonas state, Brazil (2.625833 S, 56.649722 W); this individual was selected for whole genome sequencing to obtain both sex chromosomes. The specimen was taken to an animal care facility at “Centro de Triagem de Animais Silvestres” (Cetas) of the “Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis” (Ibama) in Manaus, where blood was sampled, stored in 96% ethanol, flash-frozen on dry ice at the Laboratório de Evolução e Genética Animal (LEGAL) of the Universidade Federal do Amazonas (UFAM) in Manaus, and kept frozen until sample processing. All experimental protocols were approved by the Brazilian National Council for Control of Animal Experimentation (CONCEA) and the Ethics Committee on Animal Use of the Federal University of Pará (protocol number #7335280722). All methods were carried out in accordance with relevant guidelines and regulations and are reported following the ARRIVE guidelines (https://arriveguidelines.org). The Harpy Eagle blood sample was collected under permit #31457 from the Brazilian “Sistema de Autorização e Informação em Biodiversidade “ (SISBIO) issued to A. B. and accessed under the Brazilian “Conselho de Gestão do Patrimônio Genético” (permit CGen #A0341AE) and “Sistema Nacional de Gestão do Patrimônio Genético e do Conhecimento Tradicional Associado” (SISGEN #52231) issued to T. H. The Vertebrate Genomes Laboratory (VGL), a hub of the Vertebrate Genomes Project (VGP, https://vertebrategenomesproject.org/), performed all sequencing and assembly steps. High molecular weight (HMW) DNA was extracted using the Circulomics HMW DNA extraction standard TissueRuptor protocol with the Nanobind Tissue Big DNA Kit. After DNA extraction and sequencing library preparation, three different technologies were used for genome sequencing: Pacific Biosciences (PacBio) long-read HiFi-CCS using the Express Template Prep Kit 2.0, contact reads generated by Arima Genomics (Hi-C reads) using the v2.0 kit and sequenced on an Illumina NovaSeq 6000, and Bionano Genomics optical mapping using ultra-long DNA of high molecular weight (> 300 kb) and the Bionano Prep Direct Label and Stain (DLS) Protocol, which was run on a Saphyr flow cell.

Genome assembly

The assembly was done with the VGP pipeline v2.0 in HiFi-only mode. Overall HiFi read quality check was performed with FastQC53 and Nanoplot feature from the NanoPack2 suite54. Reads were discarded when at least 35 bp of adapter sequences (-O 35) were found in any position using CutAdapt55, accepting a maximum error rate of 10% (-e 0.1). Contig assembly was performed with Hifiasm v. 0.15.1-r334 in primary mode56,57 with default settings, followed by removal of haplotype duplicates with purge_dups v. 1.2.558 using two rounds chaining (-2). Scaffolding was performed in two steps. For the first step, the optical map was fed into Bionano’s Solve software v3.6, setting the conflict filter level for genome maps (-B) and for contig sequences (-N) as two. The second step was done with Hi-C data59 using the SALSA2 v2.3 software60. Briefly, Hi-C reads were first mapped to the previous scaffolds using the BWA mem module61, setting the mismatch penalty to eight (-B 8). Chimeric reads alignments at the 3’ end were filtered out with Arima Mapping pipeline scripts (https://github.com/ArimaGenomics/mapping_pipeline), and alignments for each pair were then combined into a single file and converted to BED format. Next, we executed the SALSA2 scaffolding pipeline with the default parameter set for the Arima Hi-C Mapping pipeline (-e GATC, GANTC, CTNAG, TTAA -m yes). The manual curation of the Hi-C contact map was done with PretextView86. The final primary assembly was named bHarHar1.0 (NCBI accession number GCF_026419915.1). Assembly quality was evaluated using Merqury62, gfastats63 and compleasm v0.2.264 software fed with Aves_odb10 and visualized with BlobToolKit65. The complete mitochondrial genome assembly was generated from the whole genome sequence data, using the MitoHifi pipeline66. The mitogenome of Bustatur indicus, the Grey-faced Buzzard, was retrieved from NCBI and used as a reference for searching the mitochondrial scaffolds. Circularization and gene completeness evaluation were also done using MitoHifi.

Transposable element and repeat annotation

We first retrieved all coding sequences (CDS) for all Accipitridae (except for the Harpy Eagle) with available assembled genomes in NCBI by September, 2023. Next, we filtered out CDS with BLAST hits to transposable element proteins deposited in the NCBI nr database, following the 80-80-80 rule (sequences are annotated as TE if they are longer than 80 base pairs, and share at least 80% sequence identity over 80% of their length; see67,68. Then, we created a custom repeat library for the Harpy Eagle genome with the EDTA pipeline v2.0.169, filtering out false positives with the previously filtered CDS dataset (–cds). We also supplemented annotation with Aves-restricted long interspersed elements (LINEs) and short interspersed elements (SINEs) from DFAM (–curatedlib). The library was fed to RepeatMasker70. We grouped all CR1, L1, and L2 LINEs into a single group (referred to as CR1) and also R1 and R2 LINEs into a single group referred to as R1. The LTR TEs found were reclassified using TEsorter to retrieve Endogenous RetroVirus (ERVs) from misclassified elements. The RepeatMasker output was parsed to obtain a non-redundant estimation of genome coverage by the repeats and the repeat landscape for the Harpy Eagle genome (https://github.com/4ureliek/Parsing-RepeatMasker-Outputs). We assumed the mutation rate of 2.3 × 10−9 substitutions/site/year, estimated from Ficedula albicollis71, recently used for Aquila chrysaetos populational studies72, and other birds73,74,75,76. TE insertion time was estimated as the divergence (Kimura-2-parameter distance) divided by twice the mutation rate per year. We generated hard and soft-masked versions of the assembled genome using the annotated repeats and BEDTools77. Individual TE sequence searches were conducted manually, using online tools to scan DFAM, InterProScan, and UniProtKB (SwissProt).

Gene annotation

We submitted the bHarHar1.0 assembly to NCBI and requested annotation with the Eukaryotic Genome Annotation Pipeline v10.178,79 Since we did not sequence RNA-seq libraries for the Harpy Eagle, the annotation pipeline was run based on orthology to other birds already deposited in the NCBI Genome database. The RefSeq genome records can be found under accession number GCF_026419915.1-RS_2022_12.

Comparative genomics

Syntenic regions between the Harpy Eagle and other species were investigated using two complementary approaches. We included in our alignments three other Accipitriformes species with available chromosome-level high-quality genomes: the Eurasian Goshawk Accipiter gentilis (NCBI accession number: GCF_92944395.1), the Golden Eagle Aquila chrysaetos (GCF_900496995.4), the California Condor Gymnogyps californianus (GCF_018139145.2), and an outgroup from the Galliformes order represented by the chicken (Gallus gallus GCF_016699485.2). While Accipiter gentilis and Aquila chrysaetos belong to the same family clade as the Harpy Eagle (Accipitridae), Gymnogyps californianus belongs to the family Cathartidae, recovered as the sister group to all other Accipitriformes26. First, collinear blocks were identified by directly aligning the genomic sequences using the minimap2 aligner80 and represented with dot plots using the D-Genies interactive visualization software81. Second, macrosyntenic blocks of at least 20 genes were also identified using the MCScanX method82 and visualized with SynVisio83. Chromosomes in all genome assemblies used here were renamed following the NCBI numbering preceded by the first letter in the genus and the first in specific epithet (e.g., chromosome 1 in Harpia harpyja deposited genome assembly was renamed as hh1, chromosome 2 as hh2, and so on). The longest sequence for any gene locus was retrieved and used as the primary isoform for orthology inference, using OrthoFinder2 with default parameters84. All chromosome translocations were validated by cross-species mapping the HiC data used to scaffold each genome assembly used in this study to the Harpy Eagle genome assembly presented here (Data available in the Genome Ark Database—entries bAccGen1, bAquChr1 and bGalGal1; and DNAzoo (entry G. californianus). Mapping was performed for each read pair file with BWA-MEM285, filtering and merging of BAMs with Bellerophon (https://github.com/davebx/bellerophon) and image generation with Pretext86.

Population dynamics

To reconstruct population dynamics through time and infer the effects of past climatic events in effective population size (Ne) dynamics, we used the pairwise sequentially Markovian coalescent (PSMC) method87. The generation time for the Harpy Eagle was calculated from estimates of age to maturity and reproductive longevity, as “age to maturity + ½ reproductive longevity”88. We considered values from Watson et al.89 (age to maturity = 4.5 years and 5.5 years, reproductive longevity = 30.5 years and 23.5 years, for males and females respectively) with an estimated mean generation time of 18.5 years. Effective population size and times were scaled using the generation time and mutation rate estimated above, 4.26 × 10−8 substitutions/site/generation. Otherwise, we used the default settings of PSMC. Briefly, we mapped HiFi reads using bwa-mem61 to the assembled genome and produced a consensus sequence with SAMtools v0.1.1990 and bcftools91. Reads that mapped to the Z and W chromosomes and unanchored contigs were excluded, as the sex chromosomes have different mutation and evolutionary rates. Confidence intervals were estimated using 100 bootstrap replicates87.