Introduction

Genome-wide association studies (GWAS) have gained importance in the arsenal of today’s genetic analyses. Whole genome sequence assemblies provide the most important support for GWAS. In this paper, we present the genome assembly CerEla1.0 for the red deer, and Cervus elaphus. CerEla1.0 could serve manifold Genome-wide association studies (GWAS), including, for instance, organ regeneration, comparative analyses of orthologous mammalian genes, their promoters, regulatory networks, SNP sets, development of chromosome-specific microsatellites for individual identifications, and DNA markers for refined population and evolutionary studies. Red deer is an emblematic member of the natural megafauna of the Northern Hemisphere, and is displayed in Neolithic cave paintings. Humans introduced and spread the species in the Southern Hemisphere. Today, the red deer is both a wild animal and also a half-domesticated farm-bred livestock for venison and velvet antler products, and is getting recognized as a model animal for bone, osteoporosis and regeneration research as well as for population and evolutionary studies. Red deer, the Royal Game of the Middle Ages in Europe, is surrounded by respect and amazement in many cultures. For instance, in historical legends, a mythological Wonder Deer led the Hungarians to their present land. Antler and skeleton remains excavated from time to time document that red deer have lived in the Carpathian Basin for at least 7000 years. Due to the very high average quality and often record size of our trophies as well as for spiritual and emotional reasons, red deer is a “charismatic” member of the Hungarian megafauna and stands in the center of attention (Bán 1998). The estimated wild population size was around 100,000 in 2010 (Szabolcsi et al. 2014).

The closest relative of C. elaphus is the sika deer Cervus nippon. The two species can interbreed and produce fertile offspring in co-occurring natural populations (McDevitt et al. 2009; Biedrzycka et al. 2012). In the Far-East, the sika deer (C. nippon) has also been in the focus of interest for a long time, including farming, meat production, and alternative medicine. In China, antler extracts have been used for tonic since ancient times.

In modern biomedical research, the antler mainly serves as a model for studies on bone development or regeneration of a complete organ (Li and Suttie 2001, 2012; Li et al. 2002, 2007; Price and Allen 2004; Park et al. 2004; Price et al. 2005; Gu et al. 2008; Kierdorf et al. 2009). The advent of next-generation sequencing (NGS) technologies added further impetus to antler research with transcriptome analyses (Yao et al. 2012; Ba et al. 2016), which were extended toward more complex developmental studies of the whole body of the deer (Jia et al. 2016).

In the past decade, we introduced deer research in our laboratory on a molecular biological basis; we cloned and functionally identified a set of genes which are differentially expressed during the rapid but controlled proliferation and robust mineralization of the developing antler, or involved in skeletal osteoporosis coupled with the annual antler cycle of the deer (Molnár et al. 2007; Gyurján et al. 2007; Stéger et al. 2010). Physiological deer osteoporosis and pathological human osteoporosis were compared, and new genes of significant roles were identified (Borsy et al. 2009; Balla et al. 2008). In another line of our deer project, we developed a highly sensitive tetranucleotide STR multiplex PCR for identification of individuals and for population and forensic studies (“DeerPlex”, Szabolcsi et al. 2014). The mtDNA sequence of the Central European red deer was established by the NGS approach (Frank et al. 2016); the mitochondrial DNA lineages of the Carpathian Basin were described (Frank et al. 2017). Our ambition of whole genome sequencing and to assemble the first draft reference genome for C. elaphus stems from these efforts, as well as from the strong tradition in bioinformatics and functional genomics at the NARIC Agricultural Biotechnology Institute, Gödöllő, Hungary (Barta et al. 2016).

Compilation of DNA markers and the application of interspecific hybridization and back-crosses provided a way to develop the basis for the classical genetic analysis for deer, a so far non-model animal. The excellence and the power of the concept were demonstrated by the construction of the C. elaphus genetic map (Tate et al. 1995; Slate et al. 2002a).

This genetic map of 621 sites (2532 cm in length with 5.7 cm average spacings) integrated modern day technologies and methods, including among others: comparative genomics and orthologous DNA marker alleles drawn from ruminants and other mammals, making interspecific F1 hybrids between milu (Pere David’s deer, Elaphurus davidianus) and red deer C. elaphus, building an F2 mapping herd of a reasonable size by artificial inseminations from the cross of F1 stags × red deer hinds, computing recombination frequencies by LOD support, map making using the MapMaker V3.0 computer program and high resolution statistics. The deer genetic map served as the origin of further studies such as among others, the evolution of ruminant genomes (Slate et al. 2002a), QTL scan (Slate et al. 2002b), genome-wide SNP search (Fisher et al. 2015; Brauning et al. 2015), and whole genome annotation and assembling pseudochromosomes (this work).

In this study, using next-generation sequencing technologies, we assembled the pseudochromosomes and annotated the resulting reference genome CerEla1.0 of the Central-European red deer, Cervus elaphus hippelaphus. Two references, the (recombination) genetic map of red deer by Slate et al. (2002a), the genome sequence of Bos taurus (an evolutionally close species, NCBI Btau_5.0.1) and the rule of co-linearity between the genetic map and the chromosome were utilized, as was an independent work published in bioRxiv by Fisher et al. (2015) and Brauning et al. (2015). The concept may be useful for genome annotation in the case of other non-model animals. Shortly before the completion of the CerEla1.0 manuscript, Johnston et al. (2017) published the very high-density genetic map of C. elaphus (based on 38,000 SNPs) in bioRxiv. We believe that this genetic map will provide a rich source for further improvement of CerEla1.0.

The reference genome sequence CerEla1.0 and the pseudochromosome complement of Central-European red deer (C. elaphus hippelaphus) may provide a basis for broader interests including, among others, conservation genetics, refined evolution, and population studies within the family Cervidae and Pecora, for identification of descents for autosomal, maternal and paternal lineages, for forensic identification, for defining allelic compositions behind phenotypes important, for example, in game management, or in biomedical research and applications, like in bone research, osteoporosis, organ regeneration and tumour biology.

Materials and methods

Collection of samples

Blood (3 × 10 ml) was taken from a living animal of 7 years of age (Fig. 1), a capital stag with shed antlers weighing 12 kg at the age of 11 years, at the Deer Farm of the Game Management and Landscape Center of Kaposvár University (Bőszénfa, Hungary) (the definition of a capital individual is: a stag of exceptionally strong antlers, whose trophy would score at least 170 CIC points, i.e., that would be at least a bronze medalist.). In this case, the stag would have been golden medalist (at least 240 CIC points) by far, significantly above 210 CIC points, the minimum for gold medal. Sample collection was performed by a trained veterinarian according to standard veterinary medical practice with a permission from the Hungarian Veterinary Chamber [Hungarian Animal Rights Law (243/1998, XII.31)]. Samples were preserved in EDTA buffer and stored at − 20 °C.

Fig. 1
figure 1

Photograph of the C. elaphus stag (crot. No. 3016). This stag donated the DNA for the CerEla1.0 genome program. The stag was born in 2003. This snapshot was taken in 2010 at the Deer Farm of the Game Management and Landscape Center of Kaposvár University (Bőszénfa, Hungary). The blood samples for DNA preparation were taken in 2011. At the age of 11, his shed antlers weighted 12.0 kg (would be above 240 CIC points highly gold medalist, i.e., above 210 CIC points)

Full size image

DNA preparation and sequencing

Total genomic DNA was extracted from the blood samples using the Duplicα Prep Automated DNA/RNA Extraction System (EuroClone S.p.A., Italy). Isolated DNA samples were sent to Aros Applied Biotechnology (Aarhus, Denmark). As a custom service, one paired (~ 500 bp) and one mate pair (1.2 kb) library were prepared and sequenced in two and one illumina HiSeq 2000 lane (2 × 101 bp), respectively.

Sequence assembly and sequence analysis

The reads were assembled into contigs and scaffolds by the Broad Institute’s ALLPATHS-LG (Gnerre et al. 2011) program using mostly the default parameters. The program ran on an SGI UV 1000 machine (1152 intel CPU core and 6 TB memory) provided by the Hungarian National Infrastructure Development Program. All the further sequence analyses were done using standard bioinformatics/genomics programs EMBOSS (Rice et al. 2000), BWA (Li and Durbin 2009), lastz (Harris 2007), BLAST (Altschul et al. 1990; Korf et al. 2003), etc., and in-house custom-made Bash, Python, and Perl scripts.

Pseudochromosome assembly

Genetic marker sequences were identified in UCSC, UNIPROT, NCBI, and ENA databases and retrieved for further analysis. Microsatellite marker sequences were extended by the surrounding regions using the genome sequence of the species from where they originated. They were mapped to both the cattle reference genomes (Baylor Btau_4.6.1, Bos_taurus_UMD3.1.1, and Btau_5.0.1) and the scaffolds using blastn (with the following options: -1e-10-outfmt 6-best_hit_score_edge 0.05-best_hit_overhang 0.25-max_target_seqs 3). In the following iterative steps, we used the MUMmer3.0 program (Kurtz et al. 2004) with nucmer option and bwa-mem together with our custom Bash scripts to map selected scaffolds and contigs into the masked bovine genomes to determine the order and orientation of scaffolds and contigs in the red deer pseudochromosomes. The final generation of pseudochromosome sequences was carried out by a custom-made Python script.

Genome annotation

Pseudochromosome sequences were used in the MAKER version 2.31.8 (Cantarel et al. 2008; Cambell et al. 2014) genome annotation pipeline. MAKER identifies and masks repetitive elements in the genome with RepeatMasker-open-4.0.5 and RepeatRunner programs. Therefore, first, an organism-specific repeat library suitable for repeat masking was created. This was achieved by the RepeatModeler-1.0.4 software using RECON version 1.08 and Repeat Scout version 1.0.5. While RepeatMasker identifies known repeats, RepeatModeler (Smit et al. 2013–2015) predicts novel repeat sequences in the genome. In the next step, MAKER generated the initial (evidence-based) gene models based on expressed sequence tag (EST), mRNA, and protein evidences. As EST evidence, cDNA from cattle (Bos taurus) was used (ftp://ftp.ensembl.org/pub/release-88/fasta/bos_taurus/cdna/). For RNA-seq evidence, we downloaded the raw RNA-seq reads from a sika deer (Cervus nippon) experiment (Yao et al. 2012) and the transcriptome using the TrinityRNAseq pipeline (http://trinityrnaseq.sf.net) (Grabherr et al. 2011). We downloaded the cattle, the human (Homo sapiens) and the sheep (Ovis aries) complete proteomes from the ENSEMBL database (ftp://ftp.ensembl.org/pub/release-88/fasta/homo_sapiens/pep/; ftp://ftp.ensembl.org/pub/releas-88/fasta/ovis_aries/pep/), and used them as protein homology evidence. For repeat masking, artiodactyl was applied as model organism for RepBase masking in RepeatMasker, together with the previously generated organism-specific library and MAKER’s internal database containing transposable elements. BLAST was used for aligning EST, mRNA, and protein sequences to the genome and repeat identification. Soft-masking rather than hard-masking was applied in BLAST. MAKER used Exonerate software to refine BLAST alignments. Next, SNAP (Korf 2004) and AUGUSTUS (Stanke et al. 2006) gene prediction programs were trained and employed to create ab-inito (non evidence-based) gene predictions. The evidence-based gene models were used to train AUGUSTUS and SNAP (Cambell et al. 2014). After training SNAP, the predicted set of genes is used to retrain (bootstrap) SNAP. To obtain the final annotation set, the evidence-based and the ab-initio gene predictions were integrated, filtered, and optimized by MAKER. The MAKER’s derived gene model is in gff3 format, which can be easily loaded into genome browsers. Finally, to identify protein functions, the InterProScan (Jones et al. 2014) software was used to identify and to further annotate protein-coding genes. Different types of RNA-coding genes were annotated using MAKER. We identified ribosomal RNAs and microRNAs with the blastn program using the small and large subunit (SSU, LSU) sequences of mammalian ribosomal RNA of the SILVA123 (Quast et al. 2013) database and 21 mammalian sequences of the miRBase (Kozomara et al. 2014) database. We searched 5S units of ribosomal RNA with barrnap-0.6 (http://www.vicbioinformatics.com/software.barrnap.shtml). We found the transfer RNAs with tRNAscan–SE-1.3.1 (Schattner et al. 2005) software by eliminating the short interspersed and the pseudo tRNAs. In the command, we used -H -f -m -o options.

Genetic variant detection

The raw reads of both the paired-end and mate pair libraries were mapped to the reference genome using bwa-mem (Version: 0.7.10-r789). The unmapped read pairs were withdrawn for further analysis. We used SAMtools software (Li et al. 2009) (Version: 0.1.19–44428cd) to detect genetic variants, with “mpileup -D -S -E -uf” command line options. The vcf files were generated with BCFtools “view -bvcg” parameters and we applied the vcfutils.pl script’s “varFilter -D 188” option to filter those variants, where the coverage was maximum 188 (3× the average genome coverage). The variants whose Phred quality score was higher than 30 were filtered with custom Perl script. The annotation of the variant’s function was performed with the ANNOVAR software (Wang et al. 2010), where we created our own gene definition database using a MAKER-derived gff3 file. To annotate the genetic variants’ functions, we used the table_annovar script.

Data availability

The raw reads have been deposited into the SRA database (SRR4013902). The reference genome sequence has been submitted to the NCBI database and can be accessed at the following accession number (MKHE00000000). The gene annotation and the variation tracks are available for browsing and downloading from the JBrowse web page http://emboss.abc.hu/wonderdeer/JBrowse.

Results

Genome sequence and assembly: reads, contigs, and scaffolds

A stag, a 7-year-old Central European red deer (Cervus elaphus hippelaphus) with capital antlers (shed antlers of 12 kg) was the source of the DNA, prepared from blood samples. The stag was bred in the Bőszénfa Deer Farm of Kaposvár University, Kaposvár, Hungary (Fig. 1).

The red deer stag has 33 pairs of autosomes, plus the X and Y chromosomes. For DNA sequencing, a paired-end and a mate pair library were constructed in Aros Applied Biotechnology (Aarhus, Denmark) as a custom service, and sequenced by Illumina technology (Illumina HiSeq2000). The combined length of all reads (2.2 × 109) added up to 222.7 Gbp, which corresponded to about 74× coverage of the sequence libraries and 62× of the haploid genome of C. elaphus, assuming that the genome size was similar to that of related ruminants like cattle (Bos taurus) or sheep (Ovis aries), i.e., around 3 Gbp. Both the initial de novo assembly and the scaffolding were carried out with the ALLPATHS-LG program. According to the ALLPATHS-LG assembly, the reads added up to 437,412 contigs with a total contig length of 1.95 Gbp. The contigs assembled to 34,724 scaffolds with a combined length of 3.4 Gbp (i.e., “total scaffold length, with gaps”). Read statistics is shown in Table 1, ALLPATH-LG report on statistics in Table S1.

Table 1 Read statistics: number of reads and bases of paired-end and mate pair libraries
Full size table

Assembly, guided by the reference genetic map of C. elaphus, mapmarker scaffolds (MMSc-s)

A genetic map of C. elaphus with 5.7 cM average spacings, derived from data of the interspecific back-cross F2 population between C. elaphus × Elaphurus davidianus (David peer’s deer) by Slate et al. (2002a), was utilized as the first step for reference-guided assembly. In accordance with the haploid chromosome number, the genetic map split into 34 linkage groups (X and Y drawn together) with 2532 cM in combined lengths. From the 621 genetic markers mapped by Slate et al., we could map 365 (Table S2). These were defined by spacious DNA marker alleles (derived from ESTs, RFLVs, STSs, and protein sequences, which were adapted mainly from Artiodactyla, less from hominids and rodents). The rest were AFLPs, with no sequence data available. We unambiguously identified the DNA sequence of these 365 map markers in the deer scaffolds and called them “mapmarker” scaffolds (abbreviated as MMSc). By this step, the MMSc-s gained a defined position in the map array: in exchange, the site of the genetic map became part of an extended and specific DNA sequence of the deer genome (i.e., as MMSc, Fig. 2 provides a typical example, C. elaphus linkage group 1).

Fig. 2
figure 2

Co-linearity of genetic map markers and map marker scaffolds (MMSc) of deer Cervus elaphus pseudochromosome and linkage group 1 (in genome program CerEla1.0) and Bos taurus pseudochromosome 15 (in genome program Btau_5.0.1). Numbers on the deer genetic map segments correspond to cM distances taken from Slate et al. (2002a). Numbering along the pseudochromosomes correspond to distances in Mbp. Black dots correspond to centromerons. Similar relations were found for pseudochromosomes Ce2/Bt29, Ce4/Bt18, Ce7/Bt23, Ce9/Bt7, Ce10/Bt25, Ce11/Bt11, Ce12/Bt10, Ce13/Bt21, Ce14/Bt16, Ce18/Bt4, Ce20/Bt3, Ce21/Bt14, Ce23/Bt13, Ce24/Bt22, Ce25/Bt20, Ce27/Bt24, Ce30/Bt12, Ce32/Bt27, CeX/BtX, and CeY/BtY. Note: Slate et al. 2002a applied the Kosambi mapping function to calculate cM map distances (Kosambi 1943) instead of the standard Haldane function (Haldane 1919)

Full size image

Co-linearity between the genetic map of C. elaphus and the sequence of the B. taurus genome

The 365 C. elaphus “mapmarker” scaffolds were probed against the two versions of the DNA sequence of the B. taurus genome (NCBI Btau_4.6.1, Btau_5.0.1). We assumed that the bovine genome as template would provide high-fidelity alignments to the C. elaphus scaffolds. Our previously developed “zoo clonings” (Gyurján et al. 2007; Stéger et al. 2010) and sequencings over a number of deer genes strengthened this belief, since 98–100% similarity was registered in exons and more than 90% in the 5′ regulatory regions/promoters. The 365 deer MMSc-s uniquely identified the corresponding orthologous positions in the bovine genome. The array of the MMSc-s on the deer genetic map and the array of the bovine orthologues in the bovine genome were co-linear along all the chromosomes of the two species. Moreover, the syntenies for the in-MMSc genes were identical (see below in the section “Intra-scaffolds and intra-contigs syntenies”).

The majority of deer chromosomes (19 autosomes plus X, Y) could be directly paired with single B. taurus homologous chromosomes (for a typical example, see Fig. 2). Interestingly, eight acrocentric deer chromosomes appear to be tandemly fused into four acrocentric bovine ones (for a typical example, see Fig. S3), and conversely, one acrocentric deer chromosome appears to be split to two acrocentric bovine ones (Fig. S1). One metacentric deer chromosome appears to be centrically split to two acrocentric ones in the B. taurus complement (Fig. 3). The structural relation between deer chromosomes 19 and 31 vs. bovine 1 (i.e., Ce19 and 31 vs. Bt1) as well as for Ce26 and 28 vs. Bt9 seemed more complex. In the case of “Ce19 and 31 vs. Bt1”, a tandem fusion of the two acrocentric C. elaphus chromosomes and a translocation are combined (Fig. 4), whereas the tandem fusion of acrocentric chromosomes Ce26 and Ce28 in acrocentric Bt9 displayed a paracentric inversion in relation of Ce28 and Bt9 (Fig. S2). All structural relations of Ce and Bt chromosomes are summarized in Table 2. These observations are in complete accordance with the comparative cytogenetic analyses by Bonnet et al. (2001) between cattle and sika deer (C. nippon).

Fig. 3
figure 3

Example for the fission of deer metacentric chromosome Ce5 in two acrocentric bovine chromosomes Bt17 and Bt19. Numbering for cM and Mbp distances is similar to that in Fig. 2. Black dots correspond to centromeres

Full size image
Fig. 4
figure 4

Fusion of two deer chromosomes Ce19 and Ce31 in one Bt chromosome Bt1 coupled with a translocation. Note, a putative break in the C276–RT6/INRA11 region could convert the Ce19 and Bt1 sequences into each other. Numbers on the deer genetic map segments correspond to cM distances. Numbering along the pseudochromosomes correspond to distances in Mbp. Black dots correspond to centromerons

Full size image
Table 2 Pseudochromosome complement of Cervus elaphus: some basic characteristics
Full size table

The co-linearity test between the deer map and the sheep (O. aries) genome (ISGC Oar_v3.1) was repeated with a very similar outcome (data not shown).

Filling up the segments of the C. elaphus genetic map with scaffolds, guided by reference genes of B. taurus (RGSc-s)

The high similarity of the DNA sequences of individual genes of cattle and deer and the co-linearity of the map markers (MMSc-s) made it possible that aligning the deer scaffolds along the bovine genome sequence as template also led to a reliable genomic order for the deer scaffolds, i.e., the co-linearity and syntenies would manifest themselves at a lower, sub-chromosomal scale.

First, bovine reference genes corresponding to the cognate genetic map segments of deer were selected (UCSC Btau_4.6.1 and Bos_taurus_UMD.3.1.1) one by one along the bovine genomic DNA segments. Next, using the bovine reference gene sequences as “bait” and the deer scaffolds as “prey”, deer scaffolds that carried the orthologous deer gene of the bovine reference (we named these RGSc, reference gene-containing scaffolds) were angled. Consequently, after probing the deer RGScaffold sequences against the bovine genome sequence, the deer (genetic) map segments (defined by the deer/bovine DNA markers/MMSc-s at the flanks) were “filled up” with the deer RGScaffolds. In this way, beyond the 365 MMSc-s, a further 6013 RGScaffolds were arrayed—following the bovine gene order—in the corresponding deer genetic map segments. The final outcome, with the scaffolds arranged along the C. elaphus pseudochromosomes, is shown at http://emboss.abc.hu/wonderdeer/JBrowse, and a detailed example for one chromosome segment is shown in Fig. 5.

Fig. 5
figure 5

Examples for in-scaffold and in-contig syntenies in the CerEla1.0 genomic segment defined by MMSc-s sc_8 and sc_511. Note, syntenies parallel in deer genomic scaffolds and in the corresponding bovine genomic segments. Numbers at left: cM distances in deer genetic map. Middle: note the perfect syntenies for in-scaffold genes (e.g., along the 2.4-Mbp-long RGSc sc_9). Similar syntenies were recorded along the entire CerEla1.0 genome (see Results: “Intra-scaffolds and intra-contigs syntenies”)

Full size image

Filling up inter-scaffold gaps

The bait/prey relation was reversed in this search, i.e., DNA sequences of C. elaphus scaffolds served as bait and the bovine genomic segments as prey. All the map marker- and the reference gene-containing scaffolds (MMSc-s and RGSc-s) were omitted here from use as baits; hence, only those carrying so far unidentified genetic elements like protein-coding genes, rRNA, tRNA, and miRNA genes remained on the screen. We called these IRGSc-s, inter-reference genes’ scaffolds. Furthermore, the IRGSc-s should be located in the bovine genome in the gaps between an MMSc and an RGSc or between two RGSc-s. Altogether, 13,748 scaffolds contained genetic elements. In-scaffold syntenies of genes were then tested for deer–bovine relation (see the section below and Fig. 5).

After aligning the MMSc-s, RGSc-s, and IRGSc-s (13,748 combined), 15,205 scaffolds longer than 2Kb still remained non-localized. To find their positions, regions of the bovine genome sequence overlapping the previously mapped 13,748 deer scaffolds were masked out, and the remaining regions served as template for probing the 15,205 still non-localized scaffolds. In this way, 9845 new scaffolds (called GFSc-s, gap filling scaffolds) gained chromosomal positions (although in bovine order) resulting in 23,593 scaffolds having homologous regions in the bovine reference genome.

Although most of these homologous scaffolds (99.6%, 23,491/23,593) could be unambiguously aligned with the bovine genome, the localization for 102 scaffolds still remained ambiguous. These scaffolds were next broken up and rearranged in 35 new scaffolds and 2582 individual contigs, and then re-aligned scaffold by scaffold and contig by contig. In this way, we could solve all the intra-scaffold contradictions and find their final positions in the red deer pseudochromosomes.

Finally, the pseudochromosome complement of the haploid red deer (C. elaphus hippelaphus) reference genome (CerEla1.0) was assembled from 23,491 plus 35 scaffolds (MMSc, RGSc, IRGSc, GFSc-s plus the 35 new ones) and 2582 individual contigs, adding up to 26,108 sequence elements and spanning 3.4 Gbp.

We were unable to locate 11,444 scaffolds, adding up to 52,989,442 bp sequence. We classified these into the unplaced category.

Intra-scaffold and intra-contig syntenies

We selected scaffolds containing more than one genetic element (i.e., syntenic genes), and tested these intra-scaffold syntenies against the bovine genome (Btau_5.0.1). In cases when uncertainty arose, we also considered the sheep genome. In all cases investigated (3422 scaffolds), we registered identical local, in-scaffold syntenies for the deer and for the cognate, orthologous overlapping bovine chromosomal segments. Hence, the syntenies appeared not only at the chromosomal level for the map marker sites, but also at the sub-chromosomal (scaffold) level (examples are shown in Fig. 5 for scaffolds sc-8, sc-9, and sc-511).

Annotation of the C. elaphus genome

To annotate the genes on the red deer pseudochromosomes, we applied the MAKER annotation pipelines. As evidences, we used the complete bovine transcriptome, the human and ovine proteome, and the de novo partial transcriptome of sika deer. Using this pipeline, we could identify 19,368 genes. As mentioned above, the syntenies and order of the intra-scaffold deer genes were identical with those of the cognate bovine genes. The number of annotated genes is shown in Table 3.

Table 3 Summary of data of denovo assembly of CerEla1.0
Full size table

We could also annotate 589 rRNA genes (LSU, SSU), adding up to 98.3 Kbp sequence (0.0029% of the C. elaphus genome CerEla1.0), 1029 5 s rRNA genes (96 Kb and 0.0028% of CerEla1.0), 2096 tRNA genes (128 Kbp and 0.0038% of CerEla1.0), and 264 microRNA genes (27.7 Kbp, 0.0008% of CerEla1.0). These figures were in very good agreement with those found in other mammalian genomes.

Repetitive sequences were annotated both as part of the MAKER pipeline and also separately with dedicated programs. We applied the RepeatMaker program, which marked 22.73% of the genome as repetitive. In aggregate, it represents a 769,492,957-bp-long DNA sequence (Table 4).

Table 4 Repetitive sequences in the CerEla1.0 genome
Full size table

It is worth mentioning that our previously developed tetranucleotide STRs for 10 sites (DeerPlex for multiplex PCR analyses, Szabolcsi et al. 2014) were all identified in the scaffolds. Eight of these were in different chromosomes, and the pair (C01–T26) which matched the same linkage group in the bovine genome (Bt16) was also linked to the homologous deer chromosome 14 in the deer genome.

SNP pattern/heterozygosity along CerEla1.0 assembly

Since we sequenced the DNA of a stag, all the autosomes provided two copies of DNA. It allowed us to find and annotate the heterozygous autosomal variations in this individual (Table 5 and http://emboss.abc.hu/wonderdeer/JBrowse). We aligned the original reads to the pseudochromosomal sequences and determined the SNVs and the small indels. We could identify 2,807,458 SNVs. This number is only slightly lower than the 3.2–3.7 million that we found when we sequenced three Hungarian Mangalica pigs (Molnár et al. 2014). We also found 364,689 indels. Using the result of the MAKER annotation pipeline, we further annotated the heterozygous SNVs. In this way, we found 17,700 nonsynonymous and 14,252 synonymous variations.

Table 5 Indels and single nucleotide variations (SNV) in CerEla1.0
Full size table

Centromeron positions in the CerEla1.0 assembly

The karyotype for C. elaphus (as well as for some other Cervidae) and for B. taurus have been well established by classical chromosome cytogenetics (Gustavsson and Sundt 1968; Fontana and Rubini 1990; Bonnet et al. 2001). The structures of the chromosome complements are regarded “primitive”, since nearly, all the chromosomes are acrocentric (A chromosomes) and only rare metacentric ones (M chromosomes) can be recorded as a result of centric fusions of acrocentrics (i.e., Robertsonian translocations). Interestingly, however, tandem fusions and fissions (i.e., two acrocentrics to one acrocentric and vice versa) are often selected karyologically during the phyletic evolution of Pecoran descendant lineages (e.g., bovide, sheep, and deer). It is worth to note here that in this work, the comparisons for the C. elaphus and B. taurus pseudogenomes displayed the same chromosomal associations and homeologies (one-to-one correspondence, fusion, scission, and translocation) as those reported in the detailed comparative studies by Bonnet et al. (2001) on the chromosome RBG banding patterns of the sika deer (C. nippon) and B. taurus, which allowed us to combine their cytogenetic results with the genomics of this study (see below, Table 2). In contrast to the cytogenetic studies, the genetic map of deer did not define the positions of the centromerons (Table S3). Unlike in deer, in the case of cattle genetic markers close to the centromeron had been defined for all 29 autosomes and the X chromosome (Ma et al. 1996; Everts-van der Wind et al. 2004, 2005). We identified these markers in the successive B. taurus genome projects, although no markings or notes were found for the centromerons in the sequences of the pseudochromosomes (NCBI Bos_taurus_UMD3.1.1, Btau_5.0.1). However, comparisons of the orthologous genetic arrays of the deer and bovine pseudochromosomes combined with the cytogenetic comparisons of deer and cattle chromosomes (i.e., alignment of banding patterns of metaphase chromosomes) made possible to assign centromerons to all deer linkage groups (i.e., 34) and to the cognate pseudochromosomes (Table 2, slots 2 and 3). Furthermore, all the bovine homologies for deer centromerons were established (Table 2, slots 4 and 5). In certain cases, comparisons of the orthologous genetic arrays of the pseudochromosomes combined with the cytogenetic comparisons of deer and cattle chromosomes (i.e., alignment of RBG banding patterns of metaphase chromosomes, Bonnet et al. 2001) made possible, alone and independently, to assign centromerons to 6 deer linkage groups and to the cognate pseudochromosomes: Ce5, 15, 19, 26, 28, and 31. Briefly, Ce5 M is proven independently by cytology and genomics (Fig. 3 of this work and Fig. 2 in Bonnet et al. 2001); Ce15 A, the centromeron corresponds to that of Bt28, proven by the correspondence of its centromeron–proximal part to Bt28, and of the distal part to Bt26 (i.e., “tandem fusion”, Fig. S1 in this work and Fig. 2 in Bonnet et al. 2001); and Ce19 A corresponds to the centromeron distal half of Bt1. The centromeron position is proven by the observation that a well-defined centromeron–proximal segment of Ce19 “translocated during the evolution” to the end of Bt1. It is worth noting that the perfect congruence of the cytology and genomics is demonstrated by Fig. 4 in this work and by Figs. 2 and 5 in the paper of Bonnet et al. (2001); Ce26 A, the position of the centromeron is proven by the correspondence of Ce26 to the distal half of Bt9 (what is revealed as if a “tandem fusion of Ce26 and 28” happened, Fig. S2 of this work and Fig. 2 in Bonnet et al. 2001); and Ce28 A, the centromeron corresponds to that of Bt9. However, putting the centromeron in place, we considered that the banding pattern of Ce28 and that of the proximal part of Bt9 can be aligned, if a large paracentric inversion occurred in evolution (as demonstrated by Bonnet et al. 2001), and if, as a further precondition, this inversion overlapped the entire region defined by the orthologous map marker sites in both species (Fig. S2); and Ce31 A, the centromeron corresponds to that of Bt1. Its position is proven by the alignment of Ce31 with the centromeron–proximal half of Bt1, furthermore by Bt1’s appearing as a “tandem fusion of Ce31 and 19” (see Fig. 4 and also at Ce19 A, above).

On physical lengths and cM distances

The genetic map of C. elaphus we used for reference is 2532 cM long (Slate et al. 2002a); the assembled CerEla1.0 reference genome, as shown in this work, spans 3.4 Gbp, i.e., 1 cM corresponds to 1.34 Mbp in CerEla1.0. This figure is significantly higher than the 1Mbp/1 cM “thumb rule” or the 0.8 Mb/1 cM value established for the bovine genome (Arias et al. 2009). It is worth mentioning, however, that the deer and bovine mapping systems were significantly different, since the deer mapping system was an interspecific back-cross population, whereas in the case of cattle, it was based on an F2 (i.e., intra-specific) outbred population. It is also noteworthy that in the recently announced high-density genetic map of C. elaphus, the cM/Mbp ratio was proven to be 1.04 in a pedigree-based mapping system (Johnston et al. 2017).

The length of the deer pseudogenome CerEla1.0, we present here is 25% longer than that published for cattle in Btau_5.0.1 (NCBI, Science), namely, 3.4 and 2.7 Gbp, respectively. It is worth noting that the 2.7 Gbp Btau_5.0.1 genome that we used for reference and for template is built up only from sequenced DNA sections (i.e., no NNNs inserted). To test the 0.7 Gbp “surplus” length of CerELA1.0 in smaller genomic distances, the homologous genomic segments in the CerEla1.0 and Btau_5.0.1 pseudogenomes were compared. These segments were the ones defined by the orthologous map markers syntenic in both species. As shown in Fig. 6, with few exceptions, all along the deer and bovine pseudogenomes, the deer segments are uniformly 1.25-fold longer than the bovine segments, which also means that the lengths of the deer scaffolds in CerEla1.0 are uniformly 1.25-fold longer than the corresponding bovine genomic segments.

Fig. 6
figure 6

Genomic segments defined by the C. elaphus map markers in CerEla1.0 are compared with the B. taurus orthologous segments in Btau_5.0.1. Note, (i) circles: the great majority of pseudogenomic distances are 1.25-fold longer in CerEla1.0 than in Btau_5.0.1 and (ii) squares: distances along the corresponding pseudochromosomes deer 11 and cattle 11 are 2.2-fold longer in deer 11

Full size image

The 25% (0.7 Gbp) virtual excess length of the deer pseudogenome CerEla1.0 (3.4 Gbp) vs. the bovine Btau_5.0.1 (2.7 Gbp) can be accounted for by the too many “NNNs” inserted between the in-scaffold contigs by the ALLPATH-LG program. Provided that the lengths of the B. taurus and of the C. elaphus genomes are essentially concordant; furthermore, taking into account that in CerEla1.0, the combined DNA sequences of the contigs add up to 1.9 Gbp (Table 3, slot 3), 0.8 Gbp for the total length for the “in-scaffold NNNs” would be more feasible than 1.5 Gbp. However, Fig. 6 also shows that the ALLPATH-LG program distributed this 1.5 Gbp of “NNNs” proportionally with the corresponding physical distances. We believe that in CerEla1.0, the contigs cover 70% of the bovine genome Btau_5.0.1. (i.e., 1.9 Gbp per 2.7 Gbp).

Genomic segments created by rearrangements

Inversions, translocations, chromosomal fusions, and splits translocate or separate the mapmarkers and make new neighbourhoods around the break points of these rearrangements. In the deer–bovine relation, these “evolution-created” rearrangements resulted in joint sequences with new deer–bovine neighbourhoods. In this CerEla1.0 project, we registered 26 “deer–bovine” rearrangements, including 18 inversions, 2 translocations, and 6 chromosomal fusions/splits that led to new linkages of the orthologous mapmarker sites (Table 6). In the six cases, when one acrocentric bovine chromosome corresponded to two acrocentric deer chromosomes, the last segment of the “front” Ce chromosome and the first region of the “rear” Ce chromosome could not be defined unambiguously from our sequencing data (of scaffolds and contigs) combined with the alignments to the bovine genome. Scaffolds and contigs aligned to the corresponding bovine regions were shared between the deer chromosomes (in CerEla1.0) by considering the recombination distances in the deer genetic map and arranged in “bovine” order. The combined DNA sequences of this type added up to 0.166 Gbp, 5% of the CerEla1.0 genome (Table 6 part A).

Table 6 Genomic rearrangements in CerEla1.0 in relation with the bovine genome
Full size table

Inversions led to 54 deer–bovine sequence transition points at the flanking regions totalling 462.9 Mbp (13.5% of the CerEla1.0) as defined by the neighbouring MMSc-s (Table 6 part B). So far, we have not completed defining the points of the deer–bovine sequence switches. At the present stage in CerEla1.0, these flanking segments were filled up with the scaffolds and contigs exclusively following the bovine order. We believe, however, that future comparisons with other mammalian genomes and with a higher density genetic map of deer will allow more precise identifications.

All in all, in 81.5% of CerEla1.0, the genes were ordered according to the bovine order (between MMSc-s) or deer order (arrays of MMSc-s and of in-scaffold syntenic genes), whereas in 18.5% (at chromosomal splits/fusions, at flanks of inversions), syntenic blocks of deer and bovine genes were combined.

Discussion

In conclusion, in this study, we sequenced the genome of the red deer Cervus elaphus with the Illumina New Generation Sequencing technology. Guided by two references, i.e., by the co-linearity of the recombination map of the red deer Cervus elaphus and by the bovine reference genome sequence, we successfully assembled the whole genome as CerEla1.0. The sequences in CerEla1.0 were assorted in the pseudochromosome complement of the red deer, one of the most valuable members of the European megafauna, especially in the Carpathian Basin. CerEla1.0 made available deer, a so far non-model animal, for Genome-Wide Association Studies. This study demonstrates the power, both in the annotation of a new mammalian genome and in the pseudochromosome set assembly, of the combination of the genetic map, based on molecular markers, of the target species (the deer) with the existing genome reference of a taxonomically related species (Bos taurus)—in the present case, both belonging to ruminants Artiodactyla and Pecora.

This deer genome (CerEla1.0) was assembled from 26,108 sequence elements (scaffolds and contigs), assorted in 33 auto plus X and Y pseudochromosomes with a total length of 3.4 Gbp (1.9 Gbp without gaps). This means that the scaffolds of the draft deer genome covered virtually the whole B. taurus genome, the contig sequences covered 70% of the same, and 90% of the bovine orthologous genes were identified along the deer pseudochromosomes.

The relationships between the deer and bovine (and also of a number of ovine) chromosomes, i.e., congruencies, fusions, fissions, and inversions, were identified in a “semi-fine” scale (Tables 2, 3; Figs. 2, 3, 4, 5, S1, S2, S3).

In its present state, the gene array in the deer pseudochromosomes is a mosaic of deer and bovine segments. The order of the “mapmarker” scaffolds (MMSc-s) and the intra-scaffold syntenies represent the valid deer arrays.

The length of the CerEla1.0 genome is 3.4 Gbp: due to the somewhat long inserted intercontig NNNs, it is 1.25-fold longer than the bovine genome, 2.7 Gbp in Btau_5.0.1. All along the two genomes, at sub-genomic, sub-chromosomal, and scaffoldic levels, the physical distances are directly correlated with each other in the same ratio of 1.25 (as shown in Fig. 6). The approximately 0.7 Gb (virtual) surplus length for the deer genome can be accounted for by the ALLPATH-LG program characteristics (i.e., insertion of NNN tracts in the scaffolds proportionally between the contigs).

The validity of this deer genome sequence and pseudochromosome complement (CerEla1.0) is supported by:

  1. 1.

    Slate’s 365 genetic mapmarkers and the 10 Deerplex STRs were also identified in the bovine genome, and the deer and bovine arrays were co-linear along entire chromosomes as well as in chromosome rearrangements (fusions, fissions and translocations) relative to each other, i.e., between deer and cattle. It is worth noting that no reciprocal translocations were found in deer vs. bovine relation.

  2. 2.

    In all those cases (3422), when two or more genes were carried in a deer scaffold, the syntenies were identical in the deer scaffolds and in the corresponding bovine genome segments.

  3. 3.

    All genes cloned previously (related to antler development and cyclic physiological osteoporosis of deer stag, Molnár et al. 2007; Gyurján et al. 2007; Borsy et al. 2009; Stéger et al. 2010) as well as the STR loci previously developed for multiplex PCR analyses (DeerPlex, Szabolcsi et al. 2014) were recognized in the scaffold/contig sequences and localized in the pseudochromosomes. Both of the two STRs, C01 and T26, which indicated linkage by linkage disequilibrium (LD) test, were localized on C. elaphus pseudochromosome 14.

  4. 4.

    Although the C. elaphus genome CerEla1.0 was arranged in a mosaic deer–bovine order, the fair correlations for the orthologous deer and bovine genomic regions indicated that the deer scaffolds and contigs covered nearly the whole deer genome. This was supported by the fact that some 90% of the deer orthologs of the bovine protein-coding genes were identified in deer (19,368 for CerEla1.0 vs. 21,427 Btau_5.0.1).

Possible further studies: The reference genome CerEla1.0 of the red deer (Cervus elaphus hippelaphus) and its annotation, in accordance with fresh data from other programs, is under continuous monitoring and updating. If the sequence data of the SNP-based map markers will be available (Johnston et al. 2017), the updating of CerEla1.0 will become possible using the approach described in this work. The sequence and the pseudochromosome complement of CerEla1.0 may provide a basis and a rich source for broader interests, including, among others: conservation genetics, refined evolution, and population studies within the family Cervidae [e.g., fallow deer (Dama dama) or roe deer (Capreolus capreolus)] as well as in a wider range of ruminants and Pecora. CerEla1.0 also provides a source for developing chromosome-specific microsatellite sets.

A large number of SNP/heterozygotic sites were identified (2.8 × 106 SNVs, 3.6 × 105 indels) and aligned to the deer pseudochromosomes. CerELA1.0 is a leading basis for future genome-wide SNP and microsatellite studies, which may shed light on inbreeding/outbreeding, may help in the identification of gene introgressions and of descents for autosomal, maternal, and paternal lineages. Forensic identification, or definition of allelic compositions underlying phenotypes important, for example, in game, management could also be possible areas of utilization. The exploration of the genetic secret of record antlers becomes possible by genome-wide association studies. Applications and utilizations in several fields of medical research (e.g., bone and osteoporosis research, organ development and regeneration, and robust tissue proliferation/tumour biology) are also feasible.