Improving Genome Assemblies and Annotations for Nonhuman Primates

Abstract

The study of nonhuman primates (NHP) is key to understanding human evolution, in addition to being an important model for biomedical research. NHPs are especially important for translational medicine. There are now exciting opportunities to greatly increase the utility of these models by incorporating Next Generation (NextGen) sequencing into study design. Unfortunately, the draft status of nonhuman genomes greatly constrains what can currently be accomplished with available technology. Although all genomes contain errors, draft assemblies and annotations contain so many mistakes that they make currently available nonhuman primate genomes misleading to investigators conducting evolutionary studies; and these genomes are of insufficient quality to serve as references for NextGen studies. Fortunately, NextGen sequencing can be used in the production of greatly improved genomes. Existing Sanger sequences can be supplemented with NextGen whole genome, and exomic genomic sequences to create new, more complete and correct assemblies. Additional physical mapping, and an incorporation of information about gene structure, can be used to improve assignment of scaffolds to chromosomes. In addition, mRNA-sequence data can be used to economically acquire transcriptome information, which can be used for annotation. Some highly polymorphic and complex regions, for example MHC class I and immunoglobulin loci, will require extra effort to properly assemble and annotate. However, for the vast majority of genes, a modest investment in money, and a somewhat greater investment in time, can greatly improve assemblies and annotations sufficient to produce true, reference grade nonhuman primate genomes. Such resources can reasonably be expected to transform nonhuman primate research.

Introduction

Nonhuman primates include the great apes—chimpanzees (Pan troglodytes), bonobos (Pan paniscus), gorillas (Gorilla gorilla), and orangutans (Pongo abelii); the lesser apes including the gibbon (Nomascus leucogenys); old world monkeys including rhesus macaques (Macaca mulatta), cynomolgus monkeys (Macaca fascicularis), pigtail macaques (Macaca nemestrina), baboons (Papio anubis), African green monkeys (genus: Chlorocebus), and sooty mangabeys (Cercocebus atys); new world monkeys including marmosets (Callithrix jacchus); and strepsirrhines including the aye-aye (Daubentonia madagascariensis), the grey mouse lemur (Microcebus murinus), and the sifaka (Propithecus coquerelli). Draft genomes for all of the species listed above are either in progress or are complete (Table 1; Marques-Bonet et al. 2009). This vast group of species represents a significant span of primate evolutionary history (Figure 1). These animals are used in two main types of studies: the study of human evolution and as models for translational medicine. Both types of study would benefit from high quality genomic information, a resource that is currently not available for any nonhuman primate.

Table 1.

Status of nonhuman primate genome sequencing projects

					Coverage
Common name	Scientific name	Sanger	Illumina	454	Solid PacBio	Status	Refs
Rhesus macaque	Macaca mulatta	6X				Complete	1
Rhesus macaque	Macaca mulatta		47X			Complete	2
Cynomolgus macaque	Macaca fascicularis		54X			Complete	2
Cynomolgus macaque	Macaca fascicularis				40X	Complete	3
Cynomolgus macaque	Macaca fascicularis			3X	3X	Complete	4
Chimpanzee	Pan troglodytes	6X				Complete	5
Chimpanzee	Pan troglodytes verus (10 samples)		9X			Complete	6
Chimpanzee	Pan troglodytes (25 samples, 4 subspecies)		23X			Complete	7
Bonobo	Pan paniscus			26X		Complete	8
Bonobo	Pan paniscus (13 samples)		27X			Complete	7
Gorilla	Gorilla gorilla	2X	60X			Complete	9
Gorilla	Gorilla gorilla (27 samples, 3 subspecies)		18X			Complete	7
Orangutan	Pongo abelii	6X				Complete	10
Orangutan	Pongo abelii/pygmaeus (10 samples)		9.3X			Complete	10
Orangutan	Pongo abelii/pygmaeus (10 samples)		27X			Complete	7
Aye-aye	Daubentonia madagascariensis		38X			Complete	11
Common marmoset	Callithrix jacchus	6X				In progress	12
White-cheeked gibbon	Nomascus leucogenys	6X				In progress	13
Olive baboon	Papio anubis	2X	85X	4.5X		In progress	14
African green monkey	Chlorocebus aethiops			18X		In progress	15
Mouse lemur	Microcebus murinus	2X	100X		7X	In progress	16, 17
Sooty mangabey	Cercocebus atys		107X		6X	In progress	17
Pigtail macaque	Macaca nemestrina		100X		6X	In progress	17
Sifaka	Propithecus coquerelli		100X		6X	In progress	17

Note: Sequences from different platforms but from the same individual are on the same horizontal line. Sequences from the same species but different individuals are indicated on different horizontal lines. When information is available on the likely platform and amount of coverage for “in progress” species is available, estimates are provided.

References: 1. Gibbs et al. (2007); 2. Yan et al. (2011); 3. Higashino et al. (2012); 4. Ebeling et al. (2011); 5. Chimpanzee Sequencing and Analysis Consortium (2005); 6. Auton et al.; 7. Prado-Martinez et al. (2013); 8. Pruüfer. et al. (2012); 9. Scally et al. (2012); 10. Locke et al. (2011);

11 Perry et al. (2012); 12. http://genome.wustl.edu/genomes/detail/callithrix-jacchus/ (accessed on July 19, 2013); 13. http://genome.wustl.edu/genomes/detail/nomascus-leucogenys/ (accessed on July 19, 2013); 14. http://www.ncbi.nlm.nih.gov/nuccore/AHZZ00000000.1/ (accessed on July 19, 2013); 15. http://genome.wustl.edu/genomes/detail/chlorocebus-aethiops/ (accessed on July 19, 2013); 16. https://www.hgsc.bcm.edu/content/mouse-lemur-genome-project (accessed on July 19, 2013); 17. Personal communication, Dr. Jeff Rogers.

Figure 1:

Phylogeny of nonhuman primates The point of divergence from human's the last common ancestor is indicated at the branching points in millions of years ago (MYA). Strepsirrhines include the aye-aye (Daubentonia madagascariensis), grey mouse lemur (Microcebus murinus), and sifaka (Propithecus coquerelli). New world monkeys include the common marmoset (Callithrix jacchus). Old world monkeys include rhesus macaques (Macaca mulatta), cynomolgus monkeys (Macaca fascicularis), pigtail macaques (Macaca nemestrina), baboons (Papio anubis), African green monkeys (genus: Chlorocebus), and sooty mangabeys (Cercocebus atys). Gibbons are lesser apes, including the white-cheeked gibbon (Nomascus leucogenys). Orangutans (Pongo abelii), gorillas (Gorilla gorilla), chimpanzees (Pan troglodytes), and bonobos (Pan paniscus) are all great apes. Bonobos and chimpanzees last shared a common ancestor about one million years ago. Draft genomes are (or will be) available for all the species listed (see Table 1).

Table 2.

Nonhuman primate transcriptome projects

Common name	Scientific name	Refs
Chimpanzee	Pan troglodytes	1, 2
Gorilla	Gorilla gorilla	2
Rhesus macaque	Macaca mulatta	1, 2
Cynomolgus macaque	Macaca fascicularis	2
Japanese macaque	Macaca fuscata	2
Pig-tailed macaque	Macaca nemestrina	2
Olive baboon	Papio anubis	2
African green monkey	Chlorocebus aethiops	1, 2
Sooty mangabey	Cercocebus atys	2
Common marmoset	Callithrix jacchus	1, 2
Owl monkey		2
Squirrel monkey		2
Mohol bushbaby	Galago moholi	1
Slow loris		1
Aye-aye	Daubentonia	1
Madagascariensis
Coquerel's sifaka	Propithecus coquereli	1
Black & white ruffed lemur	Varecia variegata	1
Ring-tailed lemur	Lemur catta	2
Mongoose lemur	Eulemur mongoz	1
Crowned lemur	Eulemur coronatus	1
Mouse lemur	Microcebus murinus	2

References: 1. Perry et al. (2012); 2. Pipes et al. (2013)

Evolutionary studies in nonhuman primates are important, not only for their intrinsic intellectual interest, but also because filtering genetic variants identified in NextGen sequencing studies, designed to determine which mutations are related to disease, requires a knowledge of evolutionary context (MacArthur et al. 2012; Torkamani et al. 2012; Norgren 2012). High quality genomes are important for these studies because the comparison of genes across species cannot yield correct results if the genes for a given species have not been annotated, or have been incorrectly annotated (Vallender 2009). To conduct these studies, genomic information from species closely related to humans like chimpanzees and species more distantly related like the lesser apes, and monkeys, are necessary for correct results.

Biomedical researchers engaged in translational research depend on nonhuman primates because these animals are most likely to recapitulate human symptoms, and offer the best hope for predicting human responses to experimental therapies. Rhesus macaques are the most frequently used animal models among the nonhuman primates. This species is used extensively for investigations into the pathogenesis of AIDS, and countermeasures including vaccines (Baroncelli et al. 2008; Shedlock et al. 2009). Sooty mangabeys and African green monkeys are also of interest to AIDS investigators because these species are naturally resistant to HIV (Bosinger et al. 2011; Chahroudi et al. 2012; Ma et al. 2013). Pharmaceutical companies use cynomolgus monkeys extensively to test drugs before advancing to clinical trials because the toxicological profile of a drug can be different in rodents than in humans and nonhuman primates (Chellman et al. 2009). Nonhuman primates are also used for studies of higher order cognitive function and behavior (Watson and Platt 2012; Fox et al. 2012; Shackman et al. 2013), reproduction (Hewitson 2004; Sparman et al. 2007), and cardiovascular disease (Vinson et al. 2011). All of these studies could benefit from NextGen sequencing approaches, especially mRNA-seq for expression analysis, and exome sequencing for investigations of genetic effects.

Most NextGen sequencing studies require a reference genome sequence against which to align the short sequences in these studies. The current draft rhesus genome has been found to be inadequate as a reference genome. For example, investigators conducting expression studies in rhesus macaques have not been able to obtain high quality data when aligning rhesus NextGen mRNA sequences against the draft genome (Dr. J Knowles, University of Southern California, Keck School of Medicine, personal communication, 2013, and Dr. N Kalin, University of Wisconsin School of Medicine and Public Health, personal communication, 2013). Attempts to use the rhesus genome for NextGen Exome studies have also revealed serious problems (Vallender 2011). Similar issues likely apply to the genomes for other nonhuman primates.

Limitations of Draft Nonhuman Primate Genomes and Annotations

The “draft” human genome was published with 7.5X Sanger coverage (Lander et al. 2001). A considerable amount of additional “finishing” Sanger sequencing was performed to produce a higher quality human genome assembly (International Human Genome Sequencing Consortium, 2004). Additional “finishing” of human chromosomes continued to 2006 (Dhand 2006). In contrast, nonhuman primates have been sequenced with, at most, 6X Sanger coverage (Table 1). Three species have received this level of Sanger coverage: rhesus macaques (Gibbs et al. 2007), orangutans (Locke et al. 2011), and chimpanzees (Chimpanzee Sequencing and Analysis Consortium 2005). It is anticipated that a similar level of coverage will be obtained by Sanger sequencing for the common marmoset (http://genome.wustl.edu/genomes/detail/callithrix-jacchus/) and white-cheeked gibbons (http://genome.wustl.edu/genomes/detail/nomascus-leucogenys/). The gorilla (Scally et al. 2012), bonobo (Prüfer. et al. 2012), and cynomologus macaque (Yan et al. 2011) were sequenced with a mixture of Sanger and NextGen methods, or NextGen methods alone (Table 1). As a result of shorter read lengths, many more NextGen sequences are needed to produce Sanger level coverage. This fact makes comparisons of coverage between the different technologies difficult. The mixed and NextGen only genomes are likely to be even less complete than the Sanger only genomes. Thus, the available nonhuman primate genomes have received far less attention to quality of assembly than the human genome.

Since the quality of the rhesus macaque genome has been independently assessed by a number of methods and it was finished to a similar level as the other nonhuman primate genomes, this species will be used as an example of the limitations of draft nonhuman primate genomes. Radiation hybrid (Karere et al. 2008) and FISH analyses (Roberto et al. 2008) have demonstrated numerous misassemblies in the rhesus genome. Using a gene-based approach (Zhang et al 2012), several types of misassemblies have been documented including: scaffolds assigned to the wrong chromosome (Figure 2A), and scaffolds placed in the wrong orientation (Figure 2B). In addition, there were many scaffolds containing entire genes or fragments of genes that were not placed on any chromosome (Figure 2B). There are documented sequencing errors that introduced apparent nonsense mutations in exons (Figure 2C). The assembly and sequencing errors cause automated annotators to incorrectly or fail to annotate many genes. Further, missing exons can result in automated annotation pipelines, which generate spurious sequences. Automated annotators may use intronic sequence to create incorrect gene models (Figure 2D). Serious errors in annotation of the rhesus genome have also been identified in the course of attempting to use it for evolutionary studies and NextGen exome analysis (Vallender 2009, 2011). It is estimated that approximately 50% of the rhesus gene annotations available at NCBI are either incomplete or incorrect (Zhang et al. 2012). It is important to note that this is not just an issue for nonhuman primate genomes. Similar problems have been observed for other draft genomes as well (Nagy 2008, 2011).

Figure 2:

Schematic diagrams illustrating assembly and annotation errors in the rhesus macaque draft genome. (A) Scaffold assigned to the wrong chromosome: The scaffold containing exons 1, 2, 4, 5, and 6 of the SRC homology 2 domain containing E (SHE) gene is correctly assigned to chromosome 1 in the rhesus draft genome. However, the scaffold containing exon 3 of the SHE gene was incorrectly assigned to chromosome X. (B) Scaffold with exon in the wrong orientation: An unlocalized scaffold from the draft rhesus genome contains exons 9-13 of the Bardet-Biedl syndrome 1 (BBS1) gene. It was not included in the rhesus chromosome 14 file with the scaffold that contains exons 1-8 of BBS1. This is likely the contig containing exon 9 was in the wrong orientation with respect to the rest of the scaffold. (C) Sequencing error results in apparent nonsense mutation: The rhesus draft genomic DNA had sequencing error in the adrenergic, beta-1, receptor (ADBR1) gene. This introduced a premature stop codon (arrow, top panel). This has resulted in this locus being labeled a pseudogene by NCBI. Our targeted sequencing of this region has revealed the correct sequence (JN589014.1 - bottom panel). (D) Missing exon results in substitution of intronic sequence: The original rhesus draft genome did not contain the sequence for exon 15 for the adenylate cyclase 3 (ADCY3) gene. Instead, intronic sequence between exons 14 and exon 16 was substituted (top panel) when this gene was annotated. This led to spurious protein sequence (original protein) and a premature stop codon. The missing exon 15 was sequenced and deposited in GenBank (HM067826.1). NCBI then corrected the rhesus ADCY3 gene model and now reports a correct protein sequence for this gene (bottom panel). Figure 2 is redrawn from Zhang et al. 2012.

Assemblies have been reported for a Chinese rhesus macaque (Yan et al. 2011) and several cynomolgus macaques (Ebeling et al. 2011; Yan et al. 2011; Higashino et al. 2012). However, because these assemblies were all dependent on the original chromosome files produced for the reference rhesus animal, they can be expected to contain many of the same misassemblies documented for the rhesus genome.

Comparing draft nonhuman primate genomes with the human genome will become increasingly difficult as NextGen sequencing is either combined with existing Sanger sequences or used separately. There are no clear standards or benchmarks for assessing the quality of draft sequences. However, there have been independent efforts at evaluating different assemblers (Narzisi and Mishra 2011; Earl et al. 2011; Salzberg et al. 2012). The N50 statistic is often reported for assemblies. The bigger the number, the longer the contigs contained within an assembly. Although it may be tempting to judge reports of bigger N50s as better, this is not necessarily the case (Narzisi and Mishra 2011; Salzberg et al. 2012). Scaffolds, and even contigs, can be misjoined, that is, sequences are placed together that should be far apart, perhaps even on different chromosomes (Salzberg et al. 2012; Zhang et al. 2012). Depending on the assembler chosen, one can get smaller N50s with more accurate assemblies, or bigger N50s with more mistakes in the assembly (Narzisi and Mishra 2011; Salzberg et al. 2012). In the final analysis, the most important quality metric for most users is the number of genes that can be correctly and completely annotated for a given genome. Unfortunately, annotation reports in most mammalian genome papers are so terse as to make it difficult to assess their completeness or correctness.

Strategies for Improving Nonhuman Primate Genomes

Finishing a draft genome, to the level of the human genome, using the same approaches would be prohibitively expensive. There is considerable interest in using much less expensive NextGen sequencing to “top off” Sanger sequencing, or replace it entirely. At this time, there is no standardized approach for using NextGen sequences to produce a high quality genome. However, for animals for where there are already Sanger sequences available, it makes sense to utilize a hybrid strategy that combines the long reads from the Sanger sequencing with NextGen Sequences (Figure 3). Ideally, the genomic DNA used for the NextGen sequencing should come from the same animal that was used for the Sanger sequencing.

Figure 3:

Flowchart describing assembly and annotation procedures The steps involved in creating a high-quality genome. Sequencing can include the conventional Sanger technique and/or several NextGen technologies including 454, Illumina, and Ion Torrent (see Table 1). Contig and scaffold assembly can utilize several assemblers including: Atlas (Havlak et al. 2004), AbySS (Simpson et al. 2009), ALLPATHS-LG (Gnerre et al. 2011), Celera assembler (Myers et al. 2000), MaSuRCA (http://www.genome.umd.edu/masurca.html) (accessed on July 19, 2013), and SOAPdenovo (Li et al. 2010). Chromosome mapping can use genetic information, radiation hybrids or fluorescence in situ hybridization (FISH). “Breaking” misassembled scaffolds and placing them on chromosomes can involve extensive manual work. Expressed sequence tags (ESTs) are usually partial transcripts obtained from Sanger sequencing. mRNA-seq is often performed with Illumina technology but can also be conducted with Ion Torrent machines.

After contigs and scaffolds are created using an assembly pipeline, they must be examined for misassemblies. It is not uncommon for scaffolds to contain fragments that belong in different locations in the same chromosome, or even on different chromosomes (Figure 2A). This is because repeat regions are scattered throughout the genome and can confuse assemblers. One test relating to the correctness of scaffolds, examines exon order within genes. Since exon order is completely conserved among mammals, exons for a given gene should be in the same order and orientation in all mammals. This information can be used not only to identify misassembled scaffolds, and correct them, but also to place scaffolds in the correct order. If two different scaffolds each contain some exons from the same gene, then they must belong adjacent to each other.

Synteny is relatively well conserved among the apes and old world monkeys. Still, there have been a number of chromosome fission, chromosome fusion, and rearrangements in the course of primate evolution (Kehrer-Sawatzki and Cooper 2008; Capozzi et al. 2012; Nie 2012; de Oliveira et al. 2012; Stanyon et al. 2012). Thus, to place scaffolds on chromosomes accurately, mapping information is necessary. Three types of maps are available for the rhesus macaque: genetic (Rogers et al. 2006), radiation hybrid (Karere et al. 2008), and FISH (Roberto et al. 2008). Although helpful, even more detailed maps would be useful when placing scaffolds correctly on chromosomes. For other nonhuman primates, the amount of mapping data varies, but is generally less than ideal.

Improved assemblies are necessary to make nonhuman primate genomes useful, but they are not sufficient. Annotations, specifications of exon ranges from named genes within chromosome sequences, are required before nonhuman primate genomes can be used for NextGen expression and exome studies. Many regions of draft nonhuman primate genomes have provisional annotations. For example, a region may be annotated with a “gene” which is designated with “LOC” prefixed to a set of numbers. Although this may be counted as an annotation, it is of little value to users. Investigators prefer that orthologs of human genes be annotated with the same names so that comparisons between human and nonhuman primate data can be easily made.

Annotations of nonhuman primate assemblies can be improved by aligning named transcripts against the chromosome files of the genome to be annotated using programs like sim4cc (Zhou et al. 2009), GMAP (Wu and Watanabe 2005), and Spidey (Wheelan et al. 2001). Due to their close evolutionary relationships, the well-characterized human RefSeq transcripts can be used to identify orthologs in most primate species. However, mRNA transcripts from the closest species to the target genome will work the best. Ideally, one would use transcripts from the same species being annotated. Some expressed sequence tags (ESTs) are available for nonhuman primates (Magness et al. 2005). However, in no case are these complete. One way to economically and efficiently obtain transcripts is to perform NextGen, mRNA sequencing from cells and tissues derived from the species of interest. (Table 2; Pipes et al. 2013; Perry et al. 2012) Transcripts can then be constructed using de novo assemblers such as velvet or oases (Zerbino and Birney 2008; Schulz et al. 2012), and Trinity (Grabherr et al. 2011).

For the vast majority of nonhuman primate transcripts, establishing orthology with human genes is straightforward. However, for highly divergent, and highly polymorphic genes, such as those contained within the MHC, annotation is much more difficult. For these regions, the first challenge is the assembly itself. It is difficult to get a good assembly of the MHC region without careful bacterial artificial chromosome (BAC) sequencing, as has been performed for the rhesus macaque (Daza-Vamenta et al. 2004). Long, accurate reads are currently not possible with NextGen sequencing, but may be available in the near future. Although this will aid assembly of the MHC region, annotation will still be a challenge. Because there is so much variation in MHC genes among individuals of any nonhuman primate species, no one annotation of an individual's genome will encompass all of the MHC genes. Extensive characterization of MHC transcripts followed by manual annotation of chromosome files will be required. For highly polymorphic immune system genes, specialized databases such as the International ImMunoGeneTics information system (http://www.imgt.org/) will be useful adjuncts to the reference genome.

Future Opportunities

High quality nonhuman primate genomes have the potential to catalyze biomedical research. A rhesus macaque GeneChip, in collaboration with Affymetrix has been developed (Spindel et al. 2005; Duan et al. 2007). Although this has proven to be a useful reagent for gene expression studies for rhesus macaques, mRNA-seq using NextGen sequencing promises cheaper and more sensitive expression studies. Further, expression microarrays are not available for most nonhuman primates. State-of-the-art expression studies can be performed in any nonhuman primate species once a high quality genome is available. These studies will likely accelerate biomedical research.

Exome studies in rhesus macaques are currently impaired by the lack of a high-quality genome (Vallender 2011). However, once one becomes available, it should be possible to economically survey colonies of rhesus macaques and identify pairs of animals with mutations in the same Mendelian recessive genes. Directed breeding could then be used to develop nonhuman primate models of human genetic disease. Such animals will be useful for preclinical therapeutic tests.