Tissue sampling

We included three samples in this study, which represent one sample from each of three species of warbling-finches: the Gray-Throated Warbling Finch (Poospiza cabanisi), Buff-Throated Warbling Finch (Poospiza lateralis) and Bay-chested Warbling Finch (Poospiza thoracica). These species inhabit montane and subtropical forests in SE and S Brazil, and they are being studied as part of a larger phylogeography study (Amaral et al. in prep). A recent phylogenetic study [28] including two of these species (P. thoracica and P. cabanisi) suggests that they are not close relatives, and that Poospiza may be polyphyletic. Poospiza lateralis and P. cabanisi, in turn, appear to have split recently, and their divergence may be among the most recent in the Neotropics (Amaral et al., unpublished data). Poospiza lateralis and P. cabanisi are allopatric, while P. thoracica is sympatric to both. Specimens were attracted in the field using playback recordings and collected using an air shotgun [29]. This research has been approved by the ethics committee of the Federal University of São Paulo (CEP 0069/12), including the adopted sampling procedure. Field expeditions and specimen collection permits for all visited localities were granted by ICMBio (permit numbers 14673 and 30840). All specimens were housed at the ornithological collection of the Museu de Zoologia da Universidade de São Paulo (MZUSP), and tissue samples deposited at the Laboratório de Genética e Evolução de Aves da Universidade de São Paulo (LGEMA-USP, see Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sample information and descriptive sequence statistics.

https://doi.org/10.1371/journal.pone.0138446.t001

Laboratory methods

We extracted total DNA from muscle samples using the Qiagen DNeasy kit (Valencia, CA) according to the manufacturer's protocol, including the suggested RNAse treatment. Sequence capture and sequencing of ultraconserved elements were performed according to the original UCE capture and sequencing protocols [16] by RAPiD Genomics (Gainesville, FL, USA). Modifications made to the original protocol consisted of the reduction from 2,560 to 650 probes (targeting 634 loci selected to cover all chromosomes of the Gallus genome), use of 100 bp paired-end Illumina Hiseq 2000 sequencing run, and use of 16 cycles in both pre- and post-capture PCR reactions. We sequenced the three specimens used here in a multiplexed batch of 48 individuals (i.e. sequencing included 45 individuals from other studies).

Sequence quality control and assembly

We separated raw sequences by individual tags using Illumina's Casava software. Sequences were initially evaluated using FastQC 0.10.1 [30]. Adapters, barcodes and low quality regions were removed using Illumiprocessor 2.0.7 [31], which allows processing of Illumina sequencing reads using the trimming tool Trimmomatic 0.32.1 [32]. The resulting reads were processed using Phyluce 1.4 [33], which is a software package that allows processing of trimmed reads into aligned loci. We performed processing in Phyluce up to the assembly stage, using the Trinity assembler v20140717 [34]. All analyses were performed using default parameters. Contigs resulting from assembly were used for downstream analyses.

mtDNA assembly, variant calling, annotation and phylogenetic inference

Contigs were compared to all avian mtDNA genomes at NCBI as of August 6^th 2014 using the standalone version of BLAST 2.2.28 [35]. Contigs with a strong match (e-value < 0.1) to avian mtDNA were used to manually assemble the mitogenome in Bioedit 7.1.3 [36]. Despite the haploid nature of mtDNA, we called SNPs in order to detect potential heteroplasmy and/or PCR errors. Paired reads processed in Illumiprocessor were resynchronized using custom scripts and then mapped to the respective mitogenome using Bowtie2 v2.2.4 [37]. PCR duplicates, which consist of identical sequences that arise due to amplification during library preparation, were removed using Samtools 0.1.19 [38], and variants and indels were called with Freebayes 0.9.15 [39] using the diploid option and standard filters (-m 30-q 20-R 0-S 0). Sequencing depth statistics were obtained using Bedtools 2.17 [40].

Complete mitochondrial genomes were annotated based on the automatic methods of MITOS [41] and DOGMA [42] using default parameters. Incongruences in the results obtained between the two methods for each species were manually adjusted based on comparisons to the annotated mtDNA genomes of Gallus gallus (NC_001323) and Emberiza pusilla (NC_021408). Final annotation of the mitogenomes was performed using Geneious 6.1.8 [43].

In order to evaluate mitochondrial sequence authenticity, we performed a phylogenetic analysis including cytochrome b sequences from the mitogenomes obtained here and public sequences from species attributed to the Poospizinae sub-family [28], which were sequenced by those authors using standard Sanger sequencing. We included representatives of the genera Poospiza, Compsospiza, Cnemoscopus, Cypsnagra, Donacospiza, Hemispingus, Nephelornis, Piezorina, Pyrrhocoma, Thlypopsis, Urothraupis and Xenospingus. We generated sequence alignments in Muscle [44] and performed model selection based on the Akaike Information criterion as implemented in MrModeltest 2 [45], which indicated HKY+I+G as the best-fit model. We performed Bayesian phylogenetic inference in MrBayes 3.2.3 [46] using the CIPRES infrastructure [47], adopting a conservative burnin of 25% and Xenospingus concolor as outgroup. MrBayes runs were performed twice to verify convergence of topology and posterior probabilities obtained from independent runs.

Identification of microsatellite loci candidates

Contigs resulting from the Trinity assembler (implemented in Phyluce) were used as input for QDD 1.3 [48], which is a software package that detects microsatellites and allows primer design. We searched for pure di-, tri-, tetra-, penta- and hexa-nucleotides with at least five repeats. Potential primers were designed using default settings in Primer3 [49] as implemented in QDD. Sequences used for primer design were compared to the corresponding species’ mitogenome using BLAST, and removed in case of matches. The same procedure was also performed using sex chromosomes in order to identify sex-linked microsatellites. Comparisons to sex chromosomes were based on both chicken and zebra finch Z chromosomes, while only the chicken W chromosome was used due to the lack of currently available zebra finch W chromosome sequences. Identification of homologous loci across species was based on alignment of sequences used for primer design in Geneious. Alignments with a minimum of two species were used to evaluate microsatellite polymorphism.

Mitogenomes

Here we show that UCE sequencing may be a low-cost (~ US$ 100 per sample) and efficient method for obtaining high-depth complete mitogenomes based solely on off-target sequences, with no need for mtDNA-specific baits. The three samples sequenced here provided complete mitogenomes along with at least 625 out of 634 target loci per sample (Table 1). Mitogenome sequencing depth was high in all cases (Table 1), with means higher than 500x when PCR duplicates were considered. Removal of the PCR duplicates considerably dropped depth, but mean depths remained high (>70x), with depth ranging from 2x to 122x. Use of skeletal muscle as source material, which has a high mtDNA/nDNA ratio, may be responsible for such high mean depths [25,50], and further tests will be important to evaluate the efficiency of this protocol with other types of tissue. Blood samples, for example, appear to be a poor source of mtDNA off-target in UCE experiments (B. Faircloth, personal communication). Our results in particular showed PCR duplication levels higher than 90% for all three species, and those values may be considerably lowered by decreasing the number of PCR cycles during library amplification (B. Faircloth, personal communication).

Mitogenome completeness of the three warbling finches is supported by the presence of all regions found in two close mitogenomes available (Thraupis episcopus, [51]; Emberiza pusilla, [52]), including one control region, two rRNAs, 22 tRNAs and 13 protein-coding genes (S1 Table). Gene synteny was observed among the three species (S1 Table), and their gene order is the one present in most oscines studied so far. Genome sizes and G+C contents were also similar to those of published avian mitogenomes, ranging between 16,768–16,773 bp and 46.6–47.1%, respectively. Five mitochondrial polymorphic sites were found (P. lateralis, T and C at position 7847 and A and T at position 15290; P. cabanisi, C and A at position 15600; P. thoracica, A and T at position 14, and T and C at position 15606), and they can be explained by PCR error or heteroplasmy, which is an increasingly common finding in mtDNA studies [53–55]. Although an early diverging numt could also explain these polymorphic sites [56], we expect that use of muscle as DNA source makes this hypothesis unlikely due to its high mtDNA/nDNA ratio. Bayesian phylogenetic inference based on cytochrome b sequences also supported authenticity of the mitochondrial sequences with maximum posterior probabilities (S1 Fig). Poospiza lateralis was sister to P. cabanisi, a finding in line with their conspecificity according to some authors (see [57] for a taxonomic review). In addition, both sequences of P. cabanisi and P. thoracica collected here were sister to conspecific sequences obtained from Genbank.

Microsatellite loci

We found a minimum of 361 microsatellites per species (Table 1), with primer design being possible for at least 64 loci per species (S1 Table). These loci numbers are higher than those recommended by Gardner et al. [58] for typical molecular ecology studies (~40 designable, unique loci). We believe that the number of microsatellites for which primers cannot be designed may decrease with use of inserts larger than the ones used here (~300 bp), which could translate into higher average assembly length and consequently longer flanking regions for primer design. We found microsatellite loci in both UCE and non-UCE data. Purifying selection may be a concern in UCE regions [59], which may include linked microsatellite loci. However, less than one fifth of the loci with flanking sequence suitable for primer design found here matched UCEs (P. cabanisi: 11 of 64; P. lateralis: 11 of 77; P. thoracica: 10 of 68). No mitochondrial or W-linked microsatellite loci were found, but a minor proportion (eight in P. cabanisi, eight in P. lateralis, and nine in P. thoracica) were Z-linked. From the microsatellite loci alignable between at least two species (23), 10 matched UCE loci and 13 consisted of non-UCE loci, with one of each category being variable. Although the small panel of individuals (2–3) may not be sufficient to reliably explore polymorphism, most of the sequences found were outside UCE regions, and thus should present levels of variation similar to those of microsatellite loci isolated using standard techniques (e.g. microsatellite enrichment).

Advantages and potential pitfalls

Obtaining mitogenomes and microsatellite loci from off-target sequences is an opportunistic approach that depends on the incomplete efficiency of sequence capture protocols. Thus, future increases in sequence capture efficiency can negatively affect any application based on off-target sequences. However, despite the rising efficiency in sequence capture, off-target sequences are still common (e.g. protocols with only 40% on target in some cases, [26]), providing abundant mtDNA sequences and microsatellite loci. Even in cases of high specificity, increasing sequencing depth may still provide sufficient off-target sequences for diverse applications at reasonable costs. When the intended number of microsatellite loci or the desired mitogenome completedeness is not met, the low cost associated with UCE sequencing (~ US$ 100/sample) makes replicated sequencing of the same sample or sequencing of multiple individuals of a given species still a cost-effective strategy. Use of mixed mtDNA and nuclear baits (as suggested by Falk et al. [60]) is certainly an exciting and promising alternative to simultaneous sequencing of mtDNA and UCEs, but may also have its downsides, such as the additional costs of design and synthesis of compatible baits at different concentrations. In addition, designing baits for fast-evolving mitogenomes may be challenging when sequencing distant species, which is a common requirement of phylogenetic studies. Finally, use of off-target sequences may facilitate locus design of non-model organisms, since they contain non-UCE microsatellite markers.