MHC-containing clone screening

To obtain the MHC region, a ~10-fold genome equivalent fosmid library containing 37 pools was constructed. Subsequently, referring to the incomplete genomic resources [23], 17 target sites were identified to isolate fosmid clones that contain MHC sequences (see the Methods section for detailed procedures). At each target site, a pair of specific screening primers (S1 Table) was designed to screen the pools and 4~10 positive pools were obtained. Also, based on the sequence of PCR production in the positive pool, a recombinant cassette was designed for each target site (S2 Table). Lastly, after electroporation, recombination reaction and ampicillin plate screening, 78 target clones were fished out. All 78 clones were end-sequenced with vector primers and identical ones were excluded. Finally, 41 clones were left to construct the MHC region (Fig 1A).

Download:

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

Fig 1. Overview of recombination-based CSA strategy.

(A) Fosmid library construction and target clones isolation. ~40 kb fragments of genomic DNA were ligated to the pCC2Fos vector. Then these loaded vectors were transduced into E. coli clones, which were divided into 37 pools. After screening the pools by PCR, recombinant cassettes were used to isolate target clones from positive pools. (B) Illumina library construction and sequencing. One small insert size library was constructed for each clone and a large insert size library was constructed with a mixture DNA of all clones. (C) Step by step assembly strategy of the target region. Each clone was sequenced and assembled separately to get the first class scaffold. First class scaffolds derived from the same recombinant site were assembled to second class scaffold by overlap relationship. Second class scaffolds from all sites were finally assembled to obtain the sequence of the target region.

https://doi.org/10.1371/journal.pone.0168744.g001

CSA

After expansion, fosmid DNA of each selected clone was isolated. One 500 bp insert size library was constructed for each clone and 2 large insert size (2 kb and 5 kb) libraries were constructed with the pool of 48 fosmid clones. Sequencing was performed on an Illumina Hiseq 2000 platform with the 100 bp paired-end sequencing strategy (Fig 1B). After filtering the vector contamination, ampR gene sequences on selection cassettes and low quality reads, 480 Mb of clean data were obtained and the average sequence coverage per clone was ~250-fold. First, each clone was assembled to a first class scaffold individually and the size of the assembled scaffolds ranged from 28 kb to 47 kb with a peak at 39 kb (S1 Fig and S3 Table). Second, clones derived from the same recombinant cassette were assembled to a second class scaffold according to overlap relationships (Fig 1C). Finally, the second class scaffolds of 17 target sites were used to construct super-scaffolds SC1, SC2 and SC3 based on overlap relationships, and 3 super-scaffolds totally spanned 488.21 kb (Fig 2A).

Download:

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

Fig 2. Assembly of the MHC region.

(A) Reference assembly. Each color represents one target site. In each target site, the diamond indicates the position of the recombination cassettes; the dashed color line indicates the scaffold of each clone; the solid color line indicates the second class scaffold of each target site. The black bold line in the middle indicates the 3 super-scaffolds of the MHC. Recombinant sites C3a and C3b came from the same recombination cassette. (B) Partial haplotype separation. Paired black lines in addition to the super-scaffolds indicate the separated regions; short bars on paired black lines indicate the SNP or Indel differences between 2 chromosomes; the dashed colored lines indicate the clones used for determining the haplotypes.

https://doi.org/10.1371/journal.pone.0168744.g002

By comparing the overlapped clones, we further identified haplotypes in the ~174 kb clone region. At the SC1 region, 4 regions that spanned 114 kb in total were separated by 76 SNPs and 10 Indels. At the SC2 region, 137 SNPs and 14 Indels were identified. Three clones (b16-15-2, b16-15 and a1-23) represented the first haplotype and other 4 clones (b16-28, b16-6, b16-28-2 and a3-14) represented another haplotype (Fig 2B).

Assessment of short-read assembly

To assess the local accuracy of CSA, we randomly sequenced a fosmid clone containing the BLB locus in the BF/BL region with the Sanger method (~20-fold coverage) and assembled it with phred-phrap-consed. Sanger and CSA assemblies were aligned using BLAT with a default parameter and the sequence identity reached 97.29% (Fig 3A), which was remarkably higher than that of WGS (50.96%; Fig 3B). Ten sequence differences between the Sanger and CSA assemblies were totally identified. Four of them were located in runs of Gs, and the rest were located in gap regions of the CSA or Sanger assembly. In summary, 6 and 4 potential misassemblies were validated in Sanger and CSA, respectively (Table 1). All information about these differences is detailed in S2 Fig.

Download:

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

Fig 3. Assessment of CSA assembly.

(A) Alignment between CSA assembly and Sanger assembly. (B) Alignment between comparisons of the WGS assemblies with the Sanger assembly. The orange blocks indicate the consistent sequences between the 2 strategies. Single-base depth was calculated by mapping the short-read to the Sanger and CSA assemblies, respectively.

https://doi.org/10.1371/journal.pone.0168744.g003

Download:

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

Table 1. Assembly differences between Sanger and CSA.

https://doi.org/10.1371/journal.pone.0168744.t001

To eliminate incorrect assembly caused by potential chimeric clones and assess the overall accuracy, we used the mate pair reads of large insert-size libraries (2 kb, 5 kb, 10 kb and 20 kb) in the crested ibis WGS project [23] and mapped them to our MHC sequences. The pair-end relationship is presented in S3 Fig, which is well mapped and supports the validity of our assemblies. We also tried to determine the organization of 3 super-scaffolds, by considering all possible connections and looking for supporting evidences from WGS pair end reads. Interestingly, 16 mate pair reads (1 pair with 10 kb and 15 pairs with 20 kb insert size) showed that SC2 and SC3 were tightly linked within BF/BL region (S3 Fig). Long PCR was also used but failed to obtain informative products to fill the gap between SC2 and SC3.

Annotation of crested ibis MHC

Genes, transposable elements (TEs), short tandem repeats (STRs) and tRNAs of the MHC region were annotated. Combining the homologous and de novo gene prediction approach, we identified 50 genes in the MHC region, corresponding to one gene per every 9.96 kb (Fig 4A). SC1 contains Blec genes, TRIM family, BTNs and BGs, SC2 ranges from COL11A2, class II genes to TAPs and LTB4R1, and SC3 represents TNXB, several class I genes and ADPRHs. Based on the nomenclature for chicken MHC, SC1 was classified into the extended BF/BL region [24], SC2 and SC3 were classified into the BF/BL region [25]. By mapping the RNA-Seq data [21] to MHC sequences, 29 genes were verified to be expressed in blood (S4 Table). In addition, 84 long interspersed nuclear elements (LINEs, 76 CR1, 3 Jocker, 2 L2, 2 R2 and 1 RTE), 32 long terminal repeats (LTRs, 19 ERVL, 6 ERV1, 4 ERVK, 2 Gypsy and 1 Pao; Fig 4B and S5 Table), 31 STRs (Fig 4C) and 30 tRNA loci (Fig 4D and S6 Table) were identified. In addition, using 500 bp non-overlapping sliding windows, the crested ibis MHC showed a high average GC content of 57.1% (Fig 4E).

Download:

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

Fig 4. Gene organization and sequence features of the crested ibis MHC.

(A) MHC gene map. Genes belong to the same family are indicated with the same color. Upper boxes indicate genes on a positive strand, and lower boxes indicate the opposite. The red bar indicates the MHC region that is first described in the present study. (B) Position of repetitive elements, LINEs and LTRs are represented by black bars. (C) Location of STRs. STRs with 2–5 bp repeat units are shown. (D) Location of tRNA loci. (E) Plot of local GC content in continuous 500 bp windows. The red line indicates the average GC content.

https://doi.org/10.1371/journal.pone.0168744.g004

Comparative analyses of the avian MHC

We compared the crested ibis MHC with reported chicken MHC-B and found that both gene family members and overall organization were similar between the 2 species. Particularly, aligning with PipMaker [26], 2 compact blocks, the TRIM cluster and BLB-BRD-DMA-DMB-BF-TAP1-TAP2 cluster, also showed high sequence identity between the 2 birds (Fig 5A). However, several genes, such as COLL11A2 and ADPRHs, were found in the crested ibis but absent in the reported chicken MHC-B sequence. Additionally, more copies of Blec and BF were found in the crested ibis than in the reported chicken MHC-B (Fig 5B). We also compared the TEs of MHCs between chicken and crested ibis. With the same annotation pipeline, 33 TEs (22 LINEs and 11 LTRs) were identified in the reported 242 kb chicken MHC-B region [15]. The crested ibis MHC showed a frequency of 0.23 TEs per kb, which was higher than that of the reported chicken MHC-B (0.13 TEs per kb) (S5 Table).

Download:

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

Fig 5. Comparative analyses of crested ibis and reported chicken MHC-B.

(A) Dot matrix analysis of the crested ibis vs. the reported chicken MHC-B sequences. The red dot indicates that the sequences are homologous between 2 species at this position. Rectangles with a solid line indicate regions that demonstrate a high degree of synteny between the 2 birds. (B) Gene organization comparison. Genes in red indicate absence in reported chicken MHC-B sequence.

https://doi.org/10.1371/journal.pone.0168744.g005

Ethics statement

With the permission of the authorities, DNA samples were collected from peripheral blood of a female crested ibis in Yangxian County Reserve, Shaanxi Province, China. Blood sample collection was performed by professionals in the reserve. Ethical approval was granted by the Institutional Review Board on Bioethics and Biosafety in BGI, China (FT14035).

Target sites selection

First, using tblastn, 41 protein sequences of chicken MHC genes (GenBank: AB268588.1) were aligned to the crested ibis genome (GeneBank: GCF_000708225.1). Homologous exon segments of 28 chicken MHC genes were found to be mainly distributed on 4 scaffolds of the crested ibis genome (S5 Fig). Second, one target site (500–1,000 bp in size) was chosen at every 30 kb region on the 4 scaffolds in the following priority order: homologous exon > intron > intergenic regions. The sequence uniqueness of the target site was tested by blastn with the database of the crested ibis genome. Totally, 17 target sites were selected and their locations are shown in S5 Fig and S1 Table.

Fosmid library and PCR-based positive pools screening

A library of genomic DNA was constructed with the CopyControl HTP Fosmid Library Construction Kit following the manufacturer’s protocol (Epicentrue). Before recombination, the entire library was pre-screened to find the positive pools for each target site. The screening primers were designed on these fragments with the Primer3 tool (http://frodo.wi.mit.edu/primer3/; S1 Table). All primers were synthesized at Invitrogen. Primer specificity and annealing temperatures were tested on crested ibis genomic DNA. For the PCR template, 50 ng of DNA that were extracted from each pool was used. PCR amplification was performed using ABI9700 with the following parameters: 94°C for 5 min; 35 cycles of (94°C for 30 sec; annealing for 30 sec; 72°C for 30 sec); with a final extension at 72°C for 5 min. To further confirm the positive pools, the PCR products were also sequenced with the Sanger method.

Construction of recombinant cassettes

To reduce the cost of the recombination step, ampicillin was used to replace blasticidin (Bsd) in the study [47]. The cassettes were produced by PCR using the ampR gene and a pair of 75 bp primers. The 75 bp primer consisted of the 50 bp homology arm of the targeted region and the 25 bp ampR amplification primer (S2 Table). The homology arm was chosen based on the following criteria: 1) G+C < 35 nt; 2) G+C difference between the F arm and R arm < 2 nt, to be in close proximity to the screening primer pair. An in-house PERL script was developed to perform the filter process. All forward primers were phosphorylated at the 5’ end to generate PO (O strands for hydroxyl), and the reverse oligos were not modified. Using the PMD-18T vector as the template of the ampR gene (TaKara), PCR amplification with Phusion DNA Polymerase (Thermo Fisher) was performed on ABI9700 with the following parameters: 98°C for 30 sec; 35 cycles at (98°C for 30 sec; 55°C for 30 sec; 72°C for 30 sec); and a final extension at 72°C for 5 min. The cassettes were purified from the PCR reaction using 1.5% agarose gel electrophoresis followed by use of a Qiagen Gel Extraction Kit.

Recombinant procedure

Nedelkova’s vectors were used, and the recombinant procedure was performed as recommended by studies [22] with the following modifications. First, 25 μl aliquots from the PCR positive pools were grown in 1 ml of LB supplemented with tetracycline (tet, 5 μg/ml) and chloramphenicol (cm, 10 μg/ml) overnight at 30°C. A 30 μl overnight culture was diluted to 1.4 ml LB with tet (5 μg/ml) and cm (10 μg/ml) at 30°C for 2 h, followed by the addition of arabinose (Sigma A3256) and rhamnose (Sigma R3875) to 15% and growth at 37°C for 45 min. Second, the cells were centrifuged, washed twice with 1 ml ice-cold 10% glycerol and resuspended in 30 l ice-cold 10% glycerol. About 300–600 ng recombinant DNA was added to 30 μl of competent cells. For each electroporation, a pre-chilled 1 mm electroporation cuvette was used at 1350 V, with a 10 μF (Eppendorf Electroporator 2510). After electroporation, the cells were resuspended in 1 ml LB and incubated at 37°C for 1 h, and then plated on LB agar supplemented with 100 μg/ml ampicillin. The plates were incubated at 37°C for 18 h to 24 h. Third, 4–10 clones per recombinant cassette were incubated in 1ml LB supplemented with 100 μg/ml ampicillin and grown overnight at 37°C. PCR validation with screening primers was used to confirm the recombination success.

Library construction and sequencing

The fosmid DNA from 41 clones was isolated with a Qiagen plasmid plus kit (Qiagen, Hilden, Germany) and sheared using a CavorisLE220 device (Covaris, Inc. Massachusetts, MA, USA). DNA fragments that were 500 bp, 2 kb and 5 kb in size were recovered. End-repair, A-tailing and paired-end adapter ligations were performed following the protocol in the kit NEBNext^® DNA Library Prep Master Mix Set for Illumina^® from New England Biolabs. All libraries were quantified by real-time PCR using a SYBR Fast Illumina Library Quantification Kit (KapaBiosystems). Next, every 16 libraries were pooled together with equal content. Paired-end sequencing (2x100 bp) was performed on an Illumina HiSeq 2000 following the manufacturer’s instructions. About 60,000 pair raw reads were generated for each library.

Data filtering

To facilitate an accurate assembly, we performed a series of checking and filtering measures. The raw reads were filtered as follows: 1) Reads were trimmed 2 bp at the 5’end and 3 bp at the 3’end; 2) Reads with adaptors and with a small size were filtered; 3) Reads with >40% base quality below Q7 were filtered; 4) Reads with >10% “N” base were filtered; 5) The duplicate reads generated by PCR amplification in the library construction process were filtered; and 6) Reads with >30 bp mapped to either bacterial sequences in the NCBI BCT database or fosmid vector sequence were filtered.

Short-read assembly and evaluation

Each fosmid clone was separately assembled with SOAPdenovo [48]. We optimized assembly performance by varying the values of Kmer (K = 64…120) to get a longer scaffold N50. First, the pair-end reads from the 500-bp library were employed to assemble the contigs. Second, reads from 2 kb and 5 kb large insert-size libraries were mapped to the assembled contigs with SOAPalign and the exclusively mapped reads were used to construct scaffolds. Also, the gaps in scaffolds were closed using Kgf and GapCloser. Third, clones derived from the same recombinant cassette were assembled into one scaffold using 2 steps: 1) taking the longest scaffold of each clone together and determining the mutual positional relationship of these scaffolds by aligning them using MAFFT with the parameter "—clustalout", 2) merging the scaffolds into one second class scaffold using an in-house script, and in the regions covered by more than 1 clone, the assembly with a smaller gap was kept. Finally, with the same strategy, the second class scaffolds of 17 target sites were assembled to super-scaffolds. To evaluate the quality of assemblies, we first aligned our CSA sequence to Sanger and other sequence versions by BLAT and then we compared the SNP or short indel difference manually. The resulting MHC sequences were deposited in GenBank (accession number: KR995141—KR995143).

Annotation

Homology and de novo prediction, as well as RNA-Seq data, were used to identify genes. First, the MHC region genes of chicken, zebra finch, mallard, peregrine falcon and pinnated grouse from NCBI were used to generate the corresponding protein sequences, which were then mapped to our crested ibis MHC sequence using tblastn with a cutoff E-value of 1e-5. All high-score segments were grouped into gene-like structures by genBlastA [49]. The intron-included homologous sequences with flanking 500 bp were aligned to the protein sequences using Genewise [50]. Second, genes were predicted using a de novo software Augustus [51] with model parameters trained from a set of high-quality homolog prediction proteins in the crested ibis genome. Third, with RNA-Seq data [23], the transcripts were clustered using TopHat [52] and Cufflinks [53], and then aligned to SwissProt/TrEMBL database [54] with a cutoff E-value of 1e-5. Finally, the identified genes were merged to form a comprehensive and non-redundant gene set following 3 criteria: 1) candidate genes clustered with a >100 bp genomic overlap; 2) one cluster per gene (precedence: homology-based model > RNA-seq model > de novo predicted model); and 3) if not met 2), the gene required a >30% aligning rate to a known protein in the SwissProt/TrEMBL database and at least 2 exons. TEs were identified using RepeatMasker with default parameters. tRNAs were identified using tRNAScan with default parameters. STRs were identified using TandomRepeatFinder. GC content was calculated in 500-bp sliding windows.