Animal ethics statement

Blood collection was done according to good practices recommended for identification of sires for paternity checking in France. This study was approved by the scientific committee of the Metaprogramme SELGEN of INRAE.

Sample information and genome sequencing

Twenty three Creole bulls representative of the INRA nucleus herd in Guadeloupe were selected for the purpose of this study. Ten out of them are sons of founder sires of the experimental flock while the remaining thirteen animals are unrelated sires (based on genealogical records and sampling sites) chosen from the local stock of Guadeloupe. Genomic DNA was extracted from whole-blood and semen samples collected between 1995 and 2015. Paired-end libraries with insert size of 500 bp were constructed for each individual and sequenced using the HiSeq 3000 platform (Illumina) in the Genome et Transcriptome (GeT) GénoToul platform (Toulouse, France), following the manufacturer’s protocol.

Sequence alignment and SV discovery

Quality control of raw sequence reads was performed using the fastQC software v.0.11.7 [26]. Trimmomatic-0.36 [27] was used to remove Illumina adapter sequences, low-quality bases and artefact sequences. Filtered sequences were then mapped against the bovine reference genome (ARS-UCD1.2) using the Burrows-Wheeler Alignment tool (bwa mem v.0.7.17) [28] with default parameters. The resulting SAM files were then converted to BAM format, sorted, and indexed using SAMtools [29]. PCR duplicates were removed using the MarkDuplicates tool from Picard version 1.88 (http://broadinstitute.github.io/picard). Only properly paired reads with a mapping quality of at least 30 were kept. Local realignment was performed using two GATK (Genome Analysis Toolkit) version -3.8-1-0-gf15c1c3ef modules, RealignerTargetCreator and IndelRealigner. We used three SV callers, Pindel (v.0.2.5), LUMPY (v.0.2.13) and DELLY (v.0.7.8). We developed a custom pipeline combining the detection results of the three aforementioned SV-finding algorithms. We retained only SV identified by at least two callers and having a length between 50 bp and 5 Mb because structural variants identification tools relying solely on a single principle were shown to generate many false positive calls [30,31]. According to their frequency within the GUA sample, we have defined four SV categories: singletons (detected in one individual), low frequency (detected in less than 25% of individuals), common (detected in more than 25% and less than 80% of individuals) and highly frequent (detected in more than 80% of individuals). The R package karyoploteR [32] was used to display SV distribution across 30 chromosomes (29 autosomes and the X chromosome). To identify novel structural variations that had not been discovered so far, the genomic regions defining the SV detected in the present study were intersected with known structural variations reported in cattle from Database of Genomic Variants archive (DGVa) using the function ‘intersect’ from BEDtools [33]. For this purpose, we defined non-overlapping SV regions for the three types of SV and we compared these regions to the 9277 SV regions reported in the DGVa database.

Functional impact of structural variants

Ensembl Variant Effect Predictor (http://www.ensembl.org/info/docs/tools/vep/index.html)(VEP) was used to provide a prediction for each transcript with which the common and highly frequent SV overlap (those whose MAF > 0.25). VEP was used to determine the location of the SV (e.g. intronic, intergenic, upstream of a transcript, in coding sequences, in regulatory regions) as well as the predicted effect of SV on the protein (e.g. stop lost, frameshift). VEP also provides an impact rating (high, moderate, low, and modifier) indicating the severity of the consequences.

Functional enrichment analysis was performed using the online tool DAVID version 6.8 (Database for Annotation, Visualization and Integrated Discovery, https://david.ncifcrf.gov/). DAVID uses thousands of annotation terms in several annotation categories, such as Gene Ontology (GO), Biological Process, GO Molecular Function and InterPro Domains. An adjusted Benjamini-corrected p-value of 0.05 was used as the criterion for statistical significance of over-enrichment of genes in one of the categories.

Sequencing, SV discovery, and validation

A total of 7,543,644,154 clean reads were generated after sequencing the complete genome of the 23 samples, representing an average depth of ~16.35 fold (min depth = 9.3; max depth = 23.77). The average alignment rate against the Bos Taurus reference genome (ARS-UCD1.2), for the different individuals varied from 94.93% to 96.51% with an average of 95,7% (S1 Table). Three classes of SV: deletions (DEL), duplications (DUP) and inversions (INV) with a size ranging between 50 bp and 5 Mb were detected. Among the three callers, Pindel exhibited the highest detection sensitivity for deletions and duplications while DELLY identified much more inversions. LUMPY, for its part, called a lower number of SVs for all the three types of structural variants (Table 1). In total, 69% and 87% of the deletions (those at a minimum overlap identity of 90%) identified by Pindel and LUMPY, respectively were also detected by DELLY but this latter detected only 36% and 49% of the deletions, identified by Pindel and LUMPY, respectively. A high percentage, 94%, of the duplications identified by LUMPY were also identified by DELLY, but only 42% of those identified by DELLY were also identified by LUMPY. Also, we found that 59% of the duplications detected by DELLY were identified by Pindel and that 69% of those identified by Pindel were revealed by DELLY. In total, 92% of the inversions detected by Pindel were also identified by DELLY but only 53% of the inversions identified by DELLY were detected by Pindel.

Download:

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

Table 1. Performance of the three SV callers in detecting different SV types within the 23 Creole cattle individuals.

https://doi.org/10.1371/journal.pone.0309411.t001

In total, we identified 32,821 ascertained SV defining 15,258 regions with at least 2 software from the 23 GUA animals. Although combining multiple callers is recommended for a higher detection accuracy of structural variants [34], it should be outlined that SV detection in the present study is likely to be altered by several factors that are related to the sequencing technology used in the present study (short read sequencing) and to the threefold ancestry of the Creole cattle from Guadeloupe. For instance, insertions longer than short reads are expected to be easily missed because they cannot align correctly with the reference genome [35]. Likewise, the use of linear reference genomes causes reference allele bias that affects genetic variation detection tools [36]. Hence, we expect that the detection of SVs that are from African taurine and indicine origin in the genome of GUA, woud be affected when aligning the mapped reads to the Hereford reference genome assembly ARS-UCD1.2. Clearly, using a personalized reference genome of GUA should improve reliability of structural variation calls.

SV distribution and statistics

In accordance with previous studies [37–39], we observe an abundance of deletions (76% of the total SV detected in the present study) compared to duplications and inversions. On the other hand, deletions have a similar range of total length to duplications (Table 2) which is due to the fact that most of the identified deletions (81.56%) were <5 kb in length, whereas more than half (~ 55%) of duplications and inversions are of large size (10 Kb <size<5 Mb) (Fig 2A). This agrees with previous findings reporting a possible correlation between SV type and size.

Download:

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

Fig 2. Statistics and individual distribution of structural variations (SVs) in the Creole cattle from Guadeloupe sample.

(a) distribution of length for each SV type, (b) frequency of SVs by type. Four categories are defined: Singletons (detected in one individual), low frequency (detected in less than 25% of individuals), common (detected in more than 25% and less than 80% of individuals) and highly frequent (detected in more than 80% of individuals), (c) SV distribution among the 23 sequenced animals. The upper part shows SV < 1 Kb, while the bottom part shows SV lengths > 1 Kb.

https://doi.org/10.1371/journal.pone.0309411.g002

Download:

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

Table 2. Statistics of the SV number detected by at least two callers for the 23 GUA animals.

https://doi.org/10.1371/journal.pone.0309411.t002

Approximately half (~ 54%) of all detected SV have low frequency (detected in less than 6 individuals; Fig 2B) and covered 1.214 Gb of the chromosomes of the ARS-UCD1.2 assembly. About a quarter of these were detected in a single individual. This is consistent with previous studies indicating that high SV diversity exists among different cattle individuals [40,41]. Accordingly, we also observe an important within-individual variation of high-confidence SV. Their number ranged between 2092 and 4002 (average = 3051 ± 455) for deletions < 1 Kb, 1793 and 3009 (2362 ± 308) for deletions > 1 Kb, between 84 and 187 (average = 131 ± 28) for duplications < 1 Kb, 282 and 736 (average = 488 ± 114) for duplications > 1 Kb, and between 61 and 115 (average = 86 ± 16) for inversions < 1 Kb, 155 and 318 (average = 242 ± 40) for inversions > 1 Kb (Fig 2C). We observe, on average, 6360 SV per GUA individual which is in the same order of magnitude of African and Indian zebu but goes well beyond the values previously reported in European and African taurine breeds [38]. Part of the differences between GUA on one hand and taurine European and African populations on the other hand may be due to population structure. Indeed, the three-way admixture of the GUA genome [24] of which more than one third has an indicine origin is likely to be behind the high number of observed SV. Several studies have reported a higher SV in indicine than in taurine breeds which is consistent with the known breed divergence and history [42,43]. The origin of the indicine ancestry in GUA population appears controversial. Some historical evidences [23] relate introductions of West African cattle in Guadeloupe between the 16th and 18th century, and it is possible that African zebu as well as African taurine have been introduced. The alternate possibility of a recent introduction of Indian zebu in Guadeloupe at the end of the 19th century (as in the other regions of America) has not been documented [24]

A small portion (7.6%) of the detected SV is present in more than 80% of the GUA individuals (Fig 2B). Chromosomal distribution of common and highly frequent SV regions and having a size > 1 Kb was not uniform and varied according to the SV type. Chromosome 25 has the lowest number of deletions and inversions (47 and 6 deletions and inversions, respectively) and chromosome 26 has the lowest number of duplications (7). By contrast, BTA5, BTA18 and BTA3 showed the highest number of deletions, duplications and inversions, respectively (S2 Table).

Overall, common and highly frequent SV (MAF > 0.25) detected in the present study covered a total length of ~ 446 Mb which corresponds to almost 17% of the ARS-UCD1.2 assembly. Analysis of the distribution of SV regions across chromosomes showed substantial variation depending among others on SV type (Fig 3). Chromosome-wide SV coverage along each chromosome varies from 6.36% on chromosome 25 to 39.4% on chromosome 28 (S1 Fig) and S3 Table) and was not correlated to the number of SV regions (S3 Table). BTX is the most densely covered with SV in terms of Megabases (32 Mb) with inversions representing 70% of its whole SV length (S3 Table). Similar finding was reported by [44]. They explained this by difficulties when mapping SV on the X chromosome especially in males where effective coverage is halved. Another possible explanation for the observed higher number of SV observed on the X chromosome in the 23 Creole cattle bulls is likely related to the high male specific contribution of the X-chromosome to individual global recombination rate, previously reported in several cattle breeds [45]. It is therefore reasonable to assume that chromosomes with higher recombination rates are more likely to show more SVs since these are generated by different recombination mechanisms.

Download:

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

Fig 3. Chromosomal distribution of nonredundant SV regions with MAF > 0.25 and size > 1 Kb.

https://doi.org/10.1371/journal.pone.0309411.g003

We identified 6065, 464 and 343 new deletion, duplication and inversion regions, respectively that have not been previously reported in the DGVa database. Together, the three types SV defined 6638 nonredundant regions that have not been previously reported in DGVa (S4 Table). Focusing on SV > 1 Kb, 1967, 238 and 221 new deletion, duplication and inversion regions, were detected respectively (S2 Fig). Among these, three deletions, located on chromosomes 12, 16 and 26 and three duplications, located on chromosomes 7, 28 and X, are highly frequent in the GUA sample (S5 Table).

Annotation of SV

Structural variants that occur in genes can alter gene expression either by changing gene dosage or interrupting coding sequences, or disturbing long-range gene regulation which could broadly influence phenotypes. To better predict the downstream effect of the detected SV on protein function, we used the Ensembl VEP tool [46]. To get potential important insights into population-level effect of SV on genes with adaptive functions in cattle breeds raised under tropical environment, we merged the common and highly frequent SV into a set of nonredundant 9734 SV regions among which 55% were intergenic (Fig 4). SV regions occurring in genic regions overlapped with 7793 genes and 12,922 transcripts. According to VEP, almost all of the detected SV were assigned to the “modifier” impact category which is not surprising since most of these SV were located within introns (Fig 4). Only three deletions, located on BTA8 (at positions: 9,754,093–9,839,766), BTA10 (at position:101,090,067–101,094,332 bp) and BTA15 (at position: 45,917,286–45,918,286 bp), were classified as having a high (disruptive) impact on the protein function. Three genes are affected by these SV: HMBOX1, FOXN3 and a novel cattle gene: ENSBTAG00000027525. The first two genes are transcription factors. HMBOX1 is a transcriptional repressor that negatively regulates IFN-γ in natural killer cells [47]. While IFN-γ production is momentarily abrogated following intense exercise which provides a window for invasion of pathogens [48], the deletion in HMBOX1 might be regarded as a compensation mechanism that substitutes the downregulation of IFN-γ during prolonged exhausting exercise. Creole cattle from Guadeloupe are mainly used for a draught of sugar cane carts or ploughing [49] which would weaken their immune system for large parts of the day. HMBOX1 might have also a protective role against splenomegaly and anaemia, which are prominent features of trypanosomiasis in cattle [50]. Unlike HMBOX1, the adaptive interpretation of FOXN3 deletion (observed in nearly 83% of our GUA sample), is less obvious because this gene is involved in a variety of physiological processes ranging from cell proliferation, apoptosis and pathogenesis in human cancer [51] to craniofacial development and fasting blood glucose and glucagon, in other non-human species [52–54]. In Angus cattle, it has been associated with chest width and skeletal development [55]. Further investigations of the functional effects of partial deletions FOXN3 on adaptive features of cattle are needed.

Download:

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

Fig 4. Variant effect predictor output from the genomic regions defined by nonredundant SV with size > 1 Kb.

https://doi.org/10.1371/journal.pone.0309411.g004

We also checked if there is any effect of SV frequency on their functional consequences by comparing common and highly frequent SV on one hand and low frequency SV on the other hand (these are defined as SV detected in less than four individuals). We found that SV of the second category tend to be in coding sequences more often than the first category (738 Vs 487, χ-squared = 51.712, p-value = 6.425 x 10⁻¹³).

We then generated a gene list including 579 genes that overlap with the genomic regions specific to GUA population (those that were not reported in DGVa) and encompassing common and highly frequent deletions and duplications. These genes were annotated according to DAVID Bioinformatics resources (https://david.ncifcrf.gov/). We found significant enrichment of the functional category ‘UP_SEQ_FEATURE: DOMAIN:C2’ (Benjamini-corrected P-value = 0.034, n = 13). Calcium transport was the most enriched biological process (BP) term (Benjamini-corrected P-value = 0.0011, n = 10) (S6 Table). C2 domains are widespread and conserved motifs that often serve as Ca2+-binding modules. Single and multiple copies of C2 domains have been identified in a growing number of eukaryotic signalling proteins that interact with cellular membranes and mediate a broad array of critical intracellular processes, including membrane trafficking, the generation of lipid-second messengers, activation of GTPases, and the control of protein phosphorylation [56]. In total, thirteen C2 domain genes located on 11 different chromosomes were impacted by SV. Among these, eight genes: CPNE4, PKN2, PIK3C2G, UNC13C, DYSF, PLCB1, SYT9, SMURF2 are affected by frequent deletions (affecting more than 65% of our GUA sample). In ruminants, several of these genes have previously been associated with growth traits, such as CPNE4 or DYSF [57,58] or with fat metabolism, such as PLCB1 [59]. Interestingly, the differentiation of some of these genes between taurine and indicine cattle have also been discussed (CPNE4, DYSF, PIK3C2G) [57,58,60]. Three of these genes (DYSF, PLCB1, SMURF2) are also known to regulate many aspects of the immune system response to pathogens. Similar to HMBOX1, SMURF2 and DYSF negatively regulate some aspects of immune response. SMURF2 gene is an essential negative regulator of TGF-β signalling and plays a role in the vascular inflammatory response in the presence of hypoxia in endothelial cells [61] while DYSF negatively regulates phagocytosis (GO:0050765). Finally, PKN2 appears essential for embryogenesis in mouse [62], and its loss causes severe cardiovascular and morphogenetic abnormalities. More research would be necessary to identify the impact of these deletions in cattle, in particular in GUA breeds.

We took a closer look at the three deletions and the three duplications that were not previously reported in the DGVa database and which are highly frequent in our GUA sample (S5 Table). These six CNV regions encompass 135 genes, most of them (67 genes) are located on chromosome X. We have focused primarily on genes for which there is evidence for local adaptation in GUA cattle. In this regard, we find that duplications on BTA28 encompass two genes, ras homolog family member U (RHOU) and actin alpha 1, skeletal muscle (ACTA1) that play a key role in muscle development and contraction [63–65]. RHOU also regulates cell-adhesion molecules during cardiac morphogenesis [66]. Creole cattle from Guadeloupe have often been used as a major labour force in sugarcane fields and are known for their draft endurance. Duplications in these genes might be a hint that allows physiological adaptation of GUA population to such strenuous activities. In humans, it has been shown that SV occurring in genes expressed in muscle or heart contribute to the variation of endurance capacity [67]. CNV observed in RHOU were also reported as involved in mammary development in Dairy Gir cattle [68]. Deletions on BTA26, overlaps with INPP5A, that have been associated with body temperature regulation in Nellore cattle, through the changes in the nervous system and regulation of inflammatory processes [69].

In other respects, we observed that almost all duplicated genes on BTA7 included olfactory receptors which are well known for their extremely frequent gene duplications and losses in vertebrates [70]. Population specific SV affecting olfactory receptor genes were also reported in cattle [71]. Likewise, we find that GUA-specific structural variants were also found to overlap with genes influencing milk as well as meat traits in cattle. For instance, duplications on BTA28 overlap with 15 genes, some of which (RAB4A, CCSAP, ENSBTAG00000048654, or URB2) have previously been associated with fat or protein metabolism involved in beef or milk production traits [72,73]. The same finding holds for GPC6, TGDS, GPR180 and SOX21 genes which overlap with deletions on BTA12 [72,74].

The 135 genes were analysed for GO term enrichment. Gene Ontology (GO) analysis showed that tRNA threonylcarbamoyladenosine metabolic process is the most enriched biological process (BP) term (GO:0070525, n = 4, Benjamini-corrected p-value = 7.03× 10⁻⁴). tRNAs are central players in translation, functioning as adapter molecules between the informational level of nucleic acids and the functional level of proteins. Modifications on tRNA structure modulate rigidity and flexibility of the transcripts and confer thermal adaptation in thermophilic as well as psychrophilic bacteria [75]. Although tRNA modifications play multi-faceted roles in several cellular processes, they remain largely unexplored in mammals. In humans, analysis of tRNA indicates that many tRNA modifications are incomplete under physiological conditions and that variation in the levels of tRNA modification should enable cellular adaptation to environmental changes [76].

Overlap between structural variants and previously identified signatures of selection in the GUA genome

We investigated the potential overlap between the highly frequent structural variants, identified in the present study and six genomic regions that have been recently shown to be under selection in the GUA population [25]. These six regions are located on BTA2 (at position: 120–120.5 Mb), BTA4 (at position: 113–113.5 Mb), BTA5 (at position: 47–47.5 Mb), BTA6 (at position: 69–69.5 Mb), BTA12 (at position: 29–30 Mb) and BTA13 (at position: 63.5–64 Mb). Within each region, we identified one or two candidate genes, based on their involvement in adaptive traits: EIF4E2 (CR on BTA2), GIMAP genes (CR on BTA4), (GRIP1 and HELB on BTA5), LNX1 and OCIAD1 (CR on BTA6), RXFP2 (CR on BTA12) and ASIP (CR on BTA13).

Overall, the six candidate regions included 17 structural variants ranging between 50 bp and 253.7 Kb in length (14 deletions and 3 inversions). GIMAP genes (GIMAP4 –GIMAP7) are located within an inversion of 253 Kb which overlapped with a previously reported CNV (nsv616158, [43]) in cattle (S7 Table). They are related to the primary immunodeficiency pathway and were also shown to play a major role in feed utilization and the metabolism of lipids, sugars, and proteins in Jersey cattle, and present a signature of selection in Asturiana de los Valles Spanish breed [77,78]. Among the SV specific to GUA individuals, we identified two highly frequent deletions of 50 bp and 1056 bp in length affecting the intronic regions of HELB and GRIP1, respectively. HELB is involved in the response to DNA damage including exposure to ultra-violet light and specific mutations carried out by indicine cattle and admixed populations have already been described. It appears to be associated with reproductive traits and yearling weight in tropical cattle, and could contribute to the adaptation of tropical cattle to their harsh environment [79]. Likewise, we found that a 5.5 Kb-length deletion (at position:120,098,203–120,103,733 bp) on chromosome 2 that affects a coding sequence of ALPI gene and have a high predicted functional impact on the protein (S7 Table). This gene is involved in diverse biological processes, including modulating host-bacterial interactions [80], mucosal defence and maintaining gut homeostasis [81]. Functional studies would need to be conducted to investigate whether the identified deletion impacts the adaptive immune system of GUA cattle which in turn shapes the diversity and the balance of gut microbiota required for immune homeostasis.

On chromosome 6, a deletion affecting 18 out of 23 GUA individuals is present in LNX1 gene, a major regulator of the presynaptic glycine transporter GlyT2 [82]. The 334-bp deletion in our sample overlaps an intronic region and have a “modifier” impact, which supposes that it has a mild effect on phenotypes in GUA cattle. Importantly, mutations in the exon 4 of GlyT2 were shown to cause recessive congenital muscular dystonia type 2 (CMD2) in Belgian Blue cattle [83]. An association between LNX1 and birth weight has been reported in the Colombian creole breed Blanco Orejinegro [84]. In humans, mutations in GlyT2 gene are responsible hyperekplexia, a condition in which affected individuals have increased muscle tone (hypertonia) and an exaggerated startle reaction to unexpected stimuli (tactile or acoustic) [85]. The impact of this deletion in GUA cattle would be interesting to evaluate, whether it affect embryo development, muscle tone or behaviour.

The closest SV to RXFP2 and ASIP (candidate genes on chromosomes 12 and 13, respectively) are two deletions located 11.5 and 14.5 Kb upstream RXFP2 and ASIP, respectively. Importantly, we found another deletion on chromosome 13 held by ~ 50% of GUA individuals, affecting EIF2S2 gene which is differentially expressed in the skin epidermis of Humans, thus playing potential role in pigmentation phenotypes [86]. Some other genes affected by SV in GUA are also associated with coat colour determinism in cattle such as DCT, OCAID1 (close to KIT gene) and ASIP [87–89] which may explain coat color variation observed in this breed.