Development and Genome Sequencing of a Laboratory-Inbred Miniature Pig Facilitates Study of Human Diabetic Disease

Summary

Pig has been proved to be a valuable large animal model used for research on diabetic disease. However, their translational value is limited given their distinct anatomy and physiology. For the last 30 years, we have been developing a laboratory Asian miniature pig inbred line (Bama miniature pig [BM]) from the primitive Bama xiang pig via long-term selective inbreeding. Here, we assembled a BM reference genome at full chromosome-scale resolution with a total length of 2.49 Gb. Comparative and evolutionary genomic analyses identified numerous variations between the BM and commercial pig (Duroc), particularly those in the genetic loci associated with the features advantageous to diabetes studies. Resequencing analyses revealed many differentiated gene loci associated with inbreeding and other selective forces. These together with transcriptome analyses of diabetic pig models provide a comprehensive genetic basis for resistance to diabetogenic environment, especially related to energy metabolism.

Graphical Abstract

Highlights

• •Bama miniature pig (BM) is one of the pig lines with the highest inbreeding coefficient
• •This atlas is a report on the chromosome-level genome assembly of miniature pig
• •Genomic analyses revealed genetic basis underlying BM's advantages to study diabetes
• •Some lncRNAs and mRNAs may be linked to resistance to diabetogenic environment

Biological Sciences; Gene Ontology; Genomic Analysis; Genomics; Omics; Sequence Analysis; Transcriptomics

Introduction

Pig (Sus scrofa) has served not only as one of the most economically important livestock but also as an important model organism used in many areas of medical research, including obesity, cardiovascular disease, endocrinology, diabetes, alcoholism, nephropathy, and organ transplantation, owing to parallels with humans in anatomy and physiology (Andersson, 2016, Ibrahim et al., 2006, Rocha and Plastow, 2006, Schook et al., 2005, Yan et al., 2018). There are over 730 distinct pig breeds worldwide, whose diverse phenotypes are shaped by the combined effects of local adaptation and artificial selection (Ai et al., 2015). However, the vast majority of pig breeds have been developed with a focus on economic benefits, rather than breeding an ideal laboratory animal, which directly resulted in almost non-existence of excellent inbred pig strains as model organisms used in biomedical research.

After several hundred years of intense artificial selection, current commercial pig breeds, represented by Duroc, have undergone drastic phenotypic changes and genetic adaptations that are economically important to the pig industry (e.g., reduction in feeding costs) and the consumer (e.g., higher production of lean meats) (Ai et al., 2015, Frantz et al., 2015, Rubin et al., 2012). In this context, a series of absolutely visible traits of current commercial pigs, such as large body size (adult individuals can reach 300–400 kg in weight), long life cycle, and weak inbreeding level (Table S1), have become obstacles in using pigs as biomedical animal models, especially in the studies of obesity and diabetes mellitus, because they result in high maintenance costs, specialized facility requirements, long experimental periods, and poor repeatability (Kleinert et al., 2018). Moreover, the different biomedical responses and performance of commercial pig breeds from those of wild boars, including severe resistance to “diabetogenic” (high-calorie and low-activity) environment (Gerstein and Waltman, 2006), also go against the construction of some pig models for human diseases. These defects mean the urgent need for a professionally experimental pig strain, which drove us to develop a laboratory Asian miniature pig inbred line— Bama miniature pig (BM) based on Bama xiang pig (BX), a primitive breed without artificial imprinting for commercial characters, whose many characteristics, such as small volume, early maturity, and long-term adaptation to inbreeding, are valuable in the construction of an ideal inbred laboratory pig line (Table S1), more than 30 years ago.

It is known that reference genome sequence is quite important to biomedical studies using pig model (Bains et al., 2016, Crawley et al., 1997). Although the genomes of some pig breeds had been published (Groenen et al., 2012, Li et al., 2017), the chromosome-level genomes with comprehensive annotation are still scarce resources to date. Moreover, most of the currently reported pig genomes are from commercial pig breeds, except, to our knowledge, only one highly fragmented draft genome sequence from an experimental inbreeding line, Wuzhishan miniature pig (Fang et al., 2012). Therefore, there is an inevitable quandary in most of the health studies involved in pig genome, that the Duroc reference genome (Sscrofa11.1, GenBank assembly accession: GCA_000003025.6) nearly becomes the only choice, no matter which kind of pig breed is selected. The presentation of BM high-quality reference genome can enrich the Sus scrofa genome database to effectively improve this dilemma, and likewise provide essential information needed to shed light on the genetic components of the BM phenotypes advantages to diabetic study, especially the relatively lower resistance to diabetic pressure, by comparative genomic analyses.

In this study, we have successfully inbred BMs to generation 19 (inbred line F19), which is, to our knowledge, the pig line with the highest inbreeding coefficient to date. Using combined technologies, we presented a chromosome-level genome sequence and the available annotation of highly inbred BM. Comparative analyses of BM and Duroc genomes revealed substantial genomic differences between them, as well as identified genetic basis underlying the BM's superior traits to study diabetic diseases. Resequencing analyses between BX and BM populations confirmed the leading inbreeding degree of BMs at the genome-wide level. Besides the positively selected genes (PSGs), selective sweep and transcriptome analyses also found some changes of energy metabolism systems, like phosphatidylinositol 3-kinase (PI3K)-Akt signaling pathway, related to diabetic resistance. This study provided not only an inbred miniature pig line to overcome these preexisting obstacles of using pig in diabetic researches but also a comprehensive molecular basis as reference to further optimize breeding of the experimental animal's phenotypes advantageous to diabetic diseases researches. Meanwhile, it systematically boosted the understanding of the mechanism of resistance to diabetogenic pressure from the genome level to the transcriptome level.

Results

Development of BM Strain and Detection of Resistance to Diabetogenic Environment

To provide a dedicated laboratory pig strain for biomedical research, we have been developing a laboratory Asian miniature pig inbred line, the BM (Figure S1), from the original subtropical BX population (2 males and 14 females selected), which is native to south China (Table S1), for more than 30 years (since 1987) (Figure 1).

Figure 1.

Establishment of Highly Inbred Laboratory BM from BX

Two male and fourteen female BX were introduced to breed a laboratory inbred line (BM) since 1987.

After 10 years (1987–1997) closed pure-bred breeding and directional selection, we had developed a closed colony (A strain) to generation 10. The F10 closed colony individuals' adult weight decreased about 10 kg (24-month-old F0: 52.78 ± 0.86 kg versus F10: 43.67 ± 0.77 kg). The ratio of individuals with uniform two-end-black coat color in closed colony increased from 70.44% (F0) to 94.73% (F10). Compared with F0 individuals (Table S1), the aggressive behaviors of F10 animals had disappeared almost completely. Subsequently, in 1997, establishment of inbred line (B strain) was initiated using four male and four female F10 closed colony (A strain) individuals (as inbred generation 0 [inbred line F0]), of which one couple has broken inbreeding bottleneck and has been bred for 19 generations (inbred line F19 maintaining all specific features of the F10 closed colony) so far.

Like other commonly used experimental animal inbred lines (Lilue et al., 2018), the BM has clear genetic background (without foreign gene flow influx), stable characters, and high homozygosity (inbreeding coefficient of inbred line F19: 0.9825) (Figure 1), which can ensure the reproducibility of conclusions of experiments based on this animal. Furthermore, it can be a valid substitute to overcome some shortcomings of the use of commercial pig breeds in biomedical studies, because of BM's observed phenotypic specificity. The BM was selected by inbreeding for a small adult body size (adult body weight: 40–50 kg) and short life cycle (<1.5 months for male sexual maturity), thereby lowering maintenance costs and reducing the duration of experiments (Table S1). In addition, breeding of the BM rigorously conforms to the standards of laboratory animal breeding husbandry, feeding a restricted food supply that provides ∼70% of energy ad libitum of miniature pig (approximately equals to digestible energy requirement for maintenance) (Table S1), to just accomplish basic energy needs for essential life activities since the start of inbreeding. In short, the laboratory-inbred BMs harbor different physiological characters from those of commercial pig breeds, which make them more useful as disease models, especially for diet (high-fat and high-carbohydrate)-induced diabetes. We exerted diabetogenic pressures on BMs (BM-induced group) and Durocs (Duroc-induced group) by high-fat and high-carbohydrate diet and limited activity space, and the resistance to “diabetogenic” environment was confirmed according to a series of parameters. The fact that 66.7% (10/15) of BMs and 13.3% (2/15) of Durocs have fasting blood glucose level >126 mg/dL (7 mmol/L), abnormal pathology, increased fasting insulin level, and decreased glucose disappearance rate at month 12 means that BMs' resistance to “diabetogenic” environment is much lower than that of Durocs (Figures 2 and S2; Tables S2 and S3).

Figure 2.

Detection of Resistance to Diabetogenic Environment

(A–D) Fifteen male BMs and fifteen male Durocs (6-month-old) were fed a high-fat and high-carbohydrate diet and had limited activity spaces for 12 months, and three male BMs and three male Durocs were selected to be fed with standard material as control groups (Table S2). Successful resistance of these pigs to diabetogenic environment was assessed by measuring changes in fasting blood glucose (FBG) at different times (A), intravenous glucose tolerance at month 12 (B), and fasting insulin (FINS) at different times (C) and by examining histopathological sections of the pancreases, kidney (Figure S2), liver, and skeletal muscle tissues after 12 months (D). (A) Changes in average FBG. Average FBG of BM-induced group rapidly increased after eighth month, in contrast to that of Duroc-induced group. (B) Intravenous glucose tolerance test (IVGTT) on individuals with FBG >126 mg/dL (7 mmol/L) conducted at month 12. The glucose clearance rate was reduced in BM-induced and Duroc-induced group individuals compared with the control group individuals. (C) Changes in average FINS. The FINS of individuals with FBG >7 mmol/L (10 BMs and 2 Durocs) were much higher than those of control group individuals (3 BMs and 3 Durocs). FINS levels are represented by box-and-whisker plots (with no box if n < 4). Boxes represent the interquartile range between the first and third quartiles and median (internal line), and whiskers denote the lowest and highest values, respectively. (D) Pathological sections of pig liver and skeletal muscle. In the liver, the liver cords exhibited a disordered arrangement, and many lipid droplets were observed in the abnormal liver tissue compared with the normal liver tissue in control group. Green and black arrows indicate liver cords and lipid droplets, respectively. In skeletal muscle, the abnormal tissues had fewer fibers, which were disordered, and significantly higher lipid content between fibers compared with normal tissues in control group. Green and black arrows indicate the muscle fibers and lipid droplets, respectively. Scale bars, 50 μm.

Sequencing, Assembly, and Annotation of the BM Reference Genome

To fill the gap of chromosome-level genome assembly of the miniature pig, we generated a reference genome assembly from a male BM. The BM has a diploid chromosome number of 38 (Figure S3) and an estimated genome size of 2.58 Gb (Figure S4; Table S4). To achieve high-quality genome assembly, we adopted a combination of sequencing methods including Illumina paired-end and mate-paired sequencing, 10× Genomics linked reads (>50 kb), Pacific Biosciences (PacBio) single-molecule real-time (SMRT) sequencing, Oxford Nanopore sequencing technology (ONT), and chromosome interaction mapping (Hi-C) sequencing. We developed a hybrid assembly pipeline to assemble this genome (Figure 3A) using a total of 932.04-Gb sequence data (equivalent to 361.25 genomic coverage). First, an initial draft genome was assembled by using 71.64-fold (184.83 Gb) 10× Genomics barcoded sequencing data and 152.87-fold (349.42 Gb) Illumina sequencing reads, with a scaffolds N50 size of 21.13 Mb (Tables S5 and S6). Second, PacBio long-reads (53 Gb; 20.54-fold), Oxford Nanopore sequences (26 Gb, 10.12-fold), and Hi-C sequence data (273.7 Gb; 106.09-fold) were used to upgrade the BM genome assembly and obtain a chromosome-scale genome with a contig N50 size of 1,010 kb, and total assembled length of 2.49 Gb, of which 97.49% was anchored to 20 chromosomes (18 autosomes and 2 sex chromosomes) ranging in length from 9,839,741 to 283,123,735 bp (Tables S6–S8). This new reference assembly has 5,723 gaps giving an estimated mean gap length of 2 kb (Figure 3B; Table 1). The high level of accuracy and completeness of BM genome assembly is demonstrated by the normal GC content (41.90% of genome), mapping of 97.69% of short sequencing reads, and Benchmarking Universal Single-Copy Orthologs (BUSCO)-based completeness assessment (Figure S5; Tables S9–S12). The BM reference genome has 43-fold higher contiguity than the previously published short-read genome assembly of Wuzhishan pig (contig N50: 23.5 kb) (Table 1) (Fang et al., 2012). In short, these combined technologies produced one of the most continuous porcine de novo assemblies to date, with chromosome-length scaffolds and the shortest gap lengths.

Figure 3.

Genome Assembly

(A) Workflow for genome construction.

(B) Ideograms of BM reference chromosome-scale pseudomolecules. The upper track shows positions of all gaps in the pseudomolecules; few of them are longer than 10 kb. More than half of the assembly consists of contigs >1 Mb, which are shown as black red bars in the lower track.

Table 1.

Comparison of the Quality between the BM and Published Inbred Miniature Pig Genome Assemblies

	BM Genome	Wuzhishan Pig Genome
Sequencing technology	10X Genomics, Illumina, PacBio, Nanopore and Hi-C	Illumina
Sequence coverage (X)	361.25	126
Assembly level	Chromosome	Scaffold
Genome size (Gb)	2.49	2.64
Contig N50 (bp)	1,010,657	23,535
Scaffold N50 (bp)	140,438,739	5,432,118
Total assembly gap length (bp)	11,642,258	–
Mean gap length (bp)	2034.2928	–

To aid genome annotation, we also sequenced the transcriptomes of 10 tissues (brain, liver, heart, spleen, lung, kidney, pancreas, stomach, skeletal muscle, and adipose) from the BM. A total of 21,334 protein-encoding genes were annotated using both de novo and homologous-based predictions (Figures S6, S7, and S9; Tables S13 and S14). Moreover, it was found that the BM genome is composed of 37.32% repetitive elements, fewer than that (40.55%) of Duroc genome, and encodes functionally important noncoding RNAs (Figures S8 and S9; Tables S15–S17).

Comparison of the BM genome with the human, and three common experimental animal (macaque, mouse, and dog), genomes unveiled three gene families, including ARF1 and IGHD, shared between the BM and human genomes but absent in macaque, mouse, and dog genomes (Figure S10). These genes may play roles in Alzheimer disease, pituitary dwarfism, and growth failure (from database “DisGeNET”). The presence of these genes in the BM potentially facilitates research on the above-mentioned diseases using this animal model. Moreover, BM has fewer unique genes compared with the Duroc (1,303 versus 1,531) (Figure S10), and the genes specific to BM were significantly enriched in the “steroid hormone biosynthesis” Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway (p = 0.00908), which is associated with sex hormone secretion, male testicles development, and rapid maturation of sperm.

Porcine endogenous retroviruses (PERVs) are ancient viral sequences integrated into the pig genome and transmitted vertically to the offspring, which has made them difficult to eliminate in the field of organ transplantation (Niu et al., 2017). BM has significantly fewer PERV gene copies (including the gag, pol, and env genes), with a total of 45 copies of potential virus-derived genes in the BM genome assembly, compared with 171 copies in the Duroc genome (Table S18). This reduction in PERV genes included complete loss of the PERV-A and PERV-C env genes, and the latter was further validated by general env genotyping (Figure S11). Fewer PERVs in BMs alleviate the requirement of knocking out PERV-C genes for pig-to-human xenotransplantation, especially for the organ transplantation of the tissues that diabetic diseases mainly target on, like liver, heart, kidney, and pancreas.

Structural Variants between the BM and Duroc Genomes

Although BM and Duroc share a high degree of chromosomal collinearity (Figure S12), we found that these two genomes still have numerous blocks of DNA sequence variations. To identify the large-block sequence variations between BM and Duroc, we conducted a comprehensive survey of structural variants (SVs), including genome-wide deletions, duplications, insertions, and inversions, in the BM by alignment of the BM genome with the Duroc reference genome assembly (Sscrofa 11.1). We focused on identifying these SVs >50 bp because of their severe effect on gene function. Our genome-wide alignment identified 59,373 SVs in the BM genome, most of which (98%) were located in intergenic and intronic regions (37,080 [62.45%] within intergenic regions and 21,092 [35.52%] within intronic regions) and few (622 or 0.01%) were located within exons (Table S19), indicating that few of the SVs affected the coding sequences of genes, with most located in noncoding regions.

During inbreeding, BM maintained the early sexual maturity of BX (BM/BX can produce mature sperm for insemination at the age of 45 days) circumventing the inherently long life cycle of the commercial pig (180–240 days for Duroc). We found that some of the SVs between BM and Duroc are located within the exonic regions of genes related to male sperm maturation (Figure 4; Table S20). The functions of these genes cover many aspects of sperm development, including formation of the oviductal sperm reservoir in the pig (AQN1) (Dostalova et al., 1994, Sanz et al., 1992); protective effect on boar sperm functionality (PSP1 and PSP2) (Garcia et al., 2006) (Figure 4A); epididymosomes and sperm plasma membrane development (ADAM7) (Oh et al., 2009) (Figure 4B); processes in meiosis, germ cell apoptosis, and male infertility (HSP70-2) (Dix et al., 1996, Dix et al., 1997); idiopathic male infertility (UBE2B) (Suryavathi et al., 2008); and regulation of meiotic pachytene progression during spermatogenesis (OVOL1) (Li et al., 2005) (Table S21).

Figure 4.

Genome Structural Variation Related to Sperm Maturity between BM and Duroc

(A) An ∼67-kb inversion on chromosome 14 (9486427–9553719) in the BM genome (annotated red lines) revealed by PacBio sequencing.

(B) A large-scale deletion (∼29kb) resulted in removal of the putative coding exons of ADAM7 from the BM genome.

(C) Microstructures of BM testes (30 days of age). The red arrows indicate spermiogenesis of spermatids. Scale bar, 2 μm.

(D) Section of BM epididymis (48 days of age). The red arrow indicates the appearance of mature sperm. Scale bar, 50 μm.

SVs identified influence genes involved in metabolic disorders. Three genes, AHNAK, ADGRF5/GPR116, and ATP10D, reported to regulate obesity (Ramdas et al., 2015), were affected by both deletion and duplication events within exonic regions (Table S21). Knocking out the AHNAK gene in mice results in protection from diet-induced obesity, ADGRF5/GPR116 affects insulin sensitivity via modulation of adipose function (Nie et al., 2012), and ATP10D is involved in endoplasmic reticulum-to-Golgi ceramide processing and regulation of obesity (Kengia et al., 2013). These SVs might be correlated with that BMs are more unbearable to diabetogenic pressure compared with Durocs (Figures 2 and S2; Tables S2 and S3).

Evolutionary Status and Small Body Size of BM

It is necessary to determine the medically applied scope of laboratory animals by dissecting the genetic relationship between them and humans. To detect the exact phylogenetic position of BM and Duroc, we constructed a highly resolved phylogenomic tree. BM and Duroc clustered within one clade as expected, but the branch length of Duroc is longer than that of BM (Figure 5A), suggesting that the commercial pigs have undergone rapid evolution, which gives rise to less genetic divergences between humans and BM than between humans and Duroc.

Figure 5.

Comparative and Evolutionary Genomic Analysis and Regulation in Terms of Difference of Resistance to Diabetogenic Environment between BM and Duroc

(A) Phylogenomic tree showing expansion (green, +) and contraction (red, −) of gene families in BM and Duroc, and other nine Eutherian species. MRCA, most recent common ancestor.

(B) Geographic origin of BMs and Durocs. The BMs originated from low latitudes but the Durocs from high latitudes.

(C) An unrooted neighbor-joining tree constructed using LILR family genes identified in BM and Duroc.

(D) Control from CNS to intracell to glucose balance. When mammals eat, the CNS responds to the hormonal and sensory signals produced by the food to regulate ingestive behaviors (upper part). The box in purple indicates the PSGs of BM, and the other box in light yellow indicates the PSGs of Duroc involved in this process. After food intake, the PI3K-AKT pathway plays an important role in the transport of glucose from blood to inside the cell to regulate glucose homeostasis (lower part). The light yellow box indicates the AKT1 and AKT2 PSGs of Duroc, and the Duroc-specific amino acid substitutions have resulted in variation in the protein structure of AKT2 (the dashed box) (Figure S18). Specifically, compared with the BM and human proteins, substitutions in amino acids 13 and 14 in AKT2 of Duroc have affected the protein's β-fold (shown in left bottom). In contrast to the absence of amino acids 15 and 16 in AKT2 of BM and human, the α-helix structure has disappeared from the structure of AKT2 of Duroc (shown in left bottom). Colors in the heatmap (the dashed box) represent the degree of sequence identity of the AKT proteins at the amino acid level between BMs/Durocs and humans (shown in right bottom).

Among the 11,368 single-copy orthologous genes, the number of orthologous genes, which are closer to counterparts of humans than those of mouse, is higher in BM than that in Duroc (8,547 in BM versus 8,240 in Duroc) (Figure S13). The specific genes, more similar to genes of humans than mice, of BM relative to Duroc were associated with a wide range of physiological processes. These genes were significantly (p < 0.01, t test) enriched in a major energy metabolic KEGG pathway, “insulin secretion” pathway, which is tightly intertwined with diabetes mellitus and other comorbidities (such as atherosclerosis). We further surveyed gene sets pertinent to eight common human diseases (obesity, Curtasu et al., 2019; type 2 diabetes mellitus [T2DM], Okitsu et al., 2004; nonalcoholic fatty liver disease, Yamada et al., 2017; atherosclerosis, Natarajan et al., 2002; Parkinson disease, Danielsen et al., 2000; Huntington disease, Yan et al., 2018; Alzheimer disease, Holm et al., 2016; and amyotrophic lateral sclerosis, Chieppa et al., 2014) suitably studied by pig model and found that these human-encoded genes are better conserved in the BM than in Duroc and mouse (Figure S14). Taken together, these genome-level analyses illuminated better similarities in physiological genetic background between BM and humans than between Duroc and humans, suggesting that the BM may be more appropriate for analyses of some common diabetic diseases.

Next, we focused on insights into gene family evolution, during which expansion or contraction of gene families have contributed to the phenotypic evolution of animal (Kim et al., 2016, Nowoshilow et al., 2018). To study whether expanded and contracted gene families are responsible for the biological phenotypic changes in BMs relative to commercial pigs, we calculated the number of gene families that diverged along different branches with marked changes (expansion or contraction) (Figure 5A). BM displayed relatively less events of gene family expansion (283 versus 313) and contraction (535 versus 592) compared with Duroc.

Compared with commercial pigs, small body size is one of the most visible characteristics of BM, probably as an ecological response called out by the need of relatively large heat dissipation area formed through body volume loss to a long-term warm-temperature environment in low latitudes (Figure 5B) like other species (Forster et al., 2012, Sheridan and Bickford, 2011). Our analysis identified that the contraction of the LILRA and LILRB subfamilies (from 11 and 4 copies in Duroc to 5 and 2 copies in BM, respectively), belonging to the LILR gene family, which is associated with bone development, can result in pycnodysostosis characterized by osteosclerosis, short stature, clavicular dysplasia, and skull deformities in humans (Song et al., 2017) (Figure 5C). This may explain why BM has a relatively short body length with only 19–20 thoracic and lumbar vertebras, less than that in commercial pigs (21–23 thoracic and lumbar vertebras in Duroc) and low body height with a dramatically shorter fibula than that in Duroc (Figure S15; Table S1).

Bidirectional Selection in BM and Duroc

To look for rapidly evolving genes that underlie different adaptive traits between BM and Duroc under divergent selective conditions, we identified 789 and 990 PSGs in BM and Duroc by estimating ω values (nonsynonymous/synonymous rate ratio [Ka/Ks]), respectively (Tables S22 and S23). Genes related to organ development and morphology in Duroc appear to have undergone rapid evolution (correct p < 0.05; Figures S16 and S17; Table S23). Many PSGs of Duroc were significantly enriched in the categories “focal adhesion” (17 PSGs), “Hippo signaling pathway-fly” (5 PSGs), and “extracellular matrix-receptor interaction” (8 PSGs), all of which play important roles in tissue and organ morphogenesis (Hynes, 2009) (Figure S16; Table S23). This information coincides with the findings that BMs exhibit an organ weight that is more comparable with that in humans than that in commercial pigs (Table S24), supporting that BMs provide better donors, including liver (Shah et al., 2016), heart (Mohiuddin et al., 2016), kidney (Higginbotham et al., 2015), spleen (van der Windt et al., 2009), and lung (Kubicki et al., 2015), for xenotransplantation.

Notably, although both these PSGs in BM and Duroc are over-represented in the candidate gene set related to energy homeostasis processes (“PI3K-Akt signaling pathway” and “glycerolipid metabolism” KEGG pathways shared in both BM and Duroc [correct p < 0.05]; “AGE-RAGE signaling pathway in diabetic complications” KEGG pathway unique to Duroc [correct p < 0.05]), the specific gene contents within these sets differ drastically between the two pigs (Figure S16; Tables S22 and S23). Rapidly evolving energy metabolism genes in BM are involved mainly in adipose deposition (such as AGPAT2, AGPAT4, AWAT1, and FABP6) and the growth and development of cardiac and skeletal muscles (such as FGFR2, FGFR4, and IGFBP4), reflecting a need to enhance the efficiency of biomass production under the nutrient-restricted feeding condition. Conversely, energy metabolism PSGs in Duroc are involved mainly in diabetic diseases, including eight PSGs playing important roles in resistance to diabetes (GAPT4, ZNF608, and BBS2) (Nishimura et al., 2001, Speliotes et al., 2010), atherosclerosis (SERPINE1) (Koch et al., 2010), insulin secretion (UCP2) (Bordone et al., 2006), and energy expenditure (GPAM, PRKCA, and NCOA3) (Han et al., 2017, Yu et al., 2017).

We additionally found different PSGs potentially involved in the neurocircuitry control of peripheral metabolism between BM and Duroc. Unlike the enrichment of the “dopaminergic synapse (12 PSGs)” KEGG category for Duroc PSGs (correct p < 0.05), the PSGs of BM were significantly enriched in KEGG pathway related to the functional regulation of the central nervous system (CNS), “GABAergic synapse (8 PSGs)” (correct p < 0.05) (Figures 5D and S16; Table S22). BMs, in contrast to commercial pigs feeding ad libitum, have undergone long-term limited feeding (over their entire breeding history of more than 30 years) without any flavor supplementation. As a result, BMs, similar to humans under restrained eating condition (Laessle et al., 1989), consumed ∼2.46-fold energy beyond their need-based requirements when they were implemented free food intake (Table S1), which was supported by a large number of distinct PSGs involved in synapse in the CNS between two breeds. Duroc PSGs are enriched in “dopaminergic synapse” affecting “food reward” mechanisms (overconsumption of rewarding palatable foods, often in quantities exceeding energetic needs) (Clemmensen et al., 2017). The BM PSGs over-represented in “GABAergic synapse,” which release an inhibitory projection (GABA) onto MC4R (Kleinridders et al., 2009, Myers and Olson, 2012), are involved in promoting food intake and hyperphagia, even when the basic energy needs are met. These two genetic changes in the CNS may interact to affect the eating behavior of BMs different from Durocs by increasing food intake and reducing sensitivity to dietary excess (Figure 5D). Thus, with free access to enough nutrients despite less appealing food, BMs are still willing to consume excess nutrients beyond their basic physiological needs due to decreased satiation (Table S1). After food digestion in the stomach or gut, cellular signaling pathways involved in energy metabolism, such as PI3k-AKT pathway, play a key role in the peripheral glucose homeostatic loop. Insulin receptor and insulin-like growth factor receptor 1 promote the phosphorylation of receptor tyrosine residues (pY), leading to the recruitment and phosphorylation of the insulin receptor substrate (IRS). These recruit PI3K, which activates AKT by targeting its pleckstrin homology domain indirectly to control glucose transporter 4 (GLUT4) translocation to the plasma membrane and thus cellular uptake of glucose (Hribal et al., 2002, Kleinridders et al., 2009, Manning and Toker, 2017, Myers and Olson, 2012). In contrast to BMs, Durocs have undergone positive selection for AKT1 and AKT2 (Figures 5D and S18); especially the AKT2-encoded protein has multiple alterations in pleckstrin homology domain, which likely affects cellular uptake of glucose and regulation of blood glucose levels. The differential positive selection of genes involved in the CNS and cellular energy metabolism may be responsible for the lower resistance to “diabetogenic” environment in BM compared with Duroc.

Effects of Inbreeding and Directional Selection on BM

Inbreeding of the BM from BX population was also accompanied by the selection of some features (e.g., smaller body size, tamer behavior) favorable for use as an animal model (Figure 1). To detect genomic footprints left by selection, we conducted population resequencing analyses to measure genome-wide variants between BMs and BXs and found over 16 million single nucleotide polymorphisms (SNPs) in these two populations (9,169,662 in BXs versus 7,051,076 in BMs).

Among all SNPs detected in the BMs and BXs, we found 6,040,262 (58.9%) unique heterozygous variations in BXs, six-fold more than the number in BMs (928,628 or 9.1%), and the number of fixed homozygous variations was significantly smaller in BXs (167,992, 1.6%) than in BMs (3,161,040, 30.9%) (Figure 6A), confirming that long-term inbreeding has dramatically decreased the degree of heterozygosity in BMs. Chromosome-wide comparison of nucleotide diversity (Pi, π) for fixed and unique variants revealed fewer variants in BMs than in BXs on all autosomes and X chromosome (Figure 6B). These indicate high genetic similarity or homology among individual BMs and confirmed that the BM inbred line has already become genetically stable, which was also further verified by genetic structural analyses of microsatellite data (Figures S19 and S20; Table S25).

Figure 6.

Single Nucleotide Divergence and Genomic Regions with Strong Selective Sweep Signals in BMs and BXs Revealed by Population Resequencing

(A) Classification of single nucleotide variation between BMs and BXs. The ∼17 million single nucleotide differences between the two Bama pig lines were classified into three subclasses. The overlapping regions represent heterozygous variations shared between BMs and BXs. U, unique heterozygous variations evident in each species; F, the number of fixed homozygous variations in each species.

(B) Population diversity (Pi) values between BMs (n = 50) and BXs (n = 50). Data are represented by box-and-whisker plots. Boxes represent the interquartile range between the first and third quartiles and median (internal line). Whiskers denote the lowest and highest values within 1.5 times the range of the first and third quartiles, respectively.

(C) Z score of heterozygosity (ZHp) patterns in BMs and BXs within 100-kb windows across the genome.

(D) Distribution of Hp ratios and Fst values. Data points located to the left and right of the left and right vertical dashed lines, respectively, and above the horizontal dashed line were identified as selected sweep regions for BM and BXs, respectively.

Besides, a genome-wide screen to determine the Z score of heterozygosity (Hp) of BMs showed a skewed distribution and low average value (0.365), in contrast to the normal distribution and higher average value (0.414) in BXs (Figure 6C). This suggests that BM is already a non-natural population after the 30-year inbreeding period, which is coincident to the differences of some biological characters of BMs from BXs. To reveal the genomic basis underlying these biological differences, we further calculated the fixation index (F_ST), a measure of genetic differentiation, and identified convincing selective sweep regions using significant high F_ST values and low/high Hp ratios as cutoff values in sliding windows of 100 kb with a 50-kb step size along the entire genome, in BMs and BXs (Figure 6D). Even under such a stringent criterion, we found eight times as many genomic regions with strong selective sweep signals in BMs (271) as in BXs (41), quantitatively suggesting that the selection in BMs is more powerful than that in BXs in shaping the genome, resulting in rapid changes in the phenotypic or behavioral traits of BMs (Figure 6D).

We subsequently extracted the annotated protein-coding genes from these regions (282 genes in BMs and 36 genes in BXs) to examine the precise correlation between selection and the altered physiological traits of BMs relative to BXs (Table S26). One of the greatest challenges during the over 30-year period of BM breeding was overcoming inbreeding bottleneck (Charlesworth and Willis, 2009). Our inbreeding practices resulted in increased genetic homozygosity, and individuals carrying homologous deleterious recessive alleles are at risk of reduced survival and fertility, the severity of which intensifies with each generation (only a few individuals from one of the four pair of original parents were bred to the 19th generation; Figure 1). Our population-level analysis revealed clearly selective footprints on inbreeding depression for six autosomal recessive disease-related genes involved in autosomal recessive cone dystrophy (CACNA2D4) (Wycisk et al., 2006), autosomal-dominant familial Meniere disease (DTNA) (Requena et al., 2015), early coronary disease and metabolic risk (LRP6) (Mani et al., 2007), autosomal recessive autism spectrum disease (MYO1A) (Talebi et al., 2018), limb defects associated with epicanthus inversus syndrome and Möbius syndrome (SOX14) (Hargrave et al., 2000), and autosomal dominant retinitis pigmentosa (SPP2) (Liu et al., 2015). Thus, selective elimination of recessive deleterious mutations within these genes concurs with our inbreeding practice of weeding out the individuals that lack fitness traits, allowing the healthiest individuals to continue breeding.

Selection of individuals with a small body size has been a major focus during BM inbreeding. We found four genes related to bone development, showing evidence of a strong signature of selective sweeps. ATP6V1H regulates the growth and differentiation of bone marrow stromal cells (Zhang et al., 2017). CHMP5 controls bone turnover rates by decreasing nuclear factor-κB activity in osteoclasts (Greenblatt et al., 2015). GPR55 is a putative cannabinoid receptor that regulates osteoclast function and bone mass (Whyte et al., 2009). UHMK1, as a bone mineral density-related protein, regulates osteoblasts and osteoclasts (Choi et al., 2016). These genes encoding factors associated with bone development may explain the smaller body size of the BM compared with the BX (Table S1).

A visible trait of BM is its uniform two-end-black fur color. Its coat color differs from that of the BX, in that it lacks the black spots of different sizes on the shoulders, back, and waist characteristic of BXs (Table S1). The SNAI2 gene, deletion of which results in human piebaldism characterized by congenital patches of skin and hair from which melanocytes are completely absent (Sanchez-Martin et al., 2003), exhibits strong selective sweep signals. We infer this is why the black spots disappeared from BM skin (Table S1).

Another key artificially selected trait in BM inbreeding was tamer behavior for adaptation to a captive environment, as opposed to the anxiety-associated aggressive behavior of the primitive population. Four BM genes predominantly involved in anxious behaviors in humans or animal models were identified in the regions with strong selective sweep. GPR55 receptor agonists and antagonists modulate anxiety-related behaviors in rats (Rahimi et al., 2015). ASIC1 is highly expressed in patients with a panic disorder characterized by unexpected, recurrent panic attacks, associated with a fear of dying and worry about possible future attacks or other behavioral changes as a consequence of the attacks (Gugliandolo et al., 2016). RXRG is associated with mouse anxiety and human bipolar disorders (Alliey-Rodriguez et al., 2011, Ashbrook et al., 2015). CRY1 directly influences cognitive function and anxiety-related behaviors (De Bundel et al., 2013). In short, BMs have broken the inbred bottleneck and some phenotypes of them have changed, which are consistent with the altered genomic regions resulting from long-term intense selection.

Difference in Resistance to Diabetogenic Environment between BM and Duroc

It is speculated that the divergent feeding conditions contribute to different evolution of energy metabolism system between BMs and commercial pigs, so as to weaken diabetogenic pressure endurance of BMs relative to Durocs (Figures 2 and S2; Tables S2 and S3). To investigate the molecular mechanism underlying difference of tolerance to diabetogenic pressures between these two breeds, we first conducted selective sweep analysis based on F_ST values and Hp ratios to dissect selection in BM and Duroc for adaptation to the divergent feeding conditions (Figure S21). Population-level analysis revealed that the energy metabolism systems of BM and Duroc were under distinct selections. The BM energy metabolic genes embedded in selected regions belong mainly to categories that are related to energy deposition (Gene Ontology [GO] term: “lipid storage,” “positive regulation of lipid storage,” “regulation of lipid storage,” p < 0.05) and diabetic disease (KEGG pathway: “Maturity onset diabetes of the young,” correct p < 0.05) (Figures S22 and S23). Conversely, we identified that five Duroc lipid-related genes were overrepresented in “response to lipid” and “cellular response to lipid” GO term (p < 0.05) (Figure S23). These distinct selective sweep events related to energy metabolism are coincident with feeding difference between BM and Duroc, which potentially facilitates the BM's predisposition of diabetes.

Currently, long noncoding RNAs (lncRNAs), their target genes, and diabetes have drawn increasing attention among researchers (Knoll et al., 2015). As an important post-transcriptional pathogenesis of diabetes, lncRNAs and their associated orchestrated networks are implicated in mediating complex pathological mechanisms of diabetes (Kato et al., 2016, Liu et al., 2014). To delineate the influence of lncRNAs and mRNAs on the different resistance between BM and Duroc to diabetogenic environment, we next generated transcriptome sequencing by using two key glycogen-metabolizing tissues—liver and skeletal muscle of BMs (BM-induced group) and Durocs (Duroc-induced group) in “diabetogenic” environment (Figures 2 and S2; Tables S2 and S3), respectively. After RNA sequencing (RNA-seq) and a series of lncRNA identification steps, 5,186 and 6,552 lncRNAs were identified in liver and skeletal muscle, respectively, and subjected to subsequent analyses (Figure S24). As is characteristic of lncRNAs, our identified lncRNAs had fewer exons, a shorter average length, and lower expression levels compared with mRNAs (Figure S25). We found a total of 258 (20 upregulated and 238 downregulated) and 227 (26 upregulated and 201 downregulated) lncRNA transcripts and 477 (279 upregulated and 198 downregulated) and 277 (140 upregulated and 137 downregulated) mRNA transcripts differentially expressed in liver and skeletal muscle, respectively, between the BM-induced and Duroc-induced groups (Figure S26).

GO analysis of the potential targets of differentially expressed lncRNAs revealed diabetes-associated terms among the top five significantly enriched terms (p < 0.01) (Tables S27 and S28), such as “insulin-like growth factor binding” (GO: 0005520) in liver and “glycerol-3-phosphate and alditol phosphate metabolic process” (GO: 0006072 and GO: 0052646) in skeletal muscle. We also found a total of nine and eight significantly enriched KEGG pathways among the potential targets of the differentially expressed lncRNAs in liver and skeletal muscle, respectively (p < 0.05) (Tables S29 and S30), of which the AMPK and PI3K-Akt signaling pathways were related to diabetes. Next, GO and KEGG analyses of the significantly dysregulated mRNAs in liver and skeletal muscle also indicated significant enrichment of transcripts related to the AMPK signaling pathway (Tables S31–S34).

Furthermore, we selectively analyzed the lncRNAs and their target genes that (1) were both significantly differentially expressed between BM-induced and Duroc-induced groups and (2) should be associated with diabetes based on functional enrichment. We detected a total of six pairs fulfilling these criteria involving six lncRNAs and two target genes. In the liver, we found that XLOC_006422 and XLOC_026958 were correlated (trans-acting) with PGC-1α and XLOC_044402 with PEPCK. In skeletal muscle, XLOC_003015 and XLOC_023027 were correlated with PEPCK through trans activity, and XLOC_026564 was correlated with PGC-1α. Among these, the target genes PGC-1α and PEPCK have been directly linked to T2DM previously (Samuel et al., 2009, Sawada et al., 2014, Soccio et al., 2015), and their expression was significantly higher in BM-induced liver and skeletal muscle than in Duroc-induced tissues (p < 0.05; Figure S26). Furthermore, KEGG pathway analysis showed that PGC-1α and PEPCK were both enriched in the “AMPK signaling pathway” (p < 0.05) related to energy metabolism. The selected lncRNAs and target gene expression in RNA-seq were validated by qRT-PCR analysis (Figure S27; Table S35). Given these results, we suspect that these lncRNAs probably participate in the regulation of resistance to “diabetogenic” environment by influencing the expression of diabetes-related genes such as PGC-1α and PEPCK, although the underlying mechanisms require additional investigation.

Discussion

The establishment of BM line with high inbreeding level is an important action, which can strengthen the laboratory pig-resource gene pool and enrich the pig diversity, as well as directly compensate the disadvantages of using pig in diabetes study. The chromosome-level reference genome sequence of BM reported here is a high-quality miniature pig genome that offers the needed information to expedite current efforts in developing BM as an ideal experimental animal to reveal the genetic basis of diabetes. Comparing the genome sequence of BM with that of commercial Duroc provided insights into the distinct evolutionary scenarios, especially in the CNS and cellular energy metabolism, which leads to the different resistance to diabetogenic pressure between them, occurring under inbred selection for experimental use and artificial selection for commercialization. Besides, genomic comparison of these two pig breeds also identified many genetic loci related to body size and sexual maturation, which may not only promote the miniaturization and prematurity of experimental animals but also provide possible selection markers for the breeding of commercial pigs. The whole-genome resequencing analysis between BX and BM revealed genomic loci that have been under selection during BM inbreeding, maximizing the scientific value of these BM populations as a reference for improving the inbreeding practices of other inbred strains. The dissection of different resistance to diabetogenic environment between BM and Duroc by selective sweep, transcriptome, and comparative genome analyses help to improve current understanding of different diabetes susceptibility in humans. In conclusion, the data in this study provide a valuable resource and tool for functional genomic studies on BM as well as increases use of the BM as an animal model in broader field, particularly human diabetes.

Limitations of the Study

In this study, we developed a laboratory miniature pig inbred line with advantages for translational medicine, and presented the chromosome-scale BM genome. Although the genomic analyses revealed that BMs strongly resemble human beings, it also demonstrated that attention should be paid to the interspecific differences when selecting BMs for use as human disease models. Besides, functional research should be performed to further validate the candidate genes involved in susceptibility to diabetes.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: September 27, 2019