SNP-based breeding for broiler resistance to ascites and evaluation of correlated production traits

The goal of this study was to evaluate marker-assisted selection (MAS) in broiler chickens using previously mapped gene regions associated with ascites syndrome incidence. The second-generation MAS products were assessed for impact on ascites phenotype and whether there were associated changes in important production traits. Previously, we used whole genome resequencing (WGR) to fine-map 28 chromosomal regions as associated with ascites phenotype in our experimental ascites broiler line (Relaxed, REL) based on a hypobaric chamber challenge. Genotypes for single nucleotide polymorphisms (SNPs) in mapped regions on chromosomes 2 and 22, were used for MAS in our REL line. After two generations, birds homozygous for the genotypes associated with resistance for both chromosomal regions were established. The MAS F2 generation was then compared to the REL line for ascites susceptibility and 25 production traits. Selection based on SNPs in the carboxypeptidase Q (CPQ, Gga2) and leucine rich repeat transmembrane neuronal 4 (LRRTM4, Gga22) gene regions resulted in a sex- and simulated altitude- dependent reduction of ascites incidence in two F2 cohorts of the MAS line. Comparisons of the F2 MAS and REL lines for production traits when reared at ambient pressure found no significant negative impacts for feed intake (FI), feed conversion ratio (FCR), or deboned part yields for either sex for two F2 cohorts. There were, however, improvements in the MAS for full-trial body weight gain (BWG), FCR, absolute and relative tender weights, and relative drumstick weight. These results validate the mapping of the 28 chromosomal regions and demonstrate that fine mapping by WGR is an effective strategy for addressing a complex trait; it also stands as the first successful SNP-based selection program against a complex disease trait, such as ascites. The MAS line is comparable and, in some instances, superior, in growth performance to the REL control while being more resistant to ascites. This study indicates that MAS based on WGR can provide significant breeding potential in agricultural systems.

hypertension is driven by increased oxygen demands of a rapidly growing body supplied by an inadequate cardiopulmonary system [16][17][18]. The incidence can be amplified as a result of an oxygen-reduced environment, or increased thermoregulation in cold environments [9,15]. The bird's body increases blood pressure in an attempt to respond to metabolic demands, which leads to incomplete gas exchange in the lungs [15,[19][20][21]. Semi-oxygenated blood is then sent to the organs with substantial detrimental effect on the liver, causing the accumulation of abdominal fluid. It has been estimated that ascites is the cause of up to 8% of broiler mortality and accounts for up to $100,000,000 in economic losses annually, making it both a significant animal welfare and economic concern [6,14,15,[21][22][23][24][25].
Previous research at the University of Arkansas on the genetic basis of ascites involved the development of three research lines from a commercial elite line through divergent selection for ascites resistance when exposed to simulated high elevation conditions [11]. The base population (Relaxed, REL) for the selection study was derived from a commercial elite line in the 1990s and maintained through random mating without selection. Sib-selection based on ascites phenotype assessed through a 6-week hypobaric chamber challenge produced ascites resistant (RES) and ascites susceptible (SUS) lines. Rapid response in divergent selection with successive generations suggested a limited number of major genes. A series of genome-wide association studies (GWAS) using SNP panels identified a few candidate SNPs as associated with ascites phenotype, but subsequent MAS-based breeding projects were unsuccessful in validating these few loci [41][42][43][44]. More recently, whole genome resequencing (WGR) identified 28 genomic regions where SNP clusters (100s to 1000s of SNPs) showed frequency bias with respect to ascites phenotype [45,46]. Two of these regions were validated by further genotyping of additional DNA samples and found to have potential epistatic interaction. One region spanned more than 120 kbp on chromosome 2 including the 3′ end of the gene for carboxypeptidase Q (CPQ). The second was an approximately 50 kbp region on chromosome 22 spanning the 3′ end of the gene for leucine-rich repeat transmembrane neuronal 4 (LRRTM4). Both these genes have been associated in human GWAS with blood traits, heart rate, and blood pressure consistent with factors contributing to ascites incidence in poultry.
Therefore, the current study reports on whether MAS based on SNP genotypes for the regions of both CPQ and LRRTM4 can produce offspring with greater innate ascites resistance. Since ascites susceptibility could potentially be linked to important production traits, we also assessed the impact of selection on important broiler production traits.

Methods
All breeding, hatching, grow-out, and processing took place at the University of Arkansas Poultry Research farm. All animal procedures were approved by the University of Arkansas Institutional Animal Care and Use Committee (Approval Numbers 18083 and 18088) and performed in accordance with relevant guidelines and regulations. This study is reported in accordance with ARRIVE guidelines (https:// arriv eguid elines. org).

Bleeding, genotyping, and husbandry of breeder stock
Birds used for breeding were genotyped by collecting 10 μl of blood from the brachial vein which was further processed using a rapid DNA extraction method [47]. These DNAs were then genotyped using exonuclease assays run in triplicate for both CPQ and LRRTM4 genes, as described [45,46]. Selected breeders were then moved into individual breeder cages for insemination and production of MAS offspring. At 18 weeks of age, the birds were put on a lighting schedule to induce egg production: a starting schedule of 12 h light:12 h dark that progressed for 4 weeks to 16 h light:8 h dark, which was maintained through the insemination/egg collection period. Insemination occurred two times weekly, and eggs were collected daily and were labeled by hen. All eggs were stored at 18 °C and 60% relative humidity until sufficient numbers were reached to begin hatching the next generation.

Hatchery protocol
Sets of eggs were placed in a setting incubator (Jamesway Incubator Co., Cambridge, Ontario, Canada) at 99.6 °F and 85% relative humidity for 18 days. On d 18, the eggs were candled, infertile eggs removed, fertile eggs placed into hatch baskets, and transferred by mating combination to a hatching incubator (Jamesway Incubator Co.) at 98.0 °F and 88% relative humidity. On d 21, hatched chicks were wing-banded using different band colors for each sub-population, and individual band numbers recorded.

Hypobaric trials
For both hypobaric cohorts, birds of each line (MAS and REL) were mixed in one of 40 battery cages (measuring 0.6 × 0.6 × 0.3 m) with appropriate numbers from each line to maintain even distribution throughout all the cages. No mortality data other than wing band number was collected for the first week so that final mortality rates would not be confounded by chicks that failed to start. For the remaining 4 weeks of the trial, the husbandry and necropsy methods followed those described previously [11,41,44]. As the birds grew, bird densities were reduced in cages to maintain compliance with animal welfare requirements. Birds were initially chosen for removal due to observation of clinical ascites phenotype through palpation of the abdomen. All birds removed were euthanized and subsequently necropsied for ascites phenotype. Once all birds showing ascites phenotype had been removed, additional birds were chosen at random to meet welfare requirements and maintain consistent numbers for both lines. All birds remaining at the end of 5 weeks were euthanized and necropsied to determine ascites phenotype.

Floor trials
Hatches were placed in floor pens that were top-dressed with fresh pine shavings. Similar numbers for each line were placed at similar densities. Surplus chicks were placed in a separate pen. Cardboard trays for feed were placed and remained in the pens for the first 7 days of the trial. Feed and water were provided ad libitum throughout the trial. One continuous water line per row of pens was adjusted as needed for bird height. Health inspections occurred twice daily at a minimum. All mortalities were removed upon discovery and wing band number, pen of origin, body weight, and any clinical observations including ascites incidence were recorded. If available, the bird was replaced with another of the same sex and genetic line from the surplus birds. Feed was formulated to Cobb-Vantress, Inc., industry recommended standards (formulations can be found in Supplemental Table 1

Processing
The day prior to processing, a subset of each cohort consisting of 25 males and 25 females per genetic line (n = 100 per cohort) were randomly selected. These birds were removed from the pen first on the morning of processing and were removed from the processing line prior to evisceration to be necropsied for organ weights of liver, lungs (set), spleen, and heart, as well as right ventricle to total ventricular weight (RVTV). These birds were not chilled or deboned. Also on the day prior to processing, a second subset from each cohort were randomly chosen, consisting of 50 males and 50 females per genetic line (n = 200 per cohort). These birds were wing-banded to be assessed for MAS impact on meat quality measurements of the breast fillet. Feed was removed 10 h prior to processing to ensure feed passage. All birds were collected the morning of processing and transported to the University of Arkansas Poultry Pilot Processing Plant. At the processing plant, back dock live weight was collected prior to the birds being electrically stunned and exsanguinated, followed by a scalding water bath to loosen feathers, then feather, head, and paw removal. Carcasses were then eviscerated, and the hot carcass and fat pad weights were collected. Carcasses were chilled for 3 h and deboned thereafter for the determination of absolute weight and relative (to back dock live weight) yield of wings, breasts, tenders, thighs, and drumsticks. The 200 carcasses marked for further evaluation were processed as described but also evaluated for muscle quality traits including breast fillet weight, color, and pH at 4-and 24-h chill time. The deviation between the 4-and 24-h fillet weights were used to calculate drip loss. Color and pH were measured using a Minolta CR-400 handheld model with PC-linked Spectramag-icX software and Testo model 205 handheld spear-tip probe, respectively. Color readings were taken on the dorsal surface of the breast, while pH was measured in the cranial region of the breast. The breasts were then frozen until the completion of both trials for cooking and shear force measurements. Shear requirements were calculated based on four measurements in the cranial region using a TA.XTPlus equipped with a Meullenet-Owens Razor Shear head attachment.

Statistical methods
All statistical analyses other than the survival model were conducted in R with statistical significance denoted by a P-value ≤0.05.
Hypobaric mortality data were analyzed using a generalized linear model (RStudio Team, 2016) of final ascites mortality as well as a survival model which showed the effect of genetic line and sex on the probability of survival over time.
Live performance data were analyzed using two-way ANOVA between the main effects of trial and genetic line. Parts weights, organ weights, heart characteristic, and meat quality characteristic data from processing were analyzed using three-way ANOVA between the main effects of trial, genetic line, and sex. All means were separated by Tukey's Honest Significant Difference (HSD) test.

First-generation breeding
Breeders for the first parent generation were selected from REL breeder stock in July of 2018. After bleeding and genotyping, two P 0 crosses were then created: one consisting of birds with all non-reference SNPs (compared to galGal6, [46]) for each gene region, designated P 0 -1, and the other consisting of birds with nonreference SNPs for CPQ and heterozygous for SNPs for LRRTM4, designated P 0 -2. P 0 -1 consisted of 10 males and 13 females; P 0 -2 consisted of 12 males and 24 females. Separately, ungenotyped REL birds were used to breed generation 1 of the control population. Semen was collected from all the males of each P 0 group, pooled, and used to artificially inseminate each of the females from the same P 0 group.

Second-generation breeding
After hatching, the F 1 progeny produced from P 0 -1 and P 0 -2 were kept in floor pens and managed as breeders until they reached 18 weeks of age, at which point they were then bled and genotyped. Breeders for each F1 population were placed in individual breeder cages and photo stimulated. In May of 2019, breeders from the F 1 were selected based on being homozygous for the nonreference SNPs for both genes; though all breeder SNP genotypes from this point forward were the same, the F1 populations were kept separate in order to complete reciprocal matings between them. Breeders from F 1 -1 (from P 0 -1) consisted of 12 males and 31 females; F 1 -2 (from P 0 -2) consisted of 12 males and 37 females. To produce the control group, 24 males and 72 females from the REL line were also utilized. Insemination began at the same time, when the MAS breeders were 22 weeks of age and the REL control breeders were 20 weeks of age. For the REL, pooled semen from all 24 males were used to artificially inseminate all 72 REL females. For the MAS, pooled semen from the F 1 -1 males was used to artificially inseminate the F 1 -2 females and similarly the F 1 -2 males were used to inseminate the F 1 -1 females. This reciprocal mating scheme produced the F2 generation of birds which would possess only the non-reference SNPs for the CPQ and LRRTM4 genes. Sets of eggs for hatching included at least 250 eggs each from REL and the F 2 of the MAS. At transfer, all eggs were candled and infertile or eggs with embryonic mortality were removed and stored for breakout on hatch day along with eggs that did not hatch; no significant difference (P > 0.05) in hatchery breakout was found between the two lines. After hatch, birds received a wing band that represented their genetic line (either MAS or REL). Sets of eggs were produced every 2 weeks for four total hatches. The first and fourth hatches were subjected to 5-week hypobaric challenges. The second and third hatches were placed for 8-week floor pen trials to evaluate change in production traits associated with MAS.

Hypobaric challenges
F2 chicks for challenge in the hypobaric (Hypo) chamber were placed on November 28, 2019 (Hypo1) and January 8, 2020 (Hypo2). Hypo1 was maintained at 9000 ft. simulated altitude while Hypo2 was initially set at 9000 ft. simulated altitude, then after 2 weeks increased to 11,000 ft. simulated elevation to induce a higher incidence of ascites. In Hypo1, all hatched birds were placed in the chamber (n = 578) whereas in Hypo2, similar numbers of birds were placed from each line and fewer total birds were placed (n = 433) which reduced the number of birds that would need to be culled for compliance with bird density regulations.
The hypobaric challenge results indicate a sex-and elevation-dependent reduction in ascites incidence in both cohorts. Hypo1 saw an overall decrease (P = 0.041) in ascites mortality between the MAS and REL birds, with a 27.3% reduction for ascites in MAS males and a 39.8% in MAS females [ Table 1]. For Hypo2, there was an overall numerical, although not statistically significant (P = 0.162), decrease in ascites mortality between the lines, with reductions of 23.4% in males (P = 0.126) and only 5.2% reduction in females [ Table 2]. There was no significant difference (P > 0.05) between the lines for the right ventricle to total ventricle (RVTV) ratio or body weight. The Kaplan-Meier survival model curve visually echoes these trends, however the analyzed P-values from this model are only numerically different (P > 0.05) [ Fig. 1].

Floor trials
Birds reared for the floor trials were placed on December 12, 2019 (Floor1) and December 28, 2019 (Floor2). On the day of placement, the total number of hatched birds was counted for each line and the number of birds placed per pen determined based on the smallest group; for Floor1 this was 14 birds per pen (0.133 m 2 /bird) and for Floor2 13 birds per pen (0.143 m 2 /bird) with all excess birds placed in spare pens for mortality replacement up until the beginning of the FCR measurement period from d 49 to d 55. As both floor trials were run concurrently in the same barn, though offset by 2 weeks, we considered whether the live performance and processing data should be analyzed as completely separate trials with separate analyses or together by adding the main effect of trial. After consultation with a professional agricultural statistician and colleagues at the University of Arkansas, the latter option was chosen. However, the interaction between line and trial was assessed for each measurement, and any traits identified as having an interaction between line and trial were noted and are discussed. Conclusions about the overall affect of MAS on that particular measurement were not drawn in those cases.
A number of significant differences were found between trial and sex. These were live weight (P = 0.035), hot carcass weight (P = 0.028), fat pad weight (P = 0.019), chilled carcass weight (P = 0.036), absolute and relative tender weight (P = 0.010, females, and the lowest in males from both trials. The mean weights of thighs and drumsticks were separated into four distinct groups, from largest to smallest being Floor1 males, Floor2 males, Floor1 females, and Floor2 females. The relative weight of liver was found to be the largest in females from both trials and the smallest in males from both trials. For both the shear force requirement and the 1:3 shear area, the measurements from Floor1 females were found to be the largest, the Floor1 males were found to be the smallest, and both sexes in Floor2 were intermediate.  Finally, there were interactions found between genetic line and sex. These were live weight (P = 0.011) hot carcass weight (P = 0.012), chilled carcass weight (P = 0.019), breast weight (P = 0.050), thigh weight (P = 0.006), drumstick weight (P = 0.032), and 24-h b* color measurement (P = 0.025). For live weight and the weights of hot carcass, chilled carcass, thighs, and drumsticks, MAS males were found to be the largest, REL males were intermediate, and females from both the MAS and REL lines were found to be the smallest. Breast weight was the largest for males from both the MAS and REL lines, with the smallest weight in the females from both lines. The 24-h b* color measurement was found to be the highest in MAS females, with the three other groups having comparable lower b* measurements. There was also a single three-way interaction (P = 0.039) found for the processing measurements, which was fat pad. In order from largest to smallest, the means were separated as follows: Floor1 MAS females, Floor1 REL females, Floor2 females from both lines, Floor1 REL males, Floor1 MAS males, Floor2 MAS males, and Floor2 REL males.

Discussion
Ascites has been reported to have a significant sexdependent incidence in flocks [39,48]. In the REL line, females show an earlier onset and higher overall incidence of ascites [11]. However, a reduction of ascites incidence was observed between MAS females of the first and fourth cohorts for hypobaric challenges where the simulated altitude differed. At moderate elevations (9000 ft) in Hypo1, ascites incidence was reduced by nearly 40% for the MAS females, while at higher simulated altitude (11,000 ft) in Hypo2, there was only a 5% reduction of female mortality from 61.9% in the REL line to 58.7% in the MAS line. For males, incidence was reduced in the MAS line by 27% in Hypo1, while being reduced by 23% in Hypo2 with the increased simulated altitude. This further demonstrates the sex-linked nature of ascites resistance or susceptibility and suggests more research into epistasis with genes on the sex chromosomes are warranted. Additional research into the impact of MAS on each sex is also warranted. The timeline of this study, and the capacity of the hypobaric chamber, did not allow for single-sex experiments to allow for greater sample sizes. However, even with relatively small cohort sizes, significant reduction of ascites for the MAS line was observed. An additional point of success in these results comes from the lack of impact on production traits. If MAS negatively impacted economically important traits in any major way, then MAS would be much less appealing to the industry. From the live production data, a significant improvement of the MAS was found in body weight gain from the full trial and in FCR. A limited number of traits showed significant differences between overall MAS and REL averages, all of which were improved in the MAS birds; these were absolute and relative tender weights and relative drumstick weight. Additionally, there were some improvements in the MAS that were only seen in one sex. These were live weight, hot carcass, chilled carcass, thigh weight, and drumstick weight, all of which were larger in MAS males over REL males while the MAS and REL female measurements in each of these cases were statistically equivalent. Given these improvements, MAS appears to be not only capable of reducing ascites incidence, but to have a positive impact on some growth characteristics. Notably, there was a significant three-way interaction between trial, genetic line, and sex for fat pad weight relative to back-dock live weight. From the trend in this characteristic, we observed that the females from each cohort had larger relative fat pad weights than the males, but as this did not have an impact on the trends of the deboned parts data, we are less concerned about this outcome.
We also identified differences between MAS and REL for breast color measurements. The measurement of each breast fillet was broken into three components: L* represents the lightness from 0 (black) to 100 (white); a* represents the color spectrum from − 60 (green) to + 60 (red); b* represents the color spectrum from − 60 (blue) to + 60 (yellow) (American Meat Science Association, 2012). A significant difference was found in the L* and a* measurements between the two genetic lines, and in the L* measurement between the sexes. The difference in L* measurements between the two genetic lines is especially interesting as this means that MAS line breast fillets were consistently lighter in color than the REL. While the classification of Qiao et al. [49] considers all of the L* measurements from both cohorts except for MAS female to be "normal" in lightness (MAS female would be classified as "lighter than normal"). This color difference is difficult to visually discern, however fillets with higher L* values may have higher