Genomic Selection Extends Shrimp Survival Under Ammonia Stress

Pacific white shrimp, Litopenaeus vannamei, dominate global aquaculture, accounting for roughly 80 percent of farmed shrimp production and an annual yield exceeding four million tons. Yet the intensive farming systems that make this species so productive carry a persistent threat: the progressive accumulation of ammonia nitrogen in the water, released directly by the shrimp themselves and generated by microbes breaking down leftover feed and feces. Long-term exposure to elevated ammonia suppresses growth, damages immunity, accelerates molting in harmful ways, injures shells, and triggers oxidative stress pathology, all of which raise mortality and economic losses. A new study published in Genetics Selection Evolution reports that genomic selection can meaningfully improve shrimp tolerance to this stressor, and that a cost-saving hybrid breeding strategy makes the approach practical for commercial hatcheries.

Researchers led by Shuo Fu and Jianyong Liu set out to test whether genomic selection, which uses genome-wide markers to estimate each animal’s breeding value, could outperform traditional pedigree-based evaluation for ammonia resistance, a trait no previous study had targeted with genomic methods in shrimp. In March 2023, the team established 51 families through an incomplete diallel crossing design using breeders from a core population descended from five geographic sources. From these, 22 full-sib families sired by 14 males and 13 females, including 12 half-sib families, were selected for challenge testing. Thirty individuals per family were tagged with visible implant elastomer tags and reared together in a single tank to minimize environmental variation, yielding 566 shrimp, 23 to 27 per family, for the stress experiment.

The challenge test itself was severe. Based on preliminary trials, the researchers set ammonia nitrogen at 85 mg/L, the concentration they had determined to be the 96-hour semi-lethal level. Shrimp were placed in a 1.8 by 1.8 meter test pool, and after a five-day holding period the concentration was adjusted with analytical-grade ammonium chloride. Water was replaced daily, temperature was held at 25 plus or minus 1.5 degrees Celsius, pH at 8.1 plus or minus 0.2, salinity at 30 plus or minus 1, and dissolved oxygen at 6 to 8 mg/L. Mortalities were checked every hour, and survival time was recorded for each animal. Death was defined as lying motionless at the bottom of the pool without responding to touch. Initial mortality appeared at 9 hours, and all shrimp had died by approximately 160 hours post-exposure.

Across the population, the median lethal time was 84 hours, rising to 93 hours within the sequencing subset. Family means for survival time ranged from 63.67 to 105.33 hours, with an overall mean of 89.16 plus or minus 12.96 hours, a maximum inter-family gap of 41.66 hours that pointed to real genetic differentiation in the trait. For genotyping, the team ranked all shrimp by tolerance and sampled 200 individuals across low, intermediate, and high phenotypic strata, excluding two families whose average body weights deviated beyond two standard deviations of the family mean. Whole-genome resequencing on the MGISEQ-2000 platform at 10-fold depth, aligned to the reference genome, initially produced more than 20 million raw SNPs; after stringent quality control on missingness, minor allele frequency, and Hardy-Weinberg equilibrium, 15,928,425 high-quality SNPs remained.

The genetic parameter estimates revealed a striking gap between pedigree-based and genomic models. Heritability of ammonia tolerance estimated from the pedigree relationship matrix in the PBLUP model was 0.26 plus or minus 0.09, close to earlier published values of 0.13 and 0.17 for this trait. Every model incorporating genomic information yielded markedly higher estimates, ranging from 0.56 to 0.58: GBLUP gave 0.57, ssGBLUP 0.57, and the four Bayesian variable-selection models, BayesA, BayesB, BayesC-pi, and Bayesian LASSO, returned 0.58, 0.58, 0.56, and 0.58 respectively. The authors attribute this difference to genomic relationship matrices capturing realized genetic relatedness and Mendelian sampling variation among full-sib and half-sib relatives, which pedigree matrices based on recorded ancestry cannot, thereby separating additive genetic variance from environmental and residual noise more cleanly.

Predictive performance told a similar story. Using five-fold cross-validation repeated 20 times for a total of 100 validation runs, all six genomic models beat PBLUP, improving prediction accuracy by 32.45 to 39.25 percent, with ssGBLUP and Bayesian LASSO showing the largest gains. The genomic models were also highly concordant with one another, with correlations between their estimated breeding values ranging from 0.87 to 1.00, and the four Bayesian models nearly identical at above 0.98. Correlations between PBLUP and the genomic models were lower, at 0.68 to 0.78, and dropped further, to 0.30 to 0.56, when the analysis was restricted to the top 30 percent of individuals by survival time, the animals a real breeding program would actually select. Among genomic models, correlations among top-ranked individuals stayed high at 0.75 to 1.00, meaning the different genomic approaches converge on the same elite candidates while pedigree-based rankings could diverge substantially.

The team was candid about the limits of cross-validation. Because partitioning was done at the individual level rather than the family level, closely related animals could appear in both training and validation sets, which can inflate observed accuracy and produce more optimistic estimates than would hold in truly independent populations. Prediction accuracy from cross-validation also depends on the trait’s heritability, complicating comparisons across studies. To address this, the researchers moved beyond cross-validation to an independent progeny test, a step many genomic selection studies skip. Because the genomic models agreed so closely, they chose ssGBLUP for validation and, instead of genotyping all candidate parents, calculated average genomic estimated breeding values for families in the reference population.

The progeny design was deliberately practical. The top 20 percent of families by average GEBV, four families, formed the high-resistance group, and the bottom 20 percent formed the sensitive group. Each group was propagated through complete diallel crosses among its four families, producing 16 families per group that were then mixed and reared under standardized conditions. In the first-stage stress test at 60 mg/L ammonia nitrogen, 200 shrimp per group were challenged with mortality recorded every 6 hours. For the second stage, 120 shrimp per group were tagged, co-reared in a common environment for three months to eliminate environmental effects, and then challenged at 160 mg/L with observations every 2 hours. Genotyping of the progeny used a self-developed 50K liquid-phase breeding chip, with SNPs shared between the resequencing and chip datasets used to estimate GEBVs, since pedigree information for the validation population was unavailable.

The results confirmed the genomic predictions. In both stages, the high-resistance group survived significantly longer than the sensitive group, by 11.82 percent in the first stage and 11.38 percent in the second. Kaplan-Meier survival curves and Log-rank tests showed highly significant differences: in the first stage, the high-resistance group averaged 35.01 hours of survival versus 31.32 hours for the sensitive group, with a hazard ratio of 1.118 and a test statistic of 7.1901 at p equals 0.0073; in the second stage, mean survival was 50.94 hours versus 45.72 hours, with a hazard ratio of 1.114 and p equals 0.0062. The correlation between GEBVs and actual survival time in the progeny was statistically significant but weak to moderate, at r equals 0.306 with p below 0.05. The authors note the improvement is smaller than gains reported for genomic selection against vibriosis in Pacific oysters or visceral white-nodules disease in large yellow croaker, but statistically robust across both independent stages.

The broader significance lies in the economics. Shrimp are small, hard to tag early, and a single female can produce up to 200,000 eggs at once, making it difficult to decide which individuals to genotype and prohibitively expensive to genotype entire candidate populations. The hybrid strategy tested here, combining conventional BLUP selection with genomic evaluation of families rather than individuals, cuts genotyping costs dramatically while retaining most of the accuracy benefit, and the authors argue it may also reduce the negative impact on other traits that intense selective breeding can impose. Given current sequencing costs and the absence of a comprehensive multi-trait breeding system in L. vannamei, they position this combined approach as a practical, transitional route from conventional selection to full genomic selection in commercial shrimp breeding, with the study data deposited in NCBI under accession PRJNA1308829.

Subject of Research: Genomic selection for ammonia nitrogen tolerance in Pacific white shrimp

Article Title: Genomic selection improves survival time under high ammonia nitrogen stress in Litopenaeus vannamei

Article References: Fu, S., Zhang, Y., Wu, G., Guo, C., & Liu, J. (2026). Genomic selection improves survival time under high ammonia nitrogen stress in Litopenaeus vannamei. Genetics Selection Evolution, 58(1), Article 63. https://doi.org/10.1186/s12711-026-01081-6

Image Credits: AI Generated

DOI: 10.1186/s12711-026-01081-6

Keywords: genomic selection, Litopenaeus vannamei, ammonia nitrogen tolerance, aquaculture, heritability, ssGBLUP, GEBV, progeny testing, selective breeding, shrimp farming, SNP markers, survival time