16 Chapter 16 — Genetics of Disease Resistance and Molecular Epidemiology

Applied Epidemiology in Aquaculture — Part VII: Genetics

Where this sits — Prevent (primary prevention), cutting across all five phases. Breeding fish that resist disease is one of the most durable Prevent tools — it reduces risk every generation, for free, once established. And tracking the pathogen’s own genetics (genogroups, lineages) sharpens Detect, Respond and vaccine choice. Genetics is the thread that runs through the whole framework.

This chapter is built in self-contained sections so collaborators can deepen either the resistance-breeding or the molecular strand without disturbing the rest of the book.

16.1 Overview

Two genetic stories matter for fish health. The host’s genetics determine how susceptible the fish is — and if resistance is heritable, it can be selected, building disease control into the animal itself. The pathogen’s genetics determine how it behaves — and tracking its lineages tells us about virulence, treatment, spread and vaccine fit. This chapter covers both: estimating heritability and breeding value from a challenge test, and the molecular epidemiology of P. salmonis genogroups.

We use srs_challenge_families.csv: 40 sire families, ~50 challenged offspring each, with days survived.

16.2 16.1 Preparation

pacman::p_load(rio, here, tidyverse)   # lme4 / sommer for animal models; ape / ggtree for phylogenetics
fam <- import(here("data", "srs_challenge_families.csv"))

16.3 16.2 Is resistance heritable?

Selection only works if resistance is passed from parents to offspring — captured by heritability (h²), the share of phenotypic variation that is additive-genetic. A challenge test exposes families to the pathogen and measures an outcome (here, days survived). In a half-sib design (families sharing a sire), h² comes from partitioning the variance between and within families:

h² = 4 × σ²(sire) / σ²(total)

a <- aov(days_survived ~ sire, data = fam)
# between-family and within-family mean squares -> variance components
# h2 (half-sib) = 0.27

Heritability is 0.27 — resistance to SRS is moderately heritable, squarely in the range reported for salmonid disease resistance (Barría et al., 2020; Lhorente et al., 2014). That single number licenses a breeding programme: a quarter of the variation in survival is heritable, so selecting resistant parents will shift the next generation.

16.4 16.3 Breeding values and genetic gain

Which fish to breed from? The estimated breeding value (EBV) ranks each family’s genetic merit — here, the sire family means, centred:

fam |> group_by(sire) |> summarise(mean_days = mean(days_survived)) |>
  mutate(ebv = mean_days - mean(mean_days)) |> arrange(desc(ebv))
# top sires: Sire24 (+1.8 d), Sire30 (+1.7 d), Sire16 (+1.3 d)

Select the top sires and the expected genetic gain per generation follows the breeder’s equation (gain = selection intensity × accuracy × additive-genetic SD):

# selecting the top 20% of sires, accuracy 0.7 -> gain ~1.8 days/generation

About 1.8 extra days of survival per generation, compounding every cycle. Genomic selection — using thousands of DNA markers instead of pedigree alone — raises the accuracy term and so accelerates gain, which is why commercial salmon breeding now routinely selects for SRS, IPN and sea-lice resistance (Yáñez, Barría and colleagues). One caution from your keynote: resistance to one pathogen genogroup may not transfer to another, so challenge tests must use field-relevant strains.

16.5 16.4 Molecular epidemiology of the pathogen

The pathogen has genetics too, and typing it is operationally useful. P. salmonis falls into two main genogroups with genuinely different behaviour:

	LF-89-like	EM-90-like
Clinical course	more chronic; kidney/liver lesions	acute haemorrhagic septicaemia; higher, faster mortality
Field marker	no specific lesion	liver nodules
Hosts	all farmed salmonids	Atlantic salmon (Chilean isolates)
Antimicrobials	mostly reduced florfenicol susceptibility	generally fully susceptible
Distribution	~94% of Chilean isolates; Europe, Canada	mainly Chile

(Rozas-Serri et al., 2017, 2023.) Genotyping therefore informs prognosis (expect faster disease with EM-90-like or mixed infection), treatment (interpret antibiograms by genogroup), tracing (follow introductions between sites), surveillance (detect shifts in the pathogen population), vaccine fit (current vaccines are EM-90-based while LF-89-like dominates) and breeding (strain-specific resistance). Multiplex and genogroup-specific qPCR now make typing feasible directly from tissue (Isla et al., 2021), and lineage typing methods extend this to fine-scale tracking (Mora-Salas et al., 2023).

The molecular toolkit in R — ape and ggtree for building and plotting phylogenies, and phylodynamic packages that link genetic distance to transmission — turns sequence data into an outbreak tree, connecting this chapter back to the network spread of Chapter 14. A cautionary note closes the loop: cross-reactive assays mean a “P. salmonis-positive” may be another species entirely (P. nova in Europe), which is why validating assays against circulating strains is both a Detect and a genetics problem (Rodger et al., 2025).

16.6 16.5 Interpretation on the farm

Genetics is the most durable and the most cross-cutting lever in the book. On the host side, a moderately heritable resistance means every breeding cycle can lower baseline risk — a Prevent investment that pays indefinitely. On the pathogen side, routine genogroup and lineage typing sharpens almost every other phase: it refines prognosis and treatment (Respond), tracks spread and population shifts (Detect and Anticipate), and tells you whether your vaccine still matches what is circulating (Prevent). Reading both genomes — the fish’s and the bacterium’s — is how modern aquatic animal health moves from reacting to anticipating.

16.7 16.6 Exercises

Try it yourself (16.1). Estimate heritability from the challenge data using the half-sib ANOVA. What does an h² of 0.27 tell you about the prospects for a resistance-breeding programme?

16.2 Compute the EBVs and select the top 20% of sires. Estimate the genetic gain per generation and explain each term in the breeder’s equation.

16.3 Explain why genomic selection achieves faster gain than pedigree-based selection, in terms of the accuracy term.

16.4 Give three operational decisions (one each for Respond, Detect and Prevent) that change depending on whether an isolate is LF-89-like or EM-90-like.

16.5 A vaccine is formulated against EM-90-like isolates, but ~94% of field isolates are LF-89-like. Using Chapter 10’s effectiveness logic, explain why field VE might be low, and how genogroup surveillance would reveal the mismatch.

16.8 Summary & what’s next

Host resistance to SRS is moderately heritable (h² ≈ 0.27), so it can be selected; genomic selection accelerates gain.
Breeding values rank genetic merit; gain compounds every generation (~1.8 days here).
Pathogen genogroups (LF-89-like vs EM-90-like) differ in virulence, susceptibility and vaccine fit; typing informs prognosis, treatment, tracing, surveillance and breeding.
Genetics is cross-cutting — it strengthens every phase of the framework.

This completes the book’s journey through Anticipate → Prevent → Detect → Respond → Safeguard. The appendices that follow collect an R cheat-sheet, a formula reference, the dataset catalogue, and worked exercise solutions.

16.9 References & further reading

Barría A, et al. Genetic and genomic architecture of resistance to P. salmonis in Atlantic salmon. 2020.
Lhorente JP, et al. Genetic variation in resistance to P. salmonis. PLoS ONE. 2014.
Rozas-Serri M, et al. P. salmonis genogroups LF-89-like and EM-90-like. J Fish Dis. 2017; Pathogens. 2023;12:450.
Isla A, et al. Multiplex genotyping PCR for P. salmonis. Front Microbiol. 2021;12:673216.
Mora-Salas N, et al. Method for lineage typing of P. salmonis. 2023.
Rodger H, et al. P. nova and assay cross-reactivity. Bull Eur Assoc Fish Pathol. 2025.

Files for this chapter: data/srs_challenge_families.csv and scripts/make_srs_challenge_families.R (new).