Traceability of mussel (Mytilus chilensis) in Southern Chile using microsatellite molecular markers and assignment algorithms: Exploratory survey Presented by Carmen, Recheta, and Soliven

Outline ● Background ○ Marine mussels ○ Genetic traceability and geographic origins ● Methods ● Results ● Discussion ● Conclusion

Marine Mussels - One of the most cultivated and marketed bivalves with the Mytilus genus used widely in prepared food products In Southern Chile, mussel culture is an important economic activity yielding 14.9% of the world’s Mytilidae aquaculture production in 2010 (FAO,2012). The international seafood trade has adopted the food chain or “from farm to fork” concept in terms of standards and regulations regarding food quality

3 Levels of Seafood Traceability: ● Identification of the species ● Geographical origin determination ● Supply chain tracking and tracing

3 Levels of Seafood Traceability: ● Identification of the species ● Geographical origin determination ● Supply chain tracking and tracing DNA-based methods

Genetic Traceability - DNA-based methods can verify the accuracy of traditional identification methods such as product labeling - Determine species present in the product (DNA barcoding) - Identification of breeds, local populations - Employing different DNA markers - Coding regions - Non-coding regions (more commonly used)

First level traceability: genetic species identification ● Employed to investigate commercial fraud by species substitution ● PCR-based assays allows for identification of various Mytilus species ○ RAPD (randomly amplified polymorphic DNA) ○ FINS (forensically relevant nucleotide sequencing) ○ AFLP (amplified fragment length polymorphisms

Second level of traceability: Geographical Origin ● Cannot be established merely by identifying the species due to worldwide distribution of Mytilus ● Equivalent to identifying its biological population (Ogden, 2008) ● Various chemical markers have been used (trace elements coupled with volatile compound analysis) ● DNA-based methods to study genetic diversity and population structure ○ Allozymes ○ Mitochondrial COI gene sequence ○ SNPs ○ Microsatellites

Assignment methods ● Assignment methods use genetic information to ascertain population (predefined) or cluster (non-predefined) membership of individuals ● Frequentist or likelihood methods ○ Frequentist - statistical hypothesis testing, giving a p value derived from a frequency distribution ○ Likelihood - assumes that observed data arises from a probabilistic model with unknown parameters ● Bayesian approach ○ Derives the posterior distribution, taking into account prior probability and a likelihood function derived from the observed data, allowing for incorporation of existing knowledge into data analysis

Objective To evaluate the potential of microsatellite markers combined with allocation algorithms for assigning Mytilus individuals from southern Chile to their geographical origin

MATERIALS & METHODS

Mussel Sample Collection and SSR Genotyping - Samples of mussels (n=50 by location) were collected in Southern Chile in six sites (inc. 1 wild population and 5 seed collection centers) - DNA extraction by phenol-chloroform method - Amplification of 9 SSR loci - Genotyping: polyacrylamide gels (6%) with silver staining

Data Analysis ● Micro-Checker - used to test for the presence of null alleles, stuttering and large allele dropout ● GENEPOP 4.0.10 - used to test for Hardy Weinberg equilibrium and estimate Wright’s fixation index (FST )

Data Analysis Four assignment methods to assign reference population as possible origin of individuals: ● Genetic distance-based criterion ● Frequency-based method ● Bayesian-based method - operating in an “unsupervised” mode ● Bayesian-based method - operating in a “supervised” mode

Data Analysis ● GeneClass2 - used for genetic distance and allele frequency - assigning a population of origin for each individual ● Structure 2.3.3 - used for Bayesian method - assignment of individuals were conducted under the mixture model - 50,000 initial burn-in, 100,000 Markov chain iterations - K = 6

Data Analysis Evaluation of the performance of all four assignment methods: ● Sensitivity (S) = # of individuals correctly assigned to their original location/ total # of individuals sampled from that location - Reflects how good a test is at correctly assigning individuals to a location ● Specificity (E) = # of individuals properly excluded from the population/total # of individuals who do not belong to the population - Reflects how good the test is at correctly excluding individuals who do not belong to that location ● Average probability assignment score (AP) - average of the likelihood of each successful re-assignment to the respective location ● Likelihood ratio (LR +) = s/(1-E) - How many times more (or less) likely it is that individuals belonging to a location will be assigned to this location as compared to individuals belonging to another location

RESULTS & DISCUSSION

Results and Discussion ● All 9 loci are polymorphic and a total of 68 alleles are observed ● Global FST - genetic differentiation between the 6 sites was 0.042 - Only 4.2% of total allele frequency variance lies among sample sites and is highly significant (P<0.001) - 95.8% of the variance between allele frequencies is explained by variation within sites - In other words, the groups (as defined by sampling sites) are very similar to one another

Results and Discussion: Assignment performance

35% Accuracy for Bayesian method without prior information

50.7% Accuracy for distance based method

50.7% Accuracy for frequency based method

95.3% Accuracy for Bayesian method with prior information

Results and Discussion Power of assignment depends closely on the level of population structure In this study, global Fst is low at 0.042

Results and Discussion The study performed better than: The study performed poorly compared to: Blue marlin between more separated locations (41% accuracy) Golden humped tench (64-92% accuracy) Based on literature, assignment accuracy for farmed populations is better than those found in wild populations (likely due to artificial selection and gene flow being less likely between farms)

CONCLUSION & RECOMMENDATION

Conclusion ● The frequency-based algorithm showed the best performance in matching Mytilus individuals from Southern Chile to their geographical origin - 50% of correct assignment in a challenging scenario with low genetic differentiation among locations (Global Fst = 0.042)

Recommendation ● Increase assignment performance by increasing the number of microsatellite loci and/or adding SNPs

THANK YOU!