Cattle breeds assignment test using STR loci:
a comparison between two different approaches.
R. Ciampolini¹, E. Ciani¹, V. Cetica¹, E. Mazzanti¹, C. Sebastiani 2, M. Biagetti 2, S. Presciuttini 3, and D. Cianci¹
¹Dipartimento di Produzioni Animali, Pisa; 2 Istituto Zooprofilattico Umbria Marche, Perugia; 3 Centro di Genetica Molecolare e Clinica, Pisa, Italy.
 

Aim of the work

A commercial problem of increasing concern is the traceability of the breed origin of individual beef cuts. Current methods, based on paper certifications, are vulnerable to fraud. We report the use of DNA profiles from STR loci for inferring the breed source of given specimens, using both our published Breed Genomic Formula (BGF) method (Ciampolini et al., 2000) and an alternative approach based on likelihood ratio statistics.

A total of 386 unrelated animals (in at least three controlled generations) belonging to six beef cattle breeds (Chianina, 67; Marchigiana, 60; Romagnola, 51; Piemontese, 60; Limousine, 67; Charolaise, 69) and 66 Holstein Friesian subjects were investigated using nineteen microsatellite markers.Two distinct assignment tests were performed.

Marker
Sequence
 Chromosome
INRA 5
(GT)13
12
INRA 6
(CA)12
3
INRA 11
(CA)8T(AC)10
1
INRA 13
(TG)20
16
INRA 16
(TG)18
27
INRA 25
(GT)16
17
INRA 27
(TG)14
27
INRA 31
(TA)26
21
INRA 32
(AC)/(CA)9T(AC)14
11
INRA 35
(TG)16
16
INRA 37
(TG)12
4
INRA 50
(AC)23
15
INRA 53
(CA)13
7
INRA 63
(AC)13
18
INRA 64
(CA)12
23
INRA 72
(TG)5(GT)13G(TG)5
4
ETH 131
(CA)19
21
ETH 152
(TG)20
5
ETH 225
(CA)18
9

Tab.1.Analyzed microsatellites and their polymor-phisms.

 

 

Fig. 1. Neighbour-joining tree of individuals from four breeds, constructed using the PHYLIP software. The paired distance matrix was computed according to Mountain and Cavalli-Sforza (1997).

The first is based on the allele-shared BGF method. A BGF is obtained by selecting the two most frequent alleles from each marker of each breed, thus composing a reference multilocus genotype. An individual to be tested is then compared with the BGFs of the breeds under evaluation, and the number of alleles it shares with each BGF is determined. The subject is assigned to the breed with which it shares the highest number of alleles. In case of tied sharing values, a second set of formulas (the Pairwise BGFs) is applied.
The second method is based on a likelihood ratio (LR) approach. For a given tested individual, the probabilities of the observed genotype at each locus are computed based on the allele frequencies calculated in each breed under evaluation, and the ratio of these probabilities (in each pair of breeds) is multiplied over all loci, in the end obtaining the likelihood ratio that each particular animal originated from one breed rather than the other. We used the assignment test implemented in the Arlequin software. Main results are presented in the tables below:
%
CHI
ROM
MAR
PIE
FRI
CHI
95.92
0.00
3.33
0.00
0.00
ROM
2.04
98.04
8.33
0.00
0.00
MAR
2.04
1.96
88.33
0.00
0.00
PIE
0.00
0.00
0.00
100.00
0.00
FRI
0.00
0.00
0.00
0.00
100.00

Tab. 2. Percentages of animals of each breed assigned to the five breed with the BGF method.

-
CHA
CHI
FRI
CHI
10-3
-
-
FRI
10-4
10-5
-
LIM
10-2
10-3
10-4

Tab. 3. Non-overlapping confidence limits (converted into probability values) of the log(LR) distributions attained by each pair of breeds.

Fig. 2. Plots of all possible pairs among the four breeds. Points represent the log-likelihood that an individual of a given breed belongs to its true breed versus the log-likelihood that it belongs to the other breed.

Conclusions

The present work establishes the feasibility of assigning individuals from four relevant breeds to their own breed using 19 microsatellites. The BGF assignment test was not completely accurate, but in the case of Chianina (a breed requiring a high level of protection against fraud), produced 100% of true positives, whereas only 1.5% of Limousine were misattributed to it. Given its simplicity, the BGF test could be used as a first-choice approach. On the other hand, the LR approach, as implemented in the Arlequin package, was very powerful in our case; however, this program is based on a critical assumption concerning the frequency of the alleles absent from a given breed, and the consequences of different assumptions should be further investigated. Future work will be devoted to optimizing the cost/benefit ratio of microsatellite-based assignment tests. In particular, a substantial reduction in the number of total typed genotypes should be achieved by using a sequential test, based on the consecutive typing of several sets of markers (say, five markers per set), while keeping sensitivity and specificity of the test constant. This procedure could eventually produce a cost-effective, routine-implemented test.