Research Article |
Corresponding author: Francesca Davoli ( francesca.davoli@isprambiente.it ) Academic editor: Klaus Henle
© 2018 Francesca Davoli, Mario Cozzo, Fabio Angeli, Claudio Groff, Ettore Randi.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Davoli F, Cozzo M, Angeli F, Groff C, Randi E (2018) Infanticide in brown bear: a case-study in the Italian Alps – Genetic identification of perpetrator and implications in small populations. Nature Conservation 25: 55-75. https://doi.org/10.3897/natureconservation.25.23776
|
Sexually Selected Infanticide (SSI) is thought of as a male reproductive strategy in social mammalian species, because females who lose cubs may quickly re-enter oestrus. SSI has rarely been documented in non-social mammals and, in brown bears, SSI has been studied mainly in an eco-ethological perspective. The authors examined the first genetically documented infanticide case which occurred in May 2015 in brown bears in Italy (Trentino, Central-Eastern Alps). The infanticide killed two cubs and their mother. Hair samples were collected from the corpses as well as saliva, through swabs on mother’s wounds, with the aim of identifying the genotype of the perpetrator. The samples were genotyped by PCR amplification of 15 autosomal microsatellite loci, following the protocol routinely used for individual bear identifications within the Interregional Action Plan for Brown Bear Conservation in the Central-Eastern Alps (PACOBACE). Reliable genotypes were obtained from the mother, cubs and putative perpetrator. The genotypes were matched with those populating the PACOBACE database and genealogies were reconstructed. Both mother and perpetrator genotypes were already present in the database. Kinship analyses confirmed mother-cubs relationships and identified the father of the cubs. In this study, for the first time, the authors used the open-source LRmix STUDIO software, designed to analyse human forensic genetic profiles, to solve a case in wildlife. Through LRmix STUDIO, those alleles that do not belong to the victims were isolated and, finally, the perpetrator was identified. This study presents a method that allows, through the application of different models, the genetic identification of the conspecific perpetrator with the highest probability. The identification of the infanticidal male is relevant for the better management and conservation of wild populations with small effective population size (Ne) and low population growth rate, especially in the case of recently established populations in human-dominated landscapes. This procedure will have predictably wide applications, supplying important data in the monitoring of small and isolated populations.
Conservation genetics; LRmix STUDIO; Low template DNA; Small population; Ursus arctos
Infanticide, the killing of dependent offspring by conspecifics, has been thought of as a component of intersexual conflicts in social mammals (
In this paper, the authors report the first observation of infanticide in brown bears documented through genetic analysis in Italy. On 10 May 2015, as part of the field and genetic long-term monitoring of the brown bear population re-introduced in Trentino, central-eastern Italian Alps (
Since an infanticidal bear can affect the growth rate of a small and isolated population (
In 2015, the brown bear population, re-introduced in central-eastern Italian Alps, extends mainly in the western part of the Trento Autonomous Province (PAT), across an area of ca. 20794 km2 (including movements of young dispersal males). The females permanently occupy a smaller area (1303 km2) entirely located within the PAT. The Extent of Occurrence is estimated as 100% minimum convex polygon, delimited by all the validated indices of presence (
This population has been continuously monitored during the last 17 years by both genetic and direct observation procedures (AA VV 2010,
In the period between 2003 and 2015, the authors collected and analysed ~7800 biological samples, of which ~7700 (hair samples, scat, saliva samples, urine and blood samples on snow) were collected through non-invasive techniques during the monitoring programmes and ~100 were of an invasive origin (tissue, hair samples, teeth and bone sample). Tissue, teeth and bone samples were taken during necropsies of animals which had died of natural causes or had been killed by traffic or poaching, while hair samples came from animals live-captured for radio-tracking studies.
During the genetic monitoring carried out in 2015, 45 different bears (24 females and 21 males) were identified of which 25 were adults: 13 females (>3 years old, reconstructed on the basis of field data and genetic pedigrees) and eight males (>4 years old). In addition to these bears, the authors considered in the population even those bears that had been sampled during 2012–2015 (the last sample collected not before than 2012) and not known as dead in 2015. Thus, the total number of individuals present in 2015 was 57 (of which 14 were adult females and 12 were adult males).
Following the infanticide case, in 2015 two types of biological samples were collected: three samples of hairs from the corpses and four saliva swabs, by swabbing the mother’s injuries, hoping to isolate the DNA of the perpetrator. The samples were preserved dry until the DNA extraction.
DNA from hairs was extracted using the ZYMO Research ZR-96 Genomic DNA™ – Tissue MiniPrep Kit (CA, U.S.A.) and DNA from swabs was extracted using the QIAGEN QIAamp® DNA Investigator Kit (Hilden, Germany), following the manufacturer’s protocol. The amplification and analysis of microsatellites were carried out by updating the protocol described by
Low genetic variability in the sampled population and small numbers of markers used in genotyping, might lead different individuals to show the same multilocus genotype. This shadow effect (
To ensure as much as possible the proper genetic reconstruction of the cubs’ pedigree, the paternity probabilities were calculated by comparing the results of two software taking into account the allelic frequencies of the population and the error rate per locus: COLONY v.2.0.5.0 (http://www.zsl.org/science/software/colony) and FRANz v.2.0.0 (http://www.bioinf.uni-leipzig.de/Software/FRANz/). COLONY implements a maximum likelihood method to assign sibship and parentage jointly, using individual multilocus genotypes at a number of co-dominant or dominant marker loci. FRANz reconstructs pedigrees (family trees) using polymorphic, co-dominant markers. [See Suppl. material
Two different statistical models were used aiming to obtain reliable results from the analysis of the Low Template-DNA (LT-DNA) from the swab samples: the “classical” biological model (
Following the biological model, the genetic profiles from the four swabs were interpreted, not individually, but in an integrated manner, by comparing the results obtained from the four independent replicates of each swab. Scientific literature describes two main approaches, both of which were applied to the evaluation of the genetic profiles of the single trace:
• The consensus method (
• The composite method (
The results from the biological model analyses were compared with the results obtained by the statistical model. There are three groups of statistical methods to evaluate the Weight-Of-Evidence from the traces through the calculation of the likelihood value (Likelihood Ratio – LR), based on different algorithms and classified as: binary models (traditional methods of calculation), semi-continuous models and continuous models (
Specifically in the case of conspecifics, the low amount of DNA mixtures can be treated as the DNA traces usually found in a crime scene. For the interpretation of the low-level complex DNA mixtures with the statistical model, an open-source software was used: LRmix STUDIO (version 2.1.3-CommunityEdition, 2013–2016 Netherlands Forensic Institute, freely available at http://lrmixstudio.org). This software is dedicated to the semi-continuous approach and explicitly accommodates for uncertainty in the DNA profile from the allelic drop-out (ADO) and drop-in (contaminations) phenomena. LRmix STUDIO estimates these quantities from the available data and uses those estimates to generate LR. LRmix STUDIO was used to compute the LR for each suspected male (reference DNA profile) and to compare the global consensus profile and the global composite profile, obtained from the comparison of each trace, with all the reference profiles. The authors performed:
• A LR calculation defining the prosecution (Hp) and the defence (Hd) hypotheses. Under each hypothesis, the authors defined the contributors (Hp: victims and 1 suspect, Hd: victims and 1 unknown), the drop-out probabilities (victims: 0.01, suspect and unknown: 0.6, afterwards replaced with the average value of drop-out, calculated over all suspects, by the drop-out estimation of the sensitivity analysis), the Pr(C) and the rare alleles frequency (the default values: 0.05 and 0.001 respectively), the allelic frequencies of the population (calculated by GenAlEx on the reference database) and the Theta correction (0.03 for small and isolated populations);
• A Sensitivity Analysis (SA) that plots the log10 LRs along with the likelihoods of the Hp and Hd. The SA, showing the variation of the LR value when 0 ≤ Pr(D) ≤ 0.99 , allows the verification of the range of the most likely values of Pr(D) (from the 5th to 95th percentile) using a Monte-Carlo simulation method (
• A Non-contributor test for better understanding the case specific LR (
Reliable genotypes were obtained from all the hair samples collected from the three corpses. The DNA profiles obtained from each victim were compared with the profiles in the reference database: 93 genotypes (9 founders and 84 offspring in 15 years) of which 45 are females and 48 are males. This comparison allowed the identification of the mother (a female called BJ1), while the cubs (one male and one female) were unknown and these were added them to the reference database, with the names of M33 and F22 (Table
Sample | cxx20 | G10M | G10P | G10X | G1D | Mu11 | Mu15 | Mu23 | Mu50 | Mu59 | G10C | G10H | G10L | Mu09 | Mu10 | |||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
M22 | 134 | 134 | 123 | 123 | 165 | 171 | 133 | 139 | 106 | 108 | 80 | 86 | 136 | 146 | 118 | 120 | 102 | 106 | 123 | 123 | 94 | 106 | 253 | 253 | 151 | 151 | 193 | 195 | 118 | 130 |
Positive Control | 134 | 134 | 123 | 123 | 165 | 171 | 133 | 139 | 106 | 108 | 80 | 86 | 136 | 146 | 118 | 120 | 102 | 106 | 123 | 123 | 94 | 106 | 253 | 253 | 151 | 151 | 193 | 195 | 118 | 130 |
BJ1 | 118 | 118 | 117 | 119 | 151 | 171 | 139 | 143 | 102 | 116 | 88 | 88 | 136 | 142 | 120 | 122 | 98 | 104 | 103 | 123 | 94 | 106 | 253 | 253 | 151 | 151 | 177 | 185 | 128 | 130 |
Victim 1 (mother) | 118 | 118 | 117 | 119 | 151 | 171 | 139 | 143 | 102 | 116 | 88 | 88 | 136 | 142 | 120 | 122 | 98 | 104 | 103 | 123 | 94 | 106 | 253 | 253 | 151 | 151 | 177 | 185 | 128 | 130 |
Victim 2(male cub – M33) | 118 | 118 | 117 | 123 | 171 | 171 | 133 | 139 | 102 | 106 | 78 | 88 | 136 | 146 | 120 | 122 | 98 | 98 | 111 | 123 | 94 | 94 | 253 | 253 | 151 | 151 | 185 | 185 | 118 | 128 |
Victim 3(female cub – F22) | 118 | 118 | 119 | 123 | 171 | 171 | 133 | 143 | 102 | 116 | 88 | 88 | 136 | 146 | 120 | 122 | 102 | 104 | 101 | 103 | 94 | 94 | 253 | 253 | 151 | 153 | 185 | 193 | 118 | 130 |
BJ1 was genetically identified as the mother of M33 and F22. Moreover, both software used for parentage analysis agree in identifying a known male, MJ4, as the father of the cubs (Table
(a) | ||||
---|---|---|---|---|
Offspring ID | Inferred Mum | Prob. Mum | Inferred Dad | Prob. Dad |
M33 | BJ1 | 1.000 | MJ4 | 1.000 |
F22 | BJ1 | 1.000 | MJ4 | 1.000 |
(b) | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Offspring | Parent 1 | Parent 2 | LOD | Posterior | Mismatches | n_f | n_m | Pair LODParent 1 | Pair LODParent 2 | Posterior Parent 1 | Posterior Parent 2 |
M33 | BJ1 | MJ4 | 2.104E+01 | 1.000 | 0 | 14 | 12 | 9.377E+00 | 9.835E+00 | 1.000 | 1.000 |
F22 | BJ1 | MJ4 | 2.298E+01 | 1.000 | 0 | 14 | 12 | 1.126E+01 | 7.903E+00 | 1.000 | 1.000 |
Due to the small quantities of DNA extracted from the swabs, only 10 loci (PIDunb=3.0×10-10 and PIDsibs=1.3×10-04) were typed. Although drop-in phenomena cannot be excluded, the presence in most of the STR loci of more than two alleles suggests a genetic mixture, produced from organic material from at least two individuals. The presence of the allelic signal Y (male-specific), appreciable both in Amelogenin and in SRY, in three out of four swabs, suggests that the contribution to the formation of the traces is derived from at least one male, in addition to the dominant genetic component originating from the blood of the mother. As a whole, for the extrapolation of genetic profiles, all the replicates of the traces were considered useful and these were all used for comparison (Table
Results of typing of the four replicates for each swab sampled on the female’s injuries.
Genetic profiles | Sexing | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
cxx20 | G10M | G10P | G10X | G1D | Mu11 | Mu15 | Mu23 | Mu50 | Mu59 | AMG | SRY | ||
I SWAB | I | 118/118 | 117/119 | 151/171 | 139/143 | 102/116 | 88/88 | 136/142 | 120/122 | 98/104 | 103/123 | 158/212 | 75/75 |
II | 118/118 | 117/119 | 151/171 | 139/143 | 102/116 | 88/88 | 136/142 | 120/122 | 84/98/104 | 103/123 | 158/212 | 75/75 | |
III | 118/118 | 117/119 | 151/171 | 139/143 | 102/116 | 88/88 | 136/142 | 120/122 | 84/98/104 | 103/123 | 158/212 | 75/75 | |
IV | 118/118 | 117/119 | 151/171 | 139/143 | 102/102 | 88/88 | 136/142 | 120/122 | 84/98/104 | 103/123 | 158/212 | 75/75 | |
Biological model results | I Consensus | 118/118 | 117/119 | 151/171 | 139/143 | 102/116 | 88/88 | 136/142 | 120/122 | 84/98/104 | 103/123 | 158/212 | 75/75 |
I Composite | 118/118 | 117/119 | 151/171 | 139/143 | 102/116 | 88/88 | 136/142 | 120/122 | 84/98/104 | 103/123 | 158/212 | 75/75 | |
II SWAB | I | 118/118 | 119/123 | 171/171 | 133/143 | 102/102 | 88/88 | 136/146 | 120/122 | 102/104 | 101/103 | 212/212 | 0 |
II | 118/118 | 119/123 | 171/171 | 133/143 | 102/116 | 88/88 | 136/146 | 120/122 | 102/104 | 101/103 | 212/212 | 0 | |
III | 118/118 | 119/123 | 171/171 | 133/143 | 102/116 | 88/88 | 136/146 | 120/122 | 102/104 | 101/103 | 212/212 | 0 | |
IV | 118/118 | 119/123 | 171/171 | 133/143 | 102/116 | 88/88 | 136/146 | 120/122 | 102/104 | 101/103 | 212/212 | 0 | |
Biological model results | II Consensus | 118/118 | 119/123 | 171/171 | 133/143 | 102/116 | 88/88 | 136/146 | 120/122 | 102/104 | 101/103 | 212/212 | 0 |
II Composite | 118/118 | 119/123 | 171/171 | 133/143 | 102/116 | 88/88 | 136/146 | 120/122 | 102/104 | 101/103 | 212/212 | 0 | |
III SWAB | I | 118/118 | 111/117/119 | 151/171 | 139/143 | 102/102 | 88/88 | 0 | 120/122 | 98/104 | 0 | 158/212 | 0 |
II | 118/134 | 111/117/119 | 151/171 | 139/143 | 102/102 | 78/88 | 0 | 120/122 | 84/98/104 | 103/103 | 158/212 | 75/75 | |
III | 118/118 | 111/117/119 | 171/171 | 139/143 | 102/102 | 78/88 | 0 | 122/122 | 84/98/104 | 103/103 | 212/212 | 75/75 | |
IV | 0 | 111/117/119 | 151/171 | 139/143 | 102/102 | 78/88 | 0 | 120/122 | 84/98/104 | 103/103 | 158/212 | 0 | |
Biological model results | III Consensus | 118/118 | 111/117/119 | 151/171 | 139/143 | 102/102 | 78/88 | 0 | 120/122 | 84/98/104 | 103/103 | 158/212 | 75/75 |
III Composite | 118/134 | 111/117/119 | 151/171 | 139/143 | 102/102 | 78/88 | 0 | 120/122 | 84/98/104 | 103/103 | 158/212 | 75/75 | |
IV SWAB | I | 118/118 | 117/123 | 171/171 | 133/139 | 102/106 | 78/88 | 136/146 | 120/122 | 98/98 | 111/123 | 158/212 | 75/75 |
II | 118/118 | 117/123 | 171/171 | 133/139 | 102/106 | 78/88 | 136/146 | 120/122 | 98/98 | 111/123 | 158/212 | 75/75 | |
III | 118/118 | 117/123 | 171/171 | 133/139 | 102/106 | 78/88 | 136/146 | 120/122 | 98/98 | 111/123 | 158/212 | 75/75 | |
IV | 118/118 | 117/123 | 171/171 | 133/139 | 102/106 | 78/88 | 136/146 | 120/122 | 98/98 | 111/123 | 158/212 | 75/75 | |
Biological model results | IV Consensus | 118/118 | 117/123 | 171/171 | 133/139 | 102/106 | 78/88 | 136/146 | 120/122 | 98/98 | 111/123 | 158/212 | 75/75 |
IV Composite | 118/118 | 117/123 | 171/171 | 133/139 | 102/106 | 78/88 | 136/146 | 120/122 | 98/98 | 111/123 | 158/212 | 75/75 |
Comparison of the results obtained using the biological model provides quantitative information regarding the degree of concordance or discordance between each genetic profile of comparison (potential infanticidal males) and the genetic profiles from the traces, based on the consensus and the composite methods (Table
Classical biological model. Biological model results obtained in terms of percentage of concordances and discordances for each suspected male.
About the results obtained using the statistical model, the Hp was tested by comparing the values of Log10(Pr(E|Hp)) and Log10(Pr(E|Hd)) for every suspect (Figure
Test for the veracity of the prosecution hypothesis. Likelihood ratio of the prosecution hypothesis (Log10(Pr(E|Hp))) versus the defense hypothesis (Log10(Pr(E|Hd))) for each suspected male. The Δ value gives an idea of the veracity of the hypothesis tested.
A comparative analysis of the multilocus genotypes of the 12 adult males considered present in the PAT during 2015 and the genetic profiles found in the analysed traces (biological model) has highlighted a variable percentage of discrepancies depending on the suspect, both in the consensus and in the composite profile (from 0.35 to 0.72). In particular, for some subjects (DG2, MJ2G1, M1, M4, M8 and M9), there is clear evidence about the absence of a genotype consistent with the alleles of the suspects on almost all of the STR loci from the traces (percentage of divergent alleles >0.60). These findings are an important support to the hypothesis of exclusion of these subjects as contributors to the genetic traces found on the swabs. As expected, alleles from MJ4, the father of the killed cubs, have been detected in the biological traces, not because the infanticidal male is the father (which is highly unlikely,
LRmix STUDIO (statistical model) estimates the likelihood and the Weight-Of-Evidence by comparing two hypotheses: the accusatory hypothesis (Hp) and the defensive hypothesis (Hd). Each one of the 12 adult males considered in the population during 2015 was tested individually. LRmix STUDIO tests what is the probability that DNA of the suspect contributing to the formation of the traces (DNA mixture extracted from swabs). The alternative hypothesis (Hd), is that the subject did not contribute to the formation of the traces. The calculation of the LR value is the result of the statistical analysis and it gives an estimate of the weight of the two hypotheses that were explored: the subject is present in the traces or the subject is absent. A high LR value indicates that Hp is much more reasonable than Hd; on the contrary, a low LR value indicates that the Hd is preferred. The LR values, obtained from the statistical analysis, range between 5.94×10-4 and 1.87 in the case of Global Consensus (Pr(D) = 0.65) and between 1.2×10-5 and 2.07×10-1 in the case of the Global Composite (Pr(D) = 0.55). The only suspect that obtains values of LR greater than 1 is M7. These values, considered with the results of the Sensitivity Analysis and Non-contributor test (see Suppl. material
The use of the statistical model is certainly preferable to the use of the classical model, as it eliminates the bias due to the presence of the father of the killed cubs between the suspected bears. Numerically, the LR value can range between 0 (absolute non-involvement) and +∞ (certain identification) and can express three consequences: (i) LR>1 means “strong support for the hypothesis of identification”; (ii) LR~1 means “neutrality” (the result of the genetic analysis does not allow support for either Hp or Hd, since it has not yielded useful results, i.e. it was inconclusive); (iii) LR<1 means “strong support for the hypothesis of exclusion”, in a manner much more accentuated as LR tends to 0; if LR=0, the non-identity between the suspect and the perpetrator can be assured.
Finally, in light of the evaluations expressed, the question can be answered: Who is the perpetrator of the killing of BJ1 and her cubs? The genetic analysis conducted on the four swabs showed the presence of a very small amount of genetic material, resulting from the contribution of more subjects, which led to considering the samples in complex analytical conditions. The genetic typing, carried out with multitube protocol procedure on the traces, allowed the authors to obtain four genetic profiles largely overlapping amongst them and, on the whole, suitable for comparisons. The comparison was carried out for each adult male considered in the bears’ population during 2015 and the genetic results were obtained from the traces, interpreting the results on the basis of both the biological model and the statistical model, in accordance with the strictest and updated protocols of interpretation, drawn from international scientific literature. The outcome of this comparison excludes the hypothesis that the genetic material of MJ4, MJ5, DG2, KJ2G2, MJ2G1, DJ1G1, M1, M3, M4, M8 and M9 is present in the analysed traces and that, therefore, these subjects may have contributed to the genetic traces; on the contrary, the overall assessment of the interpretative analysis carried out supports the hypothesis that the genetic material of M7 is present in the traces. Therefore, M7 is probably the killer of M33, F22 and BJ1.
Infanticide occurs in brown bear populations and it is an important cause of mortality, which can affect even the demographic evolution of the population (
Analogously to what happens in bear populations subject to hunting pressure (
Future monitoring actions should allow the supervision of the behaviour of infanticidal males (e.g. using radiotelemetry) and, in the case of risk of repeated infanticide, should facilitate suitable conservation actions (e.g. deterrence plans that can include some level of active and passive dissuasion activities). In small and isolated populations, in fact, behaviour that leads to the killing of cubs and adult females could lead to a further decrease in the Ne and a potential reduction in the population growth rate. Wildlife managers should be cautious when dealing with small populations of vulnerable and threatened species. The small populations, in fact, must be studied to understand their dynamics. The monitoring of litters is a fundamental tool for the management of bear populations: it has allowed the authors to genetically confirm the existence of cases of infanticide and in the future may facilitate the retrieval of information necessary to assess the impact of SSI on demographic trends. In the Italian Alps, although infanticide does not seem to be a serious problem and the population seems to be in progressive and continuous growth, it is imperative to continue to gather further information.
Mario Cozzo was founded by the Autonomous Province of Trento, Forests and Wildlife Service, Division of Large Carnivores [project R0029605]. Laboratory costs were covered by PAT. We thank Michele Bruni for reporting the first cub’s corpse. Special thanks go to Daniele Asson, Natalia Bragalanti, Luca Pedrotti, Renato Rizzoli and Paolo Zanghellini for the support.
Text S1. Parameters used for parentage analysis
Data type: statistical analysis
Text S2. Detailed results of LRmix STUDIO for each suspected male: Global Consensus (ADO 0.65)
Data type: statistical analysis
Text S3. Detailed results of LRmix STUDIO for each suspected male: Global Composite (ADO 0.55)
Data type: statistical analysis
Table S1. Detailed results of the biological model (consensus and composite)
Data type:statistical analysis