Implications for wolf monitoring and research across Europe
Whereas the obtaining wolf-dog hybridization rates remains a central issue in wolf monitoring and management, relying on non-standardized microsatellite-based analysis of non-invasively collected samples has so far hampered the comparability of regional data, resulting in a lack of over-regional, European-wide hybridization rate estimates [2]. The application of this novel panel would solve the technical issues that prevent us from obtaining data that are comparable across regions.
We found an overall low population signal in this study. Nevertheless, our results show the importance of including samples from the relevant populations. Indeed, including reference samples from wolves from Central or Northern Europe, the Iberian Peninsula and Italy and/or the Alpine region when testing for admixture in these regions is of critical importance. In contrast to microsatellites, obtaining reference data can be easily achieved through extracting genotypes from already available genome-wide SNP or sequence data.
Laboratories that have already established the Fluidigm genotyping workflow could offer genotyping services to other institutions, or provide assistance in establishing those protocols. We assume that for most national wolf monitoring programs, only one or two 96 sample array runs per year would be sufficient to screen for potential hybridization events on a routine basis, which produces consumable costs of around 800 € (without tax) per array plus a couple of working days for one staff member [24, 25, 48].
Wolves and dogs have co-existed for millennia. Even if dog genomic introgression into wolves is more common than initially appreciated in studies using a small number of markers, our results show that wolves have kept their genetic distinctiveness, in agreement with genome-wide studies [19, 20, 23]. In addition to a correct management of dogs, maintaining viable population sizes of wolves and limiting human disturbance on wolf pack structure is probably the best way to minimize the risk of hybridization. Wolves play an important ecological role and perturbations to wolf social structure by removing individuals, particularly advanced backcrosses to wolves, could in some cases be detrimental and promote further hybridization. Plans to routinely monitor hybridization in Europe should be initiated to help identify areas where actions may be directed to better control feral dogs and to promote measures that would support ecological separation of dogs and wolves. Standardized, concerted assessment of hybridization rates across Europe may serve as a basis for further research aiming at understanding regional differences in hybridization rates and degrees of dog introgression in wolf populations.
The designed 96 SNP panel is a highly discriminative new tool that could be used in routine wolf monitoring to detect wolf-dog hybrids up to third-generation backcrosses to wolves. We demonstrated a high genotyping success rate for all sample types, including different types of non-invasive samples commonly collected in monitoring practices and even museum samples, making the panel suitable for various types of studies. Moreover, the developed SNP panel is applicable at a European-wide scale, making it possible to produce comparable results of hybridization rates across the continent, as long as all the potential reference populations are included in the analyses. Extensive collection of wolf and dog reference samples is not required, as already published genotypes of wolves and dogs can be added to the analyses. The study reduces the gap between genomic research and real-world application by developing a fast and affordable method for wolf monitoring and management purposes.
Our initial SNP panel consisted of 300 wolf-dog ancestry-informative markers (AIMs) obtained from Harmoinen et al. (in prep). The SNPs were initially selected from a total of 173,662 SNPs on the CanineHD Whole-Genome BeadChip microarray (Illumina, Inc., San Diego, California, USA) which was used to genotype wolves sampled in most of their Eastern European range (Finland, Sweden, Russia, Estonia, Latvia, Poland, Belarus, Ukraine, Slovakia, Croatia, Bulgaria and Greece; n = 180) and dogs from 58 different breeds (collected in Finland, n = 352). In the study, unlinked (r2 < 0.2) data with MAF > 0.1 was used to select SNPs with the highest FST between wolves and dogs as AIMs (FST 0.67–0.86). Due to strict pruning, SNPs were evenly distributed across the 38 autosomal chromosomes. We then excluded SNPs located near another polymorphic site (minimum separation distance 100 base pairs; based on the dog genome, [49], with UCSC Genome Browser, [50]) to avoid problems in the interpretation of results and to simplify primer design (n = 63 excluded). This resulted in 237 markers, from which we selected 192 markers with the highest FST values for downstream testing using microfluidic arrays (Table S12).
SNPtype™ genotyping assays were designed for the 192 selected AIMs and tested on microfluidic 96.96 Dynamic Arrays™ (Fluidigm Corp., South San Francisco, USA) following the recommendations and testing scheme in vonThaden et al. [25, 48]. The Fluidigm platform uses chips containing integrated fluidic circuits (IFCs), harbouring nanoscale PCR reaction chambers that allow the simultaneous genotyping of 96 samples and 96 loci [51]. We chose samples with high DNA concentration (n = 92, tissue and concentrated buccal swabs; ~ 20–80 ng/μl DNA) for the initial assessment of the 192 AIMs following in silico design. Samples included wolves (n = 51), non-pedigree dogs (n = 30), known hybrids (n = 7) and three species that may be a source of DNA contaminations in non-invasively collected samples (red fox, Vulpes vulpes, n = 1; golden jackals, Canis aureus, n = 2; and red deer, Cervus elaphus, n = 1; see next section for more information on the samples). All 192 AIMs were initially run without a multiplexed pre-amplification step to exclude primer interference as a cause of potential performance failure. Results were then examined to exclude markers that either: (i) produced ambiguous genotype clusters or fluorescence for non-template controls (n = 38); or (ii) showed genotype disagreements compared to the genotypes generated with the Illumina CanineHD chip (n = 6). Subsequently, the best performing 96 SNPs were selected and tested on the same reference sample set, but now including a multiplexed pre-amplification step (specific target amplification; STA) according to the manufacturer’s protocol, which is recommended for samples with moderate DNA concentration. In subsequent runs of samples with low DNA quality and quantity, we adjusted the manufacturer’s STA protocol as indicated in vonThaden et al. ([25]; i.e., 3.2 μl instead of 1.25 μl DNA template and 18 instead of 14 PCR cycles in the STA step).
Ewalt-Evans said that means test results are unreliable. “Within each generation of back-crossing to dog,” she wrote, “after the original wolf x dog breeding, there is random shuffling of genetic material such that a subject may end up with more or less wolf content. The biggest challenge to wolf-hybrid genetic analyses are for hybrids with high ‘dog’ content. The analyses easily detect higher ‘wolf’ content.”
To do that, scientists compare genetic data from known dog breeds to genetic data collected from wolf populations in the wild. They focus on a number of known genetic variations between the two — things like Y chromosome differentiation (in males) and DNA STR (short tandem repeat) markers that have variants specific to wolves. Results are then reported back as “dog,” “wolf” or “hybrid.”
“They don’t actually test for ‘wolf’ DNA,” Addams said. “They’ll just report, ‘This dog does not match any breeds,’ which could mean it’s one of the breeds they don’t have samples for, an unusual mutt, a mutant, or a wolf, who knows? It’s possible to send a DNA sample out to a research laboratory for testing, I’m sure, but the research labs don’t (yet) have a really wide database of many, many wolves — usually they have a few individuals from local populations.”
To find out, I emailed Becky Ewalt-Evans at the UC Davis Veterinary Genetics Laboratory. They’re the lab running the test for DEEP. She said proving an exact percentage of a hybrid’s wolf ancestry is basically impossible. “Wolves and dogs are fairly closely related,” Ewalt-Evans wrote. “Fixed genetic differences (that allow determination of origin) between wolves and dogs are relatively few.”
Connecticut’s Department of Energy and Environmental Protection has ordered genetic testing for seven hybrid “wolfdogs” found in the state. But if all dogs come from wolves, can a DNA test actually tell us how much “wolf” and how much “dog” is in a hybrid?
Discriminating power of the selected SNP panel
We developed a 96 SNP panel from which 93 SNPs were finally selected based on performance (three SNPs were dropped as they had low genotyping success rate, < 0.7 across samples). The 93 selected SNPs allowed for reliable discrimination of wolves, dogs and their hybrids. This high discriminating power is due diverging allele frequencies in the wolf and dog groups, accompanied by the presence of private alleles for dogs. For all loci, alle frequencies were > 0.69 for one of the groups. While this panel was chosen to maximize the differentiation between wolves and dogs, significant differentiation between the wolf populations was detected. However, panels designed specifically to study population differentiation are available and better suited for this purpose (e.g. Illumina CanineHD chip, Affymetrix Canine SNP array or specifically designed SNP chips).
The fact that golden jackals and foxes had high amplification success and were not distinguishable from wolves requires caution. However, there are several genetic methods for differentiating these species from wolves that could be applied in routine laboratory analyses. Stronen et al. [28] have shown that only 11 microsatellite markers are sufficient to differentiate golden jackals from wolves. Even more convenient is to sequence a targeted region of mtDNA that allows to differentiate between the two species, e.g. cytochrome oxidase I (the barcoding gene), cytochrome b [29] or control region [30]. Amplifying the targeted mtDNA sequence a priori would not require much resources and could be implemented routinely for all non-invasive samples before SNP genotyping. As golden jackals were about as distinguishable from dogs as wolves were, this SNP panel could potentially also be used for detecting hybrids between these two species, albeit that would require further testing. Golden jackals have been shown to rarely hybridize with domestic dogs in the wild [30], which might be more common in the future, as golden jackals are expanding extensively throughout Europe [31], particularly if suitable mates are scarce, as seen for wolves [32]. Golden jackals and dogs have also been bred intentionally to develop a new breed (Sulimov dog) with good olfactory capabilities [33]: however, although used for narcotic detection at the Sheremetyevo Airport in Moscow, their superior olfactory skills have been questioned [34].
Although the discriminating power between wolves and dogs with this SNP panel was high (100% for F1 and F2, 99% for BC1w), and we were able to assign even third-generation backcrosses to wolves to the right category with high accuracy (89–92%), the assignment accuracy for second-generation backcrosses to wolves was slightly lower (81–82%). This hybrid category’s lower assignment accuracy is due to the fact that reliably distinguishing between second- and third-generation backcrosses is difficult; most of the incorrectly assigned hybrids from this category were assigned as third-generation hybrids (the remaining two or three individuals were assigned as first-generation hybrids). However, unless the criteria for defining a hybrid requires the distinction between these two hybrid categories, the lower assignment accuracy in this category is not relevant for management as the individuals would be anyway categorized as advanced hybrids. The software could not assign any third-generation backcrosses to dogs into the right category, possibly because the analysis was hampered by the large variation in allele frequencies in dogs. The amount of genetic variation is higher in dogs than in wolves when all dogs are combined, but variation within each single breed is less than that found in wolves [35]. In this SNP set, variation in dogs is emphasized by the fact that the SNPs are selected from the Illumina CanineHD Chip, the SNPs of which are in turn selected from the dog reference genome. Somewhat higher variation in dogs, wolves and different hybrid categories can be observed in a study testing 100 SNPs chosen from the Affymetrix Canine Mapping SNP Array 2.0, with SNPs also originally chosen from the dog reference genome [21]. Here we attempted to develop an efficient and reliable genotyping method that would allow to detect wild wolf-dog hybrids during routine wolf monitoring based on samples with low DNA quality and quantity. Therefore, reliable discrimination of dogs from backcrosses to dogs falls beyond the scope of this study.