The cougar in Michigan has been considered extirpated since the 1930s (Baker, 1983) and few historical specimens exist of cougars known to have been from Michigan. We examined records at Michigan museums and gained access to 1 pre-settlement skull that might yield cougar DNA. The skull was from a Native American burial pit excavated (Foster and Hagge, 1975) in 1966 in Saginaw County in the Lower Peninsula. Its position in the burial pit, which also included the remains of four children, suggested that the skull was incorporated as part of a hide used as a robe or as a medicine bundle. This indicated to Foster and Hagge (1975) that it was an authentic artifact of a Michigan cougar rather than an imported object obtained through inter-tribal trade. Baker (1983) considered this skull the earliest cougar record in Michigan, possibly placed in the burial pit as early as pre- Columbian times. However, no datable artifacts were uncovered.

The advent of PCR has allowed researchers to better study rare, cryptic and elusive species, such as cougars, by using low copy DNA (lcDNA) sources such as scat or hairs to determine presence/absence of specific species (Höss et al., 1992; Gerloff et al., 1995; Morin et al., 2001; Parson, 2001; Ernest et al., 2002; Lucchni et al., 2002; Palomares et al., 2002). Since the intial uses of lcDNA (e.g., Pääbo, 1990; Höss et al., 1992) several pitfalls have been reported when trying to amplify lcDNA, especially when the molecular markers used are microsatellites. One of the most common problems associated with lcDNA is allelic dropout or the failure of an allele to amplify in an individual during a specific PCR reaction due to stochastic sampling of the targeted locus. An additional problem arises when the source of DNA, such as feces, contains PCR inhibitors which produce false alleles, or "misprinting." These two problems produce significant genotyping errors, with misprinting rates occuring in up to 11% of loci sampled and allelic dropout found in up to 39% of the loci analyzed using lcDNA (Creel et al., 2003). These errors are especially problematic when the molecular data are used to try and estimate census population sizes. Given the frequency of errors associated with lcDNA several methods have been developed to minimize the impact of the lcDNA errors, including such methods as running many PCR reactions per sample (Taberlet et al., 1996), using maximum likelihood methods to determine genotype reliability (Miller et al., 2002), quantifying the amount of amplifiable DNA present in a sample (Morin et al., 2001) and allowing some some samples with non-identical genotypes to be counted as identical (Creel et al., 2003). Typically, these problems are considered to be more common in nuclear genes than mitochondrial genes because of the lower frequency of nuclear genes in a cell relative to mitochondrial genes, although errors are associated with high copy DNA as well ( Jeffery et al., 2001).

Use of mtDNA reduces, but does not eliminate, the problems associated with analysis of lcDNA. The discover of numts, pieces of mtDNA translocated into the nuclear genome, has called into question the ability to use mtDNA as a tool for assesing evolutionary relationships (Thalmann et al., 2004). Numts evolve independently of the mtDNA sequence from which they are derived (Actander, 1995) and can often assume multiple different sequences within the nuclear genome (Greenwood and Pa¨a¨bo, 1999), confounding analyses of evolutionary relationships.

This study analyzed available samples from Michigan’s pre-settlement and modern-day cougars to determine genetic relationships with other North American cougars. Accordingly, we sought genetic clues as to the status and origin of Michigan cougars. Our study avoids many of the problems associated with using lcDNA by using a modification of Foran et al.’s (1997) method of determine the species responsible for producing the scat based on the size of the fragment amplified and the presence of specific restriction sites rather than an individuals genotype or base pair sequence. No variation in size of the amplified cytochrome b/D-loop fragment or in the presence of the restriction sites has been found for these species (Foran et al., 1997).

Field crews generally consisted of two or three biologists who walked dune ridges, beaches, logging roads or trails in or near large swamps used as winter yarding areas by >100 white-tailed deer (Odocoileus virginianus). Scats consistent with cougar droppings in size and shape were collected, most often within about 30 d of spring snow melt. This was usually March 10–April 10 in the Lower Peninsula and April 20–May 20 in the Upper Peninsula. Scats were wrapped in tissue, placed in plastic bags, and cooled or frozen prior to transport to the laboratory where they were stored at -20 C.

The 297 scat samples were extracted in quadruplicate from different regions of each scat using Qiagen Stool Kits (Qiagen Inc. Valencia, California) and each extraction was amplified twice. This produced 8 PCR reactions from each sample, exceeding the seven independent amplifications suggested for non-invasive samples (Taberlet et al., 1996). All of the extractions were performed in a fume hood only used for the scats and which was kept completely closed when not in use. Samples were removed from each scat with a unique razor blade, which was then discarded in a lidded sharps container approximately 15 feet away from the work area. Between working with individual samples all surfaces were wiped down with with DNA Displace (Fisher Scientific, Pittsburgh, Pennsylvania), allowed to dry, wiped down with molecular grade water and allowed to dry again. Each extraction was then amplified at a 500–1000 bp region containing portions of the cytochrome b and D-loop regions from the mtDNA (Kocher et al., 1989; Foran et al., 1997) using a modified reverse primer (5'TACACCAGTCTTGTAAACC 3'; D. Foran, pers. comm.). DNA was not quantified as it would not be possible to partition the total amount of DNA between bacterial, fungal, parasite and the animal producing the scat. Tissue samples from three known bobcats, cougars, wolves, and coyotes were extracted using Qiagen Tissue Kits (Qiagen Inc. Valencia, California) to serve as standards for comparison.

PCR was performed on each extraction in a 20 µL cocktail containing 5 µl of extracted DNA, 250 µM dNTPs, 0.5 µM fluorescently labeled dCTP (ChromaTide Alexafluor 488-7, Molecular Probes, Eugene, OR), 0.16 µM of each primer, 1x HotMaster Taq buffer (Brinkman Instruments, Inc., Westbury, New York) and 1.5 units of HotMaster Taq polymerase (Brinkman Instruments, Inc., Westbury, New York). Every PCR reaction was run with a negative control to test for contamination. Amplification was conducted on an Eppendorf MasterGradient Thermocycler (Brinkman Instruments, Inc., Westbury, New York) and consisted of an initial denaturation step of 2 min at 94 C followed by 3 cycles of 20 s at 94 C, 20 s at 54 C and 15 s at 72 C. This was followed by 35 cycles of 15 s at 94 C, 15 s at 54 C and 15 s at 72 C followed by a terminal extension step of 4 min at 72 C for the cytochrome b/D-loop. Each PCR set of scat samples was jointly run with DNA from known bobcats, cougars, coyotes and wolves with a total of 75 ng of template DNA.

Following amplification, we visualized all samples on an ABI 310 Genetic Analyzer using GeneScan Analysis 3.1.2 (Applied Biosystems Foster City, California) software with collection time set at 60 min. TAMRA 2500 (Applied Biosystems Foster City, California) was run as an internal size stand with all samples. Scat samples were identified as canid or felid based on multiple PCR products in the felid samples and the smaller size of the fragments amplified in canids (Foran et al., 1997). All samples that amplified and produced a felid banding pattern were further processed by separate restriction analysis using Rsa1 in conjunction with restriction digest of the amplified DNA from the known animals. Restriction digests followed conditions included with the restriction enzymes (New England Biolabs, Beverly, Massachusetts) and allowed to digest for 3 h. We analyzed the restriction fragments by running the samples on an ABI 310 Genetic Analyzer using GeneScan Analysis 3.1.2 software with collection time set at 60 min.

A single premolar was removed from a cougar skull found in a Native American burial mound (Michigan State University Museum Catalog Number 13178) and sectioned by the Michigan DNR. The saw was thoroughly cleaned with DNA Displace and rinsed with molecular grade water (Fisher Scientific, Pittsburgh, Pennsylvania). Prior to sectioning of the tooth swab samples were taken from the blade of the saw and the water reservoir. During sectioning of the tooth the water reservoir was filled with molecular grade water to prevent contamination. The tooth was cut just below the jaw line, cutting off the two roots associated with the tooth. Visual examination under a 1000x magnification showed that there were still cells in the pulp cavity. DNA was extracted using Qiagen Tissue Kits following Qiagen’s published protocols for tissues (Qiagen, 2000). We amplified the extracted DNA at three mtDNA loci, 16s rRNA, NADH-5 and ATPase-8 (Johnson et al., 1998; Johnson and O’Brien, 1997), which Culver et al. (2000) used to investigate the evolutionary relationship of North and South American cougars. Our PCR cocktail and thermal program were based on Culver et al.’s (2000) methods and was run for a total of 40 cycles of 30 s at 94 C, 30 s at 50 C and 60 s at 72 C following an initial denaturation of 94 C for 2 min. A terminal elongation period occurred for 4 min at 72 C. The PCR products were sequenced in both directions using Prism Big Dye Termination kits (Applied Biosystems Foster City, California) following the included instructions and sequenced on an ABI 310 automated sequencer and compared to sequences published by Culver et al. (2000).

Our method was a conservative procedure when attempting to determine the species of origin from scats. We only assumed species identity if the sizes of the amplified DNA and the products of the restriction digest matched samples from known species, reducing the impact of independently evolving numts. Two possibilities exist for numts to produce false readings. First, a felid numt could be a reduced version of the mtDNA region in which we were interested, producing a canid pattern. However, if this were the only band produced we would conclude that the animal was a canid, not the questionable cougar. Alternatively, a numt could be amplified in addition to the repeated felid mtDNA regions, producing a banding pattern that did not agree with any of the controls used and would be assumed to be neither canid nor felid. It is unlikely that numts from a canid would falsely produce a felid pattern because the numt would have to be larger than the mtDNA sequence and inserted multiple times.

It would also be possible for a numt to alter the banding pattern produced from the restriction digest. However, this would also make it unlikely that we would falsely conclude that a sample came from a cougar. Even with variation found in numts (Greenwood and Pa¨a¨bo, 1999), it is unlikely that we would see mutations producing restriction sites such that similar sized fragments were produced. Given this, and not needing to worry about allelic dropout or misprinting since we were not working with microsatellites, we are con- fident that our results accurately represent the species which produced the 12 scats we were able to identify.

With the exception of two samples from the same sites in Roscommon and Dickinson Counties, the shortest distance between locations of cougar scats was 56 km (Dickinson and Menominee Counties), with all other samples being separated by >90 km. While juvenile cougars often disperse >60 km (Scott, 1996) and adult home ranges vary from 41 km²–461 km² (Lindzey, 1984), it is unlikely that we could have found scat dropped by the same cougar (dispersing or resident) at any two locations given our limited sampling. Our restriction enzyme analysis suggests that there were at least 8 cougars in Michigan, over the three years of this study, with at least one of them of North American origin. There are several methods for more accurately estimating population size using molecular data, such as genotype mark-recapture (e.g., Creel et al., 2003). Sampling genotypes from hair pulls over a given region allows the production of density estimates with confidence intervals. Individual capture events are recorded by how often novel genotypes are found relative to previously sampled ones using traditional mark-recapture estimators (e.g., Otis et al., 1978; Seber, 1982). However, these methods still face the problems associated with lcDNA sources and thus need to be adjusted for factors such as allelic dropout and misprinting (Creel et al., 2003). The genotypic mark-recapture methods however are costly and time consuming which put them beyond the scope of this project.

The age of our scat samples is the most likely explanation for our inability to amplify the ctochrome b/D-loop region we used in our species determination; we did not use a systematic searching of a given area to obtain the freshest scats. Our sampling method was a haphazard sampling of areas and we collected all scat samples regardless of age. This collection technique resulted in our collecting many scats that had been exposed to the environment for unknown, but often, long periods of time. Thus, the majority of our scat samples had likely been exposed to several freeze-thaw cycles and extensive amounts of precipitation and UV radiation, all of which degrade DNA (Kohn et al., 1995; Foran et al., 1997). A more systematic and repeated searching of the same location over shorter periods of time would likely to produce more scats with amplifiable DNA.

One of the most contentious debates regarding cougars in Michigan is if the current animals are released pets, natural recolonization (Heist et al., 2001), or a relict population and if there is a breeding population. Our methods were unable to determine the population of origin for the cougars in Michigan. Other techniques, such as microsatellite assignment based methods, would allow more accurate determination of where the Michigan cougars originated. However, this would require an extensive database of microsatellite genotypes from multiple locations throughout the cougar’s range as well as a solid database for current or historical Michigan cougars. Similarly, it is not possible to distinguish between released cougars that originated in North America and natural recolonization or residual populations of cougars without more extensive data. If the Michigan cougar population has the same mtDNA lineage of the rest of the North American cougars, it would only be possible to determine if an animal was a released pet if it had a non-North American genotype. Pets raised from females originating from North American female breeding stock would all show the M-genotype, as would natural recolonizers and relict populations. Evidence of breeding status would require microsatellite genotypes and finding scat from both a mother and her offspring, as well as allele frequencies for the population, none of which were available for this study. While non-invasive DNA sampling makes it easier to find evidence of cryptic, rare or elusive species, it typically requires major expenditures of time and money to produce accurate estimates of population size and breeding status.

Our suggestion that at least eight animals were in Michigan over the 3 y of this study would seem to cast doubt on the suggestion that all the animals were released pets. However, in 2000, Michigan passed the Large Carnivore Act, prohibiting the ownership of large carnivores such as cougars. Individuals who owned animals prior to the passage of the act were allowed to keep their animals although they were subjected to expensive and intrusive regulations. Thus, it is possible that more individuals dumped their animals to avoid having to upgrade the facilities of their cougars to meet the new requirements. Since most pet cougars have their front claws removed it is unlikely they would still be surviving today if they had been released during the last several years. While we were able to show that cougars are present in the Michigan further work is needed. An additional study, which collects scats in a more uniform and timely fashion, would produce higher quality DNA and allow the use of microsatellite DNA to produce more detailed conclusions regarding the status of the cougar in Michigan.

Acknowledgments.—We thank N. Gagnon, J. Gregory, D. Millions, M. Rusz, A. Veselenek and M. Zuidema for aid in scat collection, P. Fredrick for the sectioning of the tooth sample. We are indebted to H. Ernest and one annonymous reviewer for valuable comments and insights on this paper. This project was supported by the Michigan Wildlife Conservancy.