River Otter Population Size Estimation using Noninvasive Latrine Surveys Author(s) :Rebecca A. Mowry, Matthew E. Gompper, Jeff Beringer, Lori S. Eggert Source: Journal of Wildlife Management, 75(7):1625-1636. 2011. Published By: The Wildlife Society-2 URL: http://www.bioone.org/doi/full/10.1002/jwmg.193

BioOne (www.bioone.org) is a a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research.

The Journal of Wildlife Management 75(7):1625–1636; 2011; DOI: 10.1002/jwmg.193

Research Article

River Otter Population Size Estimation Using Noninvasive Latrine Surveys REBECCA A. MOWRY,1 Department of Fisheries and Wildlife Science, University of Missouri, 303 Anheuser-Busch Natural Resources Building, Columbia, MO 65211, USA MATTHEW E. GOMPPER, Department of Fisheries and Wildlife Science, University of Missouri, 303 Anheuser-Busch Natural Resources Building, Columbia, MO 65211, USA JEFF BERINGER, Resource Science Center, Missouri Department of Conservation, 1110 S. College Avenue, Columbia, MO 65201, USA LORI S. EGGERT, Division of Biological Sciences, University of Missouri, 226 Tucker Hall, Columbia, MO 65211, USA

ABSTRACT Across much of North America, river otter (Lontra canadensis) populations were extirpated or

greatly reduced by the early 20th century. More recently, reintroductions have resulted in restored populations and the recommencement of managed trapping. Perhaps the best example of these river otter reintroductions occurred in Missouri, regarded as one of the most successful carnivore recovery programs in history. However, abundance estimates for river otter populations are difficult to obtain and often contentious when used to underpin management activities. We assessed the value of latrine site monitoring as a mechanism for quantifying river otter abundance. Analyses of fecal DNA to identify individual animals may result in an improved population estimate and have been used for a variety of mammal species. We optimized laboratory protocols, redesigned existing microsatellite primers, and calculated genotyping error rates to enhance genotyping success for a large quantity of river otter scat samples. We also developed a method for molecular sexing. We then extracted DNA from 1,421 scat samples and anal sac secretions (anal jelly) collected during latrine site counts along 22–34-km stretches representing 8–77% of 8 rivers in southern Missouri in 2009. Error rates were low for the redesigned microsatellites. We obtained genotypes at 7–10 microsatellite loci for 24% of samples, observing highest success for anal jelly samples (71%) and lowest for fresh samples (collected within 1 day of defecation). We identified 63 otters (41 M, 22 F) in the 8 rivers, ranging from 2 to 14 otters per river. Analyses using program CAPWIRE resulted in population estimates similar to the minimum genotyping estimate. Density estimates averaged 0.24 otters/km. We used linear regression to develop and contrast models predicting population size based on latrine site and scat count indices, which are easily collected in the field. Population size was best predicted by a combination of scats per latrine and latrines per kilometer. Our results provide methodological approaches to guide wildlife managers seeking to initiate similar river otter fecal genotyping studies, as well as to estimate and monitor river otter population sizes. ß 2011 The Wildlife Society. KEY WORDS fecal DNA, Lontra canadensis, microsatellites, Missouri, Nearctic river otter, noninvasive genetic sampling.

River otters (Lontra canadensis) are apex predators in aquatic ecosystems across North America. By the early 20th century, populations of this species had been extirpated or greatly reduced from overharvesting, prompting numerous reintroduction events over the last 30 years (Raesly 2001). Monitoring the status of these reintroduced populations has proven difficult due to the low density and elusive nature of otters (Melquist et al. 2003). This difficulty, in combination with human-otter conflicts and the reinstitution of legal harvests in some regions, has further increased the need for the development of accurate, efficient methods of otter population size assessment to guide management activities. Received: 15 July 2010; Accepted: 13 February 2011; Published: 26 July 2011 1

E-mail: [email protected]

Mowry et al.  River Otter Population Survey

Among carnivore reintroductions, the Missouri river otter restoration project has been widely regarded as the most successful (Breitenmoser et al. 2001). The river otter was believed extirpated from the state by the mid-1930s (Bennitt and Nagel 1937). In 1982, the Missouri Department of Conservation (MDC) initiated recovery efforts, translocating 845 otters to 43 sites statewide over the next 10 years (Hamilton 1998). In 1996, in response to high population estimates obtained from models based on the survival rates of the translocated otters (Erickson and McCullough 1987), as well as public speculation that otters were impacting some wild and farm pond fish populations, MDC initiated a statewide trapping season that resulted in a harvest of 1,054 otters. The Missouri Department of Conservation then began utilizing samples obtained from harvested otters to estimate population sizes based on reproductive rates and age and sex structure. These population analyses (e.g., Gallagher 1625

1999) have produced inconsistent results, and thus Missouri’s early trapping seasons were controversial (Goedeke and Rikoon 2008). The abundance of otters is a critical information gap that needs to be filled to better address such controversies, both in Missouri and elsewhere. Scat samples (including fecal samples and anal sac secretions or jelly) of river otters are easy to find due to the exposed nature of the communal latrines, the tendency for multiple otters to use a latrine, and the general restriction of the animal to the immediate banks of the river transect. Many studies have capitalized on the value of latrines to monitor otter occupancy and activity (e.g., Guter et al. 2008, Jeffress et al. 2010). Since 2001, MDC has been counting latrine sites along Ozark rivers (Roubidoux Creek, Big Piney River, West Piney Creek, Niangua River, Osage Fork of the Gasconade River, Current River, Courtois Creek, and Maries River) to estimate relative otter densities (Fig. 1). River otters use latrine sites for defecation and intraspecific communication, with social groups typically using sites together and returning to the same sites throughout a season, year, or lifetime (Melquist and Hornocker 1983, Macdonald and Mason 1987, Gallant et al. 2007). However, abundance estimates based on counts of latrine sites may not accurately represent river otter populations, because use of latrine sites may vary both seasonally and yearly, the number of latrines tends to plateau as a result of multiple otters using the same latrines (Gallant et al. 2007), and individual otters may use the same latrine multiple times during a sampling period (Ben-David et al. 2005). Thus there is a need to test the value of latrine counts for estimating river otter populations (Guter et al. 2008, Calzada et al. 2010). Analyses of fecal DNA to identify individual animals may facilitate an improved population estimate and have been

used to survey a variety of mammal species (e.g., Tuyttens et al. 2001, Eggert et al. 2003, Kays et al. 2008). However, fecal DNA is typically degraded and exists in small quantities compared to DNA from blood or tissue samples, making extraction and polymerase chain reaction (PCR) amplification problematic (Schwartz et al. 2006). Novel laboratory techniques have been developed to alleviate these problems, and field work has emphasized the importance of obtaining the freshest of samples to reduce the extent of DNA degradation. However, extracting and genotyping DNA from otter (Lontra and Lutra spp.) scat remains notably problematic. Dallas et al. (2003) hypothesized that Eurasian otter (Lutra lutra) DNA degrades at a faster rate than DNA extracted from feces of other carnivores, and Prigioni et al. (2006) suggested the small size of otter scat may contribute to extraction problems. Genotyping success may also be affected by the humid environments typical of the streamside and riparian habitats used by otters (Farrell et al. 2000). Our objective was to use a molecular approach to estimate the population size of river otters and assess the value of scat and latrine counts for surveying otters. We predicted that different types or ages of scat may affect genotyping success, and that river otter scat counts at latrines may provide an accurate estimate of abundance.

STUDY AREA Field collection of scat samples occurred between 6 January and 23 April 2009 along stretches of 8 rivers in south-central Missouri (Fig. 1). These rivers (transect length [km] and proportion of total river length [%] in parentheses) were Big Piney (23.5 km, 13%), Courtois (22.4 km, 40%), Current (27.4 km, 8%), Maries (27.2 km, 16%), Niangua (29.0 km,

Figure 1. Study area and 8 rivers we sampled for river otter scat in winter and spring of 2009 in Missouri, USA. Small circles delineate approximate latrine site locations. Large circles and star represent locations of major regional cities. 1626

The Journal of Wildlife Management  75(7)

13%), Osage Fork of the Gasconade (31.7 km, 35%), Roubidoux (34.4 km, 39%), and West Piney (24.8 km, 77%). We chose these rivers due to their high recreational and sport fishing value, accessibility for sampling, variable estimated otter densities in past years, and liberal trapping regulations, which collectively made them particularly relevant for determining population size for management decisions. In addition, in 2003 MDC operationally removed otters from the Roubidoux River, resulting in a substantial decrease in scat counts and heightened interest in the subsequent population processes occurring there.

METHODS Sample Collection River otters have been shown to decrease movement through their home ranges during winter (Gallant et al. 2007), maximizing the likelihood of system closure during this period. In addition, river otters increase their use of latrine sites for scent-marking during the breeding season (Dec– Apr; Hamilton and Eadie 1964, Stevens and Serfass 2008), and cold conditions can increase genotyping success rates (Arrendal et al. 2007). We selected 31 March as the cutoff between winter and spring. We sampled all rivers and sections during the winter except the second section of the Maries. Due to adverse weather conditions, we were not able to collect a winter sample from that region until 3 April. We divided each river into two sections of approximately equal length. A team of technicians scouted both banks of each river, marking latrine sites and clearing all scat. After 6 days, we returned to each latrine site and collected all scats. We marked samples estimated to have been deposited within 1 day as ‘‘fresh’’ and samples deposited between 1 and 6 days previous as ‘‘old.’’ We based these classifications on moisture

content, appearance, and odor, and we acknowledge the possibility of overlap and miscategorization. We recorded anal jelly (anal sac secretions) samples separately and did not categorize them as fresh or old. We then collected each scat sample in a separate sealable plastic bag. Upon returning to the field station, we stored all samples at 208 C. We sampled the Courtois and Current rivers again on day 12 to allow simple mark-recapture estimations for comparison with our CAPWIRE estimates (data not shown), selecting these rivers in response to positive weather forecasts and because they represented examples of both large and small sample sizes (Table 1). Also, we sampled the Big Piney, West Piney, and Roubidoux rivers in both winter and spring to assess seasonal changes in population size. We also obtained matched fecal and tissue samples from 34 river otters harvested in Missouri during the 2007–2008 trapping season. We removed fecal samples from carcasses and extracted them once per day for up to 7 days to test for differences in DNA extraction success for different-aged scats. We refrigerated samples between sampling days. Optimizing Microsatellite Loci and Multiplex PCR We selected 10 microsatellite loci identified by Beheler et al. (2004, 2005), choosing those with no obvious deviance from Hardy–Weinberg equilibrium, high allelic diversity, and low to no frequency of null alleles. We designed new primers for 9 of the 10 loci to amplify shorter fragments of target DNA, amplifying less of the flanking region surrounding the repeat region (Table 2). R. Crowhurst performed DNA extractions from the 34 matched scat and tissue samples in the lab of J. Koppelman (MDC) using DNeasy Blood and Tissue Kits and QIAamp Mini Stool Kits (QIAGEN, Valencia, CA), using the manufacturer’s protocols. Using DNA extracted from tissues, we tested each microsatellite locus individually along an

Table 1. Locations and sampling dates for river otter scat samples collected in winter and spring, 2009, in Missouri, USA. We show genotyping success rates (proportion of genotypes which were complete for 7 loci), and total scat samples collected (in italics) for each river and time period. For rivers sampled twice within a season (e.g., Courtois, Current), we give both scat collection dates. Sampling dates River Big Piney Courtois Current Maries Niangua Osage Fork Roubidoux West Piney

Winter

Section

Winter

Spring

6 days

12 days

01 02 01 02 01 02 01 02 01 02 01 02 01 02 01 02

22 Jan 19 Feb 23 Mar 24 Mar, 3 Apra 6 Mar, 12 Mar 5 Mar, 11 Mar 26 Mar NA 17 Feb 23 Feb 9 Mar 16 Mar 12 Jan, 21 Jana 14 Janb, 20 Jan 2 Feb 18 Feb

2 Apr 5 Apr NA NA NA NA NA 14 Apr NA NA NA NA 15 Apr 22 Apr 6 Apr 7 Apr

0.308, 13 0.500, 20 0.400, 15 0.375, 8 0.342, 38 0.250, 76 1.000, 4 NA 0.027, 37 0.560, 9 0.219, 105 0.356, 216 0.111, 18 0.550, 20 —, 0 0.333, 18

NA NA NA 0.368, 19 0.304, 56 0.290, 69 NA NA NA NA NA NA 0.077, 13 NA NA NA

Spring—6 days 0.273, 22 0.211, 133 NA NA NA NA NA 0.167, 24 NA NA NA NA 0.145, 131 0.144, 284 0.308, 13 0.184, 49

a

Due to weather, the re-sampling for these sections was delayed by 3 days. Although this places the second sampling of section 2 of the Courtois in the spring, it was analyzed as a winter session. b Eleven scat samples were also collected during the initial winter survey (0 days), with a genotyping success rate of 0.182. Mowry et al.  River Otter Population Survey

1627

Table 2. Microsatellites from Beheler et al. (2004, 2005) used for genotyping river otter scat samples collected in winter and spring of 2009 in Missouri, USA. Loci ending in ‘‘R’’ or ‘‘R2’’ indicate primers that we redesigned for shorter product lengths, expressed in base pairs (bp). For error testing, we performed polymerase chain reactions (PCRs) at the optimal annealing temperature (AT) for each locus, but we performed all PCRs at 598 C during multiplexing of field-collected scat samples. F ¼ forward and R ¼ reverse. Locus RIO01R2 RIO02R RIO04R RIO06R RIO07R RIO08R RIO11 RIO13R RIO15R RIO16R

a

Primer sequence

PCR product (bp)

Optimal AT (8C)

F: Ned-TGAGGTATGGATAGAAGATTGATGA R: GCTTGACCTTGAGCAACTTACTT F: Vic-TAGAGTGGGGCGCCTAAGTT R: TTACTCGCCAATGGTTCAGC F: Pet-TCTGCCTTTTCAAATTCTCCA R: CCCTTTTCTCCCTTTTCTCTC F: Ned-TCCTGTTTCACAAAATCAAACAA R: AAAGACCAATAGTTCATCCAGTTC F: Fam-AAGCACTTCCAGATATCAGTTGC R: CCGCCTCCCTGTTAGAAGTT F: Vic-TCCTGAGGCATAAGGAAGACA R: ACTTGCCTGCTGACATTGAA F: Fam-TCTTCCACTTTTCAATTTAGGTA R: GCCCAAAGGTTCACTATCAAG F: Fam-GCACATGGGCTTTTATGAAGAa R: GCACACGTGGTAAGATGAGC F: Ned-CTGACCCAAAATGAATAACAGAA R: TTCTGCTTGGTTCAGTGCAT F: Vic-GCCCGTGGTCACTTTACCTa R: CACAGTAGAGGGACATTTGCAC

146–154

59

117–135

59

98–116

59

126–138

59

87–101

56

104–114

59

150–160

56

144–168

59

137–141

59

149–161

59

We switched labels for field sample genotyping to condense loci into 2 multiplex reactions while preventing overlaps in allele ranges.

annealing temperature gradient to determine the optimal annealing temperature. We performed all PCRs in a hood that was decontaminated with ultraviolet light between uses, and we used aerosol barrier pipet tips to prevent contamination. We performed PCRs in 25-ml volumes containing 1 PCR Gold buffer (Applied Biosystems, Foster City, CA), 2.0 mM deoxyribonucleotide triphosphates (dNTPs), 0.4 mM each unlabeled forward and reverse primers, 0.8 bovine serum albumin (BSA), 2.0 mM MgCl2 solution, 0.5 units AmpliTaq Goldß DNA polymerase (Applied Biosystems), and 1.0 ml (15–50 ng) DNA from tissue of one harvested river otter. The PCR profile included an initial cycle of 958 C for 10 min; followed by 35 cycles of denaturation at 958 C for 1 min, a variable annealing temperature gradient (53–608 C) for 1 min, and primer extension at 728 C for 1 min; followed by a final extension cycle of 728 C for 10 min. With each reaction, we included a negative control to detect contamination. We then tested all loci for polymorphism on DNA extracted from the tissues of 7 harvested otters. All loci amplified well at an annealing temperature of 598 C. We designed 2 multiplex reactions of 5 loci labeled with fluorescent dyes (Table 2) for amplifying and genotyping the field samples. We performed PCRs in 10-ml volumes containing 5.0 ml Master multiplex mix (QIAGEN), 0.5 mM diluted primer mix, 0.8 BSA, and 1.2 ml fecal DNA extract. The PCR profile consisted of an initial cycle of 958 C for 15 min; followed by 40 cycles of denaturation at 948 C for 0.5 min, primer annealing at 598 C for 1.5 min, and primer extension at 728 C for 1 min; and a final extension cycle at 608 C for 30 min. With each reaction, we included a positive control to standardize allele scoring and a negative control to detect contamination. 1628

We amplified DNA from the 34 matched scat and tissue samples individually for each locus, combining PCR products with different fluorescent labels for fragment analysis (Table 2). Fragment analysis was performed at the University of Missouri DNA Core Facility in a 3730 96-capillary DNA Analyzer with Liz 600 size standard (Applied Biosystems). We analyzed results using GeneMarkerTM AFLP/ Genotyping Software (Softgenetics LLC, State College, PA) and assigned genotypes manually. We computed rates of successful amplification, allelic dropout, and false alleles across time (scat age 0–7 days) and among microsatellite loci for the matched scat and tissue samples. We calculated rates of allelic dropout and false alleles by dividing the number of amplifications with these errors by the total number of genotypes (Broquet and Petit 2004). We tested for significant deviations from heterozygosity values expected under Hardy–Weinberg equilibrium and for linkage disequilibrium in GENEPOP (Raymond and Rousset 1995). We also used the program Prob_id5 to calculate the probability of identity (PID) and the PID for randomly sampled siblings (PIDsib) to determine the power of the set of microsatellite loci to differentiate individuals (Paetkau and Strobeck 1994, Waits et al. 2001). DNA Extraction, Genotyping, and Sexing of Field Samples We extracted DNA from the field samples in a separate extraction room using equipment and supplies dedicated to noninvasively collected samples. We selected approximately 180 mg of scat using either razorblades or forceps, choosing pieces of scat from various areas of each sample, especially the ends (Fike et al. 2004, Stenglein et al. 2010). To increase DNA yields, we followed the extraction protocol The Journal of Wildlife Management  75(7)

recommended in the QIAamp Mini Stool Kit (QIAGEN) for isolation of DNA from stool for human DNA analysis, with the following modifications: 1) after addition of the Inhibitexß tablets, we centrifuged samples for 5 min instead of 3 min and 2) we extended the incubation period for the final elution from 1 min to 10 min. For every 49 samples (one QIAGEN kit), we included a negative extraction to control for reagent contamination. We stored extractions and the remainder of the scat immediately at 208 C. We tested each extraction using 2 microsatellite loci (RIO07R and RIO16R) that yielded small and large fragment sizes, respectively, and exhibited high amplification success rates during error testing (Table 3). We used the PCR protocol for individual loci, increasing the number of cycles to 45. We then repeated PCRs at all 10 microsatellite loci for samples that amplified at 1 or both of the screening loci using the multiplex protocol. Each set of reactions included a positive control to standardize allele scoring and a negative control to detect contamination. To generate consensus genotypes, we used the comparative method (Frantz et al. 2003, Hansen et al. 2008), confirming heterozygotes after 2 matching genotypes and homozygotes after 3. Occasionally, we repeated genotyping for up to 5 PCR runs to confirm a genotype. The same researcher assigned all genotypes to avoid bias. We discarded samples that failed to generate consensus genotypes across 7 loci from further analysis (based on PIDsib calculations; see Results Section). Following the elimination of failed samples, we compared genotypes manually for identification of unique individuals and recognition of recaptures. We redesigned primers originally designed to amplify the male-specific SRY gene of the Eurasian otter (Dallas et al. 2000) to amplify the Nearctic river otter SRY gene (SRY2F: 50 -GAGAATCCCCAAATGCAAAA-30 and SRY2R: 50 CTGTATTCTCTGCGCCTCCT-30 ) resulting in a 111base pair fragment. We used this primer set in conjunction with primers designed for Eurasian otters (using primers ZFXYRb [Mucci and Randi 2007] and P1-5EZ [Aasen and Medrano 1990]) to amplify the zinc-finger protein gene (ZFX/ZFY) in both males and females. Combining these 2 methods resulted in amplification of the SRY gene in males, as well as amplification of the larger ZFX/ZFY gene in both sexes to confirm positive amplification and eliminate

false female calls (e.g., electrophoresis of fragments reveals 2 bands for males, and 1 band for females). We performed sexing PCRs in 25-ml volumes, using the same protocol for primer redesign and optimization, except that we increased the number of cycles to 50, used an annealing temperature of 578 C, and used 3.0 ml of DNA extract. Each reaction included a negative control to detect contamination and 2 positive controls (DNA from tissue samples of a known male and female) to confirm successful PCR amplification. We confirmed females after 3 positive PCR runs showing amplification of the ZFX/ZFY fragment only, and we confirmed males after 2 positive PCR runs showing amplification of both fragments. Population Estimation We determined the minimum population size for each river by counting the number of unique genotypes. We used GENEPOP to test for deviations from expected heterozygosity values under Hardy–Weinberg equilibrium and for linkage disequilibrium, HP-RARE (Kalinowski 2005) to calculate allelic diversity corrected for sample size, and ARLEQUIN 3.11 (Excoffier et al. 2005) to calculate observed and expected heterozygosities for each population. We used the computer program CAPWIRE (Miller et al. 2005, Petit and Valiere 2006, Arrendal et al. 2007) to estimate abundance in the entire river segment (i.e., not each individual section) based on a single sampling session. Using the program, we conducted likelihood ratio tests to determine the presence of capture heterogeneity. Where we confirmed heterogeneity, we used the Two Innate Rates Model (TIRM) to estimate population size. If we did not confirm capture heterogeneity, we used the Even Capture Probability Model (ECM). When we sampled rivers in both winter and spring, we calculated population size separately for each season. We used an information-theoretic approach to contrast the performance of a series of relative abundance indices to predict river otter population size. Because the models needed to reflect indices that could be easily collected during field sampling, we included the following variables in the models (Table 4): number of active latrines (latperkm), total scat samples (scatperkm), average scats per latrine (scatperlat), number of anal jellies (jellyperkm), number of fresh

Table 3. Results of error testing from matched river otter scat and tissue samples collected in the 2007–2008 trapping season in Missouri, USA. We provide amplification success rates by locus for scat and tissue samples. We give errors as a percentage of total successful amplifications (polymerase chain reactions that could be assigned a genotype) and include allelic dropout and false alleles. Locus RIO01R2 RIO02R RIO04R RIO06R RIO07R RIO08R RIO11 RIO13R RIO15R RIO16R Mean

Number of alleles

Success rate (tissue)

Success rate (scat)

Total errors

Allelic dropout

False alleles

7 7 6 4 7 7 7 5 3 4 5.7

0.882 1.000 0.941 0.588 0.971 1.000 1.000 0.794 0.882 0.882 0.894

0.620 0.564 0.570 0.324 0.682 0.665 0.553 0.570 0.698 0.721 0.596

0.092 0.044 0.030 0.013 0.032 0.085 0.023 0.031 0.110 0.094 0.059

0.028 0.037 0.022 0.000 0.026 0.052 0.023 0.023 0.045 0.037 0.031

0.064 0.007 0.008 0.013 0.006 0.033 0.000 0.008 0.065 0.057 0.028

Mowry et al.  River Otter Population Survey

1629

Table 4. Model sets and rankings for the 8 a priori hypotheses predicting population size of river otters in Missouri, USA during winter and spring of 2009. For each model we provide the number of parameters (K), Akaike’s Information Criterion corrected for small sample size (AICc), the difference between the AICc value of the model and the best fit model (DAIC), the Akaike weight of the model (wi), and the coefficient of determination (r2). Scatperlat ¼ scats per latrine, latperkm ¼ latrines per kilometer, jellyperkm ¼ anal sac secretions (jelly) per kilometer, freshperkm ¼ fresh (1–6 days) per kilometer. Hypothesis H8 H5 H7 H1 Hglobal H2 H6 H4 H3 Inter

Description

K

AICc

DAIC

wi

r2

Scatperlat þ latperkm Scatperlat Scatperlat þ jellyperkm Latperkm All variables Scatperkm Jellyperkm þ freshperkm þ oldperkm Freshperkm Jellyperkm Intercept only

3 2 3 2 7 2 4 2 2 1

123.68 120.99 118.44 116.85 114.26 109.15 104.87 101.74 92.67 87.14

0.00 2.17 4.71 6.31 8.90 14.01 18.29 21.42 30.49 36.54

0.67 0.23 0.06 0.03 0.01 0.00 0.00 0.00 0.00 0.00

0.76 0.73 0.71 0.68 0.75 0.58 0.56 0.44 0.22 NA

(1.0 and the weight of data points with lower than average genotyping success was between 0 and 1.0). We generated linear models for all hypotheses, and we used Akaike’s Information Criterion corrected for small sample size (AICc; Burnham and Anderson 2002) to evaluate the relative support of each model. Following identification of the model with the highest Akaike weight (wi), we applied the resulting equation to the removed outlier to determine the ability of the model to predict population size of that river section. We also tested 20 random combinations of river segments to evaluate the predictive power of the 2 top models at variable sample sizes.

(RIO15R), with a multi-locus average of 0.059. Rates of allelic dropout and false alleles were similar (0.028 and 0.031, respectively). Analysis using GENEPOP indicated that the observed genotypes did not deviate from those expected under Hardy–Weinberg equilibrium, and there was no linkage disequilibrium among loci. Multilocus PID ¼ 4.33  1014, and PIDsib ¼ 2.25  105. Using PID and PIDsib values, we determined that 7 loci were needed to differentiate siblings (PIDsib ¼ 5.00  104). For fecal samples collected from the 34 harvested river otters, the freshest samples (0 days) yielded the lowest amplification success rates and the highest error rates (0.491 and 0.126, respectively; Fig. 2). Amplification success increased steadily thereafter, peaking at day 5 (0.760). Error rates decreased beyond day 0 and showed a slight increase following day 3. Genotyping of Field Samples and Population Estimation We collected 1,421 samples across all rivers (104 anal jelly samples, 375 fresh, 930 old, 12 unknown fresh or old). Overall genotyping success rate of field samples across all rivers, seasons, and scat types was 24% (based on the number of samples for which we assigned multi-locus genotypes at 7 loci). Similar to the pattern observed with the harvested otter samples, we observed a difference in amplification success between fresh and old samples. The genotyping

RESULTS Optimizing Microsatellites All 10 loci were polymorphic in Missouri river otters (Table 3). Overall error rates ranged from 0.013 (RIO06R) to 0.110 1630

Figure 2. Error rates and amplification success (with SD) across time as proportion of total polymerase chain reaction (PCR) attempts, averaged across all loci, for matched river otter fecal and tissue samples collected from 34 harvested otters in Missouri, USA during the 2007–2008 trapping season. The Journal of Wildlife Management  75(7)

Table 5. Minimum (from genotyping, with sex ratio), CAPWIRE, and model (H8; see Table 4) estimates of river otter population size in the winter and spring of 2009 in Missouri, USA, by river and season. We rounded the predicted densities from the model to the nearest whole number. For the CAPWIRE model used, TIRM ¼ Two Innate Rates Model and ECM ¼ Even Capture Probability Model. Genotyping estimate, sex ratio (M:F)

River Big Piney Winter Spring Total Courtois Current Maries Niangua Osage Fork Roubidoux Winter Spring Total West Piney Winter Spring Total

Min. density (otters/km)

CAPWIRE estimate, 95%CI

CAPWIRE model used

Model estimate

0.255 0.511

9, 6–16 17, 12–26

TIRM TIRM

5 9

0.134 0.403 0.110 0.069 0.442

3, 3–3 11, 11–11 3, 3–3 2, 2–2 14, 14–15

ECM TIRM TIRM TIRM TIRM

3 9 5 6 16

6, 4:2 10, 8:2 11, 8:3

0.174 0.291

6, 6–6 11, 10–13

ECM TIRM

6 19

3, 1:2 3, 2:1 5, 2:3

0.121 0.121

5, 3–10 3, 3–3

TIRM ECM

3 3

6, 2:4 12, 6:6 14, 6:8 3, 3:0 11, 8:3 3, 3:0 2, 2:0 14, 9:5

success rate of anal jelly was higher (71%) than that of scat (analysis of variance [ANOVA], F1, 1411 ¼ 176.45, P < 0.001), and there was a difference between the success rates of old (24%) and fresh (12%) scat (ANOVA, F1, 1310 ¼ 27.16, P < 0.001). Genotyping success rates were higher in winter (Jan–Mar; 26–31%) than in spring (Apr; 18%); however, this effect was not statistically significant.

We identified 63 individuals (41 M, 22 F) across the 8 rivers, ranging from 2 otters in the Niangua River to 14 otters in the Big Piney River and Osage Fork of the Gasconade River (Table 5). No population showed significant deviation from Hardy–Weinberg Equilibrium (Table 6), and no significant linkage disequilibrium existed for any loci in any population. The number of alleles per locus for each population differed primarily due to population size

Table 6. Number of alleles (A), number of alleles corrected for sample size (Ar), and observed (Ho) and expected (He) heterozygosity values for each microsatellite locus and river otter population sampled in winter and spring 2009 in Missouri, USA. n refers to the number of otters identified in each river and the subsequent number of multilocus genotypes used for the analyses. A River (n) Locus RIOO1R2 RIO02R RIO04R RIO06R RIO07R RIO08R RIO11 RIO13R RIO15R RIO16R All loci

Ho

He

A

Ar

Big Piney (14) 2.1 2.4 1.3 2.4 2.8 2.3 2.5 2.9 1.9 2.6 2.3

0.714 0.571 0.143 0.714 0.643 0.571 0.714 0.714 0.357 0.769 0.591

0.537 0.614 0.140 0.643 0.746 0.574 0.691 0.794 0.421 0.677 0.584

4 3 3 3 4 3 3 4 2 2 3.1

A

Ar

Ho

He

A

Niangua (2) 3 3 1 3 4 3 4 3 1 2 2.7

3.0 3.0 1.0 3.0 4.0 3.0 4.0 3.0 1.0 2.0 2.7

0.500 1.000 NA 1.000 1.000 1.000 1.000 0.500 NA 1.000 0.875

Ho

He

A

Courtois (3)

4 6 3 4 6 5 4 6 3 6 4.7

River (n) Locus RIOO1R2 RIO02R RIO04R RIO06R RIO07R RIO08R RIO11 RIO13R RIO15R RIO16R All loci

Ar

Mowry et al.  River Otter Population Survey

4 9 2 3 6 4 4 7 2 3 4.4

Ho

He

A

Ar

Current (11)

Ho

He

Maries (3)

3.2 2.6 2.8 2.6 3.2 2.3 2.3 3.0 1.9 1.9 2.6

0.667 0.667 1.000 1.000 1.000 0.333 1.000 0.667 0.667 0.667 0.767

0.867 0.733 0.800 0.733 0.867 0.600 0.733 0.800 0.533 0.533 0.720

4 5 2 3 4 3 4 7 2 5 3.9

2.4 2.7 1.5 2.4 2.8 1.9 2.4 3.1 1.9 2.8 2.4

0.662 1.000 0.273 0.778 0.800 0.455 0.818 0.818 0.364 0.900 0.684

0.662 0.740 0.247 0.680 0.768 0.437 0.645 0.840 0.520 0.774 0.631

2 3 1 3 5 3 3 4 2 3 2.9

2.0 2.3 1.0 2.6 3.6 2.6 2.6 3.2 1.9 2.6 2.5

1.000 0.667 NA 0.667 1.000 0.667 1.000 0.667 0.667 1.000 0.815

0.600 0.600 NA 0.733 0.933 0.733 0.733 0.867 0.533 0.733 0.719

Ar

Ho

He

A

Ar

Ho

He

A

Ar

Ho

He

Osage Fork (14) 0.833 0.833 NA 0.833 1.000 0.833 1.000 0.833 NA 0.667 0.854

Ar

2.7 3.2 1.8 2.2 3.0 2.5 2.6 3.0 1.3 2.1 2.4

0.769 0.857 0.429 0.714 0.786 0.786 0.692 0.857 0.143 0.786 0.682

Roubidoux (11) 0.726 0.849 0.476 0.595 0.815 0.667 0.717 0.794 0.138 0.542 0.632

5 5 3 2 2 3 3 4 2 3 3.2

2.7 2.4 1.5 1.9 1.5 2.3 2.3 2.2 1.3 2.0 2.0

0.900 0.909 0.273 0.727 0.273 0.909 1.000 0.636 0.182 0.636 0.645

West Piney (5) 0.737 0.645 0.255 0.485 0.247 0.628 0.636 0.567 0.173 0.507 0.488

5 5 4 4 3 3 3 4 2 3 3.6

2.8 3.1 2.7 3.1 2.5 2.5 2.5 2.8 2.0 2.3 2.6

0.800 0.800 0.600 0.000 0.600 0.200 1.000 0.600 0.200 0.667 0.547

0.756 0.822 0.733 0.857 0.689 0.689 0.711 0.778 0.556 0.600 0.719

1631

variation (2–14 otters per river); corrected for sample size, the allelic diversity did not differ significantly among rivers. The 2008–2009 otter trapping season occurred from 15 November to 20 February. Of 1,374 otters reported harvested across the entire state, 399 (29%) were harvested during our sampling period (J. Beringer, Missouri Department of Conservation, unpublished data). However, in the transect locations and sampling periods for each river, trappers did not harvest any otters, with the exception of the Roubidoux, where one trapper harvested 6 otters during that river’s winter scat sampling period (between 6 and 16 Jan 2009) and 3 more in the 2 weeks following (up to 4 Feb 2009). With this exception, the harvest would not have violated the assumption of closure or impacted our population estimates. Average density across all rivers was 0.239 otters/km, with the highest winter density occurring in the Osage Fork River (0.442 otters/km) and the lowest density occurring in the Niangua River (0.069 otters/km). Over both seasons, the highest density was in the Big Piney River (0.511 otters/km). In 2 of the rivers, we observed a seasonal shift in density; densities for the Big Piney and Roubidoux rivers averaged 0.215 otters/km in winter and 0.401 otters/km in spring. Genotyping success rates per river varied greatly, from 2.7% in the Niangua River to 100% in the Maries River (Table 1). We obtained complete genotypes for 56 of the otters (genotyped at all 10 loci); 4 individuals lacked genotypes at 1 locus, two lacked genotypes at 2 loci, and one lacked genotypes at 3 loci. Recapture rates ranged from 1 to 24 per otter. Across all rivers, the average number of recaptures was 4.5  3.7 (SD) per otter. Of 63 total genotypes identified, 13 (21%) were captured only once. Using CAPWIRE we estimated the same population sizes as the minimum in 7 of 11 analyses, and 5 of those had variances of 0 otters (Table 5). To further assess differences in genotyping success among different types of scat, we evaluated the number of otters that would have been detected if only those scat types had been collected. Of 63 total otters identified in all 8 rivers, if we collected only fresh samples then we would have detected 22 individuals (34%); if we collected only anal jelly samples then we would have detected 33 individuals (52%); and if we collected only old samples then we would have detected 58 individuals (92%). We would have detected 40 individuals (63%) by collecting only fresh scat and anal jelly samples and 60 individuals (95%) by collecting only old scat and anal jelly samples. Thus the collection of fresh and anal jelly samples added few additional individuals to the census results. Model Selection Of 8 a priori hypotheses, the top ranked model was H8 (Table 4), which used scats per latrine and latrines per kilometer to predict otter density (Fig. 3; r2 ¼ 0.7619, P < 0.001). The regression for H8 generated the equation Otter density ¼ 0:01574 þ ð0:03103  scats per latrineÞ þ ð0:18036  latrines per kmÞ: Overall, the average deviation from the genotyping estimates was 1.46  1.37 (SD) otters. When applied to Big 1632

Figure 3. Relationships between the 2 most supported candidate variables and river otter density during winter and spring 2009 in Missouri, USA, excluding the removed outlier (Big Piney, section 2, spring). The top model predicting otter density, H8, incorporated both scats per latrine and latrines per kilometer, whereas the next model H5 used only scats per latrine.

Piney, spring, section 2 (removed from regression analysis), the model estimated a total of 8 otters, three below the total identified by genotyping. The maximum underestimate was 3.6 otters in Roubidoux, winter, section 2 (6 otters detected, 2.4 predicted; genotyping success 55%), and the maximum overestimate was 6.2 otters in Roubidoux, spring, section 1 (3 otters detected, 9.2 predicted; genotyping success 15%). The model was robust in predicting population size for both sections combined, accounting for the appearance of individuals in both sections. Average deviation from the minimum genotyping estimate was 2.4  2.1 (SD) otters, the maximum overestimate was 7.7 otters in Roubidoux, spring (11 detected, 18.7 predicted, genotyping success 14%), and the maximum underestimate was 2.6 otters in Big Piney, spring (12 detected, 9.4 predicted, genotyping success 24%). We also tested the top 2 models on random combinations of river sections (excluding Niangua; Fig. 4). Both models overestimated total population size, H5 slightly more than H8. Average deviation from the genotyping estimate was 5.7 otters for H8 and 6.7 otters for H5. There was no indication that different sample sizes produced greater deviations.

DISCUSSION Although the overall genotyping success of our samples was low (24%), it is higher than that observed in many similar studies, especially when considering the number of microsatellite loci used. Ben-David et al. (2004) reported a 33% success rate for river otter scat described as fresh, genotyped The Journal of Wildlife Management  75(7)

Figure 4. Relationship between the observed number of otters (N) based on genotyping and the predicted number using model H8 applied to 20 random combinations of river transect data collected during winter and spring 2009 in Missouri, USA.

for up to 7 PCR runs on only 1 microsatellite locus (RIO05). Hansen et al. (2008) showed a 56% success rate for amplification at 1 locus, but that rate dropped to 8% for generating consensus genotypes across 4 loci; these authors also focused on collection of the freshest available samples. Guertin et al. (2010) achieved a 12% success rate for fresh and anal jelly samples genotyped across 7 loci. For Eurasian otters, Ha´jkova´ et al. (2009) collected only fresh samples and anal jellies, repeated PCRs up to 16 times for 10 microsatellite loci, and reported 60% genotyping success on 9 loci. We discarded samples that did not provide a consensus genotype on 7 loci after 4–5 PCR runs to prevent errors in otter identification, even though it inherently decreased the overall success rate. We believe that the increase in genotyping success we observed, compared to previous Nearctic river otter studies, is likely a result of the redesigned primers, which amplified smaller fragments of DNA. Genotyping success rates can be influenced by a variety of factors. For otters, Fike et al. (2004) determined that storage method, individual microsatellite used, and type of scat influenced amplification success and frequency of genotyping errors. Other studies described the effects of ambient temperature at collection and storage time (Ha´jkova´ et al. 2006), age of scat (Dallas et al. 2003), DNA extraction method (Lampa et al. 2008), and location of DNA on the scat sample (Stenglein et al. 2010). Sieving feces to remove prey remains and homogenize unequal epithelial cell distribution from the scat surface (Kohn et al. 1995, Hansen et al. 2008) or using storage buffers or silica desiccant at collection (Foran et al. 1997) might have improved our success rate but were not practical for >1,400 samples. Due to time and funding limitations, we did not re-extract failed samples or perform additional PCR runs to attempt to recover them. We observed highest amplification and genotyping success in anal jelly samples, consistent with previous studies (Coxon et al. 1999; Fike et al. 2004; Ha´jkova´ et al. 2006, 2009; Lampa et al. 2008). However, contrary to suggestions from other otter studies using otter scat (Ben-David et al. 2004, Prigioni et al. 2006, Ha´jkova´ et al. 2009) as well as the Mowry et al.  River Otter Population Survey

general consensus for fecal-based molecular ecology studies (e.g., Foran et al. 1997, Wasser et al. 1999), our results suggest that collection of only very fresh samples from the field may not improve genotyping success rates. In both field and carcass-collected scat samples, fresh scats had lower genotyping success rates than older fecal samples (1– 6 days after collection). Had we collected only anal jelly and fresh samples, only 63% of the total individuals would have been detected, whereas analysis of only old samples would have represented 92% of the total counted population. Addition of anal jelly samples to the old samples would only have increased the population size by 2 individuals. Therefore, despite the high amplification success of anal jelly samples, we found that the information they provided was largely redundant. To determine why we observed lower amplification success for fresh samples, we tested several post hoc hypotheses (R. A. Mowry, University of Missouri, unpublished data), including testing for suboptimal DNA concentration (too high or too low; Mangiapan et al. 1996) of failed samples with a spectrophotometer and testing for PCR inhibition by substances in failed samples by spiking PCRs of these samples with DNA from a positive control. We also evaluated the possibility that the DNA-containing cells of these samples rubbed off onto the plastic bag before freezing by re-extracting them using material rinsed off the interior surface of the bag. None of these hypotheses were supported. Farrell et al. (2000) observed a substantial decrease in amplification success of canid and felid scat in the wet season in Venezuela compared to the dry season, suggesting that collection of fresh samples in plastic bags may trap moisture, creating a humid environment inside the bag which may encourage growth of mold or bacteria during the hours before the sample is frozen. The activity of these organisms may hasten the rate of DNA degradation, act to inhibit PCR amplification, or result in a non-uniform distribution of DNA on the surface of the scat. Using a storage buffer, silica desiccant, or paper bag for fresh samples, while storing drier (and presumably older) samples in plastic bags, may be a suitable compromise to enhance genotyping from as many samples as possible while maintaining the collecting pace necessary for large-scale projects. Overall, the methods we developed were effective for processing the large number of samples we collected at latrine sites and led to river otter density estimates comparable to those reported elsewhere. Previous studies of river otters using traditional field methods (e.g., radio telemetry, snow tracking, and radioisotopes) generated river otter density estimates of 0.17–0.37 otters/km (average 0.26 otters/km) in western Idaho (Melquist and Hornocker 1983), 0.26– 0.45 otters/km of shoreline on an Alaskan coastline (Bowyer et al. 2003), and a predicted maximum density of 0.40 otters/km of river or shoreline in the interior west (Melquist and Hornocker 1983, Melquist et al. 2003). In 2 study areas in Missouri, Erickson et al. (1984) observed densities of 0.13–0.25 otters/km. Genotyping studies of Eurasian otters found densities of 0.18–0.20 otters/km in southern Italy (Prigioni et al. 2006), 0.45–0.83 otters/km in 1633

Slovakia and the Czech Republic (Ha´jkova´ et al. 2009), and 0.17 otters/km along the Drava River in Hungary (Lanszki et al. 2007). The only published report of density estimation for North American river otters using genetic methods (Guertin 2009) found densities between 0.37 and 0.63 otters/km in a coastal population on Vancouver Island, British Columbia, Canada. Generally, our estimates fell within these ranges. For the Big Piney River, sampled in both winter and spring, the population size and density nearly doubled in the spring, despite the lower genotyping success rates that were typical of that time period and the tendency for otters to increase latrine visitation during the winter breeding season (Stevens and Serfass 2008). The population size estimate of the Roubidoux River, also sampled in both seasons, nearly doubled as well, despite the harvest during winter sampling and between winter and spring. Furthermore, the sex ratio data indicate that the Big Piney, Roubidoux, and West Piney rivers became more male-biased in spring. Male-biased sex ratios for river otters have been reported frequently (see Melquist and Hornocker 1983 for summary). Hamilton and Eadie (1964) observed equal sex ratios in winter with a shift toward a male-bias in spring, and Blundell et al. (2002) suggested that males may increase home range size to increase female encounters during mating season. In addition, Lauhachinda (1978) observed a male bias when examining river otter fetuses from harvested pregnant females (173 M:100 F). Thus, the increase in population size and male-biased sex ratios we observed in spring may reflect an increase in adult male abundance (possibly from movement of neighboring males into the study transects), birth of young, or both. In addition, if female otters had given birth during or just prior to sampling, they may have decreased latrine visitation or restricted use to a few sites close to the den (Melquist and Hornocker 1983), which would have decreased our ability to detect them at least once and caused the population to appear male-biased. In most cases, the population estimates derived from CAPWIRE analyses were not significantly different than the minimum size observed through genotyping (Table 5). However, for the Niangua River, the extremely low sample sizes (2 individuals with only 5 captures) combined with the low genotyping success rates for the river (13%) suggest an underestimate. The model based on latrine site density and scats per latrine reflected this uncertainty by predicting a higher number of otters (Table 5), but CAPWIRE results did not. Results derived from CAPWIRE estimations also predicted a greater population size than we observed, with wider confidence intervals, for the Big Piney River in both winter and spring. Scat abundance can act as an index to population size. The relationship between scats per latrine and latrines per kilometer reflected the number of otters using those sites (by individual section as well as in combined sections), and only slight overestimates were produced when we combined randomly selected river sections (Fig. 4). The model implies that population size is predicted not so much by the number of scats occurring across the study area, but rather by the 1634

distribution of those scats at communal latrine sites (scats per latrine) and the total number of latrines in the landscape (latrines per kilometer). This result reflects that scent-marking by river otters plays an important role in intra-specific communication (Kruuk 1992), particularly among males (Rostain et al. 2004). The model we developed detected possible underestimates of population size due to low genotyping success rates (e.g., Roubidoux and Niangua rivers) and did not show bias relating to sex ratio skew (Big Piney River). Although anal jellies were the most consistent in providing positive genotyping results, these samples provided little additional information for estimating population size, as virtually all individuals we identified from anal jellies we also identified from fecal samples. We also observed a substantial sex bias in rates of anal jelly deposition; of 69 anal jelly samples that we could assign to a particular otter, 80% were deposited by males. This evidence that anal jellies are produced by only a portion of the total population was reflected in the model selection process, which indicated that numbers of anal jelly were of little value for predicting otter population density. The sex ratio was biased toward males for all rivers except the Big Piney. Three rivers (Courtois, Maries, and Niangua), which also had the lowest minimum population sizes, contained males only. The Courtois and Maries rivers were not stocked with otters during the initial reintroductions. Blundell et al. (2002) reported that male otters in Prince William Sound, Alaska disperse further than females; thus, the low population size and male-biased sex ratios observed in the Courtois and Maries rivers may indicate recent colonization by dispersing males.

MANAGEMENT IMPLICATIONS Fecal genotyping can be applied to estimate population sizes for otter populations, and the genetic work we conducted indicates that even without genotyping, surveys that quantify numbers of scat and numbers of latrines accurately reflect the abundance of otters. Such surveys, which are straightforward to conduct in the field, should focus on numbers of feces and latrines, as counts of anal jellies are largely redundant and male-biased. These sign surveys can be used to contrast the relative abundance of otters, or if absolute abundance estimates are desired, field surveys can be combined with laboratory genetic work such as we conducted to assess the relationship between signs of otters and true numbers of otters. However, if a genetic component is being used to quantify the regional accuracy of the relationship between numbers of scat and latrines versus the actual number of otter genotypes occurring on the landscape, managers should recognize that the freshest scats (those approx.