Genomic correlates for migratory direction in a free-ranging cervid

Abstract

Animal migrations are some of the most ubiquitous and one of the most threatened ecological processes globally. A wide range of migratory behaviours occur in nature, and this behaviour is not uniform among and within species, where even individuals in the same population can exhibit differences. While the environment largely drives migratory behaviour, it is necessary to understand the genetic mechanisms influencing migration to elucidate the potential of migratory species to cope with novel conditions and adapt to environmental change. In this study, we identified genes associated with a migratory trait by undertaking pooled genome-wide scans on a natural population of migrating mule deer. We identified genomic regions associated with variation in migratory direction, including FITM1, a gene linked to the formation of lipids, and DPPA3, a gene linked to epigenetic modifications of the maternal line. Such a genetic basis for a migratory trait contributes to the adaptive potential of the species and might affect the flexibility of individuals to change their behaviour in the face of changes in their environment.

1. Introduction

Migration, defined as the seasonal movement between home ranges [1], is critical for the persistence of species in variable environments. Animal migrations influence nutrient deposition across ecosystems and affect seed dispersal [2,3]. Migratory behaviour has been identified across all major branches of the animal kingdom and is a complex phenomenon that involves the interaction between environmental and genetic cues that can influence how individuals respond to selection [4]. Migration also is one of the most threatened ecological processes [5], and declines in migratory behaviour due to climate change and habitat loss have been observed, leading to the reduction in the variation in migratory behaviours exhibited in nature [6–8].

A wide range of migratory behaviours occurs in nature, ranging from short altitudinal migrations in birds and bats [9,10] to the annual approximately 1000 km journey of wildebeest across East Africa [11,12]. Variation in migratory behaviour exists within species [13], between populations [14,15] and among individuals [16,17]. Variation in migration propensity, timing and movement patterns can have critical implications for ecosystems connectivity [2]. While variation in migratory behaviour is largely driven by the environment, which has been the focus of most research on migratory behaviour [18,19], a significant amount of phenotypic variance can be explained by genetic variation [20–22]. Genes associated with migration direction and timing have been identified in birds [23–25] and fishes [26], but large mammals remain less studied (e.g. [27]). Understanding the genetic mechanisms driving migration can help uncover the evolutionary origins and constraints of migration. This would help quantify the adaptive potential—the ability of a population to evolve adaptively in response to selection—of migratory species to environmental change. Furthermore, as migratory behaviour continue to disappear across the globe, we could see a reduction in genetic variance contributing to migratory behaviours, thereby reducing overall adaptive potential of migratory species [28].

Ungulates (hooved mammals) show substantial phenotypic variation in migratory behaviour. Broadly, variation in migration propensity occurs within partially migrating populations [29]. Variation exists in migration timing and migration distance, with some species of ungulates migrating thousands of kilometers [3] while others migrate less than 100 km [30,31]. Studies have primarily attributed migration in ungulates to variation in the abiotic and social environment [32–34]. Recently, Cavedon et al. [20] identified genetic markers associated with migration propensity in caribou (Rangifer tarandus), and Gervais et al. [35] found evidence that movement (i.e. speed) and space-use behaviours are heritable in roe deer (Capreolus capreolus), which suggests genetic underpinnings for migratory behaviour in ungulates.

This study aims to identify genes associated with a migratory trait by undertaking pooled genome-wide scans on a natural population of migrating mule deer (Odocoileus hemionus). Migration is important for survival and reproduction in this species. Mule deer have a wide geographical range and typically migrate from high altitude, productive summer ranges to low altitude winter ranges. Migrations typically match changes in resource availability, with mule deer attempting to optimize migratory timing relative to both plant productivity and weather on their summer range (e.g. [36]). Mule deer can migrate a wide range of distances [37], while also showing variation in migration timing [27]. Mule deer are the subject of extensive management programmes throughout North America, due to their importance as a game species; their annual migrations are of interest to wildlife managers as recent anthropogenic development and climate change may threaten migratory routes [36,38]. Migratory plasticity may be one way for the population to buffer against the effects of rapid environmental change, however, Sawyer et al. [37] found that mule deer exhibited little to no migratory plasticity in terms of whether or where to migrate, meaning they could be less resilient to change. Consequently, adaptation to a rapidly changing environment may require microevolutionary change, for which a genetic basis for migratory variation is required. Therefore, the existence of genetic correlations to variation in migratory traits in mule deer could provide insight into the effectiveness of management programs and aid managers in making decisions in the context of the evolutionary processes that maintain variation in natural populations.

We use a system consisting of four study areas of migrating mule deer in the Piceance Basin of northwestern Colorado, USA, that all share the same winter range but migrate to two geographically distinct summer ranges (figure 1; electronic supplementary material, appendix S1 figure S1). The Piceance Basin supports one of the largest populations of migratory mule deer, and one of the largest natural-gas reserves in North America. Natural gas development density varies across summer ranges in the Piceance Basin, with some areas being free from development and other areas having high levels of development [38]. Mule deer in Colorado have seen a protracted decline over the last 30 years with the ultimate causes remaining uncertain [39]. Variation in migration timing has already been linked to genetic differentiation and mitochondrial haplotype in this population of mule deer [27], suggesting a genetic basis for variation in migration strategies. The variation in migration to one of two distinct summer ranges that exists within this system presents an opportunity to assess the genomic basis for a migratory trait. An association analysis allows us to locate regions of the genome that likely contribute to phenotypic differences among individuals or populations by correlating marker variants (i.e. divergent regions of the genome) with variation in a trait of interest [40]. Using DNA extracted from the blood samples of radio-tracked individuals, and whole genome sequencing, we assess divergent genomic regions correlated with different migratory behaviour and identified candidate genes associated with migration direction (e.g. [27]).

Figure 1.

(a) Winter range areas for Ryan Gulch (■), South Magnolia (▲), North Magnolia (◆) and North Ridge (●). Symbols represent study area, and colour represents migratory direction, with blue indicating north–south migration and orange indicating east–west migration. Points shown are the centroid of all GPS locations per individual while on the winter range. (b) Eastern and southern summer ranges are designated by the orange and blue outlines, respectively. (Online version in colour.)

2. Material and methods

(a) . Data collection and migratory phenotypes

Adult (greater than 1 year old) female mule deer were captured using helicopter net gunning on the four winter range study areas and fitted with store-on-board GPS radio collars (Advanced Telemetry Systems, Insanti MN, USA) with three different relocation schedules (5 h, 60 min and 30 min) depending on the individual. Deer were spotted visually by the helicopter capture crew and captured using a net gun. Deer were then blindfolded, hobbled and administered 0.5 mg kg⁻¹ of Midazolam (and 0.25 mg kg⁻¹ of Azaperone intramuscularly) to alleviate capture-related stress (dose of both drugs based on an average weight of 75 kg). Deer were transported to a central processing site typically within 2 km of the capture site, where they were weighed, measured for chest girth and hind foot length, and blood samples were collected for genetic analysis. We also obtained a body condition score by palpating the rump and measured the thickness of subcutaneous rump fat and the depth of the longissimus dorsi muscle using ultrasound [41–45]. We used the body condition score and ultrasound measurements to estimate the percent ingesta-free body fat of each deer [42,43]. Deer were released at the processing site immediately following blood sample collection and collar attachment.

The four winter range study areas located in the Piceance Basin are: (1) North Ridge, in the northeastern portion of the Basin; (2) Ryan Gulch in the southwestern portion of the Basin; (3) North Magnolia; and (4) South Magnolia in the central portion of the Basin. North Ridge and Ryan Gulch are both geographically separated from the other study areas by topography. It was initially thought that the deer in the Magnolia region were one contiguous group, however, GPS radio-collar data from the first year of the study indicated that individuals were split between the northern and southern half of the winter range, with most individuals from the two groups migrating to different summer ranges. We split our study area into North Magnolia and South Magnolia and assigned deer to an area based on where they spent the majority of the winter using the proportion of GPS radio-collar locations in each area [38]. Therefore, some overlap in space may occur between individuals in North and South Magnolia study areas. During spring migration, mule deer from the North Ridge and North Magnolia area move east–west towards the Flat Top Mountain Range, where they remain at high elevation for the summer. However, some mule deer whose GPS radio-collar data indicate that they spend the majority of the winter in North Magnolia and the mule deer from the South Magnolia and Ryan Gulch migrate north–south to high elevations along the Roan Plateau [36]. Elevation on the winter study areas ranged from 1675 to 2285 m and from 2000 to 2800 m on the summer study areas. Habitat characteristics (i.e. vegetation type) occurring along the migratory paths of mule deer was similar among study areas [46], however, individuals that migrate east–west on average travel greater distances than individuals migrating north–south. To test for differences of body condition that may account for genetically linked differences between the groups, we compared body weight, chest girth, hind leg length and ingesta-free body fat between north–south and east–west migratory individuals. We visually examined the GPS data for each individual to determine which summer range each mule deer migrated to on the year they were collared. Mule deer tend to show little plasticity in terms of their migratory behaviour, resulting in high fidelity to their migratory routes [37,38]. We then grouped individuals with the same migratory direction according to study area resulting in five groups as some individuals from North Magnolia migrate east–west while others migrate north–south (figure 1). There is no genetic subdivision among study areas [27].

(b) . DNA extraction and sequencing

We extracted DNA from each individual's blood sample using the DNeasy Blood and Tissue Kit (Qiagen, Valencia, CA, USA), following the manufacturer's protocol. We pooled the DNA of individuals from each of the five groups for sequencing using a pooled sequencing approach (Pool-seq [47]). Equal quantities of DNA (100 ng sample⁻¹) were combined into representative pools North Ridge (east–west migrating, n = 50); North Magnolia (east–west migrating, n = 50), North Magnolia (north–south migrating, n = 33); South Magnolia (north–south migrating, n = 50); and Ryan Gulch (north–south migrating, n = 50) to a final concentration of 20 ng µl⁻¹ of combined DNA for each pool. Pool-seq produces accurate allele frequency estimates with pools of approximately 50 individuals, thereby making this a cost-effective approach to sequencing [47]. The Pool-seq approach is also valid for smaller sample sizes (less than 40 individuals) as analysis is carried out on genomic windows containing multiple SNPs instead of individual SNPs [47]. Each pool was sequenced across five lanes of an Illumina HiSeqX platform to a desired 50X coverage (total of 5 lanes) at the Centre for Applied Genomics (Toronto, ON) (electronic supplementary material, appendix S2, table S1).

(c) . Genome alignment and SNP calling

Mule deer and white-tailed deer can hybridize [48], so we opted to use the long-read-based draft genome of white-tailed deer (Odocoileus virginianus) (Accession No. JAAVWD000000000) that was recently annotated [49]; we note that the currently available mule deer genome is simply a consensus sequence from reads mapped to earlier versions of the white-tailed reference [50]. We performed initial quality filtering for all reads using fastqc and trimmed reads for quality and adaptors using the default Trimmomatic v. 0.36 settings [51]. We aligned the pooled sequences to the unmasked white-tailed deer genome using BWA-mem v0.7.17 [52]. We used samtools v. 1.10 [53] to merge and sort all aligned reads into five files representative of each pool. We filtered for duplicates using Picard v. 2.20.6 [54], obtained uniquely mapped reads with samtools, and conducted local realignment using GATKv. 3.8 [55] before SNP calling. We called SNPs using samtools mpileup (parameters = -B -q20). We filtered out the masked regions and indels with 5 bp flanking regions and removed all scaffolds < = 50 kb.

(d) . Genome scan for population differentiation in migratory direction

We used an empirical F_ST approach [14,56] to identify potential SNPs relating to migratory phenotype between east–west-migrating and north–south-migrating pools. The trait of interest does not need to be under selection in order to locate divergent regions of the genome associated with phenotypic differences. We used a sliding window approach in calculating F_ST using the Popoolation2 software suite [57]. We used 2500 bp sliding widows, and a minimum covered fraction of 0.8 and specified a minimum overall minor allele count of 3 for each pool. We had 5 pools that provided 10 pairwise comparisons, 6 opposite phenotype comparisons (east–west versus north–south), and 4 same phenotype comparisons (1 east–west versus east–west, 3 north–south versus north–south). There is debate regarding the identification of outliers and the potential for false positives, specifically with regards to how the selection of an outlier cut-off level influences the process of identification [58,59]. To address this, we defined an outlier window as being within the top 1% of F_ST values in at least 4/6 of the opposite phenotype comparisons and not within the top 1% of F_ST values in all 4 of the same phenotype comparisons. Outlier windows within the top 1% of F_ST values in same phenotype comparisons are likely differentiated due to drift. By removing any outlier windows within the top 1% of F_ST values in opposite phenotype comparisons that are also highly differentiated in the same phenotype comparison, we are maximizing the probability windows are associated with differences in migration direction and not some unaccounted for population variation. We also ran the same analysis using less conservative criteria and identified outlier windows from the top 1% of at least 3/6 of the opposite phenotype pairwise comparisons, and those results can be found in the electronic supplementary material (electronic supplementary material, appendix S3). F_ST calculations account for sample size variation [60], and for pooled samples are minimally affected by differences in sample size, even at small sample sizes [61,62]. We are using a window-based approach, requiring multiple SNPs within a region to be divergent, therefore our measurements of F_ST are unlikely to be a result of sampling [63].

We evaluated genetic differentiation between east–west-migrating and north–south-migrating pools by conducting a principal component analysis (PCA) using allele frequencies for the SNPs within outlier regions. We conducted one PCA using the major allele frequencies of all available biallelic SNPs across the whole genome and one using the major allele frequencies of the biallelic SNPs identified within the outlier regions identified through the F_ST approach. We expect that the PCA using the outlier SNPs putatively associated with the trait would show clear separation between the two migratory types, while the PCA using all SNPs would not. We also examined the PCA using the major allele frequencies of biallelic SNPs found within the top 1% of F_ST values of same migratory type comparisons as we would expect that these would not show separation between the two migratory types. For all pools, we examined the proximity to genic regions for all SNPs identified within outlier regions using SnpEff v. 4.3t [64]. For this, we generated VCF files using bcftools v. 1.9 [52] and bed files containing the coordinates of all outlier SNPs. All SNP locations were characterized as being in an intergenic/intragenic region, 25 kb up/downstream of a gene (i.e. regulatory), intron or exon. We also assessed the putative proximity to the nearest gene for every identified SNP.

(e) . Gene ontology

To identify shared gene pathways among outlier SNPs, we used an analysis of gene ontology (GO) terms. We used the program Gowinda v. 1.12 [65] to determine GO term enrichment while accounting for gene length biases. We created a .gtf version of our annotation by removing duplicated genes, retaining only the longest version of each gene, which resulted in 15 395 unique genes in the annotation [49]. All outlier SNPs that SnpEff identified as being on or within 25 kbp of genic regions were used to analyze gene ontology and compared to all outlier SNPs in every qualifying window. Finally, we plotted the GO results with WEGO (Web Gene Ontology Annotation Plot) to visualize annotations following the vocabularies and classifications provided by the GO Consortium [66].

3. Results

Body weight, hind leg length, chest girth and ingesta-free body fat percentage did not differ between east–west and north–south migrating individuals (electronic supplementary material, figure S2).

(a) . Differentiated SNPs associated with migration

After filtering for coverage, INDELS and minimum allele counts, 194 533 windows each 2500 bp wide were evaluated for each of the migratory-study area groups, respectively. The genome-wide mean F_ST value was 0.026, with 0.027 mean F_ST across each of the same migration direction comparisons and 0.026 mean F_ST across all opposite migratory group comparisons. Based on our pairwise comparisons, we identified 19 windows within the 99^th percentile in at least 4/6 pairwise comparisons of opposite migratory types and not within the 99^th percentile in the 4 same migratory type comparisons (figure 2a). The average F_ST for these windows was 0.074 in the opposite migratory type comparisons and 0.027 in the same migratory type comparisons. Within these windows, 2903 SNPs were identified as on or within 25kbp of genic regions (figure 2b). The PCA analysis using major allele frequencies from biallelic SNPs across the whole genome (n = 21 944 208), and the major allele frequencies from biallelic outlier SNPs (n = 463) revealed separation of the migratory groups along the PC2 axis (figure 2c). Additionally, we confirmed that no such separation of migratory groups occurred for SNPs within the 99^th percentile in the 4 same migratory type comparisons through a PCA of the major frequency of those biallelic SNPs (n = 8128; electronic supplementary material, figure S3).

Figure 2.

(a) Manhattan plot for average F_ST values across the whole genome, for all opposite migration comparisons (top) and same migration comparisons (bottom). Each point represents a 2500 kbp window. The highlighted points indicate outlier windows associated with migratory direction, determined to have high levels of differentiation in at least 4/6 of the opposite migration comparisons and low levels of differentiation in all same migration comparisons. (b) Distribution of 2903 outlier SNPs across chromosomes identified as being on or within 25 kbp of genic regions. Bars indicate the proportion of SNPs found on each chromosome distributed between exons, introns, 25 kbp up- and downstream cites, and all other genic cites. Numbers above bars indicate the total number of outlier SNPs found on each chromosome. (c) Principal component analysis plot of major allele frequency values across 21 944 208 biallelic SNPs on the whole genome (I) and 463 biallelic outlier SNPs differentiated based on migratory direction (II). Arrows represent the five study group-migration direction groups with east–west migrators shown in orange and north–south migrators shown in blue. (d) Gene ontology (GO) assignment plots based on white-tailed deer annotations. Functional groups (x-axis) were in three functional categories: cellular component, molecular function and biological process. Number and percent of genes within a given functional category performing a specific function are indicated on the y-axis. Some genes belong to more than one functional group, which may result in a sum exceeding 100% in a category. Plot was created using WEGO (Web Gene Ontology Annotation Plot; [66]). (Online version in colour.)

(b) . Eleven genes associated with migration

The results from GOWINDA showed that the non-intergenic outlier SNPs were found on or within 25 kb 11 genes: the genes were ADGRB1, BSG, DPPA3, EMC9, FITM1, IKZF3, ITSN2, MAN2A2, PSME1, RNF216 and SETX. Two of note were FITM1, a gene linked to the lipid synthesis in humans, and DPPA3, a gene linked to epigenetic modifications of the maternal line. GO analysis yielded 214 GO terms which were classified into 31 functional groups belonging to three functional categories: cellular components (11 groups), molecular functions (10 groups) and biological processes (10 groups) (figure 2d; electronic supplementary material, appendix S4 table S2). Some genes belonged to more than one functional group (e.g. protein binding and cell differentiation), which sometimes resulted in a sum exceeding 100% in a category (e.g. cellular components). Among the genes categorized as cellular components, 100% were classified as cell parts. Most of the genes with molecular functions were associated with protein binding (90.9%), and most of the genes categorized as having biological processes were involved in cellular processes (90.9%), biological regulation (81.8%) and metabolic processes (72.7%).

4. Discussion

Our results show genomic regions associated with migration direction in mule deer; such a genomic basis for a migratory trait can have implications for our understanding of the adaptive potential of migration and the flexibility of individuals to change their behaviour in the face of stochastic changes to their environment. Migratory behaviours continue to disappear globally, largely due to climate change and anthropogenic alteration of the environment [67]. The disappearance of migratory routes can affect prey abundance, impact terrestrial and aquatic nutrient cycling, and could reduce genetic variation across species [3]. When there is an underlying genetic basis for a migratory trait, this contributes to the additive genetic variance influencing migratory behaviours. A reduction in that variance can directly affect the adaptive potential of a species to cope with environmental change [35]. The amount of additive genetic variance associated with migratory traits might affect the response of a population to new selection regimes, and the rate at which adaptive evolution can occur [68]. Consistent directional selection on migration traits could lead to distinct changes in migration variation within populations over a relatively short period of evolutionary time. Uncovering a genomic basis for migration can provide managers with information on the levels of genetic variation for this adaptive behaviour, which could be used to prioritize populations for protection or make decisions regarding which individuals to translocate to boost diversity in a declining population [69,70].

The tiered window selection criteria seem to extract primarily functional outliers compared to other pooled genome-wide association studies using statistical thresholds (e.g. [71]). These outliers were localized to 15 out of 32 chromosomes which is expected with polygenic traits [72]. Of the 11 genes on which outlier SNPs were found, ADGRB1, BSG, PSME1, SETX, RNF216, ITSN2 and IKZF3 are associated with immunity, host-virus interaction, and cell differentiation. The ability of an animal to maintain its immune system or mount an immune response can depend on its nutritional health and energetic condition. Variations in immune function and condition have been linked to migration in bats [73] and stopover behaviour along a migration route in birds [74]. FITM1 is of note as it is related to the formation of lipid droplets and has been associated with fat storage in humans [75]. Body fat is an indicator of fitness in ungulates [76]. It can influence the annual survival of adult females [77], pregnancy and twinning rates [78] and the probability of a female rearing a fawn through summer [78,79]. The link between body condition and migration performance has been documented in several taxa, including migratory shorebirds [16], giant tortoises [80] and impalas [81]). Lipid metabolism rather than deposition may be the driving mechanism that supports FITM1 as a potential modulator of mule deer migration in this system given that we found no significant differences between ingesta-free body fat percentage between the migratory groups (electronic supplementary material, figure S2) and genetic links to body fat and variation in migration timing were suggested for this mule deer population previously [27].

A final noteworthy gene was DPPA3 that is related to epigenetic modification of the maternal germ cell line. Epigenetics refers to changes in gene expression that occur without changes in the DNA sequence but through, for example, chemical modifications to the DNA (e.g. DNA methylation). Epigenetic variation can increase the phenotypic range encoded by a single genome and there is increasing evidence to show that such phenotypic plasticity can be inherited [82,83]. Heritable epigenetic mechanisms that lead to increasing phenotypic variation may increase the chance of offspring being able to cope with stochastic environments [83,84]. Our dataset consisted entirely of female deer, and while ungulate migration behaviour is thought to have a learned component (e.g. white-tailed deer fawns follow their mother's migration route [85]), previous work in this system has identified links between migratory timing and mitochondrial haplotype [27]. This, in conjunction with our results, could indicate that variation in migration behaviour is being inherited at least partly epigenetically through the maternal line, which could be beneficial for coping with future environmental change.

Investigations targeted at these outlier regions that we identified and individual genotypes appear warranted, and we suggest screening for runs of homozygosity [86] and assaying differential methylation [82] may help elucidate how these genetic differences are being maintained within these populations. By validating these putative migration loci, it is conceivable that a gene panel could be developed for characterizing the genetic profiles of migrating populations and be used in wildlife monitoring, such as to quantify genetic variation or population structure [69,87]. It has been established that differences in migratory behaviour exist between genetically distinct populations [14,88]. Typically, genome-wide association studies of migratory behaviour using natural populations have two very distinct groups that are often geographically separated [13]. We have shown that it is possible to detect genomic variation associated with migratory phenotypes without such large-scale comparisons. We were able to identify distinct differences in the genomes of individuals who overlap in space on their winter range while migrating to two different summer ranges, demonstrating that genomic differentiation between migratory strategies is detectable at a fine scale. It is likely that similar detectable patterns may exist in other taxonomic groups with population variation in migration behaviour. Many migratory ungulate species are considered at risk (e.g. caribou), raising concerns regarding population isolation and loss of genetic diversity. Screening for similar genetic associations in more imperiled ungulate populations may help shed light on local population dynamics, evolutionary potential, and better inform management decisions as migration routes continue to be affected by environmental and anthropogenic change.

Ethics

All procedures were approved by the Colorado State University (protocol ID: 10-2350A) and Colorado Parks and Wildlife (protocol ID: 15-2008) Animal Care and Use Committees.

Data accessibility

Raw sequence data FastQ files are available on the Sequence Read Archive (Accession: SRX15798759). All bioinformatic and analytical code are available on GitLab (https://gitlab.com/WiDGeT_TrentU).

Authors' contributions

M.B.: conceptualization, formal analysis, methodology, writing—original draft, writing—review and editing; S.J.A.: conceptualization, methodology, writing—review and editing; C.R.A.: data curation, writing—review and editing; G.W.: data curation, writing—review and editing; J.M.N.: conceptualization, data curation, methodology, supervision, writing—review and editing; A.B.A.S.: conceptualization, methodology, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant (A.B.A.S. and J.M.N.); NSERC Vanier PhD Fellowship (MB); Compute Canada Resources for Research Groups (A.B.A.S.). Mule deer capture and monitoring was funded by Colorado Parks and Wildlife (CPW), ExxonMobil Production/XTO Energy, Williams Companies/WPX Energy, Shell Exploration and Production, EnCana Corp., Marathon Oil Corp., Federal Aid in Wildlife Restoration (grant no. W-185-R), the Colorado Mule Deer Foundation, the Colorado Mule Deer Association, Safari Club International, Colorado Oil and Gas Conservation Commission and the Colorado State Severance Tax.