Allele frequencies and genetic diversity at four polymorphic loci (CK-A2*, G3PDH*, GPI-A*, sMDH-B1,2*) in wild brook trout populations in 56 southwest Virginia streams, grouped by drainage.
Population genetic evidence suggests differentiation among evolutionarily significant units of southern and northern Appalachian brook trout, with the zone of contact in southwestern Virginia. Before this differentiation was recognized, brook trout of northern origin were stocked throughout the southeastern United States. In order to determine this differentiation, established allozyme markers were used to classify 56 southwest Virginia populations as southern, northern, or introgressed. Variation at 4 polymorphic loci, including the diagnostic creatine kinase (CK-A2*) locus, indicated that 19 populations were of southern origin, 5 of northern origin, and 32 of mixed genetic origin. Data compiled among genetic studies of brook trout in the southern Appalachians showed that the southern/northern break is sharp, occurring at the New/Roanoke-James watershed divide. New River drainage populations exhibited the southern allele at high frequency, suggesting their historic native character as southern, with presence of northern alleles due to stocking or stream capture events. In conclusion, the present study suggests that management of southern Appalachian brook trout should include: (1) genetically cognizant planning of stocking events, (2) management of populations on a stream-by-stream basis, (3) prioritized conservation of pure southern brook trout populations, and (4) use of southern Appalachian hatchery stocks in restoration efforts.
- southern Appalachian brook trout
- population genetics
- trout management
Brook trout, Salvelinus fontinalis, is the only salmonid native to the southern Appalachian Mountains, and it is distributed across eastern North America from Canada to Georgia . This species was once abundant in coldwater lakes and streams throughout its range, but environmental disturbances such as deforestation, development, and pollution: and the introduction of non-native rainbow trout (Oncorhynchus mykiss) and brown trout (Salmo trutta) have drastically reduced the number and sizes of wild populations .
Beginning in the mid-1800s, fishery managers began stocking hatchery-reared brook trout extensively. However, hatchery-reared brook trout often exhibit lower growth, yield, survival, and natural reproduction than locally adapted wild populations [3, 4]. Further, the hybridization of hatchery-derived fish with wild populations can compromise the genetic integrity and fitness of receiving populations by introducing foreign alleles and breaking up locally adapted gene complexes [5, 6]. The stocking of northern-derived hatchery brook trout is of particular concern in the southern part of its range due to significant population genetic differentiation between southern and northern lineages of brook trout. Genetic differences between the two lineages may be large enough to justify distinction at the subspecies level [7, 8]. In addition, screening of allozyme [7–16], mitochondrial DNA [17–19], and microsatellite nuclear DNA [20, 21] markers has uncovered smaller scale genetic variation throughout the geographic range of brook trout. Differentiation at smaller geographic scales may reflect different colonization histories, as well as differential effects of selective and non-selective population genetic processes.
Native southern Appalachian brook trout (SABT) populations share several biological characteristics . Food availability being a limiting factor in these systems, adult fish are typically small (<229 mm total length) and life span seldom exceeds 3 years [23, 24]. Native SABT and introduced northern-lineage brook trout differed in terms of survival in the laboratory and diet in a natural stream . Comparison of external microbial assemblages suggested that SABT exhibit greater ability to inhibit microbial growth in their epidermal mucus than do northern brook trout of hatchery ancestry . Demonstration that SABT are genetically distinct from northern-origin hatchery stocks led management agencies to assess the heritage of populations within their jurisdiction, for example, in the Great Smoky Mountains National Park [8, 13], Tennessee , North Carolina [12, 16, 27], and Georgia . Molecular and adaptive differentiation may warrant management of brook trout populations or groups of populations as evolutionary significant units , although some of their population genetic differentiation may reflect stocking history.
The zone of contact between the southern and northern lineages of Appalachian brook trout is roughly at the New River watershed [14, 15, 29]. Against the background of decline of the southern form and history of stocking with non-native strains, genetic characterization of brook trout populations at the zone of contact is needed to support informed management decisions and conserve the native form of the species. The objective of this study was to use established allozyme markers to wild Appalachian brook trout populations at the zone of contact in southwest Virginia as southern or northern lineages or introgressed.
Seventy-eight historic wild brook trout streams from the New, James, Holston, and Yadkin river drainages  were sampled by backpack electrofishing. Brook trout tissue samples were collected from 916 individuals from 56 streams (Table 1). Sample sizes ranged from 8 to 26 individuals per stream. Fish were anesthetized, and two samples of dorsal muscle tissue (from fish greater than 120 mm TL) were collected non-lethally using an 18-gauge Monopty Biopsy Instrument (C.R. Bard, Inc., Covington, GA) and immediately placed on dry ice. Anesthetized fish were fully revived in fresh water prior to release. A limited number of fish of <120 mm total length were sacrificed to sample streams from which few adults were collected. Samples were stored at −80°C.
|Charles Creek, NC||5||1.00||1.00||1.00||1.00|
|Paint Bank Hatchery||16||1.00||0.44||0.56||1.00||1.00|
|Holston River drainage|
|James River drainage|
|New River drainage|
|Big Horse Creek||18||1.00||1.00||1.00||1.00||0.25||1.3||0.011||0.011|
|Big Laurel Creek||11||0.05||0.95||0.09||0.91||1.00||1.00||0.00||1.0||0.000||0.000|
|Big Reed Island Creek||20||0.08||0.93||1.00||1.00||0.95||0.05||0.50||1.5||0.068||0.066|
|East Fork Cove Creek||14||0.11||0.89||1.00||0.93||0.07||1.00||0.75||1.8||0.145||0.133|
|East Fork Crooked Creek||20||0.03||0.98||1.00||0.98||0.02||1.00||0.50||1.5||0.089||0.084|
|East Fork Dry Run||20||1.00||1.00||1.00||1.00||0.50||1.5||0.025||0.025|
|East Fork Little Reed Island||10||1.00||1.00||1.00||1.00||0.00||1.0||0.000||0.000|
|Little Indian Creek||19||0.79||0.21||1.00||1.00||0.95||0.05||0.50||1.5||0.125||0.101|
|Little Snake Creek||8||1.00||1.00||1.00||1.00||0.50||1.5||0.132||0.111|
|Little Stony Creek||14||0.11||0.89||1.00||0.96||0.04||1.00||0.00||1.0||0.000||0.000|
|Little Wilson Creek||19||0.21||0.79||0.03||0.97||1.00||0.82||0.18||0.50||1.5||0.071||0.067|
|Middle Fox Creek||12||0.04||0.96||1.00||0.04||0.96||0.58||0.42||0.00||1.0||0.000||0.000|
|NB Elk Creek||14||0.25||0.75||1.00||1.00||1.00||0.75||1.8||0.250||0.168|
|NF Stony Creek||21||0.02||0.98||1.00||0.98||0.02||1.00||0.50||1.5||0.147||0.128|
|No Business Creek||20||0.20||0.80||0.03||0.98||1.00||0.90||0.10||0.50||1.5||0.024||0.024|
|Pearis Thompson Branch||17||1.00||0.15||0.85||1.00||0.91||0.09||0.50||1.5||0.155||0.119|
|Sulfur Springs Branch||10||0.30||0.70||1.00||1.00||1.00||0.00||1.0||0.000||0.000|
|Upper West Fork Dry Run||10||1.00||1.00||1.00||1.00||0.00||1.0||0.000||0.000|
|West Fork Dry Run||19||1.00||1.00||1.00||1.00||0.25||1.3||0.063||0.057|
|West Fork Furnace Creek||17||0.12||0.88||1.00||1.00||0.97||0.03||0.50||1.5||0.044||0.068|
|Yadkin River drainage|
|South Fork Stewarts Creek||24||1.00||1.00||1.00||1.00||0.00||1.0||0.000||0.000|
2.2. Protein analysis
Genetic analysis was performed using cellulose acetate gel electrophoresis to observe variability at nine loci encoding five polymorphic enzymes: creatine kinase (CK-A2*), aspartate aminotransferase (sAAT-1,2*), glycerol-3-phosphate dehydrogenase (G3PDH*), glucose-6-phosphate isomerase (GPI-A*, GPI-B1,2*), and malate dehydrogenase (sMDH-B1,2*). Muscle tissue was homogenized in 200 μl of 0.09 M tris-HCl (pH 8.0), and subjected to electrophoresis in tris-glycine buffer (pH 7.5 or 8.0) for 45 min, followed by staining for enzyme activity. Electrophoretic conditions and histochemical staining procedures were modified from those described by Hebert and Beaton  and Galbreath et al. . Individuals from the Paint Bank Hatchery in Virginia were included in the analysis as a northern reference population because the hatchery is known to culture the northern lineage. The North Carolina Wildlife Resource Commission provided tissue samples from individuals from Charles Creek of the North Toe River drainage, a known SABT population, for use as a reference population.
2.3. Data analysis
Allele frequencies for CK-A2*, G3PDH*, GPI-A*, and MDH-B1,2* were calculated for all populations using the Excel Microsatellite Toolkit . Allele frequencies could not be calculated for sAAT-1,2* and GPI-B1,2* using that program because both enzymes are encoded by isoloci (i.e., duplicated loci with alleles of overlapping mobility). Since genotypes among heterozygous individuals could not be determined with certainty for sAAT-1,2*, phenotype frequencies were calculated using the program FDASH . The GPI-B1,2* isoloci contain multiple alleles that could not be assigned to either locus with confidence; hence, they were treated as a single tetraploid locus and allele frequencies were estimated using the program AUTOTET . Initially, allele frequency data from all nine marker loci were used to calculate genetic distance, population differentiation, contingency-table analysis of heterogeneity among populations, and hierarchical cluster analysis using the program BIOSYS-1 . The same statistics then were calculated using only the five marker loci with unambiguous interpretation of allelic expression (i.e., omitting data from sAAT-1,2* and GPI-B1,2*), to determine any effect of omitting these data from analysis. Similar conclusions were drawn from analysis of both data sets. Here, we report results based on analysis of the reduced dataset only.
Initial characterization of the genetic origin of each population was based on allele frequencies at the diagnostic CK-A2* locus. Allele frequencies at the other markers were compared to those observed in northern and SABT populations characterized in previous studies [7–16]. Individual heterozygosity and polymorphism were calculated across five loci to assess levels of genetic diversity within each population . Arlequin  was used to test for departures from Hardy-Weinberg equilibrium and to perform analysis of molecular variance (AMOVA) to characterize the distribution of the genetic diversity within and among populations and river basins. Cluster analysis using the unweighted pair-group with arithmetic averaging algorithm (UPGMA, ) was performed using BIOSYS-1 , and a dendrogram was built based on Nei’s unbiased genetic distance .
Allele frequency data from previous studies of brook trout population genetics were compiled and combined with the results from this study to gain a better understanding of the geographic distribution of SABT in Virginia, as well as the genetic composition of brook trout populations throughout the Appalachian portions of the native range.
Of 56 wild brook trout populations from 4 major river drainages analyzed in this study, 19 were fixed for the diagnostic CK-A2*100 allele, and were designated as pure SABT populations (Table 1). Five populations fixed for the CK-A2*78 allele were designated as northern, and 32 populations exhibiting variation at the CK-A2* locus were designated as introgressed. The three James watershed populations exhibited alleles characteristic of northern-form brook trout. Populations in other watersheds were characterized as southern (n = 19), northern (n = 2), or introgressed (n = 32).
Only the Cabin Creek population (New River drainage, Grayson County) deviated significantly (p < 0.05) from Hardy-Weinberg equilibrium at the CK-A2* locus. No other deviations from Hardy-Weinberg equilibrium were detected, indicating that the respective populations were in reasonable conformance with assumptions underlying the model. The proportions of polymorphic loci (P), the mean number of alleles per locus (A), and mean heterozygosities (H) for each population are listed in Table 1. Observed mean P and H0 values were lowest in the putative southern populations (P = 0.05, H0 = 0.004; Table 2). The introgressed populations exhibited the highest means for metrics of genetic variability (P = 0.48, H0 = 0.099), and the northern populations exhibited intermediate means (P = 0.20, H0 = 0.053). Grouped by drainage, Yadkin River populations had the lowest means (P = 0, H0 = 0), followed by James River (P = 0.08, H0 = 0.007), New River (P = 0.34, H0 = 0.064), and Holston River (P = 0.29, H0 = 0.100) populations. Atlantic-slope populations exhibited lower mean percent polymorphic loci and heterozygosity values (P = 0.05, H0 = 0.004) than Gulf of Mexico drainage populations (P = 0.33, H0 = 0.068). Analysis of molecular variance showed that approximately 34% of the total genetic diversity resulted from variation within populations, 18% among populations within drainages, and 48% among drainages. Most of the total limiting variance was attributed to the CK-A2* locus, meaning that most of the variance that we measured with allozyme markers was due to differentiation among northern and southern lineages of the species.
|Holston River drainage||6||0.29||1.3||0.100||0.102|
|James River drainage||3||0.08||1.1||0.007||0.012|
|New River drainage||45||0.34||1.4||0.064||0.058|
|Yadkin River drainage||2||0.00||1.0||0.000||0.000|
|Atlantic Ocean drainages||5||0.05||1.1||0.004||0.007|
|Gulf of Mexico drainages||51||0.33||1.4||0.068||0.063|
There was no apparent pattern regarding where populations characterized as southern, northern, or introgressed were located geographically within the New, Holston, Yadkin, and James drainages (Figure 1). Cluster analysis of unbiased genetic distances  among all populations showed that all populations of northern origin or with a high frequency of the CK-A2*78 allele clustered together; these included populations from the James River drainage (Barbours Creek, Ewin Run, and Pickles Branch), the Holston drainage (Henshew Creek), the New River drainage (Pearis Thompson and Little Indian Creek), and Paint Bank Hatchery. The Roaring Fork population in the Holston drainage had a high frequency of the northern allele, but did not cluster closely with the other northern populations due to a high frequency of a rare allele at the GPI-A* locus. Cluster analysis of unbiased genetic distances  among populations showed no geographic patterns of genetic variation among the populations of putative southern Appalachian origin.
4.1. Decline of brook trout
We sampled 78 streams that historically contained brook trout populations, but found the species in only 56 of them . The range of brook trout is shrinking  for several reasons, including habitat alteration, overexploitation, competition with introduced rainbow trout (O. mykiss) and brown trout (S. trutta) and more recently, climate change.
4.2. Duplicated isozyme loci in brook trout
Certain allozyme markers posed complications to interpretation of underlying genotype. Brook trout show a high incidence of duplicated enzyme loci due to the tetraploid ancestry of salmonids . Duplicated loci (termed isoloci) are genetically independent, but exhibit alleles of similar electrophoretic mobility that cannot be unambiguously assigned to either locus. Three of the five enzymes that we screened were encoded by isoloci (i.e., MDH-B1,2*, sAAT-1,2*, and GPI-B1,2*). Ambiguous interpretation of the banding patterns of two of these isoloci, sAAT-1,2* and GPI-B1,2*, led us to eliminate them from statistical analysis . Precise estimation of genetic diversity and differentiation metrics require data from many loci [41, 42]. Information from only four markers clearly limited the power of statistical analysis of genetic differentiation, especially with small sample sizes for some of the populations . Genotypic data from more markers likely would reveal genetic differentiation not detected with only four loci. Ongoing screening of additional, more highly polymorphic markers, such as microsatellite DNA markers, will increase the ability to quantify population genetic differentiation.
4.3. Geographic distribution of SABT in southwest Virginia
Based on fixation for the diagnostic allele at the CK-A2* locus and allele frequency differences at three other marker loci, 34% (n = 19) of the brook trout populations analyzed in this study were of southern Appalachian origin, 9% (n = 5) were of northern origin, and 57% (n = 32) were of mixed genetic origin (Tables 1 and 2). The level of certainty for precise characterization of a population is directly related to sample size. That is, any population observed to be fixed for the common allele actually may harbor the alternate allele at a low, undetected frequency. For example, with a sample size (s) of 20, our likelihood (p) of detecting an allele with a frequency (pa) of 5% is 36% (i.e., p = (1−pa) s = 0.9520, ). Therefore, there is a non-zero likelihood that some populations characterized as “pure” southern Appalachian are of mixed genetic origin. Similarly, sample size also affects estimation of within-population diversity statistics such as P and H0. Sampling of a limited number of populations in a watershed also would affect estimates of between-population genetic variability.
Of the six populations from the Holston drainage, four were of mixed genetic origin, with the southern allele at frequencies ranging from 0.44 to 0.95. The Grassy Branch population was characterized as southern Appalachian, and the Henshew Branch population was characterized as pure northern. Results from earlier genetic studies [8, 11, 14] and its geographic location suggest that the Holston River historically contained the southern Appalachian lineage, so the presence of the northern allele is likely due to stocking.
The Yadkin (upper Pee Dee) River is an Atlantic-slope watershed. Despite the common presumption that Atlantic-slope drainages would contain native northern-form brook trout [8, 12, 15], two pure southern Appalachian populations (Pauls Creek, South Fork Stewarts Creek) were found in the Yadkin drainage. Although no early sampling efforts are known from the upper Pee Dee in Virginia , the section of the river that flows through North Carolina was excluded from the range of brook trout originally described by Smith . However, several stream capture events have been inferred in this region, suggesting that these populations are descendants of brook trout captured from the New River . Inspection of stocking records showed that both Pauls Creek and South Fork Stewarts Creek were stocked in the recent past, implying that the “native” southern strain persisted despite stocking.
Earlier genetic study  and geographic location suggest that the James River historically contained northern-form brook trout. Three populations from the James River screened in this study were characterized as northern form. This finding leaves little doubt that the New River is the boundary between northern and southern Appalachian brook trout populations.
In this study, 16 populations from the New River drainage (36%) were characterized as southern Appalachian brook trout. No geographic patterns of genetic variation were observed among the populations of putative pure southern origin. Interestingly, two of these “pure southern” populations (Crooked Creek and West Fork Dry Run) were stocked in the recent past with northern-derived hatchery fish. Crooked Creek is a “put-and-take” fishing area, and 5000 brook trout are stocked annually, yet it maintained an apparently pure southern population. Sixty-three percent of the populations from the New River drainage were of mixed origin, with the southern allele at frequencies ranging from 0.21 to 0.98. Although stocking records are limited, only two of these (Howell Creek and Little Indian Creek) are known to have been stocked with northern-derived hatchery fish. Only one population (Pearis Thompson Branch) in the New River was characterized as pure northern.
In addition to the 56 populations characterized in this study, we compiled data from all known genetic studies of brook trout populations in southwest Virginia [12, 14, 15]. Forty-seven percent (n = 39) of all 83 populations characterized in southwest Virginia were of mixed genetic origin (Table 3); however, many of these introgressed populations were largely southern. In addition, the “pure” southern populations (n = 26) that remain provide opportunities for restoration of southern Appalachian brook trout in Virginia.
|Stream||River drainage||County||N||% Southern allele||Source|
|Green Cove Creek||Holston||Washington||19||95|||
|Little Laurel Creek||Holston||Smyth||16||100|||
|Hanks/EF Chestnut Creek||New||Grayson||10||70|||
|Middle Fork Helton||New||Grayson||20||100|||
|NF Elk Creek||New||Grayson||19||100|||
|NP Buckhorn Creek||New||Carroll||25||100|||
|Big Stony Creek||Roanoke||Bedford||10||0|||
|Little Stony Creek||Roanoke||Bedford||6||0|||
|Rock Castle Creek||Roanoke||Patrick||25||36|||
4.4. Range-wide geographic distribution and genetic affinity of New River brook trout populations
With the zone of contact between the northern and southern forms lying roughly at the New River watershed, it is unknown whether the New River historically contained the pure southern Appalachian form, or whether it was a zone of intergradation among southern and northern Appalachian lineages. Interpreting data across this study and the three studies noted above [12, 14, 15], the New River drainage contains 20 pure southern populations, suggesting that the presence of northern alleles could be due to either stocking or stream capture events. However, a large proportion (64%) of populations from the New River are of mixed genetic origin, suggesting either that hatchery fish persisted in the New watershed or that the New River is a zone of natural intergradation. To gain a better understanding of the geographic distribution of southern Appalachian brook trout, we compiled allele frequency data from all known genetic studies of brook trout populations throughout the native range (Table 4). Frequencies of the CK-A2*100 (i.e., southern) allele were weighted based on sample size and averaged across all populations in each river drainage. Figure 2 shows the frequency of the southern allele in each of the major river drainages from which data were collected.
|River drainage||State||Position1||# of streams||# of individuals||% Southern2||Source(s)|
|New||VA/NC||West||101||1999||85||[14, 15, current]|
|Yadkin||VA/NC||East||37||691||58||[8, 12, 15, current]|
|Holston||VA/TN||West||24||320||91||[8, 11, 14, current]|
|Nolichucky||NC/TN||West||51||1058||64||[7, 8, 11]|
|French Broad||NC/TN||West||80||1281||73||[8, 11, 16]|
|Little Tennessee||NC/TN||West||49||886||82||[8, 13]|
All river drainages north of the New River were characterized as pure northern, with the exception of the Roanoke River drainage that contained a single population with a low frequency of the southern allele, likely due to the transfer of individuals from another location or stream capture. The frequency of the southern allele in river drainages south of the New River ranges from 29% in the Broad River of North Carolina to 100% in the Coosa River of Georgia. Genetic characterization of individuals from 111 populations in the New River drainage showed an 85% frequency of the southern-form allele. Figure 2 shows that the south/north break is sharp and that this break occurs at the New/Roanoke-James watershed divide. This weakens the hypothesis that the New River is a zone of natural intergradation between the southern and northern forms of brook trout, and supports the hypothesis that the presence of northern alleles is due to either stocking or stream capture. However, it is important to qualify this inference by noting that genetic characterization is based on variation at a single locus. Ongoing screening of New River populations using microsatellite DNA markers will provide further insights into patterns of population genetic differentiation, shedding light on the native character of New River brook trout populations. In particular, microsatellite variation may clarify whether northern alleles observed in populations examined are characteristic of particular hatchery stocks or of native regional variation.
4.5. Management implications
Brook trout is the only salmonid native to the southern Appalachian region. The American Fisheries Society Southern Division Trout Committee developed a position statement  expressing the importance of SABT and presenting recommendations for conservation-oriented management of this regional resource. Our results contribute to the recommended completion of genetic inventory of critical populations using non-lethal sampling methods. In this context, we frame the management implications for management of SABT populations.
Results from this and other studies demonstrate that stocking of non-native genotypes poses long-term genetic impacts and interferes with efforts to conserve southern Appalachian brook trout. Although negative effects of stocking have become well known, some fisheries management agencies maintain imprecise stocking records. Further, hatchery personnel often substitute one brook trout stock for another based on availability. We recommend that all stocking and transfers of brook trout be well planned with cognizance of genetic conservation objectives and thoroughly and accurately documented.
Management units—that is, populations that are demographically independent of one another—may be defined functionally as populations that have substantially divergent allele frequencies at many loci . We had but limited ability to estimate levels of genetic diversity and differentiation among regional brook trout populations using allozyme markers. The results of ongoing screening of microsatellite DNA markers will be used to quantify differentiation among native populations, providing the basis for defining defensible management units. Results to date support the view that southern Appalachian brook trout populations should be managed on a stream-by-stream basis.
Those populations characterized as pure SABT should be given conservation priority. The stocking and transfer of non-native genotypes into these populations should be prohibited. Harvest should be allowed only in those populations that are demographically able to sustain themselves. We recommend that introgressed populations that contain less than 5% admixture from northern-strain brook trout be treated as ‘pure’ southern. However, we caution that the level of introgression in these populations may be higher than allozyme frequencies suggest; hence, individuals from these streams should not be transferred into streams that contain pure SABT populations. Hatchery brook trout should be stocked only into those streams that contain pure northern-strain populations and those with greater than 5% admixture.
We caution that any negative consequences of stocking also would apply to native northern-strain populations (i.e., in the James and Roanoke river drainages). Allozyme markers do not provide enough resolution to differentiate between native northern and hatchery populations, and so we recommend that all brook trout populations should be screened and characterized using microsatellite or single nucleotide polymorphism markers. Until we know more about the genetic composition of these populations, it may be wise to stock only infertile triploid brook trout .
Southern Appalachian brook trout hatchery stocks are being established in conservation-oriented hatchery programs (,
This work was funded through the Federal Aid in Sport Fish Restoration Project F-128-R, administered by the Virginia Department of Game and Inland Fisheries, and is based on the Master’s degree research of Joanne (Davis) Printz. We thank George Palmer and Cliff Kirk for assistance with fieldwork and collection of genetic samples. Ray Morgan of the Maryland Department of Natural Resources kindly provided unpublished data from Maryland populations, and Doug Besler of the NCWRC generously provided tissue samples. Finally, we thank Chris Printz of ATS International, Inc. of Christiansburg, VA for his assistance in the design and production of the maps. Funding for EH’s participation in this work was provided in part by the Virginia Agricultural Experiment Station and the Hatch Program of the National Institute of Food and Agriculture, U.S. Department of Agriculture.