Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 37
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine 5 Quality of the Data BACKGROUND Reservations have been voiced about the studies on Maine Atlantic salmon reported to date. Some relate to the sampling regime in the rivers, some to the reality of population delineation, some to the laboratory assays being used, and others to the statistical analyses. Some issues are more serious than others, but all deserve comment. First, however, we discuss the importance for these studies of taking into account the consequences of the overlapping generations of salmon, in particular the temporal variability that is introduced. OVERLAPPING VERSUS DISCREET GENERATIONS Although the genetic processes in populations with overlapping and nonoverlapping generations are identical in many respects, the temporal allele-frequency dynamics are different. Without proper consideration of such differences, it can be difficult to interpret population data (Ryman 1997). When considering temporal changes in populations, textbooks on population genetics refer almost exclusively to organisms with discrete (nonoverlapping) generations. However, unlike organisms with nonoverlapping generations, those with overlapping generations have a demographic age structure with the following features: (1) the total population is made up of individuals that belong to different age classes; (2) a restricted set of age classes participates in reproduction, and their relative contributions to the newborns of the next
OCR for page 38
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine year vary; (3) as a consequence, the parents of a particular year-class (cohort) do not represent a random sample from the entire population of the previous year; and (4) as a further consequence, allele- frequency differences among the newborns of consecutive years do not necessarily indicate temporal frequency changes for the total population (Jorde and Ryman 1995). In an age-structured population, the collection of individuals for genetic analysis frequently focuses on a restricted set of age classes. For anadromous Atlantic salmon, commercial catches at sea will not include the youngest age classes, whereas electrofishing on the nursery grounds in streams will primarily sample subadult cohorts. Comparisons of different collections (say commercial catches and nursery-ground collections) are not comparisons of comparable age classes. Allele-frequency dynamics within a generation are such that comparisons among cohorts must be made with care. To illustrate the different genetic dynamics of populations with discrete versus overlapping generations, Ryman (1997) designed a model of typical Atlantic salmon populations and simulated the effects over 200 years (Figure 5a). The most important observation from Figure 5a is that a population with overlapping generations displays considerably larger allele-frequency shifts from year to year than does a population with discrete generations of the same effective size. Moreover, there is a tendency for temporal correlation in the overlapping generation population; allele frequencies tend to fluctuate in a cyclical fashion that is not expected when generations are discrete. Clearly, population size is not the only determinant of the amount of temporal allele- frequency change when generations overlap, because the total population does not represent a single genetically homogeneous unit. Rather, it consists of several age classes (six in the Ryman 1997 simulation; Figure 5b) that are produced, partly or completely, from different sets of parents, and, therefore, the different cohorts might exhibit different allele frequencies. The amount of temporal noise in the allele frequencies of the total population thus depends on the age structure (the relative proportions of the different age classes). Likewise, periodicity is introduced, because each cohort of progeny (young fish) is similar not to the cohort of progeny in the previous year but instead to the cohort to which most of the breeding fish of the previous year belonged. Those breeders were young fish several years previously. The magnitude of the variation and the length of the periodic cycle are functions of the number of cohorts present among the breeders (in any given year) and their relative contributions to reproduction. The amount of temporal allele-frequency noise also depends on the age-specific birth and death rates of the particular population (Jorde and Ryman 1995).
OCR for page 39
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine For a population with overlapping generations, allele-frequency homogeneity among age classes is not expected for a neutral locus. The random component of temporal variation is increased—especially in small populations—by genetic variation among cohorts (Waples 1989, Jorde and Ryman 1995, Ryman 1997). Conversely, the observation of statistically significant differences should not automatically be viewed as evidence of gene flow, natural selection, nonrandom sampling (say of family groups), or an alarmingly small effective population size (Waples 1989, Ryman 1997). Several reports have been published comparing multiple samples from the same population for allele-frequency heterogeneity for genetic loci (Ryman 1997, and references therein). Too often, rejecting the null hypothesis of allele-frequency homogeneity (statistically significant heterogeneity) has led investigators to the conclusion that something is “wrong” with the population or with the samples. Explanations have ranged from very small effective population sizes (genetic drift) to natural selection on the loci examined, to straying among populations, and to familial sampling. Such conclusions might not be warranted, given the temporal heterogeneity expected with overlapping generations. FAMILIAL STRUCTURING OF SAMPLED FISH All routine statistical tests assume that any collection of fish gathered as a single population sample (a location and a year, a sampling event) represents a genetically random draw. Basically, the assumption is that the collection of individuals in the sample is as likely to be a genetic draw of individuals from that year and location as any other draw from that year and location. The sample is considered to be exchangeable with other samples. All the fish within a population will eventually become related to some degree or another, but that is not the issue. The question is whether an exchangeable sample, relative to the blend of genotypes available, is drawn. If the fish sampled in one event are more closely related to each other (family members) than an exchangeable draw, then subsequent analysis will yield a higher estimate of the genetic divergence among samples, because the genetic divergence shown by closely related fish from a single sample will be less than that shown by unrelated fish from different samples (Jorde and Ryman 1995, Hansen et al. 1997). Concern has been expressed that some of the samples collected in the past might not have met the exchangeability criterion (Kornfield 2000, Gold 2000).
OCR for page 40
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine FIGURE 5 Simulated temporal allele-frequency shifts of a selectively neutral allele in populations of identical effective size (Ne=124) with overlapping and discrete generations. The initial allele frequency is p=1/2: (a) total population results (and those for age class 1 of the model with overlapping generations), each population simulated or 200 years; (b) contrast of single-cohort frequencies, followed through time,for years 100–120; (c) contrast of age classes 1 and 2 for years 100–120. Source: Ryman 1997. Reprinted with permission from ICES Journal of Marine Science; copyright 1997, Academic Press.
OCR for page 41
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine Females typically lay several nests within a single redd,1 along a short linear segment of stream (=3 meters). Hatching fry from a single redd, representing a blend of full- and half-siblings, tend to school in the vicinity of that redd. Parr are territorial and tend to spread out a bit more from their hatching locality. Still, within a small stretch of stream, they may be related to some degree. Smolts are solitary and are more likely to spread out over longer stretches of the stream. Adults, collected either at the spawning sites or downstream at the weirs, can be viewed as representative, relative to the array of genotypes drawn by homing to a particular natal tributary. Many of the samples studied in previous work, involving fry, parr, smolts, and adults, have been collected by electrofishing, accomplished by walking along the stream with an electric wand that is effective over a small area (1–2 square meters). The total area of coverage can be large, but if fish within the treated area have a higher probability of being close relatives than a random draw of fish from the total population, any interpopulation measure of genetic divergence will be increased. Thus, fry collected by electrofishing might yield overestimates of the degree of population differentiation. Collections of parr are probably less affected, and collections of smolts and adults are probably not affected to any significant degree. Given the sorts of mixed cohort sampling that have been routine in the past, the distinctiveness measures among populations might be slightly increased, but the observed degree of divergence among sampled populations appears to be too large for familial sampling to be a major cause of its magnitude. We view the results as (possibly) minimally inflated but as nevertheless credible and compelling. We suggest that fry and parr should be collected from as wide a spatial area as possible within each local population as a means of avoiding undue sampling of any one family. POPULATION DESIGNATIONS The results to date for Atlantic salmon in Maine make clear that there is genetic divergence among tributaries within large watersheds. This divergence is a common result for fishes of the salmon family. The results for the Penobscot River (Table 4), for example, are clear that the entire watershed does not contain a cohesive, randomly mating population. The key to this situation is that genetic organization of salmon populations is hierarchical and organized by watershed. Straying occurs, but it is most likely from one spawn- 1 A group of egg pits, or nests, prepared by a single female. The eggs in each nest are fertilized by one or more males—not necessarily the same male for each nest.
OCR for page 42
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine ing segment of stream to the next, a bit less from one tributary to the next, and even less from one watershed to the next. One could imagine situations where a large watershed might best be treated in tributary-specific fashion, as in Pacific salmon species (e.g., Waples and Smouse 1990). That might make some sense within the Penobscot and within the Machias-East Machias complex, but some degree of lumping within a watershed seems unavoidable, particularly if brood stock are to be reared en masse at the hatchery. It also is obvious that watersheds within a region are divergent from those of the next region (on average), but what does regional divergence mean, when the regions are defined by political (rather than biophysical) criteria? For example, the Saint John River includes two political units but is assigned to the regional (political) unit that contains the mouth of the river. An intriguing question is whether the current (politically delimited) regions make optimal biological sense and whether the management of Maine, New Brunswick, and Nova Scotia salmon might be coordinated profitably. Toward that end, it would be productive to extend the assignment analyses of Maine fish to include the relevant watersheds in New Brunswick and Nova Scotia. Having said that, it is clear that the collective Maine population is distinct from the collective New Brunswick and Nova Scotia populations. LIMITATIONS OF GENETIC ASSAYS The laboratory aspects of genetic assays have received some criticism as well (Kornfield 2000, Gold 2000). All the genetic assays in current use have limitations, most associated with scoring the genetic markers on electrophoretic gels. This technique separates variant DNA fragments or proteins on a gel when they migrate at different rates in response to an electric charge. The basic operating premise of such work is that if two individuals each have a band (a scorable element) at the same position on the gel, the two bands represent the same thing. For allozyme protein markers, different alleles (slightly different amino acid sequences) can move the same distance on the gel under the influence of an electric charge. Allozymes are dealt with as classes, each of which is defined by its (joint) position on an electrophoretic gel. Although the formal population genetic analysis is not sensitive to the hidden variation within bands (i.e., the variation that is not separated by electrophoresis), there is a natural frustration with the inability to see everything. The availability of DNA markers has greatly improved the resolution of genetic variation, and most laboratories have now switched to DNA markers,
OCR for page 43
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine because the allozyme classes can be separated into the sequences associated with different protein alleles. To screen a population involving thousands of fish, the required DNA sequencing would be too expensive. What has been done instead—with a variety of molecular-genetic techniques, such as RFLPs and RAPDs, with minisatellites, and (most important for this report) with microsatellites—is to revert to gel-scoring techniques for DNA fragments, this time separated by size (number of nucleotides in the fragment). The lumping (different DNA sequences of similar size) is less than it is for allozymes (different protein sequences of the same charge), but the principle that equal movement implies genetic identity is the same, even though the principle implies some loss of resolution. As mentioned earlier, the population genetic theory is impervious to the nuances of lumping. Some variation is missed but what is seen is clearly present. Microsatellite methods reveal substantially more genetic variation than earlier allozyme methods. Interestingly, however, the conclusions drawn from the two sorts of data are qualitatively similar. Kornfield (2000) rightly called attention to a related issue with the early microsatellite results of King et al. (1999). With microsatellites, adjacent bands are very close together on the gel. There are so many alleles (fragments with different lengths) for some of the loci that it can be difficult to make small distinctions on the gel, and one is forced to “bin”2 alleles, placing alleles of almost the same size into allelic classes and reducing the number of alternatives under consideration. Even though there is variation within the allelic classes, it is necessary to construct such classes. This is again the issue of lumping. Although it is tolerable, it is exacerbated, because alleles between two adjacent bins (sizes) can be misallocated. A variety of related scoring problems can beset such work, particularly with large-scale population-screening efforts. Microsatellite markers provide a rich source of genetic information and permit population assessments that were previously impossible, but the best laboratory control available must be used. Fortunately, quality control continues to improve, and the obstacles are all surmountable with some extra attention given to detail in the laboratory. The more recent published reports (King et al. 2001, Spidle et al. 2001) have profited from Kornfield’s (2000) timely reminders. 2 The binning approach to the analysis of genetic information has been widely used in forensics (e.g., NRC 1992).
OCR for page 44
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine STATISTICAL CONSIDERATIONS Previous work has involved many statistical tests. A trio of statistical issues have been raised about such tests, all of which must be dealt with in future work (Kornfield 2000, Gold 2000). First, some sample sizes have been small, particularly in view of the number of genetic characters examined (Smouse and Chevillon 1998), calling into question the precision and power of the resulting statistical tests. In some cases, small sample sizes are unavoidable, because population sizes are themselves small, but whenever possible, efforts should be made to obtain sample sizes of at least 100 fish per collec tion.3 Second, the analysis of rare alleles or haplotypes is particularly sensitive to sample-size effects. Given a collection of S alleles or haplotypes from the population being sampled, the expected number recovered in a sample of size N is computable (Chakraborty et al. 1988), but because the number of fish sampled from a single population is sometimes smaller than the number of potential alleles or haplotypes, no more than a few of the latter can be seen. In general, the common (high frequency) types are seen consistently with a smattering of the rare types. Although the pattern of rare types is intriguing, too much has been made of that information in view of available sample sizes; any inference attributed to the presence or absence of rare types in small samples is unreliable. The genetic resolution for measuring population divergence is in the high-frequency genetic markers, and the results in Tables 1–4 are insensitive to rare alleles or haplotypes. The third criticism of much of the previous work is that too many hypotheses were being tested. The nominal probability levels are too generous, and it would seem that more conservative Bonferroni or even stepwise Bonferroni procedures (Rice 1989) would be better. Moreover, with K population samples, there are K(K-l)/2 pairwise tests of divergence but only (K-1) independent contrasts, and the full set of pairwise tests is highly intercorrelated. For multipopulation comparisons, a standardized test criterion (Smouse and Williams 1982) might be better. For broad survey work, a small number of independent tests, such as those available from F-statistics, Amova, or assignment procedures, would be best. The available population genetic data depart 3 There are many more fry in the rivers that can be sampled than returning adults, so this is a reasonable expectation.
OCR for page 45
Genetic Status of Atlantic Salmon in Maine: Interim Report from the Committee on Atlantic Salmon in Maine so far from the null hypothesis (of homogeneity) that substantial divergence is obvious by inspection. USEFUL INFERENCE A statistically significant difference is not necessarily biologically meaningful. With large-enough sample sizes, any difference could be statistically significant. The results of the studies reviewed here are highly statistically significant. The question is whether the differences are large enough to be biologically useful. The assignment tests in Tables 3–5 show that the multiple-locus gene pools of the various populations are substantially nonoverlapping. The differences are large and strongly suggest biologically important genetic isolation among the populations.
Representative terms from entire chapter: