King et al. (2006; hereafter KEA) produced a second independent set of genetic data on the Preble's meadow jumping mouse (formerly Zapus hudsonius preblei) and their analyses thereof led to the opposite conclusions of Ramey et al. (2005) (hereafter REA) regarding the uniqueness of that subspecies. KEA argued that their different conclusions result from sampling design, tissues used (museum specimens vs. fresh ear punches), amount of molecular genetic data used [longer and more mitochondrial DNA (mtDNA) sequences, more microsatellite loci], analytical methods used and criteria for defining subspecies. We respectfully disagree with their interpretation of differences between the studies and their wholesale portrayal of our work as inaccurate. We find the difference between conclusions is largely a function of basic conceptual and philosophical differences. While the KEA sampling represented a notable effort, it fell short in both sampling design and strength of inference that they attempted to present so forcefully.
KEA chose not to use several key sources of contrary information that did not support their conclusions. These included a range wide study of over 9000 specimens of Zapus (Jones 1981), which did not find support for recognizing any subspecies of Zapus hudsonius. Also not cited was a literature review conducted by one of the coauthors of KEA (Cryan 2004), which could not find any published literature or reports suggesting adaptive or ecological differences among putative subspecies; a literature that spans 107 years of study. KEA also ignored the fact that we retested the original quantitative basis of Krutzsch's (1954) description of Z. h. preblei and found no support for his results (that were based on measurements of just three skulls). That alone would be sufficient basis to reject the taxonomic separation of Z. h. preblei. Rather than acknowledging that finding, KEA dismissed the use of morphology in general because it might not reflect genetic differences, an argument they attempted to support by selective use of references. However, by that same argument, these mice should never have been listed under the United States Endangered Species Act (US-ESA) — a listing that KEA now strive to defend. Of some significance to this issue is the fact that Krutzsch, who originally described Z. h. preblei, no longer accepts it as a valid subspecies (Ramey et al. 2006).
Conceptual basis and thresholds for uniqueness
We based our analyses on a definition of subspecies provided by Ball & Avise (1992) to avoid the long history of taxonomic subspecies decisions having no definitional basis (Cronin 2006). Ball & Avise (1992) proposed that subspecies should represent major subdivisions in the gene-pool diversity of species. By that definition, subspecies are similar to evolutionary significant units (ESUs) as discussed by Moritz (1994a) in requiring deeper historical phylogenetic separation, an important criterion of Crandall et al. (2000), for recognizing distinct populations. Although KEA claimed to use the same conceptual basis, they actually employed a far lower threshold for subspecies than REA that appears to be equivalent to what Moritz (1994b) defined as management units. Their null hypothesis was that subspecies of Zapus husdonius comprise a single homogeneous unit. The evidence that KEA considered adequate to reject that hypothesis was ‘significant phylogeographical separation of mtDNA alleles between subspecies combined with congruent phylogeographical structure for nuclear loci’. The issue is: what constitutes a significant difference?
KEA failed to acknowledge the ways in which the molecular results of REA were similar to their results. These include: (i) shallow levels of evolutionary divergence found among putative subspecies for mtDNA and microsatellites; (ii) support in mtDNA analyses for a Z. h. pallidus/luteus clade and a Z. h. preblei/campestris/ intermedius clade; (iii) not even near reciprocal monophyly among putative subspecies; and (iv) few unique microsatellite alleles in Z. h. preblei despite a larger sample size for this putative subspecies. These similarities are important because of their bearing on how different conceptual approaches to subspecies affected differences in conclusions.
Statistical significance versus biological significance
In their null-hypothesis test of genetic homogeneity, KEA equated statistical significance with biological significance, an analytical approach that deviates from REA. It is well known that with larger sample sizes it is possible to find statistical significance in almost any comparison, especially when intervening geographical variation is ignored. As pointed out by Hedrick (2001): ‘the statistical power for determining differentiation between groups is closely related to the number of independent alleles, so that even for a few highly variable microsatellite loci, there can be extremely high statistical power. When there is such high statistical power, extremely small molecular genetic differences between groups become statistically significant.’
Although KEA found a high level of statistical significance in their comparisons (using 27 microsatellites), the degree of differentiation among Zapus hudsonius preblei, Z. h. campestris and Z. h. intermedius were the lowest of any of the pairwise comparisons for mtDNA and microsatellites. The low degree of differentiation is illustrated by the fact that only four unique alleles were reported in Z. h. preblei (out of 279 in total), the lowest number of unique alleles for any subspecies sampled.
KEA reported high levels of correct assignment to subspecies using the program structure. These authors attribute this to Z. h. preblei having ‘considerable evolutionary differentiation’ from other putative subspecies and to shortcomings of REA. However, KEA failed to acknowledge that this high level of correct assignment could also be an artifact of sampling design and number of loci surveyed (Rosenberg et al. 2005). KEA's findings raise a valid critique of all such studies — use of assignment tests such as structure may not be an appropriate tool for evaluating taxonomic separation because of the sensitivity of these tests to the number of loci employed. Future efforts employing these types of analyses may need to establish threshold assignment probabilities for a set number of loci with a given amount of variation per locus to allow comparability between studies.
KEA sampled seven populations of Zapus hudsonius preblei but only one or two populations from each of the other putative subspecies. In contrast, REA sampled many populations, but few individuals per population, across the range of each putative subspecies. An ideal study design with unlimited resources would incorporate both approaches, as well as sampling across the entire species; however, this is not often practical because of logistical and funding constraints. Given the choice, which strategy provides the most objective test of subspecies uniqueness?
KEA claim their sampling strategy allows more appropriate statistical testing, but they do not acknowledge that their approach created artificial gaps in the distribution of genetic variation, leading to an ‘isolation-by-sampling design’ effect among all five of the subspecies. This sampling strategy predisposes the results to an exaggeration of genetic distances among putative subspecies. This assertion is supported by the fact that genetic distances from KEA were strongly correlated with geographical distances for the Z. h. preblei/campestris/intermedius lineage (R2 = 0.82), including a sample from southeastern Wyoming that is intermediate between Z. h. preblei and Z. h. campestris (see KEA Fig. 3).
Sources of material
KEA criticized REA for using museum specimens, claiming a wide variety of problems associated with such tissue, based mostly on literature for ancient DNA samples, not museum specimens collected within the past 45 years. In truth, Ramey et al. (2005) used a mixture of museum specimens and frozen tissue. However, all of the Zapus hudsonius preblei specimens used by KEA are subject to the same issues they raise regarding museum specimens because these ear punches were obtained without bleaching the ear punch between samples or wearing laboratory gloves to reduce cross contamination (R. Taylor, personal communication entered into the record on 7/6/06; Riggs et al. 1997).
If only newly trapped individuals are used, as advocated by KEA, investigations will be limited in the extent to which current patterns of variation might be parsed relative to historical natural processes vs. recent anthropogenic effects (e.g. bottlenecks).
KEA assert that the shared mtDNA haplotypes found by REA between Zapus hudsonius preblei and Z. h. campestris are the result of contamination of museum specimens, and they present a reanalysis of these specimens in support of their assertion. While this may be a point well taken, KEA did not consider any alternative explanations (e.g. nuclear paralogs, heteroplasmy, different amplification primers and conditions, and template quality). Additionally, the use by KEA of different primers, annealing temperatures (48 rather than 60 degrees) and buffers suggests conditions that could lead to different amplification success and results.
Despite efforts to portray our work as inaccurate, KEA changed several key methods and details of results at the proof stage after we pointed out those errors. First, KEA originally claimed that we amplified with primers L15926 and H16498, then changed this in the proofs to L15320 and ZAP5PLr (which they claimed did not work for disputed museum specimens). We actually used L15320 and ZAP5PLr, and we then used nested primers L15398 and H16498 for nested polymerase chain reaction (PCR) for a subset of weakly amplifying specimens (Ramey et al. 2005). Second, Table 1 of Appendix B in KEA clearly mixed up REA's mtDNA haplotypes and sample numbers but was changed by KEA in the proofs.
Even if the mtDNA sequences in question are excluded from analysis, it does not alter the basic conclusions of REA. That is because our critical tests did not rely on any sharing of haplotypes among putative subspecies. Instead, our subspecies test relied on: (i) morphological analyses to test the original quantitative basis of Z. h. preblei; (ii) mtDNA reciprocal monophyly, amova; and (iii) the proportion and frequency of unique microsatellite alleles and pairwise FST values. With the samples in question excluded, amova results just exceed our threshold but Z. h. preblei is still not even close to being reciprocally monophyletic. Similar results are obtained if all mtDNA sequences obtained by REA using nested PCR are excluded from analyses and KEA's data are substituted for REA's Z. h. campestris sequences.
Are KEA's mtDNA sequences ‘diagnostic’?
KEA claim their mtDNA results are ‘diagnostic’ in support of Zapus hudsonius preblei as a subspecies and as it being on ‘its own independent evolutionary trajectory’. Although they present additional mtDNA sequence data, Z. h. preblei remains paraphyletic with low bootstrap support.
Standardization between studies
KEA misrepresented that they had not obtained necessary samples from the Denver Museum of Nature and Science. The submission of a proposal along with a request is a standard requirement for destructive sampling of any museum specimens. KEAs request was fulfilled after King provided a proposal for the use of specimens.
We concur with KEA that the amova criterion that we proposed for mtDNA may not be an ideal measure with which to test the uniqueness of subspecies or distinct populations. As found by KEA, if there are slight differences among mtDNA haplotypes, but those haplotypes are fixed or nearly fixed in populations, this will have a substantial effect on the value of ΦST. That could lead to the erroneous designation of a weakly differentiated population as a subspecies or distinct population segment (DPS).
As noted in REA, thresholds for identifying subspecies and DPS's below the level of subspecies have been lacking. It is legitimate to debate thresholds, but the need for them is obvious — there are many endangered taxa and not enough resources to conserve them. If conservation effort is allocated to nondistinct or weakly differentiated populations, other more unique taxa (e.g. full species) will lose out. Hypothesis testing relative to these thresholds can provide objective assessments of degree of uniqueness and a basis for prioritizing the allocation of conservation effort.
The publicly available record shows that KEA changed their interpretation of results at least twice, from subspecies being ‘weakly differentiated’ to ‘evolutionarily distinct’; and the number of potential DPS's of Zapus hudsonius preblei changed from two, to three, to none. This subjectivity is symptomatic of an approach that lacks clearly defined thresholds, and epitomizes the problem that REA attempted to address. We believe that few if any subspecies or DPSs proposed for US-ESA listing or delisting will be falsifiable under the general approach and low level of genetic differentiation that KEA used to accept subspecies. General application of that approach may move the allocation of conservation effort outside the realm of scientific inquiry.
Rob R. Ramey is a conservation geneticist and field biologist with 26 years of research experience on the taxonomy and population structure of endangered wildlife in North America, Central America, Africa, and Asia. He consults on endangered species conservation issues in the U.S.A. and abroad. For the past 33 years John D. Wehausen has researched numerous aspects of the population biology and nutritional ecology of bighorn sheep in high mountain and deserts ranges of California. As a Research Scientist at the University of California, San Diego, he is also engaged in morphometric and genetic research concerning taxonomic questions of North American wild sheep. Hsiu-Ping Liu is a Research Assistant Professor at the University of Denver where she conducts genetics research on endangered invertebrates. Clinton W. Epps is a postdoctoral researcher at the University of California, Berkeley, where he is studying habitat fragmentation and connectivity of large mammal populations in the United States and Africa. Lance M. Carpenter is a biologist with the Colorado Division of Wildlife working on the conservation of non-game species.