Scott et al. (Reports, 27 November 2020, p. 1086) suggest, on the basis of conclusions obtained from a desert tortoise reintroduction program, that higher genomic heterozygosity should be used to identify individuals for successful translocation. I contend that this recommendation is questionable given these relocated tortoises’ unknown origin, their high mortality, insufficient data on resident tortoises and other components of fitness, and potential allelic dropout.
Translocations of endangered species have been advocated for genetic reasons, both to introduce adaptive variation through assisted gene flow (1) and to alleviate inbreeding depression through genetic rescue (2). It does not appear that either assisted gene flow or genetic rescue was the reason for the translocations of desert tortoises examined by Scott et al. (3). In this study, 9105 tortoises were translocated to a 100-km2 site in southern Nevada after they were “salvaged from harmful anthropogenic activity and habitat destruction” with the majority from captive situations and “many from Nevada’s free pet tortoise pickup program.” Scott et al. genomically examined living and dead samples of these translocated tortoises.
Their recommendation to translocate tortoises with higher heterozygosity is based on a higher genomic heterozygosity in tortoises that survived translocation. However, given their unknown origin, this association might be confounded with other factors. For example, low-heterozygosity tortoises might have been bred in captivity (about half were juveniles), may have spent most of their life in captivity, or may represent groups of related tortoises, reducing their ability to survive in the wild. The distribution of heterozygosity before mortality was not known in this study, making it impossible to determine how translocation might have influenced that distribution in survivors.
There was unusually low survival across all the tortoises (190 survivors out of 9105 translocated tortoises, 2.1% survival) while at two nearby sites, the average estimated population reduction over an 11-year period was only 61%. As a result, it appears that the area to which the tortoises were translocated had particularly unsuitable conditions during this period. The actual causes of mortality are not known; thus, mortality at this site during this time is probably not a good barometer for translocation success elsewhere. Infectious diseases such as mycoplasmosis, often introduced by releases from captivity, have caused high mortality in other desert tortoise populations (4). With such high mortality, differences due to adaptive variation might be obscured and not easily documented.
One possible approach to determining the factors involved in mortality is a comparison between the estimated 1450 initial adult resident tortoises in this area and the translocated tortoises. Such a comparison could have examined whether the surviving translocated tortoises had more or less heterozygosity than, or were genetically similar to, the surviving resident tortoises.
Only one part of overall fitness was examined: adult survival. Although adult survival after translocation is fundamentally important, other components of fitness—juvenile survival, mating success, and fecundity—are important for subsequent contribution to the population. For example, translocated male desert tortoises in California had no detectable mating success and did not contribute genetically to the population (5). Reproduction of the translocated tortoises could be documented by examining the genotypes of any new recruits to identify their parents and determine the relationship between reproductive success of translocated tortoises and their heterozygosity.
The surviving tortoises sampled by Scott et al. might not be a random sample of all surviving tortoises, and/or the sampled dead tortoises might not be a random sample of all dead tortoises. There were proportionally many more surviving tortoises sampled [79 of the estimated 190 survivors (41.6%)] than dead tortoises sampled [87 of 8915 dead tortoises (0.98%)], hence the surviving percentage/dead percentage ratio is 42.4 because many more tortoises that died were not sampled. This suggests a much greater potential for nonrandomness, and therefore biased estimates of heterozygosity and survival, in the sample of dead tortoises.
The difference in estimated genomic heterozygosity between the survivors (HS = 0.00180) and those that died (HD = 0.00146) suggests that the dead tortoises had an inbreeding coefficient of ~0.189 [f = (HS – HD)/HS] compared to the survivors. This is much higher than expected in a random-mating population and between the level expected for progeny of full sibs (f = 0.25) and of half sibs (f = 0.125), which suggests that a number of the dead tortoises might have been inbred.
The heterozygosity levels given for surviving and dead tortoises are the observed heterozygosity (HO) values that can be compared with the expected heterozygosities (HE). For example, HE could be at the same level for both the surviving and dead tortoises as for the HO for the surviving tortoises. In this case, the lower HO in the dead tortoises could be attributed to inbreeding in this group. Or HE could be elevated relative to HO among the sampled dead tortoises as a result of population structure.
It is possible that the dead translocated tortoises had lower heterozygosity than the surviving translocated tortoises for some technical reason. Although the authors extensively discussed their efforts to examine this potential in their supplementary materials, it is still possible that allelic dropout in the samples from the dead tortoises might have occurred, thereby reducing quantified heterozygosity. One approach to examining this would be to compare heterozygosity for samples taken from the same individual tortoises when alive and later when found dead—a type of ground truth for these results.
Finally, Scott et al. found no association between translocation distance, measured as the straight-line distance between the estimated individual tortoise origin and the translocation area, and survival. A distance that incorporates obstacles (such as inhospitable mountains) and uses a probable tortoise movement distance between two points would be more appropriate. In addition, it would be useful to designate sites by habitat characteristics such as soil type, rainfall, plant community, temperature, etc., and to examine the association of a calculated habitat distance with heterozygosity.
The authors concluded that “one of the advantages of individual heterozygosity is that it can now be easily and economically measured” and that their findings “suggest that an optimal strategy of assisted migration could be to prioritize moving the most genomically variable individuals, rather than current practice based solely on geographic or genetic similarity.” Such genomic data could be used to supplement other biological data, but it should not be used as the primary determination of individuals for translocation, particularly when there are concerns about the data and their interpretation as discussed here.
Acknowledgments: I thank M. Kardos, R. Fleischer, and T. Edwards for comments.