Although it was previously shown that the size and shape of osteocyte lacunae vary between compact and cancellous bone in one studied taxon [11
], variation in osteocyte measurements between numerous different bones of the skeleton had remained unexplored until the present study. Furthermore, the measurement method employed here should be more accurate, and less prone to human error, than the techniques of previous studies, as areas of all lacunae are automatically calculated using a computer program rather than by measuring the length and width by hand [3
Our results unmistakably show that there is statistically significant variation in osteocyte lacuna size across the skeleton of extant vertebrates. With this established, the important question is: to what degree does this variation affect genome size estimates? Are C-value estimates relatively robust to variation in lacunae size among different bones, or is predicting genome size based on bone cell size doomed to intractable error? This is critical, because genome size is a correlate of many attributes of organismal physiology, such as metabolic rate, that are notoriously difficult to assess in fossil organisms [12
Although there is demonstrable variation in the volume and area of osteocyte lacunae in the four vertebrates sampled, genome size estimates are often accurate despite this variability. The tiger salamander and rock pigeon measurements performed the best, with the vast majority of skeletal elements, including all bones in the rock pigeon, predicting the measured C-value within 2 s.e.m. The genome size estimations for the woodchuck and the Chinese alligator, on the other hand, were not as accurate, as the majority of skeletal elements incorrectly predicted the genome size. This illustrates that genome size estimates based on osteocyte size may not always be accurate on a taxon-by-taxon basis, which raises serious doubts about the reliability of genome size reconstruction. This is not surprising, however, given that Organ et al
] found an overall significant correlation between osteocyte lacunae size and genome size in living vertebrates based on a large dataset of 26 species, but several cases of individual taxa in which this relationship does not hold. As a result, their regression analysis found that osteocyte size predicted only 32 per cent of the variation in genome size, meaning most of the variation is not described by this relationship. It is important to note, however, that phylogenetic regressions still offer a vast improvement in the accuracy of genome size estimations over simple linear regressions [3
Variability in lacunar size is also a vexing problem. First, there is no systematic pattern in osteocyte lacuna size across the skeleton of all vertebrates. In other words, there is not one bone or bone type that always results in the smallest or largest osteocyte measurements in the four taxa sampled. Instead, it seems as if osteocyte size is essentially randomly variable across the vertebrate skeleton. Therefore, consistently measuring osteocytes from the same bone would not be expected to improve resolution, standardize data collection or reduce measurement error.
In addition, even within bones, osteocyte size can be highly variable: each histological slide contains abundant osteocyte lacunae, so there will always be a standard error of measurement. Problematically, the commonly used phylogenetic estimation methods do not take this variability into account, because only one value for ‘measured osteocyte size’ can be entered. This is clearly a quandary for genome estimations, because the standard errors of the lacunae volume calculations are disproportionally large, often more than 50 per cent of the average calculated volume for that bone. Therefore, using only an average measure to estimate the genome size effectively fails to propagate uncertainty by not taking standard errors into account. We do find that average osteocyte size in many bones, despite their variability, still accurately predicts genome size, at least in the sense that the 95 per cent confidence interval of the estimation includes the measured value in the Animal Genome Size Database. Accuracy, however, must not be mistaken for precision, and this should be remembered when estimating genome size based on highly variable osteocyte sizes.
In summary, considerable variability in osteocyte lacunae measurements across the skeleton shows that lacunae size are an imperfect, but in some cases still useful, proxy for genome size estimation. We have found three important results: (i) the size of osteocyte lacunae in compact bone vary within an individual; (ii) lacunae size measured on most bones are often an accurate method for predicting genome size, but this is not true for each taxon; and (iii) variability in osteocyte size is often large, meaning that genome size estimations are imprecise. Together, these qualms mean that genome size estimates may not always be accurate, and are usually far from precise, but in the absence of genetic material in fossil organisms, they still prove our only means to trace the large-scale trends of genomic macroevolution in deep time. Coarse questions, such as whether Mesozoic theropod dinosaurs had relatively small genomes on par with those of avians [3
], should be tractable with this method, but genome size estimates for individual taxa should not be considered accurate or precise.