Results References Materials and Methods

1. 
   Materials and Methods
   
2. 
   Results
   
3. 
   References
   

1. 
   Materials and Methods
   

The method of osteocyte lacuna size measuring is carried out using an automated approach. We utilize scanning electron microscopy to photograph the thin section slides to obtain the finest available resolution. Since lacunae are so small, a difference of a few pixel widths can make a large difference in calculating the area of an osteocyte lacuna. The threshold of the image is altered in the program ImageJ (Abramoff et al. 2004), so that the osteocyte lacunae are easily distinguishable from the rest of the bone matrix. The “Analyze Particles” feature is utilized to outline and automatically measure the lacunar areas in the photograph. The length and width calculated by this algorithm were subsequently used to calculate cell volume using the equation (4/3 x π x width axis radius² x length axis radius) like in Organ et al. (2007).

The calculated volume values for each bone of each of the four studied taxa were then added to the most complete lacunar and genome size data set from Organ et al. (2009). As opposed to estimating the missing genome size values using the program BayesTraits (as used in Organ et al. (2007), Organ & Shedlock 2009, and Organ et al. (2009), a similar web based program called PhyloPars was employed. PhyloPars is extremely user friendly, able to easily handle missing data, and uses a maximum likelihood based method for inferring missing values (see Bruggeman et al. 2009 for more details). Additionally, the genome sizes used to check the accuracy of the estimated C-values from PhyloPars were obtained from the Animal Genome Size Database. To be consistent, only C-values obtained using the method of Feulgen Densitometry were averaged if more than one genome size was available for the given taxon.

The phylogenetic tree used in analysis was the extant taxon tree from Organ and Shedlock (2009). Each taxon in this study was added to this tree using Mesquite v. 2.73(Maddison & Maddison 2007) and the StratAdd package (Faure et al. 2006). Details of how this tree was assembled can be found in the electronic supplementary information of Organ et al. (2009).

In this sensitivity analysis, the “ln genome” for the 35 extant taxa was replaced with the minimum and maximum C-values reported in the Animal Genome Size Database. Subsequently, all of the analyses were rerun for each the minimum C-value and maximum C-value. It was shown that no matter which C-value (min, max or average) was used, the estimated genome size for the unknown value did not significantly differ (Table S3). Additionally, using the 3^rd quartile of the cell size measurements did not make a significant difference in the estimated genome size as compared to using the mean (Table S4). It is important to note that Organ et al. (2011) has presented an improved multiple regression method for estimating the genome sizes of amphibians- that have inordinately large genomes that improves the accuracy of the estimation. We were not able to use this method due to the fact that the program we used, PhyloPars, does not support multiple regressions. When estimating genome size for a salamander or related amphibian, the multiple regression method takes into account the different relationship between osteocyte lacunae size and genome size, so it should be more accurate.

2. 
   Results
   

Table S1: A summary of measured values from the 44 slides. The natural log of the mean lacuna volume presented here is the value that was used in reconstructing the genome size for that skeletal element in the program PhyloPars. The column “n lacunae” is how many lacunae were automatically measured per slide.

Table S3: A summary of estimated values from the sensitivity analysis, where maximum, minimum and average C-values were used in three separate sets of analyses to determine if the value selected for C-value from the Animal Genome Size Database affected the estimated genome size for the one unknown value. The difference in mean estimated genome size is not significant with the standard error taken into account.

Table S4: Due to the fact lacunae are 3-dimensional objects, the thin section preserve only one plane, and the most accurate representation of the size of that cell should be the largest 2-dimensional slice. Regression analysis of cell volume vs. genome size, however, does not demand an absolute, precise measurement of average cell volume for each taxon, but rather a consistent proxy for cell size in each taxon. This proxy, as long as it is consistently measured in all taxa, can then be regressed against genome size to obtain an equation to predict genome size from cell volume. Our study utilizes the size of all lacunae as a proxy for lacunar volume, because this method should be less prone to observer bias than arbitrarily averaging only a selected number of the largest osteocytes (as has been done in previous studies: Organ et al. 2007, 2009, 2011). However, to ease comparison with previous studies, we here present a supplementary analysis in which we use the average volume measurement of the largest 25% of osteocytes as a proxy for cell volume. The table below lists the average predicted genome size for each of the four taxa we studied, based on both the mean volume of all osteocyte lacunae and the mean volume of only the largest 25% of lacunae. As is shown, these estimates are statistically indistinguishable when standard deviation is taken into account, indicating that both methods of measuring cells return approximately the same estimated genome values.

Figure S1: An example of the intraspecific variation (95% CI) of reported C-values in the Animal Genome Size Database (www.genomesize.com). The variation in the database is most likely caused by different modes of measurement and associated error. Regardless of the source of this variation (measurement error or biological), the important result is that the spread of estimated genome size values (both the overall spread of estimates returned from all of the bones of a single taxon and the 95% CIs for the measurements of each individual bone) are still greater than the range of values presented in the database. In other words, estimated genome sizes using different bones are less precise than measured genome sizes. (Note: Not all species have multiple entries; the two plots below show our results for the pigeon and tiger salamander).

3. 
   References
   

Abramoff, M.D., Magelhaes, P.J., Ram, S.J. 2004. Image processing with ImageJ.

mammal-like genome size in the ancestral amniote and derived large genome size in

amphibians. Journal of Evolutionary Biology 24, 372-380.

Organ, C. L., Brusatte, S. L. & Stein, K. 2009. Sauropod dinosaurs evolved moderately sized genome unrelated to body size. Proceedings of the Royal Society B 276, 4303-4308.

Organ, C. L. & Shedlock, A. 2009. Palaeogenomics of pterosaurs and the evolution of small genome size in flying vertebrates. Biology Letters 5, 47-50.

Organ, C. L., Shedlock, A. M., Meade, A., Pagel, M. & Edwards, S. 2007. Origin of avian genome size and structure inc non-avian dinosaurs. Nature 446, 180-184.