Sexual maturity in growing dinosaurs does not fit reptilian growth models

Lee & Werning 1010.1073/pnas.0708903105

Supporting Information

Files in this Data Supplement:

SI Text
SI Figure 4
SI Figure 5
SI Figure 6
SI Figure 7
SI Figure 8
SI Table 2
SI Table 3

SI Text

Methods

Histological Sectioning and Imaging. The histological sections that we report in the current study were produced as parts of three broader ontogenetic analyses (1-3). The histological sections were produced by using standard hard-tissue techniques (4) and all were taken from standardized locations at mid-diaphysis. Photomicrographs were captured without filters or specialized lighting through an Optiphot2-Pol light transmission microscope (Nikon) and a D70 DSLR camera (Nikon). Brightness and contrast were improved uniformly for aesthetics by using autolevel and brightness/contrast functions in Photoshop CS2 (Adobe).

Age Estimation. Because all three specimens preserve growth marks or lines of arrested growth (LAGs) (1-3), we were able to estimate their age at death. We assumed that LAGs were deposited annually, which is generally the case in amniotes (5). However, a simple count of the LAGs preserved within a specimen likely will underestimate age because the earliest LAGs are lost to medullary cavity expansion and osteonal remodeling. Thus, to obtain an age estimate for each specimen, the number of preserved LAGs must be summed with the number of estimated missing LAGs. The number of missing LAGs can be estimated by using methods that fall into two major classes.

The first class uses the preserved LAGs in an ontogenetic series of specimens to approximate the number of missing LAGs in any given specimen (2, 5, 6). This class of "overlapping" retrocalculation requires that cross-sections are taken from a standardized location (e.g., mid-diaphysis) and assumes minimal variation in size at a given age. In each specimen, the relative order of LAGs and their circumferences were recorded by using ImageJ (NIH). These data were plotted in Excel (Microsoft) to produce an ontogenetic series of partial growth trajectories. The partial trajectories were aligned manually until (i) a least-squares regression using both a linear and a power model showed a maximum coefficient of determination (R²) and (ii) the composite growth trajectory produced a reasonable neonatal bone circumference (13-25 mm) (2). This class of methods works best when a fairly complete ontogenetic series of bones is analyzed. It, however, is less robust in the absence of neonate and juvenile specimens. Because this situation is so frequent, other complementary methods of LAG retrocalculation are used.

The second class of methods analyses each specimen individually and uses the distances between adjacent LAGs within each specimen to retrocalculate the number of missing LAGs. In each specimen, successive distances between preserved LAGs (zone thickness) were measured and used to calculate the maximum, mean, penultimate, and average incremental change in thickness. Those calculated values were extrapolated into the medullary cavity to retrocalculate the number of missing LAGs in each specimen. As with the overlapping methods, we assumed neonatal bone circumference to be 13-25 mm. However, unlike the overlapping methods, these incremental methods tend to maximize individual variation in size at a given age.

To account for uncertainty in estimate of age we used several methods from both classes (when possible) of retrocalculation. For Tenontosaurus (OMNH 34784), three age estimates using maximum, mean, and penultimate zone thicknesses were used. For Allosaurus (UUVP 5300), five age estimates were generated by: overlapping ontogenetic data fitted to a power model, overlapping ontogenetic data fitted to a linear model, maximum thickness, mean thickness, and the average incremental change in thickness within individual specimens. Estimated ages for Tyrannosaurus (MOR 1125) were gathered from a published study (1) and comprise two estimates using mean zone thickness, two estimates using maximum zone thickness, and one estimate using the average incremental change in zone thickness. Because the different estimates of age are not expected to be normally distributed, we calculated the median age estimate for each specimen. We generated 95% confidence intervals for each median age estimate by bootstrapping 10,000 replicates of the available age estimates.

Growth Curve Reconstruction. We used previously reported age and mass data (2, 3, 7) to reconstruct the growth curves of Tenontosaurus, Allosaurus, and Tyrannosaurus. A previous reported estimate of age for OMNH 34784 (3) fell within the nominal 95% confidence interval of the current estimate, so a reanalysis of the original data was not necessary. However, this was not the case for UUVP 5300 (2), so we applied the current methodology of age estimation to the original Allosaurus data. For all three dinosaurs, mass was estimated by using a scaling relationship between femoral circumference and body mass (8). For Tenontosaurus and Tyrannosaurus, only terminal age and mass data were reported (3,7). However, for Allosaurus, measurement of LAG circumferences (and subsequent mass transformation) allowed a finer sampling of mass accumulation during growth. SI Table 2 lists specimens and their mass accumulation data. Growth curves for each species were generated by using a nonconstrained three-parameter logistic regression (9) in Prism (Graphpad), which uses an iterative least-squares criterion.

Results

Diagnosis of MB. The endosteal tissue that we refer to MB was found in two specimens. The first specimen (Tenontosaurus, OMNH 34784) was recovered with both femur and tibia in association and both bones have the endosteal tissue. The second specimen (Allosaurus, UUVP 5300) is a tibia recovered from a bonebed (Cleveland-Lloyd Dinosaur Quarry), which preserves the disarticulated remains of tens of individuals (10). The femur and tibia from OMNH 34784 and the tibia from UUVP 5300 were well preserved and completely intact. Upon initial sectioning, we observed that the medullary cavities were completely filled with calcite, which prevented gross examinations by hand sample (in contrast to ref. 11). However, histological examinations revealed well preserved microstructure. That and the context of robust ontogenetic series allowed us to establish homology by using three accepted criteria.

First, the histological structure of the endosteal tissue in OMNH 34784 and UUVP 5300 is generally consistent with MB. In both specimens, the endosteal tissue forms a woven network of spicules (SI Fig. 4) similar to that found in avian MB (11, 12) (SI Fig. 5). Like avian MB, vascular canals and sinuses that surround the spicules are abundant and radiate toward the center of the medullary cavity. The network of spicules, however, is not as extensive as those reported in avian MB and only extends up to 1 mm into the medullary cavity in OMNH 34784 and up to 3.2 mm in UUVP 5300. The relatively smaller thickness of this layer compared to that of living birds may be the result of phylogenetic differences in tissue deposition. Alternatively, it may indicate incomplete resorption of MB after the previous laying cycle.

Second, the position of this tissue corresponds to MB. In both OMNH 34784 and UUVP 5300, the woven endosteal tissue is internal to several endosteal lamellae (EL). These lamellae clearly separate the woven endosteal tissue from the cortical bone (SI Fig. 4). In addition, the endosteal tissue occurs in both the femur and tibia of OMNH 34784. This broad distribution in OMNH 34784 suggests an organismal response characteristic of MB formation (7). Whether the distribution of the endosteal tissue in UUVP 5300 is also broad cannot be currently tested because the bones from that individual were disarticulated and scattered within the bonebed.

Third, development inferred from tissue stratification suggests an endosteal origin. Specimens smaller and younger than OMNH 34784 and UUVP 5300 show that endosteal surface is resorptive (i.e., it has a scalloped surface from medullary cavity expansion and lacks EL) (SI Fig. 6). In OMNH 34784 and UUVP 5300, however, EL line that surface and mark the extent of medullary cavity expansion. Furthermore, at least in OMNH 34784, EL cross-cut several secondary osteons, which suggests that some cortical remodeling must have occurred before the deposition of EL (SI Fig. 4A). Remodeling of the innermost cortex continued during and after deposition of the EL; some secondary osteons with intact lamellae are flattened against the EL (suggesting simultaneous formation), whereas the presence of other secondary osteons that have eroded partly into the EL shows that remodeling continued after EL deposition. Despite that remodeling, secondary osteons do not cross-cut the woven endosteal tissue. This suggests that the woven endosteal tissue is relatively younger than the surrounding cortical tissue.

We investigated whether the woven endosteal tissue might be the product of metaphyseal reduction (13). During long bone growth, cancellous bone from the metaphyses of earlier stages of growth become reshaped and compacted into the diaphyses of later stages. None of the ontogenetically younger or older specimens in the growth series showed compacted cancellous (or woven endosteal tissue) at mid-diaphysis (SI Fig. 6). That both younger and older specimens lack the woven endosteal tissue suggests that the woven endosteal tissue in OMNH 34784 and UUVP 5300 represents a special depositional event.

The woven endosteal tissue does appear to differ from avian MB in one way. A second set of EL lines the surfaces of the spicules in both dinosaurs (SI Figs. 4 and 7). This second EL has not been reported in the MB of either T. rex or living birds. It suggests that after deposition, the MB in OMNH 34784 and UUVP 5300 was not completely resorbed for use in shelling eggs. Rather, it remained long enough for the surfaces to be "finished" with EL in a similar fashion as medullary cavities. This minor idiosyncrasy requires the investigation of other interpretations for these endosteal tissues that we call MB.

Pathology is not a viable interpretation for the development of these endosteal tissues. In living birds, virally induced avian osteopetrosis causes rapid deposition of woven bone with radially oriented vascular canals and sinuses on both the periosteal and endosteal surfaces (14-16). Histologically, MB can resemble the endosteal portion of osteopetrotic tissue, but the externalmost periosteal tissue in OMNH 34784 clearly does not have osteopetrotic characteristics (SI Fig. 8A), and is otherwise histologically identical to other Tenontosaurus samples of similar age (3). In localized areas of UUVP 5300, the externalmost periosteal bone (SI Fig. 8B) does somewhat resemble osteopetrotic tissue. However, these areas are thin, and although similar tissues occur in the externalmost periosteal bone of other nonavian dinosaurs (15), none shows comparable endosteal deposition. The histology of the periosteal region, in combination with its restricted deposition, is most consistent with reactive bone deposited as a result of localized trauma (17). Additionally, the potential decoupling of growth along both periosteal and endosteal surfaces suggests that the externalmost periosteal bone in UUVP 5300 was deposited independently of the endosteal bone tissue.

We reject the interpretation of pathology and support the interpretation of MB. The depositional forms seen in OMNH 34784 and UUVP 5300 further suggest that the appearance of MB may be more diverse than previously thought even when bone histology shows identical patterns of deposition.

Reproductive Scaling Across Organisms. The scaling of mass and maturation time are strongly correlated in organisms (P < 0.0001). Size alone explains 96% of the taxonomic variation in maturation time. This result is interesting because it suggests that the relationship generally holds for both asexual and sexual organisms. We acknowledge that some taxa deviate from that relationship (e.g., animal and plant viruses), but despite those deviations, predictions calculated from this relationship for dinosaurs coincide with skeletochronological estimates of RM (SI Table 3). The two overlap considerably for Tenontosaurus and Allosaurus. However, for two of the smallest mass estimates of Tyrannosaurus, scaling predictions precede skeletochronological estimates of RM. This is not totally unreasonable because skeletochronological estimates of RM are upper bounds.

1. Horner JR, Padian K (2004) Proc R Soc London Ser B 271:1875-1880.

2. Bybee PJ, Lee AH, Lamm E-T (2006) J Morphol 267:347-359.

3. Werning S (2005) Thesis (University of Oklahoma, Norman).

4. Wilson JW (1994) in Vertebrate Paleontological Techniques, eds Leiggi P, May P (Cambridge Univ Press, New York), Vol 1, pp 329-339.

5. Castanet J, Francillon-Vieillot H, Meunier FJ, de Ricqlès A (1993) in Bone: Bone Growth, ed Hall BK (CRC, Boca Raton, FL), pp 245-283.

6. Chinsamy A (1993) Mod Geol 18:319-329.

7. Erickson GM, Makovicky PJ, Currie PJ, Norell MA, Yerby SA, Brochu CA (2004) Nature 430:772-775.

8. Anderson JF, Hall-Martin A, Russell DA (1985) J Zool 207:53-61.

9. Zullinger EM, Ricklefs RE, Redford KH, Mace GM (1984) J Mammal 65:607-636.

10. Madsen JH (1976) Utah Geol Min Surv Bull 109:1-163.

11. Schweitzer MH, Wittmeyer JL, Horner JR (2005) Science 308:1456-1460.

12. Dacke CG, Arkle S, Cook DJ, Wormstone IM, Jones S, Zaidi M, Bascal ZA (1993) J Exp Biol 184:63-88.

13. Enlow DH (1963) Principles of Bone Remodeling: An Account of Postnatal Growth and Remodeling Processes in Long Bones and the Mandible (Charles C. Thomas, Springfield, IL).

14. Sanger VL, Holt JA (1965)Can J Comp Med Vet Sci 29:245-252.

15. Campbell JG (1966) J R Microsc Soc 85:163-174.

16. Boyde A, Banes AJ, Dillaman RM, Mechanic GL (1978) Metab Bone Dis Rel Res 1:235-242.

17. Ogden JA (1981) Skeletal Radiol 6: 237-253.

SI Figure 4

Fig. 4. Cross-sectional views of endosteal bone tissues in the mid-diaphyseal tibia. (A) Tenontosaurus. (B) Allosaurus. Schematics indicate the level at the mid-diaphysis where these elements were sampled (dashed line). CB, cortical bone; EL, endosteal lamellae; MB, medullary bone; EL2, internalmost endosteal lamellae, deposited last. (Scale bar: 0.5 mm.)

SI Figure 5

Fig. 5. Cross-sectional view of medullary bone in the mid-diaphysis of a Pelecanus (pelican) femur. Specimen CMC 104. CMC, Creighton Medical Center, Creighton University, Omaha, NE; CB, cortical bone; EL, endosteal lamellae; MB, medullary bone. (Scale bar: 0.5 mm.)

SI Figure 6

Fig. 6. Cross-sectional views of the endosteal surface in younger tibiae. (A) Tenontosaurus, OMNH 63525. (B) Allosaurus, UUVP 40304. CB, cortical bone. (Scale bar: 0.5 mm for A, 1.0 mm for B.)

SI Figure 7

Fig. 7. Detail of bony spicules in the medullary bone of Allosaurus. The bony spicules of the medullary bone are finished with lamellar bone. The center of the medullary cavity is toward the right of the image. (Scale bar: 0.5 mm.)

SI Figure 8

Fig. 8. Cross-sectional views of periosteal bone tissues in the mid-diaphyseal tibia. (A) Tenontosaurus. (B) Allosaurus. The periosteal surface is at the top of both images. In Tenontosaurus, normal cortical bone (CB) occurs up to the periosteal surface, but in Allosaurus, localized pathologic bone (PB) tissue is found periosteal to the cortex in one area. (Scale: 0.5 mm.)

SI Table 2

SI Table 3