THE ANATOMICAL RECORD 293:2044–2055 (2010)

Calcified Cartilage Shape in Archosaur Long Bones Reflects Overlying Joint Shape in Stress-Bearing Elements: Implications for Nonavian Dinosaur Locomotion MATTHEW F. BONNAN,1* JENNIFER L. SANDRIK,1 TAKAHIKO NISHIWAKI,1 D. RAY WILHITE,2 RUTH M. ELSEY,3 AND CHRISTOPHER VITTORE4 1 Functional Morphology and Evolutionary Anatomy (FMEA) Working Group, Department of Biological Sciences, Western Illinois University, Macomb, Illinois 2 School of Veterinary Medicine, Auburn University, Auburn, Alabama 3 Louisiana Department of Wildlife and Fisheries, Rockefeller Wildlife Refuge, 5476 Grand Chenier Hwy, Grand Chenier, Louisiana 4 Rockford Memorial Hospital, 2400 Rockton Avenue, Rockford, Illinois

ABSTRACT In nonavian dinosaur long bones, the once-living chondroepiphysis (joint surface) overlay a now-fossilized calcified cartilage zone. Although the shape of this zone is used to infer nonavian dinosaur locomotion, it remains unclear how much it reflects chondroepiphysis shape. We tested the hypothesis that calcified cartilage shape reflects the overlying chondroepiphysis in extant archosaurs. Long bones with intact epiphyses from American alligators (Alligator mississippiensis), helmeted guinea fowl (Numida meleagris), and juvenile ostriches (Struthio camelus) were measured and digitized for geometric morphometric (GM) analyses before and after chondroepiphysis removal. Removal of the chondroepiphysis resulted in significant element truncation in all examined taxa, but the amount of truncation decreased with increasing size. GM analyses revealed that Alligator show significant differences between chondroepiphysis shape and the calcified cartilage zone in the humerus, but display nonsignificant differences in femora of large individuals. In Numida, GM analysis shows significant shape differences in juvenile humeri, but humeri of adults and the femora of all guinea fowl show no significant shape difference. The juvenile Struthio sample showed significant differences in both long bones, which diminish with increasing size, a pattern confirmed with magnetic resonance imaging scans in an adult. Our data suggest that differences in extant archosaur long bone shape are greater in elements not utilized in locomotion and related stress-inducing activities. Based on our data, we propose tentative ranges of error for nonavian dinosaur long bone dimensional measurements. We also predict that calcified cartilage shape in adult, stressbearing nonavian dinosaur long bones grossly reflects chondroepiphysis shape. C 2010 Wiley-Liss, Inc. Anat Rec, 293:2044–2055, 2010. V

Key words: cartilage; morphometrics; alligator; bird; dinosaur; locomotion

Grant sponsor: Western Illinois University URC (University Research Council); Grant number: 3-30185; Grant sponsor: College of Arts and Sciences Graduate Student grant. *Correspondence to: Matthew F. Bonnan, Department of Biological Sciences, Western Illinois University, Macomb, IL 61455. E-mail: [email protected] C 2010 WILEY-LISS, INC. V

Received 3 March 2010; Accepted 4 August 2010 DOI 10.1002/ar.21266 Published online 2 November 2010 in Wiley Online Library (wileyonlinelibrary.com).

JOINT SHAPE IN ARCHOSAURS

Given their large average body size (>1 metric ton) (e.g., Farlow et al., 1995), many studies of nonavian dinosaurs have focused on locomotion and weight-support (e.g., Christiansen, 1997, 1999a,b; Carrano, 2001; Hutchinson and Garcia, 2002; Gatesy et al., 2009). Unfortunately, the epiphyseal joint cartilage (chondroepiphysis) that capped the ends of dinosaur long bones is lost during fossilization (Chinsamy-Turan, 2005). These missing data are problematic given that all models concerning dinosaur locomotion are ultimately derived from inferences of limb bone articulations. How much size and shape data are lost, and how these factors influence dinosaur locomotor models, remains poorly understood. Nonavian dinosaur long bones grew like those of their closest living archosaur relatives, crocodylians and birds, wherein the epiphyses remained cartilaginous throughout life, forming no secondary ossification center as in mammals and lizards (Haines, 1969; Carter and Beaupre´, 2001; Horner et al., 2001; Chinsamy-Turan, 2005). As archosaur long bones grow in length, the chondroepiphysis grows by addition of new chondrocytes whereas older cells accumulate and calcify, forming a calcified cartilage zone (Chinsamy-Turan, 2005). It is this calcified cartilage atop the bone shaft that is utilized in lieu of the chondroepiphysis to infer locomotion and range of motion in nonavian dinosaurs (Chinsamy-Turan, 2005). Previous preliminary data (Holliday et al., 2001) showed that chondroepiphysis removal from alligator and bird long bones significantly altered their linear dimensions: original length was truncated (range, 6%–10%) and marked shape changes occurred (Holliday et al., 2001). More recently, a well-preserved humerus from the sauropod dinosaur Cetiosauriscus shows thick and extensive fossilized epiphyseal cartilage on its distal end (Schwartz et al., 2007). However, although long bone size may change dramatically after chondroepiphysis removal, it is not clear whether the same trend holds quantitatively for calcified cartilage shape: whereas long bone dimensions will truncate after chondroepiphysis removal, calcified cartilage shape may remain unaffected. For nonavian dinosaurs, the preserved calcified cartilage might still retain significant information on their articular surfaces to act as a reliable proxy for joint shape. Here, we test the null hypothesis that calcified cartilage shape in extant archosaur long bones will not differ significantly from the overlying chondroepiphysis. An ontogenetic series of long bones with intact epiphyses from American alligators, helmeted guinea fowl, and a sample of hatchling and juvenile ostriches were measured and digitized for geometric morphometric (GM) shape analyses before and after chondroepiphysis removal. Magnetic resonance imaging (MRI) scans of an adult ostrich were used to constrain patterns observed in the hatchlings and juveniles.

MATERIALS AND METHODS Specimens To test our hypothesis, we selected taxa that comprise the extant phylogenetic bracket (EPB) of nonavian dinosaurs, Crocodylia and Aves (Brochu, 2001; Hutchinson, 2006). For Crocodylia, we selected Alligator mississippiensis due to its availability, its nonendangered status, and the fact that the anatomy of this crocodylian is generalized (Meers, 2003). Thirty-six wild American alliga-

2045

tor (Alligator mississippiensis) specimens (femur length: range, 37–160 mm) were collected and euthanized by Louisiana Department of Wildlife and Fisheries biology staff under general scientific collection permits on the state-owned Rockefeller Wildlife Refuge (RWR; Grand Chenier, Louisiana) as part of an annual harvest. Specimens used in this study were salvage from other studies. Given that Alligator mississippiensis grows indeterminately, identification of adult and juvenile specimens is exceptionally difficult. We addressed this difficulty by dividing the Alligator sample into a ‘‘small’’ and ‘‘large’’ group using the median length of the femur. Given our sample size (n ¼ 36; range ¼ 37–160 mm), the median was calculated to be 86 mm: those specimens under the median were categorized as small, and those above were categorized as large. Although not ideal, this avoided problems associated with determining whether an individual had become sexually mature (data not readily available to us from these salvaged specimens) at a smaller or larger size. We are also following a precedent set in other studies (e.g., Bonnan et al., 2008) where, because of similar constraints, specimens were grouped by size rather than sexual maturity. For Aves representatives, we selected one volant species (Numida meleagris) and one nonvolant species (Struthio camelus). Numida were selected because they are readily available and spend time both walking on the ground and flying, and because they utilize their forelimbs in vigorous activity. Twenty-nine deceased domestic, free-range guinea fowl (Numida meleagris) specimens from hatchling to adult size were donated as salvage from Tim Piper of Macomb, Illinois. For our nonvolant representative, we selected Struthio camelus because it bears significant weight on its hindlimbs, and its forelimbs are used for display only. We purchased 25 dead hatchling to juvenile ostriches and 1 adult (Struthio camelus) as salvage from Freedom Sausage Ostrich Farm in Earlville, Illinois. All specimens were curated at Western Illinois University under a United States Federal Salvage Permit to MFB. Ideally, a large and varied set of archosaurs would be examined and compared with one another, but a number of difficulties prevented such an approach. First, among crocodylians, we had access to a large number of salvaged Alligator mississippiensis specimens, and this made this taxon the most viable option for obtaining a reasonably sized sample of specimens for statistical analysis. Second, certain domestic birds, such as Gallus gallus or Meleagris gallopavo, have been bred for human consumption and many such birds spend their lives relatively immobilized. Moreover, although available commercially in great numbers, the aforementioned species are usually culled before maturity, and therefore complete ontogenetic series are difficult to come by. Third, a sample of ostrich chicks and juveniles, rather than a complete ontogenetic series, was the most practical approach for obtaining data from a free range terrestrial bird and provided us with a statistically meaningful sample that could be compared with those of Numida and Alligator. Although a sample which included adult and subadult Struthio would certainly have been ideal, obtaining a large enough sample size, and controlling the maceration of the bones (see below) would have been impractical. However, recognizing that a sample of

2046

BONNAN ET AL.

Fig. 1. Measurement and qualitative analysis of long bone shape before and after chondroepiphysis removal in selected archosaurs. Measurement scheme adapted from Bonnan (2004, 2007) and Bonnan et al. (2008). Not illustrated: craniocaudal measurement of greater trochanter

width on bird femora. Laser scans of humeri (cranial view) and femora (caudal view) from Alligator, Numida, and Struthio (juvenile only) show shape differences before and after chondroepiphysis removal. Numbered points indicate digitized landmarks (detailed in Table 1).

hatchling and juvenile Struthio specimens would provide results skewed toward smaller individuals, we utilized MRI scans of an adult ostrich to confirm the trends suggested by our Struthio sample for larger individuals.

saurs (Farlow et al., 2005; Bonnan et al., 2008) and are used here as a proxy for body size. The shoulder, elbow, hip, and knee joints of each specimen were dissected, measured (Fig. 1), and photographed before and after chondroepiphysis removal. Photographs were taken of each limb element before and after chondroepiphysis removal for GM shape analysis. Given that hyaline cartilage is a water-rich tissue, exposure to air will result in its dehydration, shrinkage, and thus artificial distortion of dimensional and shape data. To avoid data distortion, we developed a method that allowed for controlled maceration of muscles and hydration of the chondroepiphysis. First, frozen limbs were thawed and dissected at the joints, and dimensional measures were made with digital calipers as the cartilage was exposed. We note that freezing of epiphyseal cartilage does not result in significant cell destruction and is in fact less damaging to cartilaginous tissues than chemical fixation (e.g., Hunziker

Measurements and Data Collection We chose to focus on shape changes in the chondroepiphysis of the humerus and femur because these elements are usually the most common, least distorted, and most often compared bones in dinosaur studies. These two limb elements provide much of the mechanical support to the forelimbs and hindlimbs, and many of the largest muscles, which control locomotor movements of the entire limb, insert or originate on the humerus or femur (Bonnan, 2007). Moreover, humerus and femur length are highly correlated with body size in mammals (e.g., Christiansen, 1999a), dinosaurs, and other archo-

2047

JOINT SHAPE IN ARCHOSAURS

TABLE 1. Landmark Points Digitized on Each Element Element Humerus

Femur

Number Alligator

Numida

Struthio Medial extent of humeral head Lateral extent of humeral head Lateral constriction of midshaft Lateral epicondyle Intersection of medial and lateral condyles Medial epicondyle

1 2 3 4 5

Medial extent of humeral head Lateral extent of humeral head Deltopectoral crest Lateral constriction of midshaft Lateral epicondyle

Medial extent of humeral head Lateral extent of humeral head Deltopectoral crest Lateral constriction of midshaft Lateral epicondyle

6

Intersection of medial and lateral condyles Medial Epicondyle Medial constriction of midshaft Medial extent of femoral head Lateral extent of femoral head Fourth trochanter Lateral constriction of midshaft Lateral epicondyle Intersection of medial and lateral epicondyles Medial epicondyle Medial constriction of midshaft

Intersection of medial and lateral condyles Medial Epicondyle Medial constriction of midshaft Femoral head Greater trochanter Lateral extent of proximal end Lateral constriction of midshaft Lateral epicondyle Intersection of medial and lateral epicondyles Medial epicondyle Medial constriction of midshaft

7 8 1 2 3 4 5 6 7 8

Medial Constriction of Midshaft N/A Femoral head Greater trochanter Lateral extent of proximal end Lateral constriction of midshaft Lateral epicondyle Intersection of medial and lateral epicondyles Medial epicondyle Medial constriction of midshaft

See Fig. 1 for illustrations of these landmark points.

et al., 1984). Although freezing may compromise the mechanical abilities of epiphyseal cartilage (e.g., Kennedy et al., 2007), we are not aware of any published studies showing that chondroepiphysis shape is adversely affected. Following our measurements, large muscles and other obstructing tissues were removed, and the partially defleshed limbs were submerged in water heated to 60
C, a temperature at which cartilage shape has been shown to remain unaltered (Wright et al., 2005). This water bath allowed for soft tissue removal and exposure of the chondroepiphysis without damaging its shape. To ensure that our method did not cause a significant effect, we remeasured the bone dimensions after 24 hr of immersion. After photographing the limb elements in a standardized orientation, water temperature was increased to 100
C to remove the chondroepiphysis. Finally, an additional series of post chondroepiphysis removal measurements and photographs were taken. C desktop laser scanner was used to A NextEngineV scan the surfaces of one set of juvenile (small) and adult (large) Alligator and Numida elements, and one set of juvenile Struthio elements, before and after chondroepiphysis removal. This allowed us to generate threedimensional models to compare with our statistical data. Care was taken to moisten the epiphyseal cartilage between scans with water to prevent dessication and distortion of shape, and we followed the maceration procedure described above. MRI scans of an adult Struthio camelus were performed at the Rockford Memorial Hospital. We used a 1.5 Tesla (Signa; General Electric, Milwaukee, WI) scanner to obtain proton density-weighted images with frequency specific, fat-suppression (see Novelline, 2004). This type of imaging sequence accentuates the signal intensity from hyaline cartilage. The hindlimb of the adult ostrich specimen was scanned intact, but was dissected from the body wall just before freezing and transport to the MRI facility. Given that frozen tissues are nearly opaque to MRI scans, the ostrich hindlimb was thawed before its analysis. Although it is conceivable that our dissection of the limb may have caused some distortion

of the cartilage tissues in the adult ostrich, we also scanned dismembered specimens of a juvenile ostrich, a juvenile and adult guinea fowl, and an adult alligator. In these other specimens, MRI scans showed our predicted match-up between the underlying calcified cartilage and the chondroepiphysis. In other words, we saw nothing different in MRI scans of these taxa that we did not already observe in our morphometric data and surface laser scans.

Morphometric Analyses SPSS software (v. 16) was used for statistical analysis of all linear data and the MANOVA and principal components analysis (PCA) of the partial warps data generated from GM analyses. All linear measurements of bone dimensions were made with digital calipers, log10 transformed to normalize their distribution (Zar, 1999), and tested for normality using the Kolmogorov–Smirnov test. Repeated measures ANOVAs tested for significant linear differences both in the method and after chondroepiphysis removal. For GM analysis, we used two-dimensional thin-plate splines (TPS) because this technique is ideal for analyzing a set of objects (limb bones) that are similar in overall morphology and where the detection of more subtle shape differences is desired (Zelditch et al., 2004; Slice, 2005). In a TPS analysis, homologous landmark coordinates of all specimens are aligned, rotated, and scaled into a grand mean reference form via generalized least squares Procrustes superimposition (Zelditch et al., 2004; Slice, 2005). Measuring the sum of squared Procrustes distances in the homologous landmark coordinates of each specimen against the reference form reveals shape differences, which can be analyzed mathematically and visualized as a deformation grid or TPS (Bookstein, 1991). Normalized shape coefficients generated from the sum of squared Procrustes distances (partial warps) are correlated, dependent variables that collectively describe shape and are analyzed with

2048

BONNAN ET AL.

TABLE 2. Test of Normality and Percentage of Dimensional Truncation after Chondroepiphysis Removal in Alligator mississippiensis

Element

Measurement

Humerus Length (n ¼ 35) Proximal breadth Deltopectoral crest Distal breadth Femur Length (n ¼ 36) Proximal breadth Fourth trochanter Distal breadth

TABLE 3. Test of Normality and Percentage of Dimensional Truncation after Chondroepiphysis Removal in Numida meleagris

KS % of remaining normality after truncation Test (P) all/small/large 0.200

93/92/94

0.200 0.200 0.200 0.200

91/90/91 90/88/93 90/90/90 94/93/95

0.200 0.200 0.200

92/92/92 91/89/93 96/92/94

Komolgorov–Smirnov (KS) tests of normality are reported for each measurement.

standard multivariate statistics (Zelditch et al., 2004; Slice, 2005). Limb bones were digitized and analyzed using the TPS program suite developed by Rohlf (TPSUtil, TPSDig2, TPSRelw; 2008). The landmarks selected for digitization followed standard landmarks detailed elsewhere (Bonnan, 2004, 2007; Bonnan et al. 2008) (Fig. 1). Changes in limb bone morphology associated with landmarks are indicated by numbers in parentheses in the text. Sliding semilandmarks were also digitized to capture the outline of the articular surfaces (see Zelditch et al., 2004). As partial warp scores are dependent variables that together describe shape, a MANOVA of these variables was used to detect significant differences in limb bone shape before and after chondroepiphysis removal. A PCA of the partial warps calculated so-called relative warps or components of maximum shape variation (Zelditch et al., 2004; Slice, 2005). Deformation grids of the significant principal components of shape (PRINs) were generated by the tpsRelw program (Rohlf, 2008) and used to visualize these shape changes. See Table 1 for landmark descriptions. A potential limitation of this study is the compression of three-dimensional long bone shapes into two-dimensional coordinates. Although three-dimensional TPS applications are available (Zelditch et al., 2004), the time required to capture such coordinate data was difficult given that long-term exposure of the chondroepiphysis results in its dehydration. For this reason, rapid, two-dimensional photography was preferred. Certainly, we recognize the value in capturing three-dimensional long bone shape data and methods for digitizing threedimensional shape data from wet bones are currently being explored.

RESULTS Linear Data Linear data from all taxa except Numida showed normal distributions (Tables 2–4). For the Numida sample, the non-normal signal (Komolgorov-Smirnov P < 0.05 for all measurements) is explained by the fact that the smallest juveniles and full-grown adults cluster closely together at the lower and upper ends of the size range,

Element

Measurement

Humerus Length (n ¼ 29) Proximal breadth Deltopectoral crest Distal breadth Femur Length (n ¼ 29) Proximal breadth Greater trochanter Distal breadth

KS % of remaining normality after truncation test (P) all/juveniles/ all/juveniles adults 0.004/0.032

92/86/98

0.014/0.114 0.008/0.132 0.004/0.069 0.021/0.105

86/78/94 86/79/91 84/75/93 93/90/98

0.030/0.141 0.027/0.200 0.008/0.104

88/76/99 80/70/97 86/77/96

Komolgorov–Smirnov (KS) tests of normality are reported for each measurement. TABLE 4. Test of Normality and Percentage of Dimensional Truncation after Chondroepiphysis Removal in Struthio camelus (juveniles)

Element

Measurement

Humerus Length (n ¼ 25) Proximal Breadth Deltopectoral Crest Distal Breadth Femur Length (n ¼ 25) Proximal Breadth Greater Trochanter Distal Breadth

KS % of remaining normality after truncation test (P) all/smallest/largest 0.200

85/84/86

0.200 0.073 0.200 0.200

58/52/55 77/75/79 53/51/55 86/85/88

0.053 0.200 0.200

80/76/83 80/77/81 72/69/75

Komolgorov–Smirnov (KS) tests of normality are reported for each measurement.

respectively. This was established statistically. When only Numida juveniles were tested, a normal distribution was reported for all variables except humerus length (Kolmogorov-Smirnov P > 0.05; Table 3). Repeated measures ANOVA tests show no significant effect of the method on long bones of Alligator and adult Numida (Table 5). Significant effects did occur with the juvenile Numida and Struthio samples, but these are due to swelling of the chondroepiphysis, an effect that did not appear to alter our shape data (see below). Removal of the chondroepiphysis causes significant truncation of all humerus and femur dimensions in Alligator and juvenile birds (Table 5). For adult Numida, significant truncation of all humerus dimensions occurs, but only the greater trochanter width is significantly shortened on the femur (Table 5). On average, Alligator long bone dimensions are truncated from 4%–10% of the original length (Table 2), those of Numida are truncated from 7%–20% (Table 3), and Struthio juveniles show nearly 50% truncation in some dimensions (Table 4). Although these data agree generally with previous preliminary data (Holliday et al., 2001) on alligators and birds, the averages mask an intriguing but intuitive trend: in both Alligator and the bird taxa, larger individuals show less truncation than smaller individuals (Fig. 2).

2049

JOINT SHAPE IN ARCHOSAURS

TABLE 5. Test of Maceration Method and Effect of Chondroepiphysis Removal Using Repeated Measures ANOVA Taxon Alligator Numida (All)

Maceration of humerus

Maceration of femur

None All

None Length, greater trochanter width, distal breadth Distal Breadth None

Numida (Juveniles) Numida (Adults)

Distal breadth None

Struthio (Juveniles)

Distal Breadth

Length, proximal breadth, greater trochanter width

Chondroepiphysis removal: humerus

Chondroepiphysis removal: femur

All All

All All

All All

All Greater trochanter width All

All

Measurements, which were significantly different (P < 0.05) after maceration and/or chondroepiphysis removal, are denoted in bold. All indicates every measurement was significantly effected either by the maceration method or chondroepiphysis removal.

For Alligator the trend is subtle, with large individuals showing about 1%–5% less truncation in their long bones after chondroepiphysis removal than small specimens (Table 2 and Fig. 2). For Numida the trend is stark: on average, adults show at least 8% less difference in truncation compared with juveniles, and in some cases the difference between adult and juvenile dimensional loss is much greater (>10%; Table 3 and Fig. 2). Struthio juveniles show greater differences between smaller and larger individuals (range, 2%–7%) than Alligator and are missing the most significant amount of dimensional data from the proximal and distal breadth of the humerus at any size (Table 4). However, these major losses in data are probably due to the immature stage of these animals and do not reflect a trend that continues into adulthood. Bivariate plots of the humerus and especially the femur of Struthio show a clear trend toward diminishing epiphyseal cartilage thickness as size increases (Fig. 2). Moreover, for the femur, Struthio specimens show the steepest, most positive slope of the three examined taxa, indicating that even small changes in body size have a significant, negative effect on cartilage thickness. In fact, MRI scans of the Struthio adult show that as in Numida adults only a thin layer of epiphyseal cartilage is present on the humerus and femur (Fig. 4). It should be noted that the truncation in length does not appear to happen equally at the proximal and distal ends of the long bones in these archosaur taxa. For example, truncation is greater for the distance measured between the deltopectoral crest and humeral head than for overall humerus length (Tables 2–4), a result explained by a thicker proximal chondroepiphysis. If both proximal and distal epiphyses were of equal thickness, little or no difference in this measurement would be predicted.

GM Data GM shape comparisons via TPS of the humerus and femur before and after chondroepiphysis removal revealed several patterns. For the humerus of Alligator and both elements in Numida, it is the second principal component (PRIN) 2, rather than PRIN 1 that describes epiphyseal shape. Because shape PRINs record the maximum amount of variation, PRIN 1 often picks up individual variation. For both Alligator and Numida, individual nuances in deltopectoral crest orientation in the humerus accounted for the most shape variation in

the sample. This is not unprecedented, as a similar trend appears in dinosaur taxa for the same landmark (Bonnan, 2007). For the femur of Numida, differences in shaft bending among individuals accounted for most of the shape variation in PRIN 1. Examination of shape changes associated with PRIN 2 in these taxa and for these elements clearly showed that chondroepiphysis shape was linked strongly with this component. In Alligator, humerus shape changes significantly across the sample as might be expected (Table 6), and deformation grids show a loss in condyle distinctness (landmarks 5–7 and their associated semilandmarks) and humeral head curvature (landmarks 1–2 and their associated semilandmarks; Fig. 3). However, although small Alligator show significant femur shape differences (again associated with reduced condyle prominence; landmarks range 5–7 and their associated semilandmarks), large Alligator femora show none (Table 6). For Numida, juveniles show significant humerus shape differences after chondroepiphysis removal (flattened condyles; landmarks range 5–7 and their associated semilandmarks), but surprisingly no significant difference is reported for adults (Table 6). Moreover, both juvenile and adult Numida individuals show no significant difference in femur shape (Table 6). As expected, the sample of hatchling and juvenile Struthio show significant shape changes in all humerus and femur landmarks (Table 6). When the principal components (PRINs) of shape change are plotted, a general size-associated trend is discernable for all taxa. For Alligator, humerus and femur shape differences diminish before and after chondroepiphysis removal with increasing size (Fig. 3), a trend also apparent in both bird taxa (Fig. 3). These data show that with the exception of the significant result for the Alligator humerus, as size increases the shape of the underlying calcified cartilage more closely approximates that of the overlying chondroepiphysis. Although we did not have a complete ontogenetic series of Struthio, MRI imaging of an adult specimen confirm that this pattern of shape convergence continues into adulthood for this taxon (Fig. 4).

DISCUSSION Summary of Results and Implications for Extant Archosaurs The results of our linear data support previous analyses in general (e.g., Holliday et al., 2001): size and linear

2050

BONNAN ET AL.

Fig. 2. Bivariate plots of element length against the percent of the original element length after truncation for (A) the humerus and (B) the femur in Alligator, Numida, and Struthio. In the graphs, Alligator are represented by open circles, Numida by X symbols and Stuthio by þ signs. Note that all taxa show trends toward less element truncation with increasing specimen size. For both the humerus and femur, Alligator shows a gradual trend toward a decreasing amount of truncation. In contrast, the bird taxa show both a greater range of dimensional loss and a more rapid trend toward decreasing truncation with increasing size. In particular, note the steep slope for Struthio juveniles with increasing femur size. All data were measured in millimeters and log10 transformed.

dimensions of the humerus and femur are significantly truncated after chondroepiphysis removal. However, our data also show a negative correlation between size and chondroepiphysis thickness. Without exception, less truncation occurs in humerus and femur dimensions for larger individuals than in smaller specimens for Alligator, Numida, and Struthio. Our GM results show that a surprising amount of shape information is retained in the calcified cartilage, especially in larger individuals. Overall, two general trends can be discerned from our shape data. First, significant differences in shape between the chondroepiphysis and underlying calcified

cartilage occur most often in juveniles, especially in the humerus. Second, shape differences tend not to be significant in adults, especially in the femur. We find it significant that shape differences before and after chondroepiphysis removal diminish as specimen size increases. Notably, the femur, which is weight-bearing and the main locomotor element in most archosaurs, shows nonsignificant shape differences in large Alligator and across all Numida. Even in juvenile Struthio femora, a clear trend is observed toward a closer approximation of calcified cartilage with the chondroepiphysis as size increases. These shape data complement our reported linear trend of diminishing relative chondroepiphysis thickness with increasing size: the calcified cartilage and chondroepiphysis become more intimately associated in larger individual archosaurs in our sample. For the humerus, only adult Numida show a nonsignificant difference between the calcified cartilage and chondroepiphysis shape, whereas the effects in Alligator and juvenile birds are significant. Our linear and shape data suggest collectively that long bones experiencing the primary stresses associated with locomotion or wing-beating will have thinner epiphyseal cartilage and a calcified cartilage shape, which approximates that of the overlying chondroepiphysis. Conversely, our data suggest that long bones, which are not primarily involved in locomotion or flight, will have thicker epiphyseal cartilage and thus show greater differences in shape between the calcified cartilage and chondroepiphysis. Overall, our linear and shape data agree with general trends reported for long bones: epiphyseal cartilage is less thick and more closely associated with calcified cartilage in regions of greatest stress (Frost, 1990a,b,c,d; Carter and Beaupre´, 2001). An apparent exception to the trend we report occurs in Alligator humeri. We note that the humerus, which is used extensively for locomotion, shows significant shape differences in even the largest specimens. This result is perhaps explained by a combination of tail drag and semiaquatic habits in these archosaurs. In crocodylians, drag force from the heavy tail exerts a significant pull during locomotion, and the hindlimb experiences greater stress than the forelimb as it overcomes locomotor and tail drag forces (Willey et al., 2004). Essentially, the center of mass in these archosaurs is shifted toward the hindlimbs. It is also significant that older crocodylians spend more time in water where their hindlimbs and tail play a greater role in locomotion than their forelimbs (Ross and Garnett, 1989). These ecological factors could help explain why the humerus, although loaded and stressed during locomotion, still shows a significant difference in shape between the calcified cartilage and chondroepiphysis into the largest Alligator specimens in our sample. We also acknowledge that avian long bones grow rapidly and cease growth when adulthood is reached, in contrast to the slow, indeterminate growth of Alligator (Horner et al., 2001; Chinsamy-Turan, 2005). These different growth patterns could contribute to the signals we see in our data. For example, cessation of chondroepiphyseal growth in adult Numida will result in more ossification and calcification underneath this region, a pattern somewhat reminiscent of determinate long bone growth reported for mammals (Carter and Beaupre´, 2001). However, a nonsignificant shape difference was

2051

JOINT SHAPE IN ARCHOSAURS

TABLE 6. PCA and MANOVA of Partial Warp Scores to Test for a Significant Difference in Shape before and after Chondroepiphysis Removal in Selected Archosaur Taxa Taxon

Element

Alligator

Humerus Femur Humerus Femur Humerus Femur

Numida Struthio

Shape PRIN (%) PRIN PRIN PRIN PRIN PRIN PRIN

2 1 2 2 1 1

(25%) (37%) (20%) (31%) (50%) (47%)

F-Statistic

P (Juveniles)

P (Adults)

15.978 4.440 1.604 2.587 20.541 34.829

0.019 0.050 0.026 0.550