Determinate growth is predominant and likely ancestral in squamate reptiles

Body growth is typically thought to be indeterminate in ectothermic vertebrates. Indeed, until recently, this growth pattern was considered to be ubiquitous in ectotherms. Our recent observations of a complete growth plate cartilage (GPC) resorption, a reliable indicator of arrested skeletal growth, in many species of lizards clearly reject the ubiquity of indeterminate growth in reptiles and raise the question about the ancestral state of the growth pattern. Using X-ray micro-computed tomography (µCT), here we examined GPCs of long bones in three basally branching clades of squamate reptiles, namely in Gekkota, Scincoidea and Lacertoidea. A complete loss of GPC, indicating skeletal growth arrest, was the predominant finding. Using a dataset of 164 species representing all major clades of lizards and the tuataras, we traced the evolution of determinate growth on the phylogenetic tree of Lepidosauria. The reconstruction of character states suggests that determinate growth is ancestral for the squamate reptiles (Squamata) and remains common in the majority of lizard lineages, while extended (potentially indeterminate) adult growth evolved several times within squamates. Although traditionally associated with endotherms, determinate growth is coupled with ectothermy in this lineage. These findings combined with existing literature suggest that determinate growth predominates in both extant and extinct amniotes.

. Specification of BiSSE and HiSSE models used for ancestral state analysis

Supplementary Results
Outcomes of parsimony and likelihood models are summarized, and data gathered in this study are provided. Figure S1. Ancestral state reconstruction of growth type in squamates, maximum parsimony model Figure S2. Ancestral state reconstruction of growth type in squamates, AIC-weighted average of two best fit maximum likelihood models Figure S3. Ancestral state reconstruction of growth type in squamates, AIC-weighted average of all five acceptable maximum likelihood models Figure S4. Mid-diaphyseal transverse cross-sections of the femur in four species of adult lizards with resorbed GPC. Table S2. Epiphyseal state in the proximal epiphysis of the femur in the examined specimens representing clades Gekkota, Scincoidea and Lacertoidea Video File S1. The visualization of transversal cross-sections of the proximal part of the femur by µCT in adult Timon tangitanus (ID 120).

µRTG and µCT examinations
The Bruker SkyScan 1275 µCT scanner was used for scanning of large samples, while a custom-built µCT system utilizing large-area photon counting detectors based on Timepix technology [1] was used for smaller samples, since higher resolution and higher contrast-to-noise ratio could be achieved using this set-up [2]. The scan parameters were adjusted for each sample individually to reflect its size and attenuation properties (for more technical details see [3,4]. The voxel-size of the reconstructed slices was within the range of 4 -13 µm. Data analysis was carried out using Bruker CTVox [5] and Fiji [6].

Ancestral state reconstruction
We fit a total of 12 models of trait evolution and diversification using the R packages hisse [7], including the full HiSSE model, subsets of the full HiSSE model with different constraints on transitio n rates, BiSSE-like models without hidden states and null-models with diversification rates independent of the observed states (all input data are in Dataset 2). BiSSE-like model refers to a model in which there are no hidden states, turnover and extinction rates vary across states and there are two transition rates (i.e., state 0 to 1 and 1 to 0). Full HiSSE model refers to a HiSSE model in which each observed state is associated with two hidden states (i.e., states 0A, 0B, 1A, 1B), turnover and extinction rates vary across all four states, and there are eight character transition rates, with dual transitions (e.g., state 0A to 1B) not allowed. All other models are subsets of the full HiSSE model with various constraints on diversification rates and transition rates, or "null" models with diversification rates independent of the states but still allowing for varying diversification rates (for hidden states). All models are described in detail in Table S1 below.
In all analyses, we accounted for incomplete species sampling by setting the sampling fraction of species in each state of the observed trait. We estimated the total number of recent species with a given state for each high-level clade, multiplying the number of species in the clade included in the Reptile database [8] by the fraction of species with the given state in our dataset, assuming that our sampling gives a good approximation of the states proportion. For clades not represented in our data, we assumed a conservative proportion of 50% for each state. For the phylogenetic relationships we used a timecalibrated phylogeny of 4162 squamate species [9]. The best model fit was selected based on the Akaike information criterion (AIC) and the composite models were created with AIC weighted average of the model fits with Δ AIC < 4 (2 best models) or 10 Δ AIC < 10 (all acceptable models).

Histological examinations
We analysed histology of femoral bones in four fully grown individuals representing phylogenetically distantly related species, namely at least 9-year old female of Yellow-throated plated lizard Gerrhosaurus flavigularis (ID 99); at least 5-year old male of Balkan green lizard Lacerta trilineata major (ID 640); nearly 12-year old male of Kuhl's flying gecko Ptychozoon kuhli (ID 358) and at least 7-year old female of Common leopard gecko Eublepharis macularius (ID 187). All these individuals were captive bred and featured complete resorption of the GPC. Samples were fixed in ethanol, decalcified in 8% nitric and 8% hydrochloric acid solution for 7 hours, dehydrated in graded ethanol series and embedded in paraffin [10]. Mid-diaphyseal regions of femoral bones were transversally sectioned at a 15μm thickness by a rotary microtome. Diaphyseal cross-sections were mounted on glass slides, stained with Ehrlich's haematoxylin and examined under bright field illumination, phase contrast, Nomarski interference contrast and in polarized light at 400x magnification. A minimum of thirty sections per bone were examined.

Ancestral state reconstruction
Maximum parsimony model suggests that the last common ancestors of squamate reptiles was a determinate grower ( Figure S1), as do the maximum likelihood models (Figures 2, S2, S3). The best-fit likelihood model (the HiSSE equal rate model with state-dependent diversification rates and transition rates equal across observed and hidden states) support determinate growth as the ancestral state for Squamata (85.5%, Figure  2), the same is true for model-averaged reconstructed states for the two best models (ΔAIC < 3; 88%, Figure  S2) and all acceptable models (ΔAIC < 10; 85%, Figure S3). Figure S1. Ancestral state reconstruction of growth type in squamates, maximum parsimony model. A circular tree depicting the growth plate cartilage (GPC) state in whole Squamata as revealed by µRTG and µCT examination of the proximal part of femoral bones. Ancestral state reconstruction method was employed using parsimony model to uncover the evolution of growth type (determinate vs. indeterminate) in Squamata. GPC present (green) and absent (red) is suggesting extended (potentially indeterminate) vs determinate body growth. Tuatara (Sphenodon punctatus), as a sister group of Squamata, was included as an outgroup. The state of tuatara is according to the presence of external fundamental system and recapture growth data suggesting the determinate type of body growth [11,12]. Species marked with asterisk were scored according to the GPC state from literature [13]. Species marked with † were very old individuals (for details of age see references [3] and [4]).

Figure S2. Ancestral state reconstruction of growth type in squamates, AIC-weighted average of two best fit maximum likelihood models.
A circular tree depicting the growth plate cartilage (GPC) state in whole Squamata as revealed by µRTG and µCT examination of the proximal part of femoral bones. Ancestral state reconstruction method was employed using maximum likelihood with hidden state speciation and extinction models to uncover the evolution of growth type (determinate vs. indeterminate) in Squamata. GPC present (green) and absent (red) is suggesting extended (potentially indeterminate) vs determinate body growth. Tuatara (Sphenodon punctatus), as a sister group of Squamata, was included as an outgroup. The state of tuatara is according to the presence of external fundamental system and recapture growth data suggesting the determinate type of body growth [11,12]. Species marked with asterisk were scored according to the GPC state from literature [13]. Species marked with † were very old individuals (for details of age see references [3] and [4]).

Figure S3. Ancestral state reconstruction of growth type in squamates, AIC-weighted average of all five acceptable maximum likelihood models.
A circular tree depicting the growth plate cartilage (GPC) state in whole Squamata as revealed by µRTG and µCT examination of the proximal part of femoral bones. Ancestral state reconstruction method was employed using maximum likelihood with hidden state speciation and extinction models to uncover the evolution of growth type (determinate vs. indeterminate) in Squamata. GPC present (green) and absent (red) is suggesting extended (potentially indeterminate) vs determinate body growth. Tuatara (Sphenodon punctatus), as a sister group of Squamata, was included as an outgroup. The state of tuatara is according to the presence of external fundamental system and recapture growth data suggesting the determinate type of body growth [11,12]. Species marked with asterisk were scored according to the GPC state from literature [13]. Species marked with † were very old individuals (for details of age see references [3] and [4]).

Histological examinations
Transverse cross-sections through mid-diaphysis of the femur stained with Ehrlich's haematoxylin were examined in four individuals with fully resorbed GPCs and known age to assess whether arrest of longitudinal bone growth is associated with arrest of bone growth in girth. In three out of four femoral bones examined, the lines of arrested growth (LAGs) were not clearly visible ( Figure S4 ac), which might reflect aseasonal growth in captive bred individuals. Yet, tightly spaced rings of laminar bone depositions were observed in the outer cortex of these bones ( Figure S4 a-c), a clear indication of decelerated or ceased periosteal growth. In one bone only, we were able to observe LAGs forming the external fundamental system (EFS) (Figure S4 d). Because we know that this individual was at least 7 years old, it seems that this individual has stopped growing in the fourth or fifth year of life. Thus, complete resorption of the femoral GPC is coupled with well-developed EFS in this animal. Taken together, histological examinations performed in this study strongly suggest that periosteal growth is decelerated, if not arrested, in animals with arrested longitudinal growth.

Technical considerations
There are several issues that need to be considered for the sake of an unbiased interpretation of the data presented here. First and foremost, we use the disappearance of the femoral GPC as an indicator of whole-body longitudinal growth cessation, assuming that growth of the femur is synchronized with growth of the axial skeleton. While the timing of GPC degradation is not necessarily synchronous in different bones or the two epiphyses of a single bone [13,14], this assumption still seems reasonable. Length of the femur is highly correlated (r 2~0 .93) with snout-vent length (SVL) in lizards [15], which is in line with a tight coupling and synchronization of growth between the axial and appendicular skeleton.
Second, animals included in our analyses are assumed to be skeletally mature but not senescent. To select adult animals, we preferentially examined individuals that achieved at least 80% of the maximum body length reported in the literature, since 75% is the mean relative size at maturity found in lizards [16]. We were able to collect many fully or nearly fully-grown animals with SVL rel > 80%, with some analysed individuals exceeding the largest size ever reported for a given sex-species category (Table S2). In case of captive bred animals, we confirmed the approximate age and reproductive history of each investigated individual with the breeders. While extensive use of captive bred animals (they constituted two-thirds of the examined specimens) enabled us to exclude immature individuals, growth curves are likely different in wild populations, with captive animals exhibiting faster growth rates and reaching maturity sooner due to controlled environmental and dietary conditions (e.g., [17][18][19]). Whether this affects the timing of GPC degradation is unknown, but it should not change the inherent growth type. Indeed, our examination of wild and captive bred individuals yielded consistent results.
Finally, it has to be noted that the presence of the GPC in adults cannot be taken as conclusive evidence of indeterminate growth. While the presence of GPCs indicates the potential for longitudinal growth, it does not necessarily mean that the animal is actually growing. Because GPC degradation is triggered by exhaustion of the proliferative potential of growth plate chondrocytes, it has been suggested that complete GPC degradation might not precede, but rather follow the cessation of body growth [20,21]. Hence, our analysis may overestimate the number of species actually exhibiting extended (potentially indeterminate) adult growth, as we cannot differentiate cases where growth permanently stops while the GPC is still at least partly preserved. Nevertheless, a synchronous timing of GPC degradation, EFS development and body growth arrest has been recently demonstrated in a mammal [22]. Similar complex studies are needed to elucidate the timing and relationship between growth arrest and GPC degradation in squamate reptiles.