A comprehensive database of thermal developmental plasticity in reptiles

How temperature influences development has direct relevance to ascertaining the impact of climate change on natural populations. Reptiles have served as empirical models for understanding how the environment experienced by embryos can influence phenotypic variation, including sex ratio, phenology and survival. Such an understanding has important implications for basic eco-evolutionary theory and conservation efforts worldwide. While there is a burgeoning empirical literature of experimental manipulations of embryonic thermal environments, addressing widespread patterns at a comparative level has been hampered by the lack of accessible data in a format that is amendable to updates as new studies emerge. Here, we describe a database with nearly 10, 000 phenotypic estimates from 155 species of reptile, collected from 300 studies manipulating incubation temperature (published between 1974–2016). The data encompass various morphological, physiological, behavioural and performance traits along with growth rates, developmental timing, sex ratio and survival (e.g., hatching success). This resource will serve as an important data repository for addressing overarching questions about thermal plasticity of reptile embryos.


Background & Summary
Conditions experienced early in life are known to impact phenotypes in profound ways that can have longlasting effects on fitness 1,2 . Understanding the extent to which developmental environments impact phenotypes is important for addressing many fundamental questions in ecology and evolutionary biology [3][4][5] , as well as predicting the effect environmental change will have on populations both locally and globally 1,6 . Ectothermic vertebrates in particular are sensitive to variation in early developmental temperature, which often is mediated by local climatic conditions, landscape features as well as maternal nest site choice or maternal basing behaviour. In reptiles (turtles and tortoises, tuatara, lizards, snakes and crocodilians), there is a growing empirical literature testing the effects of early thermal environments (i.e., incubation temperatures) on the phenotypic development of a broad range of physiological, morphological and performance traits [7][8][9][10][11] . Nonetheless, there is currently no database collating and summarising this vast literature in a way that is amenable to updates as the literature grows or that can be expanded to address not only questions on the impacts of temperature, but also other environmental conditions (e.g., moisture, pH) that may be relevant to phenotypic development and survival.
Here, we describe a large database on the effects of incubation temperature on phenotypic traits in oviparous reptiles. Our database differs from others (e.g., BioTraits 12 ), in that it focuses primarily on thermal developmental plasticity by collating studies manipulating temperatures experienced during pre-hatching developmental periods only. Furthermore, ours is the first database on thermal developmental plasticity to provide an updatable platform summarising phenotypic effects of incubation temperatures. A smaller, preliminary version, of the dataset was thoroughly analysed in a related manuscript 1 and future plans are to expand the data to capture other environmental drivers of phenotypic variation, such as moisture, pH, and oxygen concentrations. In addition, although the database focuses on oviparous species, it can be expanded to include environmental effects on embryos of viviparous species.
As the database grows, we believe that it can be used to address a wide variety of questions. For example, some of the questions that are currently or have previously been addressed with the database include: • What are the overall magnitude of effects of incubation duration on phenotypic development?
Qualitative syntheses of this research area have provided an unclear picture of both the magnitude of effect temperature has on phenotypic development, and whether complex patterns alluded to in these reviews [13][14][15] can be explained by species-specific or study-level attributes. Using aspects of these data Noble et al. 1 have shown strong overall effects, independent of temperature differences between studies, but found little support for the hypothesis that much of the variation in effects could be explained by phylogeny. Nonetheless, more robust phylogenetic analyses may provide greater insight. suggests that thermal environments can affect the phenotype later in life, however, a more detailed analysis of what predicts variation among species will be worthwhile.
• Do extreme developmental temperatures elicit developmental stress? Novel environmental conditions, including extreme temperatures, are predicted to affect phenotypic variation 17 and can lead to compromised survival 1 through developmental stress. We are currently using the database on a more targeted set of traits explore the generality of this prediction.
• What are the shapes of developmental reaction norms? While previous analyses suggest that many thermal reaction norms for traits exhibit the expected 'thermal performance curve' shape 1 , we are currently exploring the database in more detail on sub-samples of the data to understand the shapes of reaction norms for specific traits that are highly represented in the data (e.g., body size and mass).
• How realistic thermal fluctuations change the impact thermal developmental environments have on phenotypic development? Previous analyses suggest that more natural, fluctuating conditions decrease the magnitude of phenotypic effects. However, as the database grows and more detailed and realistic thermal conditions are applied experimentally, resolving uncertainty surrounding this question will be possible. It will also be important in establishing the impact thermal conditions have in nature, where temperatures typically vary on daily and longer time scales.
• How does thermal developmental plasticity facilitate or impede invasion success and adaptation to changing climatic conditions? Phenotypic plasticity is predicted to play an important role in early stages of adaptation to changing environmental conditions 17,18 or novel environments (as encountered by invasive species 19 ) and thermal plasticity is expected to feature strongly in this process for ectotherms. As the database grows and thermal reaction norms are more thoroughly characterized in more species these questions maybe feasibly addressed. The Reptile Development Database can be accessed freely online via a user-friendly webpage (www.repdevo.com) that stores and lodges all versions of past databases in addition to the most up-todate version. This ensures reproducibility of analyses as the database is updated and evolves to include new types of data. Unpublished data can be submitted through downloaded data templates, and queries can be sought by emailing the team (contact details are on the webpage).

Methods
We searched for published literature  describing experiments that manipulated incubation temperature (i.e., pre-hatching developmental period only) in reptiles in Web of Science (v5.13.2) using the following 'title' or 'abstract' search terms: temperature* AND incubat*, along with one of the following: reptil*, lizard*, squamat*, snake*, turtle*, chelon*, testudin*, crocodil*, alligator*, tuatara*, sphenodon*. In addition, we considered all citations in three major reviews of the topic [13][14][15] , and included any additional papers from these sources not identified in our searches. For data currently included in this database, the studies had the following attributes: 1) research on an oviparous reptile (Class Reptilia; excluding birds); 2) employ an experimental manipulation of incubation temperature of eggs; 3) present data on hatching success, incubation duration, or post-hatching phenotypes; 4) consist of eggs that did not receive exogenous hormone application and yolk removal; 6) there was not a substantial delay between oviposition and experimental incubation (e.g., >48-hours). In some cases, papers could not be accessed and/or were in languages that were non-translatable. While we attempted to translate where possible, those that could not be were excluded from the database (approximately 8 studies). We included 684 publications from the primary scientific literature that were relevant based on the title of the paper. If exclusion of papers was not possible based on the title of the publication, we assigned a unique identification number to each publication and considered the abstract and full text for exclusion/inclusion criteria. 'Citations.csv' (Table 1; [Data Citation 1]) details the publications considered based on their fulltext, and we provide reasons for their exclusion (if excluded) along with full reference information.
Following exclusion / inclusion based on the criteria we had a dataset of 300 publications from which we extracted complete or partial data. From each publication, we extracted data into 'Database.csv' as Column Description paper_no Unique number assigned to each manuscript data was extracted from. This paper_no column matches the data_no column in Database.csv.

first_author_surname
The surname of the first author on the manuscript data was extracted from authors All authors on the manuscript, and ordered as specified when published year Year of manuscript publication Excluded (Y/N) for another reason, that is not specified above. More details can be found in the 'comments' column as appropriate.
comments Text to specify comments for any other reason the paper was excluded, or to provide more details about the manuscript. outlined in Table 2 (available online only) [Data Citation 1]. This included the focal information of incubation temperature regime, phenotypes of study, and summary results [mean, error, sample sizetype of error (e.g., standard error, standard deviation or 95% confidence interval) is indicated using a separate columnsee Table 2 (available online only)] for each phenotype and incubation treatment. We also noted the age of the specimens (assumes to be age 0, hatchlings, if not indicated) in the sample, the temperature fluctuation of treatments, the sex of the sample (assumes to be mixed if not indicated), whether the data were raw or adjusted (e.g., least square means) phenotypic means and the location of the population sample. Data were taken from text or tables from each manuscript, however, when data were provided in figures we extracted key information from these figures (assuming they were clear and readable) using DataThief 20 . In addition, we recorded contextual information regarding species, study design, cofactors included, and comments. Taxonomic naming was standardised using TimeTree.org 21 . Genus and species names not identified (N = 15 species) in TimeTree were cross checked with the EMBL Reptile Database 22 . Missing data were coded as 'NA'. All traits measured were classified into one of eight Trait Categories (see Table 2, (available online only)). Taxonomic representation of data was highest in the Orders Squamata (lizards and snakes -N = 75 species) and Testudines (turtles and tortoises -N = 69 species) in terms of raw numbers of studies (Fig. 1), although representation relative to species richness was highest in Orders Rhynchocephalia (tuatara, 1 species) and Crocodilia (alligators and crocodiles, - Traits classified under the category morphology were the most-commonly collected data (Fig. 2). These methods, as well as further exclusion criteria, were used for and described in a meta-analysis of a subset of the data 1 .

Code Availability
Code for technical validation (see below) can be found on the Zenodo archived repository [Data Citation 1].

Technical Validation
We have implemented a number of tests that check the database prior to new releases and major updates by using the testthat 23 package in R. The tests check the database for structural integrity (i.e., its internal organization), variable consistency (i.e., correct naming rules) as well as data integrity checks (i.e., outliers, correct data types). Additionally, sub-samples of the data have been thoroughly checked by multiple collectors prior to the release of v1.0.0. The online database can also easily be expanded and corrected if errors are identified. Data will be updated annually with new studies and any errors identified corrected.

Usage Notes
Regularly-updated versions of the Reptile Development Database (RepDevo) can be found at, and downloaded from: www.repdevo.com. The Database.csv, Citations.csv and Metadata.csv files are located on external servers hosted by GitHub and are fully version controlled. Releases will be lodged on GitHub and version releases will be provided with a unique, stable DOI and permanently stored for improved reproducibility as new updates and errors are documented. This is achieved using Zenodo's DOI and versioning capabilities (https://doi.org/10.5281/zenodo.1188482). Users can cite within the manuscript the specific version used along with its DOI if necessary, which will be provided with the version  downloaded from the webpage (www.repdevo.com). However, minimally the specific version should be specified when the data are used to ensure reproducibility of any resulting analyses.
We have been fairly inclusive in our database and many studies report or manipulate a multitude of factors at once, including moisture conditions, oxygen concentrations or measure phenotypes at different ages, temperatures or post-incubation treatments. Our database contains all these data, and so, any future work should take care to extract relevant data based on the question of interest. We have been careful to indicate the population, species, moisture conditions and any post hatching manipulations (e.g., temperature) or measurements (e.g., age) that data are derived from. Some studies confound incubation treatment with year and we have tried to be careful in identifying the specific year in which the experimental manipulation took place to ensure that comparisons of temperatures across different years is not undertaken. In addition, experimental designs can vary substantially among studies (e.g., random allocation of eggs to treatment, split-clutch designs, etc.). This variation could influence results and care should be taken when calculating effect sizes and comparing results of studies with different designs. To account for this, we have also categorized experimental designs that can be considered by users.