INTRODUCTION
The conservation of mammalian diversity is an urgent issue globally (Bowyer et al., 2019; Crooks et al., 2017), but population declines have been particularly precipitous in Australian marsupials (Baker & Dickman, 2018; Fisher et al., 2014; Woinarski et al., 2015; Ziembicki et al., 2015). A high proportion of marsupials reached high levels of vulnerability in the last century, making them a particular conservation concern (Woinarski et al., 2011). Consequently, high priority conservation efforts are underway for over one hundred threatened Australian mammals (Legge et al., 2018).
One of the challenges of current conservation efforts is the determination of within-species diversity. Determining population units for management plans ensures the preservation of evolutionary potential in endangered species (Crandall et al., 2000; Moritz, 1994) and the functioning of ecosystems (Des Roches et al., 2018). Population units are largely determined through molecular data (Allendorf, 2017) – e.g., endangered species of squirrels (Finnegan et al., 2008), jaguars (Wultsch et al., 2016) and wolves (Hindrikson et al., 2017) have all relied on genetics to identify their population diversity for conservation purposes. This genetic management has established links between diversity metrics and population fitness; however, it does not assess the phenotypic variation within a species and is therefore blind to organismic diversity within a population (Wanninger, 2015). This results in the potential for serious disjuncts between phenotypic intraspecific variation and genotype variability (Boyko et al., 2010; Le Rouzic & Carlborg, 2008; Vogt et al., 2008).
Understanding the phenotypic diversity among fragmented populations can provide valuable information to conservation management. In particular – in analogy to the interpretation of genetic distances – morphological differences between populations may indicate local adaptation (Colangelo et al., 2012; Meloro, 2011; Meloro et al., 2017). Current conservation studies of endangered taxa rarely use morphological data to determine phenotypic differentiation below the species level (e.g., Dierickx et al., 2015; Wilting et al., 2015); and most quantitatively rigorous assessment of phenotypic differentiation remains the domain of taxonomic studies (Celik et al., 2019; Meloro et al., 2017; Nicolosi & Loy, 2019; Senczuk et al., 2018; Sveegaard et al., 2015). Therefore, quantifying morphological variation within a species represents a largely untapped potential for understanding the phenotypic variation between taxonomic units; test hypotheses of adaptation and relatedness within a species; and provide a valuable tool for management, for example in assessing if population units may be too morphologically divergent to be cross bred in outbreeding conservation efforts. It can also inform predictions of morphological change during future species’ fragmentation events, which is a common consequence of human activity (Bennett & Saunders, 2010; Haddad et al., 2015; Hansen et al., 2013).
The anatomical complex with the most comprehensive amount of quantifiable morphological information is the mammalian skull. This is reflected in a long tradition of using linear skull measurements for taxonomic purposes (Baker et al., 2015; Cardini, 2013; Travouillon, 2016; Van Dyck, 2002). The shape of the mammalian skulls contains information on animal function (Hanken & Hall, 1993), such as masticatory loading (Herring et al., 2001), acting as a proxy for dietary behaviours in mammals (Maestri et al., 2016; Marroig & Cheverud, 2005; Nogueira et al., 2009), including marsupials (Mitchell, Sherratt, Ledogar, et al., 2018; Wroe & Milne, 2007). This is particularly relevant in the context of marsupial mammals, whose skull might not be as adaptable as that of placental mammals due to a development constraint on skull shape variation (Goswami et al., 2012; Porto et al., 2013; Sánchez-Villagra et al., 2008; Weisbecker et al., 2008, 2019). This is because marsupials are born at an extremely immature (altricial) state, but with a highly developed oral apparatus adapted to immediate and extensive feeding at the mother’s teat. This seems to reduce the potential of the oral region to diversify, both developmentally (Goswami et al., 2016) and evolutionarily (Porto et al., 2013; Sánchez-Villagra et al., 2008; Weisbecker et al., 2008). Such a developmental constraint may reduce the ability of marsupials to adapt their skull morphology at the level of within-species variation, leaving adaptation through changes in size as the only avenue of heritable adaptive shape (Marroig & Cheverud, 2005, 2010; Porto et al., 2013; Shirai & Marroig, 2010).
In this study, we use geometric morphometric analyses to provide the first population-level study of morphological population variation in the skull of a mammal of particular conservation concern. We focus on the endangered Northern quoll (Dasyurus hallucatus : Gould, 1842), a guinea pig-sized carnivorous marsupial with well-understood genetic differentiation between populations (Cardoso et al., 2009; Firestone et al., 2000; Hill & Ward, 2010; Hohnen et al., 2016; How et al., 2009; Woolley et al., 2015) but no information on morphological adaptation. Northern quolls appear to have had a pre-colonial distribution over 5000 km across northern Australia (Braithwaite & Griffiths, 1994). They are now separated by major biogeographic breaks into four mainland populations with no apparent gene flow (Hill and Ward, 2010) and several island populations (Woinarski et al., 1999). Northern quolls are also a suitable study system for this investigation because they inhabit a wide range of habitats, ranging from rainforests to deserts (Begg, 1981; Moore et al., 2019; Oakwood, 2002). They are opportunistic foragers of small vertebrates, invertebrates, fruit and carrion (Dunlop et al., 2017). The species is also expected to evolve quickly because, as a semelparous species, most males die off in their first year after mating (Oakwood et al., 2001).
We capture fine-scale morphological differences of the cranium using 3D geometric morphometrics, which differs from traditional taxonomic morphometrics (Baker & Van Dyck, 2015; Travouillon et al., 2019) by being agnostic to expected shape differences and by allowing the shape variation of the whole skull to be described in high detail. This process also has the ability to provide statistical significances of shape variation patterns alongside visualisations of exactly what the shape variation in question is, thus permitting a finely resolved dissection of the drivers of shape variation that is not possible with conventional linear measurements.
We examine several potential drivers of northern quoll skull shape variation. We expect shape differences between populations to be a main part of overall skull variation. If variation relates to heritable adaptation to local environments, for example through dietary differences between high and low rainfall areas (Dunlop et al., 2017), we expect differences between populations to increase with either increased geographical distance, according to how individuals are related, or depending on local environmental differences (here accounted as rainfall and temperature). Alternatively, if local adaptation is prevented by a constraint due to the quoll’s early birth, most shape variation is expected along a size gradient or should be unexplained. Lastly, it is also possible that most variation relates to the biomechanical use of the cranium in feeding and – particularly in males – biting. This would result in a mostly uniform distribution of shape variation across geographic range of northern quolls.