Basal tolerance but not plasticity gives invasive springtails the advantage in an assemblage setting

Abstract As global climates change, alien species are anticipated to have a growing advantage relative to their indigenous counterparts, mediated through consistent trait differences between the groups. These insights have largely been developed based on interspecific comparisons using multiple species examined from different locations. Whether such consistent physiological trait differences are present within assemblages is not well understood, especially for animals. Yet, it is at the assemblage level that interactions play out. Here, we examine whether physiological trait differences observed at the interspecific level are also applicable to assemblages. We focus on the Collembola, an important component of the soil fauna characterized by invasions globally, and five traits related to fitness: critical thermal maximum, minimum and range, desiccation resistance and egg development rate. We test the predictions that the alien component of a local assemblage has greater basal physiological tolerances or higher rates, and more pronounced phenotypic plasticity than the indigenous component. Basal critical thermal maximum, thermal tolerance range, desiccation resistance, optimum temperature for egg development, the rate of development at that optimum and the upper temperature limiting egg hatching success are all significantly higher, on average, for the alien than the indigenous components of the assemblage. Outcomes for critical thermal minimum are variable. No significant differences in phenotypic plasticity exist between the alien and indigenous components of the assemblage. These results are consistent with previous interspecific studies investigating basal thermal tolerance limits and development rates and their phenotypic plasticity, in arthropods, but are inconsistent with results from previous work on desiccation resistance. Thus, for the Collembola, the anticipated advantage of alien over indigenous species under warming and drying is likely to be manifest in local assemblages, globally.


Introduction
Although assemblages lie within a metacommunity setting, their dynamics are significantly mediated by the physical environments they encounter and by interactions among species. Because physiological traits modulate the effects of the environment on populations (Helmuth et al., 2005), knowing the range of trait variation for local assemblages, or significant components of them, can provide much insight into assemblage structure and dynamics (Albert et al., 2012;Leibold and Chase, 2018). The development of such understanding is especially important given the growing need to understand the mechanistic basis of a globally common pattern of high local turnover through time without large changes in the richness of assemblages (Blowes et al., 2019). Whether such dynamics into the future will include the rising dominance of assemblages by alien species, owing to the expectation that changing climates will generally benefit them (Hulme, 2017), is of much interest given the economic and conservation significance of biological invasions. Such invasions are among the most significant conservation concerns globally (McGeoch and Jetz, 2019).
Assessments of trait variation for significant proportions of the species in local assemblages are uncommon for animals (Chown and Gaston, 2016). Rather, the species compared are typically selected from different localities. Where assemblage investigations have been undertaken, the outcomes can be quite different to those involving interspecific comparisons (compare, for example, Diamond et al., 2012with Kaspari et al., 2015. Hence, the general macroecological insight that interspecific and assemblage-level investigations are different and may provide complementary or even contrary insights (Chown and Gaston, 2016).
For plants, explorations of the extent to which indigenous and alien species differ in their characteristics at the assemblage level are increasing (e.g. van Kleunen et al., 2018;Mathakutha et al., 2019;Sandel and Low, 2019). These comparisons explicitly test the 'ideal weed' and 'plasticity' hypotheses, proposing that invasion success of a non-native species depends on its specific traits or enhanced phenotypic plasticity, respectively (Enders et al., 2020). By contrast, studies exploring whether the indigenous versus alien components of assemblages vary consistently in one or more physiological traits remain rare for animals. Most of the work on trait differences between indigenous and alien species is based on interspecific studies from animals collected across a wide range of localities (e.g. Moyle et al., 2013;Bradie and Leung, 2015;Jarošík et al., 2015;Allen et al., 2017;Janion-Scheepers et al., 2018) or for only a small component of a local assemblage (Stachowicz et al., 2002;Chown et al., 2007). These studies do not consider a range of species from a local setting. Yet, they are frequently used as a basis to forecast rising success of alien species under changing climates (e.g. Janion-Scheepers et al., 2018). Indeed, because of the availability of the data, interspecific comparisons remain among the most common macrophysiological approaches adopted. Thus, insights into whether assemblages might be dominated by alien species as climates continue to change, and what mechanisms might lie at the heart thereof, may at best be incomplete and at worst inaccurate. In consequence, much need exists to determine whether predictions made from interspecific studies are borne out at the assemblage level.
Here, we therefore examine the extent to which empirical outcomes from interspecific studies of the trait differences among indigenous and invasive animal species, i.e. tests of the ideal weed and plasticity hypotheses (Enders et al., 2020), are borne out by a comprehensive investigation of a local assemblage. We use Collembola as a model group. Springtails are important in belowground systems and mediate aboveground ecological outcomes (Bardgett and van der Putten, 2014). Understanding of physiological trait diversity in the group is growing rapidly (Van Dooremalen et al., 2013;Ellers et al., 2018;Jensen et al., 2019). How this diversity might be partitioned among indigenous and invasive species has been the subject of recent attention at the interspecific level (Janion-Scheepers et al., 2018). Recent work has been spurred by concerns over the extent of soil invasions globally, including among Collembola, and by suggestions that anthropogenic change will exacerbate the impacts of invaders on soil systems (Cicconardi et al., 2017;Coyle et al., 2017;Geisen et al., 2019).
We consider five physiological traits that have significant influences on fitness and are therefore frequently incorporated into models of the likely impacts of environmental change on organisms. These are critical thermal minimum and maximum (and the derived trait of tolerance range), desiccation resistance and egg development rate (Birkemoe and Leinaas, 2000;Kearney et al., 2013;Sinclair et al., 2016;Rozen-Rechels et al., 2019). The assemblage is that of sub-Antarctic Macquarie Island. We use this particular springtail assemblage because it is well surveyed both in terms of the species present and their abundances, is characterized by a range of alien species and is representative with regard to Collembola invasions of several islands globally (Cicconardi et al., 2017;Baird et al., 2019). Moreover, because it is an island assemblage, local factors are likely to be more important in determining dynamics than regional biotic influences (Leibold and Chase, 2018). The general climate of Macquarie Island and its change over the past 40 years are also relatively well understood (Adams, 2009;Bergstrom et al., 2015).
Specifically, we test two predictions based on general expectations (Daehler, 2003;Davidson et al., 2011;Hulme, 2017;Enders et al., 2020) and previous work on springtails (Chown et al., 2007;Janion et al., 2010;Janion-Scheepers et al., 2018). Compared with their indigenous counterparts, alien species should have (Prediction 1) greater basal physiological tolerances (for their definition see Chown and Nicolson, 2004) as suggested by the ideal weed hypothesis and (Prediction 2) more pronounced phenotypic plasticity as suggested from the phenotypic plasticity hypothesis.  Table S1). The island is small (12 800 ha) with a cool (mean air temperature range 3.8 • C to 6.6 • C), wet (∼954 mm annual precipitation) and windy oceanic climate; vegetation varies from coastal tussock to higher elevation fellfield areas (Selkirk et al., 1990).
Springtail collection involved beating into a tray and aspirating individuals from a variety of vegetation types and from the soil surface into 70 ml plastic pots with a saturated Plasterof-Paris and charcoal powder base (9:1 mixture) (hereafter lined plastic pots). Vegetation from the collection site was placed in to the pots as a food source. Initial sorting into additional lined plastic pots in the laboratory at Macquarie Island was undertaken to initially separate species and ensures densities of ∼75-100 animals per pot. Food sources were also placed into these new pots. Turf samples (10 cm 2 surface area, 5 cm deep) were also collected to ensure that springtails from all layers of soil were included. Springtails in pots and turf samples were maintained at ∼5 • C and on a 12:12 light:dark (L:D) cycle during the 2-week transportation back to the laboratory in Melbourne. Here, springtails were extracted from turf samples into lined plastic pots over 3 weeks using Berlese-Tullgren funnels.
Springtails were identified using current available keys for Macquarie Island Collembola (e.g. Greenslade, 2006) [further verified with DNA barcoding (see Supplementary  Table S2)] and sorted into species. DNA barcoding, involving the extraction and sequencing of 658 bp of the mitochondrial cytochrome oxidase subunit I gene (COI) was undertaken by the Canadian Centre for DNA Barcoding (CCDB, http://www. ccdb.ca/) at the Centre for Biodiversity Genomics, University of Guelph, Canada through the Barcode of Life Data Systems (BOLD, http://www.boldsystems.org/; Ratnasingham and Hebert, 2007) (see also Janion-Scheepers et al., 2018). A total of 91 individuals from 16 species was sequenced, with a minimum of three individuals for any one species (Supplementary Table S2). These sequences are available at the BOLD (www.boldsystems.org) under the larger project 'sub-Antarctic Collembola'. Species were classified as either indigenous to the island or introduced by human activity (alien) based on previously published information (Greenslade, 2006;Phillips et al., 2017). Most of the alien species are widespread on the island and hence considered invasive (Terauds et al., 2011;Phillips et al., 2017).

Colony maintenance
Springtail colonies were maintained in a controlled temperature room at 10 • C ([10.15 ± 0.23 • C], verified with iButton Hygrochron ® data loggers, Maxim Integrated, San Jose, USA) on a 12:12 L:D cycle. Individuals were maintained by species at intermediate density (75-100 individuals) in 70 ml lined plastic pots to maintain a humid environment (>99% relative humidity). They were fed weekly on a diet of Platanus sp. bark (Hoskins et al., 2015) to allow for self-selection of nutrients, with the bark also providing some shelter for individuals. Pots were randomly re-arranged in the controlled temperature room during feeding and during experiments to minimize shelf effects.
For experiments measuring thermal and desiccation resistance, springtails were assessed at the F0 and the F2 generations. F0 springtails were used to ensure that as much information on the assemblage could be captured as possible, including springtail species that we failed to rear successfully under laboratory conditions. The F2 generations of springtails were examined to ensure that carry-over effects from the environment of collection, including parental effects, were minimized while also minimizing adaptation to laboratory conditions (Hoffmann and Sgrò, 2017). For investigations of egg development rate, only the F2 generation was used. Eggs were removed weekly from parental pots (F0 individuals) and transferred to new pots to establish the F1 generation. The same process was then used to generate the F2 generation from F1 parents (following Janion-Scheepers et al., 2018). In each case, eggs from multiple adults were randomly combined within generations to maintain genetic diversity. F2 springtails reached adulthood between 5 and 16 months after field caught (F0) springtails entered the laboratory.

Critical thermal limits
Critical thermal limits provide a proxy for survival in active adult organisms (Lutterschmidt and Hutchinson, 1997), including in springtails (e.g. Everatt et al., 2013). The critical thermal maxima (CT max ) and critical thermal minima (CT min ) were determined for 16 species of springtails at F0 (9 alien, 7 indigenous, Supplementary Table S3), after they had been held at 10 • C for 1 week to examine differences in basal thermal tolerance between the indigenous and alien groups (Prediction 1). At the F2 generation, 10 species were investigated (7 alien, 3 indigenous, Supplementary Table S3). These F2 species were also examined for adult (shortterm, non-developmental) plasticity in critical thermal limits (Prediction 2). Adult phenotypic plasticity was assessed by acclimating F2 springtails to one of five temperature treatments for 7 days prior to experimentation (Supplementary Table S4). Three stable and two variable temperature acclimations were used. Much interest exists in understanding the extent to which fluctuating versus constant temperatures may alter estimates of phenotypic plasticity (Colinet et al., 2015). Recently, the importance of the influence of extreme temperature events on the evolution of thermal tolerance has been further emphasized, with the idea that extreme events disproportionately drive changes in such traits (Hoffmann, 2010;Kingsolver and Buckley, 2017). Stable temperatures were set at 5 • C, 10 • C and 15 • C, and variable temperatures were set at 10 • C with either a high (25 • C) or a low ( 3 extreme temperature spike that occurred for 1 h each day, with a 30-min temperature ramp up/down either side of the temperature extreme. The temperature spikes were based on extreme event temperatures from a long-term soil surface temperature record for the island (Leihy et al., 2018). Acclimation treatments were completed in controlled temperature cabinets (MIR-154-PE, Panasonic, Osaka, Japan) and rooms. Adults were held at the acclimation temperatures for 1 week (following Janion-Scheepers et al., 2018;Jensen et al., 2019).
Critical thermal limits were determined for individual adult springtails using established protocols (Janion-Scheepers et al., 2018). Springtails were contained within custombuilt thermal stages (Monash University Instrument Facility, Clayton Campus, VIC, Australia) that were heated or cooled with programmable water baths (Grant Instruments, Cambridge model TXF200) at 0.05 • C per minute. This rate was chosen for its environmental relevance, reflecting a commonly encountered rate of temperature change under microclimatic conditions (Allen et al., 2016). The floor of the stage was lined with saturated Plaster-of-Paris to minimize desiccation of springtails during experiments, and temperature of the stage floor was recorded using Omega thermometers (model: RDXL 12SD, Omega Engineering, Norwalk, USA) with type K thermocouples. A starting temperature of 10 • C (rearing temperature) was used for all experiments. Springtails were observed every ∼5 • C until a behavioural change occurred (e.g. moving faster/slower) after which they were monitored every 1 • C and then every 0.5 • C after the CT max /CT min was reached for the first individual in the experiment. CT max and CT min were defined as the temperature at which a loss of righting response occurred for each individual (Everatt et al., 2013;Janion-Scheepers et al., 2018). Different sets of individuals were used for the CT max and CT min experiments. Three replicates, typically of 10-15 individuals, were completed per species per treatment (sample sizes in Supplementary Table S3). Because variation in critical thermal limits may be affected by differences in body mass (Ribeiro et al., 2012), a mean body mass for each species was determined by weighing a random sample of 50 adult individuals of each species using a high-resolution (0.1 μg) microbalance (Mettler-Toledo XP2U, Switzerland) (Supplementary Table S5).

Desiccation resistance
Desiccation resistance was determined for 10 species of springtails at the F0 generation (5 alien, 5 indigenous, Supplementary Table S3) to examine differences in absolute desiccation resistance between alien and indigenous species (Prediction 1). In the F2 generation, 8 species were used (6 alien, 2 indigenous, Supplementary Table S3) to investigate plasticity in desiccation resistance in a cross-tolerance framework with temperature (Prediction 2). Short-term temperature acclimation has previously been shown to alter desiccation resistance in indigenous and alien springtails unequally, to the alien species' advantage (Chown et al., 2007). Here, the effects of short-term temperature acclimations on desiccation resistance were examined at two acclimation and two test temperatures. F2 springtails were acclimated at either 10 • C or 20 • C in controlled temperature rooms for 7 days prior to the desiccation experiment that was conducted at either 10 • C or 20 • C.
An experimental protocol for desiccation resistance, measured as survival time at 76% relative humidity, was established based on previous methods (Kaersgaard et al., 2004;Chown et al., 2007). Individual springtails were contained within glass vials covered with fine mesh, which were then housed in sealed, 70 ml plastic pots containing 15 ml of saturated NaCl solution as a desiccant. Saturated NaCl was used as it provides a consistent relative humidity of 76% from 0 • C to 20 • C. Furthermore, it has been shown that springtails can survive between 1 and 24 h at this humidity (Chown et al., 2007). Each pot contained two glass vials with ∼5 springtails per vial and an iButton Hygrochron ® data logger (Maxim Integrated, San Jose, USA) to verify temperature and relative humidity. Throughout the experiment, conducted in controlled-temperature rooms, springtails were examined every 10 min under a Leica M80 microscope (Leica Microsystems Pty Ltd, Wetzlar, Germany), and time to death (minutes) was recorded for each individual. Typically, four replicates of 10 individuals were used per experiment, with some exceptions for F0 experiments due to low numbers of springtails available (see Supplementary Table S3 for sample sizes). Following the experiment, springtails were dried at 40 • C for 24 h and then weighed in groups by replicate using a high-resolution (0.1 μg) microbalance (Mettler-Toledo XP2U, Switzerland) to obtain an estimate of individual dry body mass.

Egg development and hatching success
Egg development time and hatching success were determined for eight species, including six alien and two indigenous species (Supplementary Table S3), at seven temperatures ranging from 0 • C-30 • C, in 5 • C increments (Supplementary  Table S4) (Predictions 1 and 2) following previous protocols (Birkemoe and Leinaas, 2000;Janion et al., 2010). Eggs laid by F2 adults at 10 • C were collected and transferred to each respective development temperature within 24 h of laying. Eggs were transferred to 70 ml lined pots and kept in controlled temperature cabinets (MIR-154-PE, Panasonic) or rooms for the duration of development. Five replicate pots per temperature with 10 eggs per pot were used to provide a sample size of 50 eggs per temperature for each species. Eggs were checked daily for hatching. Days to hatching for each egg, and hatching success, measured as a percentage of eggs hatched within each pot, were recorded. Eggs were classified as unviable/dead if they were either visibly dead (shrivelled, dissolved or extremely discoloured) or if they had not hatched within 10 days (at 10 • C-30 • C) or within 14 days of the previously hatched egg within the same pot (at 0 • C and 5 • C).

Statistical analyses
All analyses were undertaken using R v. 3.5.2 (R Core Team, 2018) implemented in R Studio v. 1.1.463. All code and data files are archived in the Monash Figshare repository (doi: 10.26180/5e17b874b125c and doi: 10.26180/5e17c3bc55197).

Critical thermal limits
To test Prediction 1, basal critical thermal limit differences (CT max , CT min , CT range [difference between species mean CT max and mean CT min ]) among alien and indigenous species at the F0 generation (10 • C acclimation) were assessed using two approaches. The first explored differences in CT max and CT min among individuals from the two groups excluding species identity, assuming that individual trait variation is important in a community dynamics context (Albert et al., 2012), using a linear model with status as a fixed factor. Because individuals were not weighed, mass was not included as a covariate. Differences between the alien and indigenous groups were then assessed using species means (or differences in means for CT range ) within a phylogenetically explicit framework using phylogenetic generalized least squares (PGLSs) (Garland and Ives, 2000), as implemented in the caper v. 0.5.2 package (Orme et al., 2013), including species mean mass as a covariate. Given small numbers of species, a Brownian motion model of evolution was used (Cooper et al., 2016) and a maximum likelihood approach estimated Pagel's λ (Pagel, 1999) to indicate the degree of phylogenetic influence in the data. The phylogenetic tree used for the analyses was based on Janion-Scheepers et al. (2018) and with mtCOI data used to infer species relationships. For the final tree, branch lengths were assigned using Grafen's method (Grafen, 1989), and the tree (Supplementary Figure S1) is available as a Newick file in the Monash Figshare repository (doi: 10.26180/5e17b874 b125c). Density plots made using the package ggplot2 were used to illustrate the range of variation in CT max and CT min for each species and across individuals in the full assemblage investigated.
Because differences in traits among F0 and F2 adults might arise for various reasons (Hoffmann and Sgrò, 2017), means of the critical thermal limit traits in the F0 and F2 generations, each acclimated for 1 week at 10 • C, were compared among the 10 species common to both sets of trials. A PGLS approach using a reduced tree was initially used. Because Pagel's λ was estimated as zero for CT max , CT min and CT range and because of the likely measurement variation of the traits, a ranged major axis model (RMA, Legendre and Legendre, 2012), implemented in the lmodel2 package, was used for each trait to determine whether the slope differed from 1 and the intercept from zero in each case by examining the 95% confidence intervals of the estimated values.
To test Prediction 2, the effects of acclimation to 5 • C, 10 • C and 15 • C were examined for the F2 CT max and CT min trials by calculating an acclimation response ratio (ARR) ( • C/ • C) (Gunderson and Stillman, 2015). The ARR was calculated from the slope of the intraspecific relationship between acclimation temperature and critical thermal limits trait for each of the 10 species investigated based on individual data for each acclimation temperature. Systematic differences between the indigenous and alien species were investigated using a PGLS approach as above.
The impacts of a high (25 • C) or low (−5 • C) temperature spike for 1 h on a daily basis as an extreme event acclimation treatment were compared for each of the 10 species using linear models with temperature as a fixed factor and Tukey's honest significant difference (Crawley, 2013).

Desiccation resistance
Desiccation resistance was measured as individual survival time resulting in data that are bounded to the left at zero and right skewed (Supplementary Figure S2). To assess whether time to death differed between the F0 and F2 generations, five species for which F0 and F2 data were available at both acclimation and test temperatures of 10 • C were each compared using a generalized linear model (GLM) assuming a quasipoisson distribution and a log link function because of the form of the data (O'Hara and Kotze, 2010). Because substantial differences between the F0 and F2 generations were found in one of the species, data from the F0 and F2 generations were not pooled for comparisons among indigenous and alien species, even though the two data series did not overlap completely in the available species.
To test Prediction 1, comparisons of the indigenous and alien species were made in two ways using the F0 data, in keeping with the previous approach. In the first, a GLM (assuming a quasipoisson distribution and a log link function) was used to compare the alien and indigenous assemblages (fixed factor), including an estimate of log 10 dry body mass for each individual (from the individuals weighed at the end of the study) as a covariate. Thereafter, differences between the alien and indigenous groups were assessed using species means (here log 10 of time to death to account for the skew in the data) using PGLS as described above, including species log 10 mean dry mass as a covariate.
To test Prediction 2, the effects of thermal acclimation (fixed factor) on desiccation resistance were analysed for each species separately using a GLM assuming a quasipoisson distribution and a log link function. Acclimation at higher temperatures was expected to afford an extended survival time to the alien but not the indigenous species (Chown et al., 2007). Patterns of acclimation were compared visually for each of the groups and then status (alien or indigenous) included in a model (as above) with all species.  CT min ( • C) −2.8 ± 1.0 −3.9 ± 1.6 86.7 1528 <0.0001

Indigenous Alien
Mean ± SD (range) Mean ± SD (range) (1/days to hatching). Because single individuals were not examined across a range of temperatures (different eggs were assessed at each temperature), a function-valued trait approach (Gomulkiewicz et al., 2018) was not implemented. Mean values for development rate were obtained for each species at each temperature and plotted against temperature.
Maximum development rate (U max ) and the temperature at which this rate was realized (T opt ) were extracted from the means data. Prediction 1 was tested by inspecting the curves and selecting the appropriate mean values and temperatures, following previous approaches which have sought not to fit curves to the empirical data (Jarošík et al., 2015;Sørensen et al., 2018). Further to test Prediction 1, hatching success (as a proportion) was plotted against rearing temperature. This revealed that hatching success did not decline to zero at the lowest temperatures investigated in all of the species. Therefore, low temperature variation in hatching success was not investigated. Rather only the high temperature at which hatching success declined to 50% of the sample population, known as the upper lethal temperature 50 (ULT50) was estimated using a GLM assuming a binomial distribution and using a logit link function, with ULT50 values calculated from the fitted models using the mass package (Crawley, 2013).
To test Prediction 2, for each species the slope of the relationship or the temperature sensitivity of development was calculated in two ways. First, a linear model was used to estimate the slope of the relationship between mean rate (1/days to hatching) at a given temperature and that temperature ( • C) for each species. Data above the optimum temperature of the relationship [i.e. the temperature at maximum rate (see Sørensen et al., 2018)] were not used. Second, following a range of previous approaches (e.g. Dell et al., 2011), the natural logarithm of rate was plotted against 1/kT,  where k = Boltzmann's constant (8.617 * 10 −5 ev.K −1 ) and T is temperature in Kelvin.
A PGLS approach, as described previously, was implemented to investigate differences between alien (six species) and indigenous groups (two species) in each of these four traits (slope, U max , T opt , ULT50). Here, Pagel's λ was always estimated as zero. Because one of the alien species, Hypogastrura purpurescens, was found to be quite different to the others with regards to these variables, linear models used to assess differences between the alien and indigenous groups excluded this species.

Critical thermal limits
The alien assemblage had, on average, a higher CT max and lower CT min than the indigenous assemblage (Table 1), although the alien assemblage was bimodal for CT max (Fig. 1), largely owing to low values for Proisotoma sp. (Supplementary Figure S3). The PGLS models, based on species means (Table 2), revealed significant and substantial differences among the alien and indigenous species in CT max (4.1 • C) and CT range (5.1 • C), but not in CT min , with substantial phylogenetic signal in CT min only (Table 3).
Species mean mass was not a significant covariate for any of the traits and was omitted in the final models.
ARRs (in • C/ • C) did not differ between the alien and indigenous groups for CT min , and only marginally so for CT max (Tables 1, 3). Thus, the ARR for these two traits is similar for the two groups of species, though the ARR for CT min (0.065 ± 0.038 • C/ • C) is significantly larger than the ARR for CT max (0.020 ± 0.031 • C/ • C) (linear model F (1,18) = 8.29, P = 0.01) (summary data in Supplementary Table S6).
The extreme event treatments of either a low (−5 • C) or high (25 • C) temperature spike for 1 h each day had limited and variable effects across the species, especially compared with either the constant 5 • C acclimation in the former case and the constant 15 • C in the latter (Table 4; Supplementary  Figures S4 and S5).
For the F0 generation and using individual data, large and significant differences were found between the alien and indigenous species in time to death [alien mean: 326 ± 413 min (median: 140); indigenous mean: 115 ± 112 (median 60)], including with dry mass as a covariate (Table 5; Fig. 2). The PGLS models using species means also showed significant and substantial differences among the alien and indigenous species in time to death but with no phylogenetic signal in the data (Tables 3, 6).
Acclimation treatments in the F2 generation revealed that, as expected, time to death was shorter at the higher test temperatures, but that pre-exposure to an acclimation of 20 • C frequently resulted in improved desiccation resistance either at 20 • C (three species) or at 10 • C (two species), although in two species no effects of acclimation were found ( Fig. 3; Table 7). Similar responses were found among the alien and in the indigenous species [e.g. in Fig. 3

Egg development and hatching success
Despite considerable differences in the form of the development rate-temperature curves among species (Fig. 4) in the PGLS analyses only T opt differed between the alien and indigenous groups, with indigenous species having the lower value (Table 3). Excluding H. pupurescens, an outlier among the alien species (Table 8) In all of the species, hatching success had declined to zero by 30 • C, the highest temperature investigated. On average, the temperature at which hatching success had declined to 50% (ULT50) ( Table 8), was significantly lower (by ∼5.7 • C) for the indigenous than for the alien species (Table 3), with H. purpurescens an outlier among the alien species.

Discussion
In this springtail assemblage from Macquarie Island, the outcomes of the tests of the two predictions are clear. Prediction 1 [from the ideal weed hypothesis (Enders et al., 2020)], of greater basal physiological tolerance in the alien than in the indigenous species, is supported. On average, CT max is higher, CT range is broader, desiccation resistance is greater and egg development T opt and U max and the ULT50 for egg hatching success are higher in the alien than in the indigenous species. Only CT min is indistinguishable between these two T. bisetosa 0.6 ± 0.2 • C, t = 3.1, P = 0.020 0.1 ± 0.2 • C, t = 0.5, P = 0.985 F (4,148) = 4.77, P = 0.001 groups at the species level. At the individual level, however, the difference in CT min between alien and indigenous species is clear. By contrast, Prediction 2, of greater phenotypic plasticity in the alien than in the indigenous species [from the phenotypic plasticity hypothesis (Enders et al., 2020)] is not supported. Acclimation responses for CT max , CT min and desiccation resistance and the slopes of the rate-temperature relationships for egg development do not differ between the alien and indigenous species groups.
Differences in basal tolerance, but not in phenotypic plasticity, of critical thermal limits specifically, are largely in keeping with previous work. The most extensive interspecific study to date (Janion-Scheepers et al., 2018) found that indigenous springtail species are characterized by critical thermal maxima that are on average 3 • C lower than those  the differences in CT max and CT range are slightly larger (4 • C and 5 • C, respectively), but otherwise the findings accord closely. Moreover, irrespective of the acclimation conditions, we were unable to effect much change in the value of CT max in any of the species we investigated, reflected also by the low ARR values for this trait. Longer-term laboratory selection experiments have also been unable to do so in both alien and indigenous springtail species (Janion-Scheepers et al., 2018).
These findings of limited plasticity and adaptability in CT max over the shorter term are in keeping with previous investigations of other organisms (Hoffmann et al., 2013;Gunderson and Stillman, 2015;MacLean et al., 2019). Such similarity does not help to explain, however, why such substantial interspecific variation exists in the basal CT max of springtails. Here, for example, the largest difference in CT max among species is 8.6 • C, whereas the largest difference in CT min is 5.1 • C. In the broad-scale interspecific study (Janion-Scheepers et al., 2018), CT max varied among species by 11.6 • and CT min by 13 • C, only slightly more. Variation in basal CT max that either exceeds or is similar to variation in basal CT min in springtails (see also Jensen et al., 2019 for among-population variation) is different to findings for many insects and for other terrestrial ectotherms generally, but not unlike the situation found for marine ectotherms (Sunday et al., 2011;Araújo et al., 2013). Clearly, some of this difference must reside in the reasons for the evolution of much higher CT max in springtail species that succeed when introduced outside their native range. One reason may be that such species tend also to experience regular disturbances, which might be associated with broader tolerance ranges (Coyle et al., 2017). Another may be that variation in CT max at the assemblage level is much greater than interspecific analyses tend to reveal (e.g. Kaspari et al., 2015;Kühsel and Blüthgen, 2015), which would have substantial implications for assessments of response to global climate change. Yet a third might be that the introduced species all come from regions, such as continental Europe, where thermal variation is much greater and much more predictably so than for the sub-Antarctic , resulting in greater physiological tolerance ranges than in the indigenous species. This latter hypothesis requires further investigation with information that enables the exact localities of origin of the introduced species to be identified-information that is slowly becoming available (e.g. Baird et al., 2020).
In the case of desiccation resistance, the variation found among species in the time to death at 76% humidity is largely consistent with findings from other species, often examined under less extreme desiccating conditions (e.g. Kaersgaard et al., 2004;Elnitsky et al., 2008;Sørensen and Holmstrup, 2011). That we found a positive effect of thermal acclimation at 20 • C accords with the only previous investigation of such cross-tolerance effects for springtails (Chown et al., 2007). However, here, a similar effect for the indigenous M. caeca and no effects for the indigenous P. insularis and the alien P. fimata, differ from the outcomes of that work. There, acclimation at 5 • C tended to improve performance of the indigenous species at that temperature, whereas acclimation to 15 • C generally reduced it, with little difference among acclimation treatments at a 15 • C test temperature. By contrast, acclimation to 15 • C improved desiccation resistance at both the 5 • C and 15 • C test temperatures. Here, no such consistent differences among the indigenous and alien species were found. Basal desiccation resistance (measured as survival time) was on average, however, higher in the alien species, contrary to the previous work that found no such differences (Chown et al., 2007). Thus, differences in desiccation resistance among indigenous and alien springtail species cannot yet be generalized.
Intriguingly, despite the importance of water balance in determining the activity and distribution of ectotherms, and especially arthropods Rozen-Rechels et al., 2019), and forecasts for substantial changes in water availability globally (Sarojini et al., 2016), few studies have focussed on determining the extent of differences among indigenous and alien species in traits related to water balance. In mosquitoes, desiccation tolerant eggs are associated with species that have become alien but not those that are invasive (Juliano and Lounibos, 2005). By contrast, in freshwater molluscs, the two groups of species do not differ in desiccation resistance (Collas et al., 2014). Variation in development rate parameters has been the subject of two major studies contrasting indigenous and alien species. In the first (Jarošík et al., 2015), the sum of effective temperatures [1/slope of the rate-temperature relationship, SET)] and the lower development threshold (LDT) for insects across either part of the life cycle, or the full cycle, were compared among indigenous and invasive species. No significant differences were found for SET, but LDTs were lower for the invasive species (Jarošík et al., 2015). In the second study, of seven springtail species (Janion et al., 2010), the slopes of the rate-temperature relationships for egg development rate did not differ, although on average development rates were higher for the alien species, with lower hatching success at the higher temperatures for the indigenous species. Our results are largely consistent with these outcomes, so extending the findings for assemblage-level assessments. We found no significant differences between groups in the slope of the egg development-rate temperature relationship but a higher T opt , higher U max (when excluding H. purpurescens) and higher ULT50 in the indigenous compared with the invasive species. Although we did not calculate LDT (simply: -intercept/slope of the linear part of the rate-temperature relationship-see Jarošík et al., 2015),   Table S8) suggests that no differences between the groups are to be expected. Thus, the egg development work here supports suggestions that the temperature sensitivity of development, a form of phenotypic plasticity (Ghalambor et al., 2007), does not differ between the two groups of species and in magnitude is in keeping with the variation previously found for arthropods (Irlich et al., 2009;Dell et al., 2011). Yet, it also shows that alien species typically have increased capacity to complete their development at higher temperatures and to do so at faster rates than their indigenous counterparts.
The insights presented here on assemblage level variation in traits among indigenous and alien species are necessary to clarify (i) how interactions among species will play out (Leibold and Chase, 2018) and (ii) how future changes to systems because of either local disturbances (such as urbanization, see e.g. Diamond et al., 2017), or global climate alterations (Sarojini et al., 2016), will differ from place to place and therefore assemblage to assemblage, so affecting indigenous-alien species interactions (Hulme, 2017). In this regard, two caveats apply to our study. First, although we were able to investigate the most abundant species in the assemblage (see Terauds et al., 2011), we were not able to investigate the entire assemblage. At most, we included species representing 85% of the indigenous and >95% of the alien assemblage by abundance (Terauds et al., 2011). Nonetheless, the Macquarie Island assemblage now consists of 22 indigenous and 12 alien springtail species. That we found little difference between the F0 and F2 generations in critical thermal limits, and for most species also little difference among generations in desiccation resistance, suggests that examining recently captured individuals may not be as significant a concern as has been suggested (Hoffmann and Sgrò, 2017), especially when attempting to estimate the full suite of assemblage traits. The importance of investigating additional, and especially rare, species will depend on how important they are in the structure and functioning of the system in question, with evidence from other systems suggesting that rare species can be important and should be considered (Winfree et al., 2018;Dee et al., 2019).
Second, we did not differentiate between species that live above ground, in the litter, or deeper in the soil or among the major orders of springtails: the Symphypleona, Poduromorpha and Entomobryomorpha (Bellinger et al., 2019), as is often done (Janion et al., 2010;Bokhorst et al., 2012;Ellers et al., 2018). In part, we did not have sufficient species to undertake a full factorial design to enable us to do so, though the PGLS analyses mitigated these effects to some extent. We also think, however, that consideration at the assemblage level as a whole of how the distribution of individuals with different trait values might play out into the future is important at the local scale. Changes in response traits (Naeem and Wright, 2003) will take place through either differential survival or differential reproduction of individuals, altering the overall composition of the assemblage and its effects on ecosystem structure and functioning. Such whole-of-assemblage considerations of individuals from a trait perspective are becoming more common (e.g. Salo et al., 2020). Theory (e.g. Albert et al., 2012;Leibold and Chase, 2018) also suggests that they require further consideration when the outcome of the interactions between indigenous and alien species in a particular assemblage, under expected conditions of change, is being investigated.
Overall, our investigations have revealed that while basal trait values differ on average between the indigenous and alien species groups of Collembola, with the latter having the advantage under higher temperatures and drier conditions, phenotypic plasticity does not differ between them. These

15
outcomes suggest that as local climates become warmer and, in some places, drier with global change, the conservation problems associated with biological invasions (McGeoch and Jetz, 2019) will increase, especially in soils (Coyle et al., 2017).

Supplementary material
Supplementary material is available at Conservation Physiology online.