Living on the Edge: Changes in the Foraging Strategy of a Territorial Ant Species Occurring with a Rival Supercolony – a Case Study

Territorial strategy in animals is characterized by the monopoly of resources inside a protected area, the territory. The presence of territorial species considerably alters the behavior of co-occurring submissives, as it is known in several submissive ant species living on the territories of red wood ants in temperate regions. On the other hand, as a rule, territorial species cannot share the same territory and usually exclude each other. However, this ‘rule’ is inferred from the almost complete lack of data on the coexistence of rival territorials, and not from observations regarding the effective exclusion or behavioral inhibition of one territorial species by other. In the frame of this study, we investigated the foraging strategy of the territorial red wood ant Formica pratensis that occurred inside a large polydomous system of another territorial ant species, F. exsecta. Formica pratensis colonies outside the supercolony served as control. Within the F. exsecta supercolony, F. pratensis showed the characteristics of a subordinate species characterized by low discovery and exploitation success of artificial food sources. On the contrary, control colonies outside the polydomous system clearly behaved like typical territorials as they successfully monopolized the majority of the baits. In addition, submissive species were more successful around the F. pratensis nests within the supercolony, than outside of it. As suggested by our results, territorial species could co-occur with other territorials given certain plasticity in their behavior and small colony size. Nevertheless, exclusion also happens as proven by our field observations.

shifts in their requirements, thus minimizing niche overlap as has been observed also in ants (see Cerdá et al. 2013 for a review). Territoriality is the most effective strategy to ensure the control of resources in ants (Hölldobler and Wilson 1990;Adams 2016). Territorial ant species usually dominate their foraging grounds, wherein they exert competitive pressure on the cohabiting ant species Vepsäläinen 1988, 1989;Pisarski and Vepsäläinen 1989;Petráková and Schlaghamerský 2011;Czechowski et al. 2013;Adams 2016;Trigos-Peral et al. 2016;Ślipiński et al. 2018).
In temperate regions, the composition of an ant assemblage is mostly determined by intra-and interspecific competition for food and suitable nesting places (Savolainen and Vepsäläinen 1988;Braschler and Baur 2003). Clear division of food resources contributes to the decrease in competitive pressure (Levings and Traniello 1981;Sorvari and Hakkareinen 2004) and can be related to morphological and behavioral differences (Davidson 1977;Gordon 2010). The cohabitation of two morphologically similar species can be mediated by behavioral differences. For instance, if a species can dominate in the contest competition due to its aggressive behavior, the other will adjust its foraging strategy by being faster in retrieving the prey and/or more efficient in discovering food supplies (Reznikova 1981;Savolainen 1991;Gordon 2010). However, other factors can also play an important role in determining species co-occurrence (Vepsäläinen and Pisarski 1982;Savolainen and Vepsäläinen 1989;Czechowski et al. 2013;Adams 2016;Stuble et al. 2017).
Territoriality in ants sometimes involves the development of multi-nest polydomous systems, consisting of interrelated nests formed by nest fissions, so-called 'budding', with intense worker and information exchange. The extreme form of such systems, where several hundred or thousands of nests form the cooperative network is called supercolony (Debout et al. 2007;Erős et al. 2009;Csata et al. 2012;Ellis and Robinson 2014;Adams 2016;Schultner et al. 2016;Wiezik et al. 2017). Such polydomous species efficiently exploit the available food supplies (Debout et al. 2007;Erős et al. 2009;Csata et al. 2012), and because of high worker density, they have a strong negative effect on other ant species (e.g. Czechowski et al. 2013, Trigos-Peral et al. 2016. Under normal circumstances, a lower-ranked species can cohabit with a species of a higher level, but when resources diminish, the competition between them increases to an extent that could end coexistence (Savolainen and Vepsäläinen 1988;Markó and Czechowski 2004;Cerdá et al. 2013;Czechowski et al. 2013;Markó et al. 2013;Ślipiński et al. 2014). On the other hand, as a rule, in the temperate region, territorial species exclude each other. However, there is a lack of empirical evidence concerning territory sharing of top dominant ant species. Such conditions, when two different territorial species share to some extent space and persist for a given period together, are rarely or not even observed in nature (see e.g. Petráková and Schlaghamerský 2011;Czechowski et al. 2013). Discovery of two colonies of the territorial ant Formica pratensis inside a large F. exsecta polydomous system tightly surrounded by nests of the rival, created the possibility for a unique case study. The questions related to their ways of co-existence immediately came up: (a) do they share the same territory? (b) do the two territorial species exert different impacts on other co-occurring ant species? and finally, (c) can we observe changes in the foraging strategy of F. pratensis that would ensure its coexistence with the supercolonial species at least for a certain period of time?

Species and Site
Both Formica (Formica) pratensis Retzius, 1783 and Formica (Coptoformica) exsecta Nylander, 1846 are relatively common, Pan-Palearctic, mound-building, territorial ant species. Formica pratensis usually occurs in open areas, on grasslands from the plain areas to mountain pastures, up to 1500 m in the Alps (Seifert 2007). Generally, it forms monodomous and monogynous colonies, but polydomous systems are also known. Nests are usually large, they can reach a diameter of 300 cm (Seifert 2007), containing tens of thousands of workers . It keeps a large territory around its nests with a diameter of several tens of meters and maintains trunk trails leading to aphid colonies, which can persist even for 15 years (Seifert 2007). With the help of these trunk trails and by a large number of foragers, F. pratensis very efficiently covers its territory.
F. exsecta inhabits open areas in mixed and deciduous forests, mostly forest edges, forest clearings or mountain pastures (Seifert 2000). Large polydomous colonies are known from the Alps, Central Europe, and European Russia which consist of several hundreds of nests (see  for a review; Wiezik et al. 2017). Usually, nests of F. exsecta are smaller than those of F. pratensis. Nest mound diameter ranges from 10 cm in incipient to above 1 m in mature nests (Seifert 2000;Erős et al. 2009;Csata et al. 2012;Wiezik et al. 2017), and nests contain several hundred to several hundred thousand individuals (Sorvari 2009;Wiezik et al. 2017). Territories of monodomous colonies are also smaller, than those of F. pratensis, covering from a few square meters to 60 m 2 (Pisarski 1982;Sorvari 2009;Erős et al. 2009;Csata et al. 2012).
The investigations were carried out in Central Romania, in the southern part of the Giurgeului Depression in the Eastern Carpathians in July 2009. The depression is one of the coldest regions in Romania; frosts are still common in May and can occur as early as September. There are nine large polydomous systems of F. exsecta in the southern part of the depression, in addition to several monodomous colonies . The largest polydomous system selected for the purpose of our study (46°36'N, 25°36'E, 780 m a.s.l.), is a genuine supercolony comprised of more than 3,400 nests stretching over more than 20 ha ). This polydomous system is located in a fen meadow with Molinia caerulea, Deschampsia caespitosa, Festuca pratensis, Briza media, Nardus stricta, Succisa pratensis, Stachys officinalis, and Cirsium palustre as the most abundant or characteristic plant species. The area is overgrown with scattered small trees and saplings of Betula pubescens, Picea abies, Frangula alnus, and Salix spp. The meadow is fairly intensely grazed by cows for most of the year. The structure of the F. exsecta polydomous systems as well as intraspecific and mutualistic relationships of F. exsecta to other species are well studied (Goropashnaya et al. 2007;Erős et al. 2009;Martin et al. 2009;Csata et al. 2012;Trigos-Peral et al. 2016).
Altogether four F. pratensis nests were used in the study, two of which were located inside the large F. exsecta polydomous system (further on I-1 and I-2; Fig. S1) and the other two nests were outside the polydomous system, but still close to its border (further on O-1 and O-2; Figs. S2-3). The latter two nests served as controls, as it could be assumed that here the effect of F. exsecta would be weak, if any. The distance between nests I-1 and I-2 was 10.05 m and they were closely surrounded by 24 F. exsecta nests (see Fig. S1). Since there were no observed visible trails connecting them, we assumed that they were separate colonies. The average distance between nest I-1 and the F. exsecta nests surrounding it was 12.4 m (SD ± 3.39, min 8.05 m), and 9.04 m for I-2 (SD ± 3.35, min 8.05 m). Both colonies tended eight separate aphid colonies at a mean distance of 2.95 m (SD ± 1.43, min 0.4 m, max 4.6 m) and 2.64 m (SD ± 1.24, min 0.4 m, max 3.8 m), respectively.
The distance between nests O-1 and O-2 was > 100 m. No trails were detected between them or other F. pratensis nests, thus they could be also treated as monodomous colonies. A single F. exsecta nest was in the vicinity of O-1 located at 10.75 m. In addition, there were three F. (Raptiformica) sanguinea Latr. nests around O-1 with a mean distance of 8.03 m (SD ± 1.9, min 10.75 m; see Fig. S2). This facultative slave-maker species is known to apply a strange dominant strategy. It is, generally, weaker than F. pratensis or F. exsecta (see Trigos-Peral et al. 2016). In O-2, there were two F. exsecta nests at 10.9 and 18.31 m, and three nests of the submissive F. (Serviformica) cunicularia species with a mean distance of 10.83 m (SD ± 3.34, min 10.9 m) (see Fig. S3). O-1 tended 48 aphid colonies with a recorded average distance of 12.19 m (SD ± 8, min 0.8 m, max 30 m), whereas O-2 exploited 23 aphid colonies with an average distance of 6.42 m (SD ± 4.1, min 0.6 m, max 14.1 m).

Study Methods
Transects of observation plots (20 × 20 cm quadrats) were set up to determine under undisturbed conditions the distribution and abundance of ant species. The 1st transect inside the supercolony connected the two focal F. pratensis nests (both being surrounded by more F. exsecta nests), whereas outside the supercolony each focal F. pratensis nest with the nearest F. exsecta nest. The transects also extended in the other direction from the nests along the same axis, whereas the 2nd transect was placed perpendicular to the axis of 1st transect (see Fig. S1-3). The setup allowed us to trace the transition between the territories of the two focal species. There was 1 m between the centers of two consecutive observation plots. At nests I-1 and I-2, the 1st transect consisted of 17 plots with 9 plots being established between the two colonies, whereas the 2nd transects contained 16 plots (altogether 28 and 29 obs. plots, respectively; see also Fig. S1). At nests O-1 and O-2, the 1st transects consisted of 17 plots and connected the focal nests with the nearest F. exsecta colonies that were located at 10.75 and 10.9 m, respectively, whereas the 2nd transects consisted of 16 plots (altogether 33 and 33 obs. plots, respectively; see Fig. S2-3).
Baits are widely accepted means of studying ant foraging behavior around their nests. However, the presence of large food sources enhance and thus alter the activity of ants in the specific area in which a bait is placed (e.g. Vepsäläinen and Pisarski 1982;Savolainen and Vepsäläinen 1988;Vepsäläinen and Savolainen 1990;Sanders and Gordon 2003;Markó and Czechowski 2004;Petráková and Schlaghamerský 2011;Markó and Czechowski 2012;Czechowski et al. 2013;Ślipiński et al. 2018). For this reason, we recorded the distribution of foragers in the absence (further on 'nudum' observations) and in the presence of baits as well, in order to obtain an appropriate view of the foraging strategy of a specific ant species (see also Markó and Czechowski 2004;Czechowski and Markó 2005;Markó and Czechowski 2012;Czechowski et al. 2013;Ślipiński et al. 2018). Interspecific conflicts were also recorded. Observations were carried out on sunny days with fair weather conditions for two periods in a day involving the activity peak of the ant species in this habitat: between 9 AM and 12 AM, and between 4 PM and 7 PM. During mid-day, due to elevated temperatures, the activity of the species dropped significantly with only a few observed active workers on the trunk trails. Within each period, each plot was checked for one min at 20 min intervals, which yielded a total of 18 observations per plot per day. First-day observations were carried out in the absence of baits (nudum), which were followed by bait observations next day. To avoid effects due to seasonal variation in food preferences, baits contained both animal protein (tuna flakes) and carbohydrate (mixed-flower honey) Czechowski 2004, 2012;Czechowski and Markó 2005;Petráková and Schlaghamerský 2011;Czechowski et al. 2013;Ślipiński et al. 2018). Bait portions of ca. 3.5 cm in diameter were placed on 10 cm diameter green plastic plates in the center of each plot. The plates and portions were set-up 10 min before the first observation at the beginning of each observation period, retrieved at the end of each period and then rinsed with water.

Statistical Analysis
The distribution of most frequent species around the focal F. pratensis nest in nudum observations (both inside and outside the polydomous system) were analyzed using Generalized Linear Mixed Model approach (GLMM, Poisson error, maximum likelihood) with colonial identity as the input factor, while observation time and the identity of the observation plots were included as random factors.
The interaction between the four most abundant species (two focal species and two subordinates -Myrmica spp. and F. cunicularia) and the effect of distance from the focal F. pratensis colony (further on distance) were analyzed also using GLMM approach (Poisson error, maximum likelihood) separately for nudum and bait observations. Distance from the focal nests and abundance of rival foragers were included as covariates, while observation time, colonial identity and the identity of the observation plots as random factors. Data for each observation period was used separately (N inside supercolony = 1026; N outside supercolony = 1188). To test the predictability of the baits' exploitation patterninferred from the abundance of ants at baits -, we also included the abundance of the respective species recorded during nudum observations as a covariate in the models testing their occurrence at baits.
The normality of datasets was regularly tested with Kolmogorov-Smirnov test but none proved to be of normal distribution. In model over-dispersion, negative binomial error term was applied. In every model, automated model selection was carried out, and the effects of different explanatory factors and variables were averaged across the best models with delta < 2 (see Grueber et al. 2011). Since F. exsecta individuals were not recorded at baits outside the polydomous system, the number of F. exsecta individuals were not included in these GLMM analyses. The tables containing the GLMM models included in the average best-models are stored as digital supplementary material to this article (Tables S1-4).
All statistical analyses were carried out in an R Statistical Environment (R Core Team 2018). GLMMs were performed using the glmer function in the lme4 package (Bates et al. 2013), automated model selection with the help of the dredge function in the MuMIn package (Bartoń 2013). The lsmeans function from lsmeans package was used in order to carry out posthoc sequential comparisons among factor levels when performing GLMM analyses (Russell 2016) and used the Tukey method (p.adjust function) to correct the p values for multiple comparisons.

Species Occurrence in Natural Conditions
Both territorial species were present in relatively high abundance inside the supercolony in the absence of baits, but F. exsecta's presence was considerably more reduced at the control colonies (Fig. 1). Generally, more ant species occurred outside the polydomous system, and the submissive Myrmica spp. and F. cunicularia occurred in considerable numbers both inside and outside the supercolony (Fig. 1).
The abundance of F. pratensis was higher at the control nests (GLMM z ≥ 5.29, p < 0.001), with no significant differences between them (z = 2.09, NS), while considerable differences were recorded between the inside nests (z = 5.29, p < 0.001). The abundance of F. exsecta was higher inside the supercolony (z ≥ 8.82, p < 0.001), and it also differed between both inside and outside nests (z ≥ 3.2, p < 0.001). The abundance of F. cunicularia was also higher inside (z ≥ 5.3, p < 0.001), and significant differences were recorded among both inside and outside nests (z ≥ 4.23, p < 0.001). In Myrmica spp. no consistent pattern could be revealed, there were considerable fluctuations in their abundance at the different nests. Thus significant differences were recorded in their abundance between the control nests, but also between the inside nests (z ≥ 3.33, p < 0.004), while with the exception of O-1 and I-2 nests (z = 0.034, NS), all other pairwise comparisons yielded significant differences in the favor of control nests (z ≥ 3.35, p < 0.004).
The average best GLMM model regarding the distribution of territorial species retained each of the original variables, with the exception of the model on F. exsecta's distribution at the control nests, which retained only the distance from these nests (Table S1). Inside the supercolony, the abundance of F. pratensis decreased with distance from the focal nests (z = -4.56, p < 0.001), whereas the opposite was valid for F. exsecta (z = 5.62, p < 0.001). There was no significant relationship between the abundance of the two species (z < -0.76, NS), however. In a similar manner, at the control nests, the activity of F pratensis decreased with distance from their nests (z = -4.84, N = 1188, p < 0.001), but this time F. exsecta was not influenced by the distance (z = 0.48, NS). No significant relationship was revealed between the two territorials in this case either (z = 0.42, NS).
All variables were retained in the average best model concerning the distribution of submissive species in nudum (Table S2). Inside the supercolony, the abundance of F. cunicularia was not affected by any of the included variables (z ≤ 1.16, NS). On the other hand, the abundance of Myrmica spp. was negatively affected by the abundance of F. exsecta (z = -2.16, p < 0.05). The other variables did not have any significant effect on the abundance of Myrmica spp. (z < 0.28, NS) either. At the control nests, the abundance of F. cunicularia foragers was affected by the distance from the F. pratensis nests (z = 1.94, p = 0.05) and the abundance of F. exsecta workers (z = 2.69, p < 0.01). In Myrmica spp., the abundance of F. pratensis had a significantly negative effect on their abundance (z = -2.37, p = 0.01). The other variables did not have an effect on any of the subordinate species (-1.88 < z < 0.82, NS).

Under Experimental Conditions: Discovery and Exploitation of Baits
Outside the supercolony, F. exsecta was almost entirely absent at baits (with only two individuals present), while the abundance of F. pratensis was quite high compared to other ant species (Fig. 2). This species was the most successful in bait discovery and exploitation around these nests (Figs. 2 and 3) and its abundance decreased with distance (full model) from the colonies (z < -5.21, p < 0.001; Table S3).
In the absence of F. pratensis, submissive species were able to exploit baits, but also encounter species from the genera Lasius and Tetramorium at some baits ( Figs. 2 and 3). However, the exploitation pattern was not predictable in any of the species (z < 0.95, NS; Tables S3-4 ). The average best model for F. cunicularia contained all the variables of the full model except for the distance (Table S4), but neither of the included variables had a significant effect (z < 0.89, NS). In the model concerning Myrmica spp., only the abundance of F. pratensis was included (Table S4) which had a negative influence on Myrmica spp. (z = -3.92, p < 0.001).
As expected, within the supercolony F. exsecta was the most abundant at baits followed by F. pratensis (Fig.  2). The majority of baits were mostly discovered and dominated by F. exsecta. The exploitation pattern of baits was predictable based on the nudum observations only in F. exsecta (GLMM z = 2.05, N = 1026, p < 0.05), and not in the other species (z < 0.87, NS; Tables S3-4). Contrary to F. exsecta, F. pratensis never succeeded in monopolizing baits. In some cases, both territorials co-occurred with the submissive F. cunicularia. The full model was the best in both F. pratensis and F. exsecta (Table S3). The abundance of F. pratensis decreased significantly with distance from its nests (z = -5.82, p < 0.001), while that of F. exsecta increased (z = 5.48, p < 0.001). The two territorials had a significant negative influence on each other's abundance (z < -2.84, p < 0.01).

Interactions Among Species
During nudum observations, a single interspecific conflict was recorded in which one F. pratensis forager attacked and drove away a Myrmica spp. individual at nest O-2. However, at baits, several interspecific conflicts occurred (Table 1). Inside the supercolony, many aggressive interactions were recorded between the territorial and the submissive species (Table 1) but did not record any direct interactions between the two territorial rivals. Outside the supercolony, the two territorial species never occurred together, or with F. cunicularia. On the other hand, F. pratensis co-occurred with the submissive Myrmica spp. resulting in several aggressive interactions (Table 1). On the contrary, the facultative slave-maker F. sanguinea was not observed to co-occur at baits (neither at those situated close to its nests) with F. pratensis. Workers of the territorial species chased away both workers of those of the slave-maker and its F. fusca slave (Table 1).

Discussion
According to the competitive exclusion theory (Pisarski and Vepsäläinen 1989) the territories of territorial ant species cannot overlap. However, even within this theoretical framework, it is specified that the competitive strategy of a given species is context-dependent which is determined by the age and size of the colony (Savolainen and Vepsäläinen 1988;Zakharov 1991;Gallé 1994). Supercoloniality is the extreme territorial strategy where high nest density and intensive exchange of individuals among colonies (see Erős et al. 2009;Csata et al. 2012;Robinson 2014) could result in almost total exclusion of any species (see Czechowski et al. 2013). Since rival territorials are hardly ever found within a polydomous system, the general validity of this rule can be accepted. Under these circumstances, the peculiar natural setup discovered, where F. pratensis nests were found in the middle of a F. exsecta supercolony, offered the rare possibility to study the impact of the supercolonial species on the foraging strategy of F. pratensis. However, since there were only two F. pratensis colonies that could be studied, Fig. 1 Abundance of ant species in nudum around the nests inside (a) and outside (b) the supercolony (median, quartiles, min-max values, outliers) therefore the possibilities for generalization of the results should be handled cautiously.
As it was expected from an ant species living within a supercolony of another species, within the supercolony F. pratensis was present in quite low numbers around their colonies. A few meters farther from its colonies, the abundance of F. pratensis dropped sharply, while that of F. exsecta increased. The boundary between their territories was quite fuzzy without obvious 'no ant's land' as known from other observations on territorials neighboring each other (see Hölldobler 1979).
Other lower-ranked species found on the territory of the supercolony, like Myrmica spp. and F. cunicularia, are known to have adapted to living with territorial species. Thus, Serviformica species are fast in discovering new resources, and select smaller pieces of food in order to escape easier when harassed by the dominants (Reznikova 1981;Savolainen 1991). Myrmica species shift their activity period in a different time in order to avoid territorial Formica species foraging (Savolainen and Vepsäläinen 1989). Such adaptations make possible the long-term coexistence of submissives with the dominant species even inside a polydomous system (see Cerdá et al. 2013). Such changes could have been applied by F. pratensis as well. However, these strategies could also result in a lower amount of food intake that, could seriously limit or slow down the growth of colonies especially in larger-bodied species like F. pratensis.
While F. pratensis inside the supercolony was poorly performing, the control F. pratensis colonies outside the supercolony performed well and could be characterized by standard features of territorial species such as having high worker density around their colonies, a large number of aphid colonies tended that are crucial for the sustainability of strong colonies in territorial Formica ants, and competitive exclusion of lower-ranked species (Sorvari and Hakkarainen 2004). Moreover, tended aphid colonies were quite far from the nests thus contouring the borders of large territories in both control colonies (see Fig. S2-3). The presence of other ant species inside the F. pratensis territory was limited to those parts where F. pratensis was mainly absent, and their exploitation success was quite low which is consistent with the results of several other studies Vepsäläinen 1988, 1989;Czechowski 2004, 2012;Czechowski and Markó 2005;Ellis and Robinson 2014;Adams 2016). Inside the supercolony, however, F. pratensis behaved quite differently which is similar to submissive species.
What remains unclear is the extent of plasticity in territorial ant species. That is whether such 'unexpected' coexistence could persist for a longer period. We visited the hereby studied F. pratensis colonies in the following years in order to confirm or dismiss the long-term coexistence hypothesis. While the control nests were healthy and thriving, signaling that for F. pratensis as such the general habitat conditions were still fair, those located inside the supercolony went extinct. Therefore, the loss of the nests located inside the supercolony would be consistent with the competitive exclusion theory, although we cannot rule out other mechanisms due to the small sample size.
The results of our study suggest that the strategies applied even by a territorial species could be contextdependent. They may be shaped by many other factors like the presence of other territorials, stronger competitors, and/or the abundance of available resources (e.g. Savolainen and Vepsäläinen 1988;Cerdá et al. 2013;Adams 2016;Johansson and Gibb 2016;Ślipiński et al. 2018;Stuble et al. 2017), and also most probably by nest size, making it possible for them to survive within a polydomous system at least for a while. The final disappearance of the studied colonies is consistent with the competitive exclusion, but also adds important nuances: colony foundation and development to a certain size is possible even for territorials within rival supercolonies.
Acknowledgements This work was supported by a mobility project of the Romanian Ministry of Research and Innovation, CNCS-UEFISCDI, project number PN-III-P1-1.1-MCT-2018-0038, within PNCDI III. We are grateful for the help of Enikő Csata, Enikő Német and Csilla Gál during fieldwork. The suggestions of László Gallé and Wojciech Czechowski greatly improved the quality of the manuscript. Róbert Gallé and Attila Torma kindly offered their help during data analysis. During fieldwork, we enjoyed the hospitality of the Apáthy István Field Station run by the Apáthy István Association. B.M.'s work was supported by the Bolyai János Scholarship of the Hungarian Academy of Sciences during manuscript preparation.

Compliance with Ethical Standards
Competing Interest The authors declare that they have no competing interests. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, F. sanguinea as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.