Phenotypic variations among Laudakia stellio populations in Anatolia: Colour, colour pattern, and phylogeny (Reptilia: Agamidae)

We examined the phenotypical intra- and inter-populational variation of Laudakia stellio, one of the most-studied agamids, which is widely distributed in Anatolia in order to reveal the qualitative and quantitative aspects of the morphological differences among populations. We used for the quantification of colouration of different body parts colour palettes with the RGB (Red-Green-Blue) colour scale and prepared colour histograms. The populations were compared with Principal Component Analyses (PCA). We found that the 1st dorsal pattern which is the first of 4-5 star-like pattern throughout the vertebral line, provides information about the sex of the individual and the geographic location of the population. PCA graphs revealed the colouration differences between localities of dorsal, lateral, ventral head and dorsal body regions. A phylogenetic analysis of the mitochondrial ND4 marker revealed four different lineages, namely Hatay, Adana-Mersin-Konya, Burdur-Denizli and Manisa-İzmir-Çanakkale. Noting the need for new studies with different genes, our results show that colouration and quantitative morphology are supported by phylogeny in L. stellio populations, especially in terms of subspeciation.


Introduction
Studies on geographical variation provides knowledge about an organism's phenotype and genotype as well as the adaptations that serves for the changing environment. In particular, the adaptations of ectotherms, which must adapt their body temperature to their environment, provide insight into adapting with dramatic environmental changes such as climate change or habitat degradation (Gunderson & Leal, 2012;Phillips et al., 2016;Senior et al., 2021). In this regard, Laudakia stellio, the Agamidae member with the widest distribution in Anatolia would give insight about the geographical variations based on the morphological knowledge of intra-and inter-populational variations (Kumlutaş et al., 2004, Gül et al., 2010. Laudakia stellio (Hardun or starred agama) has a wide distribution in North Africa, Southeast Europe and Southwest Asia (Terentjev & Chernov, 1965;Baig, 1992). In Turkey, it is distributed in Western, Southern, Central and South-eastern Anatolia Zones, but is absent from Thrace. Some populations are also known from the Black Sea Zone (Sindaco et al., 2000;Uğurtaş et al., 2000;Kumlutaş et al., 2004).
For the discrimination of Anatolian populations and subspecies of L. stellio, quantitative morphological characters were used including head-body length, head length, head depth, head width, member lengths, finger lengths, number of sub-digital lamellae, tail length, tail base width, as well as the ratio obtained using these and indexes (e.g., Göçmen et al., 2003;Almog et al., 2005;Kumlutaş et al., 2015;Bülbül et al., 2021). Additionally, assessments of colour and colour pattern variations have been applied in several studies on Anatolian populations of L. stellio (e.g., Sindaco et al., 2000;Göçmen et al., 2003;Almog et al., 2005;Gül., 2011;Kumlutaş et al., 2015). According to these morphological results, Sindaco et al. (2000) showed that the western populations of Turkey belong to Laudakia stellio daani, while the southern, central and south-eastern populations belong to L. stellio stellio. Göçmen et al. (2003) stated that Hatay populations, which they compared with Cyprus populations, belong to the nominate subspecies. According to Kumlutaş et al. (2004), Gül et al. (2010) and Gül and Tosunoğlu (2011) all Anatolian populations except for the Hatay population belong to the L. stellio daani subspecies. Almog et al. (2005) stated that the distribution of L. stellio daani extends over Anatolia, while the south-eastern populations are mixed in terms of subspecies. Kete and Yılmaz (2006) stated that the subspecies L. stellio daani is found in the Western Mediterranean region of Turkey and Urfa, the nominate subspecies in eastern Amanos, whereas a hybrid (transitional form) is found in the western Amanos. However, the results of some of these assessments remained insufficient as these traits were not assessed quantitatively.
With the development and widespread use of molecular methods, studies on taxa have begun to be carried out using mitochondrial and nuclear markers. Özdemir et al. (2011) found two mitochondrial lineages of L. stellio in Anatolia. Subsequent studies based on mitochondrial and nuclear markers (ND2, RAG1) by Melville et al. (2009) and Edwards and Melville (2011) updated Laudakia as a monophyletic group. Baig et al. (2012) summarized the results of the aforementioned studies in a morphology-based revision of Laudakia. This taxonomic approach was later criticized by Pyron et al. (2013), who confirmed that Laudakia is monophyletic using a super-matrix approach. In a recent update of Speybroeck et al.'s (2020) species list of European herpetofauna, the species was re-named L. stellio, and this change has been recognized by new studies (e.g., Karameta et al., 2022). Karameta et al. (2022) showed that L. stellio stellio, L. stellio daani and Western Taurus, Symi-Kaş, Anatolian and Lebanese-Hatay lineages are found in Anatolia.
The method of creating colour palettes in Red-Green-Blue (RGB) triple system was used e.g. in some optical studies (e.g., Tedore & Johnsen, 2017), however, there is no study on the use of reptilian colouration in the determination of inter-population variations. We aimed to use this system to give us a detailed comparison of the colours of certain body parts of different sexes and different populations, with precise RGB values. This system changes the colouration, which is mainly an observational characteristic into a quantitative data that is used to asses precise discrimination (e.g., Bergeron & Fuller, 2018).
In the former studies, the phenotypic variations of Anatolian L. stellio populations has not been assessed in detail in terms of quantitative colour-pattern arrangements and phylogeny. Based on this shortcoming, the present study aims to reveal the qualitative and quantitative aspects of the morphological differences in L. stellio populations distributed in different zones of Turkey.

Material and Methods
We examined a total of 81 individuals from 12 provinces and 22 localities from different parts of Anatolia between June and August 2018 and 2022 ( Figure 1). Tissue samples were taken from the tail end. All the individuals were released to their habitats. Detailed data of field studies are given in Table 1. The samples from Mersin, Antalya and Afyon were only used for phylogenetic analyses.
Morphometry. Males and females were distinguished by the preanal glandular and ventral glandular scales that only males have. Eleven different body measurements were taken from the captured individuals (Almog et al., 2005;: Head Length (HL), Head Width (HW), Forelimb Length 1 (FLL1, from shoulder to elbow, proximal half), Forelimb Length 2 (FLL2, from elbow to wrist, distal half), Forelimb Distance (FLD, between axilla), Body Width (BW, widest line of the body), Hindlimb Distance (HLD, between inguinal), Hindlimb Length 1 (HLL1, proximal half), Hindlimb Length 2 (HLL2, distal half), Snout-Vent Length (SVL), Tail Base Width (TBW). These measurements were taken with a tape measure for total SVL and with a calliper (Mitutuyo Absolute AOS Digimatic, CD-20AXWW, 200 mm, Kanagawa, Japan) for other lengths. For the statistical analysis, the measurements were transformed into the following indexes. Head Index (HI ((100xHL)/HW)), Head Length Index (HELI ((100XHL)/SVL), Forelimb Index (FLI FLL1/FLL2), Hindlimb Index (HLI (HLL1/HLL2)) and Limb Index (LI (FLL1+FLL2) /(HLL1+HLL2)). For pholidolial counts and comparisons, the supralabialia, sublabialia, forelimb 3 rd finger sub-digital lamellae (FL3SDL) and hindlimb 4 th toe sub-digital lamellae (HL4SDL) were counted. Colour and Pattern. On the brink of moulting, lizards experience dulling of skin colour and less pronounced patterns (Maderson, 1985). For this reason, field studies were carried out for each population in the same period of the same year, within one or two days, and trying to coincide with the resting phase of moulting in order to make a healthy colour pattern analysis. It was ensured that the individuals were under the same light (Exoterra, natural light / full spectrum daylight bulb, 25W, Hagen Group, USA) when photographs to be used for colour comparison were taken. Canon EOS 6D (WG) camera, Canon Zoom Lens (24-105 mm) and Canon Macro Lens EF (100 mm) were used for photography and video shooting. After separating the photographs of male and female individuals, colour determination was made for four different body regions: Dorsal head (Dh), Lateral head (Lh), Ventral head (Vh) and  Table 1) (Adapted from Atalay, 2002).
Dorsal body (Db) was considered as a whole and colour palettes consisting of nine different colours were created for male and female individuals of each population from this area. Advanced colour studies were performed with CorelDRAW X6 and Adobe Photoshop CS6 programs. As a result of the created palettes, the percent representations of the colours in the female/male individuals of the population were extracted. RGB (Red, Green, Blue) values of colour palettes and colour histograms were analyzed by ImageJ (Image Processing and Analysis in Java, Vers. 1.53s) program. Principal Component Analysis (PCA) was performed with the PAST (PAleontological Statistics, Vers. 4.10) program. For pattern analysis, the ventral head patterns were examined. Dorsal body generally has 4-5 star-like pattern throughout the vertebral line (Gül, 2011) ( Figure  4). The first of these patterns, which lies between shoulders, is called the "1 st dorsal pattern". DNA extraction and sequencing. Mitochondrial NADH-dehydrogenase-4 (ND4) marker was used in phylogenetic studies. DNA isolation was performed with the High Pure PCR Template Preparation (Roche, Basel, Switzerland) kit. Agarose gel electrophoresis was performed on a mini gel containing 1% agarose (Merck, Sigma-Aldrich, Germany) with 2.5 µl of GelRed. The integrity of the obtained genomic DNAs was checked by agarose gel electrophoresis. Purity controls and quantitation were performed spectrophotometrically with Thermo Scientific-Nanodrop 2000c device. For this purpose, A260/A280 ratio of 1.5µl DNA sample was determined. The components to be added to the amplification tube and the reaction steps are arranged according to the kit. The primer used for the ND4 marker is: CACCTATGACTACCAAAAGCTCATGTAGAAGC (Arévalo et al., 1994). PCR conditions are as follows: Denaturation 95°C, Annealing 57°C, Extension 72°C, Final extension 72°C.
The ND4 sequence of Laudakia caucasica was taken from Pang et al. (2003) (GenBank accession number: AY053998) Phylogenetic analyses. Alignments were done by ClustalX (Thompson et al., 1997) and a 660 bp final alignment was obtained. MegaX software was used (Kumar et al., 2018) to select a model

Results
Morphometry. The body measurements HL, HW, FLL (FLL1+FLL2), HLL (HLL1+HLL2) and TBW showed significant differences between male and female individuals (p≤0.05). Among the indices examined, FLI, HLI and LI did not show statistically significant differences (p≥0.05), whereas HI and HLI did (p≤0.05). Accordingly, HI is greater in females and HLI is greater in males. With regard to pholidosis, the only scale series that showed a significant difference between the sexes was sublabialia (p≤0.05). The number of scales in the sublabialia is higher in females than in males. Supralabialia, FL3SDL and HL4SDL did not show any sex-related significance (p≥0.05).
Among the indices, HI, FLL1+FLL2, HELI, HLI, HLL1/HLL2 values were found to vary significantly between localities (p≤0.05) ( Table 2). HL, HW, and SVL gradually increase from east to west. However, this increase is not linear; it gradually increases and decreases, reaching the highest value in the Manisa population ( Figure 5). The numbers of FL3SDL and HL4SDL are much higher in the easternmost Hatay locality than in other localities. The supralabial, sublabial and FL3SDL values varied significantly between localities (p≤0.05). Colouration. The colour palettes created were examined on an areal basis. All palettes and their descriptions are given from the east to the west ( Figure 6). -Dh1: Although the blue colour is seen in 17% of the palette in Hatay in males, it is concentrated in the Mediterranean Transition Zone in both sexes. -Dh2: Blue colour is represented differently in males in all populations (e.g., Hatay East 17%, Çanakkale 67%). Blue colours in females are beginning to be seen in Central Anatolia. -Dh3: The earth colours and light greens seen in Hatay turn to dark brown and orange tones in Adana and Konya. Blue tones are dominant in Central Anatolia (for females 67%, for males 84%, both in Burdur). -Lh1: While there is a general transition from green tones to light yellows in   hand, it is seen that blue tones are intense in Burdur and Denizli, and earth colours in Manisa, İzmir and Çanakkale (Figures 6, 7, 8, Sup1). Principal Component Analysis (PCA). In the PCA graphs comparing the RGB values of the colour palettes of the three easternmost localities, i.e., Hatay East, Hatay West and Adana, the dorsal head (Dh) of females was distinctly different from each other in all three localities. In Adana and Hatay West, colours are clustered in a narrower and darker area, while Hatay West has a wider and lighter colour range. Lateral head (Lh) colouration showed wide distribution in the graph in all localities. Despite this, Adana's colour palette diverges from Hatay and is concentrated under the component 2 axis. In the ventral head (Vh) region, Adana and Hatay West populations are clearly clustered in the dark and light regions of the diagram, respectively. The separation of Hatay East is similar to the West. In terms of dorsal body (Db) colouration, Adana is clearly different from Hatay. When the male individuals in the same three localities were examined, it was observed that the lateral head colouration in the Adana population was concentrated in the dark region of the diagram and diverged from the other two localities. Ventral    In the PCA graphs of Konya and Burdur, the dorsal and lateral head colouration of females is clearly differentiated. Although there is no definite divergence in ventral head colouration, Konya is clustered in a narrower area and is different from Burdur. In the dorsal body, the separation becomes sharper again. With respect to male individuals, there is a clear distinction between the populations in terms of the dorsal and lateral head regions. In terms of ventral head colouration, it was observed that Burdur clustered in the lighter region of the graph and diverged from Konya (Sup2).
The PCA graphs of Burdur and Denizli show that the two localities are clearly differentiated in terms of lateral and ventral head in females, but they are scattered in terms of dorsal head and dorsal body. Male individuals also differ in terms of lateral and ventral head colouration, as in females (Sup2). All RGB values taken from all the colour palettes of the entire sample can be found in Sup3.
İzmir and Çanakkale PCA graphs show that female individuals are clearly differentiated in terms of lateral and ventral head regions but have a scattered pattern in terms of dorsal head and dorsal body. In males, the separation is found only in the lateral head   Colour pattern. Basically, two shapes were determined in the ventral head patterning of L. stellio. These are lines and spots ( Figure 10). The lines are indented, starting from the back of the sublabial and descending towards the gular area. Their number is 7±1 on average in all individuals which are examined and have this pattern. The spots were observed in two different sizes. The first is small and frequent (mean 41±5), the second is large and less frequent (mean 12±2). Spotted pattern was considered as mosaic pattern.
The most common pattern of the ventral head is a mosaic pattern on dark background with 36% of all the examined specimen. This is followed by a mosaic on a light background with 27% and a striped pattern on a light background also with 27%, and the least common is a striped pattern on a dark background with 10% . In females, the most common pattern is a mosaic pattern on light background with 41%, and the most common pattern in males is a mosaic pattern on dark background with 43% (Table 3).  These patterns did not differ significantly between populations (p≥0.05). Similarly, they were also found to be independent of sex (p≥0.05) within and between populations. Both small and frequent dark spots and fine longitudinal lines were observed in juveniles.
In all populations studied, the 1st dorsal pattern was primarily divided into two, according to whether it was distinct (regular) in shape or indistinct (irregular). Irregular shapes are defined as the scattered presence of dark and light scales that continue along the back without forming any distinct pattern. Regular shapes are divided into three as circular, rectangular, and stellate ( Figure 11). A circular shape is a pattern that is approximately equal in width and length and has no vertebral extensions on its sides. A rectangular shape is a pattern that is longer than its width and has short extensions or no extensions at all. Stellate is a pattern that is equal in length or shorter than the width, with long and prominent vertebral appendages. No significant differences were found between populatons (p≥0.05). Similarly, there was no statistical significance in the figures evaluated according to gender within the populations (p≥0.05). Inter-population comparison revealed that the circular pattern decreases from east to west, is rarely seen in the Mediterranean transitional population (66% of Denizli females only), and not at all in the Aegean population. Similarly, it was understood that the prominent rectangular pattern decreases from east to west, is rarely observed in the Mediterranean transitional population (only in 34% of Burdur females and 40% of males) and not at all in the Aegean population. In contrast, the distinct stellate pattern appears to increase from east to west. Accordingly, this pattern is present only in 25% of Adana females in the Mediterranean population and 20% of Konya males in the Central Anatolian population. In the Mediterranean Transition population, it is found at different rates in males and females of all localities. There is a 100% stellate pattern in all males and females in the Aegean population. Irregular body dorsal pattern is rarely seen among all samples. It is represented by 66% in Western Hatay males, 50% in Adana males and 20% in Konya males, and is absent in Mediterranean Transitional and Aegean populations. While the regular circular pattern was found in 84% of females, the regular rectangular pattern was seen in 71% of males. The regular stellate pattern was observed at similar rates in males (47%) and females (53%). Irregular pattern was observed in all (100%) males. The regular circular pattern was seen by 56% in the Mediterranean population, 30% in the Central Anatolian population, and 14% in the Mediterranean Transition population. Similarly, the regular rectangular pattern was seen by 68% in the Mediterranean population, 25% in the Mediterranean Transition population, and 7% in the Central Anatolian population. Irregular pattern was seen by 85% in the Mediterranean and 15% in Central  Anatolia. All three of these patterns are absent in the Aegean population. On the contrary, regular stellate was seen by 66% in Aegean population and 28% in Mediterranean Transition population. The incidence rate in Central Anatolia (2.2%) and the Mediterranean (2.7%) is very low and close to each other ( Figure 12). Phylogeny. Considering the Hatay lineage, Kırıkhan differs from other Hatay localities. Çanakkale, İzmir and Manisa together form a western lineage. Differentiating from Denizli and Burdur, Adana, Mersin and Konya constitute an inner-south lineage (Figure 13).

Discussion
Morphometry. Sexual dimorphism was found in six characters as a result of the statistical analysis of the metric and meristic characters of all populations. Accordingly, head length (HL), head width (HW), fore limb length (FLL1+FLL2), hind limb length (HLL1+HLL2) and tail base width (TBW) were longer in males than in females by 17%, 23%, 14%, 16.5% and 21%, respectively. The head index (BI) is 5% smaller in males than in females. No difference was found between the sexes in terms of pholidosis characteristics. In this respect, the results are consistent with Gül (2011).  Gül (2011) stated that all populations, i.e., West-Anatolian, Mediterranean, Central-Anatolian, Black Sea, Hatay and South-east Anatolian populations were similar to L. stellio daani in terms of pholidosis and morphological ratio-index. These results were consistent with Almog et al (2005). Yet, in this latter study, head length (HL) was measured from the tip of the nose to the border of the ear, and the measurements were HL≤HW (Almog et al., 2005). Therefore, this measurement should not be compared with other studies that take the HL as the longest part of the head. In the same study, the mean head width (HW) was not different between L. stellio daani and L. stellio stellio. Almog et al. (2005) found that head index (HI), fore limb length (FLL1+FLL2), snout vent length (SVL), hind limb length (HLL1+HLL2) was shorter, and the number of hind limb 4 th finger sub-digital lamella (HL4SDL) were fewer in L. stellio stellio subspecies. However, in that study, head index (HI), fore limb length (FLL1+FLL2) and hind limb length (HLL1+HLL2) were given only for male individuals. In contrast, Baig et al. (2012) found that the number of HL4SDL was higher in L. stellio stellio (18-(20.5)-25) than L. stellio daani (16-(19.7)-23). In our data, the HL4SDL count in the Hatay population is in the range of 19-25, which is higher than in other populations and is consistent with Daan (1967) and Baran and Öz (1985). Furthermore, this range is consistent with Kete and Yılmaz (2006) Karameta et al. (2022), the most comprehensive and recent study on the phylogeny of L. stellio and differ only in terms of the Burdur-Denizli-Afyon lineage. Karameta et al. (2022) divided Anatolian L. stellio into L. stellio stellio (including the localities Adana, Mersin, Konya), L. stellio daani (including the inner and western Aegean), Symi-Kaş, and Hatay lineages. The reason why the Inner Aegean appeared as a different lineage in our study may be because of the single marker analysis or a point mutation.
Different results have been obtained regarding subspeciation in Hatay in the literature; the general common point of these studies is that, whether with a genetic background or not, they vary morphologically (e.g., Baran & Öz, 1985;Kete & Yılmaz, 2006). We found with the ND4 marker a divergence between the samples of Kırıkhan town in the north-east of the Amanos (Nur) mountains and Arsuz town, which is the southernmost tip of Hatay and the Anatolian diagonal. However, contrary to expectations, the samples taken from the southeast of Amanos did not differ from the samples from the southwest. The reason for this may be that the distance between the east and west of the mountain at the south end decreases to 20-30 km and the elevation decreases to 100-150 m, thus preventing the isolation. Colouration. Daan (1967) stated that the dorsal head colour is blackish grey similar to the back in L. stellio daani and yellow to red in contrast to the brown back in L. stellio stellio. Similarly, Beutler and Froer (1980) distinguished these two subspecies primarily based on colouration. Baran and Öz (1985) and Tok (1999) stated that the dorsal body of L. stellio in Western Anatolia is blackish grey, the ventral part is usually stained, and the blue colouration of the dorsal body of the males is more than that of the female individuals. Baig et al. (2012) reported that in L. stellio daani, the dorsal head colouration is similar to the dorsal body, with usually 4-5 yellowish vertebral spots on the back, usually on a dark grey background, the ventral body is whitish, and the ventral head is dark mottled on a whitish background. The authors also stated that the coloura-tion of L. stellio stellio is quite variable, and there is a wide variation of colour combinations due to the physiological colour change, but in general, there are 4-5 yellowish vertebral spots on the dark grey background on the dorsal body. Pale yellow to brownish grey colours on the dorsal head, pale yellow tones on the ventral body, and typical yellow to red colours on the dorsal head in males during breeding or defending the territory were noted.
Recent studies in Anatolia have provided new information about new localities of L. stellio and their subspeciation based on morphology. For example, Kumlutaş et al. (2015) stated that in individuals from Bursa, Yozgat and Sivas, dorsal body background colour is blackish grey and there is blue colouration on the dorsal head. These features characterize L. stellio daani as stated in previous studies. Bülbül et al. (2021) found that the back colour of L. stellio samples from the Black Sea (Zonguldak) is generally blackish grey, there is blue colouration on the dorsal scales, the spots on the back are yellowish brown, the anterior members are light blue, and the posterior members are dark blue. The same study also emphasized that the ventral body was yellowish brown in all samples and all these findings pointed that the samples belonged to L. stellio daani subspecies.
Lateral head colouration has so far not been used to distinguish subspecies. We used nine different area of the lateral head for the study of colouration and found that the colouration of this region is mostly in congruence with the dorsal head. In addition, it was observed that bright orange, yellow and copper tones around the eyes and eardrums were dominant in Adana and Konya males, which are thought to represent L. stellio stellio, intense blue tones in the Mediterranean Transition Zone, and earth colours in the Aegean Zone, which were thought to represent L. stellio daani. Almog et al. (2005) stated that the dorsal head of L. stellio stellio is lighter in colour than the dorsal body and, unlike it, in yellow-red colours, whereas in L. stellio daani the dorsal head is similar to or darker than the dorsal body that has blackish grey background and added this character to the subspecies key. Subsequent studies used these characters for distinguishing the subspecies in different parts of Anatolia (e.g., Gül, 2011;Kumlutaş et al., 2015;Bülbül et al., 2021). Our results from Hatay, Adana and Konya, in terms of dorsal head colouration, are consistent with the L. stellio stellio definition of Almog et al. (2005).
Comparison of PCA graphs created according to colour palettes and histogram results revealed the colouration differences between localities of dorsal, lateral, ventral head and dorsal body regions. When the colouration of male and female individuals is compared, Adana, which is geographically very close to Hatay and under the same climatic-ecological (Mediterranean) environmental conditions, is clearly differentiated from Hatay with much darker colours. Colour pattern. The results of the dorsal body 1 st dorsal pattern analysis showed that this pattern provides information about the sex of the individual and the geographic location of the population. These results are in line with Gül (2011) who stated that transversal extensions are present in this pattern (stellate) in Western and South-eastern Anatolian populations, but they are not seen or are unclear in other populations. Based on these results, although there are exceptions, it can be said that the regular circular and regular rectangular patterns indicate L. stellio stellio subspecies, while the regular stellate pattern indicates L. stellio daani subspecies. Similarly, it can be argued that the irregular pattern is distinctive for males of the L. stellio stellio subspecies. Baran and Öz (1985) found that the ventral head pattern was striped in 60% and mosaic in 30% of samples from Hatay. Tok (1999) found that the ventral head pattern consists of a mosaic on black background in 25% and black longitudinal dashed lines on a yellowish background in 70% of the individuals in the West Anatolian population. Göçmen et al. (2003) showed that the ventral head pattern was mosaic in 63.2% of the Hatay population and with dark spots in the neck in 36.9%. Almog et al. (2005) reported that the ventral head pattern in L. stellio daani is mostly mottled black and rarely mosaic, while it is characterized by a plain or weak mosaic pattern in L. stellio stellio. The findings of the present study are consistent with these previous ones in the literature.

Supplementary Material
Supplementary Material is given as a Supplementary Annex, which is available via the "Supplementary" tab on the article's online page.