Morphological convergence in a Mexican garter snake associated with the ingestion of a novel prey.

Morphological convergence is expected when organisms which differ in phenotype experience similar functional demands, which lead to similar associations between resource utilization and performance. To consume prey with hard exoskeletons, snakes require either specialized head morphology, or to deal with them when they are vulnerable, for example, during molting. Such attributes may in turn reduce the efficiency with which they prey on soft-bodied, slippery animals such as fish. Snakes which consume a range of prey may present intermediate morphology, such as that of Thamnophiine (Natricinae), which may be classified morphometrically across the soft-hard prey dietary boundary. In this study, we compared the dentition and head structure of populations of Thamnophis melanogaster that have entered the arthropod-crustacean (crayfish)-eating niche and those that have not, and tested for convergence between the former and two distantly related crayfish specialists of the genus Regina (R. septemvittata and R. grahamii). As a control, we included the congener T. eques. Multivariate analysis of jaw length, head length, head width, and number of maxillary teeth yielded three significant canonical variables that together explained 98.8% of the variance in the size-corrected morphological data. The first canonical variable significantly discriminated between the three species. The results show that head dimensions and number of teeth of the two Regina species are more similar to those of crayfish-eating T. melanogaster than to non-crayfish-eating snakes or of T. eques. It is unclear how particular head proportions or teeth number facilitates capture of crayfish, but our results and the rarity of soft crayfish ingestion by T. melanogaster may reflect the novelty of this niche expansion, and are consistent with the hypothesis that some populations of T. melanogaster have converged in their head morphology with the two soft crayfish-eating Regina species, although we cannot rule out the possibility of a morphological pre-adaptation to ingest crayfish.

INTRODUCTION

Some of the most compelling test cases for adaptive evolution involve morphological convergence (Schluter, 2000), which is predicted to evolve when organisms experience similar functional demands on their phenotype (Schluter, 2000; Vincent, Brandley, Herrel, & Alfaro, 2009). If resource use or other environmental factors impose demands on performance, morphological convergence is predicted to occur (Ruber & Adams, 2001; Winemiller, Kelso‐Winemiller, & Brenkert, 1995).

Snakes are very good subjects for studying feeding morphology because their head is directly involved in feeding (Cundall & Rossman, 1984; Dwyer & Kaiser, 1997). They consume their prey whole; therefore, some morphological attributes of their typical prey should be associated with trophic morphology (Hampton, 2011, 2013; Mori & Vincent, 2008; Vincent et al., 2009). To consume prey with external body features such as the hard exoskeletons of arthropods, snakes require specialized morphology (e.g., piercing teeth) or behavior such as targeting arthropods when they are vulnerable, such as when molting, as the hard exoskeleton is both slippery to grasp and hard to pierce.

The tribe Thamnophiine (family Natricinae) comprising North American semi‐aquatic snakes, includes Thamnophis melanogaster (Mexican black‐bellied garter snake), an aquatic dietary specialist that is sympatric with a freshwater crustacean, the crayfish Cambarellus montezumae, but eats crayfish only in 3.0% of the area of sympatry (Manjarrez, Macías Garcia, & Drummond, 2013; Figure 1). The 35% of prey consumed by T. melanogaster were crayfish eaten only when recently molted, so with the exoskeleton as yet unhardened (Manjarrez et al., 2013). Extensive dietary studies of Thamnophis species have failed to reveal crayfish ingestion, except in a rare record for T. proximus (0.8% of individuals with crayfish in stomachs; Hampton & Ford, 2007). Therefore, the rarity of crayfish ingestion in the focal cluster of populations of T. melanogaster (Alfaro & Arnold, 2001; de Queiroz, Lawson, & Lemos‐Espinal, 2002) suggests crayfish eating represents a niche invasion that has not yet expanded to more populations.

Figure 1

Tula and Lerma drainages where snake Thamnophis melanogaster (Natricinae Thamnophiine) consumes crayfish, Cambarellus montezumae. Black dashed lines are watershed boundaries; thin dotted lines are 500‐m contour lines, and gray continuous lines are rivers

Two Thamnophiine species of the genus Regina eat newly molted crayfish, which are soft‐bodied (Gibbons & Dorcas, 2004; Godley, 1980; Mushinsky, Hebrard, & Vodopich, 1982). Therefore, we hypothesized that the seemingly recent dietary convergence of some populations of T. melanogaster with (allopatric) Regina (Hibbitts & Fitzgerald, 2005; McVay & Carstens, 2013) may have led to morphological convergence associated with the demands of finding and capturing hidden soft crayfish.

We explored possible morphological differences in dentition and head structure within T. melanogaster by comparing individuals from crayfish‐eating versus non‐crayfish‐eating populations and included in this comparison both soft crayfish‐eating Regina species and the aquatic generalist Thamnophis eques (Mexican garter snake). Thamnophis eques is sympatric with T. melanogaster over most of its range (Rossman, Ford, & Siegel, 1996) and represents a control for geographic determinants of head morphology. We predicted a morphological convergence between crayfish‐eating T. melanogaster and Regina species that specialize in eating soft crayfish.

Consuming soft crayfish may not require specialized teeth, but because of their vulnerability during the molt, crayfish seek refuge and must be sought in burrows and crevices, which would impose different demands on the head morphology of a snake that often preys in the open and guides its strikes visually (Drummond, 1983; Macías Garcia & Drummond, 1995). Dwyer and Kaiser (1997) proposed that Thamnophiine species might be classified morphometrically across the soft–hard prey dietary boundary. They concluded that the soft crayfish‐eating species of Regina have skulls of similar dimensions to those of two Thamnophiine species of Nerodia, which feed mainly on soft prey (fish; Mushinsky & Hebrard, 1977; Mushinsky & Lotz, 1980), whereas the skulls of hard crayfish eating of Regina were different (larger/thicker). It has been proposed that the elongated skull morphology of garter snakes (Thamnophis) is associated with the ingestion of soft prey (Britt, Clark, & Bennett, 2009; Savitzky, 1983).

Whereas R. septemvittata (Queen snake) and R. grahamii (Graham's crayfish snake) eat newly molted crayfish, congeners R. alleni and R. rigida primarily eat hard, nonmolted crayfish (Franz, 1977). Both eaters of newly molted crayfish have shorter and narrower heads than their hard crayfish‐eating congeners (Dwyer & Kaiser, 1997; Nakamura & Smith, 1960), and their teeth are sharp, curved, and oriented backwards as in most generalist relatives Thamnophiine (Myer, 1987; Nakamura & Smith, 1960), contrasting with the more rounded back teeth (to hold hard prey) of R. alleni and R. rigida (Nakamura & Smith, 1960; Rossman, 1963).

Thamnophis melanogaster is a snake that specializes in underwater foraging and feeds mainly on soft‐bodied aquatic prey such as fish (ca. 50%), tadpoles and leeches. It has a narrow head (Rossman et al., 1996), similar to that of other species that feed on aquatic soft prey (Dwyer & Kaiser, 1997; Hibbitts & Fitzgerald, 2005), and curved, pointed, and backward‐directed maxillary teeth, suitable for piercing through soft skin (Rossman et al., 1996). This species is located within the monophyletic group of garter snakes, whereas Regina is polyphyletic with respect to other thamnophiines (Alfaro & Arnold, 2001; Guo et al., 2012; McVay & Carstens, 2013; de Queiroz et al., 2002). This suggests that crayfish ingestion has arisen independently among Regina species via evolutionary convergence associated with the ingestion of soft‐versus‐hard crayfish. Thamnophis eques feeds on soft prey, primarily leeches, fishes, and frogs (Table 1; Macías Garcia & Drummond, 1988; Drummond & Macías Garcia, 1989; Rossman et al., 1996).

Table 1

Mean snout–vent length (SVL ± 1 SD, range) of the species/morphs Regina grahamii, Regina septemvittata, Thamnophis eques, and two dietary morphs of Thamnophis melanogaster (Natricinae Thamnophiine) and their reported prey

Species	n	Snout–vent length ± SD, (range)	Prey	Reference of prey reported in the diet
Regina grahamii	19	30.8 ± 19.8 (18.0–77.0)	Newly molted soft crayfish	Burghardt (1968), Mushinsky and Hebrard (1977), Godley, McDiarmid, and Rojas (1984)
Regina septemvittata	81	29.6 ± 14.9 (11.5–65.0)	Newly molted soft crayfish	Burghardt (1968); Godley et al. (1984)
Thamnophis melanogaster
Noncrayfish eating	88	39.5 ± 9.7(19.3–59.5)	Leeches, worms, fish, tadpoles	Manjarrez et al. (2013)
Crayfish eating	80	38.4 ± 11.1 (15.0–56.5)	Leeches, worms, fish, tadpoles, crayfish	Manjarrez et al. (2013)
Thamnophis eques	42	55.8 ± 11.5 (31.5–79.0)	Leeches, frogs, fish, and salamanders	Macías Garcia and Drummond (1988), Drummond and Macías Garcia (1989), Manjarrez (1998)

Regina and Thamnophis are two genera of semi‐aquatic North American snakes (Natricinae Thamnophiine). Thamnophis snakes were collected at ponds and rivers in two watersheds in Central Mexico, while Regina were museum specimens (see Materials and Methods).

MATERIALS AND METHODS

We measured 80 crayfish‐eating T. melanogaster individuals from 10 populations (Manjarrez  al., 2013) and 88 non‐crayfish‐eating individuals from 29 populations adjacent to the crayfish‐eating populations (Table 1). All snakes were captured in the wild. In addition, we examined 19 specimens of R. grahamii and 81 of R. septemvittata at the Florida Museum of Natural History, University of Florida (Table 1). We also included 42 T. eques (Table 1). Snakes were mostly adults or of a size close to that of the adults (Table 1; Appendix 1).

Four variables were used to characterize head structure: (1) jaw length (distance from the posterior edge of the posterior‐most supralabial scale to the anterior tip of the rostrum; King, 2002), (2) head length (distance from the snout tip to the posterior‐most portion of the parietal bone), (3) head width (widest part measured while applying pressure on the posterior portion of the head to spread the quadrates and mandibles laterally; Miller & Mushinsky, 1990), and (4) number of maxillary teeth. Although often used in similar studies (King, 2002; Miller & Mushinsky, 1990), we did not use gape index in our analyses because this is a composite of several of the above measures. An exploratory analysis showed that gape index (computed as the area of an ellipse with major and minor axes equal to jaw length and head width; Miller & Mushinsky, 1990) is highly correlated with the three head measures in T. eques and in the two feeding morphs of T. melanogaster (Appendix 2). Accordingly, this index does not add information to the analysis beyond that provided by jaw length, head length, and head width (King, 2002). We also measured snout–vent length (SVL, Table 1) and recorded the snake gender.

Statistical analyses

We ascertained whether head measurements differed between sexes. As the head variables are influenced by snake size, sexes were compared using one ANCOVA for each species (n = 5) and head measurement (n = 4), entering SVL as a covariate (n = 20 ANCOVAs; see Appendix 3). In general, these comparisons did not indicate differences between sexes (ANCOVA F values range from 0.005 to 3.4, with p values from .06 to .94), except in only three of the 20 comparisons (Appendix 3). Consequently, in the multivariate tests described below, the two sexes were pooled.

To verify whether tooth number and head shape in crayfish‐eating populations of T. melanogaster are similar to those of R. septemvittata and R. grahamii, we conducted a discriminant function analysis with stepwise selection of variables. The initial explanatory variables were the residuals from linear regressions of head length, head width, and jaw length (all log‐transformed because of the lack of homoscedasticity and skewed distributions) and number of teeth, on SVL. The grouping variable was snake species/dietary morph (R. septemvittata, R. grahamii, T. eques, T. melanogaste crayfish‐eating, T. melanogaste noncrayfish eating). We compared the canonical variates among groups using one‐way ANOVAs and explored the distribution of the groups' means within the multivariate morphological space. Tests were performed using with Statistica software (ver. 8.0 StatSoft, Tulsa, Oklahoma, USA) and NCSS 10 Statistical Software (2015; NCSS, LLC. Kaysville, Utah, USA).

RESULTS

The residuals of number of teeth and jaw length, head length, and head width on SVL contributed significantly to the discriminant function (all p < .000001), which correctly classified 64.8% of the 310 snakes on the basis of four significant canonical variables which explained 100% of the variance in the data (Table 2). As the means in canonical variable (CV) 1 (59% of variance explained; Table 3) for the two Regina species are virtually identical, the overall canonical analysis did not distinguish between them; it classified all individuals as R. septemvittata, except for one individual of each species which, interestingly, were classified as crayfish‐eating T. melanogaster. Sixty‐six percent of T. eques were correctly classified, and the rest were assigned indistinctly to R. septemvittata and to the two feeding morphs of T. melanogaster (Table 2). Among non‐crayfish‐eating T. melanogaster 53.4% of individuals were correctly classified, 16% were mistakenly classified as Regina, and only 4.5% were mistaken for T. eques; the equivalent figures for their crayfish‐eating congeners were 59%, 8.8%, and 5%, respectively (Table 2). Only about one‐quarter of T. melanogaster individuals were incorrectly classified as belonging to the alternative dietary morph (26% and 27.5% for noncrayfish eating and crayfish eating respectively), compared to only 18% of Regina individuals being incorrectly assigned to the wrong species.

Table 2

Number of snakes classified as Regina grahamii, R. septemvittata, Thamnophis eques, crayfish‐eating and non‐crayfish‐eating T. melanogaster (Natricinae Thamnophiine) by a discriminant function analysis performed using the residuals from linear regressions of number of teeth and three log‐transformed head shape variables, on SVL

True species	n	Classified as
		R. grahamii	R. septemvittata	T. melanogaster		T. eques
				Crayfish eating	Noncrayfish eating
Regina grahamii	19	0	18	1	0	0
Regina septemvittata	81	0	80	1	0	0
Thamnophis melanogaster
Crayfish eating	81	0	7	47	23	4
Noncrayfish eating	88	0	14	23	47	4
T. eques	42	0	6	5	4	27

Wild‐caught Thamnophis and museum Regina specimens were used (see Materials and Methods).

Table 3

Canonical coefficients from a discriminant analysis to assort individual snakes belonging to Regina grahamii, R. septemvittata, Thamnophis eques, and T. melanogaster (Natricinae Thamnophiine) from two dietary morphs; crayfish eating and noncrayfish eating (see Table 2)

Morphological variable	CV 1	CV 2	CV 3
Head width	−0.178	−0.260	−0.575
Head length	−0.114	0.259	−0.021
Jaw length	0.296	−0.473	0.136
Number of teeth	−0.195	−0.049	0.159
Eigenvalue	1.005	0.555	0.120
Proportion of variance explained	59.1	32.6	7.1
Cumulative variance explained	59.1	91.7	98.8
One‐way ANOVA F (df)	29.8 (16, 923)	21.9 (9, 737)	10.5 (4, 608)
p	<.001	<.001	<.001

CVs are linear functions of the original morphological variables (jaw length, head length, head width, and number of maxillary teeth), each multiplied by a canonical coefficient. Measures are from wild‐caught Thamnophis and museum Regina specimens (see Materials and Methods).

Values of CV1 obtained from the discriminant function analysis of morphological variation among means of snake species (which explained 59% of the variance) increased with jaw and head length, and decreased with head width and number of maxillary teeth, according to the coefficients shown in Table 3, and they differed significantly between genera and between Thamnophis species, but not between Regina species, nor between T. melanogaster dietary morphs (Figure 2a; Table 4). Values of CV2, which explained about one‐third (32.6%) of the variance, increased with head length and decreased with jaw length and head width (Table 3). Thus, high values of CV2 depict a slender‐headed snake with a small mouth; hence, it separated (with very large, negative values) stout‐headed T. eques from the rest (Table 3). Values in the third canonical variable (CV3), which explained 7% of the variance, decreased with head width while increasing with number of teeth and jaw length (Table 3). On CV3, the two dietary morphs of T. melanogaster differed significantly (Table 4, Figure 2b), with crayfish eating also being significantly different from R. septemvittata and noncrayfish eating also differing significantly from R. grahamii (and from T. eques; Table 4). Because differences in CV3 (or CV2) are not significant between the two species of Regina, it is possible in the plot of the second and third canonical variables to define a morphological space that is shared by both species of Regina and by the crayfish‐eating populations of T. melanogaster (Figure 2b).

Figure 2

Principal canonical variates obtained from a discriminant function analysis of morphological variation among snake species Regina septemvittata, Regina grahamii, Thamnophis eques, and Thamnophis melanogaster (Natricinae: Thamnophiine) with two dietary morphs, crayfish eating and noncrayfish eating. Thamnophis snakes were captured in the wild at two Mexican drainages, while Regina were museum specimens (see Materials and Methods). (a) Principal canonical variables (CV)1. Equal letters represent statistical similarity when we compared the canonical variates among groups (one‐way ANOVA). (b) Plotting the Principal canonical variables 2 versus 3 reveal morphological proximity between the crayfish‐eating morph of Thamnophis melanogaster, and the two species in genus Regina which also prey on newly molted crayfish

Table 4

Canonical means used in the classification of individual Regina grahamii, R. septemvittata, Thamnophis eques, crayfish‐eating and non‐crayfish‐eating T. melanogaster (Natricinae: Thamnophiine) by a discriminant function analysis based on jaw length, head length, head width, and number of maxillary teeth (see Tables 2 and 3)

Snake species/morph	CV 1	CV 2	CV 3
Regina grahamii	1.320	0.634	0.458
Regina septemvittata	1.399	0.198	−0.060
Thamnophis melanogaster
Crayfish eating	−0.962	0.350	0.414
Noncrayfish eating	−0.679	0.243	−0.458
Thamnophis eques	−0.041	−1.847	0.0817

Thamnophis (wild‐caught) and Regina museum specimens were used (see Materials and Methods).

After correcting for body size, we found no difference in the number of maxillary teeth between the two Regina species, but we found difference between the two morphs of T. melanogaster. Crayfish‐eating T. melanogaster had 2.28 more teeth than non‐crayfish‐eating conspecifics (Student‐t 187 = 2.92, p = .001), which themselves had 6.0 more teeth than R. grahamii and 6.3 more than R. septemvittata (F 1,190 = 227.6, p = .0001). We found no difference in the number of maxillary teeth between T. eques and either morphs of T. melanogaster.

DISCUSSION

We found a large overlap in head morphology and number of teeth between the several species/morphs examined, yet we also found evidence consistent with the hypothesis that the head morphology of soft crayfish‐eating T. melanogaster should more closely resemble that of the two soft crayfish‐eating species of Regina than that of non‐crayfish‐eating conspecifics.

Crayfish ingestion in only some locations can be explained by subtle environmental differences between localities (Arnold, 1981), for example, spatiotemporal availability of crayfish or differences in use of microhabitats by T. melanogaster. However, a sampling suggests that, if anything, crayfish are more abundant in ponds where snakes do not eat them that in ponds where they do (Appendix 4).

Although significant, the magnitude of the apparent morphological convergence between crayfish‐eating T. melanogaster and the two Regina species is small. This may be because invasion of this dietary niche is recent, thus even if challenging, crayfish consumption has not had time to shape head and tooth morphology. Alternatively, the selective pressures from soft crayfish predation on head/tooth morphology could be weak, for instance because crayfish‐consuming populations mostly feed on other prey such as fish, tadpoles, and leeches (cf., Forsman & Shine, 1997; Manjarrez et al., 2013). Additionally, other adaptive demands on head morphology may be more important (Rossman & Myer, 1990), while optimal capture and handling of crayfish may require only minor morphological modification (both in T. melanogaster and R. septemvittata and R. grahamii). Indeed, both Regina species have been described as having head and tooth morphologies similar to those of generalist Thamnophiinae snakes (Dwyer & Kaiser, 1997), suggesting that specializing on crayfish does not induce major morphological adaptation.

Snakes preying on soft crayfish may occasionally attack slightly harder ones as these occupy the same refuges and their surface chemicals are capable of eliciting a predatory response (Manjarrez, 2003). If occasionally successful, these attacks could select for morphological adjustments to profit from such encounters. Weak selective pressure of this kind may be operating in both soft crayfish‐eating Regina species and in soft crayfish‐eating T. melanogaster, slowly yielding minor convergence.

The small effect size of our evidence for convergence may reflect the novelty of this niche expansion by T. melanogaster (Arnold, 1981). No phylogeographic analysis has been made, but the restricted geographic expansion of crayfish ingestion (only 3% of the total area of sympatry of crayfish and T. melanogaster; Manjarrez et al., 2013) and its location close to the southern limit of the snake's distribution (the Natricinae originated further north) suggests that crayfish ingestion by T. melanogaster is a recent development (Lozoya, 1988).

It has been proposed that dental morphology in snakes is associated with dietary preferences (e.g., Britt et al., 2009). Thamnophis melanogaster has maxillary teeth that are curved, pointed, and oriented to pierce soft prey such as vulnerable molting crayfish. Only a few snake species ingest hard preys, and they have specialized teeth. For example, R. alleni and R. rigida have maxillary teeth with rounded tips for handling hard crayfish (Dwyer & Kaiser, 1997), whereas Storeria has long maxillary teeth that allow the extraction of land snails from their shells (Rossman & Myer, 1990). The higher number of maxillary teeth in crayfish‐eating T. melanogaster (34.1 ± 3.9 teeth) compared with their congeners (32.2 ± 4.9 teeth) and soft crayfish Regina is unlikely to be an adaptation to ingest soft crayfish per se, as this runs against the trend of fewer maxillary teeth. We should, however, not dismiss too readily the possibility that having more teeth is adaptive when preying on soft crayfish, because different combinations of teeth number, head, and jaw morphology may represent equivalent mechanical solutions to the same problem (see also Arnold, 1993).

The limited scope of morphological microevolution associated with adopting a crayfish diet could also be interpreted as evidence for T. melanogaster being morphologically pre‐adapted to ingest crayfish. Our multivariate analysis supports this hypothesis because in relation to CV2, which explained a third of the variance in the original variables, Regina species and T. melanogaster cluster together and away from T. eques (Figure 2). Thamnophis is a monophyletic group that originated in the Mexican highlands ~5–6 million years ago (Mao & Dessauer, 1971; de Queiroz et al., 2002), whereas Regina is a polyphyletic group first found in North America 4–5 million years ago (Guo et al., 2012), making it more recently evolved than Thamnophis. Consequently, crayfish consumption by T. melanogaster could represent recent dietary convergence (analogy) with Regina rather than homology resulting from the common ancestor of Regina and T. melanogaster and more primitively shared with T. eques. The rarity of soft crayfish ingestion within populations T. melanogaster supports the hypothesis of analogous behavior, and it is more likely a phenomenon of invasion of a new feeding niche in an aquatic diurnal species (Hibbitts & Fitzgerald, 2005).

In conclusion, our analyses suggest that T. melanogaster shows morphological convergence in head and tooth parameters with two Regina species, potentially associated with the ingestion of a novel prey, newly molted crayfish, by the genus Thamnophis.

CONFLICT OF INTEREST

The authors have no conflict of interest on the manuscript.

Species/morph	Juvenile	Adult	Threshold SVL
Thamnophis melanogaster
Crayfish eating	24 (30%)	56 (70%)	33
Noncrayfish eating	21 (24%)	67 (76%)	33
Thamnophis eques	3 (7%)	39 (93%)	39
Regina grahamii	11 (58%)	8 (42%)	31
Regina septemvittata	48 (59%)	33 (41%)	35.5

Snake species/morph	df	Jaw length	Head length	Head width
Regina grahamii	17	0.008	−0.166	0.032
Regina septemvittata	79	0.191	0.195	0.105
Thamnophis melanogaster
Crayfish eating	78	0.960*	0.733*	0.866*
Noncrayfish eating	86	0.957*	0.721*	0.821*
Thamnophis eques	40	0.984*	0.944*	0.967*

Snake species/morph	df	Jaw length	Head length	Head width	Number of maxillary teeth
Regina grahamii 8 males: 11 females	1, 16	0.32 (0.58)	0.13 (0.73)	0.10 (0.76)	0.28 (0.87)
Regina septemvittata 28 males: 30 females	1, 78	0.90 (0.35)	3.0 (0.06)	0.50 (0.47)	0.92 (0.34)
Thamnophis melanogaster
Crayfish eating 54 males: 40 females	1, 77	0.68 (0.41)	1.50 (0.21)	3.40 (0.07)	0.005 (0.94)
Noncrayfish eating 52 males: 40 females	1, 85	2.30 (0.13)	0.83 (0.36)	7.04 (0.01)	0.38 (0.53)
Thamnophis eques 23 males: 19 females	1, 39	4.40 (0.04)	4.70 (0.03)	1.0 (0.33)	0.45 (0.50)

Thamnophis melanogaster	Mean crayfish abundance in pond ± SD	Number of locations sampled	Student's t test
Crayfish eating	0.66 ± 1.2	9	t = 0.58 p = .23
Noncrayfish eating	2.08 ± 3.03	8