Ontogenetic plasticity in cranial morphology is associated with a 1 functional change in the food processing behavior in Alpine newts 2

5 Background : Salamander morphology changes substantially during metamorphosis, prompting the 6 hypothesis that larvae need a different processing mechanism than post-metamorphic adults. 7 Salamandrid newts with facultative metamorphosis are suitable for testing this hypothesis, because 8 paedomorphic and metamorphic adults often coexist in the same population. Facultative paedomorphic 9 newts provide a direct comparison of form-function relationships with specimens of the same species 10 sharing similar body size and feeding on overlapping prey spectra, whilst having divergent food- 11 processing morphologies. 12 Methods : We use high-speed videography to quantify the in vivo movements of key anatomical 13 elements during food processing in paedomorphic and metamorphic Alpine newts ( Ichthyosaura 14 alpestris ). Additionally, we use micro-computed tomography (μCT) to analyze morphological differences 15 in the feeding apparatus of paedomorphic and metamorphic Alpine newts and sort them into late-larval, 16 mid-metamorphic and post-metamorphic morphotypes. Results : Paedomorphic and metamorphic individuals exhibited clear morphological differences of their 18 feeding apparatus. Regardless of the paedomorphic state being externally evident, different 19 paedomorphic specimens can conceal different morphotypes (i.e., late-larval and mid-metamorphic 20 morphotypes). Though feeding on the same prey under the same (aquatic) condition, food processing 21 kinematics differed between late-larval, mid-metamorphic and post-metamorphic morphotypes. may have later paved the way for terrestrial feeding mechanisms. hyobranchial elevator in the PMM. Depression of the hyobranchial system is achieved by a combination of rectus cervices activity and the ligamentous and muscular suspension of the hyobranchial skeleton to the skull. Please the course of the ligaments was obtained from other morphological descriptions [32,34,47] and could not be verified in this study.


Results 81
Functional morphology of the food-processing apparatus 82 Detailed descriptions of the cranial anatomy of Ichthyosaura alpestris and other salamandrids 83 can be found elsewhere [25,31,45-51] and we focus on structures relevant for processing and on 84 specific differences between morphotypes. 85

Cranial osteology 86
The feeding apparatus of the Alpine newt consists of an osseous skull and mandible, and a 87 complex, partially cartilaginous hyobranchial system (i.e., hyobranchial in larval or hyolingual in 88 metamorphosed salamanders, respectively) (see Fig. 1) and prominent muscles (Fig. 2). We group the 89 specimens into three distinct morphotypes: basibranchial; cb 1 -4, ceratobranchial 1 -4; ch, ceratohyal; d, dentary; eo, exoccipital; fr, frontal; hh, hypohyal (also referred to 101 as radial); hb 1 -2, hypobranchial 1 -2; m, maxilla; n, nasal; os, orbitosphenoid; p, parietal; pa, prearticular; pf, prefrontal; pm, 102 premaxilla; ps, parasphenoid; pt, pterygoid; q, quadrate; sq, squamosal; uh, urohyal; v, vomer. Arrows connecting different 103 morphotypes (rows) highlight significant structural differences. Arrows ending in the space between morphotypes marked with † 104 indicate the reduction of the structure. branchial system. In contrast, in the MMM and the PMM, the insertion of the BHE shifted to the antero-144 ventral part of ceratohyal and therefore acts as a protractor of the branchial arch. The SR1 extends from 145 the antero-ventral side of the cartilaginous ceratobranchial I to the medio-lateral region of the 146 hypobranchial I in the LLM, acting as antagonist of the BHE by depressing the tip of the hyobranchial 147 system, while in the MMM and the PMM the SR1 extends from the medial part of the ceratobranchial I 148 to the medio-lateral part of the ceratohyal to act as a protractor of the branchial arch. The most prominent 149 muscle of the hyobranchial system in all morphotypes is the RC that originates from the ventral 150 abdominal trunk muscles and inserts onto the basibranchial. Due to its course and the ligament and 151 muscle suspension of the hyobranchial apparatus on the skull (hyomandibular or hyoquadrate ligament 152 and levator hyoideus), the RC facilitates retraction and depression of the hyobranchial apparatus.

163
Intraoral food processing 164 After initial ingestion via suction feeding, one or two transport movements were used by all 165 morphs to position prey prior to a consecutive set of processing cycles. The mean total processing 166 cycles were 5.7 ± 3.2 (mean ± S.D.) for the larval, 5.6 ± 2.4 for the mid-metamorphic, and 5.9 ± 2.5 for 167 the post-metamorphic morphotypes. A processing cycle was defined from start of gape opening until 168 the next start of gape opening. Processing involved the cyclical opening and closing of the jaw (i.e., 169 arcuate mandible movement), elevation and depression of the hyobranchial apparatus (i.e., the tongue) 170 and, in the metamorphic morphotype only, additional rhythmic flexion and extension of the neck (vertical 171 cranial movement) (Fig. 3 C). During these movements, prey debris and blood were occasionally 172 expelled from the oral cavity, indicating that the behavior caused significant prey disintegration. After a 173 processing bout (i.e., a series of processing cycles), water flows induced by hyobranchial movement 174 transported the food backwards, after which it was either repeatedly processed or swallowed.

Kinematics of intraoral food processing 183
Intraoral food processing cycles were clearly distinguishable from food transport in that 184 hyobranchial elevation accompanied gape opening during processing, whereas during transport 185 hyobranchial depression accompanied gape opening. During processing, at the onset of gape opening, 186 the LLM initiated hyobranchial elevation, which continued past peak gape opening and reached its peak 187 coincident with complete gape closure. Then, in a returning motion, the hyobranchial apparatus was 188 depressed while the mouth remained shut (i.e., stationary phase). The MMM started elevating the 189 hyobranchial apparatus at the onset of gape opening. Both movements peaked approximately at the 190 same time, after which simultaneous gape closing and hyobranchial depression (i.e., resetting 191 movements) occurred ( Fig. 4 B Table 1 shows the kinematic parameters of food processing in the three morphotypes. The 206 stationary gape phase in the LLM clearly differed from the other two morphotypes (compare Fig. 4A with 207 B and C) as did the cranial flexion of the PMM (compare Fig. 4F with D and E). Vertical cranial flexion 9 Ventral (°) n/a n/a n/a n/a 12.61±6.51 0.39 10 Dorsal (°) n/a n/a n/a n/a 12.30±6.46 0.40 11 Ventral duration (s) n/a n/a n/a n/a 0.10±0.05 0.41 12 Dorsal duration (s) n/a n/a n/a n/a 0.17±0.07 0.33 13 Cycle duration (s) n/a n/a n/a n/a 0.   Cycle duration 3.56 1.00 n/a n/a n/a n/a n/a n/a 0.88 n/a n/a n/a n/a n/a n/a Statistical analysis was calculated using Kruskal-Wallis 1-way ANOVA and only performed on parameters present in all 217 morphotypes. P-values were Bonferroni adjusted to account for multiple testing; significant p-values are indicated by asterisks.
218 Some significant changes concern the duplication of the vertical hyobranchial magnitude of the PMM 219 compared to the MMM (compare with Fig. 4H and I), the duplication in gape magnitude from the MMM 220 to PMM (compare Fig. 4B and C), and the significantly higher mean mandible acceleration from peak 221 gape opening to reaching maximal gape-closing speed in the LLM compared to the MMM. The durations 222 of the gape and vertical hyobranchial movement cycle are the same across all morphotypes.

Ordination analysis of processing kinematics 224
A principal component analysis (PCA) was performed to analyze how the processing kinematics 225 of the three morphotypes relate to each other and to visualize differences. Distribution of the chewing 226 cycles among the processing modes and morphotypes on the first two principal components axes Figure  227 5, and the loadings of the kinematic parameters on principal component 1 and 2 (i.e., PC1 and PC2) 228 are given in Table 3. Hyobranchial kinematics load more strongly on PC1 while mandible kinematics 229 loaded more strongly on PC2. Processing in PMM and LLM are separated in kinematic space with no 230 overlap, but MMM processing overlaps with both LLM and PMM. 231 The coefficient of variation (CV) was calculated for each kinematic parameter (Table 1)   249

Stomach content analysis 250
Post-metamorphic newts used in stomach content analysis applied suction feeding to ingest 251 lake fly larvae (Chironomidae). After ingestion, the newts used cyclic processing movements involving 252 ventral cranial flexion and mouth opening accompanied by hyolingual elevation. Microscopic 253 examinations of the processed lake fly larvae extracted from the stomachs of freshly euthanized newt 254 specimens revealed clear lesions and other structural damage. Lesions were recognized by intensified 255 methylene blue staining, which gradually attenuated along the unharmed part of the prey (Fig. 6B -D). We found distinct intraoral food processing kinematics (Fig. 5) and feeding apparatus 267 morphologies ( Fig. 1 and 2) in the three heterochronic morphotypes of the Alpine newt. Thus, this study shows that externally similar animals can have different internal anatomies, which in turn may result in 269 different food processing mechanisms. 270 It was recently shown that metamorphosed salamandrid newts use loop-like movements of their 271 hyobranchial apparatus (i.e., tongue) to translate food across the palatal dentition (i.e., tongue-palate 272 rasping) [20]. It has also been suggested that salamandrids with a larval morphology cannot employ the 273 same processing mechanism as metamorphic animals because of morphological constraints, including 274 the lack of a movable tongue and diverging dentition patterns in larval morphotypes [14,15] Our data support this hypothesis, as we observed many differences in prey processing behavior across 285 heterochronic morphotypes (Table 1 and Fig. 5). First, gape excursion, which consists of gape opening 286 and closure, is lowest in the LLM, mid-range in the MMM, and greatest as well as significantly different 287 from the aforementioned in the PMM (Table 1 and 2). Similarly, the duration of the gape excursion is 288 shortest in the LLM, mid-range in the MMM, and largest in the PMM -whereby that of MMM and PMM 289 differ significantly from that of the LLM (Table 1 and 2). The extent and duration of the gape opening, 290 which are roughly on the order of the extent and duration of the gape closure, loaded most heavily on 291 PC2 (0.872 and 0.699 respectively) -which separates the kinematics of the three processing modes in 292 kinematic space (Fig. 5 and Table 3). Second, the LLM was the only morphotype to exhibit a stationary 293 phase after the gape cycle and the PMM was the only morphotype to show cranial flexion during 294 processing ( Fig. 4 and Table 1). Third, movements of the hyobranchial apparatus in the PMM were 295 significantly greater than either in the LMM or MMM (Table 1 and 2). In sum, the MMM differed in 6 out 296 of 12 kinematic parameters from the LLM, the PMM in 8 out of 12 kinematic parameters from the LLM 297 and the PMM differed in 6 out of 12 kinematic parameters from MMMsuggesting that they apply 298 different food processing mechanisms. 299 The PMM of the Alpine newts used its tongue to cyclically and rhythmically drive the food against 300 the vomerine dentition on the palate (Fig. 3C and Video 3), very similar to movement patterns reported 301 for the crested newt [20], that uses tongue-palate rasping to process prey. In fact, our stomach content 302 analysis revealed that processing in I. alpestris caused substantial mechanical damage to the food 303 objects (Fig. 6). Tongue-palate rasping in the PMM was characterized by relatively stereotypical 304 movements of the mandible (CV between 0.21-0.32, except for the relatively variable gape closure 305 duration and acceleration with 0.44 and 0.45 respectively) and relatively flexible tongue movements (CV 306 between 0.31-0.53, except for the relatively stereotypical hyobranchial cycle duration with 0.23), which 307 may indicate that the tongue movements need to be flexibly adjusted during processing. Similar to the 308 PMM, the MMM showed evidence of a tongue-palate rasping mechanism being used ( Fig. 3 B and Video 2) with a modified tongue motion pattern (compare Fig. 4B and H to C and I) and small and 310 sporadic cranial movements (Fig. 4E). Tongue-palate rasping in the MMM was characterized by 311 relatively stereotypical movements of the tongue (CV between 0.28-0.35, excluding the relative flexible 312 duration of hyobranchial elevation with 0.45) and relative flexible mandible movements (CV between 313 0.38-0.57, excluding the relative stereotypical gape cycle duration with 0.27). With regard to the switch 314 from chewing to tongue-palate rasping from the LLM to the MMM respectively, this could suggest that a 315 relatively stereotypical motion sequence is used first when mastering a new behavior pattern, while this 316 motion sequence can become more flexible during ontogenesis (as seen in the PMM). 317 The LLM used a processing mechanism with a restricted amount of gape opening, which 318 prevented us from determining how food was processed. However, we could distinguish the post-319 ingestion behavior (i.e., jaw and hyobranchial movements) into prey transport (characterized by 320 hyobranchial depression during gape opening [16-18]) and rhythmic food processing (characterized by 321 hyobranchial elevation throughout or during some of the gape opening cycle [20,22]). Food-processing 322 kinematics in the LLM involved the highest mean gape-closure acceleration (Fig. 4A -C and Table 1). 323 As the mandibles of all morphotypes are of approximately the same size and therefore likely have 324 approximately the same mass, the finding that the LLM showed the highest mean gape-closure 325 acceleration, might suggest that they exhibit the highest bite force. This, in turn, could indicate that the 326 dentition of the mandible could be directly involved in intraoral food processing (i.e. chewing). We use 327 the term bite force to describe the result of the action of the mandible elevator muscles modified by the 328 craniomandibular biomechanics [55] and thus the force that the mandible can transmit onto an object in 329 the oral cavity (therefore not merely equivalent to adductor muscle force). Additionally, one of the most 330 striking characteristics of the LMM cranio-mandibular anatomy is its overbite, causing dentition on the 331 mandible to occlude between the two functional upper jaw systems, creating an effective shearing bite 332 against the palate (Fig. 1 B and C). Consequently, the morphology of the LLM supports the idea that it 333 chews its food using the tooth-bearing mandible (Fig. 1) to pierce the prey against the palate (i.e., 334 'mandible-palate clenching') while the tongue and dentition on both functional upper jaws hold the prey 335 in place. The kinematic profiles support this assumption as initial gape opening is followed by 336 hyobranchial elevation, which potentially act to position and hold the prey in the area of the occlusal 337 surface on the palate, before the mandible accelerates towards the palate ( Fig. 4A and G) to bite the 338 prey. Additionally, since the prey occasionally protruded far out of the mouth in the LLM, we were able 339 to observe how the jaws acted on it in a clenching manner (Video 1). 340 Externally, the processing behavior of LLM showed striking similarities with the chewing 341 behavior of another paedomorphic salamander, Siren intermedia. It was shown using high-speed x-ray 342 analyses that S. intermedia use its mandible to rasp the prey across the dentition of the palate [22]. 343 Larval Alpine newts, however, chew their food using simple arcuate movements of the mandible (i.e. 344 opening-closing), and switch from chewing to tongue-palate rasping during ontogeny. Tongue-palate 345 rasping appears to become the main food processing mechanism before the tongue is completely 346 remodeled during metamorphosis (Fig. 2 B,  It has been previously hypothesized that salamanders show a phylogenetic trend of evolving 365 tongues with greater protrusion potential, increased freedom of the branchial arch in relation to the hyoid 366 arch and that tongue prehension might have evolved from a manipulative function [27]. In line with that 367 idea, we found a concurrent ontogenetic process of remodeling in the tongue apparatus. In which the 368 tongue develops from a bulky relatively inert system (i.e., hyobranchial system) with small protrusion 369 ability in the LLM (Fig. 1D and 2A) to a delicate and relatively mobile system (i.e., hyolingual system) 370 with greater protrusion ability in the PMM (Fig. 1L and 2C). The LLM hyobranchial system has a 371 muscular anatomy that creates motion-potential in all directions of the median plane. However, tongue 372 protraction is limited to geniohyoideus and hyomandibularis which act as the primary tongue protractor 373 complex in larval salamanders ( Fig. 2A). During the metamorphosis in the MMM, the branchiohyoideus 374 externus and subarcualis rectus 1 are rearranged to functionally suspend the branchial arch on the 375 paired ceratohyal (i.e., hyoid arch) (Fig. 2B). This muscle rearrangement enables a more effective 376 protraction of the branchial arch, since it can now be moved by the suspension on the hyoid arch and 377 thus pulled further anteriorly (Fig. 2B). This secondary tongue protractor complex allows the tip of the 378 tongue to be ejected out of the mouth which has been described for post-metamorphic salamandrids 379 [32] and in turn is the functional basis for tongue-palate rasping [20]. Our data suggest that aquatic 380 salamandrid larvae begin to use their tongue for processing ( Fig. 3B and Video 2) as soon as the 381 mandibular reorganization prevents them from chewing but their tongue morphology enables improved 382 protraction during development. Thus, we hypothesize that salamanders that are able to protract their 383 tongue effectively and have a metamorphic palatal dentition are potentially able to combine these 384 elements to achieve tongue-palate rasping. Consequently, it is likely that tongue-palate rasping is the 385 general processing mechanisms in salamanders with a metamorphic feeding apparatus morphology. 386 Additionally, our data support Regal's hypothesis that tongue prehension likely evolved from a 387 processing or manipulation function of the tongue [27] as our animals mastered tongue-palate rasping 388 before they were apt to leave the water and thus before they used their protractible tongue to catch prey. 389 Mid-metamorphic Alpine newts develop the ability to rasp a food item against the palatal 390 dentition and engage in tongue-palate rasping due to rearrangements of the branchiohyoideus externus