Functional Diversity of Evolutionary Novelties: Insights from Waterfall-Climbing Kinematics and Performance of Juvenile Gobiid Fishes

Synopsis The evolution of novel functional traits can contribute substantially to the diversification of lineages. Older functional traits might show greater variation than more recently evolved novelties, due to the accrual of evolutionary changes through time. However, functional complexity and many-to-one mapping of structure to function could complicate such expectations. In this context, we compared kinematics and performance across juveniles from multiple species for two styles of waterfall-climbing that are novel to gobiid fishes: ancestral “powerburst” climbing, and more recently evolved “inching”, which has been confirmed only among species of a single genus that is nested within the clade of powerburst climbers. Similar net climbing speeds across inching species seem, at first, to indicate that this more recently evolved mode of climbing exhibits less functional diversity. However, these similar net speeds arise through different pathways: Sicyopterus stimpsoni from Hawai’i move more slowly than S. lagocephalus from La Réunion, but may also spend more time moving. The production of similar performance between multiple functional pathways reflects a situation that resembles the phenomenon of many-to-one mapping of structure to function. Such similarity has the potential to mask appropriate interpretations of relative functional diversity between lineages, unless the mechanisms underlying performance are explored. More specifically, similarity in net performance between “powerburst” and “inching” styles indicates that selection on climbing performance was likely a limited factor in promoting the evolution of inching as a new mode of climbing. In this context, other processes (e.g., exaptation) might be implicated in the origin of this functional novelty.

tribute substantially to the diversification of lineages. Older functional traits might show greater variation than more recently evolved novelties, due to the accrual of evolutionary changes through time. However, functional complexity and many-to-one mapping of structure to function could complicate such expectations. In this context, we compared kinematics and performance across juveniles from multiple species for two styles of waterfall-climbing that are novel to gobiid fishes: ancestral ''powerburst'' climbing, and more recently evolved ''inching'', which has been confirmed only among species of a single genus that is nested within the clade of powerburst climbers. Similar net climbing speeds across inching species seem, at first, to indicate that this more recently evolved mode of climbing exhibits less functional diversity. However, these similar net speeds arise through different pathways: Sicyopterus stimpsoni from Hawai'i move more slowly than S. lagocephalus from La R eunion, but may also spend more time moving. The production of similar performance between multiple functional pathways reflects a situation that resembles the phenomenon of many-to-one mapping of structure to function. Such similarity has the potential to mask appropriate interpretations of relative functional diversity between lineages, unless the mechanisms underlying performance are explored. More specifically, similarity in net performance between ''powerburst'' and ''inching'' styles indicates that selection on climbing performance was likely a limited factor in promoting the evolution of inching as a new mode of climbing. In this context, other processes (e.g., exaptation) might be implicated in the origin of this functional novelty.

Synopsis
FRENCH: Diversit e fonctionnelle des innovations evolutives: l'exemple de la cin ematique et des performances de grimpe des chutes d'eau des juv eniles de gobies R esum e L' evolution de nouveaux traits fonctionnels peut contribuer significativement a la diversification des lign ees. Les traits fonctionnels les plus anciens peuvent montrer plus de variabilit e que les plus r ecents du fait de l'accumulation de changements evolutif au cours du temps. Cependant, ces pr edictions peuventêtre complexifi ees par la diversit e des fonctions et par l'implication de plusieurs structures dans une même fonction. Dans ce contexte, nous avons etudi e la cin ematique et les performances de grimpe des chutes d'eau de plusieurs esp eces de gobies utilisant deux styles de grimpe originaux au sein de cette famille: le mode « powerburst » plus ancestral et le mode « inching » qui a evolu e plus r ecemment. Le mode inching n'a et e confirm e que pour deux esp eces du même genre incluses au sein du clade des powerburst. Des vitesses de grimpe similaires entre les esp eces utilisant le mode inching paraissent indiquer que ce mode de grimpe, qui a evolu e plus r ecemment, pr esente une diversit e fonctionnelle moins elev ee. Toutefois, la similarit e des vitesses de grimpe entre les deux esp eces s'explique par des processus diff erents: le Sicyopterus stimpsoni d'Haiwaï se d eplace plus lentement que le S. lagocephalus de La R eunion mais passe plus de temps en mouvement. La production de performances similaires, r esultant de processus diff erents, refl ete un ph enom ene semblable a celui de l'implication de plusieurs structures dans une même fonction. Si les m ecanismes sous-jacents ne sont pas explor es, ces similarit es peuvent perturber l'interpr etation des diff erences relatives de diversit e fonctionnelle entre les lign ees.

Introduction
The evolution of novel functional abilities is widely regarded as a key factor that can influence the ecological diversification of lineages (Liem 1973;Wainwright 2007). Through the evolution of novel functional capacities, opportunities to exploit new resources or habitats may open, promoting diversification through either the advent of new specializations, or the radiation of taxa within a lineage (e.g., Konow et al. 2008). Alternatively, some functional novelties may have little impact on diversification (e.g., Price et al. 2010;Wainwright and Price 2016), or even reduce aspects of functional diversity in a lineage (e.g., Higham et al. 2015).
Numerous factors might affect how novel traits contribute to ecological and functional diversification (Vermeij 2001). For example, older traits might show greater functional variation due to the accrual of evolutionary changes through time (Wainwright and Price 2016). Alternatively, novel functional capacities could emerge through evolutionary transitions that pass through a flexible, intermediate range of performance between lower and higher extremes (Blob 2001). In such cases, recently evolved novelties might show greater variation in performance than older traits. Beyond such considerations, the complexity of many functional systems (Wainwright 2007) leads to a potential for multiple structural arrangements of features to produce similar functional performance (many-to-one mapping of structure to function: Alfaro et al. 2005;Wainwright et al. 2005). Such circumstances could further complicate expectations for the functional diversity that would be exhibited by novel evolutionary traits.
The waterfall-climbing abilities found in multiple species of gobiid fishes represent a remarkable functional novelty, through which potential patterns of diversification in functional performance can be tested. Waterfall-climbing is common among amphidromous juvenile gobies that are returning to adult stream habitats. After completing a marine larval phase, climbing is facilitated by a ventral sucker comprised of the fused pelvic fins (Schoenfuss and Blob 2003;Blob et al. 2006;Schoenfuss et al. 2011). Previous observations identified two distinct modes of climbing used by different species of gobies. ''Powerburst'' climbers use pectoral fin adduction and bursts of axial undulation to propel themselves upwards between periods of attachment to the substrate with the pelvic sucker. In contrast, ''inching'' climbers sequentially detach and reattach the pelvic sucker and a novel, oral sucker to climb, with little axial movement. Inching has been confirmed only in the genus Sicyopterus (Schoenfuss and Blob 2003;Blob et al. 2006;Schoenfuss et al. 2011;Maie et al. 2012), which is phylogenetically nested within several powerburst outgroups (Taillebois et al. 2014; Fig. 1). Thus, powerburst climbing appears to be ancestral, and inching more recently evolved (Cullen et al. 2013).
Comparisons among powerburst species from Hawai'i and the Caribbean have shown a wide range of performance within this climbing mode. However, inching performance has only been measured in one species, S. stimpsoni from Hawai'i (Schoenfuss and Blob 2003;Blob et al. 2006;Schoenfuss et al. 2011). To evaluate whether more recently evolved novel traits, like inching, might show less functional diversity than ancestral traits, like powerburst climbing, we compared our previous data on climbing kinematics and velocity from Hawaiian and Caribbean species (Schoenfuss and Blob 2003;Blob et al. 2006;Schoenfuss et al. 2011) to new data from two additional species from the Indian Ocean island of La R eunion: the inching climber S. lagocephalus, and the powerburst climber Cotylopus acutipinnis.

Materials and methods
Experiments were conducted under St. Cloud State University Institutional Animal Care and Use Committee guidelines (protocol 12-07). Sampling in La R eunion was conducted under permit N 15-002/DEAL/SEB/UPEMA issued by Direction de l'Environnement de l'Am enagement et du Logement de La R eunion.
Juvenile S. lagocephalus and C. acutipinnis were collected by electroshocking from the St. Etienne River in La R eunion during April 2015. Fish were collected near the river mouth to ensure that they were recent, postmetamorphic recruits that had not previously climbed. Fish were transported by car to the Hydro R eunion laboratory facility, where they were maintained in aerated stream water at ambient temperature ($20 C) and a 12:12 h cycle of room lighting. Fish were acclimated a minimum of 24 h prior to testing, which proceeded for 1-2 days after collection. Rocks were placed in enclosures to provide cover and a source of algal growth for food.
To facilitate comparisons across species, kinematic and performance data from La R eunion gobies were collected and analyzed following methods of previous studies as closely as possible (Schoenfuss and Blob 2003;Blob et al. 2006;Schoenfuss et al. 2011). Briefly, the kinematics of individual climbing cycles were recorded in ventral view using highspeed video (100-200 Hz, Fastec Highspec 2G) as groups of fish ascended an inclined panel of Plexiglas (52 , lightly roughened with sandpaper) with water flowing as a sheet over its surface (180 mL s À1 ). Lighting was provided by light-emitting diode (LED) fixtures or ambient room lighting. Measurements were performed by tracking landmarks on the head, fins, body axis and suckers (Fig. 2) using DLTdv5 for MATLAB (Hedrick 2008), and using the coordinate data for these landmarks to calculate kinematic variables in MATLAB (Schoenfuss and Blob 2003;Schoenfuss et al. 2011). Performance trials were recorded in dorsal view with Sony Handycam mini-DV cameras (60 Hz), over a standard distance of 20 cm ($10 body lengths), as fish ascended an inclined (70 ), textured plastic chute with water flowing down its surface (Blob et al. 2006). Video records were used to determine the portion of time each fish spent moving versus resting as it ascended the 20 cm distance, from which we calculated climbing speeds while moving and net climbing speeds including rest time. Performance metrics were compared across species via Kruskal-Wallis analyses with Dunn's post-hoc tests corrected for multiple comparisons, using the Prism 6.1 statistical package (Graph-Pad software).

Results
Inching S. lagocephalus from La R eunion showed similar kinematic profiles to patterns observed previously in S. stimpsoni from Hawai'i ( Fig. 3A; Supplementary Movies S1-S2). The area of the oral sucker initially decreases as the front lip advances    Supplementary Table S1. early in the cycle. However, the area of the oral sucker then increases as fish reattach it to the substrate while the pelvic sucker advances up the surface. Ranges of values for kinematic variables are also similar between the species, with the slightly longer upward advance of the oral and pelvic suckers during each cycle in S. lagocephalus corresponding to the slightly larger average body size of this species (25 mm standard length) compared to S. stimpsoni (22 mm; Table 1; Schoenfuss and Blob 2003). Kinematic profiles across powerburst climbers also show generally similar patterns, with moderate ranges of variation in kinematic variables ( Fig. 3B; Supplementary Movies S3-S4). In Hawaiian powerburst climbers, Sicydium punctatum from the Caribbean and C. acutipinnis from La R eunion, body segments near the head and tail show the greatest amplitudes during axial undulation, with caudal body segments showing the greatest angles to the direction of travel (Fig. 3B). However, larger C. acutipinnis (16 mm) and S. punctatum (19 mm) exhibit larger values for both of these variables than the smaller (14 mm) Hawaiian taxa.
In contrast to kinematic comparisons, the two climbing styles show different patterns of variation in performance across taxa. With regard to net climbing speed over 20 cm, including periods of motion and rest, the two species of inching climbers show similar performance to each other (and to three of the four powerburst climbers: mean 0.10-0.14 BL s À1 ; Table 1; Fig. 4A). In contrast, the powerburst climber C. acutipinnis from La R eunion (mean 0.06 BL s À1 ) is significantly slower than two of the other powerburst climbers (Awaous stamineus and Lentipes concolor), as well as one of the inching species (S. lagocephalus) (Dunn's post-hoc test P < 0.05; Table 1; Fig. 4A). Thus, powerburst climbers include species with at least two distinct levels of net climbing performance, whereas inching climbers include only one such group. Across all of the variables that we compared, the average coefficient of variation was also greater for powerburst climbing species (45 ± 3.1 standard error [SE]) than for inching species (33 ± 3.2 SE) (Mann-Whitney Utest, P < 0.01). However, this apparent greater performance variation among powerburst climbers is a product of the variables that underlie this performance. Comparisons of movement speed (i.e., speed restricted to periods of actual movement) show significant differences between the two inching climbers, whereas the powerburst climbers group similarly (Table 1; Fig. 4B). It is noteworthy, though, that C. acutipinnis uses the slowest movement speeds among powerburst species (Table 1; Fig. 4B). Comparisons of the time spent in motion over 20 cm show distinctions between the powerburst (<25%) versus inching ($50%) climbers (P < 0.05), with species within each climbing style Table 1. Means (± SD) of morphological and performance variables for juveniles of waterfall climbing stream gobies from Hawai'i (A. stamineus, L. concolor, S. stimpsoni), Dominica (S. punctatum) and La R eunion (C. acutipinnis, S. lagocephalus), with numbers in parentheses indicating the sample size of locomotor events from which data were collected with each cycle from a different individual  (11) (Blob et al. 2006), data for S. punctatum from (Schoenfuss et al. 2011), data for species from La R eunion new in this study. 2 Superscript letters (a, b, c) indicate groupings of significantly different species (P < 0.05), based on Dunn's post-hoc tests corrected for multiple comparisons, conducted after Kruskal-Wallis analyses. 3 Data arcsine transformed prior to Kruskal-Wallis analysis. n/a, not applicable; BL/s, body lengths per second.
Functional novelties in gobiid fishes grouping together (Table 1; Fig. 4C). However, both C. acutipinnis and S. lagocephalus from La R eunion showed lower mean values than other species using their respective climbing styles. As a result, the similar net speeds of inching climbers seem to arise through different pathways: S. stimpsoni from Hawai'i move more slowly than S. lagocephalus from La R eunion, but may spend more time moving. In contrast, the cumulative effect of seemingly minor differences in both movement speed and the portion of time spent moving, differences that were not initially recognized as significant, led to distinctly slow net speeds for C. acutipinnis among powerburst species.

Discussion
Recently evolved novelties have had less time to accrue evolutionary changes that might contribute to functional diversity than older traits. As a result, older traits might be expected to show more variation than recent evolutionary novelties (Wainwright and Price 2016). However, our results from climbing gobies illustrate how the equivalent performance of alternative functional pathways can influence interpretations of the relative diversity exhibited by functional traits. Comparisons of net climbing speed between inching and powerburst species gave an initial indication that more recently evolved climbing by inching was largely similar in performance across species, whereas the older trait of powerburst climbing displayed greater variation across the taxa that retain this climbing mode (Table 1; Fig. 4A). In contrast, comparisons of the mechanisms that underlie the performance of both styles present a different picture, in which powerburst climbing species might show more consistent average patterns than inchers (Fig. 4B). These patterns result because, within each climbing style, species appear to use largely similar kinematics in different ways-for example, by spending more (or less) time moving compared to resting (Fig. 4C). Thus, without considering the mechanics that underlie performance, the functional diversity climbers on the right). Colors represent differences in locality, with white boxes showing data from Hawaiian species derived from Blob et al. (2006), gray boxes data from Caribbean species derived from Schoenfuss et al. (2011), and orange boxes new data from R eunionese species. Different boldface letters above each box plot indicate significant differences between groups, determined by Kruskal-Wallis analyses with Dunn's posthoc tests, corrected for multiple comparisons (P < 0.05). As, A. stamineus; Lc, L. concolor; Sp, S. punctatum; Ca, C. acutipinnis; Ss, S. stimpsoni; Sl, S. lagocephalus. Original data and sample sizes are reported in Supplementary Table S2. present within more recently evolved inching might have been masked. Such masking could lead to a potentially oversimplified conclusion that functional diversity is lower in the more recent evolutionary novelty.
The complexity of many functional systems opens opportunities for similar performance to be produced through multiple pathways. One variety of this general phenomenon that has received considerable attention is the many-to-one mapping of structure to function, in which multiple anatomical arrangements of structures may produce equivalent functional performance Wainwright et al. 2005). Our data from waterfallclimbing gobies provide an example of the broader phenomenon of which many-to-one mapping of structure to function is a part, in which alternative functional pathways-such as different climbing styles, or different patterns of motion versus rest within a climbing style-can produce equivalent locomotor performance. For example, three of the four powerburst species compared show net climbing speeds that can be statistically grouped with those of both inching species observed, despite using different propulsive mechanisms to scale waterfalls (Table 1; Fig. 4A). In addition, among inching species, equivalent climbing performance can be achieved by longer durations of slow movement (e.g., S. stimpsoni) and shorter durations of faster movement (e.g., S. lagocephalus; Table 1; Fig. 4).
The ultimate similarity of net climbing performance between the two broad styles of climbing in gobies is striking. Out of the six species we compared, from islands in three different ocean basins (Pacific, Caribbean, and Indian), five showed net climbing speeds between 0.10 and 0.14 BL s À1 . Only C. acutipinnis used markedly slower speeds (0.06 BL s À1 ), which were a product of moderately slower movement and a smaller proportion of time spent moving, but which still overlapped with the net speeds of some of the other five species (Fig. 4). Given that climbing behaviors are components of a migratory phase of the life cycle, to what extent might differences in performance impact features such as the in-stream distributions and upstream penetration of species? In La R eunion, C. acutipinnis has a higher failure rate during attempted climbs than faster S. lagocephalus (Lagarde et al. 2018), and it takes C. acutipinnis 20-25% more days to migrate to adult upstream habitats (Lagarde et al. 2020). Although this could allow faster juvenile S. lagocephalus to reach suitable habitat before C. acutipinnis, both species ultimately penetrate to comparable upstream distances (Teichert et al. 2014). Hawaiian species of amphidromous gobies show even more limited correlations between climbing performance and in-stream distribution. Although the two Hawaiian powerburst species exhibit nearly identical net speeds, one species (A. stamineus) penetrates the shortest distance upstream and the other species (L. concolor) penetrates the furthest, with inching S. stimpsoni reaching intermediate distances between the two Blob 2003, 2007). Given that climbing performance does not predict ecological factors such as habitat distribution, changes in net performance appear unlikely to have formed a selective basis for the evolutionary origin of inching as a functional novelty in climbing style. Instead, the origin of inching with an oral sucker may have been tied to non-locomotor pressures, potentially reflecting an exaptation of feeding adaptations for locomotor roles (Cullen et al. 2013).