Phenotype Variation in Niphargus (Amphipoda: Niphargidae): Possible Explanations and Open Challenges

Abstract

Understanding phenotype variation is among the central topics in biology. We revise and reanalyze studies of the amphipod genus Niphargus to confront two potential mechanisms driving its phenotype variation, namely, cladogenesis and adaptive evolution. We found evidence for both mechanisms. Reanalysis of a subset of traits using molecular phylogeny showed moderate phylogenetic signal, consistent with the hypothesis that overall phylogenetic variation increases with phylogeny. The phylogenetic signal in Niphargus traits seems to be stronger at the tips of the phylogeny than at basal splits. Indirect evidence suggests that much of the phenotype variation can be attributed to adaptive evolution. Both lines of evidence are consistent with the hypothesis that Niphargus evolved in several adaptive radiations, where theory predicts that most of the phenotype variation evolves early, when ecological niches are vacant. As the niches fill up, the rate of phenotype variation slows down and becomes associated with cladogenetic events. This hypothesis can explain the high level of trait-convergence and unresolved taxonomy above the species level. The main caveats to these hypotheses comprise lack of experimental evidence for trait function and nonquantified heritable component of trait variation. Promising venues towards better understanding of phenotypic variation include studies of ontogenetic variation, functional interactions between traits, and genome–phenotype associations.

1. Introduction

Understanding the origin of phenotype variation is one of the central topics in biology. Human fascination over diversity of phenotypes predates naturalist explorations and origin of science, as evidenced by societies of indigenous hunters exploiting the information conveyed by outer appearance of organisms [1]. With the beginnings of science, many fields of biology were driven by observations of phenotype variation, such as taxonomy [2], developmental biology, and evolutionary ecology [3,4]. To explain mechanisms driving phenotypic variation, the variation is nowadays being studied on different time scales and at different hierarchical levels of biological organization, between individuals and species [5]. Two commonly invoked and mutually nonexclusive mechanisms explaining phenotypic variation are cladogenesis and natural selection, captured in the narratives of phylogeneticists and adaptationists, respectively. Both mechanisms jointly shape phenotype variation in the wild [6], although their relative importance is hard to assess. We use the term “phenotype” in a broad sense, referring to a trait that can be observed and measured. Although both phylogeneticists and adaptationists assumed that traits are heritable, we do not have prejudice regarding what mechanism underlies the expression of a trait [7]. We limit this discussion to morphological phenotype.

The central idea of phylogeneticists is that descendants inherit traits from ancestors; thus, closely related species are more similar to each other than to distantly related species. History of cladogenetic events both increases and explains most of the observed phenotypic variation of the clade. Phenotypic traits thus presumably convey information on species’ relatedness [8]. Based on this premise, phylogeneticists employed phenotypic variation to reconstruct phylogenetic relationships. The procedure of phylogenetic inferences using phenotype variation was formalized by Willi Henning. He suggested that shared ancestry can be inferred from synapomorphies, traits derived from the common ancestor, and shared among all descendants [9]. Phylogenetic relationships among species can be then reconstructed with the help of a character matrix using the principle of parsimony [8]. Ideally, characters subjected to parsimony analysis are selectively neutral and evolve under a model of neutral evolution at low pace [10]. Assuming a homology of aligned sequences, these premises were subsequently extended to and upgraded with the models of molecular evolution in the fields of molecular phylogenetics and DNA barcoding [8]. An important methodological challenge of phylogeneticists using morphological variation is determining how to assess primary homology of characters and character states, and how to account for the within- and between-species variation [11].

By contrast, adaptationists assumed that much of phenotype variation evolves in response to natural selection, driving heritable trait towards its adaptive optimum in subsequent generations [4]. Many authors uncritically treated all traits as adaptive, and circularly defended this view using post hoc explanations of trait function and trait adaptive value. This practice was criticized by Gould and Lewontin, who demanded that the term “adaptation” be used upon meeting criteria of heritability, tested function, and association between trait value and individual’s fitness [12]. It has been shown that selection-driven phenotype variation can result in substantial between-population differences within a few generations [13], and is associated with well-known phenomena, such as character displacement [14,15,16,17]. Such traits often evolve convergently and thereby mask true phylogenetic relatedness [18].

The motivations of phylogeneticists and adaptationists are clearly different. The former are interested in neutrally evolving traits, mirroring phylogenetic relationships. Any within-species variation in phylogenetically informative traits introduces uncertainty into the phylogeny reconstruction. By contrast, research of adaptationists depends on the study of variation itself, where phylogenetic contingency in traits due to shared ancestry represents an unwanted noise and violation of assumption of statistical independence [19]. However, to understand net-phenotypic variation in nature, we need to ask to what extent can variation be explained by neutrally evolving traits and adaptive traits, corresponding to degree of phylogenetic relatedness and change of selective regime, respectively. The nowadays availability of molecular phylogenies makes this task easier, under the assumption that molecular phylogenies accurately depict the hierarchy of branching events. We discuss possible answers in this article using the case of the amphipod genus Niphargus.

Niphargus is a predominantly subterranean genus of amphipods, living in the Western Palearctic, and with more than 400 species it comprises one of the largest genera among freshwater amphipods [20,21,22] (Figure 1). These species inhabit virtually all subterranean habitats, ranging from springs and waters close to the surface [23,24,25,26], to fissure systems, cave streams, and deep phreatic lakes [27,28], as well as acid [29], sulfidic [30,31,32,33], brackish [34], and thermal waters [35,36] (Figure 1 and Figure 2). The species feed on detritus and scavenge and prey on other invertebrates, ranging from crustaceans to insect larvae [37,38,39]. Cannibalism seem to be common [40]. The main competitors seem to be other Niphargus species and, to a lesser extent, other amphipods [39,41], although the evidence for competition is indirect [23,32,42]. The genus originated 50–60 Mya in the area that nowadays belongs to Western Europe [43], from where it spread towards the east. With the onset of Alpine and Dinaric orogenesis and subsequent karstification, a number of subterranean habitats emerged (representing the so-called ecological opportunity) in the region of the Western Balkans. This prompted adaptive radiations with pronounced variation in functional ecology [29,44,45]. The taxonomy of the genus has been notoriously challenging, with high intraspecific variation and small differences among species. Molecular phylogenies have shown that cryptic species are common and that any nominal name not examined by molecular tools on average comprises 2–3 genetically divergent molecular operational taxonomic units (MOTUs) [46]. Although MOTUs were often considered as reasonable proxies for species [45], they should not be equated to species without additional evidence [47]. Noteworthy, the degree of crypticism was only rarely analyzed [48,49,50]; therefore, it is likely that some MOTUs could be morphologically diagnosable. Many authors attempted to split the genus into smaller, morphologically defined species groups or genera, but neither of the proposed splits have been successful so far [51,52]. Despite description of several smaller genera [53,54,55], a large majority of the species has remained attributed to the genus Niphargus.

The aim of this review is to reconsider phenotypic variation within the context of molecular phylogenies, assess the importance of phylogenetic and functional-adaptive drivers, and provide a hypothesis explaining the phenotypic variation of Niphargus. We identify the main gaps in our knowledge and provide suggestions for future research.

2. Terminological Note

Literature interchangeably uses two terms, traits and characters. The term “trait” is a more general term, commonly employed in functional ecology, and encompasses discrete and continuous properties of an organism’s body. The term “character” and related term “character state” were adopted in cladistics. Even if the term describes the continuous traits, for the needs of parsimony analysis, it is more or less arbitrarily categorized into discrete character states (e.g., small–large). Throughout the text, we use the term “trait” with exception to parts describing cladistic analyses, where we stick to widely used jargon and use the term “character”.

3. Phylogenetic Signal

We name phylogenetic signal a property of a trait that conveys information about the ancestor–descendant relationship, i.e., a measure of neutrality/adaptivity ratio of a trait used to assess the degree of phylogenetic relatedness and change of selective regime that can be measured using different metrics (see below) [56,57]. When strong, descendants from the same ancestor are similar to each other. Phylogenetic signal in morphological traits is crucial for accuracy of parsimony analyses, inference of synapomorphies [8], and taxonomic implications [58]. The degree of phylogenetic signal can be assessed from the distribution of a trait on a molecular phylogeny and hierarchical branching [56]. Here, we review studies that were explicitly or implicitly based on phylogenetic signal: attempts of splitting the genus into subgenera or species groups, and results of cladistic analyses and tests of the phylogenetic signal.

Niphargus has been long known for its intra- and interspecific variation. Despite several species being extremely similar to each other, the extent of morphological variation across all 400 species is impressive (Figure 1). In spite of this variation, there are genus-level traits such as clefted telson, uropod III with reduced endopodite, uropod I without facial spines, mitten-form gnathopods, and uni- or biarticulated accessory flagellum, which make Niphargus well distinguishable. This overall “Niphargus appearance” is phylogenetically conserved and can be seen already in 50–30 My old specimens from Baltic amber [59,60,61]. A minor number of species was erected to separate genera [53,54,55,62], whereas most of the species have been assigned to the genus Niphargus. In this review, we treat these genera as part of Niphargus. Several authors attempted to assign Niphargus species into their own subgenera or species subgroups. Sub-generic taxonomy encountered two frustrations. Firstly, subgenera or species groups were defined on a one-by-one basis, leaving the rest of the species unclassified. The most complete subgeneric structure of 12 species groups was proposed by Milan Straškraba; however, even his classification did not avoid the “taxonomic leftover”, the 13th category incertae sedis [51]. The second frustration was poor diagnosability of the subgenera. Attempts to diagnose subgenera were rare, mostly made by Stanko Karaman (but see [63]). In a series of partial reviews of Niphargus, he classified a fraction of genus diversity into subgenera using diagnoses, all complemented with a disclaimer that diagnostic combination is not exclusive [64,65,66,67]. Hence, Gordan Karaman and Sandro Ruffo in their review of Niphargus diversity referred to subgeneric division reluctantly [52].

To our knowledge, the first discussion on traits within the cladistic framework was written by Sket [25]. He contrasted the morphology of N. valachicus against the rest of Niphargus, hypothesizing that the species likely resembles Niphargus ancestor. Niphargus valachicus lives close to the surface in Pannonian lowlands, i.e., the remains of Paratethyan Sea. Sket proposed that many of the species’ traits could be plesiomorphic, based on the observation that N. valachicus lives in a region of putative biogeographic origin of the genus (inferred from high species diversity) and that life close to surface represents ancestral ecology of the genus. However, he did not formally test his hypothesis [25]. Hypothetical plesiomorphies should be treated with caution, as there is some evidence that the genus originated in the west [43] and that the subterranean–surface boundary might have been colonized secondarily, from the subterranean environment [26,68].

Figure 2. Latest Niphargus molecular phylogeny comprising 562 molecular operational taxonomic units (MOTUs) and reconstructed using Bayesian inference [69]. The phylogeny was used in calculations of phylogenetic signal (Table 1). Tips colored according to MOTU ecology (Herrera-Alsina et al., submitted).

Figure 2. Latest Niphargus molecular phylogeny comprising 562 molecular operational taxonomic units (MOTUs) and reconstructed using Bayesian inference [69]. The phylogeny was used in calculations of phylogenetic signal (Table 1). Tips colored according to MOTU ecology (Herrera-Alsina et al., submitted).

Cladistic analyses are few, all of them suggesting a high degree of homoplasy in characters. Sket and Notenboom studied relationships among 12 species of the so-called transitivus species group [63], using a priori polarized characters (apo- vs. plesiomorphic) and differently restricted optimizations (reversals allowed vs. not allowed), yet without outgroup. By contrast, Fišer et al. tested for the monophyly of the large-bodied species belonging to the orcinus species group using unpolarized data for 33 ingroup and 13 outgroup species, and unconstrained parsimony analysis [70]. The third analysis studied 112 species and confronted results with molecular-data-based phylogenetic analysis [71]. This study coded continuous variables using two alternative methods, subjective eye-detection of the largest gaps in the continuous variation, and step-matrix-gap weighted method. The latter does not make assumptions about homology of character states, but uses raw data and converts differences among them into weights employed in parsimony analyses, i.e., transition among two very different values is penalized more than transition between two near-identical values [11].

The conclusions of all these studies are similar: all three studies detected high degree of homoplasy in characters, as inferred from consistency indices for each trait. Consistency index is a ratio between realized number of transitions between character states on a phylogeny, and minimal possible number of transitions for the respective character (e.g., one step in binary character). Its values ranged between 0 (maximal homoplasy) and 1 (minimal number of transitions) [8]. Consistency index was in all analyses calculated on morphology-based phylogenies. In the most restrictive analysis not allowing reversals, the consistency indices reached rather high value of 0.75 [63], whereas in unconstrained analyses, values ranged from 0.2 to 0.5 [70].

All presumed species groups or subgenera turned out to be poly- or paraphyletic [70,71]. In order to detect hierarchic level below which species morphology accurately matches molecular phylogenetic structure, Fišer et al. employed the entropy equation [71], and showed that the accuracy of the match between morphology and molecular phylogenetic structure declines when clade size exceeds two to three species. Considering the estimation that every morphological species covers on average 2–3 cryptic species [46], we tentatively suggest that morphology accurately delimits only clades of closely related species, i.e., at most up to ten species per clade (three morphologically described species multiplied by three undiagnosable or yet unidentified species) share synapomorphies.

To test for phylogenetic signal of the selected traits (recognized as functional traits in Niphargus; see below), we reused the available molecular phylogeny comprising 311 molecular operational taxonomic units [69] (Figure 2) and morphometric data [72] (a summary of analysis in the caption of Table 1, data and scripts are available at Zenodo, doi: 10.5281/zenodo.12107269). The studied traits showed variable strengths of phylogenetic signal between different traits (Table 1). We used two methods for a test of phylogenetic signal, Blomberg’s K [56] and Pagel’s λ [57]. Both methods assume Brownian motion (BM) as evolutionary model. Blomberg’s K compares trait variance in observed data to trait variance that would be expected under Brownian motion (tests fit to BM model), while Pagel’s λ evaluates the strength of phylogenetic signal by scaling phylogenetic tree branch lengths to find the value of λ that produces a phylogenetic tree best explaining the observed data. Blomberg’s K = 1 indicates a perfect fit to the BM model and strong phylogenetic signal, while K < 1 and K > 1 indicate less and more phylogenetic signal than expected, respectively. Pagel’s λ is distributed between 0 and 1, where 0 indicates no phylogenetic signal and 1 indicates a strong phylogenetic signal. In Niphargus, Blomberg’s K ranged from 0.14 (gnathopod I carpus length) to 0.51 (body length) (Table 1), suggesting that all tested traits bear less phylogenetic signal than expected by Brownian motion, i.e., that closely related species are more dissimilar than expected under BM. On the other hand, Pagel’s λ ranged between 0.63 (antennae II length) and 0.91 (body length), indicating a moderate to strong phylogenetic signal in tested traits (Table 1). While the strength of signal apparently varies, it is noteworthy that both metrics consistently suggest that phylogenetic signal is smaller than expected.

Table 1. Phylogenetic signal in selected functional traits of Niphargus. For input data, we used the latest available molecular phylogeny comprising 562 molecular taxonomic units (MOTUs) from across the distribution range [69] pruned to 311 MOTUs with available morphological data [72]. We processed the morphological data as described in [72], and calculated phylogenetic signal using R package phytools: function phylosig() with default settings and additional hypothesis test for K and λ [73]. Data, code, and comments for performing the analyses are available at https://doi.org/10.5281/zenodo.12107269. K = Blomberg’s K [56], λ = Pagel’s λ [57].

Trait	K	p Value (K)	λ	p Value (λ)
Gnathopod I carpus length	0.14	0.240	0.67	0.000
Antenna II length	0.17	0.062	0.63	0.000
Gnathopod II carpus length	0.17	0.069	0.72	0.000
Antenna I length	0.20	0.040	0.66	0.000
Pereopod VII basis width	0.23	0.003	0.72	0.000
Pereopod VII length	0.24	0.016	0.79	0.000
Gnathopod II perimeter	0.25	0.015	0.84	0.000
Pereopod V basis width	0.25	0.010	0.71	0.000
Pereopod VI basis width	0.26	0.007	0.69	0.000
Coxa II depth	0.26	0.006	0.76	0.000
Pereopod V length	0.28	0.005	0.85	0.000
Body length	0.51	0.001	0.91	0.000

Table 1. Phylogenetic signal in selected functional traits of Niphargus. For input data, we used the latest available molecular phylogeny comprising 562 molecular taxonomic units (MOTUs) from across the distribution range [69] pruned to 311 MOTUs with available morphological data [72]. We processed the morphological data as described in [72], and calculated phylogenetic signal using R package phytools: function phylosig() with default settings and additional hypothesis test for K and λ [73]. Data, code, and comments for performing the analyses are available at https://doi.org/10.5281/zenodo.12107269. K = Blomberg’s K [56], λ = Pagel’s λ [57].

Trait	K	p Value (K)	λ	p Value (λ)
Gnathopod I carpus length	0.14	0.240	0.67	0.000
Antenna II length	0.17	0.062	0.63	0.000
Gnathopod II carpus length	0.17	0.069	0.72	0.000
Antenna I length	0.20	0.040	0.66	0.000
Pereopod VII basis width	0.23	0.003	0.72	0.000
Pereopod VII length	0.24	0.016	0.79	0.000
Gnathopod II perimeter	0.25	0.015	0.84	0.000
Pereopod V basis width	0.25	0.010	0.71	0.000
Pereopod VI basis width	0.26	0.007	0.69	0.000
Coxa II depth	0.26	0.006	0.76	0.000
Pereopod V length	0.28	0.005	0.85	0.000
Body length	0.51	0.001	0.91	0.000

4. Functional Morphology

The seminal study of amphipod functional morphology was made by Dahl [74]. Amphipods of gammaridean type have a laterally compressed body, where flattened pereopod coxae, broad bases of pereopods V–VII and ventrally elongated epimeral plates form a “ventral channel” (Figure 3a). In the anterior part of this ventral channel (on the pereon) are gills and oostegites in females. In the posterior part (pleon segments I–III) are broom-like pleopods that generate water current flowing along the entire channel. Water current enters the channel anteriorly and exits from it posteriorly. The currents bring chemical cues, organic particles, and oxygen aerating the gills and female marsupium. When the animal is bent, the entering and exiting currents cancel each other out, whereas when the animal stretches, the current turns into jet propulsion [74].

While this elegant body plan had been known for a long time, the detailed functions of individual traits were not. A separate line of Niphargus research searched for the association between the phenotype and its function, such as habitat exploitation, locomotion, sensory ecology, trophic ecology, antipredation mechanism, and reproductive biology. Some traits can contribute to diverse functions; we call them multifunctional. Below, we review the traits and their presumed function (Figure 3b). The studies were of a comparative nature in which morphological variation was associated with habitat use (habitat-related traits, life history traits), stable isotopes (feeding-related traits), or simple behavioral observations (locomotion-related traits). Neither of these studies was designed experimentally, calling for the caution in result interpretation.

Body size is an important multifunctional trait. It can be measured as body length or as body mass. Both measures covary, although the relationship between the two depends on body shape; the selection of body size measure can affect the results of a study [75]. Body size corresponds to habitat properties. Species from interstitial and fissure systems are, on average, small, whereas species from streams and lakes are large [28]. Large-bodied species are, on average, faster [76]. Males and females may differ in body size. Large males apparently evolve in response to male–male rivalry [77], while large females are more fecund [78]. Body size allometry is important for female fecundity. Larger females have more eggs than smaller females of the same species [41]. Moreover, species with large females have more eggs and larger eggs than species with small females. After controlling for the body length, the number of eggs per female no longer differed among the large and small species. Moreover, the reproductive effort, defined as the ratio between clutch volume and volume of female, is significantly smaller in large-bodied females. In other words, species with large females have more offspring in absolute terms and larger eggs despite the lower reproductive effort [78].

Body shape is a multifunctional trait that varies between stout and slender shape. It can be inferred from the depth of pereopod coxal plates and the width of pereopod bases. In small-pore habitats, species are either small and stout or large and slender [79]. In noninterstitial habitats, species from streams and springs are, on average, more slender than species from caves [28]. It has been suggested that deeper ventral channels are associated with better swimming capacity [76], but likely also with higher breathing capacity. Indeed, several species living in sulfidic lakes are stout. Water rich in sulfide is anoxic and oxygen is diluted only in the upper millimeters of water. Some species from sulfidic water use water tension to cling upside down close to the surface and thus exploit the thin layer of oxygenated water for breathing [80,81].

Antennae are probably the main chemosensory and tactile organ. Each antenna comprises a peduncle and a flagellum composed of a number of articles. Flagellum articles on antenna I bear aesthetascs, specialized setae with apical pore, likely used for the detection of chemical cues in water. Some species with long antennae have two or more aesthetascs per article [82,83]. Longer antennae have more articles and more chemosensory setae and likely better allocate food. While behavioral evidence for this hypothesis is lacking, the length of antenna seem to be associated with development of chemosensory-processing lobes of the brain [84]. Longer antennae are likely beneficial in stagnant water, but not in currents. Indeed, species from streams have shorter antennae [28].

Mouthparts comprise a pair of mandibles, two pairs of maxillae, and maxilliped. Mouthparts are engaged in food processing [85,86,87]. Studies of the functional significance of mouthparts in Niphargus are lacking [39]. In general, Niphargus mouthparts indicate ubiquitous feeding biology and little variation among the species. However, there are a few noteworthy deviations in the shape of mandibles and maxilla I. On the mandible, all Niphargus species have a row of hairy setae between the molar and pars incisiva. In some species, this row is long and dense, suggesting that it may contribute to filtering [88,89]. Maxilla I comprise palp and two lobes. Generally, the outer lobe bears seven spines, the inner one pectinate, and the remaining ones have one to three denticles. In some species, the number of spines exceeds seven, and/or all spines are pectinate [30,53,80,88]. The latter two types of maxillar spination were found in interstitial species, in species from cervices, and from sulfidic lakes. It seems that these types of mandibles and maxillae filter smaller particles, such as bacteria. The hypothesis remains to be tested.

Gnathopods are the first two pairs of trunk appendages, with enlarged propodus (6th article) and subchelate dactylus. In general, amphipods use gnathopods for grooming, feeding, burrowing, and mating, when the male holds female in a precopula. As Niphargus do not form long-lasting precopula [90,91], the main function of gnathopods is feeding. Correlations between nitrogen stable isotopes (δ15N) and gnathopod sizes suggest that species with larger gnathopods occupy higher trophic positions than species with smaller gnathopods. Variation in shape (inclination of palm) correlates with carbon stable isotopes (δ13C) and suggests a division of trophic niche within trophic position [37]. Usually, the first pair of gnathopods is smaller, except in some alleged predators in interstitial, suggesting that animals in sand need to have the first pair designed to grab the prey [79,92]. One species, N. trullipes, has shovel-like gnathopod dactylus, and Sket observed that these species burrow into the clay (pers. comm.).

Pereopods and pleopods contribute to locomotion. The length of pereopods correlates with speed [76], as usual in crustaceans [93]. Bases of pereopods build part of the ventral channel and may be in some broad species. Pleopods are broom-like appendages, generating water currents. The function of pereopods and pleopods should be studied in greater detail, including the gait analysis in crawling and swimming, as well as in explosive movement.

Uropods are three the most posterior pairs of appendages, used in locomotion and sensory function. In some species, the third pair of uropods is shorter and flattened [70,80,88], and may be used in steering while swimming, although this hypothesis has never been tested.

Dorsal spines in Niphargus are not true cuticular extrusions, but modified thick and rigid setae at dorso-posterior margins of the pleon [94]. It seems that these structures are homologous with hairy and flexible setae, found in all other species. Spiny pleons were noted in species that live in lakes with the olm (Proteus anguinus). Spatial co-occurrence of the olm and spiny Niphargus, as well as simulations of spine evolution, suggest that these structures could be an antipredation mechanism [95]. Unpublished observations of the spiniest species, N. balcanicus, bred in the laboratory suggest that these structures appear early in the development and that this trait is not induced by the presence of the olm.

Formation of ecomorphs is complex. In an early study, it was shown that body size, body shape, and appendage length covary with habitat properties: species from interstitial and crevices are small, while species from streams and lakes are larger. Species from cave streams are slender, short-legged, and with short antennae, whereas species from cave lakes are stout and long-legged [28]. A subsequent analysis showed that some of the functional traits are interconnected, and that only a subset of all possible trait combinations is adaptive [96]. Moreover, ecomorphs within the habitat differ with respect to their trophic niches. Species occupying higher trophic positions have longer appendages regardless of the habitat, implying their greater agility. However, species occupying higher trophic positions in interstitial increase in their slenderness, whereas species occupying higher trophic positions from the streams are stouter. This suggests that predators are more mobile species than species from lower trophic levels. Slender species in interstitial enhance their crawling capacity through the voids in a sand, whereas the stouter species from cave streams and lakes allegedly enhance their swimming capacity [96].

5. Intraspecific Variation

Intraspecific variation is a common phenomenon, emerging on all levels of biological organization. While many authors acknowledge the high intraspecific variation of Niphargus species [97,98,99,100], the phenomenon itself has rarely been studied. The within-species variation emerges due to age, sex, and environmental conditions (Figure 3b).

Ontogenetic variation was studied in differently sized individuals of nine species [101]. The analysis suggests that antennae, pereopods, and uropods grow allometrically (see also [99]), that gnathopods during development remarkably change shape, and that individual spines and/or setae develop in subsequent instars. Importantly, the sequence of individual developmental events differs between the species. A developmental event is called a distinct event in organismal development, such as accelerated/slowed growth of the appendage or time of spine appearance. The analysis of these events suggests high between-species heterogeneity in developmental sequences, that studied traits were development-wise independent, and that heterochrony [3] may be an important evolutionary mechanism in Niphargus.

Sex dimorphism is clearly expressed in some species, but absent in others [100]. In sexually dimorphic species, males are usually larger than females, and have either elongated inner ramus of uropod I, elongated distal article of uropod III, or elongated both uropods. Sexual dimorphism is attributed to male–male rivalry, as larger males with overdeveloped uropods easier chase away smaller ones. Degree of sexual dimorphism is directly and indirectly linked to habitat properties. Species close to the surface and species from cave streams have roughly similar number of males and females, whereas in species from cave lakes, females are often more numerous than males [102]. In female-biased species, male–male competition for females is weakened, and sexual selection maintaining hypertrophied male phenotypes is relaxed. Moreover, females of these species are under strong fecundity selection favoring females with large bodies [78]. Such females no longer differ in size from males. Hence, the degree of sexual dimorphism diminishes from the surface towards the more stable environment in deeper parts of the karstic massif [77]. Sexual dimorphism was not studied in interstitial species. At least one interstitial species, N. dimorphopus from the UK and Netherlands [103], has sexually dimorphic gnathopods, suggesting that males and females separate in their trophic niches [104].

Adaptation to local conditions either due to different environmental factors or due to interspecific competition drives phenotype variability on a between-population level [14,16]. Niphargus croaticus is an extreme ecomorph, with extremely long appendages and antennae [105]. The species mostly lives in cave lakes, but was occasionally found in cave streams. In part of its range, the species co-occurs with its sister species, N. subtypicus. An analysis of appendage lengths showed that populations living in streams have shorter appendages than populations living in lakes, suggesting that water current affects the appendage length. Moreover, lake populations that co-occur with N. subtypicus have longer appendages than those that live in allopatry, suggesting competition-driven character displacement [42].

6. Discussion

6.1. Phylogenetic Contingency or Selection Acting on a Functional Trait or Both

A phylogenetic perspective on Niphargus phenotype variation is compatible with a functional one. Niphargus evolved in a series of adaptive radiations [29]. The theory of adaptive radiation predicts initially fast diversification that slows down as the ecological niches fill up [106]. Speciation in the early phase of diversification is associated with a substantial ecological and phenotype divergence [107,108]. Many of the hitherto studied traits in Niphargus are functional and evolve in response to natural selection. Rapid changes of phenotypes, especially in the early phase of adaptive radiation, can explain a mismatch between phenotype similarity and molecular phylogenetic structure. In the late phase of adaptive radiation, when ecological niches fill up and ecological diversification-driven speciation slows down, phenotype similarity more accurately depicts molecular phylogenetic relationships, consistent with previous observations [71]. The attempts of splitting genus into subgroups or subgenera failed because Niphargus evolved in multiple radiations that resulted in convergent ecomorphs [29]. The similarity of the presumed subgenera is, thus, not an outcome of the shared ancestry, but a result of environmental selection [28] that is additionally reinforced by the between-trait interconnection [96].

Overall, it seems that phenotype variation of the herein discussed Niphargus traits has been shaped predominantly by selection on functional traits. This conclusion has a potential caveat. Many traits have not yet been examined in depth, and we expect that variation of a subset of unknown or at least unstudied traits might have been driven predominantly by phylogenetic relatedness. A long list of traits awaits to be tested [109]. An unsolved issue is determining how between-trait-interconnection interferes with phylogenetic signal. Some traits, for example, might have low phylogenetic signal because they evolved along with the traits under selection, and vice versa, some traits may be maladaptive because of their connectedness with phylogenetically conserved traits.

Finally, we recognize that—at least in theory—phylogenetic informativeness could be hidden in functional traits. Natural selection maximizes phenotype performance. Performance is an outcome of concerted action of many functionally linked traits [110,111]. For example, faster locomotion can be achieved by longer legs or faster muscles. Trait integration into complex functional traits could be phylogenetically channelized, e.g., members of one clade could elongate their appendages by tiny elongation of all appendage articles, whereas members of another clade could keep the length of all but one article fixed. To assess whether or not such phylogenetic channelization occurs, complex traits (e.g., listed in [72]) should be decomposed into subordinate traits that would be the subject of detailed observation [112] and analyzed within the comparative framework.

6.2. Future Directions

We are only beginning to understand mechanisms driving phenotype variation of Niphargus. We foresee several directions of future research. These include study of phenotype variation on a subspecies level, experimental testing of trait function, quantitative genetics, and genome-wide association studies (Figure 3c).

A handful of studies on variation below the species level suggest that interindividual and interpopulation variation is a rich yet unexploited source of information. Differences in growth patterns [101] could help in the recognition of phylogenetically informative traits. Even if the adults of different species look similar, growth patterns of traits may differ and point to convergent origin of adult phenotypes. Additionally, between-population variation may emerge as a response to local selection [42]. If natural selection shapes phenotypes, we can expect similar trends in variation on a population and species level.

All herein cited studies of functional traits are correlative, and the evidence for trait function is indirect. It is timely to experimentally quantify traits performance, a challenging task that relies on breeding a high number of individuals and their manipulation [113].

Also challenging will be the assessment of trait heritability to discriminate adaptive evolution from phenotypic plasticity. Phenotypic plasticity is a ubiquitous phenomenon in nature that can evolve [7]. This mechanism allows organisms to colonize a new environment [7], such as a subterranean environment [114], whereas in stable environments, it becomes reduced [7]. Niphargus is an old inhabitant of the subterranean environment [29] and we hypothesize weakened phenotypic plasticity. Understanding the genetic basis of trait variation is important for both adaptationists and phylogeneticists, as both study heritable traits. Assessment of heritability requires breeding of a large number of individuals subjected to common garden experiments. Amphipod breeding is challenging, both for subterranean [115] as well as epigean species [116]. To our knowledge, only one study has tested for heritability of traits in subterranean amphipod Gammarus minus [117].

Finally, entire genomes of nonmodel organisms are becoming more and more achievable [118]. This, in combination with automated phenotypization [119], may accelerate data acquisition and routinize studies that associate genome variation with phenotype variation [120].