Among living marine mammals, toothed whales (Odontoceti) are the only clade with multiple lineages that have independently evolved obligate freshwater habits from marine ancestries. These lineages, termed “river dolphins,” are a paraphyletic grouping of odontocetes that are broadly endemic to estuarine or exclusively freshwater river systems. Traditionally, this grouping has been delimited to the living genera Inia, Pontoporia (which is predominantly coastal to estuarine in habitat), Platanista, and the recently extinct Lipotes (Simpson, 1945; Rice, 1998; Hamilton et al., 2001; Nikaido et al., 2001; Pyenson, 2009). Sotalia fluviatilis, a delphinid that exclusively lives in the main Amazon River system, also qualifies as a “river dolphin,” although it is generally excluded from the group that Simpson (1945) classified. With the exception of Pontoporia, the aforementioned living genera are distributed in freshwater systems with independent tectonic and geomorphologic histories in Asia and South America (Fig. 1). Phylogenetic analyses using molecular (Hamilton et al., 2001; Nikaido et al., 2001; May-Collado and Agnarsson, 2006; Steeman et al., 2009), total evidence (Geisler et al., 2011), and recent morphological (Geisler et al., 2012) datasets confirm four independent invasions of freshwater river systems, including Sotalia in the Amazon River system (Cunha et al., 2005).
Prior to recent phylogenetic work, the shared morphological specializations among “river dolphins” created confusion about their genealogical relationships. This suite of morphological features, which is mostly not represented in other extinct and living odontocetes, include traits such as: flexible necks with unfused cervical vertebrae; reduced or absent dorsal fins; reduced orbits and eyes; elongate rostra with lingual accessory cusps on the posterior dentition; and wide, triangular, paddle-like flippers (Muizon, 1988a, 1994; Simpson, 1945; Arnason and Gullberg, 1996; Heyning, 1997; Messenger and McGuire, 1998; Rice, 1998; Cassens et al., 2000; Hamilton et al., 2001; Sanchez and Berta, 2012). Not all of these features, however, are necessarily shared among all “river dolphins.” For example, although the osteology of Inia geoffrensis has been known since the 19th century (Flower, 1867), Klima et al. (1980) described an unusual structure in its pectoral girdle that consists of the inclusion of the sternum as part of the shoulder girdle (Fig. 2). This condition, which allows a broader range of rotation for the forelimb, is unique among living mammals. Klima et al. (1980) suggested that this joint was an adaptation for high maneuverability at low swimming speeds in complex and obstructed aquatic environments, such as the flooded forests and small tributaries of the Amazon and Orinoco basins.
Understanding the evolution of these unusual morphologies in “river dolphins” has been hindered by a sparse fossil record for stem members of Platanista and Lipotes (Pyenson, 2009). However, Inia and Pontoporia, two endemic and exclusively South American genera that are sister taxa (forming the clade Inioidea sensu Muizon 1988a), possess a comparatively rich fossil record (Fordyce and Barnes, 1994; Fordyce et al., 1994; Messenger and McGuire, 1998; Cozzuol, 2010). Fossil pontoporiids have been identified in Neogene marine deposits in both the Southern and Northern hemispheres (Muizon, 1984, 1988b; Lambert and Post, 2005; Pyenson and Hoch, 2007; Godfrey and Barnes, 2008; Gibson and Geisler, 2009; Gutstein et al., 2009; Lambert and Muizon, 2013). Fossil Iniidae, by contrast, appear restricted mostly to South America, with work by Cozzuol (1985, 1996, 2010), Muizon (1984, 1988b), Cozzuol and da Silva (1996), recognizing three fossil genera in South America (Fig. 1): Inia, Saurocetes, and Ischyrorhynchus, with the latter two being extinct and restricted to the late Miocene Ituzaingó Formation of the Entre Ríos Province, Argentina. Goniodelphis hudsoni, described originally by Allen (1941) from the late Miocene to early Pliocene of Florida, USA, has been considered as a probable basal iniid by Cozzuol (2010; see also Morgan, 1994). Recently, the discovery of Meherrinia isoni and its phylogenetic position confirms the presence of the family in the late Miocene of North Carolina, USA. (Geisler et al., 2012). Presently, Inia lives in rivers of the Amazon and Orinoco basins (Best and da Silva, 1989). We follow the Society of Marine Mammalogy's Ad Hoc Committee on Taxonomy in arguing that the phylogenetic separation among these populations satisfies the criteria for recognizing two different living species, I. boliviensis and I. geoffrensis (Hamilton et al., 2001; Banguera-Hinestroza et al., 2002; Hollatz et al., 2011), as opposed to three subspecies within I. geoffrensis.
From the late Miocene Ituzaingó Formation, Ischyrorhynchus vanbenedeni is the best-known species, represented by several partial skulls and fragments of rostra and mandibles with teeth (Pilleri and Gihr, 1979; Cozzuol, 1985, 1989, 1996). Saurocetes gigas is known by a proximal fragment of a mandibular symphysis and isolated teeth, and S. argentinus is known from several specimens including incomplete mandibles, and a partial skull with the palatal region preserved (Cozzuol, 1985, 1989, 1996). Cranial fragments have also been described from the late Miocene Urumaco and Solimões formations (Cozzuol, 1996, 2006, 2010; Linares, 2004). Here, we described a single specimen collected from the lower levels of the Ituzaingó Formation (Fig. 1), which consists of a well-preserved humerus. This specimen exhibits hallmark iniid features and provides important ecomorphological information for the antiquity of riverine adaptations in this group.
MATERIALS AND METHODS
We conducted osteological comparisons among specimens of I. geoffrensis. Vouchered specimens of I. boliviensis are rare and poorly represented in natural history collections around the world, but any osteological differences between these two congeners would not impact the results of our investigations. We examined physically mature appendicular skeletons from the Instituto Nacional de Pesquisas Amazônicas, in Manaus, Amazonas, Brazil (MA 002, 004, 010, 019, 034, 104, 105, 107, 109, and 110, all of which were collected in the Amazon Basin), from the Museo de Ciências Naturais of the Pontifícia Universidade Católica de Minas Gerais, Belo Horizonte, Minas Gerais, Brazil (MCN-M 32 and 89, both form the upper Araguaya River), and from the National Museum of Natural History, Smithsonian Institution, Washington, DC (USNM 396166). For general comparison, it was also observed and photographed a Delphinidae humerus of Tursiops truncatus (USNM 504096). Sex identification for most of these specimens in Brazilian collections were lacking but there was little variation of humerus and sternum morphology and size. Sexual dimorphism has been reported in Inia based on body size (Martin and da Silva, 2006); nevertheless, it has not been found on osteological features. Nomenclature for anatomical musculature follows Strickler (1978) and Klima et al. (1980).
Fossil material (MLP 76-5-7-3, Departamento Científico de Paleontología de Vertebrados, Museo de La Plata, La Plata, Argentina) consists of an isolated right humerus with the greater tubercle broken. This specimen was collected from the so-called “conglomerado osífero,” at the base of the Ituzaingó Formation, from the cliffs of Paraná River in Entre Rios Province, Argentina. Cione et al. (2000, 2005) stated that the “conglomerado osífero” should not be older than Tortonian in age (late Miocene, 11.6–7.2 Ma; Gradstein et al., 2012) based on its stratigraphic position overlying the Paraná Formation. Based on the faunal assemblages from the base of the Ituzaingó Formation that correlate with the Huayquerian South American Land Mammal Age, a late Miocene age may be considered more precise (∼9–6.5 Ma; Flynn and Swisher, 1995; Cione et al., 2000).
The incomplete humerus (MLP 76-5-7-3) shares the same basic shape with that of Inia, bearing similar proportions (Table 1), but longer in length and more slender in thickness. It is generally dorsoventrally compressed, with a large and semispherical head. The lesser tubercle is well-developed, but smaller and less separated from the humeral body than in other odontocetes. Despite breakage leading to the loss of the greater tubercle, it is clear that this structure was merged from the greater tubercle, as it is in Inia. Notably, in other odontocetes, these two tubercles are separated (Figs. 3 and 4). The absence of articular facets on the distal, ulnar border of the humerus reflects the absence of an olecranon of the ulna, which is a character found in a minority of odontocetes, such as Inioidea, Platanista, Monodontidae, and the extinct delphinidan Atocetus (Muizon, 1988a).
Table 1. Morphological measurements comparing Inia geoffrensis (average and standard deviation, n = 11) and the fossil iniid specimen from the Ituzaingó Formation
Width at 50% length
Average of Inia geoffrensis
99.28 ± 4.96
37.39 ± 4.48
54.91 ± 2.95
23.25 ± 1.45
Percentage difference (fossil / extant sample)
On the dorsolateral surface (Fig. 2), the proximal end is dominated by the humeral head, and a separate, but poorly developed, lesser tubercle. The reduced greater tubercle is mostly missing. The head of the humerus is connected with the body by a narrow neck. The lesser tubercle is aligned with the radial border, and it is separated from the head of humerus by a shallow groove. It is posteriorly continuous with the humeral head (Fig. 4F), permitting its articulation with the sternum extension of the glenoid cavity, which Klima et al. (1980) described as the joint cavity of the sternum in I. geoffrensis (Fig. 2).
Distally, on the flat surface of the humerus body, there are two relatively deep oval-shaped depressions. The proximal depression lies at the base of the greater tubercle near the radial border. It is likely that the supraspinatus and, perhaps one head of the mastohumeralis muscle, attached there. The distal depression lies at the ulnar border and is interpreted here as the attachment for the glenohumeral ligament (proximally) and, perhaps, the teres major muscle (distally); both of these attachments are separated by a low crest. Of the two depressions, the proximal one is the deepest, which may indicate that the corresponding ligament was stronger and perhaps larger than in Inia (Figs. 3 and 4). Near the anterodistal border of the humerus, there is a broad and low tuberosity for the attachment of the deltoideus muscle. At the distal end of the bone is its broadest portion, forming a V-shaped profile that corresponds to the convergence of the ulnar and radial articular surfaces. The radial edge is blunt, rugose, and slightly concave at the attachment for the infraspinatus muscle.
The medioventral surface (Fig. 4) of the humerus is smooth and gently convex, and the muscle attachments are not very evident, appearing as quite shallow depressions. The head of the humerus is abraded by postmortem processes along its posteroventral side. Also, the lesser tubercle is shifted anteroventrally. Ventral to this latter process, a shallow depression indicates the attachment for the subscapularis muscle. Proximally, close to the ulnar border, there is an attachment surface for the latissimus dorsi muscle. Distally along the ulnar border, there is a very shallow depression interpreted as the attachment for the teres minor, as well as a scar on the distal border that we interpret as insertion for the pectoral major muscle. Finally, in cranial view, the head of humerus is more than two times larger than the lesser tubercle, which is similar to Inia and many delphinidan odontocetes (Fig. 3). The anterior margin of the lesser tubercle is a polished, somewhat convex surface, which corresponds to the attachment of subscapularis muscle.
Despite the absence of iniid postcrania in their fossil record, MLP 76-5-7-3 shares clear similarities with extant Inia, such as its proportions, dorsoventral compression, its lesser tubercle with a rounded medioventral surface, and a similar configuration of muscle attachment scars on the bone surface. The overall similarity, in our view, permits us to assign MLP 76-5-7-3 to Iniidae. There are, however, important differences that separate the fossil specimen from Inia, including: a comparatively reduced lesser tubercle; a deeper glenohumeral ligament insertion; a thinner and sharper ulnar border determining that muscle insertion that occurs on the humerus radial, and ulnar borders in Inia are dorsolateral or medioventral in MLP 76-5-7-3 (Fig. 4).
Klima et al. (1980) found that the arrangement of the shoulder girdle and its musculature in I. geoffrensis was quite different from other odontocetes, including its sister taxon, Pontoporia, whose shoulder joint Strickler et al. (1978) dissected in detail. Klima et al. (1980) linked the unusual shoulder joint in I. geoffrensis (Fig. 2) with its high maneuverability, a functional interpretation backed by long-standing observations of its swimming behavior. Both species of Inia have flexible bodies with a low swimming speed and high maneuverability (Fish, 2002), and Inia is the only odontocete known to swim backwards (Klima et al., 1980). The scapulohumeral sternal joint has not been observed in any other living or extinct cetacean, and the participation of sternum in the shoulder joint is also a unique character among living mammals as well (Klima et al., 1980). Given that the morphology of Inia's humerus is directly linked with this shoulder rearrangement, it is fair to assume the fossil humerus (MLP 76-5-7-3) belonged in an iniid with a similar joint and range of movement.
We propose here that the presence of the scapular-sternal joint, the pattern of muscle attachments on the humerus, and their functional consequences should be considered as general Iniidae characters, and not solely restricted to Inia. Because there are no other reported or known postcranial elements for any fossil Iniidae and because MLP 76-5-7-3 was not collected in association with any diagnostic cranial elements, we cannot provide any more precise taxonomic determination. However, we think that it is reasonable to infer this specimen belonged to Ischyrorhynchinae (sensu Cozzuol, 1996) because this subfamily exclusively dominates the relative abundance of collections from the “conglomerado osífero” at the base of the Ituzaingó Formation.
Moreover, given the late Miocene age of this unit, we can use MLP 76-5-7-3 to constrain a minimum age for the acquisition of a postcranial ecomorphology that is unique among all odontocetes. This timeframe is also broadly consistent with the estimated molecular divergence time between Inia and Pontoporia (17.6–8.8 Mya; Banguera-Hinestroza et al., 2002). Specifically, we argue that this ecomorphology is uniquely associated with maneuverability in the turbid, obstructed, and shallow rivers, lakes, and flooded forests that form the habitat for extant Inia (Martin and da Silva, 2004; Martin et al., 2004). High turbidity is a consequence of massive sediment load, relatively shallow water depth, low energy, and frequent root obstruction by driftwood and aquatic vegetation, which all require morphological specializations to enhance swimming control at low speeds (Fish, 2002). We parsimoniously infer that the ancestral state of this character in iniids was similar to the current shoulder girdle articulation in Pontoporia, non-inioid Delphinida, and basal odontocetes, but the sparse iniid fossil record is currently unrevealing in this regard.
Nonetheless, our ecomorphological inference about the antiquity of flooded forests in the Ituzaingó Formation is consistent with fluvial interpretations of the depositional environment, based on sedimentological and stratigraphic data (Cione et al., 2005). The Solimões Formation in Acre, Brazil, which may extend to all of western Amazonia (under several formational names), part of Venezuela (the Urumaco Formation) and Paraná Basin (“conglomerado ossífero” at the base of Ituzaingó Formation) in Argentina, also preserves fossil iniids. Cozzuol (2006) interpreted its depositional environment as similar to the present Pantanal of the southwestern Brazil. These sedimentological systems are almost exclusively dominated by seasonally flooded forests, rivers, swamps, grasslands, with broadly heterogenous ecotones (Cozzuol, 2006; Hoorn et al., 2010; Latrubesse et al., 2010). Interestingly, some features in the skull of Ischyrorhynchus vanbenedeni described by Pilleri and Gihr (1979), such as relatively reduced orbits and pneumatized maxillary crests, point to turbid environment ecomorphologies that are also observed in unrelated riverine taxa, such as Platanista. This parallel highlights how the rate of morphological evolution for cranial features may be decoupled from postcranial ones, at least in the context of adaptation to riverine environments.
The authors thank M. Reguero from Museo de la Plata for the loan of the specimen MLP 76-5-7-3 from the collection under his care. We thank S. J. Godfrey and one anonymous reviewer for comments that improved this manuscript. The authors also thank V. M. F. da Silva from Instituto Nacional de Pesquisas Amazônicas and E. M. Valadares Câmara, Museu de Ciências Naturais, Pontifícia Universidade Católica de Minas Gerais, for the permission to use specimens of Inia geoffrensis in the collections under their care. Lastly, the authors thank C. W. Potter and J. G. Mead for access to collections at the Smithsonian under their care.