The nearshore cradle of early vertebrate diversification

See allHide authors and affiliations

Science  26 Oct 2018:
Vol. 362, Issue 6413, pp. 460-464
DOI: 10.1126/science.aar3689

Shallow-water diversification

Most of what we know about the relationship between diversification and environment in ancient marine environments has come from invertebrates. The influence of habitat on vertebrate diversification thus remains a persistent question. Sallan et al. studied fossil vertebrates spanning the mid-Paleozoic, including both jawed and jawless fish (see the Perspective by Pimiento). They found that diversification occurred primarily in nearshore environments, with diversified forms later colonizing deeper marine or freshwater habitats. Furthermore, more robust forms remained in the nearshore, whereas more gracile forms moved to deeper waters. This split is similar to current relationships between form and environment in aquatic habitats.

Science, this issue p. 460; see also p. 402


Ancestral vertebrate habitats are subject to controversy and obscured by limited, often contradictory paleontological data. We assembled fossil vertebrate occurrence and habitat datasets spanning the middle Paleozoic (480 million to 360 million years ago) and found that early vertebrate clades, both jawed and jawless, originated in restricted, shallow intertidal-subtidal environments. Nearshore divergences gave rise to body plans with different dispersal abilities: Robust fishes shifted shoreward, whereas gracile groups moved seaward. Fresh waters were invaded repeatedly, but movement to deeper waters was contingent upon form and short-lived until the later Devonian. Our results contrast with the onshore-offshore trends, reef-centered diversification, and mid-shelf clustering observed for benthic invertebrates. Nearshore origins for vertebrates may be linked to the demands of their mobility and may have influenced the structure of their early fossil record and diversification.

The ancestral habitat of vertebrates has long been debated, with opinions ranging from freshwater to open ocean habitats (13). Inferences have been derived from either the evolutionarily distant modern fauna or qualitative narratives based on select fossils. Early records of vertebrate divisions, such as jawed fishes and their relatives (total-group gnathostomes), consist of long gaps between inferred origination and definitive appearances (ghost lineages), punctuated by suggestive microfossils (47). Vertebrates, apart from toothlike conodont elements, were restricted in Ordovician ecosystems as trivial components of the Great Ordovician Biodiversification Event (4, 5, 7). Ancestral habitat is a critical factor in determining both pattern and mode of diversification, potential mismatches between biodiversity and available habitat area, and the source of apparent relationships with changing sea level (6). A lack of early vertebrate fossil data and habitat information in compendia has limited quantitative approaches (4), preventing resolution of this outstanding issue in vertebrate evolution.

We developed a database of total-group gnathostome occurrences (~480 million to 360 million years ago) (4, 5, 8) during their mid-Paleozoic diversification (n = 2827) (9) (fig. S1). Data collection focused on all occurrences from the interval encompassing the five oldest localities for each major clade (n = 188) (Fig. 1 and figs. S1 and S2) and phylogenetically constrained genera within jawless groups (n = 785) (Figs. 2 and 3 and figs. S1, S3, and S4) for use with Bayesian ancestral state reconstruction. We used environmental, lithological, and invertebrate community information from the literature and available databases to assign occurrences to benthic assemblage zones (10) (Fig. 1). Benthic assemblage zones are categorized and ordered as fresh water (BA0); intertidal above typical wave base (BA1); shallow subtidal and/or lagoon (BA2); deeper subtidal, including the start of tabulate coral–stromatoporoid reef systems (BA3); middle to outer shelf (BA4 and BA5); and shelf margin toward the bathyal region (BA6). These zones have been widely used in studies of mid-Paleozoic paleocommunities (1, 1012) (Fig. 1).

Fig. 1 Mid-Paleozoic vertebrates preferentially originated in shallow marine habitats.

(A) Intertidal (BA1) to subtidal (BA2 and BA3) ancestral habitats for total-group gnathostome clades (n = 188), assuming placoderm paraphyly and Silurian first occurrence for chondrichthyans. Full results are shown in figs. S2 to S5. Wen., Wenlock; Lud., Ludlow; Pri., Pridoli. (B) Silurian and Lochkovian marine distributions for PBDB fossiliferous strata (n = 858), richness (n = 6980), and occurrences (n = 30,004); conodont richness (n = 505) and occurrences (n = 7447); paleocommunities (n = 2401); and gnathostome occurrences (n = 1035) show mid-Paleozoic records peaking on the mid-shelf (BA3 and BA4) with few records in marginal marine settings, in contrast to the shallow-water preferences of early gnathostomes. (C) Early and overall occurrences for total-group gnathostomes (n = 2827), jawed fishes (n = 1343), and jawless fishes (n = 1484) show that early occurrences were significantly more concentrated in shallow marine settings (BA1 and BA2) than overall or later occurrences. See data S2 and figs. S1 and S5 to S9.

Fig. 2 Macromeric, robust jawless fishes exhibit shallower-water diversification and greater habitat restriction.

Ancestral states for (A) heterostracans and Ordovician (Ord.) stem-gnathostomes (n = 316), (B) galeaspids (n = 112), and (C) osteostracans (n = 158) show that macromeric genera preferentially originated in very shallow waters (BA0 to BA2), with the exception of more streamlined forms. Full results are shown in figs S6 to S8. (D) Early and overall habitat distributions for macromeric clades (n = 1123), showing significant shifts toward shallower water subsequent to their origination. Full distributions are shown in figs. S19 and S20 and data S1.

Fig. 3 Micromeric, gracile jawless fishes exhibit deeper-subtidal later diversification and easier dispersal.

Ancestral states for (A) thelodonts (n = 99) and (B) anaspids (n = 100), showing diversification of genera in deeper subtidal waters (BA3) during their evolutionary history. Full results are shown in figs. S9 and S10. (C) Early and overall occurrences for micromeric jawless fishes (n = 353) show a rapid shift to deeper waters following nearshore origination. Full distributions are shown in figs. S21 and S22 and data S1.

We applied Bayesian threshold models to phylogenies of occurrences using prior probabilities of residence in each benthic assemblage zone. This methodology allowed positive inference of both ancestral habitats and amount of evolutionary change required to move between zones (“liability” values) (13). All major clades, from the first skeletonizing jawless fishes (astraspids, arandaspids) to jawed bony fishes (osteichthyans), originated within nearshore intertidal and subtidal zones (~BA1 to BA3), centered on BA2, over a period of more than 100 million years (Fig. 1A and fig. S3). This area is relatively shallow, includes lagoons in reefal systems, and is located entirely above the storm wave base in the mid-Paleozoic (11) (Fig. 1).

We appraised whether nearshore origination in gnathostomes resulted from environmental bias in the record through comparison with habitat distributions for other facets of the mid-Paleozoic captured in independent datasets, including fossiliferous strata, regional paleocommunities, and global occurrences and richness (number of genera) (Fig. 1B and figs. S11 to S16) (10, 14). Analysis of mid-Paleozoic strata in the Paleobiology Database (PBDB) (14), binned by distinct habitat categories (n = 4437), produced a distribution clustered on deep subtidal and reef environments [equivalent to BA3 and BA4 (10)], with many fewer records in freshwater-marginal marine (BA0 and BA1) and the basin and slope (~BA5 and BA6) zones (Fig. 1B and figs. S11 and S12). PBDB records of occurrences (n = 111,364) or genera (n = 24,211) provide distributions that show even greater clustering on the mid-shelf but are highly correlated with sampled strata (linear regression: r2 = 0.96, P = 0.0004 and r2 = 0.94, P = 0.0008, respectively) (fig. S12). Silurian and Lochkovian regional paleocommunities (10) are also centered on BA3 and BA4 (Fig. 1B and fig. S13). These records suggest a global, mid-shelf center for sampling and diversity and a null expectation of originations in deep subtidal and reef environments [more so than expected from previous studies focused on reef-bearing facies (15)]. This is in stark contrast with shallower gnathostome ancestral habitats (Fig. 1) and is thus unlikely to result from global sampling bias.

To evaluate whether apparent nearshore origination resulted from preservational biases in different habitats, we compared gnathostome distributions to PBDB records for conodonts. Conodonts are the sister group of extant jawless cyclostomes or the vertebrate total-group, largely known from phosphatic oral elements (4), which serve as an independent preservational proxy. Conodonts are stratigraphic index fossils and common along the marine depth gradient during the mid-Paleozoic (Fig. 1B and fig. S14). Conodont occurrences (n = 11,915) show a different distribution from other PBDB records (chi-squared P < 0.0001), exhibiting a peak in BA2 and more occurrences in BA5 and BA6 (figs. S14 and S15). Conodont richness (n = 1308) is more clustered around BA3 and BA4, particularly in the Silurian-Lochkovian (n = 505) (Fig. 1B and figs. S14 and S15). This pattern argues against early gnathostome restriction resulting from preservational bias, as does the plurality of vertebrate occurrences in deeper waters from the early Silurian (Fig. 1C and fig. S1).

Jawed and jawless fish distributions are highly clustered in BA0 to BA2 early in clade history (n = 478), in the Silurian and Lochkovian (n = 1035), and over the mid-Paleozoic (n = 2147) (Fig. 1 and figs. S1 and S16 to S18). We recover no significant or strong positive correlations between this gnathostome pattern and other fossil records (linear regression r2 range: −0.90 to 0.27, P range: 0.41 to 0.9) (Fig. 1B and fig. S16).

Ancestral states show that gnathostomes originated preferentially near shore, even as the diversity of species and body forms increased (Fig. 1A and fig. S2). Early occurrences are significantly different from later records within groups (chi-squared P = < 0.00001) (Fig. 1C and fig. S18); gnathostomes as a whole, as well as jawed and jawless fishes specifically, exhibit greater clustering in shallow marine settings (BA1 and BA2), independent of exact time of first appearance in the mid-Paleozoic (Fig. 1C and fig. S18). Shallow ancestral habitats are always supported by our analyses despite variation in first appearances of jawed fishes (e.g., inclusion of potential Ordovician “chondrichthyan” material) (15); placoderm monophyly or paraphyly (8); and even increasing the minimum prior probability of occurrences in all zones to a minimum of 5 or 10% to account for potential of false absence, missing records, or other sampling issues (Fig. 1A, figs. S2 to S5, and table S1). Gnathostomes continued to show a strong tendency to diverge in shallow marine waters long after the invasion of deeper and fresh waters by older lineages, including after the origin of jaws.

Threshold liability values suggest that shifts within the nearshore waters required little evolutionary change and were common, as was invasion of fresh water (Table 1 and Fig. 1C). Dispersal into deeper waters, including the forereef, shelf, and open ocean (BA4 to BA6), was more restricted (Table 1), complicated by a short-term tendency to return to the ancestral shallows [Ornstein-Uhlenbeck deviance information criterion (OU DIC) weight = 1; phylogenetic half-life in Table 1] (16). Yet, threshold values also suggest rapid dispersal across the offshore shelf (BA4 and BA5) once lineages managed to depart BA3, even though shifts into open waters (BA6) had much higher requirements (Table 1). However, if sampling probabilities in all bins is increased a priori, shallow-water restriction of early gnathostomes is explained by ever-higher thresholds for continued movement offshore, starting at BA2 (Fig. 1A, figs. S2 to S5, and table S1).

Table 1 Best-fit model parameters for ancestral habitats.

ancThresh (Ancestral Character Estimation Under The Threshold Model Using Bayesian MCMC) (13) holds the threshold for exiting BA0 constant at 0 and the value for exiting BA6 as infinity (Inf). Values for parameters (log-likelihood and alpha) are means after excluding “burn in.” See figs. S2 and S10 and (29) for ancestral states. Phylogenetic half-lives are given in million years. mil. gen., million generations.

View this table:

Next, we determined the association between body form and dispersal ability within major groups. Clades were categorized into two body forms: (i) macromeric, which are mostly robust and armored with large bony plates (e.g., heterostracans, osteostracans, and galeaspids) (17) (Fig. 2), or (ii) micromeric, which are mostly gracile and either naked or covered in small scales (e.g., thelodonts and anaspids) (17) (Fig. 3). These robust or gracile forms can be approximated as having benthic or pelagic and nektonic lifestyles, respectively, given their gross similarity to living fishes (18, 19).

Analysis of all gnathostome early occurrences shows that both micromeric and macromeric forms originated around shallow water (BA2) (Fig. 1A and fig. S2). However, group-level analyses suggest that slight shifts shoreward or seaward preceded the later diversification of these groups. Genus-level diversification of macromeric jawless lineages was centered in the shallows (BA1 and BA2) and fresh water (BA0) throughout their multimillion-year existence (Fig. 2 and figs. S6 to S8, S19, and S20). Later occurrences were significantly more clustered in shallow and freshwater settings than the earliest members of these clades (chi-squared P < 0.0001) (Fig. 2C and figs. S19 and S20). Threshold values indicate that moving into deeper waters was more difficult for robust groups than gnathostomes as a whole (Table 1 and tables S1 and S2), and these featured a strong tendency to return to the shallows (OU DIC weight range = 0.99 to 1; phylogenetic half-life in Table 1).

The diversification of micromeric gnathostomes was centered in deeper subtidal waters (BA3) following their origination in BA2 (Figs. 1A and 3 and figs. S9, S10, S21, and S22). Early occurrences of these clades show a significantly greater concentration in BA1 and BA2 than later forms (chi-squared P < 0.0001) (fig. S21 and S22). A handful of early Silurian thelodont taxa were already resident in deeper waters (BA3 to BA5) following their Late Ordovician appearances in BA1 and BA2 (fig. S21A). Early dispersal into deeper waters reflects low threshold parameters (Table 1) and may be a general pattern for gracile clades. Jawed fishes show a significant shift onto reefs and deeper settings in the later Devonian (chi-squared P < 0.0001) (Fig. 1C and figs. S1 and S18), after the appearance of most subclades. Robust jawless groups contain exceptions that may prove this rule: A few subclades with fusiform bodies (e.g., tremataspid osteostracans) originated in BA3 and register deeper-water occurrences than their relatives by the mid-Silurian Fig. 2 and figs. S6 to S8).

Dispersal in multiple directions appears to have been enabled by body-form evolution and did not precede the origin of new phenotypes in new habitats. These shifts affected subsequent survival. Freshwater habitats were marked by the persistence of robust clades such as osteostracans and gracile forms such as anaspids, without further changes to their gross body plan (Figs. 2 and 3). Sometimes, identical deep-water lineages appear short-lived and did not exhibit apparent further diversification, even on reefs (Fig. 1) (20). Jawless gnathostomes show a significant shift in distribution (chi-squared P < 0.00001) back into the ancestral nearshore habitats and adjacent estuarine areas following a peak in distribution across the depth gradient in the Silurian to Early Devonian (Fig. 1C and figs. S1 and S18). This shift occurred just as jawed fishes moved out of nearshore habitats in the Devonian (Fig. 1A and fig. S18) (4, 21). This pattern is reflected in the greater representation of benthic forms in later marine jawless fishes versus nektonic forms in jawed vertebrates (22).

Overall, our results show that the nearshore served as the cradle of early vertebrate taxonomic and gross morphological diversification (Figs. 1 to 3). Specific body forms evolved in coastal waters subsequently favoring expansion into shallower (e.g., macromeric jawless fishes) or deeper areas (e.g., micromeric jawless fishes, jawed fishes). This mirrors observations within living fishes of repeated splits into benthic or pelagic and nektonic forms (18, 23), as well as the gross division of fish phenotype-environment associations (19).

A persistent diversification center within the shallows may explain features of the early vertebrate record (7, 24). Ordovician gnathostomes are primarily represented by microfossils restricted to a small subset of nearshore facies (fig. S1) subject to wave action (11), despite worldwide distributions (4, 7, 17, 24). Ghost lineages for gnathostomes might be caused by environmental endemicity, low abundance, and/or a relative lack of marginal marine strata (figs. S1 and S11 to S13). Alternatively, a relationship between Ordovician diversity and sea level (6) might have a common cause in changing shallow habitat area; reduction in such environments would have delayed apparent diversification and increased extinction risk (6, 25, 26).

Endemicity in coastal waters may have later promoted origination of new clades. Biogeographic patterns suggest that body-form divergence occurred in multiple shallow settings, increasing overall diversity. Micromeric forms occur alongside macromeric astraspids in the Ordovician of Laurentia, whereas robust galeaspids existed alongside gracile chondrichthyans in the early Silurian of Gondwana (47, 15, 17, 24, 27, 28). Nektonic body plans developed in these hotspots enabled dispersal across deep early Silurian oceans, away from local competition, leading to further diversification in nearshore settings elsewhere (1, 15, 28). In contrast, benthic groups showed structured geographic patterns (27), moving along coastlines and inshore, perhaps toward nutrient inputs essential to their likely bottom-feeding and filtering lifestyles and away from increased competition. Thus, continuous origination in shallow waters shaped the evolution of vertebrates, at least during their first phase of diversification.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S23

Tables S1 to S3

References (3061)

Data S1 to S3

References and Notes

  1. Materials and methods are available as supplementary materials.
  2. Data downloaded from the Paleobiology Database ( 6 to 12 June 2018.
Acknowledgments: We thank Z. Min, for assistance with galeaspid occurrences; L. Revell, G. Lloyd, and J. Mitchell for advice on phylogenetic methods; S. Wang for assistance with statistical tests; N. Tamura for providing the reconstructions used in Figs. 1 to 3; and D. Fraser and three anonymous reviewers for comments. Funding: This work was supported by the University of Pennsylvania (L.S.), a Palaeontological Association Undergraduate Research Bursary (C.M.B.), and the University of Birmingham (I.J.S). Author contributions: L.S. and I.J.S. designed the study, assembled the figures, interpreted results, and drafted the manuscript. L.S., R.S.S., C.M.B., and I.J.S. contributed data. L.S. performed analyses. L.S., M.F., R.S.S., C.M.B., and I.J.S. participated in designing analyses, discussing results, and editing the manuscript. Competing interests: None declared. Data and materials availability: All data are available in the supplementary materials and at (29).

Stay Connected to Science

Navigate This Article