Phylogenomic resolution of lampreys reveals the recent evolution of an ancient vertebrate lineage

Abstract

Jawless vertebrates once dominated Palaeozoic waters, but just two lineages have persisted to the present day: lampreys and hagfishes. Living lampreys are a relatively small clade, with just over 50 species described, but knowledge of their evolutionary relationships has always been based on either a few mitochondrial genes or a small number of taxa. Biogeographers have noted the disjunct antitropical distribution of living lamprey families. Here, we present a comprehensive phylogenomic analysis of living and fossil lampreys, sampling 36 species with phylogenomic data and 46 in total with genetic data. We present new divergence time estimates based on comprehensive nuclear data and analysis of their diversification dynamics. Our analysis indicates a central role for extreme global warming during the Late Cretaceous Cenomanian–Turonian Boundary Event as a likely cause for the antitropical distribution of living lampreys, and a notable increase in lineage diversification in Northern Hemisphere lampreys during the Miocene corresponding with a period of global cooling.

1. Background

The lampreys (Petromyzontiformes) are one of two ancient surviving lineages of jawless vertebrates, with an evolutionary history stretching back hundreds of millions of years. Lampreys are celebrated for their cultural and ecological importance [1], and as one of the earliest vertebrate lineages, they reside in a key phylogenetic position that has been challenging to resolve [2]. They have diverse life history strategies, but all species spend most of their lives buried in the substrate during their larval stage, called ammocoetes. As adults many species feed on other fish as ectoparasites with their oral disc, such as the parasitic sea lamprey (Petromyzon marinus), whose invasion utterly devastated the Great Lakes ecosystem over the last century [3]. However, more than half of all lamprey species are non-parasitic, and many native lamprey species provide key ecosystem functions and are of conservation concern [4]. Despite their ecological and evolutionary significance, our understanding of lamprey phylogeny is limited, primarily informed by one to four genes [2,5–8], a small number of morphological characters [9] or limited taxonomic sampling [10]. Poor understanding of lamprey phylogeny in turn hampers our understanding of their fascinating and unique biology.

Extant lampreys have an antitropical distribution, with the depauperate families Geotriidae (2 spp.) and Mordaciidae (3 spp.) in the Southern Hemisphere, and the more speciose Petromyzontidae (47 spp.) in the Northern Hemisphere. The origin of this disjunct distribution is unclear, and conflicting hypotheses have been proposed. Unfortunately, lampreys and hagfishes both have a sparse fossil record, with the first definitive stem lamprey with a recognizably modern body plan from the late Devonian [11]. This is followed by a small number of fossils from the Carboniferous [12,13], two recently described from the Jurassic [14], and one in the Early Cretaceous [15]. Stem lampreys from the Carboniferous Mazon Creek Formation indicate that lampreys were not always absent from equatorial environments [16]. These fossil representatives are critical for understanding vertebrate evolution but provide few data points for the past distribution of lampreys.

Given their ancient origins and a classic Gondawanan and Laurasian distribution, the break-up of Pangea is an appealing and intuitive hypothesis to explain modern-day lamprey distribution [6]. This hypothesis rests on two key pieces of evidence: monophyly of Northern and Southern Hemisphere lampreys and an Early Jurassic divergence between Northern and Southern Hemisphere lineages. Morphological data have been unable to convincingly resolve lamprey relationships [2,9] and some molecular analyses have recovered the Southern Hemisphere families as paraphyletic [14], though recent genome-scale data have supported Southern Hemisphere monophyly [10].

However, vicariance via tectonic drift is not the only possible explanation for disjunct antitropical distributions. Physiological limitations to tolerating higher temperatures can exclude taxa from tropical habitats, and lampreys in their ammocoete larval form do not tolerate high temperatures [17–19]. Ammocoetes are a derived feature of lamprey evolution, and Palaeozoic lampreys inhabiting equatorial environments did not possess this developmental stage [20]. Climate oscillations may allow temperate-adapted taxa to disperse through equatorial provinces during cooler periods [21]. A recent study that did not resolve Southern Hemisphere families as monophyletic obtained a younger 78 Ma age estimate for crown Petromyzontiformes [14]. They hypothesized that lamprey antitropicality arose when lampreys dispersed out of the Southern Hemisphere during a cooler period between the extreme heat of the Cenomanian–Turonian Boundary Event (CTBE, maximum = ca 93 Ma) and the Palaeocene–Eocene Thermal Maximum (PETM, maximum = ca 56 Ma) [14].

Reconciling molecular age estimates and a well resolved phylogeny for crown Petromyzontiformes are critical for differentiating between these hypotheses. To date, divergence time estimation for lampreys has relied exclusively on just three mitochondrial genes, potentially misleading age estimates [22]. Phylogenomic methods that can capture and sequence hundreds to thousands of genes for non-model species are now commonplace [23] and can provide a more comprehensive dataset to test these hypotheses and estimate divergence times with nuclear DNA.

We present a phylogenomic analysis for extant lamprey species that resolves phylogenetic relationships and clarifies taxonomic uncertainties. By combining two morphological matrices with extant and extinct taxa and genome-scale data, we provide a total-evidence dating analysis to explore divergence times among major lamprey lineages, including testing for the hypothesis that the breakup of the supercontinent Pangea explains the antitropical distribution of extant lampreys. Finally, we explore diversification rates among extant lampreys to determine the macroevolutionary dynamics that underlie the highly disparate diversity patterns among lamprey clades.

2. Material and methods

(a). Taxonomic sampling

We obtained 81 tissue or DNA samples from 36 of the 48 currently recognized lamprey species and one undescribed species [24], plus five hagfish outgroup samples. We also incorporated Geotria autralis, Mordacia mordax and a hagfish outgroup sample from published transcriptomes [25]. To provide a complete lamprey tree, we used cytochrome b mitochondrial sequences (cytb) from GenBank to add 10 species not included in our phylogenomic dataset (see below). Specimen information and Sequence Read Archive (SRA) accession numbers are listed in electronic supplementary material, table S1; cytochrome b accession numbers are in electronic supplementary material, table S2.

(b). Exon probe design and sequencing

We designed new exon-capture probes from the sea lamprey (P. marinus) genome annotation for protein-coding genes [26]. Lampreys systematically eliminate a portion of their genome from their somatic cells, with the complete genome only found in germline cells [27]. As we were using somatic tissue from vouchered museum collections across Petromyzontiformes, searching only the somatic sea lamprey genome was more appropriate. Using HMMER3 [28], we searched a set of 1104 single-copy exons designed for actinopterygiian phylogenomics against P. marinus and extracted matching sequences [29,30]. Out of 1104 nuclear exons, 397 were identified in the sea lamprey genome. Probes 100 nt in length were designed and filtered for repetitive sequences, hybridization, and melting temperature by Arbor Biosciences (Ann Arbor, MI, USA) based on these reference sequences. Target capture and paired-end sequencing on an Illumina HiSeq 2500 were also performed by Arbor Biosciences. Probe sequences are available on Dryad.

(c). Matrix assembly and phylogenomic analysis

The resulting raw Illumina reads were quality-trimmed with Trimmomatic v. 0.33 [31]. Trimmed sequencing reads were assembled by locus using a pipeline designed to assemble ray-finned fish exon capture data [26], modified with P. marinus reference sequences. This pipeline uses aTRAM 2.0 to iteratively assemble loci [32], using the Trinity v. 2.13 assembler [33]. Published transcriptome sequencing data were processed through this pipeline for M. mordax and Geotria australis [25]. Target-captured hagfish (Myxini) outgroups were initially assembled under the same parameters but yielded few assembled loci (<10 per sample) owing to the ancient divergence between these living jawless fish lineages. However, shotgun assembly of hagfish samples with Trinity v. 2.13 [33] and searching for the subset of exon loci with nHMMER identified 68–74 exons of our set of 397 in these hagfish samples [29]. Sequences from eight ray-finned fish genomes were incorporated as additional outgroups (Anguilla anguilla, Lepisosteus oculatus, Danio rerio, Gadus morhua, Oryzias latipes, Oreochromis niloticus, Gasterosteus aculeatus and Tetraodon nigroviridis). Exons were aligned with MACSE v. 2.03 [34] and inspected by eye. These alignments were filtered for potential contamination following a strict quality control pipeline following the steps of Arcila et al. [35]. This included removing any identical sequences identified with BLAST+ found across different clades, correlating constrained gene-tree and concatenated-tree branch lengths and excluding sequences in individual genes with excessively long gene-tree branches [36], and applying the TreeShrink algorithm [37]. Alignments with fewer than 10 sequences were excluded, giving a final total of 355 exons included for analysis.

We conducted a concatenated analysis in IQ-Tree v. 2.2.0, fully partitioned by gene and codon position with the best-fitting models and partitioning scheme tested with ModelFinder [38–40], and 1000 ultrafast bootstrap replicates were conducted to assess branch support [41]. Given the long branches leading to Southern Hemisphere lampreys, there is the potential for heterotachy—rate heterogeneity through time which might make these lineages prone to long-branch attraction artefacts. Thus, the data were also analysed under the GHOST model [42], which accounts for heterotachy, with a reduced dataset of one individual per species given the increased complexity of this model. For analysis under the multispecies coalescent (MSC) in ASTRAL-III [43], we first estimated individual gene trees with IQ-Tree, partitioned by codon position and the best-fitting models and partitioning scheme identified by ModelFinder, and 1000 ultrafast bootstrap replicates. We ran ASTRAL with these gene trees as input, and a second analysis where nodes in gene trees with less than 30% bootstrap support were collapsed. Using the individual gene trees, we also estimated the gene concordance factors (gCF) and likelihood-based site concordance factors (sCFL) in IQ-Tree based on our Maximum Likelihood (ML) topology [44,45].

(d). Total evidence dating

For total evidence dating, we combined two morphological matrices containing extinct and extant cyclostomes. From a large study of vertebrate lineages containing 168 characters [2], we subsampled the variant characters for cyclostomes. We incorporated an additional 32-character lamprey-only matrix [9], with fossil codings recently added for †Mesomyzon mengae [6]. Overlapping characters between the two datasets were merged into a single character, and character states were re-coded starting from zero for consistency. Three five-state characters were re-coded to three states, as BEAST2 partitions morphological characters by the number of character states. These characters are noted in the NEXUS file, along with their descriptions. In total, our combined matrix contained 65 morphological characters. We analysed the matrix under a new technology search in TNT with an initial addition of 1000 sequences and 100 ratchet iterations [46]. Recently published Jurassic lampreys were not included in both datasets, and thus are not analysed here, but we discuss similarities and differences between our results.

A subset of 148 exons with reduced missing data and just one individual per species were selected for divergence time analysis. Entosphenus folletti was consistently found nested in Entosphenus tridentatus, and Entosphenus similis was consistently found nested in Entosphenus minimus. These species were excluded from this analysis. The best-fitting substitution model was selected for three partitions, defined by codon position. A total-evidence matrix combining DNA and morphology was analysed under the Fossilized Birth–Death Model (FBD) in BEAST2 under the Optimized Relaxed Clock Model (ORC) [47–50]. The uniform Origin Time (x₀) prior is based on the age of the Late Devonian †Priscomyzon riniensis, the oldest definitive member of Petromyzontida [11]. We set a minimum age of 359 Ma, slightly older than the fossil, and a maximum age of 636.1 Ma based on the recommendation of Benton et al. [51]. Given that the fossil record of lampreys is very ancient but also very sparse, we set an exponentially distributed prior (mean = 0.25) on the sampling probability (ρ), giving higher weight to lower values. The same logic and prior distribution were used for diversification rate (d). Turnover (r) was assigned an uninformative uniform prior. The monophyly of hagfish and lamprey lineages was enforced at the root of the tree. The Cretaceous lamprey †Mesomyzon mengae has been phylogenetically placed as both a stem and a crown lamprey [2,6,14]; we did not constrain its placement in our tree. We executed two independent runs for at least 600 000 000 generations, with an initial 10 million generation pre-burn-in and sampling every 1000 generations. Convergence was assessed via ESS values >200 of combined runs after excluding an additional 25% burn-in fraction.

We conducted an additional BEAST2 analysis to increase taxonomic sampling to include 10 species not included in our phylogenomic dataset. We incorporated cytochrome b mitochondrial sequences (cytb) from GenBank [5,7] (electronic supplementary material, table S2) and initiated a second BEAST run with node priors based on the first nuclear DNA-only phylogenomic results. Given the high proportion of missing data for cytb-only taxa, we constrained their position in the tree by genus. Each of these constraints was also given a uniform prior on divergence times based on the 95% highest posterior density (HPD) intervals of the phylogenomic-scale combined BEAST runs. This was to minimize the effect of the mitochondrial loci on the overall divergence times, except where it was the only available data.

(e). Macroevolutionary analysis of diversification rates

A sample of 10 000 trees from the posterior distribution of this run was used in downstream comparative analyses. Lineage through time (LTT) plots based on the consensus of those 10 000 posterior trees were visualized with the R packagess ape and phytools [52,53]. We simulated 1000 trees under the null expectation of a Yule Model (pure birth) and calculated the gamma statistic [54].

We explored diversification rates using BAMM v. 2.5.0 and the BAMMtools R package [55,56]. BAMM uses a reversible-jump Markov chain Monte Carlo (MCMC) to estimate rate heterogeneity among branches on a phylogeny. We used the speciation–extinction option to detect rate shifts along branches. We ran four chains for 10 million generations, sampling every 1000 generations. We generated priors using the ‘setBAMMpriors’ function in the BAMMtools package and explored the effects of priors by varying the ‘expectedNumberOfShifts’ parameter to 0.1, 1 and 10. Because we have complete taxon sampling, we could directly analyse our Maximum Clade Credibility (MCC) tree resulting from our BEAST2 analyses. Fossil taxa were pruned from the tree using the ‘drop.tip’ function in the phytools R package. We confirmed convergence of MCMC runs, acceptable effective sample sizes of the log-likelihood, and number of rate shifts using the ‘effectiveSize’ function in the R package CODA [57] and discarded 10% of the posterior sample as burn-in. We determined shift configuration with the highest posterior probability using the ‘getBestShiftConfiguation’ function in BAMMtools.

3. Results

(a). Phylogenomic resolution

We obtained a strongly supported genomic phylogeny with 75% complete taxonomic sampling, with all major nodes consistent and highly supported across both concatenation (figure 1a; figures S1-S2) and species-tree methods (figure 1b; figures S3-S4). Our tree resolves the relationships among the three extant families, with Geotriidae and Mordaciidae as sister taxa in all analyses, including high support under the GHOST model incorporating heterotachy (figure S2). Concordance factor analysis returned a gCF value of 50.64, indicating that roughly 50% of gene trees supported the monophyly of Geotriidae and Mordaciidae. Lower levels of discordant trees were found in roughly equal proportions for the alternative topologies. The sCFL value was 75.1 for this node (figure 1a).

Figure 1.

(a) Maximum likelihood phylogeny of Petromyzontiformes, with outgroups excluded from the figure owing to long branch lengths. Nodes without support values have >95% ultrafast bootstrap support. The gCF and sCFL values for the sister relationship between Geotriidae and Mordaciidae are shown at that node. Species pairs that are not reciprocolly monophyletic in the ML tree are shown as a collapsed clade with both species names. Taxon names surrounded by dashed boxes indicate species formerly considered in Eudontomyzon. (b) Multispecies coalescent species tree estimated in ASTRAL-III. Some species pairs are resolved as monophyletic species in ASTRAL and are shown as separate clades. Nodes with grey dots indicate local posterior probability (LPP) values <0.95; nodes without values have >0.95 support.

We present a novel resolution for the position of western North American lampreys in the genus Lampetra (L. hubbsi, L. pacifica, L. ayersii and L. richardsonii) as the sister to all European lampreys in the genera Lampetra and Eudontomyzon. Differences between topologies were toward the tips and limited to recent species-pairs (figure 1a). Non-parasitic Ichthyomyzon gagei samples (n = 6) are not monophyletic in our concatenated analyses, with parasitic Ichthyomyzon castaneus individuals (n = 3) nested within. However, individuals from these two species were reciprocally monophyletic in all ASTRAL-III analyses (figure 1b), suggesting a role for incomplete lineage sorting in this young species divergence. There is similar uncertainty in the relationships of individuals among other parasitic/non-parasitic species-pairs, including Entosphenus macrostoma and E. tridentatus; Ichthyomyzon fossor and Ichthyomyzon unicuspis; Ichthyomyzon bdellium and Ichthyomyzon greeleyi. Relationships among species within young lamprey genera typically had lower support, especially within Lethenteron. The only sampled individual of E. folletti was consistently nested in Entosphenus lethophagus samples, and the two individuals sampled from E. similis were consistently nested in E. minimus samples. These two species were dropped from downstream analysis in BEAST.

(b). Divergence times

The New Technology Search in TNT yielded three equally parsimonious trees with a length of 113. The strict consensus tree did not resolve relationships among several lineages, resulting in a polytomy for Mordacia, Geotria, †Priscomyzon, †Myxineidus or †Mesomyzon (electronic supplementary material, figure S5). Our Bayesian total-evidence analysis supported all fossil lampreys as stem lineages, including †Mesomyzon mengae (figure 2a). The topology for all fossil taxa obtained under a Bayesian search was identical to that of Miyashita et al. [2], but our ages are substantially younger than previous estimates (figure 3). While the total group stretches back to the Ordocivian, we estimate that crown lampreys emerged in the Cretaceous, with a mean age estimate of 93.8 Ma (figure 2b; HPD: 65.3–123.4 Ma). The split between the Southern Hemisphere families Mordaciidae and Geotriidae dates to a mean of 67.1 Ma (HPD: 43.9–90.5 Ma). The Northern Hemisphere lampreys are significantly younger, with an origin at the Oligocene–Miocene boundary, 24.3 Ma (HPD: 16.4–33.7 Ma), with most genera established in the Miocene or Pliocene. Entosphenus is the youngest genus, radiating at just 1.6 Ma in the Pleistocene.

Figure 2.

(a) Divergence times (Ma) of Petromyzontiformes and their stem relatives estimated in BEAST. Box indicates the region of the tree that is expanded in (b). Mordaciidae and Geotriidae are abbreviated as ‘Mor.’ and ‘G.’ respectively. (b) Divergence times of crown Petromyzontiformes in detail. Blue bars at nodes indicate the highest posterior density (HPD) intervals. Species names in bold indicate individuals sampled with phylogenomic nuclear data; all other samples are only represented with cytochrome b sequences.

Figure 3.

Comparison of divergence times from this study using phylogenomic-scale nuclear DNA data with three previous studies [2,6,14] that relied on mitochondrial DNA. Wu et al. [14] did not obtain Southern Hemisphere lampreys as a monophyletic group; thus no age is shown for that node.

(c). Macroevolutionary analysis of diversification rates

Our results rejected the null hypothesis of constant diversification rates through time (gamma = 5.4506, p = 0.0). Instead, we observe that living lampreys accumulated lineages slower than expected under the null model until a dramatic increase during the last 20 million years (figure 4a). Our diversification analysis indicated a single rate shift on the stem lineage of Petromyzontidae as the best rate configuration. Visualization of rate variation on the phylogeny shows diversification rates are relatively slow in Southern Hemisphere lampreys compared with Northern Hemisphere lampreys. Rates also appear to have accelerated over time in Northern Hemisphere lampreys, with the highest rates near the tips of the phylogeny (figure 4b). This pattern is confirmed by the LTT plots, which indicate accelerating rates towards the present day.

Figure 4.

(a) Lineage through time (LTT) plot for Petromyzontiformes, shown with a black line over 1000 simulated trees under the Yule model shown in grey. Global average temperature is plotted in red for the same time interval [58]. The vertical black arrows indicate the split of Petromyzontidae from Geotriidae+Mordaciidae coinciding with the extreme global warming at the Cenomanian–Turonian Boundary Event. (b) Diversification rates (lineages per million years) across the phylogeny inferred with BAMM under the preferred model with a single rate shift along the branch subtending Petromyzontidae.

4. Discussion

(a). Phylogenomics clarifies the taxonomy of living lampreys

Our results provide a well resolved backbone topology with near-complete taxonomic sampling for Petromyzontiformes. Higher-level relationships are resolved with strong support, including a sister relationship between the Southern Hemisphere families Geotriidae and Mordaciidae. A previous study based on whole-genome sequencing found high bootstrap support but low site concordance factors for this relationship [10]. However, those values were based on a parsimony calculation that can be influenced by homoplasy [45]. Using the updated likelihood calculation for this metric, we obtain a much higher site concordance factor value for our dataset.

The genus-level taxonomy of Petromyzontidae has long been contentious [59], and other recent studies have noted the need for a revision [7,10]. For the first time, we resolve the position of Pacific Lampetra as sister to all other Atlantic Lampetra+Eudontomyzon. Previous studies using only mtDNA have found this clade branching earlier in the phylogeny, rendering Lampetra non-monophyletic, and those authors have generally designated this group as ‘Lampetra’ pending further investigation [5]. Given our results, we make two simple taxonomic recommendations to reconcile lamprey genera in our phylogenomic tree: (i) reclassify Eudontomyzon morii to Lethenteron morii, and (ii) sink the remaining European Eudontomyzon species into Lampetra, uniting the genus into a young, holarctic lineage. Previous authors have also suggested Eudontomyzon as a subgenus of Lampetra [59]. These recommendations are summarized in electronic supplementary material, table S3 and are used in figure 2b. Uncertainty remains in disentangling relationships within these extremely young lamprey genera, for which our highly conserved exon set is not well suited. These groups would benefit from denser, range-wide sampling of individuals to clarify these relationships.

Detailed, range-wide genome-scale studies are needed to determine species boundaries between parasitic and non-parasitic species-pairs. These paired species with different life histories may represent ecotypes rather than true species, particularly in a young genus like Entosphenus. Paired Lampetra planeri and Lampetra fluviatilis can certainly hybridize resulting in substantial gene flow in parts of their range [60]. Our results indicate that incomplete lineage sorting may also play a role in the non-monophyly of some Ichthyomyzon species. Living most of their lives as larval ammocoetes burrowed in the substrate, lamprey biodiversity is likely under-described, and new species are still being discovered [61].

(b). Antitropical distribution arose during extreme Late Cretaceous global warming

The hypothesis that lamprey antitropicality was driven by the rifting of Pangea relies on two key pieces of evidence: monophyly of Southern Hemisphere lampreys, which our dataset strongly supports, and an Early Jurassic divergence time for crown Petromyzontiformes. Our mean age estimate for crown Petromyzontiformes is firmly in the Cretaceous at 94.3 Ma, fully 70 million years younger than prior estimates in the Jurassic [6,62] (figure 3), though older than another recent estimate that did not obtain the monophyly of Southern Hemisphere families [14]. Major differences in our age estimates can be attributed to multiple factors, including the position of †Mesomyzon on the lamprey stem, and relying exclusively on mitochondrial DNA, which has been demonstrated to overestimate divergence times owing to saturation [22]. These younger divergence time estimates reject the role of Pangean rifting to explain the distribution of lampreys, as the northern and southern landmasses were well separated in the Late Cretaceous. Given our strong support for the monophyly of Southern Hemisphere lampreys, our data do not clearly support a recent hypothesis that lampreys dispersed out of the Southern Hemisphere during cooler climate windows [14].

Instead, this divergence time for crown lampreys coincides with extreme global warming during the Cenomanian–Turonian boundary event (CTBE; figure 3), the warmest period on Earth since the end-Permian extinction [58]. After a relatively cool period, temperatures climbed rapidly beginning around 107 Ma, with global average temperatures rising by over 7°C by the apex of the Cenomanian–Turonian Boundary Event at 93 Ma, not returning to previous levels until 75 Ma [58]. The CTBE resulted in major extinctions in marine invertebrates [63], large marine reptiles including ichthyosaurs and pliosaurs [64], and freshwater fishes [65]. Average tropical temperatures would have climbed above 34°C at the peak of the CTBE [58]. This exceeds the temperatures known to be lethal to ammocoetes [17] and likely would have made equatorial latitudes uninhabitable for larval lampreys. Vicariance due to rising palaeotropical temperatures has been hypothesized as a mechanism to explain the distributions of other antitropical temperate-adapted fishes [66,67]. We propose that inhospitable temperatures in the tropics during the CTBE drove vicariance between Northern and Southern Hemisphere lampreys. It is worth noting two caveats: first, with long, old branches our HPD interval is quite large at the crown lamprey node (figure 2b); second, lamprey fossils are rare and thus far have only been found as stem lineages, so the distribution of crown lampreys during this time is not known.

(c). Diversification of extant lampreys in a cooling world

Extant lamprey lineages have thrived in the cooler periods of the Cenozoic. In the Southern Hemisphere, Geotriidae and Modaciidae split during the dramatic cooling associated with the Cretaceous–Palaeogene (K-Pg) extinction event (figure 2b), a much younger age than previous estimates in the Mesozoic era [2,6,68]. Rather than a steady accumulation of lamprey lineages over time, our results indicate that most extant lamprey diversity is due to an acceleration of the diversification of Petromyzontidae (figure 4b), the family that contains over 90% of extant species. Like a recent study with fewer loci and taxa [14], we also find a significantly young age for this clade (figure 3). Our analysis puts the origin of this clade at the Oligocene–Miocene boundary, rapidly diversifying during a period of significant global cooling (figure 3). Temperate flora and fauna expanded during this period [69]. Several other holarctic clades of fishes with high rates of diadromy evolved during the early Miocene, including Alosa (shads) [70,71], the salmonid genera Onchorynchusand Salvenius [72] and the Gasterosteidae (sticklebacks) [73]. It is possible the glacial activity during this period promoted diversification in diadromous clades [70,71]]. Alternatively, the rapid evolution of diverse life history strategies in Northern Hemisphere lampreys may have facilitated the increase in species diversity [74]. In contrast, the Southern Hemisphere families remain species-poor, and did not experience a shift in diversification during this period. While the Northern Hemisphere has large continental landmasses in temperate latitudes, the Southern Hemisphere is primarily dominated by open ocean—offering significantly less ecological opportunity for these families to diversify.

In addition to global cooling, major ocean gateways both closed and opened during the Miocene. The deepest split in Petromyzontidae separates an exclusively Atlantic-drainage clade (with one Caspian Sea endemic) separating Caspiomyzon+Petromyzon+Ichthyomyzon from all other genera, which contain both Pacific and Atlantic taxa. Caspiomyzon split from Petromyzon+Ichthyomyzon in the middle Miocene (15.5 Ma, HPD = 10.2–21.2 Ma), coinciding with the final closure of the Eastern Tethys Seaway, which separated the Atlantic from the Indian Ocean and isolated the Paratethys [75]. Subsequent reconnections in the Tortonian between the Mediterranean and the Eastern Paratethys [69] would explain the distribution of Caspiomyzon species in the Caspian Sea and the Mediterranean, the present-day Caspian Sea being a relic of the Eastern Paratethys. Primarily Pacific-drainage clades form the early branching lineages of the second major northern lamprey group, with multiple apparent dispersals back into the Atlantic. With several modern species inhabiting Arctic drainages, particularly in Lethenteron, dispersal via the Arctic Ocean seems a likely route for reaching the Atlantic.

5. Conclusions

Lampreys are a group of ancient jawless fishes that persist today, but our understanding of their evolutionary relationships and pattern of diversification has been limited owing to the lack of a comprehensive phylogeny. We estimated a well supported tree that resolves long-standing ambiguities in lamprey relationships and provides a framework for a revised generic classification of the group, sinking Eudontomyzon into the genus Lampetra. We estimated divergence dates using a total-evidence dating scheme with morphological data combined from two recent studies, the first analysis of this kind to use nuclear DNA. Our age estimates for the origin of crown Petromyzontiformes suggest a role for the extreme heat of the Cenomanian–Turonian Boundary Event causing vicariance and driving the disjunct antitropical distribution we see today. We also support a young Oligocene–Miocene boundary age for Northern Hemisphere lampreys, with a significant shift in diversification in this clade as the climate cooled in the Miocene.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

Raw sequence data can be accessed under NCBI Bioproject PRJNA1154577. Alignments, tree files, and code for exon-capture assembly have been made available on Dryad [76].

Supplementary material is available online [77].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors’ contributions

L.C.H.: conceptualization, data curation, formal analysis, investigation, methodology, visualization, writing—original draft, writing—review and editing; D.D.B.: conceptualization, data curation, formal analysis, funding acquisition, investigation, project administration, resources, visualization, writing—original draft, writing—review and editing; K.R.P.: conceptualization, resources, supervision, writing—review and editing; N.L.: conceptualization, resources, supervision, writing—review and editing; R.L.M.: conceptualization, data curation, resources, supervision, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

Funding was provided by National Science Foundation grant no. DEB-1754627 to D.D.B. We acknowledge the Willi Hennig Society for their funding of the TNT software.

Dedication

We dedicate this paper to the late Philip Cochran, whose passion for lampreys lives on among his colleagues and friends.