Detecting hybridization between sister species of Terebratulina (Brachiopoda, Cancellothyridoidea) in the North Atlantic: morphology versus molecules

Investigating samples of the cancellothyridid brachiopod Terebratulina collected during the IceAGE (Me85/3) expedition of RV METEOR at the continental shelf around Iceland with both morphometrical and molecular methods, we were for the first time able to detect a hybridization event between brachiopod sister species, which are thought to have separated 60 MYA. Terebratulina retusa and T. septentrionalis can clearly be distinguished on the basis of consistent species-specific molecular signatures in both mitochondrial and nuclear markers, whereas morphometrical analyses proved to be less reliable for species determination than previously thought. Two out of 28 specimens were identified as offspring of a one-way hybridization event between T. retusa eggs and T. septentrionalis sperm. Whereas the fossil record of Terebratulina in the North Atlantic region is too fragmentary to reconstruct the history of the hybridization event, the different life history traits of the two species and current oceanographic conditions around Iceland offer plausible explanations for the occurrence of crossbreeds in this common brachiopod genus.

made specific identification and description impossible" (ref. 15: p. 9). This potential cline linking the two species could either be interpreted as the morphospace of a single Atlantic species of Terebratulina with its two extremes erroneously described as different species (see refs 18-20) or as the product of hybridization 15 . Curry & Endo 16 , unable to find such cline in their data, concluded that hybrids do not exist and that their morphometric approach clearly discriminated between two Atlantic species of Terebratulina. This was also confirmed by allozyme data and mitochondrial RFLP analysis 21 and by sequence analyses of mitochondrial DNA 7 .
The geographical distribution of both species as reconstructed by Curry & Endo 16 provided another interesting aspect: whereas T. retusa was restricted to the East Atlantic and the Mediterranean in their analysis, T. septentrionalis seemed to be more widespread from North America to Greenland, Iceland and the Norwegian coast with a possible ice age relict population in Finnmarken (North Norway). Despite partial sympatry no morphological overlap existed in the data of Curry & Endo 16 between T. retusa and T. septentrionalis supporting the hypothesis of two valid species of Terebratulina inhabiting the North Atlantic. However, the intermediates observed by Wesenberg-Lund 13-15 remained elusive and her dubious specimens have never been subjected to a rigorous analysis.
In 2011 the IceAGE project collected marine benthos around Iceland with the German research vessel METEOR. Among the samples were 28 specimens of Terebratulina suitably preserved for molecular analysis from the geographical region where the unidentifiable specimens of Wesenberg-Lund 15 had been collected. Sitting half way between mainland Europe and Greenland on top of the mid-Atlantic ridge, Iceland is crucial to understand the biogeographical distribution of T. septentrionalis because it represents the most prominent contact zone with its East Atlantic sister taxon T. retusa. The IceAGE material enables us to show not only that T. retusa in Iceland is much more variable in shell ornamentation than Curry & Endo 16 suggested, but also that Wesenberg-Lund legitimately struggled to identify her confusing specimens. In particular, we found molecular evidence for hybridization between T. retusa and T. septentrionalis.

Results
Morphology. As in the study of Curry & Endo 16 , length, width and dorso-ventral height of the Terebratulina shells as well as ratios between the three size measurements did not discriminate the species because these data formed a gradient from small to large specimens simply reflecting changing size with age (not shown). However, the average rib width over a defined transect (see below and 16 ) showed a discontinuity in the resulting plot ( Fig. 1A) at about 0.35 mm seemingly reflecting the species boundary between T. septentrionalis and T. retusa. When displayed as a box-and-whisker plot (Fig. 1B) it was even more obvious that the 40 included specimens (27 IceAGE,13 MfN brachiopod collection) fell into two significantly different groups, thereby corroborating previous results. The rib width mean value of 0.298 mm representing the group with narrower ribs (=the putative T. septentrionalis specimens) was almost identical with the mean value given in Curry & Endo's study for this species. However, the rib width mean value of the group with broader ribs (=the putative T. retusa specimens) was conspicuously smaller than that given by these authors. Nevertheless, the rib width of all our putative T. retusa specimens identified by morphology only fell into the variability range of the specimens assigned to T. retusa by Curry & Endo, i.e. the distribution of our specimens within this variability range was slightly shifted towards smaller rib widths.

Molecular analysis.
All Terebratulina specimens collected during the IceAGE expedition yielded suitable amounts of DNA for sequence analysis. The analysed mitochondrial markers (n = 26) resulted in 6.07% (12 S) and 8.10% (16 S) sequence divergence between two groups of specimens clearly defining them as separate taxa. As we were interested in testing the IceAGE specimens for signs of hybridization, we needed an additional nuclear marker (28 S rRNA) in which potential recombinations may have happened in the past leading to conserved discordancies between the mitochondrial and nuclear genotypes. Thus, we analyzed a 1020 bp long fragment of this marker (n = 25) showing a much lower, but still measurable sequence divergence of 2.25% between the same two groups of specimens. These results confirm that T. septentrionalis and T. retusa both have species-specific nucleotide signatures, which are intra-specifically conservative and allow for species discrimination on the basis of molecular markers alone.
The critical question now was whether all molecular identifications agree with the clear morphological differentiation on the basis of rib width measurements or whether single specimens showed a combination of sequence identity with one species and morphological affinity to the other species. A second aspect was to look at the combination of both mitochondrial and nuclear markers to check for potential hybrids with discordant species-specific signatures. The sequence alignments of all three markers are given as additional files [see Supplementary alignments S1-S3].
Morphological and molecular data combined. When combining all results from morphological and molecular analyses (Table 1) it became clear that four different combinations of characters exist among the 25 Terebratulina specimens from Iceland successfully sequenced for all three markers in this study. The majority of specimens clearly belonged to either T. septentrionalis (first cluster) or T. retusa (second cluster) being consistent in all molecular and morphological characters for one or the other species. These were four T. retusa (one from station 1034, three from station 1047) and eleven T. septentrionalis (one from station 1034, four from station 1047, and six from station 1213). The third cluster comprised 8 specimens (two from station 1034, six from station 1047) which in both mitochondrial and nuclear markers were clearly T. retusa, but which showed very narrow shell ribs shifting them into the T. septentrionalis morphospace. These specimens, on the basis of morphological characters alone, would certainly have been misidentified in a morphometrical analysis. The fourth cluster is the most interesting one, represented by only two specimens. The rib width of these specimens would clearly identify them as T. retusa, as do the mitochondrial markers. But the nuclear marker clearly has the T. septentrionalis The same data set depicted as box-and-whisker plots showing a significant difference between two groups (A: mean rib width 0.416 mm ± 0.078 SD, and B: 0.289 mm ± 0.037 SD; Mann-Whitney U-test: p ≤ 0.0001), which according to Curry & Endo 16 should represent the two Atlantic species T. retusa (group A) and T. septentrionalis (group B). (C) Rib widths of T. retusa (0.345 mm ± 0.05 SD, n = 15) and T. septentrionalis (0,282 mm ± 0.038 SD, n = 10) are still significantly different (Mann-Whitney U-test: p = 0.003), when species are identified based on mitochondrial sequence data (12 S and 16 S rRNA), but the difference is less obvious than in (B). Same as in (C), but species identification based on nuclear sequence data (28 S rRNA) leading to non-significant differences (Mann-Whitney U-test: p = 0.154) between rib-widths of T. retusa (0.335 mm ± 0.049 SD, n = 12) and T. septentrionalis (0.298 mm ± 0.059 SD, n = 12).
signature. This mixture of the otherwise highly conserved and species-specific molecular sequences in both mitochondrial and nuclear markers can only be explained by interspecific hybridization. Both specimens with this character mix were homozygous for T. septentrionalis in their 28 S rRNA, i.e. they almost certainly do not represent the F1 of a recent hybridization event, but are the result of backcrossings of a hybrid with a parental species, in this case with T. septentrionalis. As mitochondrial genes are inherited from the female parent we can even conclude that the original hybridization event happened when a T. retusa egg was fertilized by a T. septentrionalis sperm. The mitochondrial T. retusa signature was then passed on to following generations through the female line.
Consequently, molecular data enable us to divide the samples into bona fide T. septentrionalis versus T. retusa, which partly contrasts with the previous grouping of specimens "with narrow shell ribs" versus those "with broad shell ribs". Once the species are identified on the basis of their molecular sequence markers, the resulting box-and-whisker plots of rib widths show that the clear difference between the two morphologically defined clusters becomes much smaller and when using the nuclear sequences for species identification this significant morphological difference even collapses (Fig. 1C,D). This means that the difference between narrow-ribbed and broad-ribbed Terebratulina specimens from Iceland is arbitrary and does not reflect a significant difference between T. septentrionalis and T. retusa. T. retusa around Iceland seems to be more variable in this shell character than elsewhere, i.e. some individuals have narrower ribs than the typical T. retusa from North Atlantic coasts of mainland Europe.

Discussion
Species identification and the hybridization event. As has been shown in a previous study 7 , the molecular markers used here are highly conserved intraspecifically, i.e. identification of either T. retusa or T. septentrionalis is possible based on sequence information alone. Even seemingly small genetic differences (e.g. 2.25% as in the 28 S rRNA fragment) are sufficient to tell the species apart. The occurrence of sequence information of both species in one individual can be explained if a mixture of these otherwise highly conserved sequences happened in the past. In nature this can be achieved either vertically through hybridization or horizontally through lateral gene transfer. Analyses of full mitochondrial genomes of brachiopods showed no signs of lateral gene transfer [22][23][24][25] , so that our observations can only be explained by hybridization between the two species. The role of temperature preferences and taxon sampling. The failure to discover the hybrid zone of T. retusa and T. septentrionalis on the southwestern shelf off Iceland in the otherwise meticulous and well-structured study of Curry & Endo 16 is simply due to bad luck as their impressive number of specimens analysed supposedly did not contain enough specimens from this critical region. However, their Iceland specimens coded as "b" in their principal component analysis ( Fig. 1 in ref. 16) were all identified as T. retusa and cluster conspicuously close to the border between the two clouds identified in their data set. This may be interpreted as a hint that T. retusa in Iceland has no "typical" morphology, something we could clearly show for at least some of the specimens collected at Meteor stations #1034 and #1047 in the southwest of Iceland. But why is this region so critical? When looking at the temperature regimes around Iceland it is obvious that the southwestern region is influenced by the warm North Atlantic Current leading to temperatures of 7-10 °C throughout the year. In contrast, the water in the northeast of Iceland is much colder due to the East Icelandic Current carrying polar waters southward, leading to annual temperatures of about 5-7 °C (temperature measurements for both regions at 50 m depth by the Icelandic Marine Research Institute in 2011 [26][27][28] ). According to Curry & Endo 16 T. retusa prefers temperate water conditions (see also ref. 29), whereas T. septentrionalis prefers colder climate (but see ref. 30), the latter being in line with our results that station #1213 revealed only T. septentrionalis specimens. Potential hybridization can only occur at water temperatures, which are suitable for both species and this seems to exclude the northeastern region off Iceland. According to the reconstruction of water currents around the island 31 current mediated transport of T. septentrionalis sperm along the clockwise running Iceland coastal current towards the southwestern region may be possible, but seems unlikely due to generally low sperm survival rates in open waters 32 and the low salinity and variable flow velocity of this coastal current 27 . Rather, our results show that adults of T. septentrionalis were found at Meteor stations #1034 and #1047 in the southwestern region, albeit in low frequency, offering the opportunity of crossbreeding through sympatric distribution of the two species.
The potential influence of life history traits on hybridization. Apart from their specific temperature preferences (see above), T. retusa and T. septentrionalis also differ in their reproductive biology. T. septentrionalis is a brooder, retaining its embryos within the mantle cavity until they have reached an advanced stage of development 33,34 (see also ref. 35 for the related Pacific species T. unguicula). In contrast, T. retusa is a free spawner with both sexes shedding their gametes into the surrounding water. Thus, for a T. retusa egg the probability is rather high to be hit by a T. septentrionalis sperm, especially if in sympatric populations in Europe the temperature dependent reproductive season of both species is isochronic. On the other hand, larval brooding in the mantle cavity as in T. septentrionalis is only possible if (i) sperm is washed into this cavity by the adult's inhalant feeding current and (ii) spawned eggs ready to be inseminated are retained in the same place. As has been shown in several studies, the inhalant current enters the mantle cavity from left and right sides of the articulate brachiopod shell [36][37][38] , leading the water through the network of tentacles of both lophophoral arms to filter planktonic particles. Only the filtered and clean water passes through the tentacle network into the mantle cavity and leaves the animal as the exhalant current at the mediofrontal margin of the shell. The brooding female of T. septentrionalis must have a sorting mechanism to differentiate between sperm and food as the sperm has to pass the lophophoral tentacles to enter the mantle cavity for inseminating the ripe eggs. This sorting may be accomplished by size selection of the captured particles, as the maximum efficiency for particle capture in T. retusa applies to food particles sized 7-8 µm 39 , whereas sperm diameter in Terebratulina does not exceed 1.5 µm 40 . If in addition this sorting mechanism is able to differentiate between conspecific and other gametes, T. retusa sperm randomly entering the inhalant current of a T. septentrionalis female may be doomed. This could be an explanation for finding only descendents of a hybridization event between a T. retusa egg and a T. septentrionalis sperm in our data. One-way hydridization as assumed here on the basis of different life history traits may be characteristic for this species pair. However, as we have found only two hybrids among our samples, this prediction needs future verification.
Does the fossil record help?. The oldest fossil brachiopods attributed to the genus Terebratulina date back to the Late Jurassic of Europe, California, and New Zealand 41, 42 basically representing terebratulids with shell ornament. Fossil specimens similar to or identified as T. retusa have been reported only from Upper Oligocene to Pleistocene strata in France, Italy, Hungary, Rhodes, and Algeria 16, 43-45 . As described above, extant specimens of T. retusa and T. septentrionalis are difficult to tell apart on the basis of shell morphology alone, i.e. distinguishing these species as fossils may be even more problematic. This might be the reason for T. septentrionalis being absent altogether from the records in the Paleobiology Database 46 . With regard to the unusually high longevity of 60 MY of the two species according to a previous molecular clock approach 7 , the fossil record is, therefore, not suitable to reconstruct the speciation history and palaeogeographic distribution of Terebratulina species in the North Atlantic.

Conclusions
Our results provide the first evidence of hybridization in brachiopods, i.e. between the species T. retusa and T. septentrionalis. We demonstrated that T. retusa in Iceland is much more variable in its shell ornament than previously thought, blurring the clear morphological disparity between the two species assumed by Curry & Endo 16 . The intermediate specimens of Wesenberg-Lund 13-15 although triggering the search for true hybrids do not necessarily represent descendants of a hybridization event between the two species of Terebratulina, but may just reflect the morphological variability of T. retusa in Iceland. The scarcity of fossil representatives of the two species and the demonstrated difficulties to identify T. retusa and T. septentrionalis beyond doubt on the basis of morphometrical characters alone hampers the reconstruction of the speciation event leading to the extant Terebratulina species in the North Atlantic. Whereas the genus Terebratulina based on the oldest known fossils is of Tethyan origin, its modern representatives in the North Atlantic almost certainly came into being through a vicariance event caused by the opening of the Atlantic Ocean. However, the origin of the Norwegian Finnmarken specimens described as T. septentrionalis (see Fig. 3 in ref. 16) remains elusive until molecular data for this isolated population are available.

Methods
The  Fig. S2). For each shell, number of ribs and rib width within the 4 mm transect were measured ten times and mean/standard deviation was calculated. To enlarge the morphometrical dataset we additionally measured 16 Terebratulina specimens from the brachiopod dry collection of the Museum für Naturkunde (acronym: ZMB) covering the biogeographical distribution of both, T. retusa and T. septentrionalis (Supplementary Table S2).

Molecular analysis.
Tissue (lophophore, gonad or musculature) of all 28 Iceland specimens was dried and dissolved in a CTAB mastermix (0.5% 2-mercaptoethanol and 3% proteinase K in CTAB buffer) to extract the mitochondrial and nuclear DNA. Proteins were precipitated with chloroform/isoamyl alcohol, and nucleic acids were precipitated in EtOH with sodium acetate, dried and re-dissolved in 0.1x TE buffer. Mitochondrial sequences were amplified with PCR (GenAmp ® PCR system 2700) using primers 12SF1091, 12SR1478, 16SF2510 and 16SR3080 47,48 , which after sequencing and editing yielded 357 bp fragments of the 12 S rRNA and 429 bp fragments of the 16 S rRNA (n = 26, two samples did not amplify), respectively. All mitochondrial sequences were aligned against the full mt-genome of T. retusa (Genbank acc. no. NC_000941.1) 22 . Additionally, nuclear sequences were obtained with specific primers 28SF680, 28SF700, 28SF1062, 28SR1460 and 28SR1797 (see refs 49 and 50 and B.L. Cohen pers. comm.) resulting in an edited alignment of 920 bp (n = 25, with three samples not amplifying) against the 28 S sequence of T. retusa published in ref. 50. PCR products were purified using a Nucleospin Kit (Macherey-Nagel, Düren) and commercially sequenced at Services in Molecular Biology GmbH, Rüdersdorf. Sequencing results were edited with BioEdit sequence alignment editor v7.0.0. Sequence alignments of both mitochondrial and nuclear markers were compared between specimens based on single bp comparison across their entire length. This method yielded robust data for species identification even when comparing Terebratulina populations across large spatial scales 7 . For PCR primers used see Supplementary Table S3. All sequences obtained were submitted to NCBI and can be identified by their respective Genbank accession numbers according to Supplementary Table S4.

Statistics.
To test for significant differences between morphometric measurements (length, width, thickness and rib width) we used the nonparametric Mann-Whitney U-test (threshold: 5% with p ≤ 0.05) which is appropriate to compare differences between two independent groups when the dependent variable is not normally distributed. Calculations were done with XLSTAT, ver. 2013.5.05 (Addinsoft 1995-2013).