Origin of the Eocene planktonic foraminifer Hantkenina by gradual evolution

Hantkenina is a distinctive planktonic foraminiferal genus characterized by the presence of tubulospines (robust hollow projections) on each adult chamber, from Middle and Upper Eocene marine sediments worldwide. Here we illustrate its evolutionary origin using c. 150 specimens from 30 stratigraphic intervals in two sediment cores from Tanzania. The specimens, which span an estimated time interval of 300 ka, show four intermediate steps in the evolution of the tubulospines that amount to a complete intergradation from Clavigerinella caucasica, which does not possess them, to Hantkenina mexicana, which does. Stable isotope analyses indicate that the transitional forms evolved in a deep planktonic habitat not occupied at that time by other species of planktonic foraminifera. We discuss the morphogenetic constraints involved in the evolutionary transition and propose an ecological/adaptive model for the selective pressures that resulted in the evolution of tubulospines. We compare our record with similar, recently described assemblages from Austria and Italy, and we update the biostratigraphy and systematic taxonomy of the key morphospecies involved in the transition.

T H E mineralizing plankton of the oceans have provided some of the best examples of evolutionary transition in stratigraphically ordered sequences because they are abundant, broadly distributed and can be found in continuously deposited deep-sea sediments (Norris 2000;Lazarus 2011). Most studies published to date demonstrate relatively simple changes in size or shape, for example (in planktonic foraminifera) the degree of test compression, the number of chambers per whorl or the development or loss of peripheral thickenings (e.g. Malmgren and Kennett 1981;Malmgren et al. 1983;Norris et al. 1996;Kucera and Malmgren 1998;Hull and Norris 2009). Here we examine the evolutionary origin of the fossil planktonic foraminifer Hantkenina, focusing on the record from two sediment cores obtained by us from Tanzania and in the light of other recently published records from Austria Egger 2010, 2011) and Italy (Coccioni and Bancal a 2012). The example is interesting in that it encompasses the evolution of a relatively complex structure known as a tubulospine de novo that demands an adaptive explanation.
Planktonic foraminifera are heterotrophic protists that produce a test ('shell') of calcite that is usually composed of successive chambers arranged in a spiral form with interconnecting foramina. The genus Hantkenina (reviewed most recently by Coxall and Pearson 2006) is a highly distinctive, albeit minor, component of Middle and Upper Eocene assemblages. Unlike the primary chamber wall, the tubulospines of hantkeninids are externally smooth or striated and lack the mural pores that generally occur on the walls of planktonic foraminifera (Fig. 1). The internal canals of the tubulospines sometimes open out at the distal end via a small circular aperture ( Fig. 1M-P). The first species to occur widely in the fossil record, Hantkenina mexicana, commonly possesses flaring calcite pustules at the end of each tubulospine, arranged in a so-called coronet structure (Ramsay 1962, p. 82; Blow 1979Blow , p. 1168see Fig. 1A, K). Structures similar to the tubulospines of hantkeninids are known from other unrelated genera in the Cretaceous and have been cited as an example of iterative evolution (Coxall et al. 2007).
Blow (1979, p. 1165) noticed that the tubulospines of some early hantkeninids taper into the chambers at their bases and commonly show a transition over a few microns from the nonporous tubulospine surface to the fully porous chamber wall (see also Fig. 1K). He concluded from this that they are structurally contiguous with the primary chamber wall rather than being separate constructions, as is the case with the acicular 'true spines' of some modern species. From this, he inferred that the structures must have evolved by constriction and extension of the primary chamber wall, loss of pores along the tubular constrictions and other modifications such as the development of the coronet structure. Despite these inferences, the evolution of Hantkenina in the Eocene was for many years regarded as abrupt and something of a mystery, and several very different ancestors have been proposed (Cushman and Wickenden 1930;Shokhina 1937;Blow and Banner 1962;Benjamini and Reiss 1979;Blow 1979;Banner and Lowry 1985;Pearson 1993). On the basis of comparative ontogeny and test ultrastructure as revealed by dissections and X-rays, Coxall et al. (2003) showed that the origin of Hantkenina must have been in the genus Clavigerinella as had been earlier proposed by Shokhina (1937) and Benjamini and Reiss (1979) and not Pseudohastigerina as had been proposed by Blow (1979), Banner and Lowry (1985) and Pearson (1993), among others. Coxall et al. (2003) also illustrated eight apparently 'transitional' specimens from Austria that had originally been collected by K. Gohrbandt and F. Aberer in the 1950s, which had been donated to us by F. R€ ogl (Museum of Natural History, Vienna).

PREVIOUS INVESTIGATIONS
Very well-preserved assemblages of 'early' hantkeninids were described by Blow and Banner (1962), Ramsay (1962) and Blow (1979) from a locality near Kilwa Masoko in Tanzania (see Fig. 2 for palaeogeography of this and other key sites). In 2002, shortly after we sent the paper of Coxall et al. (2003) to the press, we drilled a borehole (Tanzania Drilling Project (TDP) Site 2) at this classic site using continuous coring to a total depth of 92.78 m. One of the objectives of this drilling was to attempt recovery of the postulated evolutionary origin of Hantkenina, and in this, we were successful. The lithostratigraphy, biostratigraphy and chemostratigraphy of this site were described by Pearson et al. (2004). Initial investigation of the foraminiferal assemblages from TDP Site 2 (Pearson et al. 2004) led us to propose that the immediately ancestral species of Hantkenina was a little-known form that had been described by Subbotina (1958) from the Belaya River section in the Caucasus as Clavigerinella caucasica. This was instead of C. eocanica, as we had previously suggested (Coxall et al. 2003), the two species having generally been regarded as synonyms (e.g. Blow 1979;Pearson 1993). Clavigerinella caucasica has somewhat pointed chamber ends, reminiscent of an airship in shape, whereas in C. eocanica, they are bulbous and rounded. Pearson et al. (2004) illustrated two specimens of C. caucasica and three specimens labelled as 'transitional Hantkenina' from TDP Site 2. Despite containing these apparently transitional specimens, the biostratigraphy of TDP Site 2 was problematic for us because there was a substantial mismatch between  fig. 23). C, Hantkenina liebusi, TDP 13/16/2, 50-60 cm (re-illustrated from Coxall and Pearson 2006, plate 8.9, fig. 15). D, Hantkenina dumblei, ODP 865B/4H/3, 100-108 cm (re-illustrated from Coxall and Pearson 2006, plate 8.7, fig. 10). E, Hantkenina dumblei, ODP 865C/5H/6, 110-112 cm, Tanzania. F, Hantkenina australis, ODP 647/50/5, Labrador Sea (note recurved tubulospines; re-illustrated from Coxall and Pearson 2006, plate 8.5, fig. 14). G, Hantkenina dumblei, TDP 13/ 15/2, 30-40 cm, Mkazambo, Tanzania (re-illustrated from Coxall and Pearson 2006, plate 8.7, fig. 16 Markwick et al. (2000) showing key localities discussed in the text. the planktonic foraminifer and nannofossil biostratigraphies in relation to the then standard correlation schemes (e.g. Berggren et al. 1995;Luterbacher et al. 2004). The basic problem was that the Hantkenina first appearance, although convincing enough in itself (to us), was stratigraphically far higher than expected, only just below the first occurrence of the planktonic foraminifer Globigerinetheka and within the range of the nannofossil Chiasmolithis gigas. The mismatch between foraminifera and nannofossil ages was about three million years, which at the time we thought barely possible. To help resolve this problem, we suggested that: (1) there may have been a cryptic hiatus at TDP Site 2 just above the first appearance of Hantkenina; and (2) C. gigas may have appeared much earlier in hemipelagic environments than in the open ocean (Pearson et al. 2004, p. 49). We now know from subsequent studies by others of several sections in Europe (described below, but see especially Payros et al. 2007) that it was the standard correlation that was at fault, so there is now no need to invoke anomalous ranges and a cryptic hiatus to explain the record in Tanzania.
As a contribution to the Atlas of Eocene Planktonic Foraminifera (Pearson et al. 2006), we reviewed the taxonomy of all species of Clavigerinella and Hantkenina (Coxall and Pearson 2006). We illustrated Subbotina's holotype of Clavigerinella caucasica in SEM for the first time. We also illustrated six more transitional specimens from TDP Site 2 and, based on these specimens, suggested that the evolutionary transition first involved: (1) the development of pointed chamber ends, as in Clavigerinella caucasica; (2) the evolution of a 'terminal nub' or constriction on the chamber ends; (3) the extension of the chamber ends into a roughly tubular structure, still porous like the rest of the chamber surface, which we called a 'proto-tubulospine'; and (4) the development of this into a true tubulospine by its becoming smooth and nonporous and in some cases having a coronet structure, as in Hantkenina mexicana. We noted that for stage 2, the 'distal chamber ends can be inclined in an anterior, posterior or dorsoventral direction', and for stage 3, proto-tubulospines 'are commonly bent' (Coxall and Pearson 2006, p. 252). We named a new species, Hantkenina singanoae Pearson and Coxall, to accommodate transitional forms that possess either terminal nubs or proto-tubulospines, choosing not to oversplit what was evidently quite a plastic population of forms. We assigned our new species to Hantkenina rather than Clavigerinella on the grounds that, in our view, the terminal nubs and proto-tubulospines are essentially homologous with, and 'ancestral' to, true tubulospines, while at the same time noting that the morphological transition appears gradual, and hence, taxonomic divisions are necessarily arbitrary to some extent. We assigned one of the transitional specimens from Austria that was previously illustrated by Coxall et al. (2003) to H. singanoae and the others to C. caucasica.
Meanwhile, investigations by others were devoted to studying the biostratigraphy and magnetostratigraphy of several sections in Europe with a view to establishing a stratotype for the base of the Lutetian Stage (and, hence, of the Middle Eocene). Bernaola et al. (2006) presented the first results from a spectacular section at Gorrondatxe in the Basque country of the western Pyrenees, Spain, which has clear magnetostratigraphy. In that section, they recorded the first Hantkenina (which they called H. nuttalli, regarded here as a synonym of H. mexicana) well within the stratigraphic range of the nannofossil Chiasmolithus gigas. As in Tanzania, this was a much higher level than expected, but Bernaola et al. (2006) noted that the species is rare in the section and so the level might not record its true first occurrence. However, further consideration of the problem by the same research group (Payros et al. 2007) with reference to various sections around the world, including Tanzania, led to a radical new proposal: the true level of origin of Hantkenina was within nannofossil zones CP13b and Chron C20r as opposed to CP12a and the base of C21r as previously thought (e.g. Lowrie et al. 1982;Napoleone et al. 1983;Berggren et al. 1995;Luterbacher et al. 2004;Berggren and Pearson 2005). This solution was further evidenced from other stratigraphic sections in the Basque region (Payros et al. 2009) and now seems to have been accepted by all involved (see especially Molina et al. 2011;Wade et al. 2011).
Larrasoaña et al. (2008) and Ortiz et al. (2008) presented two related papers on the detailed magnetostratigraphy and biostratigraphy of another excellent Early-Middle Eocene section from Spain, this time at Agost in the Betic Cordillera. Once again, the first occurrence of hantkeninids was found to be within the range of Chiasmolithus gigas and within Chron C20r. Just below the first occurrence of Hantkenina (recorded as H. nuttalli by Larrasoaña et al. 2008, andH. mexicana by Ortiz et al. 2008; see the systematic taxonomy for a discussion of these names), they recorded a single sample containing 'Hantkenina cf. singanoae', which they described as showing 'gradual transition to H. nuttalli ' (Larrasoaña et al. 2008, p. 401). They also recorded the presence of Clavigerinella eocanica and Clavigerinella jarvisi at a similar level. These authors did not illustrate the transitional Hantkenina, nor did they indicate why it was recorded as 'cf. singanoae'. Nevertheless, the presence of transitional specimens is important because it implies that the Agost section approximates the true first occurrence and not a later immigration. Egger (2010, 2011) published two papers on the Clavigerinella-Hantkenina transition from the Hol-zh€ ausl section near Mattsee in Austria, the same area as was originally collected and described by Gohrbandt (1967). The first paper (R€ ogl and Egger 2010) indicated a complex evolutionary scenario, in which Clavigerinella eocanica gave rise to C. jarvisi (an evolutionary dead end), and C. caucasica. Clavigerinella caucasica then gave rise to Hantkenina singanoae (another dead end), and a transitional form called Hantkenina nov. spec. (not formally named in that paper) which then gave rise to 'Hantkenina nuttalli' (their quotation marks). This taxon was the supposed ancestor of H. mexicana and another form, H. cf. mexicana. These hypotheses of ancestry were supported by some illustrations (two of C. caucasica, one of H. singanoae and three of Hantkenina nov. spec.). The various taxa were not formally described or differentiated, but the authors did comment that 'H. singanoae is here considered to be an evolutionary side branch to the Clavigerinella-Hantkenina transition, since it is unlikely that the bent chamber ends of this species have developed into straight tubulospines' (R€ ogl and Egger 2010, p. 25). Although they did not specifically make the point, their concept of Hantkenina is polyphyletic with the terminal nub evolving twice: in a 'bent' form in H. singanoae and straight in their H. nov. spec.
This scenario was further elaborated in the second paper (R€ ogl and Egger 2011), in which 12 more transitional specimens were illustrated in SEM and three in thin section. The species Hantkenina gohrbandti was formally described for the 'H. nov. spec.' of R€ ogl and Egger (2010). This species was distinguished from H. singanoae because 'the straight and pointed chamber ends differ clearly from the cylindrical, hood-like chamber ends in H. singanoae' (R€ ogl and Egger 2011, p. 5). R€ ogl and Egger (2010) also suggested that fully formed tubulospines first appeared in the juvenile stages of some specimens of Hantkenina gohrbandti.
Coccioni and Bancal a (2012) have also investigated the evolutionary origin of Hantkenina, using samples from the Contessa Highway section near Gubbio, Italy, which is especially useful because the section contains rhythmic sedimentary variations related to Earth's orbital cycles and hence provides an astronomical calibration for the critical interval. They recognize the following relevant species: Clavigerinella eocanica, C. caucasica, Hantkenina gohrbandti (morphotypes 1-3), H. singanoae, H. cf. mexicana 'forma nuttalli Toumarkine' and H. mexicana. They illustrated 32 specimens distributed among these species and morphotypes, which they regarded as representing a single gradual evolutionary transition with one side branch to H. singanoae. Like R€ ogl and Egger (2010,2011), they described H. gohrbandti as 'the real ancestor of Hantkenina' (Coccioni and Bancal a 2012, p. 71), considering it unlikely that straight tubulospines could have evolved from bent chamber ends. Also like R€ ogl and Egger (2010,2011), the genus Hantkenina is polyphyletic in their scheme, with both H. gohrbandti and H. singanoae descended separately from C. caucasica. They also noted that specimens with long tapering proto-tubulospines like those illustrated by Coxall and Pearson (2006) from Tanzania were not found in Italy or Austria and so might be interpreted as a local ecological adaptation in Tanzania.
The various suggestions of R€ ogl and Egger (2010,2011) and Coccioni and Bancal a (2012) are revisited below in the light of our new and much more extensive evidence from Tanzania. The systematic taxonomy of all relevant specimens that have been illustrated is then reviewed in the final section.

PALAEOCLIMATE, HABITAT AND ISOTOPE PALAEOECOLOGY
The peak of Cenozoic warmth was in the Early Eocene, whereupon a gradual global cooling trend set in that (notwithstanding some plateaus and reversals) persisted until the Early Oligocene when a large ice cap first appeared on Antarctica (Zachos et al. 2008). The cooling was predominantly in the high latitudes but also affected deep-ocean water masses sourced at those latitudes (Pearson et al. 2007;Bijl et al. 2009). The cooling of oceanic deep waters would have caused gradual changes in the structure of the oceanic water column worldwide. One aspect of this cooling that may have been important for pelagic niche partitioning relates to the rates of remineralization of sinking organic matter. In the very warm Early Eocene, organic carbon would have been much more efficiently remineralized in the upper water column because of more active bacterial metabolism (Olivarez Lyle and Lyle 2007;John et al. in press). This would have produced a shallower, more focused oxygen minimum zone with a more vertically condensed pelagic ecosystem above it (John et al. in press). We suggest that as subthermocline waters cooled in the Middle Eocene, bacterial metabolic rates would have slowed at depth allowing sinking organic particles to descend further, on average, than they did before, and in consequence, the oxygen minimum zone would have become less intense and more spread out. This process could have opened up new niches for deep-dwelling zooplankton that were able to tolerate relatively low levels of oxygen. We speculate that the evolution of Clavigerinella and Hantkenina was related to this global cooling process. Pulses of deep-water anoxia may also have spurred this evolution (Coccioni and Bancal a 2012).
A number of studies have published stable isotope ratios of hantkeninid and clavigerinellid tests in relation to other species to investigate the depth habitats of the group (Boersma et al. 1987;Pearson et al. 1993Pearson et al. , 2001van Eijden 1995;Coxall et al. 2000;Wade and Pearson 2008). Coxall et al. (2000) showed that the earliest hantkeninids (Hantkenina mexicana) lived in a deep planktonic environment, although subsequent species migrated to shallower depths. The development of elongate chambers has occurred several times independently in the long history of planktonic foraminiferal evolution, seemingly always in deep, oxygen-poor environments; the basic reason for this may be a general tendency for foraminifera in those habitats to optimize food harvesting and to maximize their surface-area-to-volume ratio to more efficiently take up oxygen (Coxall et al. 2007;Coccioni and Bancal a 2012).
Modern digitate planktonic foraminifera also live in deep planktonic habitats (Bradshaw 1959;Hemleben et al. 1989). Hull et al. (2011) recently described the first detailed observations of living deep-dwelling planktonic foraminifera based on observations and collections made by a remotely operated vehicle (ROV) in Monterey Bay, California. They found Hastigerinella digitata and another related species living at depths of 150-500 m, with a clear peak in abundance at c. 320 m above the core of the local oxygen minimum zone. These species are carnivorous, preying on deep-dwelling copepods that also peak in abundance at this depth, as does mesopelagic biomass in general (Hull et al. 2011). Hull et al. (2011, p. 573) described Hastigerinella as 'planktivorous sit-and-wait predators' that prey mainly on copepods year round with a peak population density of about one individual per 5 m 3 of water. Not all deep-dwelling planktonic foraminifera are this specialized, however; others feed on sinking phytodetritus or are omnivorous (Hemleben et al. 1989).
As part of this study, we subjected the tests of transitional hantkeninids (H. singanoae) to stable isotope analysis for the first time alongside a suite of co-occurring planktonic species and a single sample of benthic forami-nifera ( Fig. 3; all data are presented in Table 1). The various species were divided into depth habitats following the classification of Pearson et al. (2001), and their carbon isotope ratios were interpreted following the model of Pearson and Wade (2009). Oxygen isotope palaeotemperatures were determined using the equation of Kim and O'Neil (1997)  250 m. These data suggest that the evolution of Clavigerinella and Hantkenina was driven primarily in response to challenging environmental conditions as they invaded a new mesopelagic habitat rather than by direct competition with other species of planktonic foraminifera.
Longitudinal striations on tubulospines and adjacent chamber surfaces in H. mexicana and other hantkeninids indicate, by analogy with modern species, that they were remoulded in life by external cytoplasm flowing along their length (see Fig. 1K, L). From this, we propose that the tubulospines anchored large rhizopodial networks that would have streamed outward from the distal ends, probably with bidirectional flow to maintain a balanced distribution of cytoplasm (Hemleben et al. 1989, pp. 60-61). Analogous structures in which rhizopodia stream from terminal openings are seen in some living benthic foraminifera (e.g. R€ ottger and Kr€ uger 1990). If H. singanoae was carnivorous like modern Hastigerinella, in which rhizopodia stream from the tips of true spines (Hull et al. 2011, p. 563), these networks would have been useful for catching and securing large prey such as copepods. Alternatively, if they were herbivorous, they could have been used for scavenging small food particles, such as sinking phytodetritus, from the surrounding water, which could then have been passed into the shell through the primary or even terminal aperture, if one was present. The diameter of the terminal aperture is always very small (c. 3-10 lm; see Fig. 1P); hence, any food passed into the test via that route must been very small also, suggesting bacteria or minute algae. Although the ends of tubulospines are commonly found closed off by calcite, such closure may have occurred late in the life cycle associated with reproduction (gametogenesis) when, again by analogy with modern species (Hemleben et al. 1989, pp. 152-154), the test surface and fine-scale features were probably modified. Alternatively, in some species or specimens, closure of the canals may have been the normal state. Observation of specimens by X-ray shows that the internal canals of tubulospines extend to their very tips, even in the early species H. mexicana and H. liebusi, suggesting that they were probably open in life (Fig. 4). Another possible function for the tubulospines has been suggested to us by Susan Goldstein (pers. comm. 2012) by analogy with species of benthic foraminifera (Dahlgren 1964;Sliter 1965), namely that they could have been used to increase the efficiency of gamete dispersal at reproduction. This possibility is not in conflict with the feeding hypothesis as both functions could have occurred at different stages in the life cycle as is the case in some canal-bearing benthic foraminifera (R€ ottger and Kr€ uger 1990, p. 424). Observations of many species by X-ray (Coxall 2000; see also Fig. 4) indicate that tubulospines only occur on the neanic and adult chambers and not in the earlier juvenile stage. Tubulospines were evidently key to the life strategy of all the various species of Hantkenina and Cribrohantkenina because they were never lost, which we speculate would have been 'easy' in an evolutionary sense, through heterochrony. The structures do, however, vary considerably between individuals and especially between species, in some of which they are long and slender, and in others short and stubby; they can be tapered or curved, and they may or may not possess a distal coronet structure (see Fig. 1; see also Coxall and Pearson 2006, for many more examples).

MORPHOGENETIC CONSTRAINTS
Chamber formation in a variety of modern spinose and nonspinose planktonic foraminifera has been studied in the laboratory Hemleben et al. 1989). The sequence was summarized by Hemleben et al. (1989) as follows: (1) extrusion of a cytoplasmic bulge from the aperture; (2) organization of the bulge and extrusion of radiating rhizopodia to define the position of a perimeter called an outer protective envelope; (3) further extrusion of the cytoplasmic bulge to the line of the outer protective envelope and the construction of a primary organic membrane; and (4) calcification on both the inside and outside of the primary organic membrane. This mechanical extrusion of cytoplasm may explain why many foraminifera have relatively simple, globular, chamber shapes (Tyszka 2006). But from a consideration of these constructional aspects, it can be seen what a challenge it must have been for a foraminifer to produce tubulospines with hollow canals of almost machine-like precision and, in some species at least, minute terminal apertures. This presumably involved much more precise delineation of the outer protective envelope, and hence the resulting site of calcification (primary organic membrane), than is normal for most planktonic foraminifera. A simple cytoplasmic bulge would not easily suffice: a more complex scaffold of rhizopodia and, we speculate, rigid microtubules (e.g. Gull 2001) would seem to be necessary. Another problem for the foraminifera is that the distal chamber ends were laid down far away from the preexisting test and aperture, which is inevitably an environment more prone to turbulence than the area immediately around the test.
With these theoretical constraints in mind, we turn to examine the full collection of transitional specimens from Tanzania.  (2004) and Nicholas et al. (2006), respectively. Both holes penetrated from Middle Eocene (planktonic foraminifer Zone E9 and calcareous nannofossil Subzone NP15c) at the surface to Lower Eocene (planktonic foraminifer Zones E6/7 undifferentiated and calcareous nannofossil Subzones NP14b/NP15a undifferentiated) at total penetration depth. The primary lithology is relatively soft hemipelagic silty mudstone with occasional limestone turbidite interbeds. Bathyal benthic foraminiferal assemblages indicate a depositional depth of >350 m water depth (Nicholas et al. 2006), possibly as much as 1000 m (John et al. in press). Planktonic foraminifer shell preservation is generally excellent, although some levels are affected by infilling and overgrowth by calcite cements. As discussed above, we were confused by the apparently anomalous stratigraphic ranges of Hantkenina, Globigerinatheka and the nannofossil Chiasmolithus gigas, which led us to infer a possible hiatus in these cores (Pearson et al. 2004;Nicholas et al. 2006). We now know that the standard correlation schemes were almost certainly in error, and the successions are, as far as we can tell, continuous, although TDP Site 2 is substantially thicker in the critical interval.

THE COMPLETE COLLECTION FROM TANZANIA
The Clavigerinella-Hantkenina group is always rare in early Middle Eocene foraminifer assemblages, probably because their deep-dwelling habitat did not support large numbers of individuals. For this reason, we cannot avail ourselves of the very large collection sizes that have been made in some other evolutionary investigations of foraminifera (e.g. Malmgren and Kennett 1981;Malmgren et al. 1983;Hull and Norris 2009). Moreover, the shells are also very delicate and prone to breakage during standard extraction procedures. The total collection described here consists of c. 150 'transitional' specimens from 30 core samples, plus more numerous undisputed representatives of Hantkenina mexicana and H. liebusi from higher in the core. Some of these specimens are fragmentary chambers that may belong to the same individual. A representative selection of these specimens is illustrated to a common scale in Figures 5-8. The entire collection of transitional forms is illustrated in Figures S1-S10 (available in the online Supporting Information). The images are assigned to morphospecies in Appendix S1 (available in the online Supporting Information), which also indicates which are the type specimens and other comments.
Despite full morphological intergradation across the transition, we have assigned specimens to four qualitatively recognizable stages: 1. Clavigerinella caucasica with radially elongate and slightly tapering chambers. 2. Hantkenina singanoae Type 1 (first transitional stage), with tapering chambers that end with a distinct terminal nub or constriction on at least one chamber. When well developed, these constrictions commonly appear slightly folded and often lean in an anterior or posterior direction. 3. Hantkenina singanoae Type 2 (second transitional stage), with radially elongate chambers that gradually taper into a 'proto-tubulospine' on at least one chamber, which is a cylindrical protuberance that is porous along its length but may show a reduction in pore density towards the distal end. 4. Hantkenina mexicana with radially elongate chambers that rapidly taper into at least one smooth or finely striated and largely nonporous cylindrical tubulospine. Tubulospines commonly end in a terminal coronet structure and may open at the end via a terminal aperture. Populations may contain individuals of more than one of these stages or 'morphospecies' without implying the existence of more than one biospecies at any time (Pearson 1998). We note also that during growth, as chambers are added, a single individual may traverse more than one of these stages. In Figure 9, we show histograms of the four morphological stages against stratigraphic position in the more densely sampled drill core, TDP Site 2. Although considerable variability exists between individuals, we do not see any obvious morphological subdivisions at any stratigraphic level, and consequently, we regard the whole population as a single evolving lineage.

EVOLUTIONARY MODEL
We note the strong similarity of the Tanzanian forms to specimens from Europe Egger 2010, 2011; Coccioni and Bancal a 2012) and regard the entire known global population as a single genetically connected lineage inhabiting deep-water niches worldwide. Studies of modern planktonic foraminifera have shown that cryptic genetic variation is common, some of which may indicate cryptic species with substantial (multi-million year) divergence times in the past (e.g. DeVargas et al. 1999;Darling and Wade 2008). However, if genetic isolates occurred during the evolutionary origin of Hantkenina, they must have been very minor and short-lived in comparison, because the transitional morphotypes of C. caucasica and H. singanoae do not persist in the fossil record.
Considering the depth habitat, functional constraints and the morphoseries illustrated in Figures 4-8 and Figures S1-S10, we propose an adaptive scenario for the origin of the tubulospines (Fig. 10). The ancestral species, which we recognize as Clavigerinella caucasica, possessed radially elongate chambers with a tendency to taper at their ends, rather than ending in rounded, bulbous extremities as occurs in other members of the genus (e.g. C. eocanica, C. akersi). Individuals of C. caucasica with more pointed chamber ends must have had a slight selective advantage with respect to their mode of life. This could have been because cytoplasm flowing over the outer surface of the shell would have flowed together and been more firmly anchored at the chamber tips, hence supporting a larger or more directed pseudopodial network than in other species. A large pseudopodial network may have been analogous to the trailing tentacles commonly found in modern mesopelagic and bathypelagic plankton. Food supply in deep planktonic habitats is usually intermittent and scarce; hence, individuals with flattened, stellate tests (having, consequently, high surface-area-to-volume ratios) and robust psuedopodial networks, further extending their reach, could have been able to harvest the surrounding water more efficiently at low metabolic cost and possibly exploit new food sources such as mesopelagic copepods (if carnivorous) or sinking aggregates of phytodetritus (if herbivorous). Food would have been passed over the chamber surface and through the primary aperture in the normal way. Once these selection pressures were initiated, the process may have led to further evolution in this direction among the population that would ultimately give rise to Hantkenina. We infer this from the apparently directional nature of the trend for tubulospine development, the fact that intermediate morphologies do not persist in the fossil record, and the observation that there are other species of Clavigerinella with rounded chamber ends that survived independently for many millions of years without undergoing this specialization. In our scenario, the process of pseudopodial anchoring favoured elongation of the chambers and the development of a constriction or terminal nub at the chamber tips, leading to stage 2 described above. These terminal constrictions often lean in anterior, posterior or even sideways direction. We regard the bending of the chamber ends as adventitious, resulting from the position adopted by the primary organic membrane far from the site of cytoplasmic bulge extrusion. The exact shape of the membrane was difficult for the foraminifer to constrain and at the mercy of environmental turbulence before it was strengthened by calcification. The fact that no two specimens or even chambers are exactly alike during the evolutionary transition supports this idea. Extreme morphological plasticity is also seen in popula-tions of the homeomorphic but unrelated Cretaceous digitate planktonic foraminiferal genera Leupoldina and Shackoina (Verga and Premoli Silva 2002). Both Leupoldina, which possesses bulbous, nub-like or proto-tubulospine-like tipped extensions of the chambers, and its probable descendant Shackoina, characterized by hantkeninid-like hollow tubulospines, show immense variability in the morphology, orientation and ontogeny of their chamber modifications suggesting similar challenges of test construction.
A continuation of these selection pressures, whether constant or intermittent, would have favoured more extreme degrees of chamber elongation. The chambers began to extend distally into long, hollow, tubular and porous extensions (stage 3 above). Their considerable variability of form suggests that the morphogenesis of these structures was still less precisely controlled than was later to be the case for the true tubulospines of subsequent hantkeninids. These 'proto-tubulospines' could have contained only a small volume of cytoplasm; hence, the pores for chemical exchange would have been redundant, and moreover, they structurally weakened the proto-tubulospine and so gradually disappeared with time. However, it was always mechanically necessary that the structures were hollow, in order for the organism to be able to lay down the organic template from the inside.
Foraminiferal shells often show irregular external pustules and protrusions that are secreted onto their outer surfaces from external cytoplasm. Exaggeration of surface roughening around the tubulospine ends would have helped fan the rhizopodial network away from the tips and possibly helped divide ingoing and outgoing flows. This, we suggest, is the most likely function of the coronet structure in H. mexicana (see Fig. 1K). At some point, the chamber ends became open via a small circular aperture. These apertures could have allowed the foraminifera to bring small food particles such as bacteria, some algae or possibly the digested remnants of large prey, directly into the test without taking the longer route via the primary aperture. This may have given bearers of this structure an immediate advantage.
The selection pressures that drove this evolution were evidently occurring during the adult phase of the life  cycle because it is always on the last one to three chambers that the innovations are observed. Earlier chambers, by contrast, retain a simple morphology throughout the evolutionary transition (see, for example, Figs 5P, S, T, 6A, J, O, 7J, 8G, J, L). In later hantkeninids, however, tubulospines occur on up to six adult chambers (e.g. Fig. 1D), suggesting either that the instructions for their formation were passed forward in ontogeny, in a heterochronic process known as predisplacement, or that the lifespan was extended allowing extra chambers to be laid down (hypermorphosis; see McNamara (1986) for the terminology of heterochrony). Both processes probably occurred, and it would take further detailed studies of Xrays or dissections to determine the pattern in detail; however, hypermorphosis seems especially likely around the time of the first H. mexicana because specimens are typically larger than the transitional forms.

COMPARISONS WITH OTHER RECORDS
Professor F. R€ ogl has kindly showed us micropalaeontological residues from the Holzh€ ausl section and demonstrated the outcrops in the field. The section is monotonous marlstone about 16 m thick. Although currently poorly exposed along a wooded hillside, it was cleaned by trenching for the studies of R€ ogl and Egger (2010,2011). The transitional hantkeninids are from several samples taken through about 1 m of stratigraphy near the top of the section (R€ ogl and Egger 2010, fig. 2). The illustrated specimens, which are from seven closely spaced samples, belong (according to our taxonomy) either to Clavigerinella caucasica, the Type 1 morphotype of Hantkenina singanoae, or to H. mexicana. The more 'advanced' Type 2 morphotype of H. singanoae has not been recorded from Austria.
The Italian Contessa Highway section is composed of marly limestones and calcareous marls bundled together into astronomical couplets. Clavigerinella caucasica occurs over approximately 2.5 m of section, but transitional hantkeninids occur only in about 0.5 m of stratigraphy.
Coccioni and Bancal a (2012, pp. 78-89) (2012) noted the absence of specimens with elongated final chambers similar to those illustrated by us from Tanzania (that is, corresponding to the Type 2 H. singanoae of this study). They envisaged the evolution of the 'real tubulospine' (by which they mean, like us, a nonporous cylindrical projection) to have occurred first on a single early chamber, then on several chambers in the final whorl and finally (in true Hantkenina mexicana) on all adult chambers. This hypothesis is similar to that of R€ ogl and Egger (2010,2011). However, almost all the illustrated specimens of both these sets of authors have the delicate chamber tips broken off making it difficult or impossible to determine which morphologic variant many of the specimens belong to. We suggest that the few specimens that have been illustrated with true tubulospines in the early chambers and broken final chambers (R€ ogl and Egger 2011, pl. 2, fig. 4; Coccioni and Bancal a 2012, pl. 2, fig. 1) are most likely attributable to Hantkenina mexicana rather than to any transitional form. We have seen no such morphology in the more  The new Tanzanian material indicates that the evolution of tubulospines was directly preceded by a prototubulospine stage affecting the final 1-3 chambers (Hantkenina singanoae Type 2). The fact that no specimens like these have been illustrated from either European section suggests to us that either those records are incomplete or the delicate morphology of such forms has not been preserved via the processing of the relatively hard lithologies (especially Contessa Highway). Another possibility is that the evolutionary transition is more fully represented at tropical sites in the centre of the geographical range of these forms. We note also that the evolutionary transition phase is much more expanded in the Tanzanian section (about 15 m of stratigraphy in TDP Site 2 compared with about 1 m in Austria and 0.5 m in Italy).
The new evidence also leads us to reject the view of R€ ogl and Egger (2010,2011) and Coccioni and Bancal a (2012) that the bent chamber terminations could not have evolved into straight tubulospines, and hence, Hantkenina singanoae should be regarded as an evolutionary offshoot from the main lineage. Our collection does not show two distinct populations with straight and bent chamber ends, respectively. As discussed above, the bending of the chambers is not likely to have been a heritable genetic feature; the pattern of bending is seldom similar on adjacent chambers of the same specimen, so it is much more likely adventitious, caused by the relatively loose morphogenetic constraints in these evolving forms. Only after some time were the hantkeninids able to reliably calcify straight and well-formed tubulospines on a relatively fixed template. For this reason, we regard Hantkenina gohrbandti to be a subjective junior synonym of Hantkenina singanoae (see Systematic Palaeontology). The best estimate for the duration of the evolutionary transition from Clavigerinella caucasica to Hantkenina mexicana is from the astronomically calibrated section at Contessa Highway at a little less than c. 300 kyr (Coccioni and Bancal a 2012). If the hantkeninids had monthly lifespans as is the case with some modern deep-dwelling planktonic foraminifera (Hemleben et al. 1989), this would indicate the passage of approximately 3-4 million generations as tubulospines were evolved.

DISCUSSION
Following their evolutionary origin, the hantkeninids were a successful component of pelagic ecosystems for over 15 million years until their eventual extinction at 33.7 Ma during the climatic changes that occurred at the end of the Eocene (Wade and Pearson 2008). Eleven distinct morphospecies (including one of Cribrohantkenina) have been described (Coxall and Pearson 2006), with a wide range of chamber morphologies, all of which were descended from the short-lived transitional forms  . Isotopic evidence suggests that by the late Middle Eocene, most descendant species lived in surface water habitats (Coxall et al. 2000) in contrast to the deep planktonic environment in which the genus evolved. However, the tubulospines evidently served useful purpose(s) throughout the history of the clade because they were retained by all subsequent species of Hantkenina. It has sometimes been remarked that stratigraphic information is, in principle, of no use in inferring phylogenetic relationships and that tracing lines of ancestry and descent in the fossil record can never be justified (e.g. Gee 2000, p. 147). Our study demonstrates the opposite and that in favourable circumstances, the proactive sampling of a critical stratigraphic interval can produce new evidence and help resolve a disputed question of phylogeny. The evolutionary origin and phylogenetic position of Hantkenina have now been demonstrated unequivocally by targeted investigation of a precise stratigraphic interval by several research groups. The stratigraphic ordering provided by these studies gives us constraints on the pathway taken by evolution, its timescale and geographical range. The presence of 'transitional' Hantkenina in both Europe and East Africa suggests that the evolution did not occur in a peripheral isolate but rather across a broad area, as is perhaps to be expected in populations of oceanic plankton (Lazarus 2011). The current lack of evidence of transitional hantkeninids from the Pacific Ocean cannot be regarded as good evidence of absence because the relatively few sites that might yield such forms have not so far been sampled in sufficient detail (although we note in passing that our own detailed sampling of equatorial Pacific ODP Site 865 failed to yield transitional forms, possibly because of a small hiatus at the expected level).
The evolution of tubulospines by the hantkeninids must have involved a variety of developmental genetic innovations. Chief among these was the ability to lay down a precise organic template on which long, hollow, chamber extensions could be calcified with precision. It is likely that this mechanism involved a rigid scaffold-ing of microtubules. We cannot envisage that the genetic innovations necessary for tubulospine formation arose solely by drift or chance without the agency of selection. Experiments with diverse model organisms have shown the remarkable efficacy of artificial selection in producing rapid responses in populations in factors such as body size in mice, or bristle number in Drosophila, but the effectiveness of selection in producing such changes generally diminishes with time. Typically a barrier to directional evolution is reached as the variation produced by genetic recombination becomes exhausted (Bell 1997). Novel gene sequences are then required, presumably through mutation, and this process requires much longer periods of time. The fossil record can help determine how long real novelty, as opposed to allometric variations, takes to evolve in natural conditions. In the case discussed here, we note the largely unoccupied (at least by other planktonic foraminifera) but challenging deep planktonic niche in which Hantkenina evolved and the very long timescale of c. 3-4 million generations. This indicates that natural selection can produce distinct trends over vast periods of time when the innovations required are not straightforward.
We are aware that adaptive explanations for biological structures are notoriously difficult to prove because the myriad selective events that produced them are lost to history and a certain odium sometimes associates with 'adaptationist' thinking (following Gould and Lewontin 1979, although not necessarily in the spirit of that paper). Nevertheless, constraints on the likely evolutionary pathway and selective agencies are found when intermediate phases in the evolutionary process are discovered, as is the case in this study. We emphasize that our explanation for the evolution of tubulospines involves the interplay between natural selection in a challenging environment and morphogenetic constraints relating to chamber construction in foraminifera (the 'spandrels' of Gould and Lewontin). Our account of the origin of tubulospines is essentially similar to that predicted by Blow (1979), who was not aware of these intermediate forms and whose suggestion can therefore be regarded as having been supported and extended. We also note that the evolutionary model presented here could be falsified in various ways, most obviously by the unlikely discovery of an Early Eocene foraminifer with tubulospines homologous to those of Hantkenina.

SYSTEMATIC PALAEONTOLOGY
Remarks. Fuller synonymy lists for some of these taxa were provided by Coxall and Pearson (2006). Here we   Coxall and Pearson (2006, p. 222) and also below. BouDagher-Fadel (2012, p. 151) recently resurrected Eoclavatorella as a distinct genus without discussion, but we stand by our earlier synonymy.

Figures 5D-O
See also Figures S1.2-S1.3, S1.5-S1.11, S1.15-S1. Diagnostic features. The species is distinguished from other species of Clavigerinella by having adult chambers that taper into an acute or pointed tip (Subbotina 1958;Coxall and Pearson 2006). It is distinguished from Hantkenina singanoae principally by lacking a terminal constriction ('terminal nub'), which must be present on at least one adult chamber. The chambers are generally cylindrical rather than equatorially flattened and triangular as is common in H. singanoae. The species intergrades with both Clavigerinella eocanica and Hantkenina singanoae.
Discussion. This long-neglected species was included in the synonymy of C. eocanica by Blow (1979). Coxall and Pearson (2006) separated the two species on the basis of the distal chamber shape, which is rounded in C. eocanica, but slightly acute in C. caucasica. They illustrated the holotype of C. caucasica by SEM in three views for the first time, revealing the wall texture. They regarded it as a very short-ranging form which gave rise to Hantkenina singanoae by equatorial flattening of the chambers and the evolution of a distinct terminal constriction or 'nub' at the end of at least one chamber. Coxall and Pearson (2006) included in synonymy several specimens from Description. Wall usually smooth, normal perforate, sometimes weakly cancellate; possibly spinose. Test planispiral or pseudoplanispiral, evolute, laterally compressed biumbilicate or showing a subtly raised spiral side and very shallow umbilicus; 4-4 1 / 2 rapidly enlarging chambers in the final whorl; early chambers rounded, final 2-3 adult chambers conspicuously elongated into club-shaped (clavate) extensions; highly lobular peripheral outline; distal chamber ends smoothly rounded; equatorial higharched aperture, symmetrical or slightly asymmetrical, bordered by a smooth broad imperforate lip, relict apertural lips often preserved along sutures; sutures are shallow, straight, becoming curved in later stages, short relative to total chamber length (modified from Coxall and Pearson, 2006, p. 222).
Diagnostic features. This species is distinguished from others of the genus by possession of rounded to slightly bulbous clavate chambers.
Discussion. This was the first species of Clavigerinella to evolve and seems to have quickly given rise to a minor radiation of species including the Clavigerinella caucasica-Hantkenina lineage. According to Coxall et al. (2003), it evolved from Parasubbotina inaequispira via an intermediate species Parasubbotina eoclava. The evolution involved lowering of the height of the trochospire until a planispiral or near-planispiral arrangement was achieved and the development of extended clavate chambers in the adult.
Parasubbotina was a spinose genus (Olsson et al. 2006), and it is possible that Clavigerinella or even Hantkenina retained a sparse distribution of true spines, but this has never been demonstrated, and it is more likely that spines were lost around the time of the evolution of Clavigerinella (Coxall and Pearson 2006). Discussion. The synonymy of the various genera and subgenera listed above was discussed by Coxall and Pearson (2006, pp. 229-230). BouDagher-Fadel (2012) resurrected Aragonella and raised it to the rank of genus, but we regard this as unnecessary. Description. Wall smooth, normal perforate and probably nonspinose; tubulospines imperforate or with small, scattered pores, smooth or finely striated with a well-defined zone of demarcation between the tubulospines and the chamber wall. Test planispiral, evolute, biumbilicate or showing a slightly raised spiral side and very shallow umbilicus; laterally compressed with 4-5 rapidly expanding chambers in the final whorl; final whorl chambers radially elongate or digitate, well separated, inflated peripherally and more compressed within the umbilical region; some specimens, including the holotype, exhibit anterior flexure of the final chamber; peripheral outline distinctly stellate; each chamber of the adult whorl extends gradually into a hollow tubulospine; aperture is a narrow, elongate equatorial arch bordered by an imperforate flaring lip, often with a crenulated and/or pustulose margin, relict apertural lips are sometimes preserved as 'webs' along the sutures; sutures straight, becoming curved in the final stages, only partially contacting adjacent chamber; tub-ulospines variable in form, broad based and stout or long and slender, positioned centrally with respect to the radial chamber axis, directed radially between anterior and posterior chamber shoulders, distal ends commonly possess terminal finger-like projections (coronet structure of Ramsay, 1962) (description modified from Coxall and Pearson, 2006, p. 245).
Diagnostic features. This species is distinguished from later species of Hantkenina by the stellate outline and radially directed tubulospines that are positioned centrally on the chamber. It is distinguished from H. singanoae by possessing true nonporous tubulospines with a distinct demarcation in texture from the primary chamber wall.
Discussion. Coxall and Pearson (2006, pp. 243-246) introduced the synonymy of Hantkenina mexicana Cushman and H. nuttalli Toumarkine. The specimen chosen by Toumarkine (1981) as the holotype does not fit the type description well and has broken chamber ends on the last three chambers, making it impossible to determine their original state and hence base a reliable species concept. The four specimens from Austria described as 'primitive' Hantkenina nuttalli by Coxall et al. (2003) include one specimen that we assign tentatively to Hantkenina singanoae and three that we assign tentatively to Hantkenina mexicana. The reason for the uncertainty is that all the specimens are broken, and the original nature of the tubulospines or proto-tubulospines is uncertain. R€ ogl and Egger (2010,2011) illustrated several specimens under different names that we assign to this species: the reason for including two of the specimens assigned to Hantkenina nov. spec. (initially, in 2010) and then to Hantkenina gohrbandti (in 2011, as paratypes of that taxon) is that both apparently show a nonporous tubulospine on an early chamber, but other chambers in the final whorl are broken or eroded so it is not possible to determine whether they possessed terminal nubs, proto-tubulospines or tubulospines; we think the latter is most likely given the tubulospines on the earlier chamber. Three specimens labelled as 'Hantkenina gohrbandti -H. mexicana transition' by R€ ogl and Egger (2011) apparently each possess a nonporous tubulospine on the one complete chamber remaining of each specimen and so are assigned to this species. Similarly, the morphotype recorded as 'Hantkenina gohrbandti R€ ogl and Egger morphotype 2' by Coccioni and Bancal a (2012) is probably a broken specimen of H. mexicana insofar as an earlier chamber appears to possess a true nonporous tubulospine according to the description of those authors.
Description. Wall layered and perforate with a smooth or weakly cancellate surface. Test planispiral or pseudo-planispiral, laterally compressed, 4-5 rapidly expanding chambers in the final whorl; final chamber and up to three of the preceding chambers radially elongate, at least one of these ending in either a terminal nub or a cylindrical projection ('proto-tubulospine'); proto-tubulospines are smooth and distinctly porous (although pore density may be reduced at the distal ends) in contrast to the true tubulospines of H. mexicana and subsequent species, which are usually imperforate; distal chamber ends can be inclined in an anterior, posterior or dorsoventral direction, and proto-tubulospines are commonly bent; aperture is a high equatorial arch with a smooth, broad lip; lips of relict apertures are commonly observed along the sutures of the final whorl; sutures are shallow and usually curved in a posterior direction (description modified from Coxall and Pearson, 2006, p. 252).
Diagnostic features. This species is distinguished from all other Hantkenina by lacking true nonporous tubulospines and possessing instead a terminal nub to the chamber or a porous proto-tubulospine in the sense of Coxall and Pearson (2006). Subdivision of H. singanoae from C. caucasica relies on the identification of the slightest terminal constriction on at least one chamber (see also Egger 2011 andCoccioni andBancal a 2012).
Discussion. Following Coxall and Pearson (2006), we take a broad view of this morphospecies, including within it specimens similar to Clavigerinella caucasica that possess a terminal constriction or 'nub' (which we call Type 1; the holotype is one such specimen) and those that possess long, slender proto-tubulospines (Type 2). We see little value in splitting the taxa formally based on minor details given that the populations are so variable during the evolutionary transition. The holotype and most of the paratypes of Hantkenina gohrbandti R€ ogl and Egger are of Type 1, but only just, insofar as the terminal constriction is very small on these specimens, and they are hence very similar to C. caucasica.
Authors' contributions. PNP made the collection and wrote the manuscript. HKC produced the X-rays and map and contributed to the interpretation and writing.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article: Figs S1-S10. The complete collection of specimens from Tanzania arranged in stratigraphic order, shown to a common scale. Numbers next to specimens are the collection numbers that are linked to Appendix S1.
Appendix S1. List of specimens illustrated in Figures 5-14 from the stratigraphic collection at TDP Sites 2 and 20.