The first size reduction (FSR) in the Reticulofenestra-Gephyrocapsa-Emiliania (RGE) lineage (order Isochrysidales), which occurred in the early Oligocene (~32 Ma), is of great significance for understanding the Lilliput effect that has affected coccolithophore communities from the late Eocene to this day. We conducted a morphologic analysis on the coccoliths of Reticulofenestra species that lived during the late middle Eocene to early Oligocene (~40–31 Ma), using marine sediments from the South Atlantic Ocean. Our data show increasing size and decreasing abundance of the large species during the late Eocene, leading to their disappearance at the FSR, and a concurrent decrease in the size variability of the small- to medium-sized coccoliths whose central opening diameter had become very reduced. Although the cosmopolitan late Paleogene through Neogene size decrease in coccolithophores has been linked to the concomitant long-term decline in global pCO2, we suggest here that the FSR was the result of environmental destabilization caused by the expansion of eutrophic environments following the late Eocene establishment of overturning circulation associated with ice buildup on Antarctica. This study also leads us to propose a hypothetical model that links coccolith morphology of species of the RGE lineage and trophic resources in the upper ocean: the small- to medium-sized, r-selected coccolithophores with smaller coccolith central openings live in nutrient-rich waters where they rely mostly on photosynthesis and little on mixotrophy, whereas the larger, K-selected species with larger coccolith central openings live in oligotrophic waters where they are more dependent on mixotrophy.

Coccolithophores are unicellular algae, with the cell surface covered by calcite scales (coccolith). Coccolithophore size has become progressively smaller from the late Paleogene to the present. The first size reduction (FSR) occurred ~32 million years ago in the most successful modern lineage, the order Isochrysidales. To decipher the driving mechanism(s) behind the FSR, we conducted a morphologic analysis on the fossils of coccolithophores that lived during the late middle Eocene to early Oligocene (~40–31 million years ago), using marine sediments from the South Atlantic Ocean. Our data show a strategic divergence between groups of large and small coccolithophores. The group of larger coccolithophores increased in size by ~4 μm before their extinction, while the size range of the group of smaller coccolithophore contracted. In addition, the central opening on the coccolith decreased in all species from the late Eocene to early Oligocene. We infer that the increased nutrient availability during the Eocene/Oligocene climate transition may be responsible for the morphologic evolution. We hypothesize that larger species with larger central openings on their coccoliths were more dependent on mixotrophy, making them better adapted to the nutrient-poor conditions from the middle to late Eocene. On the other hand, smaller species whose central opening dwindled were better adapted to nutrient-rich conditions during the early Oligocene. Our inference supports the idea that the diversity of trophic strategies among phytoplankton is underestimated.

Coccolithophores are calcifying, unicellular, planktonic algae that contribute up to 20% of the organic carbon fixation in the upper ocean (Poulton et al. 2007) and about half of the deep-sea carbonate precipitation (Milliman 1993). Because of their rapid evolution, the products of their calcification, coccoliths and coccospheres, have long been used in Earth sciences, not only to date sediments (e.g., Martini 1971; Backman et al. 2012) but also to characterize major biotic events since the Late Triassic, when they became common in marine sediments (e.g., Bown et al. 2004; Suchéras-Marx et al. 2019). Following the Cretaceous/Paleogene mass extinction, the coccolithophores underwent bursts of diversification, reaching their highest diversity during the early Eocene climatic optimum (EECO). Subsequently, global cooling and associated oceanic eutrophication fueled the diversification and spreading of opportunistic (r-selected) species, leading to an early Oligocene bottleneck of the entirety of coccolithophore diversity (Bown et al. 2004; Aubry and Bord 2009; Hannisdal et al. 2012; Henderiks et al. 2022). While the major K-selected taxonomic groups became sequentially restricted, species of the genus Reticulofenestra increased in abundance, leading to their dominance in the Oligocene coccolithophore communities (Aubry 1992). Reticulofenestra appeared during the EECO (Agnini et al. 2006), and except for one species, disappeared in the late Pliocene. It gave rise to Gephyrocapsa in the late Miocene, from which Emiliania evolved in the Pleistocene (Perch-Nielsen 1985; Gibbs et al. 2005), both of which dominate living communities (e.g., Poulton et al. 2017).

The Reticulofenestra-Gephyrocapsa-Emiliania lineage (RGE) is of great interest, because it shows the same Lilliput effect that affected all coccolithophores during the Neogene (Aubry 2007, 2009; Fig.1)—a size decrease along lineages that is not yet fully understood. The RGE lineage belongs to the family Noelaerhabdaceae Jerkovic, 1970 (Young and Bown 1997) of the order Isochrysidales Pascher, 1910, and the taxonomically complex Reticulofenestra group is often referred to as reticulofenestrids. Although marked by pervasive iterative evolution (Matsuoka and Okada 1990; Young 1990; Bendif et al. 2019), the first selection for small size in the lineage marks the Eocene/Oligocene transition, when species with coccoliths ≥14 μm disappeared (Backman and Hermelin 1986). No species of the RGE lineage with comparable size would evolve thereafter. Thus, in light of their Neogene evolution toward a small size, a documentation of this first, distinctive event, which is dubbed here the “first size reduction” (FSR), holds the potential to shed light on size and shape selection forcing mechanisms in Reticulofenestra and to confirm or nuance existing interpretations of size change forcing mechanisms among species of the RGE lineage through the Neogene and up to the present (Fig. 1).

We present here the results of a morphometric analysis of the Reticulofenestra coccoliths preserved in the upper Eocene through lower Oligocene chalks from the South Atlantic Ocean, focusing on three morphologic parameters— coccolith length, central opening/coccolith length ratio, and circularity. The forcing mechanisms that possibly drove changes in these morphologic traits are discussed in the context of the Eocene–Oligocene climatic changes at a million-year temporal resolution. We conclude with a hypothetical ecological model that attempts to relate morphology of the RGE coccoliths with the ecological and/or physiological requirements of their species and that may well be applicable to other coccolithophores.

Material

We analyzed 32 samples taken at intervals of ~50 cm in cores from a 100-m-thick section of upper Eocene through lower Oligocene chalks from Deep Sea Drilling Project (DSDP) Site 363 in the eastern South Atlantic Ocean (19°39′S, 09°03′E; Bolli et al. 1978; Fig. 2) between Core 5, section 5, 80 cm, and Core 10, section 5, 45 cm (~114–212 m below seafloor [mbsf]). Coring was discontinuous at this site, but a preliminary analysis showed that the FSR is well represented in the continuously recovered Core 7 (Proto Decima et al. 1978) and that the section yielded consistent information.

Calcareous nannofossils were abundant and well to moderately well preserved in all samples except in the bottom of Core 7 (155 mbsf) and in the upper part of Core 8 (~165 mbsf), which were omitted from the analysis. Reticulofenestra coccoliths were abundant in the chalks sampled for this study, representing ~60–90% of the assemblages (Supplementary Fig. S1). Their preservation was good, although with minor evidence of overgrowth that we have quantified (Supplementary Plate S1).

Reticulofenestra Coccoliths

Reticulofenestra coccoliths are elliptical to circular placoliths. Their structure was first comprehensively described by Stradner in his emendation of the genus (Stradner in Stradner and Edwards 1968), with further information given by Driever (1988), Gallagher (1989), and Young (1989); they are abundantly illustrated in numerous studies (e.g., Perch-Nielsen 1971; Wise 1983). The Reticulofenestra placoliths consist of two superposed shields that delineate a tubular central opening similar in shape to the coccolith. The central opening is closed by a thin, porous layer of anastomosed elements, which extend from the elements of the proximal shield. The central opening may also be spanned distally by a coarser structure that exhibits a diversity of configurations. The variable and often transitional structure of the central area has resulted in a complex and sometimes controversial taxonomy of the genus Reticulofenestra, with one to four generic names in common use depending on the authors. Species taxonomy is based on the morphology of the coccoliths and the structure of their central area, implying that species concepts in paleontology are based on coccolith morphotypes. Even so, the delineation of paleontological species in Reticulofenestra is rendered difficult by overlapping characters and iterative evolution, and size is often used as a primary specific character. Thus, for the purpose of this study, we examine all (paleontological) species collectively except Dictyococcites, whose central opening is difficult to delineate accurately for measurement. The species composition is constant throughout the studied interval, although the loss of the large species Reticulofenestra umbilicus Martini and Ritzkowski 1968 and Reticulofenestra hillae Bukry and Percival 1971 at ~32 Ma constitutes a prominent event that is part of the FSR. Backman and Hermelin (1986) showed from nearby Site DSDP 523 that the large species R. umbilicus originated in the middle Eocene following a progressive size increase between ~45 and 43 Ma and that a subsequent shrinking of the central opening led to the occurrence of the ecophenotypic species R. hillae. The abrupt and synchronous extinction of both species marks the boundary between Biozones NP22/NP23 (Martini 1971) and Biozones CNO2/CNO3 (Agnini et al. 2014). It is one of the most prominent events of the Oligocene timescale, although a comprehensive morphologic study of this event at ~32 Ma has been lacking. The species with smaller (~6 μm), subcircular placoliths that dominated Oligocene Reticulofenestra communities after the FSR belong to Reticulofenestra floridana (Roth and Hay, 1967 in Hay et al. 1967) Theodoridis 1984. For further information on Reticulofenestra taxonomy, the reader is referred to Nannotax3 (www.mikrotax.org/Nannotax3, to which the reader is also referred for all other taxa cited in this work).

Sample Preparation

All samples were prepared from dispersion in water of a small amount of chalk deposited on a cover slip and mounted on a glass slide after evaporation of the water (Corliss et al. 1984). The smear slides were analyzed at a magnification of 1000× using a Zeiss Axioplan 2 polarized light microscope coupled with Zeiss software AxioVision40 v. 4.8.2.

Morphometric Analysis

Morphometric Measurements

Morphometric measurements were performed by randomly taking photographs under bright-field conditions of 10 to 15 fields of view per smear slide at 1000× magnification. A field of view typically contained 5 to 10 Reticulofenestra placoliths. For each sample, 100 specimens were measured. To avoid distortion caused by different orientations of the coccoliths as seen in light microscopy, only specimens horizontally oriented with the microscope stage were measured.

Measurements were conducted manually using the biometric analysis software Fiji v. 1.52 P (Schindelin et al. 2012) by dragging ellipses over the contours of each placolith in order to measure its length and width. The long and short diameters of the central opening were likewise measured. Our procedure has resulted in a database of 6400 measurements for the whole study.

To estimate the uncertainty in our measurements, we selected three specimens from three different smear slides, ranging from the smallest to the largest placoliths, and measured the images of each one 20 times. The error was ~0.14 μm for the largest specimens (mean = 13.3 μm) and ~0.15 μm for the smallest specimens (mean = 4.5 μm). In addition, we randomly took five samples and measured 180 specimens in each. The results show minor differences (≤0.2 μm) in mean values between the two sampling sizes. The accuracy of the measuring method thus fulfills the needs of this study (Supplementary Fig. S2).

We undertook a morphometric analysis with the double objective of quantifying the morphologic features of the placoliths at any stratigraphic level and analyzing their change through time. Three characters yield meaningful biological and physiological information (Fig. 3):

  1. Length of coccolith. In middle Eocene through early Miocene Reticulofenestra species, the coccolith size is linearly proportionate to cell size (Henderiks 2008), so the length of a coccolith constitutes a reliable proxy for the diameters of the cells and coccospheres. Thus, for example and regardless of geological age, a cell ~6 μm in diameter, can secrete several ~6-μm-long coccoliths and form a ~9 μm coccosphere (Supplementary Fig. S3). However, it is unknown whether there is a linear correlation between coccolith size and cell size (Henderiks 2008) for the large category (e.g., R. umbilicus and R. hillae in this study), as there are no data available for coccoliths larger than 10 μm.

  2. Central opening/coccolith length ratio:

    Length ratio = (length of the central opening/length of the coccolith) × 100%.

With a few exceptions, a central opening is present in all extinct/extant Reticulofenestra species, regardless of differences in the morphology of their shields. This strongly suggests that the role of the central opening is primordial and has remained unchanged along the RGE lineage.

  1. Circularity of the coccolith:

    Circularity = width of coccolith/length of coccolith.

Higher circularity means a more circular shape, while lower circularity means a more elliptical coccolith shape. Circularity reflects species diversity.

Statistical Analysis

To verify whether the observed morphologic variations between different stratigraphic intervals are significant, we applied analysis of variance (ANOVA) by grouping the samples into three groups (see details in the “Results”). Two different test methods were performed on the datasets. For the normally distributed datasets with similar variances, an ANOVA Tukey's pairwise post hoc test was applied. The Kruskal-Wallis and Mann-Whitney pairwise post hoc tests were used to test for equal means of nonnormally distributed datasets. The null hypothesis is that all the samples have the same mean values. If p < 0.01, then the hypothesis can be rejected, meaning that the tested datasets are significantly different. We added five samples, with 180 measurements in each, as replicates in the difference significance tests. The tests were conducted using PAST 4.06 (Hammer et al. 2001).

Age Model

The age model in this study is constrained by biostratigraphy. Data from calcareous nannofossil and planktonic foraminifera are consistent (Proto Decima et al. 1978; Toumarkine 1978; Keller 1983), and we verified the location of the calcareous nannofossil datums (Supplementary Fig. S4, Supplementary Table S1). We used the timescale of Gradstein et al. (2012).

The 100-m-thick section analyzed here represents an ~9 Myr interval, extending in terms of calcareous nannofossil biochronology from middle Eocene Biochron NP17 to early Oligocene Biochron NP23, with estimated ages of ~40 Ma and 31 Ma for the bounding horizons at ~212 and 114 mbsf. An age–depth plot (Supplementary Fig. S4) shows that the section is not temporally complete but includes several important hiatuses. The sediments across the Eocene/Oligocene boundary were eroded, with an inferred unconformity at ~186.6 mbsf (Core 9, section 3); the associated hiatus is ~0.9 Myr. Two additional unconformities truncate the middle to upper Eocene interval in Core 10. The oldest inferred unconformity at ~204.9 mbsf (in section 2) represents a ~2.8 Myr hiatus between ~37 and 39.8 Ma; the younger unconformity, at ~201 mbsf (top of Core 10), represents a ~2.1 Myr hiatus between 36.9 and 34.8 Ma. However, the lower Oligocene interval is continuous between ~187 and 109 mbsf (Cores 9 to 5), and was deposited at an estimated accumulation rate of 3.8 cm/kyr (Supplementary Fig. S4).

The discontinuous sedimentary record could hamper a comprehensive analysis of the evolution of Reticulofenestra through the late middle to late Eocene. However, the morphometric data show minor morphologic changes occurring from the middle through the late Eocene, although this interval is marked by long hiatuses in our section. Importantly, the FSR, which occurred in the early Oligocene (~32 Ma), is recorded in a demonstrably continuous stratigraphic interval (see Fig. 4). Further details can be found in Supplementary Fig. S4 and Supplementary Table S1.

Variations in the Three Parameters

The coccolith length of Reticulofenestra between 212 and 114 mbsf at DSDP Site 363 comprises between ~3 and 20 μm, the central opening/coccolith length ratio between ~15% and 65%, and the circularity between 0.6 and 1 (Fig. 4AC, small forms are in gray, and large forms are in orange). Based on size distribution, two size groups can be separated out at 9 μm (Fig. 5). Although there are substantial overlaps, the central opening/coccolith length ratio among the large forms is slightly higher than among the small to medium forms (Fig. 4B). All measured coccoliths range from long elliptical (<0.75) to circular (>0.98), although the most circular coccoliths (width/length = 1) are <9 μm in length (Fig. 4C). Eight species of Reticulofenestra were identified, of which 15 specimens were measured for each for comparison with the species’ holotypes (Table 1). There is a structure to the morphologic variations, as shown by the integrated data, with three easily delineated intervals from the middle Eocene to lower Oligocene (Fig. 5).

The lower interval (Core 10, ~212 to 201 mbsf, middle Eocene Zone NP17 to lowermost upper Eocene Zone NP19-20, ~40–37 Ma; Fig. 5) is marked by contracted size distribution. Small to medium-sized coccoliths, ~3–9 μm long, dominate all assemblages, with the largest mean value of ~6.8 μm reached in the youngest stratigraphic level (~201 mbsf, NP18 zonal interval). Larger forms (coccolith length = 9–20 μm) occur consistently as well, although in notably smaller amounts. The central opening/coccolith length ratio varies from ~40% to 50% and markedly declines to ~38% at the top of this interval at ~201 mbsf (Fig. 5B). The broadest range in circularity (coccolith width/length) occurs in this interval, where minimal values of <0.6 and a mean value of 0.84–0.96 are recorded (Figs. 4, 5).

In the middle interval (Cores 9–7, ~190–145 mbsf, upper Eocene Zone NP19-20 to lower Oligocene NP22, ~37–32.2 Ma; Fig. 5), coccoliths with lengths of ~3–9 μm are predominant, and larger forms are considerably fewer than in older stratigraphic levels. In addition, there is a progressive upward size increase (by ~4 μm) in the large forms (Fig. 4A,D,G). Also, the size continuum observed in the lower interval is disrupted in the upper part of this interval, where ~9-to 10-μm-long coccoliths are scarce. The central opening/coccolith length ratio decreases progressively from ~40% to 35% on average, and the standard deviation in each sample is steady at ~5% (Fig. 5B). The circularity increases from ~0.9 to ~0.94 from the lower to the upper part of the interval, with one exceptionally low mean value occurring at ~187 mbsf (Fig. 5C).

There are no large forms (Figs. 4, 5) in the upper interval (Cores 6 and 5, ~130 to 114 mbsf, upper lower Oligocene Zone NP23, ~32.2–31.5 Ma; Fig. 5). The mean values of coccolith length are only ~5.6–5.7 μm, and the amplitude of variation (4–8 μm) of the coccolith length is more contracted than in lower stratigraphic levels. In addition, this interval is characterized by the lowest central opening/coccolith length ratios (<35% on average; Fig. 5B) and the highest average circularity (>0.92; Fig. 5C) in the small- to medium-size group.

Significance Difference Test on the Three Parameters

Because the coccolith length does not follow a normal distribution, the Kruskal-Wallis and Mann-Whitney pairwise post hoc tests were used to test for the three intervals delineated. The results show that the lower (~40–37 Ma) and middle (~34.5–32.2 Ma) intervals differ significantly: a Kruskal-Wallis test for equal means shows p(same) ≤ 0.01, and a Mann-Whitney test shows p < 0.01. The middle and the upper (~32.2–31.5 Ma) intervals show mathematically nonsignificant differences (p > 0.05), which is possibly due to the low abundance of the large forms in the middle interval. Nevertheless, from a micropaleontological perspective, these two intervals differ greatly, as the large forms (>9 μm) are absent from the upper interval, having disappeared at ~32.2 Ma. As the central opening/coccolith length ratio follows a normal distribution, a Tukey's pairwise test was conducted. It shows that the three intervals differ significantly (p < 0.01). For the circularity, the one-way analysis of similarities (ANOSIM) test also shows significant differences between the three intervals (p < 0.01).

Characterization of the FSR

The stratigraphic distribution of the coccolith length, central opening/coccolith length ratio, and circularity from the upper middle Eocene through the upper Oligocene unambiguously shows a major change in the size composition of the Reticulofenestra assemblages ~1.7 Myr after the Eocene/Oligocene boundary (Fig. 4). It is marked by the loss of coccoliths larger than 9 μm, together with a narrowing of the amplitude of variation of both the central opening/coccolith length ratio and the circularity (Fig. 5B,C). Before the FSR (~32 Ma), small- to medium-sized coccoliths (~3–9 μm) dominated the assemblages, and larger coccoliths (~9-–20 μm) were consistently present, even though their number decreased through time. Above the NP22/NP23 boundary, these large coccoliths had disappeared, and at the same time, the size variations of the small to medium coccoliths had become more restricted and the length of the central opening had decreased. The FSR describes this early Oligocene biotic event, dated at ~32.2 Ma based on our age model, which affected the size of Reticulofenestra coccoliths. On this basis, we infer that the FSR also describes a considerable reduction in the cell/coccosphere size of Reticulofenestra species.

Collective Adaptive Morphology of the Reticulofenestra Species during the FSR

As highlighted in this study of Site 363, until 32.3 Ma, the late middle Eocene to early Oligocene, Reticulofenestra coccoliths were segregated into two main size categories, with small- and medium-size coccoliths distinct from larger forms (Fig. 6). One category, rather stable, consisted of small- to medium-sized coccoliths with a mean length of ~6 μm, a 15–60% central opening/coccolith length ratio, and ~0.70 to 1.00 in circularity; the other category ranged from ~9 to 20 μm in length, 25% to 55% in the central opening/coccolith length ratio, and ~0.80 to 0.95 in circularity. These marked differences imply the existence of two groups of species, possibly with different environmental preferences, as often seen in the species of the RGE lineage (e.g., Hagino et al. 2000; Beaufort and Heussner 2001) and extinct Reticulofenestra (e.g., Imai et al. 2015).

The diversity of the small- to medium-sized coccoliths was driven by a broad range in the central opening/coccolith length ratio and circularity. The central opening/coccolith length ratio decreased from the late middle Eocene through the early Oligocene, while the abundance of circular coccoliths increased. The important moment in this transition was near the NP18/NP19-20 biochronal boundary (~37 Ma), immediately after which the central opening/coccolith length ratio had shifted from ~45% (before) to ~36%, and when a minimum circularity of 0.84–0.9 was reached. This was reflected by a notable increase in circular Reticulofenestra coccoliths with a small central opening across the Eocene/Oligocene transition (see also Wei and Wise 1990; Dunkley Jones et al. 2008). To explain this evolutionary history, we hypothesize that during the middle to early late Eocene (Biochron NP17 to early Biochron NP19-20, 40–34.5 Ma), the surface ocean was populated by species bearing coccoliths of diverse ellipticity with a central area opened to varying degrees. By ~34.5 Ma (late Biochron NP19-20), the proportion of species bearing (sub)circular coccoliths with a small central opening had increased. These achieved dominance in the early Oligocene (Biochron NP22), a dominance they maintained through Biochron NP23.

The large (>9 μm) Reticulofenestra coccoliths underwent a different evolutionary history. They form a diverse category of subcircular to broadly elliptical coccoliths, most with a large central area (>25%). The abundance and mean size of this category were rather constant during Biochrons NP17 and NP18. This constancy suggests a stable ecological niche. In contrast, from Biochron NP19-20 to Biochron NP22, the abundance of these large species decreased steadily at the same time as the length of their coccoliths (and by implication, the diameter of the cells) increased, indicating that the ecological niche might have begun to change. This decrease in abundance concomitant with an increase in cell size continued until the end of Biochron NP22, when a shift toward dominance of the larger species (~18 μm in cell diameter) with a smaller central opening (<35%) occurred in Reticulofenestra hillae, which suggests a broadening of the habitat to adapt to changing environments. The extinction of all the large species followed, which is the most visible component of the FSR.

In summary, as recognized through morphologic analysis of coccoliths and broadly converting these data into cell size, the late middle Eocene through early Oligocene evolutionary history of Reticulofenestra species reveals two distinct strategies leading to the FSR. Species with cells of small to medium size maintained a constant mean size, although their total size became greatly constrained after the FSR at the same time as species with circular placoliths and small central openings became dominant. In contrast, the species with large cells decreased in abundance at the same time as their mean diameter increased. The strategic divergence between the two size groups of species probably began in the late Eocene, increasing progressively during the early Oligocene to reach a maximum immediately before the FSR. The FSR is thus a complex event that involved different responses of the Reticulofenestra species depending on cell size.

Climatic Forcing on Size Evolution of Coccolithophores of the RGE Lineage

Although the ecology of individual species of living coccolithophores remains poorly known, a general understanding of the relationship between cell size and physiology or ecology is emerging (Linnert and Mutterlose 2015; Sheward et al. 2017; Faucher et al. 2020; Beaufort et al. 2021). Geographic distribution and temporal evolution indicate that different size groups (morphotypes) can have specific depth or seasonal niches (e.g., Beaufort and Heussner 2001; Suchéras-Marx et al. 2010), while culture studies have shown that environmental factors, including temperature, light, carbon dioxide (CO2), and nutrients, have a significant control on cell size (e.g., Müller et al. 2008; Aloisi 2015).

It is difficult to constrain a driving mechanism for coccolith/cell size evolution from fossil records because of the multiple environmental factors involved. However, the long-term size adaptation of the coccolithophores during the Cenozoic is currently attributed to long-term changes in atmospheric pCO2 (e.g., Henderiks and Pagani 2008; Erba et al. 2010; Hannisdal et al. 2012; Bordiga et al. 2015; Guitián et al. 2020; Henderiks et al. 2020). Atmospheric pCO2 is regarded as a main evolutionary driver, because the lower volume/surface area ratio (V:SA) of large cells is seen as a weakness for the uptake of carbon resources, which leads to a failure in survival in low pCO2 conditions (Henderiks and Pagani 2008). The threshold behavior of the atmospheric pCO2 found to be responsible for the Antarctic glaciation and the Eocene–Oligocene global climate transition (DeConto and Pollard 2003; Pearson et al. 2009; Hutchinson et al. 2021) could account for the evolutionary events across this transition, as described here. In addition, the occurrence of large Reticulofenestra during the middle Eocene (~40 Ma; Backmann and Hermelin 1986), when atmospheric pCO2 levels were at the highest of the late Paleogene (Rae et al. 2021), would seem to support a long-term control of pCO2 on the coccolithophore size. The relationship between coccolith size and pCO2, however, is not a simple one. A short-term control is, in fact, not easily applicable to the rapid evolution of Reticulofenestra size during the Neogene, when repeated size divergences and extinctions occurred without considerable fluctuations in CO2 concentrations (e.g., Young 1990; Gibbs et al. 2005; Imai et al. 2015). More importantly, Quaternary fossil records have shown that the large species in the RGE lineage always become larger and reach a size maximum before their extinctions (e.g., Matsuoka and Okada 1990). The so-called Matsuoka and Okada cycle (Bendif et al. 2019) refers to very typical evolutionary pattern that is also seen in the Neogene (e.g., Young 1990; Imai et al. 2015) and Paleogene Reticulofenestra (this study; Fig. 5D). The pre-extinction size increases may represent the initiation of each recurring evolutionary event. However, these events are not always synchronous with CO2 variations. Our data show that the changes in morphology and abundance patterns of Reticulofenestra species between ~40 and 31 Ma were not synchronous with pCO2 changes (Fig. 7). Substantial morphologic changes in Reticulofenestra species, including enlargement of the larger cells and a ~10% decrease in the central opening of all coccoliths, began ~2–3 Myr earlier than the distinct drop in pCO2 in the earliest Oligocene (Pagani et al. 2011; Zhang et al. 2013; Anagnostou et al. 2020; Fig. 7). It would seem, therefore, that an additional forcing is involved in the size evolution of the RGE lineage. We propose that nutrient concentration in the upper ocean was that forcing.

Studies of cores from different oceans at different latitudes show that coccolithophores responded to the global nutrient concentration increases during the Eocene/Oligocene transition, as the (inferred) oligotrophic taxa were vastly replaced by (inferred) eutrophic taxa (Dunkley Jones et al. 2008; Villa et al. 2014, 2021; Bordiga et al. 2015; Jones et al. 2019). The changes in coccolith length and central opening/coccolith length ratio that we describe here from Site 363 on the Frio Ridge portion of the Walvis Ridge were synchronous with increased nutrient availability and productivity during the late Eocene and early Oligocene (~34.5–32 Ma), as determined from proxy-reconstructed nutrient concentration (P accumulation rate) and export productivity (biogenic Ba) in the South Atlantic Ocean (Anderson and Delaney 2005) (Fig. 7D,E). Micropaleontologic/paleoceanographic studies have proposed that smaller Reticulofenestra are better adapted to upwelling and nutrient-rich conditions (Bolton et al. 2010; Imai et al. 2015) or simply to more dynamic environments (Holcová 2013). During the late Eocene, DSDP Site 363 underwent intensified surface- and deep-water circulation (Corliss et al. 1984). This was likely related to the deep-sea overturning circulation established during the same time, bringing Antarctic Intermediate Water (AAIW) into the southern Atlantic Ocean at mid- to low latitudes (Katz et al. 2011; Egan et al. 2013; Coxall et al. 2018; Fig. 7A). Early Oligocene overturning circulation meant a worldwide expansion of oceanic eutrophy at the expense of oligotrophy. The upwelling system in the eastern South Atlantic Gyre that was sustained by the Benguela Current (Lutjeharms and Meewis 1987) may be the dynamic that brought nutrients from the AAIW (~750 m; Sarmiento et al. 2004; Katz et al. 2011) to the surface ocean during the early Oligocene. On this basis, it can be speculated that the increase in small Reticulofenestra cells indicates the expansion of oceanic eutrophic conditions, while the extinction of large cells indicates the reduction/loss of oligotrophic conditions (Fig. 7). The synchronous decrease in the central opening of the coccoliths (Fig. 7D,E) may also be the result of ocean eutrophication, which is worth exploring from the perspective of adaptive morphology.

Adaptive Morphology in the RGE Lineage

Previous studies dealing with forcing on Reticulofenestra evolution have centered on the size of coccoliths and cells, often ignoring the shape of the former (e.g., Beaufort 1992; Gibbs et al. 2005; Henderiks and Pagani 2008; Imai et al. 2015; Guitián et al. 2020), although theoretically adaptive morphology is important (Aubry 2007). An exception concerns the study by Backman and Hermelin (1986), who calculated the ratio of the diameter of the central opening to the length of the coccolith. They showed a progressive decline in the central opening/coccolith length ratio in the large forms during the middle to late Eocene, and attributed it to a simultaneous oceanic cooling documented by Shackleton (1975). Our study confirms the finding by these authors that the size of the central opening is as important as the length of the coccoliths in describing morphologic evolution in Reticulofenestra.

Like the living species of the RGE lineage (see www.mikrotax.org/Nannotax3), the coccospheres of Reticulofenestra species would have consisted of interlocking coccoliths with small gaps between them allowing contact between seawater and the cell itself. It may be inferred that in species possessing coccoliths with no or tiny central openings, the surface area (and volume) represented by the gaps was much greater than the surface (volume) represented by the central openings, as can be seen in the recently extinct Gephyrocapsa caribbeanica Boudreaux and Hay, 1967 in Hay et al. 1967. In contrast, in species possessing coccoliths with a large central opening, the surface (volume) represented by the openings would have exceeded the surface (volume) represented by the gaps and would have markedly increased the porosity of the coccosphere (as can be seen in the living species Gephyrocapsa oceanica Kamptner, 1943). By inference, the central opening in the large Reticulofenestra species of the FSR would have been a primary contact between the cell and its environment. It is important to recognize that a thin, organic, pattern-less base plate likely underlined the proximal side of the Reticulofenestra coccoliths, as in Emiliania huxleyi (Billard and Inouye 2004). Such a base plate is seen to constitute a physical barrier below the central opening and its perforated proximal grid. However, in the absence of evidence to the contrary, and considering its reactivity during intracellular calcification (Paasche 2001; Marzec et al. 2019), it may be speculated that the base plate in the RGE lineage may not have constituted a chemical barrier to nutrients penetrating the cell through the central opening.

In small as in large Reticulofenestra coccoliths, the diameter of the perforations (as measured on scanning electron microscope (SEM) images of Reticulofenestra umbilicus, Reticulofenestra lockeri, Reticulofenestra dictyoda, etc. available on Nannotax3) of the proximal grid across the central opening is almost constant (~0.2 μm), which implies that the pores (and the central opening) have a very specific function (Fig. 3). We concur with Young (1987), who, noting the ubiquity of perforations in coccoliths of extinct and living species, inferred a physiologic role for them in interactions between the cell and seawater. Although this speculation is technically difficult to prove, there is support from both environmental and laboratory studies. For instance, Hagino et al. (2000) observed that in the modern equatorial Pacific Ocean, the morphotypes of Gephyrocapsa with a wide central opening were better adapted to a more stable environment and nutrient-poor conditions. Bartal et al. (2015) found that the coccospheres of E. huxleyi provide stable, enclosed microenvironments that benefit cells during the uptake of nutrients in turbulent water columns. This may be interpreted by analogy with studies of the role of a skeleton in diatoms. For example, using numerical modeling, Mitchell et al. (2013) suggested that when pulses of water pass the cells, nutrient concentration equilibrates quickly. Then, the closed surface structure of the skeleton produces asymmetric diffusion resistance. The highly limited out-diffusion from the cell to the surrounding waters permits the retention of a high nutrient concentration surrounding the membrane (Mitchell et al. 2013), which is essential for nonmotile unicellular organisms (Falkowski and Oliver 2007). We hypothesize that the coccosphere had a similar function in modulating nutrient transport, the smaller central opening being beneficial for coccolithophores’ autotrophic nutrient uptake, especially in turbulent water.

In contrast, our data suggest that large species with broadly open coccoliths were better adapted to the oligotrophic ocean. A possible explanation is that larger openings facilitated nutrient uptake through mixotrophy. Autotrophy has been seen as the rule among coccolithophores, but mixotrophy has been speculated based on morphologic/structural analysis of coccoliths and coccospheres (Aubry 2009; Young et al. 2009; Gibbs et al. 2020), and there is increasing experimental and field evidence that this may be the case. Coccolithophores are able to both prey on (e.g., Parke and Adams 1960; Rokitta 2011; Avrahami and Frada 2020) and absorb (e.g., Benner and Passow 2010; Godrijan et al. 2020, 2022) organic components from the environment. In fact, it is now recognized that mixotrophy, as an intermediate physiology between autotrophy and heterotrophy, occurs in all ecosystems and across the whole eukaryotic phylogeny (Selosse et al. 2017). The cell of planktonic algae is surrounded by a zone enriched in its discarded metabolic products. These attract bacteria, which contribute their own metabolic components, some of which can in turn be used as nutrients by the algae (Smriga et al. 2016; Seymour et al. 2017). This beneficial dual strategy is embodied in the tentative concept of a phycosphere (Bell and Mitchell 1972; Liu et al. 2022), and Seymour et al. (2017) have suggested that the degree of symbiotic mixotrophy in E. huxleyi can be determined by the size of the phycosphere where diffused organic matter is utilized via bacterial chemotaxis. We hypothesize, therefore, that a large central opening, allowing enhanced plankton–bacteria interactions, may have favored mixotrophic activities in the oligotrophic ocean of the Eocene.

At present, the oligotrophic upper photic zone is inhabited by large, usually K-selected species (Hagino et al. 2000; Kinkel et al. 2000). By comparison, we suggest that while the small to medium Reticulofenestra were r-selected species and lived in more eutrophic waters, the large Reticulofenestra species were K-selected and inhabited oligotrophic environments (Fig. 8).

To summarize, we propose an ecological model that relates nutrient forcing to morphology in Reticulofenestra (Fig. 9). Species with different cell sizes and lengths of the central openings of their coccoliths adapted differently to varying degrees of nutrient availability. The larger species with a large central opening that would have facilitated mixotrophy were better adapted to the warm, oligotrophic ocean; smaller species with a smaller central opening were better adapted to the eutrophic ocean, where they depended less on mixotrophy. This tentative model explains the divergent adaptive morphology of Reticulofenestra species around the Eocene/Oligocene transition, and it may help in conceptualizing changes in coccolith morphology in the RGE lineage through time in terms of adaptive evolution in an evolving ocean.

The sudden disappearance of large Reticulofenestra species from the early Oligocene coccolithophore communities has been known for a long time (Martini 1971; Backman and Hermelin 1986). We term it the “first size reduction” (FSR, ~32 Ma) of the Reticulofenestra-Gephyrocapsa-Emiliania (RGE) lineage and describe it quantitatively in the double context of the long-term (~40–31 Ma) morphologic history of the coccoliths of this lineage and the important paleoceanographic changes that were associated with global cooling, Antarctic glaciation, and atmospheric pCO2 decline during the Eocene/Oligocene transition (Zachos et al. 1996; Coxall and Wilson 2011). We show that the decrease in pCO2 postdates by ~2–3 Myr the first indication of destabilization of size structure among late Eocene Reticulofenestra communities. All Reticulofenestra species, small and large, were simultaneously affected by progressive establishment of global oceanic eutrophication starting in the late Eocene, although their response in terms of adaptive morphology varied with coccolith length. The larger species (>9 μm) became extinct at the FSR, ~3–5 Myr after the first evidence of destabilization of the Reticulofenestra community, and we hypothesize that this extinction reflected the disappearance of a warm oligotrophic surface ocean. From this, we infer that the large FSR Reticulofenestra species were K-selected and with ecological requirements like those of extant coccolithophores with large coccoliths. In this scenario, these Reticulofenestra species, which consistently possessed coccoliths with wide central openings, relied on absorbotrophic mixotrophy via interactions with bacteria living in the nutrient-enriched waters immediately surrounding the coccosphere. By contrast, the smaller species (≤9 μm) flourished after the FSR, when the relative size of the central opening of the coccoliths decreased. We infer that these species, dominated by Reticulofenestra floridana, were r-selected and adapted to the more eutrophic surface waters that would have prevailed after the FSR. We propose that the reduction in the central opening/coccolith length ratio was an adaptation to these more eutrophic conditions, suggesting that mixotrophy was not favored for the duration of Biochron NP23 (~32–30 Ma), when coccolithophore diversity was at its lowest level of the Paleogene (Aubry 1992).

We add nutrient availability in the photic zone to variations in atmospheric pCO2 to account for the long-term evolution of size in the coccolithophores. This is not to claim that pCO2 has played no evolutionary role. We propose, however, that the adaptive morphology of coccoliths, not only cell size, should be considered to decipher the long- (and short-) term evolution of the species of the RGE lineage, and of coccolithophores in general. Our interpretation also echoes the view that the diversity of trophic strategies among coccolithophores is vastly underestimated (Segev et al. 2016; Taylor et al. 2017; Godrijan et al. 2020, 2022; Henderiks et al. 2022) and that this diversity underpins the biodiversity of coccolithophores.

We would like to thank W. A. Berggren, L. Beaufort, L. Collins, and H. Zhang as well as fellow students C. Li and D. F. V. Hincapié for insightful discussions and constructive comments that have improved the article. Samples were provided by the International Ocean Discovery Program. This study was funded by the National Natural Science Foundation of China (41876046 and 41930536), and one of us (R.M.) would like to thank the Chinese Scholarship Council for the financial support during his 18-month visit at Rutgers University, where this study was conducted. We are very grateful for the constructive comments of three anonymous reviewers and of associate editor L. Collins, whose thoughtful comments and careful editing have greatly improved the quality of this paper.

The authors declare no competing interests.

Data and supplementary material (Supplementary File: Supplementary Figs. S1–S4, Supplementary Plate S1, and Supplementary Table S1) are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.cjsxksnbm.

This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.