Adaptive ecological niche migration does not negate extinction susceptibility

Extinction rates in the modern world are currently at their highest in 66 million years and are likely to increase with projections of future climate change. Our knowledge of modern-day extinction risk is largely limited to decadal-centennial terrestrial records, while data from the marine realm is typically applied to high-order (> 1 million year) timescales. At present, it is unclear whether fossil organisms with common ancestry and ecological niche exhibit consistent indicators of ecological stress prior to extinction. The marine microfossil record, specifically that of the planktonic foraminifera, allows for high-resolution analyses of large numbers of fossil individuals with incredibly well-established ecological and phylogenetic history. Here, analysis of the isochronous extinction of two members of the planktonic foraminiferal genus Dentoglobigerina shows disruptive selection differentially compounded by permanent ecological niche migration, “pre-extinction gigantism”, and photosymbiont bleaching prior to extinction. Despite shared ecological and phylogenetic affinity, and timing of extinction, the marked discrepancies observed within the pre-extinction phenotypic responses are species-specific. These behaviours may provide insights into the nature of evolution and extinction in the open ocean and can potentially assist in the recognition and understanding of marine extinction risk in response to global climate change.

Current extinction rates are estimated to be at least eight times higher than the background Cenozoic (< 66 Ma) average 1 and understanding the impacts of rapid climate change on global biodiversity is of critical importance for creating a sustainable future (https:// sdgs. un. org/ goals). An increasing body of evidence suggests climate state variability is potentially more important than the direction of temperature change with respect to heightened extinction rates 2,3 . As such, we look to the Cenozoic marine sedimentary record, which allows us to assess the impacts of high variability in climate state on extinct biodiversity 4,5 . In this study, we focus on the planktonic foraminifera, single-celled marine protists with a global distribution and the most complete Cenozoic specieslevel fossil record 6 . Their calcareous skeletons, or tests, preserve not only their entire life history, but also a biogeochemical expression of the surrounding water column (e.g., 7 ). These features allow for high-resolution species-specific quantification of physiological and ecological adaptation through periods of climate variability (e.g., [8][9][10][11][12][13][14][15][16][17][18][19][20][21]. Our analysis investigates the response of the planktonic foraminiferal genus Dentoglobigerina, of which two species (Dentoglobigerina altispira and Dentoglobigerina baroemoenensis) undergo an isochronous extinction at ~ 3.04 Ma, during a period associated with increasing climate state variability 22,23 . Through high-resolution (~ 5 kyr) paired morphometric and geochemical analyses, we demonstrate that despite the two species occupying the same ecological niche space and sharing close phylogenetic affinity, they exhibit species-specific ecological and morphological responses prior to extinction.
Morphological data enables the assessment of the relationship between body size and shape parameters likely to record long-and short-term morphometric trends in response to global climate 14 (see Methods for designation of size and shape parameters). Single-specimen planktonic foraminiferal geochemical records allow us to place species within discrete ecological niches or "ecogroups" (see 6 ), wherein the investigation of stable oxygen (δ 18 O) and carbon (δ 13 C) isotope ratios can be used to determine the relative degree of bathymetric and ecological separation within extant and extinct species 7,24 .
The methods and hypotheses tested in this study highlight the utility of the marine micropaleontological record in assessing the pre-extinction ecological response of organisms at high-resolution during intervals of global climate variability.

Statistical analyses.
Our data and linear models (see SI, Tables S1-4) identify trends in dentoglobigerinid morphology and ecology that support the designation of distinct ecological "Phases" in the final ~ 30 kyrs of our Dentoglobigerina species. Phase 1 is typified by stable morphological and geochemical background conditions from 3.466-3.071 Ma (Figs. 1,2). At this point the Phase Transition commences, identified by the morphological excursion of D. altispira (Figs. 1, 2). Finally, 10 kyr later Phase 2 initiates, marked by the coeval end of the period of diminished size in D. altispira, and the geochemical excursion representing a shift in its ecological niche (Figs. 1, 2, 3). For D. baroemoenensis, samples sourced from the Phase Transition are notably lacking in specimens, and enhanced size increase and morphological stochasticity is observed through much of Phase 2, (Figs. 1, 2).
Models tested whether the abiotic conditions of the paleo-water column (signified by the stable isotopic signature of the present extant species with known living depths 6,24 ), responded in or out of tandem with the dentoglobigerinid signature. Throughout Phase 1 the dentoglobigerinids have isotopic signals that are consistent with a surface mixed-layer species which hosts photosymbiotic algae 6,24 (Figs. 1, 3) and the two dentoglobigerinid species δ 18 O signals respond in tandem (Tables S1 & S2). In Phase 2, there is a marked decoupling in the δ 13 C signal of the two species (Tables S3 & S4) potentially indicating a disruption of photosymbiosis for D. altispira, but not for D. baroemoenensis, which maintains its ecological affinity up to the sample preceding extinction (Fig. 1i, j). Additional linear models testing the relationship between dentoglobigerinid morphology and environment show that as bottom water δ 18 O becomes more positive and δ 13 C more negative (e.g., as mean global temperature decreases/ice extent increases) dentoglobigerinid test area increases throughout Phase 1 (Tables S1  & S2). However, for D. altispira, this signal is lost in Phase 2, and for D. baroemoenensis the signal switches to increased test area mirroring more negative bottom-water δ 18

Discussion
The Cenozoic planktonic foraminiferal fossil record documents a strong positive correlation between test size and the degree of global marine latitudinal and vertical temperature gradients 22,27 , however the largest intraspecific test sizes tend not to be analogous with species' ecological optima 28,29 . In Phase 1, dentoglobigerinid test size data shows a gradual relative increase through time (Figs. 1; S1) likely representing a response to the development of temperature gradients associated with the intensification of northern hemisphere icesheets 22,23 . At the initiation of the Phase Transition (~ 3.071 Ma) stepwise disruptions in both morphology and geochemical signatures are likely indicative of disruptive selection and "bet-hedging" 30,31 , a typical response to the propagation of terminal stress levels preceding extinction 32,33 . Previous studies 15,16 document increasing growth asymmetry and morphological trait variance as responses to abiotic forcing, wherein species produce offspring with high inter-individual phenotypic variability during unfavourable environmental conditions to improve mean population fitness 15,[34][35][36] .
In the case of D. altispira, the ecological end-result following the Phase Transition approaching the termination of the record appears to be permanent ecological niche migration from the surface mixed-layer down to the subsurface, supported by Mann-Whitney U test results (Table 2) on δ 13 C signals between Phases 1 and 2 (Figs. 1, 3). Significantly, D. baroemoenensis displays an almost total absence during the Phase Transition ( Fig. 1 www.nature.com/scientificreports/ www.nature.com/scientificreports/ supporting the prevalence of environmental conditions detrimental to dentoglobigerinid ecology. Upon its return, it shows a dramatic increase in lateral and umbilical area (Figs. 1; S1), where Mann-Whitney U test and z-test results indicate significantly different body size and range to Phase 1 ( Table 2). These changes, which we term "pre-extinction gigantism", are antithetical to the "pre-extinction dwarfing" 10,37 previously documented in several other species, and may represent a response to the steepening of vertical and latitudinal water column temperature gradients associated with cryosphere development 22,23 , typified by more distinct spacing between the geochemically assigned ecological habits during Phase 2 (Fig. 3).
The minor general trend of increasing body size range in both species (Fig. 1a, c) potentially infers long-term mitigation of external environmental pressures expressed through rising polymorphism. Further research is required; however, such behaviour may be characteristic of temporally long ranging species 38 when subjected to global climate state variations which deviate far from their ancestral ecosystem 39,40 .
The study species share close phylogenetic and ecological affinity 38 , maintain high-order morphological likeness from speciation, and undergo isochronous extinction, yet the phenotypic responses recorded prior to extinction are species-specific. Rapid within-clade character change, cladogenesis, and extinction during periods of detrimental environmental change are likely common-place within the history of life 32 , and phylogenetically and ecologically adjacent taxa can exhibit similarities in selection pressures which do not necessarily trigger an adaptive response in the same direction 32,33,39 .
For D. altispira, migration from the surface mixed layer to the subsurface may be compounded either by a photosymbiont reduction/suppression 18 , or adoption of a facultative symbiotic ecological strategy, recognized to enhance flexibility of nutritional sources through minimal energetic investment 41,42 . This proposed adaptation is suggested over total algal photosymbiont "bleaching" 9,11,18,21,43 , as D. altispira continues to present δ 13 C enrichment higher than values observed in the asymbiotic, subsurface-dwelling taxa analysed in this study 42 , yet similar δ 18 O values (Figs. 1, 3).
For D. baroemoenensis, water column temperature gradient dynamics associated with thermocline shallowing 22,44 are a potential trigger for the brief vacation and apparent "pre-extinction gigantism" exhibited www.nature.com/scientificreports/ through Phase 2. The final sample of this species' record may be marked by symbiont bleaching wherein, despite a size increase, specimens exhibit a reduction in δ 13 C values (Fig. 1). Alternatively, D. baroemoenensis may also be recording rapid migration down to the subsurface, or a significant change in the extent of test calcification just prior to extinction 24,45,46 . Photosymbiont bleaching driven by extreme heat stress has been recorded in extant groups such as corals 47 and larger benthic foraminifera 48,49 , and previously, records of the potential bleaching of algal photosymbionts within fossil planktonic foraminifera have been confined to early Cenozoic hyperthermals (e.g., 9,11,18,21,43 ). Despite bleaching amongst modern corals being primarily driven by increasing temperatures 47,50,51 , a multitude of environmental stressors are associated with bleaching responses 52,53 , and the prospective bleaching of D.  www.nature.com/scientificreports/ baroemoenensis may indicate that this pre-extinction response in symbiont-bearing taxa may be more common than previously thought during intervals not characterised by elevated temperatures.
Model results (Tables S1-4) lend further support to our interpretations, in which both morphological and ecological responses display discrete signatures between phases, characteristic of disruptive selection 30,31 . The interpreted behaviour of these organisms raises some interesting questions. One of the most pressing and fundamental issues for palaeoceanography, is whether fossil organisms identified via their external morphology which are used for the inference of paleoclimatic data maintain ecological uniformitarianism for the entirety of their stratigraphic range. Our study, alongside other novel research on modern and fossil populations 18,[54][55][56][57][58][59][60][61][62][63] suggests not, and as such deriving environmental interpretations from fossil taxa, particularly during intervals of climate variability, should be treated with caution. Whether the documentation of these behaviours indicate failed efforts at stress mitigation via water-depth associated parapatric anagenesis is currently undetermined, but further high-resolution comparable investigations through speciation events may help to understand the fundamental mechanisms driving evolution and extinction in an ecosystem with limited vicariance potential such as the open ocean.

Summary
The studied section exhibits a high-resolution record of the pre-extinction biotic response of two members of the planktonic foraminiferal genus Dentoglobigerina during major global palaeoceanographic changes associated with the development of northern hemisphere ice sheet formation. Despite the species' phylogenetic and ecological affinities, documented phenotypic responses are species-specific, wherein both D. altispira and D. baroemoenensis exhibit evidence of permanent adaptive ecological niche migration and photosymbiont reduction. In addition, D. baroemoenensis documents "pre-extinction gigantism", and potential photosymbiont bleaching. This study highlights the importance of high-resolution analyses when investigating biological responses and extinction dynamics. The unparalleled resolution of the marine microfossil record allows us to identify and evaluate past occurrences of morpho-ecological stochasticity indicative of disruptive selection and niche adaptation. However more comprehensive studies utilising multiple localities are required to improve our understanding and identification of the potential for pre-extinction signals to better recognise extinction risk in response to rapid climate change.

Methods
Site selection. Material was sourced from Integrated Ocean Drilling Program Expedition 321 Site U1338 (Hole 1338A) (2°30.469′N, 17°58.162′W) situated in the East Equatorial Pacific, which was drilled to 410.0 mbsf through Holocene-early Miocene pelagic sediments 64 . At ~ 3 Ma, the site was in a deep-water pelagic environment of similar water depth and paleolatitude 65 to the modern. The primary lithologies represented are calcareous, diatom and radiolarian nannofossil oozes and chalks. Despite the deep-water settings and primarily calcareous nature of the sediments, excellent microfossil preservation has been recorded throughout this core interval 66 . A preliminary assessment of core U1338A was carried out to determine the approximate position of the extinction of the dentoglobigerinids (~ 3 Ma) based on tropical biostratigraphy 67 , and shipboard paleomagnetic data 64 .
Assemblage analysis. Sediment volumes of 20-40 cc were collected and washed with de-ionised water over a 63-µm sieve; the residues were dried in an oven at 40 °C and split. All samples were examined using a Zeiss Stemi 305 Compact Stereo Microscope. We identified planktonic foraminifers following the taxonomy of Kennett and Srinivasan 68 , Schiebel and Hemleben 27 , and Wade et al. 38 and performed assemblage counts on 300 individuals from > 63 µm splits.  69 using QAnalySeries software 70 to better constrain the timing of preextinction responses compared to using palaeomagnetic data alone.
Morphometrics and repeatability. The first 50 (where present) complete specimens of the genus Dentoglobigerina were picked and mounted in umbilical position on card slides pierced with a fine needle to accommodate the variably spired nature of species in the genus 38 . Specimens were imaged umbilically using a Zeiss Axio Zoom V16 microscope with attached Canon EOS 100D camera at × 19.4 magnification. All specimens were then rotated 90° laterally, and imaged whilst propped onto their penultimate chamber. Images were processed using the image analysis software Image Pro Premier, and the "size" trait parameters: test area (Area, µm 2 ), and test size range (minimum test diameter (Dmin)-maximum test diameter (Dmax)), and "shape" trait parameters: aspect ratio (AR, ratio between maximum test height and width), roundness (perimeter 2 (µm)/4π.test area), and circularity (4.test area)/(π.MaxFeret 2 ), were captured from both orientations, extracted, and databased (see SI).
To determine whether the size and shape parameters were repeatably valid measurements, trait repeatability 71,72 was performed by removing, remounting, and reimaging 200 specimens of each analysed dentoglobigerinid species (100 umbilical orientations, and 100 lateral). Measurements of the repeated runs (Fig. S2) are plotted using continuous frequency distributions (kernel density estimates with a Gaussian kernel and bandwidth h = 1.06*sn 1/5 following Silverman 73 , with s the standard deviation of trait measurements per species and n the number of analysed individuals). Results were evaluated through Wilcoxon signed-rank test using R software 74 .
Where mean run rank differences deviated significantly, measured traits were deemed non-repeatable (Fig. S3).
All measured size and shape trait parameters were deemed repeatable for D. altispira, whereas for D. baroemoenensis, roundness and circularity were not repeatable, and were subsequently removed from further interpretations. Repeatable traits were then subjected to power analysis 75 using the 'pwr' package in R 76 to determine the minimum number of individuals required to detect mean sample trait changes of 5, 10, 15, 20, 25, and 30% with power > 0.9 and a significance level of p = 0.01 as suggested by Brombacher et al. 71 (Fig. S4; Table S5).

Stable isotope analysis.
For stable isotope analysis, species-specific size fractions were screened and picked for exceptionally preserved specimens of Dentoglobigerina altispira (> 200 µm), and Dentoglobigerina baroemoenensis (> 200 µm) exhibiting "excellent" and "glassy" preservation 77 . The same screening process was performed for nominate taxa representing specific ecological habits through the water column: Globigerinoides ruber (212-350 µm, surface mixed-layer), Neogloboquadrina incompta (212-350 µm, subsurface), Globorotalia tumida (> 300 µm, thermocline/photic zone base, corrected for a 1.0‰ δ 13 C enrichment due to this species occupying the shallow oxygen minimum zone and consequential effects of reduced ambient pH 24 ), Hirsutella scitula (212-300 µm, subthermocline), and Cibicidoides wuellerstorfi (> 212 µm, bottom-water) (see 6,24,78,79 ). Single specimens of dentoglobigerinids, and multiple specimens of all other foraminifer species were analysed using an Elementar IsoPrime Dual-Inlet Isotope Ratio Mass Spectrometer in the Cohen Geochemistry Laboratory, University of Leeds, and data are reported to the Vienna Pee Dee belemnite (VPDB) scale using an Elemental Microanalysis Carrera marble standard where analytical precision was better than 0.07 and 0.13 ‰ for δ 13  Statistical analysis. Prior to statistical modelling, all sample mean isotopic and morphometric measurements were log standardised and the sample first differences generated per time step. All statistical analyses and linear modelling were carried out using R software 74 . For linear models, statistically significant relationships were identified between variables, and residual standard mean errors (RSE) were calculated to determine the model fit (Tables S3 & S4). The normality of dentoglobigerinid morphometric and geochemical parameter data was assessed using a Shapiro-Wilk test 80 to determine whether parametric or non-parametric tests were applicable. Dentoglobigerinid sample-mean morphological and geochemical parameters were tested to identify study section outliers through Grubb's test 25 using the 'outliers' package 81 . Comparisons between all dentoglobigerinid morphological and geochemical parameters between the two phases (i.e., specimens before and after the defined Phase Transition) were performed by Mann-Whitney U 26 using the 'asht' package 82 , and z-tests using the 'BDSA' package 83 . www.nature.com/scientificreports/