Vegetation response to exceptional global warmth during Oceanic Anoxic Event 2

The Cenomanian–Turonian Oceanic Anoxic Event (OAE2; ~94.5 million years ago) represents an episode of global-scale marine anoxia and biotic turnover, which corresponds to one of the warmest time intervals in the Phanerozoic. Despite its global significance, information on continental ecosystem response to this greenhouse episode is lacking. Here we present a terrestrial palynological record combined with marine-derived temperature data (TEX86) across an expanded OAE2 section from the Southern Provençal Basin, France. Despite high TEX86-derived temperature estimates reaching up to 38 °C, the continental hinterland did support a diverse vegetation, adapted to persist under elevated temperatures. A transient phase of climatic instability and cooling during OAE2 known as Plenus Cold Event (PCE) is marked by the proliferation of open, savanna-type vegetation rich in angiosperms at the expanse of conifer-dominated forest ecosystems. A rise in early representatives of Normapolles-type pollen during the PCE marks the initial radiation of this important angiosperm group.

P ast time intervals of exceptional climatic warmth, typically associated with elevated pCO 2 , had profound impacts on floral compositions and biogeographic patterns of continental vegetation [1][2][3] . Within the overall greenhouse climate characterizing the Mesozoic, the Late Cretaceous Oceanic Anoxic Event (OAE) 2, which spans the Cenomanian-Turonian boundary [94.1 million years ago (Ma)], marks the onset of an extreme phase in ocean temperatures known as the "Cretaceous thermal maximum" [4][5][6] . This phase is characterized by one of the highest (>35°C) proxy data-derived sea-surface temperature (SST) estimates of the last 150 Myrs, which are recorded by both planktonic foraminifera δ 18 O and archaeal membrane lipid-based TEX 86 data 6 . A significant rise in low-and mid-latitude open ocean SSTs (2-4°C) and mid-latitude shelf-sea temperatures (4-5°C) accompanied the onset of OAE2, resulting in the Late Cenomanian-Turonian hothouse 5,[7][8][9] . Besides the exceptional thermal conditions, the OAE2 (lasting 700-800 kyrs 10,11 ) is associated with widespread formation of organic-rich deep-water deposits 12,13 , a major positive carbon isotope excursion (CIE) in carbonate and organic carbon reflecting massive burial of 13 Cdepleted carbon 14,15 , and major biotic turnover in marine ecosystems 16,17 .
Despite the outstanding position of OAE2 as one of the most remarkable events of the Mesozoic 18 , the responses of terrestrial ecosystems and continental flora to the changes of the global climate system remain largely unexplored. To date, just a few isolated plant macrofossil discoveries have been reported from the organic-rich Bonarelli Level, which is the sedimentary expression of the OAE2 in Marche-Umbria, Italy 19 . Microfloral evidence is essentially lacking due to the overwhelming predominance of amorphous kerogen in most marine OAE2 black shales, diluting any continent-derived spore-pollen signal. However, based on a shift in the δ 13 C signature of leaf-wax n-alkanes, a change from C3 to C4-dominated low-latitude vegetation triggered by a pCO 2 drop has been proposed 20 for the early phase of OAE2. According to theoretical considerations 21 , the exceptional warmth that prevailed during Late Cenomanian-Turonian times may have exceeded the heat tolerance of continental ecosystems, which potentially resulted in widespread vegetation dieback.
Here we present a high resolution and taxonomically differentiated spore-pollen record across a stratigraphic interval correlating to the OAE2. The record is from an expanded marine section (Cassis) from the Southern Provençal Basin (SPB) of S' France ( Fig. 1) and has been analyzed for biostratigraphy, palynology, TEX 86 and stable carbon isotopic composition of carbonates and organic materials. Stratigraphic assignment of the section is based on existing ammonite data 22 combined with new information from planktonic foraminifera, calcareous nannoplankton, and carbon isotopes of bulk carbonate, organic matter and plant wax-derived long-chain n-alkanes.

Results
Stratigraphy. The~235 m thick Cassis section represents a heterolithic succession with its lowermost part being composed of clays and marls overlain by (in part) turbiditic sandstones (0.0-19.7 m). An exposure gap in the outcrop separates the siliclastic facies from a conspicuous limestone slope deposit characterized by intense slumping and disintegration (33.4-46.1 m). Above another exposure gap, the main part of the section is continuously accessible (55.8-236.0 m) and consists of homogenous marls with intercalated bundles of nodular limestone together representing basinal facies. Existing ammonite data combined with new biostratigraphic results constrain this section to the Upper Cenomanian-Lower Turonian 22 (Supplementary Note 1). This age assignment is further corroborated and refined by carbon isotope data, which show a characteristic positive CIE revealing an initial high-amplitude positive peak (a) followed by a trough interval, which gives way to a plateau with two consecutive peaks (b, c) (Fig. 2). The characteristic pattern of the CIE is mimicked in the δ 13 C org signature of bulk organic matter (Supplementary Fig. 1) and corroborated by the isotopic composition of longchain n-alkanes derived from plant leaf-wax. Backed by highresolution biostratigraphic data, the CIE evolution can be closely correlated with the reference carbon isotope record from Eastbourne, UK 23,24 (Fig. 2). The significantly increased δ 13 C carb values of the CIE peak (a) at Cassis (6.0‰) compared to Eastbourne (4.8‰) may reflect a shift in the dominant carbonate source in the SPB during this particular interval. In fact, the Calcaires du Corton Fm. corresponds to a calcareous unit composed of upper slope deposits, which can be traced toward the shoalwater carbonate platform bordering the SPB to the north where similarly high δ 13 C carb signatures are observed in OAE2equivalent platform limestones 25 . Carbonate carbon isotope signatures > 6.0‰ for the OAE2 CIE are also described from the epeiric Iberian seaway 26 and may reflect variations in the aragonite content of the periplatform ooze exported from the adjacent platform 27 . The correlation thus reveals very good stratigraphic coverage of the uppermost Cenomanian-Turonian at Cassis with an expanded~200 m thick OAE2 interval (Fig. 2). High sedimentation rates accompanied by increased input of continental organic debris in a transtensive geotectonic setting 28 may have prevented the accumulation of black shales enriched in marine-derived organic carbon in the SPB.
Palynology. At Cassis, high numbers of inaperturate and bisaccate gymnosperm pollen (avg. = 48.2%) are essentially produced by Araucariaceae and Cupressaceae-type plants (Araucariacites spp.; Inaperturopollenites spp.) and Cheirolepidiaceae (Classopollis spp.) with subordinate contributions from Pinaceae (Cerebropollenites spp.) and Podocarpaceae (Podocarpidites spp.). Other gymnosperms, including cycads-ginkgophytes (Cycadopites spp.), gnetaleans (Ephedripites spp.), and seed ferns (Alisporites spp.) are quantitatively of minor importance (avg. = 2.5%). Spores produced by a diverse assemblage of ground ferns and fern allies occur in moderate quantities (avg. = 12.8%). The assemblage contains a variety of eudicotyledonous angiosperm pollen represented by various tricolporate and triporate types, the majority of which can be assigned to early forms of the Here, the strong dominance of a few abundant conifer-derived pollen (with Inaperturopollenites spp. reaching up to 31.8%; avg. = 20.4%) causes substantially higher G/A-ratios (avg. = 0.77) and higher dominance indices (avg. = 0.12). Calculated origination and extinction records show very low values and the rangethrough diversity comparatively high and stable values throughout the studied interval, which are illustrating the absence of major land plant extinction during OAE2. Increased values at the base of the origination record and at the top of the extinction record, as well as reduced values at the base and the top of the range-through diversity trend are interpreted as consequences of edge effects.
The overall rich and diverse palynological association obtained from Cassis is taxonomically distinctive for a position within the Late Cretaceous Normapolles phytogeographic province 29 . Stratigraphic trends in the distribution of the abovementioned groups and indices are considered to primarily reflect changes in the vegetation structure of the adjacent continental hinterland of the SPB.
TEX 86 temperature estimates. SST estimates based on TEX 86 show continuously high temperatures >35°C for the SPB during Late Cenomanian-Turonian times with maximum SSTs reaching 38°C (Fig. 3). These warm SSTs conform with existing TEX 86 studies from early Late Cretaceous mid-and low-latitude sites 5 23,24 . Calcareous nannofossil zonation of Eastbourne after ref. 56 . Planktic foraminiferal biozonation of Eastbourne after ref. 69 . Lowercase letters correspond to principal carbon-isotope peaks identified in the Eastbourne record 24

Discussion
Critically high temperatures have been put forward as a mechanism for the suppression of terrestrial plant and animal life during the end-Permian extinction and basal Triassic events [30][31][32] . For these intervals, exceptional climatic warmth is considered a first order control for the well-documented turnover in continental vegetation 3,33 . During the Cretaceous thermal maximum, exceptional warmth and associated heat stress have also been suggested to significantly affect plant growth by inhibiting photosynthesis to a certain extent at daytime temperatures ranging between 35°and 42°C 21 . Continental temperatures in this range are considered life-limiting to plants, resulting in serious thermal stress and potential die-off 34,35 . However, the rich and diverse flora reconstructed from palynological datasets across the OAE2 does not support the idea of widespread heat-induced die-off in continental vegetation-at least not in the hinterland of the studied SPB. Despite maximum TEX 86 -derived SSTs of up to 38°C reconstructed for the SPB, the hinterland supported a rich and diverse flora, which is roughly similar in composition to the type of vegetation thriving in Late Cenomanian to Early Turonian times in other parts of the palaeo-European archipelago 36,37 . Mid-latitude terrestrial plant ecosystems were apparently well adapted to the exceptional warm conditions prevailing during the Cretaceous thermal maximum. The potential effects of temperature extremes might have been more aggravated in continental interior regions of the subtropics and close to the equator, where even higher mean annual surface temperatures are to be expected 38 . Despite the absence of major plant extinction associated with OAE2, variations in the relative contribution of individual taxa and groups reflect compositional changes in vegetation. High contents of Araucariaceae and Cupressaceae-derived pollen indicate the presence of forest communities composed of large trees forming a dense emergent cover 39,40 . An arborescent habit of Cretaceous araucarioids is supported by abundant fossil wood finds from midlatitudes 41 and in line with the growth habit of modern Araucaria types 42 . Together with high pollen abundances produced by Pinaceae, Podocarpaceae and xerophytic Cheirolepidiceae, this type of assemblage (AZ II, IV, and VI) is interpreted to reflect mixed forest dominated by arborescent conifers with an only moderate angiosperm component, which grew in mesic habitats with moderate availability of moisture. Such conifer-dominated ecosystems prevailed during episodes of increased δ 13 C values including the first build-up (peak a, AZ II), the second build-up (peak b, AZ IV) and upper plateau phase (AZ VI) of the CIE. During the trough-shaped CIE segment and between peaks b and c, conifer forests were replaced by more diverse, angiosperm-rich vegetation with increased contents of Normapolles-derived pollen (AZ III and V). Late Cenomanian taxa assigned to Atlantopollis and Complexiopollis are the first representatives of the Normapolles complex 36,43 , an angiosperm group, which dominated mid-latitude northern hemisphere assemblages during the Late Cretaceous and Early Cenozoic. Based on in-situ findings in fossil flowers, pollen of the Normapolles-type is related to core Fagales 44 . The thick, often multi-layered exine and complex apertures indicate anemophily 45 and are considered adaptions to open vegetation with widely spaced trees under more dry to seasonally dry climates 46 . During the late Cenomanian, Normapolles-producing angiosperms were probably small nonwoody plants or shrubs forming part of a xerophytic savanna-type Fluctuations in the overall climatic patterns during OAE2 have been reported earlier based on marine-derived proxy data. Exceptional climatic warmth associated with OAE2 is considered to have caused accelerated hydrological cycling, enhanced weathering and a generally more humid climate 9,50-52 . Phases of enhanced moisture availability may have fostered the spread of mesic conifer-dominated forest ecosystems during the onset and later phase of OAE2. Such warm and humid conditions were punctuated by a series of pronounced climatic coolings during the so-called Plenus Cold Event, PCE 53,54 . The PCE represents a phase of climatic instability and is characterized by several shortlasting SST drops and increases in the range of 2.5-11°C 5-13 and paralleled by the southward migration of boreal fauna 55 . Climatic instability was accompanied by a shift towards significantly drier conditions in northern hemisphere mid-latitudes 9 . Stratigraphically, the main pulse of the PCE is located above the first CIE build-up (peak a) at Eastbourne 54 , where it corresponds approximately to the boundary between the R. cushmani and W. archaeocretacea zones and UC3 to UC4 transition, respectively 56 . At Cassis, the PCE is expressed in a series of several transient drops of TEX 86 -derived SSTs below 32°C with the strongest SST decline occurring within the upper part of the CIE trough interval (Fig. 3), similar to what has been observed elsewhere 5 . This part of the CIE with a shift towards lower δ 13 C values has been related to improved bottom-water oxygenation, enhanced organic carbon remineralization and significant fluctuations in atmospheric pCO 2 13,54,57,58 . In continental mid-latitude settings such as Cassis, this phase of climatic upheaval is characterized by the proliferation of open, savanna-type vegetation rich in angiosperms at the expanse of conifer-dominated forest ecosystems (Fig. 4). Once established, the savanna-type biome was able to persist even during recurrent phases of exceptional warmth in the course of the PCE.
The onset of the Cretaceous thermal maximum including OAE2 took place against the background of a major turnover in global land plant vegetation, namely the rise of angiosperms towards ecological dominance. In mid-latitude settings of North America and Europe, this floral change is marked by the Late Cenomanian radiation of Normapolles-producing plants related to core Fagales 43,45 . In the Cassis record, early representatives of this group show a pronounced increase in the uppermost Cenomanian assemblage zone III, subsequent to their Mid-to Late Cenomanian origination 36 . Within the overall warm and humid greenhouse conditions characterizing OAE2, a transient shift towards relatively drier climates and punctuated cooling during the PCE may have fostered a first spread of Normapolles-type angiosperms, which henceforward became a very diverse and successful group dominating mid-latitude plant ecosystems throughout the Late Cretaceous to Early Cenozoic for more than~50 Ma.  Residues were sieved over a 11-μm mesh and mounted on microscope slides, which were analyzed at ×200 and ×1000 magnification. All samples were productive and studied for their particle content (palynofacies) and sporomorph assemblage (spores and pollen), respectively. For the quantification of the spore-pollen assemblage, a minimum of 250 grains were determined per slide (avg. 272 grains) and one entire slide was scanned for rare elements. Five samples did not provide enough sporomorphs to reach the target count of 250 grains. Variations in the spore-pollen assemblage represent normalized frequencies and are reported in percentage [%] of the total assemblage. Thermally unaltered preservation of organic matter is indicated by the virtually unchanged coloring of the sporomorphs and shows a thermal alteration index (TAI) < 2 59 . Preservation of the individual pollen grains varies from moderate to good. Sporomorphs are assigned to floral groups according to their botanical affinity 60,61 .

Methods
Biomarker analysis. For TEX 86 analysis, a total of 69 powdered and freeze dried samples (5-10 g dry mass) were extracted with an accelerated solvent extractor, using a 9:1 (v/v) dichlormethane (DCM):MeOH solvent mixture, 3 times for 5 min. at a pressure of ca. 7.6 × 10 6 Pa and a temperature of 100°C. The obtained total extracts were rotary evaporated and separated over an activated Al 2 O 3 column using 9:1 (v/v) hexane:DCM and 1:1 (v/v) DCM:MeOH solvent mixtures into an apolar and polar fraction, respectively. The polar fraction, containing the glycerol dialkyl glycerol tetraether lipids (GDGTs), was dried under a pure N 2 flow, dissolved ultrasonically in a 99:1 (v/v) hexane:isopropanol mixture at a concentration of 2 mg/ml and filtered over an 0.45 μm mesh PTFE filter (Ø 4 mm) prior to HPLC/MS analysis. High-performance liquid chromatography/atmospheric pressure chemical ionization mass spectrometry (HPLC/APCI-MS) analyses was be performed on an Agilent 1100 series/Hewlett-Packard 1100 MSD series machine equipped with an auto-injector and HP Chemostation software following 62 . GDGTs were quantified by integration of peak areas and the TetraEther indeX of tetraethers composed of 86 carbon atoms (TEX 86 ) was determined 63 . In order to assess the influence of soil-derived GDGTs on TEX 86 values, the BIT index 64 was applied. Samples with BIT index > 0.3 were excluded from SST reconstruction 65 . Absolute SSTs were reconstructed for 64 samples using TEX 86 H core top calibration equations 66 , which has a calibration error of 2.5°C. The analytical error of TEX 86 H -based SST estimates was 0.07°C based on duplicate analysis. We used the TEX 86 H calibration rather than other calibrations 67 in order to remain consistent with previous literature 6 and as this has recently been shown to match independent temperature estimates in tropical regions during the Eocene hothouse, the warmest period in the Cenozoic 68 . Although the uncertainty of the calibration used is 2.5°C, comparisons with other proxy datasets 6 suggests that TEX 86 sometimes can overestimate SSTs, or have a seasonal bias, e.g., toward summer temperatures. Nevertheless, even considering this, the here presented SST estimates of up to 38°C do suggest hot tropical temperatures well above 30°C. A level-bylevel comparison of TEX 86 values and G/A-ratios based on 42 samples supports a general link between SSTs and gross vegetation patterns (Supplementary Fig. 3).
For compound-specific carbon isotope measurements, apolar fractions were treated using an Ag-Silica column with hexane and subsequently ethyl acetate to gain a mostly n-alkane pure fraction, which was evaporated under N 2 stream followed by dissolution in 25-50 µl hexane. For measuring the δ 13 C composition of individual n-alkanes, 1 ml was injected into the gas chromatograph isotope ratio mass spectrometer (Agilent 6890N GC coupled to a Thermo Delta V advantage IRMS), equipped with a fuse silica capillary column coated with CP-Sil5, with helium used as a carrier gas. The oven was programmed at a starting (injection) temperature of 70°C, which rose to 130°C at 20°/min and then 320°at 4°/min, at which it was maintained for 20 min. Values are expressed in conventional delta notation relative to the Vienna-Pee Dee Formation belemnite (VPDB) international standard, in per mil (‰). All biomarker analyses were carried out at the NIOZ (Royal Netherlands Institute of Sea Research, The Netherlands).

Data availability
All data generated and/or analyzed in this study are included in this published article and its supplementary information file, and are also available from the corresponding author on reasonable request.