Bone histology of acipenseriform fishes reveals seasonality during the final years of the Mesozoic.


 The Cretaceous-Paleogene (KPg) mass extinction ~66 million years ago (Ma) was triggered by the Chicxulub impact on the present-day Yucatán Peninsula. This event caused the extinction of circa 76% of species, including all non-avian dinosaurs, and represents one of the most selective extinctions to date. The timing of the impact and its aftermath have mainly been studied on millennial timescales, leaving the season of the impact unconstrained. Here, we demonstrate that the impact that caused the KPg mass extinction took place during boreal spring. Osteohistology and stable isotope records of exceptionally preserved dermal and perichondrial bones in acipenseriform fishes from the Tanis impact-induced seiche deposits reveal annual cyclicity across the final years of the Cretaceous. These fishes ultimately perished in boreal spring. Annual life cycles, involving seasonal timing and duration of reproduction, feeding, hibernation, and aestivation, vary strongly across latest Cretaceous biotic diversity. We conclude that the timing of the Chicxulub impact in boreal spring significantly influenced selective biotic survival across the KPg boundary.

3D and disclose the osteohistology of acipenseriform bones. These tomographic data show 69 that impact spherules associated with the paddlefish skeleton are exclusively present in its gill 70 rakers 2 and are absent elsewhere in the preserved specimen ( Figure 1). The absence of impact 71 spherules outside the gill rakers demonstrates that they did not yet proceed into the oral cavity 72 or further down the digestive tract, nor that they impacted the fish remains during perimortum 73 exposure. Impact spherule accumulation in the gill rakers and arrival of the seiche must thus 74 have occurred virtually instantaneously 2 , indicating that the acipenseriforms were alive during 75 the bolide impact and the last minutes of the Cretaceous. 76

77
The degree of preservation of sampled acipenseriform bones was assessed using 78 micro X-ray fluorescence (µ-XRF, see Methods; Supplementary information 2), which is 79 capable of revealing potential taphonomic elemental exchange that may have affected the 80 primary stable isotope composition. The µXRF maps show that Fe-and Mn-oxides are present 81 in the bone vascular canals and surrounding sediments, but have not invaded the bone apatite. 82 Detrital sediments, characterised by high concentrations of K and Si, remain restricted to the 83 sediment matrix. The bone apatite conserves a highly homogeneous distribution of P and Ca, 84 which corroborates the unaltered preservation of these apatitic tissues. Skeletal remains of 85 these paddlefishes and sturgeons thus experienced negligible diagenetic alteration, likely due 86 to their rapid burial and possibly aided by early Mn and Fe oxide seam formation 17,18 . The 87 three-dimensional preservation of delicate structures, including non-ossified tissues that 88 originally enveloped the brain ( Supplementary Information 3), further demonstrates the 89 excellent preservation of the fossil and absence of taphonomic reorganisation 19 . 90 Like their modern-day relatives, the latest Maastrichtian paddlefishes of Tanis were 91 filter feeders that presumably consumed copepods and other zooplankton 20,21,22 . These fishes 92 likely experienced an annual feeding pattern, determined by fluctuating food availability, which 93 peaks between spring and autumn 20 .
Paddlefish dentaries consist of perichondrial bone and sturgeon pectoral fin spines 95 consist of dermal bone -an intramembranous skeletal tissue that formed in the mesenchyme 96 (mesodermal embryonic tissue) and both, unlike endochondral bone, did not originate through 97 ossification of cartilaginous precursors 23,24,25 . Fish dentaries are formed by perichondrial 98 ossification around the Meckel's cartilage 26,27 . Both dermal and perichondrial bone growth 99 exclusively proceeds through incremental bone matrix secretion and apposition by a row of 100 osteoblasts 24,25,26,27,28,29 . The cumulative thickness of one annual growth mark spans a thick 101 (favourable) growth zone, a thinner (slowly-deposited) annulus, and ultimately a line of 102 arrested growth (LAG) 28,29,30 . All bones studied here, representing six acipenseriform 103 individuals, exhibit osteocyte lacunar densities towards their periosteal surfaces that 104 consistently remained lower than highest densities recorded during previous years 31 . This 105 indicates bone growth at the time of death that was not yet slowing down (Figure 2 and 106 Supplementary Materials 1). 107 The inferred growth histories are independently corroborated by a stable carbon 108 isotope (δ 13 C) archive that recorded several years of seasonal dietary fluctuations in growing 109 bone. During maximum productivity, the zooplankton is expected to enrich the growing 110 skeleton of filter-feeding fishes with 13 C relative to 12 C 32,33 . As such, the cyclically elevated 111 13 C/ 12 C ratios in paddlefish X-2724 ( Figure 2) reflect distinct episodes of high food availability. 112 Carbon isotope records across the growth record of Paddlefish X-2724 indicates that peak 113 annual growth rate was not yet attained and the feeding season had thus not yet climaxed -114 the conclusive signature of a boreal spring death. 115 The catastrophic end-Cretaceous bolide impact during the Northern Hemisphere 116 boreal spring places the event in a particularly sensitive phase of biological life cycles. For 117 many taxa, annual reproduction and growth take place during spring. A spring impact would 118 therefore have rendered such taxa particularly vulnerable to extinction. This is especially the 119 case for species with small numbers of offspring and/or long incubation times, such as birds 34,35,36 . A spring impact confirms the earlier hypothesis by Wolfe 37 who presented 122 palaeobotanical indications for a June impact. The last boreal spring of the Cretaceous quickly 123 gave way to an impact winter, as large amounts of dust and aerosols reduced solar irradiation 124 for decades 38 . Ecological networks collapsed from the bottom up, as floral extinctions directly 125 affected species dependent on primary producers more so than animals that were able to 126 explore alternative resources, such as birds 39,40 . Alongside e.g. the end-Cretaceous 127 distributions of ecological flexibility and metabolic strategy, the timing of the impact therefore 128 potentially contributed to the still poorly understood patterns of selective survival across the 129

KPg boundary 130
The seasonal timing of the Chicxulub impact will aid in explaining the selectivity of 131 the KPg extinction. We further posit that a spring impact will have affected biota in the Northern 132 Hemisphere to a greater extent than an autumn impact in the Southern Hemisphere. This may 133 have caused asymmetry in extinction and recovery patterns between the two hemispheres, 134 partially due to the unequal distributions of terrestrial and marine environments. Although the 135 Northern Hemisphere accommodated most of the terrestrial biota, the differential impact of the 136 extinction on Northern vs. Southern hemisphere ecosystems remains underexplored. The 137 diverse Latest Cretaceous ecosystem recorded at the uniquely-constrained Tanis site will help 138 in reconstructing the environmental, climatological, and biological conditions that locally 139 prevailed when the Mesozoic terminated. 140 141 References 142 3 Smit, Jan, and Jan Hertogen. An extraterrestrial event at the Cretaceous-Tertiary 147 boundary. Nature 285, 198-200 (1980). hemisphere. Geol. Soc., London, Spec. Pub., 181, 43-54 (2000). York, USA, 196-229 (1989). The excavation for this study took place from August 10 until August 20, 2017, at the Tanis 237 locality, in southwestern North Dakota. In a team lead by Robert DePalma, we excavated the 238 acipenseriform fishes. Sections of dentaries of paddlefishes and pectoral fin-spines of 239 sturgeons were collected in the field for the histological study.

Thin sectioning 241
The majority of the samples were separated from the sediment matrix. This included all 242 sturgeon pectoral fin spine samples: X-2743M, X-2744M, MDX-3 and one of the paddlefish 243 dentaries (X-2724). Paddlefish dentary samples X-2733a and X-2733b were fractured. Their 244 exact orientation was unknown and the material was highly unstable. To avoid further fracturing 245 the specimens were embedded in an epoxy resin prior to thin sectioning. All specimens were 246 cut with a diamond saw and polished to obtain the thin sections (~50-micrometer thickness) 247 and thick-sections for micromilling (~200-micrometer thickness). 248

Micro X-ray Fluorescence 249
Fragments of the paddlefish and sturgeon samples that remained after thin sectioning have 250 been analyzed by Micro X-ray Fluorescence (μXRF) for elemental mapping at 20 μm/5 251 milliseconds at 50 kV 600μA 2 spectrometers Rh (rhodium) with no filter at the Vrije Universiteit 252 Brussel (VUB). 253

Micromill 254
The growth increments were sampled from the thick sections as accurate as possible using a 255 Micromill (Merkantek). Drill transects were assigned in the accompanying software and per 256 transect every individual sample was collected and the drill bit was cleaned with ethanol. Not 257 all thick-sections were suitable for the micromilling, the lobed structure of the sturgeon 258 specimens proved too curved for the micromill, and paddlefish specimens X-2733a and X-259 2733b had only a few growth lines that proved too narrow for micromilling. Sturgeon sample 260 was directed into a cold trap where the CO2 of the sample was frozen with liquid nitrogen for 2 266 minutes. After this 2-minute trapping, an accurate low-amplitude measurement was performed 267 with a Thermo Finnigan Delta + mass spectrometer connected to a Thermo Finnigan 268 GasBench II at the Earth Sciences Stable Isotope Laboratory (Vrije Universiteit, Amsterdam). scintillator optically coupled to a PCO edge 4.2 CLHS sCMOS camera. The resulting average 282 detected energy was then 132 keV. In order to obtain some propagation phase contrast, the 283 distance between the sample and the detector was set at 5m. 205 scans, each of 5000 284 projections of 7ms were performed with a vertical displacement of 1.4mm for a vertical field of 285 view of 2.8mm, ensuring a double scan of the complete samples. Scans were performed in 286 half-acquisition mode to enlarge the lateral field of view. The volume was reconstructed using 287 single distance phase retrieval algorithm coupled to filtered back-projection as implemented in 288 the ESRF software PyHST2. Vertical concatenation, 16 bits conversion and ring artefacts 289 corrections were performed using inhouse developed matlab scripts. Subsequently the gill 290 region and impact spherules were scanned at 13.67 μm voxel size (0.4mm of molybdenum, 291 6mm of copper, LuAG:Ce scintillator of 500 microns, leading to detected energy of 166 keV, propagation distance of 2.5m, two columns of 77 scans each of 4998 projections in half-293 acquisition with exposure time of 0.05s that were laterally concatenated after reconstruction). 294 Furthermore, samples from the paddlefish dentaries and sturgeon pectoral fin spines were 295 scanned at 4.35 µm voxel size for osteohistological analysis (3.5mm of Al, 11 bars of 5mm in 296 diameter of Al, LuAG:Ce scintillator fo 500 µm, leading to a detected energy of 92 keV, 297 propagation distance of 1.5m, one single column of 22 scans each of 4998 projections in half-298 acquisition with exposure time of 60 ms). Subsequent digital extraction of the bones and 299 impact spherules were performed on VG Studio MAX 3.2. VG-studio MAX 3.2 additionally 300 allowed for the creation of an artificial thin-section of the histological samples by using 'thick 301 slab-mode', which can be created with the maximum, minimum, or average grey-values 302 combined. These thin sections were produced at 0.1 mm thickness on average grey-values 303 after careful 3D alignment of the volume in order to maximize the visibility of the annulus and 304 lines of arrested growth (LAGs) and at 0.2 mm thickness on minimum grey-values in order to 305 maximize the visibility of the osteocyte lacunae. All scanning data is available at the 306 http://paleo.esrf.eu database.