The response of water column and sedimentary environments to the advent of the Messinian salinity crisis: insights from an onshore deep-water section (Govone, NW Italy)

Abstract During Messinian time, the Mediterranean underwent hydrological modifications culminating 5.97 Ma ago with the Messinian salinity crisis (MSC). Evaporite deposition and alleged annihilation of most marine eukaryotes were taken as evidence of the establishment of basin-wide hypersalinity followed by desiccation. However, the palaeoenvironmental conditions during the MSC are still a matter of debate, chiefly because most of its sedimentary record is buried below the abyssal plains of the present-day Mediterranean Sea. To shed light on environmental change at the advent and during the early phase of the MSC, we investigated the Govone section from the Piedmont Basin (NW Italy) using a multidisciplinary approach (organic geochemical, petrographic, and carbon and oxygen stable isotope analyses). The Govone section archives the onset of the crisis in a succession of organic-rich shales and dolomite-rich marls. The MSC part of the succession represents the deep-water equivalent of sulphate evaporites deposited at the basin margins during the first phase of the crisis. Our study reveals that the onset of the MSC was marked by the intensification of water-column stratification, rather than the establishment of widespread hypersaline conditions. A chemocline divided the water column into an oxygen-depleted, denser and more saline bottom layer and an oxygenated, upper seawater layer influenced by freshwater inflow. Vertical oscillations of the chemocline controlled the stratigraphic architecture of the sediments pertaining to the first stage of the MSC. Accordingly, temporal and spatial changes of water masses with different redox chemistries must be considered when interpreting the MSC event.


Introduction
During Messinian time (7.25-5.33 Ma) the Mediterranean area underwent progressive environmental and hydrological changes, which culminated c. 5.97 Ma ago with the Messinian salinity crisis (MSC; Hsü et al. 1973;Roveri et al. 2014;Camerlenghi & Aloisi, 2020). During this event, the Mediterranean Sea turned into the youngest Salt Giant of Earth history (Warren, 2010) as a response of its isolation from the global ocean caused by the restriction of the Mediterranean-Atlantic gateways (Flecker et al. 2015;Krijgsman et al. 2018;Capella et al. 2019). One of the most striking effects of the hydrological changes that affected the basin was the intensification of water-column stratification and consequent development of bottom-water anoxia starting at c. 6.7 Ma . The deterioration of palaeoenvironmental conditions was recorded by the deposition of precession-paced alternations of organic-rich shales, diatomites and marls Sierro et al. 1999;Krijgsman et al. 2002). Starting from 5.97 Ma (Manzi et al. 2013), the shale-diatomite-marl successions were replaced at the basin margins by couplets of sulphate evaporites and organic-rich shales of the Primary Lower Gypsum unit (PLG; Roveri et al. 2008), marking the first stage of the MSC (5.97-5.60 Ma; Roveri et al. 2014). The deposition of evaporites has been considered as compelling evidence of the development of hypersaline conditions, which resulted in the demise of most marine eukaryotes (e.g. Bellanca et al. 2001), but promoted the rise of halophilic prokaryotes (Turich & Freeman, 2011;Birgel et al. 2014). The PLG unit passes laterally in intermediateto deep-water settings (> 200 m depth) into evaporite-free successions, composed of shales alternating with carbonates and/or dolomite-rich marls barren of calcareous microfossils (Manzi et al. 2007Dela Pierre et al. 2011;Roveri et al. 2014;Natalicchio et al. 2019).
After the formulation of the 'deep desiccated basin model' in the 1970s (Hsü et al. 1973), the palaeoenvironmental conditions of the Mediterranean water column and seafloor during the MSC are still a matter of debate. The scarcity or lack of body fossils in MSC sediments, the lack of modern analogues for the Messinian evaporites and the inaccessibility of the offshore evaporites have hampered the development of a widely accepted scenario (Rouchy & Caruso, 2006;Ryan, 2009;Roveri et al. 2014;Camerlenghi & Aloisi, 2020).
The analysis of molecular fossils (lipid biomarkers) preserved in Messinian evaporites and their intermediate-to deep-water lateral equivalents provides fundamental information for the reconstruction of the Mediterranean hydrological cycle and palaeoenvironments (Vasiliev et al. 2017;Natalicchio et al. 2019;Sabino et al. 2020). This approach is a valuable tool to unveil the conditions in the water column during the MSC. Organic geochemical investigations targeting abyssal evaporites (Christeleit et al. 2015) and Messinian salts (Isaji et al. 2019b) revealed that evaporites were deposited under a stratified water column during the latest MSC phases. These results agree with recent hydrological models (de Lange & Krijgsman, 2010;Simon & Meijer, 2017;García-Veigas et al. 2018). In contrast, only a few geochemical data are available for the sediments deposited during the early stages of the MSC (Kenig et al. 1995;Sinninghe Damsté et al. 1995b;Isaji et al. 2019a). These studies all describe organic-rich shales of the PLG unit, depicting a stratified basin that received freshwater from rivers (Natalicchio et al. 2017Sabino et al. 2020) and/or low-salinity water from the Paratethys (Grothe et al. 2020).
To shed new light on the response of the water column and sediments to the advent of the MSC, we studied sedimentary strata exposed in the Govone section (Piedmont Basin, NW Italy). In this section, the onset of the crisis is archived in a sequence of organicrich shales and marls, representing the deep-water equivalents of primary sulphate evaporites deposited at the basin margins (Gennari et al. 2020;Sabino et al. 2020). These sediments have recently been investigated to reconstruct the palaeoclimate and palaeohydrologic variability in the northern Mediterranean across the onset of the MSC (Sabino et al. 2020). This previous study, which was based on inorganic geochemical proxies and on carbon and hydrogen stable isotope composition of lipids from terrestrial plant waxes, revealed fluctuations between more humid (shales) and more arid (marls) climates and an evolution towards moister conditions after the onset of the MSC (Sabino et al. 2020).
Here, we focus on the same succession studied by Sabino et al. (2020), integrating stratigraphic, petrographic, carbon and oxygen stable isotope analyses, and the study of molecular fossils to reconstruct the palaeoenvironments. This approach allows us to reconstruct the environmental changes across the onset of the MSC, and reveals how these changes influenced the stratigraphic architecture of the sediments deposited during times of change in the Mediterranean realm.

2.a. The Messinian succession in the Piedmont Basin
The Piedmont Basin (NW Italy; Fig. 1a) is a wedge-top basin located on the inner side of the SW Alpine arc filled with upper Eocene -Messinian sediments (Rossi et al. 2009;Mosca et al. 2010). The Messinian succession is exposed on the uplifted southern and northern margins of the basin and starts with outer shelf to slope marls and shales (Sant'Agata Fossili Marls; Tortonianlower Messinian; Sturani & Sampò, 1973). These deposits, characterized by the repetition of shale and marl couplets, record progressively more restricted conditions heralding the onset of the MSC (Sturani, 1973;Sturani & Sampò, 1973). The lithological cyclicity was controlled by climate fluctuations driven by precession, with shales representing more humid conditions at precession minima and marls recording more arid conditions at precession maxima Sabino et al. 2020). At the basin margins, this unit is overlain by shale and gypsum couplets belonging to the PLG unit (Fig. 1b). In settings of intermediate water depth, the gypsum beds of the lowermost PLG cycles are transitional to carbonate-rich layers and marls hosting fossilized microbial mats of putative sulphide-oxidizing bacteria, in turn passing in deep-water settings into dolomite-rich marls Natalicchio et al. 2017Natalicchio et al. , 2019Sabino et al. 2020;Fig. 1c). In the basin depocentre, only shales and marls are found (Irace et al. 2009). The PLG unit and its deeper-water equivalents are overlain by chaotic and clastic gypsum facies (Valle Versa Chaotic complex; Irace et al. 2005;Dela Pierre et al. 2007), which are correlated to the Resedimented Lower Gypsum unit deposited during the second stage of the MSC (5.60-5.55 Ma; Roveri et al. 2014). The Messinian succession is completed by fluvio-deltaic and lacustrine sediments, referred to as the Cassano Spinola Conglomerates (Sturani, 1976;Dela Pierre et al. 2011Fig. 1c), recording the third stage of the MSC (5.55-5.33 Ma; Roveri et al. 2014).
The age model for the Govone section (Gennari et al. 2020) had already been adopted by Sabino et al. (2020). The position of the MSC onset (5.97 Ma; Manzi et al. 2013) was defined through the identification of diagnostic planktonic foraminifer bioevents, the ages of which were calibrated through the correlation with the astronomically tuned Perales section (Sierro et al. , 2003Manzi et al. 2013;Fig. 2). The main bioevents identified by Gennari et al. (2020) are: (1) the first abundant occurrence (FAO) of Turborotalita multiloba in cycle Gm12, occurring in cycle UA15 in Perales and dated 6.415 Ma ); (2) the left/right coiling change of Neogloboquadrina acostaensis in cycle Gm14, identified in Perales at cycle UA17 and dated 6.36 Ma ; (3) the first influx of Globorotalia scitula in cycle Gm17, dated 6.29 Ma in the Perales section (cycle UA20; Sierro et al. 2001); and (4) a second influx of Globorotalia scitula in cycle Gm24, which falls within an acme interval of T. multiloba. For the Perales section, Sierro et al. (2001) reported this acme between cycles UA23 and UA30, allowing Gennari et al. (2020) to correlate the second influx of G. scitula in the Govone section with the second influx recognized in cycle UA29 in the Perales section and dated at 6.10 Ma. Since the second influx of G. scitula occurs six precession cycles below the onset of the crisis in the Perales section , the MSC onset in Govone was placed at the base According to the adopted model, the uppermost six Sant'Agata Fossili Marls cycles below the first local gypsum bed (Gg1 cycle) represent the deep-water counterparts of the lowermost PLG cycles (Fig. 2), with marls representing the time equivalents of the shallow-water marginal gypsum (Gennari et al. 2020;Sabino et al. 2020). In this study we investigate four pre-MSC (Gm26-Gm29) and four MSC (Gm30-Gm33) cycles, representing a time interval of c. 150 ka. Younger strata of the Govone section containing gypsum (Fig. 1c) are not studied here.

3.a. Petrography and mineralogy
A total of 33 fresh, unweathered samples were excavated from cycles Gm26 to Gm33 (on average, 4 samples per cycle); the Govone sedimentary rocks tend to be well preserved and continuously exposed due to ongoing erosion in the river bed. A total of 18 slabs and thin-sections were obtained from representative samples (10 from pre-MSC cycles and 8 from MSC cycles) and studied at the Department of Earth Sciences, University of Turin by optical (transmitted, reflected and UV light) and scanning electron microscopy (SEM). SEM analyses were performed on carboncoated stubs for morphological analyses and on carbon-coated, polished thin-sections for semi-quantitative elemental analysis and backscattered electron imagery, using a JEOL JSM IT300LV scanning electron microscope equipped with an energy-dispersive EDS Oxford Instrument Link System microprobe (University of Turin).
X-ray diffraction (XRD) analyses were performed on all 33 samples using a Panalytical X'Pert PRO diffractometer (CuKα radiation, 40 kV, 40 mA, step size 0.0167, 5 s per step) at the Department of Geodynamics and Sedimentology, University of Vienna. The samples were loaded in the sample holders as oriented powders. The X-ray diffraction patterns were interpreted using the Panalytical software 'X'Pert High score plus' to determine the carbonate phase mineralogy.

3.b. Carbon and oxygen stable isotopes
Carbon (δ 13 C) and oxygen (δ 18 O) bulk-rock stable isotope analyses were performed on 9 samples (pre-MSC cycles, 1 sample; MSC cycles, 8 samples). These analyses add to those of Bernardi (2013) who analysed 20 samples, 12 from pre-MSC cycles and 8 from MSC cycles. Analyses were performed at the MARUM stable isotope laboratory (University of Bremen). The 9 new samples were measured at 75°C on a Finnigan MAT 252 gas isotope ratio mass spectrometer connected to a Kiel III automated carbonate preparation device. The instrument was calibrated against an inhouse standard (ground Solnhofen limestone), which in turn was calibrated against the NBS 19 calcite standard. Over the measurement period, the standard deviations of the in-house standard were 0.04‰ for δ 13 C and 0.06‰ for δ 18 O values. Data are reported in delta-notation versus V-PDB. When dolomite was the only carbonate phase, the δ 18 O values were corrected for −0.8‰ for Environmental change at the onset of the Messinian salinity crisis

3.c. Total inorganic and organic carbon contents
Total inorganic (TIC) and organic (TOC) carbon contents were measured at the Institute for Geology of the University of Hamburg. After drying (at 50°C for 24 h), the samples were manually ground with an agate mortar. The powders were split in two aliquots: one fraction was heated to 1350°C and total carbon (TC) contents were measured using a LECO SC-144DR Carbon Analyser equipped with an infrared detector. The second fraction was first heated to 550°C for 5 h to remove organic carbon (OC) and then heated to 1350°C to measure TIC contents. Prior to and after sample analyses, a Synthetic Carbon Leco 502-029 (1.01 ± 0.02 carbon%) standard was measured. TOC contents were determined using the formula TOC = TC-TIC.

3.d. Lipid biomarker analyses
Lipid biomarker analyses were performed on 21 samples (at least 2 samples per cycle) using the procedure described in Sabino et al. (2020). Briefly, after a modified Bligh and Dyer extraction, samples were separated in an n-hexane-soluble and a dichloromethanesoluble fraction. The former was further separated into four sub-fractions: (a) hydrocarbons, (b) ketones, (c) alcohols and (d) carboxylic acids. The alcohol fraction was derivatized for 1 h at 70°C by adding pyridine and N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) (1:1; v/v). The ketones and carboxylic acids did not contain any indigenous compounds and are not discussed further. Alcohol and hydrocarbon fractions were dried and redissolved in n-hexane for analyses. Compounds were identified using a Thermo Scientific Trace gas chromatograph (GC) Ultra coupled to a Thermo Scientific DSQ II mass spectrometer (MS) through comparison of retention times and published mass spectra. Quantification was achieved with a Fisons Instruments GC 8000 series equipped with a flame-ionization detector (FID). Internal standards used for quantification were 5α-cholestane for the hydrocarbon fraction and 1-nonadecanol and DAGE C 18-18 for the alcohol fraction. The carrier gases were helium and hydrogen for the GC-MS and GC-FID analyses, respectively. Both devices were equipped with an Agilent HP-5MS UI fused silica column with a length of 30 m, a diameter of 0.25 mm and a film thickness of 0.25 μm. The GC temperature programme was: 50°C (3 min); from 50°C to 230°C (held 2 min) at 25°C/min; then from 230°C to 320°C (held 20 min) at 6°C/min. An aliquot of the hydrocarbon fraction was used to isolate branched and cyclic compounds from the n-alkanes. The aliquot was treated with 300 mg of 5 Å molecular sieve and 2 mL cyclohexane, was shaken well, then extracted by ultrasonication for 2 h. The extract was filtered and the mole sieve was repeatedly washed with cyclohexane and filtered once again to release all branched and cyclic compounds. After drying with molecular nitrogen, the extract was re-dissolved in n-hexane for compound identification and quantification. The GC temperature program was: 50°C (3 min); from 50°C to 230°C (held 2 min) at 25°C/min; then from 230°C to 325°C (held 25 min) at 6°C/min.

3.d.1. Desulphurization with nickel boride
Desulphurization was performed for the dichloromethane-soluble fractions (asphaltenes), applying a procedure slightly modified from Schouten et al. (1993) and Blumenberg et al. (2010). Briefly, the asphaltenes were dissolved in 8 mL tetrahydrofuran/ methanol (1:1, v/v), then 200 mg of each anhydrous nickel chloride and sodium borohydride were added. After 1 h reaction time, the samples were centrifuged and the supernatant was collected. The solid residue was extracted twice with dichloromethane/methanol (1:1, v/v), then centrifuged again and the supernatant was combined with the previous extract. An aqueous solution was added and the organic layer collected and dried with a rotary evaporator. The residual organic phase was filtered through dry sodium sulphate to completely remove water. Column chromatography was performed using a silica gel column and 2.5 mL of n-hexane/ dichloromethane (9:1, v/v) as eluent to separate the released hydrocarbons after desulphurization from the polar compounds. The desulphurized hydrocarbon fraction was re-dissolved in n-hexane and the compounds identified through GC-MS analyses using the following program: from 60°C (1 min) to 150°C at 15°C/min and then up to 320°C (held 40 min) at 4°C/min. Due to low contents, the desulphurized hydrocarbon fractions were also run in single ion mode (SIM) with the masses m/z 133 and 546, specific masses of di-aromatic carotenoids and isorenieratane, respectively.

3.d.2. Glycerol dibiphytanyl glycerol tetraether analyses
Glycerol dibiphytanyl glycerol tetraethers (GDGTs) were obtained from 10% vol. of the total lipid extract (TLE) according to the method of Hopmans et al. (2000) and modified by Baumann et al. (2018). The aliquot was dissolved in n-hexane and an internal standard (C 46 GDGT; 12 mg/L) was added. The analyses were performed using a Varian MS Workstation 6.91 high-performance liquid chromatography (HPLC) system coupled to a Varian 1200 L triple quadrupole mass spectrometer. The compounds were separated on a Grace Prevail Cyano column (150 × 2.1 mm; 3 μm particle size) and a guard column, held at 30°C. The following gradient was applied: linear change from 97.5% A (100% n-hexane) and 2.5% B (90% n-hexane: 10% 2-propanol; v/v) to 75% A and 25% B from 0 to 35 min; then linearly to 100% B in 5 min and held for 8 min; and thereafter, back to 97.5% A and 2.5% B to re-equilibrate the column for 12 min. The total run time was 60 min and the solvent flow was kept constant at 0.3 mL/min during the entire run time. The identification of GDGTs was achieved using a mass spectrometer equipped with an atmospheric pressure chemical ionization (APCI) interface operated in positive ion mode. The APCI parameters were: molecular nitrogen as nebulizing gas with a pressure of 60 psi; temperature fluctuating between 35°C and 40°C; 50°C as API housing temperature; 200°C for the drying gas of the API with a pressure of 12 psi; and 400°C and 18 psi for the APCI auxiliary gas temperature and pressure, respectively. The injection volume was 10 μL. The scanned spectral range was set at m/z 500 to 750 and 950 to 1500. The response factors were evaluated after every 4 to 5 samples using a standard mixture containing synthetic archaeol (1,2-Di-O-phytanyl-sn-glycerol; CAS 99341-19-2), DAGE C 18:18 (CAS 6076-38-6), DAGE C 18:18-4ene (1,3-Dilinoleoyl-rac-glycerol; CAS 15818-46-9) and synthetic C 46 GDGT (CAS 138456-87-8). The response factors between synthetic archaeol and C 46 GDGT were usually around 1.5:1. Relative lipid abundances were determined by selecting individual base peaks with the target mass m/z, including ions symmetrically with ±1.0 of target m/z. Environmental change at the onset of the Messinian salinity crisis 4. Results

4.a. Petrography and mineralogy
The studied interval is characterized by 8 lithological cycles of shales and argillaceous marls (referred to as marls) up to 1 m thick (Fig. 3a, b). The inorganic carbon contents range from 0.4% to 6.0%, with higher contents in marls (Fig. 3a). XRD analyses indicate that the inorganic carbon in the Govone section is composed of calcite and dolomite (see online Supplementary Material, available at http://journals.cambridge.org/geo). Dolomite coexists with calcite in cycles Gm26 and Gm27, while it represents the only carbonate phase in cycle Gm28 (except for sample Gm28.1) and above (Fig. 3a).

4.a.1. Shales
Shales are laminated, dark-green-coloured (Fig. 3c) and rich in organic matter (TOC > 2%). They are characterized by an alternation of discontinuous, sub-millimetric, cream-coloured (type A) and dark-grey-coloured (type B) laminae, irregularly interrupted by layers up to 1.5 cm thick composed of terrigenous silt-to sand-sized grains (Fig. 4a). These layers are considered to represent the products of more intense fluvial discharge to the basin (Dela . Type A laminae show a distinct bulbous to pinch-and-swelltype character ( Fig. 4a; sensu Pilskaln & Pike, 2001). They are formed of two varieties of fluorescent peloids: (1) flattened to lens-shaped peloids up to 0.5 mm across with well-defined boundaries; and (2) irregular aggregates, up to 1 mm across, with irregular, diffused boundaries (Fig. 4a-c). In cycles Gm26 and Gm27, the peloids are composed of scattered silt-sized siliciclastic grains (quartz, mica flakes and feldspars), black organic matter particles and coccoliths, very abundant in the lens-shaped peloids (Fig. 4d). In cycles Gm28 to Gm33, peloids are barren of coccoliths and mostly formed of dolomite microcrystals up to 10 μm across, scattered pyrite framboids and accessory silt-sized siliciclastic grains (Fig. 4e). Dolomite microcrystals show prevalently globular, hemispherical or cauliflower-like (sensu Warthmann et al. 2000) habits ( Fig. 4e-g), and sometimes exhibit a central hollow (Fig. 4g).
Type B laminae are only a few microns thick, laterally discontinuous (Fig. 4a) and composed of silt-to clay-sized terrigenous grains (mainly quartz, feldspars and mica flakes), black amorphous organic matter debris enriched in sulphur (as revealed by EDS analyses) and abundant pyrite framboids, up to 10 μm across. These laminae are interpreted to represent the product of enhanced terrigenous input into the basin, under the control of short-term, probably seasonal, climate change (Dela . Both types of laminae contain scattered planktonic foraminifer tests (only in cycles Gm26 and Gm27; Fig. 4h, i), fish scales and vertebrae. Tiny (<10 μm) pyrite framboids (Fig. 4f-h), locally filling foraminifer tests (Fig. 4i), are abundant.

4.c. Total organic carbon contents and lipid biomarkers
The TOC contents vary from 1.0% to 3.1% ( Fig. 6; online Supplementary Material, available at http://journals.cambridge. org/geo) and follow lithological cyclicity, with higher contents in shales (average 2.4%) and lower contents in marls (average 1.6%), particularly in pre-MSC cycles (as low as 1.0%).
Lycopane, isolated with a molecular sieve from the n-alkanes, was found in most of the 18 samples (except for marls in cycle Gm26), yielding varying contents (up to 25.0 μg/g TOC).

830
M Sabino et al.

832
M Sabino et al. Damsté et al. 2003), although petrographic and other organic geochemical evidence suggests the opposite (see Section 5.c below). This inconsistency most likely results from the high flux of leaf-wax-derived long-chain n-alkanes relative to lycopane, which can significantly affect the ratio and make the use of its absolute values unreliable (cf. Sinninghe Damsté et al. 2003). We therefore decided not to apply the absolute values of the ratio, as suggested by Sinninghe Damsté et al. (2003), but refer to trends between the lithologies to infer relative variations in palaeo-oxicity. Tetrahymanol occurs in all samples with contents ranging from 1.1 to 26.5 μg/g TOC ( Fig. 6; online Supplementary Material, available at http://journals.cambridge.org/geo). Tetrahymanol is more abundant in shales than in marls, although the difference between the two lithologies is less pronounced in the MSC interval than in the pre-MSC part of the section (Fig. 6

4.c.3. Hydrocarbons liberated after desulphurization
The desulphurization of asphaltenes released mostly n-alkanes (C 14 to C 40 ), a pentacyclic C 30 sulphide (cf. Poinsot et al. 1998) and, especially in shales, phytane. As well as these compounds, we identified C 27 to C 29 steranes, C 31 to C 35 hopanes, and, only in samples from cycles Gm28 to Gm30, the diaromatic carotenoid isorenieratane. In most cases (with the exception of sample Gm30), isorenieratane was identified in single ion mode (SIM) only (m/z 133 and 546). This is in accordance with very low contents, rendering quantification impossible.

5.a. Water-column stratification and euxinia across the onset of the MSC
The restriction of the Mediterranean leading to the MSC was associated with an intensification of water-column stratification ) that persisted during the entire MSC event (Christeleit et al. 2015;Simon & Meijer, 2017;García-Veigas et al. 2018). The same evolution has been reconstructed for the Piedmont Basin (Dela Pierre et al. 2011;Bernardi, 2013;Violanti et al. 2013;Lozar et al. 2018), especially by the study of the Pollenzo section, a section about 20 km SW from the Govone section (Natalicchio et al. 2017. For the Pollenzo section, the most compelling evidence of increasing water-column stratification after the MSC onset is the appearance of tetrahymanol. This compound is produced by organisms commonly thriving at chemoclines (Wakeham et al. 2012), namely bacterivorous ciliates, anoxygenic phototrophic bacteria and aerobic methanotrophic bacteria (Kleemann et al. 1990;Harvey & McManus, 1991;Rashby et al. 2007;Eickhoff et al. 2013;Banta et al. 2015;Cordova-Gonzales et al. 2020); tetrahymanol is therefore considered a robust indicator of water-column stratification (e.g. Sinninghe Damsté et al. 1995b).
Tetrahymanol was found throughout the section at Govone, suggesting that, in the more distal settings of the Piedmont Basin, stratified conditions had already been established before the onset of the MSC. The higher tetrahymanol contents in shales suggest that stratification was more intense at precession minima, most likely in response of enhanced riverine runoff (cf. Natalicchio et al. 2019;Sabino et al. 2020). Enhanced riverine runoff in turn promoted episodes of high productivity and phytoplankton blooms. The latter are testified by the very abundant lens-shaped peloids and irregular aggregates, which have been interpreted as faecal pellets and marine snow-flakes, respectively (cf. Dela , both main components of so-called marine snow (e.g. Alldredge & Silver, 1988;Alldredge et al. 2002). Interestingly, in the shales of the uppermost pre-MSC cycles (Gm28-30), tetrahymanol is accompanied by isorenieratane, derived from the degradation of isorenieratene sourced by green sulphur bacteria (Repeta et al. 1989;Van Gemerden & Mas, 1995). These phototrophic bacteria require hydrogen sulphide to perform anoxygenic photosynthesis (e.g. Van Gemerden & Mas, 1995) and are found in modern stratified basins when sulphidic conditions extend into the photic zone (e.g. Black Sea, Sinninghe Wakeham et al. 2007;Cariaco Basin, Wakeham et al. 2012), making isorenieratene and its derivatives biomarkers for photic zone euxinia (e.g. Repeta & Simpson, 1991;Kuypers et al. 2002). The co-occurrence of tetrahymanol and isorenieratane in shales therefore supports pronounced water-column stratification at Environmental change at the onset of the Messinian salinity crisis precession minima already before the onset of the MSC, occasionally accompained by photic zone euxinia. The lower contents of tetrahymanol and the lack of isorenieratane in the MSC shales apparently suggest the weakening of stratification, which is however in contrast with evidence from the neighbouring Pollenzo section (Natalicchio et al. 2017. This inconsistency could reflect unfavourable environmental conditions at precession minima for the producers of isorenieratene and tetrahymanol after the MSC onset in the distal part of the basin (Govone), rather than the weakening of water-column stratification. Interestingly, in modern stratified basins, ciliates are found to feed, among other organisms, on green sulphur bacteria (Wakeham et al. , 2012. The trophic relationship between autotrophic green sulphur bacteria and heterotrophic ciliates, already inferred for the MSC sediments of the Northern Apennines (Vena del Gesso Basin, Sinninghe Damsté et al. 1995a), can potentially explain the trend of compound contents in the Govone section, where tetrahymanol peaks coincide with the presence of isorenieratane. Adverse conditions for green sulphur bacteria were most likely related to an intensified input of terrigenous material by rivers after the MSC onset, as evidenced by increasing contents of terrestrial plant waxes Sabino et al. 2020). Enhanced riverine runoff might have increased turbidity and input of oxygen in the upper water column, both deleterious factors for phototrophic green sulphur bacteria (Van Gemerden & Mas, 1995), in turn causing a decrease of the ciliate population. Evidence of persistent stratification and sulphidic conditions in the water column at precession minima after the MSC onset, although not extending into the photic zone, is provided by the occurrence of small pyrite framboids and sulphur-enriched organic matter. The small size of the super-abundant pyrite framboids (< 10 μm) may suggest that they formed within an euxinic water column (cf. Wilkin et al. 1996;Passier et al. 1997;Bond & Wignall, 2010;Tagliavento et al. 2020). Organic sulphur compounds, such as the pentacyclic C 30 sulphide, are believed to form during a very early diagenetic stage, possibly already in the water column when reduced-sulphur species (e.g. hydrogen sulphide) exceed reduced iron and consequently bind to settling organic matter (Sinninghe Damsté & de Leeuw, 1990;Wakeham et al. 1995;Poinsot et al. 1998).
Compared with shales, marls reveal lower contents of tetrahymanol. Such a pattern reflects the weakening of stratification at precession maxima in response to more efficient mixing of the water column (e.g. Sierro et al. 2001;Violanti et al. 2013). Mixing was promoted by drier climate conditions and reduced fluvial discharge, in accordance with higher element/Al ratios and a decrease of contents in the sediments of terrestrial plant waxes, which show deuterium enrichment (Sabino et al. 2020). However, water-column stratification never completely ceased to exist, since the occurrence of small pyrite framboids and sulphur organic compounds suggests that sulphidic conditions were intermittently present in the water column. The trend to slightly higher tetrahymanol contents in MSC marls compared with pre-MSC marls suggests that water-column stratification was more intense at precession maxima after the onset of the MSC compared with precession maxima during the pre-MSC, in agreement with the appearance of tetrahymanol in time-equivalent more marginal sediments (Pollenzo section; Natalicchio et al. 2017). This was possibly due to the establishment of a wetter climate in the northern Mediterranean (cf. Natalicchio et al. 2019;Sabino et al. 2020).

5.b. Persistence of marine conditions in surface waters after the onset of the MSC
The Govone section is characterized by the progressive decline and final disappearance of calcareous plankton approaching the MSC onset (Bernardi, 2013). Such ecological change is recorded in other sections across the Mediterranean and was taken as evidence of progressively harsher conditions in surface waters, with fluctuations of salinity up to high levels not tolerated by most marine biota (e.g. Sierro et al. 1999;Blanc-Valleron et al. 2002;Manzi et al. 2007). The establishment of hypersaline conditions during the MSC is also reflected by the archaeal di-and tetraether lipid assemblages of MSC carbonates from Sicily and Calabria (the so-called 'Calcare di Base'); in these deposits, C 20-20 archaeol, C 20-25 archaeol (extended archaeol) and caldarchaeol dominate the isoprenoid diphytanyl glycerol diether (DGD) and GDGT inventories (Turich & Freeman, 2011;Birgel et al. 2014).
In contrast, the isoprenoid DGD and GDGT distribution found for the Govone section is significantly different from that of the Table 1. Archaeol, GDGTs (both in % relative to all GDGTs plus archaeol), lycopane and tetrahymanol abundances in pre-MSC and MSC Govone sediments. Max = maximum content, Min = minimum content, GDGT-0 = caldarchaeol, GDGT-5 = crenarchaeol.

Max (%)
Min ( Environmental change at the onset of the Messinian salinity crisis Calcare di Base and does not agree with widespread hypersalinity after the onset of the MSC. In fact, archaeol does not dominate over the total GDGTs (c. 25% of the total GDGT and DGD lipid inventory) and is never accompanied by extended archaeol, allowing the exclusion of a significant contribution from halophilic archaea (cf. Teixidor et al. 1993;Dawson et al. 2012). Possible alternative sources for archaeol are non-halophilic archaea, such as methanogens, methanotrophs and marine Euryarchaeota thriving in epi-to mesopelagic zones (De Rosa & Gambacorta, 1988;Blumenberg et al. 2004;Turich et al. 2007;Sollai et al. 2019). Methanogenic and methanotrophic archaea, on the other hand, can be excluded for the Govone section, since the distribution of isoprenoidal diand tetraether lipids varies from those observed in methane-seep environments, the latter typified by the dominance of archaeol, hydroxyarchaeol and GDGT-1 and 2 over the other dibiphytanyl tetraethers (De Rosa & Gambacorta, 1988;Blumenberg et al. 2004;Wakeham et al. 2007). Consequently, marine Euryarchaeota thriving in the epi-to mesopelagic zones of a normal marine water column are the most likely source of archaeol (cf. Turich et al. 2007;Schouten et al. 2008;Elling et al. 2015;Sollai et al. 2019).
Interestingly, it has been suggested that mesopelagic marine Euryarchaeota from the group III (MGIII) are important producers of archaeol, especially in oxygen-deficient zones where they can even dominate the archaeal community (Belmar et al. 2011;Sollai et al. 2019). Since episodical oxygen deficiency seems to have characterized the water column of the Piedmont Basin during the MSC, mesopelagic marine Euryarchaeota are the most likely source of archaeol in the Govone section, indicating the persistence of marine rather than hypersaline conditions across the onset of the MSC. Marine conditions are in accordance with the distribution of the GDGTs and the caldarchaeol/crenarchaeol ratio (c. 1), which mirror the distribution and the ratios reported for open-ocean settings (e.g. Turich et al. 2007;Kim et al. 2010;Pearson & Ingalls, 2013;Schouten et al. 2013). Under such conditions, the main contributors to the GDGT pool are planktonic marine Euryarchaeota of the group II (MGII) and Thaumarchaeota (marine group I, MGI), dwelling in the epi-to mesopelagic zones (Schouten et al. 2013;Santoro et al. 2019). Despite some controversy (Lincoln et al. 2014;Schouten et al. 2014), MGII archaea are thought to source mostly acyclic GDGTs (caldarchaeol), whereas MGI archaea synthesize both acyclic and polycyclic GDGTs, especially crenarchaeol (Elling et al. 2017;Zeng et al. 2019). This pattern results in the dominance of caldarchaeol and crenarchaeol over the other GDGTs and a caldarchaeol/crenarchaeol ratio of c. 1 (Sinninghe Schouten et al. 2013 for a review), which closely mirror the tetraether archaeal lipid distribution and the ratio of marine planktonic Thaumarchaeota growing in cultures at normal salinity (c. 35‰ salinity; Elling et al. 2015).
The persistence of marine conditions in surface waters challenges the idea of an establishment of a completely hypersaline water mass early on during the MSC (e.g. Bellanca et al. 2001). The lack of pervasive hypersaline conditions agrees with (1) the presence of marine fossils (foraminifers, calcareous nannoplankton, diatoms, fishes) above the MSC onset in more marginal sections of the Piedmont Basin (Violanti et al. 2013;Dela Pierre et al. 2014;Lozar et al. 2018;Carnevale et al. 2019) and (2) palaeosalinity estimates from gypsum fluid inclusions, indicating that the parent brine had a salinity possibly even lower than seawater . In this light, the disappearance of calcareous plankton in the Govone section two cycles below the onset of the MSC most likely reflects diagenetic dissolution of calcareous skeletons rather than harsh environmental conditions in superficial waters .

5.c. Fluctuations in oxygen concentration of bottom waters across the onset of the MSC
Intensification of water-column stratification in the Mediterranean Basin shortly before the onset of the MSC was associated with widespread depletion of oxygen in bottom waters, which is recorded by the precession-driven deposition of organicrich sediments Krijgsman et al. 2001Krijgsman et al. , 2002 and by the replacement of oxyphilic benthic foraminifera with stress-tolerant taxa (e.g. Kouwenhoven et al. 1999;Blanc-Valleron et al. 2002;Gennari et al. 2018). Such a trend was confirmed for the Piedmont Basin by changes in the assemblages of benthic foraminifers in the pre-MSC sediments (Bernardi, 2013;Violanti et al. 2013). However, very little information is available for the microfossil-poor or microfossil-barren MSC strata.
The Govone section archives the fluctuation of oxygen contents at the seafloor driven by precessional forcing. Humid climate and enhanced riverine runoff at precession minima Sabino et al. 2020) favoured intense water-column stratification and bottom anoxia, which is witnessed by the deposition of laminated, organic-rich shales (TOC > 2%). Evidence of bottomwater anoxia during the deposition of shales includes their high contents of lycopane. Studies of modern Sinninghe Damsté et al. 2003) and ancient sediments (e.g. Freeman et al. 1990;Kenig et al. 1995;Behrooz et al. 2018;Löhr et al. 2018) have shown that preservation of lycopane is highest when bottom waters are anoxic.
Lower TOC contents (mostly < 2%) and the sharp drop of lycopane contents (≤0.4 μg/g TOC) in marls indicate higher oxygen levels at the seafloor than at times of shale deposition, reflecting precession-driven climate change (Sabino et al. 2020). However, changes can be observed across the onset of the MSC, possibly recording different degrees of bottom-water ventilation before and after the advent of the crisis. In particular, pre-MSC marls show bioturbation traces and contain few stress-tolerant benthic foraminifers (Bernardi, 2013;Dela Pierre et al. 2016), suggesting that the seafloor was at least episodically oxygenated. In contrast, the absence of bioturbation and the slight increase of lycopane contents in MSC marls compared with pre-MSC marls indicate that bottom waters became progressively oxygen depleted, agreeing with an intensification of water-column stratification at precession maxima during the MSC relative to precession maxima in pre-MSC times (Section 5.a above). Despite such a trend to lower oxygen levels, the only low lycopane contents at precession maxima compared with precession minima (shale deposition) suggest that bottom waters were not fully anoxic during the deposition of marls (cf. Sinninghe Damsté et al. 2003).

5.d. Primary Lower Gypsum: implications on depositional environments and stratigraphic architecture
The integration of geochemical, sedimentological and petrographic data indicates that during the early phase of the MSC the water column in the Piedmont Basin was stratified and comprised (1) an oxygenated upper layer, typified by mostly marine conditions and receiving input of freshwater from rivers Sabino et al. 2020) and/or low-salinity water from the Paratethys (Grothe et al. 2020); and (2)  The two water masses were separated by a pycnocline, at which chemical gradients established with time and a chemocline developed, as observed for modern basins (Wakeham et al. , 2012 and reconstructed for ancient basins with stratified water masses (e.g. the Badenian basin of Eastern Europe; Babel, 2004;Babel & Bogucki, 2007). Stratigraphic data reveal a lateral facies change in the PLG unit from the margin to the depocentre of the Piedmont Basin for sediments deposited during precession maxima (Dela Pierre et al. 2011). Bottom-grown selenite gypsum is observed only in the shallow, marginal part of the basin (Arnulfi section; Fig. 7). Moving towards the depocentre, gypsum makes lateral transition to marls and carbonates with filamentous fossils (Pollenzo section; Fig. 7)interpreted to represent chemotrophic microbial mats (Dela )and finally to dolomite-rich marls (Govone section; Fig. 7). Conversely, at precession minima, organic-rich shales were deposited across the entire Piedmont Basin (Dela Pierre et al. 2011Pierre et al. , 2016Fig. 7). We suggest that these stratigraphic patterns reflect vertical chemocline oscillations controlled by precession (Fig. 7).
The widespread seafloor anoxia recorded by shales indicates that the chemocline was located above the sea bottom across most of the basin during precession minima. These conditions favoured organic matter preservation in the water column and led to an increased sedimentation of organic matter, which, in turn, favoured heterotrophic, bacterial sulphate reduction and promoted bacterially mediated precipitation of early diagenetic dolomite in anoxic, organic-rich sediments (Fig. 7a). This inference agrees with (1) the negative bulk-rock δ 13 C values (cf. Petrash et al. 2017); (2) the moderate negative correlation (r = −0.7) between TIC contents, represented by dolomite, and bulk-rock δ 13 C values (high TIC contents coincide with low δ 13 C values); and (3) dolomite crystal habits (Figs 4f, g, 5d-f; cf. Warthmann et al. 2000;van Lith et al. 2003;Sanz-Montero et al. 2008;Bontognali et al. 2010;Han et al. 2016).
The change towards a less humid climate at precession maxima (Sabino et al. 2020) resulted in the deepening of the chemocline and the thinning of the anoxic zone (Fig. 7b). Under these circumstances, selenite grew only in the shallower part of the basin (Arnulfi section, Fig. 7b), where low-salinity waters enriched in calcium and sulphate ions, possibly derived from leaching of former evaporites , occurred above an oxygenated sea bottom (e.g. García-Veigas et al. 2018). In more distal settings, the absence of gypsum was possibly related to low-oxygen conditions, which favoured bacterial sulphate reduction and resultant gypsum undersaturation ( Fig. 7b; cf. Babel, 2004;de Lange & Krijgsman, 2010;García-Veigas et al. 2018). The parts of the basin where the chemocline impinged on the seafloor (Pollenzo section, Fig. 7b; Natalicchio et al. 2017) were covered by microbial mats of putative sulphide-oxidizing bacteria (Natalicchio et al. 2017). In modern settings, these bacteria thrive under hypoxic conditions where the chemocline intercepts the seafloor and steep gradients between electron acceptors and hydrogen sulphide occur (e.g. the Black Sea; Jessen et al. 2016). In contrast, in deeper parts of the basin (Govone section; Fig. 7b) the sea bottom was in contact with more reducing waters. These conditions favoured the preservation of organic matter in the water column, enhanced the deposition of organic matter at the seafloor and led to an increase in bacterial sulphate reduction in the sediments, which triggered the widespread precipitation of early diagenetic microbial dolomite (Fig. 7b).

Conclusions
Hydrological change affecting the Piedmont Basin during the early stages of the Messinian salinity crisis (MSC) led to an intensification of water-column stratification. The water column was divided into denser, more saline and oxygen-depleted bottom waters separated by a chemocline from an upper water layer consisting of oxic seawater influenced by freshwater inflow. No evidence of a sharp increase of salinity across the MSC onset was found. Vertical oscillations of the chemocline exerted control over the stratigraphic architecture of the Primary Lower Gypsum unit and its deeperwater equivalents deposited during the first stage of the MSC. This study documents how temporal and spatial changes of water masses with different redox chemistries must be carefully considered when interpreting the MSC event. Environmental change at the onset of the Messinian salinity crisis 837