Group I mesophilic Thaumarchaeota fix dissolved inorganic carbon (DIC), accompanied by a biosynthetic fractionation factor of ~20‰. Accordingly, the δ13C signature of their diagnostic biomarker crenarchaeol was suggested as a potential δ13CDIC proxy in marine basins if input from nonmarine Thaumarchaeota is negligible. Semi-enclosed basins are sensitive to carbon-cycle perturbations, because they tend to develop thermohaline stratification. Water column stratification typified the semi-enclosed basins of the Mediterranean Sea during the late Miocene (Messinian) salinity crisis (5.97–5.33 Ma). To assess how the advent of the crisis affected the carbon cycle, we studied sediments of the Piedmont Basin (northwestern Italy), the northernmost Mediterranean subbasin. A potential bias of our δ13CDIC reconstructions from the input of soil Thaumarchaeota is discarded, since high and increasing branched and isoprenoid tetraether (BIT) index values do not correspond to low and decreasing δ13C values for thaumarchaeal lipids, which would be expected in case of high input from soil Thaumarchaeota. Before the onset of the crisis, the permanently stratified distal part of the basin hosted a water mass below the chemocline with a δ13CDIC value of approximately 3.5‰, while the well-mixed proximal part had a δ13CDIC value of approximately 0.8‰. The advent of the crisis was marked by 13C enrichment of the DIC pool, with positive δ13CDIC excursions up to +5‰ in the upper water column. Export of 12C to the seafloor after phytoplankton blooms and limited replenishment of remineralized carbon due to the stabilization of thermohaline stratification primarily caused such 13C enrichment of the DIC pool.

Marine group I (MGI) mesophilic Thaumarchaeota are among the dominant archaea in marine environments, where their occurrence can be traced by crenarchaeol, the thaumarchaeal-specific isoprenoid glycerol dialkyl glycerol tetraether (iGDGT; e.g., Besseling et al., 2020). In modern marine sediments, the δ13C values of the cyclohexane ring–containing tricyclic biphytane (Bp-cren) derived from ether cleavage of crenarchaeol typically fall between −23‰ and −18‰ (δ13CBp-cren; Schouten et al., 2013). These values originate from autotrophic fixation of dissolved inorganic carbon (DIC) as bicarbonate, with a biosynthetic fractionation factor (ε) of ~20‰ (Könneke et al., 2012). Accordingly, δ13CBp-cren was suggested as a potential (paleo-)δ13CDIC proxy, assuming that ancient MGI Thaumarchaeota had the same metabolism as their modern heirs (Schouten et al., 2013). However, the uptake of organic carbon or the input of crenarchaeol from marine Euryarchaeota was claimed to compromise δ13CDIC estimates (Pearson et al., 2016). Such concerns were recently discarded (cf. Pearson et al., 2019; Besseling et al., 2020), although it has been put forward that ε values could depend on growth rate and carbon dioxide concentration, resulting in deviations of up to ±2‰ between the measured and the reconstructed δ13CDIC values of modern seawater (cf. Hurley et al., 2019). Another potential bias for δ13CDIC reconstructions is riverine input of crenarchaeol sourced by soil- and river-dwelling Thaumarchaeota to marine basins (Pearson et al., 2016). Elling et al. (2019) disputed the effect of such bias, even under high input of soil-derived crenarchaeol to ancient marine sediments. Therefore, δ13CBp-cren values could, indeed, be a powerful (paleo-)δ13CDIC proxy, as demonstrated by successful applications in tracing past carbon-cycle perturbations (Kuypers et al., 2001; Schoon et al., 2013; Elling et al., 2019).

Carbon-cycle perturbations are common for modern semi-enclosed basins, which tend to develop thermohaline stratification, altering the δ13CDIC of the water column (e.g., Fry et al., 1991). Water masses below the chemocline tend to become 13C depleted due to pronounced organic matter recycling, resulting in δ13CDIC values as low as −19‰ (e.g., van Breugel et al., 2005). Conversely, seasonal eutrophication and preferential 12CO2 loss via enhanced degassing in surface waters can produce 13C enrichment, with δ13CDIC values as high as +16.5‰ in lacustrine environments (e.g., Stiller et al., 1985; Oren et al., 1995). Akin to modern basins, thermohaline stratification may have caused the alteration of δ13CDIC pools also in ancient semi-enclosed basins (e.g., Schoon et al., 2013).

In the late Miocene, the Mediterranean Basin developed thermohaline stratification (García-Veigas et al., 2018) due to its close-to-complete tectonic isolation from the global ocean during the Messinian salinity crisis (MSC; 5.97–5.33 Ma). Although the global carbon cycle was likely affected by this event (e.g., Capella et al., 2019), the effect of the advent of the MSC on the δ13CDIC pool of the late Miocene Mediterranean water column remains unassessed. To fill this gap, we conducted a case study on the Piedmont Basin (northwestern Italy), the northernmost subbasin of the late Miocene Mediterranean Sea (Dela Pierre et al., 2011). There, MGI Thaumarchaeota represented the main group of planktonic archaea at the onset of the MSC (Natalicchio et al., 2017; Sabino et al., 2021). After assessing potential biases in the applicability of δ13CBp-cren values as paleo-δ13CDIC proxies, we reconstructed the δ13CDIC of the stratified water body in proximal (Pollenzo section) and distal (Govone section) sectors of the margin of the Piedmont Basin. The approach used in this study shows great promise to decipher ancient carbon-cycle perturbations.

We analyzed five lithologic cycles (shale-marl couplets) deposited across the MSC onset (5.97 Ma) at a water depth >200 m in a more proximal (Pollenzo; 43 samples) and a distal (Govone; 14 samples) position along the southern margin of the Piedmont Basin (Figs. 1 and 2; for details, see the Supplemental Material1). The cycles reflect astronomically driven moister (shales) and drier (marls) climate oscillations (cf. Natalicchio et al., 2019; Sabino et al., 2020). In the studied sections, the MSC onset was placed at the base of marls of cycles Pm5 (Pollenzo; Natalicchio et al., 2019) and Gm30 (Govone; Gennari et al., 2020; Fig. 2). An aliquot of the total lipid extract of the 57 samples was used to obtain data on iGDGTs and branched GDGTs (bGDGTs), analyzed through high-performance liquid chromatography–mass spectrometry. We calculated the branched and isoprenoid tetraether (BIT) index and the iGDGT-2/iGDGT-3 ratio ([2]/[3] ratio) to assess input from soil Thaumarchaea and the relative contributions of upper-water-column versus deeper-water-column thaumarchaeal communities, respectively (Hopmans et al., 2004; Kim et al., 2015). Ether-bound isoprenoids were released from iGDGTs by treating another aliquot of the extracts with HI/LiAlH4, while the asphaltene fraction of 21 samples (Pollenzo: 7 samples; Govone: 14 samples) was desulfurized to release sulfur-bound compounds from macromolecules. The hydrocarbon fractions obtained from ether cleavage and desulfurization were analyzed using gas chromatography–mass spectrometry. The δ13C analyses of ether-cleaved biphytanes (Bp) and desulfurized phytane and C27 to C29 steranes obtained after column chromatography were performed on 41 samples (Pollenzo: 27 samples; Govone: 14 samples). The δ13C values are reported in per mil (‰) versus the Vienna Peedee belemnite (V-PDB) standard, and the average analytical standard deviation was 0.4‰ (see the Supplemental Material for details).

δ13CBp-cren Values as a Robust δ13CDIC Proxy

The BIT index suggests high (0.4–1.0, proximal sector, Pollenzo) to moderate (0.3–0.7, distal sector, Govone) soil-derived organic matter input into the Piedmont Basin at the advent of the MSC (Fig. 2A). A high BIT may imply high input of crenarchaeol sourced from group I soil Thaumarchaeota to marine sediments (Weijers et al., 2006), hampering the applicability of crenarchaeol and its derivative Bp-cren as a δ13CDIC proxy in this case. The δ13CBp-cren values from MGI Thaumarchaeota range from −23‰ to −18‰ in modern marine sediments (Schouten et al., 2013, and references therein). In contrast, the δ13CBp-cren values of soil Thaumarchaeota are approximately −30‰ in soils dominated by C3 plants and −23‰ in soils where C4 plants prevail (Weijers et al., 2010). Because C3 plants dominated the Piedmont Basin hinterland in the late Messinian (Bertini and Martinetto, 2011), a bias from soil archaea should result in high and increasing BIT values corresponding to low and decreasing δ13CBp-cren values. However, the reverse trend was found for the studied samples, with lower δ13CBp-cren values occurring when BIT values decreased, and vice versa (Figs. 2B and 3). The input of soil thaumarchaeal lipids was therefore apparently negligible. The bGDGTs were likely rather sourced from bacteria dwelling in rivers, coastal marine sediments, and in oxygen-deficient waters (e.g., Liu et al., 2014; Crampton-Flood et al., 2021). This agrees with increasing BIT values coinciding with higher river discharge and an expansion of oxygen-deficient waters toward the margin of the Piedmont Basin after the MSC onset (Natalicchio et al., 2019; Sabino et al., 2020, 2021). Because Thaumarchaeota have also been reported to dwell in rivers (e.g., Kim et al., 2007; Yang et al., 2013), river-derived crenarchaeol with increasing fluvial discharge can also affect δ13CDIC reconstructions of seawater. Today's rivers draining the former catchment area of the Piedmont Basin are typified by a mean δ13CDIC value of −8‰ (Marchina et al., 2016). Assuming a similar value for the late Miocene and an ε = 20‰ (Könneke et al., 2012) for fluvial Thaumarchaeota, a trend toward 13C depletion of Bp-cren with increasing river discharge is likely. Although river discharge increased in the Piedmont Basin after the MSC onset (cf. Natalicchio et al., 2019; Sabino et al., 2020), Bp-cren became increasingly 13C enriched (Fig. 2B). Hence, we infer negligible input also from fluvial Thaumarchaeota to the sedimentary pool and conclude that the measured δ13CBp-cren values reflect the signature of MGI Thaumarchaeota. These values can consequently be used as δ13CDIC proxy, despite the moderate to high BIT index (see the Supplemental Material for details on the acyclic and cyclic biphytanes).

Impact of Water-Column Stratification on the δ13CDIC Pool

The Piedmont Basin developed water-column stratification approaching the MSC onset (Dela Pierre et al., 2011), following the geodynamic and oceanographic evolution of the Mediterranean Basin (Roveri et al., 2014). Before the MSC onset, the δ13CBp-cren values from the distal sector reflected 13C depletion relative to the proximal sector (mean Δdist-prox = −2.7‰; Figs. 2B and 3), close to the lowest δ13C values reported for crenarchaeol and Bp-cren for modern and ancient marine sediments (~−24‰ ± 1‰; Figs. 2B and 3; Schouten et al., 2013, and references therein; Schoon et al., 2013; Elling et al., 2019). Smittenberg et al. (2005) explained 13C depletion of this magnitude in thaumarchaeal lipids as the result of the uptake of 13C-depleted DIC when the chemocline rises to the base of the photic zone. We propose that the δ13CBp-cren values from Govone reflect a similar mechanism. Considering the mean δ13CBp-cren value of −23.5‰ from samples indicating photic zone euxinia (presence of isorenieratane; Fig. 2; Sabino et al., 2021), and applying an ε = 20‰ (Könneke et al., 2012), we estimate a δ13CDIC of −3.5‰ and a possible ±2‰ deviation (maximum value; cf. Hurley et al., 2019) for the DIC pool below the chemocline in the Piedmont Basin. In the proximal Pollenzo sector, the higher δ13CBp-cren values (−20.8‰ on average; Figs. 2B and 3) probably reflect mixing (Natalicchio et al., 2017), redistributing the remineralized carbon and maintaining a δ13CDIC = −0.8‰ ± 2‰, close to that of modern seawater (−1‰ < δ13CDIC < +2‰; Schmittner et al., 2013).

The MSC onset coincided with intensification of water-column stratification in the Piedmont and other Mediterranean subbasins (Natalicchio et al., 2017; García-Veigas et al., 2018). This change is mirrored in the studied samples by 13C enrichment in thaumarchaeal lipids and algal steranes and phytane (see also the Supplemental Material; Fig. 2B). Excursions toward δ13CBp-cren values as high as −15‰ in the proximal Pollenzo sector coincide with [2]/[3] ratios lower than 3 (Figs. 2B and 2C). Remarkably, such low ratios are also found in modern sediments of the Mediterranean Sea for which thaumarchaeal lipids are derived dominantly from the upper water column (<200 m; cf. Kim et al., 2015). Accordingly, the pattern in the proximal sector is likely due to lipids sourced from Thaumarchaeota dwelling in the upper water column and fixing carbon from a DIC pool reaching δ13C values as high as +5‰ ± 2‰ during the earliest MSC phase. The drier climate in the Mediterranean during the MSC (Roveri et al., 2014) makes preferred 12CO2 degassing via enhanced seawater evaporation a possible explanation for such 13C enrichment (e.g., Horton et al., 2016). However, in the northern Mediterranean, humid conditions persisted (Bertini and Martinetto, 2011), and the climate was only relatively dry during some excursions (cf. Natalicchio et al., 2019; Sabino et al., 2020). Therefore, 12CO2 degassing is unlikely to have been the sole mechanism responsible for the positive δ13C excursions. Interestingly, similar positive excursions have been reported for the DIC pool of the superficial water of the modern Dead Sea (above the pycnocline; +2.5‰ ≤ δ13CDIC ≤ +5‰; cf. Oren et al., 1995). Such δ13C excursions are caused by phytoplankton blooms preferentially taking up 12C after episodes of enhanced runoff, which supplied additional nutrients to the basin (Oren et al., 1995). Similarly, periodic increases of riverine runoff and nutrient input were shown to have led to phytoplankton blooms in the Piedmont Basin (Natalicchio et al., 2019). We therefore suggest that massive export of 12C to the seafloor following productivity pulses resulted in the observed positive excursions in the DIC pool of the upper water column (Fig. 4). Concurrently, stabilization of thermohaline stratification limited the replenishment of remineralized carbon to the upper water column, agreeing with the coinciding 13C enrichment of algal lipids (Fig. 2B).

In contrast, the MSC sediments of the distal Govone section lack abnormally high δ13CBp-cren values and show [2]/[3] ratios constantly higher than 3 (Fig. 2C). These ratios hint at input of lipids from MGI Thaumarchaeota dwelling in deeper waters (Kim et al., 2015), probably fixing at least partially carbon from the DIC pool below the chemocline. This scenario explains the lower δ13CBp-cren values (~−21.5‰ on average), the lack of δ13C excursions, and the partial mismatch with the trends observed for algal lipids (Fig. 2B). Periodic input of 13C-depleted DIC into the upper water column is, however, testified by the lowering of the δ13C values of algal lipids (Fig. 2B). Given that mixing was suppressed by the stable thermohaline stratification, fluvial discharge was the most probable source of 13C-depleted DIC (δ13CDIC = −8‰, see above; Fig. 4) at these times. In conclusion, permanent stratification of the water column after the onset of the MSC promoted the development of strong vertical chemical gradients, establishing a 13C-enriched upper water column overlying a 13C-depleted water mass in the Piedmont Basin.

The carbon stable isotope signature of crenarchaeol-derived tricyclic biphytane (δ13CBp-cren) from marine Thaumarchaeota allows us to trace the composition of the pool of DIC in the Piedmont Basin—the northernmost late Miocene Mediterranean subbasin—~6 m.y. ago at the advent of the Messinian salinity crisis. The water column was typified by a stagnant, deeper body with a δ13CDIC of approximately −3.5‰ toward the basin depocenter before and during the advent of the crisis. The onset of the event was marked by significant 13C enrichment of the DIC pool in the upper water column (δ13CDIC as high as ~+5‰) due to the basinwide stabilization of thermohaline stratification. The 13C enrichment is interpreted to primarily reflect preferential export of 12C to the seafloor after phytoplankton blooms and limited replenishment of remineralized DIC.

We thank S. Beckmann (Universität Hamburg, Germany) for technical support during organic geochemical analyses. M. Sabino was funded by a doctoral scholarship provided by the Landesgraduiertenförderung of the state of Hamburg. The article is further based upon work from COST (European Cooperation in Science and Technology) Action “Uncovering the Mediterranean salt giant” (MEDSALT). Insightful comments by Vanni Aloisi, Marcus Elvert, and an anonymous reviewer helped to improve the manuscript.

1Supplemental Material. Geological setting, methodology, and lipid abundance, carbon stable isotopes, and sources (Tables S1–S4; Figures S1–S3). Please visit https://doi.org/10.1130/GEOL.S.15832095 to access the supplemental material, and contact editing@geosociety.org with any questions.
Gold Open Access: This paper is published under the terms of the CC-BY license.