Messinian bottom‐grown selenitic gypsum: An archive of microbial life

Primary gypsum deposits, which accumulated in the Mediterranean Basin during the so‐called Messinian salinity crisis (5.97–5.33 Ma), represent an excellent archive of microbial life. We investigated the molecular fossil inventory and the corresponding compound‐specific δ13C values of bottom‐grown gypsum formed during the first stage of the crisis in four marginal basins across the Mediterranean (Nijar, Spain; Vena del Gesso, Italy; Heraklion, Crete; and Psematismenos, Cyprus). All studied gypsum samples contain intricate networks of filamentous microfossils, whose phylogenetic affiliation has been debated for a long time. Petrographic analysis, molecular fossil inventories (hydrocarbons, alcohols, and carboxylic acids), and carbon stable isotope patterns suggest that the mazes of filamentous fossils represent benthic microbial assemblages dominated by chemotrophic sulfide‐oxidizing bacteria; in some of the samples, the body fossils are accompanied by lipids produced by sulfate‐reducing bacteria. Abundant isoprenoid alcohols including diphytanyl glycerol diethers (DGDs) and glycerol dibiphytanyl glycerol tetraethers (GDGTs), typified by highly variable carbon stable isotope composition with δ13C values spanning from −40 to −14‰, reveal the presence of planktic and benthic archaeal communities dwelling in Messinian paleoenvironments. The compound inventory of archaeal lipids indicates the existence of a stratified water column, with a normal marine to diluted upper water column and more saline deeper waters. This study documents the lipid biomarker inventory of microbial life preserved in ancient gypsum deposits, helping to reconstruct the widely debated conditions under which Messinian gypsum formed.

gypsum having accumulated in the Mediterranean Basin during the Messinian salinity crisis (MSC; Ma; Roveri et al., 2014), when the Mediterranean was turned into the youngest salt giant of Earth history (e.g., Hsü et al., 1973). The Messinian gypsum deposits are grouped into three stratigraphic units (CIESM, 2008): the Primary Lower Gypsum unit, representing the first phase of the MSC (5.97-5.60 Ma); the Resedimented Lower Gypsum unit, consisting of large blocks of the Primary Lower Gypsum unit emplaced by gravity flows during the second phase of the MSC (5. 60-5.55 Ma;Roveri et al., 2014); and the Upper Gypsum unit, deposited during the final phase of the crisis (5. . The environmental conditions controlling the deposition of Messinian gypsum are still debated, especially due to the lack of modern analogs. The onshore Messinian primary gypsum deposits are largely composed of bottom-grown selenitic (i.e., larger than 2 mm and up to some meters in size) crystals.
Among the best examples of bottom-grown laminated, selenite gypsum is the so-called "stromatolitic" gypsum deposits (sensu Rouchy &Monty, 2000 andAllwood et al., 2013) of Cyprus and Crete. Their formation has been explained by a combination of abiotic and microbially driven processes (cf. Allwood et al., 2013).
Intricate networks of filamentous microfossils (Dela Pierre et al., 2015;Panieri et al., 2010;Rouchy, 1982;Rouchy & Monty, 1981Schopf et al., 2012;Vai & Ricci Lucchi, 1977), also described as "shoestring-" and "spaghetti-like" structures (Vai & Ricci Lucchi, 1977), are the most striking feature of many gypsum crystals. The phylogenetic affiliation of the microfossils has not been unequivocally determined. It was proposed that the filaments represent brine shrimp fecal pellets (Schreiber & Decima, 1976), algae (Vai & Ricci Lucchi, 1977), or cyanobacteria (Rouchy & Monty, 2000). An assignment to cyanobacteria was further substantiated on the basis of the extraction of 16S rRNA from fossiliferous gypsum of northern Italy (Panieri et al., 2010). Later on, Schopf et al. (2012) suggested that the filaments may rather represent the remains of colorless sulfideoxidizing bacteria like Beggiatoa or Thioploca. The dense fossil accumulations have not resulted from settling of planktic filamentous organisms, but represent former benthic assemblages of microbial filaments (Dela Pierre et al., 2015). The recognition of abundant aggregates of microcrystalline pyrite and associated polysulfide-the latter representing a diagnostic criterion for this clade of bacteriawithin the filaments agrees with the assignment of the filamentous fossils to the colorless sulfide-oxidizing bacteria (Dela Pierre et al., 2015). Based on this circumstance, an unequivocal identification of the phylogenetic affiliation of these microfossils would allow the reconstruction of the depositional environment of bottom-grown Messinian gypsum. Should the assignment to phototrophic cyanobacteria be correct, gypsum would have been deposited within the photic zone, that is, at a maximum depth of 200 m . In contrast, colorless sulfide-oxidizing bacteria do not provide any depth constraints-these prokaryotes can live at any depth from bathyal to peritidal settings (Bailey et al., 2009)-but can shed light on the redox conditions at the seafloor and the presence of an active biogeochemical sulfur cycle in Messinian brines (cf. Schulz & Jørgensen, 2001).
Here, we provide a comprehensive overview on the inventory of molecular fossils preserved in filament-bearing, primary bottomgrown selenitic gypsum formed during the first stage of the MSC in four marginal basins across the Mediterranean. The excellent preservation of molecular fossils, along with the determination of compound-specific carbon stable isotope compositions, enables us (1) to recognize the main groups of micro-organisms inhabiting Messinian aquatic ecosystems at times of gypsum formation and (2) to contribute to the identification of the enigmatic filamentous microfossils.

| G EOLOG I C AL S E T TING S OF THE SAMPLED S EC TI ON S
Messinian primary gypsum from five sections was collected and analyzed for this study. The samples are from four Mediterranean peripheral basins, (1) the Nijar Basin (southeastern Spain) at the western margin of the Mediterranean, (2) the Vena del Gesso Basin (Italy; two sections) located in the northern sector of the Mediterranean, and the (3) Heraklion (Crete) and (4) Psematismenos (Cyprus) basins in the eastern Mediterranean ( Figure 1). The studied samples are from various stratigraphic levels, but they share the presence of densely packed filamentous fossils.

| Nijar basin
The Nijar Neogene sedimentary succession fills a SW-NE elongated basin located in the internal zone of the Betic Cordillera (southeastern Spain; Fortuin & Krijgsman, 2003). The Messinian succession comprises (a) pre-MSC marine strata of the Abad Marls (a member of the Turre Formation; Van de Poel, 1991), (b) sulfate evaporites of the Yesares Formation (e.g., Fortuin & Krijgsman, 2003), which is considered as the local equivalent of the Mediterranean Primary Lower Gypsum unit (sensu Roveri et al., 2008), and (c) post-evaporitic continental and lagoonal deposits of the Feos Formation (Fortuin & Krijgsman, 2003;Omodeo Salé et al., 2012). The studied gypsum was sampled from the Yesares Formation exposed in the Gafares section (37°01′28″N, 1°59′17″W); here the Yesares Formation is approximately 70 m thick and consists of a cyclic alternation of at least eight gypsum and laminated marl couplets (Lu, 2006). The Gafares section starts with large selenitic gypsum crystals (the coarse and palmate twinned selenite facies of Lu, 2006) and shows an upward trend toward smaller gypsum crystals (grass-like selenite facies of Lu, 2006). The filament-rich selenitic gypsum was sampled at the base of the Yesares Formation. A dominantly marine origin has been suggested for the Yesares selenites based on 87 Sr/ 86 Sr ratios, trace element distribution, and the marine microfossil assemblages of the intercalated marls (Lu & Meyers, 2003;Lu et al., 2001Lu et al., , 2002.

| Vena del Gesso basin
The Vena del Gesso Basin is located in the northern Apennines (Italy), exposing a complete succession of the Primary Lower Gypsum unit along a ridge (Figure 2a). This succession consists of 16 cycles composed of gypsum and shale couplets Reghizzi et al., 2018;Vai & Ricci Lucchi, 1977), cut at the top by the Messinian erosional surface (Monticino quarry). The ideal cycle starts with bituminous shales, followed by stromatolitic limestones (Rouchy & Monty, 1981Vai & Ricci Lucchi, 1977), which directly underlie the filament-bearing massive selenite facies (Figure 2b; Lugli et al., 2010). The size of the selenite crystals decreases upward, with the appearance of banded selenite and the branching selenite (the latter starting from the 6 th cycle) facies completing an ideal cycle. Two samples of filament-bearing selenite gypsum were taken from the basal part of the third Primary Lower Gypsum cycle in the Monte Tondo quarry (44°15′04″N, 11°40′13″E) and from the second cycle in the Monticino quarry (44°13′29″N, 11°45′43″E).
These deposits belong to the Lower Evaporites (Delrieu et al., 1993), which are the equivalent to the Primary Lower Gypsum. According to Rouchy (1982), they formed in a very shallow hypersaline lagoon.
In the Tsangaraki area, the 2-3 m thick stromatolitic gypsum interval is located at the base of a gypsum unit (up to 80 m thick) containing different types of gypsum (selenite crystals, nodular and laminated gypsum, and gypsarenites). This unit overlies a chaotic interval of laminated fossiliferous carbonates and a rhythmic succession of laminated marls, diatomites, and fine-grained carbonates.

| Psematismenos basin
The Psematismenos Basin, located in the southeastern part of Cyprus, is a Neogene sedimentary basin surrounding the Troodos ophiolitic massif (Manzi et al., 2016). In this basin, Messinian evaporites comprise a thick gypsum unit resting on an alternation of pre-

| ME THODS
Thin sections and thin platy gypsum crystals were studied with transmitted light on a Leica DM4500 P LED optical microscope equipped with a digital camera (Leica DFC 450 C). A FEI Inspect S equipped with an energy dispersive X-ray detection unit (EDAX Apollo XV) was used for scanning electron microscopy (SEM) and semi-quantitative element recognition. For SEM analyses, thin  Rouchy and Caruso (2006) gypsum slices were treated with deionized water to partly dissolve the selenitic gypsum. Mineralogical composition was determined with X-ray diffraction using a Panalytical X'Pert PRO diffractometer (CuKα radiation, 40 kV, 40 mA, step size 0.0167, 5 s per step; Table 1). The samples were loaded into the sample holders as oriented powder.
One representative gypsum sample from each location (approximately 200 g each) was selected for lipid biomarker analysis.
The samples were carefully cleaned with acetone and ground with pestle and mortar to a fine powder. The extraction and separation procedure was carried out following Hoffmann-Sell et al. (2011).
The powders were saponified in 6% KOH/MeOH (3 h, 80°C) and three times extracted with a microwave extraction system (MARS X, CEM Corporation) at 80°C and 300 W with dichloromethane (DCM):MeOH (3:1). The resulting total extracts were cleaned by separation into an n-hexane soluble fraction (maltenes) and a DCMsoluble fraction (asphaltenes). Maltene fractions were separated into four fractions of increasing polarity by column chromatography (hydrocarbons, ketones and esters, alcohols, carboxylic acids). In a second step, the remaining extracted gypsum powder was dissolved in clean water, enriched in NaCl, which had been annealed before, and subsequently extracted and fractionated as described above.
The extracts of (1) gypsum powder (average 71 wt% of all extracted lipids) and (2) gypsum powder after dissolution (average 29% of all extracted lipids) yielded the same lipid distribution; for this reason, the total amount of identified compounds (Table 2)  Compound-specific carbon isotope analysis (irm-GC/MS) was performed with a Thermo Electron GC-combustion-III-interface linked to a Thermo Electron Delta-plus XP mass spectrometer at the MARUM, University of Bremen. The δ 13 C values of alcohols and carboxylic acids were corrected for the addition of trimethylsilyl and methyl derivatives, respectively. Each measurement was calibrated using several pulses of carbon dioxide gas with known composition at the beginning and the end of the run. The precision of measurements was checked with a mixture of n-alkanes (C 15 -C 29 ) with known isotopic composition. The analytical standard deviation was smaller than 0.4‰.
Glycerol dibiphytanyl glycerol tetraethers (GDGTs) were prepared according to the procedure described in Hoffmann-Sell et al.

| Lipid biomarker inventory
The lipid biomarker inventory is largely similar among the various studied samples (see Table 2), although some differences can be recognized. Generally, the western and northern Mediterranean samples (Nijar and Vena del Gesso) show lower total molecular fossil contents, ranging from 1,594 ng/g rock (Monticino) to 3,348 ng/g rock (Nijar). In contrast, the samples from the eastern Mediterranean (Crete and Cyprus) are typified by higher contents (up to 5,179 ng/g rock in the Crete sample; Table 2; Figure 6a). Carboxylic acids are the most abundant molecular fossils, representing more than 60 wt% of the total amount of lipids ( Figure 6a).

| Hydrocarbons
The average contents of hydrocarbons range from ca.120 ng/g rock in the Monte Tondo and Monticino samples to ca. 420 ng/g rock in Nijar, Crete, and Cyprus. The n-alkanes are the dominant compounds in the hydrocarbon fractions, peaking either at n-C 18 or n-C 31 . Some variations in the distribution of short-chain (n-C 17-21 ) and long-chain (n-C 26-33 ) n-alkanes are observed. While Nijar, Crete, and Cyprus samples are characterized by prevailing short-chain n-alkanes, the Vena del Gesso samples show equal contents of short-and longchain n-alkanes (Table 2). Among the short-chain n-alkanes, the n-C 17 alkane is not enriched compared to neighboring n-alkanes, representing 8 to 11% of all hydrocarbons. The head-to-tail linked C 19 isoprenoid pristane and the head-to-tail linked C 20 isoprenoid phytane are the only isoprenoid hydrocarbons detected.

| Alcohols
Gypsum from all study sites contains n-alcohols, although in different proportions; short-chain n-alcohols comprise 16% of the alcohol fraction in Crete and Cyprus samples and about 57 and 55% in Nijar and Monte Tondo samples, respectively. Among all fractions, the molecular fossil inventories vary most from site to site for the alcohol fractions. While in the Nijar sample, the n-alcohols peak at n-C 16 , n-C 18 is most abundant in the Monte Tondo sample. In the Cyprus sample, n-C 24 is most prominent. Apart from n-alcohols, isoprenoid alcohols account for 20% (Nijar) to 63% (Crete) of the compound inventory of the alcohol fractions, also including diphytanyl glycerol diethers (DGDs), glycerol dibiphytanyl glycerol tetraethers (GDGTs),   Among the carboxylic acids, δ 13 C values were only obtained for few compounds. As for the n-alcohols, the n-fatty acids yielded about 4‰ higher values than the n-alkanes; only in the Crete sample, δ 13 C values of n-fatty acids are in the same range like n-alkanes.

| Compound-specific carbon stable isotopes
The δ 13 C values of the predominant short-chain n-alkanes range from −30 to −22‰. Only few data were obtained for monounsat-

| DISCUSS ION
Modern marine gypsum deposits typically form in shallow-water hypersaline environments, including coastal mudflats (sabkhas), lagoons, and human-made salterns. In such settings, biological diversity is limited by high salinity (>110 PSU), which is lethal for most eukaryotes; consequently, these environments are inhabited by highly specialized prokaryotes, such as cyanobacteria and halophilic     Euryarchaeota (Jebbar et al., 2020;Oren, 2002;Ventosa et al., 2014), and few eukaryotes including the brine shrimp Artemia salina and the green alga Dunaliella salina (Oren, 2005). Some prokaryotes also live as endoliths in gypsum deposits (Oren et al., 1995;

(c) (d) (e)
A significant input of lipids from planktic biota is not to be expected in case of salterns, salinas, and lagoons.

| DGDs and GDGTs indicate a diverse aquatic archaeal community
The  (Birgel et al., 2014;Grice et al., 1998). The large Δδ 13 C DGD-sterol could be possibly explained by the heterotrophic lifestyle of halophilic archaea, involving the uptake of relatively 13 Cenriched carbon from carbohydrates and proteins by halophilic archaea in contrast to photoautotrophic algae assimilating dissolved carbon dioxide (Oren, 1994). Carbohydrates and proteins are generally less 13 C-depleted than other organic substrates (e.g., lipids; De Niro andEpstein, 1977, Grice et al., 1998).
Phytanol and the two phytanylglycerol monoethers extracted from Crete and Cyprus gypsum revealed similar δ 13 C values (−22 to −13‰) to archaeol and extended archaeol (cf. Birgel et al., 2014;Ziegenbalg et al., 2012); most likely the former compounds represent degradation products of DGDs (cf. Liu et al., 2018). The isoprenoid hydrocarbon phytane, however, is more 13 C-depleted (δ 13 C: −33 to −30‰) than other isoprenoids with phytanyl moieties (Figure 7). Phytane probably derives either from other heterotrophic archaeal communities (Wakeham et al., 2003) or from a secondary, possibly terrigenous source, and is unrelated to the other compounds with phytanyl moieties. To sum up, the lipid biomarker inventory of Messinian gypsum confirms the prominent presence of halophilic archaea in the depositional paleoenvironment-not a surprising result if the gypsum is interpreted as a product of fully hypersaline conditions. Halophilic archaea are commonly thriving in high salinity environments like in some modern shallow-water microbial mats (Bühring et al., 2009;Jahnke et al., 2014). However, uncultured members of the genus Halobacteria have also been recognized in aquatic systems with low salt concentrations such as at sulfur-rich springs (Elshahed et al., 2004), estuaries (Purdy et al., 2004;Singh et al., 2010), and the Black Sea, where the salinity of bottom waters is ca. 20 PSU (Jessen et al., 2016). In the latter environment, The observed GDGT assemblages are much different from those reported for modern shallow-water hypersaline settings (Petrick et al., 2019). Halophilic archaea are not known to produce GDGTs (Schouten et al., 2013;Teixidor et al., 1993). Although methanogens isolated from hypersaline lakes have been shown to produce GDGT-0 and GDGT-1, they do not produce extended archaeol (Bale et al., 2019). Most of the GDGTs found in marine sediments are thought to be produced by planktic, ammonia-oxidizing Thaumarchaeota (Brochier-Armanet et al., 2008;Könneke et al., 2005;Schouten et al., 2013;Spang et al., 2010;Wuchter et al., 2006). Crenarchaeol, which is the most abundant GDGT in the studied Messinian gypsum (Figure 5c), is also the predominant GDGT of Thaumarchaeota in the marine epipelagic realm (Schouten et al., 2013;Turich et al., 2007).
It is also produced by soil-dwelling Thaumarchaeota (Sinninghe Damsté et al., 2012) and occurs in marine hydrothermal settings (Pearson et al., 2004;De La Torre et al., 2008). In contrast to soildwelling Thaumarchaeota, marine Thaumarchaeota produce not only crenarchaeol, but also the acyclic GDGT-0 (caldarchaeol) and with certainty, yet methanotrophic archaea can be excluded since this clade reveals different GDGT patterns with the predominance of GDGT-2, 13 C depletion, and other isoprenoids not observed in the Messinian gypsum (Birgel et al., 2006Natalicchio et al., 2012). A possible group from which the 13 C-depleted biphytanes could derive are methanogenic euryarchaea (Hoffmann-Sell et al., 2011;Koga et al., 1998). Among other factors, the isotope composition of lipids synthesized by methanogens depends largely on the metabolized substrate (Londry et al., 2008). Unfortunately, the isotopic composition of GDGT-derived biphytanes has not yet been determined in culture experiments with methanogens. Since archaeol, interpreted to be derived from halophilic archaea, yielded

| The affiliation of filamentous microfossils
The affiliation of the superabundant filamentous microfossils preserved in Messinian gypsum has not yet been resolved with certainty. At first, they were interpreted to represent fecal pellets of brine shrimps (Schreiber and Decima, 1976), and later, they were assigned to algae (Vai & Ricci Lucchi, 1977), cyanobacteria (Panieri et al., 2010;Rouchy and Monty, 2000) or sulfide-oxidizing bacteria (Dela Pierre et al., 2015;Schopf et al., 2012). An attribution to fecal pellets does not agree with the partial hollowness of the studied filaments ( Figure 4e). Hollowness is not expected in fecal pellets, but is strong evidence of a microbial origin of the filamentous microfossils (Andreetto et al., 2019, and reference therein). Such origin is in accord with the curved shape of filaments and their internal segmentation (Dela Pierre et al., 2015). The most striking argument for an assignment to cyanobacteria was the extraction of 16S rRNA from samples of the Vena del Gesso gypsum, matching with several cyanobacterial clones and in particular with those of the genus Geitlerinema, typical of coastal shallow marine environments (Panieri et al., 2010). The finding of genetic material in several million-year-old geological samples requires excellent preservation of organic matter, supposably caused by early permineralization in gypsum prior to cell decay and disintegration (cf. Schopf et al., 2012). Molecular fossils, which are much more stable than 16S rRNA, should therefore confirm the suggested cyanobacterial origin of filaments. Cyanobacteria synthesize lipid biomarkers (Table 3), including diplopterol (Bühring et al., 2009;Rohmer et al., 1984) and bacteriohopanepolyols Summons et al., 1999;Talbot et al., 2008). Other diagnostic cyanobacterial lipids are short-chain n-alkanes, but especially heptadecane and heptadecenes, as well as methylated hepta-and octadecanes Hefter et al., 1993;Jahnke et al., 2014;Kozlowski et al., 2018;Wieland et al., 2008). Some of these compounds or their degradation products tend to be preserved for millions of years in the geological record (Heindel et al., 2015;Summons et al., 1999).
Among these cyanobacterial biomarkers, only n-C 17 -alkane (heptadecane) is present in the studied Messinian gypsum (Table 3), yet with moderate contents only, accounting on average for 7% of the total hydrocarbon fraction and corresponding to n-C 17 /n-C 18 ratios below 1. In cyanobacterial mats from modern hypersaline environments, n-C 17 accounts for approximately 80% of the total hydrocarbon fraction and the n-C 17 /n-C 18 ratio is significantly above 1 (e.g., Bühring et al., 2009;Jahnke et al., 2004). Despite heptadecanes being prone to degradation (Hefter et al., 1993;Wieland et al., 2008), in cases of 16S rRNA preservation, these cyanobacterial-derived hydrocarbons should be present as well. Although such negative evidence can obviously not exclude the presence of cyanobacteria in the Messinian depositional environment, the scarcity of cyanobacterial biomarkers (see Table 3) renders unlikely an assignment of the superabundant filamentous microfossils to cyanobacteria.
The other previously suggested hypothesis is that the interwoven filaments in the Messinian gypsum represent the remains of colorless sulfide-oxidizing bacteria like Beggiatoa and Thioploca (Dela Pierre et al., 2015;Schopf et al., 2012). Microbial mats dominated by these large prokaryotes are found in coastal upwelling areas (Arning et al., 2008;Bailey et al., 2009), in stratified basins (e.g., the Black Sea; Jessen et al., 2016), as well as in sulfide-rich marine sediments associated with methane seepage Zhang et al., 2005). The lipid biomarker inventory of mats of sulfide-oxidizing bacteria is typified by high contents of saturated (C 16:0 and C 18:0 ) and monounsaturated (C 16:1 and C 18:1 ) fatty acids (Arning et al., 2008;Jacq et al., 1989;Jiang et al., 2012;McCaffrey et al.,1989;Zhang et al., 2005). The high abundance of these compounds in the majority of the studied Messinian samples is consistent with this hypothesis (Table 3). Unfortunately, other organisms including many bacteria (Elvert et al., 2003;Londry et al., 2004) and diverse phytoplankton (Viso & Marty, 1993;Wakeham, 1995) are known to produce these largely unspecific fatty acids too.
Given the low specificity of the lipids synthesized by sulfideoxidizing bacteria (Arning et al., 2008), the origin of the filaments can be assessed further with the overall lipid assemblage and the carbon stable isotope patterns. Compounds synthesized by sulfatereducing bacteria such as iso-and anteiso-C 15:0 and C 17:0 fatty acids (Kaneda, 1991;Perry et al., 1979;Taylor & Parkes, 1983) are present in gypsum from Nijar, Crete, and Cyprus. In the case of Cyprus and Crete gypsum, which are the samples with the highest total lipid abundance ( Table 2), even a MAGE-a biomarker of sulfate-reducing bacteria (Grossi et al., 2015;Rütters et al., 2001;Ziegenbalg et al., 2012)-is present despite the generally low stability of MAGEs phytoplankton (dinosterol, cholesterol, cholestanol;Volkman, 2003) are less 13 C-depleted (average −21‰; Figure 7) on the other hand.
Consequently, a predominant origin of these fatty acids from terrestrial plants is unlikely. An origin from phytoplankton is instead possible, since culture experiments revealed that the offset between the δ 13 C values of fatty acids and sterols can be as high as 8‰ . Although phytoplankton cannot be excluded as a source of these compounds, we suggest that sulfide-oxidizing bacteria represent the most likely source of the C 16:0 , C 18:0, C 16:1, and C 18:1 fatty acids, agreeing with petrographic observations (i.e., hollowness, internal segmentation, and curved shape; Andreetto et al., 2019;Schopf et al., 2012) and the presence of iron sulfide inclusions in the filamentous fossils (Figure 4e). Such inclusions can represent the product of early diagenetic transformations of sulfur globules stored within the cells (Dela Pierre et al., 2015), which is a clade diagnostic feature of colorless sulfide-oxidizing bacteria (Schulz & Jørgensen, 2001). Interestingly, a tight association of sulfide-oxidizing bacteria with sulfate-reducing bacteria, as suggested by the biomarker inventory of some of the Messinian gypsum, has been reported for different modern environments (e.g., Arning et al., 2008;Fukui et al., 1999;SamKamaleson et al., 2021).

| Paleoenvironmental implications
Provided that the assignment of the filamentous fossils to benthic sulfide-oxidizing bacteria like Beggiatoa and Thioploca is correct and given the association of filamentous body fossils with molecular fossils of sulfate-reducing bacteria-especially in the samples with the highest total lipid contents (i.e., Nijar, Cyprus and Crete)-the basin floor where the Messinian gypsum grew was covered by a benthic assemblage of chemotrophic bacteria involved in an intense sulfur cycle. Such interpretation suggests suboxic to anoxic bottom water conditions and production of hydrogen sulfide through intense bacterial sulfate reduction, in turn sustaining sulfide-oxidizing bacteria (cf. Bailey et al., 2009;Teske et al., 2009). Sulfate-reducing bacteria live in suboxic to anoxic environments, using low molecular weight organic compounds or molecular hydrogen as electron donors (Grossi et al., 2015;Vinçon-Laugier et al., 2016). In case of the Messinian depositional environment, the electron donor was probably a mixture of organic substrates deriving from (1) (Thomas et al., 2019), which represents the most abundant organic carbon source in this saline basin (Oren, 1999).
The precipitation of gypsum probably caused the rapid entrapment of the benthic assemblage of microbial filaments. The paleoenvironmental conditions behind the formation of the Messinian gypsum are still discussed (e.g., García-Veigas et al., 2018;Grothe et al., 2020;Natalicchio et al., 2014). Data from fluid inclusions in bottom-grown selenitic gypsum challenged the idea that gypsum formed from hypersaline brines resulting from seawater evaporation. Instead, these data indicated that gypsum formed from low salinity waters (i.e., with low contents of Na + and Cl − ions) equivalent to 1.6 wt% sodium chloride on average (Costanzo et al., 2019;Evans et al., 2015;Natalicchio et al., 2014). The water masses probably represented a mixture of marine water, freshwater from Mediterranean rivers (Reghizzi et al., 2018), and brackish water from the Paratethys (Grothe et al., 2020). Interestingly, studies on modern gypsum suggest the involvement of prokaryotes in gypsum formation (van Driessche et al., 2019;Lepinay et al., 2018;Mansor et al., 2018;Thompson & Ferris, 1990). In particular, the oxidation of reduced sulfur species by sulfide-oxidizing bacteria may supply, at least in part, the sulfate needed for gypsum precipitation (e.g., Lepinay et al., 2018;Mansor et al., 2018). Such microbial oxidation has been suggested to promote gypsum supersaturation in bottom waters and to favor the formation of gypsiferous thrombolites in a hypersaline lagoon in Venezuela (Petrash et al., 2012). Since we interpret the fossilized filaments in Messinian gypsum as the remains of sulfideoxidizing bacteria, a possible microbial involvement in gypsum precipitation should be taken into account in future studies.

| CON CLUS IONS
Primary Messinian bottom-grown gypsum deposits, formed during the first stage of the MSC (5.97-5.60 Ma), preserve biosignatures of ancient microbial communities, including diverse molecular fossils and mazes of filamentous microfossils. The molecular fossil assemblages of gypsum from different Mediterranean subbasins are unlike typical assemblage of modern marine gypsum deposits forming in shallow-water hypersaline settings. The abundance of lipids of planktic halophilic archaea, planktic thaumarchaea, and a poorly constrained community of benthic archaea confirms that gypsum formed in a stratified basin typified by a normal marine to diluted upper water column and more saline deeper waters. Petrographic F I G U R E 7 Compound-specific δ 13 C values (‰ vs. V-PDB) of biomarkers from gypsum. For n-alkanes, n-alcohols, and n-carboxylic acids, filled and empty bars indicate the range of values for longchain (carbon atoms >26) and shortchain (carbon atoms <21) compounds, respectively. i, iso and mineralogical observations, the overall biomarker inventories, and compound-specific carbon stable isotope patterns agree best with the interpretation that the superabundant filamentous microfossils enclosed in Messinian gypsum are the remains of sulfideoxidizing bacteria. Sulfide-oxidizing bacteria and sulfate-reducing bacteria were probably the main constituents of chemotrophic microbial mats on the Messinian seafloor.

ACK N OWLED G M ENTS
We thank Jean Marie Rouchy (Paris) for providing gypsum sam- Comments by Subject Editor Jake Bailey and two anonymous reviewers helped improve the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare that there is no conflict of interest regarding the publication of this article.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data of this article are available in PANGAEA Data Publisher.