Samples representative of the bedrock were collected from the ancient Etruscan necropolis of Tarquinia, a UNESCO World Heritage Site (Viterbo, Italy), in which more than 200 painted hypogeal tombs (dated from the 7th to the 2nd century BC) were discovered. The tombs were excavated in a sedimentary bedrock belonging to a middle- to upper-Pliocene formation known as macco consisting of yellowish bioclastic calcarenites interbedded with hybrid sandstones (Fig. S1 in the Supplement shows the map of the necropolis in Tarquinia and the sampling locations). From each location, rock samples were collected and kept in plastic bags on ice and transported to the laboratory for analysis. With a sterile hammer and chisel, surface material and samples from outdoor and indoor rocks were first removed to a depth of 3 to 5 cm. The interior of the rock samples was processed for DNA extraction or geological experiments.

To study the organomineralisation in the laboratory, bacterial strain cultures are maintained in an LB (Luria Bertani) liquid medium (1 % Bacto Tryptone, 0.5 % di yeast extract, 0.5 % NaCl, and 0.1 % NaOH 1N). The carbonatogenic activity of bacterial strains has been assessed in a solid complete medium YPDuc plates, containing 1 % Bacto Peptone, 1 % yeast extract, 2 % glucose, 4 % urea, 2.5 % CaCl2, and 2 % agar, adjusted to pH 8.0. To assess the carbonatogenic activity of the microbial community, 0.1 g of crushed calcarenite was inoculated in a BPuc medium (0.72 % Bacto-peptone, 4 % urea and 2.5 % CaCl2, pH 8.0) for 1 week at 28 ∘C, at 160 rpm min-1. The experiments were performed in three technical and three biological replicas. These media induce the metabolism of urea of the microorganisms, resulting in a fast organomineralisation in plates or in liquid medium.

For each sample, rock material was collected aseptically using a sterile rock hammer or chisel and stored in sterile collection bags. All samples intended for DNA extraction were collected by discarding the top 3–5 cm layer and then crushing the inner part with a sterile rock hammer and further reducing it to a powder by grinding with a sterile mortar and pestle. Genomic DNA extraction was performed using the DNeasy PowerMax Soil Kit (QIAGEN) following the manufacture manual method using about 10 g of collected material. Spectrophotometric quantification was performed using a Thermo Scientific NanoDrop spectrophotometer (Thermo Scientific), and DNA purity was assessed through the evaluation of 260/280 and 260/230 absorbance ratios. PCR amplification was performed on about 50 ng of DNA from each sample, as described by Grottoli et al. (2020). Amplicon sequencing was performed on PCR products deriving from the 16S rRNA sample regions. PCR products were sequenced through Oxford Nanopore Technologies (ONT) by a MinION sequencer. A total amount of 200 ng of PCR products for samples was used for sequencing. Rapid barcoding of PCRs was carried out following the protocol released by ONT (SQK-RBK004) and sequenced by a Flongle Flow Cell (FLO-FLG001). The total run produced ∼15 Mb of data (∼ 34 k reads), including ∼ 2.8 Mb (∼ 4 k reads) for moonmilk of the Tomba Maggi 2, ∼ 2.8 Mb (∼ 4 k reads) for moonmilk of the Tomba degli Scudi, ∼ 2.7 Mb (∼ 11 k reads) for hybrid sandstone, and ∼ 6.4 Mb (∼ 15 k reads) for calcarenite, respectively. The Kraken2 system (Wood et al., 2019) produced a taxonomic classification of 98.83 %, 96.65 %, 44.02 %, and 62.29 % of total reads for each sample, moonmilk of the Tomba Maggi 2, moonmilk of the Tomba degli Scudi, hybrid sandstone, and calcarenite, respectively.

Polarised light microscopy (PLM) observations of rock samples and moonmilk speleothems have been carried out on thin sections at the Dipartimento di Scienze della Terra of the Sapienza University of Rome by means of Olympus BX50 (Japan) and Zeiss AxioPhot (Germany) transmitted polarised light microscopes, under plane-polarised light and crossed polars, at 25×, 40×, 100×, 200×, and 400× magnifications. Microscopic images were obtained by using a digital camera. Polished thin sections (28 mm × 48 mm) were prepared from the sample vacuum impregnated with epoxy resin before cutting and thinning. Density and porosity were measured using an Ultrapyc 5000 helium pycnometer from Anton Paar with an accuracy of 0.02 % and repeatability of 0.01 %. Bulk density was obtained by dividing the dry mass of the sample by its total volume. Grain density resulted from the calculation of the mass/measured volume ratio of the pulverised matrix. Both total and effective (open) porosity was measured. The latter was obtained by dividing the difference between the geometric volume and the volume measured by the pycnometer of the sample (Ruggieri and Trippetta, 2020; Trippetta et al., 2020). All laboratory measurements were made in the Earthquake Physics Laboratory at Sapienza Earth Sciences Department. Bulk rock major and trace-element compositions were obtained by lithium metaborate–tetraborate fusion (inductively coupled plasma–atomic emission spectroscopy (ICP-AES) and ICP–mass spectrometry (ICP-MS)), at Activation Laboratories Inc. (Ontario, Canada) according to the Code 4Litho code package on solutions prepared with lithium metaborate fusion. Loss on ignition (LOI) was measured according to standard gravimetric procedures. Details on the precision and accuracy of the analyses are reported in https://actlabs.com/ (last access: 22 November 2022). The calcium carbonate content was assessed by gasometric measurements using a Dietrich–Frühling calcimeter measuring 1 g of the bulk sediment, following Siesser et al. (1971).

Scanning electron microscopy (SEM) was performed on the Moonmilk samples and on the rock thin sections using the field emission scanning electron microscopy (FESEM) Zeiss Auriga 405, with a chamber room that maintains a pressure of about 10–5 to 10–6 mbar. Before mounting the samples inside the microscope, the specimens were coated with 20 nm of chromium using a Quorum Q150T sputter. Chromium has a high X-ray Kα value (5.145 keV), so it does not interfere with lighter elements during the EDX (energy dispersive X-ray) analysis. EDX spectra were obtained using a Bruker QUANTAX detector in point mode for 30 s, with the electron microscope acceleration voltage set at 10 kV and a working distance of 6 mm to optimise the number of the incoming X-ray signals.

The area of the Monterozzi necropolis offers a favourable chance to investigate the role of bedrock in the genesis of the moonmilk. To this purpose, we analysed samples of bedrock taken from different areas of the necropolis with special focus on the sites where the Tomba dei Vasi dipinti, Tomba Maggi 2, Tomba delle Pantere, and Tomba degli Scudi were carved (Fig. S1). The hypogeal tombs of the Monterozzi necropolis, located on a flat elevated area, are excavated within a sedimentary substrate of middle- to upper-Pliocene age known as macco formation consisting of two main lithofacies showing lateral and vertical heterotopic relationships. Macco s.s. lithofacies is a bioclastic calcarenite represented by packstone to rudstone and floatstone composed of a small volume (< 5 %–10 %) of micrite matrix, coralline algal branches, bryozoans, bivalves (pectinids and oysters), echinoids, benthic foraminifers, and skeletal debris. Rare non-carbonate grains are present. In the inner walls of the intergranular voids, microsparite cement precipitation and/or recrystallisation often occurs in a phreatic marine environment. The second lithofacies is represented by a poorly cemented, crudely stratified, hybrid sandstone. It is fine to medium grain-sized and grain-supported, with a small amount of micrite matrix (< 10 % vol), and nearly devoid of carbonate cement. This lithofacies is characterised by abundant bioturbation, the skeletal assemblage is dominated by small benthic foraminifers, echinoids, serpulids, and bivalves (mainly oysters). Planktonic foraminifers are common. The terrigenous fraction mainly consists of monocrystalline grains of quartz, sedimentary lithoclasts, and subordinate detrital micas and feldspar along with rare glaucony grains and opaques. Calcimetric analyses from the two lithofacies revealed that calcium carbonate contents range from 90 % to 98 % in the macco s.s. calcarenite and from 49 % to 59 % in the hybrid sandstone. Moreover, whole-rock major oxide and trace-element compositions highlight the geochemical difference between the two lithofacies (Table 1), mostly related to the higher proportion of terrigenous fraction in the hybrid sandstone. Helium pycnometry revealed a high open porosity in both lithofacies (∼ 43% for macco s.s.; ∼ 42 % for hybrid sandstone). The macco s.s. lithofacies show a dominance of vuggy porosity and abundant intraparticle, interparticle, and mouldic porosity. The main porosity of hybrid sandstone is represented by interparticle porosity, rarely by mouldic porosity. Thus, despite the differences in composition, the calcarenite and the hybrid sandstone show two essential characteristics required for moonmilk formation: high calcium content, which activates the microbial metabolism leading to organomineralisation, and high porosity, necessary for the exchange of fluids and nutrients in an oligotrophic environment, also providing the space for microbial colonisation. Noteworthily, the characteristics of bedrock porosity (vuggy and mouldic) indicate that the dissolution processes prevail on those of inorganic carbonate precipitation; indeed, meteoric cements as well as speleothems in the largest cavities are absent. Moreover, the bedrock where the tombs are carved is located in the shallow vadose zone (a few metres below the surface). The high magnesium calcite, the dominant mineralogy of the main components, the coralline algae, is a metastable mineral phase of calcite. This phase, upon exposure to meteoric water, dissolves, partially increasing the availability of Ca2+ for the microbial metabolism and for the biogenic carbonate precipitation. It is notable that the area under investigation lies on a high flat relief, where the infiltrated water mostly derives from rainfall, without any groundwater input. Consequently, the infiltrated water can be reasonably assumed to be undersaturated with respect to calcite.

Mostly, moonmilk biogenic deposits develop indifferently within the two distinct lithofacies constituting the bedrock of the Etruscan necropolis of Monterozzi (Mura et al., 2021). Figure 1a shows an example of the hypogeal walls of tombs carved in calcarenite and hybrid sandstone and the walls covered by moonmilk in the Tomba degli Scudi and Tomba Maggi 2 (Fig. 1b). The moonmilk layer originating from hybrid sandstone is thinner than the one observed on a calcarenite substrate, but the scanning electron micrographs of the moonmilk sampled from the walls of the Tomba degli Scudi and Tomba Maggi 2 showed the same nanofibre structure (Fig. 1c). X-ray powder diffraction analysis (XRD) revealed that moonmilk is composed of calcite (Mura et al., 2020).

So far, the moonmilk has only been considered to be a deposition covering rock surfaces (Borsato et al., 2000), but the analysis with transmitted polarised light microscopy of thin sections showed that the moonmilk is present inside the calcarenite in the vuggy and mouldic porosity of the calcarenite in the Tomba Maggi 2 (Fig. 2a) and in intergranular and mouldic pores of the hybrid sandstone bedrock collected inside the Tomba degli Scudi (Fig. 2b). The presence of moonmilk inside the rocks is a general phenomenon because it is observed in all samples, irrespective of the type of rock, calcarenite or hybrid sandstone, and collection site (outdoor or indoor) (Figs. S2, S3, and S4). These results are also supported by the discovery of moonmilk deep inside a calcarenite rock sampled at the entrance of the Tomba dei Vasi Dipinti (Fig. S5).

The analysis of the rock substrate sampled outside of the Tomba dei Vasi Dipinti also suggests that the moonmilk may be contributing to the authigenic carbonate growth in the host rock, covering the inner walls of the voids (Fig. 3).

If the moonmilk observed inside the rocks (a location that was not reported before) is of biogenic origin, traces of organomineralisation would be expected. Indeed, the SEM analysis on the thin sections of calcarenite sampled outside of the Tomba dei Vasi Dipinti revealed many structures (nanofibres that originate from bacteria encased in a calcite structure) corresponding to organomineralisation (Figs. 4 and S6). Bacterial organomineralisation was also detected in the sandstone sampled inside the Tomba degli Scudi (Fig. S7) and in calcarenite sampled outside of the Tomba delle Pantere and inside of the Tomba Maggi 2 (Fig. S8), suggesting that this is a common phenomenon. Such bacterial organomineralisation is also known as “entombment” (Barton and Northup, 2007), and it is easily observed in laboratory settings when bacterial strains are subjected to environmental conditions favouring calcium carbonate precipitation (Fig. S9a, b, c). In plates, the precipitation of calcium carbonate occurs even at a considerable distance from the bacterial colony, possibly by the diffusion of extracellular enzymes known to be involved in calcium carbonate metabolism (Dhami et al., 2014; Rodriguez-Navarro et al., 2019) (Fig. S9d).

It remains unclear how the moonmilk nanofibres are produced in natural environmental conditions because to date it has been impossible to reproduce their formation in the laboratory. In fact, bacterial strains cultured from rocks represent only a negligible fraction of the total microorganisms present in the rocks. Instead, the entire microbial community, with a metabolism that sustains the growth in the rock environment, is needed to precipitate and/or dissolve calcium carbonate. Indeed, under laboratory conditions, we have evidence that the ground calcarenite, with its entire microbial community, when present in a medium containing urea and CaCl2, produced calcite. In the same conditions, calcium carbonate is not produced with sterile (autoclaved) ground calcarenite (Fig. S10) (Benedetti et al., 2023). These and previous results (Banerjee and Joshi, 2014) showed that inactivated (dead) cells were unable to precipitate CaCO3 in the laboratory, suggesting that cells need to be metabolically active for calcification and that cell structure alone is not sufficient to promote bioprecipitation. The most studied bacterial metabolism for CaCO3 precipitation is the ureolytic metabolic pathway. This process involves the production of carbamate (NH2COOH) by urea hydrolysis, which spontaneously hydrolyses to form ammonia (NH3) and carbonic acid (H2CO3). These products react with water to form carbonate (CO32-), ammonium ions (NH4+), and hydroxyl ions (OH-), finally resulting in an increase of pH. In an alkaline environment, the presence of calcium ions and the bacterial cells, as a nucleation site, allow for the precipitation of calcium carbonate (Anbu et al., 2016; Hammes et al., 2002; Nigro et al., 2022). Thus, it is not the presence of the bacterium alone, as a structure, that provides evidence of precipitation, but also high pH, metabolism, and negative membrane charge. In laboratory conditions (Fig. S9), if urea and calcium are present in the medium, the mechanism is stimulated and accelerated.

In natural environments, bacteria are centres of nucleation for nanofibre formation. They do not control the mineralisation process directly but induce the precipitation of calcium carbonate by changing the chemistry of the environment as a consequence of their metabolic activity and also serve as nuclei for crystallisation. This mechanism is a result of pH increase, a direct effect of the negatively charged bacterial surface, and the presence of a metabolic ureolytic process (Omoregie et al., 2021).

Overall, our results underscore the role of microorganisms in promoting moonmilk deposition, contributing to the rock formation processes. Nevertheless, to propose moonmilk as part of a geological process, the microbial communities of the rocks and those contributing to moonmilk deposition should have a similar composition. Aiming to identify the rock microbial communities, samples from calcarenite and hybrid sandstone were analysed together with the corresponding moonmilk samples from the Tomba Maggi 2 and Tomba degli Scudi. The results of the 16S small subunit (SSU) rRNA amplicon sequencing showed a high abundance of Actinobacteria, Bacteroidetes, Cyanobacteria, Firmicutes, and Proteobacteria (Fig. 5 and additional data). Of note, the Firmicutes phylum is abundant, and several members, such as the Lysinibacillus genus, have extremely high urease activity and therefore greatly enhance carbonate precipitation (Banerjee and Joshi, 2014; Benedetti et al., 2023; Zhu and Dittrich 2016), see also Fig. S9d.

Bacterial community diversity was measured by an inverse Simpson index and Shannon index for moonmilk (Tomba degli Scudi and Tomba Maggi 2) and rocks (calcarenite and hybrid sandstone). The indices do not show any significant differences between the samples (Mann–Whitney test, P>0.05) (Fig. S11). These results show that in moonmilk and rocks the microbial composition is similar, irrespective of rock type (calcarenite and hybrid sandstone) or the environment where the samples were collected (outdoor or indoor). It should be noted that 16S SSU rRNA analysis does not provide information about metabolic activity; thus, these data do not identify microorganisms that are active in CaCO3 deposition, but the overall data demonstrate that the endolytic community of the rocks is promoting moonmilk deposition. The results presented also revealed the presence of organomineralisation and calcite nanofibres that originate from bacterial entombment, not only on the surface but also inside the rocks. The presence of a resident microbial community deep within the rocks possibly evolved with the rocks through geological time. Therefore, no habitat should be considered as extreme for the resident microbial community, and the rocks should not be considered as a “refuge” for escaping extreme environmental conditions. Biological research should focus on microbial community evolution with respect to the geologic substrate in which they are living, considering the natural co-evolution of microbes and rocks, and possibly abstaining from the consideration of the microbial metabolism as an adaptation to adverse environmental conditions.

The search for traces of extra-terrestrial life is a complex task often ending with inconclusive results. The co-evolution of minerals and microorganisms has implications for the quest of evidence for life on other planets. The discovery of minerals of undisputable biological origin, rather than organic remains, may provide the most robust signs of biological activity (Hazen et al., 2008). Co-evolution of life and minerals throughout Earth's history lays the foundation for an inclusive search for the presence of life, not only because rocks arose from life but also because life itself may have formed from rocks (Bizzarri et al., 2021; Marshall, 2020; Saladino et al., 2018). Thousands of Earth's minerals owe their existence to the development of life on the planet, and calcium carbonate phases that are massively produced on Earth by microorganisms are the best example (Hazen et al., 2008).

In this work, we have focused on the calcium carbonate nanofibres (moonmilk). Given the tight association and co-evolution between rocks and microbial communities that results in the observed organomineralisation, calcite nanofibres are of interest in the field of astrobiology and are considered to be a potential sign of life. The moonmilk production contributed to rock formation by filling the pores and the cracks in the rocks, while the rock composition and the porosity shaped a microbial community that copes with high calcium content by producing calcite nanofibres. The moonmilk is mainly found in karst caves, but there are also examples of moonmilk bioprecipitation in hypogeal environments carved in different geologic substrates, such as granitoid rocks or sandstone (Miller et al., 2018; Saladino et al., 2018). The moonmilk has also been found in lava tubes, where the microbial communities are similar to those present in the moonmilk that originated from calcarenite (Gonzalez-Pimentel et al., 2021; Miller et al., 2020), raising the possibility of positing the moonmilk as a biosignature also beyond the Earth's calcium carbonate rocks.