Dubiofossils from a Mars‐analogue subsurface palaeoenvironment: The limits of biogenicity criteria

The search for a fossil record of Earth's deep biosphere, partly motivated by potential analogies with subsurface habitats on Mars, has uncovered numerous assemblages of inorganic microfilaments and tubules inside ancient pores and fractures. Although these enigmatic objects are morphologically similar to mineralized microorganisms (and some contain organic carbon), they also resemble some abiotic structures. Palaeobiologists have responded to this ambiguity by evaluating problematic filaments against checklists of “biogenicity criteria”. Here, we describe material that tests the limits of this approach. We sampled Jurassic calcite veins formed through subseafloor serpentinization, a water–rock reaction that can fuel the deep biosphere and is known to have occurred widely on Mars. At two localities ~4 km apart, veins contained curving, branched microfilaments composed of Mg‐silicate and Fe‐oxide minerals. Using a wide range of analytical techniques including synchrotron X‐ray microtomography and scanning transmission electron microscopy, we show that these features meet many published criteria for biogenicity and are comparable to fossilized cryptoendolithic fungi or bacteria. However, we argue that abiotic processes driven by serpentinization could account for the same set of lifelike features, and report a chemical garden experiment that supports this view. These filaments are, therefore, most objectively described as dubiofossils, a designation we here defend from criticism and recommend over alternative approaches, but which nevertheless signifies an impasse. Similar impasses can be anticipated in the future exploration of subsurface palaeo‐habitats on Earth and Mars. To avoid them, further studies are required in biomimetic geochemical self‐organization, microbial taphonomy and micro‐analytical techniques, with a focus on subsurface habitats.


| INTRODUC TI ON
Microbial communities in deep fractured bedrock on Earth serve as model systems in the search for life on Mars, where habitable conditions may be restricted to the subsurface (e.g., Stevens & McKinley, 1995). We will not be able to sample any extant deep biosphere on Mars in the near future, but the fossil remains of an ancient deep biosphere might now be exposed at the surface. On Earth, however, the fossil record of rock-hosted (i.e., "endolithic") subsurface habitats is sparsely sampled and poorly understood.  Ivarsson et al., 2020). Although they commonly lack organic matter, the morphological, textural, and compositional characteristics of these filaments are usually found to meet established "biogenicity criteria" for discriminating true fossils from pseudofossils. They have accordingly been interpreted as fossils and recommended as important analogues in the search for fossils on Mars (e.g., Hofmann & Farmer, 2000;Hofmann et al., 2008;Onstott et al., 2019).
Specific biogenicity criteria applicable to fossilized fracturedwelling (chasmoendolithic) and cavity-dwelling (cryptoendolithic) filamentous organisms were listed by McLoughlin et al. (2010). The additional criteria outlined by Johannessen et al. (2020) for recognizing fossil Fe-oxidizing filamentous bacteria in hydrothermal cherts may also be relevant in these settings. However, it was recently shown that Fe-oxyhydroxide tubules produced in simple, abiotic chemical gardens meet many published biogenicity criteria; for example, they possess circular cross-sections; curving, branching growth trajectories, and swellings (McMahon, 2019;cf. McLoughlin et al., 2010). In general, there are good reasons to doubt that "biogenicity criteria" can be reliably formulated on our present state of knowledge about the formation of lifelike microstructures in geochemical systems; more work in this area is needed to facilitate life-detection in the rock records of early Earth and Mars (Garcia-Ruiz et al., 2002;McMahon, 2019;Rouillard et al., 2021).
Despite these doubts, some reports of fossil assemblages from the deep biosphere are supported by organic geochemical evidence that is difficult to discount (Ivarsson et al., 2012(Ivarsson et al., , 2020. One particularly compeling example-a population of organically preserved archaeal colonies containing biomarker lipids-was recently reported from Cretaceous calcite-brucite veins in drill core extracted from the Iberia Margin (Klein et al., 2015). These fossils were localized to the gradational contact between two rock types, serpentinite and overlying ophicalcite, representing a redox boundary beneath the palaeo-seafloor. Serpentinite results from the hydrolysis and hydration of Fe-and Mg-silicate minerals in mafic and ultramafic rocks ("serpentinization"), while ophicalcite is a strongly oxidized, pervasively calcite-veined alteration product of serpentinite, which forms when serpentinite is fractured and carbonated by seawater mixing with the strongly alkaline fluids generated by serpentinization . This water-rock reaction generates important electron donors for the deep biosphere (H 2 and CH 4 ) while also fracturing the rock and driving the precipitation of pore-and fracture-filling minerals, notably carbonates (Kelley et al., 2005;Klein & McCollom, 2013;McKinley et al., 2000;Stevens & McKinley, 1995). Serpentinization thus enhances both habitability and the potential for fossil preservation in fractured mafic and ultramafic rocks. The discovery of fossils in serpentinization-associated carbonates is of particular astrobiological interest because serpentinization is known to have occurred in some places on Mars, producing serpentine and carbonates detectable from orbit and in meteorites (e.g., Blamey et al., 2015;Changela & Bridges, 2010;Ehlmann et al., 2010). It may continue today, perhaps contributing to sporadic emissions of CH 4 from the martian surface/subsurface to the atmosphere (Fonti & Marzo, 2010;Giuranna et al., 2019;Krasnopolsky et al., 1997;Mumma et al., 2009;Webster et al., 2015).
Here, we report the results of a search for fossils in calcite veins from a Jurassic subseafloor serpentinizing system now exposed subaerially in Liguria, NW Italy. Our sampling strategy focused on the gradational serpentinite-ophicalcite contact, which in Liguria is geochemically and structurally analogous to the younger Iberia Margin material in which Klein et al. (2015) found fossiliferous veins (e.g., Schwarzenbach et al., 2012Schwarzenbach et al., , 2013. Veins in our samples did not apparently contain organically preserved fossils, but did contain metalliferous mineral filaments (and other microstructures) similar to those previously reported from other ancient fractures and pore spaces (Milliken, 2001(Milliken, , 2002Ivarsson et al., 2008;Ivarsson et al., 2011;Ivarsson et al., 2012;Ivarsson et al., 2013).

| Sample localities and acquisition
Samples were collected inland from the coastal towns of Deiva Marina and Bonassola in Eastern Liguria, NW Italy ( Figure 1). The "classic" ophiolite sequence exposed in this region represents the oceanic lithosphere of the Ligurian-Piedmont Ocean, a western segment of the Tethys Ocean, which opened in the Jurassic as a consequence of left-lateral shear between the African and European plates (Bortolotti et al., 2001). At some point in the Jurassic, faulting and unroofing of the oceanic basement allowed seawater to percolate through upper mantle rocks (lherzolitic peridotite), generating the basal unit of the ophiolite sequence, a green-and-black serpentinite of undetermined thickness. Dykes cross-cutting these altered mantle rocks have been dated at 165 million years old (Bigazzi et al., 1972). The serpentinite is locally overlain by an ophicalcite unit known as the "Rosso Levanto", comprising maroon and violet oxidized serpentinite pervasively veined with calcite and variably brecciated (hence, the unit is often called the "Levanto breccia"). Much of this breccia is thought to have been exposed directly at the seafloor; it contains some micritic sediment and at the top grades into a sedimentary breccia (the "Framura Breccia"; Treves & Harper, 1994).
The ophiolite sequence is capped with a Jurassic-Palaeocene sedimentary succession that shows a shallowing-upward trend recording the convergence and closure of the ocean basin. The sequence was fully emplaced onto the continental crust by the middle Eocene (Bortolotti et al., 2001).
Calcite veins were sampled from gradational contacts between serpentinite and ophicalcite exposed in and around road cuts. As such, samples represent the lowermost reaches of the ophicalcite unit and probably formed at least tens of metres beneath the palaeoseafloor. The samples described in the present study were found at two localities from which calcite-brucite veins were sampled and thin sectioned (Figure 1). At "Locality 1", material was sampled from a ~ m-wide, heavily veined float block of serpentinite +ophicalcite at the base of a cliff at 44°11′51″N 9°35′25″E, near the town of Bonassola. This locality is within a 100 m laterally of the Cava Galli quarry, in whose walls the gradational contact between serpentinite and ophicalcite spans a vertical thickness of 30 m (Schwarzenbach et al., 2013). At "Locality 2", material was collected from a similar float piece at 44°13′51″N 9°32′60″E, near the town of Deiva Marina, about 4 km from the first sample and just outside the perimeter of the Cava La Sfinge quarry.

| Optical and petrographic microscopy
Double-polished, uncoated petrographic thin sections of 150 µm thickness (unless otherwise stated) were prepared at the Open University Thin Sectioning Laboratory, Milton Keynes. Optical microscopy was undertaken with a Leica DMLP Reflected/transmitted light polarizing microscope with DFC 420C camera and Leica Application Suite v 4.00, which was also used to measure the diameters of metalliferous filaments. Only filaments with well-defined, non-diffuse boundaries were measured.

| Fluid inclusion paleothermometry
Fluid inclusion studies were performed on doubly polished wafers using a Linkam THMS-600 heating-freezing stage mounted on a Nikon Labophot transmission light microscope. The instrument, equipped with a range of objective lenses including a 100 Å~ lens, was calibrated against synthetic H 2 O (374.1 and 0.0°C) and CO 2 (−56.6°C) standards (Synthetic Fluid Inclusion Reference Set, Bubbles Inc.). The petrography of fluid inclusion assemblages was first examined at low magnifications using a NIKON Eclipse E600 microscope equipped with both transmitted white and incident F I G U R E 1 Geological map showing the localities of filament-bearing calcite vein samples collected inland from Bonassola (1) and Deiva Marina (2). Adapted from Strating and Wamel (1989)

Sample locality Town
Thrust fault ultraviolet light (UV) sources. Ultraviolet light, with an excitation wavelength of 365 nm, was provided by a high-pressure mercury lamp with a 420 nm barrier epi-fluorescence filter that allows only the long-wavelength UV to reach the sample.

| Raman spectrometry
Two Raman instruments were used to investigate sample compositions, and more particularly to determine whether "D" and "G" bands attributable to disordered carbonaceous material could be measured. A Renishaw inVia Raman microscope at the University of Edinburgh was used to acquire Raman spectra from uncoated thin sections. The 514 nm, 2 mW laser beam was fo- Maps were acquired with the spectral center of the detector adjusted to 944 cm -1 , with a motorized stage allowing XYZ displacement with precision of better than 1 μm. Spectral decomposition and subsequent image processing were performed using WITec Project FOUR software, with baseline subtraction using a third or fourth order polynomial. Hematite maps were created by integrating over the ~290 cm -1 hematite Raman band. All analyses were conducted on material embedded below the surface of the thin section to avoid artefacts in the Raman spectra resulting from polishing and/or surface contamination.

| Scanning electron microscopy
Scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDS) of carbon-coated polished, uncovered thin sections were undertaken at the University of Glasgow using a Zeiss Sigma analytical SEM equipped with an Oxford Instruments X-Max silicon drift energy dispersive X-ray detector and an operating voltage of 20 kV.

| Focused ion-beam milling and transmission electron microscopy, elemental mapping and electron diffractometry
TEM samples were prepared using a FEI Helios Nanolab G3 CX dual beam (FIB-SEM) instrument in the CMCA at UWA. Electron beam imaging was used to identify microstructures of interest in the polished thin sections coated with c. 10 nm of gold, allowing site-specific TEM samples to be prepared. The TEM sections were prepared using a slightly modified version of the protocol for microstructure extraction given in Wacey et al. (2012), with milling and imaging parameters optimized to suit the specific type of sample (i.e., calcium carbonate matrix). Briefly, regions of interest (ROI) were covered with a protective (c. 2 μm-thick) platinum layer. Initial large trenches were milled either side of the ROI with a 21 nA Ga+ion beam, and the trench faces cleaned up using a 9.3 nA beam. Element mapping within the FIB-SEM using energy-dispersive X-ray spectroscopy (EDS) was performed on some cleaned trench faces to gain a preliminary understanding of the chemistry of the filaments and their surrounding matrix. SEM-BSE imaging was also performed on cleaned trench faces during the thinning process. On reaching a thickness of c. TEM and STEM (scanning transmission electron microscopy) data were obtained using an FEI Titan G2 80-200 TEM/STEM with ChemiSTEM technology operating at 200 kV equipped with a Gatan SC1000 camera located in the CMCA at UWA. Crystal orientation and mass/density difference data were gained from high angle annular dark-field (HAADF) and bright field (BF) STEM imaging. Energydispersive X-ray spectroscopy (EDX) via the ChemiSTEM system provided elemental maps and spectra. Lattice spacings of crystals to help determine mineralogy were obtained via selected area electron diffraction (SAED) using an aperture that selected a 600 nm field of view.

| Synchrotron-radiation X-ray tomographic microscopy (SRXTM)
Synchrotron-radiation X-ray tomographic microscopy (SRXTM) was carried out at the X02DA TOMCAT beamline at the Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland. The sample selected for analysis was sawed from a thin section to proper size of about 3 × 3 mm square. A LuAG:Ce 20 µm scintillator was used during the scanning process. For optimal penetration the energy of 28 keV was applied. A total of 1,501 projections were acquired, during rotation of the specimen over 180°, post-processed and rearranged into flat-and darkfield-corrected sinograms. Exposure time per single projection was 1,900 ms. Tomographic reconstruction was performed using a highly optimized algorithm based on the Fourier method , and the obtained tomographic volumes were visualized using Avizo 9.5.0 (FEI Company).
With the 40× lens used, the resulting voxel size was 0.1625 μm.

| Chemical garden experiments and analyses
A soluble polymetallic sulfate mineral (a member of the copiapite family) was obtained from the efflorescent deposits formed by pyrite oxidation associated with acid mine drainage at Baia Sprie, Romania (Buzatu et al., 2016). A small (0.096 g) sample of this mineral was dissolved in a beaker containing 20 ml of 0.9 M Na 2 SiO 3 (aq) at room temperature. Chemical gardens formed in minutes after the addition of the mineral. These were sampled and sputter-coated with a 20 nm layer of iridium. The coated samples were imaged using a FEI Nova 400 field emission scanning electron microscope (SEM) with an accelerating voltage between 10 and 15 kV. All compositional analyses were performed by energy dispersive X-ray (EDX) spectroscopy using an Oxford 100 mm 2 UltimMax SDD EDS X-ray detector. Intensity correlations between pairs of elements were obtained using an in-house MatLab script on coarsened image data (coarsening factor of 44). Image areas with very low element abundance were ignored in this analysis.

F I G U R E 2
Photomicrographs (a-i) and synchrotron X-ray microtomography renderings (j-l) of filament assemblages in calcite veins. All photomicrographs were taken in plane-polarized transmitted light. (a) Calcite vein with filaments from Locality 1, showing somewhat blocky calcite with irregular median suture (arrowed). (b) Opaque filaments clustering on both sides of Calcite-1 within a vein mantled by brown, fibrous Calcite-2 (Locality 1). (c) Curving, connected filaments (Locality 1). (d) Filaments from Locality 2. (e) A curved, possibly branched filament spanning the thickness of Calcite-1 (Locality 1). (f) A branching filament that swells towards the tip (Locality 1). (g) A branching filament (Locality 1). (h) Branching filaments with patchy preservation; arrow indicates gap (Locality 1). (i) An unusually dense population of filaments (Locality 1). (j) A three-dimensional view of the filaments in I showing curved trajectories; image is a cropped still frame from the supplementary movie. (k) An alternative view of the same region as i and j, showing numerous branches and interconnections; three contiguous regions are highlighted in color. (l) Another view of the same region as i, j, and k; white arrow indicates a conspicuous gap in one filament. This field of view shows an "X"-shaped pseudo-junction (yellow arrow) where filaments grazing each other are cemented together (and/or merged by the process of computing a surface through the rather diffuse filament margins). Some apparent connections seen in panel 2K may also result from these effects. Scale bars: 500 µm (a); 100 µm (b-i); 50 µm (j, k); 20 µm (l)

| Petrography and fluid inclusion paleothermometry
In samples from both localities, calcite veins containing metalliferous filaments vary up to about 2 mm in thickness, and contain two distinct generations of calcite ( Figure 2a) as well as some minor brucite. The remainder of the host rock comprises ~90% serpentine, with the rest dominated by relict clinopyroxene, orthopyroxene, and olivine, together with hematite, chlorite, and minor secondary minerals. The calcite veins appear to correspond to the "ribbon" veins described from the same ophicalcite unit by Treves et al. (1995), who noted the presence of "scattered hematite crystals". The inner part of every filament-bearing vein is composed of colorless, cloudy, inclusion-rich, blocky or weakly fibrous calcite ("Calcite-1") with well-  Table S1; Figure S1).

| Filaments: morphology and composition
At both localities, Calcite-1 contains slender, curving, opaque filaments composed of metalliferous minerals. These filaments are connected to the vein walls, to the contact between Calcite-1 and Calcite-2, to patchy opaque wall-lining cements, and to each other where they occur on both sides of the vein, usually oriented nearly orthogonally to the walls. One filament was observed bridging the full width of the Calcite-1 portion of a vein ( Figure 2e). Filaments range up to a few hundred µm in length within the field of view afforded by thin sections. Of 86 filaments measured, the median thickness was 4.7 µm; the mean was 5.9 µm. 55% were between 2 and 5 µm; the rest were thicker (up to a maximum of 19.0 µm; the distribution is positively skewed). Thickness is typically constant along the length of each filament, but a few swell towards the tip (Figure 2f).
Branching is common both towards and away from the vein walls and occurs at various angles, typically >45°. (Figure 2b,c,e-k). Rarely, filaments occur in densely packed clusters (Figure 2i). SRXTM revealed that filaments within these clusters are strongly interconnected by apparent branches (Figure 2j,k; Video S1). Many filaments are patchy or diffuse (e.g, Figure 2h), and some are interrupted by short gaps (e.g, Figure 2h,l). Some neighboring filaments glance each other tangentially but can misleadingly appear to merge because of their diffuse microcrystalline margins, particularly in SRXTM renderings that interpolate a surface around these margins. Thus, close scrutiny of the apparent "X"-junctions in these renderings shows that the filaments do not penetrate each other centrally (e.g., Figure 2l). Many of the branches appear genuine, however.
Reflected-light optical microscopy, SEM/STEM imaging and EDS mapping reveal that the filaments are composed of two minerals present in two or three concentric layers, with a circular or slightly

| Timing of vein and filament formation
Previous work has established that the calcite veins investigated here formed in the subseafloor during the carbonation of serpentinite in the Jurassic; a similar process is responsible for the stockwork of carbonate veins underlying the Lost City Hydrothermal Field Kelley et al., 2005;Schwarzenbach et al., 2012Schwarzenbach et al., , 2013Treves et al., 1995). The inclusion-rich weakly fibrous nature of Calcite-1 is typical for marine carbonate cements. The inference of a subseafloor origin is also supported by petrographic comparisons between our samples and carbonate veins previously reported from drill cores through the oceanic crust (e.g., Eickmann et al., 2009;Ivarsson et al., 2019;Milliken, 2001Milliken, , 2002. This interpretation is suggested by the following observations. Firstly, calcite-1 has a blocky texture and variably distinct median suture, consistent with syntaxial growth by cavity infill. Secondly, all the observed filaments either visibly attach to the vein walls or are oriented such that they probably attach to it outside the plane of the thin section, that is, they are not simply "floating" in the middle of the veins or penetrating through them in the manner of recent microboring endoliths. One filament is observed to span the whole width of the Calcite-1 vein-fill, connecting the opaque cements on both sides (Figure 2d). Thirdly, some filaments (e.g., Figure 2g) are interrupted by short gaps without any concomitant interruption in the calcite, implying that the calcite post-dates the filaments (and calcite precipitation may have disrupted them). Fourthly, bright-field STEM images show that individual carbonate grains terminate at the edge of the filaments rather than continuing through them, again implying that the calcite post-dates the filaments ( Figure S2D). Fifthly, the hematite and Mg-silicate that make up the filaments are microscopically disseminated through the serpentine outside the calcite veins.
Fe-oxyhydroxides (potential precursors to hematite) and talc-like F I G U R E 3 Optical and electron microscopy and energy-dispersive X-ray spectroscopy of filaments in calcite veins (Locality 1). (a) Thick filaments intersecting the surface of a polished thin section, imaged in transmitted light ("TL"), reflected light ("RL") and by means of backscattered electron microscopy ("BSE"), revealing a concentric structure that is typical of most filaments at both localities. Element maps obtained by energy-dispersive X-ray spectroscopy show the distribution of Fe (present as hematite), and Mg + Si (probably talc) within the filaments. Al is also slightly enriched in the Mg + Si phase (not shown).

| Are the filaments fossils?
The filaments observed in our thin sections might be interpreted as permineralized fungal hyphae or bacteria that lived as part of the deep biosphere in a Jurassic subseafloor hydrothermal system. This hypothesis explains why the filaments have circular cross-sections, curved, and irregular (non-crystallographic) growth trajectories, and lifelike branching patterns; these are expected features of filamentous microorganisms. The filaments are at least as lifelike in these respects as some broadly analogous assemblages that have been interpreted as fossils (e.g., Bengtson et al., 2017;Ivarsson et al., 2008a, Ivarsson et al. 2008bLindgren et al., 2010;Ivarsson et al., 2015). The filaments are thicker than the Fe-mineralized sheaths of most modern Fe-oxidizers (typically 1-2 µm, to our knowledge; e.g., Vesenka et al., 2018) although these sheaths can accumulate flocculent ferric coatings at least 10 µm thick (e.g., Schmidt et al., 2014).
The large size and branching behavior of many of the filaments in our samples are perhaps more suggestive of fungal hyphae, which typically range from 2 to 27 µm in diameter and branch at high angles De Ligne et al., 2019;Nicolson, 1959). The interpretation of the filaments as fossil fungi has numerous precedents in other mineralized filament assemblages from subseafloor settings (e.g., Ivarsson et al., 2011Ivarsson et al., , 2012Ivarsson et al., , 2013. The hematite in the filaments and vein-wall-lining cements might be thought to have originated as Fe oxyhydroxides precipitated by Fe-oxidizing bacteria. Marine biofilms can also mediate the formation of magnesium silicate minerals such as stevensite (a potential talc precursor), which replaces bacterial filaments in some thrombolites (Burne et al., 2014). However, since talc and hematite are expected to form during the aqueous alteration of serpentine (Geilert et al., 2020;Schwarzenbach et al., 2012), and since they are not restricted to the filaments or veins in our samples but occur commonly  et al., 2008a, 2011, 2012, 2013). It is important to note that filaments in some Eocene calcite veins in these seamounts consist largely of remnant organic matter, including chitin, which can be detected with fluorescent stains (Ivarsson et al., 2012). This observation implies that these organically preserved specimens are fossilized chasmoendolithic fungal hyphae. The fact that they are variously associated with, coated, and replaced by Fe oxides and clay minerals also implies a biological origin for morphologically and contextually similar filaments that are preserved inorganically and lack organic carbon (Ivarsson et al., 2008a). Purported filamentous microfossils consisting of an iron-rich core mantled by a silicate/aluminosilicate phase have also been reported from the non-marine deep biosphere (e.g., Lindgren et al., 2010;Drake et al., 2017). Drake et al. (2017) describe probable fungal hyphae from a deep granitic borehole that offer a snapshot of the fossilization process in action; some filaments are still organic in composition, while others are replaced to varying degrees by Fe oxide (forming a "central strand") and Fe-, Mg-, and Carich clay minerals (forming a "margin").
In carbonate veins from subseafloor drill-cores, Frutexites structures are commonly found in close proximity to filaments (e.g., Eickmann et al., 2009;Ivarsson et al., 2019). Although more finely laminated and fractal-like, some of these Frutexites are broadly similar to the globular structures described in the present study (Figure 4), even showing alternating bands of Fe-oxides and Mg-silicates with minor Al . Frutexites is generally interpreted as a kind of "microstromatolite" formed by the growth of Fe-oxidizing bacterial biofilms, although the precise role of abiotic versus biotic processes in forming these structures is unclear (e.g., Ivarsson et al., 2020;Rodríguez-Martínez et al., 2011).
The simple morphology of the globular, laminated structures in our samples is less suggestive of microbial growth than those described by, for example, Eickmann et al. (2008), Bengtson et al. (2014), and Ivarsson et al. (2019), and more reminiscent of abiotic processes.

| Are the filaments pseudofossils?
The hematite + Mg-silicate filaments might also be interpreted as fossilized metal-silica membrane structures related to "chemical gardens": abiotic self-organized structures that grow through chemiosmotic reaction-diffusion-precipitation processes under far-from-equilibrium conditions. In a well-known laboratory demonstration, chemical gardens are produced by dissolving "seed" grains of transition-metal salts into alkaline solutions rich in carbonate or silicate. The resulting acid-base reaction produces a semipermeable envelope of colloidal, hydrous carbonate or silica around the dissolving seed. Inflow of water increases the osmotic pressure until the envelope ruptures, expelling a thin, acidic jet around which new membrane material immediately forms. The resulting finger-like tubes rapidly acquire an inner coating of metal (oxyhydr)oxide, which gradually thickens and can ultimately occupy most of the interior space, forming bimineralic tubes and filaments of concentrically organized amorphous silica and metal oxyhydroxides (Barge et al., 2015;Kotopoulou et al., 2021;McMahon, 2019).
Similar "chemobrionic" processes are involved in diverse phenomena, from the production of carbonate and sulfide chimneys at alkaline (serpentinization-driven) and acidic hydrothermal vents (respectively) to the development of tubules and "whiskers" during the corrosion of iron and steel (Barge et al., 2015). In nature, such processes may occur wherever reactive metalliferous mineral particles or solutions come into contact with carbonate-and/or silicarich alkaline fluids (Barge et al., 2015;McMahon, 2019). Recent work confirms that highly alkaline, silica-rich fluids produced during the natural alteration of serpentinite generate filamentous chemical gardens when introduced experimentally to appropriate metal salts (Garcia-Ruiz et al., 2017). Thus, chemobrionic chemistry provides an especially appropriate abiotic explanation for the presence of metalliferous oxide +silicate filaments in our samples, which formed during the alteration of serpentinite. One might speculate that the "seed" material from which filaments grew could have been ferrous sulfide phases that were oxidized to Fe 3+ and SO 4 2during the ophi-  (2019), which were grown at room temperature and pressure from finely sieved (<63 µm) grains of hydrated ferrous sulfate placed into a sodium silicate solution ( Figure 5).
Both the Ligurian filaments and these laboratory-grown chemical gardens show positively skewed diameter distributions, with few filaments <2 µm or >10 µm in diameter, and modal ranges of 3-4 µm.
The Ligurian filaments are structurally similar to abiotic chemical gardens produced from Fe-salts dissolved in silica, in as much as they might be said to consist of silicon-rich "tubes" with iron oxide on the interior (Figure 3a; cf. McMahon, 2019; Kotopoulou et al., 2021). However, the Ligurian filaments also contain Mg and Al as major and minor constituents (respectively) of the silicate layer, but not of the Fe-oxide layer. To our knowledge, chemical gardens have not previously been described in the literature from polymetallic Fe-Mg-Al seed material.
We, therefore, conducted a simple experiment to determine whether Mg and Al would partition into the Si-rich layers or the Fe-oxhydroxiderich layers of a chemical garden. We produced tubular chemical gardens by dissolving Fe, Mg, Al-sulfate salt (a member of the copiapite family) in sodium silicate solution, and examined them in cross-section ( Figure 6). EDX counts and maps of the tube walls showed clear chemical zonation into discrete layers dominated by Fe and Si, respectively.
Mg and Al partitioned preferentially into Si-rich layers within the tube walls, while Fe precipitated internally and on outer surfaces (Figure 6a).
For one tube, we computed spatial correlations between different elements using the coarsened EDX maps (Figure 6b). We also cannot exclude the possibility that the Ligurian filaments are alteration products of asbestiform serpentine or clay minerals. It has been observed that fibrous crystals are long and flexible, and can somewhat resemble endolithic microbial filaments (Muscente et al., 2018); they can also radiate from vesicle walls in altered marine basalt (Kurnosov et al., 1995). The filaments in our samples do not consist of single or bundled crystal fibers, as seen in asbestiform minerals, but might perhaps derive from the breakdown and recrystallization of fibers. Speculatively, such an alteration process might soften the edges of serpentine fibers to produce rounded and diffuse cross-sections like those observed in our filaments. Contiguity with the wall-lining cements could be explained by the concomitant alteration of the non-fibrous serpentine in the fracture walls. However, the constant-thickness branching and rare "swellings" observed in many filaments are difficult to explain by this mechanism, rendering an asbestiform mineral origin less likely.
The Frutexites-like globular and botryoidal laminated structures in our samples are morphologically simple and might also be explained by abiotic self-organization processes. Indeed, stromatolite-like forms characterized by convex laminations are mimicked by a variety of abiotic growth processes including diffusion-limited aggregation and oscillating chemical reactions (Chan et al., 2019;Johannessen et al., 2020;McLoughlin et al., 2008;Papineau, 2020). Fe-oxide "plumes" in agates can show very similar features to Frutexites (including microscopic laminations) and have generally not been interpreted as microbial in origin, although these materials are very poorly understood.

| Biogenicity criteria
Several lists of biogenicity criteria have been developed to assess the origins of candidate biosignatures. Many workers have emphasized that the (palaeo) environment must be habitable for the The apparent absence of carbonaceous residues in the filaments is a clear failure to meet a widely accepted biogenicity criterion, and makes their biogenicity more doubtful than several other examples of similar filament assemblages that do contain organic matter (e.g., Ivarsson et al., 2012Ivarsson et al., , 2018. Nevertheless, this absence does not refute the hypothesis that the structures in our samples are biogenic, for (at least) the following three reasons.
Firstly, the mineral filaments and sheaths excreted by modern Feoxidizing bacteria commonly contain only a little organic matter and are vacated by cells during life (Chan et al., 2011;Emerson & Moyer, 2002;Picard et al., 2015). Secondly, the abundance of hematite in our samples indicates oxidizing conditions that would not have been conducive to the preservation of reduced carbonaceous matter; the total organic carbon content in the Ligurian ophicalcites is typically only a few hundred ppm (Schwarzenbach et al., 2013). The lack of organic carbon in some purported bioalteration textures in subseafloor volcanic glass (it is present in others) has likewise been explained as a consequence of decomposition (Thorseth et al., 2003). Thirdly, it is possible that small amounts of carbon were actually present in our samples but escaped detection, perhaps partly because mineral phases such as hematite dwarfed and obscured the diagnostic carbon bands in Raman  Figure S3. Top right: coarsened, normalized element maps show the boxed region. These were used to calculate correlation coefficients; each pixel provides one point in the scatterplots, below spectra (e.g., Brolly et al., 2016), and only a few filaments could be examined by STEM. In any case, even if organic matter had been detectable in our filaments, it would not have shown conclusively that they are biogenic (although analysis of its composition might do so). Fischer-Tropsch-type synthesis occurs in serpentinizing systems and can generate organic solids (Milesi et al., 2016).
Abiotic chemical garden filaments tend to adsorb organic matter and promote its condensation (Kotopoulou et al., 2021). Even if there were demonstrably biogenic organic matter in our samples, it might have been incorporated into abiotic filaments, such that the structures themselves would still not be morphological fossils.
It has been argued that fossil chasmoendolithic filaments can be to the surfaces on which they grew; a slight preference for this direction is retained even in the most densely clustered assemblages ( Figure 2h). This habit of surface-normal orientation has previously been noted in fracture-dwelling filamentous microorganisms but is not a reliable biosignature since abiotic mineral growth processes commonly also proceed normal to the nucleation surface (and might well follow chemical gradients). Thus, the spatial orientation of our filaments does not resolve their biogenicity.

| Dubiofossils
Having considered the morphology, spatial organization, chemical composition and mineralogy of the filaments in our samples, as well as the geochemical context in which they formed, we remain unable to reject either the hypothesis that they are fossils or the hypothesis that they are abiotic pseudofossils. Both scenarios are plausible, and in our view equally interesting. Established "biogenicity criteria" do not settle the question, although they do suggest ways in which the Ligurian filaments represent less compelling evidence for life than they might have done, for example, had they contained organic carbon.
Historically, there have been two schools of thought on how to evaluate and present such ambiguous materials in the literature. One popular view is that they simply "should not be accepted as being of biological origin until possibilities of their non-biological origin have been tested and can be falsified" (Brasier et al., 2004). This approach treats these non-biological "possibilities" as "null hypotheses", and places the burden of evidence on those proposing biogenicity . However, this strategy can actually exacerbate the very risk of "false positives" that it was designed to avert.
We suspect that the available "abiotic null hypotheses" very often do not fit the facts-and so are "rejected" in favor of biogenicitybecause they are the wrong null hypotheses. In other words, many fossil-like microstructures are likely to be the products of abiotic processes that have not yet been discovered or brought to the attention of paleontologists; the right "null hypotheses" are thus not yet available. Perhaps the most unusual aspect of the present case study is precisely the fact that the far-from-equilibrium, highly alkaline geochemical context furnishes a fairly obvious "null hypothesis", that is, that the structures we describe are ancient chemical gardens.
Arguably, however, the same mechanism has been missed in other cases, leading to dubious claims to have successfully "rejected" an abiotic origin for purported fossils (e.g., Dodd et al., 2017).
The alternative strategy recommended by, for example, Hofmann (1972) and Buick (1990), is to describe inconclusively fossil-like features as "dubiofossils" (Hofmann, 1972). The use of this neutral term avoids shoehorning ambiguous objects into favored categories on the basis of inconclusive (or cherry-picked) evidence or the falsification of inappropriate "null hypotheses". Brasier et al. (2004) expressed concern that "the dubiofossil concept may encourage fuzzy thinking in an area (the prebiotic-biotic boundary) where this now urgently needs to be avoided." Admittedly, anything resembling a fossil is ultimately a fossil or a pseudofossil, and we should not elide this distinction. But the term "dubiofossil" remains a helpful stopgap, like the paleontological term "problematicum". If desired, the dubiofossil category can be made more precise by scoring the quality of the evidence using biogenicity criteria, even if they do not finally compel either a biotic or an abiotic interpretation (Buick, 1990;McLoughlin & Grosch, 2015;Neveu et al., 2018). Thus, in Buick's (1990) (2015) present a hierarchical scheme for evaluating the biogenicity of metavolcanic-and ultramafic-hosted microalteration textures.
These criteria concern, in the following order, "textural context and syngenicity," "morphology," "geological setting," "geochemistry," and "putative growth pattens,"; with further subdivisions addressing different aspects of each of these themes. On our reading (adjusting for the fact that McLoughlin and Grosch focus on bioalteration textures and encrusted spheroids rather than filamentous body fossils), the Ligurian filaments reported here would qualify as "tentative biosignatures" on this scheme, since they are morphologically complex but lack geochemical evidence for life. We prefer "dubiofossils" to "tentative biosignatures" since "dubio-" (i.e., doubtful) seems a more apt qualifier than "tentative" (i.e., provisional) in this case, but the underlying concept is clearly similar.
Regardless of the preferred nomenclature, if dubiofossils are found on Mars, they should be presented neutrally and neither overinterpreted as fossils nor prematurely dismissed as pseudofossils until scientific advances resolve their biogenicity satisfactorily. If known processes of pseudofossil formation are relevant to the object of study then by all means they should be investigated and falsified if possible, but this is not necessarily enough to show that the object is or is not a fossil. Ambiguous results are to be expected in lifedetection research and must be handled with appropriate objectivity (Cleland, 2019).

| CON CLUS I ON: DUB I OFOSS IL S ON MAR S?
The occurrence of dubiofossils-and lack of definitive fossils-in the serpentinization-associated calcite veins of the Ligurian ophiolite highlights the risk that similarly ambiguous microstructures may be found in samples returned from geologically analogous targets on Mars. Serpentinisation-associated carbonates occur in the Northeast Syrtis region of Mars, which was a shortlisted landing site for the Mars 2020 rover (Perseverance) partly in the hope of obtaining samples from rock-hosted subsurface palaeohabitats (e.g., Onstott et al., 2019). Although this site was rejected in favor of Jezero Crater, recent work has highlighted the presence of olivine-carbonate units in Jezero Crater, which may result from serpentinization driven by magmatic or impact-related heating (Brown et al., 2020). Recovery of samples from this material, if it can be accessed, may well be worthwhile as part of a comprehensive biosignature search strategy, but may lead to agnostic conclusions like those of the present study. To avoid such an impasse, more work is still needed to advance our understanding of abiotic mineral morphogenesis, microbial taphonomy and micro-analytical paleobiology. The resulting data will inform the development of new, more definitive and robust protocols for biogenicity determination, for example, multiparametric statistical comparisons that show quantitatively whether particular dubiofossil assemblages more closely resemble true fossils or pseudofossils (Rouillard et al., 2021). In the meantime, we recommend that fossillike structures whose biogenicity cannot be determined should be reported as such, and not overinterpreted either as biological or non-biological objects. analyses. We thank three anonymous reviewers for their comments, which greatly improved the manuscript.

CO N FLI C T O F I NTE R E S T
No competing financial interests exist.