On the biogenicity of Fe‐oxyhydroxide filaments in silicified low‐temperature hydrothermal deposits: Implications for the identification of Fe‐oxidizing bacteria in the rock record

Microaerophilic Fe(II)‐oxidizing bacteria produce biomineralized twisted and branched stalks, which are promising biosignatures of microbial Fe oxidation in ancient jaspers and iron formations. Extracellular Fe stalks retain their morphological characteristics under experimentally elevated temperatures, but the extent to which natural post‐depositional processes affect fossil integrity remains to be resolved. We examined siliceous Fe deposits from laminated mounds and chimney structures from an extinct part of the Jan Mayen Vent Fields on the Arctic Mid‐Ocean Ridge. Our aims were to determine how early seafloor diagenesis affects morphological and chemical signatures of Fe‐oxyhydroxide biomineralization and how extracellular stalks differ from abiogenic features. Optical and scanning electron microscopy in combination with focused ion beam‐transmission electron microscopy (FIB‐TEM) was used to study the filamentous textures and cross sections of individual stalks. Our results revealed directional, dendritic, and radial arrangements of biogenic twisted stalks and randomly organized networks of hollow tubes. Stalks were encrusted by concentric Fe‐oxyhydroxide laminae and silica casings. Element maps produced by energy dispersive X‐ray spectroscopy (EDS) in TEM showed variations in the content of Si, P, and S within filaments, demonstrating that successive hydrothermal fluid pulses mediate early diagenetic alteration and modify the chemical composition and surface features of stalks through Fe‐oxyhydroxide mineralization. The carbon content of the stalks was generally indistinguishable from background levels, suggesting that organic compounds were either scarce initially or lost due to percolating hydrothermal fluids. Dendrites and thicker abiotic filaments from a nearby chimney were composed of nanometer‐sized microcrystalline iron particles and silica and showed Fe growth bands indicative of inorganic precipitation. Our study suggests that the identification of fossil stalks and sheaths of Fe‐oxidizing bacteria in hydrothermal paleoenvironments may not rely on the detection of organic carbon and demonstrates that abiogenic filaments differ from stalks and sheaths of Fe‐oxidizing bacteria with respect to width distribution, ultrastructure, and textural context.


| INTRODUC TI ON
Banded iron formations and jaspers record extensive marine deposition of iron in the Precambrian (e.g., Bekker et al., 2010;Klein, 2005).
A large proportion of the iron in these deposits occur as oxidized Fe(III), despite the early atmosphere being virtually devoid of oxygen (Lyons, Reinhard, & Planavsky, 2014 and references therein). The exact mechanisms responsible for the precipitation of ferric mineral phases in the early oceans are yet to be established, but micro-organisms have been suggested to facilitate Fe(II) oxidation (Chi Fru et al., 2013;Holm, 1987;Kappler, Pasquero, Konhauser, & Newman, 2005;Konhauser et al., 2002;Planavsky et al., 2009). Microaerophilic neutrophilic Fe-oxidizing bacteria thrive in suboxic conditions and may have contributed to the generation of iron deposits after the Great Oxidation Event in the Palaeoproterozoic (Chan, Emerson, & Luther, 2016;Crosby, Bailey, & Sharma, 2014;Planavsky et al., 2009;Slack, Grenne, & Bekker, 2009;Slack, Grenne, Bekker, Rouxel, & Lindberg, 2007), or possibly even earlier in the Earth's history (Dodd et al., 2017). The metabolic byproducts of microbial microaerophilic Fe(II) oxidation are extracellular Fe(III)-oxyhydroxide stalks, which are easily recognizable because of their distinct twisted and branching morphologies (Ghiorse, 1984). The morphological uniqueness and the direct connection between twisting and branching traits and fundamental cell behavior, that is, motility and cell division, make the extracellular stalks ideal biosignatures of microbial Fe oxidation in the rock record (Chan, Fakra, Emerson, Fleming, & Edwards, 2011).
As the stalk-forming Fe oxidizers are highly sensitive to O 2 concentrations and therefore require a strong redox gradient for growth, the stalks also function as palaeo-environmental indicators (Krepski, Emerson, Hredzak-Showalter, Luther, & Chan, 2013). Communities of Fe-oxidizing bacteria produce organized microbial mats, which are characterized by directional filamentous textures and alternating sparsely and strongly mineralized bands, reflecting the flux of key substrates and the prevailing hydrodynamic conditions (Chan, McAllister, et al., 2016).
Fe-oxyhydroxide deposits in modern low-temperature hydrothermal environments on the seafloor commonly host Fe-oxidizing Zetaproteobacteria (e.g., Emerson & Moyer, 2002;Forget, Murdock, & Juniper, 2010;Kato, Kobayashi, Kakegawa, & Yamagishi, 2009;Li et al., 2012;McAllister et al., 2019;Scott, Breier, Luther, & Emerson, 2015), which actively contribute to the growth of the deposits through biomineralization and stalk formation (e.g., Alt, 1988;Boyd & Scott, 2001;Edwards et al., 2011;Juniper & Fouquet, 1988;Karl, Brittain, & Tilbrook, 1989;Kennedy, Scott, & Ferris, 2003;Langley et al., 2009). These deposits may be similar to the sedimentary precursors of jaspers and iron formations, both with respect to mineralogy and genesis. Silicified hydrothermal Fe-oxyhydroxide precipitates are particularly valuable as analogues for ancient Fe deposits because biominerals produced by Fe-oxidizing bacteria have a higher potential for preservation when encased in silica. Compelling examples of biogenic stalks with characteristic twisted and hollow morphologies have been reported from modern, silicified hydrothermal deposits (Dekov et al., 2015;Sun et al., 2015Sun et al., , 2012 as well as from ophiolite-hosted Proterozoic and Phanerozoic jaspers (Boyce, Little, & Russell, 2003;Duhig, Davidson, & Stolz, 1992;Little, Glynn, & Mills, 2004;Little & Thorseth, 2002), suggesting that microbial Fe oxidation has remained an essential process in hydrothermal environments over a protracted period of geological time. Other than twisted morphology, biosignatures that have been advanced for microaerophilic Fe-oxidizing bacteria in the rock record include: constant filament widths, narrow width distributions, strong filament directionality, filament direction changes, bending, and bifurcation Hofmann, Farmer, Blanckenburg, & Fallick, 2008;Krepski et al., 2013). However, the morphologically and texturally based identification of Fe-oxidizing bacterial remains is far from unambiguous. Over time, seafloor Fe deposits typically experience diagenetic recrystallization, which obscures the distinct twisted and branching morphological traits that link filament growth to biomineralization by Fe-oxidizing bacteria. Furthermore, abiotic processes have been advocated to explain some of the common textural features in hydrothermal Si-Fe deposits, including dendrites and moss agates (Hopkinson, Roberts, Herrington, & Wilkinson, 1998;Little et al., 2004) and twisted and striated silica filaments (Park, Lee, Cheon, & Park, 2001;Sokolev & Kievsky, 2005), as well as tubes with particulate iron interiors (García-Ruiz, Nakouzi, Kotopoulou, Tamborrino, & Steinbock, 2017) have been generated experimentally by using entirely abiotic processes. Thus, one of the key remaining challenges in the search for microbial Fe oxidation in the rock record is to confidently distinguish true iron biominerals, from abiotic chemical precipitates, in Fe deposits that have experienced diagenetic alteration.
The Jan Mayen Vent Fields (JMVF), located along the southernmost segment of the Mohns Ridge, comprise both actively forming low-temperature Fe-oxyhydroxide deposits (Johannessen et al., 2017;Moeller et al., 2014;Pedersen, Thorseth, Nygård, Lilley, & Kelley, 2010; and extinct chimneys and mounds characterized by interlayered silica and siliceous Fe-oxyhydroxides. Molecular work conducted on the active, low-temperature portion of the JMVF has revealed a high abundance of Zetaproteobacteria, including the stalk-forming Fe-oxidizing bacterium Mariprofundus ferrooxydans, suggesting filaments differ from stalks and sheaths of Fe-oxidizing bacteria with respect to width distribution, ultrastructure, and textural context.

K E Y W O R D S
biomineralization, biosignature, Fe-oxidizing bacteria, microbial textures, twisted stalks that microbial Fe oxidation plays a key role in the formation of the Fe-oxyhydroxide deposits (Johannessen et al., 2017;Vander Roost et al., 2017). The association of deposits with different silica contents provides a unique opportunity to study early taphonomic processes and improve the understanding of how extracellular Feoxyhydroxide stalks are preserved with time. The Si-Fe deposits from the extinct field at the JMVF are ideal targets for reevaluating the robustness of morphological biosignatures, because their excellent preservation aids in the identification of stalks produced by Fe-oxyhydroxide biomineralization. Furthermore, these deposits are formed in a natural hydrothermal environment similar to the inferred depositional setting of Phanerozoic and Proterozoic jaspers (e.g., the Løkken Jasper; Grenne & Slack, 2003) and have experienced silicification and early diagenetic alteration. Through textural analyses and focused ion beam preparation of filament cross sections from the extinct Si-Fe mounds and chimneys, we aim to document how early seafloor diagenesis and silicification affect the ultrastructure and chemical composition of extracellular stalks and how microfossils of Fe-oxidizing bacteria can be distinguished from fossil-like abiotic precipitates.

| Geological setting and sampling
The Troll Wall forms the northernmost hydrothermal field within the Jan Mayen Vent Fields and is located at 71°N and 6°W along the southwestern segment of the ultraslow-spreading Mohns Ridge in the Norwegian-Greenland Sea (Figure 1a). Active venting of 270°C warm, white smoker-type fluids takes place along the fault-bounded margin of a rift that transects the axial volcanic ridge (Pedersen et al., 2010). Extensive areas in the center of the rift are dominated by diffuse, low-temperature venting and microbially mediated deposition of laminated, siliceous Fe-oxyhydroxides and Mn oxides (Johannessen et al., 2017;Moeller et al., 2014). An extinct vent field was discovered ca. 500 m east of the active hydrothermal field in 2011. The extinct field is situated on the crest of the axial volcanic ridge at a depth of ca. 460 mbsl and hosts laminated mounds and columnar chimneys composed of silica and Fe-oxyhydroxides (Figure 1b The Si-Fe chimney was sliced with a chain saw and collected using the robotic arm of the ROV. Samples were air-dried on board after recovery.

| Petrographic microscopy
Sample fragments were embedded in epoxy, and thin sections were prepared following standard procedures. The thin sections were studied under a Nikon Eclipse LV100 polarizing microscope located at the University of Bergen, Norway. Images were obtained by a DS-Fi1 color camera connected to the NIS-Elements BR 2.30 software.
Areas with abundant filaments were selected for analysis of width distribution and were photographed under 200X magnification in plane-polarized light. Filament widths were measured from thin section images using the ImageJ software. For selection of filaments for width measurements, a 10 x10 grid was superimposed on the image. Measurements were undertaken while moving progressively from left to right between the horizontal lines. All filaments intersecting the vertical lines of the grid were included in the analysis and were measured along segments residing in the plane of focus.
One optical photomicrograph of the Si-Fe mound sample and one image of the chimney sample were selected for directionality analysis. Both showed filaments with apparent preferred orientation. The orientation distribution was measured using the local gradient approach of the Directionality plugin in ImageJ (developed by Dr. Jean-Yves Tinevez). Orientations ranging from −90° to 90° were recorded and the results were grouped into 90 bins of 2° each.

| Scanning electron microscopy (SEM)
Textures and detailed morphology of individual filaments were studied using a Zeiss Supra 55VP Field Emission Scanning Electron Microscope (FE-SEM) at the University of Bergen. Prior to analysis, thin sections were coated with C using an Agar Turbo Carbon Coater. Mound and chimney fragments were attached to Al stubs and coated with Ir using a Gatan 682 Precision etching coating system. Thin sections were studied in backscatter electron (BSE) mode, and phases were identified with a Thermo Noran Six Energy Dispersive Spectrometer (EDS). Analyses were performed at a voltage of 15 kV and a working distance of 8 mm. Mound fragments were investigated by using the secondary electron (SE) detector at 5 kV and with an average working distance of 3 mm. The SEM was used to preselect individual filaments for focused ion beam preparation of thin section samples. Intact filaments located in depressions of the thin section were preferred targets.

| Focused ion beam-transmission electron microscopy (FIB-TEM)
Focused ion beam (FIB) preparation of samples for transmission electron microscopy (TEM) analysis was performed at the NTNU NanoLab in Trondheim, Norway. Prior to FIB milling, thin sections were coated with a ca. 30-nm-thick layer of Au. Cross sections of targeted filaments were cut perpendicular to their long axis using a FEI Helios NanoLab 600 DualBeam FIB-SEM. Protective layers of Pt (e-beam assisted deposition) and C (ion-beam assisted deposition) were deposited on the surface of the selected area prior to milling, and the material on each side of the protective strip was removed by sputtering, leaving a sample with a size of ap- Centurio SDD for energy dispersive X-ray spectroscopy (EDS). In order to remove surficial hydrocarbons from the samples, the FIB foils were gently plasma cleaned with a shielding holder in a gas mixture of 75% Ar and 25% O 2 for 30 s prior to analysis. Electron diffraction was performed on both Fe-rich and Si-rich phases to determine crystallinity. Elemental compositions were measured by EDS. The EDS maps were processed in MatLab using the jet color scale from dark blue to dark red with matrix values adapted to the intensity of each set of maps. Images were acquired in both TEM and STEM modes. Minimum and maximum filament diameters were measured using the ImageJ software.

| Laminated Si-Fe mound
The sample fragments from the Si-Fe mounds showed irregular, nodular geometry and displayed clearly defined Si-and Fe-rich layers ( Figure 1d). Each layer was composed of a series of microscopic laminae. Internal layers were primarily composed of fibrous silica with local bands of massive silica. The outer parts of the sample fragments consisted of laminated Fe-rich precipitates with alternating porous and densely mineralized textures. Aligned fibers were evident in some Fe-rich parts of the hand specimen. Small internal cavities were locally present and were particularly prominent in the silica-rich layers (see Figure 1d).

| Si-Fe chimney
The chimney fragments were composed of consolidated, yet porous material and lacked systematic layering showing an apparently random distribution of intermixed silica and Fe-rich precipitates ( Figure 1e). Large channels and concentric layers were not observed, but smaller cavities and crevasses, ranging in size from <1 to 5 mm in diameter, were abundant and could locally be traced from the base to the top of the sample fragment (see Figure 1e). Clusters of channels were evident in some areas and alternated with denser fibrous textures and massive silica-rich domains.

| Laminated Si-Fe mound
Optical microscopy and BSE-SEM of thin sections from the Si-Femounds revealed alternating Fe-rich and silica-rich laminae with complex filamentous and banded textures. Three main, morphologically distinct filament types were evident in Fe-rich laminae ( Figure 2). These filament types were typically confined to separate areas, but were locally found to co-occur. The most prevalent filamentous texture comprised clusters of irregular, gently curved, and branching filaments showing semi-parallel or radiating orientation over lateral distances of <2-3 mm (Figure 2a). The filaments were locally associated with directional dendritic features, measuring 500-1,000 µm across and ca. 500-1,500 µm in the branching direction (Figure 3). BSE-SEM demonstrated that the dendrites were composed of dispersive and radially growing filaments, some of which possessed indications of twisting ( Figure 3h). Networks of randomly oriented, narrow, straight filaments appeared to be nearly equally abundant as the irregular forms (Figure 2c), but the two morphotypes were largely confined to separate laminae. Branching did not appear to be a ubiquitous trait associated with the straight morphology, but the filaments were found to attach to other straight filaments, constructing complex networks ( Figure S1a). Irregular, curved filaments were commonly minor components in domains dominated by straight, randomly oriented filaments, while the opposite association was less frequently observed. Shorter single and multiple branching filaments were present along with both the irregular and straight forms (Figure 2e), and their trunks were typically attached to neighboring filaments. Fe-rich precipitates lacking clearly discernible structure were present in variable amounts on the surfaces of all filament morphotypes. Heavy Fe mineralization partly obscured the filamentous fabric in some areas. Fe-rich bands, ranging in thickness from 5 µm to several mm, were evident in the interface between domains of different texture. The Fe-rich bands locally occurred as stacked, dome-shaped structures ( Figure S1b was locally encrusting Fe-rich filaments. The extent of silica mineralization appeared to increase toward silica-rich laminae, from interstitial precipitates via <100-nm-thick, pitted crusts and >1-µm-thick, bulbous casts to massive domains where filaments were fully embedded in silica (see Figure 2b).
Silica-rich laminae were mainly composed of wide, straight silica filaments, which were lacking Fe-rich cores ( Figure 2g; Figure   S4). These filaments showed strong directionality across distances of more than 2-3 mm, generating hair-like fabric ( Figure S4f). The filamentous fabric locally graded into bands of massive silica. SEM confirmed that large areas of the silica-rich sample fragments were composed of aligned, ca. 2-to 15-µm-wide, hollow tubes with wall thicknesses varying from ca. 200 nm to 2 µm ( Figure 2h). Cross sections of individual tubes were found in the massive silica bands.

| Si-Fe chimney
The textures in the Si-Fe chimney samples differed from the mound textures with respect to the size, directionality, and morphology of filamentous features. Major parts of the chimney samples were composed of radially oriented, semi-linear filaments intersected by narrower, orthogonal filaments that appeared to outline arcs ( Figure 4a; Figure S5). The radial filaments were composed of silica with cores of Fe-rich, particulate material and showed local branching, which increased in intensity toward 10-to 100-µm-wide, silicified, Fe-rich bands and dendrites (Figure 4b remarkably consistent directionality ( Figure 4e). Both filaments with solid and porous interiors were evident. Botryoidal silica surfaces were visible within some pore spaces ( Figure 4f). Tabular planes composed of closely spaced silica strands were locally found at high angle to the main set of branching filaments (see Figure 4e). These planes were strongly mineralized by botryoidal silica. Fe-rich globular aggregates and fibrillar precipitates were evident on the surfaces of silica filaments in some areas, locally forming rounded nests and ca. 5-to 10-µm-wide, hollow tubes with porous, fibrillar walls ( Figure 4g-h, Figure S6).

| Widths and directionality of filaments
The widths of the three most prominent filament types in the Fe-   to have a higher oxygen tolerance than their stalk-forming counterparts (Chan, McAllister, et al., 2016;Fleming et al., 2013). An analogous ecological relationship has been documented in freshwater Fe seeps (Emerson & Revsbech, 1994;Fleming, Cetinic, Chan, King, & Emerson, 2014). In our JMVF mound samples, different geochemical requirements between stalk-forming and sheath-forming Fe-oxidizing species could explain why the twisted and straight filamentous forms occur in spatially separate domains.
The short, branching, tubes we identified are easily recognizable in thin section because of their distinct outward-pointing, arborescent branches (Figure 2e). Although this morphology is highly abundant, it does not form distinct domains like the twisted stalks and straight sheaths, but appears infrequently and always co-occurs with the other forms. Branching tubes have been found in a large number of hydrothermal Fe-oxyhydroxide deposits (Boyd & Scott, 2001; Siliceous Fe precipitates occupy the innermost core region, while the outer core is composed of concentric layers with variable P, C, S, and Si contents. Note that the detection limit for C is approximately 0.1%. The intensity scale ranges from 0 to 256, with blue color representing low element counts and red color representing high element counts the mound. Silicification clearly aids the preservation of extracellular stalks in Fe-rich horizons. We therefore assume that the silicarich layers never contained appreciable amounts of Fe and hence that the tubes are unrelated to microbial Fe oxidation. Instead, we propose that the strict parallel alignment of tubes suggests inorganically controlled precipitation. Chemical gardens are wellknown for complex inorganic tube structures that are produced by the simple addition of a multivalent metal salt to an alkaline silica solution (e.g., Barge et al., 2012;Cartwright, García-Ruiz, Novella, & Otálora, 2002). Considering the high content of both silica and metal ions in the low-temperature fluids emanating from seafloor hydrothermal systems, these are prospective candidates for natural growth of chemical gardens.

| Fe mineralization and adsorption
This study is the first to use FIB-TEM to investigate the interiors of silicified extracellular stalks of Fe-oxidizing bacteria and thereby adds another dimension to the biogenicity assessment. STEM-EDS mapping and TEM diffraction analyses ( Figure S8) confirm that the stalks in the Jan Mayen mound deposits are composed of an amorphous, siliceous Fe phase. The total lack of crystalline Fe phases contrasts with results obtained from cultures of Gallionella ferrugina and Mariprofundus ferrooxydans showing elongate akaganeite crystals associated with organic fibrils in stalks cores (Chan et al., 2004;Chan, Fakra, Edwards, Emerson, & Banfield, 2009) and bladed lepidocrocite crystals radiating from the surfaces of freshly grown stalks (Byrne, Schmidt, Gauger, Bryce, & Kappler, 2018;Chan et al., 2011;Comolli et al., 2011). Co-precipitation promotes the formation of immobile linkages between silica and ferrihydrite particles, which retard the transformation of ferrihydrite to crystalline iron oxides (Cornell, Giovanoli, & Schindler, 1987). It is therefore not surprising that the biomineralized stalks in the silica-rich hydrothermal deposits at the JMVF are primarily composed of an amorphous Fe phase.
Scanning electron microscopy observations of typical ultrastructural properties of stalk, tube, and sheath interiors ( Figure S3 Figure S7). An additional type of cross section is identified by its large, silica-fringed central pores (Figure 7). The jagged inner surface and elliptical outer morphology of the FIB-TEM cross section correspond well with the SEM observations of branching tubes (see Figure S3e, f). STEM imaging and EDS mapping of the Fe-rich portions of the stalk cross sections reveal distinct core and rim components that differ in terms of grain packing, homogeneity, and chemical composition. The twisted stalks display densely packed, homogenous cores with elevated concentrations of P along the margins (see Figure 6d,e). Studies of freshwater Fe-oxidizing bacteria have demonstrated that P is evenly distributed in the extracellular Feoxyhydroxide stalks (Suzuki et al., 2011). The uneven distribution of P in the stalk cross sections from the JMVF suggests that P adsorption is a more significant process than co-precipitation during stalk growth at this site. Phosphate can be mobilized in low-temperature hydrothermal fluids at depth as a result of Feoxyhydroxide dissolution (Johannessen et al., 2017) promoted by microbial dissimilatory iron reduction (Jensen, Mortensen, Andersen, Rasmussen, & Jensen, 1995) or chemical reduction by H 2 S under anoxic conditions (Afonso & Stumm, 1992;Kraal, Slomp, Reed, Reichart, & Poulton, 2012). Phosphate groups are efficiently resorbed by fresh Fe-oxyhydroxide particles in the oxic zone near the sediment surface. We postulate that progressive filling of sorption sites on the stalk surface immediately following stalk growth could account for the elevated P content evident around the innermost core region. Fe-oxyhydroxides on the stalk surface following the initial stalk excretion and biomineralization. Stepwise thickening of stalks is consistent with results obtained in laboratory experiments, which show that abiotic oxidation contributes considerably to Fe-oxyhydroxide precipitation on the stalk surface over time (Byrne et al., 2018;Krepski et al., 2013). Fe is present throughout, but is strongly enriched in the thin, crystalline bands located in the center. A slight K-enrichment is also evident in these bands. The abundance of Si is lower in the core region than in the surrounding matrix. The intensity scale ranges from 0 to 256

| Degradation of organic carbon template
Stalk-forming Fe-oxidizing bacteria produce polysaccharide templates as a strategy to localize the deposition of Fe-oxyhydroxides and avoid cell encrustation , and it has been demonstrated that the organic matrix is retained in extracellular stalks exposed to experimental diagenesis (Picard, Kappler, Schmid, Quaroni, & Obst, 2015). Both SEM and TEM imaging of the JMVF twisted stalks clearly show 10-to 50-nm-wide cavities in the core region (see Figure 6d; Figure S3a,b). EDS mapping (Figure 6e) indicates neither elevated concentrations of C in association with these cavities nor concentrations above background level in the remaining innermost core, as would be expected if the stalks contain remnants of organic polymers. We cannot exclude that the exceedingly low content of organic C is a primary feature of these stalks (c.f., Bennett, Toner, Barco, & Edwards, 2014), but we find it equally conceivable that the conspicuous cavities in the stalk cores represent empty remains of an organic framework. If the cavities initially accommodated organic templates, the absence of an elevated C EDS signal indicates that templating polysaccharides are easily lost during early diagenesis in natural hydrothermal settings.
One of the studied stalks displays a clearly elevated C signal in the concentrically laminated surface precipitates (Figure 6e).
Contamination by epoxy during sample preparation can be excluded as a source of the C enrichment because the EDS maps do not reveal excess Cl in association with the enhanced C signal (see Figure   S9 for EDS spectrum of epoxy). C-enriched parts of the sample furthermore display slightly elevated N concentrations (see Figure S8e).
The C enrichment may either result from small-scale migration of organic polymers from the stalk core to the periphery during hydrothermal alteration or from the addition of externally sourced organic compounds after stalk formation. Bennett et al. (2014) found that organic carbon derived from cell death and exudation may be remobilized in solution and adsorbed to the surfaces of Fe-oxyhydroxide stalks. The JMVF Fe deposits are exposed to periodic fluxes of hydrothermal fluids, which alter the chemistry of the precipitates and may redistribute organic C from deeper parts of the system. We therefore caution that the presence of C in the vicinity of putative Fe-oxidizing microfossils is not always robust evidence of biogenicity, and that when interpreting microtextures in ancient jaspers, for example, from the 3.77 Gyr Nuvvuagittuq supracrustal belt (Dodd et al., 2017), all possible sources of carbon need to be explored.

| Potential biogenic controls on filament growth
A biological origin for the filaments in the chimney fragments is considerably more challenging to discern than for the filaments found in the silicified mounds, as distinct morphological biosignatures of Fe-oxidizing bacteria are absent. Width may serve as a useful tool for determining whether the chimney filaments are stalks of microaerophilic Fe-oxidizing bacteria or abiotic artifacts, because stalk width is constrained by the size of the cell and typically lies within a narrow range (Krepski et al., 2013). The filamentous structures in the chimney samples are typically between 6 and 12 µm wide and are thus significantly wider than the stalks and sheaths in the laminated mound samples. Additionally, the range of measured widths of chimney filaments is notably larger than that of mound filaments (Figure 5a,b). The diameter of extracellular stalks depends on the extent of surface mineralization and shows substantial variation between modern samples from different environments and fossil Fe deposits (Hofmann et al., 2008;Krepski et al., 2013;Little et al., 2004). However, the range of stalk widths within samples from a single locality appears to be relatively narrow. In contrast, abiotic filamentous artifacts may display a large range of widths and a skewed distribution (Hofmann et al., 2008;Rouillard, García-Ruiz, Gong, & Zuilen, 2018). The differences in width distribution in Figure Figure S5c). The lack of distinct sequential bifurcation does not, however, preclude a biogenic origin. Microfossils assigned to the enigmatic genus Frutexites, which is known primarily from Ferich microbialites, calcareous stromatolites, and subterranean fractures and veins (e.g., Bengtson et al., 2014;Maslov, 1960;Walter & Awramik, 1979) To our knowledge, Frutexites is not commonly found in association with arrays of linear and strongly directional filamentous features.
Based on these fundamental geometric discrepancies, the chimney dendrites appear to be unrelated to Frutexites.
While stalks were not detected in the chimney samples by SEM, Fe-rich precipitates with globular and vermicular morphology were observed. Aggregates of globular Fe-oxyhydroxides are generally assumed to form by abiotic oxidation (Boyd & Scott, 2001), but the associated nest and tube features display morphological traits that may be linked to microbial activity. The fibrillar, rounded nests found in low abundance in the JMVF chimney ( Figure 4g) resemble capsules formed by freshwater Fe oxidizers belonging to the Siderocapsa group (Hanert, 2006). Similar nests have also been detected in low-temperature hydrothermal Fe deposits at the Loihi Seamount and ascribed to Fe-oxidizing secondary colonizers (Chan, McAllister, et al., 2016).

| Potential abiogenic controls on filament growth
Based on filament width distribution, internal distributions of Fe, branching patterns, and morphological characteristics of Fe-rich precipitates, we infer that stalk-forming, Fe-oxidizing bacteria do not play a major role in facilitating Fe mineralization in the hydrothermal Si-Fe chimneys. In the absence of obvious bio-mediated mechanisms for filament growth, purely inorganic processes must be considered. Areas with abundant multiple branching filaments and dendrites commonly display closely spaced bands of decreasing Fe content oriented perpendicularly to the branching direction of the filaments (Figure 4b,c). Some of the bands that are located at the termination of the branching features appear to be completely devoid of Fe. Banded, self-assembled biomorphs have been generated experimentally by autocatalytic co-precipitation of silica and carbonate under locally fluctuating pH conditions (Montalti et al., 2017), and oscillating reaction chemistry is known to produce banded features in natural settings as well (e.g., Liesegang rings; Ortoleva, Chen, & Chen, 1994). Abiotic, non-equilibrium processes have previously been invoked to explain the growth of dendritic iron oxide moss agates produced by oxidizing, silica-rich fluids percolating through hydrothermal sulfide debris (Hopkinson et al., 1998), and manganese dendrites are regarded as products of diffusion-limited reaction processes (e.g., Chopard, Herrmann, & Vicsek, 1991). The lateral variations in the relative contents of Fe and Si in the JMVF chimney dendrites could similarly reflect cyclic variations in microenvironmental conditions. Filament and dendrite growth may be facilitated by cycles of Fe oxidation, Fe-oxyhydroxide supersaturation, nucleation, and temporary depletion controlled by the diffusive transport of Fe(II) to an oxidizing reaction front. We suggest that the concentration of Fe(II) in the reaction solution at the growth front may exert a primary control on the mode and intensity of filament branching.

| Effects of environmental parameters on texture development
Textural and paragenetic evidence indicates disparate temperature regimes and formation processes for the JMVF mound and chimney deposits. Based on the observations of twisted stalks, sheaths, and branching tubes, we suggest that the mound deposits originated as microbial mats of Fe-oxidizing bacteria on the seafloor. Considering silica precipitation at temperatures between ca. 32 and 71°C (Dekov et al., 2015;Herzig, Becker, Stoffers, Bäcker, & Blum, 1988;Stüben et al., 1994;Sun et al., 2012). This range does not overlap with the reported upper temperature limit of 30°C for growth of M. ferrooxydans (Emerson et al., 2007).

| Evaluating filamentous biosignatures in Fe deposits
Morphology is a key property when evaluating putative traces of early life on Earth. The twisted shape of extracellular iron stalks is a prime example of a diagnostic trait that is linked to physiology and may promote the recognition of Fe-oxidizing micro-organisms in the rock record . It is, however, well-known that abiogenic processes are capable of producing filamentous features that may easily be mistaken for remnants of bacteria or biominerals (e.g., Rouillard et al., 2018  Organic carbon is a key component of life and the detection of C in microfossils has successfully been applied to support the identification of microbes in ancient environments harboring photosynthetic and heterotrophic organisms (e.g., Wacey et al., 2013;Wacey et al., 2012). Considering the lack of preserved C in the studied stalks from the mildly altered JMVF mounds, organic C does not appear to be a useful biosignature in the search for microaerophilic Fe-oxidizing bacteria in hydrothermally affected deposits, despite its presence in stalks from fresh environmental samples (Chan et al., 2009;Suzuki et al., 2011) and cultures exposed to elevated temperatures (Picard et al., 2015). The paucity of C either reflects a primary lack of organic substances in the stalks (c.f., Bennett et al., 2014) or suggests remobilization of C by hydrothermal fluids, a process that may have affected many of the ancient seafloor habitats for Fe-oxidizing bacteria.
The width distribution of filament assemblages is considered useful for evaluating the biogenicity of iron microfossils because microaerophilic Fe-oxidizing bacteria produce extracellular stalks with fairly uniform widths (Krepski et al., 2013). Individual twisted stalks and sheaths typically display minor changes in width from filament inception to termination, but the thickness may increase toward branching points and decrease abruptly following bifurcation . Our JMVF work supports that abiogenic filamentous features show a wider range of sizes than biogenic filaments, as noted by Hofmann et al. (2008), and furthermore demonstrates that heterogeneous oxidation and abiogenic Fe-oxyhydroxide and silica mineralization exert a major effect on stalk growth and final widths in hydrothermal systems. Measurements of absolute widths should therefore be considered together with width distributions.
Fe-oxidizing bacteria preferably grow toward zones of higher oxygen concentrations. Consequently, extracellular stalks commonly display directional orientation (Chan, McAllister, et al., 2016;Krepski et al., 2013). The preferred stalk orientation has a clear physiological

| CON CLUS IONS
We have used optical and scanning electron microscopy in combination with focused ion beam-transmission electron microscopy to characterize the differences between biogenic and abiogenic Fe-oxide filament growth in early diagenetic low-temperature Si-Fe deposits from the Jan Mayen Vent Fields. Overall, the results of our study indicate that hydrothermal fluid flow and diffusion-limited reaction processes may produce filamentous features in abiotically controlled systems, albeit with some notable differences with respect to microbial textures.
F I G U R E 9 Proposed criteria for the distinction of extracellular filaments of Fe-oxidizing bacteria from abiotic artifacts in the rock record. (a) Ultrastructures and textures that may be associated with stalk-and tube-forming Fe-oxidizing bacteria and proposed developing mechanisms. Biosignatures 2, 3, and 4 and their connection to bacterial physiology have been described previously by, for example, Chan et al. (2011), Comolli et al. (2011, Krepski et al., (2013) and Chan, McAllister, et al. (2016). (b) Ultrastructures and textures that may be associated with abiotically grown filaments and proposed mechanisms of formation Abiogenic filamentous features can be distinguished from silicified extracellular stalks of Fe-oxidizing bacteria by their larger diameters, strongly preferred orientation, intense non-bifurcating branching patterns, Fe concentration bands, and particulate interiors. Our study shows that dendrites may originate from local radial growth of stalks of Fe-oxidizing bacteria or by inorganic non-equilibrium reactions, with the latter giving rise to systematically increasing branching intensity in the growth direction. The absence of primary carbon in the silicified Fe deposits at the JMVF demonstrates the high mobility of organic material in hydrothermal environments and casts doubt on whether carbon can be retained in stalks through diagenetic and metamorphic reworking. These findings do not categorically exclude that carbon may be preserved in situ during natural diagenesis of extracellular stalks, but on the basis of our results we would not recommend the detection of organic carbon as a prerequisite for identifying stalk-forming Fe-oxidizing bacteria in the rock record. In this regard, microfossils of Fe-oxidizing stalk formers differ from bona fide organic microfossils for which the presence of organic carbon is a necessary requirement.