Deep Microbial Colonization in 2-Billion-Year-Old Ultramafic Rock from the Bushveld Complex

35 Archean cratons may provide stable microbial habitats in the deep subsurface, as 36 evidenced by the discovery of billion-year-old crustal fluids 1,2. However, the 37 long-term habitability of these cratonic environments is uncertain, as 38 polymetamorphic evolution in most cratons typically destroys microbial habitats 39 through mineral reactions and porosity loss 3,4 . Preservation of deep microbial 40 habitats is more likely where mantle-derived magma intruded the craton after 41 metamorphic overprinting 4 . Here we report the discovery of dense microbial 42 colonization at 814 m depth within the 2.05-billion-year-old, unmetamorphosed 43 Bushveld Igneous Complex intrusion, South Africa 5. Using advanced 44 contamination-control protocols6,7 and synchrotron-based X-ray spectroscopy, we 45 identified indigenous microbial cells localized at the rims of phlogopite, a hydrous 46 phyllosilicate mineral. Our study reveals that microbial colonization is associated 47 with Fe(III) derived from the structure of phlogopite, where the dehydrogenation 48 likely oxidizes Fe(II) to Fe(III) coupled to H 2 generation8. Despite the absence of 49 fracture-driven fluid ingress in the unfractured rock matrix, aqueous alteration 50 evidenced at the rims by potassium removal indicates a self-sustaining habitat 51 driven by an internal redox gradient 9. These findings demonstrate that aqueous 52 alteration of ultramafic rocks can sustain isolated microbial life over geological 53 timescales, significantly expanding the potential for long-term habitability on both 54 Earth and Mars4,10. 55 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 3 56 Main 57 Archean cratons are old, stable components of the continental lithosphere that are 58 preserved in many places around the globe 11. Recently, it has been found that crustal 59 fluids have remained isolated in deep cratonic rocks for 1.2 and 1.7 billion years, 60 respectively, of the Kaapvaal Craton in South Africa 1 and the Laurentia Craton in 61 Canada2. Given that microbes and abiotically generated organic compounds (e.g., acetate 62 and formic acid) have been found in isolated crustal fluids 12,13, Archean cratons may 63 therefore have provided stable microbial habitats throughout the history of life on Earth. 64 Despite the abundance of energy sources such as H 2, CH 4, and organic acids 12,13, 65 microbial cell densities in these crustal fluids are extremely low14,15. This low cell density 66 may result from the scarcity of dissolved electron acceptors, such as nitrate, sulfate, and 67 dissolved inorganic carbon in the fluids 12,13. Another possible explanation for this low 68 cell density is that the rocks reveal evidence for multiple tectono-thermal metamorphic 69 events16, which may have substantially reduced rock porosity and the bioavailability of 70 solid-state oxidants3. 71 Following metamorphic overprinting, Archean cratons commonly experienced 72 magmatic intrusions of mafic-ultramafic magmas from the underlying mantle 4, which 73 resulted in the formation of thick cumulate layers rich in olivine and pyroxene. The 74 Rustenburg Layered Suite of the Bushveld Igneous Complex in the Kaapvaal Craton in 75 South Africa is the largest known intrusion of this kind 5, and it has not suffered 76 significant metamorphic alteration since its emplacement in the Paleoproterozoic 77 Transvaal Supergroup approximately 2.05 billion years ago (ca. 2.05 Ga) 17,18. We 78 studied a freshly-drilled rock core sample obtained from a depth of ca. 814 m by the 79 Bushveld Complex Drilling Project (BVDP) of the International Continental Scientific 80 Drilling Program (ICDP) 19 at the Marula Platinum Mine, South Africa (24.50906°S, 81 30.08757°E). 82 83 Microbial colonization in unfractured pyroxenite 84 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 4 In an earlier study of the BVDP, we collected a drill-core sample from a depth of ca. 15 85 m that was composed of norite (a rock dominated by plagioclase and orthopyroxene) 86 and with abundant fractures to test on-site procedures to monitor drilling-fluid 87 contamination using fluorescent microspheres with a diameter range from 0.25–0.45 μ m, 88 similar to the size of microbial cells 6,20. We also established subsequent laboratory 89 procedures for decontamination, for counting fluorescent microspheres and microbial 90 cells, and for visualization within a rock section6. This sample contained microbial cells 91 located around mineral-filled fractures that were free of fluorescent microspheres, 92 showing they were uncontaminated. 93 We applied the same on-site and laboratory procedures to a study of the deeper 94 sample (ca. 814-m), which was not intersected by fractures (Fig. 1a). Under ultraviolet 95 (UV) illumination, the blue fluorescence of microspheres on the core surface was 96 intense (Fig. 1b; Extended Data Fig. 1a), demonstrating proper delivery of the tracer to 97 the bottom of the borehole through the circulating drilling fluid. The core was broken 98 open with a sterilized hammer, and the fresh surface was observed under UV 99 illumination and found to be partly contaminated at the edges only (Fig. 1c). The 100 interior and exterior of the core were then separated using a sterilized rock trimmer. 101 Fluorescent microspheres were abundant in the core exterior (4.9 ± 1.3 × 10 6 102 microspheres cm −3) but significantly depleted in the core interior (3.1 ± 0.3 × 10 4 103 microspheres cm −3). Given the concentrations of the microspheres (3.7 ± 1.8 × 10 8 104 microspheres mL−1) and microbial cells (3.2 ± 0.1 × 106 cells mL−1) in the drilling fluid, 105 the number of contaminant microbial cells in the core interior is calculated to be less 106 than 3.0 × 10 2 cells cm −3. This low level of contamination demonstrated that the core 107 interior was suitable for subsequent microbiological analysis. Petrographic observation 108 in a thin section revealed that the sample was pyroxenite, with orthopyroxene and 109 clinopyroxene contents of ca. 60% and ca. 30%, respectively, and with <10% combined 110 phlogopite and quartz (Fig. 1d). 111 From the core interior and the core exterior, including the outer core surface, we 112 cut 3 mm thick sections with a precision diamond band saw without lubricants (e.g., 113 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 5 water or oil) in a clean fume hood with air circulated through a high efficiency particulate 114 air (HEPA) filter (Fig. 1e; Extended Data Fig. 1b). Although microspheres were 115 observed at the outermost edge of the exterior section (Extended Data Fig. 1c), no 116 microspheres were detected in the interior section by fluorescence microscopy (Extended 117 Data Fig. 1d). The interior section was then stained with SYBR Green I and examined 118 using an optical photothermal infrared (O-PTIR) spectroscope equipped with a 119 fluorescence microscope. Microbial signals, indicated by peaks attributed to amide bonds 120 in proteins at 1,530 and 1,640 cm −1, were detected at mineral grain boundaries, which 121 coincided with greenish regions stained with SYBR Green I (Fig. 1f–i). In combination 122 with the absence of fluorescent microspheres (Extended Data Fig. 1d), our results 123 demonstrate the presence of indigenous microbial cells at mineral grain boundaries. 124 125 126 Fig. 1 | Rock description, contamination discrimination, and in situ microbial 127 detection. a, b, Core sample under visible ( a) and UV light ( b), showing a crack (red 128 arrowhead) made with a sterile hammer. Blue fluorescence indicates contamination on 129 the edge. c, Broken surface under UV light showing contamination where blue 130 fluorescence is visible only at the core edge (red arrowhead). d, Thin section (crossed 131 polarizers) displaying orthopyroxene (opx), clinopyroxene (cpx), phlogopite (phl), and 132 quartz. e, Prepared rock section with analytical points (yellow arrows). f–h, Fluorescence 133 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 6 microscopy (SYBR Green I) of green regions, with analytical spots marked. i, O-PTIR 134 spectra of the rock (colored squares) compared to control microorganisms ( E. coli, N. 135 aerobiophila, and M. sedula). 136 137 138 Extended Data Fig. 1 | Assessment of drilling fluid contamination using fluorescent 139 microspheres. 140 a, High-magnification fluorescence micrograph of fluorescent microspheres within the 141 drilling fluid. b, Photograph of a rock section extending from the outer edge to the 142 center of the drill core, prepared using a precision diamond band saw. The blue dashed 143 line indicates the outer edge of the core sample. Colored arrows indicate the areas 144 observed in c (magenta) and d (green). The orange rectangle corresponds to the area 145 shown in Fig. 1e. c, d, High-magnification fluorescence micrographs of the rock section 146 at the locations indicated in b. White arrows denote the presence of fluorescent 147 microspheres. 148 149 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 7 We also observed regions with strong greenish signals at or near the edge of the 150 interior section (Fig. 2a), where individual microbial cells were visible with 151 high-magnification fluorescence microscopy (Figs. 2b,c). To obtain independent 152 evidence of the presence of microbial cells, the thin section was studied using scanning 153 fluorescence X-ray microscopy (SFXM) with a soft X-ray beam line at the NanoTerasu 154 synchrotron radiation facility in Japan. Micro-X-ray fluorescence (μ -XRF) mapping of C, 155 N, P, and S with a ca. 3 μ m diameter beam revealed multiple loci with co-enrichments 156 of C, N, P, and S at and near the edge of the section (Figs. 2d,e). In addition, the regions 157 with C-, N-, P-, and S- enrichments (Points 1 and 2 in Figs. 2d,e) and a non-enriched 158 control location (Point 3 in Fig. 2d) were subjected to N K -edge X-ray absorption 159 near-edge structure analysis (XANES) using a ca. 5 × 25 μ m beam. Since this analysis 160 requires a strong X-ray beam to be irradiated at one location for ca. 20 min, we first 161 evaluated the possibility of beam damage to the reference materials (biological materials 162 and inorganic and organic reagents). We confirmed negligible damage for the first and 163 second measurements (Extended Data Fig. 2). The N K-edge XANES spectrum from 164 the C-, N-, P-, and S- enriched loci (Point 1) was very similar to those from cultured 165 bacterial cells ( Escherichia coli ) and markedly different from those of the other 166

materials, based on the relative intensities of peaks attributed to the imine 167 group at 398.8 eV, pyridine-type heterocyclic compounds at 399.7 eV, NH 4 + at 400.9 168 eV, and pyrrole-type heterocyclic compounds at 401.2 eV (Fig. 2e,g) 21,22. Although the 169 N K-edge XANES spectrum from the C- and N-enriched loci (Point 2) was similar to 170 that of cultured bacterial cells, the signal-to-noise ratio was low. This result is consistent 171 with the low enrichment levels of C, N, P, and S. Since the control location showed no 172 detectable X-ray absorption features for N, the XANES spectra cannot be attributed to 173 the rock matrix or instrumental artifacts. Taking these results together, we conclude that 174 the unfractured ultramafic rock is colonized by indigenous microbes in the deep 175 subsurface. 176 177 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 8 178 Fig. 2 | Microbial signals within the rock interior characterized by 179 synchrotron-based soft X-ray spectroscopy. a, Fluorescence micrograph of a rock 180 section stained with SYBR Green I. b, c, High-magnification fluorescence micrographs 181 showing greenish signals from microbial cells (white arrows); the areas correspond to 182 yellow rectangles in a. d, e, Elemental maps of C, N, P, and S obtained via micro-X-ray 183 fluorescence (μ -XRF) analysis; the areas correspond to light blue rectangles in a. Colored 184 arrows and yellow rectangles indicate locations for X-ray absorption near-edge structure 185 (XANES) analyses. f, N K-edge XANES spectra of the sample and reference materials. 186 Spectral colors correspond to the arrows in d and e. Reference materials include cultured 187 bacteria (Escherichia coli ), protein (albumin), DNA, humic acid, NH 4Cl, and SYBR 188 Green I. Vertical dotted lines indicate peak positions at 398.8 eV (red), 399.7 eV (blue), 189 400.9 eV (green), and 401.2 eV (orange). g, Magnified N K-edge XANES spectra 190 corresponding to the yellow highlighted area in f. Absorption peaks represent the imine 191 group at 398.8 eV (red)²¹, pyridine-type heterocyclic compounds at 399.7 eV (blue)²¹, 192 interstitial NH4 + at 400.9 eV (green)²², and pyrrole-type heterocyclic compounds at 401.2 193 eV (orange)²¹. 194 195 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 9 196 Extended Data Fig. 2 | Evaluation of beam damage through repeated N K-edge 197 XANES measurements. 198 N K-edge XANES spectra of reference materials (cultured E. coli, albumin, DNA, 199 humic acid, NH4Cl, and SYBR Green I) from three consecutive scans. Due to negligible 200 beam damage in the initial two scans, they were merged to produce the spectra shown at 201 the bottom and in Fig. 2f. Vertical dotted lines denote peaks at 398.8 (red), 399.7 (blue), 202 400.9 (green), and 401.2 eV (orange). 203 204 Microbial habitat around pyroxene grain boundaries with phyllosilicate minerals 205 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 10 To characterize the habitat of the detected microbes, we performed mineralogical and 206 geochemical analyses of the regions where the SFXM data demonstrated the presence of 207 microbial colonization. Environmental scanning electron microscopy (ESEM) coupled 208 with energy-dispersive X-ray spectroscopy (EDS) revealed microbial signals located at 209 orthopyroxene grain boundaries in contact with intercumulus grains interpreted to be 210 phlogopite (Mg-rich mica) based on the mineral’s EDS spectra (Extended Data Fig. 211 3a,b). We also found microbial signals spatially correlated with the rims of the 212 phlogopite grains , where the EDS spectra indicated K depletion (Extended Data Fig. 213 3c). In addition, a region enriched with iron and sulfur was observed near the phlogopite 214 grains (Extended Data Fig. 4a,b). We performed S K-edge XANES analysis for the Fe- 215 and S-enriched region, and the spectra obtained from this region were identical to those 216 of pyrrhotite (iron monosulfide) (Extended Data Fig. 4c). 217 218 219 Extended Data Fig. 3 | Mineralogical characterization of microbially colonized 220 regions. 221 a, Back-scattered electron (BSE) image and elemental maps (Mg, Al, Si, K, and Fe) 222 obtained via environmental scanning electron microscopy with energy-dispersive X-ray 223 spectroscopy (ESEM-EDS). The mapped area corresponds to Fig. 2a and Fig. 3a,b. 224 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 11 Symbols indicate analytical locations. b, Representative ESEM-EDS spectra for points 225 1–5 in a. c, Binary plot of K versus Si+Al (atomic %) showing K-depletion along the 226 rims of phlogopite-like grains. Plot symbols correspond to those in a. 227 228 Extended Data Fig. 4 | Characterization of an iron- and sulfur-enriched region 229 near phlogopite grains. a, Element maps of C, O, K, Ca, S, and Fe acquired 230 via environmental scanning electron microscopy with energy dispersive X-ray 231 spectroscopy (ESEM-EDS). White arrows indicate the location of the iron- and 232 sulfur-enriched region. The maps cover the same area shown in Fig. 3a,b and Extended 233 Data Fig. 3a. b, EDS spectra extracted from the iron- and sulfur-enriched region. c, S 234 K-edge XANES spectra of the iron- and sulfur-bearing grain and reference 235

(pyrrhotite, troilite, marcasite, pyrite, elemental sulfur, Na 2SO3, and Na2SO4). 236 Vertical dotted lines mark peak positions at 2471.4 eV (red), 2473.2 eV (blue), 2476.3 237 eV (green), and 2485.3 eV (orange). 238 239 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 12 We also employed powder X-ray diffraction (XRD) analysis of the whole rock 240 to identify the mineralogy of the sample. In an XRD pattern of the whole rock, peaks 241 were attributed to orthopyroxene, clinopyroxene, quartz, and phlogopite (Extended Data 242 Fig. 5a). An XRD pattern of the clay-sized fraction of the sample showed peaks at 10.0 Å 243 and 14.3 Å (Extended Data Fig. 5b). After treatment with ethylene glycol, no peak shift 244 was observed for the peaks at 10.0 Å and 14.3 Å, indicating the absence of 245 smectite-group minerals23. As the 14.3 Å peak was shifted to 10.0 Å after heat treatment 246 at 500°C, the 14.3 Å peak was identified as a vermiculite-group mineral23. 247 248 249 Extended Data Fig. 5 | Mineral identification by X-ray diffraction analysis. 250 a, Powder X-ray diffraction (XRD) patterns of the core sample and reference minerals. 251 Peak labels are as follows: E, enstatite; D, diopside; Q, quartz; P, phlogopite; T, talc; and 252 V, vermiculite. b, Powder XRD patterns of the oriented clay-size fraction from the core 253 sample before and after ethylene glycol (EG) treatment and heating (500 °C for 1 hour). 254 Hydrobiotite (an interstratification of Fe-rich mica and vermiculite) 66 and clinochlore 255 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 13 (Mg-rich chlorite) are shown as references to demonstrate the diagnostic collapse of the 256 vermiculite interlayer spacing upon heating. Vertical lines indicate the peak positions at 257 10.0 Å (blue) and 14.3 Å (orange) 258 259 To further constrain the identity of the phlogopite grains and the K-depleted 260 areas along their rims, which had a spatial association with microbial signals, we 261 performed in situ mineralogical analysis using SFXM with a soft X-ray at the SPring-8 262 synchrotron radiation facility in Japan. μ -XRF mapping of Al with a ca. 2- μ m diameter 263 beam (Fig. 3b) confirmed the locations where previous SFXM and ESEM-EDS 264 analyses indicated the presence of microbial cells (Fig. 2) in association with the rims 265 (Extended Data Fig. 3). Al K -edge XANES spectra are known to be distinct among 266 rock-forming minerals24, but there has been no systematic study to compare Al K-edge 267 XANES spectra from various phyllosilicate minerals. In this study, we found that Al 268 K-edge XANES are also sensitive to different phyllosilicate minerals (Fig. 3c). The Al 269 K-edge XANES spectra of the phlogopite grains were similar to that of phlogopite (Fig. 270 3c). In addition, spectra identical to that of hydrobiotite, a mineral consisting of 271 interstratified vermiculite and biotite (Fe-rich mica), were obtained from the K-depleted 272 rims of the phlogopite grains with microbial signals (Fig. 3c). As vermiculite is 273 generally transformed from phlogopite by loss of interlayer K by hydrothermal alteration 274 and/or weathering9, it is reasonable to expect to obtain microbial signals from the 275 K-depleted rims of the phlogopite grains. The presence of vermiculite in the sample is 276 supported by the presence of vermiculite and phlogopite peaks in the powder XRD 277 pattern of the sample (Extended Data Fig. 5). 278 Fe L3-edge XANES analysis was performed on the same locations as the Al 279 K-edge XANES analysis to distinguish the valence state of iron based on the spectra at 280 ca. 707 and ca. 710 eV 25. The relative peak intensities obtained indicated that Fe(III) 281 was abundant at the phlogopite grains, whereas Fe(II) dominated at the adjacent 282 pyroxenite grains (Fig. 3d). We also performed Fe L3-edge XANES analysis from the 283 interior to the exterior of the phlogopite grains. The relative peak intensities indicated 284 the relative abundance of Fe(II) at the phlogopite rims (Fig. 3d), which is consistent 285 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 14 with the microbial reduction of Fe(III) mediat ed at the phlogopite rims associated with 286 microbial signals (Fig. 2). 287 288 289 Fig. 3 | Mineralogical and geochemical features associated with microbial 290 colonization. 291 a, Element maps of Al and K obtained by environmental scanning electron microscopy 292 with energy dispersive X-ray spectroscopy (ESEM-EDS). b, Synchrotron-based 293 micro-X-ray fluorescence ( μ -XRF) map of Al for the same region shown in a. The 294 mapped area corresponds to Fig. 2a. c, d, X-ray absorption near-edge structure 295 (XANES) spectra obtained from the rock section and reference minerals at the Al 296 K-edge ( c) and Fe L 3-edge ( d). Pre-edge structures (red arrows in c) indicate the 297 presence of vermiculite in hydrobiotite. Spectral colors correspond to the analyzed 298 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 15 points in a and b. Reference minerals include phlogopite, hydrobiotite, talc, saponite, 299 nontronite, montmorillonite, hematite, and diopside. 300 301 As there was the possibility that Fe(II) in the phlogopite grains and their 302 vermiculite-bearing rims was oxidized by air exposure during sample preparation and 303 subsequent analyses26, we prepared a new 3 mm thick rock section using a precision 304 diamond wire saw placed inside an Ar-purged glove box (Fig. 4a). Fe K-edge XANES 305 analysis of the rock section in an He-purged sample holder was performed using SFXM 306 with hard X-rays at SPring-8. μ -XRF mapping of K revealed the presence of 307 phlogopite-like grains (Fig. 4b). The valence state of iron around the phlogopite-like 308 grains was determined based on the Fe K-edge position27. The Fe K -edge position was 309 shifted toward that of an Fe(III) standard 28 from the exterior to the interior of the 310 phlogopite grain (Fig. 4c,d,e). The mineral identity was confirmed by Al K-edge 311 XANES analysis with soft X-rays (Fig. 4f). This result is consistent with the shifts in 312 the relative peak intensities of Fe(II) and Fe(III) in the Fe L3-edge XANES spectra from 313 the air-exposed rock section (Fig. 3d). Given that the oxidation of iron in biotite in air 314 typically proceeds at temperatures above 400°C 30, it is unlikely that the presence of 315 Fe(III) in the phlogopite grains and their vermiculite-bearing rims was an artifact 316 introduced after core processing and subsequent characterizations of the rock sections. 317 318 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 16 319 Fig. 4 | Internal variation in the iron valence state inside and outside a phlogopite 320 grain determined after minimal exposure to air. 321 a, Photograph of the rock section prepared by a diamond wire saw in an Ar-purged glove 322 box. b, Element map of K obtained by synchrotron-based micro-X-ray fluorescence 323 (μ -XRF) analysis. The area corresponds to the rectangle in a. c, Higher magnification 324 map of K obtained by the same μ -XRF analysis. The area corresponds to the rectangle 325 in a and b. d, Element map of Al obtained using environmental scanning electron 326 microscopy with energy dispersive X-ray spectroscopy (ESEM-EDS). The area 327 corresponds to the rectangle in a and b. e, Fe K-edge X-ray absorption near-edge 328 structure (XANES) spectra obtained inside and outside a K- and Al-enriched grain in the 329 rock section and references. The spectral colors correspond to points indicated 330 in c and d. FeCl2 29 and nontronite28 were used as the references for Fe(II) (blue dotted 331 line) and Fe(III) (red dotted line), respectively. f, Al K-edge XANES spectra obtained 332 from the K- and Al-enriched grain and reference materials (phlogopite and diopside). 333 334

335 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 17 Magmatic-to-aqueous evolution forming microbial habitats 336 The presence of microbial life at the grain boundaries of hydrous-altered phlogopite 337 suggests that the suitability of the habitat depends on the interaction of magmatic 338 minerals and fluids. Our results suggest the following possible scenario, as illustrated in 339 Fig. 5. 340 Stage 1: Crystallization and accumulation of primary pyroxene crystals from magma 341 (e.g., <ca. 1,200°C) 31. 342 Stage 2: Crystallization of intercumulus phlogopite from a hydrous interstitial melt 343 (<750°C) 32. 344 Stage 3: Transformation of phlogopite to vermiculite by hydrous alteration below the 345 critical point of water in saline fluid (ca. 400°C) 33. Based on the cooling history of the 346 intrusion, this temperature was reached about 10 million years after the intrusion 347 occurred34. 348 Stage 4: Further cooling to the present temperature. The rock sample studied is spatially 349 associated with brackish groundwater inflowing into the drilled borehole through 350 fractures at various depths. The temperature of the brackish groundwater was measured 351 to be ca. 37°C in the borehole19. 352 Between Stages 2 and 3, cooling of the phlogopite is expected to have driven 353 the dehydrogenation reaction, in which Fe(II) oxidation in the crystal lattice and H 2 354 generation proceed at 640°C–750°C8: 355 [Fe2+ + OH−]phlogopite → [Fe3+ + O2−]phlogopite + ½H2 356 Fe(III) in the phlogopite structure is subse quently maintained, even after the phlogopite 357 is transformed into vermiculite 35. This scenario is supported by the presence of Fe(III) 358 throughout the phlogopite grains (Fig. 4c–f), because Fe(II) oxidation by O 2-bearing 359 fluid tends to oxidize the grain rim. This aqueous process is unlikely to have occurred, 360 given the low permeability of the unfractured host rock 19. Additional evidence against 361 fluid-driven oxidation is the preservation of redox-sensitive iron monosulfide 362 (pyrrhotite) found near the phlogopite grains (Extended Data Fig. 4). Given that 363 pyrrhotite is rapidly oxidized and dissolved by O 2 in fluid 36, it is unlikely that the rock 364 was exposed to O2-bearing fluid during Stage 4. 365 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 18 Since abiotically generated microbial energy sources such as H 2, CH4, and 366 organic acids are commonly present as microbial energy sources in the deep 367 subsurface37, microbes can survive by oxidizing these energy sources using the 368 structural Fe(III) in phlogopite and/or vermiculite as oxidants. In the case of the 369 oxidation of H2 coupled to the reduction of Fe(III) in vermiculite, H 2O is produced as a 370 waste product. This metabolic pathway is commonly linked with chemolithoautotrophic 371 growth, during which CO 2 is taken up by microbial cells. Pyroxenite lacks fractures or 372 veins that would allow fluid transport of metabolic nutrients and the removal of waste 3. 373 Given that the rims of phlogopite grains appear to serve as a low-permeability habitat, 374 microbial survival requires the prevention of pore occlusion by metabolic byproducts in 375 addition to the redox gradients exploited by microbial metabolisms 3. As Fe is retained 376 within the crystal lattice of vermiculite after the reduction of Fe(III), pore occlusion by 377 Fe(III) reduction is unlikely 38. Given that H 2 is available at grain boundaries without 378 needing to be transported by fluid flow, it is reasonable to find microbial colonization at 379 the vermiculite-bearing rim in association with Fe(III) in solid rock without fractures or 380 veins. 381 Although determining the precise timing of microbial colonization remains 382 challenging, the lack of metamorphic overprinting and the absence of fracture-driven 383 fluid flow suggest that the internal redox gradient was established after the intrusion 384 cooled 2.05 billion years ago (Stage 3). The preservation of the internal redox gradient 385 with no equilibration by any subsequent fluid flow provides a geochemical constraint 386 that limits the potential for recent biological ingress and drilling-fluid contamination. 387 The longevity of this isolated habitat raises questions regarding the energetic limits of 388 life. We propose that Bushveld ultramafic rock provides a continuous, albeit minimal, 389 energy flux for cell maintenance, rather than cell growth from the reduction of 390 Fe(III)-bearing phyllosilicate minerals39,40. 391 392 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 19 393 Fig. 5: Conceptual model for the formation of deep microbial habitats in pyroxenite 394 of the Bushveld Igneous Complex. The model comprises four stages. Stage 1 : 395 Orthopyroxene (opx, grey) and clinopyroxene (cpx, green) crystallize, forming a 396 cumulate layer with interstitial residual melt enriched in water and incompatible elements 397 (e.g., K). Stage 2: Phlogopite (phl, yellow) crystallizes from residual, hydrous melt in the 398 intercumulus space and undergoes dehydrogenation upon cooling. Stage 3: The rims of 399 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 20 phlogopite react with aqueous fluid, forming vermiculite while retaining Fe(III). Stage 4: 400 Following cooling to temperatures below the limit of life, microbes colonize the altered 401 phlogopite rims, with energy provided by the reduction of Fe(III). 402 403 404 Analytical advancement in biosignature detection from ultramafic rocks 405 Scientific drilling in the search for deep microbial life has repeatedly targeted ultramafic 406 rocks in the oceanic and continental crust 41 because aqueous alteration of ultramafic 407 rocks provides not only chemical energy but also mineral catalysts for prebiotic synthesis 408 and the emergence of life42 and for living and fossil rock-hosted microbes3. A number of 409 techniques have been developed to detect biosignatures. Raman spectroscopy has been 410 applied to hydrogarnet grains in ca. 1-million-year-old (Ma) seafloor ultramafic rocks 411 from the Mid-Atlantic Ridge43 and brucite veins in 120 Ma subseafloor ultramafic rocks 412 at the Iberian Margin 44. Despite the presence of sharp peaks attributed to aliphatic 413 compounds and functional groups, such as amides, usually associated with biopolymers, 414 such as proteins and lipids, Raman biosignatures from ultramafic rocks differ from those 415 of cultured microbial cells. These differences are attributed to the type and metabolic 416 activity of the cells, along with preservation with aging43. Raman spectra from microbial 417 filaments morphologically preserved in calcium carbonate veins in ultramafic rocks in the 418 ca. 100 Ma Samail Ophiolite, Oman45, and in ca. 1 Ma seafloor ultramafic rocks from the 419 Mid-Atlantic Ridge46, lack sharp peaks from biopolymers. Furthermore, they show broad 420 peaks attributed to organic material without structural information, which could reflect 421 the poor preservation potential of biopolymers in calcium carbonate minerals and/or 422 interference by autofluorescence from calcium carbonate minerals. Autofluorescence 423 interference is also a problem affecting the use of Raman spectroscopy to obtain organic 424 signals from microbial cells associated with phyllosilicate minerals47. 425 Fourier transform infrared (FT-IR) spectroscopy has been widely used to detect 426 organic molecules associated with phyllosilicate minerals in ultramafic rocks. In 427 ultramafic rocks from the ca. 2 Ma Atlantic Massif, an Mg-rich phyllosilicate mineral, 428 saponite, is associated with abiotic amino acid synthesis, as characterized by 429 synchrotron-based FT-IR spectroscopy with a 5 × 5 μ m aperture42. In our study, spectra 430 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 21 comparable to those from FT-IR were obtained from the Bushveld rock using O-PTIR 431 spectroscopy. In this technique, thermal expansion by infrared illumination is detected by 432 a green laser (532 nm) with a beam diameter of ca. 0.5 μ m48. This O-PTIR spectroscopy 433 can achieve a high spatial resolution equivalent to Raman spectroscopy without 434 autofluorescence interference (Fig. 1i). The O-PTIR spectra obtained from the pyroxene 435 grain boundaries in this study are similar to those of cultured microbial cells (Fig. 1i) and 436 distinct from abiotic amino acids 49. Given that phyllosilicate minerals preserve organic 437 molecules from Archean siliciclastic microbial mats 50, the possibility that O-PTIR 438 spectroscopy detected fossil microbial cells that were exceptionally well preserved by 439 coexisting phyllosilicate minerals is not excluded but is unlikely. 440 We further advanced analytical techniques by applying SFXM to validate the 441 biogenic origin of fluorescence microscopic and O-PTIR spectroscopy signals in the 442 deep ultramafic rock, as well as the phyllosilicate mineral identity estimated by bulk 443 XRD and ESEM analyses. In our previous SFXM-based study51, we revealed N-bearing 444 organic matter in spatial association with the reduction of Mn(IV) and Ce(IV) at a 445 deep-sea ferromanganese crust surface. This organic matter is characterized by the 446 enrichment of aromatic and carboxylate groups in N K-edge XANES spectra, which are 447 clearly distinct from microbial cells. To our knowledge, the present study is the first 448 demonstration that SFXM can be used to detect the biosignatures of rock-hosted 449 microbes. A similar synchrotron-based method called scanning transmission X-ray 450 microscopy (STXM) is commonly used to detect microbial cells based on N K-edge 451 XANES spectra 52,53. STXM detects X-ray absorption after an incident beam is 452 transmitted through a sample, whereas in SFXM, the X-ray absorption is detected using 453 fluorescence X-rays emitted from a sample without transmission 54. An STXM sample 454 needs to be thinned to ca. 100 nm, which can be problematic for void spaces around 455 mineral grains loosely filled with microbe–mineral assemblages. In contrast, an SFXM 456 sample only requires a flat surface prepared by cutting with a precision diamond band 457 and/or wire saw, which enables the preservation of fragile rock habitats. 458 Our synchrotron-based analysis is also critical to support the indigenous nature 459 of the observed microbial colonization. N K-edge XANES spectra identical to microbial 460 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 22 cells (Fig. 2) were obtained at the mineral grain boundaries, where the aqueous 461 alteration and the internal redox gradient essential to microbial survival in an 462 unfractured rock matrix were reliably localized by μ -XRF analysis and Al and Fe 463 K-edge XANES analyses (Figs. 3 and 4). The spatial congruency, in combination with 464 our rigorous contamination-control protocols using fluorescent microspheres, strongly 465 argues against drilling-induced contamination (Fig. 1b,c and Extended Data Fig. 1). 466 467 Implications for the search for rock-hosted life on Mars 468 Our findings demonstrate that the ancient, unmetamorphosed ultramafic rock in the 469 Kaapvaal Craton hosts an isolated microbial habitat. This discovery challenges the 470 conventional view of the deep-rock biosphere in which microbial life is strictly 471 dependent on nutrient delivery through fracture networks3,36. Instead, the self-sustaining 472 nature of this habitat—driven by the intrinsic redox potential of Fe-bearing 473 phyllosilicate minerals—suggests that life can persist in a state of extreme isolation for 474 as long as mineral–water reactions continue to yield chemical energy. Such a 475 mechanism has profound implications for the search for life on Mars. 476 Due to a lack of plate tectonics, metamorphic overprinting is very limited in 477 Martian rocks. Based on the similarity to the Bushveld Igneous Complex, the Noachian 478 (3.7–4.1 Ga) Columbia Hills on Mars have been regarded as a layered intrusion 10. This 479 formation was exhumed by a meteorite impact that formed the Gusev Crater in the 480 Hesperian (3.7–3.9 Ga) and underwent aqueous alteration55. The presence of olivine- and 481 pyroxene-rich cumulates has also been reported in the Jezero Crater 56, with varying age 482 estimates from 1.4–3.45 Ga 57. Considering these time periods, our results from the 483 Bushveld Igneous Complex suggest that the Martian ultramafic rocks may still harbor 484 ancient microbial lineages or their detectable biosignatures. The Perseverance rover, 485 equipped with a deep-UV Raman spectrometer with a beam diameter of 350 μ m, detected 486 one of the broad Raman bands attributed to C=C stretching vibrations at ca. 1,600 cm −1 487 from the pyroxene-rich cumulate from the Jezero Crater58. In the Raman spectra with the 488 broad band at ca. 1,600 cm −1, a peak at ca. 1,080 cm −1 was also present and was 489 provisionally assigned to silicate, due to an inability to resolve silicate phases 59. For 490 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 23 future missions to search for Martian life, the use of O-PTIR spectroscopy would enable 491 better characterization of organic molecules in aqueously altered ultramafic rocks, as 492 demonstrated in our application on Earth, which was confirmed by synchrotron-based 493 spectroscopy. 494 495 Microbial survival with minimal evolution in ultramafic intrusions older than 2 Ga 496 This study provides a new geological template for searching for ancient microbes from 497 long-term habitats on Earth. Archean cratons other than those previously mentioned 498 also host unmetamorphosed ultramafic layered intrusions, such as the Munni Munni 499 Complex (ca. 2.93 Ga) in the Pilbara Craton in Australia 60,61, the Great Dyke (ca. 2.58 500 Ga) in the Zimbabwe Craton in Zimbabwe 62,63, and the Näränkävaara layered igneous 501 complex (ca. 2.44 Ga) in the Karelian Craton in Finland 64. In all of these ultramafic 502 intrusions, the formation of intercumulus phlogopite is evident. Our study strongly 503 suggests that deep microbial life that depends on the aqueous alteration of 504 Fe(III)-bearing phlogopite may be ubiquitous in unmetamor phosed ultramafic layered 505 intrusions formed in the Neoarchean to Paleoproterozoic eras. In addition to crustal 506 fluid isolated for 1.2 billion years, the Kaapvaal Craton harbors a subsurface bacterial 507 species that has undergone minimal evolution for 55–165 million years in fractured 508 metamorphic rocks65. Our Bushveld research targets microbial genomes that potentially 509 preserve evolutionarily primitive features for billions of years in the geologically and 510 tectonically stable subsurface environment. 511 512

513 Drilling and on-site core handling procedures 514 A rotary core barrel with a diamond bit was used for drilling. The drilling fluid was 515 composed of locally sourced water and a high-molecular-weight polymer viscosifier 516 (AMC CAP 21TM, AMC/IMDEX Ltd., Balcatta, WA, Australia) to reduce friction and 517 improve core recovery. Samples for microbiological study were collected every 100 518 meters during drilling, or approximately once every 10 days. Before the planned 519 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 24 sampling, fluorescent microspheres (Invisible Blue, DayGlo Color Corp., pigment 520 SPL-594NXC, 0.25–0.45 μ m) were added to the drilling fluid in a tank. The analyzed 521 sample from ca. 814 m was drilled and collected on August 5th, 2024. The core sample 522 (Sample ID: 5067_3_B_236R_2_WR:51-87, IGSN ICDP5067EXFA001) was rinsed 523 with deionized water three times shortly after reaching the surface and then lightly flamed 524 with a gas torch, following procedures developed in previous studies 6,7. Blue 525 fluorescence from fluorescent microspheres on the core surface was observed with a 365 526 nm-UV light (Nichia Corp., Tokushima, Japan). The decontaminated core sample was 527 stored at -18°C in an Al-lined plastic bag filled with N2 gas. Drilling fluid was collected 528 from the mud tank and stored at -18°C. 529 530 Microscopic enumeration of fluorescence microspheres and microbial cells 531 The core sample was cracked with a flame-sterilized hammer, and the intrusion of 532 fluorescent microspheres into the core interior was visualized by observation of the 533 freshly fractured cross-section. Fluorescent microspheres in the drilling fluid were 534 collected on a 13-mm-diameter polycarbonate filter (0.2-μ m pore size; Merck Millipore, 535 Darmstadt, Germany) and counted in triplicate ( n = 3) using a fluorescence microscope 536 (Olympus BX51, Tokyo, Japan) equipped with a CCD camera (Olympus DP71) and 537 image processing software (Lumina Vision, Mitani Shoji, Tokyo, Japan). Microbial cells 538 in the same drilling fluid were stained with SYBR Green I (Takara-Bio, Inc., Shiga, 539 Japan) and counted ( n = 3). For rock analysis, the interior and exterior portions were 540 aseptically separated using a sterilized rock trimmer (Iwamoto Mineral Co., Ltd., Tokyo, 541 Japan) and ground into powder with a sterilized titanium-cylinder mortar. Fluorescence 542 microspheres were extracted by suspending 0.5 cm³ of the powder in 2 ml of deionized 543 water, followed by 30 s of ultrasonication. A 0.05-ml aliquot of the suspension was 544 diluted to 2 ml with deionized water, passed through a 13-mm-diameter polycarbonate 545 filter (0.2-μ m pore size), and counted using the fluorescence microscope system (n = 3). 546 The detection limit, defined as the mean ± three standard deviations (SD) of five replicate 547 blanks (n = 5), was 0.75 ± 1.3 × 10 3 microspheres cm−3. The rock interior was cut into a 548 3-mm-thick section using a precision diamond band saw (DWS 3500P; Meiwa Fosis 549 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 25 Corp., Tokyo, Japan) in a clean booth flushed with HEPA-filtered air and observed via 550 fluorescence microscopy. 551 552 Detection of microbial cells in the rock core interior 553 The 3-mm-thick rock section was stained with SYBR Green I and examined for microbial 554 cells using a fluorescence microscope system. Optical-photothermal infrared (O-PTIR) 555 spectroscopy equipped with fluorescence microscopy (mIRage-LS, Photothermal 556 Spectroscopy Corp., Santa Barbara, USA) was used to detect greenish fluorescent signals 557 from microbial cells stained with SYBR Green I, from which O-PTIR spectra diagnostic 558 of microbial cells were automatically obtained using the feature finder function. A 559 continuous-wave 532-nm laser was used as a probe beam with submicron spatial 560 resolution, whereas a tunable quantum cascade laser (950–1800 cm -1; 2 cm -1 spectral 561 resolution; 10 scans per spectrum) served as the pump beam to obtain the mid-IR spectra. 562 O-PTIR spectra were also obtained from reference materials such as cultured cells 563 of Escherichia coli (E. coli ; NBRC13168), co-cultured cells of Nanobdella 564 aerobiophila (N. aerobiophila) strain MJ1, Metallosphaera sedula ( M. sedula ) strain 565 MJ1HA (JCM33617), and SYBR Green I. 566 Micro-X-ray fluorescence ( μ -XRF) mapping was performed using a scanning 567 fluorescence X-ray microscope (SFXM) at the BL13U beamline of NanoTerasu. The 568 incident soft X-ray beam was focused to a diameter of ca. 3 μ m using a Wolter mirror. 569 The energy and intensity of the fluorescent soft X-rays emitted from the sample were 570 measured using a silicon drift detector in partial fluorescence yield (PFY) mode. 571 Excitation energies of 3000 eV and 700 eV were applied for the simultaneous mapping of 572 P/S and C/N, respectively. Subsequently, N K-edge X-ray absorption near-edge structure 573 (XANES) spectra were acquired from the rock section in the 390–420 eV range with a 0.2 574 eV step. Reference materials, including E. coli cells, bovine serum albumin (fatty 575 acid-free, FUJIFILM Wako), salmon sperm DNA (FUJIFILM Wako), International 576 Humic Substance Society (IHSS) Nordic Aquatic Humic Acid (1R105H), SYBR Green I, 577 and ammonium chloride (NH4Cl, FUJIFILM Wako), were similarly measured. The beam 578 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 26 size for N K-edge XANES was ca. 5 μ m (vertical) × 25 μ m (horizontal). To evaluate 579 beam-induced damage, N K-edge XANES spectra were acquired three times at the same 580 spot on the reference materials. 581 582 Petrologic, mineralogical, and geochemical characterizations 583 A petrographic thin section with a thickness of ca. 30 μ m was prepared from the 584 rock core sample after embedding in LR White resin (London Resin Co. Ltd., 585 Aldermaston, England) and then examined under a polarized microscope (BX51-P; 586 Olympus Co. Ltd., Tokyo, Japan). For X-ray diffraction (XRD) analysis, the rock core 587 sample was pulverized using a tungsten carbide mortar and pestle. The clay-sized fraction 588 was isolated by dispersing the powder in deionized water, followed by centrifugation at 589 3000 rpm for 5 min and freeze-drying of the supernatant. Both whole-rock and 590 clay-fraction samples, along with reference minerals, were analyzed using a RINT-2100 591 X-ray diffractometer (Rigaku Co. Ltd., Tokyo, Japan) operated at 40 kV and 30 mA with 592 Cu K α radiation ( λ = 1.5406 Å). For randomly oriented samples, XRD patterns were 593 collected from 5° to 90° at a scanning speed of 10° min -1. Oriented samples of the clay 594 fraction were scanned from 5° to 20° at the same speed. Ethylene glycol solvation and 595 thermal treatment (500 °C for 1 h) were applied to the clay fraction and selected reference 596 minerals to identify phyllosilicate phases. Reference minerals included enstatite, diopside 597 (N’s Mineral Co. Ltd., Niigata, Japan), quartz (FUJIFILM Wako, Osaka, Japan), 598 phlogopite (N’s Mineral Co. Ltd., Niigata, Japan), talc (Crown Talc PP JP Grade; 599 Matsumura Sangyo Co. Ltd., Osaka, Japan), clinochlore, and a JCSS reference 600 hydrobiotite (JCSS-5501; 601 ((K0.55Na0.08Ca0.06)[Mg2.34Fe3+ 0.39Fe2+ 0.08Ti0.07][Si3.02Al0.96Fe3+ 0.02]O10(OH)2)66), which is 602 an interstratified phlogopite/vermiculite mineral. 603 The 3-mm-thick rock section used for microbial characterizations was also 604 examined by environmental scanning electron microscopy coupled with 605 energy-dispersive X-ray spectroscopy (ESEM-EDS) without applying a conductive 606 coating. ESEM-EDS analysis was done with a JSM-6510LA ESEM (JEOL, Tokyo, 607 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 27 Japan) equipped with an EDS system, operated at 15-kV accelerating voltage and 608 60–80-Pa chamber pressure. The raw X-ray intensities were collected with an 609 acquisition time of 60 seconds and processed using the ZAF correction method via the 610 JEOL Analysis Station software. For semi-quantitative analysis, standardless-based 611 procedures were applied. 612 μ -XRF mapping was performed using an SFXM at the beamline BL17SU at 613 the SPring-8. At the BL17SU, the incident soft X-ray beam was focused to ca. 1.5 μ m 614 (vertical) × 3.5 μ m (horizontal) using a Wolter mirror as a focusing component, and the 615 fluorescent X-rays emitted from the sample were detected with a silicon drift detector in 616 the PFY mode 67–69. An excitation energy of 2000 eV was applied for the elemental 617 mapping of Al (Fig. 2b). After μ -XRF mappings, Al K-edge, Fe L3-edge, and S K-edge 618 XANES spectra were collected from the rock section and reference materials, including 619 phlogopite, hydrobiotite, talc, diopside, hematite (FUJIFILM Wako, Osaka, Japan), 620 elemental sulfur (FUJIFILM Wako, Osaka, Japan), sodium sulfite (Na 2SO3; FUJIFILM 621 Wako, Osaka, Japan), and sodium sulfate (Na 2SO4; FUJIFILM Wako, Osaka, Japan). 622 Excitation energy ranges for the Al K-edge, Fe L3-edge, and S K-edge were ca. 623 1550–1600 eV, 700–715 eV, and 2460–2490 eV with energy steps of 0.47 eV,0.15 eV, 624 and 0.22 eV, respectively. The S K -edge XANES spectra of troilite, marcasite, and 625 pyrite were obtained from the ID21 Sulfur XANES spectra database at the European 626 Synchrotron Radiation Facility (ESRF) 70. The XANES spectra were calibrated and 627 normalized using the Athena software71. 628 To determine the iron valence state without being exposed to the atmosphere, we 629 prepared a new 3-mm-thick rock section using a precision diamond wire saw (DWS 630 3400; Meiwa Fosis Corp., Tokyo, Japan) inside an Ar-purged glove box. The rock 631 section was loaded into a He-purged sample holder at the beamline BL36XU of the 632 SPring-8 for hard X-ray SFXM analysis. Oxygen levels in the glove box and the sample 633 holder were less than 0.5 ppm. The incident hard X-ray beam was focused to ca. 1.6 μ m 634 (vertical) × 0.3 μ m (horizontal) with a KB mirror as a focusing component, and the 635 fluorescent X-rays emitted from the sample were detected with a silicon drift detector 636 (Vortex-ME4; Hitachi High-Tech Corp., Tokyo, Japan) in the PFY mode 72. An 637 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 28 excitation energy of 8000 eV was applied for the elemental mapping of K. After μ -XRF 638 mappings, Fe K-edge XANES spectra were collected from the rock section and 639 nontronite (NAu-2; ((M + 0.97)[Si7.57Al0.01Fe0.42][Al0.52Fe3.32Mg0.7]O20(OH)4)28,73 (Fe(III) 640 standard). An excitation energy range for the Fe K-edge was 7105–7120 eV with an 641 energy step of 0.35eV. The Fe K-edge XANES spectrum of FeCl2 (Fe(II) standard) was 642 obtained from the XAFS Standard Sample Database at BL14B2 in SPring-8 29. The Fe 643 K-edge XANES spectra were also calibrated and normalized using the Athena software. 644 To assess chemical homogeneity, reference minerals without external 645 certification (enstatite, diopside, talc, phlogopite, and clinochlore) were analyzed using 646 an ESEM-EDS system under high-vacuum conditions (Extended Data Fig. 6). SEM 647 specimens were prepared by embedding the samples in LR White resin, followed by 648 polishing and carbon coating. 649 650 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 29 651 Extended Data Fig. 6 | Chemical homogeneity of reference minerals. Stacked 652 energy-dispersive X-ray spectroscopy (EDS) spectra from 10 grains each of enstatite, 653 diopside, phlogopite, talc, and clinochlore demonstrate compos itional homogeneity. 654 Peak intensities for major elements are consistent with the expected stoichiometry of 655 each mineral. 656 657 Data availability 658 All data needed to evaluate the conclusions in the paper are present in the paper and/or the 659 Supplementary Materials.660 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 30

661 1 Warr, O. et al. 86Kr excess and other noble gases identify a billion-year-old 662 radiogenically-enriched groundwater system. Nat. Commun. 13, 3768 (2022). 663 2 Holland, G., Lollar, B. S., Li, L., Lacrampe-Couloume, G., Slater, G. F. & 664 Ballentine, C. J. Deep fracture fluids isolated in the crust since the Precambrian 665 era. Nature 497, 357–360 (2013). 666 3 Onstott, T. C. et al. Paleo–rock-hosted life on Earth and the search on Mars: a 667 review and strategy for exploration. Astrobiology 19, 1230–1262 (2019). 668 4 O’Driscoll, B. & VanTongeren, J. A. Layered intrusions: from petrological 669 paradigms to precious metal repositories. Elements 13, 383–389 (2017). 670 5 Cawthorn, R. (2015). The Bushveld Complex, South Africa. In: Charlier, B., 671 Namur, O., Latypov, R., Tegner, C. (eds) Layered Intrusions. Springer Geology. 672 Springer, Dordrecht. 673 6 Suzuki, Y., Webb, S. J., Kouduka, M. et al. Subsurface microbial colonization at 674 mineral-filled veins in 2-billion-year-old mafic rock from the Bushveld Igneous 675 Complex, South Africa. Microb. Ecol. 87, 116 (2024). 676 7 Suzuki, Y. et al. Deep microbial proliferation at the basalt interface in 33.5–104 677 million-year-old oceanic crust. Commun. Biol. 3, 131 (2020). 678 8 Zema, M., Ventruti, G., Iacalamita, M. & Scordari, F. Kinetics of 679 Fe-oxidation/deprotonation process in Fe-rich phlogopite under isothermal 680 conditions. Am. Mineral. 95, 1458–1466 (2010). 681 9 Ma, T., Sun, H., Peng, T. & Zhang, Q. Transformation process from phlogopite 682 to vermiculite under hydrothermal conditions. Appl. Clay Sci. 208, 106094 683 (2021). 684 10 Francis, D. Columbia Hills—an exhumed layered igneous intrusion on Mars? 685 Earth Planet. Sci. Lett. 310, 59–64 (2011). 686 11 Bedle, H., Cooper, C. M. & Frost, C. D. Nature versus nurture: preservation and 687 destruction of Archean cratons. Tectonics 40, e2021TC006714 (2021). 688 12 Nisson, D. M. et al. Radiolytically reworked Archean organic matter in a 689 habitable deep ancient high-temperature brine. Nat. Commun. 14, 6163 (2023). 690 13 Sherwood Lollar, B. et al. A window into the abiotic carbon cycle: acetate and 691 formate in fracture waters in 2.7 billion year-old host rocks of the Canadian 692 Shield. Geochim. Cosmochim. Acta 294, 295–314 (2021). 693 14 Nisson, D. M. et al. Hydrogeochemical and isotopic signatures elucidate deep 694 subsurface hypersaline brine formation through radiolysis-driven water–rock 695 interaction. Geochim. Cosmochim. Acta 340, 65–84 (2023). 696 15 Lollar, G. S. et al. ‘Follow the water’: Hydrogeochemical constraints on 697 microbial investigations 2.4 km below surface at the Kidd Creek Deep Fluid and 698 Deep Life Observatory. Geomicrobiol. J. 36, 859–872 (2019). 699 16 Hawkesworth, C., Cawood, P. A., Dhuime, B. & Kemp, T. Tectonic processes 700 and the evolution of the continental crust. J. Geol. Soc. 181, jgs2024-027 (2024). 701 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 31 17 Becker, M., Harris, P. J., Wiese, J. G. & Bradshaw, D. J. Mineralogical 702 characterisation of naturally floatable gangue in Merensky Reef ore flotation. Int. 703 J. Miner. Process. 93, 246–255 (2009). 704 18 Buick, I. S., Maas, R. & Gibson, R. Precise U–Pb titanite age constraints on the 705 emplacement of the Bushveld Complex, South Africa. J. Geol. Soc. 158, 3–6 706 (2001). 707 19 Allwright, A. J., de Lange, S., Lubbe, R., Mbonambi, L., Vivier, K. & 708 Witthuser, K. T. Research-based exploration of deep groundwater within the 709 eastern limb of the Bushveld Igneous Complex for hydrogeological 710 characterisation and potential future water resource identification. Final report. 711 Water Research Commission, Pretoria (2025). ISBN 978-0-6392-0706-3. 712 20 Friese, A. et al. A simple and inexpensive technique for assessing contamination 713 during drilling operations. Limnol. Oceanogr. Methods, 15, 200–211 (2017). 714 21 Leinweber, P., Kruse, J., Walley, F. L., Gillespie, A., Eckhardt, K. U., Blyth, R. 715 I. & Regier, T. Nitrogen K-edge XANES—an overview of reference compounds 716 used to identify ‘unknown’ organic nitrogen in environmental samples. J. 717 Synchrotron Radiat. 14, 500–511 (2007). 718 22 Gillespie, A. W., Walley, F. L., Farrell, R. E., Regier, T. Z. & Blyth, R. I. 719 Calibration method at the N K-edge using interstitial nitrogen gas in solid-state 720 nitrogen-containing inorganic compounds. J. Synchrotron Radiat. 15, 532–534 721 (2008). 722 23 Moore, D. M. & Reynolds, R. C. Jr. X-ray diffraction and the identification and 723 analysis of clay minerals, 2nd edn (Oxford Univ. Press, 1997). 724 24 Li, D., Bancroft, G. M., Fleet, M. E., Feng, X.-H. & Pan, Y. Al K-edge XANES 725 spectra of aluminosilicate minerals. Am. Mineral. 80, 432–440 (1995). 726 25 van Aken, P. A. & Liebscher, B. Quantification of ferrous/ferric ratios in 727 minerals: new evaluation schemes of Fe L /i5 ,/i5 electron energy-loss near-edge 728 spectra. Phys. Chem. Miner. 29, 188–200 (2002). 729 26 Van Groeningen, N., ThomasArrigo, L. K., Byrne, J. M., Kappler, A., Christl, I. 730 & Kretzschmar, R. Interactions of ferrous iron with clay mineral surfaces during 731 sorption and subsequent oxidation. Environ. Sci. Process. Impacts 22, 732 1355–1367 (2020). 733 27 Berry, A. J., O’Neill, H. S. C., Jayasuriya, K. D., Campbell, S. J. & Foran, G. J. 734 XANES calibrations for the oxidation state of iron in a silicate glass. Am. 735 Mineral. 88, 967–977 (2003). 736 28 Qian, Y., Scheinost, A. C., Grangeon, S., Greneche, J.-M., Hoving, A., Bourhis, 737 E., Maubec, N., Churakov, S. V. & Fernandes, M. M. Oxidation state and 738 structure of Fe in nontronite: from oxidizing to reducing conditions. ACS Earth 739 Space Chem. 7, 1868–1881 (2023). 740 29 Ofuchi, H., Matsumoto, T. & Honma, T. Construction of XAFS standard sample 741 database at BL14B2 in SPring-8. Radiat. Phys. Chem. 218, 111581 (2024). 742 30 Hogg, C. S. & Meads, R. E. A Mössbauer study of thermal decomposition of 743 biotites. Mineral. Mag. 40, 79–88 (1975). 744 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 32 31 Cawthorn, R. G. & Walraven, F. Emplacement and crystallization time for the 745 Bushveld Complex. J. Petrol. 39, 1669–1687 (1998). 746 32 Li, C., Ripley, E. M., Sarkar, A., Shin, D. & Maier, W. D. Origin of 747 phlogopite-orthopyroxene inclusions in chromites from the Merensky Reef of 748 the Bushveld Complex, South Africa. Contrib. Mineral. Petrol. 150, 119–130 749 (2005). 750 33 Wagner, W. & Pruß, A. The IAPWS formulation 1995 for the thermodynamic 751 properties of ordinary water substance for general and scientific use. J. Phys. 752 Chem. Ref. Data 31, 387–535 (2002). 753 34 Scoates, J. S., Wall, C. J., Friedman, R. M., Weis, D., Mathez, E. A. & 754 VanTongeren, J. A. Dating the Bushveld Complex: timing of crystallization, 755 duration of magmatism, and cooling of the world’s largest layered intrusion and 756 related rocks. J. Petrol. 62, 1–39 (2021). 757 35 de la Calle, C. & Suquet, H. Vermiculite. In Hydrous Phyllosilicates (Exclusive of 758 Micas) (ed. Bailey, S. W.) 455–496 (De Gruyter, 1988). 759 36 Janzen, M. P., Nicholson, R. V. & Scharer, J. M. Pyrrhotite reaction kinetics: 760 reaction rates for oxidation by oxygen, ferric iron, and for nonoxidative 761 dissolution. Geochim. Cosmochim. Acta 64, 1511–1522 (2000). 762 37 Beaver, R. C. & Neufeld, J. D. Microbial ecology of the deep terrestrial 763 subsurface. ISME J. 18, wrae091 (2024). 764 38 Dong, H., Zeng, Q., Sheng, Y. et al. Coupled iron cycling and organic matter 765 transformation across redox interfaces. Nat. Rev. Earth Environ. 4, 659–673 766 (2023). 767 39 Price, P. B. & Sowers, T. Temperature dependence of metabolic rates for 768 microbial growth, maintenance, and survival. Proc. Natl Acad. Sci. USA 101, 769 4631–4636 (2004). 770 40 Hoehler, T. & Jørgensen, B. Microbial life under extreme energy limitation. Nat. 771 Rev. Microbiol. 11, 83–94 (2013). 772 41 Templeton, A. S. & Caro, T. A. The rock-hosted biosphere. Annu. Rev. Earth 773 Planet. Sci. 51, 493–519 (2023). 774 42 do Nascimento Vieira, A., Kleinermanns, K., Martin, W. F. & Preiner, M. The 775 ambivalent role of water at the origins of life. FEBS Lett. 594, 2717–2733 776 (2020). 777 43 Ménez, B., Pasini, V. & Brunelli, D. Life in the hydrated suboceanic mantle. Nat. 778 Geosci. 5, 133–137 (2012). 779 44 Klein, F., Humphris, S. E., Guo, W., Schubotz, F., Schwarzenbach, E. M. & Orsi, 780 W. D. Fluid mixing and the deep biosphere of a fossil Lost City-type 781 hydrothermal system at the Iberia Margin. Proc. Natl Acad. Sci. USA 112, 782 12036–12041 (2015). 783 45 Lima-Zaloumis, J., Neubeck, A., Ivarsson, M., Bose, M., Greenberger, R., 784 Templeton, A. S. et al. Microbial biosignature preservation in carbonated 785 serpentine from the Samail Ophiolite, Oman. Commun. Earth Environ. 3, 231 786 (2022). 787 46 Ivarsson, M., Bach, W., Broman, C., Neubeck, A. & Bengtson, S. Fossilized life 788 in subseafloor ultramafic rocks. Geomicrobiol. J. 35, 460–467 (2018). 789 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 33 47 Suzuki, Y., Koduka, M., Brenker, F. E., Brooks, T., Glamoclija, M., Graham, H. 790 V. et al. Submicron-scale detection of microbes and smectite from the interior of 791 a Mars-analogue basalt sample by optical-photothermal infrared spectroscopy. 792 Int. J. Astrobiol. 24, e1 (2025). 793 48 Li, X., Zhang, D., Bai, Y., Wang, W., Liang, J. & Cheng, J. X. Fingerprinting a 794 living cell by Raman integrated mid-infrared photothermal microscopy. Anal. 795 Chem. 91, 10750–10756 (2019). 796 49 Ménez, B. et al. Abiotic synthesis of amino acids in the recesses of the oceanic 797 lithosphere. Nature 564, 59–63 (2018). 798 50 Hickman-Lewis, K., Cuadros, J., Yi, K. et al. Aluminous phyllosilicates promote 799 exceptional nanoscale preservation of biogeochemical heterogeneities in 800 Archaean siliciclastic microbial mats. Nat. Commun. 16, 2726 (2025). 801 51 Tokumaru, A. et al. Co-enrichment of Ce and organics in microbe-like 802 structures at the deep-sea ferromanganese crust surface. Phil. Trans. R. Soc. A 803 383, 20240432 (2026). 804 52 Chan, C. S. et al. Lithotrophic iron-oxidizing bacteria produce organic stalks to 805 control mineral growth: implications for biosignature formation. ISME J. 5, 806 717–727 (2011). 807 53 Suga, H. et al. Spatially resolved distribution of Fe species around microbes at 808 the submicron scale in natural bacteriogenic iron oxides. Microbes Environ. 32, 809 283–287 (2017). 810 54 Attwood, D. Soft X-rays and extreme ultraviolet radiation: principles and 811 applications. Cambridge Univ. Press (2000). 812 55 Ming, D. W., Mittlefehldt, D. W., Morris, R. V., Golden, D. C., Gellert, R., Yen, 813 A. et al. Geochemical and mineralogical indicators for aqueous processes in the 814 Columbia Hills of Gusev crater, Mars. J. Geophys. Res. Planets 111, E02S12 815 (2006). 816 56 Liu, Y., Tice, M. M., Schmidt, M. E., Treiman, A. H., Kizovski, T. V., Hurowitz, 817 J. A., Zorzano, M. P. et al. An olivine cumulate outcrop on the floor of Jezero 818 crater, Mars. Science 377, 1513–1519 (2022). 819 57 Farley, K. A., Stack, K. M., Shuster, D. L., Horgan, B. H. N., Hurowitz, J. A., 820 Tarnas, J. D. et al. Aqueously altered igneous rocks sampled on the floor of 821 Jezero crater, Mars. Science 377, eabo2196 (2022). 822 58 Hurowitz, J. A., Tice, M. M., Allwood, A. C., Cable, M. L., Hand, K. P., 823 Murphy, A. E., et al. Redox-driven mineral and organic associations in Jezero 824 Crater, Mars. Nature 645, 332–340 (2025). 825 59 Hollis, J. R., Abbey, W., Beegle, L. W., Bhartia, R., Ehlmann, B. L., Miura, J., 826 Monacelli, B., Moore, K., Nordman, A. & Scheller, E. A deep-ultraviolet Raman 827 and fluorescence spectral library of 62 minerals for the SHERLOC instrument 828 onboard Mars 2020. Planet. Space Sci. 209, 105356 (2021). 829 60 Barnes, J., McIntyre, J. R., Nisbet, B. W. & Williams, C. R. Platinum-group 830 element mineralisation in the Munni Munni Complex, Western Australia. 831 Mineral. Petrol. 42, 141–164 (1990). 832 61 Barnes, S. J. & Hoatson, D. M. The Munni Munni complex, Western Australia: 833 stratigraphy, structure and petrogenesis. J. Petrol. 35, 715–751 (1994). 834 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 34 62 Mukasa, S. B., Wilson, A. H. & Carlson, R. W. A multielement geochronologic 835 study of the Great Dyke, Zimbabwe: significance of the robust and reset ages. 836 Earth Planet. Sci. Lett. 164, 353–369 (1998). 837 63 Wilson, A. H. & Prendergast, M. D. Platinum-group element mineralisation in 838 the Great Dyke, Zimbabwe, and its relationship to magma evolution and magma 839 chamber structure. S. Afr. J. Geol. 104, 319–342 (2001). 840 64 Järvinen, V., Halkoaho, T., Konnunaho, J., Heinonen, J. S. & Rämö, O. T. 841 Parental magma, magmatic stratigraphy and reef-type PGE enrichment of the 842 2.44 Ga mafic–ultramafic Näränkävaara layered intrusion, northern Finland. 843 Mineralium Deposita 55, 1535–1560 (2020). 844 65 Becraft, E. D. et al. Evolutionary stasis of a deep subsurface microbial lineage. 845 ISME J. 15, 2830–2842 (2021). 846 66 Kikuchi, R. & Kogure, T. Structural and compositional variances in 847 ‘hydrobiotite’ sample from Palabora, South Africa. Clay Sci. 22, 39–52 (2018). 848 67 Oura, M., Suga, H., & Suzuki, Y. Introduction to scanning fluorescence X-ray 849 microscopy and its applications for earth and planetary science. Jpn. Mag. 850 Mineral. Petrol. Sci. 55, gkk.250920 (2026). (in Japanese). 851 68 SPring-8·SACLA Annual Report FY2023, 99－ 102. 852 69 SPring-8·SACLA Annual Report FY2024, 99－ 103. 853 70 ID21 Sulfur XANES spectra database, European Synchrotron Radiation Facility 854 (ESRF). Created by E. Chalmin and ID21 users. Available at: 855 https://www.esrf.fr/home/UsersAndScience/Experiments/ID21/php.html 856 (Accessed: 18 February 2026). 857 71 Ravel, B. & Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: data analysis 858 for X-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Radiat. 12, 859 537–541 (2005). 860 72 Uruga, T., Tada, M., Sekizawa, O., Takagi, Y., Yokoyama, T. & Iwasawa, Y. 861 SPring-8 BL36XU: Synchrotron radiation X-ray-based multi-analytical 862 beamline for polymer electrolyte fuel cells under operating conditions. Chem. 863 Rec. 19, 1444–1456 (2019). 864 73 Keeling, J. L., Raven, M. D., & Gates, W. P. Geology and characterization of 865 two hydrothermal nontronites from weathered metamorphic rocks at the Uley 866 graphite mine, South Australia. Clays Clay Miner. 48, 537–548 (2000). 867 868 Acknowledgments 869 We are grateful to Master Drilling for their support during the drilling operations. 870 The authors would like to acknowledge Mpho Molautsi, Katja Heeschen, and Kwena 871 Mathopa for their on-site assistance. We thank Eoghan Dillon (Photothermal 872 Spectroscopy Corp.) for insightful discussions and technical advice regarding 873 O-PTIR measurements. We also thank the beamline scientists 874 at NanoTerasu (BL13U) and SPring-8 (BL17SU and BL36XU) for their technical 875 support. 876 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 35 μ -XRF and XANES measurements were performed at BL13U of 877 NanoTerasu with the approval of the Japan Synchrotron Research Institute 878 (JASRI) (Proposal No. 2025A9013), and at BL17SU and BL36XU of SPring-8 with 879 the approval of RIKEN (Proposal No. 20250069) and JASRI (Proposal No. 880 2025A1617). ESEM-EDS analysis was supported by the "Advanced Research 881 Infrastructure for Materials and Nanotechnology in Japan (ARIM)" of the Ministry 882 of Education, Culture, Sports, Science and Technology (MEXT) (Proposal Nos. 883 JPMXP1224UT0317 and JPMXP1225UT0023). 884 The Bushveld Drilling Project (BVDP) was supported by the International 885 Continental Scientific Drilling Program, the National Research Foundation of South 886 Africa, the Deutsche Forschungsgemeinschaft (DFG) (Grant No. 684792 to J.K.), 887 and the South African Council for Geoscience. This work was also supported by 888 the JSPS/NRF Bilateral Joint Research Project (Grant No. JPJSBP120246501 to Y.S. 889 and J.C.), the Astrobiology Center Program of National Institutes of Natural Sciences 890 (NINS) (Grant No. AB0502 to Y.S.), JSPS KAKENHI (Grant Nos. JP25K22489 to 891 Y.S. and JP25KJ1046 to T.K.), and JST SPRING (Grant No. JPMJSP2108 to T.K.). 892 The author would like to thank Enago (www.enago.jp ) for the English language 893 review. 894 895 Author Contributions 896 F.R., R.K., L.A., S.W., and R.T. contributed to the successful application and 897 implementation of the Bushveld ICDP project. J.C., J.K., K.M., S.H., A.A., S.W., 898 T.K., and Y.S. prepared and performed on-site sample collection. J.K. planned and 899 set up the on-site contamination control, and C.N., T.M., F.R., and M.M. performed 900 the geological curation of the examined rock sample. T.K., M.K., and Y.S. conducted 901 the experimental work, including data collection and analysis. H.K. assisted with 902 O-PTIR analysis. H.S., T.K., and T.W. assisted with synchrotron-based experiments 903 at NanoTerasu. H.S., T.I., T.K., T.U, and M.O. assisted with synchrotron-based 904 experiments at SPring-8. F.R., R.K., L.A., R.T., T.K., and Y.S. drafted the 905 manuscript. All authors discussed the results and approved the final manuscript. 906 907 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint 36 Competing Interest Declaration: All authors declare that they have no competing 908 interests. 909 910 Correspondence and requests for materials should be addressed to Y.S. 911 912 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted April 14, 2026. ; https://doi.org/10.64898/2026.04.13.717956doi: bioRxiv preprint