Author: admin

What do the rumblings of Iceland’s volcanoes have in common with the now peaceful volcanic islands off Scotland’s western coast and the spectacular basalt columns of the Giant’s Causeway in Northern Ireland?

About sixty million years ago, the Icelandic mantle plume—a fountain of hot rock that rises from Earth’s core-mantle boundary—unleashed volcanic activity across a vast area of the North Atlantic, extending from Scotland and Ireland to Greenland.

For decades, scientists have puzzled over why this burst of volcanism was so extensive. Now, research led by the University of Cambridge has found that differences in the thickness of tectonic plates around the North Atlantic might explain the widespread volcanism.

The researchers compiled seismic and temperature maps of Earth’s interior, finding that patches of thinner tectonic plate acted like conduits, funnelling the plume’s molten rock over a wide area.

Iceland , which is one of the most volcanically active places on Earth, owes its origin largely to the mantle plume. Beyond volcanism, the Iceland Plume’s influence even extends to shaping the seafloor and ocean circulation in the North Atlantic and, in turn, climate through time. Despite its global significance, many aspects of the plume’s behaviour and history remain elusive.

“Scientists have a lot of unanswered questions about the Iceland plume,” said Raffaele Bonadio , a geophysicist at Cambridge’s Department of Earth Sciences and lead author of the study.

Bonadio set out to explain why the plume’s volcanic imprint was much more widespread sixty million years ago—before the Atlantic opened—forming volcanoes and lava outpourings stretching over thousands of kilometres. The pattern could be explained by the mantle plume spreading outward in a branched, flowing formation, Bonadio explained, “but evidence for such flow has been scarce.”

In search of answers, Bonadio focussed on a segment of the North Atlantic Igneous Province to better understand the complex distribution of volcanoes in Scotland and Ireland. He wanted to know if the structure of Earth’s tectonic plates played a role in the surface expression of volcanism.

Using seismic data extracted from earthquakes, Bonadio created a computer-generated image of Earth’s interior beneath Britain and Ireland. This method, known as seismic tomography, works similarly to a medical CT scan, revealing hidden structures deep within the planet. Bonadio coupled this with seismic thermography measurements—a new method developed by the team—which reveal variations in the temperature and thickness of the tectonic plate.

He found that northwest Scotland and Ireland’s volcanoes formed in areas where the lithosphere (Earth’s rigid outer layer that makes up the tectonic plates) is thinner and weaker.

“We see ancient volcanoes concentrated within this corridor of thin lithosphere beneath the Irish Sea and surrounding areas,” said Bonadio. He thinks the hot plume material was preferentially funnelled along this corridor, ponding in the thin plate areas due to its buoyancy.

Previously, some scientists had put forward alternative, non-mantle plume origins for the volcanic activity, said Bonadio. But his new research shows the scattering could be explained by the magma being diverted and re-routed to areas of thinner lithosphere.

Sergei Lebedev, from the University of Cambridge said, “this striking correlation suggests that hot plume material eroded the lithosphere in this region. This resulting combination of thin lithosphere, hot asthenosphere and decompression melting likely caused the uplift and volcanic activity.”

Previously, the authors have found a close link between the uneven distribution of earthquakes in Britain and Ireland and the thickness of the lithosphere, showing how the scars left by the mantle plume influence seismic hazards today.

Bonadio and Lebedev are also using their methods to map geothermal energy resource potential. “In Britain and Ireland, the greatest supply of heat from the Earth’s mantle is in the same places where volcanoes erupted sixty million years ago, and where the lithosphere is thinner,” said Lebedev. He and Bonadio are working with international colleagues to apply their new seismic thermography methods to global geothermal assessment.

Phylogenetic framework and HMMs of sulfur-cycling proteins

To discern sulfur-cycling genes or proteins from their functionally divergent homologues, phylogenetic analysis was conducted for 116 sulfur-cycling proteins (Supplementary Table 1). For each protein family, sequences of enzymes with biochemically validated functions, including related sequences of enzymes with divergent functions (outgroups), were identified by literature surveys and recovered from SwissProt⁶¹. Additional homologues of experimentally validated proteins in KEGG prokaryotic genomes were retrieved using KEGG BLAST Search (https://www.genome.jp/tools/blast/; E value: 10⁻⁴). Distant homologues that did not align properly with biochemically characterized proteins (alignment length covered <50% of query and target length) were removed. The resulting homologues were de-replicated using CD-HIT v4.8.1 (ref. ⁶²), with longest sequences retained as representatives. For computational efficiency, different clustering identity thresholds (75–95%) were chosen for de-replication to ensure the total number of representative sequences for phylogenetic analysis did not exceed 500. The genome context of the representative KEGG homologues was analysed by retrieving genes located in a distance of fewer than n genes (n = 7–15), followed by annotation using biochemically characterized gene clusters based on BLAST analysis⁶³. All representative KEGG homologues were further aligned with biochemically validated proteins and outgroups using Muscle v3.8.1551 (ref. ⁶⁴). Poorly aligned regions were excised using TrimAl v1.4.rev15 (ref. ⁶⁵). Protein phylogeny was inferred from the trimmed alignment using FastTree v2.1.7 (ref. ⁶⁶) with -wag and -gamma options. Statistical support for each branch of the tree was estimated by nonparametric bootstrap (n = 100).

Information on reference sequences from biochemically verified proteins (for example, ingroup/outgroup, conserved residues or motif) and genomic contexts of all homologues were mapped on the tree. To identify monophyletic, orthologous clades within each tree, interior nodes of the annotated tree were scrutinized using the following criteria: (1) bootstrap support over 70%; (2) presence of at least one biochemically verified ingroup protein and absence of outgroup proteins; and (3) consistent gene neighbouring patterns and biochemical traits (thatis, catalytic residues and PFAM domain composition) among its members. All descendants of the identified clade were regarded as functional orthologs of the biochemically verified protein. If possible, existing definitions of orthologous clades from previous phylogenetic analysis of sulfur-cycling proteins was preserved, including the well-recognized clades in the phylogeny of DsrAB¹⁰ and Sqr⁶⁷. For proteins for which the biochemically validated ingroup proteins formed polyphyletic groups, multiple monophyletic clades were proposed to fulfill our criteria.

To leverage our phylogenetic framework for large-scale homology searches, sequences from the defined monophyletic clades of sulfur-cycling proteins were used to build HMMs. A cut-off that optimizes the sensitivity and specificity of homology search was calculated for each HMM using receiver operating curve (ROC)⁶⁸. This cut-off was embedded in the HMMER profile HMM file as the gathering threshold of the model (HMMER User’s Guide, p. 108; ref. ⁶⁹). The performance of the newly developed HMMs was compared with that of six published sets of HMMs for sulfur metabolism genes, including those from KoFam⁷⁰, TIGRFAM⁷¹, PFAM⁷², metabolicHMM⁷³, DiSCo⁷⁴, Teng et al.⁷⁵ and HMS-S-S⁷⁶. This was accomplished by querying each HMM against the phylogeny-curated protein dataset using hmmsearch in HMMER v3.2.1 with a predefined cut-off (http://hmmer.org/). The performance of the various HMM sets in detecting sulfur-cycling genes and proteins was assessed in terms of specificity, sensitivity, and F score (Supplementary Text). F score balancing both precision and recall of the homology detection was calculated using F score = 2 × (precision × recall) / (precision + recall).

Sulfur-cycling genes in bacterial and archaeal genomes

To provide a comprehensive overview of sulfur metabolism across bacteria and archaea, the phylogeny-derived HMMs were searched against all genomes in GTDB release 95 (ref. ²³) using hmmsearch with the –cut_ga option. Each retrieved homologue was then searched against the full set of phylogeny-derived HMMs using hmmscan with –cut_ga, and annotated as the HMM showing the highest score. For initial screening, a subset of genes (n = 42) was selected as markers for specific sulfur metabolisms if the gene: (1) has been widely recognized as a marker for a specific sulfur metabolism, (2) encodes a catalytic subunit essential for the activity of enzymatic complex; or (3) on its own confers a specific sulfur redox transformation (see justification for each of selected genes in Supplementary Table 1). The retrieved homologues were further curated using our reference phylogeny of sulfur proteins. Specifically, the GTDB homologues were aligned with sequences contained in our reference phylogeny using Muscle. A maximum-likelihood tree was reconstructed from the alignment trimmed by TrimAl. The tree was overlaid with biochemical information and data on the genomic context of sulfur genes, and visualized using ggtree⁷⁷. The physiological role of the GTDB homologues was interpreted on the basis of their evolutionary relationship with biochemically validated proteins and genome context. To predict the dissimilatory iron(iii) reduction potential, GTDB genomes were screened for marker genes involved in EET on the basis of homology search and/or motif analysis. Homologues of iron(iii) reduction genes with established HMMs in FeGenie database (https://github.com/Arkadiy-Garber/FeGenie/tree/master/hmms/iron/iron_reduction) were retrieved using hmmsearch from HMMER v3.2.1, with cut-off recommended by FeGenie (https://github.com/Arkadiy-Garber/FeGenie/blob/master/hmms/iron/HMM-bitcutoffs.txt). Additionally, homologues of MmcA gene, which is involved in dissimilatory iron(iii) reduction in Methanosarcina acetivorans⁷⁸, were extracted using BLASTP on the basis of an e-value of 10⁻⁴. The outer membrane MHCs responsible for EET with metal oxides in anaerobic methanotrophs⁷⁹ and putative electroactive bacteria⁸⁰ were recognized on the basis of the following: (1) the presence of four or more haem-binding motifs (CXXCH); and (2) their predicted outer membrane or extracellular localization, as determined by DeepProLoc v1.0 (ref. ⁸¹).

Annotation and metabolic reconstruction of specific sulfur-cycling microbial lineages

The genomes of microbial lineages of interest were downloaded from the GTDB database. The protein-coding genes were predicted from the genome using Prodigal v2.6.3 with default setting. The predicted genes were annotated using KoFam⁷⁰, PFAM⁷², and the EggNOG⁸² database. Additional metabolic pathways were predicted using HMMs (Supplementary Table 6) downloaded from dbCAN⁸³, metabolicHMM⁷³, CANT-HYD⁸⁴, MicRhoDE⁸⁵ and FeGenie⁸⁶. For HMM-based annotation, the HMMs were used as queries to search against microbial genomes using hmmsearch from HMMER v3.2.1, with cut-off recommended by each database (-T, -domT or -cut_ga options). The cellular localization of the protein was predicted using Signalp v6.0 (ref. ⁸⁷). The completeness of the KEGG metabolic pathway was calculated on the basis of the definition of each module. The KEGG module is defined with a logic expression of K numbers that records the composition of enzymes in the pathway. A particular metabolic module was considered to be present in the genome when: (1) the diagnostic/marker genes of the module were detected; and (2) the overall completeness of the pathway module was >70%. The environmental distribution of the GTDB species was retrieved by searching their GTDB species name in the Sandpiper database⁸⁸. The occurrence of the GTDB species across biomes was downloaded as CSV from Sandpiper (https://sandpiper.qut.edu.au/) and further visualized with R v4.1.0.

Thermodynamic modelling

The Gibbs free energy associated with iron(iii)-dependent sulfur oxidation at environmentally relevant conditions was estimated by following the guidelines described previously⁸⁹. In brief, the actual Gibbs free energy of reaction (ΔG_r) was calculated using:

where ΔG_r⁰ refers to the standard Gibbs free energy of reaction, given in kJ mol⁻¹; R and T are the universal gas constant (8.314 J K⁻¹mol⁻¹) and the temperature in Kelvin, respectively; and Q_r is the reaction quotient. ΔG_r⁰ values were calculated from the values of the standard Gibbs free energy of formation (ΔG_f⁰) of reactants and products (Supplementary Table 7). Values of Q_r were determined from the activity (a_i) and the stoichiometric coefficient (v_i) of the i^th chemical species involved in the reaction using:

The activity of the solvent (that is, pure water) and the solids (that is, ferrihydrite and FeS) were taken to be 1. The activity of dissolved ions was related to the concentration (C_i) using:

where γ_i denotes the activity coefficient; C_i⁰ represents the standard state concentration (usually 1 M). γ_i for cations (that is, Fe²⁺) and anions (that is, HS⁻, S₂O₃²⁻ and SO₄²⁻) in solutions of different ionic strength were retrieved from Amend et al.⁸⁹. Sulfide speciation in aqueous phase across a range of pH was determined from the pH, and pK_a1 (7.04) and pK_a2 (11.96) of hydrogen sulfide.

Synthesis of ferrihydrite and poorly crystalline FeS

Synthetic ferrihydrite was prepared by titrating 1 M NaOH (Sigma Aldrich) into 0.1 M aqueous solution of FeCl₃ 6H₂O (Carl Roth) under vigorous stirring until pH 7.5 was reached, as described⁹⁰. The suspension was centrifuged (Centrifuge 5804 R, Eppendorf) at 4 °C, 12,857g and the ferrihydrite nanoparticles were washed thoroughly with deionized water to remove traces of chloride. The pellets were then freeze-dried (Alpha 1-4 LSCbasic, Christ) and stored at −20 °C for no longer than 3 weeks before use. The mineralogy was determined by LabRAM HR800 Raman microscope (Horiba Jobin-Yvon) equipped with a 532-nm neodymium-yttrium aluminium garnet laser and either 300 or 600 grooves/mm diffraction grating. Iron monosulfide (FeS; 30 mM) was prepared by mixing equal volume of 60 mM aqueous solution of Na₂S.9H₂O (Acros Organics) with 60 mM aqueous solution of FeCl₂.4H₂O (Sigma Aldrich) in an anaerobic chamber (Coy Lab) with 95% N₂ and 5% H₂ (O₂ < 1 ppm) atmosphere. The initially precipitated FeS is often designated as ‘amorphous FeS’ or ‘poorly crystalline FeS’⁹¹. The dissolved sulfide in the FeS stock is less than 50 µM. The FeS solution was freshly prepared and used on the same day.

Cultivation of D. alkaliphilus DSM 19089

D. alkaliphilus (DSM 19089, ATH2) was purchased from the German Collection of Microorganisms and Cell Cultures GmbH (DSMZ). The bacterium was cultivated at room temperature in an alkaline mineral medium (pH 9.3) containing 3 g NaCl (Carl Roth), 0.25 g K₂HPO₄ (Merck), 6.5 g Na₂CO₃ (Carl Roth), and 15 g NaHCO₃ (Sigma Aldrich) per liter of medium. After autoclaving, the medium was cooled down under N₂ atmosphere and supplemented aseptically with 1 ml liter⁻¹ of following components (all stored under anoxic conditions): 4 M NH₄Cl (Sigma Aldrich), 1 M MgCl₂ (Sigma Aldrich), trace element solution, Se-W solution, and four different vitamin solutions (DSMZ medium 1104). The culture was routinely grown under nitrate-reducing, sulfide-oxidizing conditions in 500 ml Schott bottles²⁷, with 2 mM Na₂S 9H₂O and 1.2 mM KNO₃ (Sigma Aldrich). This yielded a culture with an optical density at 600 nm (OD₆₀₀) of ~0.040, corresponding to a cell density of ~1.3 × 10⁸ cells per ml. To test alternative growth modes, five incubation experiments were conducted, each supplemented with different electron donors and acceptors (details provided below). For all experiments, regularly maintained cultures (30 ml) that have been depleted in sulfide (< 100 µM) and nitrate (< 10 µM) were used as inoculum. Incubations were set up in 60 ml serum bottles and sealed with butyl rubber stoppers in the anaerobic chamber (N₂:H₂ = 95:5). Each culture was then flushed with pure N₂ to remove H₂ in the headspace, and incubated in the dark at room temperature. All incubations, abiotic and biotic controls from each experiment were set up in triplicates.

1. (1)

   Incubations with sulfide and nitrate. The incubations were set up by supplying 2 mM sulfide and 2 mM nitrate to 30 ml pre-growns cells in 60 ml serum bottles. Sulfide and nitrate was spiked using syringes flushed with pure N₂. The growth was monitored by phase-contrast microscopy and by the measurement of sulfide and sulfate over 3 days.

2. (2)

   Incubations with elemental sulfur. The incubations were initiated by adding 0.1 g elemental sulfur in 3 ml MilliQ water (Sigma Aldrich) to each of the serum bottles, followed by autoclaving at 110 °C for 60 min. After sterilization, 30 ml pre-grown cells were inoculated into the S(0) suspension (approximately 94 mM) and incubated under an N₂ atmosphere for 15 days. Microbial activity was monitored by measuring sulfide and/or sulfate.

3. (3)

   Incubations with ferrihydrite and formate. Synthetic ferrihydrite (0.2 g) was ground into fine particles with an agate mortar and pestle before being added to the culture. Assuming ferrihydrite has a composition⁹² Fe(OH)₃, the final concentration of Fe(iii) was approximately 62 mM. Formate was spiked anoxically using a syringe to a final concentration of 10 mM. To test the coupling of ferrihydrite reduction and formate oxidation, parallel cultures were set up with either ferrihydrite or formate alone. The culture activity was monitored by measuring total Fe(ii) and formate concentrations over a 15 day incubation.

4. (4)

   Incubations with ferrihydrite and poorly crystalline FeS. Synthetic ferrihydrite (0.2 g) was supplied to the cultures as described in the incubation (3). To amend poorly crystalline FeS, 1 ml stock solution of freshly prepared FeS (30 mM) was anoxically spiked to the cultures using syringes, resulting in a final FeS concentration of 1 mM. Abiotic controls were prepared using 30 ml autoclaved cells as inoculum to test for chemical reactions between ferrihydrite and poorly crystalline FeS. Biotic controls amended with either ferrihydrite or FeS were set up to assess the impacts of residual sulfide and/or nitrate on culture activity. Cultures were sampled daily over 5 days for sulfate and total Fe(ii) measurements.

5. (5)

   Incubations with ferrihydrite and dissolved sulfide. The cultures were prepared similarly as incubation (4), but with dissolved sulfide replacing FeS. Due to the rapid chemical reaction between dissolved sulfide and ferrihydrite, dissolved sulfide was anoxically spiked daily at a concentration of 1 mM using N₂-flushed syringes. Abiotic controls and the sulfide-only biotic controls received dissolved sulfide at the same concentration and frequency. To trace the transformation of S and Fe over 5 days, subsamples were taken daily for measurement of S(0), total Fe(ii), sulfate, and Cline-extractable sulfide before the addition of sulfide. The consumption of dissolved sulfide in the cultures was monitored by sampling at 2, 10, 20, 35, 50 and 70 min after the spike of sulfide. The kinetics of sulfide consumption were modelled as a first-order reaction. The rate constant was estimated using the exponential decay model in the drm() function from the drc R package⁹³. To compare sulfate formation patterns with and without sulfide, cultures incubated with ferrihydrite and sulfide were sampled for sulfate measurement following two phases after the 1st sulfide spike. During phase I, detectable sulfide was present in the culture, and the samples were collected at 0, 11, 21, 37, 53 and 70 min of the incubation. Phase II, spanning the next 23 h, began once sulfide was depleted, with samples taken at 3, 5.5, 8.33, 12.33, 20.25 and 24 h. As a control for phase II, cells were incubated with chemically sulfidized ferrihydrite. Specifically, 1 mM sulfide was firstly added to 0.2 g ferrihydrite (approximately 62 mM) with 30 ml autoclaved cultures for chemical reaction. After 70 min, the reaction mixture was centrifuged (12,857g; room temperature) under anoxic conditions, and 30 ml of active cells were inoculated to resuspend the solid phase compounds (for example, FeS and S(0)) produced by chemical reaction between sulfide and ferrihydrite. Samples were collected from cultures for sulfate measurement at the same time intervals as those in phases I and II.

To test whether the microbial process can outperform the chemical process in transforming sulfide with ferrihydrite, the incubation (5) was repeated using ca. 50 µM sulfide instead of 1 mM. In this experiment, a small amount of sulfide was spiked three times at 1.5-h intervals into ferrihydrite-amended cultures, abiotic controls, and sulfide-only biotic controls. After each spike, subcultures (~ 0.3 ml) were collected at 2, 5, 10, 15, 20, and 25 min for dissolved sulfide measurements. Two biological replicates were performed for each treatment. To verify the reproducibility of the observed sulfide consumption pattern, incubations were conducted using inocula at different cell densities (OD₆₀₀ of 0.042, 0.075, and 0.086). To quantify the transformation of spiked sulfide during the incubation, independent cultures were set up using an inoculum with an OD₆₀₀ of 0.072 and supplied with eight spikes of sulfide. Ferrihydrite-amended cultures, abiotic controls, and sulfide-only controls received ca. 50 µM sulfide at 1.5-h intervals, whereas ferrihydrite-only biotic controls were spiked with anoxic water. Three replicate incubations were performed for each treatment. Subsamples were taken every three hours for concentration measurement of S(0), total Fe(ii), sulfate and Cline-extractable sulfide.

Chemical analysis of metabolites

To monitor the dynamics of metabolites in the incubation experiments, subsamples of the culture were taken periodically with sterile syringes flushed with pure N₂ as described above. HCl-extractable Fe(ii) was determined by adding 0.1 ml sample aliquots to 0.2 ml 0.75 N HCl. The sample was immediately centrifuged for 15 min at 12,044g. Fe(ii) in the resulting 0.5 N HCl was measured using the ferrozine assay. Previous studies have shown the 0.5 N HCl treatment allowed quantitative extraction of the solid phase Fe(ii) associated with the surface of iron oxides, Fe(ii) from FeS, and the dissolved Fe(ii) in the Fe/S system^7,94. Therefore, we referred to HCl-extractable Fe(ii) as total Fe(ii).

Aqueous and total sulfide were determined using spectrophotometric methods. To measure dissolved sulfide, approximately 0.3 ml subculture was filtered through a 0.2 µm membrane (CHROMAFIL). The dissolved sulfide in the filtrate (0.1 ml) was fixed by 0.25 ml 3% w/v zinc acetate dihydrate (Sigma Aldrich), followed by quantification using the Cline method⁹⁵. The filtered sample from the incubation with ferrihydrite and 1 mM dissolved sulfide showed black colour, indicating the formation of FeS particles smaller than 0.2 µm. The sulfide associated with this FeS fractionation was approximated as HCl-extractable Fe(ii), assuming a 1:1 stoichiometry. The total sulfide was determined as Cline-extractable sulfide. The Cline reagent contains 6 N HCl that dissolves some solid sulfides (for example, freshly formed FeS), and thus the Cline-extractable sulfide comprises dissolved sulfide and HCl-reactive solid phase sulfide. Total sulfide in the Fe/S system is typically determined as acid volatile sulfide. Acid volatile sulfide was not analysed here owing to the large uncertainties inherent to this methodology^91,96.

Sulfate and formate concentrations in the incubations were determined by capillary electrophoresis techniques. Sample preparation for sulfate measurement involved fixation of 100 µl subsample with 10 µl 3% w/v zinc acetate, dilution with 890 ul MilliQ water, filtration through a 0.2 µm membrane, and addition of 1 mM chlorate as the internal standard. The standards were prepared by adding defined amounts of sulfate (Sigma Aldrich) to the alkaline medium, followed by the same treatment procedure as described for samples. The sulfate content in the prepared samples/standards was measured using an Agilent 7100 capillary electrophoresis system (Agilent Technologies), equipped with a capillary (72 cm × 72 µm internal diameter; Agilent Technologies) and a diode array UV-vis detector (DAD). Electrolytes for anion separation contains 2.25 mM pyromellitic acid (Sigma Aldrich), 1.6 mM triethanolamine (Sigma Aldrich), 0.75 mM hexamethonium hydroxide (Sigma Aldrich), and 6.5 mM NaOH⁹⁷ at pH 7.8 ± 0.1. Anion separation was implemented at a voltage of −30 kV. The data were acquired through indirect UV detection at a wavelength of 350 nm with a bandwidth of 60 nm, and a reference wavelength of 245 nm with a bandwidth of 10 nm. For the formate measurement, 900 µl MilliQ water was added to 100 µl samples/standards (Sigma Aldrich), which were then filtered through a 0.2 µm membrane. l-malate (Sigma Aldrich) was added to the filtrate as the internal standard. Organic Acids Buffer for capillary electrophoresis (pH 5.6; Agilent Technologies) was used as electrolytes, and the separation conditions, including DAD and capillary electrophoresis settings, were configured according to manufacturer’s instructions. All electropherogram data were analysed with the Agilent ChemStation.

Elemental sulfur was measured using high performance liquid chromatography (HPLC). One-hundred microlitres of sample was fixed with 10 µl of 3% w/v zinc acetate. Then, 300 µl chloroform was added, and the mixture was shaken at 500 rpm for 1 h. The elemental sulfur in chloroform phase was then measured using a Dionex UltiMate 3000 UPLC system, equipped with an UltiMate 3000 pump (0.2 ml min⁻¹), a column Compartment (25 °C), a column Waters ACCQ-TAG ULTRA C18 1.7 µm × 2.1 × 100 mm, and an UltiMate 3000 Variable Wavelength Detector (UV) (wavelength 254 nm). The isocratic elution with 100% methanol was applied. With these adjustments, the peak appeared after 3.4 min. Data were analysed with Dionex Chromeleon software.

Microscopy of D. alkaliphilus incubated with ferrihydrite and sulfide

For scanning electron microscopy (SEM), transmission electron microscopy (TEM) and fluorescence microscopy, cultures incubated with ferrihydrite and sulfide (daily spike of 1 mM) for 5 days were fixed in 2% glutaraldehyde or 2.3% formaldehyde, respectively. For SEM imaging, solid iron phase iron was allowed to settle without centrifugation, carefully washed with MilliQ water, and transferred to 100% ethanol. Samples were then dried using rapid chemical drying with hexamethyldisilazane and mounted on aluminium stubs with double-sided sticky carbon tape and sputtered with Gold (JEOL JFC-2300HR). The images were taken with a Scanning Electron Microscope (JEOL IT 300 LAB6EOL) with Secondary Electron Detector (SED) and Backscattered Electron Detector (BED-C) at 20 kV.

For TEM imaging, cultures were treated with a solution containing 50 g l⁻¹ sodium dithionite, 0.2 M sodium citrate and 0.35 M acetic acid (hereafter termed dithionite solution) as previously described⁹⁸. After dissolution of solid iron phase, cells were pelleted at low speed (2,300g) to minimize shear forces and washed with MilliQ water before suspending cells in MilliQ water. For negative staining, 4 µl of sample was incubated for 1 min on a formvar-filmed and carbon-coated grid (200 mesh, Cu) and excess liquid was removed with a filter paper. A drop of stain (2.5% gadolinium acetate) was applied and immediately removed. Samples were examined in a TEM EM 900 N (Zeiss) at 80 kV.

For fluorescence microscopy, the formaldehyde-fixed cultures were resuspended and a subsample was filtered onto a 0.2 µm pore size polycarbonate membrane (Millipore). Cells on the filter were stained with a 1× SYBR Green solution, and images were acquired using a epifluorescence microscope (Zeiss Axio Imager M1 with an AxioCam MRm).

Monitoring the growth of D. alkaliphilus during incubation experiments

Growth was monitored by cell counting for cultures incubated under 4 different conditions: (1) ferrihydrite (approximately 62 mM Fe equivalent) and periodic spike of approximately 50 µM dissolved sulfide (sulfide was spiked 40 times over 5 days, with one spike every hour and 8 times per day); (2) ferrihydrite (approximately 62 mM Fe equivalent) and daily spike of 1 mM sulfide; (3) ferrihydrite (approximately 62 mM Fe equivalent) and periodic spike of FeS (approximately 1 mM Fe equivalent); and (4) nitrate (4 mM) and 2 spikes of sulfide at concentration of 1–2 mM. The setup of the cultures and controls was the same as described in ‘Cultivation of alkaliphilus DSM 19089’ except that a lower starting cell density (3–5 × 10⁷ cells per ml) was used. During each of the incubation experiments, subcultures (450 µl) were sampled periodically and preserved in 2.3% formaldehyde (final concentration). Before counting, 500 µl dithionite solution was added to 50–100 µl of fixed cells to dissolve the FeS and ferrihydrite particles. After dissolution of solid iron phase (within 10–15 min), 100 µl of each sample was diluted in 900 µl of 1× phosphate-buffered saline (PBS). The suspension was then sonicated using a SONOPULS ultrasonic homogenizer (Bandelin, Berlin, Germany) at 25% power with a cycle setting of 2 for a total of 30 s. Cells were subsequently stained with SYBR Green 1× (ThermoFisher) and incubated for 10 min at room temperature in the dark. Flow cytometric analysis was performed using a CytoFLEX S flow cytometer (Beckman Coulter) equipped with a blue 488 nm laser. SYBR Green fluorescence was detected using a 525/40 nm bandpass filter. A fluorescence threshold was applied on the SYBR Green signal to exclude background events. For each sample, 80–100 µl was measured. Data were gated on SYBR Green–positive cells displaying fluorescence shifts relative to unstained controls to identify the target population (Supplementary Fig. 15). Data were acquired and analysed with the CytExpert 2.6 software (Beckman Coulter). The specific growth rate (k; day⁻¹) was estimated via linear regression analysis of ln(Cell_t/Cell₀) versus time (day) over an apparent exponential growth phase. Here, Cell_t is the cell concentration (in cells per ml) at sampling time t (day).

¹³C-bicarbonate labelling experiments and isotope analysis

To probe for autotrophic carbon fixation during MISO growth conditions, ¹³C-labelled bicarbonate (98 atom% ¹³C; Sigma Aldrich) was added to ferrihydrite-incubated cultures receiving dissolved sulfide (1 mM) or FeS (ca. 1 mM S equivalent), to reach a 10 atom% ¹³C in the inorganic carbon pool. The dissolved sulfide or solid phase FeS were spiked in the same frequency as for the growth experiment. Abiotic controls for each culture were set up using autoclaved inoculum. To detect ¹³C content in bulk biomass and in single cells, subcultures were sampled, fixed by formaldehyde (2.3% final concentration), and analysed using elemental analyser-isotope ratio mass spectrometry (EA-IRMS) and NanoSIMS. For EA-IRMS, 1.5 ml of fixed samples that included ferrihydrite and cells were pelleted by centrifugation and washed with MilliQ water, followed by overnight treatment by 0.1 M HCl to remove residual carbonates. The dried cells attached to ferrihydrite particles were weighed (4–6 mg) and transferred to tin cups. Bulk cell carbon isotope ratios (¹³C:¹²C) were measured by EA-IRMS (Delta V Advantage) coupled by a ConFlo IV interface to an elemental analyser (EA-Isolink, all Thermo Finnigan). Sample ¹³C contents were calculated as atomic percentage of ¹³C in total carbon, following ¹³C atom% = ¹³C/(¹³C + ¹²C) × 100%. The analytical precision of replicate analyses of isotopically homogeneous international standards was ±0.0001% for ¹³C atom% measurements.

For NanoSIMS analysis, 0.1 ml formaldehyde-fixed samples that included ferrihydrite and cells were mixed with dithionite solution as described above and incubated for 2 h. After complete dissolution of ferrihydrite, 400 µl of the suspension was transferred onto gold-coated polycarbonate filters (GTTP type, 0.2 µm pore size, Millipore). The filters were gold-coated by physical vapour deposition, utilizing an Agar B7340 sputter coater (Agar Scientific) equipped with an Agar B7348 film thickness monitor (Agar Scientific) for precise adjustment of the coating thickness (150 nm). The filters were incubated for 2 h in 0.1 M HCl to remove residual carbonates and then washed twice in MilliQ water and then air-dried. Filter sections were attached to antimony-doped silicon wafers (7.1 ×7.1 mm, Active Business Company) with a commercially available glue (SuperGlue Loctide).

NanoSIMS measurements were carried out on a NanoSIMS 50 L instrument (Cameca) at the Large-Instrument Facility for Environmental and Isotope Mass Spectrometry at the University of Vienna. Prior to data acquisition, analysis areas were pre-conditioned in situ by rastering a high-intensity, slightly defocused Cs⁺ ion beam for removal of surface adsorbates and establishment of the steady state secondary ion signal intensity regime with minimum sample erosion. For this purpose, the following sequence of high and extreme low Cs⁺ ion impact energy (EXLIE) was applied: high energy (16 keV) at 100 pA beam current to a fluence of 5 × 10¹⁴ ions cm⁻²; EXLIE (50 eV) at 400 pA beam current to a fluence of 5 × 10¹⁶ ions cm⁻²; high energy to an additional fluence of 2.5 × 10¹⁴ ions cm⁻². Data were acquired as multilayer image stacks by scanning of a finely focused Cs⁺ primary ion beam with 2 pA beam current at approximately 80 nm physical resolution (probe size) over areas between 60 × 60 and 62 × 62 µm² with 512 × 512 pixel and 1,024 × 1,024 pixel image resolution and a per-pixel dwell time of 5 ms and 1.5 ms, respectively. The detectors of the multicollection assembly were positioned for parallel detection of ¹²C₂⁻, ¹²C¹³C⁻, ¹²C¹⁴N⁻, ³¹P⁻ and ³²S-secondary ions. Secondary electrons were detected simultaneously for gaining information about the sample morphology and topography. The mass spectrometer was tuned to achieve a mass resolving power ((MRP) = M/ΔM) of >10,000 for detection of C₂⁻ secondary ions.

Measurement data were processed using the WinImage software package provided by Cameca (WinImage V4.8) and the OpenMIMS plugin in the image processing package ImageJ (V1.54p). Prior to data evaluation, images were corrected for detector dead-time and positional variations emerging from primary ion beam and/or sample stage drift. Carbon isotope composition images displaying the ¹³C/(¹²C + ¹³C) isotope fraction, given in atom percent (atom%), were inferred from the C₂⁻ secondary ion signal intensity distribution images via per-pixel calculation of ¹³C¹²C⁻/(2 × ¹²C₂⁻ + ¹²C¹³C⁻) intensity ratios. For numerical data evaluation, regions of interest, referring to individual cells, were manually defined on the basis of the ¹²C¹⁴N⁻ and ³¹P⁻ secondary ion maps as indicators of biomass and verified by the topographical/morphological appearance in the secondary electron images. Biomass aggregates, in which an unambiguous identification of single cells was not feasible, were rejected.

Cells were assessed as being significantly enriched in ¹³C after incubation in the presence of ¹³C-bicarbonate if (1) the ¹³C isotope fraction value was higher than the mean plus 3 standard deviations (σ) of the values determined on the cells from the control (on day 0) and (2) the statistical counting error (3σ, Poisson) was smaller than the difference between the considered ¹³C enriched cell and the mean value measured on the cells from the control. The Poisson error was calculated from the secondary ion signal intensities (given in counts per region of interest) via

On the basis of these two criteria, all individual cells measured in the ¹³C incubated sample showed a significant enrichment in ¹³C.

RNA-seq and transcriptomics

D. alkaliphilus cultures grown under five incubation conditions (as described in ‘Cultivation of D. alkaliphilus DSM 19089’), each in four replicates, were used for comparative transcriptomic analysis. Cultures (30 ml) showing metabolic activity (for example, Fe(ii) production, sulfide consumption or production) were collected in the middle to late stage of incubation experiments. Cells were collected by centrifuging (12,857g; room temperature) under anoxic condition using oak ridge tubes (Thermo Fisher Nalgene) with replacement O-rings for sealing cap (Thermo Fisher Nalgene). The cell pellets were resuspended with 1.5 ml supernatant, and distributed to three lysis matrix E tubes (MP Biomedicals), each with approximately 0.5 ml. The collected cells were immediately frozen with liquid N₂, and stored at −80 °C before subsequent analysis. The total nucleic acids were extracted following a phenol-chloroform protocol as described previously^99,100. In brief, the sample was lysed for 30 s at a speed of 5.5 m s⁻¹, after mixing with hexadecyltrimethylammonium bromide extraction buffer and phenol-chloroform-isoamyl alcohol (25:24:1) (pH 8.0). The aqueous phase was extracted by centrifugation, and the phenol within was removed by mixing with chloroform-isoamyl alcohol (24:1). The total nucleic acids in the aqueous phase were then precipitated with polyethylene glycol 6000, followed by centrifugation. The pelleted nucleic acids were washed with ice-cold ethanol and dried before resuspension in diethyl pyrocarbonate-treated water. DNA from the total nucleic acids were removed using the TURBO DNA-free kit (Thermo Fisher Scientific).

RNA-sequencing was performed at the Joint Microbiome Facility of the Medical University of Vienna and the University of Vienna (JMF) under project IDs JMF-2311-14 and JMF-2405-05. Sequencing libraries were prepared from rRNA depleted (Ribo-Zero Plus rRNA Depletion Kit, Illumina) RNA samples (NEBNext Ultra II Directional RNA Library Prep Kit for Illumina, New England Biolabs) and sequenced in 2× 100 bp paired-end mode (NextSeq 6000, Illumina), yielding 74.2–303.7 million raw reads per sample. Individual read libraries were quality checked using fastQC v0.12.1 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and quality statistics were merged using multiQC v1.21 (ref. ¹⁰¹). Adapters were trimmed and phiX contamination was removed using BBDuk (part of BBMap v39.06). Reads were k-trimmed from the right with a kmer of 21, minimum kmer of 11 and hamming distance of two along with the tpe and tbo options. Quality trimming was performed from the right with a q-score of 28 to a minimum of 50 bases in length (https://sourceforge.net/projects/bbmap/). The quality filtered reads were aligned to the reference genome of D. alkaliphilus (NC_014216.1) using BBMap with a mapping identity of 99% and with ambiguous reads assigned to the best location (that is, counted only once for duplicated genes). FeatureCounts (part of SubRead 2.0.6 (ref. ¹⁰²)) with reverse-stranded and –countReadPairs were used to generate counts tables with the resulting alignments based on gene call locations by prodigal v2.6.3 (ref. ¹⁰³). Counts tables were analysed using DESeq2 release 3.19 (ref. ¹⁰⁴) to calculate the RPKM and to determine statistical significance of differential transcription between treatment groups. All P values are adjusted for multiple comparisons using the Benjamini–Hochberg method¹⁰⁵.

Quantitative PCR with reverse transcription (RT–qPCR) was performed to verify the upregulated transcription for the MHC gene DA_402 under iron-reducing conditions. Primers DA_402_998F (5′-TTCCCAATCGGGGCGAATAC-3′) and DA_402_1081R (5′-TGGCCTCGGTATAGAGGGTC-3′) were used to target DA_402. Primers recA_79F (5′-TTCGGCAAAGGCTCCATCAT-3′) and recA_221R (5′-TCCGGCCCATATACCTCGAT-3′) were used to quantify the transcription level of the house-keeping gene recA (DA_1926) encoding the DNA recombination protein. Primers for both genes were newly designed using Primer-Blast¹⁰⁶. For RT–qPCR, DNA-free RNA was first reverse transcribed to cDNA using SuperScript III reverse transcriptase according to the manufacturer’s instructions. The absolute abundance of transcripts from DA_402 and recA were quantified by quantitative PCR using cDNA as a template. Purified PCR products of gene DA_402 and recA amplified from genomic DNA of D. alkaliphilus were used as quantitative PCR standards. The PCR reactions were prepared in triplicates and run at 95 °C for 3 min, followed by 40 cycles of 95 °C for 30 s, 60 °C for 30 s, and 72 °C for 45 s, on the Thermal Cycler with CFX96 Real-Time System (Bio-Rad). The RT–qPCR data were acquired and analysed using CFX Maestro software (Bio-Rad). The transcription level of DA_402 was compared between treatments after normalization with that of recA. The statistical significance of differential transcription between treatments were determined via Student’s t-test.

Structure prediction and phylogenetic analysis of multi-haem c-type cytochromes in D. alkaliphilus

D. alkaliphilus proteins with more than one haem-binding motifs (CX_nCH; n = 2 to 5) were considered MHCs¹⁰⁷. The haem-binding motifs in protein sequences were counted using regex expressions in the Python re package. The subcellular localization of all putative MHCs (n = 46) from D. alkaliphilus was predicted using PSORTb v3.0 (ref. ¹⁰⁸) and DeepLocPro v1.0 (ref. ⁸¹). Prediction from DeepLocPro was used for the proteins for which PSORTb returned ‘Unknown’. The transcription levels of MHCs were compared between different incubation experiments on the basis of RPKM values. The statistical significance of differential transcription was assessed as described in the ‘RNA-Seq and transcriptomics’ chapter. The most highly transcribed extracellular MHC (DA_402) during MISO was further selected for structure prediction and phylogenetic analysis. The structure of the DA_402 monomer and oligomer were predicted using AlphaFold2 v2.3.2 at the COSMIC² science gateway. The leading signal peptide, predicted using SignalP 5.0 (ref. ¹⁰⁹), was cleaved from the protein sequence before structure modelling. For comparison, the cryo-EM structure of OmcS from G. sulfurreducens was retrieved from the Protein Data Bank (PDB) database (6EF8). The protein sequence of DA_402 was aligned to OmcS using the T_coffee alignment tool¹¹⁰. The structure-structure similarity between DA_402 and OmcS was calculated using an online TM-align tool and DaliLite.v5 (ref. ^111,112,113). The haem-binding sites in DA_402 and OmcS were visualized using MacPyMOL v.1.7.4 (https://pymol.org). The haem was docked to the target haem-binding site in DA_402 using AutoDockTools v1.5.7 (ref. ¹¹⁴) and AutoDock Vina 1.1.2 (ref. ¹¹⁵). To conduct phylogenetic analysis of DA_402, homologues of DA_402 were retrieved from the KEGG database using Blastp with an E value of 0.01. The retrieved homologues were then de-replicated with CD-HIT at 70% identity, aligned with Muscle, trimmed with trimAl (–gt 0.1). The resulting sequence alignment was used to reconstruct the maximum-likelihood tree using RAxML v8.2.12. The clustering pattern and decoration of the tree were performed using iTOL v6 (ref. ¹¹⁶).

Environmental distribution of Desulfurivibrionaceae with genomic potential of MISO

The metabolic potential of members belonging to the Desulfurivibrionaceae family was analysed using publicly available genomes recovered from different environments. Metagenome-assembled genomes (MAGs) classified as Desulfurivibrionaceae were obtained from GTDB r214 (n = 121), NCBI (n = 9), JGI IMG (n = 68) and GMGC (n = 7). The environmental origins of these genomes were retrieved from the metadata in the respective databases (Supplementary Table 8). The taxonomy of collected genomes was assigned using GTDB-tk version 2.3.2 with database release 214 (ref. ¹¹⁷). The phylogenomic tree of Desulfurivibrionaceae was reconstructed from a concatenated alignment of 120 single-copy genes with FastTree v2.1.10 (ref. ⁶⁶). The protein-coding genes in the genomes were called using Prodigal v2.6.3 (ref. ¹⁰³), and the resulting proteomes were screened for proteins involved in dissimilatory sulfide oxidation (DsrAB) and iron oxides reduction (that is, OmcS, OmcZ, porin–cytochrome complex and OmcE). DsrAB was detected using HMMs and the phylogenetic framework established in this study, while proteins involved in dissimilatory iron reduction were identified with HMMs from FeGenie⁸⁶. Additional proteins likely involved in dissimilatory reduction of iron oxides—that is, extracellular MHC DA_402 and PilA—were retrieved from Desulfurivibrionaceae genomes by hmmsearch or BLASTP. Homologues of PilA were extracted by searching TIGR02532 HMM model against the Desulfurivibrionaceae genomes using hmmsearch with –cut_ga option. For DA_402, homologues were collected from the Desulfurivibrionaceae genomes using BLASTP with an E value of 1e-10, followed by prediction of the subcellular localization and counting of haem-binding sites. The extracellular homologues containing multi-haem-binding sites (n > 3) were then placed into a reference tree created through phylogenetic analyses of DA_402 (see above) with the RAxML evolutionary placement algorithm (EPA). The alignment for EPA was generated using MAFFT v7.407 with –add option. The homologues that were placed with accumulated probability over 0.95 to the OmcS-like clade were considered as functional orthologs of DA_402. For visualization purposes, Desulfurivibrionaceae genomes (n = 119) encoding both dissimilatory iron and sulfur metabolism were de-replicated on the basis of relative evolutionary divergence (RED). RED was calculated for each internal node of the Desulfurivibrionaceae phylogenomic tree following the procedure described previously¹¹⁸. The tree was then collapsed at a RED value of 0.9 and one representative was chosen randomly from the collapsed clades, yielding 53 representative members that were visualized in the tree.

Statistics and reproducibility

The physiological experiments showing the ability of ferrihydrite-incubated D. alkaliphilus to oxidize formate (Fig. 3a), FeS (Fig. 3b), 1 mM sulfide (Fig. 3c,d,g) or ~50 µM sulfide (Fig. 3h) were repeated independently at least three times, all yielding consistent results. The sulfide removal kinetic experiments at low sulfide concentration were replicated independently for two times, and all results are present in the Extended Data Fig. 1. The experiment showing the transformation of sulfide and ferrihydrite with periodic supply of ~50 µM sulfide was performed independently twice, and both yielded similar results. The growth experiments of ferrihydrite-incubated cells with periodic addition of 1 mM sulfide (Fig. 4a), FeS (Fig. 4b) or 50 µM sulfide (Fig. 4c), and the experiment with nitrate and sulfide (Fig. 4d) were conducted once with three biological replicates per treatment and control. The ¹³C-bicarbonate labelling experiment and bulk ¹³C analysis of cells incubated with ferrihydrite and either dissolved sulfide (Fig. 4e) or FeS (Fig. 4f) were independently repeated for two times, both yielding comparable outcomes. Samples for NanoSISM analysis were chosen randomly among biological replicates collected at day 0 and 5, and representative field views are present in Fig. 4g–i. Transcriptomic analysis of cells growing under five different conditions was conducted once, with four biological replicates per condition (Fig. 5 and Extended Data Fig. 4).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Mice

B6.Cg-Gt(ROSA)26Sor^{tm14(CAG-tdTomato)/Hze}/J (iTdTomato) (007914), C57BL/6J (CD45.2) (000664) and B6.SJL-Ptprc^a Pepc^b/BoyJ (CD45.1) (002014) mice were purchased from The Jackson Laboratory. Nestin–GFP mice⁴⁸ were bred in our facility. Cdh5-creER, Cdh2-creER, Cxcl12^fl/fl, Kitl^fl/fl, Tpo^−/− and Tg(Alb-Tpo) mice were provided by R. H. Adams, L. Li, T. Nagasawa, S. J. Morrison, F. J. de Sauvage and W. S. Alexander, respectively. Unless indicated otherwise, 8–10-week-old mice of both sexes were used for experiments. All these mice were backcrossed with C57BL/6J mice for more than ten generations and maintained in pathogen-free conditions under a 12 h–12 h light–dark cycle, at a temperature of 21 ± 1 °C and humidity of 40–70%, and were fed with autoclaved food and water. This study complied with all ethical regulations involving experiments with mice, and all experimental procedures performed on mice were approved by the Institutional Animal Care and Use Committee of Albert Einstein College of Medicine. No randomization or blinding was used to allocate experimental groups.

Femoral bone transplantation

Femurs with intact periosteum were isolated from 8–10-week-old donor mice and preserved on ice in PBS (21-040-CV, Corning) until they were implanted in recipient mice. For transplantation of a single femur, non-conditioned recipient mice that were age and sex-matched with donor mice were anaesthetized with ketamine and xylazine, and a small incision was made at their unilateral thoracic region. Subsequently, the preserved femur was implanted subcutaneously, and the wound was closed. For transplantation of six femurs, small incisions were made at the bilateral cervical, thoracic and pelvic regions of recipient mice, and then one femur was implanted in each area, followed by wound closure. A sham operation was performed by making small incisions on the same area of skin as the control bone transplantation group and closing them.

Parabiosis

Parabionts were generated by making an incision in the skin from the elbow to the knee of mice on opposite sides of each mouse. The elbows and knees were paired together by s.c. suturing. The skin was then matched from one mouse to the other, sutured together and secured with wound clips.

Splenectomy

After mice were anaesthetized with ketamine and xylazine, a longitudinal incision was made in the skin and peritoneum on the left dorsolateral side of the abdomen, caudal to the last rib. The splenic artery was ligated and the spleen was removed. The abdominal wall was then closed, and the skin was sutured. A sham operation was performed by exteriorizing the spleen and then reinserting it into the abdominal cavity.

In vivo treatment

For G-CSF treatment, G-CSF (NEUPOGEN/Filgrastim; 300 µg ml⁻¹, purchased from Jack D. Weiler Hospital of Albert Einstein College of Medicine) was injected s.c. at a dose of 125 μg kg⁻¹ twice a day (eight divided doses) beginning in the evening of the first day. When used in bone transplantation experiments, G-CSF was administered to all groups at 1 month after the femurs were implanted or a sham operation was performed unless otherwise indicated. When HSC mobilization was checked, blood was collected at 3 h or 7 days after the final morning dose. For induction of CreER-mediated recombination, 8–10-week-old Cdh5-creER;iTdTomato mice were injected intraperitoneally with 2 mg tamoxifen (T5648, Sigma-Aldrich) dissolved in corn oil (C8267, Sigma-Aldrich) for five consecutive days (10 mg in total per mouse). Then, 4 weeks after the injection, these mice were used as hosts, or their femurs were isolated for transplantation. In experiments examining the overlap of Cdh2⁺ cells and MSCs, 8–10-week-old Cdh2-creER;iTdTomato or Cdh2-creER;iTdTomato;Nestin-GFP mice were injected with tamoxifen and subjected to analyses 4 weeks after the injection. In experiments using Cdh2-creER;Cxcl12^fl/fl or Cxcl12^fl/fl mice as hosts, tamoxifen was administered at 2 months after the femurs were implanted or a sham operation was performed in these mice. In parabiosis experiments, each mouse of the parabionts was injected with 2 mg tamoxifen for five consecutive days (20 mg in total per parabiont) 3 weeks after the surgery. Then, 4 weeks after the injection, the parabionts were subjected to analyses. In experiments using Cdh2-creER;Kitl^fl/fl or Kitl^fl/fl mice as hosts, tamoxifen was administered to four to five-week-old mice before femurs were implanted or a sham operation was performed.

Whole-mount imaging of host femurs and femoral grafts

Antibodies used for immunofluorescence staining of femoral grafts and host femurs are CD31 (PECAM1) Alexa Fluor 647 (MEC13.3, 102516) and CD144 (VE-cadherin) Alexa Fluor 647 (BV13, 138006) from BioLegend. For all imaging experiments, these antibodies (5 μg each) were injected into mice through the retro-orbital plexus for the vasculature staining, and mice were euthanized 10 min after injection. Femoral grafts and host femurs were then isolated and fixed in 4% paraformaldehyde (PFA; 15710, Electron Microscopy Sciences) overnight at 4 °C. For cryopreservation, the bones were incubated sequentially in 10%, 20% and 30% sucrose/PBS at 4 °C for 1 h each, embedded and flash-frozen in SCEM embedding medium (C-EM002, SECTION-LAB) and stored at −80 °C. For whole-mount imaging, bones were placed at −20 °C overnight and shaved with a Cryostat (CM3050, Leica) until the BM cavity was fully exposed. The sections were carefully collected from the melting embedding medium, rinsed with PBS and post-fixed with 4% cold PFA for 10 min followed by permeabilization in 0.5% Triton X-100/PBS for 3 h at room temperature (20–25 °C) and incubation with 2 µg ml⁻¹ 4′,6-diamino-2-phenylindole (DAPI; D9542, Sigma-Aldrich) for 30 min. Images were acquired at room temperature using the Zeiss Axio examiner D1 microscope (Zeiss) with a confocal scanner unit (Yokogawa), and reconstructed in three dimensions with SlideBook 6 (Intelligent Imaging Innovations), Photoshop 26 (Adobe) and Fiji build of ImageJ 2 (National Institute of Health, NIH) software.

Cell preparation

For analyses of haematopoietic cells in host femurs and femoral grafts, BM cells in these bones were flushed and dissociated using a 1 ml syringe with PBS through a 21-gauge needle. For analyses of haematopoietic cells throughout the mouse body, BM cells in endogenous and grafted femurs, tibias, humeri and pelvis were collected by flushing and dissociating, and radii, skull, spine, sternum and ribs were minced into small pieces with scissors, crushed with a mortar and pestle and filtered through a 70 µm cell strainer. Splenic cells were obtained by gentle grinding with slide glasses and passing through a 70 µm cell strainer. Cells in the liver were obtained by gentle grinding with slide glasses followed by digestion at 37 °C for 30 min in 1 mg ml⁻¹ collagenase type IV (17104019, Gibco), 2 mg ml⁻¹ dispase (17105041, Gibco) and 50 μg ml⁻¹ DNase I (DN25, Sigma-Aldrich). Peripheral blood was collected by retro-orbital bleeding of mice anaesthetized with isoflurane and mixed with EDTA to prevent clotting. The data from the bones above, spleen, liver and blood (assumed to be 2 ml per animal) were summed to determine the total HSC numbers in the mouse body. For analyses of BM stromal cells, intact flushed BM plugs were digested at 37 °C for 30 min in 1 mg ml⁻¹ collagenase type IV, 2 mg ml⁻¹ dispase and 50 μg ml⁻¹ DNase I in Hank’s balanced salt solution with calcium and magnesium (21-023-CV, Gibco). These single-cell suspensions were then subjected to red blood cell lysis with ammonium chloride and washed in ice-cold PEB (PBS containing 0.5% BSA and 2 mM EDTA).

Flow cytometry analysis and cell sorting

Cells were surface-stained in PEB for 30–60 min at 4 °C. Antibodies used for flow cytometry analyses and sorting were as follows: anti-CD45 APC-eFluor 780 (30-F11, 47-0451-82), anti-TER-119 APC-eFluor 780 (TER-119, 45-5921-82), anti-CD31 PE-Cyanine7 (390, 25-0311-82), anti-CD51 biotin (RMV-7, 13-0512-85), anti-CD140a (PDGFRA) PE (APA5, 12-1401-81), anti-CD140a PE-Cyanine7 (APA5, 25-1401-81), anti-Ly6A/E (SCA-1) FITC (D7, 11-5981-82), anti-Ly6G/Ly6C (GR-1) FITC (RB6-8C5, 11-5931-85), anti-Ly6G/Ly6C APC-eFluor 780 (RB6-8C5, 47-5931-82), anti-CD11b PE (M1/70, 12-0112-83), anti-CD11b PE-Cyanine7 (M1/70, 25-0112-82), anti-CD11b APC-eFluor 780 (M1/70, 47-0112-82), anti-B220 APC-eFluor 780 (RA3-6B2, 47-0452-82), anti-CD3e APC-eFluor 780 (145-2C11, 47-0031-82), anti-CD48 PerCP-eFluor 710 (HM48-1, 46-0481-85), anti-CD48 PE-Cyanine7 (HM48-1, 25-0481-80), anti-CD41 PerCP-eFluor 710 (MWReg30, 46-0411-82), anti-CD34 eFluor 660 (RAM34, 50-0341-82, 1:50 dilution), anti-CD135 (FLT3) PerCP-eFluor 710 (A2F10, 46-1351-82), anti-CD115 APC (AFS98, 17-1152-82) and anti-CD45.1 PE-Cyanine7 (A20, 25-0453-82) from eBioscience; anti-CD62E PE (10E9.6, 553751) from BD Biosciences; anti-KIT PE-Cyanine7 (2B8, 105814), anti-CD117 Brilliant Violet 421 (2B8, 105828), anti-CD150 PE (TC15-12F12.2, 115904), F4/80 PE (BM8, 123110) and anti-CD45.2 APC (104, 109814) from BioLegend; and anti-CD3e PerCP-Cyanine5.5 (145-2C11, 65-0031-U100) from Tonbo Biosciences. Streptavidin FITC (11-4317-87) and Streptavidin PerCP-eFluor 710 (46-4317-82) were purchased from eBioscience. Unless otherwise specified, all antibodies, Streptavidin FITC and Streptavidin PerCP-eFluor 710 were used at a 1:100 dilution. Flow cytometry analyses were carried out on the BD LSRII (BD Biosciences) system, and cell sorting experiments were performed using BD FACSAria (BD Biosciences). Dead cells and debris were excluded by forward scatter, side scatter and DAPI staining (1 µg ml⁻¹) profiles. Data were analysed using FACS Diva 6.1 (BD Biosciences) and FlowJo 10 software. Gating strategies are shown in Supplementary Fig. 1.

Cell cycle analysis

Single-cell suspensions were stained for cell surface markers, and subsequently fixed and permeabilized with BD Cytofix/Cytoperm solution (554714, BD Biosciences) according to the manufacturer’s instructions. The cells were then stained with DAPI (Sigma-Aldrich) at 5 μg ml⁻¹ and anti-Ki-67 PerCP eFluor 710 antibody (SolA15, 46-5698-80, eBioscience) or anti-Ki-67 eFluor 660 antibody (SolA15, 50-5698-82, eBioscience) at 1:100 dilution for 30 min at 4 °C. After washing, the cells were analysed on the BD LSRII (BD Biosciences) system. A DAPI^lowKi-67^low fraction was designated as the G0 phase of the cell cycle.

Blood cell analysis

Peripheral blood was diluted in PBS, and blood parameters were determined with the Advia120 Hematology System (Siemens).

Competitive BM and HSC transplantation

Competitive repopulation assays were performed using the CD45.1/CD45.2 congenic system. CD45.1 recipient mice were lethally irradiated (12 Gy, two split doses at least three hours apart) in a caesium mark 1 irradiator (JL Shepherd & Associates). For BM repopulation assays, 1 × 10⁶ CD45.2 donor-nucleated BM cells were transplanted into irradiated recipients together with 1 × 10⁶ CD45.1 BM cells. For HSC repopulating assays, 200 HSCs (CD45.2) were sorted from BM cells and transplanted into irradiated CD45.1 recipients together with CD45.1 competitor BM cells calculated to contain 200 HSCs (1:1 HSC ratio). For secondary BMT, 3 × 10⁶ BM cells from primary recipient mice were transplanted into newly irradiated (12 Gy) CD45.1 recipients. CD45.1/CD45.2 chimerism of the myeloid (CD11b⁺), B (B220⁺) and T (CD3ε⁺) lineages in recipient blood was analysed up to 5 months after BM or HSC transplantation using a flow cytometer, and that of BM cells was checked at 5 months after BM or HSC transplantation, at which the mice were euthanized.

Ex vivo HSC culture

Ex vivo HSC cultures were performed using F12-PVA-based culture conditions as previously described⁴⁰. In brief, HSCs were sorted into 96-well flat-bottom plates containing 200 µl HSC medium and expanded at 37 °C with 5% CO₂ for up to 28 days. Medium changes were made every 2–3 days. Cells were split at a 1:3 ratio into new plates when reaching 80–90% confluency. After expansion, the cells were used for non-conditioned transplantation.

Non-conditioned HSPC transplantation

HSCs were purified from CD45.2 mice and expanded, as described above. Expanded HSPCs (10⁶ LSK cells per recipient mouse) were then transferred into non-irradiated tamoxifen-administered Cdh2-creER;Kitl^fl/fl mice (backcrossed with CD45.2 mice for more than 10 generations) after the transplantation of one or six WT femurs, split into three doses over consecutive days.

Targeted limb irradiation

Animals were anaesthetized by isoflurane before irradiation using the Small Animal Radiation Research Platform, SARRP (XStrahl). The orthovoltage X-ray unit operates at 220 kVp and 13 mA. Before irradiation, a static X-ray scan was acquired using 50 kVp and 0.7 mA tube current with Al filtration. Mice were maintained in a circular lucite jig with whole-body lead shielding (to protect the individualized compartments from unwanted irradiation) and ports through which secured four limbs protruded and were irradiated to 20 Gy in a single fraction.

RNA extraction and RT–qPCR analysis

A total of 2 × 10³ MSCs or HSCs were sorted directly into lysis buffer and stored at −80 °C. mRNA was extracted using the Dynabeads mRNA DIRECT Purification Kit (61012, Invitrogen) according to the manufacturer’s protocols. Conventional reverse transcription (RT) with random hexanucleotide primers was then performed using the RNA to cDNA EcoDry Premix (639549, TaKaRa) in accordance with the manufacturer’s instructions. Quantitative PCR (qPCR) was performed in 384-well plates with FastStart Universal SYBR Green Master Mix (04913914001, Roche) on the QuantStudio 6 Flex Real-Time PCR System v.1.7.2 (Applied Biosystems). The PCR protocol started with one cycle at 95 °C (10 min) and continued with 40 cycles at 95 °C (15 s) and 60 °C (1 min). All mRNA abundance was calculated relative to the corresponding amount of Actb (encoding β-actin) using the ΔC_t method. A list of the primer sequences is provided in Supplementary Table 1.

ELISA

For analysis of BMEF, the BM of one femur or pelvis was flushed out using 1 ml of PBS, and the cells were subsequently pelleted by centrifugation. The resulting supernatant was transferred to another tube and stored at −80 °C until analysis. For analysis of serum, blood was allowed to clot at room temperature, and serum was separated by centrifugation and stored at −80 °C until analysis. Cytokine levels in BMEF and serum were then measured using mouse IL-1β (BMS6002), IL-6 (KMC0061) ELISA kits (Thermo Fisher Scientific) and TNF (MTA00B-1), CXCL12/SDF-1α (MCX120), SCF (MCK00) and TPO (MTP00) Quantikine ELISA kits (R&D Systems) according to the manufacturer’s protocols.

Statistics and reproducibility

All data are presented as mean ± s.e.m. n represents the number of mice in each experiment, as detailed in the figure legends, and experiments presented were successfully reproduced in at least three biological replicates. No statistical method was used to predetermine sample sizes, and sample sizes were determined by previous experience with similar models of haematopoiesis, as shown in previous experiments performed in our laboratory^{13,14,16,19,20,47}. Sample exclusion was only done as a result of premature mouse death. Statistical significance was determined by an unpaired, two-tailed Student’s t-test to compare two groups or a one-way ANOVA with Tukey’s multiple-comparison tests for multiple group comparisons. Data presentation and statistical analyses were performed using Prism 10 (GraphPad), Excel 16 (Microsoft), SlideBook 6 (Intelligent Imaging Innovations), Photoshop 26 (Adobe) and FlowJo 10 software.

The data in Fig. 2j,k were obtained in the same experiments, and data from the sham-operated mice were reused in each of these figure panels. The data in Extended Data Fig. 7b,c were obtained in the same experiments, and data from the sham-operated mice were reused in each of these figure panels. The data in Extended Data Fig. 8i,j were obtained in the same experiments, and data from the sham-operated Cxcl12^fl/fl and Cdh2-creER;Cxcl12^fl/fl mice were reused in each of these figure panels. The data in Fig. 3h,i were obtained in the same experiments, and data from the sham-operated Cxcl12^fl/fl and Cdh2-creER;Cxcl12^fl/fl mice were reused in each of these figure panels. The data in Fig. 4c,d were obtained in the same experiments, and data from the sham-operated Kitl^fl/fl and Cdh2-creER;Kitl^fl/fl mice were reused in each of these figure panels. The data in Fig. 5c,d were obtained in the same experiments, and data from the sham-operated Tpo^+/+, Tpo^+/− and Tpo^−/− mice were reused in each of these figure panels. The data in Fig. 5h,i were obtained in the same experiments, and data from the sham-operated WT and Tpo-Tg mice were reused in each of these figure panels.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.