Minerals function as signals to modulate microbial adaptation

Microbial adaptation and evolution occur rapidly in response to changing environments. Minerals, as fundamental components of terrestrial ecosystems, play a pivotal role in shaping these environments and structuring microbial communities. The impact of minerals is often attributed to their compositions, providing essential nutrients and energy sources (in the form of electron donors or acceptors), along with physical structures that shelter microbial growth^1,2,3. This assertion is supported by a substantial body of evidence showing that different minerals selectively enrich distinct bacterial communities over time^4,5, a phenomenon attributed to the provision of essential nutrients by certain minerals for specific microorganisms, and the creation of microenvironments favoring the growth of particular taxa. Examples of this selective influence include the strong correlation between apatite dissolution and the abundance of Burhkolderia colonizing the mineral’s surfaces⁶, the preferential association of Glomeromycota with ferrihydrite⁵, and instances where specific chemolithotrophic bacteria utilize redox-active minerals as electron donors or acceptors, depending on their metabolic requirements⁷. At species level, the mineral effect may modify interactions between organisms and alter the evolutionary trajectories of the interacting species. This, in turn, contributes to changes in assembly, composition, and functioning at the community level⁸.

However, in the absence of specific environmental stressors, such as nutrient/energy deficiencies or harmful radiation, the roles of minerals in sculpting microbial ecosystems are largely enigmatic and underappreciated. This knowledge gap may be partially rooted in our inability to decipher the effect of non-nutrient and energy-less minerals on microbial development. Literature abounds with examples unveiling that minerals without nutrient and energy values can exert a subtle but powerful impact on microbial behavior. Dating as far back as the 1940s, Zobell⁹ noted that inert solids such as chalk, sand, kaolin, and charcoal had long been employed in liquid media to promote the growth of bacteria. The author suggested that, based upon his own study showing the metabolic activity increased in bacterial suspension if glass beads were present, the inert surfaces benefit microbial growth by adsorbing nutrients. However, more recent studies^10,11 revealed that the organisms, once surface-bound, often adjust their cell physiology, signifying the effect may extend beyond nutrient enrichment. The most recent research discovered that phyllosilicate minerals like kaolinite and montmorillonite can inhibit quorum sensing of certain bacteria by adsorbing their communication signals¹². More interestingly, it was documented that certain minerals, such as montmorillonite, can either enhance or inhibit the horizontal transfer of antibiotic resistance genes (ARGs), depending on the mineral’s concentration¹³. These examples, on the one hand, showcase the diverse and nuanced capabilities of minerals to alter the growth and functions of microorganisms, and on the other hand, enforce the previous acknowledgement that the underlying principles regarding mineralogical effect on microbial activities remain far from clear¹⁰.

One way to expand our knowledge base regarding mineral effect on microorganisms is to investigate the functional changes within the context of evolving community compositions. Because a biological community’s functional potential does not exclusively correspond to its composition, this approach offers us a window, via cross-linking shifts in composition and functional profile, into the driving forces behind the effects of non-nutrient and non-energy minerals and the potential of geomaterials to moderate microbial evolution. Extending from the broad hypothesis that minerals with distinct geochemical properties exert selective pressures, favoring the emergence of specialized bacterial communities with unique metabolic capabilities, we propose that, at the gene expression level, the mineral effect may act as a regulatory mechanism to modulate the transcription of specific functional genes, thereby enabling bacteria to adapt to changing environmental conditions. We expect to observe these changes not just in the composition revealed by 16S rRNA gene sequencing, but also in the functional profiles constructed through metatranscriptome analysis to uncover genes differentially expressed between treatments.

We test our hypotheses and approach using a soil bacterial consortium in a nutrient-rich and aerobic environment. We treat the community with kaolinite (Al₂Si₂O₅(OH)₄), a nearly insoluble phyllosilicate with no particular nutritional or energetic benefit, and olivine ((Mg,Fe)₂SiO₄), a nesosilicate that readily releases ferrous ions through dissolution. By comparing the results, we aim to determine whether the effect of mineral-bound nutrients and energy, specifically Fe²⁺, outweighs that of the inert mineral kaolinite. Unlike conventional methodology that incubates mineral mesh bags/containers with soils, the present work employs a serial passage evolution experiment to mimic the natural selection processes that drive microbial adaptive evolution over multiple generations. Here, the soil-derived microbial consortium faces the influence of the tested minerals with minimal environmental stresses. By comparing the communities cultured in the absence and presence of the two minerals for composition, diversity, and functional repertoire, we aim to unearth the cryptic imprints minerals leave on microbial development. Findings from this work may help us gain new insights into the dynamic landscapes of our planet’s hidden ecosystems in soils, with potential implications for fields ranging from agriculture, bioremediation, and microbe-mineral coevolution.

Alteration of bacterial communities in the presence of minerals

Dissimilar mineral effect on the growth of microbial consortium was soon observed in the first passage of the serial passaging experiment. This was evidenced by samples collected at different stages showing varied growth dynamics in response to the treatment by kaolinite or olivine (Fig. S1). Specifically, the introduction of kaolinite led to a substantial reduction, relative to the control (CK), in cell density at various times along the growth curve. For instance, a 53% decrease was observed at 58 h (log phase) and a 47% decrease at 80 h (stationary phase). In contrast, olivine treatment only exhibited a minimal inhibitory effect, resulting in a 0.4% and 0.5% cell density decrease at the comparable stages.

Cells collected in the late logarithmic phase after 50 passages were analyzed for morphology and growth using scanning electron microscopy (SEM). In the CK run, cells of various shapes tightly clustered into aggregates (Fig. 1). In contrast, the mineral-treated cells, while remaining intact morphologically, dispersed among mineral grains, forming mineral-cell aggregates. Furthermore, 16S rRNA gene sequencing analyses revealed a substantial change in the diversity and composition of the mineral- treated consortium (Fig. 2). In specific, the presence of minerals unevenly decreased the Shannon diversity index with kaolinite causing a stronger reduction (~30%) than olivine (~<10%) (Fig. 2a) relative to that in CK. Principal coordinates analysis (PCoA) explained 99.22% of the variation within the first two axes (Fig. 2b), showing the samples formed three distinct clusters corresponding unequivocally (ANOSIM test significance P < 0.05) to CK and the two mineral treatments. Additionally, the total numbers of amplicon sequence variants (ASVs) increased from 128 in CK to 137 and 140 in olivine and kaolinite treatment although the ASVs unique to each experiment, 64, 62, and 65, respectively, stayed close (Fig. 2c). Furthermore, a total of 40 ASVs were shared across three groups, with more ASVs (23) shared between the two mineral treatments than between CK and either olivine (9) or kaolinite (15).

a CK, no mineral presence; b kaolinite treatment; c olivine treatment. Cells in mineral treatments were false-colored for visual identification.

Full size image

a Shannon diversity changes after subculturing with kaolinite and olivine supplements. b PCoA based on weighted UniFrac distance shows bacterial community clusters in three distinct groups (ANOSIM, R = 0.9, P = 0.001) corresponding to the mineral/rock supplement in the media. c Differences in the numbers of shared and unique bacterial ASVs among control and mineral treatments. d Bacterial community composition characterized to the genus level. There were four replicates for each treatment, and the taxa with low relative abundance (<1%) in all three treatments are clustered as others. e LEfSe taxonomic cladogram highlighting significantly discriminant taxa and genus-level biomarkers. Taxon nodes are colored, and branch areas are shaped based on the group with the highest representation of the taxon. Taxa with no significant differences among groups are shown in yellow. Logarithmic LDA score = 4.0.

Full size image

Further examination of the community composition (Fig. 2d) revealed that, at the genus level, the control group was dominated by Enterobacter, Acinetobacter, Pantoea, Pseudomonas, and Sphingobacterium, accounting for 86% of the total ASVs. Upon the introduction of minerals, however, only Enterobacter, Acinetobacter, and Pantoea (accumulative relative abundance of 87% and 71% in kaolinite and olivine treatment) maintained their dominance while the proportion of Pseudomonas and Sphingobacterium decreased by 6- and 205-fold in kaolinite and 2- and 126-fold in olivine treatment. Furthermore, the mineral presence significantly promoted the growth of Acinetobacter, Clostridium, and Lysinibacillus as the relative abundance of the three genera increased by 3- to 54-fold relative to that in CK. Linear discriminant analysis (LDA) effect size (LEfSe) confirmed that Clostridium and Lysinibacillus were enriched by olivine, whereas Enterobacter, Acinetobacter, Comamonas and Pantoea were enriched by kaolinite (Fig. 2e).

Additional investigation through metatranscriptomic analyses revealed that >97% of the transcripts were shared in all treatments including CK (Fig. S2), but the active members appeared to differ after mineral acclimation with several genera, such as Enterobacter, Acinetobacter, Serratia, and Enterococcus that increased their transcription activities while Ochrobactrum and Peptoniphilus became less active (Fig. 3a). It is worth noting that Streptomyces, a minor active member in control experiments, became a significant player in mineral-treated groups after increasing its activity by 3.54- and 2.42-fold in association with kaolinite and olivine, respectively (Fig. 3a).

Metatranscriptomic analysis revealing a changes in the relative abundance of active microbial communities at the genus level and b the numbers of differently expressed unigenes after serially passaged batch cultures with mineral addition. Genes with thresholds of P-value  <  0.05 and absolute log₂(FC) ≥ 1 were regarded as differently expressed unigenes.

Full size image

Differentially expressed genes and pathway enrichment analysis

Metatranscriptomic analyses of the samples collected after 50 passages (Fig. 3b) indicated alteration of gene expression in the bacterial community. In particular, 21,631 (78% of the total unigenes) and 20,892 (76%) genes in the kaolinite and olivine experiment, respectively, were differentially expressed (DEGs). Unigenes refer to a non-redundant set of unique sequences that represent distinct genes in the dataset. Among these, the presence of kaolinite led to 11,686 unigenes (42%) significantly upregulated, while 9945 (35%) notably downregulated (Fig. 3b). In the case of olivine, 15,421 (55%) upregulation and 5471 (20%) downregulation were detected.

Mapping these DEGs data to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database indicated that, at the broad functional category level (Level 1 in Fig. 4), both mineral treatments exhibited significant enrichment in metabolic pathways. At increasingly specific category levels (Levels 2 and 3), however, the enrichment appeared to be mineral-specific. For example, the energy metabolism (Level 2) influenced by the minerals was confined to carbon fixation pathways in prokaryotes (Level 3) for kaolinite, but an additional enrichment in sulfur metabolism pathway (Level 3) was observed for olivine. In addition, each mineral had six unique pathway enrichments, i.e., “Streptomycin biosynthesis,” “Polyketide sugar unit biosynthesis,” “Chloroalkane and chloroalkene degradation,” “Pyrimidine metabolism,” “Histidine metabolism,” and “Ascorbate and aldarate metabolism” for kaolinite, and “Selenocompound metabolism,” “Sulfur metabolism,” “Biosynthesis of siderophore group nonribosomal peptidesand,” “Glutathione metabolism,” “Glycerophospholipid metabolism,” and “Riboflavin metabolism” for olivine.

a Kaolinite treatment; b olivine treatment. The pathways listed are of the upregulated DEGs. The dot size represents the number of DEGs involved in each pathway, and the dot color indicates the P-value of each pathway calculated based on the hypergeometric distribution. The pathways with P-value  ≤  0.05 were significant enriched pathways. Rich factor is the ratio of DEGs numbers annotated in this pathway term to all gene numbers in the same pathway term. GIP genetic information processing, EIP environmental information processing, CP cellular processes.

Full size image

In addition to metabolism, kaolinite treatment enriched two genetic information processing (GIP, Level 1) pathways in “DNA replication” and “Homologous recombination” (Level 3). Olivine treatment, on the other hand, enriched one GIP pathway in “Mismatch repair” (Level 3), along with two cell motility-related cellular processes (CP, Levels 1 and 2) in “Flagellar assembly” and “Bacterial chemotaxis” (Level 3), and two environmental information processing pathways (EIP, Level 1) in “ABC transporters” and “Two-component system” (Level 3) for membrane transport and signal transduction (Level 2), respectively.

Mineral impact on the expression of streptomycin biosynthesis and antibiotic resistance genes

Streptomycin biosynthesis was one of the enriched pathways identified by KEGG in kaolinite treatment (Fig. 4). A detailed examination of the expression levels of the 13 genes involved (Fig. 5) suggests kaolinite promoted significant upregulation ranging from 1.6 to 5.7 on the log₂(FC) (i.e., fold change) scale for all except rfbD. The glk, pgm, and iolG genes displayed particularly strong upregulation, with log₂(FC) = 5.3, 5.7, and 5.7, respectively (Fig. 5). Surprisingly, most genes, with the exceptions of stsC, strB1, and rfbB, also experienced significant upregulation in olivine treatment, albeit to a lesser extent (log₂(FC) ranging from 1.1 to 4.4) in comparison to that in the kaolinite treatment.

The significantly upregulated genes (log₂-fold change higher than 1) are indicated in red, and insignificantly changed genes are indicated in gray. The specific values of log₂(FC) were shown in the squares. glk glucokinase, INO1 myo-inositol-1-phosphate synthase, suhB myo-inositol-1(or 4)-monophosphatase, iolG myo-inositol 2-dehydrogenase, stsC scyllo-inosose aminotransferase, stsE scyllo-inosamine 4-kinase, strB1 scyllo-inosamine-4-phosphate amidinotransferase 1, pgm phosphoglucomutase, rfbA glucose-1-phosphate thymidylyltransferase, rfbB dTDP-glucose 4,6-dehydratase, rfbC dTDP-4-dehydrorhamnose 3,5-epimerase, rfbD dTDP-4-dehydrorhamnose reductase.

Full size image

A further comparison with CK revealed that the elevated expression was not limited to the genes involved in streptomycin biosynthesis. Instead, this heightened expression extended to the total expression of antibiotic resistance genes (ARGs), which exhibited collective upregulation of 1.42- and 2.85-fold in response to kaolinite or olivine treatment, respectively (Fig. 6a). For individual genes, more than 80% of ARGs, including macB, carA, mexQ, mdtN, mdtL, otrC, and nmcR, were significantly upregulated under mineral influence (log₂(FC) > 1) (Fig. 6c). The most abundant ARGs were associated with multidrug and macrolides resistance, followed by those conferring resistance to tetracycline, beta-lactams, glycopeptides, peptides, aminoglycosides, fluoroquinolones, chloramphenicol, bacitracin, and streptogramin A (Fig. 6b). Although the enrichment of streptomycin biosynthesis pathway was identified specifically for kaolinite by KEGG (Fig. 5), olivine treatment actually induced more pronounced increase for the expression of most ARGs (Fig. 6b). As a comparison, the expression of macrolide and multidrug resistance genes was enhanced by 2.00- and 1.24-fold by kaolinite, but increased to 5.86- and 2.94-fold in the olivine experiments.

a Total expression level of ARGs in the consortium for each treatment group; b expression level of ARGs categorized by drug classes; c different expression of specific ARGs induced by mineral treatments depicted as log₂(FC) relative to the control.

Full size image

Mineral effect on the expression of organic matter degradation genes

Detailed examination of gene expressions related to carbohydrate-active enzymes (CAZymes) biosynthesis, specifically targeting soil organic matter (SOM) decomposition, revealed upregulation of glycoside hydrolases (GH) and auxiliary activities (AA) family genes (Fig. 7). Those families are responsible for degrading plant-derived (lignin, cellulose, hemicellulose), fungal-derived (chitin, glucan), and bacterial-derived (peptidoglycan) components of SOM. Key family members responsible for the degradation of lignin and cellulose (GH1, GH4, GH5, GH9, GH12, GH10, GH36, GH43, and AA1, AA2, AA4, AA6) all exhibited log₂(FC) of greater than 1 compared to the control. Similar observations were made for families responsible for degrading fungal- and bacterial-derived components. These families include chitinase (GH18), glucosaminidase (GH20), glucanase (GH17, GH81), lysozyme (GH24, GH108), and peptidoglycan lytic transglycosylase (GH102, GH103). Interestingly, again, with a few exceptions such as GH9 for GH81, olivine induced significantly higher log₂(FC) values than kaolinite did (Fig. 7).

The transcriptional expression levels were normalized to the expression level of the CK.

Full size image

In addition, specific families of carbohydrate-binding modules (CBMs, non-catalytic protein domains that enhance the degradation process by directing CAZymes toward specific SOM substrates) associated with cellulose (CBM3, CBM5, CBM16, CBM37) and xylan (CBM13) were also significantly upregulated by olivine treatment. A similar assertion can be made for kaolinite, restricted to CBM3, CBM16, and CBM37 only.

The role of mineral surface chemistry in microbial adaptation

Microbial community assembly is typically shaped by complex ecological stressors. However, in minimal synthetic media under controlled conditions, it was shown that microbial communities exhibit predictable convergence in coarse-grained taxonomic structure (e.g., family-level composition) and community-level functions, governed by nutrient availability and stabilized by metabolic cross-feeding interactions¹⁴. In nutrition-rich media like LB culture employed in this study, the adaptative response within the bacterial consortium is expected to be minimal. While mineral treatments introduce new interfaces in the culture medium, potentially leading to some community shifts^15,16, these changes are not anticipated to be mineral-specific under nutrient-replete conditions. Yet, the consortium treated with the two minerals not only diverged from the control group, but also exhibited a clear difference between the treatments (Fig. 2). One may argue that the minerals, through selective adsorption, may have removed surface-specific ASVs that would have otherwise appeared in the control treatment. This argument, however, cannot account for the occurrence of mineral-specific ASVs. The presence of ASVs unique to kaolinite or olivine in the communities (Fig. 2c) indicates that the evolutionary trajectory shift is not merely driven by the presence of grain surfaces but instead by the intrinsic characteristics of the minerals and their cell-mineral interfacial reactions.

The nature of this driving force is not immediately clear but may be explored from multiple angles, ranging from the minerals’ crystal structure, surface electric properties, to microbial extracellular electron transfer (EET). Kaolinite (Al₂Si₂O₅(OH)₄) is a phyllosilicate with alternating sheets of Si-O tetrahedra (T-layer) and Al-O octahedra (O-layer). Reactive surface sites (>Si-OH and >Al-OH, where > denotes surface-bound) are chiefly located at the edges of the TO stacking, as the basal planes usually carry little unsaturated chemical bonds (unless there are substantial isomorphic substitutions of Si or Al in the T- or O-layer). An approximate value of 4–7 for the ratio of edge to basal plane sites was reported¹⁷ with edge surfaces having a site density of ~10⁻⁷mol/m^{2 18}. Olivine (Mg,Fe)₂SiO₄, on the other hand, is a nesosilicate with isolated SiO₄ tetrahedra bound to each other by interstitial cations through ionic bonds. The lack of corner-O sharing among the SiO₄ tetrahedra indicates that the distribution of reactive sites (i.e., >Si-OH) on olivine crystals is likely not crystallographic direction-specific. Estimated site density of olivine is on the order of 10⁻⁵mol/m², significantly higher than that for kaolinite¹⁹. A more important difference between the two minerals is their surface electric properties. While kaolinite has a point-of-zero-charge pH (pH_PZC) of ~4.0²⁰ and therefore carries a small amount of negative surface charges at the experimental pH (~6.5), olivine with pH_PZC of 10²¹ is strongly positive-charged. This dichotomy may be crucial as cell surfaces or extracellular polymeric substances (EPS) usually carry a large amount of negatively charged functional groups (e.g., carboxyl, carbonyl, hydroxyl, and aldehyde) and hence exhibit a stronger affinity toward olivine. Microbial EET is another factor to consider, as certain members of the tested community may prefer to extract electrons from surface ferrous species instead of aqueous Fe²⁺ ions in the presence of olivine. This effect may be minor as transferring electrons from minerals may be less efficient compared to transferring electrons from aqueous ions. In totality, while the occurrence of mineral-specific ASVs (Fig. 2c) and organic matter degrading CAZymes (Fig. 7) requires further explanation, the preferential expressions of genes related to ARGs (Fig. 6) on positively charged, more reactive, and Fe-bearing olivine surfaces are unlikely to be coincidental.

Other physicochemical factors that can lead to bacterial adaptive response may include size difference between olivine and kaolinite grains, and the metal effect by impurity in olivine, particularly Al (0.14%), Cr (0.13 %), and Ni (0.24%) (Table S2). The sizes of mineral particles may potentially play a role in influencing cell-mineral interactions as the contact area between the two can vary by orders of magnitude for μm- vs. nm-sized grains. Since the particle sizes of the two minerals (mostly 2–5 μm in diameter, specific surface area ~10.17 m²/g for kaolinite and ~4.69 m²/g for olivine) in the present work are comparable, we don’t anticipate a strong SSA effect. Metal’s toxicity to microorganisms can be partially assessed by its concentration. Reported ‘threshold’ levels for these metals to exert visible biological effect on microorganisms are above 5 μM in general^22,23,24. Using the compiled literature rate of ~10⁻¹⁴mol/cm²/s for olivine dissolution at neutral pHs²⁵, we estimated a release rate of <0.1 μM/day for these metals in the present study, corresponding to a maximum of 0.7 μM accumulated concentration in our longest single experiment (about a week). Assuming the dissolved metals stayed completely solubilized in the culture, their levels should be nearly an order of magnitude lower than the reported ‘toxic’ concentrations, and hence unlikely to be significant for the current work.

Potential pathways of mineral enhancement to the expression of antibiotic synthesis/resistance genes and CAZyme genes

Microbial synthesis of antibiotics is tightly controlled by a complex regulatory network in response to environmental stresses of either abiotic (e.g., redox and pH) or biotic origin (e.g., nutrient competition, quorum sensing, and physical interaction)^26,27,28. For instance, iron competition with Myxococcus was found to activate the actinorhodin (a redox-active antibiotic) biosynthetic gene cluster in Streptomyces²⁹. In addition to conferring stress resistance, antibiotics also exhibit multifunctional capabilities, such as acting as signaling molecules to regulate the homeostasis of microbial communities³⁰, and as redox agents to solubilize phosphorus through reductive dissolution of iron oxides³¹. We cannot pinpoint the causes and mechanisms behind the heightened transcriptional activity for Streptomyces in the presence of minerals (Fig. 5), but speculate it may be related to physical stress such as weakened quorum sensing signal due to adsorption and the increase of cell-to-cell distances (Fig. 1) and/or chemical stress originated from mineral dissolution. Previous studies have shown that the presence of minerals in culture can reduce the efficacy of signaling molecules through physically blocking or adsorption^12,32. Relative to free-floating cells in liquid culture, bacteria grown on solid surfaces have limited mobility and hence rely more strongly on signaling molecules for interspecies communication. When the communication becomes obstructed due to signaling molecules losing activity, the organisms can use antibiotics to either reduce competitors responsible for ‘stealing’ signaling molecules or carry out communication tasks directly. Another possibility for minerals to stimulate bacterial synthesis of antibiotics is the occurrence of toxic surface species. Surface sites exposed on kaolinite are centered at either Al or Si. While Si does not have known capability to interfere with cellular activities, Al ions are shown to be toxic to microorganisms^33,34. Although kaolinite is commonly regarded as insoluble near neutral pH³⁵, research has shown that kaolinite-rich clays exhibit antibacterial activities, with aluminum playing a central role in their effectiveness^33,34, suggesting surface-bound Al may be perceived as a stress factor by microorganisms. It then follows that bacteria may release antibiotics to resist the toxicity or excrete specific substances to passivate the mineral surfaces. The less pronounced increase of gene expression for antibiotic synthesis in olivine treatment may be due to the absence of Al stress in the mineral when compared with kaolinite.

An interesting and surprising finding of the present study was the elevated expression of ARGs in the presence of either kaolinite or olivine (Fig. 5). Whereas an increased presence of streptomycin in the media may trigger the synthesis of antibiotic-resistant proteins or enzymes, it is unreasonable to attribute the upregulation of the total ARGs to a single antibiotic. This is particularly evident given that ARGs possess multifaceted functions besides protecting microbial cells from antibiotic attack. Taking the multidrug resistance (MDR) gene as an example, the ubiquitous MDR efflux pumps in microorganisms are involved in numerous processes such as detoxification of metabolic intermediates, metal homeostasis³⁶, augmentation of microbial virulence, and signal trafficking³⁷. Other known ecological roles of ARGs in bacterial physiology include maintaining the reproductive fitness and facilitating evolutionary adaptation³⁸. Collectively, we suspect the mechanisms underlying minerals’ stimulation of antibiotic resistance gene expression likely involve deeper levels of regulation.

One possible mechanism by which minerals may directly influence the spread of ARGs involves either the horizontal transfer of ARGs or the adsorption of extracellular ARGs onto mineral surfaces. The most recognized mechanisms of horizontal transfer of ARGs are conjugation, transformation, and transduction. Studies have shown that the presence of hematite (Fe₂O₃) and goethite (FeOOH) can either promote or inhibit the conjugative transfer of ARGs, depending on the content (5–100 mg/L vs. 1000–2000 mg/L) of the mineral³⁹. More detailed work suggested the threshold concentration of hematite for the promoting effect may be <25 mg/L, as experimental data indicated no effect when the mineral presence went beyond 50 mg/L⁴⁰. These studies theorized that the effect of hematite was derived from the iron-containing minerals’ ability to boost intracellular ROS levels and thereby increase cell membrane permeability. For phyllosilicates, it was observed that low concentrations (<0.1 g/L) of montmorillonite enhanced the transformation of antibiotic-resistant plasmids (pUC19) into competent Escherichia coli cells, but inhibited this process at higher concentrations¹³. Different from hematite, the inhibition effect was attributed to the adsorption of competence-stimulating factor (a signaling molecule of cell competence) and membrane damage³². The potential effect of adsorption was also found around microplastics, where the abundance of ARGs varied within the plastisphere⁴¹.

The current experiments were performed in liquid media with a mineral concentration of 10 g/L for both kaolinite and olivine. This level of mineral presence is much higher than that shown to promote the horizontal transfer of ARGs¹³. We therefore suspect that mechanisms other than horizontal gene transfer may be responsible for the observed upregulation of ARGs expression. Consider the other intriguing finding revealing the minerals’ capacity to stimulate the expression of CAZyme genes involved in SOM breakdown (Fig. 7). This observation may reflect a pre-existing perception within the microbial community of the complexity of organic matter and its coexistence with minerals, indicative of a form of memory responses. The term “cellular memory” or “history-dependent behavior” is used to describe the phenomenon by which microbial cells are able to adapt to a new environment or disturbance using prior experience^42,43. A direct implication of the memory response is that, when a microbe experiences a stimulus, the cell treats it as a signal for all upcoming events associated with the prompt and subsequently generates related anticipatory responses^44,45. That is to say, the non-genetic cellular memory can influence the behavior of progeny cells that have not directly experienced the initial condition. For example, exposing Escherichia coli to an immediate temperature increase from ambient (<30 °C) to 37 °C triggers a transcriptional response mirroring that of oxygen deprivation, even in the presence of maximal oxygen level, analogous to the transcriptional changes E. coli cells undergo upon entering the warm, oxygen-deprived mammalian gut⁴⁶. Similarly, the observation that E. coli exposed to lactose activates genes required for maltose utilization is thought to mimic the scenario of bacteria passing through the digestive tract, where lactose is typically encountered before maltose⁴⁷.

In soil ecosystems, microbes, SOM, and minerals are closely associated with each other to form a local ecosystem termed mineralosphere². It is conceivable that microbes derived from specific soils have memories of the coexisting SOM and minerals in the original ‘microcosm’ environment. Additionally, it is well known that the structure of organic matter in soil is highly complex. If microorganisms can, through cellular memory effects, prepare the enzymes required for organic matter degradation in advance, it would significantly enhance their environmental adaptability. For the current study, we theorize that the upregulated gene expression for CAZymes is the bacteria’s memory response to the mineral-SOM aggregates prevalent in the original environment from which the consortium was isolated. The stronger upregulation of CAZyme genes when treated with olivine than with kaolinite may be due to the ability of Fe-containing minerals protecting SOM⁴⁸. Simply put, the weathering of olivine may generate Fe-oxide coatings (e.g., goethite, ferrihydrite) that, in natural soils, are known to sequester SOM and limit enzymatic access. While our experimental system contained labile carbon sources (e.g., yeast extract), the transcriptional response could reflect an evolved microbial strategy to sense Fe-mineral surfaces as proxies for SOM occlusion in environmental settings. The rationale for this proposition is grounded not only in the low solubility of ferric (hydr)oxide minerals and the widely established association of SOM with Fe-minerals in soils, but also in its consistency with the iron metabolism in our experiments. Among the Fe-related genes (Table S1), a comparison between mineral treatments reveals that those related to siderophore biosynthesis/regulation and iron stress resistance experienced noticeable upregulation in the presence of olivine, whereas those related to iron transport and redox did not exhibit obvious discrimination toward either of the two minerals. For instance, of the nine genes associated with siderophore biosynthesis/regulation, five were not differentially expressed in kaolinite treatment. In contrast, all were upregulated (log₂(FC) ranged from 2.53 to 6.06) in the olivine experiments. Given the abundance of nutrients, including iron in the LB medium, it seems unlikely that siderophores are required to prevent iron loss (through ferric hydroxide precipitation). In fact, the elevated expression of Fe stress-resistant genes (KetE, ydbK, and rssxC) suggests an overabundance of iron in the environment. Furthermore, it is expected that the precipitation of ferric minerals may trap a certain amount of nutrient carbon compounds due to the formation of mineral-SOM aggregates. However, it is unclear why organisms would rely on these C molecules (hence synthesizing siderophores to liberate them) given the ample carbon resource in the medium. A plausible explanation may be a memory effect where the presence of olivine is perceived as a signal of impending carbon limitation, prompting a preemptive response.

We suspect the upregulated expression of genes associated with streptomycin biosynthesis and ARGs also represents memory-based responses. Soil microbes predominantly reside on mineral surfaces, where the organisms encounter SOM (plant- or microbe-derived) and various environmental stresses. Therefore, minerals, SOM, and environmental stressors frequently co-occur as stimuli in the cellular memory of soil microbial community. The microorganisms in the experimental consortium may perceive minerals as a signaling substance, and subsequently pre-activate genes related to SOM degradation, streptomycin biosynthesis (for coping with stress and competition), and antibiotic resistance (for combating antibiotics and enhancing environmental adaptation) altogether.

The non-genetic cellular memory in theory differs among microbial species, leading to memory responses being organism-specific. However, the metatranscriptomic technique employed in the present study probes the overall transcriptional level of the community, therefore providing information concerning the collective memory response of the consortium to minerals. Despite lacking signals from individual species, the possible occurrence of cellular memory shown by the bacterial community may have provided the first evidence that minerals can act as triggers to activate memory responses in soil microorganisms. This finding may expand our understanding of mineral-microbe interactions and the relevant implications for soil ecology and ecosystem functioning.

Potential of minerals as signaling molecules for gene regulation

Soil microbes are extremely versatile producers of natural bioactive chemicals, with Actinobacteria alone producing two-thirds of all known antibiotics used in the clinic today, along with a vast array of anticancer compounds, immunosuppressants, anthelmintics, herbicides, and antiviral compounds⁴⁹. However, we are still far away from realizing the full potential of natural product-producing microorganisms due to the lack of knowledge required to activate the expression of biosynthetic gene clusters (BGCs) in laboratory^49,50. For example, analysis of sequenced Streptomyces genome data revealed that ∼90% of BGCs are silent or cryptic under standard laboratory growth conditions⁵¹. Current practice of microbial culture in laboratory and industrial settings typically grows soil microbes in isolation with ample nutrients and resources (as the CK treatment of this study), largely ignoring the complexity and rapid environmental changes of the natural habitat. We suspect the absence of mineral triggers or signals ubiquitous in soils may be a possible reason why a good portion of BGCs remain poorly expressed or silent in artificial systems⁴⁹. Our findings that minerals, regardless of their nutrient and energy benefit, can trigger the upregulation of genes for SOM degradation and streptomycin biosynthesis may suggest a potential mechanism for broader expression of functional genes in microorganisms. Our results further highlight the multifaceted roles of minerals in natural ecosystems, extending far beyond their known functions as sources of nutrients or energy and physical habitats for microbial growth.

Experimental material description

Top soil (0–20 cm) of a corn field (43°49ʹ N, 125°23ʹ E, northeast China) were sampled and analyzed for its chemical properties: total organic carbon content 1.38%, available nitrogen content 140 mg/kg, available phosphorus content 36 mg/kg, available potassium content 127 mg/kg, and pH (H₂O) 8.37. After the removal of gravel and plant roots, fresh soil samples were sieved using a 10-mesh screen and then sent to laboratory for further study. The sampling region has a temperate continental monsoon climate with an annual mean temperature of 5.5 °C and an annual mean rainfall of 650 mm.

The powders of olivine and medical-grade kaolinite were acquired commercially (Langfang Mineral Separation Company (China) and Aladdin Chemical Reagent Company (China), respectively). The minerals were characterized for their elemental compositions (Table S2) and purity (Fig. S3) using inductively coupled plasma (ICP) after mixed acid digestion (HF-HCl-HNO₃-HClO₄) and X-ray diffraction, respectively. In addition, scanning electron microscopy (SEM) was used to analyze the particle size and morphology. Kaolinite typically presents a platy shape with an average length of less than 2 micrometers, while olivine occurs as irregular prismatic grains with an average length of 2–5 micrometers (Fig. 1). The specific surface area (SSA) of the minerals used in this study (kaolinite and olivine) was determined using the Brunauer–Emmett–Teller (BET) nitrogen adsorption method.

Luria-Bertani (LB) medium, a nutrient-rich broth commonly used for culturing bacteria, has the following ingredients: 10 g tryptone, 5 g yeast extract, 10 g NaCl, 1 L distilled water. The nutrient content of LB medium can vary depending on the manufacturer of the tryptone and yeast extract. To ensure the LB medium used in the present study is consistent with those used in most laboratories, we selected the classic OXOID brand. In addition, we added HEPES to buffer the pH during the experiment. The pH of the LB medium is adjusted to 6.5.

Preparation of soil-borne bacterial consortium

Soil-derived bacterial consortium was used as inoculum in the serial passage experiments. The consortium was obtained through a pre-enrichment process. Briefly, three grams of soil was suspended in 30 mL sterile distilled water and vigorously vortexed. An aliquot of 4 mL suspension was transferred to 150 mL Erlenmeyer flasks containing 40 mL of sterilized LB medium and incubated under continuous shaking with 170 rpm at 30 °C. In order to obtain a stable community composition in the consortium for serial passage experiment, an acclimation process was conducted: the subculturing step was repeated for another two cycles in the successive transfer of 0.5 mL aliquot into 50 mL sterile fresh LB medium.

Serial passage evolution experiment (50 transfers) in the presence of minerals

Serial passage experiments were conducted to assess the evolution of community composition and ecological functions of the soil consortium under the influence of different minerals. Three experimental groups were set up for the serially passaged batch cultures in liquid LB media, including the control treatment (CK, no mineral addition, replaced with an equal weight of water), kaolinite treatment (K, mineral content 10 g/L), and olivine treatment (O, 10 g/L). The liquid media for all three treatments were autoclaved before inoculation, and each treatment had four replicates. The serial passages were carried out for 50 iterations under the same culture conditions (continual shaking with 170 rpm at 30 °C). Consortial cells at the late log growth phase (approximately 65 h after inoculation) were inoculated into the corresponding fresh media at 1% inoculation ratio. Because organisms have varying growth rates, the fastest-growing populations and the late log growth phase of a microcosm are likely enriched with the dominant microorganisms. The cell growth was determined using the viable plate counting method and quantitative PCR (with bacteria-specific primers). Microbial consortia for each treatment were harvested at the end of the 50 passage cultures by centrifugation (10,000 rpm) for 10 min. A schematic diagram of the experimental design for the serial passage culture is shown in Fig. S4.

DNA extraction and 16S rRNA gene sequencing

DNA was extracted by using E.Z.N.A. Soil DNA kit (Omega, USA) following the manufacturer’s instructions. The extracted DNA concentration and quality were determined by the NanoDrop instrument and 1% agarose electrophoresis. The V3–V4 hypervariable region sequences were amplified by using the primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′). Samples were sequenced on an Illumina NovaSeq 6000 platform (2 × 250 bp) provided by LC-Bio Company (Hangzhou, China). Paired-end reads were merged using FLASH. Quality filtering on the raw reads was performed to obtain the high-quality clean tags according to the fqtrim (v0.94). Chimeric sequences were filtered using Vsearch software (v2.3.4). The DADA2 method of Quantitative Insights into Microbial Ecology2 (QIIME2) software (v2019.4) was used to optimize the original sequences⁵². 16S rRNA amplicon sequence variants (ASVs) were defined at 100% sequence homology using DADA2. After dereplication using DADA2, the final output, which is an ASVs table containing feature abundance and sequence, was used for downstream bioinformatic analysis. The bacterial taxonomy was assigned to ASVs using classifier of the SILVA reference database. Shannon index was calculated with QIIME2.

Beta diversity was assessed according to the differences in microbial community structures among groups using PCoA based on weighted UniFrac distance, and the differences among groups were analyzed by Analysis of similarities (ANOSIM). Significant taxonomic differences and biomarkers among control and kaolinite/olivine supplement treatments were identified using Linear discriminant analysis effect size (LEfSe) analysis with P = 0.05 by Kruskal–Wallis test, P = 0.05 by pairwise Wilcoxon test and logarithmic LDA score = 4.0.

RNA extraction and metatranscriptomic analysis

Total RNA was extracted by the Trizol Reagent kit (Invitrogen) following the manufacturer’s instructions. Total RNAs were treated with DNase (Promega) and Ribo-Zero rRNA depletion kit to remove DNA and ribosomal RNAs. The quantity and quality of the RNA were assessed using Qubit 2.0 Fluorometer (Life Technologies, USA) and Agilent 2100 bioanalyzer system (Agilent Technologies, USA), respectively. The high-quality mRNA was fragmented and subsequently constructed cDNA library. The paired-end sequencing was conducted on the Illumina HiSeq 4000 platform by LC-Bio Company (Hangzhou, China).

Raw data reads with adapter contamination and low-quality bases were removed by Cutadapt⁵³ and Fqtrim software (version 0.94). Quality control processing was conducted using FastQC software. The filtered reads were then mapped to the SILVA database to remove the ribosomal RNA reads. Metatranscriptome contigs for each sample were constructed by de novo assembling clear reads with Trinity⁵⁴. Each sample had an average of 7.56 Gb of raw data in the metatranscriptomic analysis. The Q20 value of all samples exceeded 98.3%, indicating the sequences had high accuracy. Metatranscriptome contigs of all samples were then integrated, and redundant ones were removed by CD-HIT-EST (identity threshold set to 0.95) to obtain unigenes⁵⁵. Cluster Database at high identity with tolerance was used to merge and remove redundancy with a similarity of 0.95 and lowest coverage of 0.9, and the longest sequence was used as the representative sequence of the Unigene to construct Unigene set.

For taxonomic annotation, all unigenes were compared with the NCBI Non-redundant protein (Nr) database using BlastP (BLAST version 2.25). Gene expression levels of unigenes were estimated by Transcripts per million (TPM) based on the number of aligned reads. Only unigenes with log₂ (fold change) > 1 and P-value  <  0.05 were accepted as differentially expressed unigenes (DEGs). To gain a deeper understanding of the biological functions associated with the DEGs identified from pairwise comparisons, we performed KEGG pathway enrichment analysis. This analysis involved mapping the DEGs to KEGG pathway identifiers and identifying pathways that were significantly enriched compared to the entire transcriptome background. The KEGG pathway enrichment analysis was conducted using the hypergeometric test, a statistical method that assesses whether a particular KEGG pathway is overrepresented among the DEGs compared to its expected representation in the background gene set. This analysis revealed a set of significantly enriched KEGG pathways, indicating the potential functional implications of the DEGs. These enriched pathways provide insights into the biological processes and molecular mechanisms underlying the observed differential gene expression patterns. The KEGG pathway enrichment analyses were performed with online tools of the LC-BIO company (https://www.omicstudio.cn/index).

To obtain comprehensive gene function information, the unigene sequences were aligned with the Comprehensive Antibiotic Resistance Database (CARD) and Carbohydrate-active enzymes (CAZyme) database using the sequence alignment annotation software blast2.25 (E-value = 0.0001). The corresponding annotations of antibiotic resistance and carbohydrate-active enzyme were assigned to the Unigenes. Then, the expression levels of the Unigenes corresponding to the antibiotic resistance functions were calculated in each sample to obtain the abundance of the antibiotic resistance function. The gene expression abundance of the CAZyme was then calculated by using the sum of the expression levels of the Unigenes corresponding to the CAZyme in each sample.

Morphological characterization

The cell samples were collected after 50 passages of cultivation and subsequently fixed using 2.5% (vol/vol) glutaraldehyde in phosphate-buffered saline, followed by dehydration using serial ethanol dilutions (20, 40, 60, 80, 95, and 100% [vol/vol] at a rate of 10 min per step). The dried samples were coated with platinum and imaged by scanning electron microscopy (SEM; Zeiss Sigma 500). To better identify microbial cells in mineral treatments, the cells in mineral treatments were marked with false colors using Adobe Photoshop software.

Reporting summary

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