Microbial assembly and metabolic restructuring following CO₂ injection in subsurface oil reservoir

Abstract

Carbon dioxide (CO₂) injection into oil reservoirs is a well-established and promising strategy for enhancing oil recovery while simultaneously sequestering carbon, whose effectiveness is increasingly recognized to be influenced by microbial-mediated mechanisms and biogeochemical processes. However, the impact of CO₂-injected and storage on functional microbial communities and their metabolic potential remains poorly understood. The study employed high-throughput sequencing to analyze the composition and diversity of microbial communities in produced water from three production wells with a long-term (> 20 years) history of CO₂ injection and two adjacent water-flooding ones in Xinli Oilfield, China. The results showed that CO₂-injected wells had significantly higher microbial diversity and distinct metabolic potential relative to the adjacent water-flooding wells. A metabolic difference in the archaeal communities towards hydrogenotrophic and heterotrophic metabolisms, alongside an enrichment of bacterial taxa associated with sulfur and nitrogen cycling was observed. Correlation analysis revealed that microbial differentiation was strong associated with geochemical alteration (especially pH and some inorganic ions), with NH₄⁺ and S²⁻ identified as key factors in this process. Metagenomic analysis based on high-quality metagenome-assembled genomes (MAGs) further demonstrated that CO₂ injection led to a different profile of microbial metabolic functions relative to the water-flooding condition, characterized by significantly enhancing hydrogenotrophic methanogenesis, dissimilatory sulfate reduction, and denitrification, while diminishing methylotrophic methanogenesis and Wood-Ljungdahl pathway activity. These findings provide new insights into the microbial mechanisms driving carbon transformation in CO₂-flooded oil reservoirs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13568-026-02066-w.

Introduction

The extensive combustion of fossil fuels, which powers rapid economic growth, has resulted in a marked surge in atmospheric CO₂ levels. Consequently, addressing these emissions has become an urgent global challenge (Xia et al. 2024; Ignacz et al. 2025). While crucial for reducing CO₂, renewable energy and efficiency measures face inherent constraints. At the same time, the global economy remains deeply dependent on fossil fuels. For these reasons, researchers now are increasingly advocating Carbon Capture, Utilization, and Storage (CCUS) as a foundational climate solution (Li et al. 2025; Wang et al. 2025b; Leonzio 2026). As a promising CCUS strategy, CO₂-enhanced oil recovery (CO₂-EOR) leverages the injection of CO₂ into oil reservoirs to increase crude oil production. This process enhances oil recovery efficiency through several key mechanisms: 1) reducing the viscosity of crude oil and inducing oil swelling, which acts to increase mobility; 2) lowering the oil–water interfacial tension and improving formation permeability, further boosting extraction. Therefore, it has become a widely recognized solution with the capacity to enhance oil recovery while storing carbon dioxide underground (Davoodi et al. 2024).

Generally, after being injected into subsurface reservoirs, part of the CO₂ is sequestered by mechanisms such as dissolution, diffusion, mineralization, and microbial transformation, with the non-sequestered portion being recaptured and recycled for repeated injection (Zhao et al. 2025a). Furthermore, oil reservoirs constitute unique extreme environments characterized by high temperature, high pressure, high salinity, and multiphase fluids. These conditions support distinctive microbial communities that form a significant part of the Earth’s deep biosphere (Zhou et al. 2024b). Simultaneously, CO₂ injection stimulates subsurface microbial communities to mediate in-situ methanogenesis and bio-mineralization, thereby directly influencing the fate of the injected CO₂ (Mu and Nazina 2022; Tyne et al. 2023). This interdependence makes it necessary to conduct a comprehensive analysis of the microbial communities and geochemical parameters induced by long-term CO₂-EOR (Yin et al. 2023).

Understanding the impact of CO₂ injection on the microbial community structure and diversity in oil reservoirs is the key and foundation for the application of CCUS (Tyne et al. 2021; Xiong et al. 2023; Zhou et al. 2024a). Previous research has shown that microbial persistence in CO₂-enriched subsurface environments, including taxa tolerant to prolonged CO₂-derived stress, relies on physiological and ecological flexibility. Morozova et al. reported that CO₂ exposure induced a transient inhibitory phase for subsurface microbial communities, marked by an initial 100-fold decline in sulfate-reducing bacteria (SRB) alongside proliferating methanogens, with SRB abundance recovering to pre-injection levels within five months (Morozova et al. 2010). This indicated the adaptive resilience of these microbial communities following perturbation. Furthermore, multiple studies indicated that CO₂ injection initially reduced the abundance of indigenous microbial communities, subsequently restructured population dynamics, and ultimately selected for assemblages adapted to the altered environmental conditions (Liu et al. 2015; Gwosdz et al. 2016; Stemple et al. 2022). By comparing microbial communities in CO₂-injected and control reservoirs via functional gene analysis, Liu et al. observed that long-term CO₂ exposure led to increased or stable abundance of key functional taxa, such as methanogens (Methanoculleus and Methanolinea), acetogens (Acetobacterium), sulfur-oxidizing bacteria (Halothiobacillus), and hydrogen-producers (Clostridium and Thermotoga) (Liu et al. 2015). These findings suggested that the indigenous microbiota metabolically adapted to the CO₂-enriched environment. Stemple et al. identified the dominant genera (including Desulfotomaculum, Sulfurospirillum, Halanaerobium, Acetobacterium and Pseudomonas) in the produced water from a Michigan Basin CO₂-EOR site. These taxa correspond to functional groups known for sulfide/acid production, biomineralization, and corrosion-related biofilm formation. Collectively, these indicated a subsurface condition selection for functionally resilient and CO₂-tolerant microbial consortia (Stemple et al. 2022).

Although prior investigations have examined the impacts of CO₂ injection on reservoir microbial communities and elucidated their selective adaptation to geochemical condition changes, the understanding of the long-term CO₂-injected effects remains limited. To achieve these goals, this study provides new insights into the subsurface microbial communities of long-term CO₂-injected reservoirs. Furthermore, by combining high-throughput sequencing with metagenomics and functional analysis, it offers new perspectives on changes in microbial composition and metabolic functional potential in extreme environments.

In the present work, a comparative analysis of the composition and metabolic potentials of the microbial communities in produced water samples from long-term (> 20 years) CO₂-injected oil reservoirs and adjacent water-flooding ones was performed. High-throughput sequencing of the bacterial and archaeal communities with concurrent physicochemical and metagenomic analyses was employed. The objective of this work is to: (1) resolve compositional differences in the microbial communities between CO₂-injected oil reservoirs and adjacent water-flooding ones, (2) analyze the environmental drivers of these variations and (3) elucidate concomitant adaptations in metabolic functional potentials. This work provides a deep understanding of microbial community compositions, environmental drivers, and potential metabolic pathways in oil reservoir environments after CO₂ injection.

Materials and methods

Sampling sites and sample collection

The sampling site is located in the Xinli Oilfield (Songyuan, Jilin province, China). The production wells investigated are located in Xinli Anticline, which lies in the westernmost part of the central depression in the Songliao Basin (Fig. 1a). The target reservoir intervals of these wells are between 1,520 and 1,545 m in depth, with an average recorded bottom temperature of 66 ℃. Produced water samples were collected from five production wells in the same geological block. The samples from five sampling wells were divided into two groups: 1) C56—6, C52—6 and C54—8, from CO₂-injected production wells with a 20-year flooded history; 2) N50—08 and N50—10, from long-term water-flooded production wells in the same block as controls. All detailed characteristics of the five wells were acquired from PetroChina Jilin Oilfield Company, as presented in Table 1.

Fig. 1.

(a) Map of the Xinli Oilfield sampling well locations and the analytical protocols. Bacterial community compositions at the phylum (b) and genus (c) levels. Archaeal community compositions at the phylum (e) and genus (f) levels. Comparison of bacterial (d) and archaeal (g) community compositions at the genus level, with significance levels p < 0.05 ( ∗)

Table 1.

Details of specific parameters for sampling wells

Well ID	Depth (m)	Temperature (℃)	Water content (%)	Mineralization (mg/L)	Porosity (%)	Permeability (μm2)	pH	CO2 injection volume (m3)
C56—6	1545	66	83.8	1619	13. 51	6.49	7.21	1710
C52—6	1520	66	88.5	3063	16.11	9.74	7.11	1140
C54—8	1520	66	47.8	3321	13.06	6.07	7.84	1710
N50—08	1530	66	66.2	3321	12.23	5.14	7.26	/
N50—10	1540	66	70.3	3321	13.07	4.02	7.41	/

Produced water samples (25 L per well) were collected directly from the wellhead sampling valves for DNA extraction. Each sample was immediately transferred into presterilized plastic bottles, which had been pre-cooled on dry ice to maintain a low temperature for inhibiting microbial activity. The bottles were filled with water and sealed with butyl rubber stoppers to ensure anaerobic conditions. All samples were maintained on dry ice and transported to the laboratory within four hours for subsequent analysis.

Chemical analyses

After being separated from oil, the produced water was pretreated via centrifugation and filtration for impurity removal and analyte enrichment. Approximately 40 mL of the aqueous phase was centrifuged at 12,000 × g for 10—15 min at room temperature. The clarified supernatant was then filtered through a 0.22 μm pore-size polycarbonate membrane (Whatman Nuclepore™ 47 mm diameter) to remove particulate matter. The anions and cations were quantitatively determined by ion chromatography (Dionex DX-600, Thermo Fisher Scientific, USA) equipped with an AS11-HC analytical column (4 × 250 mm, 5 μm) for anions and a CS120A column (4 × 250 mm, 5 μm) for cations, according to a previously established protocol (Liu et al. 2020). The pH was measured using a calibrated pH meter (PHS-3C, Shanghai Leici, China) fitted with a composite glass electrode.

DNA extraction

Microbial cells were harvested from 25 L of produced water by low-temperature centrifugation (4 ℃, 12,000 × g and 20 min). The pellet was collected after careful removal of the supernatant. Genomic DNA was subsequently extracted using the AxyPrep™ Bacterial Genomic DNA Maxiprep/Miniprep Kit (Axygen Biosciences, USA) following the manufacturer’s protocol for cell lysis and DNA purification. The concentration and quality of the extracted DNA were assessed using the Qubit 1 × dsDNA HS Assay System (Life Invitrogen Corporation, USA), and the DNA was stored at -80 ℃ for subsequent analysis.

High-throughput and metagenomic sequencing

To characterize the overall microbial community structure, DNA extracted from all five well samples (three CO₂-injected and two non-injected controls) was subjected to 16S rRNA gene amplicon sequencing. The 16S rRNA gene libraries for the bacterial and archaeal communities were constructed by through the polymerase chain reaction (PCR) amplification of the extracted DNA. Amplification was carried out with the primers 515F/806R for bacteria and the primers 524F10extF/Arch958RmodR for archaea (Xu et al. 2019). The PCR amplification was initiated with a 5-min denaturation at 94 ℃ to fully denature the template DNA and activate the hot-start polymerase. This was followed by 30 cycles, each comprising denaturation at 94 ℃ for 30 s, primer annealing at 52 ℃ for 30 s, and extension at 72 ℃ for 30 s. A final extension at 72 ℃ for 10 min was then performed to ensure the completeness of amplified products, after which the products were held at 4 ℃. All amplifications were carried out in a Bio-Rad S1000 Thermal Cycler to ensure temperature uniformity, with a heated lid employed to minimize evaporation. To account for amplification variability, three replicates were performed for each sample, and the resulting PCR products were pooled for subsequent downstream analyses.

For metagenomic analysis, the DNA passing quality control from five production wells (three CO₂-injected and two non-injected) was used to prepare sequencing libraries with the ALFA-SEQ DNA Library Prep Kit (Guangdong Magigen Biotechnology Co., Ltd, China), strictly following the recommended procedures. Unique indices were incorporated into each library during preparation to facilitate subsequent multiplexing. The DNA concentration was determined using a Qubit 4.0 Fluorometer (Thermo Fisher Scientific, Waltham, USA). Metagenomic sequencing was ultimately performed on an Illumina HiSeq 2500 platform (Guangdong Magigen Biotechnology Co., Ltd, China) (Holm et al. 2019). Despite the sequencing of all libraries, metagenome-assembled genome (MAG) recovery was achieved only for samples N50—10 and C52—6. This limitation was primarily attributed to the inherent low microbial biomass and variable DNA yields characteristic of deep oil reservoir fluids, which resulted in insufficient sequencing depth in the remaining three samples, rendering them unsuitable for high-quality binning.

Bioinformatics analyses

The raw high-throughput sequencing data were subjected to a standardized analysis pipeline. Purified PCR amplicons were processed through the DADA2 plugin within QIIME 2 (v2020.6) for quality filtering, denoising, and the resolution of amplicon sequence variants (ASVs). Subsequent taxonomic classification of the representative ASV sequences was performed using the SILVA 138 reference database, thereby enabling the quantification of the microbial community composition and abundance (Callahan et al. 2016). To account for potential cross-domain amplification by the 515F/806R primers, an in silico filtering step was performed after taxonomy assignment to ensure that only sequences belonging to the domain bacteria were retained for downstream statistical analysis. A comparative analysis of genus-level taxa between CO₂-injected and water-flooded oil reservoirs was performed using STAMP (v2.1.3), employing a two-sided Fisher’s exact test with Newcombe Wilson confidence intervals (95% CI). To control for false discovery rates in multiple comparisons, p-values were adjusted using the Benjamini-Hochberg (FDR) correlation. Additionally, Permutational Multivariate Analysis of Variance (PERMANOVA) based on Bray–Curtis distances was conducted using the adonis2 function in the R vegan package (9999 permutations) to evaluate overall community compositional differences between the two groups. Microbial α-diversity for the bacterial and archaeal communities was assessed using the Shannon diversity index, Observed ASVs (representing richness), and Pielou’s evenness index. The influence of environmental factors was evaluated through Pearson correlation analysis (for linear relationships) and Mantel tests (for matrix-level associations) using Vegan (v2.6—4 in R v4.3.2) to examine linkages between the physicochemical parameters and microbial community composition. Redundancy analysis (RDA) was conducted in Canoco 5.0 on lg (x + 1)—transformed data to identify the key environmental drivers of the microbial assembly. The overall workflow integrated DADA2-derived ASVs, included Mantel tests to account for spatial autocorrelation, and applied the ecologically interpretable Inverse Simpson index to ensure alignment with contemporary microbial ecology standards.

Raw metagenomic reads were processed with Fastp v0.20.0 (Chen et al. 2018) for quality control (Phred score ≥ 20), adapter removal, and length filtering. High-quality reads were subsequently assembled into contigs (≥ 300 bp) using metaSPAdes v3.7.0 (Nurk et al. 2017). Scaffolds were then reconstructed with N-base gaps using Bowtie2 (Langdon 2015), which utilized paired-end sequencing information. Binning was carried out independently with MaxBin2 v2.2.6, MetaBAT2 v2.12.1, and CONCOCT (Arikawa et al. 2021). Initially, a total of 242 raw bins were recovered (109 from sample N50—10 and 133 from sample C52—6). These resulting bins were integrated and dereplicated across samples using the DAS Tool to remove redundant strains, generating a non-redundant set of 134 MAGs. MAG quality was optimized with RefineM v0.0.24 and assessed using CheckM v1.0.12. To ensure high reliability for downstream functional and comparative analyses, a strict quality threshold was applied uniformly; only MAGs with completeness > 75% and contamination < 10% (with an average nucleotide identity < 97% ensuring species-level non-redundancy) were retained. Exactly 72 MAGs satisfied these strict criteria, constituting the final high-quality dataset (with 30 originating from N50—10 and 42 from C52—6 as their highest-quality representatives). Taxonomic classification was performed with GTDB-Tk v2.3.0; for MAGs containing 16S rRNA genes, sequence validation was conducted against the SILVA r138 database. Functional annotation involved translating scaffolds into coding sequences (CDS) via Prodigal v2.6 (meta-mode), followed by hierarchical annotation using GhostKOALA (priority), KofamKOALA, and EggNOG-mapper v2 (Liu et al. 2022). Metabolic pathways were identified by aligning CDS to the KEGG database using DIAMOND v2.0.13 (Leão et al. 2024) (e-value ≤ 1e-5).

Sequence data submission

The raw sequencing data in this study have been deposited in the National Center for Biotechnology Information (NCBI) under BioProject accession number PRJNA1392695.

Results

Composition of archaeal and bacterial communities in production water from CO₂-injected and water-flooding oil reservoirs

The phylum- and genus-level compositions of bacterial communities in the five production water samples were shown in Fig. 1b and c, respectively. At the phylum level (Fig. 1b and Supplementary Table S1), the bacterial communities were dominated by Firmicutes (9.97%–60.12%) and Proteobacteria (11.39%–51.47%) in all the samples. In the water-flooding oil reservoir samples (N50—08 and N50—10, as control samples), Thermotogae (6.53%–12.41%) and GBS-1 (1.10%–7.35%) were also detected at relatively high proportions. In CO₂-injected oil reservoir samples (C52—6, C54—8, and C56—6), Atribacteria (0.51%–4.61%), Epsilonbacteraeota (11.02%–22.36%) and Nitrospirae (0.95%–7.54%) were also detected, exhibiting significantly higher proportions compared to the two control samples. At the genus level (Fig. 1c and Supplementary Table S2), the dominant taxa in control sample N50—08 and N50—10 were SCADAC1—2—3 (3.02%–21.07%), Unclassified Clostridia (5.11%–19.42%) and Pseudothermotoga (6.40%–12.17%). However, the CO₂-injected samples (C52—6, C54—8, and C56—6) exhibited marked differences from the control group. They shared the dominant taxa, including Thiobacillus (6.41%–24.15%), Arcobacter (10.92%–17.37%) and Thermodesulfovibrio (0.94%–7.54).

For archaea, their community composition is considerably simpler than that of bacteria. At the phylum level, control samples N50—08 and N50—10 were dominated by Crenarchaeota (84.61% and 97.51%, respectively), followed by Thaumarchaeota (11.68% and 0.51%) and Euryarchaeota (3.71% and 1.44%) (Fig. 1e and Supplementary Table S3). In addition, CO₂-injected production water samples (C52—6, C54—8, and C56—6) exhibited the dominant archaeal phyla comprising Crenarchaeota (49.19%–79.68%), Euryarchaeota (3.69%–26.55%), and Thaumarchaeota (2.22%–24.25%). At the genus level, as illustrated in Fig. 1f and Supplementary Table S4, CO₂-injected reservoir samples (C52—6, C54—8, and C56—6) demonstrated greater taxonomic richness compared to control samples (N50—08 and N50—10). In the control samples, the archaeal communities were overwhelmingly dominated by the methylotrophic methanogen Candidatus Methanomethylicus, constituting 69.62% and 87.49% of the archaeal assemblages in N50—08 and N50—10, respectively. Concurrently, in sample C52—6, Unclassified Bathyarchaeia emerged as the dominant lineage, followed by Unclassified Methanomethylicus and Candidatus Korarchaeum. Sample C56—6 remained predominantly composed of Candidatus Methanomethylicus (44.90%), with Candidatus Caldiarchaeum as the second most abundant archaeon (18.02%). A more substantial shift was observed in sample C54—8, where the archaeal community was dominated by Candidatus Caldiarchaeum (46.38%), Methanococcus (24.11%) and Unclassified Nitrososphaeraceae (23.88%).

To elucidate the compositional differences in microbial communities between CO₂-injected and water-flooding oil reservoirs, community-level and taxon-specific analyses were performed. PERMANOVA results (Supplementary Table S5) indicated that CO₂ injection explains a substantial proportion of community variation (bacteria: R² = 0.31, p = 0.3; archaea: R² = 0.35, p = 0.2). While the p-values exceeded 0.05 due to the limited sample size (N = 5) inherent to deep-subsurface sampling, the high R² values and the separation observed in the ordination space (Fig. 3b) suggest a clear biological response to CO₂ flooded. At the genus level, although no taxa reached the strict threshold statistical significance after FDR correlation (q > 0.05), distinct taxonomic trends were identified using STAMP (v2.1.3). Based on a two-side Fisher’s exact with Newcombe Wilson confidence intervals (95% CI), several genera exhibited discernible enrichment patterns (p < 0.05, Fig. 1d and g) between the CO₂-injected and water flooding reservoirs. Bacterial different genera in the CO₂-injected reservoir samples were Arcobacter and Thermodesulfovibrio, whereas that in the water-flooding reservoir samples was Pseudothermotoga. Archaeal different phylotypes were Candidatus Caldiarchaeum and Candidatus Methanomethylicus in the CO₂-injected and water-flooding reservoir samples, respectively.

Fig. 3.

Assessment of environmental drivers on microbial community compositions. a Mantel test evaluating the correlation between environmental variables and community dissimilarity. b Redundancy analysis (RDA) triplot showing the relationship between environmental factors and the composition of (b1) bacterial and (b2) archaeal communities

Alpha-diversity of bacterial and archaeal communities

A total of 77,208–83,165 high-quality bacterial sequences and 727–4,028 archaeal sequences were obtained. To assess the ecological characteristics of these microbial communities under different CO₂ injection conditions, α-diversity indices between CO₂-injection and non-injection groups were evaluated, including richness, diversity, and evenness (Table 2 and Fig. 2). Analysis of microbial α-diversity revealed marked variability across the five produced water samples. Bacterial communities exhibited consistently higher diversity metrics, with Shannon index values ranging from 3.69 to 5.08 (Fig. 2b), observed ASVs (representing richness) between 277 and 419 (Fig. 2a), and Pielou’s evenness indices from 0.57 to 0.82 (Fig. 2c). In contrast, archaeal communities displayed lower diversity, as indicated by Shannon indices of 0.91–2.14 (Fig. 2e), Observed ASVs values of 16–39 (Fig. 2d), and Pielou’s evenness values between 0.28 and 0.75.

Table 2.

Alpha diversity and sequence numbers of microbial communities in production water samples

Type	Sample	No. of sequences	Observed ASVs	Shannon index	Pielous index
Bacteria	N50—08	83,165	299.00	4.3449	0.6998
	N50—10	77,895	277.00	3.8207	0.6109
	C52—6	77,744	419.00	3.6898	0.5693
	C54—8	77,208	349.00	5.0784	0.8167
	C56—6	80,339	353.00	4.3922	0.6931
Archaea	N50—08	4028.0	17.000	1.3936	0.4822
	N50—10	3503.0	24.000	0.91380	0.2772
	C52—6	2644.0	39.000	2.3059	0.6435
	C54—8	2749.0	14.000	1.5126	0.5456
	C56—6	727.00	16.000	2.1369	0.7542

Fig. 2.

Alpha-diversity of bacterial and archaeal communities in production water samples from CO₂-injected and non-CO₂-injected oil reservoirs. Bacterial (a–c) and archaeal (d–f) communities are assessd using us indices of richness (a, d), diversity (b, e), and evenness (c, f)

Correlation analysis between environmental factors and microbial taxa

The physicochemical properties of the five produced water samples are summarized in Table 3. Overall, the pH values of all samples were neutral, ranging from 7.11 to 7.84. The ionic composition was characterized by high concentrations of Ca²⁺ (1109–3120 mg/L) and Cl⁻ (1720–2984 mg/L), accompanied by Na⁺ (583–951 mg/L) and CO₃²⁻ (298–510 mg/L). Analysis also revealed detectable amounts of other ions, including SO₄²⁻ (3.55 mg/L) and NO₂⁻ (8.20 mg/L). K⁺ (1.65–5.91 mg/L) and Mg²⁺ (2.38–6.42 mg/L) were also detected.

Table 3.

Physicochemical parameters of the production water from the Xin li oilfields

Well ID	pH	Concentration (mg/L)
		Na+	K+	Ca2+	NH4+	Mg2+	SO42−	CO32−	Cl−	S2−	NO2−
N50—08	7.26	951	1.65	3120	0.95	3.30	< 1.70	510	2964	3.95	8.20
N50—10	7.41	787	2.69	2940	0.54	5.21	< 1.70	457	2094	3.95	< 0.99
C56—6	7.21	634	5.91	1109	3.39	6.42	1.96	298	1720	4.54	4.69
C52—6	7.11	726	1.78	2702	1.00	4.92	3.55	467	2092	5.94	< 0.99
C54—8	7.84	583	1.93	2227	6.36	2.38	< 1.70	323	1741	5.80	< 0.99

To identify the environmental drivers of microbial communities, the relationships between geochemical parameters (e.g., CO₂ injection, ions and pH) and microbial community compositions across the five well samples were systematically evaluated. As shown in Fig. 3a, six physicochemical factors (pH, S²⁻, Mg²⁺, NH₄⁺, CO₃²⁻ and Na⁺) exhibited positive correlations with bacterial communities, while three factors (Cl⁻, Ca²⁺ and K⁺) showed negative correlations. Among these, the influence of Mg²⁺ and NH₄⁺ on the bacterial communities was significant with p < 0.01 and p < 0.05, respectively. Archaeal communities were positively correlated with four parameters (pH, S²⁻, NH₄⁺ and Na⁺) and negatively with the other five factors (Ca²⁺, Mg²⁺, K⁺, CO₃²⁻ and Cl⁻). Among them, only S²⁻ showed statistical significance (p < 0.05). Additionally, Pearson correlation analysis identified strong correlations between Ca²⁺, CO₃²⁻, and Cl⁻ (r = 0.89–0.96). To further elucidate the influence of physicochemical parameters on microbial community compositions, redundancy analysis (RDA) shows the relationship between environmental variables and both bacterial and archaeal assemblages across the five well samples (Fig. 3b1 and b2). For bacterial communities, axes 1 and 2 explained 28.74% and 27.45% of the variance, respectively, collectively accounting for 56.19% of the total variance (Fig. 3b1). The clustering of control samples (N50—08 and N50—10) indicated similar environmental conditions, while the dispersed distribution of CO₂-injected samples (C52—6, C54—8 and C56—6) reflected greater environmental variability. Based on vector analysis, pH, Ca²⁺, S²⁻, K⁺, Mg²⁺ and NH₄⁺ exhibited strong influences on bacterial community structure. RDA axes 1 and 2 of the archaeal communities explained 50.16% and 21.52% of the variance, respectively, cumulatively representing 71.68% of the total variance (Fig. 3b2).

Metabolic potentials of MAGs

A total of 242 non-redundant metagenome-assembled genomes (MAGs) were reconstructed through binning and stringent quality filtering from produced water samples, including the non-CO₂-injected control (N50—10) and the CO₂-injected sample (C52—6). These MAGs were phylogenetically assigned to 31 bacterial and 4 archaeal phyla (Fig. 4 and Supplementary Table S6). Of these, 109 MAGs were recovered from the non-CO₂-injected control, while 133 MAGs were recovered from the CO₂-injected sample. Bacterial MAGs were predominantly affiliated with the phyla Bacillota B (16 MAGs), Pseudomonadota (12), Desulfobacterota (11), Patescibacteria (9), Bacillota E (5) and Nitrospirota (4). Archaeal MAGs were mainly classified within Halobacteriota (11 MAGs), Thermoproteota (10), Methanobacteriota (3) and Thermoplasmatota (1). Notably, high-abundance MAGs were primarily distributed among the phyla Desulfobacterota, Thermoproteota, Nitrospirota, and Halobacteriota.

Fig. 4.

Circular phylogenetic tree of 242 MAGs from samples N50—10 (non-CO₂-injected control) and C52—6 (CO₂-injected sample) showing the phylum-level taxonomy. The outer ring depicts the normalized abundance (TPM) of each MAG, with distinct colors representing the two sampling environments

From the obtained metagenome-assembled genomes (MAGs), a subset of 72 high-quality genomes was selected, with 42 originating from the CO₂-injected reservoir (C52—6) and 30 from the water-flooding reservoir (N50—10). Functional annotation of these high-quality MAGs was performed via comparison against the KEGG database, enabling analysis of genes involved in diverse metabolic pathways (Fig. 5 and Supplementary Table S7). Analysis of multiple high-quality genomes identified eighteen key metabolic pathways, including carbon fixation (RuBisCO, the CBB cycle and the Wood-Ljungdahl pathway), methanogenesis, fermentation to lactate and formate, nitrogen metabolism (denitrification and nitrification); sulfur metabolism (dissimilatory sulfate reduction, alternative thiosulfate oxidation, and Sulfide oxidation); as well as F-type ATPases and V/A-type ATPases (energy conversion mechanisms), yet their abundance and distribution exhibited marked differences between thewater-flooded (N50—10) and the CO₂-injected (C52—6) samples. The N50—10 metageome was notably enrich in genes responsible for Wood-Ljungdahl pathway, Fermentation to lactate and Dissimilatory nitrate reduction. Conversely, the CO₂-injected sample C52—6 displayed a higher prevalence of genes related to Methanogenesis (CO₂ and acetate), Denitrification, Nitrification, Dissimilatory sulfate reduction and Sulfide oxidation. These results demonstrate that CO₂ injection profoundly reshapes the biogeochemical cycling of carbon, nitrogen, and sulfur in the oil reservoir.

Fig. 5.

Heatmaps of MAGs based on profiles of genes related to different biogeochemical cycles and energy metabolism: N50—10 (a) and C52—6 (b)

Discussion

Distribution of dominant and core taxa in production water from CO₂-injected and water-flooding oil reservoirs

Because oil reservoirs represent extreme biospheres and unique ecosystems, understanding the impact of CO₂ injection on microbial community structure and diversity of oil reservoirs is fundamental and critical for CCUS applications. In this study, we investigate the microbial composition, abundance, and key environmental factors in produced water samples collected from a long term (> 20 years) CO₂-injected oil reservoir and adjacent water-flooding reservoir.

High-throughput sequencing analysis of the five produced water samples revealed the differences in microbial community compositions between CO₂-injected and water-flooding oil reservoir samples. At the phylum level, CO₂ injection corresponded to a distinct decrease in Firmicutes and a concurrent rise in Proteobacteria. In sample C56—6, Proteobacteria became the predominant phylum, reaching a peak relative abundance of 51.47%. Meanwhile, a prominent difference induced by CO₂ injection was the significant enrichment of Nitrospirae, with relative abundances reaching 7.54% (C52—6) and 6.35% (C54—8), values substantially higher than detected in control samples (Fig. 1b). Furthermore, the significant enrichment of Thermotogae and Atribacteria in CO₂-injected production water suggests that CO₂ injection drives strong microbial selection. At the genus level, the dominant taxa in control samples (N50—08 and N50—10) were Unclassified Clostridia, SCADC1—2—3 and Pseudothermotoga. However, the CO₂-injected samples shared a core functional group predominantly represented by Thiofaba, Thermodesulfovibrio and Arcobacter (Fig. 1c). Thiofaba, first described in 2008 (family Halothiobacillaceae), exhibits metabolic activity directly linked to CO₂ utilization. It performs edchemosynthetic CO₂ fixation fueled by the oxidation of sulfur compounds, supporting its autotrophic growth and enabling it to play a dual role in the sulfur and carbon cycles (Mori and Suzuki 2008). Intriguingly, Thermodesulfovibrio (phylum Nitrospirae) is a strict anaerobe that primarily derives energy by using hydrogen or simple organic acids as electron donors and sulfate as an electron acceptor. In co-culture experiments, Thermodesulfovibrio has been shown to engage in syntrophic relationships with CO₂-reducing methanogens to produce methane. Arcobacter, a thermotolerant and anaerobic genus within the order Campylobacterales is implicated in organic acid metabolism and hydrocarbon degradation in reservoir ecosystems. Frates et al. reported that Arcobacteraceae may form metabolic interactions with sulfate-reducing bacteria (e.g., Desulfobulbaceae and Desulfobacteraceae) by consuming organic acids (e.g., acetate and propionate) generated by fermentative bacteria (such as Fusobacteriaceae) (Frates et al. 2023). This consumption helps maintain system balance, while the H₂ produced during fermentation can be utilized by sulfate-reducing bacteria, creating a synergistic interaction that regulates carbon and sulfur cycling (Wu et al. 2024). Thus, the abundance of Arcobacter in CO₂ injection samples may be directly or indirectly linked to CO₂-oxidizing activities.

In contrast, the archaeal communities, which are less diverse than bacterial commuinities, showed significant changes in response to CO₂ injection (Fig. 1e). The phylum Crenarchaeota was the dominant lineage across all five produced water samples. However, its abundance was significantly higher in non-CO₂-injected control samples (N50—08 and N50—10) than in the three CO₂-injected samples. The most considerable reduction of Crenarchaeota occurred in sample C54—8 (approximately 49%). Notably, the decline in abundance of Crenarchaeota co-occurred with a significant increase in the relative abundance of Euryarchaeota and Thaumarchaeota exclusively in the CO₂-injected samples (C54—8 and C56—6). At the genus level, archaeal communities in the control samples (N50—08 and N50—10) were composed predominantly of the methylotrophic methanogen Candidatus Methanomethylicus (Fig. 1f). This genus is known to anaerobically utilize methylated compounds (e.g., methanol and methylamines) as electron donors for methanogenesis (Huang et al. 2024). In contrast, CO₂ injection markedly altered the archaeal community compositions. Candidatus Caldiarchaeum increased in relative abundance across multiple samples, becoming the dominant archaeon in sample C54—8 and the second most abundant in sample C56—6. Candidatus Caldiarchaeum is postulated to function as a thermophilic heterotroph, utilizing microbial metabolites such as acetate without direct CO₂ fixation. Its enrichment under CO₂ injection may be indirectly stimulated by the increased availability of organic substrates derived from enhanced autotrophic activity (Takami et al. 2015). At the same time, sample C54—8 exhibited the most pronounced community difference, transitioning to a state co-dominated by Candidatus Caldiarchaeum, and notablly, Methanococcus and Unclassified Nitrososphaeraceae. Methanococcus, a mesophilic hydrogenotrophic methanogen, likely engages in syntrophic relationships with sulfate-reducing and fermentative bacteria, consuming their metabolic byproducts (hydrogen (H₂) + CO₂) to drive methanogenesis, thereby contributing to carbon cycle (Goyal et al. 2016). Members of Unclassified Nitrososphaeraceae were mostly the ammonia-oxidizing archaea (AOA) and thus play a critical role in nitrification process. Notably, AOA are characterized by their ability to thrive under oligotrophic conditions, exhibiting a high affinity for ammonia that enable them to oxidize NH₃ even at exceedingly low concentrations and under acidic conditions (Hatzenpichler 2012).

Diversity and physicochemical factors influencing microbial community in CO₂-injected and water-flooding oil reservoirs

Beyond compositional changes, assessing variation in species diversity provides crucial insights into the stability and functional potential of the community. To address this, we quantified community structure through α-diversity analysis, using indices including the Shannon index, observed ASVs (representing richness) and Pielou’s evenness (Fig. 2 and Table 2). In general, the Shannon index, which correlates positively with community diversity, showed distinct patterns between microbial domains: bacterial communities maintained stable diversity (mean indices: 4.08 vs. 4.39 in controls), whereas archaeal communities exhibited significantly increased Shannon value (1.15 vs 1.99), reflecting enhanced diversity under CO₂ exposure. The observed ASVs, reflecting species richness, further elucidated these different responses (Deng et al. 2024). CO₂-injected samples (C52—6 and C56—6) showed marginally elevated bacterial richness compared to non-injected controls, whereas archaeal communities demonstrated minimal fluctuation. This pattern may be attributed to archaeal resilience through physiological stability or functional redundancy, contrasting with bacterial utilization of CO₂-derived substrates (Phillis 2004). Specifically, heterotrophic bacteria (e.g., Firmicutes, Bacteroidetes) may exploit dissolved inorganic carbon (HCO₃⁻) or organic complexes (e.g., humic acid) generated under CO₂ enrichment, thereby expanding their niche occupancy (Nowak et al. 2017).

Environmental conditions and microbial communities exhibit interplay within oil reservoirs (Lourdes et al. 2022). Consequently, identifying the key environmental drivers that shape microbial communities is essential for both a fundamental understanding of ecosystem mechanisms and guidance of potential practical interventions in CCUS. Distinct geochemical patterns were associated with the variations in ionic concentrations between CO₂-injected (C56—6, C52—6, C54—8) and non-injected wells (N50—08 and N50—10). The CO₂-influenced produced water samples exhibited markedly lower mean concentrations of Ca²⁺ compared to non-injected samples. This reduction can be attributed to the reaction of injected CO₂ with formation water, forming carbonic acid (H₂CO₃) that subsequently dissociates into HCO₃⁻ and CO₃²⁻ (Becke 2023). The resulting CO₃²⁻ ions combine with cations such as Ca²⁺, leading to precipitation of carbonate minerals (e.g., calcite, CaCO₃), thereby depleting these ions from the aqueous phase (Pearce et al. 2022). Conversely, CO₂-injected samples demonstrated substantially elevated concentrations of inorganic sulfur compounds (SO₄²⁻, S²⁻) and ammonium (NH₄⁺). This enrichment may result from CO₂-associated enhancement of sulfur oxidation processes, coupled with proliferation of indigenous sulfate-reducing bacteria (SRB) and hydrocarbon-degrading microorganisms under the modified geochemical conditions (Zhao et al. 2025b). The intensified microbial activity likely promotes catabolism of nitrogenous organic matter, ultimately leading to accumulation of SO₄²⁻, S²⁻ and NH₄⁺ (Wang et al. 2023). These geochemical observations align with previously described microbial community shifts, particularly the dominance of sulfur-oxidizing Thiofaba and hydrocarbon-degrading SCADC1—2—3 in CO₂-influenced samples. Meanwhile, to analyze the relationship across the five well samples, Pearson’s correlation analysis, Mantel tests and redundancy analysis were employed (Fig. 3a, b1 and b2). These analyses showed distinct clustering of non-injected control samples (N50—08, N50—10) versus CO₂-injected samples (C54—8, C56—6), indicating different environmental constraints. Meanwhile, S²⁻, CO₃²⁻, pH, Na⁺, NH₄⁺, and Ca²⁺ emerged as predominant factors of archaeal community variation. Specifically, S²⁻, CO₃²⁻, pH and Na⁺ showed positive associations with control samples but negative correlations with archaeal communities in CO₂-injected environments, suggesting a potential preference for conditions with reduced acidity and ammonium concentrations following CO₂ injection. Taken together, these results demonstrate that CO₂ injection may act as a key perturbation force, potentially influencing microbial community restructuring by altering geochemical parameters such as pH and ion concentrations. Notably, NH₄⁺ and S²⁻ emerged as particularly prominent potential factors influencing this process, highlighting the domain-specific ecological strategies employed by bacteria and archaea in response to geochemical stress (Mu and Moreau 2015).

Microorganisms in CO₂-injected and water-flooding reservoirs associated with substrate carbon, sulphur and nitrogen cycles

Initial analysis of carbon fixation pathways identified eight metagenome-assembled genomes (MAGs) encoding key enzymatic genes, present in both samples N50—10 and C52—6. Among these, four MAGs (Archaeoglobaceae_bin110, Archaeoglobus_bin121, Methanomethylicaceae_bin110 and Methanothrix_B_thermoacetophila_bin118) were associated with the RuBisCo enzyme, three MAGs (Aerophobales_bin43, Bacillota_G_bin33 and Formimnas_bin51) were linked to the Wood-Ljungdahl pathway (acsABCDE) and the remaining MAG (Bacillota_G_bin33) was assigned to the Calvin-Benson-Bassham (CBB) cycle (prkB). In sample C52—6, the four MAGs associated with the RuBisCo (prkB) pathway exhibited increase abundance. Among these, Archaeoglobaceae_bin110 (Buessecker et al. 2023) and Archaeoglobus_bin121 (Klenk et al. 1997) are thermophilic archaea typically functioning as sulfate-reducing mixotrophs/heterotrophs. They encode a type Ⅲ RuBisCo, which is not primarily involved in conventional carbon fixation but rather functions in the detoxification of highly toxic metabolic by-products such as ribulose-5-phosphate. CO₂ injection induces acid and energy stress, potentially accelerating the accumulation of toxic metabolites, thereby likely explaining the upregulation of these genes as a stress response. Methanomethylicaceae_bin110 (Kohtz et al. 2024) (a methylotrophic methanogen) and Methanothrix_B_thermoacetophila_bin118 (Prondzinsky et al. 2023) (an acetoclastic methanogen) also possess Type Ⅲ RuBisCo, which serves a similar role in mitigating formyl stress under anaerobic conditions. The acidic environment inhibits methanogenesis, leading to the accumulation of formate and other intermediates, thereby increasing the demand for RuBisCo-mediated detoxification. This further supports the interpretation that the observed increased in archaeal RuBisCo likely represents a stress response. Interestingly, comparative analysis between C52—6 and N50—10 revealed a decreased signal for the Wood-Ljungdahl pathway (acsABCDE) in three MAGs (Aerophobales_bin43, Bacillota_G_bin33 and Formimonas_bin51). Given that the situ pH of the geological environment remained circumneutral following CO₂ injection, this attenuation cannot be attributed to acidification. Instead, we attribute this pattern to the competitive exclusion of acetogenic bacteria by CO₂-stimulated methanogens. Members of the Aerophobales, Bacillota and genus Formimonas have been characterized as acetogens that utilize the Wood-Ljungdahl pathway (acsABCDE) to reduce CO₂ to acetate. This metabolic strategy places them in competition with hydrogenotrophic methanogens for shared H₂ and CO₂ substrates. Under the elevated CO₂ conditions of the reservoir, methanogens are often significantly enriched, outcompeting acetogens for available electron donors (Lever 2012). This change has been observed in other reservoir environments where activation of CO₂-reducing methanogens coincided with a marked decline in acetogen populations (Yang et al. 2016).

Six MAGs encoding key enzymes for methanogenesis were successfully assembled and identified. Five MAGs from sample N50—10 (Methanothrix_B_Thermoacetophila_bin118, Archaeoglobaceae_bin110, Archaeoglobus_bin121, Methanobacteriales_bin111 and Methanolbacteriales_bin128) were implicated in various methanogenesis pathways: Methanogenesis (mcrABG), methanol → methane (mtaABC), acetate → methane (cdhCDE) and CO₂ → methane. Notably, sample C52—6 contained all five of these MAGs plus an additional MAG, Bathyarchaeales_bin122. Compared to N50—10, the overall signal for methanogenic pathways was enhanced in C52—6, with a distinct shift in pathway preference: methylotrophic methanogenesis decreased, while acetoclastic and hydrogenotrophic methanogenesis increased. This suggests that CO₂ injection reshaped the methanogenic microbial community, potentially inhibiting methylotrophic pathways while stimulating hydrogenotrophic pathways, which are more tolerant of acidic conditions and can directly utilize CO₂. The recruitment of auxiliary microbes such as Bathyarchaeales_bin122 (Wang et al. 2025a) may supply necessary fermentation substrates, resulting in overall enhanced methanogenic activity but altered pathway dominance, consistent with previous high-throughput sequencing data.

Seven MAGs were assembled from sample N50—10 (Archaeoglobus_bin121, Bacillota_G_bin32, Geocrenencus_bin132, Korarchaeum_cryptofilum_bin123, Syntrophorhabdaceae_bin102, Thermodesulfobacterium_bin74, Thermodesulfobacterium_commune_bin64). In contrast, sample C52—6 yielded nearly twice as many MAGs, including Archaeoglobus_bin121, Arcobacteraceae_bin101, Bacillota_B_bin21, Bathyarchaeales_bin122, Calditerrivibrio_nitroreducens_bin93, Geocrenencus_bin132, Korarchaeum_cryptofilum_bin123, Syntrophorhabaceae_bin102, Thermodesulfobacterium_bin72, Thermodesulfobacterium_bin74, Thermodesulforbacteriaceae_bin96, Thermodesulfovibrio_bin77. Furthermore, the dissimilatory sulfate reduction pathway (sulfite → sulfide, reversible; dsrAB) in C52—6 involved three additional MAGs compared to the sample N50—10, with the strongest signal attributed to Thermodesulfobacterium_bin74 (Ji et al. 2024). This indicates that CO₂ enriched the microbial community responsible for dissimilatory sulfate reduction, both in diversity and abundance, with Thermodesulfobacterium_bin74 emerging as a dominant player, corroborating prior high-throughput results.

Two MAGs (Azospirillaceae_bin6 and Ozemobacter_sibiricus_bin20) were assembled from sample N50—10, whereas three MAGs (Arcobacteraceae_bin101, Azospirillaceae_bin6, Thermodesulfobacterium_bin72) were identified in sample C52—6. Notably, both the denitrification (nitrous oxide → nitrogen; nosZ) and nitrification (hydroxylamine → nitrite; hao) pathways exhibited significantly enhanced total signals in C52—6 compared to N50—10. This demonstrates that CO₂ injection did not inhibit nitrogen cycling but rather reconfigured the microbial community involved in subsurface nitrogen transformation. The introduction of functional taxa such as Arcobacteraceae_bin101 (Li et al. 2024) and Thermodesulfobacterium_bin72 (Zehnle et al. 2023) — potentially serving as a bridge between sulfur and nitrogen cycles —is of particular interest. Many sulfate-reducing bacteria (SRB) can switch to using nitrate/nitrite as electron acceptors for respiration (i.e., denitrification) under sulfate-limited conditions, suggesting a potential mechanistic link. Overall, metagenomic metabolic potential analysis revealed that CO₂ injection may influence the functional landscape of the reservoir microbiome, enhancing key processes including hydrogenotrophic methanogenesis, dissimilatory sulfate reduction and denitrification, while concurrently suppressing methylotrophic methanogenesis and Wood-Ljungdahl pathway. This functional restructuring, driven by environmental acidification and the availability‌ of inorganic carbon, reflects a community adaptively reconfiguring its metabolic network toward pathways that are energetically feasible and environmentally tolerant under elevated CO₂ conditions.

While this study provides valuable insights into microbial community and functional potential changes driven by long-term CO₂ injection, several limitations must be acknowledged. First, the analysis was based on a limited sample size (N = 5 wells). Although deep subsurface sampling is operationally challenging and the wells were strategically selected from the same geological block to ensure comparability, the relatively small sample size limits the statistical power of the broader conclusions. Second, inherent reservoir heterogeneity, including localized variations in mineralogy, porosity, fluid flow dynamics, and micro-environments may influence geochemical parameters and microbial distributions. While our comparative design (CO₂-injected vs. non-injected wells in the same strata) aimed to isolate the CO₂ effect, natural background variations cannot be entirely ruled out. Consequently, future research should leverage these insights to develop site-specific microbial management strategies, integrating these decadal-scale functional shifts into predictive biogeochemical models.

Implications for CCUS and long-term CO₂ storage integrity

A central challenge in global CCUS initiatives is ensuring that injected CO₂ is either permanently sequestered or beneficially utilized. The functional gene signatures observed in this study, specifically the shift toward hydrogenotrophic methanogenesis under long term CO₂ injection, point to the reservoir’s intrinsic capacity for subsurface biological methanation. This observation aligns with findings from other long term CO₂ EOR sites. For instance, Tyne R.L. et al. conducted a study at Olla Field (Louisiana, USA) using noble gas, stable isotope, and gene-sequencing analyses, demonstrating that microbial methanogenesis converted as much as 13—19% of injected CO₂ to methane with an additional 74% of CO₂ dissolved in groundwater, yielding an in situ methanogenesis rate of 73—109 mmol CH₄ × m⁻³ (STP) × yr⁻¹ (Tyne et al. 2021). Meanwhile, at the Cranfield CO₂-EOR site (Mississippi, USA), localized soil gas anomalies containing up to 44% CO₂ and 47% CH₄ were attributed by Anderson J.S. et al. to microbial methanogenic activity rather than well leakage, underscoring the potential scale of subsurface bioconversion (Anderson et al. 2017).

Integrating our findings into a global context reveals that microbial responses to long-term CO₂ stress are highly dependent on both site-specific geochemical and operational status. On one hand, our results are consistent with observations from established international CCUS field sites. For instance, similar to the microbial response in the Michigan Basin CO₂-EOR site in the USA (Stemple et al. 2022), our results exhibited clear and sustained microbial differentiation, particularly the enrichment of taxa associated with carbon and sulfur cycling. On the other hand, it is worth noting that not all long-term CO₂-EOR sites exhibit uniform microbial responses. In a USGS (the United States Geological Survey)-led study of a hydrocarbon reservoir 40 years post CO₂-EOR, microbial communities in CO₂-impacted and non-impacted samples did not significantly differ, although several low-abundance bacteria were found to be significantly associated with the CO₂ affected group (Shelton et al. 2018). This suggests that any compositional change induced by CO₂ flooded may have been transient, with the affected wells potentially reverting toward pre-injection conditions over the four decades since injection ceased. These inter-site differences underscore the necessity of reservoir-specific microbial risk assessments and long-term monitoring tailored to each CCUS project.

Alongside potential methanogenic conversion, the enrichment of certain microbial groups may also affect reservoir performance and storage integrity. A recent comprehensive review on microbial considerations for permanent geological CO₂ storage emphasizes that microbial communities can both compromise and enhance storage by processes including methanogenesis, sulfidogenesis, corrosion and carbonate mineralization (Nixon et al. 2025). Our study revealed an enhancement in sulfur-cycling taxa, particularly those associated with dissimilatory sulfate reduction. The activity of these sulfate-reducing microorganisms can lead to hydrogen sulfide production, which may contribute to reservoir souring and associated corrosion risks. In addition, CO₂ injection may mobilize electron acceptors from the mineral matrix, potentially stimulating further microbial activity and corrosion processes (Anderson 2017).

Conclusion

In summary, this study employed high-throughput sequencing and metagenomic approaches to extensively investigate the microbial community composition, potential environmental factors and core microbial functional profiles in produced water from both long-term (over 20 years) CO₂-EOR and non-CO₂-injected production wells in the Xinli Oil Field. Several major conclusions were drawn, with the following highlights: (1) CO₂ injection significantly increased microbial diversity and triggered a functional shift in dominant communities, characterized by a transition from methylotrophic to hydrogenotrophic and heterotrophic archaea, along with an enrichment of bacterial taxa involved in sulfur and nitrogen cycling; (2) CO₂-driven microbial reassembly was strongly correlated with changes in geochemical parameters such as pH and ion concentrations, with NH₄⁺ and S²⁻ identified as particularly prominent factors potentially influencing this process; (3) Metagenomic analysis revealed that CO₂ injection may induce functional restructuring of the reservoir microbiome, specifically, the genetic potential for hydrogenotrophic methanogenesis, dissimilatory sulfate reduction, and denitrification was enriched, whereas the genomic signatures for methylotrophic methanogenesis and the Wood-Ljungdahl pathway were attenuated. These results expand our understanding of biogeochemical and microbial processes in CO₂-impacted reservoirs, and understanding these complex biogeochemical interactions provides a critical foundation for designing microbial management strategies that optimize carbon conversion in subsurface environments.

Author contributions

Legang Chen: writing—original draft, methodology, conceptualization, investigation, visualization. Lei Zhou: methodology, conceptualization, supervision, writing—review and editing, investigation, data curation. Bowen Wang: investigation. Muhammad Javed: data curation, resources. Yifan Liu: conceptualization, supervision. Shizhong Yang: conceptualization, supervision. Jidong Gu: supervision. Bozhong Mu: conceptualization, project administration, supervision, funding acquisition, writing—review and editing.

Funding

This work was supported by the National Natural Science Foundation of China (Grant No. 42173076, 42061134011, 42473082 and 52074129), the Natural Science Foundation of Shanghai (Grant No. 21ZR1417400), Fundamental Research Funds for the Central Universities (No. JKJ01241714) and the Research Program of State Key Laboratory of Bioreactor Engineering.

Data availability

The data that support the findings of this study are available at the NCBI Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) under accession number PRJNA1392695.

Declarations

Ethics approval and consent to participate

This article does not contain any studies with human participants or animals performed by any of the authors of this investigation.

Competing interests

The authors declare no competing interests.