A basic nutrient mix was employed to promote the growth of indigenous microbes. The nutrient formulation and concentrations are shown in Table 1. All fluids used in this study, including crude oil (the oil phase of the fluid produced from the well) and produced water (the water phase of the fluid produced from the well), were obtained from the Qizhong Block oil field in Karamay, Xinjiang, China. The produced water used in the experiment was mixed with formation-water from 5 production wells. The sample well numbers were 72,602, 72,604, 72,648, 72,649, 72,659. The sample was completely filled in a 15 L sterilized bucket that was previously swept with nitrogen to prevent oxygen from diffusing into the sample. The sample was then immediately transported to our laboratory. All freshly produced liquids were either immediately employed or refrigerated at 4°C for later experiments. The viscosity of crude oil was 5.55 mpa.s (thin oil, 37°C). Ionic composition of the produced water is detailed in Supplementary Table S1.

The device incorporated four components: a 2 PB series advection pump, intermediate container, an 8 m long sand-filled tube, and a large temperature incubator. The advection pump, operating on the piston mechanism, channeled nutrient solutions from the intermediate container to the sand-filled tube. Two intermediate containers were present, each dedicated for crude oil and the nutrient solution, respectively. The sand-filled tube, constructed from stainless steel, featured 5 linear sand tubes and 4 U-shaped tubes, each equipped with a sampling site and a valve. The overall size of the sand-filled tube was D5 cm × L800 cm. A scheme of the long core microbial flooding simulation device is shown in Supplementary Figure S1. Quartz sand (30 ~ 50 μm) served as the porous medium within the tube, ensuring no voids. The gas permeability was gaged at 4831 millidarcy (md). The formation-water was fully saturated, and the pore volume (PV) had been computed to be 7,040 cm³.

The procedure of the flooding experiments aimed at mimicking the entire oil recovery process in field, including the stages of water saturation, oil saturation, water flooding and microbial flooding (tertiary oil recovery). The sand-filled tube was first saturated with formation-water under vacuum conditions (water saturation stage). Secondly, oil was injected into the sand-filled tube until the end of the uniform flow of crude oil (total saturated oil 4,550 cm³). After that, the formation-water was injected for the secondary recovery stage (water drive stage). When the water content of the end of the sand-filled tube exceeded 98%, the injection was stopped. Subsequently, the microbial flooding stage was performed. Nutrient solution was injected for water flooding. All tests maintained a constant flow rate of 1 mL/min at 37°C and persisted for 50 days. The volume of released oil and water cut were measured. Oil recovery efficiency (ORE, %) was calculated as follows (Gao et al., 2020):

“Original oil in place” (mL) is the volume of water displaced by oil saturation.

Water cut (%) was derived as follows:

Throughout this study, the intermediate container’s medium was refreshed bi-daily. Samples were taken from seven sampling sites. Each time, the switch of sampling site was turned on to drain the stored liquid. Samples were collected in a 5 mL centrifuge tube and centrifuged immediately (12,000 r, 10 min). Precipitates catered to genomic DNA extraction, while supernatants contributed to nutrient concentration determination. DNA extraction employed the AxyPrep ™ Genomic DNA Miniprep Kit (Axygen Biosciences, CA, United States), with procedures elaborated in Supplementary material.

Total sugar was performed through the phenol-sulfuric acid method (Liu et al., 2011), with the standard curve presented in Supplementary Table S2. Total nitrogen was gaged via the persulfate oxidation approach (Lima et al., 1997), whereas total phosphorus was assessed by the digestion-molybdenum-antimony method, with the required kits supplied by Tianjin Hasi Water Analysis Instrument Co., LTD. Detailed procedures reside in the Supplementary material.

Genomic DNA was extracted as previously mentioned (Gao et al., 2018). High-throughput sequencing of 16S rRNA genes was performed by Beijing Novogene Co., Ltd. All 16S rRNA genes were sequenced in QIIME2 (Quantitative Insights into Microbial Ecology) 2) according to the methods recommended in the environment for processing.¹ The double-ended sequences were spliced, and then quality control was carried out to remove primers and barcodes. In the Silva database, we checked the OTUs species information (Quast et al., 2012). The community structure of each sample was counted at the genus level. The α-diversity of the samples was analyzed using the qiime diversity function of QIIME2.² Principal Component Analysis (PCA) was used to analyze the difference of OTU composition in different sampling sites and sampling times. We used “Canoco 5” software for the redundancy analysis (RDA) of each sample. The network diagram between nutrients and microorganisms (OTU) was constructed. Network visualization and topological properties were conducted using R v3.6.3 (“psych” package) and the Gephi platform v0.9.2 (Yun et al., 2021).

To explore the function of the indigenous microbial community in the long core microbial flooding simulation device, the samples 2–21, 5–20 and 10–20 were selected for metagenomic sequencing, which was completed by Beijing Novogene Co., Ltd. Following weight removal and quality control, we integrated the metagenomic sequencing data with the species classification information, as well as the integrity and pollution degrees of all MAGS obtained. MAGS with Q50 (integrity value −5 × pollution degree) greater than 50 were identified as qualified MAGS, and the MAGS in each sample were screened for subsequent analysis (Yang et al., 2023).

Prodigal software was used to translate the genomic information into protein sequences, and the functional genes and metabolic pathway information contained in each MAG were obtained by comparison with the KEGG database. The metabolic information of MAG with sufficient quality was mapped online, and the key genes in the pathway of alkane metabolism, nitrogen metabolism, sulfur metabolism and methane generation were analyzed. The key genes of alkane metabolism were selected as alkanes monooxygenase alkB gene and long chain monooxygenase ladA gene (Yun et al., 2021). In nitrogen metabolism, nitrate reduction pathway and denitrification pathway were the main concerns, and the genes involved were napA, nirK, nosZ, etc. Sulfur metabolism was primarily focused on the sulfate reduction pathway with dsrA and dsrB being the relevant genes involved (Yao et al., 2023). The key gene in the process of methane generation was methyl-coenzyme M reductase mcrA gene (Egger et al., 2016).

Over a 50-day experimental period, nutrients were injected according to the Table 1. The effects on water cut and oil recovery are delineated in Figure 1. Primary water flooding led to a recovery efficiency nearing 32%. Following nutrient injection, the first 7 days witnessed minimal changes in oil recovery efficiency (ORE). However, between the 7th and 22nd days, ORE surged from 35 to 66%. Post day 23, stability was observed with a final ORE of 68%. It may be because the long-term nutrient injection caused the formation of large pores inside the sand-filled tube. Subsequent nutrients may flow out along the large pores (Chai et al., 2018). Consequently, nutrients could not permeate oil-rich areas, stagnating the ORE.

The change trend of total sugar (TS), total phosphorus (TP) and total nitrogen (TN) at each sampling site was consistent (Figure 2). Notably, there was a delay in the alteration of nutrient concentration at the remote end of the sand-filled tube, requiring a specific period for nutrients to move towards the depths of the tube. This may be the reason that affected the community structure of each sampling site in the early stage.

From the fifth day, residual phosphorus levels stabilized at about 1,500 mg/L (Figure 2A). A noticeable decline in total nitrogen content transpired between the 6th and 15th days, possibly linked to the enhanced ORE (Figure 1). Investigating the interplay between environmental variables and microbial communities, a redundancy analysis (RDA) was conducted, exploring the initial 30 days for effects of sugars, nitrogen, and phosphorus on the community structure at three distinct sampling sites (Supplementary Figure S2 and Supplementary Table S3). Total phosphorus had significant effects on the community structure at the front and middle end of the sand-filled tube. Total nitrogen and total phosphorus had significant effects on the community structure of the terminal sampling site. The results indicated that nutrition had an effect on microbial community structure, which was consistent with findings by Gralka et al. (2020).

The Shannon index was used to analyze α-diversity at each sampling site (Supplementary Figure S3). The α-diversity of bacterial community in the two sampling sites was almost the same. From 1 to 30 days, the α-diversity in the produced fluid decreased rapidly after nutrient migration to each sampling site. This phenomenon may result from the nutrients injection that significantly stimulated a limited number of bacteria with oil recovery capabilities, causing the decline or vanishing of other microorganisms with diverse functions. Consequently, the species richness of the ecosystem decreases, leading to a reduction of α-diversity (Yun et al., 2021). After reducing the nutrient concentration, α-diversity increased slightly, but it was still lower than before the nutrient injection (Supplementary Figure S3), which might be due to the decrease of nitrate concentration in nutrients leading to the increase of α-diversity in bacterial community (Xu et al., 2023).

Principal component analysis (PCA) elucidated temporal and spatial bacterial community dynamics (Figure 3). Temporally, the pre-injection phase demonstrated negligible microbial variations (Figure 3A). From 1 to 5 days (about 1 PV), there were some differences in the community structure among the seven sampling sites. The succession direction of the community at the back sampling site was consistent with that at the front sampling site, but later than that at the front sampling site. This may be due to the time it took for nutrients to migrate to the sampling site. Across 6–50 days, community structure stability was evident. The community succession direction was consistent on the time scale. Spatially, the community succession was congruent across multiple sampling sites (Figure 3B), regardless of their distance from injection sites.

Metagenomic analyzes of samples from days 20 and 21 focused on genes related to the oil recovery (Figure 5). Results suggested a spectrum of reservoir microbial functionalities. The alkB gene was present in Paracoccus, Marinobacter, and Citrobacter in three metagenomic samples. Many bacteria contained the ability to produce rhamnolipids, such as Pseudomonas, Flaviflexus, and Enterococcus. Additionally, bacteria including Castellaniella and Vibrio were discovered that had genes involved in the nitrate reduction and denitrification processes. The results indicated that reservoir microorganisms had the potential functions of hydrocarbon degradation, denitrification gas production and surfactant production. Genes associated with SRB and methanogens (mcrA) were not detected in this study, which may be due to the introduction of nitrate (Jami et al., 2013). As SRB have the ability to produce H₂S, nitrate is added to the injected water in numerous oil fields to prevent the growth of SRB and hence inhibit the formation of H₂S (Ren et al., 2011).

Petroleum oil reservoirs are the home to phylogenetically and metabolically diverse groups of microbial communities (Liu et al., 2010). At present, this study mainly focused on the changes of reservoir microbial communities. In forthcoming research, on the one hand, the metabolites of these microorganisms need to be further explored to supplement the “food chain” of the reservoir ecosystem. On the other hand, some functional bacteria that can improve the oil recovery need to be screened. Targeted activation of these functional bacteria is helpful to enhance oil recovery.