Elevated CO₂ Drives the Enrichment of Multidrug Resistance Genes in Paddy Soils

Abstract

Antibiotic resistance genes (ARGs) are becoming a global issue due to the emergence of superbugs. However, the impact of elevated CO₂ (eCO₂) on the soil antibiotic resistome remains largely unknown. Here, using a free-air CO₂ enrichment platform, we employed high-throughput quantitative PCR and 16S rRNA gene sequencing to investigate the effect of eCO₂ (ambient + 200 ppm) on soil ARGs and bacterial communities in a paddy ecosystem at harvest. The results showed that eCO₂ had no significant effect on rice biomass. A LEfSe analysis identified a clear taxonomic shift, with taxa such as c_Clostridia, g_Dehalobacter and g_Syntrophus being significantly enriched under eCO₂. The total relative abundance of ARGs increased 1.5-fold under eCO₂, driven by a 2.8-fold increase in multidrug resistance genes. The correlation and network analyses revealed that the proliferation of specific potential host bacteria was the primary driver of the observed ARG enrichment under eCO₂. Together, this study offers new insights into the eCO₂-driven alterations of soil antibiotic resistomes, highlighting the elevated dissemination potential of multidrug resistance genes within paddy ecosystems and their potential implications for food safety.

1. Introduction

Climate change and antibiotic resistance are converging threats to global health [1,2]. As a primary driver of climate change, elevated atmospheric carbon dioxide (eCO₂) alters plant–soil interactions by modifying plant carbon assimilation and subsequent root exudation [3,4]. Given that the soil microbiome is a major reservoir of antibiotic resistance genes (ARGs) and is exquisitely sensitive to carbon availability [5,6], it is imperative to determine how the antibiotic resistome will respond to this fundamental global change.

Atmospheric CO₂ levels have risen steadily over the past century, surpassing 420 ppm in recent years, and are projected to continue increasing due to anthropogenic activities [7,8]. Increases in the atmospheric CO₂ concentration may impact soil microbial communities, which are essential drivers of plant growth and ecosystem health [9]. Specifically, enhanced photosynthesis under eCO₂ increases the flux of labile carbon into the rhizosphere via root exudation, fundamentally reshaping ecological selection pressures [10]. While the effects on microbial richness (alpha diversity) are variable, this resource affects community composition across diverse ecosystems [9,11]. This process often favors fast-growing, copiotrophic taxa (e.g., phyla Bacteroidetes, Proteobacteria) that can rapidly capitalize on these new resources [11,12,13]. Given that the soil microbiome is the primary reservoir of ARGs [14], this predictable, plant-driven shift in the host community composition necessitates a critical re-evaluation of the soil resistome’s stability and potential risk under future climate scenarios.

ARGs are widely recognized as emerging contaminants, with soil serving as a major environmental reservoir [15]. The collective ARG repertoire in environmental compartments, termed the environmental resistome, functions as a critical interface between environmental and clinical antibiotic resistance [16,17]. The dissemination of these environmental ARG reservoirs through agroecosystems poses a distinct toxicological relevance, with potential consequences for food safety and human exposure via the soil–plant–food chain. While vertical gene transfer ensures the stable inheritance of ARGs within lineages, the rapid, cross-species dissemination of ARGs is primarily driven by horizontal gene transfer. This process is mediated by mobile genetic elements (MGEs), such as plasmids, transposons, and integrons [18]. Notably, environmental stressors, including heavy metals, and climate factors can co-select for antibiotic resistance through shared genetic mechanisms (e.g., multidrug efflux pumps, integrons) or cross-resistance pathways [19]. Studies have shown that eCO₂ can not only promote the conjugation and transfer of ARGs by regulating cell surface characteristics and plasmid transfer-related genes [20] but also might reshape the soil antibiotic resistome by modulating the microbial community via changes in plant–soil interactions and biogeochemical cycles [21]. However, the net effect of eCO₂ on the soil ARGs remains ambiguous, with existing studies presenting a complex picture. For instance, previous studies suggested that eCO₂ can inhibit the proliferation of specific ARGs under organic fertilization or sulfadiazine pressure [21,22]. In contrast, Qiu et al. [23] found that eCO₂ did not exert a significant impact on the abundance of efflux pumps genes. This complexity is further amplified by co-contaminants. For example, in a multi-antibiotic contaminated paddy soil, eCO₂ decreased sulfonamide and tetracycline resistance genes while simultaneously increasing multidrug resistance genes [24]. Furthermore, eCO₂ has been reported to enrich the relative abundance of seed ARGs and MGEs under conventional agricultural management [25]. Nevertheless, there is limited information about how eCO₂ independently modulates the soil antibiotic resistome in conventionally managed agricultural ecosystems, hindering our ability to accurately assess future environmental health risks.

Rice paddy ecosystems are crucial to the food security of more than half of the world’s population. In this study, we conducted a field-based pot experiment using the China FACE (free-air CO₂ enrichment) platform to investigate the effects of elevated CO₂ on rice productivity, soil physicochemical properties, microbial communities, and soil ARG profiles under chemical fertilizer application. Specifically, we aimed to (1) assess the variation of soil ARGs under eCO₂; (2) determine the impact of eCO₂ on soil bacterial community, rice growth and soil properties; and (3) explore the potential main factors shaping soil ARGs under eCO₂.

2. Materials and Methods

2.1. FACE System and Pot Experiment

The experiment was performed at the FACE platform situated in Zongcun village, Jiangsu Province, China (119°42′ E, 32°35′ N). The local climate is subtropical monsoonal, averaging 980 mm precipitation and 15 °C annually. The platform design, which has been described previously [26,27], included three FACE plots (eCO₂, aCO₂ + 200 ppm) and three ambient control pots (aCO₂, ~390 ppm, reflecting the current local concentration). Each eCO₂ plot was encircled by a 12.5 m diameter octagonal tube that released pure CO₂ gas over the rice canopy during the growth period. The aCO₂ plots received no CO₂ supplementation and were separated from the FACE plots by a distance of 90 m. The CO₂ release was modulated dynamically according to real-time wind direction and speed data, maintaining a stable elevated concentration of ~200 ppm above ambient levels.

The soil utilized in this pot experiment was identified as Shajiang–Aquci–Cambosols, collected from the upper 20 cm of the adjacent agricultural land. Its properties are detailed in Table S1. The collected soil was prepared by air-drying and then grinding it to a particle size of less than 5 mm. Each pot was filled with 4 kg soil. Rice seedlings (Oryza sativa L. cv. Wuyunjing 23) were transplanted into each pot on 22 June and harvested at the end of October 2017. Chemical fertilizer (N-P₂O₅-K₂O = 15-15-15) was applied at the following stages: 50% pre-transplanting, 25% tillering, and 25% heading. The total nitrogen content was 22.5 g m⁻². The water management was conducted in accordance with local rice cultivation practices. Briefly, all pots were kept submerged with a water depth of ~3 cm at the seedling stage. Manual wet–dry cycles were implemented from the late tillering stage to the early jointing stage. During this period, soil was allowed to dry naturally, and a ~3 cm water layer was replenished only upon the appearance of soil micro-fissures. Then all pots were re-submerged to maintain the initial flooding level until ten days prior to harvest. The experimental pots were assigned randomly and evenly distributed in different rings. The rice biomass produced by each pot was determined at harvest. Soil samples were simultaneously collected from each pot and stored at 4 °C for subsequent analysis.

2.2. DNA Extraction and High-Throughput qPCR

Total DNA was extracted from fresh soil samples using a Fast DNA Spin Kit (MP Biomedicals, Santa Ana, CA, USA) and quantified using a NanoDrop ND-1000 spectrophotometer (Wilmington, Waltham, MA, USA). The extracted DNA was stored at −80 °C for subsequent analysis.

Using the SmartChip Real-Time PCR system, a high-throughput quantitative PCR (HT-qPCR) analysis was conducted to quantify the abundance of ARGs and MGEs [14,28]. The 296 primers (285 ARGs, 10 MGEs and one 16S rRNA gene) and operational protocol were identical to those described by Xiang et al. [29]. Triplicate technical repeats and negative controls were utilized in the HT-qPCR assays. Only primers meeting the 90–110% efficiency benchmark were retained. A gene was considered validly detected only when it showed amplification signals in all three replicates with a threshold cycle value below 31 [30]. ARG copy numbers were normalized to 16S rRNA gene abundance to account for variations in microbial biomass.

2.3. Bacterial 16S rRNA Gene Sequencing

Amplification of the bacterial 16S rRNA gene’s V4–V5 domain was achieved using the primers 515F/907R [28]. Purified PCR products were then sequenced on the Illumina MiSeq 300 platform at Majorbio (Shanghai, China). Sequencing data were processed using QIIME (v1.9.1), and operational taxonomic units (OTUs) were clustered at 97% similarity [31,32]. This OTU-based clustering approach was deployed to facilitate direct and robust comparisons with previous historical data generated from long-term global change and FACE experimental platforms. Taxonomy was assigned using the SILVA database. The resulting OTU table was used to analyze community composition and calculate alpha and beta diversity indices. All sequencing data are available in the NCBI SRA database under accession number PRJNA758632 (SRR15686267-268, 270-279, 282-284).

2.4. Soil Physicochemical Properties Analyses

To determine the physicochemical properties, all soil samples were prepared by air-drying and sieving through a 2 mm mesh. The soil pH was measured in a 1:2.5 soil-to-water suspension. Total nitrogen and total phosphorus were quantified using the Kjeldahl and molybdenum-blue colorimetry methods, respectively [33,34]. The concentrations of copper, zinc, lead, and nickel were determined by flame atomic absorption spectrometry (Hitachi Z-2000, Hitachi, Tokyo, Japan) after digestion with HNO₃-HF-HClO₄ (5:3:3, v/v/v) [21]. These metals were selected for their established roles in co-selection of antibiotic resistance [19,35].

2.5. Statistical Analysis

IBM SPSS Statistics v26 was used to calculate data averages and standard deviations. T-tests (p < 0.05) were used to evaluate differences among samples. Linear discriminant analysis effect size (LEfSe) was applied to identify biomarkers between different treatments [36]. The LDA score threshold was set at 2.0, and the Kruskal–Wallis test p-value threshold was 0.05. The ARG distribution was visualized using Krona Charts [37]. The composition of ARG subtypes, along with their relationships to bacterial taxa, was illustrated through heatmaps produced with the aid of the R package “pheatmap”. The network analyses were performed to explore the co-occurrence patterns (Spearman’s correlation coefficient |r| > 0. 8 with p < 0.01) between ARGs and bacterial taxa (OTUs with relative abundance > 0.1%) using R (v4.1.0) with psych package and were then visualized by Gephi 0.9.2 software.

3. Results and Discussion

3.1. Rice Biomass and Soil Physicochemical Properties Subsection

The aboveground biomass and grain biomass in all the pots ranged from 57.0 to 63.4 g and from 34.5 to 37.2 g, respectively (

Table 1. Rice growth parameters and soil properties under ambient and elevated CO₂ levels. Letters a indicate statistically significant differences at p ≤ 0.05.

		Ambient CO2	Elevated CO2
Rice	Aboveground biomass (g pot−1)	57.0 ± 2.4 a	63.4 ± 5.1 a
	Grain weight (g pot−1)	34.5 ± 2.1 a	37.2 ± 2.6 a
	Tiller no.	11.0 ± 0.4 a	12.5 ± 1.2 a
	Height (cm)	61.8 ± 2.8 a	64.0 ± 2.9 a
Soil	pH	5.69 ± 0.07 a	5.57 ± 0.13 a
	Total nitrogen (g kg−1)	1.20 ± 0.05 a	1.17 ± 0.02 a
	Total phosphorus (g kg−1)	0.90 ± 0.02 a	0.91 ± 0.04 a
	Zinc (mg kg−1)	77.5 ± 0.8 a	82.3 ± 1.5 a
	Copper (mg kg−1)	23.3 ± 0.3 a	24.0 ± 0.4 a
	Lead (mg kg−1)	31.5 ± 0.2 a	30.5 ± 0.6 a
	Nickel (mg kg−1)	18.7 ± 0.5 a	18.8 ± 0.5 a

). Although eCO₂ concentrations can enhance rice biomass and yield via the CO₂ fertilization effect [38,39], the magnitude of this response varies significantly among the different cultivars [40]. In this study, eCO₂ led to a non-significant increase in the aboveground biomass, grain biomass, and plant height of the rice. This finding was consistent with Wang et al. [41], who reported no significant eCO₂ effect on rice grain yield, aboveground biomass, or harvest index. The response of the rice growth to eCO₂ was modulated by a variety of factors, such as the cultivars used, the developmental stage, the nitrogen availability and the temperature [39]. A meta-analysis of 20 years of FACE data revealed that eCO₂ reduced head rice percentage by 8%, which led to no increase in head rice yield [39]. The japonica rice cultivar was adopted in this study. Previous studies have shown that japonica rice cultivars exhibit relatively weak responses in growth and yield under elevated CO₂ [40], which may be related to its lower nitrate absorption activity [42]. An elevated CO₂ concentration slightly decreased the soil pH value, but the difference was not significant. In addition, no significant difference was observed in the soil concentrations of total nitrogen, total phosphorus, copper, zinc, lead and nickel between the aCO₂ and eCO₂ conditions. The non-significant changes in soil nutrients and trace elements may partly explain the non-significant difference in rice biomass.

3.2. Soils ARGs and MGEs

Through a high-throughput qPCR analysis, an average total of 54 ARGs and 4 MGEs were detected in all the soil samples (Table S2). The eCO₂ had no significant effect on the number of soil ARGs and MGEs, nor on the abundance of MGEs (Figure 1a). The relative abundance of the detected ARGs in the ambient and elevated treatments was 0.0053 copies/16sRNA and 0.0080 copies/16sRNA, respectively (Figure 1b). The elevated CO₂ significantly increased the relative abundance of soil ARGs, i.e., 1.5 times higher than that in ambient treatment (Figure 1b).

Multidrug and aminoglycoside resistance genes were the most abundant detected ARG types among all the soil samples (Figure 1c,d). Compared to ambient, eCO₂ increased the relative abundance of multidrug and tetracycline resistance genes by 2.8 and 1.4 times, respectively (Figure 1c). As shown in Figure 1d, eCO₂ diminished the relative abundance of blaTEM and vanHB but enriched qacEdelta1 and tetG. Notably, qacEdelta1 was the most abundant gene detected in all the soil samples, accounting for 19.4% of the total ARG abundance (Figure 2). The qacEdelta1 is a significant marker for multidrug resistance. The enrichment of qacEdelta1 may be a major contributor to the overall rise in ARG abundance under eCO₂. The influence of eCO₂ on the soil antibiotic resistome may be associated with variations in CO₂ exposure duration, pollutants, and potential microbial hosts in the soil. For instance, eCO₂ enriched multidrug resistance genes in soils contaminated with multiple antibiotics [24]. In contrast, Xu et al. [21] observed that eCO₂ had no significant effect on the abundance of multidrug resistance genes in both chemical fertilized soils and sulfadiazine-contaminated soils. Our study adds a critical dimension, revealing that even under conventional chemical fertilization without direct antibiotic inputs, eCO₂ exerted selection pressure on the soil ARGs, particularly the multidrug resistance genes. Moreover, Li et al. [25] found that eCO₂ can also enrich ARGs and MGEs in rice seeds. Collectively, these findings suggest the net effect of eCO₂ on the background soil resistome may masked or even reversed by dominant local variables, including soil chemistry and specific selective pressures. From a One Health perspective, the enrichment of multidrug resistance genes under eCO₂ is linked to a larger soil ARG reservoir, raising potential food safety concerns for agricultural ecosystems under climate change.

3.3. Soil Bacterial Communities

Elevated CO₂ did not significantly alter the Shannon index of the soil bacterial community (Figure 3a). At the phylum level, the community was dominated by Proteobacteria, Chloroflexi and Actinobacteria, which collectively constituted 56.7–65.5% of the relative abundance (Figure 3b). The LEfse analysis identified distinct taxonomic biomarkers for each treatment. Specifically, the aCO₂ treatment was enriched in c_Saccharimonadia, o_Ignavibacteriales, o_Kineosporiales and f_Roseiflexaceae (Figure 3c,d). The dominance of these taxa aligns with oligotrophic microbial ecological strategies, indicating an adaptation to resource-limited conditions. In contrast, it is widely established that eCO₂ stimulates plant photosynthesis and the consequent inputs of rhizospheric carbon [13,26,43]. This altered resource availability presumably functioned as an agent of environmental filtration, reshaping the ecological assembly through deterministic ecological selection mechanisms that favor copiotrophic lineages capable of utilizing enhanced root exudates. The taxa enriched under eCO₂ included c_Clostridia, o_Clostridiales, f_Methylophilaceae, f_Nocardiaceae, f_Desulfobulbaceae, f_Peptococcaceae, g_Dehalobacter and g_Syntrophus (Figure 3c,d). As dominant members of the paddy soil bacterial community, o_Clostridiales act as primary fermenters, breaking down complex polysaccharides into simpler compounds. Through synergistic partnerships with other microorganisms, g_Syntrophus achieves the catabolism of complex organic substrates. This process yields simple end-products, notably acetate and formate [44,45]. Furthermore, the members of f_Methylophilaceae rely on methylotrophic metabolism, characterized by the utilization of methanol and methylamine as substrates for energy generation [46]. f_Desulfobulbaceae are important sulfate-reducing bacteria with roles in organic matter degradation and nutrient cycling [47]. Collectively, these findings suggest that elevated CO₂ may accelerate the decomposition and turnover of plant-derived soil carbon.

3.4. Association of Soil ARGs with Bacterial Communities

Previous studies have established that the effect of eCO₂ on ARGs is not direct, but it is driven by plant-mediated rhizosphere effects [21,22,24]. This process reshapes the soil carbon-to-nitrogen stoichiometry and the microbial community composition, thereby regulating the abundance of ARG host bacteria and MGEs, which in turn determines the abundance and horizontal transfer potential of ARGs. The proliferation of ARGs is intrinsically linked to the population dynamics of their host bacteria [48]. A Pearson correlation analysis was employed to explore the association between microbial biomarkers and ARGs (Figure 4). The results showed that the microbial biomarkers enriched under eCO₂ were strongly associated with an increased prevalence of ARGs. Specifically, the relative abundances of total ARGs, multidrug ARGs, and tetracycline ARGs were significantly positively correlated with eCO₂-enriched taxa such as c_Clostridia, g_Rhodococcus, f_Nocardiaceae, and g_Dehalobacter. Conversely, these ARG categories were negatively correlated with the biomarkers characteristic of the ambient CO₂ treatment, such as f_Kineosporiaceae and o_SJA_15. Previous studies reported that the class Clostridia and its subordinates were ARGs hosts [49,50]. These results suggest that the ecological shift driven by eCO₂ favors a community with a higher ARGs load.

To further explore the potential microbial hosts of ARGs, a network analysis was performed between the significantly changed ARG subtypes and the bacterial genera (with relative abundances >0.01%, excluding the taxonomically unclassified groups) [48]. A total of 18 bacterial genera were significantly correlated with these changed ARG subtypes (Figure 5). Notably, several genera that were enriched under eCO₂ exhibited positive correlations with specific ARGs (Figure S1), highlighting their potential role as resistance hosts. For instance, qacEdelta1 was positively linked to the eCO₂-associated genera g_Dehalobacter and g_Syntrophus (Figure 5 and Figure S1). g_Dehalobacter can act as a host for multiple ARGs in environmental microbial communities [51]. Similarly, tetG was strongly associated with the eCO₂ biomarker g_Rhodococcus (Figure 5). Members of g_Rhodococcus have been specifically identified as hosts for various ARGs, including those conferring resistance to multiple antibiotics such as aminoglycosides, beta-lactams, tetracyclines, vancomycin, and multidrug resistance genes [52]. In addition, g_Candidatus_Accumulibacter, whose abundance was higher under eCO₂ than that under aCO₂ (Figure S1), was positively correlated with tetG but negatively with blaTEM (Figure 5). A previous study observed that g_Candidatus_Accumulibacter was positively linked to tetracycline resistance genes in the biological nutrient removal process [53]. g_Roseomonas was positively associated with blaTEM but negatively with tetG, which were significantly decreased under eCO₂ (Figure 5 and Figure S1). The bla gene (β-lactamase) is specifically mentioned as present in Roseomonas genomospecies 5, conferring resistance to β-lactam antibiotics [54]. As the detected abundance of MGEs remained unchanged, the proliferation of specific potential host bacteria may be the primary driver of the observed ARG enrichment under eCO₂. Additionally, eCO₂ may promote the increase of ARGs by intensifying the antagonistic relationships between fungi and bacteria [25]. Although elevated CO₂ induced no dramatic shifts in crop biomass, it may cause effects on plant carbon and nitrogen metabolism and belowground carbon allocation, thereby impacting soil microbial communities and their associated ARGs [55,56]. It is worth noting that due to the intrinsic ecological uncertainty of open soil systems and the potential confounding influence of unmeasured variables, these correlation-based links inherently represent shared ecological niches rather than a definitive genetic alignment [57]. Therefore, these connected bacterial taxa are rigorously interpreted as potential hosts rather than confirmed physical carriers. Further efforts using more soil types at a larger temporal scale using various metagenomics surveys will provide a better understanding of the response and mechanisms of ARG transmission under future climate change scenarios.

4. Conclusions

In summary, elevated CO₂ significantly increases the abundance of ARGs in fertilized paddy soil, despite having no significant effect on rice biomass. This ARG proliferation was primarily driven by a 2.8-fold increase in multidrug resistance genes. The correlation and network analyses revealed that the ARG enrichment was driven by a distinct shift in the soil bacterial biomarkers. Specifically, eCO₂ selectively promoted the proliferation of potential ARG hosts such as Clostridia, Rhodococcus, and Dehalobacter. These findings underscore the potential amplification risk of soil multidrug resistance genes under rising CO₂ levels. The spread of antimicrobial resistance in agroecosystems under climate change may extend the threats to global food safety, emphasizing the necessity of ARG assessment in plants and grains. Incorporating mitigation strategies for these dissemination risks into future agricultural policies is essential to protect global public health.