Association between new onset type 1 diabetes and real-world antibiotics and neonicotinoids’ exposure-related gut microbiota perturbation

This multi-center cross-sectional study collected samples from eight cities in different regions of China, including Fuzhou, Nanchang, Shanghai, Suzhou, Jinan, Taiyuan, Changchun, and Zhengzhou. A total of 70 newly diagnosed T1D patients < 18 years old with disease duration < 1 month were included in this study during the study period from January 2019 to March 2020. The diagnosis of T1D was based on the criteria of the International Society for Pediatric and Adolescent Diabetes [15]. The healthy control group was matched with the T1D group by region, sex, age, and time of visit. Cases were excluded if they were < 2 years old, received antibiotic treatment within one month, and presented with infectious diseases, chronic or acute gastrointestinal diseases, other inherited metabolic disorders, and/or serious chronic disorders (Fig. 1). This study was approved by the Ethics Committee of Children’s Hospital of Fudan University ([2019]210), and the guardians of the children provided written informed consent before information and sample collection.

The clinical characteristics including sex, age, time of visit, fasting blood glucose, HbA1c, and serum C-peptide of T1D patients and healthy control children were obtained through questionnaires administered by the doctors. Fecal samples and first morning urine samples were collected in dedicated sterile containers provided by the research team. Trained medical staff guided and supervised sample collection. Fecal and urine samples were frozen in freezers immediately after collection. Frozen fecal and urine samples were transported on dry ice to the Children’s Hospital of Fudan University within 24 hours of collection and were stored at − 80 °C until analysis.

Total bacterial DNA was extracted using a DNA extraction kit according to the manufacturer’s instructions (DNeasy PowerSoil Kit). DNA was diluted to 1 ng/μL and stored at − 20 °C until further processing. The barcoded primers and Takara Ex Taq reagent (Takara, Dalian, China) were used for polymerase chain reaction (PCR) amplification of bacterial 16S rRNA genes. The 16S rRNA V3-V4 region was amplified with universal primers 343F and 798R for bacterial diversity analysis. After purification, the final amplicon was quantified with a Qubit double-strain DNA (dsDNA) assay kit. Equal quantities of purified amplicon were pooled to prepare for subsequent sequencing. After removing low-quality sequences, paired-end reads were assembled using the fast ligation-based automatable solid-phase high-throughput system with a minimum and maximum overlap of 10 bp and 200 bp, respectively, and a 20% maximum mismatch rate [16]. The 16S rRNA sequencing data were analyzed using QIIME software (v.1.8.0), and sequences were assigned to operational taxonomic units using Vsearch software (97% similarity cut-off) [17, 18]. Representative reads were annotated and subjected to a BLAST search against the Silva database (v.123.0) using the Ribosomal Database Project classifier with a 0.70 confidence threshold [19].

The selection of antibiotics was mainly based on previous studies [20]. Briefly, 28 antibiotics and five categories of metabolites (macrolide, tetracycline, fluoroquinolone, sulfonamide, and phenicol) were analyzed. All antibiotics were divided into three categories according to use: veterinary antibiotics (VA), human antibiotics (HA), and veterinary/human antibiotics (V/HA). The types of antibiotics that children can use are limited (e.g., fluoroquinolone and tetracycline are restricted for use in children); therefore, most antibiotics in the V/HA group were more likely to be exposed to children through food and water. Thus, VA + V/HA concentrations can reflect exposure to antibiotics in addition to medication. Concentrations of antibiotics were measured by isotope dilution ultra-performance liquid chromatography coupled to quadrupole time-of-flight mass spectrometry (UPLC-Q/TOF MS) according to previously described methods [20].

Briefly, eight neonicotinoids and four typical metabolites were detected according to previous studies, with isotope dilution UPLC-Q/TOF MS specifically used for neonicotinoid measurement [21].

The concentrations of urine antibiotics and neonicotinoids were adjusted by the urine creatinine levels (Supplementary Table 1). In case of underreporting of the history of antibiotic use in the previous month, we removed data for patients with urine creatinine-adjusted HA concentrations over the 95th percentile. Ultimately, 10 T1D cases and 8 control cases were excluded. Once a measurable value of one kind of antibiotic or neonicotinoid is detected in urine, the child was considered to be exposed to that antibiotic or neonicotinoid. Children were grouped according to the kinds of antibiotics, VA + V/HA or neonicotinoids they were exposed to (no exposure, exposure to one kind, or exposure to two or more kinds). To analyze the combined effect of antibiotics and neonicotinoids exposure, children were also grouped into four groups according to whether they were exposed to one or more kind of antibiotics and/or neonicotinoids (the no ANTI & no NEO group for children without exposure to antibiotics or neonicotinoids; the ANTI & no NEO group for children only exposed to one or more kind of antibiotics; the no ANTI & NEO group for children only exposed to one or more kind of neonicotinoids; and the ANTI & NEO group for children exposed to both one or more kind of antibiotics and one or more kind of neonicotinoids). Furthermore, the association of the urine total concentrations of antibiotics, VA + V/HA or neonicotinoids with clinical characteristics and gut microbiota structure were also analyzed.

Statistical analysis was performed using R software (https://​www.​r-project.​org/​, version 4.3). For clinical variables, we used χ² tests to compare categorical variables. Student’s t test was used to compare differences between two groups for normally distributed continuous variables, and the Mann–Whitney U test was used for variables that were not normally distributed. The indices of α-diversity and β-diversity were calculated in R using the “vegan” and “mixOmics” packages using permutational multivariate analysis, principal component analysis and sparse partial least-squares discriminant analysis. To determine specific taxa, linear discriminant analysis (LDA) effect size was calculated online (http://​huttenhower.​sph.​harvard.​edu/​galaxy/​). A P < 0.05 was considered significant.

Among 51 T1D children and 67 healthy control children, 22 (43.1%) and 28 (41.8%) were female, respectively (P = 0.883), with a mean age of 7.51 ± 3.61 years and 8.10 ± 2.91 years (P = 0.325) (Table 1). The T1D group displayed severely abnormal glucose metabolism, with higher fasting blood glucose levels [17.21 mmol/L (10.55–24.60 mmol/L) vs. 4.90 mmol/L (4.72–5.16 mmol/L)] (P < 0.001) and HbA1c levels [11.80% (10.30–13.40%) vs. 5.10% (4.88–5.43%)] (P < 0.001) than the control group. The C-peptide concentration of the T1D group was 0.21 μg/mL (0.11–0.40 μg/mL).

Among the five antibiotic categories, fluoroquinolone showed the highest detection frequency at 39.8% (47/118), followed by macrolide (24.6%, 29/118), sulfonamide (12.7%, 15/118), tetracycline (10.2%, 12/118), and phenicol (8.5%, 10/118). The T1D group and the control group showed similar detection rates across categories.

Children who were exposed to one kind of antibiotics were at higher risk of T1D compared to children without antibiotics exposure [odds ratio (OR) = 2.579, 95% confidence interval (CI): 1.061–6.271; P = 0.037] (Supplementary Table 4). Furthermore, children who were exposed to two or more kinds of neonicotinoids also presented higher risk of T1D (OR = 3.911, 95% CI: 1.538–9.945; P = 0.004]. When analyzing the combined effect of antibiotics and neonicotinoids, we found that children who were exposed to both one or more kinds of antibiotics and one or more kinds of neonicotinoids had higher risk of T1D than those who were not exposed to antibiotics or neonicotinoids, with the odd ratio of 4.924 (95% CI 1.239–19.572; P = 0.024) (Supplementary Table 4).

Moreover, fasting blood glucose levels were elevated according to increases in the urine creatinine-adjusted neonicotinoids concentration (r = 0.436, 95% CI 0.720–1.853; P < 0.001), with these levels similar only when considering T1D cases (r = 0.369, 95% CI 0.201–1.393; P = 0.010) (Fig. 2). Among the neonicotinoids-positive T1D cases, the onset age was significantly negatively correlated with the urine creatinine-adjusted neonicotinoids’ concentration (r = − 0.398, 95% CI − 0.580 to − 0.063; P = 0.026), whereas the urine creatinine-adjusted antibiotics concentration showed no significant relationship with T1D-onset age (P = 0.166) or fasting blood glucose (P = 0.848).

The T1D group showed a characteristic gut–microbiota structure. Both the Shannon index (P = 0.016) and Chao1 index (P = 0.007) were significantly decreased in the T1D group relative to the control group (Supplementary Fig. 1a, b). Permutational multivariate analysis of variance indicated significant differences in β-diversity between the T1D and control groups (P = 0.032, R² = 0.022) (Supplementary Fig. 1c).

At the phylum level, the T1D group showed a significantly lower abundance of Firmicutes (P = 0.001) and a higher abundance of Proteobacteria (P = 0.001) relative to the control group. Furthermore, the T1D group was characterized by a decrease in butyrate-producing genera within Lachnospiraceae and Ruminococcaceae (i.e., Faecalibacterium, Agathobacter, Lachnospira, Roseburia, and Blautia) and an increase in opportunistic pathogens within Enterobacteriaceae (Escherichia and Shigella) (Supplementary Fig. 2).