We used data and samples collected in 2003–2006 from the SEARCH-CC study, an ancillary study to the multicenter study of diabetes in U.S. youth (Dabelea et al. 2011; Hamman et al. 2014). For POP analyses, we used serum samples that were collected following fasting and where triglyceride values were obtained. SEARCH-CC participants who were residents of Colorado or South Carolina, aged 10–22 years, were recruited from primary care offices, as described previously (Guy et al. 2009; Snell-Bergeon et al. 2010). Among SEARCH-CC participants, youths were classified as controls (no diabetes), recently-diagnosed T1D/IS, or recently-diagnosed T1D/IR (Dabelea et al. 2011). The nondiabetic status of controls was confirmed by normal fasting glucose values, normal insulin sensitivity, and absence of T1D autoantibodies. T1D was determined based on clinical diagnosis and medical record review and confirmed based on the positivity of one of three specific T1D autoantibodies measured; glutamic acid decarboxylase antibodies (GADA), insulinoma-associated 2 antibodies (IA-2A), and zinc transporter 8 antibodies (ZnT8) (Crume et al. 2020). For insulin sensitivity assessment, a clinical algorithm validated against hyperinsulinemic-euglycemic clamps was used with a cut-point of 8.15 for determining the state of insulin resistance (<8.15: insulin resistance, ≥ 8.15: insulin sensitive) (Dabelea et al. 2011).

Plasma samples from SEARCH-CC participants were shipped in frozen condition to the Laboratory for Analysis of Environmental Pollutants at the Department of Laboratory Medicine, University Hospital of North Norway, Norway, and stored at ≤ −20 °C prior to analysis for organochlorine pesticides and PCBs (Table S1). Fully validated methods were applied for sample preparation and instrumental analyses as described previously (Huber et al. 2020). A Tecan Freedom Evo 200 liquid handler (Männedorf, Switzerland) equipped with a solid-phase extraction station, a robotic manipulator arm and a shaker was used for sample preparation. Before extraction and clean-up by automated solid-phase micro-extraction, a 150 μL plasma sample was diluted and spiked with an internal mass-labelled standard solution. Instrumental analyses were performed by gas chromatography with atmospheric pressure ionisation coupled to tandem mass spectrometers (Waters, Milford, MA, USA). The atmospheric pressure ionization was conducted in positive mode under charge transfer conditions. For detection on the mass spectrometers, multiple reaction monitoring (MRM) mode with two specific transitions for the individual analytes was applied.

Masslynx and Targetlynx software (Version 4.1, Waters, Milford, MA, USA) was used for quantification by application of the isotope dilution method. For quality assurance, four blank samples, four SRM 1957 and SRM 1958 samples (both NIST, Gaithersburg, MD, USA), and three bovine serum samples (Sigma Aldrich, Steinheim, Germany) were analyzed within each batch of 96 samples to control for background and carry-over effects. The quality controls were within acceptable limits and similar to our previous data (Huber et al. 2020) with coefficients of variation (CVs) ranging from 4 to 28% for the present study. The laboratory participates successfully in the international quality control programme Arctic Monitoring and Assessment (AMAP) Ring Test for Persistent Organic Pollutants in Human Serum, organized by the Laboratoire de toxicologie, Institut National de Santé Publique du Quebec, Canada, with a satisfying performance reaching z-scores within the acceptable range of −2 > z-score < 2.

Rat insulinoma INS-1E cells (Merglen et al. 2004) were cultured in a humidified atmosphere with a 95% air / 5% CO₂ mixture and maintained at 37 °C as described previously (Ghiasi et al. 2019). Briefly, cells were grown in 11 mM glucose in GlutaMAX Roswell Park Memorial Institute (RPMI)-1640 medium (Life-Technologies, United Kingdom) supplemented with 10% fetal bovine serum, 100 IU/mL penicillin and 100 ug/mL streptomycin, 0.05 μM β-mercaptoethanol, 1 mM sodium pyruvate, and 10 mM hydroxyethyl piperazineethanesulfonic acid. At 90–95% confluence, cells (150,000 cells seeded in 24-well plates) were cultured with 11 mM glucose in RPMI-1640 medium for 45 h and treated with dimethyl sulfoxide (DMSO), p,p’-DDE (MW: 318.03, Sigma-Aldrich, Switzerland) or PCB-153 (MW: 360.88, Sigma-Aldrich, Switzerland) at environmental concentrations; 1 × 10⁻¹⁵ M (1 fM), 1 × 10⁻¹² M (1 pM), 1 × 10⁻⁹ M (1 nM), 1 × 10⁻⁶ M (1 μM), and 5 × 10⁻⁶ M (5 μM). In all treatments, the maximal concentration of DMSO did not exceed 0.025%, a dose without any effect on cell viability (data not shown). In addition, TNFα (Novus Biologicals, USA) was used as a positive control (Kim et al. 2008). After 45 h, treated cells (DMSO, POPs or TNFα) were grown in RPMI-1640 medium containing 2 mM glucose for 1 h. Then, treated cells (DMSO, POPs or TNFα) were cultured with 20 mM glucose in RPMI-1640 medium for 2 h to stimulate insulin production. Thereafter, INS-1E cells were collected for assessment of intracellular insulin content and mRNA expression whereas culture media was used to measure glucose-stimulated insulin secretion. In total, cells were exposed to DMSO, PCB-153, p,p’-DDE, or TNFα for 48 h. In another set of experiments with similar treatments, INS-1E cells were plated in a 96-well plate (50,000 cells per well) and cultured with 11 mM glucose in GlutaMAX RPMI-1640 medium for 8 days to assess cell viability.

Secreted and intracellular insulin was measured using a Rat insulin ELISA (#10-1250-01, Mercordia, Sweden) from culture media and INS-1E cells, respectively. All results were normalized to protein content measured by Pierce BCA Protein Assay Kit (Thermo Scientific, USA). Samples were analyzed in duplicates for both ELISA and BCA analyses, and the average of these technical duplicates was used.

Total RNA was prepared from INS-1E cells using the Invitrogen total RNA isolation kit (Life-technologies, USA). Complementary DNA generated with a high-capacity cDNA Reverse transcription kit (Thermo Scientific, Lithuania) was analyzed by quantitative real-time qPCR using the BioRad iQ^™SYBR^® Green supermix (CFX Maestro Version 5.0.021.0616, BioRad, USA). Efficiency corrected Cq values were normalized to Tbp expression and expressed as fold change relative to DMSO-treated cells. Oligonucleotide primers are listed in Table S2.

Cell viability was determined using the AlamarBlue^™ HS cell viability reagent (Thermo Scientific, USA) according to the manufacturer’s instructions.

The characteristics of participants are shown in Table 1. Among 442 youths, 112 youths had no diabetes (controls) whereas 330 youths had T1D; 182 and 148 youths with T1D had normal insulin sensitivity (T1D/IS) or insulin resistance (T1D/IR), respectively (Table 1). The mean age was 13.5 and 16.1 years for T1D/IS and T1D/IR, respectively (Table 1). As expected, youths with T1D/IS had a significantly lower BMI z-score, waist circumference, circulating lipids, and HbA1c, as compared with T1D/IR participants (Table 1). The duration of T1D was also longer in T1D/IR participants than in youths with T1D/IS (Table 1).

Next, we analyzed the plasma concentrations of different organochlorine pesticides and PCBs in the participants. Among organochlorine pesticides, p,p’-DDE, and hexachlorobenzene were detected in almost every participant regardless of case or control status, followed by trans-nonachlor (78% detection rate) (Table 2 and S3). Among PCBs, PCB-28 and PCB-153 were the most prevalent compounds with a detection rate close to 80% (Table 2 and S3).

After adjustment for confounders (age at visit when the sample was collected, health insurance, parental education, race, ethnicity, sex, pubic Tanner stage, and study site), there was a significant association between T1D/IS and several POPs including p,p’-DDE, trans-nonachlor and PCB-153 (Table 2). The odds of T1D/IS were significantly increased among youths in the 2nd (OR 2.0, 95% CI 1.0, 3.8) and 3rd tertiles (OR 2.4, 95% CI 1.2, 5.0) of serum p,p-DDE levels. Similarly, both the 2nd and 3rd tertiles of serum trans-nonachlor levels were associated with a significant elevation in the odds (OR 2.5, 95% CI 1.3, 5.0 and OR 2.3, 95% CI 1.1, 5.1 respectively). For serum PCB153 levels, the odds of T1D/IS were significantly increased among youths in the 3rd tertiles (OR 2.3, 95% CI 1.1, 4.6). Among POPs with detection rates between 20 and 70%, only the 3rd tertile of serum p,p’-DDT levels showed a significant increase of the odds for T1D/IS (OR 2.1, 95% CI 1.0, 4.3) (Table S4). Of note, these positive associations were not observed for youths with T1D/IR, and only heptachlor was positively associated with T1D/IR (OR 2.1, 95% CI 1.0, 4.7 for 2nd tertile) (Table 2 and Table S4). Analyses with non-lipid adjustments for POPs provided similar results (Table S5).

To better understand the impact of POPs on β-cells, we investigated whether POPs could directly regulate the production of insulin in an experimental model using INS-1E cells, a widely used β-cell model in pancreatic research (Merglen et al. 2004). To determine whether POPs from different categories could trigger different effects on INS-1E cells, we selected one PCB and one organochlorine pesticide compound. We focused our investigations on PCB-153 and p,p’-DDE because both compounds had the highest OR for T1D/IS and were highly detectable in youths (Table 2). To mimic environmental doses of exposure, we treated cells with PCB-153 and p,p’-DDE at concentrations ranging from 1 × 10⁻¹⁵ M to 5 × 10⁻⁶ M. First, we explored whether PCB-153 and p,p’-DDE could affect the two insulin genes Ins1 and Ins2, which are both involved in the production of insulin in INS-1E cells. Interestingly, both PCB-153 and p,p’-DDE robustly reduced the mRNA expression of Ins1 and Ins2 even at concentrations as low as 1 × 10⁻¹⁵ M (Fig. 1A and B). Next, we tested whether this decrease in insulin gene expression translated to a reduction in intracellular insulin level. For this, we assessed the insulin content of INS-1E cells using an insulin-specific ELISA kit. In line with the mRNA expression results, PCB-153 strongly reduced the content of insulin in INS-1E cells (Fig. 1C) whereas this effect was less pronounced following treatment with p,p’-DDE (Fig. 1C). Then, we investigated the ability of the pancreatic β-cells to secrete insulin in response to glucose. By assessing the insulin concentrations present in the culture media, we observed that the release of insulin was significantly reduced in cells exposed to PCB-153 at all tested concentrations (Fig. 1D) whereas p,p’-DDE had a marked effect only at the lowest and highest concentrations (Fig. 1D). When insulin secretion was assessed relative to intracellular insulin content (Fig. 1E), we found that PCB-153 treatment resulted in a constant ratio for most of the tested concentrations. This indicates that the observed decrease in intracellular insulin was accompanied by a similar decrease in insulin secretion in the cells. For p,p’-DDE treated cells, we observed a decreased ratio between secreted and intracellular insulin suggesting that the cells were less efficient at secreting the produced insulin compared to DMSO-treated cells (Fig. 1E). No effect on β-cell viability was observed with PCB-153 and p,p’-DDE treatment at the examined concentrations for 2 days (Fig. 1F).

To better understand the mechanistic modes of action of PCB-153 and p,p’-DDE, we assessed the mRNA expression of the glucose transporter GLUT2 (encoded by Slc2a2) and glucokinase (encoded by Gck), often referred to as “the glucose sensor”, with a critical role in the initiation of glucose-induced insulin secretion in pancreatic β-cells (Matschinsky and Wilson 2019). Interestingly, PCB-153, but not p,p’-DDE, reduced the expression level of Slc2a2 (Fig. 2A) and Gck (Fig. 2B). Further, we studied the pancreatic duodenal homeobox 1 (PDX1) and V-maf musculoaponeurotic fibrosarcoma oncogene homolog A (MAFA), which are two transcription factors for insulin gene expression and pancreatic cell development (Zhu et al. 2017). PCB-153 significantly increased the expression level of Pdx1 at all doses whereas similar effect was only observed with the lowest and highest dose of p,p’-DDE (Fig. 2C). No significant effect on Mafa expression was found, except for p,p’-DDE at the lowest dose (Fig. 2D). Finally, we observed no major impact of PCB-153 or p,p’-DDE on mRNA levels of Pcsk1/3 or Pcsk2, two proteins converting proinsulin into insulin (Fig S2A and B).

Next, we assessed genes involved in the regulation of insulin secretion. We investigated the ATP-sensitive potassium channel regulating insulin secretion, which is composed of two distinct subunits, a regulatory subunit, the sulfonylurea receptor-1 (SUR1, encoded by Abcc8) and an inwardly rectifying ion channel forming the pore (KIR6.2, encoded by Kcnj11). Of note, p,p’-DDE treatment caused a significant reduction in Abcc8 expression at all doses (Fig. 2E), which was associated with an up-regulation of Kcnj11 (Fig. 2F). In contrast to p,p’-DDE, PCB-153 reduced the mRNA expression of Abcc8 at the lowest doses only (Fig. 2E), without any increase of Kcnj11 expression (Fig. 2F). We next asked whether PCB-153 and p,p’-DDE could impair voltage-gated calcium channels (CAV). To this end, we focused on CAV1.3 (encoded by Cacna1d) and CAV2.2 (encoded by Cacna1b) and found that neither PCB-153 nor p,p’-DDE affected their expression (Fig S2C and D). Finally, we investigated genes involved in exocytosis like Stx1a, Snap25, and Sytl4 and found no significant impact of PCB-153 and p,p’-DDE (Fig S2E–G).

The destruction of pancreatic β-cells is a critical feature of T1D, for which the etiology remains poorly known (Roep et al. 2021). We therefore asked whether a chronic treatment with PCB-153 or p,p’-DDE could affect the viability of β-cells. To this end, we performed a follow-up study of INS-1E cells and assessed their viability every 2 days. Confirming our previous data (Fig. 1F), treatment with PCB-153 and p,p’-DDE for 2 days did not affect cell viability at 1 × 10⁻¹⁵ M, 1 × 10⁻⁹ M, 5 × 10⁻⁶ M, and 50 × 10⁻⁶ M (Fig. 3A and B). However, when cells were exposed to a higher dose (100 × 10⁻⁶ M) of p,p’-DDE, cell viability was significantly reduced by about 20% (Fig. 3B). By day 6, PCB-153 and p, p’-DDE at 50 × 10⁻⁶ M and 100 × 10⁻⁶ M decreased the viability of β-cells by 25–60% (Fig. 3A and B). After 8 days of exposure, almost all PCB-153 and p,p’-DDE concentrations caused a significant destruction of β-cells (Fig. 3A and B).

Major strengths of our study include the comprehensive phenotyping of a large number of youths with T1D, as well as the combination of human and in vitro studies, which allows a better determination of the causative role of POPs on T1D and the mechanisms involved. While we demonstrated the impacts of some PCBs and organochlorine pesticides on T1D, whether other PCBs and organochlorine pesticide compounds and POPs, like polybrominated diphenyl ethers (PBDEs) and perfluoroalkyl substances (PFAS), contribute to T1D in youths remain to be investigated. One potential limitation to our study is the fasting human samples, which may miss manifestation of dysglycemia in the participants (no provocative tests such as OGTT were performed). However, this issue is likely limited given the young age of the cohort.