Pediatric patients with T1D were initially divided into two age groups: ≤ 6-year-old and > 6-year-old. Two control subgroups (Con1/2; n = 6 each) of healthy volunteers matched for age, sex, body mass index (BMI), and ethnicity were also included ( Table 1 ). The detailed experimental design is illustrated in Figure 1A . The two age groups showed no significant differences in age of T1D diagnosis, BMI, sex, and systolic or diastolic blood pressure. Additionally, the time from symptom onset to T1D diagnosis between the two age groups or within the subgroups were comparable, and none of the patients had complications, such as kidney or eye damage. Given that genetic factors may play an important role in T1D development, we noted that 5 out of 12 T1D patients aged ≤ 6 year-old and 6 out of 12 T1D patients aged > 6 year-old had a family history of diabetes.

Each age group was then further categorized into two subgroups on the basis of HbA1c% status: T1D-1/2 (HbA1c% > 7.0%; n = 6) and INT-1/2 (HbA1c% < 7.0% following ≥ 3-mo insulin treatment; n = 6). Among the ≤ 6-year-old subgroups, there was a mild difference in HbA1c% values between the Con-1 (5.63 ± 0.24%) and INT-1 (6.22 ± 0.21; P = 0.312) subgroups, but both were significantly lower than that of the T1D-1 subgroup (10.95 ± 0.60; P < 0.001). Among the > 6-year-old subgroups, the HbA1c% values were comparable between the Con-2 (5.02 ± 0.13%) and INT-2 (6.30 ± 0.15%; P = 0.122) subgroups; again, both were significantly lower than that of the T1D-2 subgroup (13.60 ± 0.94; P < 0.001) ( Figure 1B ).

To improve the detection and quantification of exosomal proteins, two TMT6-labeled replicates were produced ( Supplementary Figure 1 ). Totally, we identified 1868 proteins across all six subgroups, listed in Table S1 . After removing duplicates and unnamed proteins, we cross-referenced the remaining 1035 proteins using exosomal protein database of Vesiclepedia and Exocarta, respectively (25, 26). There were 570 (55.1%) or 660 (63.8%) overlapping proteins, which included known exosome markers CD9 and heat shock protein (HSP)70 (27, 28), and 50/52 proteins were among the top 100 proteins identified in exosomal preparations, confirming the successful isolation and identification of exosomes. We also detected other unannotated proteins, comprising immunoglobulins, complement factors, lipoproteins, and other exosomal proteins, with potential specificity for pediatric T1D ( Figure 1C ).

Cluster analysis was also performed on the 1035 known proteins, revealing significant variation in the expression patterns among the subgroups in the two age groups, as depicted by the cluster areas within dotted black boxes in Figure 1D . These results suggested that the protein expression patterns in T1D are age-specific.

First, we compared exosomal proteome expression in the ≤ 6-year groups. Compared to Con-1, T1D-1 showed differential expression of 558 proteins (413 upregulated and 145 downregulated). The top 10 were RAB40A, WDR62, and COL6A5, etc ( Figure 2A ). Upregulated proteins were related to the immune system, and downregulated proteins were linked to the cellular stress response and extracellular matrix (ECM) organization ( Figure 2B ).

Next, we found 562 DEPs identified in comparison of INT-1 and T1D-1, which were apparently sensitivity to insulin treatment, including RAB40A, PDZD2, CSTA, IRF1, and COL6A5 ( Figure 2C ). Downregulated proteins were enriched in the immune system, while upregulated proteins were enriched in gene expression regulation (as the epigenetic regulations: DNA methylation, histone methylation and acetylation regulation, etc.) and cell cycle pathways (as the packaging of telomeres during cell division) ( Figure 2D ), suggesting increased cell cycle activation after insulin treatment.

Venn analysis of the aforementioned comparisons revealed 164 (112 + 52) proteins that were expressed in T1D-1 but not in INT-1, indicating that they were insulin treatment-insensitive. Insulin treatment appeared to induce 168 (68 + 100) new DEPs identified in INT-1, which were mainly enriched in the cell cycle and gene expression regulation ( Figure 2E , Supplemental Figures 2C, D ). Additionally, compared to Con-1, of the 413 up-regulated proteins found in T1D-1, 301 were down-regulated in INT-1. Similarly, of the 145 down-regulated proteins detected in T1D-1, 93 were up-regulated in INT-1. In total, these two comparisons share 394 (301 + 93) DEPs. Trend analysis of these proteins revealed that patients with HbA1c > 7% showed enhanced expression of 251 proteins (Profile 5: the higher of the protein expression level, the more severe of the disease) and decreased expression of 84 proteins (Profile 2: the lower the protein expression level, the more severe the disease). However, these changes in protein expression were mitigated control of HbA1c at < 7% (i.e., the INT-1 subgroup) ( Figure 2F ). Figures 2G, H illustrates the top 10 DEPs in Profiles 5 and 2, respectively. Furthermore, reactome pathway analysis indicated that these proteins were mainly enriched in pathways involving the immune system, cellular stress response, and ECM ( Figure 2I ).

Similar to the ≤ 6-year group, we initially compared T1D-2 with Con-2, which identified 588 DEPs, comprising 333 upregulated and 255 downregulated proteins in T1D-2. The top DEPs included COL6A5, KRT83, NEB, and TRIM29 ( Figure 3A ). Pathway analysis suggested that these DEPs were mainly enriched in hemostasis and immune system pathways ( Figure 3B ).

We further compared INT-2 with T1D-2, noting 396 DEPs. Proteins such as COL6A5, KRT83, NEB, SEMA6D, and TPM3 were apparently sensitive to insulin treatment, and were predominantly enriched in immune system and cellular stress response pathways ( Figures 3C, D ).

Combining the two aforementioned comparisons ( Figure 3E ), we identified 320 (162 + 158) proteins that were expressed in T1D-2 but not in INT-2, and were mostly enriched in homeostasis, ECM, gene expression, and signal transduction. There were 128 (50 + 78) proteins with specific differential expression in INT-2, suggesting that changes in their expression were triggered by insulin treatment. These DEPs were primarily enriched in the immune system, cellular stress response, vesicle-regulated transport, and disease ( Supplemental Figures 3C, D ).

Finally, we focused our analysis on the 268 (171 + 97) DEPs identified in both comparisons. Trend analysis indicated that nearly half (127 of 268) exhibited expression patterns belonging to Profile 5, whereas 28 belonged to Profile 2 ( Figure 3F ). Figures 3G, H illustrates the top 10 proteins in Profiles 5 and 2, respectively, which included COL6A5, KRT83, NEB, SEMA6D, TP53I3, and TTR. Furthermore, altered proteins in Profiles 5 and 2 were predominantly enriched in pathways involving the immune system, developmental processes, and signal transduction ( Figure 3I ). Overall, these results highlighted the role of differential proteins and associated pathways in the insulin therapy response and T1D pathophysiology in patients over 6 years of age.

We conducted a Venn analysis to determine whether there were common or unique protein expression patterns in the two age groups. Among the 558 DEPs identified in the comparison of T1D-1 vs Con-1 in the ≤ 6-year-old and the 588 DEPs identified in the comparison of T1D-2 vs Con-2 in the > 6-year-old group, there were 304 overlapping proteins ( Figure 4A ). Furthermore, 254 proteins showed age-specific differential expression only in the ≤ 6-year-old T1D-1, whereas 284 proteins showed age-specific differential expression only in the > 6-year-old group. Pathway analysis of these three sets of DEPs showed that they were mainly enriched in pathways involved in hemostasis, ECM organization, vesicle-mediated transport, and the innate immune system ( Figure 4B ). Additionally, Gene Ontology (GO) annotation revealed that the cellular components, biological processes, and molecular functions enriched by the DEPs were similar between the ≤ 6-year-old and > 6-year-old T1D subgroups ( Supplemental Figures 2A, B , 3A, B ). This implied that, although the protein expression patterns varied, the underlying biological pathways and functional roles may be similar between the two age ranges.

We also performed a simultaneous analysis of the 394 DEPs identified in both comparisons of ≤ 6-year-old subgroups and the 268 DEPs identified in both comparisons of > 6-year-old subgroups ( Figures 2E , 3E , 4C ). This revealed that there were 273 DEPs identified in both comparisons of ≤ 6-year subgroups that did not show differential expression in the > 6-year-old subgroups. One such protein was RAB40A. Conversely, there were 147 such proteins unique to the > 6-year-old subgroups, which included SEMA6D. The differing expression patterns of these proteins might be due to children with T1D exhibit age-specific, spatiotemporal protein expression at two different developmental stages. However, there were also 121 DEPs identified in both comparisons of the two age groups ( Figure 4C , Supplemental Figure 4 ). Of these, 49 proteins (Profile 5) showed upregulation in the T1D group patients and downregulation in the INT group patients (e.g. COL6A5), suggesting that their differential expression between T1D subgroups may be age-independent. Another 71 proteins exhibited distinct patterns, such as TTR, which was upregulated in T1D-1 and downregulated in INT-1 in ≤ 6-year-old patients, but downregulated in T1D-2 and upregulated in INT-2 in > 6-year-old patients.

Functional analysis of these 121 proteins using Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway and STRING analyses identified 11 altered proteins enriched in the pancreas (false discovery rate < 0.001), 21 linked to the innate immune system, and 12 belonging to ECM components ( Figure 4D ). Among these, TTR is known to not only originate from the pancreas, but also to participate in the immune system and contribute to ECM establishment, the multifunctionality of TTR suggests that it plays a crucial role in the progression of T1D.

We validated the data with western blotting of RAB40A, SEMA6D, COL6A5, and TTR, which were chosen on the basis of their differential expression in the dataset, assay feasibility, and T1D relevance ( Figure 4E ). RAB40A was highly expressed only in the ≤ 6-year-old T1D-1 subgroup, and it returned to levels comparable to Con-1 and INT-1. Conversely, SEMA6D was overexpressed in both age groups’ T1D-1/2 subgroups, however, insulin treatment cannot effectively reduce its protein expression levels to a degree similar to the Con-l/2 subgroup. Both age groups showed elevated COL6A5 expression levels in the T1D-1/2 subgroups that returned to near-normal levels in the INT-1/2 subgroups. Differences in TTR were evident between the subgroups, with distinct regulatory patterns in each subgroup ( Figure 4E ). These results were mostly aligned with the quantitative TMT6-labeling outcomes.

Finally, we compared our data with findings from several serum proteome studies in T1D patients (29). Table 2 shows the variance in protein expression patterns across these studies. For instance, compared to healthy controls, adiponectin (ADIPOQ) was highly expressed in sera of 1139 T1D patients aged 0~90 years (30), but downregulated in 13 children with T1D aged 2~12 years (31). Our study found that ADIPOQ was upregulated in the ≤ 6-year-old T1D-1 subgroup and downregulated in the > 6-year-old T1D-2 subgroup, suggesting age-dependent variation in expression among our pediatric T1D subgroups, potentially reconciling the discrepancies across previous studies.

Other proteins, such as APOA4, APOC4, FN1, KNG1, LUM, and TGFBI, showed no differences between the pediatric age groups here, but were differentially expressed in other T1D studies (31, 32). The expression patterns of APOE, APOM, ATXN2, BTD, CRP, GSN, ITIH1, PGLYRP2, and PPBP in both of our T1D age groups were consistent with other studies (33, 34). In other studies, TTR was downregulated in islet autoantibody-positive children (median age, 3.2 years) (32) and ~20-year-old T1D patients (34). In our study, compared to healthy controls, TTR was upregulated in the ≤ 6-year-old T1D-1 subgroup but downregulated in the > 6-year-old T1D-2 subgroup, indicating age-related variation. The other three exosomal DEPs that were examined in this study, namely RAB40A, SEMA6D, and COL6A5, have not been previously reported.

A total of 24 T1D patients, were recruited from Capital institute of pediatrics affiliated children’s hospital, and a total of 12 nondiabetic controls were recruited. The diagnostic criteria for T1D was according to the “Standardized Diagnosis and Management of Type 1 Diabetes in Chinese Children (2020)” (43): a fasting blood glucose >7.0 mmol/L; a glucose tolerance test >11.1 mmol/L; HbA1c% >7% (two tests, one day apart). In addition, positive of serum pancreatic autoantibodies and C-peptide levels (<0.3 mmol/L) and considered clinical features such as increased thirst, frequent urination, unexplained weight loss, or blurred vision for a comprehensive diagnosis of T1D (44). The INT sub-group refers to T1D patients whose HbA1c% levels have dropped to < 7% after ≥ 3-months insulin treatment, T1D patients who might not effectively regulate their HbA1c% level to < 7% consequent to irregular insulin usage or other physiological conditions were not included in our study; It’s noteworthy that there were no significant differences in HbA1c% between the T1D-1/2 patients and pre-insulin-treatment INT-1/2 patient. The selection of controls was random; they were outpatients from nondiabetic unit who came to the hospital for health check-ups, and were matched for age, gender and ethnicity. They reported no symptoms such as fever and the exclusion criteria included subjects who had conditions of bones, kidneys, liver, endocrinopathies, other autoimmune diseases or melanoma – all diseases that might influence vitamin metabolism. Specifically, the age range for all children in the ≤ 6 year-old group was from 3 to 6 years old; the age range for all children in the >6 year-old group was from 7 to 14 years old.

All human subject samples were fasted 12h beforehand, about 2 mL blood was harvested via venipuncture using a syringe pre-coated with EDTA. Then centrifuged at 4°C, 3000 x g, for 30 minutes to remove platelets and cell debris to obtain the supernatant, namely plasma. Using this method, we were able to obtain up to 1mL of plasma. The obtained plasma were then aliquoted and stored at -80°C before further processing and/or analyses. Care was taken to avoid hemolysis and if a plasma sample was hemolyzed it was not used in the study. Add 50µl of 1X PBS to 100µl of clarified plasma, followed by the addition of 30µl of exosome precipitation reagent (4484450, Thermofisher). Vortex mix the mixture and incubate at room temperature for 10 minutes. After that, centrifuge at 10000 × g for 5 minutes at room temperature. Aspirate the supernatant with a pipette and discard it, add 1X PBS to the pellet, once the pellets are resuspended, the exosomes are ready for downstream analysis.

For each subgroup, the extracted exosomes from the six individuals were pooled.Samples were enzymatically digested with trypsin before analysis. ∼100 µg of exosomal proteins underwent denaturation with 8 M urea and reduction with 10 mM dithiothreitol at 37°C for 1 h. Cysteine residues were alkylated with 40 mM iodoacetamide in the dark for 1 h at 37°C, then desalted using PD-10 columns. Trypsin (1:50 enzyme-to-protein ratio) was added, followed by a 3 h incubation at 37°C. All steps were performed using a Thermomixer R (Eppendorf) at 800 rpm. Digested samples were cleaned using C-18 SPE tubes (Discovery DSC-18, Supelco), eluted peptides were concentrated via speedvac. Dried peptides were resuspended and was quantified using the Quantitative Colorimetric Peptide Assay (Thermofisher) in triplicate. 50ug of peptide were labeled with TMT6 reagents in a 30% final concentration of acetonitrile for an hour followed by 15 minutes of quenching with 5% hydroxylamine.Then the peptide mixtures were subjected to high-pH, one-dimensional reversed-phase chromatography separation on a Dionex Ultimate 3000 HPLC system. The trypsin digests were injected onto a Phenomenex column (Gemini NX 3U C18 110A; 150 x 2.00 mm) at a flow rate of 200 μl/min. Fractions were collected every 1.5 minutes throughout the run. Subsequently, all 40 collected fractions were combined into 10 fractions, lyophilized, and stored at -80°C for further nano-HPLC/MS/MS analysis.

Samples were run on a QExactive HF mass spectrometer (Thermo Scientific, Bremen, Germany) equipped with an UltiMate 3000 RSLC nano System (Dionex, Germering, Germany). MSI data were collected in the Orbitrap at 120k resolution (m/z range 350-1350) with a 100 ms maximum injection time. Charge states 2-6 were sequenced, using a 60s dynamic exclusion window and excluding isotopes. MS2 sequencing occurred in the ion trap after quadrupole selection and CID fragmentation, with an m/z window of 400-2000.The obtained raw data were processed using Peaks X studio (Version 10.0, BSI) by searching a Homosapiens database (www.uniprot.org, UP00005610). Precursor mass error tolerance was set to 10 ppm and fragment mass error tolerance to 0.02 Da. Peptide spectral matches were validated at a 1% FDR using a percolator based on q-values. Post-translational modifications and chemical labeling settings were as followes: fixed carbamidomethylation of cysteine residues and TMT-6plex; variable methionine oxidation and protein N-term acetylation. Protein unique peptides were set to >1, and a high confidence score of -10IgP >20 indicated accurately identified proteins.

Gene ontology (GO) term annotations of differentially abundant proteins in exosome isolated from children T1D was performed using Blast2GO (version 2.6.4) with the default annotation parameters (BLAST e-value threshold of 1e-06, Gene Ontology annotation threshold of 55). Reactome pathway annotation was performed using the Server (https://reactome.org/, Version 84 released on March 29, 2023). Trend analysis was performed by Short Time-series Expression Miner (STEM) software, the correlation coefficient between the expression pattern of protein and the trend it belongs to should be greater than 0.7, after Bonferroni correction, if the adjusted P-value is less than 0.05, we consider the enrichment of this trend to be significant (45).

The exosomal protein samples underwent separation via a 4-12% SDS-PAGE and were then transferred onto polyvinylidene fluoride membranes. The membranes were blocked with 10% skim milk for one hour and incubated overnight at 4 °C with primary antibodies, which included RAB40A (bs-8366R, Bioss, China), SEMA6D (bs-2943R, Bioss), COL6A5 (bs-11047R, Bioss), and TTR (#29872, CST). Following this, the membranes were rinsed using Tris Buffered Saline with Tween 20 before being incubated with infrared-labeled anti-rabbit/mouse IgG antibody. Ultimately, the protein bands were detected using the Image Lab system (Gel Image System, Tanon, Shanghai, China), with their expression levels determined relative to the staining of β-actin (#3700, CST).

The data were analyzed using the SPSS, version 23. Normative distribution of data was tested by P-Plot. All data were normally distributed. Student t-test was used to two‐group comparison. ANOVA, followed by an appropriate post‐hoc analysis, was employed in multiple‐group (>2 groups) comparisons. P value less than 0.05 was considered statistically significant.