Diet plays a role in regulating the inflammatory conditions within the body. For example, we found that the intake of the precursor FA is associated with the plasma levels of some oxylipins in DAISY. While it may not be possible to prevent inflammation causing insults, it may be possible to identify the modifiable interventions that reduce the inflammatory response to these exposures. Anti-inflammatory nutrients, including n3 fatty acids , are associated with a reduced risk of IA. The supplementation of vitamin D in children, which has both immuno modulatory and anti-inflammatory properties, has been shown to be protective of T1D. In a case report, high-dose vitamin D combined with n3 FA have slowed beta cell destruction. On the other hand, serum levels of the FA palmitoleic acid, a pro-inflammatory nutrient, is associated with the development of IA. In DAISY, a higher total sugars intake and glycemic index—also associated with inflammation—increased the risk of progression from IA to T1D. We hypothesized that both genetic and dietary factors influenced oxylipin levels, so we adjusted for the genetic variants associated with oxylipins prior to developing the dietary pattern. The aim of this study was to develop an oxylipin-related nutrient pattern that was independent of genetics and test the relationship between the nutrient pattern and the risk of T1D in a cohort of children at risk of T1D.

We used nutrients, rather than foods,grow bag for blueberry plants to develop the intake pattern because nutrients may be more portable to different cultures and dietary practices than food. To do this, we used genetically adjusted oxylipin profiles that are associated with T1D as the response variables for a reduced rank regression model and tested the resulting nutrient patterns with the risk of T1D in the DAISY cohort.Oxylipins were quantified as described by Pedersen et al.. The extracted oxylipins were separated and quantified using a Waters i-Class Acquity UHPLC system coupled with a Sciex 6500+ QTRAP mass spectrometer operated in a negative ionization mode. Oxylipins were quantified by targeted, retention-time specific, multiple reaction monitoring ion transitions. There were quantified values for 24 n6-related oxylipins, including 14 ARA-derived and 10 LA-derived oxylipins. There were 12 measured n3-related oxylipins, which included six ALA-derived, four Docosahexaenoic acid -derived and two Eicosapentaenoic acid -derived oxylipins. The oxylipins were Box-Cox transformed using the forecast package in R and considered normally distributed for further analyses. Diet was measured using a Willett 111-item semi-quantitative food frequency questionnaire in the DAISY cohort. The FFQ was administered annually to the parents, inquiring on the previous year’s dietary intake of the child starting at 2 years of age, and was validated for use in the DAISY population. After the age of 10, the children recalled their own diets by completing the youth/adolescent questionnaire. Both questionnaires produce similar estimates of the nutrients and may be combined for analyses, as reported. As compared to a 3-day food record, the FFQ produced similar dietary patterns with a modest agreement in adolescents.

The nutrient values were residually adjusted for the total energy intake. To date, most studies on diet and T1D have examined single nutrients or food items. However, foods and nutrients are generally eaten in combination and may interact synergistically or antagonistically. Testing combinations of foods or nutrients through comprehensive patterns of intake may improve our understanding of how diet may promote or mitigate disease risk. RRR is a data-driven approach for creating dietary patterns by creating linear combinations of predictors that explain the maximum covariance of a set of response variables. RRR is useful for studying disease pathophysiology because it utilizes a priori knowledge of disease-specific markers and a posteriori information on dietary intake for creating dietary patterns for a disease outcome. RRR may be useful in leveraging T1D-specific inflammatory markers in determining a relevant dietary pattern. We used RRR to create the nutrient patterns, and this statistical method did not allow for the use of longitudinal data in identifying the optimal set of predictors for the outcome . So, we created the summary measures of oxylipins and the nutrients in order to create the nutrient pattern. We employed a five-step process in this analysis . We first created the response variable for the RRR, namely the genetically adjusted oxylipin PCs. In order to maximize the sample size, we included everyone with both exome chip and oxylipin data, which had been measured in the participants in the IA nested case control study . We had multiple measures of oxylipins, so we created a summary measure of each oxylipin. To do this, we used a linear mixed model with arandom intercept and an unstructured covariance structure with age as the predictor and the oxylipin as the outcome, as previously described. The subject-specific intercept from these models was used as an age-adjusted measure of the average oxylipin level over time. Since oxylipins share precursor FAs and enzymes, we used principal components analysis of the oxylipin intercepts to measure the oxylipin profile, as described.

Based on the screen plot of the oxylipins that loaded onto the PCs, we selected PC1 and PC2 to reflect the distinct oxylipin profiles in the DAISY children, as described. PC1 represented LA- and ALA-related oxylipins and PC2 represented ARA-related oxylipins . We previously found that these oxylipin profiles were associated with risk of T1D in DAISY . To create the nutrient measures for the RRR, we included all the participants with complete data . Four participants from the GWAS in Step 1 did not have the measured FFQ data, so n = 335 participants were included . Nutrient intake was used for the dietary variables because nutrients may be more portable across cultures and dietary practices compared to specific food items. Additionally, because RRR assumes a Gaussian distribution of variables, nutrients, which are aggregated across foods, are more likely to follow a Gaussian distribution compared to the measurements of food intake from a FFQ. We selected macro- and micro-nutrients and removed the nutrients that were components of one another for the ease of interpretation, leading to a list of 48 nutrients . We summarized multiple measures of the nutrient intake. We used a linear mixed model with a random intercept and unstructured covariance structure with age as the predictor and nutrient intake using the residual method as the outcome. The subject-specific intercept from these models was used as an age-adjusted measure of the average nutrient intake over time and was standardized to a mean of 0 and a standard deviation of 1. We tested the association between the nutrient pattern and the risk of T1D longitudinally in the full DAISY cohort using a joint Cox proportional hazards model. We included everyone with at least one measured FFQ, n = 1933, including 81 T1D cases , . Joint models allow for analyzing the repeated measurement of the exposure data with time-to-event outcomes, and it utilized the extensive dietary intake of the data available through DAISY. Joint models estimate the exposure variable profile over time,blueberry grow bag assume a smooth change over time and can accommodate the missing data or measurements taken at unequal intervals . For the joint model, the age of the FFQ measurement was used to estimate the trajectory of the nutrient pattern. We adjusted the nutrient intake for the TEI using the residual method for all the FFQs in the full cohort and then standardized the nutrient intake to a mean of 0 and standard deviation of 1. We applied the nutrient patterns to all the FFQ measures by scoring the nutrients calculated from each food record according to the weights derived from the nutrient pattern. A higher score indicated an intake closer to the nutrient pattern. We determined that a quadratic trajectory with five knots was the best fit for the data using AIC to assess the model fit. The nutrient pattern was tested as an average and as the cumulative effect. When applying an RRR dietary pattern to another population, it is recommended to implement a “simplified pattern” that consists only of foods or nutrients that load as important. Therefore, we scored a “simplified nutrient pattern,” with only nutrients with loading values >|0.2|.

The joint models were adjusted for family history of T1D, the high-risk HLA genotype, race/ethnicity and sex. We extracted two nutrient patterns using RRR. Nutrient pattern 1 explained 9.5% of the variation in genetically adjusted PC1 and 0.01% of the variation in genetically adjusted PC2. The positive loadings indicated higher intakes, whereas the negative loadings indicated lower intakes. We considered the nutrients with loadings > |0.2| to be important based on the distribution of the factor loadings. NP1 represented a diet low in beta cryptoxanthin, flavanone, vitamin C, total sugars and iron, and high in lycopene, anthocyanidins, LA and sodium . Nutrient pattern 2 explained 9.5% of the variation in genetically adjusted oxylipin PC1 and 4.7% of the variation in genetically adjusted oxylipin PC2. NP2 represented a diet low in beta cryptoxanthin, vitamin C, potassium, flavonols, magnesium, vitamin B12 and LA.Finally, we tested the nutrient pattern in the DAISY cohort, including people who did not have measured genetics or oxylipins, using a joint Cox PH model. This model estimated the trajectory of the nutrient pattern using age at the FFQ measurement and incorporated these estimates with the standard error into the Cox PH model. Here, we included n = 1933 participants, including 81 T1D cases. The T1D cases were more likely to have the high-risk HLA genotype and a first degree relative with T1D and were more likely to report non-Hispanic white ethnicity . We applied the original NP1 and the simplified NP1 to all the FFQ measurements and tested the original and simplified NP1 as the average and the cumulative value, adjusted for family history of T1D, the high-risk HLA genotype, race/ethnicity and sex. NP1 was not associated with T1D . Using the data from 335 participants in the DAISY nested case-control study, we developed two nutrient patterns that each explained the variation in an oxylipin profile suspected to play a role in the pathophysiology of T1D via pro-inflammatory or proresolving pathways. One of these nutrient patterns was inversely associated with the T1D outcome in our nested case-control study. This nutrient pattern was characterized by low beta cryptoxanthin, flavanone, vitamin C, total sugars and iron, and high lycopene, anthocyanidins, LA and sodium. While this nutrient pattern predicted a risk of T1D in the nested case-control in which it was developed, this association did not replicate in the full DAISY cohort.Both iron and vitamin C loaded negatively on NP1 . These results make sense, as it is well known that vitamin C increases iron absorption. Observationally, maternal iron may increase the risk of T1D, although the results are not consistent. High dietary iron alongside dysregulation of pancreatic iron homeostasis may accelerate the onset of T1D in a mouse model. Total sugars also loaded on to NP1 , which aligns with similar findings that total sugars intake increased the risk of the progression from IA to T1D in the DAISY. This is in line with the “overload hypothesis” that suggests that a high insulin load may trigger the onset of autoimmunity and T1D. In this analysis, we were not able to assess the difference between heme and non-heme iron, or the timing of the iron and vitamin C intake, which may impact the absorption levels. Future studies on the interplay between iron intake, vitamin C and total sugars in the context of inflammation and T1D may reveal the mechanisms of these relationships in reducing the risk of T1D. We also found that the antioxidants lycopene and anthocyanidins found in red and blue fruits and vegetables loaded positively on NP1, and, interestingly, the antioxidants beta cryptoxanthin and flavanone , generally found in yellow fruits and vegetables,loaded negatively. Anthocyanidins have been shown to reduce oxidative stress and autophagy in islet cells in vitro. Blueberries, which have a high level of anthocyanins, decreased the production of pro-inflammatory ARA-related oxylipins post-exercise. The antioxidant effect of anthocyanidins and lycopene may shift the oxylipin production to a pro-resolving, anti-inflammatory profile, reducing the risk of T1D. Sodium loaded the most strongly on NP1 and was significantly and positively associated with oxylipin PC1. In adults, salt loading increased the formation of LA-related oxylipins. Additionally, sodium has been hypothesized to inactivate fatty acid desaturases that synthesize ARA from LA and EPA and DHA from ALA.