25-Hydroxyvitamin D and Peripheral Immune Mediators: Results from Two Nationwide Danish Pediatric Cohorts

(1) Background: We aimed to examine if 25-hydroxyvitamin D (25(OH)D) was related to the peripheral immunological and inflammatory signature both at birth, and in newly diagnosed patients with childhood type 1 diabetes (T1D) and their healthy controls; (2) Methods: The birth cohort consisted of 470 patients and 500 healthy controls. Dried blood samples were collected from the neonates in the period 1981–1999. The newly diagnosed cohort consisted of 460 patients and 453 siblings. Serum samples were collected in the period 1997–2005. A variety of peripheral immune mediators were measured and compared to total 25(OH)D levels (25(OH)D2 + 25(OH)D3). For each immune mediator, the relative change (RC) in the mean level was modeled by robust log-normal regression and correction for multiple testing was performed; (3) Results: Two associations were identified; there was a negative association between 25(OH)D (10 nmol/L increase) and leptin (RC (95% confidence interval (CI)), 0.98 (0.96; 1.00)), and a positive association between 25(OH)D (10 nmol/L increase) and the chemokine, chemokine (c-x-c motif) ligand (CXCL) 8 (RC (95% CI), 1.07 (1.01; 1.13)); (4) Conclusion: CXCL8 and leptin have significant associations with levels of 25(OH)D in the newly diagnosed cohort. These results do not indicate a strong influence of 25(OH)D on the peripheral immunological or inflammatory signature.


Introduction
Type 1 diabetes (T1D) is an endocrine disease that results from autoimmune destruction of the pancreatic insulin-producing β-cells, leading to a loss of insulin secretion and symptomatic hyperglycemia. Changes in the complex interplay between genetic, epigenetic, and environmental factors are thought to be involved in T1D etiopathogenesis [1][2][3].
Vitamin D is in the spotlight, as many in vitro and animal studies indicate that vitamin D has anti-inflammatory activity and may also prevent or halt the destruction of β-cells [4,5]. Findings from epidemiological studies focusing on the association between vitamin D and T1D, both in early life and around time of diagnosis, are inconsistent [6,7]. However, to our knowledge, no large-scale studies have yet examined if there are any associations of 25-hydroxyvitamin D (25(OH)D) with peripheral immune mediators in cohorts of patients with childhood T1D and healthy controls. Therefore, we aimed to examine if 25(OH)D was associated with a wide range of peripheral immune mediators at two different time-windows, namely (1) at birth in patients who subsequently developed childhood T1D and their healthy controls and (2) in newly diagnosed patients with childhood T1D, and their healthy siblings.

Study Design, Sample Population, Data Sources and Variables
We conducted two separate case-control studies-both study samples are described extensively elsewhere including coding of covariates [7][8][9][10].

The Birth Cohort Samples
Blood sampling (dried blood spots (DBS)) was performed within a week following birth during the years 1981 to 1999. For each identified clinical case (a neonate who subsequently develops T1D before the age of 18 years) one control was selected based on having measurements on both peripheral immune mediators and 25(OH)D. Clinical cases were identified from the Danish Patient Register and further validated against the Danish Registry of Childhood and Adolescent Diabetes (DanDiabKids) [11]. Our random sample of clinical cases and controls are based on a sub-sample from a population-based case-control study comprised of 2086 clinical cases and 4172 controls [12]. In the original study the following inclusion criteria were: (1)

The Newly Diagnosed Cohort Samples
A random sample of 500 clinical cases was collected from a biobank linked to the DanDiabKids. Blood samples were taken less than three months after the onset of T1D during the years 1997 to 2005. The onset date was defined as the date of the first insulin injection. The control group comprised a random sample of 500 siblings selected from an eligible sample of approximately 2000 siblings with blood samples stored in the biobank and a sibling diagnosed with T1D. The chosen control sample was independent of participation of a sibling with T1D in the study. Randomization aimed at covering all sample months (January 1997 to December 2005). A last inclusion criterion was that all individuals had to be between 0 to 18 years of age at time of blood sampling.
Serum samples from the DanDiabKids were stored at −80 • C/−112 • F during the entire study period from 1997 to 2005.
Quality control of the analysis were made using mouse IL-6 as an internal analyte added to the extraction buffer to detect pipetting errors, and biotinylated beads to detect signal errors (more thoroughly described in Skogstrand et al. [16]). Calibration curves were used on each plate together with one high and two low controls. Samples, calibrators, and controls were analyzed in duplicate.

Assessment of 25(OH)D on Dried Blood Spots
In short, 25(OH)D status was assessed by measuring 25(OH)D 2 and 25(OH)D 3 in 3.2 mm samples (also called punches) taken from DBS cards. Sample preparation and analysis was performed using liquid chromatography-mass spectrometry (LC-MS) according to a modified method [17], described in detail elsewhere [9]. The

Assessment of 25(OH)D in Serum
Vitamin D status was measured as serum 25(OH)D by high-performance liquid chromatography (HPLC) [19]. Detection limit for 25(OH)D was 9.5 nmol/L, with a CV of 8%.

Statistical Analysis
For each peripheral immune mediator the relative change (RC) in the mean level by 10 nmol/L increase in total 25(OH)D (equivalent to 25(OH)D 2 + 25(OH)D 3 ) was modeled by a robust log-normal model regression, which takes into account: (1) that measurements are potentially both left and right censored; and (2) correlation within assay. To account for correlation within cluster inference was based on a working independence generalized estimation equation (GEE) approach. For the birth cohort the clusters are equivalent to assay. For the newly diagnosed cohort, the cluster was equivalent to family/sibling ID.
For the birth cohort, the following risk factors are included in the model: 25(OH) levels, sex, case status (T1D or control), gestational age, mothers age, birth weight, season and calendar year group. For the newly diagnosed cohort the following risk factors are included in the model: 25(OH) level, sex, case status (T1D or sibling), age at blood sampling, season and calendar year group. The coding of these variables is presented in Table 1.
Simultaneous evaluation of risk factors on all peripheral immune mediators was done using the model stacking approach detailed in Pipper et al. [20]. Subsequent adjustment for multiple testing and familywise 95% confidence bands are calculated using the single step procedure by Hothorn et al. [21]. Likelihood ratio estimates of mean ratios and accompanying confidence limits are calculated on a log scale and transformed back to the original scale. Overall functional misspecification by including 25(OH)D as a trend (linear variable) was assessed by a lack-of-fit test. Specifically, we included a quadratic term of 25(OH)D and tested its significance by a robust Wald test.
All analyses are made using the statistical software package R version 3.2.0 (the R foundation for statistical programming, Vienna, Austria) and the add-on packages survival, ggplot2, and multcomp.

Ethics
Both studies were performed in accordance with the Helsinki II Declaration. Furthermore, both studies were approved by the Danish Ethical Committee (H-4-2013-049 and H-KA-20070009). All of the patients and their parents or guardians gave informed consent. Table 1 The birth cohort contains 970 individuals with a complete set of covariates (470 patients and 500 controls). In the birth cohort, the median/first and third quartile (Q1-Q3) 25 In the newly diagnosed cohort the median/Q1-Q3 25(OH)D level was 62.0/40.1-95.5 for patients and 57.9/38.0-89.0 nmol/L for controls. It is noteworthy to mention that we have already shown that 25(OH)D levels are not associated with later risk of childhood T1D in the birth cohort and levels between newly diagnosed patients with childhood T1D and their healthy siblings do not differ either [7,9]. Levels of the peripheral immune mediators are presented in Table 2.

Association between Peripheral Immune Mediators and 25(OH)D in the Birth Cohort
We were unable to detect any linear association of 25(OH)D on the 14 examined peripheral immune mediators. The numeric results including a graphical overview (forest plot) are presented in Figure 1.
We also examined if non-linear associations existed between 25(OH)D and the peripheral immune mediators in the birth cohort, but found no sign of such associations (p = 0.25).
We also examined if non-linear associations existed between 25(OH)D and the peripheral immune mediators in the birth cohort, but found no sign of such associations (p = 0.25).
As for the birth cohort, we also examined if a non-linear associations between 25(OH)D and the peripheral immune mediators existed in the newly diagnosed cohort, but no proof of such associations were found (p = 0.26).
As for the birth cohort, we also examined if a non-linear associations between 25(OH)D and the peripheral immune mediators existed in the newly diagnosed cohort, but no proof of such associations were found (p = 0.26). We also examined if non-linear associations existed between 25(OH)D and the peripheral immune mediators in the birth cohort, but found no sign of such associations (p = 0.25).
As for the birth cohort, we also examined if a non-linear associations between 25(OH)D and the peripheral immune mediators existed in the newly diagnosed cohort, but no proof of such associations were found (p = 0.26).

Discussion
In these two large-scale, case-control studies, 14 peripheral immune mediators were measured in a birth cohort and 18 peripheral immune mediators were measured in a newly diagnosed childhood T1D cohort. 25(OH)D levels were compared with these immune and inflammatory factors, and two peripheral immune mediators, i.e., CXCL8 and leptin, were associated with 25(OH)D in the newly diagnosed cohort. These results do not indicate a strong role of 25(OH)D as an immune modulator when using peripheral immune mediators as a proxy for the child's immunological fingerprint.
We demonstrated that CXCL8 levels in our newly diagnosed cohort rises with 1%-13% for every 10 nmol/L increase in 25(OH)D. CXCL8, also known as IL-8, is a chemokine which affects leukocyte migration positively and thereby promotes inflammation [22]. Interestingly, the CXCL8 gene has been shown to be a vitamin D receptor (VDR) binding-site; hence the active form of vitamin D (1,25(OH) 2 D) can alter, e.g., up-regulate, its expression, which may help to clear infections by recruitment of specific immune cells [23,24]. In an in vitro study using hyperinflammatory macrophages from patients with cystic fibrosis, vitamin metabolites have been shown, in high concentrations (e.g., 25(OH)D >100 nmol/L), to down-regulate CXCL8 [25]. These discrepancies could be due to different cell lines used e.g., monocytic leukemia cell line (THP-1) versus hyperinflammatory macrophages, but may also depend on choice of vitamin D metabolite, vitamin D metabolite concentration, and cell culture duration [23,25]. Our results reflect an overall systemic in vivo association of 25(OH)D on CXCL8 levels. The role of CXCL8 in regards to T1D pathogenesis remains to be examined, but we previously reported no difference in CXCL8 levels between patients and healthy controls in both these cohorts [26]. Furthermore, no association between 25(OH)D and CXCL8 was found in the birth cohort.
We also found leptin levels significantly reduced with 0%-4% for every 10 nmol/L increase in 25(OH)D in the newly diagnosed cohort. Leptin is an adipokine, secreted primarily from adipose tissue. The overall action of leptin in the immune system is to activate leukocytes and mediate inflammation [27]. In a recent study that utilized the newly diagnosed cohort, we found leptin levels were 10%-40% lower in patients with childhood T1D compared to their healthy siblings, but this finding was not mirrored by higher 25(OH)D levels [7,28]. A systematic review and meta-analysis conducted on adult populations without T1D also found evidence of an inverse relationship between 25(OH)D and leptin, but well-designed clinical trials with vitamin D supplementation of more than 1000 IU/day are needed to confirm such a relationship [29].
The effect sizes of the abovementioned significant associations are both small, and may not have profound immunological importance. Importantly, we do not find any association between physiological 25(OH)D levels and well-known Th1, Th2 or Treg cytokines that have been associated with either β-cell destruction or protection e.g., IL-1β, IFNγ, TNFα, IL-10 and TGFβ [30,31].
This study has a number of strengths. Firstly, the sample size in both cohorts includes over 900 individuals-both cohorts consist of a patient group (childhood T1D) and a healthy control group (in the newly diagnosed cohorts these healthy controls are a sampling of siblings). Secondly, both cohorts are population-based and patients with childhood T1D are thoroughly validated and controls are randomly selected. Thirdly, quantification of both 25(OH)D and a broad spectrum of peripheral immune mediators, initially measured for testing other hypotheses, gave us the opportunity to conduct the current study using these unique cohorts. Our study also has some limitations. Firstly, a single measurement of both 25(OH)D and peripheral immune mediators may not detect long-term differences or reflect dynamic changes. Secondly, sample storage time may influence assay measures due to degradation, however calendar year of blood sampling was included in both models to control for this. Thirdly, one may ask if these measured peripheral immune mediators, in some part, mirror or affect the tissue-specific immunological micro-milieu in the islet of Langerhans in the pancreas. Or, it could be the case that we just looking at a more general immunological cross-sectional fingerprint and its association with 25(OH)D. Either way, we find a very small influence of 25(OH)D on these peripheral immune mediators.

Conclusions
We have examined a wide range of peripheral immune and inflammatory mediators in both neonates who subsequently develop childhood T1D, and in a separate cohort of newly diagnosed patients with childhood T1D. Two immune factors, i.e., CXCL8 and leptin, have significant associations with levels of 25(OH)D in the newly diagnosed cohort. These results indicate that vitamin D does not appear to play a major role as an immune-modulator of the peripheral immune system. To further understand a possible role of vitamin D on the human immune system, studies need to be performed in more complex systems that include genetics, repeated measures of functional immune cell populations and gene-immune mediator interactions.