Baseline Serum Vitamin A and D Levels Determine Benefit of Oral Vitamin A&D Supplements to Humoral Immune Responses Following Pediatric Influenza Vaccination

Maximizing vaccine efficacy is critical, but previous research has failed to provide a one-size-fits-all solution. Although vitamin A and vitamin D supplementation studies have been designed to improve vaccine efficacy, experimental results have been inconclusive. Information is urgently needed to explain study discrepancies and to provide guidance for the future use of vitamin supplements at the time of vaccination. We conducted a randomized, blinded, placebo-controlled study of influenza virus vaccination and vitamin supplementation among 2 to 8 (inclusive) year old children over three seasons, including 2015–2016 (n = 9), 2016–2017 (n = 44), and 2017–2018 (n = 26). Baseline measurements of vitamins A and D were obtained from all participants. Measurements were of serum retinol, retinol-binding protein (RBP, a surrogate for retinol), and 25-hydroxyvitamin D (25(OH)D). Participants were stratified into two groups based on high and low incoming levels of RBP. Children received two doses of the seasonal influenza virus vaccine on days 0 and 28, either with an oral vitamin supplement (termed A&D; 20,000 IU retinyl palmitate and 2000 IU cholecalciferol) or a matched placebo. Hemagglutination inhibition (HAI) antibody responses were evaluated toward all four components of the influenza virus vaccines on days 0, 28, and 56. Our primary data were from season 2016–2017, as enrollment was highest in this season and all children exhibited homogeneous and negative HAI responses toward the Phuket vaccine at study entry. Responses among children who entered the study with insufficient or deficient levels of RBP and 25(OH)D benefited from the A&D supplement (p < 0.001 for the day 28 Phuket response), whereas responses among children with replete levels of RBP and 25(OH)D at baseline were unaffected or weakened (p = 0.02 for the day 28 Phuket response). High baseline RBP levels associated with high HAI titers, particularly for children in the placebo group (baseline RBP correlated positively with Phuket HAI titers on day 28, r = 0.6, p = 0.003). In contrast, high baseline 25(OH)D levels associated with weak HAI titers, particularly for children in the A&D group (baseline 25(OH)D correlated negatively with Phuket HAI titers on day 28, r = −0.5, p = 0.02). Overall, our study demonstrates that vitamin A&D supplementation can improve immune responses to vaccines when children are vitamin A and D-insufficient at baseline. Results provide guidance for the appropriate use of vitamins A and D in future clinical vaccine studies.

A lack of clarity as to when and how vitamins influence immune responses toward vaccines is concerning because vitamin A and D deficiencies and insufficiencies are prevalent worldwide, affecting both developed and developing countries [11]. Low vitamin levels render individuals vulnerable to infectious diseases [9,13]. Although vitamin supplementation programs are common in developing countries where deficiencies are a known public health concern, they cannot be implemented worldwide without a comprehensive understanding of benefits and risks.
Surprisingly, although there is considerable interest in the independent effects of vitamins A and D on the immune response, the two vitamins are rarely examined together. This is despite the knowledge that vitamins A and D influence numerous cell functions, are closely related, and can be cross-regulated [8,20,24,26,27,[51][52][53][54][55][56][57][58]. Unfortunately, when clinical studies are designed to evaluate the influences of vitamin supplements on the immune response, the baseline vitamin A and/or D levels of study participants are often unknown, and the potential consequences of vitamin cross-regulation are rarely considered.
To address knowledge gaps, we designed a pediatric randomized controlled study of influenza virus vaccination that measured baseline vitamin A (using retinol-binding protein [RBP] as a surrogate [59]) and 25(OH)D levels among healthy children. Participants were stratified into two groups based on high and low baseline RBP levels and randomized to receive a vitamin A and D supplement (A&D) or a placebo control immediately prior to influenza virus vaccination. The hemagglutination inhibition (HAI) assay, a standard in the field for the assessment of influenza virus vaccines, was used to measure immune responses [60]. We discovered that baseline vitamin levels were critical parameters that determined not only antibody responses toward the influenza virus vaccine but the influence of vitamin supplementation on the immune response. . Informed consent was given by parents or guardians and assent was given by minors when age-appropriate. Comprehensive histories of previous vaccinations and influenza virus exposures were not available. Participants were excluded from study entry if they had any chronic illness, developmental delay, or neurological disorder. They were also excluded if they were known to have received an influenza vaccine for the current season or were routinely taking a daily vitamin supplement.

Blood Sample Collection
Blood samples were collected at screening (≤10 days from day 0) and on days 0, 28, and 56. Participants who completed the day 28 visit were considered evaluable. Retinol was measured by extraction from test samples under conditions of UV-blocked lighting with samples of SRM-968f human reference sera (NIST, https://www-s.nist.gov/srmors/orderingSRMs.cfm) used as controls. Samples were diluted with HPLC-grade water and then absolute ethanol and HPLC grade hexane. Tubes were vortexed, incubated for 30 min in the dark, and centrifuged. Hexane layers were removed. A second hexane extraction followed, which was combined with the first, and evaporated to dryness under nitrogen. Methanol:water:acetonitrile (10:20:70 by volume) was added followed by vortexing. Samples were transferred to UPLC vials, briefly centrifuged, and loaded into a Waters Acquity UPLC tray for injection onto a Waters Acquity BEH C-18 reverse phase column using methanol:water:acetonitrile (10:20:70) as the mobile phase. Standards of purified all-trans-retinol (0 to 0.8 pmol/10 µL) provided a standard curve (R 2 > 0.99). 25(OH)D was tested in the Pathology Department at St. Jude using the Roche Elecsys Vitamin D ELISA (Roche, Basel, Switzerland) that measures 25(OH)D metabolites of cholecalciferol (vitamin D3) and ergocalciferol (vitamin D2). Vitamin D sufficiency was defined as ≥ 30 ng/mL 25(OH)D.

Hemagglutination Inhibition (HAI) Assay
The HAI assay was conducted to evaluate antibody responses to each of the 4 components in the seasonal influenza virus vaccine. Briefly, antigen (~4 agglutination doses representing each antigen) was added to wells of a 96-well plate containing serial dilutions of test sera (initiated with a 1:10 serum dilution followed by serial 1:2 dilutions, each tested in duplicate). After 30 min incubation at room temperature, 50 microliters of 0.5% vol/vol turkey red blood cells were added to each well and plates were incubated at room temperature for an additional 30 min. Titers were recorded as the highest dilution that inhibited hemagglutination. A few participants did not return for the day 56 visit. A score of 5 (1/2 the lowest dilution tested) was given if no HAI activity was detected. If two different values were observed for duplicate samples, the geometric mean value was used for graphing and calculation purposes.

Statistical Analyses
Results for each vaccine season were tested independently due to differences in vaccine composition each year. Spearman's rank correlation was applied to evaluate relationships involving ordinal variables. A generalized estimating equation (GEE) model was used to assess the effect of influenza vaccine plus A&D over time on HAI responses to each antigen among participants with different baseline vitamin levels. Specifically, the GEE model was constructed with log2-transformed HAI titers as the response, age, race, and a three-way interaction among time, baseline vitamin levels, and study groups as covariates, and the first order autoregressive (AR1) as the working correlation structure. Data analyses were performed using SAS 9.4 (SAS Institute, Inc, Cary, NC, USA) and GraphPad Prism software (San Diego, CA, USA).

Participant Characteristics
Age, sex, race, and baseline vitamin levels are shown for all three seasons and all participants in Table 1. As shown, a different vaccine was administered to study participants each year, and FluMist ® was only used in season 2015-2016 due to revised recommendations by the CDC. As expected, baseline RBP and retinol levels were positively correlated (Spearman's correlation r = 0.7, p < 0.0001), emphasizing that RBP can serve as a surrogate for retinol. An additional test of RBP was conducted on day 56 for placebo and A&D groups, but day 56 values were not significantly different from day 0 values in either group. Baseline HAI means and ranges are shown for each season (Table 1). For HAI, data were not combined among years because vaccines and consequent immune responses differed. Full details for all participants and all years are provided in Tables A1-A6. Table 2 shows the frequencies of ≥ 4-fold increases in HAI titers by year and influenza virus antigen on days 28 and 56. As shown, the largest participant number was in season 2016-2017 (n = 44). Accordingly, results from this season were used for the primary analyses. For other seasons, descriptive analyses were performed due to limited available data.
For several vaccine components, baseline HAI titers were highly variable among study participants, a situation that may confound comparisons between participants and study groups. Fortuitously, for the B/Phuket/3073/13 vaccine in the 2016-2017 season, when enrollment was highest, baseline titers were all negative (highlighted in Table 1).

Vitamin Supplements were Beneficial to Children with Insufficient Vitamin A and D Levels at Baseline
To determine how baseline vitamin levels affected immune responses and benefits of vitamin supplementation, we characterized each participant from season 2016-2017 as being sufficient or insufficient/deficient for vitamins A or D at baseline, and accordingly assigned participants to one of four groups (e.g., the "Low A/Low D" group included individuals who were insufficient or deficient for both vitamins A and D). Cut-offs for sufficiency (termed "high") were ≥ 22,000 ng/mL for RBP and ≥ 30 ng/mL for 25(OH)D. The fold-change of HAI titers on days 28 and 56 compared to baseline were examined for each group (Table 2 and Figure 1.) The percentages of participants with a ≥ 4-fold rise in HAI activity on day 28 or day 56 compared to baseline are shown in Table 2. Of particular interest, we found that children who were insufficient or deficient for both vitamins A and D (Low A/Low D) benefitted significantly from the A&D supplement (Table 2 and Figure 1.).

Baseline RBP Correlates Positively, while Baseline 25(OH)D Correlates Negatively, with Immune Responses Toward Influenza Vaccine Components.
We next examined how baseline serum vitamin A and D levels correlated with HAI toward the four vaccine components in season 2016-2017 (A/CA/7/09 H1N1, A/Hong Kong/4801/14 H3N2, B/Phuket/3073/13, and B/Brisbane/60/2008). Correlations were evaluated between baseline RBP levels and HAI titers (including absolute HAI titers on day 28 and day 56, peak HAI titers, or changes in HAI titers between days 0 and 28 or days 0 and 56). As shown in Figure 2A and 2B, baseline RBP was positively correlated with HAI.
This result was more pronounced in the placebo group (Figure 2A) compared to the A&D group. Surprisingly, the same result was not observed for 25(OH)D. In this case, there were often negative correlations between 25(OH)D and the HAI immune response (i.e., a high 25(OH)D level at study entry associated with a low HAI response).
The negative correlation was more pronounced in the A&D group ( Figure 2D) compared to the placebo group. Individual patient data for two of these correlations are shown in Figure 2E-F. As shown, the baseline RBP value was positively and significantly correlated with the day 28 HAI response toward B/Phuket/3073/13 in the placebo group (r = 0.6, p = 0.003, Figure 2E), and the baseline 25(OH)D value was negatively and significantly correlated with the day 28 HAI response toward B/Phuket/3073/13 in the A&D group (r = −0.5, p = 0.02, Figure 2F).

Discussion
Our clinical study was designed to investigate how baseline vitamins A and D influence the immune response to influenza virus vaccines. We found that baseline vitamin A levels (scored by RBP) correlated positively with immune responses, particularly toward the B/Phuket/3073/13 vaccine component in the placebo group in season 2016-2017. However, negative correlations were observed between baseline 25(OH)D and the immune response, particularly in the group that received vitamin supplementation. The vitamin A&D supplement significantly improved immune responses toward the B/Phuket/3073/13 vaccine in season 2016-2017 but only when baseline vitamin A and D levels were insufficient at baseline. The supplement had no significant effect or weakened the HAI response when baseline RBP and 25(OH)D levels were sufficient.
Our results were somewhat surprising given that vitamins are often viewed as positively associated with immune responses. Our observation that 25(OH)D correlated negatively with vaccine-induced HAI appeared contrary to the report by Chadha et al. who observed a positive influence of vitamin D on the immune response to influenza virus in cancer patients [15]. Of note, not all researchers have agreed that vitamins are beneficial. Some authors have argued that "too much" vitamin can have a negative influence on the immune response to vaccines [5,34,40,44,45]. Lee et al. (described above), Principi et al. [41], and Kriesel et al. [64] each failed to identify an influence of vitamin D on influenza virus-specific responses. In the latter two studies vitamin D supplements were administered to influenza virus vaccine participants but supplements did not improve vaccine-induced immune responses. A study by Lin et al. [45] showed a negative association between vitamin D levels and responses to type B influenza viruses. Our data help to explain the differing results and interpretations of previous literature by showing that baseline vitamin A and baseline vitamin D levels each affect outcome; we found that the benefit of a vitamin A&D supplement was only evident when children were insufficient or deficient in both RBP and 25(OH)D at study entry.
To explain why 25(OH)D could have a negative influence on immune responses to influenza virus vaccines, we consider the vitamin's capacity to clear pathogens. Vitamin D upregulates cathelicidin, an anti-microbial peptide [65,66] that can denature and clear both bacteria and viruses, including influenza virus. Perhaps children who were 25(OH)D replete (due to diet and/or the A&D supplement) rapidly cleared vaccine antigens, thus removing the trigger for an antibody response. The influenza virus vaccine is not adjuvanted and may have been particularly vulnerable to cathelicidin-induced damage. A suggestion of rapid vaccine clearance in the presence of high vitamin levels was similarly proposed by Semba et al. when their vitamin A supplementation of 6 month old infants was found to inhibit antibody responses to the measles vaccine [44]. Another explanation for vitamin D inhibition of HAI may relate to the vitamin's capacity to alter innate and adaptive immune cell trafficking in vivo, and to inhibit B cell proliferation and survival in vitro [1,67]. One final consideration concerns the contrasting influences of vitamins A and D; vitamin D has been positively correlated with serum IgM and IgG3, whereas vitamin A has been positively correlated with other isotypes that might better support HAI [11,[68][69][70]. If vitamin D and its receptor blocked the binding of RAR to DNA response elements, vitamin D may have inhibited positive influences of vitamin A [27].
Our study had limitations because more than one factor could have influenced HAI responses and data interpretation. The phenotypes and functions of individual T cell populations, B cell populations and innate immune cells were not analyzed in our study. T cell and antigen presenting cell (APC) populations are particularly important as drivers of B cell activation, proliferation, and antibody expression. Each can be influenced by vitamin levels. In vitamin A deficient mice, for example, CD103 is significantly upregulated on dendritic cell (DC) and T cell membranes [1,12]. CD103 is the αE component of the αEβ7 integrin, which binds the epithelial cell marker E-cadherin. Changes in homing receptor expression affect DC and T cell trafficking/residence and thereby alter antigen presentation and the "help" T cells provide to virus-specific B cells [1,12,71]. The plethora of indirect and direct influences of nuclear hormones on the B cell response (including nuclear hormone receptor binding to enhancers and switch regions of the immunoglobulin heavy chain locus [28,29,58]) may explain some of the complex outcomes of clinical studies. Our primary data were from season 2016-2017 with the highest enrollment; smaller datasets from other seasons were sufficient only for observation and not for confirmatory analyses. We note differences between HAI responses to the four components in each vaccine. For example, among individuals with low baseline RBP and low baseline 25(OH)D levels in the 2016-2017 season, the most significant improvements in vaccine-induced responses in the A&D group compared to the placebo group were toward the B/Phuket/3073/13 vaccine component. In this case, our analyses of vaccine-induced immune responses benefitted from the harmonious and negative HAI responses toward B/Phuket/3073/13 among study participants at baseline; in contrast, study participants exhibited variable and confounding baseline responses toward the other three vaccine components (Table 1). Another limitation of our study was that subclinical or clinical exposures to cross-reactive viral antigens by study participants before or during vaccination may have contributed to outcomes [72,73]. One individual in the placebo group of the 2016-2017 season (ID #58, Table A2) had a confirmed influenza virus infection and one individual in the placebo group of season 2017-2018 (ID #66, Table A3) reported an exposure to family members with confirmed influenza virus infections.
Despite the limitations noted above, our data provide an explanation for conflicting results in previous vitamin supplementation studies. As in our study, responses to vaccine antigens and vitamin supplements in other studies were likely dependent on baseline vitamin levels [40]. Unfortunately, in past studies, baseline levels of vitamin A, vitamin D, or both, were often untested and/or unreported. Our study provides guidance for the suitability of adding vitamin A and D measurements to future clinical vaccine protocols to improve data interpretation and thereby improve vaccine efficacy and infection control in children. Finally, our results support a recommendation to use vitamin A and D supplements with influenza vaccines in areas where children are frequently vitamin deficient or insufficient but not in areas where children are frequently replete for vitamins A and D.     N/S = no sample.