Twelve healthy participants (11 males and one female) were recruited by local and social media advertisements. Inclusion criteria were that participants were between 18 and 50 years old, had a body mass index (BMI) between 18.5 and 30 kg/m², with a stable weight (<5% body weight change) and constant eating habits over the past 3 months. Exclusion criteria were the previous diagnosis of an inflammatory condition or disease or a history of hypothyroidism, chronic kidney or/and liver disorders, coronary artery disease, malignant hypertension, seizures, involved in intensive sports activities more than four times a week or at top sport level, regular intake of medication that may affect inflammatory response including NSAIDs, psychotic, addictive, or other mental disorders, aversion, intolerance or allergy to cruciferous vegetables and/or palm olein, dextrose, protein supplement, vanilla aroma, the use of dietary supplements with potential effects on antioxidant or inflammatory status and/or viral or bacterial infections requiring the use of antibiotics, laxatives and anti-diarrheal drugs 4 weeks prior to inclusion, excessive alcohol consumption (≥28 consumptions approx. 250 g alcohol per week), pregnancy and/or breastfeeding, reported slimming or medically prescribed diet, as well as adhering to a vegetarian or vegan lifestyle. The sample size calculation is based on a crossover study by Meijer et al. (61) [Trial NL3290 (NTR3435)] in which they examined whether broccoli seedlings could reduce glucose-induced postprandial inflammation in healthy male participants. Meijer et al. measured plasma concentrations of sVCAM-1 and sICAM-1 (primary outcomes) at different timepoints in healthy men after consumption of broccoli seedlings or lettuce (placebo). The detectable difference used for the sample size calculation is calculated based on the mean concentrations of sVCAM-1 (ng/mL). Variance = Mean difference/SD Variance = (26.7–2.4)/14.20 = 1.711.

‘Variance explained by special effect’ was set to 0.5. We calculated an effect size of 0.54. A power of 80% was implemented, the chance of having a type I error was 5% and an effect size of 0.54 (medium effect size). In present study, we have two groups (two repeated measures, within-between interaction) yielding 10 participants, and considering a 20% dropout, which results in the final sample size of 12 participants.

Commercially available broccoli sprouts BroccoCress®, a rich source of glucoraphanin (the parental glucosinolate of sulforaphane), were used as the experimental product. In total, 16 g of sprouts were utilized, equivalent to one serving (portion) of microgreens. Sulforaphane (BroccoCress®) and the placebo (Affilla Cress®) were randomly administered to each participant on separate testing days (as detailed in section 2.3). The period between two visits was 7 ± 3 days. Information about demographics, alcohol consumption, and anthropometric data were assessed on the first visit. Body mass index (BMI), total body fat and visceral fat were measured using the Omron BF511R® monitor. The same testing scheme was applied during two visits (Figure 1), i.e., each participant received a single serving of intervention/placebo, which after 90 min was followed by oral administration of the PhenFlex challenge. Blood samples were collected twice, just before intake of PhenFlex and 120 min after. All participants were instructed to come fasted to each visit, to avoid consumption of broccoli or other cruciferous vegetables 2 days before the visit and to restrain from intense physical activity on the day of the visit. During the visit, participants remained in the testing location and were allowed to drink water ad libitum. No food intake was permitted during the visit.

The broccoli sprouts were cut approximately 1 cm below the leaves (with the hypocotyl being cut below the cotyledons), weighed, and mashed with a small amount of tap water (approximately 13°C) in a kitchen blender for 30s at room temperature immediately before administration (Premium Impuls Blender Smoothiemaker; Impuls; 180 W). Subsequently, tap water (approximately 13°C) was added to the mixture to bring the total amount to 250 mL and participants were instructed to drink the entire mixture immediately. Commercially available pea sprouts (Affilla Cress®) were used as a placebo in this study since pea sprouts do not contain glucoraphanin/sulforaphane. Affilla Cress (16 g) was prepared and administered in a similar fashion. Blinding of participants was ensured by the even appearance of both drinks and the use of nasal plugs during the consumption of the investigational products. The placebo or intervention product was prepared by a researcher who was not involved in any other study procedures and data analysis. Ninety minutes after administration of the investigational products, participants were asked to drink the high-fat, high-glucose, high-caloric product (PhenFlex) (62). For the preparation of the PhenFlex (400 mL, 950 kcal) 60 g palm olein, 75 g dextrose, 20 g protein, 0.5 g artificial vanilla aroma and 320 mL tap water were used (62). In all cases, PhenFlex mixtures were freshly prepared, and the participants were instructed to consume the drink within 5 min.

Samples of venous blood were taken twice per visit from the antecubital vein for measurement of inflammatory and metabolic biomarkers. Samples were collected in 4 mL BD tubes containing K2EDTA as anticoagulant, and centrifuged for 5 min (at 3,000 g, 4°C) within 30 min after collection. Plasma was stored at ≤−80°C until the day of analysis.

Plasma samples were analyzed for inflammatory biomarkers, sVCAM-1, sICAM-1, IL-1β, IL-6, TNF-α, CCL-2, IL-8, IL-10, adiponectin, hs-CRP, and IL-12 p70 using Enzyme-linked immunosorbent assays (R&D Systems Netherlands; Supplementary Table 1). A systemic low-grade inflammation score was generated summing the z-score log-transformed inflammatory biomarkers plasma concentration (sICAM-1, sVCAM-1, IL-6, TNF-α, CRP, IL-12, CCL-2, IL-1β, and IL-8). The z-score log-transformed plasma adiponectin and IL-10 levels were subtracted from the systemic low-grade inflammation score due to their anti-inflammatory properties (51). Plasma samples were analyzed for glucose via colorimetric assay (Cayman Chemical; Glucose Colorimetric Assay Kit, Item No. 10009582) and lipoprotein A using Enzyme-linked immunosorbent assay (Abcam®; Human Lipoprotein A SimpleStep ELISA® Kit, ab212165).

All normally distributed data are presented as mean ± standard deviation (SD). The non-normally distributed data are shown as median (interquartile range). For categorical variables, frequency and/or percentages are presented. Differences between the groups were assessed by paired sample T-tests for normally distributed parameters, or Wilcoxon signed-rank tests for the data that was not normally distributed. Wilcoxon signed-rank tests were performed to check for potential carryover effects. Associations between clinical features and biomarkers were assessed using Pearson (normally distributed parameters) or Spearman’s rank (non-normally distributed) correlation coefficient. All analyses were performed, two-tailed with a p ≤ 0.05 considered statistically significant.

Between November 2021 and January 2022, a total of 12 subjects were enrolled into the study and randomly allocated to either initial administration of sulforaphane or placebo. Baseline characteristics of the study population (n = 12) are summarized in Table 1. All participants completed the study and were included in the data analysis for biochemical testing (Figure 2). The pre-test (Wilcoxon rank sum test) revealed no differences between treatment allocations for all parameters (all z < 0.00, p > 0.14). Absence of sulforaphane in placebo was verified by testing urine samples which showed significantly higher concentration of the total metabolites of sulforaphane after intake of broccoli sprouts (8.2 vs. 0.4 μmol, p < 0.001).

The single serving of sulforaphane or placebo induced no significant changes in concentrations of sICAM-1 and sVCAM-1 in healthy participants before and 2 h after the PhenFlex challenge. A general trend was seen in the enhancement of endothelial activation in the sulforaphane group, however not significant; sICAM-1 (sulforaphane 1.5 ± 10.1 vs. placebo 3.1 ± 7.8 ng/mL, p = 0.696) and sVCAM-1 (sulforaphane 3.1 ± 5.2 vs. placebo 0.9 ± 4.5 ng/mL, p = 0.431; Table 2; Figure 3).

Eleven inflammatory biomarkers, sICAM-1, sVCAM-1, IL-6, TNF-α, hs-CRP, adiponectin, IL-12 p70, CCL-2, IL-10, IL-1β, and IL-8, were measured in plasma before and 2 h after the PhenFlex challenge. Levels of sICAM-1, sVCAM-1, hs-CRP, adiponectin, CCL-2, and IL-10 and changes are listed in Table 2. Changes in CCL-2, measured as differential concentrations before and 2 h after caloric load, showed a significant change between groups, with sulforaphane causing a significant increase in this biomarker compared to placebo [1.9 (3.3) vs. 0.0 (4.8) pg./mL, p = 0.017; Figure 4]. Changes in sICAM-1, sVCAM-1, hs-CRP (sulforaphane 2.2 ± 4.3 vs. placebo −0.5 ± 2.9 pg./mL, p = 0.275), and IL-10 [sulforaphane −0.6 (26.3) vs. placebo 4.4 (5.1) pg./mL, p = 0.715] revealed an overall slight and statistically non-significant pro-inflammatory effect of sulforaphane (Table 2). No detectable levels of IL-6, TNF-α, IL-12, IL-1β, and IL-8 were quantified.

Aside from CCL-2, a more robust change was observed with the integration of the individual biomarkers in the composite systemic low-grade inflammation score (Table 3; Figure 4). In the sulforaphane group the composite score revealed a pro-inflammatory trend after caloric challenge [−0.092 (1.06) before vs. after 0.018 (1.06), p = 0.087] which was less prominent in the placebo group [−0.001 (0.81) before vs. after 0.014 (0.81), p = 0.251].

Before the PhenFlex challenge, no significant correlations between inflammation biomarkers and demographic, anthropometric, and lifestyle data were observed, with the exemption of IL-10 and smoking (p = 0.042), and the number of cigarettes per day (p = 0.042). Moreover, moderate strength correlations between the systemic low-grade inflammation score and fat percentage (r_s = −0.551, p = 0.079) and the number of cigarettes per day (r_s = 0.557, p = 0.075) were observed.

The single serving of sulforaphane or placebo induced no significant changes in concentrations of glucose and lipoprotein A in healthy participants before and 2 h after the PhenFlex challenge (Table 4). Before the PhenFlex challenge, fasting glucose levels correlated negatively with age (p = 0.003), BMI (p = 0.046) and fat percentage (p = 0.034), and positively with adiponectin concentrations (p = 0.015). Fasting lipoprotein A concentrations correlated positively with CCL-2 (p = 0.049) and IL-10 levels (p = 0.005; Table 5). Moreover, administration of sulforaphane caused changes in glucose levels in response to the caloric load which correlated positively with changes in sICAM-1 (p = 0.006), adiponectin (p = 0.048) and lipoprotein A (p = 0.005). Changes in lipoprotein A concentrations in response to the challenge after sulforaphane administration correlated positively with changes in sICAM-1 levels (p = 0.004). In the placebo group, changes in glucose levels in response to caloric loading only negatively correlated with changes in IL-10 (p < 0.001; Table 6).

In the present study, metabolic overload did not significantly affect plasma adhesion marker levels in healthy participants, as measured 2 h after caloric overload. In contrast, previous research has shown that a single administration of the PhenFlex challenge increased the levels of sVCAM-1 and sICAM-1 after 2 h in healthy volunteers (18). Additionally, Derosa et al. demonstrated significant increases in these plasma adhesion markers within 2 h after an oral glucose tolerance test (OGTT) (63). A possible explanation for the lack of effect found on these markers in our study is the fact that the subjects in the current study were given whole food products (sprouts of broccoli or pea) before undergoing the challenge. Both products contain retinol, vitamin E and ascorbic acid, which could have counteracted the expected transient disruption of post-exposure endothelial homeostasis observed in other studies. This hypothesis is supported by findings of Nappo et al., who showed that supplementation with vitamin C and E prevented an increase in sICAM-1 and sVCAM-1 in healthy middle-aged subjects after a high-fat meal (64). Furthermore, Rubin et al. found no changes in plasma adhesion markers in young participants (25 years on average), after a standardized lipid-rich meal which contained retinol (65). Thus, in this study, the effect of sulforaphane on endothelial homeostasis may have been influenced by the other nutrients present in the whole food product. Nonetheless, the associations between metabolic parameters and inflammatory biomarkers during the PhenFlex challenge, particularly between plasma sICAM-1, and glucose and lipoprotein A concentrations in the sulforaphane group, provided relevant information on the modulation of endothelial function and metabolic homeostasis by sulforaphane in response to a high-glucose, high-fat product. Consistent with our findings, Chen et al. demonstrated significant relationships between the plasma glucose and insulin responses to an OGTT and plasma sICAM-1 concentrations in healthy participants (66). The fact that these correlations were not observed in the pea sprouts group (placebo) supports the hypothesis that the other bioactive compounds in the whole food products may have blunted the transient disruption of post-exposure homeostasis expected after caloric load. This is also revealed in part by the only correlation between circulating IL-10 and plasma glucose. The presence of other nutrients in broccoli sprouts may have interfered with the strong effects of sulforaphane, which, however, were still evident as more correlations were shown in this group.

In this study, sulforaphane facilitated the development of a mild pro-inflammatory state during caloric challenge, as evidenced by a moderate increase in sICAM-1, sVCAM-1, hs-CRP, CCL-2 and decrease in IL-10. The effects of dietary intervention on chronic inflammation in other studies that used the PhenFlex challenge are inconsistent (62, 67). Kim et al. examined the effect of a single-intake microencapsulated garlic powder and/or tomato extract in healthy male smokers during the metabolic challenge. Consumption of tomato extract elicited a differential response, increasing CCL-2 and decreasing sVCAM-1 6 h after PhenFlex compared to placebo. When garlic powder was consumed, IL-13 levels decreased after 2 h and IL-1α increased 6 h after the challenge, indicating a pro-inflammatory effect of the food product. The combination of interventions elicited a mixed response, with IL-10 and CCL-7 being reduced 6 h after the metabolic challenge (67). Hoevenaars et al. investigated the effects of a 12-week whole grain wheat (WGW) intervention compared to refined wheat (RW) and observed increased CRP, IL-6, IL-8, and decreased IL-1β in RW and decreased CRP, serum amyloid A, IL-8, and IL-10 in WGW, indicating pro- and anti-inflammatory effects in respective groups (62). These inconclusive results highlight the importance of implementing integrative outcome measures to unravel the subtle, pleiotropic effects of phytonutrients. As an illustration, the study of Weseler et al. testing the effects of grape seed extract on multiple biomarkers reflecting vascular health integrated them into a vascular health index, which unveiled an improvement in overall vascular health from flavanols, which was less clear from the analysis of individual outcomes (13, 16). In addition, previous cross-sectional studies evaluated CLGI through an index that pools multiple indicators to provide a better overall picture of the synergistic changes of inflammatory biomarkers (51, 68–72).

To the best of our knowledge, this experimental study is the first to examine the effects of phytonutrients on calorie-induced inflammation as measured by a composite scoring system. Intriguingly, the score more accurately reflected the pro-inflammatory effect of broccoli sprouts than single biomarkers during phasic response. In addition, the relationships between risk factors for the development of NCDs such as high visceral fat and smoking, and inflammation became more apparent through the use of the score. Specifically, fat percentage and smoking showed a moderate inverse relationship with the score. Using the systemic low-grade inflammation score also led to another finding; five of the 11 inflammatory biomarkers were undetectable in the blood of our young and healthy population. These findings suggest that even a well-characterized scoring system may show limitations. For future research, assessing the health status or risk profile of the test population and adjusting the scoring system could be beneficial. The six biomarkers detectable under basal conditions in our study may be more suitable for challenge testing in young healthy subjects (19, 36, 73–76).

Over the past few decades, research has increasingly focused on antioxidants as the main health-promoting compounds in fruits and vegetables, leading to a gigantic array of antioxidant supplements on the market today (5–7, 12, 77–79). However, initial excitement regarding the potential health benefits of antioxidants failed to be confirmed by clinical evidence (12). There is quite a bit of debate about whether supplementing with antioxidants is healthy, ineffective, or even harmful (5, 11, 12, 22, 28, 36, 77, 78, 80–86). So far, whole foods and fresh produce have not shown a clear protective effect against the PhenFlex challenge (62, 67). In fact, dietary interventions high in hormetins facilitated the development of a mild pro-inflammatory state during caloric overload, e.g., broccoli sprouts and garlic extract (67). We hypothesize that this moderate pro-inflammatory state in the sprouts containing sulforaphane may be due to the initial pro-oxidative action (to activate Nrf2) of hormetins present in fresh produce. As a result, the exogenous antioxidant capacity of the direct antioxidants present in both sprouts is blunted in the broccoli sprouts compared to the pea sprouts. This, in turn, led to a reduced initial integrative anti-inflammatory capacity against the caloric overload demonstrated by the broccoli sprouts compared to the placebo. However, we speculate that because sulforaphane also enhances endogenous antioxidant defenses via Nrf2 activation at a later stage, whole foods that increase both exogenous and endogenous antioxidants may have more significant effects on phenotypic flexibility (Figure 5). This may explain why supplementation of direct antioxidants such as ascorbic acid and vitamin E attenuates the metabolic stress of caloric loads (64, 65), while hormetins in fresh produce, e.g., sulforaphane, diallyl sulfide, withaferin A and rutin, induce a mild pro-inflammatory effect via activation of the Nrf2-pathway (38, 49, 50, 67, 87). We hypothesize that the health effects of fruit and vegetable consumption are due to the wide variety of bioactive compounds in the food matrices and the synergy between the different mechanisms of action of these phytonutrients in the body, rather than just antioxidants (36). One aspect of synergy may be a buffering effect (88, 89). The effect of a large intake of a given nutrient may vary depending on whether it is taken in concentrated form or as part of a food matrix, e.g., the matrix may slow down the absorption of the nutrient, which lowers the likelihood of a bolus effect (89).

The limitations of our study include a small sample size and a short observation period. A longer time of observation (6, 8, 12, or even 24 h) could have helped to demonstrate that the increase in inflammatory activity caused by sulforaphane represents the initial part of the hormetic response. In fact, previous research, conducted with larger sample sizes, has demonstrated the sustained anti-inflammatory effects of sulforaphane (56, 90). At the same time, biotechnological advancements that allow continuous monitoring of certain functions (e.g., glucose) and innovative designs (e.g., n-of-1 trials) may enable more accurate research into personalized nutrition strategies in the future (91–94). Learning more about the interplay between phytonutrients may eventually reveal whether “an apple a day can keep the doctor away”—at least for a while.