Milk polar lipids favorably alter circulating and intestinal ceramide and sphingomyelin species in postmenopausal women

BACKGROUND High circulating levels of ceramides (Cer) and sphingomyelins (SM) are associated with cardiometabolic diseases. The consumption of whole fat dairy products, naturally containing such polar lipids (PL), is associated with health benefits, but the impact on sphingolipidome remains unknown. METHODS In a 4-week randomized controlled trial, 58 postmenopausal women daily consumed milk PL-enriched cream cheese (0, 3, or 5 g of milk PL). Postprandial metabolic explorations were performed before and after supplementation. Analyses included SM and Cer species in serum, chylomicrons, and feces. The ileal contents of 4 ileostomy patients were also explored after acute milk PL intake. RESULTS Milk PL decreased serum atherogenic C24:1 Cer, C16:1 SM, and C18:1 SM species (Pgroup < 0.05). Changes in serum C16+18 SM species were positively correlated with the reduction of cholesterol (r = 0.706), LDL-C (r = 0.666), and ApoB (r = 0.705) (P < 0.001). Milk PL decreased chylomicron content in total SM and C24:1 Cer (Pgroup < 0.001), parallel to a marked increase in total Cer in feces (Pgroup < 0.001). Milk PL modulated some specific SM and Cer species in both ileal efflux and feces, suggesting differential absorption and metabolization processes in the gut. CONCLUSION Milk PL supplementation decreased atherogenic SM and Cer species associated with the improvement of cardiovascular risk markers. Our findings bring insights on sphingolipid metabolism in the gut, especially Cer, as signaling molecules potentially participating in the beneficial effects of milk PL. TRIAL REGISTRATION ClinicalTrials.gov, NCT02099032, NCT02146339. FUNDING ANR-11-ALID-007-01; PHRCI-2014: VALOBAB, no. 14-007; CNIEL; GLN 2018-11-07; HCL (sponsor).

There is growing evidence showing that several factors may affect circulating SM and Cer concentrations such as drugs and lifestyle modifications, including exercise and dietary changes (15,16). SP is found in plant and animal cell membranes; the daily SP intake represents approximately 0. .4 g/day in humans (17), but the impact of such consumption on the endogenous sphingolipidome remains largely unknown. Cow milk has recently attracted more attention because it naturally contains SM and Cer (SM accounting for ~25% of PL in the milk fat globule membrane [MFGM]), including C20+22 species (i.e., C20:0, C20:1, C22:0, and C22:1 species) that are found in higher amounts in milk fat than in human blood. Recent metaanalysis and research papers highlight the beneficial cardiometabolic effects of the consumption of whole fat dairy product, which contain sizable amounts of milk PL (18,19). Given that milk PL also contain C16+18 SM and C24:1 Cer species with potential deleterious effects, understanding the impact of milk SP on circulating levels and endogenous SP metabolism is of great importance.
Preclinical studies revealed that milk SM supplementation prevents hyperlipemia, hypercholesterolemia, and low-grade inflammation and that it improves intestinal health related to gut microbiota modulations (2, [20][21][22][23][24][25][26]. Buttermilk, a by-product of butter industry, represents a natural source of the SM-rich MFGM (1-10 g of milk PL/L) (27,28). In the VALOBAB-C trial, we demonstrated recently that the 4-week consumption of milk PL-enriched test cheese decreases circulating total cholesterol (C) (primary outcome), triacylglycerols (TAG), LDL-C, and ApoB, at both fasting and postprandial states (Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.146161DS1) in postmenopausal women at risk of cardiovascular disease (CVD) (29). However, there are still open questions regarding the potential involvement of the endogenous metabolism and intestinal fate of milk SP species in these benefits. We thus explored the prespecified secondary outcomes of the VALOBAB clinical trial by analyzing SM and Cer molecular species of particular interest in several biological compartments. We aimed to determine how the 4-week supplementation with milk PL impacts the circulating and fecal SM and Cer species at the fasting state, as well as their amount in intestine-derived chylomicrons during the postprandial period, in postmenopausal women at cardiovascular risk ( Figure 1). We then verified whether this contributes to the beneficial effects of milk PL on lipid cardiovascular risk markers. In a complementary study conducted in ileostomy patients, we further aimed to identify the digestive fate of milk SP in the upper gastrointestinal tract after acute consumption of milk PL-rich meals ( Figure 1).
Milk PL decrease SM content in intestine-derived chylomicrons and impact their SM and Cer molecular profiles. The variations of plasma concentrations of chylomicron-rich fraction-bound SM (CMRF-SM) decreased in the 5 g-PL group during all the postprandial period (P group = 0.015; P CTL-5g = 0.013), and the variation of plasma CMRF-Cer concentration also tended to decrease (P group = 0.053; P PL = 0.051) (Supplemental Table 3). To focus on potential modifications of chylomicron lipid composition regardless of their circulating concentration, we also determined their enrichment in SP by analyzing the SM/ TAG and Cer/TAG ratios in CMRF particles. Milk PL significantly reduced the CMRF-SM/TAG ratio ( Figure 3A; P group = 0.00095; P PL = 0.00026; P CTL-3g = 0.001; P CTL-5g = 0.009), notably after lunch that contained the test cream cheese (240-480 min). CMRF-Cer/TAG ratio also significantly decreased in milk PL-treated groups, regardless of dose (P group = 0.071; P PL = 0.024) ( Figure 3B). Milk PL effects on SM molecular composition in intestine-derived chylomicrons were mainly mediated by a significant decrease in several CMRF-SM species content relative to CMRF-TAG including C16:0, C16:1, C18:0, C18:1, C20:1, C24:0, and C24:1 SM species (Figure 3  Data are presented as mean ± SEM, n = 10/group. Results are presented based on the assumption of sphingosine d18:1 as the major sphingoid base for determined SM and Cer species. P group represents P value associated with group effect as calculated by generalized linear model, while p PL represents P value associated with binary effect of milk PL compared with control. A,B Different superscript letters indicate statistically different intervention effects between groups as calculated by post hoc analysis. P < 0.05 was considered significant. C P value remains significant (< 0.05) after adjustment for clinical center, quartiles of volunteer age and waist circumference. See Supplemental Table 2.  we explored the molecular composition of SM and Cer species in fecal samples collected by the postmenopausal women included in the VALOBAB-C trial.

Discussion
This study is the first to our knowledge to report how the daily consumption of a significant amount of dietary SP present in milk PL impacts the endogenous sphingolipidome in the bloodstream and along the gastrointestinal tract in humans. Firstly, we reveal that the increased intake of milk SM and Cer did not increase their total amount in serum, but the molecular composition of SM and Cer species was markedly improved by the 4-week intervention with milk PL. Notably, the atherogenic C16+18 SM and C24:1 Cer species decreased significantly, despite their increased intake from the provided supplementation. These variations even correlated with the beneficial impacts of milk PL on lipid cardiovascular markers reported previously (29). In addition, the Mayo Clinic published the reference values for circulating level of C24:1 Cer species (i.e., 0.65-1.65 μM; https://www.mayocliniclabs.com). Here, we found that serum C24:1 Cer species concentration returned within the normal range after the intervention in the 5 g-PL group only (V1, 1.96 ± 0.17 μM; V2, 1.59 ± 0.16 μM; P group = 0.033). Our results demonstrate that milk PL supplementation positively impacts the endogenous sphingolipidome, with the specific decrease of serum SM and Cer species known for being associated with inflammation and metabolic disorders (31,32). Previous studies reported that high concentrations of serum C18:0, C20:0, and C24:1 Cer species are associated with type 2 diabetes, while high serum levels of C16:0 Cer and C18:0 SM species correlate with insulin resistance (33). Regardless of milk PL dose, the analysis of the relative abundance of each SM and Cer species in the bloodstream revealed a significant increase in the proportions of C20:0 SM, C22:1 SM, and C20:0 Cer species that are normally poorly detected in human blood but found in nonnegligible amount in MFGM. We also estimated to what extent the changes in major SM or Cer species could explain the relationships between milk PL consumption and enhanced lipid cardiometabolic risk factors. According to the estimated regression coefficients, assuming a mean variation of C16+18 SM species of about -22 μM, as observed in the 5 g-PL group, its mean effect is expected to be (a) -0.16 mM on LDL-C (with a global effect of -0.34 mM observed in this group) (29) and (b) -0.19 mM on total C (with an observed effect of -0.4 mM). Moreover, assuming a mean variation of the relative proportions of C24+26 Cer of about -0.83% as observed in the 5 g-PL group, its mean effect is expected to be (a) -0.032 mM on LDL-C and (b) -0.045 mM on total C. These results show that changes in serum total C and LDL-C are significantly associated with changes in serum C16+18 SM species, and to a lower extent with the modulation of the relative proportion of C24+26 Cer species (here mainly driven by the variation of C24:1 Cer species).
To investigate underlying mechanisms involved in the effects of milk PL consumption on circulating SP species, we first estimated the contribution of intestine-derived chylomicrons, which are the dietary lipid carriers secreted by the small intestine during the postprandial phase. Chylomicrons represent a major source of circulating SM, although the mechanisms by which SM is inserted into these lipoproteins have For panels A-C, and G, asterisks in bold represent correlations that remain significant after adjustment for clinical center, quartiles of volunteer age, and waist circumference. (D-F and H-K) Graphs illustrating specific Spearman's correlations between the intervention impact on C16+18 SM species and on LDL-C (D), total C (E), and ApoB48 (F); between C20+22 Cer species proportions (%) and LDL-C (H), and total C (I); between C24-26 Cer species proportions (%) and LDL-C (J), and total C (K). Apo, apolipoprotein; C, cholesterol; Cer, ceramides; CTL, control; HDL, high density lipoprotein; LDL, low density lipoprotein; NEFA, nonesterified fatty acids; PL, polar lipids; SM, sphingomyelin; TAG, triacylglycerols.

C L I N I C A L M E D I C I N E
JCI Insight 2021;6(10):e146161 https://doi.org/10.1172/jci.insight.146161 not been established (34). Dietary SM and Cer are not absorbed as such; their lipolysis products released in the small intestine can be absorbed and a small proportion of their sphingoid bases contribute to the newly formed SP ultimately found in chylomicrons (34). Herein, the 4-week milk PL supplementation decreased chylomicron total SM and Cer, especially during the second part of the postprandial period (after test cheese consumption), without change in particle size (i.e., no change in the surface/TAG core ratio) (29). Data are presented as mean ± SEM (n = 6 / group). The vertical dotted line represents the intake of the meal including the control or milk PL-rich dairy (according to group). The P group and P posthoc are shown for the postprandial period from 120 to 480 minutes. Statistical analysis was done using a linear mixed model followed by Tukey-Kramer's post hoc test. P posthoc corresponds altogether to P CTL versus 3g-PL ; P CTL versus 5g-PL and P 3g versus 5g-PL . Results are presented based on the assumption of sphingosine d18:1 as the major sphingoid base for determined SM and Cer species. Cer, ceramides; CMRF, chylomicron-rich fraction; CTL, control; PL, polar lipids; SM, sphingomyelins; TAG, triacylglycerols. See Supplemental Table 3 and Supplemental Figure 2, VALOBAB-C trial.  These modifications were also observed at species level for almost all SM and Cer species, including those whose concentrations in total serum decreased -namely C16:1 SM, C20:1 SM, and C24:1 Cer. Because SP are located at the surface of lipoproteins, this reveals a lower SP amount in the chylomicron composition. Whether this is due to decreased SM synthesis in enterocytes after intervention with dietary SP remains to be elucidated. Milk PL also modified the SM molecular profile in chylomicrons with an increase in the proportions of C20 SM species and a decrease in the proportions of C24:1 SM species -a possible precursor of C24:1 Cer via acid SMase (35). These results suggest that these modifications may originate from the gut or from enterocyte metabolism during the intestinal digestion and absorption processes.
To determine the contribution of SM and Cer in the intestine, we analyzed their molecular profiles in the gut lumen of ileostomy patients after the acute intake of milk PL, as well as in feces of the postmenopausal women after 4-week milk PL supplementation. These analyses revealed an increase of both total SM and Cer in gut contents in milk PL supplemented groups. At a molecular level, the amounts of most detected SP species of interest were significantly increased by milk PL consumption, especially C16:0, C22:0, C24:0 SM and Cer, and also C20:0 SM species. Altogether, SM and Cer species whose amounts increased in gut contents reflect species that are present in milk PL-enriched cheeses. These results are consistent with the fact that SM digestion is incomplete, as only 75%-80% of milk SM was reported to be digested and absorbed in humans (36). It has been previously reported that ileal efflux of C16:0 SM was only ~10% of ingested dose versus ~20% for C24:0 SM after intake of lower doses of SM -i.e., 50-200 mg (36) -suggesting that longer-chain saturated species of SM and Cer are less efficiently digested and absorbed. In addition, with the digestion of SM being slow and incomplete, it may induce an important increase of nondigested SM and nonabsorbed Cer in the lumen content (3,30), which may explain the present results. Moreover, fecal metabolites, including the various lipid species normally found in feces, may originate directly from food, but also from host cells or bacterial cell components, or indirectly from the molecular conversion of SP by gut microorganisms or host enzymes (37). In a recent study performed in healthy patients, plasma and fecal lipidomic analyses demonstrated that the lipid fraction of fecal samples contains significant amounts of Cer species, with only 2 SM species detected, while plasma samples commonly contain significant amounts of several SM species and lower quantities of Cer (37,38). Herein, we chose to determine the concentration of 12 SM and Cer species of interest in serum, and we were also able to quantify all these species in feces. In ileostomy patients, we report a higher total SM amount in ileal content compared with total Cer, while total Cer was largely more abundant in the fecal samples compared with total SM. The latter could be the result of several metabolic pathways such as the conversion of dietary SM species in Cer species by host enzymes present in the lumen and in enterocytes. At a molecular level, monounsaturated SM and Cer species increased in both ileal efflux and fecal samples, despite being found in minority in test cheeses compared with saturated species. A potential differential absorption process between monounsaturated and saturated species would, thus, deserve to be investigated. Interestingly, the major changes reported in the serum after the dietary intervention with milk PL mainly concern monounsaturated SM and Cer species. It may suggest that some modifications of serum and chylomicron SP profiles occur in response to changes in the SP fate in the small intestine.
The increased amount of total Cer reported in the fecal samples of milk PL-supplemented volunteers could also be the result of gut bacteria metabolism because several bacteria, including those belonging to the Bacteroides genus, were reported to be able to produce SP (39,40). Very recently, Lee et al. demonstrated in female mice that sphinganine, which is the main sphingoid base of SM and Cer present in MFGM, is assimilated by gut bacteria (41). In this study, 99% of gavaged fluorescent sphinganine was assimilated by Bacteroides spp., with the remaining 1% by Prevotella spp., Lactobacillus spp., and Bifidobacterium genus (41). Also, Bifidobacterium spp., which are known to be increased after milk SM consumption in rodents (20,26), can release free milk Cer by hydrolyzing milk gangliosides (42). In this context, the contribution of the gut microbiota SP metabolism in the effects of milk PL consumption on the intestinal and circulating SM and Cer profiles cannot be ruled out.
To further explore potential endogenous mechanisms, we analyzed whole blood cells gene expression of some key enzymes involved in SM synthesis (SM synthase 1 and 2 [SGMS1 and SGMS2, respectively]) and hydrolysis (an acid sphingomyelinase; SMase, also called SM phosphodiesterase 1 [SMPD1]). We found only slight effects on SGMS1 and SGMS2 expression (Supplemental Figure 4), while SMPD1 expression decreased in milk PL-treated groups compared with control (P group = 0.052; P PL = 0.030). As previously described, acid SMase activity in plasma is increased in acute coronary syndromes (43). However, these results were not likely to explain the changes observed in the circulating sphingolipidome. It would JCI Insight 2021;6(10):e146161 https://doi.org/10.1172/jci.insight.146161 also be relevant to consider the possible contribution of intestinal enzymes, given that the small intestine is rich in enzymes known to contribute to SP metabolism, such as alkaline SMase that converts SM in Cer (35). Unfortunately, we could not collect intestinal biopsies from healthy postmenopausal women in the present study. However, in an 8-week milk PL supplementation performed (0.9 g per 100 g of diet) in high-fat diet-fed mice (22), we observed a significant increase of jejunal expression of Enpp7, coding for the alkaline SMase, compared with the high-fat control group (1.7-fold change). Interestingly, previous preclinical studies reported opposite impacts of Cer production depending on SMase activity; Cer generated from neutral or acid intestinal SMases are more prompt to exert proinflammatory effects, while Cer generated from alkaline SMase promote antiinflammatory pathways (44,45). The conversion of exogenous SM in Cer by the alkaline SMase could also play a role in the inhibition of cholesterol absorption (45,46), which is concordant with the present findings and supports the role of SM metabolism in cholesterol absorption. Previous preclinical studies demonstrated that dietary SM are able to play a beneficial role on cholesterol levels and more largely in the prevention of cardiometabolic disorders (25,26,47,48). In mice fed high-fat diets, supplementation with egg SM lowered intestinal absorption of cholesterol and lipids with a reduction of hepatic cholesterol (49). In vitro, both SM and Cer inhibit cholesterol absorption in Caco-2 intestinal epithelial cells (45). However, it has been suggested that small SM catabolites, such as Cer and sphingosine, might be the effectors of the beneficial impact of milk SM (2). The present findings consolidate our previous clinical results, given that the observed reductions in circulating total cholesterol, LDL-C, and ApoB (29) significantly correlate with the reduction of serum proinflammatory C16+18 SM in the milk PL-treated groups. In accordance with above-mentioned studies, our findings also bring potentially new information and insights on Cer in the gastrointestinal tract as signaling molecules potentially participating in the beneficial effect of milk PL consumption on cholesterol metabolism.
The present study has several strengths, but it also has some limitations that need to be outlined. The clinical trials were performed in a targeted population (i.e., overweight postmenopausal women) known to present an important risk of CVDs, but results cannot be extrapolated to other populations. We took care to include 4-day dietary records before and after the nutritional intervention to show that volunteers of the 3 groups did not differentially modify their energy and dietary intakes (Supplemental Table 5). Many parameters of the present study were measured in a limited number of subjects. Nevertheless, we performed, for the first time to our knowledge, a broad sphingolipidomic analysis, including a large scale of measurements of SM and Cer species at both fasting and postprandial states in various biological compartments -serum, chylomicron fractions, and feces -but also in 8 h-cumulative ileal efflux from ileostomy patients. The sphingolipidome is a complex and dynamic system that encompasses several important SP families, including dihydroceramides, gangliosides, and cerebrosides (17). Considering the variations in the sphingoid bases, FA, and headgroups of SP molecules, the number of species exceeds thousands. SP are localized in cellular membranes (lipid rafts) and are carried by albumin, lipoprotein particles, blood cells, and platelets in the bloodstream (50,51). Based on the present findings, future studies should, thus, explore the sphingolipidome in other blood compartments and, potentially, epithelial cells to better understand the fate of milk SM and Cer species. Furthermore, we cannot exclude the potential contribution of other components of the PL fraction/MFGM from buttermilk concentrate and/or the lower milk TAG content in the PL-enriched cheeses in the reported metabolic effects in both trials. Putting aside these limitations, this is -to the best of our knowledge -the first time that such a wide sphingolipidomic analysis has been performed in humans in response to a controlled dietary intervention in the context of cardiometabolic disorders. The present study clearly responds to the need to identify relevant dietary strategies to improve the endogenous SP metabolism, which was highlighted in recent reviews (2,16).
The present findings uncover that milk PL supplementation providing particular SP species markedly improved the endogenous sphingolipidome by reducing serum atherogenic C16+18 SM and C24:1 Cer species in overweight postmenopausal women at risk of CVD. These reductions in SP were (a) correlated with and (b) significantly involved in the decrease of lipid cardiovascular risk markers induced by milk PL intervention. We further demonstrate that, despite a significant ingestion of SP provided by milk PL, SM and Cer concentrations decreased in intestine-derived chylomicrons, while their concentration increased in gut contents. The related differences in SM and Cer profiles between gut contents and circulating compartments suggest that small intestinal mechanisms occurred during digestion and absorption processes of milk SM and Cer, and that a contribution of the gut microbiota may be possible. Considering that milk PL are naturally found in large amounts in buttermilk, which is still poorly valued in human food, such bioactive lipids could be envisioned as promising ingredients for the development of new functional foods, providing health effects in the frame of chronic diseases. VALOBAB-C trial. Details of the VALOBAB-C study have been published previously (29). Briefly, the multicenter study used a double-blind randomized placebo-controlled parallel design and was conducted in 58 overweight postmenopausal women, without metabolic syndrome but at risk of CVD. The eligibility criteria and sample size calculation have been described previously (29). Volunteers were randomly divided into 3 groups. Randomization was performed electronically using a random number generator and supervised by a biostatistician (29). Both volunteers and investigators were kept blind regarding group allocation. Volunteers were subjected to the daily consumption of either control or milk PL-enriched cream cheese (100 g of cream cheese containing 13 g of total fat including 0 [control], 3, or 5 g-milk PL during 28 days; n = 19, 19, and 20, respectively). The strategic approach was to formulate cheeses with identical total lipid content with partial substitution of TAG by milk PL to avoid increased energy intake. The 3 g-and 5 g-PL cream cheeses were based on a butterserum concentrate rich in milk PL prepared according to Gassi et al. (52) representing a 3to 5-fold increased daily consumption of milk SM and Cer compared with an estimated intake of dairy SP in Western countries (Supplemental Table 1) (17). After the "run-in" period, volunteers were subjected to a first exploratory visit (V1), followed by 28 days of intervention and ended by a second exploratory visit (V2). During each visit, overnight-fasted participants received a breakfast meal rich in fat and carbohydrates, and 4 hours later, they consumed a standardized lunch containing the corresponding test cream cheese, thus dividing the exploratory visit in 2 specific postprandial periods (0-240 min and 240-480 min, as detailed previously) (29). Volunteers were asked to continue their usual diet and physical activity throughout the study. Participants were told to avoid the consumption of cheeses other than the test cream cheese and to avoid listed foods that may influence the gut microbiota composition. Particular attention was drawn to standardize the meal consumed the evening before each postprandial exploratory visit. Subjects recorded their food consumption for 4 days before and after the nutritional intervention. No difference in changes in energy and macronutrient intakes, fibers, alcohol, or cholesterol or FA intakes was observed between groups (29). The primary outcome was the impact of the 4-week milk PL consumption on fasting serum concentration of total C (29). The predefined secondary outcomes tested in the present study were related to the impact of the dietary intervention on serum, chylomicron, and fecal SM and Cer profiles. Considering available samples and practical/technical aspects, some analyses were performed on a subgroup of individuals only (Figure 1).

Methods
VALOBAB-D trial. The double-blind randomized controlled trial (RCT) was performed in 4 ileostomy patients following a crossover design, as previously described (29) (Supplemental Figure 5). An ileostomy is a surgical opening in the abdomen in which a piece of the ileum is brought outside the abdominal wall to create a stoma through which digestive contents leave the body and are collected in a pouch (ileal efflux). Selected patients according to eligibility criteria were invited to participate in 3 distinct exploratory visits separated by a 4-to 6-week washout period (29). During each visit, overnight-fasted patients consumed one of the test cream cheese containing 0, 3, or 5 g-milk PL, and their ileal efflux was collected over 8 hours. Sequences of meal allocation were based on a random number generator (29). Both patients and investigators were kept blind regarding meal allocation.
Isolation of CMRF. Isolation of intestine-derived CMRF was performed by ultracentrifugation from plasma collected at different time points, as previously described (29,53).
Analysis of serum phospholipids. Total lipids were extracted from 300 µL of serum with chloroform/methanol (2:1, v/v) according to the method of Folch et al. (54). After drying under nitrogen, total lipids were determined gravimetrically and were dissolved precisely with 1 mL of chloroform/methanol (2:1, v/v). This stock solution of total lipids was stored at -20°C. Phospholipid classes were then separated by high-performance liquid-chromatography (HPLC) coupled to an evaporative light-scattering detector (SEDEX LT-ELSD SOLT, HPLC DDL Sedere, Thermo Fisher Scientific) (55,56), using a silica normal-phase column (Lichrospher Si 60, 3 μm, 100 × 4.6 mm, Waters). The chromatographic separation was carried out using a linear binary gradient according to the following scheme: t0 minutes: 90%A, 10%B 0%C; t20 minutes: 42%A, 52%B, 6%C; t30 minutes: 32%A, 52%B, 16%C; t55 minutes: 30%A, 70%B, 0%C; t60 minutes: 90%A, 10%B, 0%C. Total chromatographic run time was 75 minutes per sample, which consisted of a 60-minute analysis and 15 minutes to restore initial conditions and reequilibration. Eluent A consisted of hexane/tetrahydrofuran (99:1, v/v), eluent B of isopropanol/chloroform (80:20, v/v/v), and eluent C of isopropanol/water (50:50, v/v/v). The flow rate of the eluent was 1 mL/min. Identification of phospholipids and lysophospholipids was carried out by comparison with the retention time of pure standards (Avanti polar Lipids). Calibration curves for each compound were calculated from the area values of stock solution JCI Insight 2021;6(10):e146161 https://doi.org/10.1172/jci.insight.146161 and figures. For parameters analyzed along the postprandial period, a mixed linear modeling (MIXED procedure) was performed to account for within-subject repeated measures, seeking for main effects, at least "group" or time effect and interaction. Post hoc analyses were performed following Tukey-Kramer's test to both detail main effects and control for familywise type I error. In case of residual distribution departing from normality, the analyses were performed on ranks. Global "milk PL" effect was also considered as binary factor, and statistical analysis was performed by lumping together milk PL doses in one group versus control. Spearman's correlation analyses were also performed between blood lipid markers of cardiovascular risk and serum SM and Cer species grouped in 3 subclasses (i.e., C16+C18, C20+C22, and C24+C26 SM or Cer species). In order to check for any confounding effect, these analyses were also carried out adjusting for center, age, and waist circumference quartiles. Analyses were performed on SAS v9.4 (SAS Institute Inc. ) with a 2-sided type I error set at 0.05. In order to determine to what extent the changes in SM and Cer species could explain their relationship with lipid markers of cardiometabolic risk, we performed additional analyses. We aimed to adjust the analysis of ΔLDL-C, Δtotal C, and ΔApoB variables with ΔC16+18 SM, Δ%C20+22 Cer, and Δ%C24+26 Cer species variables. We first transformed each covariate as a 4 classes ordinal variable and then checked for a linear relationship between each covariate and each response variable, seeking an almost constant effect from an ordinal class to its neighbor. Since we merely found a monotonic relationship, it allowed us to include these covariates in their original continuous form, associated with a unique and relevant regression coefficient, thereby simplifying interpretation. We then reported the magnitude of the effect of ΔC16+18 SM, Δ%C20+22, and Δ%C24+26 Cer species on Δtotal C, ΔLDL-C, and ΔApoB by estimating the coefficient of regression associated to each variable in the mixed linear general model.
Regarding VALOBAB-D, data are presented as mean ± SEM and were analyzed with GraphPad Prism 8.3. For normally distributed data (Shapiro-Wilk's test), repeated measures 1-way ANOVA was performed followed by Tukey's post hoc test. For nonnormally distributed data, a Friedman's test was performed followed by Dunn's post hoc test. The variation between groups was reported using P meal values, and post hoc analyses were added directly on corresponding figures using the letters a and b.
For all analyses performed in both clinical trials, results were considered significant for P < 0.05. All graphs and heatmaps were created using GraphPad Prism 8.3. Study approval. Both clinical trials were approved by the Scientific Ethics Committee of Lyon Sud-Est-IV and ANSM (French Agency for the Safety of Health Products) and registered at Clinical Trials (NCT02099032, NCT02146339). The clinical trials were conducted at the Human Nutrition Research Centre Rhône-Alpes (CRNH-RA; Lyon, France) and at the Human Nutrition Research Centre Auvergne (CRNH-A; Clermont-Ferrand, France) according to the Second Declaration of Helsinki and the French Huriet-Serusclat law. All data reported in the current article were obtained from samples stored in the biobank during the clinical studies, for which participants gave a written consent prior to inclusion in the study to use the samples for further metabolic analyses. All authors had access to the study data and reviewed and approved the final manuscript.

Author contributions
MLB contributed conceptualization, validation, formal analysis, investigation, data curation, writing of the original draft, and visualization; CV contributed conceptualization, methodology, validation, formal analysis, investigation, data curation, writing of the original draft, and visualization. E. Combe and LJC contributed methodology, formal analysis, investigation, and data curation. M. Lecomte contributed conceptualization, validation, formal analysis, investigation, and data curation. FJ contributed methodology, validation, formal analysis, and investigation. MT, SP, EL, AEB, KB, JD, AD, and CC contributed investigation. EM contributed methodology, validation, and investigation. CR contributed formal analysis and data curation. PG, NL, and E. Cotte provided essential resources. ABD contributed validation, formal analysis, and investigation. M. Laville contributed conceptualization. SLP contributed conceptualization, methodology, validation, and investigation. LO contributed methodology, formal analysis, and data curation. HV contributed to results interpretation and revised the manuscript. CMB contributed conceptualization and methodology. DC contributed conceptualization, methodology, validation, formal analysis, investigation, writing of the original draft, and data visualization. MCM contributed conceptualization, methodology, writing of the original draft, data visualization, project administration, and supervision and takes primary responsibility for final article content. All authors read, revised, and approved the final manuscript.