Eighty male ternary hybrid piglets (Landrace × Yorkshire × Duroc, 2 days postnatal, 1.3–2.0 kg each) used in this study were provided by Chengdu Shepherd Boy Village Agricultural Development Co., Ltd., Sichuan Province, China. Two piglets were excluded because of poor growth state. Due to the facility space limitations, the animal experiments were conducted in five separate sessions, each consisting of 16 piglets obtained from three to six different sows. All animals were housed and kept in separate cages at the KangCheng facility [Sichuan KangCheng Biotech Co., Ltd.; animal production license number: SYXK (Chuan) 2019–215] in pig cages for 30 days under controlled temperature: 22 ± 2°C, humidity: 60% ~ 70% under a 12 h-alternate light/dark cycle (lights on at 8:00 am and off at 8:00 pm). Fresh milk powder was provided to the piglets every 2 h a day. At the facility, the standard room temperature was set at 24.5°C. Initially, the temperature in the cage was higher than the standard temperature (33–35°C). Then, the piglets were housed in cages with decreased temperatures as they grew. If a piglet developed diarrhea, it was given a saline rehydration solution. At the facility, 16 two-day-old newborn male piglets per entry were randomly divided into four groups according to body weight when piglets entered the facility, four piglets for each group in every session, and the piglets were reared until 31 days postpartum (Figure 1). All experiments followed the guidelines for the Care and Use of Laboratory Animals provided by the Ministry of Science and Technology of the People’s Republic of China (PRC). All animal care and experimental procedures were performed according to the Animal Research: Reporting in Vivo Experiments (ARRIVE) guidelines (17) and were approved by the Institutional Animal Care and Use Committee (IACUC) of West China Hospital Sichuan University. The environmental breeding conditions followed the PRC national standard GB14925-2010.

In this study, all artificially reared piglets were fed milk powder provided by Inner Mongolia Dairy Technology Research Institute Co., Ltd. Piglets in the control group (n = 19) were fed the control diet, the other groups were provided with diet supplemented with milk fat globule membrane (MFGM; Lacprodan^® MFGM-10, Arla Foods Ingredients) of 1.74 g/100 g MFGM (MFGM-L, n = 19), 4.64 g/100 g MFGM (MFGM-M, n = 20), and 6.09 g/100 g MFGM (MFGM-H, n = 20). The specific diet in each group are listed in Table 1. The difference in the four groups mainly comes from the concentration of total phospholipids, phosphatidyl serine, phosphatidyl ethanolamine, phosphatidyl inositol, and phosphatidylcholine, as indicated by the bold part in Table 1. According to the nutritional needs of piglets (nutritional requirement criterion, NRC 2022), the amount of piglets’ daily dietary intake is 285 mL/kg*fasting body weight + 50 mL. The milk was freshly prepared with warm water (37–40°C) every 4–5 h, in a 1:4 (powder-to-water) ratio, and the prepared milk was placed in a feeder.

Spatial learning and memory were assessed by the spatial T-maze test, the task first used to compare the performance of different pigs by Mendl et al. in 1997 (18). The behavioral device comprises a 45 cm wide cross maze with a removable acrylic clapboard. Piglets can identify the orientation from all four corners using visual cues. Two bowls were placed at T-maze’s left (west) and right (east) arms, one with milk as a reward and another empty (Supplementary Figure 1A). Topscan software (Cleversys Co. Ltd. United States version 8, 2020) assessed the piglet’s motion trails. The nine-day T-maze behavioral test included a six-day acquisition and three-day reversal phases (19). On the morning of day 15, the piglets were fasted (a nightly 6-h food deprivation period), removed from their home cages, and carried to the maze in the adjacent behavioral room until the behavior assessments were completed. In the acquisition phase, they were randomly placed at the northern or southern entrance of the maze (Supplementary Figure 1). They aimed to find the reward milk bowl via visual cues in 60 s. Each piglet was assessed 10 times daily for 6 days between 0900 and 1700 by the same two trained experimenters, and the correct rate was recorded. Within the 3-day reversal phase, the reward milk bowl and visual cues were swapped to assess the piglets’ memory and to ensure that the piglets’ orientation discernment was not self-centered. The number of incomplete/completed tasks within 60 s and the correct latency were also recorded. All behavioral technicians were blinded to the grouping information of the intervention dietary. The test indices collected include the correct rate (piglets choose reward milk bowl to represent correct), qualification (correct rate reaching 80%), non-qualified rate (the whole rate of animal in correct rate not reaching 80% across the entire test), and the earliest qualified day (the first-day piglets achieving qualification). The trace of pigs in the T-maze was shown in Supplementary Video.

All the piglets were anesthetized by intramuscular injection of Zoletil 50 (Vibrac, French, 0.2 mg/kg). The brains were dissected and weighed. The hippocampus and prefrontal cortex (PFC) were excised from the fresh brain tissue and frozen in liquid nitrogen. Total RNA was isolated from 40 to 100 mg of hippocampal and prefrontal cortex tissues. Tissues were dissolved with 1 mL Trizol (Beyotime Biotechnology) and purified with organic solvent (Chloroform, isopropyl alcohol, and alcohol). The isolated total RNA was then diluted to 500 ng/μL. RNA was converted into cDNA with a reverse transcription kit (Vazyme Biotech Co., Ltd., R323-01). BDNF content was measured with a qPCR kit (Vazyme Biotech Co., Ltd., Q711-02). The purity and concentration of the RNA were evaluated by spectrophotometry (Thermo-Nano Drop 2000c-spectrophotometer) by measuring the mean absorbance ratio and optical density of all RNA at 260/280 nm. BDNF and β-actin primer sequences were as follows: BDNF primer sequence (F: 5’-AGCATTAGCGAGTGGGTGAC-3′; R: 5’-GGGACTTTTTCGAGGACCGT-3′; β-actin primer sequence: F: 5’-GACTGCGCCCATAAAACCC-3′, R: 5’-CACGAGCGCCAGCAATATCGT-3′). The qPCR parameters were as follows: pre-denaturation at 95°C, 30 min, 40 cycles reaction procedure: denaturation 95°C, 10 s; Annealing 55°C, 30 s; The fusion curve program was 95°C 10 s, 55°C 30 s, and 95°C 10 s. (The relative quantification of the transcription levels of the copy melting curve genes was performed by the ^-ΔΔCT method).

We weighed 40 mg of porcine hippocampal tissue and lysed in 1 mL buffer Radio Immunoprecipitation Assay (RIPA; Beyotime, P0013B) with 10 μL phenylmethylsulfonyl fluoride (PMSF; Servicebio, G2008); added 400 μL RIPA for lysis tissue (the RIPA concentration was 10 μL/mg). The tissue lysates were homogenized and lysed on ice for 20 min and centrifuged at 15000 g for 20 min at 4°C. The supernatant was collected as a total protein extract. BCA Protein Detection Kit (Beyotime, P0010S) measured the protein concentration. The total protein was added to 1 × SDS-loading buffer and boiled for 5 min at 100°C. Thirty microgram protein was loaded on 12% SDS polyacrylamide gel (GenScript, M01210C). The separated protein was transferred to the polyvinylidene fluoride (PVDF) membrane (Beyotime, FFP32) and washed with Tris Buffered saline Tween (TBST). At room temperature, the membrane was blocked in a 5% nonfat milk powder blocking buffer for 2 h. The membrane was then incubated with primary antibody [Anti-BDNF antibody (Bioss, bs-4989R, 13 kDa, China); neurotrophin-3 (NTF3) antibody (Affinity, DF7235, 31 kDa); Rabbit anti-beta-actin antibody (Bioss, bs-0061R, 42 kDa)] at 4°C overnight followed by the incubation with secondary antibody HRP-conjugated affinipure goat anti-rabbit IgG (Proteintech, SA00001-2) 1:5000 for 2 h at room temperature. For imaging, they were treated with a chemiluminescence detection kit (Millipore, P90719) and exposed to a chemistry luminescence imaging system (Fluor chem. Co. model protein simple FC3, United States). Each band was quantified using ImageJ software (version 8).

In this study, all data were normally distributed and presented as Mean ± Standard Error of the Mean (SEM). GraphPad Prism 8 software (version 3.1) was used to plot the data. The data were analyzed by SPSS 22.0 statistical software (IBM Corporation, version 26). The parameters were evaluated for variance homogeneity by the Kolmogorov–Smirnov test. Repeated measures analysis of variance (ANOVA) was performed to determine the main effects and interaction effects of body weight, food intake, correct rate, correct latency, and proportion correct at different time points. Differences among the single index, such as white matter volume and FA, were analyzed by one-way ANOVA, followed by Bonferroni post-hoc tests. The correlations between correct rate and FA in different regions were analyzed statistically by Pearson’s correlation. A p < 0.05 was considered statistically significant.

In this study, the piglets were fed with a diet containing different percentages of MFGM. In 30 days, the body weight of piglets was not affected (Supplementary Figure 2A); all body weight gain remained at normal levels of growth [F_{(3, 74)} = 0.295, p = 0.829, Supplementary Figure 2B]. Milk intake showed no difference between groups and stayed at average increases as the piglets’ development changed [Supplementary Figure 2C; F_{(3, 74)} = 0.348, p = 0.790, Supplementary Figure 2D]. As part of the total body weight, the brain weight also showed no difference between the groups. These results indicated that dietary MFGM was well-tolerated for piglets and supported growth. Besides, it was reported that formula feeding increased the risk of diarrhea (20). Unfortunately, two piglets had to be eliminated from the study due to the severity of their diarrhea. For the remaining piglets, the incidence of diarrhea was not alarmingly high. On average, about two or three piglets would experience bouts of diarrhea during each round of experimentation. Notably, this was particularly prevalent within the first 10 days after birth. It is worth mentioning that there were no discernible differences in the incidence of diarrhea among the various groups studied. Overall, dietary MFGM was well-tolerated and supported piglet growth and milk intake in 31 days.

The spatial T-maze test was performed on day 15 of feeding, and the trial was divided into two phases: the acquisition and reversal phases. The difference between the acquisition and reversal phases was the direction of visual cues (Supplementary Figure 1). The acquisition phase lasted 6 days, while the reversal phase lasted 3 days. The most important indicator of the T-maze is the correct rate in the 10 trials every day. During the acquisition phase, the correct rate of the piglets improved steadily over time. They completed the learning process by day 3, defined as achieving an accuracy rate of 80%. By day 5, most of the tested piglets met the specified criterion. On day 7, when the reversal phase started, the accuracy rate dropped sharply due to the changed position of the visual cues. Although the accuracy rate improved on the following days (8 and 9), it did not return to the comparable performance level achieved during the acquisition phase (Figure 2A). Repeated measures revealed a day effect on correct rate (F_{(8, 68)} = 10.822, p < 0.001) and no interaction effects between days and groups [F_{(24, 210)} = 0.889, p = 0.617]. Difference existed among four group [F_{(3, 75)} = 3.049, p = 0.034]. The post-hoc Bonferroni analysis showed that the MFGM-L group had the highest correct rate, much higher than the control (p = 0.011) and MFGM-H groups (p = 0.044; Figure 2A). There was a day effect of latency to choice [F_{(8, 68)} = 10.822, p < 0.001] and no interaction effects between days and groups. Significant differences in latency to make a choice existed in each group [F_{(3, 75)} = 3.049, p = 0.034]. However, the post-hoc Bonferroni analysis showed only a trend of difference between Control and MFGM-L (p = 0.083), as well as MFGM-L and MFGM-H (p = 0.054; Figure 2B). The correct proportion demonstrated that in neither the acquisition nor the reversal phase, the MFGM-L had the highest proportion correct (Figure 2C).

The qualification days of the MFGM-L group also showed the best results in the T-maze test’s acquisition and reversal phases (Figure 2D). In the MFGM-L group, all piglets were qualified (achieving a correct rate of 80%) in both the acquisition and reversal phases. However, 20% of piglets were non-qualified in 9 days. The other two groups had the same non-qualified rates, higher than MFGM-L but lower than the control (Figure 2E). The other index of T-maze was the earliest qualified day; piglets in the MFGM-L group learned faster than the other groups, but no difference in learning rate was identified among the four groups [F_{(3, 75)} = 1.199, p = 0.316; Figure 2F].

On day 30, all the piglets were recruited for an MRI scan to measure the gray and white matter volume. The results showed that the representative white and gray matter images were based on VBM, the standard accepted test method (Figures 3A,B). There were no significant differences in white and gray matter volume calculated based on VBM [F_{(3, 57)} = 0.442, p = 0.724; F_{(3, 57)} = 0.760, p = 0.521, respectively; Figures 3C,D]. The other index of brain development was the trace of fiber bundles, with the representative images of fractional anisotropy (FA) determined by DTI in the whole brain, corpus callosum, and hippocampus shown in Figure 3E. No differences in FA were found among each group both in the whole brain [F_{(3, 74)} = 2.237, p = 0.091; Figure 3F] and corpus callosum [F_{(3, 74)} = 1.665, p = 0.182; Figure 3G]; However, there was a statistically significant difference in FA in the left hippocampus among the four groups [F_{(3, 74)} = 5.882, p = 0.001]. A Bonferroni post-hoc analysis revealed significant differences between Control and MFGM-L (p < 0.001), MFGM-M (p = 0.040), and MFGM-H (p = 0.005; Figure 3H). The same trend could be seen in the right hippocampus [F_{(3, 74)} = 6.269, p < 0.001]. A Bonferroni post-hoc analysis showed that MFGM-L fiber bundles were much better than control (p < 0.001), MFGM-M (p = 0.006), and MFGM-H (p = 0.003; Figure 3I). The MRI results were consistent with T-maze tests, indicating that MFGM-L contributed to memory-improving and white matter fiber growth.

We performed a correlation test to find the relation between FA and the correct rate in spatial T-maze tests. Specifically, the FA values of the whole brain, corpus callosum, left and right hippocamps white matter fiber bundles were correlated with the area under the curve (AUC) of the correct rate. The correlation between FA in the whole brain and correct rate (AUC) was no difference (r = 0.344; p = 0.665; Figure 4A). Similarly, the correlation between FA in the corpus callosum and correct rate (AUC) was also no statistical difference but more trend than the whole brain (r = 0.750; p = 0.250; Figure 4B). However, compared to the above results, the correlation between FA in the left hippocampus and correct rate (AUC) was another story. More correlation and statistical differences were observed (r = 0.998; p = 0.003; Figure 4C). Analogously, the right hippocampus also correlated with FA, and the correct rate was observed (r = 0.986; p = 0.013; Figure 4D). This figure shows that the FA value in the hippocampus correlates with the correct rate in the T-maze task.

We further studied the mechanism of MFGM function by examining the expression of BDNF, which plays a role in the growth, maturation (differentiation), and maintenance of the neurons (21). Neurotrophin-3 (NT-3), including BDNF, nerve growth factor, and neurotrophin-4/5, play an important role in neurogenesis (22). The acquisition and consolidation of spatial memory depend on the correct functioning of both the hippocampus (23) and prefrontal cortex (24), where the highest levels of BDNF are found in the central nervous system (25), so we analyzed changes in BDNF in the hippocampus and PFC. There was a significant difference in relative BDNF mRNA expression in the hippocampus among each group [F_{(3, 58)} = 7.332, p < 0.001]. A Bonferroni post-hoc analysis showed that the BDNF mRNA expression was higher in MFGM-H than in control (p < 0.001) and MFGM-L (p < 0.001; Figure 5A), indicating that a high dose of MFGM promoted the expression of hippocampal BDNF and the development of the neurons. However, the expression of BDNF in PFC was no statistical difference but only a trend [F_{(3, 70)} = 1.590, p = 0.200; Figure 5B]. The protein expression of BDNF and NTF -3 was determined by Western blotting, mixing each sample of one group as a new sample. It showed that BDNF expression was the highest in MFGM-L and MFGM-M and lowest in control, while NTF-3 expression was the highest in MFGM-M and most lacking in control (Figures 5C,D). These results demonstrated that after the addition of MFGM, regardless of the dosage, the expression of BDNF and NTF-3 in brain tissues would increase.