A hydrolyzed lipid blend diet promotes myelination in neonatal piglets in a region and concentration‐dependent manner

The impact of early life nutrition on myelin development is of interest given that cognitive and behavioral function depends on proper myelination. Evidence shows that myelination can be altered by dietary lipid, but most of these studies have been performed in the context of disease or impairment. Here, we assessed the effects of lipid blends containing various levels of a hydrolyzed fat (HF) system on myelination in healthy piglets. Piglets were sow‐reared, fed a control diet, or a diet containing 12%, 25%, or 53% HF consisting of cholesterol, fatty acids, monoglycerides, and phospholipid from lecithin. At postnatal day 28/29, magnetic resonance imaging (MRI) was performed to assess changes to brain development, followed by brain collection for microscopic analyses of myelin in targeted regions using CLARITY tissue clearing, immunohistochemistry, and electron microscopy techniques. Sow‐reared piglets exhibited the highest overall brain white matter volume by MRI. However, a 25% HF diet resulted in the greatest total myelin density in the prefrontal cortex based on 3D modeling analysis of myelinated filaments. Nodal gap length and g‐ratio were inversely correlated with percentage of HF in the corpus callosum, as well as in the PFC and internal capsule for g‐ratio, indicating that a 53% HF diet resulted in the thickest myelin per axon and a 0% HF control diet the thinnest in specific brain regions. These findings indicate that HF promoted myelination in the neonatal piglet in a region‐ and concentration‐dependent manner.

But studies have shown that long-term cognitive deficits that result from myelin abnormalities can be reversed by enhancing myelin renewal (Chen et al., 2021;Leyrolle et al., 2022;Wang et al., 2020).
In addition, myelin deficiencies can be rescued by lipid-enriched diets that have either higher overall fat content (Dimas et al., 2019;Stumpf et al., 2019) or supplementation of specific lipid species such as cholesterol (Berghoff et al., 2017;Saher et al., 2012) or phospholipids (Fledrich et al., 2018).These data suggest that dietary lipid may act as a source for myelin membrane precursors.
Lipid-enriched diets have not only been shown to rescue myelin deficits during disease or in a state of myelin impairment but have also been shown to promote myelin development in healthy subjects (Oshida et al., 2003;Sultan et al., 2021).In humans, myelination begins around the third trimester of gestation, surges until around age 5, and continues well into adulthood (Barkovich, 2005;Lebel & Beaulieu, 2011;Miller et al., 2012).In contrast to rodents in which myelination is a predominately postnatal event, myelin development of pigs begins in utero and mirrors that of humans (Conrad et al., 2012;Conrad & Johnson, 2015;Pond et al., 2000;Sweasey et al., 1976), and is a useful model to study myelination of cortical regions (Sobierajski et al., 2023).In addition, the proportion of white matter to gray matter volume in pigs is comparable to that of humans (Conrad et al., 2012;Ryan et al., 2018).As such, the neonatal piglet serves as an important model for studying myelination during development.It was previously reported that a 53% hydrolyzed fat (HF) diet increased white matter volume in small for gestational age and average for gestational age piglets as determined by magnetic resonance imaging (MRI) analysis (Caputo et al., 2020).Similarly, neonatal piglets fed a diet supplemented with phospholipids and gangliosides exhibited increased white matter volume by MRI (Liu et al., 2014).To date, there are neither data that describe dietary lipid-induced alterations to myelin development of healthy neonatal piglets by microscopic analysis nor data that provide a sow-fed reference group to assess dietary changes to myelin.
Breast milk contains bile salt-stimulated lipase, which hydrolyzes triglyceride to produce fatty acids and monoacylglycerols (Bernbäck et al., 1990;Freed et al., 1987), providing key lipid digestion and absorption support for the developing infant GI tract (Abrahamse et al., 2012).Replacing a portion of dietary triglyceride with hydrolyzed triglyceride components may mimic the end result of bile salt-stimulated lipase content in breast milk.Here, we investigated the effects of increasing concentrations of dietary HF consisting of cholesterol, fatty acids, monoglycerides, and phospholipid from lecithin, as a percentage of total lipid, on myelin development in the neonatal piglet by MRI as well as by microscopic analyses.Our findings demonstrate that dietary HF promoted myelination during development in a region-and concentration-dependent manner and provide a basis for further investigation into how dietary lipid acts to promote myelination during early life.

| Animals and care practices
Animal care protocols were in accordance with the National Institutes of Health Guidelines for Care and Use of Laboratory Ani- Here, we used magnetic resonance imaging in conjunction with microscopy techniques to assess myelination in the neonatal piglet, a relevant model for human brain development, and found that increasing concentrations of hydrolyzed fat resulted in thicker myelin per axon in specific brain regions.Our findings indicate that altering dietary lipid composition promoted postnatal myelination of piglets in a region-and concentration-dependent manner.(Rytych et al., 2012) and randomly assigned to one of four HF diet treatment groups: 53% HF (n = 9; 5 male, 4 female), 25% HF (n = 7; 3 male, 4 female), 12% HF (n = 10; 5 male, 5 female), or Control (n = 7; 3 male, 4 female).A reference group of sow-fed piglets (n = 8; 4 male, 4 female) was included.These piglets remained with the sow until day 28 of age with free access to suckle and were weighed daily at the Imported Swine Research Laboratory at the University of Illinois at Urbana-Champaign.
To accommodate eight pigs per group, the experiment was divided into two identical cohorts.Piglets in Cohort 1 had lower starting weights and had a higher incidence of sickness compared with piglets in Cohort 2: 43% (7/16) of piglets in Cohort 1 met the criteria for euthanasia (see Section 2.2) and were not included in subsequent analysis while 12% (2/17) of piglets in Cohort 2 succumbed to illness and were not included in subsequent analysis.When possible, euthanized animals were sent to Veterinary Diagnostic Lab at Veterinary Medicine at Illinois for necropsy (Table S2).Steps were taken to encourage piglet health between Cohort 1 and Cohort 2. For instance, based on bacterial resistance analysis performed by the veterinary staff, the antibiotic was changed from SpectraGuard™ to Baytril® 100.Final sample sizes used for analysis were as follows: 53% HF (n = 5; 2 male, 3 female), 25% HF (n = 6; 3 male, 3 female), 12% HF (n = 7; 4 male, 3 female), and Control (n = 6; 2 male, 4 female), sowreared group (n = 8; 4 male, 4 female).Reduced sample size after euthanasia precluded our ability to test for sex effects.space allowance for individual piglets was 30″ deep, 23″ wide, and 18.5″ high, providing 4.8 square feet of floor space per piglet.Each piglet was supplied with a heating pad, toy, and blanket.Room temperatures were kept between 29.5 and 35°C (85-95°F) using space heaters.Animals were weighed daily and limit-fed based on body weight five times a day: 9:00 a.m., 13:00 p.m., 16:00 p.m., 19:00 p.m., and 22:00 p.m.
All experimenters were blinded to diet composition.In Cohort 1, all animals were fed 300 mL/kg of body weight.In Cohort 2, to compensate for losses remaining in the graduated cylinder due to differences in diet consistency, 53% HF diet was fed at 342 mL/kg, 25% HF diet was fed at 315 mL/kg, 12% HF diet was fed at 318 mL/kg, and Control diet was fed at 306 mL/kg of body weight.This compensation did not alter weight change between Cohort 1 and Cohort 2. To discourage microbial growth, any leftover diet at the end of a feeding shift was measured using a serological pipette and discarded.The total volume of leftover diet over the course of the day was added to the final feeding shift at 22:00 p.m. Animals were also offered water or BlueLite® electrolyte solution ad libitum.Cages and heating pads were disinfected and toys and blankets replaced daily.On postnatal days 28-29, pigs were anesthetized with telazol:ketamine:xylazine cocktail (100/50/50 mg/kg; .022mL/kg body weight i.m.) and once at an appropriate plane of anesthesia, administered an intracardiac injection of sodium pentobarbital (Fatal Plus; 72 mg/kg body weight) for subsequent tissue collection.

| Medical case and euthanasia criteria
A medical case was filed with a staff veterinarian if piglets exhibited body temperature extremes (>40.2°C or <37.2°C), severe diarrhea, anorexia or lethargy, or weight loss greater than 10%.Criteria for euthanasia were as follows: weight loss of more than 15% of peak weight, weight loss for 2 consecutive days, animal becomes too weak to stand, animal does not attempt to eat for 2 consecutive days, severe diarrhea, anorexia, and lethargy, and/or body temperature extremes (>41.1°C or <36.6°C).

| Diet formulations
Diets were formulated and supplied by Abbott Nutrition (Table S1).
Diets were formulated to meet the nutritional needs of neonatal piglets and were supplied in a premixed, ready-to-feed format.The HF blend consisted of cholesterol, soy free fatty acids, monoglycerides, and soy phospholipid from lecithin and is expressed as percent (w/w).
The composition of HF test diets was as follows: 53% HF diet contained 360 mg cholesterol, 12.78 g soy fatty acids, 14.69 g monoglycerides, and 6.39 g phospholipid from lecithin; 25% HF diet contained 180 mg cholesterol, 6.41 g soy free fatty acids, 7.05 g monoglycerides, and 3.20 g phospholipid from lecithin; 12% HF diet contained 90 mg cholesterol, 3.21 g soy free fatty acids, 3.85 g monoglycerides, and 1.60 g phospholipid from lecithin.To acclimate the piglets to the fat content, each test diet was diluted with a very low fat diet for the first 5 days of feeding as previously described (Caputo et al., 2020).

| Physiological and behavior scores
For piglets on HF diets, feeding behavior, stool consistency, and activity levels were recorded daily.Feeding behavior was evaluated by assigning a feeding score upon delivery of milk by the experimenter that indicated either no attempt at eating (score of 1), attempt at eating (score of 2), or finishing while the experimenter was still present in the room (score of 3).When possible, stool consistency was evaluated on a scale of 1-4: 1 indicating solid, 2 indicating semisolid, 3 indicating loose, and 4 indicating watery.
Activity scores were assigned according to response to stimuli, such as feeding, handling during weighing, or measuring temperature, on a scale of 1-3 as follows: 1 indicating active, 2 indicating weak, 3 indicating lethargic.

| MRI acquisition
MRI data were acquired using 3T Prisma scanner (Siemens, Erlangen) housed at the Biomedical Imaging Center at the University of Illinois.Pigs were anesthetized using TKX-a combination of 2.5 mL of xylazine (100 mg/mL) and 2.5 mL of ketamine (100 mg/mL) added to a Telazol vial and administered at a dosage of .02-.03 mL/ kg IM.Pigs were maintained on isofluorane (1%-3%) throughout the imaging and scanned in the supine position using a specialized piglet head coil (Rapid Biomed, Rimpar).During scanning, the respiration rate, heart rate, and blood oxygen levels were monitored using a LifeWindow LW9x monitor (Digicare, Boynton Beach, FL).

| Volumetric analysis
We first performed manual brain extraction, or skull stripping, as described previously (Stanke et al., 2023).Tissue segmentation was performed with Statistical Parameter Mapping software (SPM12) (Ashburner & Friston, 2005), using a publicly available 28-day piglet atlas (Conrad et al., 2014).Because the default parameters of SPM were developed for human brain, the voxel size of the piglet images and all atlas images were first rescaled by a factor of 2.5 to approximately match the size of the human brain (Casteels et al., 2006).Voxel-based morphometry (VBM) was performed by creating an average-shaped template of gray matter and white matter distributions using the DARTEL package within the SPM12 (Ashburner, 2007).This template was affinely registered with tissue probability maps provided with the 28-day piglet atlas; then, "modulated" spatially normalized images of individual gray and white matter volumes were created.In the resulting images, voxel intensity provides an estimate of local volume, and the brain shape is uniform across all piglets.These images were smoothed using a Gaussian blurring kernel of full width at half maximum (FWHM) of 6 mm in the rescaled units, which corresponds to 2.4 mm in the physical dimensions.

| Diffusion analysis
Manual brain extractions were first performed using an approach similar to that used for volumetric analysis.We rigidly reoriented the diffusion scans to approximately align them with the 28-day structural piglet atlas (Conrad et al., 2014).This was done by modifying the affine array of each scan such that the b = 0 diffusion scan was approximately aligned to the average brain template.The affine array adjustments were performed using manual adjustments in SPM12, followed by the "Coregister: Estimate" function of SPM12.
A brain mask for each piglet was then manually drawn by modifying a standard mask, as with the brain extractions of the structural images (Stanke et al., 2023).We used the "dtifit" function of the FMRIB Software Library (FSL; Oxford, UK) to calculate maps of fractional anisotropy (FA).This calculation was performed by excluding b = 0 data to minimize the partial volume bias from the cerebral spinal fluid (CSF) in the b = 0 scan.The NODDI model was fit to the diffusion data using the NODDI Toolbox (http://nitrc.org/projects/ noddi_toolbox) (Zhang et al., 2012).In this model, the spatial distribution of intracellular water diffusion is described by a Watson distribution of zero radius cylinders.The diffusivity along the cylinders is assigned to be 1.7 × 10 −3 mm 2 s −1 , and the isotropic diffusivity is assigned to be 3.0 × 10 −3 mm 2 s −1 (Zhang et al., 2012).The toolbox estimates values for the intracellular volume fraction, or neurite density index (NDI), and the orientation dispersion index (ODI) that ranges from 0 (unidirectional diffusion) to 1 (isotropic).Spatial analysis of diffusion parameter maps was performed using the tract-based spatial statistics (TBSS) image processing pipeline in FSL (Smith et al., 2006), which consisted of four steps: (1) removal of regions of spurious high apparent FA that commonly surround the brain, (2) a nonlinear registration to a publicly available piglet FA atlas (Zhong et al., 2016), (3) creation of a mean FA image and a skeletonization of the image, (4) projection of all subjects' FA maps to the mean skeleton.This step required a threshold FA value, which was taken to be .2,the default value.The nonlinear registration, skeletonization, and projection vectors calculated for FA maps were then applied to maps of the ODI and NDI predicted by the NODDI model, using the FSL function "tbss_non_FA."

| CLARITY tissue processing and immunostaining
Pigs were euthanized at postnatal days 28-29.Brains were harvested, regions dissected and cleared by clear lipid-exchanged acrylamide-hybridized rigid imaging/immunostaining/in situ hybridization compatible tissue hydrogel (CLARITY).Specifically, tissues from prefrontal cortex (PFC), caudal corpus callosum (CC), and internal capsule (IC) were post-fixed in phosphate-buffered saline (PBS, pH 7.4) containing 4% paraformaldehyde (PFA; w/v) overnight at 4°C and then incubated in PBS overnight at 4°C.Next, 1-mm sections were generated and each section was submerged in hydrogel solution modified from a previous publication (Epp et al., 2015) containing final concentrations of 3% acrylamide (Bio-Rad Cat.No. 161-0140), 3% formaldehyde (Electron Microscopy Sciences Cat. No. 19200), and .25%VA-044 thermal initiator (m/v; Wako Chemicals Cat.No. NC0632395) for 24 h at 4°C.Incubated tissues were polymerized in hydrogel solution at −90 kPa for 3 h at 37°C, washed three times with PBS, then actively cleared by electrophoresis for 24 h using X-CLARITY™ tissue clearing system (Logos Biosystems).

| Immunostaining and nodal gap length analysis
Upon sacrifice, tissues from PFC, rostral CC, and IC were dissected and fixed in PBS containing 4% PFA (w/v) for 24 h at 4°C and then cryoprotected in PBS containing 30% sucrose for 48 h at 4°C.Next, the tissues were snap frozen in dry ice, embedded in Tissue-Tek® O.C.T. Compound (Sakura #4583) in a dry ice-ethanol slurry, and sectioned using a cryostat microtome (Leica CM 1950).Twenty micrometer sections were collected onto positively coated microscope slides .Tissues were rehydrated with PBS for 20 min and then blocked and permeabilized with .3%PBST containing 5% goat serum for 1 h at room temperature.Tissue sections were then incubated in .3%PBST containing 5% goat serum and the following primaries at 4°C overnight on sequential days: anti-PLP (1:200; Abcam #ab105784) and anti-Caspr (1:1000; Abcam #ab34151).After three 5-min washes with .3%PBST, tissues were incubated with secondary antibodies goat anti-rat Alexa Fluor 488 (Thermo #A11006) and goat anti-rabbit Alexa Fluor 594 (Thermo #A11012) at 1:1000 dilution in the same solution as the primary step for 1 h at room temperature.After washing as described above, tissues were counterstained with Hoechst 33342 (1:5000; Thermo #H3570), washed again, and then mounted with Fluoromount-G® (Southern Biotech #0100-01) and cover glass (Corning #2980-245).
Images were obtained on a confocal microscope (Olympus BX51) with a 60x objective.
Nodal gap length was measured using the "Straight" line tool on ImageJ software (NIH).To avoid skewing measurements based on filament orientation, node length was measured only between paranode pairs that were oriented in straight alignment with each other and parallel to the sectioning plane, which was determined by the presence of PLP-stained internode segments flanking the paranodes.Paranodes that were asymmetrical or unpaired due to nonparallel orientation to the sectioning plane were disregarded.For 1-2 animals per diet, 10-28 measurements were averaged per region.For all other animals, 30-113 nodal gap length measurements were averaged per region.The researcher performing analysis was blinded to experimental condition.

| Serial block face scanning electron microscopy (SBF-SEM) and g-ratio analysis
Immediately upon dissection, tissues from PFC, rostral CC, and IC were drop-fixed in .15M sodium cacodylate buffer containing 2.5% glutaraldehyde and 2% formaldehyde with 2 mM calcium chloride (pH 7.4).Tissue was then post fixed in 2% osmium tetroxide in .15M cacodylate buffer and 1% thiocarbohydrazide solution.Tissues were stained with 1% uranyl acetate, dehydrated in an ascending alcohol series, and embedded in Durcupan ACM epoxy resin (Cat.No. 14040).Serial images of the resin-embedded samples were acquired by alternating between imaging the block face and shaving off a layer with a diamond knife using a Gatan 3View SBF microtome housed in a Gemini SEM column (Sigma VP, Carl Zeiss) equipped with a variable pressure secondary electron (VPSE G3) detector.Sets of 300-500 images at 50 nm steps, 8-18 k × 8-18 k pixels, and 5 nm/pixel resolution were obtained, resulting in whole datasets of 40-90 μm × 40-90 μm in the X, Y-plane and 15-25 μm in the Z direction.
ImageJ software was used to manually measure g-ratio (ratio of the axon diameter to the outer diameter of the myelin sheath) in all tissues examined.The total stack of serial EM images for each animal was divided evenly into three block portions (beginning, middle, and end).Ten continuous representative images were selected from each portion (30 total images per pig) and analyzed using a digital stylus pen and the "freehand selections" tool in ImageJ with "perimeter" as the parameter.Axons in the first image of each portion were labeled alphabetically and utilized as a reference to track the same axon through each block.Axons from different blocks were considered discrete.Only circular myelinated axons free of artifact were included in the analysis.The g-ratio was calculated for each axon as follows: axon diameter (perimeter/3.14159)was divided by outer myelin sheath diameter (perimeter/3.14159).In this analysis, thinner myelin sheaths correspond to higher g-ratios.Diet effect on both g-ratio per axon and g-ratio per animal was determined.Each axon measurement represents the average of 5-10 g-ratio measurements taken along the axon.For each animal, 5-15 axons were averaged per region (n = 5 animals per diet group).The researcher performing analysis was blinded to experimental condition.
EM datasets were also utilized to quantify the number of myelinated axons in the PFC, CC, and IC.The total stack of serial EM images for each animal was divided evenly into three block portions (beginning, middle, and end).One image from each block (three per region) was used to manually count all myelinated axons within a frame (40-90 μm × 40-90 μm in the X, Y-plane) and averaged per animal.Data are expressed as the number of myelinated axons per mm 2 .

| Statistics
For volumetric MRI analysis, statistical tests on the smoothed modulated white and gray matter images were performed using SPM12 software.The data were modeled only according to diet groups.An omnibus F-test was performed to determine voxels for which diet group was predictive of volume differences.The significance threshold was taken to be p < .05,where p has been corrected for multiple comparisons using familywise error.Analysis was limited to voxels within masks of gray matter and white matter, taken to be voxels for which the tissue probability maps from the atlas exceeded .5.For each piglet, we averaged the voxel intensities over all voxels that surpassed the significance threshold.

| HF diets caused differences in body weight gain
At postnatal day 2, facility-housed piglets were randomized into four diet groups-a 0% HF Control diet, or a diet containing 12%, 25%, or 53% HF-based on birth weight (Figure S1a).A reference group of sow-reared piglets was included and remained with the sow for the entirety of the study.By one-way ANOVA with Bonferroni correction (F (4, 27) = 3.552, p = .0188),birth weights did not differ between any facility-housed groups (0%, 12%, 25%, and 53% HF groups), but piglets assigned to be sow-reared had lower birth weights than those assigned to the 12% HF group (p = .0216;Figure S1a).No piglets met the criteria for being smallfor-gestational age.Repeated-measures mixed effects analysis with fixed effects of time and diet, matched values by time, and Bonferroni correction revealed an effect of diet (F (4, 27) = 16.76,p < .0001)and time (F (1.283, 34.39) = 938.9,p < .0001) on weight gain analyzed as percent change from baseline, as well as a diet × time interaction (F (104, 697) = 14.42, p < .0001; Figure 1a).Sowreared piglets gained more weight than those on all other diets.This effect became significant at postnatal day 8 (Figure 1a).Piglets fed a Control diet and a 25% HF diet gained less weight than sow-reared piglets until postnatal day 22 and 25, respectively, and weighed significantly more than those fed a 53% HF diet only at postnatal day 27 (Figure 1a).Compared to piglets fed a 53% HF diet, piglets fed a 12% HF diet had significantly greater weights on postnatal days 8, 15, 20-27 (Figure 1a).The overall effects of diet were consistent even when raw weight values were analyzed (Figure 1b).Despite lower body weights in the 53% HF group compared to all other diets on postnatal day 27, neither gross brain weights nor brain weight to body weight ratios were significantly different by diet (Figure S1b,c).
Feeding behavior, stool consistency scores, and activity levels were recorded daily in piglets fed HF-containing diets.The majority of piglets did not attempt to eat on experimental days 0-2 (Fig- ure 1c).Following acclimation to the diet regime, most piglets were observed eating immediately after being fed, although some still displayed no feeding attempt through experimental day 11 (Figure 1c).
After this time, the majority of piglets finished their meal while the experimenter was present, especially those fed a 25% and Control diet (Figure 1c).However, there were no differences in feeding behavior across diets by one-way ANOVA (Figure 1f).The time period from experimental days 0-11 also corresponded to the most prevalent incidence of diarrhea, regardless of diet (Figure 1d).Similarly, piglet activity level was lowest during the initial weaning period with some weak or lethargic behavior observed in all diet groups between experimental days 0 and 11 (Figure 1e).The majority of piglets exhibited normal activity levels after experimental day 11 (Figure 1e), which corresponded to the time frame where stool scores and feeding scores improved (Figure 1c,d).Like feeding behavior, stool scores and activity scores did not differ between diet groups (Figure 1f).
Taken together, these data indicate that dietary HF content did not alter feeding behavior, stool consistency, or activity levels, but had some effect on body weight gain.

| Effect of HF diets on survival
A total of nine piglets met euthanasia criteria during the study and were omitted from analysis (Table S2).Sow-reared piglets suffered no losses (Figure S1d).Piglets fed a 25% HF and Control diet both had survival rates of 85.7% (6/7) (Figure S1d).Piglets fed a 12% HF diet had a survival rate of 70.0% (7/10) (Figure S1d).The group fed a 53% HF diet had the greatest number of euthanized animals with a survival rate of 55.6% (5/9) (Figure S1d).The most common cause of illness among necropsied animals was likely porcine circovirus 3 (PCV3; Table S2), a relatively novel porcine virus that has been shown to infect the gastrointestinal tract and cause diarrhea as well as potential bladder and/or kidney issues (Klaumann et al., 2018;Saporiti et al., 2021).Thus, we tested serum from all piglets collected at postnatal days 14 and/or 28 for PCV2 and PCV3.At least one animal from each diet had a positive or suspect test result for viral presence, but of those with a positive or suspect test result only the 53% HF group had no survivors (Table S2).There was no significant effect of diet on survival based on logrank (Mantel-Cox) test (χ 2 (4) = (5.430),p = .2460).

| VBM results
The F-test of gray matter volumes revealed 1632 voxels with significantly different volumes among all diets.Of these, only 11 voxels were isolated from neighboring voxels, and the rest were grouped into 22 clusters of face-connected voxels ranging from 2 to 832 voxels in size.All significant voxels were within the left and right cortex, and the largest clusters were symmetrically distributed in lateral and superior cortical regions (Figure 2a).Analysis of average values from the significant voxels revealed differences between diet groups (F (4, 27) = 26.46,p < .001).Specifically, Bonferroni post hoc analysis revealed that sow-reared piglets exhibited lower gray matter volume than all other groups (p = .003for Sow vs. 53%, p < .001for all other pairwise comparisons), and piglets fed 53% and 12% HF diets exhibited significantly lower gray matter volumes than piglets fed Control diet (p < .001and p = .04,respectively;

Figures 2b and S2).
The F-test of white matter volumes revealed 200 voxels with significantly different volumes among all diets.Of these, only three voxels were isolated from neighboring voxels, and the rest were grouped into six clusters of face-connected voxels ranging from 2 to 130 voxels in size (Figure 2c).The voxels were symmetrically distributed in the olfactory bulb (189 voxels), the superior cortex (9 voxels), and the inferior cortex (2 voxels).Analysis of average values from the significant voxels revealed differences between diet groups (F (4, 27) = 18.45, p < .001).Specifically, Bonferroni post hoc analysis revealed that all piglets fed HF diets exhibited significantly lower white matter volumes than the sow-fed group (p = .003for Sow vs. 53%, p < .001for all other pairwise comparisons; Figure 2d).

| TBSS results
The F-test of FA values revealed one cluster of eight voxels with significantly different values among all diets.This cluster was in the right frontal region of the brain (Figure 2e).Analysis of average values from the significant voxels revealed differences between diet groups (F (4, 27) = 11.12,p < .001).Specifically, Bonferroni post hoc analysis revealed that piglets fed 53% and 12% HF diets exhibited significantly lower FA values than the sow-fed piglets (p < .001and p = .001,respectively), and the piglets fed a 25% HF diet (p = .002and p = .03,respectively; Figure 2f).Piglets fed a 53% HF diet also exhibited lower FA values than those fed a Control diet (p = .004).Measures analyzed by one-way ANOVA with Bonferroni correction are as follows: total filament surface area, total filament length, total number of branches, total number of branch points, total number of terminal points, total number of segments, total number of filaments, total number of points, mean point distance from origin, mean filament volume, mean filament diameter.Measures analyzed by Kruskal-Wallis test with Dunn's correction are as follows: mean filament surface area, total filament volume, mean filament length, mean filament straightness, mean number of branch points, mean number of branches, mean number of segments, mean number of terminal points, mean number of scholl intersections, and total number of scholl intersections.Data from individual piglets are shown (n = 5-8) and presented as box and whisker plots with individual data points.p-value * < .05,**p < .01.Full statistics described in the Results section.
Collectively these data indicate that piglets fed a 25% HF diet had increased myelination per mm 2 within the PFC.
Imaris analysis of PLP-stained CC and IC tissues was also performed (Figure S3a,d).Since these regions are major white matter tracts, individual myelinated filaments were not distinguishable by this method.Thus, the surfaces modeling module in Imaris was used.
There was no effect of diet on total surface area and total volume measurements within these regions (Figure S3b,c,e,f).

| HF diet altered nodal gap length in a concentration-dependent manner
The node of Ranvier is an axonal segment enriched with sodium channels that allow for membrane depolarization, whereas the myelinated internodal segment allows for saltatory conduction.Nodal gap length can be measured as the space between the flanking paranodal loops demarcated with an antibody against Contactinassociated protein 1 (Caspr), a transmembrane protein essential to adhering the paranodal loops to the axonal membrane (Schneider et al., 2016) (Figure 4a).Here, PFC, CC, and IC tissues were immunostained with an antibody against Caspr as well as with anti-PLP to label mature myelin (Figure 4b).The average number of nodes analyzed per animal across diets ranged from 30 to 56 in the PFC, 35 to 46 in the CC, and 34 to 57 in the IC (Figure 4c).Kruskal-Wallis test revealed differences in node length in the PFC (H (4) = 116.9,p < .0001), the CC (H (4) = 55.58,p < .0001),and the IC (H (4) = 11.07,p = .0258).In the PFC, node length was smallest in sow-reared piglets compared to all other groups (p < .001for all pairwise comparisons by Dunn's post hoc; Figure 4d).In the CC, node length was smallest in piglets fed a 53% HF diet (p < .0001for all pairwise comparisons by Dunn's post hoc; Figure 4e).In the IC, node length was smaller in piglets fed a 12% HF diet compared to those fed a Control diet (p = .0231;Figure 4f).When averaged per animal, node length was not different between diet groups (Figure 4g-i).But all node length measurements taken together were inversely correlated with percentage of HF in the CC (ρ = −.1374,95% C.I. = −.2007 to −.07291, n = 961, p < .0001)with a 53% HF diet resulting in the smallest average node length (Figure 4j).

| HF diet altered g-ratio in a region and concentration-dependent manner
Measuring g-ratio remains the gold standard for assessing myelin ultrastructure (Chomiak & Hu, 2009).Therefore, to further examine the effects of diet on myelin thickness, we performed serial block face scanning electron microscopy of PFC, CC, and IC tissues (Figure 5b) and measured g-ratio, the ratio of axon diameter to outer myelin sheath diameter (Figure 5a).The average number of g-ratios analyzed per animal across diets ranged from 64 to 81 in the PFC, 95 to 128 in the CC, and 106 to 117 in the IC.G-ratio was plotted against axon diameter and analysis of linear regression models fit to the data

| Effect of sex
Sex had no effect on weight gain (Figure S4a,b), physiological scores (Figure S4c), or survival (Figure S4d).Likewise, there were no differences in MRI measurements including gray matter volume, white matter volume, FA, and ODI between male and female piglets (Figure S4e).However, in the PFC, males exhibited higher overall myelin density compared to females as indicated by increased number of scholl intersections, total filament surface area, total filament length, total filament volume, number of branches and branch points, number of terminal points, number of segments, number of filaments and total number of points (Figure S4f).Males also exhibited longer node length in the PFC compared to females (Figure S4g), but exhibited no changes to g-ratio in any region (Figure S4h).When averaged per animal, node length and g-ratio were not different between males and females (Figure S4i,j).

| DISCUSS ION
We have examined the ability for dietary HF to cause macroscopic and microscopic changes to myelination during development in neonatal piglets.In the frontal region of the brain, piglets fed Control and a 25% HF diet had FA values that matched that of sow-fed piglets.In addition, a 25% HF diet resulted in myelin density in the PFC that exceeded that of sow-fed piglets.Finally, we found that nodal gap length and g-ratio were inversely correlated with percentage of HF in the CC, as well as in the PFC and IC for g-ratio, indicating that a 53% HF diet resulted in the thickest myelin per axon and Control diet the thinnest in specific regions.Our study is the first to demonstrate that a HF system fed to neonatal piglets during the early postnatal period promoted myelin development in a region-and concentration-dependent manner.
One strength of the current study is the inclusion of a sow-fed reference group.HF is designed to exogenously deliver the products of fat digestion that are provided in breast milk due to natural lipase activity.However, HF-induced changes to myelin were not equivalent to myelination patterns that were observed in sow-fed piglets.Our results indicate that sow-reared piglets exhibited the greatest weight gain, highest overall brain white matter volume, and highest FA value in the frontal region of the brain by MRI.Human neuroimaging studies have shown that breast-fed infants exhibit greater overall myelination and better cognitive performance later in life compared to formula-fed infants (Deoni et al., 2013(Deoni et al., , 2018)), which could be attributed to enhanced lipid composition or nutrient bioavailability in breast milk (Chiurazzi et al., 2021;Garwolińska et al., 2018;Haschke et al., 2016;Koletzko, 2016).In particular, compounds vital for myelin development such as sphingomyelin, gangliosides, and folic acid differ between breast milk and infant formula (Schneider et al., 2022).The interpretation of the data presented here may also be limited by differences in non-dietary factors, most notably rearing (sow-reared vs. individually housed in a facility) and maternal separation.Maternal separation is a known early-life stressor that disrupts OL differentiation and myelination processes of the PFC in rodents (Teissier et al., 2020;Yang et al., 2017;Zeng et al., 2020).However, we demonstrate that HF milk replacer formula promoted myelination in a concentration-dependent manner in facility-housed piglets that were subjected to maternal separation starting at postnatal day 2. Interestingly, while piglets fed a 53% HF diet exhibited the thickest myelin per axon in the PFC and CC, they also had the least weight gain by postnatal day 28.Differences in weight gain and survival may be partially attributed to sickness.The most common cause of sickness appeared to be porcine circovirus 3, a viral pathogen that infects the gastrointestinal tract.Unlike the other diet groups, all piglets fed a 53% HF diet that were positive for virus met euthanasia criteria.
Since dietary cholesterol has been shown to increase morbidity in a murine model of influenza A virus infection (Louie et al., 2022), it is possible that higher levels of cholesterol in the 53% HF diet had negative consequences on clinical outcome of infected piglets.
While lipid species including cholesterol (Jurevics & Morell, 1995;Saher et al., 2005), phosphatidylcholine (Fledrich et al., 2018;Skripuletz et al., 2015), and fatty acids (Leyrolle et al., 2022;Trapp & Bernsohn, 1978) are required for myelination, cholesterol is the most abundant lipid constituent of myelin (Norton & Poduslo, 1973), is rate limiting for myelination (Saher et al., 2005), and its supplementation alone promotes myelin development (Haque & Mozaffar, 1992) and repair (Berghoff et al., 2017).Thus, the cholesterol component of HF may have been the main underlying stimulant of concentrationdependent, increased myelin thickness.But differences in weight gain in the current study highlight the need to consider diet-induced white matter alterations in a larger context that includes diet effects that may be unrelated to myelination.
Consistent with previous findings reported by Caputo et al., VBM analysis showed decreased cortical gray matter volume in piglets fed a 53% HF diet compared to piglets fed a Control diet.In contrast to Caputo et al., the current VBM analysis showed no global change in white matter volume between these two groups.While MRI can determine the regions of interest based on gross changes to white matter, analysis of microscopic changes in myelination can be limited by resolution.To gain more insight into how myelin thickness and the degree of myelination were affected by diet, we performed immunohistochemistry and electron microscopy analyses in conjunction with MRI.In order to maximize the volume of tissue analyzed while retaining measurements of the individual fiber, we utilized a lipidcleared protein scaffold stained against PLP as a proxy for myelinated fibers in the PFC.Studies have reported correlations between MRI and histology or other staining methods such as Luxol Fast Blue that also rely on protein scaffold labeling of myelin and is typically performed on 5-30 μm thick tissue sections (Fang et al., 2005;Fanous et al., 2020).Major advantages of the CLARITY/immunofluorescence technique over conventional staining procedures are (1) the ability to analyze larger tissue volume due to lipid clearing prior to staining and (2) the non-subjective approach used to generate the data.
We measured both nodal gap length and g-ratio to assess the effect of diet on myelin thickness.Since myelin thickness affects conduction velocity and likely network activation, subtle changes may have implications for learning and memory (Fields, 2008).In addition to myelin sheath thickness, nodal gap length has been identified as a potential regulator of conduction velocity along an axon (Arancibia-Cárcamo et al., 2017).Longer node lengths often coincide with thinner myelin as determined by larger g-ratios (Cullen et al., 2021;Dutta et al., 2018).Node elongation has also been reported to coincide with reduced oligodendrogenesis (Schneider et al., 2016).Our data support this relationship between g-ratio and node length, as both measures were inversely correlated with HF percentage whereby longer node lengths and larger g-ratios were associated with lower HF percentage.It should be noted that in order to acquire an accurate measure of g-ratio, we examined myelinated axons free of artifact.While anomalous alterations of myelin have been observed in rodent models of dietary protein deficiency and disrupted amino acid signaling in developing OL, these changes are also accompanied by decreased percentage of myelinated axons and corresponding increase in unmyelinated axons (Almeida et al., 2005;Yu et al., 2022).
In the current study, the number of healthy myelinated axons did not differ between diet groups in any of the regions analyzed; thus, any myelin irregularities observed by electron microscopy were likely due to artifact rather than myelin dysregulation.
Altering myelin thickness is thought to be a mechanism of finetuning neuronal circuits (Almeida & Lyons, 2017;Fields, 2015).As such, diet-induced increase in myelin thickness should be considered in the context of functional consequences on the neuronal circuit as a whole.In addition to structure, the biochemical composition of myelin can affect function, and thus, it would be important to consider the effect of dietary lipid on the biochemical composition of myelin in future investigations.
Notably, a concentration-dependent effect of g-ratio was most evident in the PFC and CC, and seemingly extinguished in the IC.This may be attributable to the fact that the developing brain is generally myelinated in a caudal-to-cranial, posterior-to-anterior, central-to-peripheral manner (Ballesteros et al., 1993;Barkovich et al., 1988;Brody et al., 1987).As such, the sequence of myelination of the regions examined here would be IC, rostral CC, then PFC.It is possible that at the time of dietary lipid treatment, myelination of the IC was more developed than the more rostral regions and less vulnerable to environmental influence.This suggests that alterations to dietary lipid composition must occur during a critical developmental window to exert maximal effects on myelination.
After excluding data from euthanized animals, percent weight change did not differ between the 53% HF group and all other diets until the last experimental day.Nevertheless, the suboptimal weight gain of the 53% HF group is a pitfall of the current study.Although fat and energy content were designed to be well balanced among diet groups, we cannot exclude the possibility that higher HF levels, particularly cholesterol, may have resulted in effects unrelated to myelination, such as predisposition to exacerbated sickness.Another pitfall is the reduced group sample sizes due to some animals reaching euthanasia criteria before the end of the study.On the one hand, including several experimental groups with a 0% HF Control group instead of one control versus one experimental group allowed us to determine a dose effect on node length and g-ratio measures.On the other hand, smaller samples sizes per group may have limited the robustness of certain measurements, such as correlation between node length and HF percentage in the PFC, which did not reach significance.
In addition, delineating any sex × diet interactions would require bolstering sample sizes in future studies.That said, after collapsing diet groups in order to assess effect of sex, Imaris modeling of myelinated filaments in the PFC indicated greater myelin density in the PFC of males compared to females, which is in agreement with what is reported in literature regarding sex differences in cortical myelin development of humans as assessed by MRI (Corrigan et al., 2021).Finally, while HF promoted thicker myelin per axon in specific brain regions by node length and g-ratio analysis, the impact of HF on myelination varied between macroscopic and microscopic evaluations.For example, greater myelin thickness per axon does not equate to greater overall myelin density of a given region.This underscores the complex nature of dietary influences on myelination as well as the need to utilize a multimodal approach when estimating changes to myelin content.
Taken together, our data demonstrate that altering postnatal dietary lipid composition is capable of modulating myelination of neonatal piglets in a region-and concentration-dependent manner and provide a basis for further study of functional consequences of dietary lipid-induced alterations to myelin during early life.

D ECL A R ATI O N O F TR A N S PA R EN C Y
The authors, reviewers and editors affirm that in accordance to the policies set by the Journal of Neuroscience Research, this manuscript presents an accurate and transparent account of the study being reported and that all critical details describing the methods and results are present.
mals and were approved by the University of Illinois Laboratory Animal Care and Use Committee.Male (n = 20) and female (n = 21) appropriate for gestational age, Yorkshire crossbred, full-term, naturally delivered piglets were obtained from the University of Illinois Swine Farm at postnatal day 2 to allow for colostrum consumption.All piglets remained intact, but did undergo processing on the farm including iron dextran (Henry Schein Animal Health, Dublin, OH, USA) and antibiotic injection (EXCEDE®, Zoetis, Parsippany, NJ 07054, USA) per routine farm practice and according to label.Piglets were placed individually into a caging system under standard conditions as described in a previous publication Significance Cognitive and behavioral functions rely on proper myelination during neurodevelopment.Dietary lipid has been shown to alter myelination, but its effect on myelin structure and volume in healthy neonates is not well studied.
Sets of 100 serial images at 1 μm steps in the Z direction, .48μm/ pixel in the X, Y-plane, and .38 μsec pixel dwell time were acquired with the Zeiss LSM 710 Confocal Microscope and 10× objective, resulting in whole datasets of 1.06 mm in the X, Y-plane and 100 μm in the Z direction.Images were rendered into 3D with Imaris 9.3 software (Bitplane, Oxford Instruments) and analyzed with the software's automated filament tracer module for PFC and the surfaces module for CC and IC.All image stacks were analyzed in the same fashion by a rater blinded to experimental condition.Definitions of Imaris terms are provided in Table For diffusion MRI analysis, statistical tests of the skeletonized maps of FA, ODI, and NDI were implemented by voxelwise application of the general linear model (GLM) tool of FSL.Diffusion parameters were modeled as functions of diet only.Four contrasts consisting of pairwise comparisons, diet1 > diet0, diet2 > diet0, diet3 > diet0, and diet4 > diet0, were grouped into an F-contrast to provide an omnibus test for the significance of diet.Statistical significance, thresholded at p < .05,was estimated using the "Randomize" function of FSL, using 500 permutations of group assignments to select the F-statistic threshold values.For each piglet, we averaged the values of FA, ODI, and NDI, obtained from skeletonized maps, over all voxels that surpassed the F-statistic significance threshold.GraphPad Prism software (ver.9.1.0)was used to execute all other statistical analyses and to create graphs.For weight change, repeated-measures mixed effects analysis with fixed effects of time and diet, matched values by time, and Bonferroni correction for multiple comparisons were performed.Random effects of subject and residual were included in the model.For all other datasets, after testing for normality (Anderson-Darling, D'Agostino & Pearson, Shapiro-Wilk, Kolmogorov-Smirnov), one-way ANOVA with Bonferroni correction was performed if data followed Gaussian distribution and if standard deviations (SDs) were not significantly different (assessed by both Brown-Forsythe and Bartlett's test).Kruskal-Wallis nonparametric test with Dunn's correction for multiple comparisons was performed if data did not pass one of the normality tests or if SDs were significantly different.Correlations between HF concentration and g-ratios or node lengths were assessed by Spearman's correlation test because g-ratio and node length data were not normally distributed.Difference in survival curves was assessed by logrank (Mantel-Cox) test.A p-value of .05 was considered statistically significant.

F
Hydrolyzed fat (HF) diet affects body weight gain but does not alter physiological scores.(a) Percent weight change and (b) raw weight values in kg of piglets from birth until euthanization (n = 5-8 animals per group; repeated-measures mixed effect analysis with Bonferroni correction; data presented as mean ± SD). ¥ indicates sow-reared group is significantly different from all other diets on postnatal days 8-21, and from 12% HF and 53% HF groups after postnatal day 22 onward.§ indicates 53% HF diet group is significantly different from 12% HF diet group on postnatal days 8, 15, 20-25, and from all other diets on postnatal day 27.‡ indicates sow-reared group is significantly different from 53% HF diet group on postnatal days 8-27, from 12% HF diet group on postnatal days 13-14, from 25% HF diet group on postnatal days 11, 13-17, and from Control diet group on postnatal days 11, 13-14.† indicates 53% HF diet group is significantly different from Control diet group on postnatal day 27.(c) Heat map of feeding behavior scores.Red indicates no attempt to eat, light green indicates attempt to eat, dark green indicates finished meal while experimenter was present.(d) Heat map of stool scores assigned when possible on a scale of 1-4: 1 solid, 2 semi-solid, 3 loose, and 4 watery.(e) Heat map of activity scores assigned on a scale of 1-3: 1 active, 2 weak, and 3 lethargic.For c-e, white indicates no value recorded.(f) Sum of physiological score across experiment for each assessment.Data from individual piglets shown (n = 5-8; one-way ANOVA with Bonferroni correction performed for feeding and stool scores, Kruskal-Wallis performed for activity score; data presented as box and whisker plots with individual data points).The F-test of ODI values revealed 72 voxels with significantly different values among all diets.They included all eight voxels that exhibited significantly different FA values.Of these, only eight voxels were isolated from neighboring voxels, and the rest were grouped into eight clusters of face-connected voxels ranging from 2 to 16 voxels in size.The significant voxels were in the right and left F I G U R E 2 Effect of hydrolyzed fat diet on white matter volume and diffusion measures by magnetic resonance imaging.Regions of significant dietary differences are shown in red for (a) gray matter and (c) white matter, superimposed over an average brain image.Dietinduced differences in volume within regions of (b) gray matter and (d) white matter.Voxels of significant dietary differences are shown in red for (e) fractional anisotropy (FA) and (g) orientation dispersion index (ODI), superimposed over an average brain image and the skeletonized white matter tracks, shown in green.Diet-induced differences in (f) FA and (h) ODI within significant regions.Data were analyzed by one-way ANOVA with Bonferroni correction and presented as box and whisker plots with individual data points.Data from individual piglets are shown (n = 5-8).p-value * < .05,**p < .01,***p < .001.Full statistics described in the Results section.F I G U R E 3 Hydrolyzed fat diet altered myelination in the prefrontal cortex (PFC) as assessed by CLARITY and immunofluorescence.(a) Experimental schematic for CLARITY and immunofluorescence technique on PFC tissue.(b) Representative 3D renderings of PLP-stained PFC tissue.Each dataset measures 1.06 mm in the X, Y-plane and 100 μm in the Z direction.Scale bar is 100 μm.(c) Measurements generated by Imaris software Filament Tracer on 3D renderings of PLP-stained PFC tissue.

F
I G U R E 5 Hydrolyzed fat (HF) diet altered g-ratio in a region-and concentration-dependent manner.(a) Representative cross section of an axon from an electron micrograph with axon perimeter and outer myelin sheath perimeter outlined in yellow.Scale bar is .5 μm.(b) Representative electron micrographs of prefrontal cortex (PFC), rostral corpus callosum (CC), and IC tissues of each diet group.Scale bar is 5 μm.(c) Number of g-ratio measurements analyzed per animal for PFC, CC, and IC.Data analyzed by one-way ANOVA for each region.(d-f) g-ratio plotted against axon diameter in the PFC, rostral CC, and IC.Data from individual axons are shown (n = 33-38 for PFC, n = 61-73 for CC, n = 63-68 for IC) where each data point represents the average of 5-10 measurements taken along the same axon; p-value shown on graph indicates whether linear regression slopes differed significantly from each other by one-way ANOVA.(g-i) g-ratio measurements in the PFC, rostral CC, and IC.Data from individual axons are shown (sample size same as in d-f).(j-l) g-ratio averaged per animal in the PFC, rostral CC, and IC.Data from individual piglets are shown (n = 4-5 animals per group).(m-o) Number of myelinated axons per mm 2 in the PFC, rostral CC, and IC.Data from individual piglets are shown (n = 4-5 animals per group).Each data point represents the average count of three electron micrograph images selected from the beginning, middle, and end of the serial block dataset for each animal.(p) g-ratio measurements plotted against percentage of HF.The rs (or ρ) correlation coefficient and p-value shown on graph were derived from Spearman's correlation test.p-value * < .05,** < .01,*** < .001.Unless otherwise indicated, data are presented as box and whisker plots with individual data points.Full statistics described in the Results section.