Pigs were randomly assigned to one of three experimental conditions: (1) a nutritionally adequate milk replacer (CON; Purina® Multi-Species Milk Replacer, Purina Animal Nutrition LLC, North Arden Hills, MN, United States), (2) CON diet plus oral administration of DSS (DSS), and (3) CON diet with 8.3 g/kg of gamma-cyclodextrin encapsulated tributyrin (TBCD) plus oral administration of DSS (TBCD + DSS). Pigs received ad libitum access to liquid milk replacer treatments starting on PND 3 and continuing until study conclusion.

TBCD was manufactured (Shi et al., 2020) and administered (Bartholome et al., 2004) based on previously reported procedures and recommendations. DSS and TBCD+DSS treatment groups were orally administered DSS at 1.25 g/kg body weight daily from PND 14 to 18. Thirty minutes prior to scheduled milk replacer feeding each morning, pigs in the DSS and TBCD+DSS treatments received their respective dose in their milk bowl for voluntary ingestion. Health checks, feed disappearance, and body weights were obtained daily, and associated descriptions/outcomes have been reported separately (Sommer et al., 2022).

MPRAGE sequences were obtained from the tip of the snout to the cervical/thoracic spinal cord junction. Sagittal-oriented data was captured with a 173 mm × 173 mm × 153.6 mm field of view (FOV) and a 288 × 288 × 256 matrix. The following specific parameters were utilized: repetition time (TR) = 2,060 ms, echo time (TE) = 2.05 ms, inversion time (TI) = 1,060 ms, and flip angle (α) = 9° providing a final voxel volume of 0.6 × 0.6 × 0.6 mm³. Image processing followed previously described methods (Golden et al., 2023).

To estimate white matter integrity and microstructure, a diffusion-weighted echo planar imaging (DWI-EPI) sequence was utilized (Feinberg et al., 2010; Moeller et al., 2010; Xu et al., 2013). Images were acquired in transverse orientation with a 160 mm × 160 mm × 80 mm FOV and 100 × 100 × 50 matrix size. The following parameters were utilized: TR/TE = 5,100 ms/70.00 ms, α = 90°, GRAPPA accelerated by a factor of 2 in the phase encode direction, multiband factor of 1, and 3 diffusion weightings at 0, 1,000 and 2,000 s/mm² across 30 directions.

The preprocessing of DTI data involved several steps. Raw DWI images were first denoised using the Marchenko-Pastur Principal Component Analysis (MPPCA) method (Manjón et al., 2013; Neto Henriques, 2018; Veraart et al., 2016a; Veraart et al., 2016b) within the DIPY toolbox (Garyfallidis et al., 2014). Following denoising, between-volume motion correction was performed using an affine registration method (Jenkinson and Smith, 2001), also implemented through DIPY. This step corrects for subject movement between different DWI volumes, subsequently reducing motion artifacts. Additionally, manual whole brain tissue segmentation was performed by trained technicians for the delineation of brain structures.

Diffusion tensors were reconstructed using the diffusion model (Basser et al., 1994; Pajevic and Pierpaoli, 1999) as provided by DIPY. This reconstruction produces maps of fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD), which were used to assess microstructural properties of white matter. Affine registration was conducted using ANTs (Avants et al., 2014). Each subject’s FA map was registered to their corresponding MPRAGE image, allowing regions of interest (ROI) defined in MPRAGE space to be warped to diffusion space. This registration facilitated the extraction of mean FA, MD, RD, and AD values from the ROIs for subsequent statistical analysis.

An established form of myelin water imaging, called multicomponent-driven equilibrium single pulse observation of T₁ and T₂ (mcDESPOT) (Deoni et al., 2008; Deoni, 2011; Deoni and Kolind, 2015) was conducted for myelin content estimation in young pigs. Utilizing multiple sets of spoiled gradient-recalled echo (SPGR) and T₁/T₂-weighted balanced steady-state free precession (SSFP) acquired at varying flip angles (α) and with a constant TR, estimates were derived for various parameters. These parameters included T₁, T₂, water residence times, and water volume fractions for the extracellular and intracellular pools of water that comprise the lipid bilayers of myelin.

Sagittal-oriented images were acquired with a 160 mm × 160 mm × 124.8 mm FOV and 128 × 128 × 96 matrix. SSFP data were acquired with two phase-cycling increments of 0° and 180° utilizing the following parameters: TR/TE = 5.3 ms/2.7 ms, α = (11, 15, 19, 23, 27, 35, 50, and 70)°, and bandwidth = 350 Hz/Px, providing a final voxel volume of 1.3 × 1.3 × 1.3 mm³. On the other hand, SPGR data were acquired with the following parameters: TR/TE = 5.6 ms/2.7 ms, α = (3, 4, 5, 6, 7, 9, 13, and 18)°, and bandwidth = 350 Hz/Px, providing a final voxel volume of 1.3 × 1.3 × 1.3 mm³. To correct for transmit (B₁) magnetic field inhomogeneities, an additional two high resolution T₁-weighted inversion recovery (IR)-SPGR sequences were acquired with inversion times of 450 and 750 ms. IR-SPGR parameters were as follows: TR/TE = 5.6 ms/ 2.7 ms, α = 5°, and bandwidth = 350 Hz/Px, providing a final voxel volume of 1.7 × 1.7 × 2.6 mm³. Image processing followed previously described methods for pigs at 4-weeks of age (Golden et al., 2023).

To compare modulated gray and white matter concentrations across the three treatment groups, the statistical non-parametric methods (SnPM) toolbox was utilized.¹ Data were subjected to a one-way ANOVA specifying an absolute threshold of 0.2 and ANCOVA global normalization. Using the Dartel-generated final template as an inclusive mask, an uncorrected p < 0.01 and a threshold of 20 edge-connected voxels were applied.

Differences in absolute volume were observed for the whole brain and gray matter (p < 0.05) (Figure 2). Pigs in the TBCD+DSS treatment group had smaller absolute whole brain (p = 0.006) and gray matter (p = 0.013) volumes compared with both the CON and DSS treatments. Differences in absolute volume were also observed for 28 of the 29 isolated ROI (p < 0.05). A similar effect was observed where pigs in the TBCD+DSS group had lower (p < 0.05) absolute ROI volumes compared with both other treatments. Absolute brain volumes are displayed in Supplementary Table S1.

A difference in relative volume was only observed for gray matter (p = 0.041) (Figure 2). Pigs in the TBCD+DSS group were observed to have a lower (p < 0.05) relative volume of gray matter compared with the CON group. No other relative volume differences were observed. All relative volume results are displayed in Table 2. Moreover, voxel-based morphometry analyses revealed no significant cluster differences across treatment groups for either gray or white matter.

Across the four diffusion measures, several regional differences were observed between treatment groups (Table 3; Figure 3). For fractional anisotropy, these regions included: hypothalamus (p = 0.025), left and right hippocampus (p = 0.034 and 0.008, respectively), right internal capsule (p = 0.029), and thalamus (p = 0.040). For the hypothalamus and thalamus, the CON group had higher (p < 0.05) FA values compared with the TBCD+DSS group, with the DSS group as intermediary. For the left hippocampus, the CON group had higher (p < 0.05) FA compared with both DSS and TBCD+DSS treatment groups. In the right hippocampus, the same effect was observed although DSS was intermediary. In the right internal capsule, the CON and DSS groups had larger (p < 0.05) FA values compared with the TBCD+DSS group.

Conversely, in some regions where FA was observed to be higher for a treatment, across the other diffusion metrics (MD, RD, and AD) values were lower for that same treatment (displayed in Table 3; Figure 3). For example, in the left hippocampus, across all three DTI measures, the CON and DSS groups had lower (p < 0.05) values compared with the TBCD+DSS group (p < 0.05). Similarly, the same effect was observed in the thalamus for MD (p = 0.016) and RD (p = 0.005). In the right hippocampus, the CON group had lower values compared with the TBCD+DSS group, with the DSS group as intermediary for MD (p = 0.005), RD (p = 0.003), and AD (p = 0.049). Several additional regions had a treatment effect for RD. In the medulla, left superior colliculus, right cortex, and right internal capsule, the CON group was observed to have lower (p < 0.05) RD compared with the TBCD+DSS group, with DSS as intermediary. In the pons, both CON and DSS groups had lower (p < 0.05) RD compared with the TBCD+DSS group. Contrary to other measures, AD in the corpus callosum had a different trend, where the DSS group had higher (p = 0.03) AD compared with the TBCD+DSS group.

Differences in mean myelin water fraction were observed for the whole brain and both left and right cortices (Figure 4). For overall mean myelin water fraction in the whole brain (p = 0.003) and left cortex (p = 0.001), pigs supplemented with DSS (DSS and TBCD+DSS treatments) had less myelin compared with the CON pigs. For the right cortex, all three treatments differed (p = 0.001) from each other. The TBCD+DSS had the lowest (p < 0.05) mean MWF, whereas CON pigs had the highest (p < 0.05) value overall. MWF values for all other regions are listed in Supplementary Table S2.

For general assessment of region-specific volumes, anatomical images were acquired to calculate absolute and relative volume of whole brain, white matter, gray matter, cerebral spinal fluid and 29 regions of interest. Across multiple brain regions, lower absolute volumes were observed for the TBCD+DSS group. This finding can be explained by growth performance results reported by Sommer et al. (2022), where it was observed that TBCD+DSS pigs had overall decreased body weight from PND 15 to study conclusion compared with both other treatment groups. With a smaller body size in general, it is expected that absolute volume of the brain and corresponding regions would be smaller.

Once volume was considered relative to whole brain size, most treatment effects dissipated. This finding provides support that allometric brain growth was largely unaltered by treatment group, other than for relative gray matter volume. Contrary to previous work by Kornelsen et al. (2021), we observed lower relative gray matter volume in the TBCD+DSS group and no differences between the CON and DSS groups. Kornelsen et al. (2021) conducted VBM analyses and identified several gray matter clusters where patients with ulcerative colitis exhibited larger gray matter volumes compared with healthy controls. On the other hand, Zikou et al. (2014) observed decreased gray matter clusters across several brain regions in patients diagnosed with CD or UC compared with controls. These contradicting results differ from the present study where no significant VBM cluster differences were identified across treatment groups. In general, given the conflicting evidence, it is unclear what to expect with brain volume in a colitis condition, hence bolstering our justification for use of multiple neuroimaging modalities to improve interpretation of results. Therefore, investigating microstructure in conjunction with myelination specific measures was advantageous.

To investigate microstructural organization of the brain, DTI imaging was conducted. This imaging technique captures the diffusion of water molecules across imaging voxels and is used to construct four maps (FA, MD, RD, and AD) that correspond to diffusivity across fibers (Alexander et al., 2007; Chang et al., 2017). Using RD and AD maps, respective parallel and perpendicular water movement across the long axis of fibers can be captured, whereas overall average diffusivity is reflected by MD. During brain development, the tensor parameter that is of most interest is FA, which captures the orientation-specific diffusivity of water molecules within a tissue (Alexander et al., 2007; Chang et al., 2017). In general, FA can be influenced by a variety of microstructural features such as myelination, axon diameter, and axon packing density (Jones et al., 2013), and FA is repeatable and highly sensitive to neurodevelopmental status. Previously, an inverse correlation was observed for FA and RD regarding myelination (Chang et al., 2017; Scheel et al., 2013). Chang et al. (2017) observed that myelin basic protein (MBP) immunofluorescence was positively correlated with FA and negatively correlated with RD, but this dual relationship was only observed in select regions like the corpus callosum, fimbria, and anterior commissure. In other white matter regions, MBP immunofluorescence was only correlated with FA (Chang et al., 2017), highlighting the possibility of FA as a more myelin-specific measure. Even given this evidence, there is much debate on the true biological relevance of the various diffusion parameters (Jones et al., 2013). Specifically, DTI provides an estimate of fiber orientation rather than a precise measurement and parameter values can be highly influenced by crossing fibers (Jeurissen et al., 2013; Jones et al., 2013).

Several trends were observed across diffusion outcomes in the present study. In regions such as the hippocampus, FA was decreased in pigs administered DSS as compared with control pigs. However, the opposite was observed in the other diffusion outcomes, where MD, RD, and AD were increased in both DSS groups compared with control pigs. Similarly, when FA was decreased in the thalamus for the TBCD+DSS group compared with the control, the opposite directional trend was observed for MD, RD, and AD, indicating more restrictions to diffusion. These directional changes align with a previous study by Hou et al. (2020) in a group of CD patients. Patients diagnosed with CD had decreased FA paired with increased MD, RD, and AD in a subset of brain regions when comparing to results in healthy controls (Hou et al., 2020). On the other hand, work in an adolescent population with pediatric onset of CD detected higher FA in several regions and lower MD across multiple cortices in the CD group (Filip et al., 2024). However, this study was completed in CD adolescent patients in remission which may have impacted results.

In our pig study, several other regions were identified as having significant differences across the diffusion parameters, but due to inconsistent differences observed across diffusion parameters, these outcomes were more difficult to interpret. Interestingly, in the corpus callosum, a highly myelinated structure, AD was observed to be lower for the TBCD+DSS group compared with the DSS group, which was not consistent across other regions. These findings are similar to those of Zikou et al. (2014) who observed decreased AD in patients with CD and UC, although lower AD was not observed in pigs administered DSS in the current study. Not all DTI trends were consistent with previous work, which may be due to differences in age. In the present study, young pigs were chosen to represent a pediatric population whereas in human studies it is typical to recruit adult patients with chronic CD and UD (Hou et al., 2020; Zikou et al., 2014).

In general, it is possible these results reflect variance in how axons are distributed within a voxel across treatments (i.e., the architectural paradigm), which may affect overall tissue anisotropy (Jones et al., 2013). Additionally, across development, FA has been shown to steadily increase into adolescence (Barnea-Goraly et al., 2005), which follows the trend for myelination to increase throughout childhood (Croteau-Chonka et al., 2016). Still, the interpretation of DTI parameters is highly debatable, as studies have shown contradicting evidence when it comes to mapping longitudinal development (Arshad et al., 2016). Additionally, metrics like FA can be influenced by a multitude of microstructural features outside of myelination (Jones et al., 2013). Therefore, to gain better insight into how myelination is affected, more myelin specific imaging sequences were acquired.

Various myelin-specific neuroimaging methods are available to quantify myelin concentration in the brain. A highly utilized method, specifically in the field of neurodevelopment, is through steady-state-based myelin water imaging (Croteau-Chonka et al., 2016; Dean et al., 2014; Deoni and Kolind, 2015). Myelin water fraction (MWF) can be calculated utilizing the distinct signatures emitted by extracellular vs. intracellular water molecules within the myelin sheath of axons, (Dubois et al., 2014). Generally, studies have shown strong correlations between MWF and myelin histology, justifying it as a potential biomarker of myelin (Lee et al., 2021).

In the present study, across whole brain and both left and right cortices, both DSS groups were identified as having lower MWF values compared with the control group. In particular, the TBCD+DSS group was observed to have the lowest values of MWF in the right cortex. This is relatively consistent with the DTI outcomes that reflected both DSS-treated groups having lower FA values in the hippocampus. Specifically, the TBCD+DSS group was consistently observed to have the lowest FA values across multiple regions other than the right hippocampus. Previously, it has been observed that myelination is a particularly vulnerable process to DSS-induced increased intestinal permeability (Boles et al., 2024) This may explain why colitis has been repeatedly associated with demyelinating disorders such as multiple sclerosis (Scheid and Teich, 2007). Usually, demyelinating conditions manifest after long-term chronic colitis exposure, which has been commonly studied (Scheid and Teich, 2007). However, even acute exposure to DSS has been linked to deficits in recognition memory and anxiety-like behavior in mice several weeks after exposure (Salvo et al., 2020), likely indicating that early-life inflammatory bowel conditions can detrimentally impact brain development. There is limited work on how early-life colitis can impact myelin development, with most early-life colitis research completed in rodents (Boles et al., 2024; Han et al., 2018; Salvo et al., 2020) whereas others indicating no differences in myelination as assessed through neuroimaging in adolescent patients with CD (Filip et al., 2024). Therefore, the findings of this study add to the literature and support the hypothesis that early-life exposure to colitis influences myelination patterns within the developing brain.