All studies were performed at the University of Missouri School of Medicine. Pregnant C57BL/6 mice were obtained from Charles River. Litters of mice were breast-fed by the mother until days of life (DOL) 7, separated from the mother, and randomly assigned to either the control, NEC, NEC plus butyrate, or butyrate control groups. NEC was induced using a standard NEC protocol (14) (see Supplementary Material 3). The mice were gavage-fed with a 24-Fr angiocatheter for five times in a 24 h period. Three of the feeds received 0.15 ml every 3 h with a formula (15 g Similac 60/40 in 75 ml Esbilac canine milk replacer), and two of the feeds received mixed 0.15 ml every 3 h with a formula with lipopolysaccharide Escherichia coli 055:B5 (200 μl/5 g of mouse body weight, Sigma, St Louis, MO, USA). The mice were simultaneously exposed to hypoxic stress (5% oxygen and 95% nitrogen) twice daily for 10 min each of exposure. This model represents a reproducible model of mucosal injury and systemic inflammation in the postnatal period. Prior to NEC induction, the NEC plus butyrate group and butyrate control group received 10 µl of 60 mM butyrate twice a day on day of life 5 and 6; this dose was chosen on the basis of butyrate concentration in the human intestine lumen and previously reported research findings (11, 15–17). Given that butyrate has a short half-life (18), in addition to pretreatment with butyrate as noted above, formula milk was supplemented with 60 mM butyrate in the NEC butyrate group. After 3 days of NEC induction, the mice were euthanized on DOL-10 using an intraperitoneal injection of phenobarbital, and the terminal ileum and brain were harvested. The control and butyrate control pups were also sacrificed on DOL-10 for comparison with the NEC group. Brain and ileum were collected and divided into two parts: formalin for staining and RNA later for quantitative polymerase chain reaction (qRT-PCR).

Brain tissue was collected and divided into two equal parts (1) half of the brain was placed in Carnoy's (30% chloroform, 60% ethanol, 10% glacial acetic acid), embedded in paraffin blocks, and sectioned at 5 µm. Specific brain regions including the (1) cortex, (2) hippocampus, and (3) cerebellum were dissected and used for RNA and Western blot analysis.

Histologic slides were stained with hematoxylin and eosin. Slides were scanned using a Leica Biosystems Slide Scanner. A standardized 4-point scale as previously described was used to grade intestinal injury in the ileum (19), by two separate investigators, one of whom was blinded to the experimental conditions (MM and WY). Briefly, injuries are ranked on a scale of 0–4: 0, healthy mucosa; 0 = normal intestine; 1 = disarrangement of villus enterocytes, villus-core separation; 2 = significant disarrangement of villus enterocytes, villus-core separation down the sides of villi, blunting of villi; 3 = epithelial sloughing of villi, loss of villi; 4 = intestinal necrosis or perforation. A score of 2 or higher indicated the development of experimental NEC. The analysis was completed using the system's imaging software (Aperio Imagescope). One of the two personnel who assessed intestinal injury scores was blinded to treatment groups.

Pieces of the cortex, cerebellum, and hypothalamus were homogenized using a bullet blender (Midwest Scientific, St Louis, MO, USA). Total RNA was extracted using the Trizol standard protocol (Thermo Fisher, Waltham, MA, USA), and cDNA was synthesized from 1 μg of RNA using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA). The transcripts were amplified and gene expression data were collected on a ViiA7 with an SYBR Green master mix. Prevalidated KiCqStart SYBR primers were purchased from Sigma, and the following primers were used: Actin (ACTB), tumor necrosis factor alpha (TNF-α), IL-6, IL-8 (KC), interleukin-1 beta (IL-1β), and TLR4. Actin was used as the housekeeping gene. The relative gene expression of these markers was calculated by the Pfaffl method and delta-delta CT (20).

To analyze the protein levels in the cortex and cerebellum, protein lysates were obtained from mouse brain tissues after being extracted using a radioimmunoprecipitation assay (RIPA) lysis buffer containing commercially available Halt protease and phosphatase inhibitors (Thermo Fisher) that were homogenized by using the bullet blender (Midwest Scientific). The concentration of purified lysates was measured using a bicinchoninic acid Protein Assay Kit (Thermo Fisher). Equal amounts of lysates were loaded and separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis, and gels were blotted to polyvinylidene difluoride membranes. The membranes were incubated with primary antibodies, washed, and again incubated with peroxidase-conjugated secondary antibodies (Abcam, Waltham, MA, USA). The primary antibodies are as follows: Rabbit anti-Phospho-p65 (Ser536, Cell Signaling Technologies, CST, Danvers, MA, USA), rabbit anti-p65 (CST), rabbit anti-Phospho-p38 (Thr180/Tyr182, CST), rabbit anti-p38 (CST), rabbit anti-TLR4 (Abcam), rabbit anti-SIGIRR (Thermo Fisher), mouse anti-GFAP (CST), rabbit anti-IL6 (CST), rabbit anti-Cleaved Caspase 3 (CC3, CST), rabbit anti-IBA1 (Abcam), mouse anti-ICAM1 (SCBT, Santa Cruz, CA, USA), and mouse anti-β-Actin (ACTB, Sigma). Reactive proteins were detected using a chemiluminescence detection system and visualized on the imaging system iBright FL1000 (Thermo Fisher). Densitometry was performed using ImageJ software (NIH, Bethesda, MD, USA), and changes were normalized to beta-actin (ACTB) or the corresponding non-phosphorylated antibody.

A segment of the cortex, cerebellum, and hypothalamus was embedded in paraffin. The paraffin blocks were cut at 5 μm and adhered to positive-charged slides. The samples were air-dried, washed in PBS and PBS-Tween for 5 mins, blocked in 10% heat-inactivated normal goat serum (HINGS, Jackson ImmunoResearch, West Grove, PA, USA), and incubated with the primary antibodies or control overnight at 4°C. Sections were subsequently washed in PBS ×3 and incubated with a secondary antibody for 2 h at room temperature, washed, and counterstained with 4,6-diamidino-2-phenylindole (DAPI; Thermo Fisher). The primary antibodies are as follows: CNPase-2′, 3′-cyclic nucleotide 3′-phosphodiesterase (anti-mouse CNPase, Abcam) to identify myelinating oligodendrocytes; Glial Fibrillar Acidic Protein (anti-mouse GFAP, Cell Signaling) to identify glial cells; ionized calcium-binding adaptor molecule 1 (anti-Rabbit IBA1, Abcam) to identify microglia; Phospho-histone H3 S28 (anti-rabbit Phis, Abcam) to identify replicating cells in the M phase of the cell cycle. The secondary antibodies are as follows: for CNPAse, anti-mouse Alexa Fluor 647 (Abcam); for GFAP, anti-mouse Alexa Fluor 488 (Abcam); for IBA1, anti-rabbit Cyanine 3 (Abcam); for Phis, anti-rabbit Cyanine 3 (Abcam). The nucleus was stained with DAPI. Images were captured using EVOS imaging software. The images were taken at 20× and processed with Adobe Photoshop. The images were imported into the QuPath version 0.2.3 program and analyzed using the “Positive Cell Detection” software (21). Each channel (red, green, and blue) was analyzed separately using the following settings: requested pixel size (0.5 µm), Sigma (1.5 μm), threshold (25), cell expansion (5–15 µm), and mean nuclear or cytoplasmic thresholds per channel were set to (1+:10, 2+:20, 3+:30), the minimum area was set to 10 µm², and the maximum area was set to 400 µm². The parameters measured are as follows: number of cells detected, number of cells positive, centroid in the X and Y axes, number positive per mm, area in µm², and perimeter in µm.

Data was represented as mean ± SD. For all data, we initially examined whether the distribution of data was Gaussian using the D’Agostino–Pearson omnibus normality test. If data were normally distributed, then ANOVA with Bonferroni correction was used for analysis. If data did not meet Gaussian assumptions, the Kruskal–Wallis test with Dunn's correction was used for analysis. A score of p < 0.05 was considered significant for all variables. Animal considerations: each experimental group contained an n ≥ 5. RNA quantification and PCR results had two to three technical replicates. Statistical analysis was done using GraphPad Prism v9.0 (San Diego, CA, USA). An n of 5 was used for immunofluorescent staining of pup brains; specific regions included the cortex, hippocampus, and cerebellum.

To evaluate the effects of butyrate administration on intestinal and brain injury in the experimental mouse NEC model, wild-type C57BL/6 pups were fed a combination of formula milk, enteral LPS, and timed hypoxia exposure. We first examined the protective effect of butyrate in the neonatal intestine. Pups with NEC showed epithelial layer sloughing and separation of the submucosal layer and lamina propria (Supplementary Material 1A) when compared with the butyrate-treated group, which had decreased epithelial disruption and an increased amount of villi with normal morphology. A histologic evaluation of the terminal ileum based on pathological damage on a scale of 0 (normal) to 4 (bowel necrosis) showed that the NEC pups had a higher degree of intestinal injury when compared with the control and NEC butyrate groups (NEC mean 2.5 ± 0.4 vs. NEC + butyrate; mean 1.2 ± 0.3, p < 0.05, n > 5/group). These findings are consistent with previous findings published by our group (11).

Microglia are specialized macrophages that are activated by induced brain injury (9, 22). To determine the protective effect of butyrate in the neonatal brain, we first examined the population changes of microglia in different regions of the neonatal brain. Experimental NEC resulted in a greatly increased expression of the microglial marker ionized calcium-binding adaptor molecule 1 (IBA1) in the cortex, hippocampus, and cerebellum (Figure 1). Interestingly, butyrate treatment decreased NEC-induced microglia activation in several brain regions (Figure 1), especially in the cerebral cortex and hippocampus. Because activated microglia would result in a dysregulation of oligodendrocyte development (22, 23), we next examined the population of maturing oligodendrocytes. CNPase was used as the oligodendrocyte marker as it recognizes progenitors and both mature and immature oligodendrocytes (24). Consistently, NEC led to significant decreases in the oligodendrocyte population in the cerebral cortex, hippocampus, and cerebellum. Similar to microglia, butyrate treatment attenuated the decrease in the oligodendrocyte population noted with NEC in all three regions of the brain (Figure 2).

The brain of neonatal mice with NEC was smaller and weighed less (305 ± 27 mg) than that of the control pups (390 ± 49 mg, p < 0.05) (Figure 3N). Moreover, the percent brain/body weight ratio of the NEC pups was higher (7%) than that of control pups (6%, p < 0.05) (Figure 3O). The loss of brain weight in the NEC group suggests that brain cell proliferation may be suppressed. Phosphorylated histone H3, a cell proliferation marker (25), was significantly decreased in the NEC group compared with the control group (except in the cerebellum). While brain weight in the NEC + butyrate group was not substantially restored when compared with the NEC group, butyrate administration helped rescue cell proliferation deficits induced by NEC (Figures 3A–L).

TLR4 signaling is considered a key signaling pathway in NEC pathogenesis (26). Therefore, we examined elements of TLR signaling in the cerebral cortex. The markers of canonical TLR signaling, such as phosphorylated P65 (RELA, NFκB subunit) and phosphorylated P38 MAPK, were both elevated in the NEC group but repressed with butyrate pretreatment, although the expression of TLR4 was not altered (Figures 4A,B). Previously, we have reported that SCFAs induce the intestinal expression of SIGIRR (11), a major inhibitor of TLR signaling (11). Butyrate promoted SIGIRR expression in the cortex (Figures 4A,B). Butyrate treatment attenuated NEC-induced TLR-mediated inflammation, which was evident by a decreased expression of proinflammatory cytokine interleukin 6 (IL-6), cytokines C-X-C motif chemokine ligand 1 (Cxcl1), and proinflammatory cellular adhesion molecule ICAM1 either at the protein or at the mRNA level (Figure 4). Cleaved caspase 3 (CC3), a marker of apoptosis, was elevated in NEC but suppressed with butyrate treatment, indicating a partial rescue of cell death (Figures 4C,D). Induction of IBA1 and GFAP expression in the cerebral cortex seen with NEC was repressed with butyrate pretreatment (Figures 4C,D).

Because the cerebellum also showed morphological changes in response to NEC that were decreased by butyrate treatment, we examined TLR4 signaling and related proinflammatory gene expression in the cerebellum. Experimental NEC modestly induced the marker of activated TLR signaling phospho-P65 (Supplementary Material 2A,B). NEC-induced proinflammatory markers ICAM1, GFAP, and IL-6 were reduced by butyrate treatment at the protein level in the cerebellum (Supplementary Material 2C,D). The MRNA levels of Il-6, Tlr4, Cxcl1, and 1l-1b were not altered in the cerebellum with NEC (Supplementary Material 2E,H). Butyrate treatment effectively prevented NEC-induced cell apoptosis in the cerebellum, which was evident from decreased CC3 expression. Interestingly, NEC-induced changes in the cerebellum were much more modest than the changes observed in the cerebral cortex. We also noted that SIGIRR was not as strongly induced in the cerebellum as in the cerebral cortex.