Diet D12079B (Research Diets Inc., New Brunswick, NJ) with 5% fructose added to the drinking water was used as the high-fat Western diet for this study. This diet was designed to provide 500 gm/kg of carbohydrates, 198 gm/kg of proteins and 210 gm/kg of fat. The percentage of calories from carbohydrates, proteins and fat in this diet was 43, 17, and 40%, respectively. This diet also contained 39 gm/kg of a standard mineral mix (S10001A) containing calcium carbonate and calcium phosphate and other essential trace elements. Diet 5,008 (Purina) was used as the low-fat control diet. The caloric distribution from this low-fat diet was approximately 57, 27, and 16% from carbohydrates, proteins and fat, respectively. Diet 5,008 was supplemented with standard minerals (2.5%) as part of the diet formula. High-fat diet alone was considered as the control for comparison with the low-fat diet or with Aquamin® and OCA as interventions in the high-fat diet.

Aquamin® is a product rich in calcium, magnesium, and trace elements, obtained from the skeletal remains of red marine algae (27) (Marigot Ltd, Cork, Ireland). Aquamin® contains calcium and magnesium in an approximately 14:1 ratio, along with measurable levels of 72 additional trace minerals including minerals with calcimimetic activity [for example, magnesium, strontium, and trace elements from the lanthanide family (28–30)], and essentially all of the minerals concentrated from ocean water in the algal fronds. Mineral composition was established by an independent laboratory (Advanced Laboratories; Salt Lake City, Utah) using Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES). Supplementary Table 1 provides a complete list of elements detected in Aquamin® and their relative amounts. Aquamin® is sold as a dietary supplement (GRAS 000028) and is used in various products for human consumption in Europe, Asia, Australia, and North America. A single batch of Aquamin®-Soluble (citrate-malate salt) was used for this study. Aquamin®-Soluble was added to the drinking water (18 mg/mL) throughout the entire in-life portion of the study. We have used Aquamin® in previously conducted long-term mouse studies (21, 31, 32).

Obeticholic acid (OCA) was prepared as a 3 mg/mL suspension in 0.5% methylcellulose. Animals received 0.25 mL OCA (approximately 30 mg/kg) by oral gavage five times per week beginning on week-8 of high-fat feeding.

The study was conducted in a mouse model - MS-NASH mouse (formally known as FATZO [(33, 34)]) by Crown Bioscience, Inc. Three cohorts of male MS-NASH mice (9 animals per group) were maintained on the high-fat diet, beginning at 6–7 weeks of age (at weaning) and after 1 week of acclimation. One group was kept as control (high-fat diet alone) while the other groups received Aquamin® or OCA as interventions along with the high-fat Western diet. Since mineral supplementation (Aquamin®) is proposed as a preventive agent, it was included along with high-fat feeding. OCA is envisioned as a therapeutic agent for existing disease, and animals were started on OCA at week-8 (16 weeks of age). A low-fat diet control group consisted of male C57BL/6 (Envigo) mice (nine animals per group) maintained on Diet 5,008. Food and water were provided ad libitum. The animal environment was maintained at 70–74°F with a 12-h light-dark cycle. During the in-life portion of the study, health-checks were done daily. At termination (after 16 weeks on diet), animals were euthanized by carbon dioxide asphyxiation. The entire in-life portion of the study and termination (euthanasia and tissue harvesting) was carried out by Crown Bio, Inc. (at their Lafayette, LA facility) under the approved Standard Operating Procedures at the testing site. All procedures involving live animals were approved by the Institutional Animal Care and Use Committee (IACUC). Crown Bio, Inc., is an AAALAC-accredited institution.

At euthanasia, blood was obtained by cardiac puncture and used to assess liver enzyme (alanine aminotransferase [ALT] and aspartate aminotransferase [AST]) levels as a measure of injury and triglyceride (TG) levels as an indicator of metabolic function. Livers were removed and weighed. After gross examination, a piece of tissue from the left lobe was immediately frozen in liquid nitrogen and used for proteomic analysis. Another tissue piece from the left lobe was fixed in 10% buffered formalin and used for histology.

Serum levels of AST, ALT and triglycerides were assessed using a Beckman chemical analyzer. All blood chemistry measurements were made following a standard operating procedure at Crown Bioscience.

Formalin-fixed liver tissue samples were processed for histology and 3–4 μm-thick sections were stained with hematoxylin and eosin (H&E) and picrosirius red (PSR), independently. H&E-stained tissue sections were evaluated under light microscopy for evidence of steatosis, ballooning hepatocyte degeneration and inflammation using the Kleiner scoring system (35). Kleiner et al have described the standardized histological scoring assessments of these NASH-related parameters in liver sections (35). These scoring methods have previously been applied to the histological scoring of mouse models of steatohepatitis by others (36, 37) and by us (20). Under this scoring system, a score of (0–3) was given for steatosis and inflammation separately, and a score of (0–2) for ballooning degeneration of hepatocytes. From these evaluations, a NAFLD activity score (NAS) was calculated by adding these individual scores. Finally, PSR-stained liver sections were used to assess collagen deposition and scored using a scale from 0–4 as defined by Kleiner et al. (35).

After microscopic evaluation, slides were digitized using the Aperio AT2 whole slide scanner (Leica Biosystems) at 40x with a resolution of 0.5 μm per pixel with a 20x objective. The scanned images were housed on a server and accessed using Leica Aperio eSlide Manager (Version 12.3.2.5030), a digital pathology management software. These digitized histological sections were viewed and analyzed using Aperio ImageScope (Version 12.3.3.5048), a slide viewing software by Leica.

Proteomic experiments were carried out in the Proteomics Resource Facility (PRF), a core laboratory in the Department of Pathology at the University of Michigan. For protein mass spectrometry analysis, we employed liver tissue from five mice from each group and assessed each separately. A cryopreserved tissue piece from the left lobe (of each mouse liver) was weighed and homogenized in Radioimmuno-Precipitation Assay (RIPA) - lysis and extraction buffer (Pierce, # 89901; ThermoFisher Scientific) for protein isolation, as described in our previous publications (38–40). Fifty micrograms of sample protein from each liver were digested separately with trypsin and individual samples labeled with one of 11 isobaric mass tags following the manufacturer's protocol. Tandem Mass Tag (TMT) 11plex Isobaric Label Reagent Set (ThermoFisher Scientific) was utilized for this application. After labeling, equal amounts of sample (peptide) from each liver sample were mixed together. In order to achieve in-depth characterization of the proteome, the labeled peptides were fractionated using 2D-LC (basic pH reverse-phase separation followed by acidic pH reverse-phase) and analyzed on a high-resolution, tribrid mass spectrometer (Orbitrap Fusion Tribrid, ThermoFisher Scientific) using conditions optimized at the PRF. MultiNotch MS3 analysis (41) was used to accurately quantify identified proteins/peptides. Data analysis was performed using Proteome Discoverer (v 2.4, ThermoFisher Scientific). MS2 spectra were searched against UniProt mouse protein database (17,011 sequences as reviewed entries; downloaded on 05/06/2020) using the following search parameters: MS1 and MS2 tolerance were set to 10 ppm and 0.6 Da, respectively; carbamidomethylation of cysteines (57.02146 Da) and TMT labeling of lysine and N-termini of peptides (229.16293 Da) were considered static modifications; oxidation of methionine (15.9949 Da) and deamidation of asparagine and glutamine (0.98401 Da) were considered variable. Identified proteins and peptides were filtered to retain only those that passed ≤ 2% false discovery rate (FDR) threshold of detection. Quantitation was performed using high-quality MS3 spectra (Average signal-to-noise ratio of 10 and <40% isolation interference). Differential protein expression in each treatment cohort was normalized to the high-fat alone cohort. Protein names retrieved from UniProt.org, and Reactome V78 (reactome.org) was used for pathway enrichment analyses (42) for species Mus musculus. STRING database-v11 (string-db.org) was used to detect protein-protein interactions and additional enrichment analyses provide information related to cellular components, molecular functions, and biological processes by Gene Ontology (GO) annotation. It also offered WikiPathways and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases to curate pathways. Only proteins with a ≤ 2% FDR confidence of detection were included in the analyses. The individual differential protein expression was established by calculating the abundance ratio of normalized abundances of each intervention sample to the high-fat alone samples. For the group analysis, individual data points were merged by groups to obtain means and standard deviations. The initial analysis was an unbiased search of differentially-expressed proteins and significantly altered proteins. Subsequently, we searched the database, specifically, for changes in the expression of keratins and other differentiation-related proteins. Mass spectrometry-based proteomics data were placed in a data repository known as ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD030954.

Group means and standard deviations were obtained for discrete gross and histochemical features as well as for individual biochemical values (ALT, AST and TG) and individual protein values obtained in the proteomic analysis. Data generated in this way were analyzed by analysis of variance (ANOVA) followed by unpaired t-test (two-tailed) for comparison using GraphPad Prism, version 9. Pathways enrichment data reflect Reactome-generated p-values based on the number of entities identified in a given pathway as compared to total proteins responsible for that pathway. The significance (p-value) is calculated by the overrepresentation analysis (hypergeometric distribution). For STRING enrichment analysis, the whole genome statistical background was assumed and FDR stringency was ≤ 1%. Data were considered significant at p < 0.05.

Animal weights and liver weights were assessed at euthanasia (Figure 1A). As expected, there were large differences between mice fed the low-fat diet and those on the high-fat diet irrespective of intervention. Inclusion of Aquamin® along with high-fat feeding had no effect on either total body weight or liver weight, but there was a small reduction in both parameters in the cohort of mice receiving OCA as compared to control mice on the high-fat diet. Liver enzyme (ALT and AST) levels and triglyceride levels are shown in Figure 1B. Levels of both ALT and AST increased substantially with high-fat feeding as did triglyceride levels. Both liver enzymes were also increased in mice receiving either Aquamin® or OCA along with high-fat feeding. With OCA, there was a modest reduction compared to levels seen in untreated mice on the same diet. Differences between the interventions and control (high-fat diet-fed mice) were not statistically different for either ALT or AST but all three high-fat groups were different from the low-fat control group. Serum triglyceride levels (evidence of metabolic dysfunction) were significantly elevated with high-fat feeding but neither Aquamin® nor OCA had a significant effect on this parameter.

Sections of formalin-fixed liver tissue were stained with hematoxylin and eosin and evaluated for steatosis, ballooning hepatocyte degeneration and inflammation. Additional sections were stained with PSR and assessed for collagen deposition. Data and histological representations of all four groups are shown in Figure 2. Steatosis was virtually undetectable in mice on the low-fat control diet but widespread in every animal on the high-fat diet after 16 weeks of feeding. Neither Aquamin® nor OCA reduced steatosis significantly compared to the high-fat control, but both interventions reduced ballooning degeneration and inflammation to some extent. When scores from the three parameters (steatosis, ballooning degeneration and inflammation) were added together (as in citation 35) to give a NAFLD activity score, both Aquamin® and OCA were different from high-fat alone. However, the degree of elevation was small to begin with (especially with inflammation) and the overall significance of the modest reduction is difficult to gauge. PSR staining demonstrated collagen deposition within the liver parenchyma of mice on the high-fat diet while sections from low-fat control mice did not show staining except around vascular elements and in relation to the capsule. Some reduction in the collagen deposition score was noted with both interventions. With Aquamin®, the reduction in collagen deposition reached a level of statistical significance. However, as with inflammation, the overall PSR staining score was low (<1.5 out of 4) after 16 weeks on the high-fat diet.

The number of upregulated and downregulated proteins in C57BL/6 mice on the low-fat diet (relative to high-fat control mice) and the number of up- and downregulated proteins with Aquamin® or OCA in comparison to the high-fat control mice is presented as Figure 3A (2-fold cutoff) and Figure 3B (1.5-fold cutoff). It can be seen from the bar graphs that the low-fat control diet was responsible for a substantial number of protein changes. With the 2-fold cutoff and ≤ 2% FDR, 182 proteins were upregulated, and 126 proteins were downregulated with the low-fat control diet. With Aquamin® and OCA, a total of 86 and 76 proteins, respectively, were upregulated to the same extent, while only five and eight, respectively, were downregulated. When the less stringent cut-off (1.5-fold) was used, more proteins were altered with each intervention but the trends were the same.

Venn diagrams (Figures 3A,B) show the overlap between protein alterations with low-fat feeding and the other two interventions. Among the proteins elevated with either Aquamin® or OCA, ~50% (at either cut-off) were in common with proteins upregulated in the low-fat control diet. In contrast, the number of downregulated proteins with either intervention that overlapped with proteins downregulated with low-fat feeding was much lower on a percentage basis. This was seen with either cutoff but was more extreme with a 2-fold cutoff. Overall, 55 upregulated proteins were common between the two interventions (OCA and Aquamin®), while only two downregulated proteins were common between these two groups. Supplementary Tables 2, 3 provide lists of upregulated and downregulated proteins common at 2-fold among three groups (low-fat, OCA and Aquamin® on the high-fat diet) across five mice per group.

Figure 4 shows the fold-change distribution of individual proteins responsive to Aquamin® in the presence of a high-fat diet. Upregulated proteins that met the criterion of statistical significance at p < 0.05 are shown in blue while downregulated proteins are shown in red. The bold colors (blue and red) indicate proteins that were statistically different and also at least 1.5-fold up- or downregulated.

Table 1A identifies the upregulated proteins that met the dual criteria (statistical difference from the high-fat control at p < 0.05 and at least 1.5-fold increase) in response to Aquamin®. Supplementary Table 4 provides a comprehensive list of upregulated proteins that met the criterion of statistical difference from control, irrespective of fold-change. While only a small number of proteins (12 in total) met both criteria, a total of 183 upregulated proteins were statistically increased. Reactome pathway analysis was utilized as a way to help delineate how altered protein expression in response to Aquamin® influences biological function. As can be seen from the altered pathways, the upregulated proteins (for example, Apolipoprotein A-I [Apoa1]) impacted lipid metabolism and trafficking of lipids (Table 1B). Apoa1 was significantly upregulated with Aquamin® and it did not change in OCA and low-fat groups. Apoa1 is a major component of high-density lipoprotein (HDL) and may have a protective role against liver cancer (43). Supplementary Table 5 lists top pathways affected by the altered (significantly upregulated) proteins presented in Supplementary Table 4.

Aquamin®-responsive proteins that met the dual criteria of 1.5-fold decrease and statistical significance are shown in Table 1C and those that were statistically-different (and downregulated) from high-fat control, regardless of fold change, are shown in Supplementary Table 6. A total of 14 proteins met both criteria while a total of 130 were statistically different from the high-fat control, independent of fold-change. As with upregulated proteins, Reactome pathway analysis was used to help identify potential mechanisms of action (Table 1D). Proteins whose expression was reduced by Aquamin® impact, primarily, amino acid metabolism, although purine and pyrimidine metabolism also appear to be targets. Among the downregulated proteins were multiple cytochrome P450 family members (10 in all) including three (i.e., 2C54, 2C50, 2C29) that were among the most highly downregulated (Supplementary Table 6). Cytochrome P450 family members are highly-expressed in the liver and serve to metabolize and detoxify numerous chemicals, including fats (44). During detoxification, cytochrome P450 enzymes can cause oxidative stress and may play a role in cellular injury, which can lead to NASH (45). Supplementary Table 7 lists top pathways affected by all of the significantly downregulated proteins presented in Supplementary Table 6. Amino acid metabolism, bile acid and bile salt metabolism and recycling of bile acids and salts are among the significantly downregulated pathways identified.

Tables 1A,C also show for comparison how the Aquamin®-sensitive proteins responded to low-fat feeding and to OCA. It can be seen that while there was overlap between Aquamin® and OCA (as shown in Figure 3) in many of the highly upregulated proteins, the response to OCA was not as robust as the response to Aquamin®. Fold-change was smaller and only two proteins were statistically different from high-fat control, although both interventions were introduced in the presence of the same diet and the same mouse strain. In low-fat mice, only one protein met the same criteria (Table 1A). In regard to downregulated moieties, of the 14 Aquamin®-sensitive proteins that met the dual criteria, nine were also responsive to OCA by the same criteria. In contrast, there was little overlap between Aquamin®-responsiveness and response to the low-fat diet. In fact, many of the protein changes were in the opposite direction (Table 1C).

Supplementary Tables 8, 9 present proteomic enrichment data curated by the STRING database with 183 significantly upregulated proteins and 130 significantly downregulated proteins, respectively, in response to Aquamin®. These datasheets provide information related to biological processes, molecular functions and cellular components along with additional pathways curated by KEGG and Wiki (as part of STRING enrichment analysis). Among these, inflammation-related pathways were significantly downregulated with Aquamin® (Supplementary Table 9).

After analyzing data based on statistical significance, the proteomic database was searched for the most highly up- and downregulated Aquamin®-responsive proteins based on fold-change independent of statistical significance. Table 2 presents a list of the most highly upregulated proteins (2-fold or greater) with Aquamin®. As a group, epithelial cell keratins were prominent in the protein list. Of the top 22 most highly upregulated proteins, 10 were keratins (Table 2). Also, nine of these keratins were common at a 2-fold-change with the low-fat group (Supplementary Table 2). Of interest, the major keratins found in the liver – i.e., keratin 8, and 18 (46) – were not altered (up or down) with Aquamin® (Figure 5). Keratins are structural components of epithelial cells and play key roles in differentiation, morphogenesis, barrier formation and tissue integrity. They are also important signaling intermediates (47). How each of the keratins identified here functions in the liver, specifically, is not known. The effects of low-fat feeding and the effects of treatment with OCA on these same proteins are included in Table 2 (complete protein list) and Figure 5 (keratins). As can be seen, some of the same keratins were increased with low-fat feeding but the degree of upregulation was less than that observed with Aquamin®. While the low-fat diet does not provide the same trace elements as does Aquamin®, D5008 chow contains some minerals as part of its formulation. Thus, it is not unreasonable to observe some change in the keratin expression profile in livers of mice on the low-fat diet. In contrast, with the exception of keratin 79, upregulation of individual keratins with OCA was almost non-existent.

In addition to keratins, three members of the tubulin family were also upregulated in response to Aquamin® (Table 2). Tubulins constitute another set of important structural proteins within cells, as they are the major constituents of the microtubule system of the cytoskeleton (48). Of interest, the tubulin proteins were responsive to both low-fat feeding and OCA treatment as well as to Aquamin® (Table 2). Browsing these upregulated (with 2-fold) proteins there are some interesting moieties worth mentioning. Arylsulfatase A (Arsa) was upregulated with Aquamin® with no change in the two other groups. Arsa is regarded as a lysosomal enzyme that is involved in lysophospholipid metabolism and improves insulin sensitivity (49). Another study also suggests its role as a component of the extracellular matrix (50). Another protein, signal transducer and activator of transcription 1 (Stat1) protein was upregulated (at 2.5-fold) with Aquamin® and it is considered an important negative regulator in liver fibrosis by inhibiting stellate cell proliferation through attenuation of TGF-beta signaling (51).

Reactome pathway analysis was utilized to help identify the various pathways influenced by the protein changes noted in Table 2 with Aquamin®. Not surprisingly, pathways related to differentiation were the most highly affected (Table 3). Among these were keratinization, cornified envelope formation, hemidesmosome assembly, cell junction organization, gap junction formation and gap junction assembly. In addition, pathways involved in intracellular trafficking of proteins as well as protein trafficking between the cytoplasm and plasma membrane were significantly influenced by the protein changes occurring in response to Aquamin®. Protein trafficking events are well-known to depend on cations in the environment (52, 53). Similarly, plasma lipoproteins clearance and LDL clearance pathways were also upregulated. Another pathway of interest was hedgehog signaling in the off-state. Signaling through the hedgehog pathway is associated with chronic liver injury (54, 55) and hedgehog signaling may have implications in hepatocellular carcinomas. Additionally, a pathway known as regulation of PTEN stability and activity was also upregulated. Phosphate and tensin homolog (PTEN) gene is considered a tumor suppressor and its loss can contribute to liver injury (56). All of the Aquamin®-responsive Reactome pathways that met the criterial of having an entities p-value of <0.05 are presented in Table 3.

The database was also searched for proteins that were downregulated by 2-fold or greater with Aquamin®. A total of five individual proteins met this criterion. Of the proteins in the list, four of the five were also statistically significant and presented as part of Table 1C.

While the response to Aquamin® was the focus of this study, we also assessed protein changes driven by low-fat feeding or by OCA in high-fat mice. Supplementary Tables 10, 11 show proteins upregulated and downregulated, respectively, by 2-fold or greater with low-fat feeding. Reactome pathways influenced by these protein changes are shown in Supplementary Tables 12, 13. A large majority of the protein changes observed with low-fat feeding (both up- and downregulated) are related to fat and carbohydrate metabolism. This is consistent with previous proteomic assessment of diet-influenced protein profile in livers of both rats (57) and mice (58).

Supplementary Tables 14, 15 show protein changes in mice on a high-fat diet treated with OCA. Supplementary Tables 16, 17 show pathways responsive to OCA. The most interesting (if not entirely unexpected) finding was the prominent downregulation of pathways related to cholesterol metabolism and bile acid formation. Supplementary Table 18 highlights all the pathways involved with upregulated proteins (55 in total at 2-fold-change) common between the two interventions (Aquamin® and OCA). There are some similarities between the two groups mainly due to the proteins of the tubulin family. This finding provides a sharp distinction between the likely mechanistic events driven by Aquamin® and those responsive to OCA. This finding helps validate the utility of the proteomic screening approach employed here.