Blockade of High-Fat Diet Proteomic Phenotypes using Exercise as Prevention or Treatment

previous data from mice fed an isocaloric high saturated fat (SFA) or polyunsaturated fat (PUFA) diet. This identified several common changes including increased APOC2 and APOE, but also highlighted changes specific for either over-consumption of HFD (ALDOB, SERPINA7, CFD), SFA-based diets (SERPINA1E), or PUFA-based diets (Haptoglobin - Hp). Together, these data highlight the importance of early intervention with exercise to revert HFD-induced phenotypes and suggest some of the molecular mechanisms leading to the changes in the plasma proteome generated by high fat diet consumption. Web-based interactive visualizations are provided for this dataset (larancelab.com/hfd-exercise), which give insight into diet and exercise phenotypic interactions on the plasma proteome.


INTRODUCTION
Obesity is at epidemic levels in the developed world with many western countries having >25% of their population being classified as obese (1). Obesity promotes organ dysfunction and changes in circulating measures of metabolic disease (2). For example, obesity is frequently associated with non-alcoholic fatty liver disease (NAFLD) and associated premature mortality, ultimately linked to cardiovascular disease, some cancers, and type 2 diabetes (1). Obesity is often modelled in animals such as rodents through the provision of high fat diets (HFD) ad libitum. This allows the voluntary consumption of excess calories due to both the high energy density of the HFD and the increased appetite for high fat food compared to normal chow diets (3). In mice, the ad libitum consumption of HFD for 1 week is sufficient to induce liver insulin resistance (4). Usually more than 3 weeks of ad libitum HFD is enough to also induce muscle and adipose tissue insulin resistance (4), which are key pathways in the induction of type 2 diabetes. However, recent studies have shown that if mice are provided HFD in an isocaloric manner through pair-feeding with chow fed animals, the animals given isocaloric HFD quickly adapt to the altered nutrient proportions and display improved metabolic phenotypes compared to the chow animals (5). This suggests it is the excess consumption of calories and not just the high-fat content of the diet, which is responsible for the negative outcomes on metabolic health. This agrees with studies of ketogenic diets that are very high in fat content, but provide metabolic benefits (6).
Given the effects of obesity induced by the over-consumption of calories, strategies that aim to prevent and/or treat obesity-related dysfunction are of increasing interest. Exercise can attenuate adverse metabolic changes in a variety of disease contexts (7). Currently, the most effective exercise prescription (e.g.: intensity, duration, frequency) to utilise in the presence of obesity is unclear. To address this, a comparison between moderate constant-moderate endurance (END) and high-intensity interval training (HIIT) protocols, in different metabolic conditions, has been informative and shown both exercise regimes induced similar reductions in total body mass and waist circumference (8).
However, HIIT required ⁓40% less training time commitment to promote those changes. (8). Few studies have aimed to elucidate molecular pathways and mechanisms behind desirable changes induced by exercise, especially in a high fat intake environment. Recently, Groussard et al., when investigating effects of 10 weeks of END and HIIT, described tissue-specific effects in obese Zucker rats on oxidative stress modulators in white adipose tissue and skeletal muscle (9). Endurance exercise increased catalase and glutathione peroxidase activities in adipose tissue, whereas HIIT specifically increased glutathione peroxidase activity (9). These results are aligned with our research in high-fat fed mice, where 10 weeks of END versus HIIT exerted differential metabolic effects on liver, white adipose tissue and quadriceps muscle, and END specifically reduced fibrotic markers in liver, whereas HIIT increased UCP1 in adipose tissue (10), highlighting that these programs may have specific metabolic effects in an obesity context. While most relevant literature indicates that exercise in any form exerts metabolic benefits during obesity, it is unclear whether preventative exercise while obesity develops, or exercise treatment in an individual that is already obese, has similar effects or if certain exercise prescriptions have metabolic advantages (11).
In this study, we investigated the plasma proteome in mice for the effects of two aerobic exercise approaches (END and HIIT) that were provided in either a treatment regime, or preventative regime, for animals provided food ad libitum that was either a HFD, or a healthy control diet (10,12).
We identified many significant changes to the plasma proteome in response to the HFD and showed that only a small proportion of these could be blocked by exercise regimes. We showed that a preventative exercise regime is required to have significant effects and that both endurance and HIIT exercise modes provided similar benefits. By comparing to a previous dataset, we could identify HFD-induced plasma protein abundance changes that were specific to the over-consumption of HFD in our model. Furthermore, we could use this comparison to delineate the required fatty acid saturation needed to generate these changes. These data provide a comprehensive overview of the plasma proteomic phenotypes generated by HFD exposure and its interaction with several aerobic exercise regimes which are provided as an online resource (larancelab.com/hfd-exercise). USA), in groups of n = 4 mice per cage, for at least one week before entering experimental models. In the first study (exercise treatment) (10), mice at 10 weeks of age were randomly assigned into either chow, or HFD groups (n=36 per group). After 10-weeks of diet without exercise the mice in each dietary group were then randomised into one of three exercise groups: no exercise training (none), endurance treadmill running, or HIIT treadmill running (n=12 per group), which continued for 10-weeks.
In the second study (preventative exercise) (12), mice at 10 weeks of age were randomly assigned into either chow, or HFD groups (n=36 per group) and were then immediately randomised to one of three exercise groups: no exercise training (none), endurance treadmill running, or HIIT treadmill running (n=12 per group), which continued for 10-weeks.

Treadmill-based exercise regimes
Mice were acclimatized to the treadmill for 1 week and then a maximal running capacity (MRC) test was performed for each animal (10,12). The exercise intensity of the two different training programs (endurance or HIIT) was calculated using the animal's MRC. Importantly, the exercise programs were designed to be comparable in terms of exercise volume and distance covered per session. For endurance exercise a constant running session of 70% MRC (17 m/min) for 40 min was used, whereas for HIIT eight bouts (2.5 min each) at 90% of the MRC (22 m/min) intercalated by eight active rest periods (2.5 min each) at 50% of the MRC (12 m/min) (40 min total per session), was used. Each training program was performed in the morning, three times per week for 10 weeks. Nonexercised animals were not exposed to additional exercise.

Tissue collection
Mice were euthanized after an overnight period of feeding by terminal anaesthesia with isoflurane (3%) in oxygen, starting at 0900 h with all animals euthanized by 1100 h. Plasma was collected via cardiac puncture into tubes pre-coated with 0.5 M EDTA at 10% recovered blood volume. Whole blood was kept on ice and then spun at 1,500 x g for 15 minutes at 4°C, with collected plasma snap frozen in liquid nitrogen before being stored at -80°C.

Plasma sample preparation using SDB-RPS StageTips
Plasma sample preparation was performed as described previously (13). Briefly, 1 µl of plasma (70 µg protein) was added to 24 µl of SDC buffer (1% sodium deoxycholate, 10 mM TCEP, 40 mM chloroacetamide and 100 mM Tris-HCl pH 8.5) and heated to 95°C for 10 minutes. Once cooled to RT the sample was diluted 10-fold with water. LysC and trypsin were then added at a 1:100 ratio (μg/μg) and digested at 37°C for 16 h. An equal volume (250 μL) of 99% ethylacetate/1% TFA was added to the digested peptides and vortexed. Digested peptides were purified using SDB-RPS StageTips and the Spin96 as described (13). Dried peptides were resuspended in 30 L of 5% formic acid and stored at 4°C until analysed by LC-MS.

LC-MS/MS and analysis of spectra
Using a Thermo Fisher RSLCnano UHPLC, peptides in 5% (vol/vol) formic acid (injection volume 3 μL) were directly injected onto a 15 cm x 150 um C18Aq (Dr. Maisch, Ammerbuch, Germany, 1.9 μm) fused silica analytical column with a ~10 μm pulled tip, coupled online to a nanospray ESI source. Peptides were resolved over gradient from 5% acetonitrile to 40% acetonitrile over 25 min with a flow rate of 1,200 nL min−1 (capillary flow). Peptides were ionized by electrospray ionization at 2.3 kV. Tandem mass spectrometry analysis was carried out on a Q-Exactive HFX mass spectrometer (Thermo Fisher) using data-independent acquisition. The dataindependent acquisitions were performed as described previously using variable isolation widths for different m/z ranges (14). Stepped normalized collision energy of 25 +/-10% was used for all DIA spectral acquisitions.
All the RAW MS data and corresponding search outputs have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with the dataset identifier PXD018561, username: reviewer03304@ebi.ac.uk, password: Ws3E1fZc.
RAW data were analysed using the quantitative proteomics software Spectronaut Pulsar X (version 12.0.20491.11.25225 (Jocelyn)). DirectDIA analysis was used to identify peptides (15). The database supplied to the search engine for peptide identifications was the mouse UniProt database downloaded on the 15 th July 2019, containing 63,439 protein sequence entries. Enzyme specificity was set to semi-specific N-ragged trypsin (cleavage C-terminal to Lys and Arg) with a maximum of 2 missed cleavages permitted. Deamidation of Asn and Gln, oxidation of Met, pyro-Glu (with peptide N-term Gln) and protein N-terminal acetylation were set as variable modifications. Carbamidomethyl on Cys was searched as a fixed modification. To recalibrate for retention drift and intensity assignment for peaks the iRT profiling workflow was used (16). For peak list generation, interference (MS1 & MS2) correction was enabled, removing fragments/isotopes from potential quantitation if there was a presence of interfering signals, while keeping a minimum of three for quantitation. Spectra were de-isotoped based on RT apex distance and m/z spacing, de-multiplexing was not required. Each observed fragment ion could only be assignment to a single precursor peak list (15).
The FDR was set to 1% using a target-decoy approach. Spectronaut generated a custom mass tolerance and retention time tolerance for each precursor ion and its product ions. The threshold for accepting a precursor was set at a Q value <0.01 and each precursor must have >3 fragment ions. All other settings were factory default.

Experimental design and statistical rationale
The number of animals used per treatment group was established from previous studies in mouse muscle high molecular weight (HMW) adiponectin (12). We have used the values for HMW adiponectin as pilot data to provide suitable values for power calculation. At 90% power, with a worst case true difference in means of 200, a representative standard deviation of 130, probability of type I error of 0.05, and using a one-way ANOVA test the observations that are required is n=10 and we therefore used 10 animals per treatment group. For all datasets statistical analyses were performed using R (version 3.4.3) and processed data was plotted using Tableau (version 2019.2), where outliers beyond 1.5 times the inter-quartile range may have been excluded from plots to aid visualization. Fold changes for protein abundance were calculated using the median in each group. Statistical significance for changes induced by Diet condition (chow vs HFD), Exercise (none vs endurance vs HIIT), or the interaction between these was calculated using a two-way ANOVA. In addition, the HFD group only was analysed using a one-way ANOVA to examine the effect of exercise in this group alone. The resulting p-values from both tests were adjusted to control for multiple testing using the Benjamini-Hochberg correction. Significance was set at P<0.05, corresponding to a false discovery rate in the ANOVA of 5%.

Plasma proteome analysis of HFD-fed animals with exercise as treatment
The plasma proteome contains a large range of circulating factors secreted from many organs/tissues of the body, particularly the liver. The most abundant proteins in plasma are enriched in factors mediating lipid metabolism and lipid transport, which are of particular interest in relation to high fat diet studies. For the analysis of the interaction between a high-fat diet (HFD) and aerobic exercise we started with plasma samples from a standard exercise treatment model described previously (10). In this model, C57BL/6J male mice were used and the intervention was commenced at 10-weeks of age. Animals were fed either a HFD (45% kcal as fat), or normal chow (12% kcal as fat), for 10-weeks without any exercise treatment, and were then switched to a combined HFD and exercise treatment regime for the subsequent 10-weeks prior to plasma collection (Figure 1a). The treatment consisted of either: no exercise training (none); treadmill-based high-intensity interval training (HIIT); or treadmill-based endurance running (END). Exercise bouts of 40 minutes each were repeated 3 times per week. Each treatment group had 8-12 animals. We characterized the blood plasma from each animal harvested in the ad libitum fed state with undepleted plasma proteomics (13) using trypsin digestion and LC-MS/MS analysis by data-independent acquisition (DIA). This yielded the identification of 273 protein groups (Supplementary Table 1) from a total of >5,100 peptides (Supplementary Table 2 and Supplementary Table 3) across the six animal groups on each diet (Chow vs. HFD) combined with an exercise treatment (none vs. HIIT vs. END).
To determine those proteins that were significantly altered by either the diet, or the exercise treatment, we applied a two-way ANOVA to each protein detected in this proteomic dataset. The pvalues derived from this test were corrected for multiple testing using the method of Benjamini-Hochberg. This analysis also allowed us to determine if there was any significant interaction between the diet variable and the exercise treatment variable. This analysis detected 82 proteins that had a P<0.05 (<5% FDR) for a significant difference between the HFD and chow fed animals, irrespective of exercise (Figure 1b). This constitutes ~40% of the detected plasma proteome and shows the strong influence of the HFD on mammalian physiology. Comparing the different exercise regimes, we detected only 4 proteins that were significantly regulated by exercise treatment. Finally, there were no proteins that showed a significant interaction between the diet and exercise variables. Detailed analysis using a one-way ANOVA across the exercise treatment groups in only the HFD-fed animals showed that endurance exercise, but not HIIT, had a small but significant effect of reverting the HFD changes on 2 proteins: clusterin (CLU) and transthyretin (TTR).
To provide an overview of the changes induced by the treatments across all animals we plotted a heat map of the normalised LFQ intensity for all proteins detected with proteins separated by significance grouping (Figure 1c). This confirmed the consistency of response across each treatment group and the minor effect of exercise as a treatment for animals already exposed to HFD for a 10week period. Some of the proteins with the largest significant fold-change increase in response to the HFD were thyroxine-binding globulin (SERPINA7), apolipoprotein C-II (APOC2), haptoglobin (Hp), apolipoprotein E (APOE), apolipoprotein A-IV (APOA4), and fructose-bisphosphate aldolase B (ALDOB, liver-isoform) (Figure 1b,d). Among the most significantly down-regulated proteins were complement factor D (CFD -Adipsin), the major mouse 1-antitrypsin isoform (SERPINA1E), complement component C8  chains (C8b/g), and the soluble form of the leukemia inhibitory factor receptor (LIFR) (Figure 1b,d). Most proteins significantly altered by the HFD are known to be secreted from the liver (17), highlighting the importance of this tissue in the response to HFD and the liver's role in producing most of the abundant plasma proteins. Exceptions to this would include CFD and APOA4, which are mainly adipose and intestine derived proteins, respectively. A small subset of 10 proteins significantly altered by the HFD are not known to be secreted including ALBOB, PSMA2, SNX13, HSP1A1 and ERN1, which may have been released into the circulation due to liver cell death. Lastly, in this exercise-treatment cohort only 2 proteins transthyretin (TTR) and clusterin (CLU) had small but statistically significant exercise-induced effects to correct the HFD-induced 7 by guest on December 11, 2020 https://www.mcponline.org Downloaded from changes. Importantly, the exercise effects on these two proteins were only significant after endurance exercise, but not HIIT.

Plasma proteome analysis of HFD-fed animals with exercise-based prevention
We next wanted to analyze the effect of exercise as a preventative measure using plasma samples from a model previously described (12), where the animals started the HFD and the exercise regime at the same time and continued for 10-weeks in total (Figure 2a). We hypothesized that starting the exercise intervention earlier should have a much bigger effect to revert HFD-induced plasma proteome changes. Again, we used the same two-way ANOVA analysis as for the treatment model. This showed that 79 proteins had a significant difference (<5% FDR) between the HFD and chow fed animals, irrespective of exercise (Figure 2b). However, we now observed 24 proteins that were significantly regulated by exercise treatment and 18 proteins that showed a significant interaction between the diet variable and the exercise variable (Figure 2b). Detailed analysis using a one-way ANOVA across the exercise treatment groups in the HFD-fed animals only, shows that endurance exercise alone had a significant effect on 11 proteins, HIIT alone had an effect on 1 protein, and 12 proteins were significantly affected by both endurance and HIIT. The heatmap plot of the plasma proteome across the animal groups shows several subsets of proteins altered by the HFD that are modified similarly by both preventative exercise regimes (Figure 2c). The proteins with the largest significant fold-change increase in response to the HFD in this prevention model were very similar to those observed in the treatment model and included ALDOB, APOE, APOC2, Hp and serum amyloid A-4 (SAA4) (Figure 2b,d). The most significantly down-regulated proteins also showed similarity, including LIFR, EGFR, C8B/G, SERPINA1E, and CFD (Adipsin) (Figure 2b).
To directly compare the HFD response of the plasma proteome between the treatment and preventative exercise regimes we generated a scatter plot for all those proteins that showed a significant effect of HFD in both the exercise treatment and prevention with exercise models. The fold-change in both the prevention and treatment models were used as the plot x and y axes, respectively (Figure 3a). Proteins displaying the same response to the HFD would align on a 45degree line (Figure 3a -grey dashed line). This plot showed that most proteins up-regulated by the HFD had similar fold-changes (HFD/chow) across the exercise model datasets. In contrast, proteins down-regulated by HFD in general showed a reduced fold-change in the prevention model potentially associated with the decreased total exposure time to HFD (10 vs 20 weeks). Proteins that were significantly upregulated by HFD in both the preventative and treatment exercise model datasets were analysed for pathway enrichment (Figure S1a), which showed many of these proteins are involved in lipid metabolism and lipid transport and several are known to interact directly on lipoprotein particles such as HDL. Proteins that were significantly downregulated by HFD in both the preventative and treatment exercise model datasets ( Figure S1b) showed enrichment in protease inhibitors, coagulation factors, and complement factors, with many of these also known to interact in protein complexes. This comparison plot also highlights those proteins in the prevention model that showed a significant interaction between the HFD and exercise (Figure 3a). Examination of boxplots for some of these proteins highlights the differences in exercise-mediated reversion in response to the HFD (Figure 3b). For example, in both the exercise treatment and prevention models the APOE protein was induced ~2-fold by HFD alone with no exercise. Exercise treatments of either HIIT or endurance running showed a mild reversion of APOE protein abundance to chow values in the treatment model, but the response in the prevention model was significantly larger. A similar profile was also seen for histidine-rich glycoprotein (HRG) with the exercise response being even more marked in the prevention model. Conversely, both the soluble LIFR protein and SERPINA1E showed strong

Comparison of HFD-response when provided ad libitum versus an isocaloric diet study
Mice provided with HFD ad libitum (free access) will consume a similar mass of food to chow fed animals, but given the significantly higher energy density of HFD, those mice will gain weight rapidly compared to chow fed animals, as seen for both of the models examined in this study (10,12) (Figure S2). We wanted to determine the differences in the plasma proteome response (HFD/chow) when the diets were provided either ad libitum (this study) versus as an isocaloric pairing between HFD and chow-fed animals (5). In the previous isocaloric pair-feeding study of mice with plasma proteome analysis data, two different high fat diets were compared to a chow control diet (Ctrl), which were either high in saturated fat (SFA), or high in polyunsaturated fat (PUFA) (5). This allowed us to compare the effects of both diet over-consumption and diet composition on the plasma proteome. Again, we generated scatter plots comparing the fold changes (HFD/chow) between the Lundsgard et al study (5) and our treatment model data for all those proteins that showed a significant effect of HFD in our dataset (Figure 4).
Comparison of our data with either the SFA or PUFA response showed many similar responses such as increased abundance of APOC2, APOE and APOA4. Other common changes included decreased abundance of the soluble epidermal growth factor receptor (EGFR) and the complement protein 8 components (C8B/C8G). Therefore, these protein abundance changes are likely due to the consumption of food containing a high proportion of fat regardless of the amount of energy consumed. Several stand-out differences between the over-consumed versus isocaloric HFD were ALDOB, SERPINA7, and CFD (Adipsin) (Figure 4 -red points). Both ALDOB and SERPINA7 were markedly increased in plasma from animals over-consuming the HFD, whereas the response in the isocaloric HFD animals was either very small (SERPINA7), or the inverse response (ALDOB). Another protein showing an inverse response was CFD (Adipsin), which was significantly downregulated by over-consumption of the HFD, but was slightly increased in abundance in the isocaloric high fat diet dataset.
There were several proteins whose response to fatty acid composition was distinct, including 1-antitrypsin (SERPINA1E) and haptoglobin (Hp) (Figure 4 -black stars). SERPINA1E showed significant downregulation in our over-consumption HFD (composed of a mixture of saturated and poly/mono-unsaturated fatty acids) animals and the isocaloric SFA group (Figure 4a). In contrast, SERPINA1E was not significantly changed in animals of the isocaloric PUFA group (Figure 4b).
Therefore, SERPINA1E abundance is likely controlled by saturated fatty acid exposure regardless of amount consumed. Conversely haptoglobin was up-regulated in both our over-consumption HFD animals and the isocaloric PUFA group (Figure 4b), but not the SFA group, suggesting haptoglobin abundance is controlled by polyunsaturated fatty acid exposure regardless of the total high fat food consumed. Comparison of the raw data for selected proteins from each of these studies confirms these observations and highlights the dramatic change in ALDOB abundance in animals over-consuming the HFD (Figure 4c).

DISCUSSION
In this study, we used plasma proteome analysis to characterise the interaction between HFD consumption and aerobic exercise in mice. Our goal was to identify circulating proteins that were modulated by the HFD and to determine if exercise applied as either a treatment, or a preventative measure could rectify those HFD-induced changes. This study provides four key findings. First, animals fed a HFD showed marked changes in >30% of the detected plasma proteins compared to chow fed animals. Second, exercise applied from the beginning of the HFD as a preventative measure was able to revert the changes for <20% of these proteins. In contrast, the use of exercise as a treatment after an extended period on HFD only showed minor effects. Third, the type of aerobic exercise applied did not make a significant difference with both HIIT and endurance training providing similar benefits for most proteins with few exceptions. Lastly, a comparison of our ad libitum HFD response to a previous plasma proteome dataset where HFD was provided in an isocaloric regime identified several proteins that were differentially regulated. This comparison also allowed us to identify plasma proteins whose response was triggered by either saturated fatty acids, or polyunsaturated fatty acids in the diet. Together, these findings show that the interaction of dietary lipid consumption and aerobic exercise, when provided as early as possible, can revert significant physiological changes associated with these mouse models of human disease. Web-based visualizations for this dataset are available to the scientific community (larancelab.com/hfd-exercise), which allow interrogation of diet and exercise phenotypic interactions on the mouse plasma proteome.
The over-consumption of a high-fat high-calorie western diets for long periods has dramatic effects on mammalian physiology including weight gain and obesity, increased insulin resistance and 10 by guest on December 11, 2020 https://www.mcponline.org Downloaded from diabetes risk, and increased risk of cancer and cardiovascular disease (1,2). Reverting the physiological outcomes of HFD consumption can be mediated by food restriction, bariatric surgery, exercise, pharmacological intervention (e.g. GLP-1 receptor agonists), or combinations of these (18).
While exercise alone is not commonly effective in reverting these effects in humans, it does provide important improvements in insulin sensitivity and lipid metabolism, which may reduce cardiovascular event risk (19). In this study, we have examined the effect on the plasma proteome of exercise interventions applied either after the insult of the HFD (treatment), or at the same time as the deleterious HFD regime (prevention). These data clearly show a strong effect of preventative exercise on ~20% of the plasma proteins undergoing a significant response to HFD alone. In contrast, using aerobic exercise as a treatment only had minor effects. It is interesting that only ~25% of the HFDresponsive proteins had a significant response to exercise. These proteins were enriched for complement and coagulation-associated factors and acute phase proteins enriched for the GO term "response to stress" (Supplementary Figure 1c). Many of these exercise responsive proteins are largely synthesized and secreted from the liver. As has been shown previously, exercise triggers myokine signaling to the liver to mediate regulation of liver function, which in turn effects the secretion of liver-derived proteins (20). One reason the preventative exercise regime alone could provide superior beneficial changes for these proteins can be seen in the physiological data from these two models. This showed that using exercise as a treatment had no significant effect on the body weight gain caused by the HFD (10) (Figure S2), which was not unexpected and is widely observed in human clinical trials using exercise alone for weight loss. However, the prevention exercise regime had a significant effect on body weight gain, to an intermediate point between chow-fed animals and HFD-nonexercised animals (12) (Figure S2). Therefore, prevention of animals reaching the maximum weight gain possible due to preventative exercise is likely a strong contributing factor to this effect. Preventative exercise may inhibit weight gain by the increased energy expenditure induced, from both the exercise intervention itself and its associated increase in spontaneous activity (12).
When high-fat diets are provided in an isocaloric fashion based upon the energy consumption of mice fed a control chow diet, it is clear that most of the deleterious effects of HFD are removed and animals may experience beneficial effects (5). For example, liver triglycerides are significantly lower in mice fed an isocaloric HFD, which may be correlated with decreased liver abundance of de novo lipogenesis enzymes (5). A comparison of the isocaloric HFD responsive plasma proteome with the data in this study from mice provided HFD ad libitum, shows that there are several interesting differences in plasma proteome profile. Top among these is the ~8-fold increase of the liver aldolase enzyme (ALDOB) in plasma from our ad libitum HFD-fed mice compared to their chow-fed controls.
However, in isocaloric HFD-fed animals the abundance of ALDOB is not significantly altered compared to control (5). Liver aldolase is known to be increased in its plasma abundance after acute liver injury in humans and mice by acetaminophen (21), which is probably due to hepatic cell 11 by guest on December 11, 2020 https://www.mcponline.org Downloaded from death/damage leading to the leak of abundant cytoplasmic enzymes into the blood. Similarly, a recent study has shown that high liver aldolase abundance in plasma is significantly associated with NAFLD in both humans and mice (22). Therefore, the over-consumption of HFD and associated obesity is probably leading to the increase in plasma ALDOB abundance that we observed here in mice through liver damage. Consistent with this concept the plasma abundance of the alanine transaminase (ALT) enzyme, which is characteristic of NAFLD, was significantly increased in HFD mice ( Figure S2).
While the change in ALDOB abundance we observed was not significantly corrected by the preventative exercise regime, it was trending to be reduced by exercise with HIIT having the best corrective effect, which parallels a similar effect on ALT levels ( Figure S2).
Complement factor D (CFD) or adipsin is a largely adipocyte-derived protein (23), which we also observed to be differentially regulated by HFD over-consumption versus the isocaloric HFD dataset. Plasma CFD abundance was significantly decreased in HFD-fed animals by ~3-fold, whereas mice fed an isocaloric diet high in either saturated, or polyunsaturated fatty acids showed no significant change in CFD plasma abundance. Previous studies have shown that CFD is downregulated in rodent obesity models including HFD (24,25) and that it is a protective molecule for pancreatic beta cells in diabetes rodent models and humans (26,27). Our data shows that exercise provided as a preventative regime using either endurance or HIIT regimes was able to only partially revert the decrease in plasma CFD abundance due to HFD over-consumption. This may reflect its adipose tissue origin and the potential inability of exercise-based signals such as myokines and metabolite changes to correct the reduced secretion of CFD.
Some proteins detected in this study were induced by any increase in fat consumption regardless of the composition such as apolipoprotein CII (APOC2), which activates lipoprotein lipase and facilitates fatty acid uptake from the circulation (28). However, the liver and adipose tissue can respond to fatty acids of different classes through the ligand binding specificity of several transcription factors including PPARA/G, SREBP1 and HNF4A (29). Our analysis shows that two highly abundant plasma proteins, haptoglobin (Hp) and 1-antitrypsin (SERPINA1E), had differential regulation of their plasma abundance in response to the saturation of fatty acids present in their diet.
Haptoglobin was up-regulated 8-fold in the PUFA-rich isocaloric diet described previously (5), and by 4-fold in our ad libitum PUFA-containing HFD diet dataset. The major function of haptoglobin is to bind free hemoglobin in the blood, which can then be taken up by the liver for iron recycling (30).
Haptoglobin is also an acute phase protein that has a myriad of other related functions including antioxidant activity, anti-bacterial activity, and protease inhibition (30). In both humans and mice haptoglobin has been shown to increase in plasma abundance due to increased adiposity and increased adipose haptoglobin secretion (31-33). Haptoglobin has also been shown in obese rats to increase by ~2-fold when fed a diet that was 10% n-3 PUFA (34). This is in agreement with our observations of increased plasma haptoglobin in mice fed HFD ad libitum, where animals significantly increased their fat mass (Figue S2). Haptoglobin is largely secreted by the liver and increased plasma abundance of haptoglobin can result from increased liver synthesis as is known to occur in response to toxins such as turpentine (35). Adipose tissue is also another source of haptoglobin (25)(26)(27)(28)30), whose secretion is increased by many factors including TNF-alpha and IL-6 (36). These cytokines are known to be increased in their abundance when animals are placed on a high PUFA diet (37). The transcription factor peroxisome proliferator-activated receptor gamma (PPARG) in adipocytes is known to be induced by PUFA binding (38,39), however treatment of adipocytes with a strong agonist like thiazolidinediones cause decreased haptoglobin mRNA levels (40). This suggests further work will be needed to delineate the tissue source of the haptoglobin, the molecular mechanism leading to its increased plasma abundance and the consequences of its increased plasma abundance on a HFD, which may include increased cardiovascular disease risk (41).
In summary, this study provides a detailed comparison of the HFD response reflected in the plasma proteome of mice and the interaction between this diet and several exercise regimes used to correct those changes. Strengths of this study include the unbiased proteomic analysis methods used, which will sensitively detect significant changes in the more abundant plasma proteins. In addition, the C57BL/6J mouse model is a commonly used animal model for human disease including in effects of changes in macronutrient dietary composition and this study was complemented further through examination of two different exercise types, each as preventative or treatment regimes. Study limitations include that time points for analysis were fixed and few, and that the studies were undertaken in male mice only. This study has shown that a preventative endurance-based exercise regime provided the best outcome to block HFD-induced changes. In addition, we were able to contrast the HFD response in our model of diet over-consumption and obesity with an isocaloric feeding model to identify several proteins specific for over-consumption of HFD. This comparison also allowed us to delineate those changes regulated by saturated or polyunsaturated fatty acids in the diet. Future studies characterising plasma protein turnover rates would allow us to delineate if these changes are due to a change in protein secretion from the tissues of origin, or a change in the removal/degradation rate of these proteins from the plasma. We propose that the proteins identified here may be useful as markers of dietary composition and consumption history as well as providing information on the effectiveness of exercise-based interventions in humans. Ws3E1fZc. Web-based visualizations for this dataset are available to the scientific community (larancelab.com/hfd-exercise).      . (a,b) Proteins that were significantly regulated by the HFD in the exercise treatment model are plotted. The y-axis shows log 2 fold change for the HFD/chow response in the treatment model. In (a) the x-axis shows the log 2 fold change to a saturated fatty acid rich diet (SFA) compared to a chow control diet (Ctrl), which were both provided in an isocaloric manner (pair-fed) (5). In (b) the x-axis shows the log 2 fold change to a polyunsaturated fatty acid rich diet (PUFA) compared to a chow control diet (Ctrl), which were both provided in a isocaloric manner (pair-fed) (5). For (a,b) the size, shape and color of the points represents significant differences between either isocaloric versus over-consumption HFD models (red points), or the saturated versus polyunsaturated fatty acid diets (solid black stars), as indicated by the legends. (c) Box and whisker plots for specific proteins of interest. Each point represents protein abundance from an individual mouse. Asterisks placed at the top of the any plot represents statistical significance (P<0.05, or 5% FDR) across the treatment variables as indicated by the legend (n=12 per group). Figure S1. STRING Analysis of Significantly Regulated Proteins. Networks of interacting proteins within subsets of significantly regulated proteins were detected using the STRING database. Each protein is represented by a node and the lines between nodes indicate an interaction as defined in the legend. Gene set enrichment analysis has been used to identify significant and relevant protein interaction clusters, with common coloring of these nodes. (a) Proteins up-regulated in HFD compared to chow for both the treatment and preventative exercise regimes. (b) Proteins downregulated in HFD compared to chow for both the treatment and preventative exercise regimes. (c) Proteins that were either up or down regulated in response to HFD and also showed a significant effect of exercise to block these changes in the preventative exercise regime. Figure S2. Mouse Model phenotypic and metabolic measures. Box and whisker plots for various physiological measurements from the mouse models used in this study as reported previously (10,12), where each point represents data from an individual mouse. In general, HFD mice in the treatment and prevention regimes, were heavier, had greater % fat mass, higher circulating insulin and alanine transaminase (ALT) levels. Exercise regimes had the most notable effects in the preventative protocol, to attenuate the body weight gain, % fat mass and ALT levels. Asterisks placed at the top of the any plot represents statistical significance (P<0.05, or 5% FDR) across the treatment variables as indicated by the legend (n=12 per group).         (a) Proteins that were significantly regulated by the HFD in both the exercise prevention and treatment models are plotted. Each point shows the log 2 fold change for the HFD/chow response in the treatment model (y-axis) and the prevention model (x-axis). The size and color of the points represents significance of any exercise effect in the prevention model for the HFD group alone as indicated by the legend. (b) Box and whisker plots for specific proteins of interest showing a differential response to exercise applied as a either treatment, or prevention model. (c) Box and whisker plots for specific proteins that did not significantly respond to exercise. Each point in (b,c) represents protein abundance from an individual mouse. Asterisks placed at the top of the any plot represents statistical significance (P<0.05, or 5% FDR) across the treatment variables as indicated by the legend (n=12 per group).   Figure 4. Comparison of the plasma proteome response to HFD between isocaloric and over-consumption diet regimes.

Supplementary
(a,b) Proteins that were significantly regulated by the HFD in the exercise treatment model are plotted. The y-axis shows log 2 fold change for the HFD/chow response in the treatment model. In (a) the x-axis shows the log 2 fold change to a saturated fatty acid rich diet (SFA) compared to a chow control diet (Ctrl), which were both provided in an isocaloric manner (pair-fed) (5). In (b) the x-axis shows the log 2 fold change to a polyunsaturated fatty acid rich diet (PUFA) compared to a chow control diet (Ctrl), which were both provided in a isocaloric manner (pair-fed) (5). For (a,b) the size, shape and color of the points represents significant differences between either isocaloric versus over-consumption HFD models (red points), or the saturated versus polyunsaturated fatty acid diets (solid black stars), as indicated by the legends. (c) Box and whisker plots for specific proteins of interest. Each point represents protein abundance from an individual mouse. Asterisks placed at the top of the any plot represents statistical significance (P<0.05, or 5% FDR) across the treatment variables as indicated by the legend (n=12 per group).