Divergent effects of Western and Mediterranean diets on behavior and monocyte 1 polarization 2

Monocytes and macrophages, important mediators of innate immunity, health, and disease, are sensitive to diet and stress. Western diets promote inflammation and disease, while Mediterranean diets reduce inflammation and promote health, although the mechanisms through which diet alters immune function are unknown. Here, we conducted the first randomized, long-term diet trial in which macaques were fed either a Western- or Mediterranean-like diet to determine how diet and behavior interact to influence monocyte polarization, a precursor to subsequent health outcomes. Monocyte gene expression profiles differed markedly between the two diets, with over 40% of expressed genes being differentially expressed (FDR

monocyte-derived macrophages are phenotypically variable along a spectrum. In the circulation, 79 monocytes are generally defined as classical, intermediate, and non-classical, which upon 80 activation may become polarized along this spectrum which ranges broadly from 81 proinflammatory (M1-like) to regulatory/reparative (M2-like) phenotypes. An appropriate 82 balance of monocyte phenotypes is essential for a healthy immune system. Classically-activated 83 "M1" monocytes respond to proinflammatory cytokines such as tumor necrosis factor (TNF)-α 84 and interferon (IFN)-γ by becoming macrophages which propagate the inflammatory response to 85 infection (Mosser and Edwards, 2008). In contrast, M2 activated monocytes mobilize tissue 86 repair processes and release anti-inflammatory cytokines in response to interleukin (IL)-4, IL-13, 87 and transforming growth factor (TGF)-β (Mosser and Edwards, 2008). Thus, diet may alter 88 disease propensity by reprogramming the balance between these proinflammatory and anti-89 inflammatory monocyte subsets (Devêvre et al., 2015). 90 91 Diet is only one aspect of the environment that impacts health and survival by altering immune 92 function. Components of the social environment, such as social status and social integration, may 93 also affect these fitness outcomes by activating inflammatory programs in primary white blood 94 cells (Cole, 2013; Snyder-Mackler and Lea, 2018;Tung and Gilad, 2013). For instance, socially 95 subordinate rhesus macaques exhibit increased activation of the inflammatory response through 96 MyD88-dependent Th1-mediated inflammatory activation in response to a bacterial infection 97 (Snyder-Mackler et al., 2016). Given the similar molecular pathways through which diet and the 98 social environment impact health, these two environmental stimuli could compound or mitigate 99 one another. Moreover, given that changes in food intake can directly alter social behaviors 100 themselves (Kaplan et al., 1991;Warden and Fisler, 2008), dietary effects could plausibly act 101 through (or interact with) social environmental effects. 102 103 Earlier investigations of diet effects on immune function primarily focused on the effects of 104 single nutrient modifications, which precluded potentially important synergistic effects of 105 multiple nutrients in a given diet (Hu, 2002). Indeed, whole-diet modifications are more effective 106 at lowering heart rate than single-nutrient supplementation (Whelton et al., 1992) Suchanek et al., 2011). Further, no studies of whole diet manipulation in primates have probed 113 the molecular mechanisms through which diet can alter immune function-data that are critical to 114 identifying targets of future therapies and interventions. To address these gaps, we conducted a 115 long-term diet intervention study in nonhuman primates by feeding them either Mediterranean-116 or Western-like diets. After 15 months on the diet, we assessed monocyte polarization and gene 117 regulation by measuring genome-wide gene expression. As expected, proinflammatory genes 118 were more highly expressed in animals fed a Western diet relative to a Mediterranean diet, 119 indicative of shifts in monocyte phenotypes. Diet also affected gene co-expression patterns, 120 altering large modules of co-regulated genes, and influenced monkey behavioral phenotypes. 121 Western-fed monkeys became more socially isolated and exhibited more anxiety-associated 122 behaviors. Interestingly, these behavioral changes mediated some of the effects of diet on 123 monocyte gene expression. Together, these results suggest both direct and behaviorally-mediated 124 effects of diet on monocyte polarization that may contribute to chronic inflammatory diseases. 125  Figure 1. Experimental design and diet effects on body weight. A) Monkeys were housed in groups of 3-4 animals (n = 35 monkeys) and fed standard monkey chow diet for 8 months before being fed experimental diets.

G roups Established
Behavioral data were collected during the last 6 weeks of the baseline phase and the first twelve months of the experimental phase. Body weight was measured 5 months prior to, and 14 months after the start of the experimental phase. Monocytes were isolated from blood collected 15 months after the start of the experimental phase. B) Experimental diets were isocaloric with respect to macronutrients, but differed in food sources and relative amounts of micronutrients. Orange bars indicate nutrients with higher concentration in the Western diet formulation, while blue bars indicate higher levels of a given nutrient in the Mediterranean diet. See SI Table 1 for a more detailed comparison of the two diets.   Fig. 2A). There was no significant 144 difference in behavior between the two diet groups in all other measured behaviors (SI Fig. 1, SI 145 Table 2), or in any behaviors during the baseline phase (SI Table 2). 146

147
We next leveraged the fact that some behaviors were correlated with one another (SI Fig. 2), and 148

Diet alters monocyte gene expression and regulation 168
To test how diet affected circulating monocytes, we used RNA sequencing to measure genome-169 wide gene expression of purified CD14+ monocytes after 15 months on the experimental diets. 170 The first principal component of genome-wide gene expression, which explained 59.2% 171 variance, was significantly associated with diet (t(25.1) = -4.41, p = 1.7 x 10-4; Fig. 3A), and 40% 172 of the 12,240 expressed genes (SI Table 5A) were significantly differentially expressed between 173 the two diets (n = 4,900 genes, FDR < 0.05; SI Table 5B). 174

175
The number of diet-responsive genes was roughly balanced between those that were more highly 176 expressed in monkeys fed the Mediterranean diet (hereafter "Mediterranean genes"; n = 2,664) 177 and those that were more highly expressed in monkeys fed the Western diet (hereafter "Western 178 genes"; n = 2,236). While balanced in direction, the effects of the two diets differed significantly 179 in magnitude (t(3124.9) = -26.0, p = 4.9 x 10-135). The effect size of diet on Western genes was, on 180 average, 1.6-fold larger than on Mediterranean genes (Fig. 3B). Thus, the Western diet induced 181 stronger perturbations in monocyte gene expression than the Mediterranean diet. The most strongly affected Western genes included well-known inflammatory-related genes, 185 such as including interleukin-6 (βdiet = -1.66, FDR = 8.9x10-3; Fig. 3B), interleukin-1α (βdiet = -186 were more highly expressed in Western-fed monkeys ("Western genes") was, on average, 1.6-fold larger than  SI Table  220 5A,B for polarization categories). Western genes were enriched for M1-associated genes (n = 221 162 genes, 27.1% more than expected, 95% CI = +9.0%, +45.9%; Fig. 3C), but not M2-222 associated genes (n = 24 genes, 4.8% fewer than expected, 95% CI = -59.5%, +34.9%). 223 Conversely, the Mediterranean gene set was depleted for both M1-associated genes (n = 112 224 genes, 26.2% fewer than expected, 95% CI = -38.9%, -14.4%) and M2-associated genes (n = 20 225 genes, 33.3% fewer than expected, 95% CI = -60.0%, -3.3%). Together, these observations 226 indicate that a Western diet induces a more proinflammatory (M1-like) phenotype thus providing 227 a target for mechanistic follow-up studies. identified 445 gene pairs that were differentially correlated in Mediterranean-versus Western-236 fed monkeys at a relaxed FDR < 20% (SI Table 8B). The majority (97%) of these gene pairs 237 exhibited positive associations in one diet and negative associations in the other, as we have the 238 most power to detect differential correlation in gene pairs in which the direction of correlation is 239 opposite in the two diets. Some of these gene pairs were significantly correlated in reversed 240 directions (n = 23 gene pairs; nominal p-value < 0.05), suggesting that diet can reverse the co-241 expression relationship between two genes ( Figure 4A). Importantly, we identified 16 "hub" 242 genes that exhibited differential correlations with partner genes more so than expected by chance 243 (Fig. 4B, SI Table 8C). These hub genes were enriched for genes encoding transcription factors 244  4,900 diet-affected genes, 34% were also significantly associated with DAB in a univariate 258 model (n = 1,679, p < 0.05). Of these, DAB significantly mediated the effect of diet on the 259 expression of 872 genes (18% of all diet-associated genes, bootstrapped p < 0.05; Fig. 5A). 260 These DAB-mediated genes were significantly more likely to be Western genes (n = 482, 55%, 261 two-sided binomial test p = 0.0020), and were enriched in biological processes such as regulation 262 of leukocyte migration (FET pBH = 0.0042; SI Table 9A-C). We also tested the possibility that 263 diet effects on behavior were mediated through changes in monocyte gene expression but found 264 no genes that significantly mediated the relationship between diet and DAB (bootstrapped p < 265 0.05; Fig. 5B). 266

Discussion 268
Here, we showed for the first time that a controlled, whole-diet manipulation exerts profound 269 effects on monocyte gene regulation and behavior in a primate. Forty percent of the expressed 270 genes were differentially expressed between monkeys fed Western or Mediterranean diets, 271 indicating that diet dramatically altered monocyte programming. Beyond differential gene 272 expression, we identified differences in gene co-expression and enrichment of transcription 273 factor binding motifs, suggesting that diet alters gene regulatory networks. More specifically, we 274 found that the Western diet promoted a more proinflammatory monocyte phenotype, while the 275 Mediterranean diet had a more benign effect suggesting better balance in monocyte polarization. The Western diet promoted a more proinflammatory monocyte polarization relative to 291 Mediterranean diet, which is consistent with that seen in brain myeloid cells of mice fed a 292 Western diet (Yang et al., 2019). Conversely, Mediterranean diet could reduce proinflammatory 293 gene expression, as has been suggested in at least one study in humans (Camargo et al., 2012). 294 Future comparisons of each diet to the current standard diet for captive primates will help to 295 identify which of these two non-mutually exclusive hypotheses is most likely. It is also worth 296 For a subset (18%) of genes, the diet-altered behavior (DAB) phenotype mediated the effect of 313 diet on monocyte gene expression. Notably, the Western diet induced both a proinflammatory 314 monocyte phenotype, as well as social isolation and anxiety behaviors. It is, therefore, possible 315 that a Western diet contributes to inflammation by producing a more socially isolated or anxious 316 animal. Indeed, monocytes have been shown to be responsive to social stimuli, including social 317 isolation (Cole, 2019) and anxiety (Cole et al., 2015), suggesting that part of the effect of diet on 318 monocyte function may be due to diet-induced changes in socially relevant behaviors. In support 319 of this hypothesis, animals fed a Western diet exhibited significantly higher expression of pro-320 inflammatory genes involved in the "conserved transcriptional response to adversity" (CTRA; 321 Wilcox T-test p = 0.016; (Cole et al., 2015), and lower expression of antiviral-and antibody- Interestingly, Western diet induced substantial variation in multiple phenotypes, including body 337 weight, gene expression, and behavior. This heterogeneity in response to diet is consistent with 338 previous studies demonstrating that some individuals may be more resistant (or susceptible) to 339 the effects of a Western diet (Shively et al., 2009), presumably due to genetic variation or past 340 environmental experiences. However, we were unable to identify any consistencies in individual 341 responsiveness across the phenotypes (SI Fig. 7). For instance, monkeys that exhibited a strong 342 gene regulatory response to the Western diet did not also exhibit the largest increase in body In summary, we found that diet significantly alters behavior and monocyte polarization. The 351 Western diet increased the proinflammatory phenotype relative to a diet, which supports the role 352 of monocyte polarization in diet-associated chronic inflammatory diseases. Thus, avoiding a 353 Western-style diet and/or consuming a Mediterranean-style diet could be beneficial in preventing 354 or treating chronic inflammation and disease. The majority of the effects of diet are presumably 355 mediated through direct or combined actions of saturated/polyunsaturated fats, n-6:n-3 ratios, 356 pro-and anti-antioxidant characteristics, and other unique features of the protein and 357 carbohydrate constituents in the two diets. Monocyte reprogramming was also partially mediated 358 by the diet-induced changes in behavior, although the mechanisms by which this occurred are 359 unknown. Ongoing and future work will address interactions between social behavior (e.g., 360 social status) and diet to further understand how environmental stressors may impact 361 inflammation in the periphery and in the central nervous system. 362

Subjects 364
At study initiation, subjects were 43 adult (age: mean = 9.0, range = 8.2-10.4 years, estimated by 365 dentition), female cynomolgus macaques (Macaca fascicularis), obtained from Shin Nippon 366 Biomedical Laboratories (SNBL USA SRC, Alice, TX) and housed at the Wake Forest School of 367 Medicine Primate Center (Winston-Salem, NC) as described previously (Shively et al., 2019). 368 Briefly, monkeys were socially housed in groups of 3-4 and consumed standard monkey chow 369 (SI Table 1) during an eight-month baseline phase, after which they were fed either the Western 370  Table 1. 397 398

Behavioral Characterization 399
Behavioral data were collected weekly during two 10-minute focal observations, randomly 400 ordered and balanced for time of day, for 6 weeks during the baseline phase (2 total observation 401 hours/monkey) and for 12 months during the experimental phase (15 total observation 402 hours/monkey). Behaviors were collected as previously described (Shively, 1998), and combined 403 into summary behaviors (e.g., "aggression" was a combination of all total, noncontact, contact 404 aggressive events). A complete table of all behavioral data can be found in SI Table 2. In order to 405 quantify the overall impact of diet on behavior, we conducted a principal component analysis 406 using the R package FactoMineR (Lê et al., 2008). We tested for enrichment of transcription factor binding motifs within 2 kb (upstream or 471 downstream) of the transcription start sites of differentially expressed "Western genes" or 472 "Mediterranean genes" (FDR < 0.05) using the program HOMER (Heinz et al., 2010) and 473 equivalent regions around the transcription start sites of all genes expressed in these data as the 474 background set for enrichment testing. We searched for known vertebrate transcription factor 475 binding motifs and report the TF motifs passing a threshold of FDR < 0.05. 476 477

Gene-gene co-expression analysis 478
In addition to testing whether diet led to mean differences in gene expression between Western 479 and Mediterranean animals, we also tested whether diet impacted the correlation structure among 480 expressed genes (i.e., gene co-expression). Specifically, we used 'correlation by individual level  Normalization was performed by scaling expression values to mean 0 and unit variance within 487 Mediterranean and Western subsets of the data respectively, to ensure that distributional 488 differences between sample groups did not bias our results, following previously described Using this approach, we identified 445 gene pairs that were significantly differentially co-498 expressed as a function of diet at a 20% empirical FDR. Next, we performed two follow up 499 analyses to understand their biological import. First, we tested for the existence of 'hub genes', 500 defined as genes that displayed differential co-expression to their tested partner genes more so 501 than expected by chance. To define the null distribution for identifying hub genes, we randomly 502 sampled 445 gene pairs from the set of all 9730 tested gene pairs 1000 times and calculated the 503 number of partners a focal gene had in each sample; we considered a gene to be a significant 504 'hub gene' if it fell outside the 95th percentile of this distribution, which was equivalent to a focal 505 gene that displayed significant differential co-expression with 13 or more of its tested partner 506 genes. Second, we asked whether the set of 'hub genes' we identified were enriched for 507 transcription factors, relative to the background set of all 140 genes tested for differential co-508 expression. We performed this analysis because many of the proposed mechanisms to generate 509 large scale changes in gene co-expression patterns involve changes in transcription factor To test if DAB mediated the effect of diet on gene expression, we conducted mediation analyses. 516 We used a bootstrapping approach where we conducted 10,000 bootstrap iterations of two 517 models: (Model 1) the expression of each gene as a function of diet, and (Model 2) the 518 expression of each gene as a function of diet and DAB (Preacher and Hayes, 2004). For each 519 bootstrap iteration, we then calculated the mediation effect (i.e., the indirect effect) of DAB as 520 the difference between the effect size of diet in Model 1 (βdiet) and Model 2 (β'diet). We 521 considered there to be a mediation effect when the 95% confidence interval for the indirect effect 522 (βdiet-β'diet) did not include zero.