Paternal obesity and epigenetic inheritance of breast cancer: The role of systemic effects and transmission to the second generation

Background While genetics explains some familial breast cancer cases, we showed that environmentally-induced epigenetic inheritance of breast cancer can also occur in rodent models. We previously reported that paternal consumption of a high-fat diet and ensuing obesity increased breast cancer susceptibility in the offspring (F1). Nevertheless, it is still unclear whether paternal-induced programming of breast cancer in daughters is associated with systemic alterations or mammary epithelium-specific factors. It also remains to be determined whether the ancestrally programmed breast cancer predisposition in F1 progeny can be transmitted to subsequent generations. Methods Male mice (F0) were fed either a control (CO) diet or an obesity-inducing diet (OID) for seven weeks and then mated with female mice (F0) reared on a CO diet. The resulting offspring (F1), also exclusively fed CO diet, were either used for mammary gland and tumor transplantation surgeries or to generate the F2 generation. To induce the mammary tumors, female mice were treated with 7,12 dimethylbenz[a]anthracene (DMBA). Total RNA extracted from F0 or F1 males sperm was used for small RNA-Seq analysis. Results Mammary glands from F1 CO female offspring exhibited enhanced development when transplanted into OID females [OID(CO-MG)], as shown by higher mammary gland area, epithelial branching and elongation, compared to CO females that received a CO mammary gland [CO(CO-MG)]. Similarly, mammary tumors from F1 CO female offspring transplanted into OID females [OID(CO.T)] displayed improved growth with a higher proliferation/apoptosis rate. We also found that granddaughters (F2) from the OID grand-paternal germline showed accelerated tumor growth compared to COxCO granddaughters (F2). Transmission of breast cancer predisposition to the F2 generation through OID male germline was associated with alterations in specific sperm tRNA fragments (tRF) in both F0 and F1 males. Conclusions Our findings indicate that systemic metabolic and mammary stromal alterations are the most significant contributors to paternal programming of mammary gland development and cancer predisposition in female offspring rather than mammary epithelium confined factors. Our data also show breast cancer predisposition in OID daughters can be transmitted to subsequent generations and could explain some familial cancers, if confirmed in humans.


Introduction 52
Genetic predisposition explains most but not all familial diseases, including breast 53 cancer [1]. It is increasingly evident that epigenetic inheritance of disease can also occur and 54 may explain some inherited conditions. There is strong indication that, at conception, parents 55 pass more than genetic material to their offspring. They also transmit a molecular memory of 56 transplantation surgery as previously described [22,23]. The experimental design is shown in 121 and DAB-stained cells. Differences between groups were analyzed using one-way ANOVA,194 followed by post-hoc analyses. 195

Analysis of cell apoptosis 196
Cell apoptosis analysis was performed in transplanted mammary glands and tumors by 197 morphological detection. Tissues were fixed in neutral buffered 10% formalin, embedded in 198 paraffin, sectioned (5 µm) and stained with hematoxylin and eosin (H&E). Cells presenting loss 199 of adhesion between adjacent cells, cytoplasmic condensation and formation of apoptotic bodies 200 were considered apoptotic as described before [28]. Sections were photographed using an 201 Olympus IX-71 Epifluorescence microscope at 40x magnification. Twenty areas were 202 photographed randomly, and the number of apoptotic bodies counted. Images were evaluated 203 with ImageJ software (NIH, USA). Differences between groups were analyzed using one-way 204 ANOVA, followed by post-hoc analyses. 205

Mature spermatozoa collection and purification 206
CO and OID-fed males (F0) and their male offspring (F1) were euthanized and their caudal 207 epididymis dissected for sperm collection. The epididymis was collected, punctured, and 208 transferred to tissue culture dish containing M2 media (M2 Medium-with HEPES, without 209 penicillin and streptomycin, liquid, sterile-filtered, suitable for mouse embryo, SIGMA, product 210 #M7167) where it was incubated for 1 hour at 37°C. Sperm samples were isolated and purified 211 from somatic cells. Briefly, the samples were washed with PBS, and then incubated with SCLB 212 (somatic cell lysis buffer, 0.1% SDS, 0.5% TX-100 in Diethylpyrocarbonate water) for 1 hour. 213 SCLB was rinsed off with 2 washes of PBS and the somatic cell-free purified spermatozoa 214 sample pelleted and used for RNA extraction. 215

Small RNA-Seq and Gene Ontology (GO) analyses 216
Total RNA was isolated from sperm using Qiagen's miRNeasy extraction kit, according to the 217 manufacturer's instructions. One hundred ng of column-purified sperm RNA was used to prepare 218 individually barcoded small-RNA libraries. Samples were barcoded, pooled, precipitated and 219 separated on a 15% polyacrylamide gel (PAGE). The gel was stained with SYBR® gold dye 220 and the small non-coding RNA segment corresponding to transfer RNA fragments or tRFs  45 nucleotides) excised and purified using a cDNA library preparation method described 222 previously [29]. This library preparation method was demonstrated to be highly reproducible 223 using total RNA with RNA Integrity Numbers as low as 2.0 [29]. Indexed, single-ended small-224 RNA sequencing libraries were prepared. For each individual barcoded library, at least 10 225 million reads (raw data) were generated using an Illumina Hi-Seq 2500. The raw reads were 226 subjected to 3' adapter trimming and low quality filtering using Trimmomatic program [30]. The 227 high quality clean reads were aligned to the mouse genome. tRFs tags were mapped to the mouse 228 genome (GRCm38/mm10 reference genome) in order to analyze their genomic distribution and 229 expression in the different sperm RNA samples. Small RNA tags were annotated and aligned to 230 known t-RNA sequences using Ref-seq, GenBank and Rfam database using blastn with standard 231 parameters. To analyze the differential expression of tRFs between CO and OID groups, tRFs 232 were normalized to TPM (Transcripts Per Kilobase Million). tRFs with a P value less than 0.05 233 were considered significant, with an appropriate correction for multiple testing [31]. Target genes for the 5 overlapping tRFs in OID F0 and F1 males were predicted using TargetScan 235 Mouse custom seedmatch and modified miRanda algorithm (energy <= -20 and score >= 150). 236 The common predicted genes were then uploaded to PANTHER 15.0 for GO term and pathway 237 analysis, final lists were filtered by FDR < 0. 25. 238 239

Offspring of OID fathers have impaired metabolic function and altered mammary gland 241 development 242
We previously reported that paternal consumption of obesity-inducing diets (OID) at the pre-243 conception window increased female offspring's susceptibility to breast cancer [15,16]. In those 244 studies, we also described mammary gland morphological changes as well as metabolic 245 dysfunction-a phenotype also reported by others-in offspring of obese fathers [16,18,19,32]. 246 Our present results corroborate our previous findings as OID offspring (F1) displayed impaired 247 metabolic function with both F1 males and females showing significantly reduced insulin 248 sensitivity compared to CO offspring (P=0.002, P=0.011, Fig. 1a-f). In addition, mammary 249 glands of OID daughters also showed increased number of terminal end buds (TEB), higher 250 epithelial branching and elongation, although only the last parameter reached statistical 251 significance compared to CO (Table S4). Those phenotypes were not associated with body 252 weight gain (Fig. S4) as OID offspring weights either did not differ from or were lower than CO. 253

Systemic effects play a larger role in normal mammary tissue and mammary tumor growth in 255
offspring of OID fathers 256 Next, we examined the contributions of systemic alterations and mammary tissue specific factors 257 (stroma vs. epithelium) to the increased breast cancer development in offspring of obese fathers. 258 In the first experiment, female offspring of either CO or OID-fed males underwent a mammary 259 gland transplantation surgery. CO mammary glands transplanted into OID females 260 [OID(CO.MG)] exhibited accelerated development (Fig. 2a-e) as shown by higher mammary 261 gland area (p=0. 032, Fig 2b), higher mammary branching and higher epithelial elongation 262 (p=0.014; p=0.008, respectively, Fig. 2c-d), but not higher number of TEBs (Fig. 2e)

Consumption of OID alters the tRF content in sperm of fathers (F0) and their sons (F1) 283
Recent studies have suggested that sperm non-coding RNAs play a role in transmitting 284 environmentally-induced information from fathers to offspring. Transfer RNA fragments or tRFs 285 make up the majority of small RNAs in mature sperm and can recapitulate the effects of paternal 286 obesity in offspring [3]. As reported before, GlyGCC and GlutCTC were the most abundant 287 tRFs in sperm of both fathers (F0) and their male offspring (F1), representing about 70% of all 288 tRFs ( Fig. 4a-b) [8,19]. We also found that consumption of OID altered specific tRFs in both 289 father ( Fig. 4c) and sons (Fig. 4d), with five tRFs overlapping between the two generations 290

Breast cancer predisposition in OID daughters is transmitted to a second generation 297
Given the tRF alterations observed in the F1 OID offspring germline, we then asked whether 298 breast cancer predisposition in OID daughters could be inherited by a second generation of 299 females. To this question, we produced the F2 generation by mating F1 male offspring from OID 300 fathers with F1 females from either CO [OIDxCO] Fig.5a). The incidence of mammary tumors at the end of 306 the monitoring period was also significantly higher in F2 OIDxOID females compared to the 307 COxCO group (p=0.037; Fig. 5b), suggesting a synergistic effect of both the male and female 308 OID germlines. Tumor latency and tumor mortality rates in the OIDxCO group were slightly 309 shorter than in all other groups, however results did not reach statistical significance ( Fig.5c-d). We previously reported that paternal obesity increases tumorigenesis in offspring, including 319 breast cancer [15,16,18]. In this follow-up study, we showed that metabolic disturbances in the 320 F1 generation play a key role in the increased breast cancer development observed in offspring 321 of obese fathers in a mouse model. We also report that the paternal obesity leads to higher cancer 322 development in two successive generations. Transmission of the increased breast cancer 323 phenotype into the F2 generation was associated with epigenetic changes in the germline, namely 324 alterations in the abundance of tRFs present in OID F1 male sperm. 325 The first aim of our study was to dissect the distinct contributions of systemic effects and 326 mammary tissue-confined factors to increased breast cancer development in daughters of obese 327 fathers, as we had observed both metabolic dysfunction and mammary gland abnormalities in 328 previous studies [15,16]. Our results showed that systemic metabolic effects, likely acting 329 though the mammary stroma, in OID daughters play a larger role compared to the mammary 330 epithelium. Further, tumors from CO offspring transplanted into OID daughters acquired a 331 growth advantage compared to those transplanted in controls, suggesting that the stroma in OID 332 females allows for better implantation and tumor growth. It is still possible that mammary 333 epithelium confined factors play a role in the increased tumor development in OID offspring, 334 however, they play a reduced role compared to systemic and mammary stromal effects according 335 to our data. While it has been traditionally thought that the epithelium is the compartment with 336 the dominant contribution regarding breast cancer initiation and growth and mammary tissue 337 regeneration, some studies have highlighted the importance the stroma microenvironment, 338 particularly adipocytes, on normal mammary development and malignant transformation of the 339 mammary epithelium [33][34][35][36]. Our analyses are in agreement with those findings and suggest that 340 the stroma plays an important enabling role for tumor growth. 341 It is also well established in epidemiologic studies that metabolic conditions such as obesity, 342 metabolic syndrome and diabetes are important risk factors for breast cancer and other 343 malignancies [37][38][39][40] and data from animal models offer support to those findings [41,42]. In 344 line with that, we demonstrated that a milieu of metabolic dysfunction and altered stromal 345 microenvironment creates conditions for increased proliferation and survival of both normal and 346 tumorigenic mammary cells as demonstrated by our transplantation studies.
While we have not directly investigated the molecular mechanisms behind the findings reported 348 here, it is known that metabolic dysfunction contributes to cancer development via extrinsic and 349 tumor-intrinsic factors [43]. Metabolic-induced alterations in growth factors signaling, 350 inflammation and the associated microenvironment, as well as changes in tumor metabolism 351 itself are all major contributors to cell proliferation and cancer development [43]. Not 352 surprisingly, our previously reported results show that paternal obesity or malnutrition alters the 353 molecular make-up of tumors which show increased growth factor and energy sensing signaling 354 and altered amino-acid metabolism [15][16][17][18]. 355 We also examined whether the offspring's breast cancer predisposition programmed by paternal 356 obesity could be inherited by a second unexposed generation. We found that the risk of breast 357 cancer is passed down to the OID grandchildren equally via the F1 male and female germlines. 358 Our data also suggest that there is a synergistic effect when both F1 parents had an obese father, 359 with their descendants showing not only accelerated tumor growth but also higher tumor 360 incidence. As with the F1 generation, F2 females from the OID lineage showed signs of 361 metabolic dysfunction which depended whether they originated from the male or female lineage 362 or both. 363 Our study offers some insights into the potential mechanism of transmission of breast cancer risk 364 from one generation to another. Given the increased mammary tumorigenesis in the 365 granddaughters of OID males in the absence of any further exposure, transmission of this 366 phenotype conceivably occurs via F1 germ cells, which give rise to the F2 generation. In support 367 of that, we found that F1 male germline showed alteration in tRFs, a class of small non-coding 368 RNAs abundant in sperm, recently shown to transmit environmentally-induced information from 369 one generation to another [7,8]. While details on the functional role of tRFs in embryonic development are still under investigation, these small RNAs have been implicated in the 371 regulation of translation, stress granule formation, viral replication and retrotransposons [44,45]. 372 Unfortunately, the inherent technical challenge of collecting enough eggs for molecular analysis 373 precluded us from evaluating the F1 female germline. However, given that both the F1 male and 374 female OID germline were able to transmit the increased predisposition to breast cancer 375 phenotype to a second generation it is likely that we would have observed changes in the female 376 germline as well. Nevertheless, we cannot rule out that some of the effects observed in F2 377 generation are due to maternal metabolic dysfunction in pregnancies of F1 OID females. 378 Interestingly, we found overlap in tRFs altered in sperm of F1 and F0 males. This suggests either 379 that the F1 male germline is programmed by paternal obesity or that sperm non-coding RNAs are 380 re-set in the F1 generation. Although, no changes in body weight were detected in F1 OID males, 381 they did show metabolic dysfunction (impaired insulin sensitivity) later in life. However, others 382 have shown that changes in the germline of male offspring of obese fathers occur in the absence 383 of overt metabolic dysfunction [19], suggesting that F2 generation phenotypes represent true 384 epigenetic inheritance. 385 The mechanisms for how germline epigenetic programming lead to phenotypes in offspring are 386 still being investigated. However, given the short half-life of sperm small non-coding RNAs such 387 as tRFs, is likely that they act early in embryonic development, setting a cascade of molecular 388 events which biases cellular programming during subsequent divisions and culminate in disease 389 phenotypes [3,6]. Our gene ontology analysis of targets of the five overlapping tRFs in OID F0 390 and F1 OID males' sperm showed an enrichment for functions related to DNA binding, 391 transcription factor activity, transcriptional regulation, and transmembrane transporters. It is 392 possible that an imbalance in the amount of those specific tRFs in sperm can disrupt embryonic 393 development post-fertilization, programming the organism to be more to be more amenable and 394 tolerant to cellular growth which would translate in increased cancer development. The exact 395 mechanisms, however, need to be further investigated in a follow-up study. 396 In conclusion, the findings described here builds on our previous works and show that 397 paternally-induced cancer development is largely due to systemic alterations in offspring and that 398 the offspring's breast cancer predisposition, as evaluated in this study, can be transmitted to a 399 subsequent generation. While our study was conducted in an animal model, it could have 400 important implications for human health. It is well known that family history is a strong 401 predictor of cancer risk [46], yet not all familial cancers can be explained by genetic mutations [1, 402 47]. Though it is estimated the up to 30% of breast cancers cluster in families, only about one 403 third of those are due to mutations in high penetrance genes such as BRCA1 and BRCA2, leaving 404 a sizable portion of familial breast cancers without a biological explanation [48]. Our study 405 suggests that ancestral history of obesity from the paternal lineage could account for some 406 familial cancers and that some organisms may be predisposed to the tolerance of cancer cells or 407 may provide adequate conditions for their growth and development. This notion is supported by 408 our prior findings showing that maternal exposure to an endocrine disruptor or dietary fat can 409 also lead to multigenerational risk of breast cancer through both the male and female germlines 410 in rats [12]. Given that the current study was performed in mice, our findings have now been 411 confirmed in two different animal species. 412 It is also important to note that conditions such as obesity and malnutrition often occur in 413 minorities and disadvantaged populations [49]. Our findings would suggest that social 414 determinants of cancer predisposition and outcomes may be imprinted even before birth and are 415 epigenetically mediated. However, it remains to be determined whether the biological insights 416    all gender (a-b), female (c-d) and male (e-f) F1 offspring (n=7-8/gender/group) from CO and OID-fed fathers. The data are expressed as mean ± SEM. Significant differences versus the control group were determined by two-way ANOVA followed by post-hoc analysis. *P≤0.05; **P≤0.01.  The data are expressed as mean ± SEM. Significance differences between groups were analyzed by repeated measures ANOVA (mammary tumor volume) and one-way ANOVA (tumor latency, proliferation index and number of apoptotic cells) followed by post-hoc analysis. "a" indicates statistically significant difference (P≤0.05) between OID(CO.T) and CO(CO.T); "b" indicates statistically significant difference (P≤0.05) between OID(CO.T) and CO(OID.T).  in CO and OID female F2 offspring (n=8/group). Tumor incidence is shown as percentage of animals with tumors. All other data are mean ± SEM. Significant difference were determined by Kaplan-Meier analysis followed by log-rank test (tumor incidence), repeated measures ANOVA (mammary tumor volume), one-way ANOVA (tumor latency, mortality and area under curve), or two-way ANOVA (ITT) followed by post-hoc analysis. "a" indicates statistically significant difference (P≤0.05) between OIDxOID and COxCO; "b" indicates statistically significant difference (P≤0.01) between OIDxOID and COxOID.     The data are expressed as mean ± SEM. Significant differences were determined by two-way ANOVA followed by post-hoc analysis. "a" indicates statistically significant difference (P≤0.05) between OIDxCO and COxCO group; "b" indicates statistically significant difference (P≤0.05) between OIDxCO and OIDxOID; "c" indicates statistically significant difference (P≤0.05) between OIDxCO and COxCO, OIDxCO and OIDxOID.

Supplementary Figures
Figure S1  F2:     a,b a,b a a,b a,b a