Environmentally Induced Sperm RNAs Transmit Cancer Susceptibility to Offspring in a Mouse Model

Abstract DNA sequence accounts for the majority of disease heritability, including cancer. Yet, not all familial cancer cases can be explained by genetic factors. It is becoming clear that environmentally induced epigenetic inheritance occurs and that the progeny’s traits can be shaped by parental environmental experiences. In humans, epidemiological studies have implicated environmental toxicants, such as the pesticide DDT, in intergenerational cancer development, including breast and childhood tumors. Here, we show that the female progeny of males exposed to DDT in the pre-conception period have higher susceptibility to developing aggressive tumors in mouse models of breast cancer. Sperm of DDT-exposed males exhibited distinct patterns of small non-coding RNAs, with an increase in miRNAs and a specific surge in miRNA-10b levels. Remarkably, embryonic injection of the entire sperm RNA load of DDT-exposed males, or synthetic miRNA-10b, recapitulated the tumor phenotypes observed in DDT offspring. Mechanistically, miR-10b injection altered the transcriptional profile in early embryos with enrichment of genes associated with cell differentiation, tissue and immune system development. In adult DDT-derived progeny, transcriptional and protein analysis of mammary tumors revealed alterations in stromal and in immune system compartments. Our findings reveal a causal role for sperm RNAs in environmentally induced inheritance of cancer predisposition and, if confirmed in humans, this could help partially explain some of the “missing heritability” of breast, and other, malignancies.


Introduction 50
Family history is one of the most important risk factors for cancer. A number of germline 51 mutations are causally linked to familial cancer syndromes. Yet, not all familial cancer cases can 52 be explained by high penetrance genes mutations. Genome-wide association studies (GWAS) of 53 variants have been able to account for a fraction of familial cancer predisposition, but much of the 54 "missing" cancer heritability remains unexplained. 1 55 Accumulating evidence suggests that parental life experiences can affect the progeny's 56 predisposition to disease in an epigenetically inherited manner. 2,3 At conception, parents not 57 only contribute their genome but also transmit a molecular memory of past environmental 58 experiences to the offspring. 2,4 This environmentally-induced disease predisposition has been 59 shown to be transmitted to the offspring via epigenetic mechanisms through both the female and 60 male germ-lines. 5-7 Although most of the evidence for this mode of disease inheritance comes 61 from maternal exposures in pregnancy, a number of independent studies have shown that paternal 62 exposures in the pre-conception window are also important in determining disease outcomes in the 63 offspring. 8-14 64 Several recent published reports demonstrated that the sperm RNA load can transmit 65 environmentally-induced phenotypes from fathers to offspring. 2,7,15,16 Some of these studies 66 implicate specific classes of sperm small non-coding RNAs, such as miRNAs and tRNA fragments 67 (tRFs), which can recapitulate the effect of specific paternal exposures in offspring when injected 68 into normal embryos following fertilization. 15-19 69 Population studies link parental exposures to environmental pesticides with both childhood 70 and adult cancers in the progeny [20][21][22][23][24][25][26][27][28][29] . For instance, the epidemiologic analysis using the Child 71 Health and Development Studies cohort (CHDS) found that daughters of women exposed to high 72 4 levels of dichlorodiphenyltrichloroethane (DDT) in pregnancy are more likely to have increased 73 rates of breast cancer 20,30 and to be diagnosed with advanced stage tumors. 20 74 Though the use of DDT has been restricted in many countries for decades, developing 75 countries with endemic malaria, and other vector-borne diseases, continue to use it. 20,[31][32][33] In 76 countries where it has been banned, this pesticide is a persistent organic environmental pollutant 77 and an endocrine disruptor. 31,34 Because of its low degradation rates and lipophilic properties, 78 DDT bio-accumulates in the food chain, and can still be detected in fatty tissues and circulation of 79 animals and humans in these countries. 31 In the U.S., the highest concentrations of DDT 80 metabolites are found in minority populations and recent immigrants. 35,36 81 Despite strong association between parental exposures to environmental toxins and cancer 82 in the next generation, the precise mechanisms underlying this association remain unknown. Here, 83 using mouse models of breast cancer, we tested the hypothesis that paternal exposure to DDT alters 84 the sperm non-coding RNA load and increases cancer susceptibility in the next generation. We 85 also examined whether paternal DDT-induced programming of breast cancer development in 86 daughters is mechanistically linked to sperm non-coding RNAs and the underlying mechanisms. 87 88 Results 89

Paternal exposure to the environmental toxicant DDT leads to enhanced breast tumor growth 90 in offspring 91
Parental exposure to environmental toxicants such as DDT and other pesticides has been 92 associated to cancer development, including breast tumors, in the next generation. 20,22,28 To test 93 whether we could replicate these findings in an animal model, adult male mice were either exposed 94 to a vehicle-control (CO) or an environmentally relevant dose of DDT and, subsequently, mated 95 5 with unexposed females to generate the progeny (Fig. 1a). Using the well-established DMBA-96 induced mouse model of breast cancer, 8,37 we examined mammary cancer development in the 97 female progeny of CO and DDT exposed fathers. Tumor growth was significantly increased in 98 female offspring of DDT exposed fathers (referred to as 'DDT offspring' from this point on) 99 compared to CO (Fig. 1b), with significantly higher tumor burden and tumor weight at the end of 100 the monitoring period (Fig. 1c, Fig. S1). Although not significantly different, the incidence and 101 latency of mammary tumors was slightly higher and shorter, respectively, in DDT offspring 102 compared to CO (Fig. 1d-e). 103 To test whether the increase in carcinogen-induced mammary tumor growth observed in 104 DDT offspring could be replicated in a different in vivo system, we used a syngeneic orthotopic 105 mouse model of breast cancer. Consistent with our observations in the carcinogen-induced model, 106 we found that EO771 murine breast cancer cells implanted in mammary fat pads of DDT offspring 107 also resulted in significantly higher tumor growth compared to CO ( Fig. 1f-g). 108 109

Sperm RNAs contribute to DDT-induced paternal transmission of breast cancer predisposition 110
to offspring 111 The sperm RNA load, which is abundant in small non-coding RNAs, 38 was previously 112 thought to be a non-functional remnant of the spermatogenesis process. However, several reports 113 show that sperm RNAs can alter early embryonic development, and they have been recently linked 114 to transmission of phenotypes in a variety of disease models. 17-19,39 115 To assess whether sperm RNAs are associated with the increased cancer development in 116 DDT offspring, the RNA load extracted from mature sperm of either DDT-exposed or CO male 117 mice was injected into normal mouse embryos at the zygote stage. Injected embryos were then 118 6 transferred into surrogate dams to generate the DDT-sperm RNA and CO-sperm RNA offspring 119 (Fig.2a). Remarkably, DMBA-induced mammary tumors grew significantly larger in DDT-sperm 120 RNA female offspring than in the CO-sperm RNA group (Fig. 2b-c), recapitulating the original 121 phenotype observed in DDT offspring. Consistent with results in DDT offspring, we also observed 122 a non-significant increase in tumor incidence and shorter tumor latency in DDT-sperm RNA 123 offspring ( Fig. 2d-e). 124 Our findings were further confirmed in an orthotopic mouse model of breast cancer. 125 Mammary fat pad injection of EO771 cells resulted in increased mammary tumor growth in DDT-126 sperm RNA female offspring, which had significantly increased tumor volume over time with 127 increased tumor percentage growth at the end of the monitoring period compared to CO-sperm 128 7 We found that sperm of DDT-exposed males showed higher abundance of miRNAs reads 141 (7%) compared to the CO group (3%) (Fig. 3b). A total of 23 miRNAs were differentially 142 expressed (six down-and seventeen upregulated) in sperm of DDT-exposed mice compared to CO 143 ( Fig. 3c). Among the upregulated miRNAs, miRNA-10b had the highest surge, representing close 144 to three quarters of the differentially expressed miRNA normalized counts in sperm of DDT-145 exposed males compared to about half of the reads in the CO group (Fig. 3d-e). This is consistent 146 with a recent report showing that miRNA-10b is one the main miRNAs acquired by sperm cells 147 during epididymal transit. 40 148 To establish whether regulation of miRNAs in sperm was, in fact, induced by DDT 149 exposure, we treated DDT-exposed mice with phenobarbital (PB, Fig.4a), a drug known to 150 promote hepatic enzymes activation and accelerate DDT metabolism and excretion in humans 41,42 151 and animals. 43-45 Indeed, we observed a sharp increase in the hepatic phase I enzyme, Cyp2B10, 152 mRNA levels (Fig.4b) and a reduction in DDT main metabolites in DDT-exposed males also 153 treated with PB (Fig.4c). 154 Next, we investigated the effects of PB treatment on miRNAs that were differentially 155 expressed in sperm of DDT-exposed mice. For greater reliability, our analysis focused on the nine 156 differentially expressed miRNAs, which had an average of at least 100 normalized read counts 157 in our original RNA-seq data. 158 Compared to mice in the DDT-exposed group, miRNA-10b levels were significantly 159 reversed in sperm of mice treated with both DDT and PB (DDT+PB) with levels returning to those 160 observed in controls (CO and CO+PB) both by RNA-seq and q-PCR (Fig. 4d-e). 161 In addition to miRNA-10b, levels of miRNA-182-5p, miRNA-205-5p, and miRNA-375-162 3p were also reversed by PB treatment in sperm of DDT-exposed males in the RNA-seq analysis, 163 8 though these findings could not be confirmed by q-PCR (Fig. S3). The expression levels of the 164 five remaining miRNAs-including miR-6240, which had the sharpest decrease-were not 165 reversed by PB treatment (Fig. 4d and Fig. S3), suggesting that they may not be  Because tRFs have also been linked to intergenerational transmission of phenotypes, 15,16 167 we also investigated these small RNAs in sperm of DDT-exposed males via RNA-seq analysis. 168 We found nine tRFs differentially expressed (one upregulated and eight downregulated) in sperm 169 of DDT-exposed mice. However, these differentially expressed tRFs are likely not DDT-specific 170 as they were not reversed by PB treatment and could not be confirmed in the follow-up analysis 171 ( Fig. S4). 172 173 miRNA-10b embryonic injection recapitulates tumor phenotypes observed in offspring of DDT-174

exposed fathers 175
To further validate the functional importance of miRNA-10b to phenotypes observed in 176 DDT offspring, we injected synthetic miR-10b or controls (scramble-miR or vehicle solution) into 177 normal mouse zygotes. Injected embryos were then transferred into surrogate dams to generate the 178 miR-10b and CO offspring (Fig. 5a). The resulting miR-10b female offspring showed increased 179 orthotopic (EO771 cells) mammary tumor development, with higher tumor growth compared to 180 both control groups (vehicle injection or scrambled miR, Fig. 5b). The miR-10b females also had 181 higher tumor burden and weight compared to the scrambled group (Fig. 5c, Fig. S5), at the end of 182 the monitoring period. 183 Because small RNAs driven alterations in early embryonic development can have long 184 lasting consequences and affect adult phenotype 19 , we next examined whether embryonic injection 185 of miR-10b at the zygote stage reshaped the mouse transcriptome profile at the blastocyst stage 186 9 (E3.5), using bulk RNA-sequencing (Fig. 5d, Table S1). We first assessed whether differentially 187 expressed genes (DEG) were potentially regulated by miR-10b. In total, 63 genes down-regulated 188 in miR-10b injected embryos are predicted targets of miR-10b (Table S2). Next, we performed a 189 functional annotation analysis to identify the biological signatures most strongly associated with 190 DEGs in miR-10b injected embryos compared to CO. We found an enrichment for terms related 191 to embryonic, cell, tissue and organ development and cell differentiation as well as abnormal 192 hematopoietic and immune system development in miRNA-10b injected embryos (Fig. 5e).  (Fig. 6a, Fig. S6, Tables S3A-3C). 205 Our imaging mass cytometry (IMC) analysis confirmed alterations in EMT-associated 206 proteins with a decrease in the extracellular matrix protein collagen type I expression (Fig. 6b, 207 Family history is one of the most important risk factors for cancer. Yet, not all familial 223 cancer cases can be explained by germline genetic mutations. 1 While there is agreement that the 224 interplay between the environment and genetic factors could also contribute to cancer 225 predisposition, much of the "missing" cancer heritability remains unexplained. 226 Growing evidence suggests that environmentally induced epigenetic inheritance 227 occurs in mammals and that the progeny's traits can be shaped by parental environmental 228 experiences. 17,18,39 The findings presented here lend support to this concept: Our study showed 229 that pre-conception paternal exposure to an environmentally relevant dose of the pesticide DDT 230 modulates the sperm non-coding RNA load, particularly miRNAs, and programs female 231 offspring's breast cancer development and growth in a carcinogen-induced mouse model. These 232 11 results were substantiated in an orthotopic mouse model of breast cancer. We also showed that 233 sperm RNAs are functionally linked to the cancer phenotypes observed in offspring and 234 demonstrated that embryo injection of total sperm RNA from DDT-exposed males or synthetic 235 miRNA-10b, the miRNA with the highest increase in sperm of DDT-exposed fathers, recapitulated 236 the mammary tumor phenotypes observed in DDT daughters. Furthermore, published reports using the CHDS cohort showed that maternal exposure to DDT 287 increases rates of breast cancer in daughters who are also more likely to be diagnosed with 288 advanced-stage tumors. 20 Another recent population study using a historical cohort showed that 289 paternal grandfathers' nutrition leads to a transgenerational increase in cancer mortality rates in 290 grandsons. 57 291 In summary, our study shows that the environmentally induced alterations in paternal 292 sperm non-coding RNAs are functionally linked to increased cancer susceptibility in the progeny. 293 Though our results are intriguing and suggest that cancer predisposition could be determined via 294 epigenetic inheritance, many questions still remain: For instance, how sperm non-coding RNAs 295 and miR-10b specifically alter early embryonic development to disrupt offspring's systemic and 296 mammary specific development needs to be further examined. While epidemiologic studies 297 strongly suggest it to be the case, whether this phenomenon is restricted to breast tumors or whether 298 it applies to other malignancies needs confirmation in additional experimental models. 299 Our findings also highlight the potential effects of environmental inequities. According to 300 period, DDT-exposed and control male mice were mated to unexposed females to generate the 315 female offspring. Mice were kept on a standard chow diet during the breeding period, for the extent 316 of pregnancy (21 days) and after birth. To avoid litter-effect, pups were cross-fostered one day 317 after birth. Pups from 2-3 dams were pooled and housed in a litter of 8-10 pups per nursing dam. 318 All pups were weaned on postnatal day (PND) 21. The female offspring of control or DDT-319 exposed fathers were used to study mammary tumorigenesis and for tissue collection as described 320 in the following sections. All animal procedures were approved by the Georgetown University 321 Animal Care and Use Committee, and the experiments were performed following the National 322 Institutes of Health guidelines for the proper and humane use of animals in biomedical research. 323 All exposures were performed blindly. 324 15 DDT and Phenobarbital (PB) treatment: PB treatment is known to promote hepatic enzymes 325 activation, accelerating DDT metabolism and excretion in humans 41,42 and animals [43][44][45] . A sub-326 set of control-vehicle (CO) or DDT exposed male mice (as above) were concomitantly treated with 327 phenobarbital (20 mg/kg of body weight, i.p.) or saline and used for sperm harvesting. 328 Mature spermatozoa collection and purification: CO and DDT-exposed male mice as well as 329 those concomitantly treated with PB were euthanized and caudal epididymis dissected for sperm 330 collection. The epididymis was collected, punctured, and transferred to a tissue culture dish 331 containing M2 media (M2 Medium-with HEPES, without penicillin and streptomycin, liquid, 332 sterile-filtered, suitable for mouse embryo, SIGMA, product #M7167) where it was incubated for 333 1 hour at 37°C. Sperm samples were isolated and purified from somatic cells. Briefly, the samples 334 were washed with PBS, and then incubated with SCLB (somatic cell lysis buffer, 0.1% SDS, 0.5% 335 TX-100 in Diethylpyrocarbonate water) for 1 hour. SCLB was rinsed off with 2 washes of PBS 336 and the somatic cell-free purified spermatozoa sample pelleted and used for RNA extraction. Injected zygotes were transferred into recipient female mice as described above for term delivery. 384 Female offspring were used to study tumor growth or for tissue collection. sequenced on an Illumina platform and paired-end reads were generated. RNAseq raw data quality 420 was checked using FastQC (v0.11.09) and adapter trimming on raw data was performed using 421 Cutadapt (v3.5). Reads with low quality (quality score < 33, error rate > 10%) or reduced length 422 after trimming (<25 bp) were removed before alignment. We used the reference genome 423 downloaded from Ensembl GRCm38 release 101, and the reference index was built using Star 424 (v2.7.9a). Paired end trimmed read alignment and raw read count calculation was performed using 425 RSEM software (v1.3.1), and an additional bioinformatic workflow utilized Hisat2 v2.0.5 for 426 alignment to the GRCm38 genome and quantified to the Gencode vM25 transcriptome annotation 427 using Stringtie 2.2.1. For both workflows, differential expression analysis of two 428 conditions/groups was performed using the DESeq2 R package (v1.36.0), and visualized using the 429 Enhanced Volcano R package (v3.16 ). We considered the genes with q-value <0.1 to be 430 differentially expressed (DEG) in embryos and genes with a p-value <0.05 to be DEG in mammary 431 tumors to account for tissue heterogeneity. DEGs in mammary tumors were used as input for 432 Gene Set Enrichment Analysis (GSEA) (v4.2.3, Broad Institute). Both KEGG and HALLMARK 433 gene sets were selected for enrichment score calculation. DEGs in E3.5 embryos were used as 434 input for Gene Ontology (GO) functional characterization and pathway enrichment analysis. miR-435 10b target prediction in E3.5 embryos was determined using TargetScan and miRBase. 436 20

Imaging mass cytometry (IMC): Staining Procedures. After baking for 2 hours at 60 °C, FFPE 437
sections were dewaxed and rehydrated through a graded alcohol series. Heat-induced epitope 438 retrieval was conducted in a steamer at 96 °C in antigen retrieval AR9 buffer for 30 min. After 439 cooling, the sections were blocked with 3% BSA in PBS for 45 min at room temperature. Samples 440 were incubated overnight at 4 °C with metal-conjugated antibody cocktails diluted in PBS/0.5% 441 BSA (See Table S4 for antibody list and dilutions). Samples were then washed twice with 442 PBS/0.5% Tween and twice with PBS and exposed to 1:400 Ir-Intercalator   Tumor incidence is shown as percentage of animals with tumors; All other data are shown as mean (horizontal bars in scatter plots) or mean± SEM (tumor growth curves). Tumor burden data does not include tumors with fewer than two measurements. Tumor latency includes both measurable and non-measurable tumors. Tumor volume curves were analyzed using two-away ANOVA (group, timewith repeated measures). Differences in tumor incidence were analyzed using Kaplan-Meier survival curves followed by the log-rank test. Differences in final tumor burden (% growth) and tumor latency were analyzed using unpaired t-test. P values are displayed in each figure panel. Tumor incidence is shown as percentage of animals with tumors; All other data are shown as mean (scatter plots) or mean± SEM (tumor growth curves). Tumor burden data does not include tumors with fewer than two measurements. Tumor latency includes both measurable and non-measurable tumors. Tumor volume curves were analyzed using two-away ANOVA (group, timewith repeated measures). Differences in tumor incidence were analyzed using Kaplan-Meier survival curves followed by the log-rank test. Differences in final tumor burden (% growth) and tumor latency were analyzed using unpaired t-test. P values are displayed in each figure panel. (a) Schematic representation of the experimental design: Eight week-old male mice were treated orally with a vehicle-control (CO) or a DDT solution (1.7 mg/kg) for 17 days. RNA extracted from sperm of CO and DDT-exposed males was analyzed via RNA-seq.
(b) Small RNA subtype distribution (percentage of raw reads) and (c) heat-map showing differentially expressed miRNAs in sperm of CO and DDT-exposed males (n=9).
(d-e) Proportional abundance of differentially expressed miRNAs (DESeq2 normalized counts) in sperm of (d) CO and (e) DDT-exposed males (n=9). (b) Expression levels of Cyp2B10, assessed by q-PCR, in liver tissues of CO and DDT-exposed male mice treated with PB or vehicle injection (n=4-5).
(c) Levels of DDT metabolites in liver tissues of CO and DDT-exposed male mice treated with PB or vehicle injection (n=3 pools).
(d) miRNA expression levels (DESeq2 normalized counts), assessed by RNA-seq, in sperm of CO and DDT-exposed male mice treated with PB or vehicle injection (n=5-9).
(e) Expression levels of miRNA-10b, assessed by q-PCR, in sperm of CO and DDT-exposed male mice treated with PB or vehicle injection (n=8-16).
Data are shown as mean (horizontal bars in scatter plots) or mean± SEM (tumor growth curves). Tumor volume curves were analyzed using two-away ANOVA (group, timewith repeated measures). Differences in final tumor burden (% growth) were analyzed using one-way ANOVA. P values are displayed in each figure panel. (b) IMC staining for collagen type I (green), E-cadherin (red), DNA (blue, iridium intercalator) and quantification of collagen type I from IMC data in orthotopic (EO771 cells, n=9-12) and carcinogen-induced mammary tumors of DDT-derived offspring (n=6). Scale bar equals to 200 μm.