Mice deficient of Myc super-enhancer region reveal a differential control mechanism between normal and pathological growth

The gene desert upstream of the Myc oncogene on chromosome 8q24 contains susceptibility loci for several major forms of human cancer, including cancers of breast, prostate, and colon. The region shows high conservation between human and mouse and contains multiple MYC enhancers that are activated in tumor cells. However, the role of this region in normal development has not been addressed. Here we show that a 538 kb deletion of the entire MYC upstream super-enhancer region in mice results in 50 to 80% decrease in MYC expression in multiple tissues. The mice are viable and show no overt phenotype. However, they are resistant to tumorigenesis, and most normal cells isolated from them grow slowly in culture. Consistently, deletion of the 8q24 super-enhancer region perturbs Myc targets only in cultured cells, but not in vivo. These results reveal that only cells whose Myc activity is increased by serum or oncogenic driver mutations depend on the 8q24 super-enhancer region, and indicate that targeting the activity of this element is a promising strategy of cancer chemoprevention and therapy.


Introduction
Deregulated expression of the MYC oncogene is associated with many cancer types (Reviewed in (Albihn et al. 2010;Dang 2012;Evan 2012)). MYC acts primarily as a transcriptional activator that increases expression of many genes required for RNA and protein synthesis above the level that is required in resting cells. In cancer cells, aberrantly elevated levels of MYC drive global amplification of transcription rates, providing the cells with necessary resources for rapid proliferation (see, for example (Brown et al. 2008;van Riggelen et al. 2010;Ji et al. 2011;Lin et al. 2012;Sabo et al. 2014;Walz et al. 2014)).
Transcription of the MYC gene is regulated by a diverse array of regulatory elements located both upstream and downstream of the MYC transcription start site (TSS). Variants in the MYC upstream region contribute to inherited susceptibility to most major forms of human cancer, and account for a very large number of cancer cases at the population level (Amundadottir et al. 2006;Gudmundsson et al. 2007;Yeager et al. 2007;Al Olama et al. 2009;Yeager et al. 2009). For example, the polymorphism rs6983267 linked to colorectal (Tomlinson et al. 2007) and prostate (Yeager et al. 2007) cancers contributes more to cancer morbidity and mortality than any other known inherited variant or mutation, including the inherited mutations in classic tumor suppressors such as RB, TP53 and APC. Through computational and experimental analyses, we and others have shown that the risk allele G of rs6983267 creates a strong binding site for the colorectal-cancer associated transcription factor Tcf7l2 (Pomerantz et al. 2009;Tuupanen et al. 2009). This binding site is located within the Myc-335 enhancer element that is dispensable for mouse viability, but required for efficient Tcf7l2-driven intestinal tumorigenesis (Sur et al. 2012).
More recently, another enhancer element, located 1.47 Mb downstream of Myc was shown to be required for formation of acute lymphoblastic leukemia (ALL) in mice (Herranz et al. 2014). However, in contrast to the Myc-335 element, this element is also required for normal T-cell development. Thus, the mechanism by which individual Myc enhancer elements contribute to normal development and tumorigenesis is still unclear.
Several studies have shown that the 8q24 region contains a large number of additional enhancer elements (see, for example (Hallikas et al. 2006;Ahmadiyeh et al. 2010;Yan et al. 2013;Yao et al. 2014)), forming a 'super-enhancer' region that is active in many different types of human cancer (Hnisz et al. 2013;Loven et al. 2013;Zhou et al. 2015). The MYC-associated super-enhancers are activated during the process of carcinogenesis (Hnisz et al. 2013), and downregulation of super-enhancer activity leads to selective inhibition of MYC expression (Loven et al. 2013). Thus, MYC-associated super-enhancer activity is required for tumorigenesis, but the role of these elements in normal tissue morphogenesis and homeostasis has been unclear.
To address this problem, we have in this work generated multiple mouse strains deficient of regulatory elements upstream of the Myc promoter. We found that surprisingly, the entire super-enhancer region conferring multi-cancer susceptibility contributes to Myc expression in vivo, yet is not required for mouse embryonic development and viability. However, this region, is required for the growth of normal cells in culture and cancer cells in vivo, thus identifying a common mechanism of growth control between the two cellular states.

Functional mapping of the super-enhancer region upstream of Myc
To dissect functional significance of the 8q24 region during normal development, we generated series of Myc alleles in mice using homologous recombination in ES cells. These include the Myc-335 enhancer deletion allele we have described previously (Sur et al. 2012), and deletions of two additional conserved enhancer elements, , both of which are active in mouse intestine and colorectal cancer cells. In addition, we generated a point mutation that inactivates a conserved CCCTC-Binding factor (CTCF) site 2 kb upstream of the Myc TSS. This site has previously been reported to be required for MYC expression (Gombert and Krumm 2009), and to have insulator activity (Gombert et al. 2003) ( Fig. 1a). Each allele contained loxP site(s) in the same orientation to allow conditional knockouts of the enhancers, and to facilitate generation of large deletions and duplications by interallelic recombination (Wu et al. 2007). All alleles were bred to homozygosity, and resulted in generation of viable mice. Expression of Myc in the colon of Myc-196 -/and Myc-540 -/mice was not markedly altered, suggesting that these elements have little effect on regulation of Myc in the intestine under normal laboratory conditions (Fig. 1b). Myc expression level was also normal in Myc-CTCF mut/mut mouse colon despite complete loss of CTCF binding to the region proximal to the Myc promoter (Fig. 1c).

Mice lacking the Myc super-enhancer region are viable and fertile
As the individual mutations and deletions had limited effect, we next decided to generate two large deletions in the Myc locus using interallelic recombination between the Myc-CTCF mut loxP site and the loxP sites at Myc-335and Myc-540 -, 6 yielding deletions of 365 kb (GRCm38/mm10 chr15:61618287-61983375) and 538 kb (chr15:61445326-61983375), respectively (Fig. 2a). The resulting alleles, Myc Δ 2-367 and Myc Δ 2-540 , were then segregated out from the corresponding duplications, and bred to homozygosity. Given the very large regions that were deleted (Fig. 2b), we expected to see a strong phenotype. However, no overt phenotype was identified in the Myc Δ 2-367/ Δ 2-367 mice. The mice were born at the expected mendelian ratio, and both males and females were viable and fertile. Analysis of Myc expression, however, revealed a strong decrease in Myc expression in the colon and ileum of the mice (not shown).
The larger deletion, Myc Δ 2-540 , could also be bred to homozygosity, and both males and females were viable. The viability of the mice is striking, given that the region deleted contains regions linked to chronic lymphocytic leukemia, and bladder, prostate, breast, and colon cancers. To characterize the mice further, we analyzed histology and Myc expression in the tissues where these tumors originate from. This analysis revealed normal morphology of mammary gland, spleen, bladder, prostate and colon in Myc Δ 2-540/ Δ 2-540 mice (Fig. 2c).

Loss of the super-enhancer region leads to tissue-specific changes in Myc expression
Although the Myc Δ 2-540/ Δ 2-540 mice exhibited a normal phenotype, the expression of Myc was strongly decreased in colon, small intestine and prostate of these mice ( Fig. 3a and not shown). Immunohistochemical analysis of Myc expression in intestine revealed strong decrease of nuclear staining, and loss of Myc expression from the transit amplifying cell compartment. However, expression of Myc was still detected at the base of the crypt in the region where the intestinal stem cells are known to reside (Fig. 3b). These results are consistent with the role of the deleted region in tumorigenesis of colon and prostate. To analyze the effect of decreased Myc expression on the proliferation in the transit amplifying compartment, we performed immunohistochemistry (IHC) for the proliferation marker Ki-67. Both the wild-type and Myc Δ 2-540/ Δ 2-540 had similar proliferative activity in the intestinal crypts (Fig. 3b).
In contrast to colon and prostate, Myc expression was not markedly affected in the bladder, and was elevated in the spleen (Fig. 3a). To analyze the cellular composition of the spleen, we performed flow cytometric analysis of markers for hematopoietic stem cells and lymphoid lineage cells. Myc Δ 2-540/ Δ 2-540 mice had a near normal hematopoietic compartment (Fig. 2d). The only observed difference was a small reduction of B cells in the Myc Δ 2-540/ Δ 2-540 mice compared to the wild-type mice both in the spleen and the bone marrow. Although this is consistent with a role of this region in CLL development, this difference did not translate into a phenotypic consequence under normal laboratory conditions.
To compare the role of the 8q24 super-enhancer region in growth of cells in vivo and in cell culture, we isolated fibroblasts from the skin of adult Myc Δ 2-540/ Δ 2-540 and wild-type mice. Based on presence of active histone marks, and undermethylation of focal elements ( Fig. 4a and Supplementary Fig. S1), the super-enhancer region is active in fibroblasts from both humans and mice. However, the resident fibroblasts in the skin of Myc Δ 2-540/ Δ 2-540 mice appeared normal as judged by Vimentin expression (Fig. 4b). Ki-67 staining (IHC) of skin sections showed comparable proliferation levels in wild-type and Myc Δ 2-540/ Δ 2-540 mice (Fig. 4b). In contrast, most lines of fibroblasts (6 out of 7) isolated from Myc Δ 2-540/ Δ 2-540 mice grew slower in culture compared to fibroblasts isolated from wild-type mice ( Fig. 4c; p-value= 0.0256, Mann-Whitney one tailed test).

Deletion of the Myc super-enhancer region affects Myc target gene expression only in culture conditions
To understand the mechanism by which the deletion of the 8q24 superenhancer region has a differential effect on growth during normal tissue homeostasis and growth under culture conditions, we subjected both the mouse tissues and cultured cells to RNA-seq analysis. Analysis of mouse tissues confirmed the changes in Myc expression observed by qPCR (Fig. 5a). Surprisingly, despite more than 80% decrease of Myc expression in the colon, very few genes were downregulated in the tissues, and none of the significantly altered genes were known Myc targets (Supplemental Table S1). These results suggest that expression of canonical Myc target genes is not sensitive to decreases in Myc protein level during normal tissue homeostasis. In contrast to the in vivo situation, where Myc is downregulated but key target genes are not affected, in cultured Myc Δ 2-540/ Δ 2-540 fibroblasts that grew slowly in culture, the downregulation of Myc lead to a loss of expression of key target genes that drive cell growth and division. Upstream regulator analysis performed using Ingenuity Pathway Analysis revealed that the highest-ranked potential regulator for the identified gene set was Myc (Fig. 5b).
Measured by FPKM values, the cultured wild-type fibroblasts had higher Myc mRNA levels than normal tissues, whereas the cultured null fibroblasts had Myc levels that were comparable to or lower than those of normal wild-type tissues. The elevated Myc levels in cultured cells are caused by serum stimulation, as Myc mRNA levels are low in serum-starved fibroblasts and strongly induced by serum (Ref. (Dean et al. 1986) and our unpublished data). These results indicate that the 8q24 superenhancer region is dispensable for normal tissue homeostasis under conditions where Myc activity is relatively low. However, the region is required for induction of Myc activity to levels that are high enough to drive the expression of Myc target genes above their basal levels during pathological growth.

The Myc super-enhancer region is required for tumorigenesis in mice
We have shown earlier that deletion of a 1.8 kb MYC-335 enhancer sequence located at the 8q24 super-enhancer region is required for intestinal tumorigenesis in mice (Sur et al. 2012). As the super-enhancer region deleted in Myc Δ 2-540/ Δ 2-540 mice carries risk also for leukemia, and prostate, breast, and bladder cancer, we tested the susceptibility of the Myc Δ 2-540/ Δ 2-540 mice to carcinogen induced bladder and mammary tumorigenesis. The Myc Δ 2-540/ Δ 2-540 mice were not resistant to N-Butyl-N(4-hydroxybutyl) nitrosamine (BBN) induced bladder tumors. Both wild-type (n=8) and Myc Δ 2-540/ Δ 2-540 (n=8) mice developed urothelial changes ranging from hyperplasia to high grade invasive urothelial carcinoma after 5 months of BBN treatment. In contrast, comparison of median tumor-free survival times of wild-type and Myc Δ 2-540/ Δ 2-540 mice exposed to mammary-tumor inducing dimethylbenz[a]anthracene/ medroxypregesterone (DMBA/MPA) regimen revealed that the Myc Δ 2-540/ Δ 2-540 mice were partially resistant to mammary tumorigenesis ( Fig.   6a). The median tumor-free survival time for the wild-type and Myc Δ 2-540/ Δ 2-540 mice was 88 and >120 days, respectively. Together with our earlier findings, these results indicate that loss of the 8q24 super-enhancer region makes mice resistant to both genetically and chemically induced tumors.
We further tested the requirement of this region for the proliferation of cancer cell lines in cultures. We found that the corresponding region (hg19: chr8:128226490-128746456) was also required for GP5d colon cancer cell growth, as indicated by progressive loss of cells bearing a CRISPR/Cas9 induced deletion of the region during co-culture with unedited cells in the population (Fig. 6b).

Discussion
The region around the MYC gene carries inherited risk towards multiple major forms of cancer. On aggregate, this region contributes more to inherited cancer than any other locus in the human genome. The risk alleles for different cancer types are To study the role of the 8q24 region more systematically, we have in this work deleted several individual elements, and also analyzed the effect of larger deletions on normal development and carcinogenesis in mice. Our analysis of mice lacking a 538 kb region upstream of the Myc gene suggests that enhancer elements within this region cooperatively enhance Myc expression. Deletion of individual enhancers in this region has only a weak (Sur et al. 2012) or no effect on Myc expression in the mouse intestine in contrast to the deletion of the entire super-enhancer region, which leads to severe decrease in Myc expression in multiple tissues.
Myc deficient mouse embryos die due to placental defect at E9.5. When Myc is deleted only in the epiblast, the embryos grow normally and survive until E11.5, when they die due to defects in hematopoiesis (Dubois et al. 2008 GWAS has a high power to identify common variants, and most variants that are common have only a limited effect on physiological functions. This is because a variant that has strong positive or negative effect is rapidly fixed or lost, respectively.
Thus, GWAS are specifically biased to find variants that have a relatively large effect on disease, but a small effect on fitness.
Most genes in mammals do not have haploinsufficient phenotypes. Buffering could be due to mechanisms that maintain constant expression level irrespective of gene dose. However, a simpler buffering mechanism involves either expressing a gene at a very low level where it has no effect, or at a high level where it can contribute its functions even if its expression level is decreased due to transcriptional noise or loss-of-function of one allele. A similar two state mechanism where physiological TF activity levels in the relevant cell types are either too low to drive any target genes (off state), or high enough to activate all important targets (on state) could also mechanistically explain why most heterozygous null mutations of TF genes 12 have no apparent phenotype. Our analysis of the role of Myc in normal colon is consistent with such a simple buffering model (Fig. 7). Under normal physiological conditions, the system is in the off state, and a basal level of expression of the Myc target genes is maintained by a Myc-independent mechanism. The genes are thus only sensitive to an increase in Myc levels. Consistently, an 80% decrease of Myc mRNA expression does not lead to a proliferation defect, or major changes in expression of known Myc target genes. In contrast, in tumors the system is locked to an on state, where Myc targets are driven to a maximal level by Myc, and the targets are now only sensitive to a decrease in Myc activity (Fig. 7).
The requirement of MYC in tumor cells appears absolute. In transgenic animal models, overexpression of Myc leads to deregulated proliferation and tumor development in multiple tissues (Felsher and Bishop 1999;Pelengaris et al. 1999;D'Cruz et al. 2001;Jain et al. 2002;Shachaf et al. 2004). Furthermore, inhibition of MYC almost invariably causes growth arrest of cancer cells both in culture and in vivo (Soucek et al. 2002;Soucek et al. 2004;Hart et al. 2014 (Oskarsson et al. 2006). It is however required for Ras mediated tumorigenesis and growth of fibroblasts and keratinocytes in vitro (Mateyak et al. 1997;Oskarsson et al. 2006). Taken together, these results suggest that MYC is required for pathological proliferation, but is less important and in many cases dispensable for normal homeostasis of tissues. Our results are consistent with these observations. However, prior to our study, the mechanism behind this differential activity was unclear. In Mice mosaic for the deletion and duplication were backcrossed to the C57Bl/6 mice in order to segregate the chromosomes carrying the deletion. The F1 heterozygotes were intercrossed to generate mice with homozygous large deletions. Myc-335 strain has been previously described (Sur et al. 2012). All mice used in the study were on a C57Bl/6 genetic background. Mouse experiments were conducted in accordance with the local ethical guidelines. The sequences of the different primer pairs used for genotypings are given in Supplemental Table S2.
Inguinal mammary glands were removed from 8 week old virgin females and spread on glass slides. These were fixed for 4 hours in Carnoy's fixative and subsequently stained O/N with Carmine Alum. The whole mounts were rinsed and dehydrated through increasing series of ethanol and cleared in xylene before mounting with the pertex mounting medium.
Quantitative PCR analysis qPCR was performed as described previously (Sur et al. 2012). Essentially, total RNA was isolated from whole tissue by homogenizing in RNA Bee reagent (ambios AMS Biotechnology) followed by RNA isolation using Qiagen's RNA MinElute kit according to manufacturers' protocols. 0.5-1 µg of total RNA was reverse transcribed using high capacity reverse transcription kit in a 20 µl reaction (Applied Biosystems). Quantitative PCR in triplicates was performed using the SYBR select master mix (Applied Biosystems) on the LightCycler 480 instrument (Roche).
For normalization, mouse β -actin transcripts were used as internal controls.
Following primer pairs were used for quantitative PCR analysis.

ChIP-seq
ChIP-seq was performed as described in (Sur et al. 2012;Yan et al. 2013 (Langmead and Salzberg 2012). Duplicates were removed using the Bismark deduplicate function. Extraction of methylation calls was done with Bismark methylation extractor discarding first 10 bp of both reads and reading methylation calls of overlapping parts of the paired reads from the first read (--no_overlap parameter). Genomic sites with the coverage of at least 10 reads were considered and methylation ratios smoothed with loess method across 49 bp windows.
All sequencing data will be uploaded to European Nucleotide Archive (ENA, EMBL-EBI) under accession number PRJEB11397.

Immunohistochemistry and flow cytometry
Five micron paraffin embedded tissue sections were processed for immunohistochemistry as previously described (Sur et al. 2012). Rabbit polyclonal anti-Myc

Isolation and culture of mouse primary fibroblasts
Fibroblasts were isolated from adult mice by dissecting the skin to ~ 1 mm 3 pieces, and allowing the pieces to adhere to cell culture plates, followed by addition of DMEM medium supplemented with 10% FCS and antibiotics. The fibroblasts were allowed to migrate out from the explants, after which the cells were collected by trypsinization and passaged in the same media for 1-3 passages. For growth assays, 2x10 3 cells were plated per well in 96 well plates. Cells were trypsinized and counted using hemocytometer at respective time points.

Tumor induction
Mammary tumors: Six week-old female mice were implanted s.c. with        p-values were calculated from all the regulator-targeted differential expression genes using Fisher's Exact Test.