DNA methylation, through DNMT1, has an essential role in the development of gastrointestinal smooth muscle cells and disease

DNA methylation is a key epigenetic modification that can regulate gene expression. Genomic DNA hypomethylation is commonly found in many gastrointestinal (GI) diseases. Dysregulated gene expression in GI smooth muscle cells (GI-SMCs) can lead to motility disorders. However, the consequences of genomic DNA hypomethylation within GI-SMCs are still elusive. Utilizing a Cre-lox murine model, we have generated SMC-restricted DNA methyltransferase 1 (Dnmt1) knockout (KO) mice and analyzed the effects of Dnmt1 deficiency. Dnmt1-KO pups are born smaller than their wild-type littermates, have shortened GI tracts, and lose peristaltic movement due to loss of the tunica muscularis in their intestine, causing massive intestinal dilation, and death around postnatal day 21. Within smooth muscle tissue, significant CpG hypomethylation occurs across the genome at promoters, introns, and exons. Additionally, there is a marked loss of differentiated SMC markers (Srf, Myh11, miR-133, miR-143/145), an increase in pro-apoptotic markers (Nr4a1, Gadd45g), loss of cellular connectivity, and an accumulation of coated vesicles within SMC. Interestingly, we observed consistent abnormal expression patterns of enzymes involved in DNA methylation between both Dnmt1-KO mice and diseased human GI tissue. These data demonstrate that DNA hypomethylation in embryonic SMC, via congenital Dnmt1 deficiency, contributes to massive dysregulation of gene expression and is lethal to GI-SMC. These results suggest that Dnmt1 has a necessary role in the embryonic, primary development process of SMC with consistent patterns being found in human GI diseased tissue.


Introduction
The aberrant growth, or loss, of smooth muscle cells (SMCs) is a common symptom found in burdensome gastrointestinal (GI) diseases such as chronic intestinal pseudo-obstruction 1 and megacystismicrocolon-intestinal hypoperistalsis syndrome 2 as well as in vascular SMC (vSMC) found in atherosclerotic lesions 3 . Recent research has shown that inducing genomic DNA hypomethylation can change SMC phenotypic identity, growth patterns, and expression levels of contractile genes 4,5 , giving preliminary evidence that altering DNA methylation can shift the phenotypic identity of SMCs. Global changes in DNA methylation patterns are characteristic of digestive pathologies such as colorectal cancer 6 and Crohn's disease 7 .
The process of methylating DNA at cytosine residues, specifically at CpG sites, is a dynamic and reversible epigenetic modification 8 . In the mammalian genome, CpG sites can be found clustered together in regions entitled CpG islands that are often found in close proximity to, or within, transcription start sites (TSS) as well as the promoters of housekeeping and/or tissue-specific genes 9 . The addition of a methyl group to cytosine at its fifth carbon, creating 5-methylcytosine (5-mc), by DNA methyltransferases (DNMT1, DNMT3A, DNMT3B) can cause repression of associated genes as this mark, at and around the TSS, has been linked to transcriptionally repressed chromatin that denies access for transcription factors, such as serum response factor (SRF) 10,11 . Thus, as SRF is a master regulator of contractile genes in SMC 12 , the transcription of many contractile smooth muscle genes (Myh11, Acta2) 13,14 and microRNA (miRNAs; mir-133, mir-143/ 145) 15,16 is dependent on the binding of SRF to CArG (CC [A/T] 6 GG) boxes at these loci 17,18 . CArG boxes are highly conserved, have high CpG content in the region bound by SRF, are commonly found within promoter regions/close to TSS, and are recognized specifically by SRF 17 . Additionally, high CpG content sites, such as those around CArG boxes, are also well-established targets for cytosine methylation by DNA methyltransferases 19,20 . While several studies have given valuable insight into the role of DNA methylation in SMC, many of them similarly use cytidine analogs in vitro 4,5 , which are not selective in their inhibition of any specific DNMT isoform, nor do they present any efficacy of these techniques in vivo in GI-SMC.
To further this area of research, we have selectively eliminated the methyltransferase, Dnmt1, from the murine SMC genome. Dnmt1 is known as the maintenance methyltransferase as its main enzymatic function is to faithfully restore methylation marks on newly formed daughter strands of DNA 21 . Deficiency of Dnmt1 caused loss of the intestinal tunica muscularis and halted peristaltic contractions which induced intestinal dilation and death around P21. CpG methylation is significantly reduced genome-wide at promoters, exons, and introns. We also found that markers of mature SMC are lost, pro-apoptotic gene expression is increased, and Dnmt1 knockout (KO) SMCs cytoplasmically accumulate a large number of coated vesicles. Furthermore, we noted consistent patterns of dysregulation of DNMT1 and TET3, an enzyme that initiates the demethylation of DNA 22 , between KO mice and variously diseased human GI tissue, including colon/anal cancers and Crohn's disease. To our knowledge, this is the first in vivo evidence showing the requirement of a DNA methyltransferase, and the importance of DNA methylation, in the primary development and differentiation of GI-SMC.

Results
Generation, and confirmation, of deletion in SMCrestricted Dnmt1 knockout mice As global Dnmt1 -/mice are embryonic lethal 23 , it was necessary to generate a SMC-restricted Dnmt1 -/-KO line.
By crossing smMHC Cre-eGFP/+ , a mouse line that selectively expresses Cre recombinase and eGFP under the control of a Myh11 promoter 24 beginning at embryonic day 12.5, with a Dnmt1 lox/lox mouse line 25 , we were able to successfully generate congenital smDnmt1 -/-;Cre-eGFP/+ (Dnmt1-KO) mice. In smMHC Cre-eGFP/+ (Dnmt1-WT) mice, cells that express Cre recombinase also express eGFP, thus making it possible to identify Cre expression through enhanced green fluorescent protein (eGFP) fluorescence. Knockout was confirmed through genotyping with primers specific for Cre, eGFP, known loxP sites, and deletion of floxed cassette (Supplementary Figure 1). Only those mice that were positive for Cre, both genomic loxP sites, and the deletion of the floxed cassette (Dnmt1-KO) presented the following symptoms: Dnmt1-KO mice were born similar in size and body mass to their wild-type (WT) littermates, but they have significantly less body mass by postnatal day 15 (P15) continuing through to P21 (Fig. 1a, b) when death occurs. Gross anatomical images show intestinal dilation beginning at P10 that continues to progress until P21 (Fig. 1c) when dilation is severe enough to cause intestinal ischemia or perforation of the deteriorated intestinal wall, causing septic shock. Dnmt1-KO mice also have significantly shortened GI tracts in both the small intestine (Fig. 1d) and colon (Fig. 1e) by P21.
Dynamic changes of enzymes involved in DNA methylation in Dnmt1-KO mice and diseased human tissue DNA methylation is positively regulated by DNMT1/ DNMT3A/DNMT3B and is negatively regulated by the Ten-Eleven Translocation (TET) family of proteins (TET1/TET2/TET3). In the tunica muscularis of Dnmt1-WT mice, expression levels of DNMT1 and DNMT3A are developmentally dynamic while DNMT3B is constantly expressed at all time points (Fig. 2a). DNMT1 is highly expressed at embryonic day 18 (E18) and gradually decreases from birth to P20, with the opposite pattern being seen with DNMT3A. Intriguingly, temporal expression levels of TET1/TET2/TET3 are reduced in Dnmt1-KO with the reductions of TET2 and TET3 being significant by P15. DNMT1 levels were drastically reduced in Dnmt1-KO after P1 and the protein was not detected at P20 (Fig. 2a). As previously mentioned, in Dnmt1-KO smooth muscle tissue, TET2 and TET3 levels were significantly reduced (Fig. 2d), suggesting that expression of these TET isoforms may be dependent on DNMT1. Furthermore, as Dnmt1-KO smooth muscle tissue developed, potentially compensatory DNMT3A induction occurred (Fig. 2a). Also, as DNMT1 begins to be reduced in Dnmt1-KO, so are markers of SMC maturity and contractibility, including SRF, MYH11, and SM22α (Fig. 2b, d). Since MYH11 and SM22α (Tagln) are SRF targets 12 , the reduction of SRF in Dnmt1-KO could result in, or exacerbate, the reduced expression of these SRF Fig. 1 SMC-restricted congenital Dnmt1-KO mice are smaller in size/mass with shorter GI tracts that become increasingly dilated over time. a SMC-restricted congenital Dnmt1-KO mice at P20 are smaller when compared to Dnmt1-WT littermates. b No significant differences in body mass were seen between Dnmt1-KO mice and Dnmt1-WT mice until P15 when WT mice continued to gain mass and Dnmt1-KO began to lose mass. Dnmt1-KO slowly began to regain mass by P18 continuing into P21, most likely due to lack of peristaltic movement in the intestine causing buildup of food in the small intestine. c By P10, dilation can be seen in the small intestine of Dnmt1-KO mice with dilation increasing in size until P21 when death occurs. d, e By P21, Dnmt1-KO mice have significantly shortened small intestines and colons. For (b, d, e) n = 10, error bars are SD; *p < 0.005; **p < 0.001, unpaired t-test target genes and their subsequent protein products. Due to aberrant methylation patterns being identified in colorectal cancer 26 , we examined if DNMT1 is differentially expressed in the inflamed tunica muscularis from human patients with either Crohn's disease, colon/anal cancer, or diverticulitis. We detected increases in DNMT1 as well as increases in TET3 in the inflamed smooth muscle showing that DNMT1 can regulate TET3 across species (Fig. 2c). Quantitation of protein showed that the increases in DNMT1 and TET3 were significant in the inflamed smooth muscle from anal or colon cancer samples (Fig. 2e). Intriguingly, the correlated expression of DNMT1 and TET3 in the inflamed human tissue is consistent with the smooth muscle of Dnmt1-WT and Dnmt1-KO mice (Fig. 2a), suggesting a role for DNMT1 in regulating the expression of TET3.
Dnmt1-KO mice gradually lose tunica muscularis and necessary contractile proteins and accumulate coated vesicles in SMCs Hematoxylin and eosin staining revealed that Dnmt1-KO mice begin to lose both muscle layers of the Fig. 2 Enzymes regulating DNA methylation and mature SMC markers are dynamically altered in the jejunal tunica muscularis of Dnmt1-KO mice and inflamed human colonic tissue. a Western blotting of isolated jejunal tunica muscularis from Dnmt1-WT and Dnmt1-KO mice show temporal changes in expression of enzymes regulating DNA methylation (DNMT1, 3A and 3B) and demethylation (TET1-3). Note a progressive loss of DNMT1, and a massive reduction in expression of TET2 and TET3 in Dnmt1-KO mice. b A loss of DNMT1 correlates with significant reductions in mature SMC markers (SRF, MYH11, and SM22α) that become more pronounced with age indicating that GI-SMCs in Dnmt1-KO mice are indeed losing their mature, contractile status. c Human colonic tunica muscularis shows increased expression of DNMT1 in inflamed tissues along with TET3. Diverticulitis-M (marginal area); Diverticulitis-P (pouch). d, e Quantification of western blotting from Dnmt1-KO and Dnmt1-WT and human healthy and inflamed colonic tissue; n = 2-3 for mouse and human separately, error bars are SEM, *p < 0.005; **p < 0.001, unpaired t-test tunica muscularis at noticeable levels by P15 but did not have a significant loss of either layer until P21, with the more drastic tissue loss being found in the circular layer (Supplementary Figure 2). To confirm that the loss of tissue was indeed due to a loss of mature SMCs, immunohistochemical analysis was performed. As MYH11 is the most selective marker of mature and contractile SMCs 27 , antibodies targeting MYH11 were used to confirm the loss of mature SMCs alongside the endogenous eGFP reporter expression based on Myh11 expression. As expected, MYH11 levels in the tunica muscularis of Dnmt1-WT mice continued to increase from P10 to P21, while in Dnmt1-KO MYH11 levels waned, starting at P15 until there was no detectable MYH11 signal at P21, indicating a complete loss of mature SMCs (Supplementary Figure 2D). The loss of MYH11 in the SMCs of Dnmt1-KO mice results in SMCs that are no longer contractile, leading to a weakened tunica muscularis, allowing for intestinal distention. Electron microscopy of the tunica muscularis from P21 Dnmt1-KO mice revealed a loss of intercellular connections between SMCs, submucosa cells, and interstitial cells of Cajal (ICC), the necessary pacemaking cells in peristaltic contractions 28 (Fig. 3c). We also noted cellular fragmentation and the loss of nuclear/nucleolar organization within Dnmt1-KO SMCs (Fig. 3c, d). Furthermore, we were able to detect macrophages engulfing the deteriorated SMCs in Dnmt1-KO mice (Fig. 3e). Finally, we found that almost all Dnmt1-KO SMCs contained small coated vesicles (Fig. 3d, f), which were not associated with Dnmt1-WT SMC (Fig. 3a, b). Dnmt1-KO SMC appeared to accrue coated vesicles in their cytosol that congregated in distinct regions around the endoplasmic reticulum and Golgi (Fig. 3f). The expression profile of Dnmt1-KO jejunal tunica muscularis shows a higher number of unique mRNA transcripts and more genes that increased their expression than reduced their expression when compared to Dnmt1-WT. Changes in overall miRNA expression were not as dramatic. c Increases in several genes associated with apoptosis or lipid processing were upregulated in Dnmt1-KO mice. Additionally, transcript levels of SMC markers (Srf, Myh11) are reduced along with the SRF transcriptional co-activator, Myocd. d Although overall miRNA changes are not as pronounced as mRNA changes in Dnmt1-KO mice, there are select miRNAs that increase their expression level. e A gene ontology evaluation uncovered the prevalence of five general categories enriched in Dnmt1-KO jejunal tunica muscularis with many genes relating to defense, immunity, lipids, motility, and apoptosis being found Transcriptomic analysis of tissue/SMC reveals loss of mature SMC markers and miRNAs and increases in lipid, immune processing, and pro-apoptotic genes To gather a more complete understanding of the gene expression changes happening in Dnmt1-KO mice, we performed a transcriptomic analysis on the jejunal tunica muscularis of both Dnmt1-KO and Dnmt1-WT mice. Summaries and full expression profiles of messenger RNA (mRNA)/miRNA sequencing results can be found in Supplementary Table 1 and Supplementary Table 3, respectively. The transcriptome of Dnmt1-KO has 954 unique mRNA gene annotations, compared to only 530 in red stain) that is not seen in Dnmt1-WT mice. b At P21, a clear fragmentation of extracted DNA can be seen in the isolated tunica muscularis of Dnmt1-KO mice. Distinct fragmentation patterns of genomic DNA are generated due to cleavage at specific sites, a characteristic of apoptotic cells. c qPCR of isolated tunica muscularis at P15. As expected, Dnmt1, Srf, and Myh11 all have significantly reduced expression in Dnmt1-KO tunica muscularis as well as significant increases in many genes linked with lipid processes and/or pro-apoptotic pathways (n = 4, error bars are SEM, *p < 0.05, **p < 0.01, unpaired t-test). d In isolated Dnmt1-KO SMC at P15, more dramatic losses of Dnmt1, Srf, and Myh11 were observed as well as increases in the pro-apoptotic markers Gadd45g & Nr4a1 (n = 2, error bars are SEM, *p < 0.05, **p < 0.01, unpaired t-test). Ubb was used as an endogenous control in (c, d) Dnmt1-WT with approximately equal amounts of unique miRNA annotations between Dnmt1-KO (184) and -WT (170) mice (Fig. 4a). When compared to Dnmt1-WT mice, Dnmt1-KO mice have 2,829 genes that at least double their expression and only 1,319 that lose at least half their expression, with similar amounts of both upregulated (289) and downregulated (265) miRNA transcripts (Fig. 4b). The differences between the relative amount and type of mRNA/miRNA transcripts expressed exposes a remodeled transcriptome in Dnmt1-KO mice. b CpG methylation is reduced~20% in Dnmt1-KO mice when compared to Dnmt1-WT mice. c While there is an~20% reduction of CpG methylation in Dnmt1-KO mice, the patterns of methylation across each chromosome is not altered. d Globally, across the genome, CpG methylation is significantly reduced at promoters, exons and introns (**p < 0.00001, Fisher's exact test). e Pro-apoptotic marker Nr4a1 shows significant losses of methylation at various genomic locations, including the promoter and exon 1 (*p < 0.01, Fisher's exact test). f A UCSC genome browser view of the Nr4a1 gene with methylation tracks added (yellow=more methylated, red=less methylated) and the promoter region boxed in purple Dnmt1-KO mice have considerable losses of SMC marker transcripts necessary for mature functioning, including Acta2 (25.7% loss), Myocd (37.4% loss), Myh11 (53.2% loss), Srf(41.7% loss), and Tagln (19.9% loss) (Fig. 4c) as well as reductions in necessary SMC miRNAs miR-143 (31.1% loss), miR-145 (40.8% loss), and miR-133 (70.4% loss) (Fig. 4d). Since the transcription factors SRF and MYOCD are required for Myh11 expression 29 , and the previously mentioned miRNAs 30 , the reduction of Srf and Myocd exacerbates the deterioration of SMCs as Srf expression in Dnmt1-KO does not become significantly reduced until P15 (Supplementary Figure 3) and therefore is likely not the primary insult involved in the loss of GI-SMCs. We also noted increases in miRNAs associated with regulating cellular identity and proliferation including, miR-10b 31 , miR-21a 32 , miR-486a 33 , and miR-148a. MiR-148a is an established inhibitor of Dnmt1 expression 34 , thus Dnmt1-KO SMCs, through increased miR-148a, likely create even further repression of Dnmt1 expression. Furthermore, both miR-10b and miR-148a expression are known to regulated by genomic CpG methylation 31,35 . For mRNA analysis, we selected 392 overexpressed genes in the tunica muscularis of Dnmt1-KO mice and 365 of them had some level of significant demethylation (Supplementary Figure 4). Gene ontology (GO) examination revealed that the most overrepresented biological process categories included those associated with lipids, apoptosis, defense responses to stress/stimuli, and immune responses ( Fig. 4e and Supplementary Table 4). The quantitative increase in expression levels of many of these genes that were found to be upregulated in Dnmt1-KO varied widely (Fig. 4c).
We selected upregulated gene groups associated with lipids and apoptosis for further study. Oil red O staining of cross-sections from Dnmt1-KO mice confirmed an accumulation of lipids in their jejunal tunica muscularis (Fig. 5a), consistent with our mRNA-sequencing and GO term analysis. Next, we examined if Dnmt1-KO SMCs were undergoing apoptosis. DNA fragmentation is a hallmark of apoptosis 36 and was observed in Dnmt1-KO tunica muscularis at P21 (Fig. 5b). Our transcriptome represents the entirety of the tunica muscularis where the main population of cells are SMCs. However, immune cell infiltration into the tunica muscularis may be responsible for the increases in genes associated with immunity and other GO categories. With the use of flow cytometry, we removed cells from the tunica muscularis that were CD45 + , a known marker of immune cells 37 , then selected for cells that were CD45 − and eGFP + in P15 Dnmt1-KO and Dnmt1-WT. Indeed, the results of our flow cytometry identified~5-fold increase in CD45 + cells within Dnmt1-KO tissue when compared to Dnmt1-WT mice while CD45and eGFP + SMC were reduced~3-fold (Supplementary Figure 5). Sorted CD45 -, eGFP + cells represented isolated SMCs. We confirmed that isolated SMCs from Dnmt1-KO mice had almost complete ablation of Dnmt1, significant reductions in Myh11 and Srf expression, and dramatic increases in the pro-apoptotic genes Gadd45g and Nr4a1 (Fig. 5d). However, we found other genes related to apoptosis and lipids upregulated in P15 tissue that were not upregulated in isolated Dnmt1-KO SMCs, indicating that their upregulation is likely due to immune cells infiltrating into the tunica muscularis (Fig. 5c, d). While CHH sites and CHG sites did have individual sites that changed their methylation status in Dnmt1-KO mice, there was no strong genome-wide change in methylation status. c At CpG sites in Dnmt1-KO mice, there was marked hypomethylation at multiple site across the genome which was not observed in either CHH or CHG sites. Each dot represents an individual cytosine that mapped to both Dnmt1-KO and Dnmt1-WT samples; r is Pearson's correlation coefficient and n is the total number of overlapping sites for each type of cytosine site

Loss of Dnmt1 induces global genomic CpG hypomethylation
Employing reduced representation bisulfite conversion sequencing, we discovered that~90% (20,808) of annotated genes in Dnmt1-KO mice had at least one CpG site with significant demethylation in either their promoter (−1 kb of TSS), exons, or introns and~17% (4,004) of genes had at least one CpG site with significant demethylation in all sites, encompassing their promoter, exons, and introns ( Fig. 6a) with genomic hypomethylation being significant, on a genome-wide scale, across annotated promoter, exon, and intron CpG sites (Fig. 6d). None of these patterns were observed for genomic CHG or CHH (H can be A, C, or T) sites (Fig. 7a, b). Across the genome of Dnmt1-KO mice, there was a~20% loss of CpG methylation (Fig. 6c) with a majority of genes reducing their average CpG methylation levels from 0.8-0.6 to 0.6-0.4 (Fig. 6b). We also observed CpG sites that were strongly hypermethylated in Dnmt1-KO mice, but these sites were considerably outnumbered by the amount of strongly hypomethylated CpG sites (Fig. 7c). We created browser tracks that visualize and quantify the changes in CpG methylation within the University of California, Santa Cruz (UCSC) genome browser 38 for any annotated genomic site. In addition, we added transcriptome level RNA-sequencing (RNA-seq) data for both mRNA and miRNA from jejunal smooth muscle in both Dnmt1-KO and Dnmt1-WT mice (Supplementary Table 3) to the browser. The browser tracks were installed at our Smooth Muscle Transcriptome Project webpage: https://med.unr.edu/physio/transcriptome, entitled "UCSC Smooth Muscle Methylome Browser." This browser provides a comprehensive reference for transcriptomic and genomic DNA methylation status at CpG sites in the jejunal smooth muscle of Dnmt1-WT and Dnmt1-KO mice. The browser can also interact with genome-level bioinformatics (e.g., ENCODE) data publically available in the UCSC Genome Browser. In Dnmt1-KO mice, Nr4a1 has significant losses of methylation at its promoter as well as exons 1, 2, 5, and 6 and introns 5 and 6 ( Fig. 6e), as shown on the browser (Fig. 6f), suggesting the expression of the pro-apoptotic Nr4a1 could be regulated by CpG methylation levels controlled by DNMT1, agreeing with previous findings 39 . The entirety of our CpG methylomic results and analysis for every annotated promoter, exon, and intron can be found in Supplementary  Table 5.

Discussion
By eliminating Dnmt1 from the murine genome in a SMC-restricted model, we showed the in vivo necessity of Dnmt1 in the embryonic development of GI-SMCs. SMC-restricted congenital Dnmt1-KO mice have shortened GI tracts, and lose their tunica muscularis by P21. Dnmt1-KO GI-SMC have significant genomic CpG demethylation, a loss of intercellular connectivity, decreases in necessary SMC transcripts (Myh11, Srf, miR-133, miR-143/145), and increases in both pro-apoptotic genes (Gadd45g, Nr4a1) and miRNAs associated with changes in cellular identity (miR-21a, -148a, -186a, -10b). In fact, we found aberrant hypermethylation flanking a CArG box (CCATATAAGG: antisense) and SRF binding site found within the second intron of Srf in Dnmt1-KO (Supplementary Figure 6) that may have contributed to the reduction Srf expression, thus increasing the apoptotic potential of Dnmt1-KO SMC. Our results are similar to both congenital and inducible intestinal epithelial cell-specific Dnmt1-KO models. Both models show genomic hypomethylation, DNA damage, and increased apoptosis in affected Dnmt1 knockout cells 40,41 . In the congenital intestinal epithelial model, only 35% of Dnmt1 knockout mice live until P20, similar to our SMC-restricted Dnmt1-KO mice which die around P21 and none have ever survived past P24. When Dnmt1 is inducibly ablated from intestinal epithelial cells at later stages of development, the mouse does not die and sees an increase in Dnmt3b expression, which also directly compensated for the loss of DNA methylation due to the knockout of Dnmt1. This pattern of compensatory DNMT isoform induction due to loss of a separate DNMT could potentially be why we see increased DNMT3A seen during development in our Dnmt1-KO mice, as DNMT3A has been shown to be able to maintain methylation patterns in mouse embryonic stem cells lacking DNMT1 42 .
While it is clear that GI-SMCs in congenital Dnmt1-KO mice are undergoing remodeling and apoptosis, it is not clear what the causative insult of these changes is, and it is more likely a combination of several pro-apoptotic influences including genomic CpG hypomethylation, epigenome modulation, lack of cellular connectivity, upregulation of pro-apoptotic genes/miRNAs, downregulation of required SMC proteins/miRNAs, exposure to lipoproteins, and chromosomal instability. As it pertains to pro-apoptotic genes upregulated in Dnmt1-KO SMCs, Gadd45g is a marker of DNA damage, activator of the p38/c-Jun N-terminal kinase apoptotic pathway 43 and has been correlated with increased expression in a double knockout of DNMT1 and DNMT3B in vitro 44 . Nr4a1 is an orphan nuclear receptor that has been shown to be regulated by DNMT1 and the methylation level of its promoter region 39 . Expression of Nr4a1 is associated with an induction of apoptosis through multiple pathways 45 , as well as direct inhibition of the growth of SMCs 46 . As for genomic CpG hypomethylation, it is likely a causative feature involved in the transcriptomic remodeling of mRNA and miRNA expression, but probably does not account for all changes. Only about 10% of all miRNAs have been shown to change expression based on levels of DNA methylation 47 , denoting a secondary role of DNMT1 in regulating miRNA expression in GI-SMCs. DNMT1 interacts with up to 58 transcription factors 48 , and many epigenetic modifiers 49 , and thus it is probable that loss of Dnmt1 has downstream deleterious consequences outside of genomic hypomethylation. Furthermore, in Dnmt1-null cells, DNA mismatch repair is impeded and microsatellite instability increases~4-fold 50 , indicating a pivotal, methylation-independent role of DNMT1 in DNA damage repair. While previous in vitro studies of complete Dnmt1 knockout in mouse embryonic stem cells have reported more dramatic genomic CpG methylation reductions from~75% to~18% 51 or lower at specific sites 52 , cell-selective in vivo knockout of Dnmt1 in intestinal epithelial results in CpG methylation reductions from~75% to~45% 40,41 , similar to our own in vivo findings (Fig. 6c).
Additionally, the observed lack of nuclear/nucleolar organization in electron microscopy images of Dnmt1-KO could be indicative of catastrophic chromosomal instability. Of the 25 genes in the CIN25 gene signature of chromosomal instability 53 , 23 were found to be upregulated in Dnmt1-KO smooth muscle tissue (data not shown). We were also able to show that congenital Dnmt1-KO mice have reduced levels of all TET proteins, and thus it is likely that Dnmt1-KO mice cannot properly regulate 5-hmc or 5-mc levels, causing dysregulation of the entire DNA methylation apparatus. We also showed that expression levels of DNMT1 and TET3 are associated and correlated within the inflamed tunica muscularis of colorectal cancer tissue as well as in the jejunal tunica muscularis of both Dnmt1-WT and Dnmt1-KO mice. Further studies are needed to elucidate the associations and interactions between DNMT1 and TET3 in inflamed and injured smooth muscle.
Most surprisingly, Dnmt1-KO SMC show an accumulation of lipids and lipid associated transcripts within their tunica muscularis. Many of the genes upregulated in our transcriptomic analysis are commonly associated with vSMC containing atherosclerotic plaques 54 . In vSMC, lipoproteins, based on oxidation status, have been implicated in changing the phenotypic identity of SMCs into either a proliferative or apoptotic fate 55 .
In conclusion, the lack of Dnmt1 in GI-SMC resulted in an inability to maintain de novo methylation patterns and the consequences of this loss become exacerbated with each round of mitotic division. As can be surmised, further research into all observed phenotypic aspects of Dnmt1-KO is necessary to understand all cellular consequences of DNMT1 deficiency.

Mice
In a C57BL/6 background, smMHC Cre-eGFP/+ mice 24 were crossed with Dnmt1 loxllox mice 25 to make smDnmt1 -/-;Cre-eGFP/+ mice. PCR confirmed genotypes (Supplementary Figure 1). Primer sequences are found in Supplementary Table 2. Mice were bred, maintained, and euthanized following guidelines set by the Institutional Animal Care and Use Committee at the University of Nevada-Reno Animal Resources.

Human tissue
Isolated human GI smooth muscle tissue sections were donated by Dr. Kent Sasse (Sasse Surgical Associates) and Dr. Laren S. Becker (Stanford University School of Medicine). Use of human tissue was approved by the Institutional Review Board at the University of Nevada, Reno.

Tissue isolation and preparation
The GI tract was extracted into 1× Hank's Calcium Free Buffer. Tissue then underwent one of two processes: (1) smooth muscle isolation and (2) whole tissue preparation for cross-sectioning. (1) The muscle layer was peeled from the mucosa for use in extractions/experiments. (2) The tissue was fixed in 4% paraformaldehyde (PFA), a dehydration in 20% sucrose/1× phosphatebuffered saline (PBS), then placed in 1:1 OCT/20% sucrose and super-cooled by liquid nitrogen. Using a cryostat microtome (Leica, Wetzlar, Germany), 8 µM thick tissue sections were cut onto slides coated with VectaBond (Maravai Biosciences, San Diego, CA, USA), and allowed to dry. Prepared slides were used for downstream immunohistochemical/histological staining.

Isolation of total and small RNAs
Isolated muscle tissue was placed in lysis buffer and homogenized by bead beating in an air-cooled Bullet Blender Storm (Next Advance, Troy, NY, USA). Total RNAs and miRNAs were isolated from smooth muscle and sorted SMCs using mirVana miRNA Isolation Kit (Ambion, Foster City, CA, USA) as previously described 56 . Extracted total RNAs and small RNAs were used for quantitative PCR (qPCR) and RNA-seq.

Quantitative PCR
Extracted total RNA was reverse transcribed into complementary DNA (cDNA) using SuperScript III Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA), after DNA removal with DNA-free DNA Removal Kit (Ambion). cDNA samples were then quantified, diluted, and tested for quality. Standard qPCR protocol was carried out on a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) as previously described 56 . Primer sequences are found in Supplementary Table 2.
Immunohistochemistry Slides were exposed to 1× Tris-buffered saline (TBS)/ 0.1% Tween 20 followed by exposure to 4% milk/1× TBS/ 0.1% Tween 20. Slides were then incubated at 4°C with 1°a ntibody. The following day, slides were incubated with 2°a ntibody at room temperature. After washing the slides in 1× TBS, the slides were mounted and allowed to cure. See Supplementary Table 2 for antibodies/dilutions. Slides were imaged with an Olympus FluoView FV1000 (Tokyo, Japan) confocal microscope.

Hematoxylin and eosin staining
Cross-section slides were exposed to hematoxylin, rinsed with water, and then stained in eosin Y. After several dehydrating baths of increasing ethanol (EtOH) concentration (80%, 95%, 100%) and two terminal xylene baths, the slides were mounted and allowed to cure. Slides were imaged with an Olympus BX43 (Tokyo, Japan) brightfield microscope.

Oil Red O staining
Cross-section slides were exposed to pure propylene glycol, then moved to 0.5% Oil Red O solution at 60°C followed by soaking in 85% propylene glycol. slides were then rinsed and counterstained with hematoxylin, rinsed with water, then mounted with aqueous mounting reagent and allowed to cure. Slides were imaged with a Keyence BZ-x710 (Osaka, Japan) brightfield microscope.

Electron microscopy
Whole jejunum and colon tissue were fixed (3% glutaraldehyde and 4% PFA in 0.1 M phosphate buffer) for several days at room temperature. Specimens were then fixed in 1% OsO 4 at 4°C, rinsed in distilled water, then block stained with saturated aqueous uranyl acetate solution, followed by dehydration in a graded series of EtOH and finally embedded in EPON 812. Ultrathin sections were stained with uranyl acetate and lead citrate and then examined using a Hitachi H-7650 (Tokyo, Japan) microscope.

Western blot
Isolated SM tissue was ground by mortar and pestle in modified RIPA buffer. Transferred polyvinylidene fluoride membranes were pre-blocked in 5% milk, then exposed to 1°antibody at 4°C in 2% milk/0.05% Tween 20 at varying dilutions. The membranes were exposed to 2°a ntibody at room temperature. All antibodies used can be found in Supplementary Table 2. Imaging by UVP EC3 (Upland, CA, USA) Imaging system. Quantification of western blot banding patterns was done using ImageJ and histograms were produced in SigmaPlot (Systat, San Jose, CA, USA).

DNA methylome
Genomic DNA was isolated from jejunal tunica muscularis at P15 (2 males and 2 females for both Dnmt1-KO and Dnmt1-WT; n = 4) using Qiagen (Hilden, Germany) AllPrep DNA spin columns. DNA samples were pooled and shipped to Zymo Research Corporation (Irvine, CA, USA) where the samples underwent Methyl-MidiSeq which involves genomic DNA extraction, fragmentation, end modification, bisulfite conversion, limited amplification, and sequencing. After sequencing, bioinformatic processing was utilized in comparing amount of methylation at CpG, CHH, and CHG sites between Dnmt1-WT and Dnmt1-KO samples. Methylation ratios were calculated (methylated cytosines/total cytosine reads) for individual sites and then averaged across regions (exons, introns, promoters = −1 kb of TSS). These averages were then used in Fisher's exact test to quantify significant changes (p < 0.05) in methylation. The reference genome used was GRCm38/mm10. The data from our bisulfite conversion sequencing were submitted to the Gene Expression Omnibus (GEO) repository with the following accession numbers: GSM2947783, Dnmt1-WT jejunum; GSM2947784, Dnmt1-KO jejunum.

RNA sequencing
Total RNAs and small RNAs isolated from Dnmt1-KO and Dnmt1-WT jejunal tunica muscularis at P15 (2 males and 2 females for both Dnmt1-KO and Dnmt1-WT; n = 4) were shipped to LC Sciences (Houston, TX, USA) where mRNA-seq and miRNA-seq were performed. Sequencing data analysis, alignments, annotations, and bioinformatics processing were carried out as previously described 57 . The reference genome used was GRCm38/ mm10. Our RNA-seq data for both mRNA and miRNA were submitted to the GEO repository with the following accession numbers: GSM2936348, Dnmt1-WT jejunum mRNA; GSM2936349, Dnmt1-KO jejunum mRNA; GSM2936350, Dnmt1-WT jejunum miRNA; GSM2936351 Dnmt1-KO jejunum miRNA.

Cell sorting
Cells were dispersed as previously described 58 . Cells were incubated with APC-Cy7 Anti-CD45 (Biolegend, Clone 30-F11, 1.0 μg/mL; San Diego, CA, USA) followed by washing with PBS/1% FBS. Resuspended cells had Hoechst 33258 (1 μg/mL) added as a viability marker. Cells were sorted and analyzed using the BD Biosciences (San Jose, CA, USA) FACSAria II Special Order Research Product with a 130 μm nozzle with sheath pressure at 12 psi. The 355 nm laser excited Hoechst 33258 with a 450/ 50 nm bandpass filter. The eGFP was excited using a 488 nm laser with a 530/30 nm bandpass filter. A neutral density filter 2 was used on the forward scatter detector due to the high forward scatter properties. Cells that were CD45and eGFP + were sorted into PBS/1% FBS. Acquisition was performed on BD FACSDiva 8.0 and TreeStar Flowjo (Ashland, OR, USA) was used to generate figures.