MED12 mutations are associated with LM burden
Mutations in the MED12 gene are prevalent in uterine LM and thus have been implicated in LM tumor development. In order to determine whether MED12 mutation rate is affected by the LM burden of the uterus, the incidence of MED12 mutations was analyzed on a large cohort of 529 tumors (Fig. 1a). Among them, 301 LM were from uteri with more than 5 LM, which we call LM-high (LM-H), 197 were from uteri with 1-4 leiomyomas, called LM-low (LM-L) and 31 were from uteri with an unknown number of LM/uterus (Fig. 1a, Table S1). We observed that the MED12 mutations were highly enriched at c.130-131 (55.0%), followed by deletions (10.6%) and other point mutations at eight other sites (9.3%; Fig. 1b). A significantly increased MED12 mutation rate was seen in LM-H compared to LM-L (79.7 % vs. 68.0%; p<0.01; Fig. 1c). Given the noticeable difference in MED12 mutations between LM-H and LM-L, we performed a comparative analysis of all mutation types including point mutations at c.130-131, other point mutations, and insertion/deletions, and observed that point mutations at c.130-131 and insertion/deletions were higher in LM-H compared to LM-L (p<0.01; Fig. 1d). When comparing the types of missense mutations at c.130-131, G>A (53.9% in LM-H, 60.6% in LM-L), followed by G>T, and G>C, no significant differences were noted between tumor number groups (Fig. 1e). MED12 mutation patterns and distributions in association with tumor number are summarized in Fig. S1a-1b. These data demonstrate that increased LM burden is associated with higher MED12 mutation rate, implicating the importance of the microenvironment that promotes MED12 mutations.
Transcriptome analysis reveals myometrium with LM has a high ROS burden and oxidative stress response
The mechanisms associated with the development of LM tumors remain unknown. The increased MED12 mutation when LM burden is high suggests that the signals from LM promote MED12 mutations. Our previous studies demonstrated that LM are under high oxidative stress related to ROS metabolic defects [9, 18]. In order to evaluate the naturally occurring ROS burden in the myometrium, we examined a subset of age-matched myometrial tissues from women with LM (MM+LM) and without LM (MM-LM). As ROS is able to oxidize deoxyguanosine in DNA, 8-OHdG was used as a biomarker for oxidative stress. Immunohistochemistry (Fig. 2a) and immunofluorescence staining (Fig. S2a) for 8-OHdG, were performed in MM+LM or MM-LM in a tissue microarray (TMA, Fig. 2a, 2b). Significantly higher immunoreactivity for 8-OHdG was observed in MM+LM compared to MM-LM (Fig. 2c; p<0.001), suggesting that MM+LM have a higher ROS burden than MM-LM. As oxidized DNA 8-OHdG accumulation has been found to overlap with DNA damage, we examined gH2AX, DNA damage in the same samples by immunofluorescence staining (Fig. S2b). Quantification of the gH2AX immunofluorescent signals was significantly higher in MM+LM compared to MM-LM (Fig. 2d).
To evaluate the gene expression differences in myometrium with and without LM, RNA sequencing analysis was performed in age-matched MM+LM and MM-LM. Expression analysis revealed a total of 3446 genes were significantly different between MM+LM and MM-LM (1596 upregulated and 1850 downregulated with p-value < 0.05 and absolute log2 fold change > 1) (Fig. 2e, Table S5). Volcano plots illustrated gene expression in fold change and level of significance and genes in ROS, DNA repair, p53, and hypoxia were highlighted and most of them were upregulated (Fig. 2f). Pathway analysis by Gene Ontology (GO) revealed cell proliferation and cell cycle, ROS pathways, DNA repair and DNA damage response pathways were strongly associated with MM+LM (Fig. 2g). Gene Set Enrichment Analysis (GSEA) revealed that MM+LM were significantly associated with ROS pathway, hypoxia, DNA repair, p53 pathway, estrogen response, and AKT signaling (Fig. 2h). Transcriptome and pathway analyses further defined defects in the uterine with LM in response to ROS stress.
Stable ROS-mediated 8-OHdG, DNA damage response, and global gene expression in myometrium can be recapitulated in vitro
To demonstrate that ROS directly promotes DNA oxidation and damage and changes genes expression, primary myometrial cells were treated with two ROS inducers, Paraquat (PQ) , and responses were measured. Following dose-response studies, myometrial cells were treated with 100 µM PQ for 24-48 hours which showed the greatest oxidative stress with minimal cell death (Fig. 3a). Fluorescent staining was performed to detect ROS (dihydroethidium [DHE]), oxidized DNA (8-OHdG), and DNA damage (gH2AX) (Fig. S3a-3b). Myometrial spheroids treated with PQ showed high ROS level (Fig. 3a) and histologic analysis showed no significant cytohistologic change or cell death on histology evaluation, however, a significant increase in immunoreactivity for 8-OHdG, and gH2AX were found in PQ-treated cells compared to untreated controls (Fig. 3b and 3c). Since apurinic/apyrimidinic (AP) sites are closely associated with 8-OHdG in genomic DNA mediated by OGG1, the primary enzyme responsible for the excision of 8-OHdG , we performed AP site analysis. As shown in Fig. 3d, primary myometrial cells treated with PQ resulted in a significantly higher number of AP sites in genomic DNA than controls. Together, these findings indicate that exposure of myometrial cells with ROS inducers causes DNA oxidation leading to 8-OHdG and DNA damage.
To evaluate the gene expression patterns in myometrial cells treated with PQ, RNA sequencing analysis was performed. After normalizing the data with an absolute log2 fold change cutoff of 1 (for Control vs. PQ), a total of 325 genes were dysregulated in cells treated with PQ (178 upregulated and 147 downregulated, Fig. 3e and 3f, Table S5). GSEA revealed the dysregulated genes are significantly associated with the ROS pathway, hypoxia, DNA repair, p53 pathway, estrogen response, and AKT signaling (Fig. 3g), consistent with findings in myometrial tissue with LM (Fig. 2g and 2h), suggesting common mechanisms of the ROS-mediated DNA damage response to these two compounds. Upregulation of AKT pathways by PQ (Fig. 2h and 3g) was consistent with our previous functional analysis . Genes in ROS stress response and the DNA repair pathway were mostly upregulated in PQ treatment (Fig. S4a). The pathway connections and gene expression trends are summarized in Fig. S4b. These findings demonstrate a common effect of PQ on ROS metabolic pathways, with DNA damage response, specifically the base excision repair pathway were dysregulated in PQ treatment.
CRISPR/Cas9 editing of c.1308-oxodG results in DNA misrepair in myometrial cells
High ROS promotes DNA nucleotide oxidization, such as modified guanine 8-OHdG . The 8-OHdG adduct, a typical product of oxidative DNA damage, can cause misrepair or mutation of guanine . To demonstrate that oxidized guanine can indeed promote misrepair and mutations, CRISPR/Cas9 was used to replace guanine at codon44 c.130G of the MED12 gene with an oxidized derivative of deoxyguanosine, 8-oxo-dG adduct (c.1308-oxodG). Technical details are summarized in Methods and Fig. 4a. c.1308-oxodG and normal control were introduced into a myometrial cell line (myo-hTERT, Fig. 4b). c.1308-oxodG mutations were detected by deep sequencing. Among a total of 15k reads, approximately 34.3% of c.1308-oxodG showed misrepair of G>T, A, C, as demonstrated by the large peak at c.130 in the myometrial cell line (Fig. 4b, Table S4). This data shows the extent of misrepair at this particular codon. In comparison, normal controls showed a baseline misrepair rate at c.130G. To accurately calculate the misrepair rates at c.1308-oxodG, an index nucleotide change at c.138index (T>C, no amino acid change) along with c.1308-oxodG from an ssDNA donor sequence was introduced into primary cultures of myometrial cells (n=3) and myo-hTERT cells. Misrepair at c.1308-oxodG was examined by a high depth deep sequencing analysis (depth of 500k to 1 million reads /sample, Fig. 4c, Fig. S5a). As shown in Fig. 4d (Table S4), the c.1308-oxodG misrepair rate in exon sequences with c.138index was 16-46%. This is consistent with previous data from Manabu et al.  who used the TATAM (tracing DNA adducts in targeted mutagenesis) system to investigate the consequent mutations of synthetic 8-oxodG introduced into the human genome. The most common misrepair of c.1308-oxodG was G>T, accounting for 96.0%-99.4% of mutations, whereas G>A represented a small fraction (0.3%-3.8%) of total misrepair and G>C accounted for 0.1% -0.4% of misrepair (Fig. 4e). Furthermore, CRISPR/Cas9 targeted replacement of 8-oxodG at codon44 c.130 was reproducible in primary myometrial cells, but the success rate of site editing (based on presence of index c.138C) varied from case to case (ranging from 0.63 to 1.32%, Fig. S5b), consistent with published data . Of note, structural changes accounted for more than 75% of sequences (Fig. 4d, Fig. S5b). Increased point mutations at c.123-126 AAAA might relate to the repeated adenosine sequence which is around the Cas9 cut site near the PAM site. (Fig. 4b and Fig. S5a). Taken together, this data demonstrates as proof of concept the significant consequences that the replacement of G with 8-OHdG has on misrepair and mutation.
PQ treatment induces misrepair of MED12 exon 2 in myometrial cells in vitro
Thus far, we have demonstrated that the myometrium from a uterus with LM exhibited increased ROS, that ROS inducers cause oxidation of DNA, 8-OHdG, which can eventually be misrepaired and mutated as demonstrated using replacement of G with oxidized G at c.130 of the MED12 gene. Next MM cells were exposed to either acute or chronic treatment with PQ (Fig. 5a) to determine the effects on exon2 of MED12. LM samples with known MED12 mutations served as positive controls and showed MED12 mutations at c.122, c.128, c.130, and c.130 were present at rates of 58.1, 25.0, 33.3, and 22.7%, respectively, based on a total of 50k reads/sample (n=4, Fig. 5b). Sequencing of MM samples treated with an acute or chronic dose of vehicle control (DMSO; Fig. 5c) showed a low mutation rate per nucleotide (acute: 0.12%; chronic: 0.10%) through the c.100-c.204 region (Fig. 5c). Twenty-four myometrial samples were treated with PQ once (acute) and the DNA sequence of MED12 exon 2 showed a higher rate of nucleotide alterations (average 0.42% per nucleotide) (Fig. 5d, Table S4). Of note, c.127C, c.130G and c.131G alterations reached an average rate of 0.56%, 0.46% and 0.49%, respectively. Among these, 11 out of 24 myometrial samples harbored misrepair of c.127C, and 2 out of 24 harbored misrepair of c.130-131GG above cut-off (Fig. 5d). A total of 15 myometrial samples were treated with PQ five times (chronic). Deep sequencing analysis revealed that overall baseline mutations/alterations were similar to those seen with acute treatment, with no statistical difference (average 0.417% vs. 0.416%, p>0.05). To our surprise, 6 of 15 myometrial samples harbored c.130G,131G mutations above the cut-off line with an average mutation rate per nucleotide reaching 0.52% (c.130G) and 0.57% (c.131G), respectively. Three of 15 myometrial samples in this group showed a high mutation rate at c.132T above the cut-off (Fig. 5e, Table S4). Interestingly, acute and chronic treatment resulted in different MED12 mutation hotspots (Fig. 5f), suggesting a potential cell selection mechanism or non-random targeted DNA strand breaks and repair . Next, we analyzed the ratio of G/C and T/A mutations and found that the G/C mutation rate is significantly higher than the A/T mutation rate (Fig. 5f). When all guanines were extracted from MED12 exon 2 and the misrepair pattern of G>T, A, C examined, we found high G>A, followed by G>T, and G>C showed the lowest misrepair pattern (Fig. S6a). These findings suggest that in vitro ROS exposure can increase mutations in MED12 exon 2 and selected point mutations are non-randomly presented at unproportionally high rates above chance alone.
Duplex Sequencing confirms high rate of MED12 mutations in myometria with oxidative burden both in vitro and in vivo
Given the error rate of conventional deep sequencing, we used duplex sequencing technology to detect bona-fide mutations with high accuracy and low rate of error by single strand-nucleotide change, effectively eliminating the chance of PCR errors (see Methods). A new cohort of 16 myometrial cases was exposed to chronic PQ treatment and duplex sequencing was performed. A similar mutation pattern in exon 2 with high misrepair/mutations at c.130-131GG was observed (Fig. 6, Table S4). In the control group, 6 of the 16 samples detected c.130-131G mutation (Fig. 6b). In contrast, 11 of the 16 PQ-treated cells detect 0.024% to 0.73% of c.130-131G mutations (Fig. 6a). The G misrepair rate in the PQ-treated group was also higher than the control group (Fig. S6b). Altogether, these data provide strong evidence that ROS promotes MED12 c.130-131GG mutations in myometrial cells in vitro. This is a discovery of substantial value to the field as proof of concept showing hotspot mutations in MED12 caused by inducers of ROS in the myometrium. Findings support MED12 exon 2 mutations at c.130-131G can be induced by oxidative exposure in vitro and in vivo.