Parity associates with chromosomal damage in uterine leiomyomas

Mechanical forces in a constrained cellular environment were recently established as a facilitator of chromosomal damage. Whether this could contribute to tumorigenesis is not known. Uterine leiomyomas are common neoplasms that display relatively few chromosomal aberrations. We hypothesized that if mechanical forces contribute to chromosomal damage, signs of this could be seen in uterine leiomyomas from parous women. We examined the karyotypes of 1946 tumors, and found a striking overrepresentation of chromosomal damage associated with parity. We then subjected myometrial cells to physiological forces similar to those encountered during pregnancy, and found this to cause DNA breaks and a DNA repair response. While mechanical forces acting in constrained cellular environments may thus contribute to neoplastic degeneration, and genesis of uterine leiomyoma, further studies are needed to prove possible causality of the observed association. No evidence for progression to malignancy was found.

The notion that confined environment is a cause of chromosomal damage has attracted significant attention in recent years and rupture of the nuclear envelope (NE) as a mechanism has been in a particular focus1,2. NE rupture has been found to be a common phenomenon even under normal conditions, and the frequency increases dramatically in confined environments3. Known causes for NE rupture include increased actomyosin contractility often promoted by the stiffness of the extracellular matrix4,5, as well as migration through narrow pores (reviewed in6). Damage can also be caused by nuclear deformation without NE rupture, for instance by mechanical shearing of DNA or exclusion of DNA repair factors from the site of DNA damage7,8. In addition, nuclear deformation caused by migration and cell compression have been shown to cause DNA damage via increased replication stress without NE rupture9. While chromosomal damage is an important driver of tumorigenesis10, whether mechanical forces could be a significant contributor to human neoplasia has not yet been addressed.

During pregnancy uterine smooth muscle cells are subject to constant stretching and subsequently, during labor, forceful episodic contractions generated by actin and myosin11,12. Uterine leiomyomas (ULs) are extremely common benign smooth muscle tumors. Lifetime prevalence approaches 70%13. ULs are a frequent cause of abnormal bleeding and pain, and are the most common cause of hysterectomy. Occurrence of multiple synchronous tumors is typical14. Most ULs are driven by a mutation in the MED12gene15,16. The karyotype is typically relatively stable as compared to the often highly complex karyotypes seen in cancer, yet chromosomal aberrations due to DNA double-strand breaks (DSBs) – in particular losses of chromosomal regions – are commonly seen17,18,19. A particularly interesting feature is the relatively frequent occurrence of complex chromosomal rearrangements (CCRs) that resemble chromothripsis, in otherwise benign ULs17. Chromothripsis is a single event where one or few chromosomes become subject to tens to hundreds of rearrangements. It is usually associated with TP53 mutant cancers20. Multiple mechanisms may lead to this phenomenon, the formation of micronuclei and subsequent damage to the chromosomes entrapped in them being a likely key cause21. The CCRs observed in ULs involve a limited number of chromosomes, encompass chained structural rearrangements, and thus, appear to have formed in a single event as shown in17. While these features characterize also chromothripsis, the difference is that most of the CCR events detected in UL are less dramatic. One plausible explanation is that the chromothriptic events that are powerful cancer drivers are not tolerated in UL genesis, as the cellular checkpoints are still intact21 in the TP53 wild-type precursor cells17. Uterine leiomyosarcoma (ULMS) is a rare malignant smooth muscle tumor displaying prevalent copy-number changes and chromothripsis events22. While the malignant potential of UL is extremely low, the possibility of transformation of UL to ULMS has been proposed23.

In this work, we test the hypothesis that if mechanical forces acting in a constrained cellular environment are a relevant source of chromosomal damage, signs of this could be seen in myometrial tumors from women who have undergone pregnancy and delivery. We examine the association of parity to chromosomal breakage in 1946 ULs derived from 628 patients, as well as to the association to ULMS in a population-based setting.

Patient and tumor characteristics

Parity was known for 628 patients, and 1935 tumors entered the analyses. In all, 31% of the tumors were from nulliparous patients and 69% from parous patients; 79% of the tumors were MED12 mutated (referred as MED12 tumors); and 21% were wild type (WT) (referred as WT tumors) for known MED12 driver mutations. A total of 595 patients contributing 1836 tumors had information on all background variables: age at hysterectomy, menopause status (pre, post, and hormonal replacement therapy), hormonal contraception use (never/ever), and smoking status (ever/never). Summary of patient characteristics is reported in Fig. 1a and Supplementary Table 1, tumor characteristics in Fig. 1b, and relationship between parity and the background variables in Fig. 1c.

a Patient characteristics (N = 628). The distribution of the number of tumors per patient, parity, the age of the patients, and menopause status. Median is shown by blue vertical line. b Intramural, subserous, and submucous locations of uterine leiomyomas (ULs) (N = 1451). The majority of the tumors are intramural, but MED12 mutated tumors are associated with subserous location. c On the top left, age at hysterectomy stratified by parity. The y-axis is in years. Boxplots show the median and the first and third quartiles. Error bars extend up to 1.5 IQR above the quartiles. The number of patients in each class is: parity 0 = 196, 1 = 108, 2 = 213, 3 = 71, 4 = 28, 5 = 9, 6 = 1, 7 = 1. On the bottom left, the number of patients with a history of ever smoking (yes/no) stratified by parity. On the top right, the number of patients with combined oral contraceptives use (ever/never) stratified by parity. On the bottom right, the number of pre-menopausal, post-menopausal patients and patients with hormonal replacement therapy (HRT). Source data are provided.

Full size image

Chromosomal gains and losses in UL associate with parity

We studied the relationship between parity and DNA damage by utilizing single-nucleotide polymorphism (SNP)-array based allelic imbalance (AI) data created in the Finland myoma study19. An example of AI determination from SNP-array is presented in Fig. 2a and examples of different levels of DNA damage in WGS data Fig. 2b. The mean number of breakpoints, and the mean of total length of loss regions, total length of gain regions, and total combined length are reported in Fig. 2c and Supplementary Table 2. The results are presented by parity (nulliparous and parous) as well as by tumor subtype (MED12 mutant and WT). We found an increase in the mean of breakpoint number as well as total length of copy number alterations when tumors from parous patients were compared to tumors from nulliparous patients. For example, the total length of somatic losses was higher in tumors from parous patients than in tumors from nulliparous patients: the mean was 62% higher in MED12 tumors and 47% higher in WT tumors.

a Allelic imbalance (AI) segments in SNP-array data. From top-down, examples of chromosome 1 B-allele frequencies (BAF), mirrored BAF (mBAF) and Log R Ratios (LRR). Bottom, a schematic figure of the interpretation of the segments: black horizontal lines represent AI segments. Blue horizontal line represents LRR threshold −0.15. Blue dots represent the calculated breakpoints: I. When AI segment is at the end of the chromosome, only one breakpoint is counted for. II. If the LRR difference between adjacent copy-number segments is <0.1, the segments are considered contiguous and are counted as only two breakpoints. III. When the segment is not at the chromosome end and the LRR difference between both adjacent segments is >0.1, two breakpoints are calculated. b Typical allelic imbalance patterns in ULs. Representative examples of complex chromosomal rearrangement (CCR) patterns in different ULs. Only chromosomes affected by allelic imbalance changes are shown: Top: the majority of tumors with high breakpoint count are from tumors wild type (WT) for known MED12 mutations. Translocations between chr12 (HMGA2) and chr14 (RAD51B) are common. Middle: chromosomal rearrangements seen in MED12 tumors are typically more moderate. Bottom: whole-chromosome losses were predominantly detected in parous patients’ WT tumors. Blue line: deletion, green line: insertion, purple line: inversion, red line: duplication, black line: translocation, blue box: LOH/Loss, red box: Gain. c Average breakpoint number, total length of loss regions, total length of gain regions, and the combined length of loss and gain regions are presented for tumors from parous and nulliparous patients. The results are presented for all 1935 tumors (purple) and separately for MED12 (salmon) and WT (turquoise) tumors. 95% bootstrap confidence intervals are presented for the mean. In all comparisons, the mean length of tumor genome AI regions is higher in parous than nulliparous patients. Source data are provided.

Full size image

The statistical associations are presented in Table 1 and the full models in Supplementary Figs. 1–3. Briefly, the combined total length of loss and gain regions was significantly associated with parity (P < 0.001, rate ratio (RR) = 1.32, 95% CI: 1.15–1.52, Bonferroni adjusted). The association remained significant within the MED12 tumors (P = 0.007, RR = 1.39 95% CI: 1.05–1.84).

Forty tumors had at least one whole chromosome loss. Of these tumors, only two (5%) were MED12 mutated. The most common loss, observed in 19 tumors, was chr22, followed by chr13 and chr14 in seven tumors. The WT tumors displayed a significant positive association between whole chromosome losses and parity (P = 0.002, odds ratio (OR) = 1.64, 95% CI: 1.13–2.36). See Supplementary Fig. 4 for full model. Among nulliparous patients, only one tumor displayed whole chromosome losses (Fig. 3b).

a Summary of CCR statuses for all 1935 tumors (purple), stratified by patient’s parity. Similarly, CCR statuses among MED12 mutated tumors (salmon) and among WT tumors (turquoise). b Association between whole chromosome loss and parity. c, Breakpoint number in tumors with CCR, stratified by nulliparous (n = 45 tumors) and parous patients (n = 134 tumors). Orange dots depict MED12mutated tumors, turquoise WT tumors. Of note, all ULs with a high breakpoint number are from parous patients’ WT tumors. The mean number of breakpoints in CCR–nulliparous tumors is 13.0 (95% CI: 10.93–15.79) and in CCR–parous tumors 18.17 (14.74–22.14). d, An average breakpoint number, total length of loss regions, total length of gain regions and the combined length of loss and gain regions are presented for MED12 tumors with CCR (n = 52) from parous (n = 36) and nulliparous patients (n = 16). 95% bootstrap confidence intervals are presented for the mean. e An average breakpoint number, total length of loss regions, total length of gain regions and the combined length of loss and gain regions are presented for WT CCR tumors (n = 127) from parous (n= 98) and nulliparous patients (n = 29). 95% bootstrap confidence intervals are presented for the mean. Source data are provided.

Full size image

Identifying complex chromosomal rearrangements

CCR events in 38 whole-genome sequenced (WGS) ULs were readily available17. We studied the number of DSBs in WGS samples with and without CCR as well as the respective SNP-array data, and extrapolated the CCR criteria to all SNP-array samples. The SNP-array based analysis identified 179 tumors (9.3% out of 1935) compatible with CCR. The mean number of breakpoints, and the total length of copy-number losses and gains are reported in Supplementary Table 2 for both CCR tumors and tumors without CCR separated by parity and subtype. MED12 mutation-positive tumors were underrepresented among the CCR tumors (29.0% out of 179), in line with our previous study17.

SNP-array based predicted CCRs were validated in 43 tumors by long-read WGS. Three tumors (7%) did not display interconnected rearrangements, while 40 (93%) displayed at least one CCR event (Supplementary Fig. 5 and Supplementary Table 3). Three tumors with CCR and different patterns of chromosomal damage are presented in Fig. 2b.

CCRs in UL associate with parity

Parity was significantly associated with CCR events (Fig. 3a, P < 0.001, OR = 1.29, 95% CI: 1.09–1.52). The association remained significant within the MED12 tumors (Fig. 3a, P = 0.006, OR = 1.47, 95% CI: 1.07–2.0). The confounding variables—age, smoking status, menopausal status and use of oral hormonal contraceptives—were not associated with CCR. See Supplementary Figs. 6–8 for full model description. Among the 179 tumors affected by CCR, the number of breakpoints was larger in tumors from parous patients compared to tumors from nulliparous patients: all 16 tumors with more than 28 breakpoints were from parous patients (Fig. 3c). The total length of AI for CCR tumors is presented in Fig. 3d, e.

The effect of tumor location on allelic imbalance

Information on tumor location and size was available for 1451 tumors. From the 1142 MED12 tumors 76% were intramural, 15% subserous, and 9% submucous. From the 309 WT tumors 87% were intramural, 6% were subserous, and 7% were submucous (Fig. 1b). In MED12 tumors, shorter length of lost chromosomal regions was observed in subserous tumors compared to intramural tumors (P = 0.001, RR = 0.28, 95% CI 0.11–0.72, Supplementary Fig. 9 and Supplementary Table 4). Tumor size was not associated with AI length.

Mechanical force induces DNA damage in myometrial cells

Effects of mechanical stretching were studied using primary cell cultures from patient-derived uterine leiomyoma and corresponding normal myometrium. Static stretching with 26% elongation was applied to the cells cultured on flexible membranes for 24–48 h with or without a recovery period (Fig. 4a). After releasing the stretch, initial evaluation by light microscopy showed the typical elongated morphology in non-stretched control cells, while the stretched cells appeared smaller and rounder. After fixation and immunofluorescence stainings, the cells were imaged by high-content confocal microscopy for in silico analysis. Representative fields of view of the cells are shown in Fig. 4b. An average of 483 nuclei were identified in the image analysis per staining (median 282 nuclei; interquartile range of 183–571 nuclei quantified per staining; Fig. 4c; details in Supplementary Fig. 10 and Supplementary Table 5).

a Schematic overview of the experiment. HR homologous recombination. b Representative fields of view of cells under each experiment condition (stretched and non-stretched control at the 24 h stretch+24 h recovery time point). Inset shows individual nuclei at higher magnification. c Number of nuclei quantified in the in silico image analysis. On top, average numbers of nuclei per experiment condition. On bottom, average numbers of nuclei per one imaging field (bar plot is mean of means; dots are mean values per experiment). Unadjusted, two-sided Mann–Whitney Utest P-values shown between control (n = 15) and stretched (n = 20). Blue: 24 h stretch without recovery; orange: 24 h with 6 h recovery; green: 24 h with 24 h recovery; red: 48 h without recovery. Time points 24 + 24 h and 48 h share the same control (hatched red bar). d DNA damage (γH2Ax positive nuclei) and, e NHEJ repair (γH2Ax + 53BP1 positive nuclei) quantified in stretched and non-stretched control cells. d, e Data are presented as a proportion (%) of positive nuclei (top bar plots) and as a fold-change of nuclei-positivity (bottom bar plots) between stretched and non-stretched control cells; error bars show a bootstrapping 95% confidence interval of each fold-change value. Odds ratio (OR) and fold-change (FC) values >1 suggest that stretching induces DNA damage/repair. Unadjusted, two-sided P-values from either logistic regression (top) or Fisher’s exact test (bottom) are denoted by: ***P < 0.0001, **P < 0.001, *P < 0.05, not significant (NS) P ≥ 0.05. Sample sizes (n) are shown for each significance test. See Supplementary Tables 6–7 for multiple-test adjusted P-values. Source data are provided as a Source Data file.

Full size image

To evaluate the level of DNA damage, γH2Ax positive nuclei (>10 γH2Ax foci) indicating DNA DSBs were quantified in stretched cells compared to non-stretched control cells. Myometrium and leiomyoma cells treated with ionizing irradiation (IR) as a positive control displayed the expected increase in γH2Ax positive nuclei (OR = 33.1 and OR = 11.9, respectively; both P < 0.0001; Supplementary Fig. 11). Stretching resulted in a more moderate yet highly significant increase in γH2Ax-positive nuclei both in leiomyoma and myometrium cells (OR = 1.6 and P < 0.0001; Fig. 4d and Supplementary Fig. 12). The response was dependent on treatment time; significant DNA damage was first detectable after 24 h of stretching in leiomyoma cells and showed a biphasic pattern, while myometrium cells showed significant DNA damage at the two later time points (Supplementary Table 7). The increase in DNA damage coincided with an increase in DNA repair (Fig. 4e). Non-homologous end joining (NHEJ), quantified as nuclei double-positive for γH2Ax (>10 foci) and 53BP1 (>5 foci), was the major DNA repair mechanism induced by stretching in both myometrium (OR = 1.5) and leiomyoma (OR = 2.0; both P < 0.0001; Fig. 4eand Supplementary Fig. 13) cells. Nuclei double-positive for RAD51 (>5 foci) and cyclinA2 (cell cycle S/G2), indicating homologous recombination (HR), were rare in both the IR (Supplementary Fig. 11) and stretching experiments (Supplementary Fig. 14, Supplementary Fig. 15), with no statistically significant differences between the experimental conditions. Taken together, mechanical stretching induced both DNA damage and NHEJ repair (Supplementary Table 6).

Reduced cell density was observed after stretching in both myometrium and UL cell cultures (P < 0.05; Fig. 4c). This was not due to apoptosis (as measured by cleaved caspase-3 positivity) that predominantly showed a decrease, perhaps owing to additional space created by stretching (Supplementary Fig. 16). While stretching had no overall effect on cell senescence (p14ARF) (Supplementary Fig. 17), it reduced the number of proliferative cells (nuclear Ki67 foci) (Supplementary Figs. 15 and 18). There was no consistent effect on numbers of cyclin A2 positive nuclei, that is, S phase cells (Supplementary Figs. 15 and 19). Notably, even after 48 h of stretching, a considerable proportion (25–37%) of the surviving cells remained proliferative. Data on CyclinA2, Ki67, cCasp3 and p14ARF are summarized in Supplementary Table 8. Segmented nuclei imaging data revealed increased roundness in nuclei with >10 versus <10 γH2Ax foci after stretching (P < 0.0001; Supplementary Fig. 20); also nuclear area showed a slight increase in γH2Ax positive stretched cells.

Parity is not associated with uterine leiomyosarcoma risk

The population-based analysis of 399 ULMS patients and 1657 matched controls did not show a statistically significant association (OR = 1.03; 95% CI 0.96–1.11; P = 0.38) between parity and ULMS.

Confined cellular environment is a recently proposed source of chromosomal damage2,7,24. It has also been shown that many cells are viable after mechanical stress7 and thus, the damage-induced chromosomal aberrations can be passed onto daughter cells. We hypothesized that traces of this phenomenon might be detected in uterine leiomyomas from parous women, as myometrial cells are exposed to mechanical forces during pregnancy and parturition and chromosomal aberrations have the potential to drive neoplastic growth. Our data shows that chromosomal breakage in UL is strongly associated with parity. We studied both the total length of genome altered by copy-number changes, a commonly used surrogate for genomic instability25,26, and the occurrence of complex chromosomal rearrangements resembling chromothripsis. Both of these DNA damage types displayed a positive association. Considering that the mechanisms that cause structural and numerical chromosome instability may be different27, we analyzed the occurrence of whole chromosome losses and found also these to be significantly associated with parity. Because chromosomal instability is an established feature of many cancers including ULMS, we examined the association of parity to occurrence of this malignancy, with negative results. While our population-based ULMS patient series is extensive considering the rarity of the condition22,23, its power was limited to large effects (OR ≥ 1.5), and even more extensive material is required to comprehensively resolve the question. The possible increased risk of ULMS associated with parity is likely to be small, if at all present.

We examined experimentally, utilizing static stretching of primary myometrial and leiomyoma cells, whether forces similar to those observed during pregnancy have the potential to cause DNA damage. The stretched cells displayed a significant increase in DNA damage, and concomitant induction of DNA repair by NHEJ was detected. This result is compatible with previous observations of cyclic mechanical tension triggering an increase in γH2Ax foci in the inner core of the vertebral disc28. Moreover, cyclic stretching was previously shown to enhance (though not induce) nuclear rupture and DNA damage caused by lamin A knockdown in cardiomyocytes4. In our experiments, stretching resulted in reduced cell density in the absence of apoptosis; one contributing factor may be the impaired cell proliferation observed after stretching. Nevertheless, a proliferating subpopulation of cells persisted after the challenge, and there was no induction of cell senescence in these short-term experiments.

Of the main pathways repairing double-strand breaks29, NHEJ was the major repair mechanism induced by stretching, while HR-mediated repair was uncommon both after irradiation and stretching. Because NHEJ does not use a complementary template, the fusion of two broken ends of DNA duplexes, with no or limited homology, may result in deletion or insertion of base pairs and other rearrangements30,31. Thus, occasional transmission of genetic alterations to daughter cells after stretching-induced DNA damage seems conceivable, although the short-lived primary cell cultures precluded long-term follow-up. This could be involved in increasing the risk of developing ULs after mechanical stretching. Further studies are needed to clarify what mechanisms contribute to the low rate of HR in human myometrium and ULs.

The irradiated positive control cells cultured on glass showed a more robust induction of γH2Ax foci compared to the stretched cells on flexible membranes. Compatible with these observations, increasing stiffness of cell culture matrix favors DNA damage and nuclear rupture in nuclear lamin A knockdown cells4,5. In this model, cell spreading on a stiff matrix causes local sites of high nuclear curvature, promoting rupture and decreasing nuclear circularity in live imaging5. In our experimental setup where the cells were fixed after the release of stretching, the stretched cells positive for DNA damage showed an increased roundness. While these observations may be related to nuclear deformation occurring during stretching or recovery, the detailed mechanisms remain to be elucidated in future studies. Interestingly, nuclear deformation, even in the absence of nuclear envelope rupture, can cause DNA damage through replication stress in the context of cell compression or migration through tight spaces9. Taken together, cell nuclei are sensitive to several types of mechanical stresses and the responses elicited depend on the cell type, culture matrix, and the type, strength and duration of such stress.

While complete understanding of the association between parity and chromosomal damage in UL needs more work as uterine myometrium is under multiple different powerful cues, direct force acting upon myometrial nuclei during gestation and labor appears as one conceivable candidate factor contributing to our findings in primary tumors. Confined environment is known to predispose cells to DNA damage through multiple different mechanisms (reviewed in Shah et al.3): Damage from increased actinomyosin contractility – a known key contributor of nuclear envelope rupture6 – that is most prominent at labor, and deformation of nuclei and subsequent damage at nuclear envelope through stretching during gestation, are both plausible possibilities. Mechanical shearing of DNA can take place even without damage to the nuclear envelope3. Nuclear deformation and NE rupture are common under extreme conditions, and chromatin protrusion through the nuclear lamina exposes DNA to the cytoplasm, resulting in DNA damage. NE rupture also exposes DNA to cytoplasmic nucleases. A particularly interesting form of damage resembling formation of micronuclei is that after NE rupture chromatin can become fragmented and separated from the nucleus6. This could predispose to the CCR changes observed in ULs17. The chromosomal fragments following NE rupture may also be distributed unequally during mitosis. Indeed, a defective nuclear envelope has been proposed as a main cause of aneuploidy in ovarian cancer32. In addition, mechanical forces work in concert with other cues present in the myometrium; during hypoxia for example33.

While our results firmly associate parity to chromosomal damage in UL, further studies are needed to examine whether mechanical forces play a causative role in this association. While mechanical force has recently been established as a factor capable of causing DNA damage3, it has to our knowledge not been implicated in genesis of common tumors. Its potential contribution to neoplastic degeneration in myometrium as well as other tissues is an intriguing new avenue of research, towards a more complete understanding of tumorigenesis.