Nuclear envelope rupture and repair during cancer cell migration

See allHide authors and affiliations

Science  15 Apr 2016:
Vol. 352, Issue 6283, pp. 353-358
DOI: 10.1126/science.aad7297

Repairing tears in the nuclear envelope

The nuclear envelope segregates genomic DNA from the cytoplasm and regulates protein trafficking between the cytosol and the nucleus. Maintaining nuclear envelope integrity during interphase is considered crucial. However, Raab et al. and Denais et al. show that migrating immune and cancer cells experience frequent and transitory nuclear envelope ruptures when they move through tight spaces (see the Perspective by Burke). The nuclear envelope reseals rapidly during interphase, assisted by components of the ESCRT III membrane-remodeling machinery.

Science, this issue pp. 359 and 353; see also p. 295


During cancer metastasis, tumor cells penetrate tissues through tight interstitial spaces, which requires extensive deformation of the cell and its nucleus. Here, we investigated mammalian tumor cell migration in confining microenvironments in vitro and in vivo. Nuclear deformation caused localized loss of nuclear envelope (NE) integrity, which led to the uncontrolled exchange of nucleo-cytoplasmic content, herniation of chromatin across the NE, and DNA damage. The incidence of NE rupture increased with cell confinement and with depletion of nuclear lamins, NE proteins that structurally support the nucleus. Cells restored NE integrity using components of the endosomal sorting complexes required for transport III (ESCRT III) machinery. Our findings indicate that cell migration incurs substantial physical stress on the NE and its content and requires efficient NE and DNA damage repair for cell survival.

The nuclear envelope (NE), comprising the inner and outer nuclear membranes, nuclear pore complexes, and the nuclear lamina, presents a physical barrier between the nuclear interior and the cytoplasm that protects the genome from cytoplasmic components and establishes a separate compartment for DNA and RNA synthesis and processing (1). Loss of NE integrity and nuclear pore selectivity has been linked to the normal aging process and a variety of human diseases, including cancer (2). In cancer progression, key steps of tumor cell invasion depend upon deformation of the nucleus into available spaces within the three-dimensional tissue (36). Whereas the cytoplasm of migrating cells can penetrate even submicron-sized pores, the deformation of the large and relatively rigid nucleus becomes a rate-limiting factor in migration through pores <25 μm2 in cross section (4, 610). We hypothesized that migration through such tight spaces provides a substantial mechanical challenge to the integrity of the nucleus. Thus, we investigated whether cell migration through confining spaces induces NE rupture and compromises DNA integrity and how cells repair such NE ruptures during interphase.

To model cancer cell invasion with precise control over cell confinement, we designed a microfluidic device containing constrictions with fixed height and varying widths mimicking interstitial pore sizes (fig. S1, A and B) (11). We detected NE rupture using previously established fluorescent reporters consisting of green or red fluorescent proteins fused to a nuclear localization sequence (NLS-GFP and NLS-RFP, respectively) that rapidly escape into the cytoplasm when NE integrity is lost (1214). Breast cancer, fibrosarcoma, and human skin fibroblast cells displayed transient loss of NE integrity, which coincided with the nucleus passing through the constrictions (Fig. 1, A and B; fig. S1, C to L; and movie S1). NE rupture was associated with transient influx of fluorescently labeled cytoplasmic proteins into the nucleus (fig. S2) and could also be detected by accumulation of the fluorescently labeled DNA-binding proteins’ barrier-to-autointegration factor (BAF) (15) and guanosine 3′,5′-monophosphate–adenosine 3′,5′-monophosphate (cyclic GMP-AMP) synthase (cGAS) (16) at sites of NE rupture (fig. S3).

Fig. 1 Nuclear rupture during migration through confining environments.

(A) Image sequence of an MDA-MB-231 breast cancer cell that exhibited multiple NE ruptures while moving through 2- × 5-μm2 constrictions. See also movie S1. DIC, differential interference contrast; EGF, epidermal growth factor. Here and in all other figures, red arrows and lines below frames indicate beginning and duration of NE rupture(s). Scale bar, 20 μm. (B) Fluorescence intensity (arbitrary units) of nuclear and cytoplasmic NLS-GFP of the cell in (A), showing loss of nuclear signal and concomitant increase in cytoplasmic fluorescence upon NE rupture, followed by gradual reimport of NLS-GFP into the nucleus. R1 to R3 indicate NE rupture events. (C) NE rupture in an HT1080 fibrosarcoma cell coexpressing NLS-GFP and fluorescently labeled histones (H2B-RFP) migrating inside a collagen matrix (2.5 mg/ml) with MMP inhibitor GM6001. See also movie S2. White arrowheads indicate the minimal nuclear diameter. (Insets, top row) Close-up of nuclear bleb formation (red arrowhead). Scale bar, 10 μm; insets, 2 μm. (D) Minimal nuclear diameter and nuclear and cytoplasmic NLS-GFP fluorescence intensity for the cell in (C). (E) Minimal nuclear diameter in rupturing and nonrupturing HT1080 cells in collagen matrix. ***P < 0.001; n = 159 and 62 cells, respectively. (F) Incidences of NE rupture as a function of constriction sizes in collagen matrices [slope = –1.224; coefficient of determination (R2) = 0.999] and microfluidic devices during a ~12-hour period (slope = –1.219; R2 = 0.974). Regression based on HT1080 and MDA-MB-231 cells; n = 55 to 445 cells per condition. (G and H) Multiphoton image of HT1080 fibrosarcoma cells 5 days after implantation into the mouse dermis. Dashed box in (G) indicates cell with NE rupture and nuclear bleb (arrowhead) shown in (H). Collagen fibers detected by second harmonic generation (SHG); blood vessels visualized by AlexaFluor-750–labeled 70 kD dextran. Scale bar, 20 μm. (I) Fluorescence intensity of cytoplasmic (red) and total (black and gray) NLS-GFP signal in rupturing (red and black) and nonrupturing (gray) cell(s) in the same field of view. (J) Incidence of NE rupture as function of migration mode. **P < 0.01 versus thin and thick strands; n = 22 to 211 cells per condition. Error bars, SE.

We then tested whether NE rupture also occurs during cancer cell migration in biological environments. Fibrosarcoma cells and skin fibroblasts exhibited NE rupture during migration in fibrillar collagen matrices (Fig. 1, C and D; fig. S4; and movie S2), with kinetics similar to those recorded in the constriction channels (fig. S1M). NE rupture typically occurred when the minimal nuclear diameter, which closely matches the pore size encountered by the cell (6), dropped to 3 μm (Fig. 1, D and E) and so linked NE rupture to cell movement through narrow spaces. Accordingly, NE ruptures were rare (i.e., below 5% per ~12-hour observation period) for cells migrating on glass, in low-density collagen matrices, or through channels 15 × 5 μm2 wide, in line with previously reported rates of spontaneous NE rupture in cancer cell lines (13). However, when matrix pore sizes were reduced to 5 to 20 μm2 and below by increasing collagen concentration and/or by blocking the cells’ ability to cleave collagen fibers and widen pores by addition of a matrix metalloprotease (MMP) inhibitor, the incidence of NE rupture increased 10-fold and more (fig. S5, A and B). Similarly, reducing the pore size of the microfluidic channels to <20 μm2 increased NE rupture more than 10-fold (fig. S5, C and D). Irrespective of the experimental model, the incidence of NE rupture increased exponentially with decreasing pore size (Fig. 1F) and reached >90% when the nuclear height was confined to 3 μm (fig. S5E). Imaging HT1080 fibrosarcoma cells invading the collagen-rich mouse dermis in live tumors after orthotopic implantation confirmed that migration-induced NE rupture also occurs in vivo, particularly in individually disseminating cells (Fig. 1, G to J; fig. S6; and movie S3). NE rupture was less prevalent in cells moving as multicellular collective strands (Fig. 1J and fig. S6), which typically follow linear tracks of least resistance and undergo less pronounced nuclear deformations (3, 7).

NE rupture in vitro and in vivo was often accompanied by protrusion of chromatin through the nuclear lamina (Fig. 2, A and B; fig. S7, A to I; and movies S4 and S5). The incidence of such “chromatin herniations” increased significantly with decreasing pore sizes (Fig. 2C and fig. S7H). In severe cases, small pieces of the nucleus were pinched off from the primary nucleus as cells passed through narrow constrictions (Fig. 2D and movie S6), which resulted in an elevated and persistent fraction of cells with fragmented nuclei (Fig. 2E and fig. S7J). Furthermore, cells that had passed through microfluidic constrictions had more nuclear fragments positive for γ-H2AX, a marker of DNA double-strand breaks (17), than cells that had not yet entered the constrictions (Fig. 2F), consistent with recent reports that loss of NE integrity in micronuclei can cause DNA damage (14) and chromothripsis (18). Intense γ-H2AX staining could also be found at chromatin protrusions (fig. S8A). To confirm that DNA damage was caused by migration-induced nuclear deformation and NE rupture, we performed live-cell imaging on cells coexpressing NLS-GFP and fluorescently labeled 53BP1 (RFP-53BP1), another marker of DNA damage (19, 20). NE rupture, and even severe nuclear deformation alone, resulted in rapid formation of new RFP-53BP1 foci as cells squeezed through the constrictions (Fig. 2, G and H; and fig. S8, B to G), consistent with previous reports of increased activation of DNA damage response genes after compression-induced chromatin herniation and NE rupture (21).

Fig. 2 Confined migration leads to chromatin protrusions, nuclear fragmentation, and DNA damage.

(A) Representative image of an MDA-MB-231 cell coexpressing GFP-lamin C and H2B-RFP developing chromatin protrusion (arrowhead) during migration through a microfluidic constriction. Scale bar, 5 μm. (B) HT1080 cell in a collagen matrix (2.5 mg/ml + GM6001) with chromatin protrusion (arrowheads) across the nuclear lamina, stained for lamin A/C (green), DNA (red), and F-actin (turquoise). Scale bars, 10 μm; bottom inset, 2 μm. (C) Percentage of cells with chromatin protrusions as a function of collagen matrix pore size. 2D, unconfined migration on glass slide. **P < 0.01; ***P < 0.0001; n = 50 to 146 cells per condition. (D) Representative image sequence of the formation of chromatin-filled nuclear membrane blebs (white arrowheads) and subsequent nuclear fragmentation in an MDA-MB-231 breast cancer cell coexpressing NLS-GFP (green) and H2B-RFP (red) during migration through consecutive 2- × 5-μm2 constrictions. Insets are indicated by dashed lines. See also movie S6. (E) Percentage of cells with fragmented nuclei before entry into constriction channel, inside the channel, and after exit. ***P < 0.001; n = 9775, 1376, and 3072, respectively. (F) Example of an intact nucleus (left top), a nucleus with a small fragment (arrowhead) (left middle), and a nucleus with a γ-H2AX–positive fragment (left bottom). Scale bar, 10 μm. Percentage of cells with γ-H2AX–positive nuclear fragments before, inside, and after migration through constriction channels (right). ***P < 0.001; n = 1376 to 3072 cells per condition. (G) Percentage of HT1080 cells migrating through 2- × 5-μm2 or 1- × 5-μm2 constrictions that formed 53BP1-RFP foci as a function of nuclear rupture. **P < 0.01, n = 35 cells total. (H) Representative example of formation of 53BP1 foci (arrowheads) in U2OS cell coexpressing NLS-GFP and RFP-53BP1 during migration through 2- × 5-μm2 constriction and NE rupture. Scale bar, 10 μm. Error bars, SE.

To gain further insights into the biophysical processes underlying NE rupture, we analyzed the timing and location of NE rupture in regard to nuclear deformation, local membrane curvature, NE composition, and cytoskeletal forces. NE rupture occurred predominantly (~76%) at the leading edge of the nucleus (Fig. 3, A and B) and was almost always (89.9 ± 2.14%, n = 198 cells) preceded by, or coincided with, the formation of nuclear membrane protrusions (“blebs”) as the nuclei moved through the constrictions (Fig. 1, C and H; Fig. 3, A and C; and fig. S9). Nuclear membrane blebs typically (96.1 ± 1.38%, n = 178 blebs) formed at sites where the nuclear lamina signal, particularly the lamin B1 network, was weak or absent (Fig. 3, D to F, and fig. S10, A and B). These findings indicate that blebs form when segments of the nuclear membrane detach from the nuclear lamina and bulge into the cytoplasm. Depletion of the lamins A/C and lamin B2 significantly increased the likelihood of NE rupture (Fig. 3G and fig. S10C), which together with previous studies (4, 12, 13, 2123) suggest that lamins are important for stabilizing the NE.

Fig. 3 Molecular sequence of nuclear rupture.

(A) Nuclear membrane bleb formation (arrowhead) and collapse upon NE rupture. Scale bars, top and bottom rows, 10 μm; inset, 5 μm. (B) Distribution of NE rupture sites along nuclear periphery. Regions are based on the migration direction (black arrow) and denoted in the cartoon. n = 352 cells; *P < 0.05; ***P < 0.001; all comparisons relative to the hypothetical uniform distribution of 33%. (C) Kymograph of nuclear bleb formation (arrowhead) and collapse during subsequent NE rupture, corresponding to the image sequence in Fig. 1C. (D) Nuclear membrane blebs formed at sites of low GFP-lamin B1 intensity and were devoid of GFP-lamin B1 (arrowheads). Intensity along the blue line is quantified in (E). The gray area indicates the local absence of lamin B1. Representative cell out of 178 cells observed. (E) GFP-lamin B1 fluorescence intensity profile. The gray area and arrowhead indicate the section where the nuclear bleb forms. ***P < 0.001; comparing values inside and outside the gray area. (F) Percentage of nuclear blebs containing detectable amounts of lamin A, B1, or B2. ***P < 0.001; compared with the expected value of 100% for the primary nucleus; n = 178 to 199 cells per condition. (G) Incidence of NE rupture after small interfering RNA (siRNA) treatment against lamin A/C, lamin B2, or nontarget (NT) control. *P < 0.05; **P < 0.01; ***P < 0.001; n = 384, 150, and 163 cells for constrictions ≤2 × 5 μm2; n = 166, 68, and 49 for 15- × 5-μm2 constrictions. (H) Blebbistatin treatment reduced NE rupture incidence during migration in constriction channels, but not in 15- × 5-μm2 control channels. DMSO, dimethyl sulfoxide as a control. *P < 0.05; n = 286 and 194 for constrictions ≤2 × 5 μm2; n = 122, and 54 for cells in constrictions 15 × 5 μm2. (I) GFP-lamin A accumulated at sites of NE rupture, forming lamin scars (arrowheads). Gray and black bars under images correspond to line profiles at different times in (J). Scale bar, 10 μm. (J) GFP-lamin A signal intensity along a section of the nuclear rim [blue line in (I)]. (K) GFP-lamin A accumulation increases with the severity of NE rupture. P < 0.0001; n = 46 cells (slope = 1.345, R2 = 0.3945). Error bars, SE.

Consistent with previous observations (13, 21, 22, 24, 25), the expanding nuclear blebs were devoid of GFP-lamin B1 (Fig. 3, D to F) and nuclear pores (fig. S10D), and initially contained little or no GFP-lamin A and B2 (fig. S10, A and B). Upon NE rupture the blebs retracted and collapsed (Fig. 1, C and H; Fig. 3, A and C; fig. S9; and fig. S10, E to H), which suggested that hydrostatic pressure was released from the fluid-filled blebs. Although the NE contains pores, recent studies have shown that the NE can provide an effective barrier to support intracellular pressure gradients (26, 27). Nuclear pressurization could arise from actomyosin contraction at the rear of the nucleus that is required to move the nucleus through tight spaces (5, 6), which effectively mimicks cellular compression experiments (21, 22). Supporting this idea, treatment with low concentrations of blebbistatin, a myosin II inhibitor, significantly decreased the incidence of nuclear rupture (Fig. 3H) without inhibiting the ability of cells to migrate through larger channels (fig. S10I)

The transient nature of NE rupture suggests that cells can efficiently restore nuclear membrane integrity during interphase. We observed rapid (<2 min) accumulation of GFP-lamin A at the site of rupture that correlated with the severity of nuclear rupture and that often persisted for hours (Fig. 3, I to K, and fig. S11). Subsequent ruptures within the same cell occurred at distinct sites (fig. S11A), which implied local protection by these “lamin scars.” Two recent reports have shown that members of the endosomal sorting complexes required for transport III (ESCRT III) family are involved in resealing the nuclear membrane during late anaphase (28, 29). To assess whether ESCRT proteins have a similar function in interphase NE repair, we generated GFP-fusion constructs of the ESCRT III subunit CHMP4B, involved in recruiting other ESCRT III proteins and facilitating membrane scission, and the ESCRT III–associated VPS4B, which is required for disassembly and recycling of ESCRT III proteins (30). Upon NE rupture induced by confined migration or laser ablation, CHMP4B-GFP and VPS4B-GFP rapidly (in ≤2 min) formed transient foci at the site of nuclear membrane damage (Fig. 4, A to D; fig. S12; fig. S13, A to C; and movie S7). Superresolution microscopy confirmed recruitment of endogenous ESCRT III proteins to sites of NE rupture into complexes ≤160 nm in size (Figs. 4E and fig. S13, D to H). Recruitment of the ESCRT III machinery was independent of microtubules (fig. S14). Depletion of the ESCRT III subunit CHMP2A, CHMP7, or ectopic expression of a dominant-negative VPS4B mutant (GFP-VPS4BE235Q) that prevents ESCRT III subunit recycling significantly increased the time required for nucleo-cytoplasmic recompartmentalization (Figs. 4, F and G, and fig. S13, I to N), which indicated a crucial role of ESCRT III proteins in restoring nuclear membrane integrity. To assess the functional relevance of NE repair, we quantified cell viability after NE rupture. Under normal conditions, the vast majority (>90%) of cells survived even repeated NE rupture (Figs. 1A and 4H and fig. S4A). Inhibiting either ESCRT III–mediated NE repair or DNA damage repair pathways alone did not reduce cell viability, but inhibition of both NE and DNA repair substantially increased cell death after NE rupture (Fig. 4H).

Fig. 4 ESCRT III mediates nuclear envelope repair.

(A) CHMP4B-GFP was transiently recruited to the site of nuclear membrane damage (arrowhead). See also movie S7. Dashed boxes indicate areas used for measurements in (B). Representative sequence from 12 HT1080 cells total. Scale bar, 10 μm. (B) CHMP4B-GFP fluorescence intensity at rupture site (black), normalized to prerupture intensity, increased after NE rupture, indicated by increase in cytoplasmic NLS-RFP signal (gray). Red arrow indicates time of NE rupture. ΔF/F0, the change of the intensity from the original intensity before stimulation. (C) Recruitment of VPS4B-GFP to sites of NE rupture (arrowhead) in an MDA-MB-231 cell. Dashed boxes indicate areas used for measurements in (D). Representative example from 18 cells total. Scale bar, 10 μm. (D) VPS4B-GFP fluorescence intensity in the cytoplasm (black) increased rapidly after NE rupture, detected by increase in cytoplasmic NLS-RFP signal (gray). Red arrow indicates time of NE rupture. (E) Representative superresolution image (from 12 cells total) of endogenous CHMP4B accumulation at NE rupture site. Lamin A/C accumulation confirmed rupture site (red arrowhead). Blue arrowhead indicates decreased lamin B intensity at the base of the bleb. Scale bars, 5 μm; inset, 1 μm. (F) siRNA-mediated depletion of ESCRT III proteins CHMP7 and CHMP2A in HT1080 cells resulted in an increased duration of NLS-GFP in the cytoplasm after migration induced NE rupture compared with nontarget (NT) controls. ***P < 0.001; n = 137 and 107, respectively. (G) Expression of the dominant-negative mutant VPS4BE235Q-GFP increased the duration of NLS-GFP in the cytoplasm after migration induced NE rupture in HT1080 cells, which indicated impaired nuclear membrane repair. **P < 0.005; n = 17 and 10, respectively. (H) Percentage of HT1080 cells dying after migration induced NE rupture, in the absence or presence of inhibition of ESCRT III by dominant-negative VPS4BE235Q-GFP and/or DNA repair with the ataxia telangiectasia mutated kinase (ATM) inhibitor KU-55933 (ATMi). **P < 0.01, comparing ATMi+/E235Q to ATMi+/WT and ATMi/E235Q; n = 20, 32, 38, 30, respectively. Error bars, means ± SEM.

Taken together, our studies demonstrate that cell migration through confining spaces, as frequently encountered during cancer cell invasion, can challenge the integrity of the NE and DNA content, which could promote DNA damage, aneuploidy, and genomic rearrangements, and—in the absence of efficient repair—cell death (31). We propose a biophysical model in which cytoskeletal-generated nuclear pressure results in the formation and eventual rupture of nuclear membrane blebs at sites of high membrane curvature and weakness in the underlying nuclear lamina (fig. S15). These events could be particularly prominent in cells with reduced levels of lamins, whose expression is deregulated in many cancers and often correlates with negative outcomes (32, 33). Although NE rupture, and resulting genomic instability, may promote cancer progression, it may also represent a particular weakness of metastatic cancer cells and an opportunity to develop novel antimetastatic drugs by specifically targeting these cells, for example, by blocking NE repair and inhibiting DNA damage repair.

Supplementary Materials

Materials and Methods

Figs. S1 to S15

Movie Captions S1 to S7

References (3443)

Movies S1 to S7

References and Notes

Acknowledgments: The authors thank G. Lahav, A. Loewer, M. Zwerger, M. Smolka, S. Emr, N. Buchkovich, B. Burke, H. Worman, S. Young, L. Fong, D. Discher, J. Broers, J. Goedhart, H. Yu, and W. Zipfel for cells, reagents, and helpful discussion. The authors acknowledge E. Wagena for mouse surgery and cell transfection, C. Paus for generation of in vitro collagen data, and K. Zhang, E. Ghazoul, D. Huang, G. Fedorchak, G.-J. Bakker, and F. Ahmadpour for help with image analysis. The authors are grateful to M. Piel and colleagues for sharing their unpublished data and manuscript. This work was supported by awards from NIH (R01 HL082792 and R01 NS059348); the Department of Defense Breast Cancer Research Program (BC102152 and BC150580); NSF (CBET-1254846); the National Cancer Institute, NIH, through the Cornell Center on the Microenvironment and Metastasis (U54 CA143876); and an NSF Graduate Research Fellowship to A.L.M (DGE-1144153). This work was further supported by the Netherlands Science Organization (NWO-VIDI 917.10.364 to K.W. and NWO-VICI 918.11.626 to P.F.) and the Cancer Genomics Center, Netherlands. This work was performed in part at the Cornell NanoScale Science and Technology Facility, which is supported by NSF (grant ECCS-15420819). Imaging was performed in part through the Cornell University Biotechnology Resource Center, which is supported by New York Stem Cell Foundation (NYSTEM) (CO29155), NIH (S10OD018516), and NSF (1428922), and by the Radboud University Medical Center Microscopic Imaging Center and Preclinical Animal Imaging Center (PRIME). Additional experimental details and data are available in the supplementary materials.

Stay Connected to Science

Navigate This Article