Abstract
The LMNA gene encodes the A-type lamins that polymerize into ~3.5 nm thick filaments, and together with B-type lamins and lamin binding proteins form the nuclear lamina. Mutations in LMNA are associated with a wide variety of pathologies. In this study, we analyzed the nuclear lamina of embryonic fibroblasts from LmnaH222P/H222P mice, which develop cardiomyopathy and muscular dystrophy. Although the organization of the lamina appeared unaltered, there were changes in chromatin and B-type lamin expression. An increase in nuclear size and consequently a relative reduction in heterochromatin near the lamina allowed for a higher resolution structural analysis of lamin filaments using cryo-electron tomography. This was most apparent when visualizing lamin filaments in situ, and using a nuclear extraction protocol. Averaging of individual segments of filaments in LmnaH222P/H222P mouse fibroblasts resolved two-polymers that constitute the mature filaments. Our findings provide better views of the organization of lamin filaments and the effect of a striated muscle disease-causing mutation on nuclear structure.
Introduction
Nuclear lamins are the intermediate filament (IF) building blocks of the nuclear lamina on the nucleoplasmic aspect of the inner nuclear membrane of metazoan cells (Fisher et al., 1986; McKeon et al., 1986). They are classified as type V IF proteins based on their sequences (Steinert and Roop, 1988). Similar to cytoplasmic IF proteins, lamins contain a long rod domain comprised of four coiled-coil α-helical segments, termed 1A, 1B, 2A and 2B, separated by flexible linkers (Gruenbaum and Foisner, 2015). This domain is flanked by a non-helical N-terminal head and C-terminal tail domain. The C-terminal domain hosts a nuclear localization sequence, and an immunoglobulin (Ig) like fold.
The nuclear lamina primarily provides mechanical support to the cell nucleus (Maurer and Lammerding, 2019; Pfeifer et al., 2019; Sapra et al., 2019). Four main lamin isoforms are found in mammals. In humans, they are encoded by the LMNA, LMNB1, LMNB2 genes that in somatic cells encode lamin A/C, lamin B1 and lamin B2, respectively (Worman, 2012). While the B-type lamins are expressed in almost all mammalian cell types, the expression of A-type lamins is developmentally regulated and occurs primarily in differentiated cells (Constantinescu et al., 2006; Kim et al., 2011). Both types of lamins are localized to the nuclear periphery; however, small amounts of A-type lamins are also found in the nucleoplasm where they may function in chromatin organization and gene regulation (Naetar et al., 2017).
In solution, lamin dimers form a parallel coiled-coil structure between two monomers (Klapper et al., 1997). These further assemble by head-to-tail association into long polymers, which further associate laterally into the mature filaments (de Leeuw et al., 2018; Stuurman et al., 1998). However, due to the length and flexibility of the coiled-coil domain, which is ~50 nm, structural determination of lamin dimers has been a challenging task (Makarov et al., 2019; Zwerger and Medalia, 2013). A detailed atomic model of full-length lamin proteins is still elusive. However, structures of lamin A fragments (Ahn et al., 2019; Herrmann and Aebi, 2004; Kapinos et al., 2011; Lilina et al., 2019) have exemplified the interactions between lamin coil 1B to form a tetrameric filament (Ahn et al., 2019; Lilina et al., 2019). The atomic structure of the Ig-fold of human lamin A/C has also been determined, exhibiting a globular two β-sheets structure (Dhe-Paganon et al., 2002; Krimm et al., 2002). The analysis of coiled-coil fragments provided insights into the intra-organization of lamin dimers (Strelkov et al., 2004). Based on such analysis (Ahn et al., 2019) an alternative model for lamin assembly in which lateral interactions between the coiled-coil of the dimers governs the lateral assembly, thus, head-to-tail polymer is avoided. An analysis of lamin filaments in fibroblasts has shown that lamins assemble into 3.5 nm thick filaments within a ~14 nm thick meshwork attached to the inner nuclear membrane (Turgay et al., 2017). The filaments exhibit a short persistence length of <200 nm, which not only hints at their unique mechanical properties and flexibility (Sapra et al., 2019), but also impose a major challenge for structurally reconstruction of lamin filaments.
Mutations in the genes encoding lamins, particularly LMNA, cause human diseases termed laminopathies (Tatli and Medalia, 2018; Worman and Bonne, 2007). The most common of these rare laminopathies is dilated cardiomyopathy usually associated with muscular dystrophy (Maggi et al., 2016) (Nicolas et al., 2019). The LMNA p.His222Pro (hereafter H222P) missense mutation was first identified in a family with autosomal dominant Emery-Dreifuss muscular dystrophy (Bonne et al., 2000). Homozygous LmnaH222P/H222P mice develop dilated cardiomyopathy and regional skeletal muscular dystrophy that phenocopies the human disease (Arimura et al., 2005; Muchir et al., 2012). Cells from LmnaH222P/H222P mice have altered stiffness (Chatzifrangkeskou et al., 2020) and abnormalities in several cell signaling pathways (Choi et al., 2018). However, it is not clear how this mutation affects the structure of the nuclear lamina and lamin filaments. The H222P amino acid substitution is localized to the linker domain of lamin A/C, between coil 1B and 2A of the proteins. While substitution with a proline would presumably disrupt α-helix structure, it does not play a role in the predicted coiled-coil interactions.
To obtain insights into how the H222P amino acid substitution affects lamin structure, we used several modalities to analyze fibroblasts from LmnaH222P/H222P mice. Quantitative immunofluorescence and total internal reflection fluorescence (TIRF) microscopy indicated nuclear alterations in these cells, while cryo-electron tomography (cryo-ET) provided new structural insights into the nuclear lamina and the lamin filaments in situ and in ghost nuclei. We show that the nuclear area is increased while heterochromatin is proportionally reduced. This results in an increase of contrast for lamin filaments using cryo-ET. Applying averaging approaches to in silico fragmented filaments (Martins et al., 2020) produced an unprecedented view of the lamin filaments and their substructures. Moreover, mapping back the structural class averages into their original position on lamin filaments provides better resolved insight into lamin filament organization. While the structure of lamins is presumably unaffected by the pathogenic H222P amino acid substitution in lamin A/C, the chromatin organization at the nuclear lamina and overall nuclear organization is altered.
Results
The nucleus, lamins and condensed chromatin in LmnaH222P/H222P fibroblasts
At a cellular level, the impact of amino acid substitutions on A-type lamins on nuclear organization has been previously demonstrated for several point mutations (Bertrand et al., 2020; Goldman et al., 2004; Scaffidi and Misteli, 2008; Vigouroux et al., 2001). To gain insights into the impact of the lamin A/C H222P amino acid substitution on the nucleus, and in particular on nuclear lamin filaments, we examined immortalized LmnaH222P/H222P and wild type mouse embryonic fibroblasts (MEFs). Using TIRF microscopic analysis, we confirmed that both lamin A/C and lamin B1 were properly localized to the nuclear lamina in LmnaH222P/H222P MEFs similar to wild type, but the nuclei appeared larger (Fig. 1A). Quantitative analysis showed that the nuclei area of LmnaH222P/H222P MEFs was significantly increased at 1.09 times that of wild type (Fig. 1B). We used immunofluorescence microscopy to obtain signals to quantify the amounts of histone H3K9me3, a marker of heterochromatin, histone H327ac, a maker of transcriptional activity, lamin B1 and lamin A/C in wild type and LmnaH222P/H222P MEFs (Fig. S1). This analysis showed that the LmnaH222P/H222P MEFs had significantly decreased histone H3K9me3, increased histone H3K27ac, increased lamin B1 and decreased lamin A/C (Fig. 1B). When analyzed by immunoblotting, lamin B1 and lamin B2 expression was increased in the LmnaH222P/H222P MEFs; however, the histone H3K9me3 signal was slightly increased and the lamin A/C signal was not reduced (Fig. S2). These results suggest that incorporation of lamins into the nuclear lamina is not directly affected by the lamin A/C H222P amino acid substitution in linker 1-2, between coil 1 and 2. The increase in nuclear surface area likely resulted in reduction of heterochromatin levels per nuclear envelope area, rather than changes in the absolute quantity of heterochromatin. The changes in lamin expression are presumably attributed to the increased surface of the nuclear envelope that may induce an upregulation of lamin B1, while the overall chromatin and lamin A/C levels resemble their original amounts.
Nuclear lamina structure in LmnaH222P/H222P MEFs
The molecular organization of the nuclear envelope can be visualized by cryo-ET (Harapin et al., 2015; Weber et al., 2019). We cultured cells on an EM-grid prior to vitrification and focus ion beam (FIB) milling in conjunction with cryo-ET (Rigort et al., 2012). This allowed us to visualize the organization of the nuclear envelope in LmnaH222P/H222P MEFs in situ. The nuclear membranes, lamin filaments and chromatin were all seen without any apparent structural alterations; however, the lamins were better resolved in the mutant cells (Fig. 2A,B; Fig. S3). Surface rendering of the thin reconstructed sections provides a view of the nuclear envelope regions, including the adjacent cytoplasmic structures (Fig. 2C). The nuclear areas hosting lamin filament meshworks were clearly visible with less chromatin and other nuclear components shadowing the lamins in LmnaH222P/H222P MEFs compared to wild type ones (Fig. 2C, Fig. S4A,B). The reduced density of nuclear structures at the lamina detected in cryo-tomograms of LmnaH222P/H222P nuclear envelopes confirmed the fluorescent microscopy analysis and indicated a bigger gap and reduced interaction of lamins with dense chromatin.
To further enhance visualization of the nuclear lamins, we knocked down vimentin expression and removed the cytoplasm, using short exposure to mild detergent followed by nuclease treatment prior to rapid vitrification (Turgay and Medalia, 2017). This produced ghost nuclei in which lamin filaments could be readily identified and followed over longer distances and large data sets acquired from most positions along the nuclear envelope (Tenga and Medalia, 2020). Here, we applied this procedure and detected lamin filaments in both LmnaH222P/H222P and wild type ghost nuclei (Fig. 3A, B, respectively). The lamin meshworks were better identified in the ghost nuclei from the mutant cells (Fig. 3A). A surface rendering of LmnaH222P/ H222P MEFs showed lamin filaments, nuclear pore complexes and cytoplasmic vimentin filaments (Fig. 3C) that were similar to that previously reported in wild type fibroblasts (Turgay et al., 2017). The overall organization of lamin filaments in ghost nuclei from LmnaH222P/H222P MEFs resembled the wild type nuclear lamina as seen by the rendered view of two tomograms (Fig. S4C,D). While nuclease treatment only removed part of the heterochromatin from ghost nuclei of wild type MEFs, these densities were hardly detected in the ghost nuclei prepared from the LmnaH222P/H222P MEFs (Fig. S4C,D, green). These experiments and cryo-ET analysis suggest that the overall organization of the nuclear lamins is not substantially altered in LmnaH222P/H222P MEFs. Rather, the H222P amino acid substitution in lamin A/C influences the size of the nucleus and the chromatin organization at the interface of the lamina, providing a better view of the nuclear lamina.
Structural organization of lamin filaments by averaging and filament reconstitutions
Lamins dimerize in solutions and form head-to-tail polymers (Stuurman et al., 1998); however, how these long coiled-coils protein structures are further assembled into the filaments of the nuclear lamina remains unknown. The higher contrast of individual lamin filaments observed by cryo-ET of LmnaH222P/H222P MEFs encouraged us to acquire 250 cryo-tomograms of ghost nuclei to gain further insights into their organization and formation. Lamin filaments in these nuclei appeared flexible and often curved, although some filaments exhibit a straight appearance, and were occasionally decorated with globular densities (Fig. 4A). These filaments closely resembled lamin filaments extracted from wild type MEF ghost nuclei (Fig. S5A). The globular densities were presumably the Ig-like fold domains (Turgay et al., 2017). The position of these globular structures along the filament was roughly regular with some variable distance from the axis of the filaments. This was likely due to the 70 amino acid linker between the end of helix 2B and the Ig-like fold domain. The heterogeneity in the position of Ig-fold domains and heterogeneity of lamin filaments, as well as and the possibility of other proteins binding to the lamins, prevents using conventional averaging approaches to analyze their structure. In order to reduce the complexity and flexibility of these filaments, we therefore fragmented them into short segments (12 nm in length and 4.4 nm in width) in silico and calculated their 2D projections followed by a single particle classification approach (Martins et al., 2020). The restricted width of the analyzed areas allowed us to primarily focus on the coiled-coil rod domains that constructed the core of the filaments, with the risk of partial exclusion of the Ig-like fold domains. Analysis of the prominent 2D classes representing different views revealed substructures within the ~3.5 nm diameter lamin filaments, most pronounced among them two ~1.8 nm thick filamentous substructures often interacting and crossing each other and sometimes merging into one structure (Fig. 4B asterisks, Fig. S5B). We next mapped back the class averaged structures of lamin filaments to their original coordinates in the ghost nuclei. This produced a set of reconstituted lamin filaments resolved to higher resolution and with higher contrast (Fig. 4C). Two protofilaments compose the mature filaments interacting and crossing each other frequently but not in a uniform manner (Fig. 4C, asterisks). These substructures resembled in shape and dimensions the head-to-tail polymer of dimers. The presence of two polymer structures within lamin filaments and their heterogenous appearance may increase the spectrum of conformations that can be adopted by the mature filaments. This emphasizes the possible variations within the structure. The interactions between the two head-to-tail polymers may vary to allow tight interactions as well as more scarce appearance, providing additional capability for lamin filaments to fulfill various functions. Resolving the 3D structure of the lamin filaments, together with their Ig-like fold domains, could provide additional information on lamin assembly and the differences between A- and B-type lamin filaments. However, a larger data set would be needed in addition to an innovative image processing approach.
Discussion
Based on our novel data, we have generated a model of the nuclear envelope and lamin organization in wild type and LmnaH222P/H222P MEFs (Fig. 5). Nuclei of LmnaH222P/H222P MEFs have a larger 2D surface area and less densely packed heterochromatin and likely chromatin-associated factors. While the lamin filaments of the lamina of LmnaH222P/H222P MEFs are structurally similar to wild type lamin filaments, they are better visualized by cryo-ET, presumably because of decreased interactions with the densely-packed chromatin at the nuclear periphery. The contrast of the lamin filaments in nuclei of LmnaH222P/H222P MEFs is substantially higher than the filaments previously observed in wild type vimentin-null MEFs (Turgay et al., 2017). This provides us with a unique system to obtain a better resolved view of lamin filaments.
Our results suggest that the distribution of lamin filaments within the lamina of the LmnaH222P/H222P MEFs is presumably unaffected. The lamina of most somatic cells is composed of both A- and B-type lamins, although in LmnaH222P/H222P fibroblasts, we found that B-type lamin expression was upregulated relative to wild type fibroblasts. We analyzed a large number of lamin filaments, >950,000 filament segments form 233 tomograms, ensuring with a high statistical confidence the close structural resemblance of lamin A/C H222P filaments to the wild type ones. The H222P amino acid substitution in lamin A/C is situated in a linker domain, between coil 1 and 2; therefore, this region presumably can accommodate the mutation without a major structural alteration.
The assembly of lamin dimers into polymers is central for understanding the basic characteristics of the filaments. A recent publication suggested that lamins interact laterally through coil 1 and 2 to form a tetrameric structure (Ahn et al., 2019). This is similar to cytoplasmic IFs (Herrmann and Aebi, 2004). Our experiments revealed the existence of two protofilaments ~1.8 nm in thickness, which have the dimensions of a dimeric coiled-coil structure (Zaccai et al., 2011). This observation supports the notion that two head-to-tail filaments are the basic lamin components that interact to form the mature filaments (Stuurman et al., 1998). Moreover, our in silico reconstituted filaments provide evidence that the interactions between the two head-to-tail polymers of dimers are not structurally homogeneous (Fig. 5 arrowheads), allowing lamin filaments to adopt a spectrum of conformations that presumably crucial to their flexibility and the required changes upon the introduction of external forces (Cho et al., 2019; Maurer and Lammerding, 2019; Sapra et al., 2019).
Pathogenic LMNA mutations leading to many different amino acid substitutions along the lamin A/C proteins cause striated muscle disease (Briand et al., 2018; Cattin et al., 2013; Maggi et al., 2016; Perrot et al., 2009; Tatli and Medalia, 2018). These mutations appear to result in a loss of some aspect of lamin A/C function, as human patients with LMNA mutations leading to haploinsufficiency and Lmna null mice both develop striated muscle disease (Bonne et al., 1999; Sullivan et al., 1999). However, how these lamin A/C amino acid substitutions cause cardiomyopathy and muscular dystrophy is poorly understood. Among several hypotheses are that the pathogenic lamin A/C variants lead to altered chromatin organization, gene expression and cellular mechanotransduction (Osmanagic-Myers and Foisner, 2019; Schreiber and Kennedy, 2013). Our results provide structural insights into how a striated muscle disease-causing lamin A/C variant affects nuclear structure. In LmnaH222P/H222P MEFs, the overall organization of lamins in the nuclear lamina appears unaffected by the amino acid substitution. However, nuclei are slightly larger with alterations at the lamina-chromatin interface. How these alterations affect the mechanical properties of the nucleus, signaling cascades or gene expression remains to be determined. The normal lamina structure at the resolution of cryo-ET also cannot explain the alterations in nuclear structure observed at the light microscope level in fibroblasts of human subjects with striated muscle disease-causing LMNA mutations and transfected cultured cells expressing the pathogenic lamin A variants (Muchir et al., 2004; Ostlund et al., 2001; Raharjo et al., 2001).
Using cryo-ET to examine additional cells expressing different lamin A/C variants could lead to a better understanding of the structure of nuclear lamins and filament assembly. Combining these studies with in vitro assembly assays of lamin filaments will lead to a higher-resolution structural resolution of lamins and lamina formation. Examination of cells from mice and human subjects with other mutations causing striated muscle disease and other laminopathies, such as partial lipodystrophy, peripheral neuropathy or progeria, could also provide further insights into pathogenic mechanisms. Furthermore, muscular dystrophy and cardiomyopathy caused by LMNA mutations in humans is autosomal dominant. Therefore, the lamin A/C filaments or dimers in the cells of human patients could be composed of both wild type and variant proteins. The autosomal-dominant nature of most human laminopathies would therefore provide an additional challenge in scrutinizing the effects of the disease-causing variants on lamina structure and pathogenic mechanisms.
Materials and methods
Preparation and immortalization of MEFs
MEFs were isolated from E14-E15 embryos generated by crosses between LmnaH222P/+ male and female mice. Briefly, each embryo was cut into fine pieces and incubated at 37°C for 15 min in 1 ml 0.25% trypsin (Gibco, 25200-056). The tissue pieces were then sheared in an 18 gauge needle attached to a syringe and trypsin was inactivated by addition of DMEM supplemented with high glucose, sodium pyruvate, GlutaMAX, and phenol red (Gibco, 10569-010) and containing 15% (v/v) fetal bovine serum (Gibco, 26140-079). Cells were then plated in 10 cm culture dishes and allowed to adhere for 24 h. Non-adherent cells were discarded and the adherent fraction were the MEFs. To establish the immortalized lines, MEFs at passage 2 were infected with a packaged retrovirus, which expresses SV40 large T antigen (a gift from Drs. Eros Lazzerini Denchi and Larry Gerace). Stable immortalized MEF pools were established by selecting the infected cells in 400 μg/ml geneticin (Gibco, 10131035). Genotypes of the embryos and MEFs were confirmed by PCR as previously described (Arimura et al., 2005). The Institutional Animal Care and Use Committee at Columbia University Irving Medical Center approved the protocol.
Cell culture
Immortalized wild type and LmnaH222P/H222P MEFs were cultured in DMEM (Sigma-Aldrich, D5671) supplemented with 10% (v/v) FCS (Sigma-Aldrich, F7524), 1% penicillin-streptomycin (Sigma-Aldrich, P0781), 2 mM l-glutamine (Sigma-Aldrich, G7513), and 400 ug/ml geneticin at 37 °C and 5% CO2 in a humidified incubator. Confluent cells were trypsinized (Sigma-Aldrich, T4174) and seeded onto a glow discharged holey carbon covered EM-grids (R 2/1, Au 200, Quantifoil). After approximately 16 h, the grids were subject to the nuclease treatment or they were directly plunged frozen into liquid nitrogen cooled ethane for FIB-milling.
Knockdown of vimentin in MEFs
Transfection using the committal vimentin shRNAi lentiviral vector (Sigma-Aldrich) and the PolyPlus protocol (Jetprime) yielded 60-80% vimentin knockdown. At 24 h post-transfection, the medium was replaced with DMEM supplement with high glucose, 2 mM l-glutamine, 1% penicillin-streptomycin, 15% (v/v) FCS, sodium pyruvate, 400 μg/ml geneticin and 4 μg/ml puromycin. All untransfected cells died upon the addition of the medium. The transfected cells were expanded and analysed by immunoblotting.
Quantitative immunofluorescence microscopy
MEFs were seeded onto glass cover slips coated with fibrinogen (50 μg/ml, Sigma-Aldrich, 341576) for 3.5 h in 10% (v/v) FCS. Cells were synchronized by serum starvation for 16 h in 1% FCS. Cells were then fixed in 4% (w/v) paraformaldehyde (Sigma-Aldrich, 16005) for 10 min before being permeabilized in 0.1% Triton X-100 (Sigma-Aldrich, T8787) for 10 min. Permeabilized cells were incubated in a blocking buffer (2 % (w/v) BSA, 22.5 mg/ml glycine in PBS with 0.1 % Tween-20 (PBST)) for 1 h at 25 °C. The cells were incubated in the respective primary antibodies, diluted in blocking buffer, for 1 h at 25 °C. The following primary antibodies were used: lamin A/C (Santa Cruz, sc-376248, 1:100), lamin B1 (Santa Cruz, sc-6217, 1:200), H3K9me3 (Abcam, ab8898, 1:700), H3K27ac (Abcam, ab4729, 1:350). After a 3x 5min wash in PBST, the cells were incubated with the secondary antibodies, diluted in blocking buffer for 1 h at 25 °C. The following secondary antibodies were used: Alexa Fluor 488 donkey anti-goat (Jackson Immuno Research 705-545-003, 1:400), Cy3 donkey anti-rabbit (Jackson Immuno Research 711-165-152, 1:400), Cy3 donkey anti-mouse (Jackson Immuno Research 715-165-150, 1:400). Immunostained cells were washed 3x for 5 min in PBST and 3x for 5 min in PBS before being incubated with Hoechst 33342 (Sigma-Aldrich, B2261) for 15 min at 25 °C. After a final wash 3x for 5 min in PBS, the cover slips were mounted on glass slides with Dako mounting medium (Agilent, S3023), and sealed with nail polish.
Cells were analysed with an automated inverted microscope (Leica Microsystems, DMI4000 B) equipped with a fluorescence lamp and a (Leica Microsystems, DFC365 FX) monochromatic digital camera. Images for quantitative analysis were acquired with a 20x air objective (Leica Microsystems, HCX PL Floutar 20x/0.4). Image analysis was done in ImageJ (Schneider et al., 2012). Threshold of the image was applied manually in the Hoechst channel. Next, the image was used to define the nuclei, their area, and the average signal intensity of each channel obtained with the “Analyse Particles” function. The measured signal intensities for each channel was normalized to the median of the corresponding wild-type and the measured area of the nucleus. For TIRF microscopy, the slides were prepared the same way as described above. TIRF microscopy was performed on an inverted widefield microscope (Leica Microsystems, DMI6000B) equipped with an Andor iXon Ultra 897 EMCCD camera. TIRF images were acquired with an 160x oil objective (Leica Microsystems, HC PL APO 160x/1.43 OIL) and with a penetration depth of 90 nm.
Immunoblotting
For immunoblotting, MEFs were lysed in 1x cell lysis buffer (Cell Signalling Technology, 9803) with proteinase inhibitor cocktail (Sigma-Aldrich, P8340) and 1mM PMSF (Sigma-Aldrich, 93482-50ML-F). Proteins in the cell lysates were separated by SDS-PAGE. Immunoblots were performed by using anti-lamin B1 (Cance et al., 1992) 1:1,000), anti-lamin B2 (Invitrogen, 33-2100, 1:500), anti-H3K9me3 (Abcam, ab8898, 1:1,000), anti-Lamin A/C (Santa Cruz, SC-20681, 1:5,000), anti-ɣ-tubulin(Sigma-Aldrich, T-5326, 1:1,000), and anti-Gapdh (Ambion, AM4300, 1:5,000) antibodies. Quantification of blots was performed with ImageJ, normalized to loading controls as indicated, and presented as fold change over untreated or wild type MEFs.
Preparation of ghost nuclei on EM grids
Cells growing on EM-grids were rinsed in PBS supplemented with 2 mM MgCl2, permeabilized for 15-20 s in a permeabilization buffer (1X PBS, 0.1% Triton X-100, 600 mM KCl, 10 mM MgCl2 and protease inhibitors), and rinsed again. Thereafter, grids were treated with 2.5 units/μl benzonase (Merck, 71206-3) in PBS/2mM MgCl2 for 30 min, washed again prior to applying 3 μl of fiducial gold markers (Aurion, 10 nm, BSA-coated), and then plunged frozen into liquid ethane.
Cryo-FIB-scanning electron microscopy (SEM) milling
Prior to FIB-milling, the grids were coated with 5 nm Pt/C by using a Leica BAF060 system cooled to −160° C. The grids were transferred to a Zeiss Auriga 40 Crossbeam FIB-SEM. Using the gas injection system, an organometallic platinum protective layer was applied to the grids. Cells were milled with a focused gallium ion beam at a stage temperature of <-150 °C and a stage angle of 18°. The milling was controlled by the NanoPatterning and Visualization Engine software (Zeiss) and observed by SEM. Final thickness of FIB-milled lamellas were 100-200 nm.
Cryo-ET acquisition
Tilt series were acquired of FIB-milled lamellas and ghost nuclei using an FEI Titan Krios at 300 KeV electron microscope equipped with a Gatan Quantum Energy Filter and a K2 Summit direct electron detection camera. All tilt series were acquired at a magnification of 64,000x and 4-6 μm underfocus with SerialEM (Mastronarde, 2005), resulting in a 0.22 nm/pixel at the specimen level. The data covered an angular range of −60° to 60°, acquired from −30° to +60° followed by −30° to −60°, every 2°, with a total dose of 100 - 140 e−/Å2. Tilt series acquisition for FIB-milled lamellas were conducted by a dose symmetric tilt scheme (Hagen et al., 2017) from −60° to 60° every 3° with a total dose of 150e−/Å2.
Cryo-ET image processing
All tilt series were reconstructed by using the IMOD workflow. For visualization purposes, the tomographic slices picked for the visualization were reconstructed with SIRT algorithm and 8 z slices were projected in slicer window in IMOD. A dataset of 233 tomograms with −4 μm defocus and 120 e−/A2 were selected for image averaging. The tomograms were reconstructed again in MATLAB using TOM Toolbox. Thereafter, APT workflow (Martins et al., 2020) was followed. Ten manually segmented tomograms (Amira-Avizo 2019.1, Thermo Fischer Scientific) were used as a reference for the neural network segmentation of 233 tomograms. Evenly spaced coordinates every 5 nm along the filaments were picked to determine the centre of each segment along the filament axis. Approximately 967,000 sub-volumes (12×12×4.4 nm3) were reconstructed around the picked coordinates and projected to 2D images. The projected images were imported into Relion 3.0. After several rounds of classifications, 200 class averages with a total number of 350,000 particles were obtained. The final 2D class averages were mapped-back two or more sequential segments along a filament and were selected and mapped back to their x-y positions. Surface rendering images of wild type and LmnaH222P/H222P MEFs tomograms were generated using the Amira-Avizo 2019.1 software package (Thermo Fischer Scientific).
Authors contributions
H.J.W. and O.M. conceived the project and provide the funding. M.T. and R.T. conducted the experiments and together with M.E. analysed the data. G.B. generated the Lmna H222P mice. W.W. bred and cared for mice, generated immortalized MEFs and performed immunoblotting experiments. J-Y. S. generated immortalized MEFs. O.M and H.J.W wrote the manuscript with contributions from all authors.
Supplementary Materials
Acknowledgements
This work was funded by grants from the Swiss National Science Foundation Grant (SNSF 31003A_179418) and the Mäxi Foundation to O.H. and a grant from the U.S. National Institutes of Health (R01AR04897) to H.J.W. We thank the Center for Microscopy and Image Analysis at the University of Zurich (ZMB), Dr. Yagmur Turgay for technical help in the initial stages and Drs. Eros Lazzerini Denchi and Larry Gerace for reagents and advice on immortalizing MEFs. The authors declare no competing financial interests.