Vimentin regulates nuclear segmentation in neutrophils

Significance

Neutrophils, the most abundant immune cells in the body, exhibit heavily segmented nuclei, but the mechanism responsible for this subcellular morphology is incompletely understood. This study utilized cryo-electron tomography to image mouse neutrophils and identified vimentin, a cytoplasmic intermediate-filament protein, as a key protein involved in nuclear segmentation. The genetic deletion of vimentin reduced the number of nuclear lobes of neutrophils and eosinophils, mimicking the Pelger–Huët anomaly. These findings offer insights into the unique nuclear morphology of neutrophils, which may have important implications for understanding their immune functions in various pathophysiological contexts. In addition, this study highlights the potentially broad involvement of cytoplasmic intermediate-filament proteins in shaping the structure of nuclei, opening an exciting research avenue in cell biology.

Abstract

Granulocytes are indispensable for various immune responses. Unlike other cell types in the body, the nuclei of granulocytes, particularly neutrophils, are heavily segmented into multiple lobes. Although this distinct morphological feature has long been observed, the underlying mechanism remains incompletely characterized. In this study, we utilize cryo-electron tomography to examine the nuclei of mouse neutrophils, revealing the cytoplasmic enrichment of intermediate filaments on the concave regions of the nuclear envelope. Aided by expression profiling and immuno-electron microscopy, we then elucidate that the intermediate-filament protein vimentin is responsible for such perinuclear structures. Of importance, exogenously expressed vimentin in nonimmune cells is sufficient to form cytoplasmic filaments wrapping on the concave nuclear surface. Moreover, genetic deletion of the protein causes a significant reduction of the number of nuclear lobes in neutrophils and eosinophils, mimicking the hematological condition of the Pelger–Huët anomaly. These results have uncovered a new component establishing the nuclear segmentation of granulocytes.

Granulocytes, i.e., neutrophils, eosinophils, and basophils, are the most abundant immune cells, among which neutrophils account for approximately 50% of white blood cells in the body (1). Granulocytes are essential for immune defense against various pathogens, e.g., bacteria, fungi, or parasites, via cellular events of degranulation or phagocytosis (2). Also, they exert indispensable roles in tissue repair after pathological insults by clearing debris and releasing specific cytokines (3). On the other hand, granulocyte malfunctions may contribute to the disruption of tissue homeostasis leading to lethal infection, allergy, or cancer (4).

Notably, unlike other immune or nonimmune cell types, the nuclei of granulocytes, particularly neutrophils, are heavily segmented into multiple lobes during their maturation. Studies have suggested that such a unique feature of subcellular morphology is evolutionarily conserved (4, 5), which may facilitate the migration, extravasation, phagocytosis, and other immune functions of neutrophils (6–8). In addition, recent works implicated that this distinct process of nuclear segmentation is associated with chromatin remodeling or the formation of neutrophil extracellular traps (9–12).

Although the phenomenon of neutrophil nuclear segmentation has long been documented, its underlying mechanism is yet to be entirely understood. Previous studies indicated that mutations of the human lamin B receptor (LBR), an integral protein of the inner-layer nuclear membrane, cause the hyposegmentation of neutrophil nuclei, a hematological condition commonly known as the Pelger-Huët anomaly (OMIM *169400) (13). However, how the concave regions of segmented nuclei are established in neutrophils remains unclear. In particular, whether any cytoplasmic component(s) may participate in nuclear segmentation has been elusive for the past two decades.

Results

We sought to image neutrophils by cryo-electron tomography (cryo-ET) technology in the hope of identifying novel molecular component(s) involved in nuclear segmentation (SI Appendix, Fig. S1). Neutrophils were FACS (fluorescence-activated cell sorting)-sorted from C57BL/6 wild-type mice untreated or treated with lipopolysaccharides (LPS), which is known to facilitate their maturation (14). Blood-smear analyses showed that the number of nuclear lobes in neutrophils increased following the LPS treatment (Fig. 1A; untreated: 3.90 ± 0.06, LPS-treated: 4.15 ± 0.06; mean ± SEM). Also, 200-nm-thick lamellae of neutrophils prepared by cryo-focused ion beam (cryo-FIB) exhibited a similar trend of the increased number of nuclear lobes (Fig. 1B and SI Appendix, Fig. S2; untreated: 2.15 ± 0.12, LPS-treated: 2.32 ± 0.11; mean ± SEM), though the average numbers became lower due to the reduced thickness of cryo-FIB samples. We then chose the concave nuclear regions of neutrophils for tilt-series acquisition and tomographic reconstruction by cryo-ET (Fig. 1 C and E).

Fig. 1.

Molecular architecture of the nuclear segmentation of neutrophils. (A) Blood smears were prepared from C57BL/6 wild-type mice without (−) or with (+) the LPS treatment. The number of nuclear lobes in each neutrophil was counted (200 cells per condition), and the percentage of neutrophils with 2, 3, 4, 5, or 6 nuclear lobes (L2, L3, L4, L5, or L6) was calculated. (B–G) Neutrophils FACS-sorted from C57BL/6 wild-type mice were prepared for cryo-ET analyses. (B) The number of nuclear lobes in each neutrophil observed in cryo-ET lamellae was counted (48 cells for LPS- condition; 59 cells for LPS+ condition). The percentage of neutrophils with 1, 2, 3, 4, or 5 nuclear lobes (L1, L2, L3, L4, or L5) was calculated. (C and E) Representative cryo-EM images of lamellae showing neutrophils with segmented nuclei. Boundaries of the nuclear envelope are denoted by white dashed lines, with the nuclei (nuc) highlighted in purple. Green boxes indicate the regions targeted for tomogram collections in D and F. Asterisks denote the positions of ice on lamellae. (Scale bars, 2 µm.) (D and F) Representative cryo-ET tomographic slices of the concave nuclear regions of neutrophils. The nuclear envelope (NE), cytoplasmic granules, and nuclear pore complex (NPC) are highlighted in different colors. (Scale bars, 200 nm.) Insets exhibit the high-magnification side view or cross-section view of perinuclear filamentous structures in the regions indicated by orange boxes. (Scale bars, 50 nm.) (G) Three-dimensional (3D) rendering of the area shown in F, with identifiable subcellular structures annotated in different colors. (Scale bar, 200 nm.)

A total of 49 cryo-ET tomograms of the concave nuclear regions of neutrophils were obtained. 3D reconstruction enabled the visualization of subcellular structures at molecular levels (Fig. 1 D and F and Movie S1). A plethora of membraned organelles corresponding to granules and secretory vesicles were clearly identified, and nuclear pore complexes (NPCs) on the nuclear envelope (NE) could be observed. Also, microtubules, actin filaments, glycogens, and macromolecules such as ribosomes and proteasomes were visualized in the cytosol (Fig. 1G), together validating the high quality of the cryo-ET tomograms. It is worth noting that no lamin filaments were detected underneath the inner-layer nuclear membrane, though such structures could be discernable by cryo-ET as previously reported (15), which might reflect the reduced expression of lamin proteins during neutrophil maturation (16).

Significantly, in most of the cryo-ET tomograms of neutrophils (32 out of 49; Fig. 1 D and F), we observed the cytoplasmic enrichment of filamentous structures on the concave regions of nuclei, implying their potential involvement in nuclear segmentation. We thus computationally segmented such filamentous structures and their adjacent nuclear membranes for detailed analyses (Materials and Methods). The number of filaments within each perinuclear bundle varied between 4 and 39, and the filaments appeared parallel to each other and traversed without any branching through each tomogram (Fig. 2A and SI Appendix, Fig. S3). Of importance, we noticed that the bending directions of the NEs and filaments were highly correlated, especially those in the closest proximity (Fig. 2A and SI Appendix, Fig. S3). Moreover, the distance between outer-layer and inner-layer nuclear membranes was always compressed in the regions with higher membrane curvature (Fig. 2 B–E and SI Appendix, Fig. S4), suggesting the possibility that a contractile force originating from the cytoplasmic side might contribute to the formation of concave nuclear surface.

Fig. 2.

Cryo-ET topological analyses of perinuclear filaments in neutrophils. (A) Representative 3D renderings of the concave nuclear regions of neutrophils. The NEs are presented as transparent gray surfaces. The overall shapes of NE and its nearest filament are highlighted by black dashed lines. Each perinuclear filament is colored based on its nearest distance from the outer-layer membrane of the NE. The Inset model depicts a circular bundle of filaments (yellow) wrapping the concave region of the NE (transparent gray), with the purple box and the green arrow denoting the area and the perspective shown in A. (Scale bars, 50 nm.) (B–D) Three orthogonal views of 3D renderings exhibited in A. The inner-layer membranes of the NE are colored according to their distance from the outer-layer membrane. (Scale bars, 50 nm.) (E) The distance from the outer-layer membrane (purple lines) and the curvature (gray lines) of the inner-layer membranes shown in B–D are calculated. The plotting range of each sample includes ±0.5 µm from the most curved point of the inner-layer membrane, and the shadow of lines represents ±1 SD.

We next examined the molecular identity of those perinuclear filaments. We extracted a total of 34,248 subtomograms from 410 filaments, with a box size of 60 cubic pixels. Those subtomograms were then projected onto the plane of each filament axis for two-dimensional (2D) classification. The resulting averages revealed the tube-like structures with varying diameters of approximately 10 nm (Fig. 3 A and B), reminiscent of the reported structural characteristics of cytoplasmic intermediate filaments (IF) (17, 18). For most of the averages, luminal densities were observed (Fig. 3A), which corroborated the cross-sectional views made from the raw tomograms (Fig. 1 F, Inset). We utilized the subtomograms of the most distributed class for 3D reconstruction, which produced a cylindrical tube with an internally dense core (Fig. 3C). This structure was again consistent with recently resolved intermediate-filament structures (17, 18).

Fig. 3.

Vimentin filaments wrap the concave nuclear regions of neutrophils and nonimmune cells. (A–C) Filamentous structures on the concave nuclear regions of neutrophils are IFs. (A) Class averages of cryo-ET density projections reveal tube-like filament structures with varying diameters and luminal density. The proportion of each class average is labeled as indicated. The white dashed line indicates the orientation of profile statistics shown in B. (B) Radial profiles of the filament structures are plotted, with the line colors corresponding to the dot colors assigned in A. The average diameter of filaments is calculated. (C) 3D reconstruction of subtomograms of the first (i.e., most distributed) class of filaments in A, with the side view (Left) and the clip view (Right) shown. (D) Expression profiling of intermediate-filament genes in the neutrophils of C57BL/6 wild-type mice without or with the LPS treatment by RNA-seq analyses. FPKM, fragments per kilobase per million. (E and F) Cytoplasmic enrichment of vimentin in the concave nuclear regions of neutrophils. (E) Representative image of immuno-electron microscopy. The white box denotes the area shown at high magnification in F. (G–I) Vimentin is sufficient to form cytoplasmic filaments wrapping the concave nuclear surface in nonimmune cells. Green Fluorescent Protein (GFP)-labeled vimentin was exogenously expressed in Hela cells, and the NE of cells was simultaneously marked by mCherry-labeled lamin A. (G) Representative image of fluorescence microscopy. The yellow box indicates the area targeted for the lamella preparation of cryo-CLEM analyses. (H) Representative overlay of fluorescence light microscopy and electron microscopy images by cryo-CLEM. The blue box denotes the region targeted to collect a tomogram. (I) Representative tomographic slice showing the bundle of filaments in the concave nuclear region. The NE, NPC, nucleus (nuc), and IF are highlighted by different colors. The Inset shows the side view (Left) and the clip view (Right) of 3D-reconstructed filaments.

To determine the precise protein identity of those IFs enriched on the concave nuclear regions of neutrophils, we performed the expression profiling of neutrophils FACS-sorted from the untreated and LPS-treated mice. In particular, we focused on the expression levels of intermediate-filament genes (Fig. 3D). The cytoplasmic intermediate-filament protein vimentin (Vim) exhibited the highest expression among all the genes examined without or with the LPS stimuli. In addition, the nuclear intermediate-filament protein lamin B1 appeared as the second-highest expressed. This observation implicated vimentin being responsible for the perinuclear filament structures observed in the cryo-ET analyses. Therefore, we conducted the conventional immuno-electron microscopy of mouse neutrophils using anti-vimentin antibody. The gold particles were primarily concentrated in the concave nuclear regions where the perinuclear filaments were identified (Fig. 3 E and F and SI Appendix, Fig. S5), confirming vimentin as the molecular component of such filament structures.

To further support the involvement of vimentin in nuclear segmentation, we expressed GFP-tagged vimentin in HeLa cells that normally do not have segmented nuclei. The exogenous expression of vimentin led to a significant appearance of concave nuclear morphology (SI Appendix, Fig. S6 A–D). We then employed cryo-correlative light and electron microscopy (cryo-CLEM) technology to target those concave regions for tomographic data acquisition (SI Appendix, Fig. S1). GFP-tagged vimentin evidently assembled a circular bundle on the concave nuclear surface in HeLa cells (Fig. 3 G and H and SI Appendix, Fig. S6). Moreover, the cryo-CLEM analyses detected the cytoplasmic enrichment of perinuclear filaments similar to those in neutrophils (Figs. 1D and 3I and SI Appendix, Fig. S6 F–M). The detailed structural reconstruction of those filaments gave rise to the identical features obtained from neutrophil samples as above (Fig. 3 C and I, Inset). These results proved that vimentin is sufficient to form the filament structures wrapping the concave nuclear surface in nonimmune cells.

Finally, we determined the impact of the genetic deletion of vimentin (Vim^−/−) on the nuclear segmentation of neutrophils (SI Appendix, Fig. S7 A–C). Vim^−/− mice were viable and exhibited no significant defect in organ development or overall growth. FACS analyses showed that the cell numbers of neutrophils, eosinophils, and basophils were comparable in the blood samples of Vim^−/− and control Vim^+/+ littermates (Fig. 4A), suggesting that vimentin is dispensable for the differentiation of granulocytes. In addition, the blood counts of monocytes, natural killer (NK) cells, B cells, and CD4⁺ T cells were unaffected by the vimentin deletion (Fig. 4A). On the other hand, we consistently observed fewer CD8⁺ T cells in the blood of Vim^−/− mice (Fig. 4A), although this putative role of vimentin in adaptative immunity awaits future investigations. We then imaged the neutrophils FACS-sorted from Vim^−/− mice by the cryo-ET workflow. Filament structures were entirely lost at the nuclear concave regions of Vim^−/− neutrophils (12 out of 12 cryo-ET tomograms; Fig. 4 B–D and SI Appendix, Fig. S7), verifying the essential role of vimentin in establishing such perinuclear structures. We further examined the neutrophils of Vim^−/− and control Vim^+/+ littermates by blood smears. Of importance, the number of nuclear lobes in Vim^−/− neutrophils markedly decreased compared to those of Vim^+/+ cells (Fig. 4 E and F; Vim^+/+: 3.94 ± 0.05, Vim^−/−: 3.09 ± 0.06; mean ± SEM; P < 0.001). We also assessed eosinophils in parallel in the blood smears of Vim^−/− mice. Similar to that occurring in neutrophils, there were fewer nuclear lobes in Vim^−/− eosinophils than those in Vim^+/+ counterparts (Fig. 4G; Vim^+/+: 1.97 ± 0.05, Vim^−/−: 1.78 ± 0.05; mean ± SEM; P = 0.010). These results showed that the genetic deletion of vimentin in mice recapitulates the central phenotype of the Pelger–Huët anomaly.

Fig. 4.

Genetic deletion of vimentin impairs the nuclear segmentation of neutrophils. (A) The genetic deletion of vimentin in mice does not affect the overall differentiation of granulocytes. Immune cell types in the blood of Vim^−/− or control Vim^+/+ littermates were examined by FACS analyses. mean ± SEM; n.s., not significant; *P = 0.012. (B) A representative cryo-EM image of lamella showing neutrophil with segmented nucleus. Boundary of the NE is denoted by white dashed lines, with the nucleus highlighted in purple. Green box indicates the region targeted for tomogram collection in C. Asterisk denotes the position of ice on the lamella. (C) Cryo-ET tomographic slice of the concave nuclear region of neutrophil. The NE is highlighted in red. Inset exhibits the high-magnification image of the orange boxed region. (D) 3D rendering of the area shown in C, with subcellular structures annotated in different colors. (E and F) Vimentin deletion impairs the nuclear segmentation of neutrophils. (E) Representative blood-smear images of the neutrophils of Vim^−/− or control Vim^+/+ mice. (F) The number of nuclear lobes in each neutrophil was counted (200 cells per condition), and the percentage of neutrophils with 2, 3, 4, or ≥5 nuclear lobes (L2, L3, L4, or L5+) was calculated. Vim^+/+: 3.94 ± 0.05, Vim^−/−: 3.09 ± 0.06; mean ± SEM; P < 0.001. (G) Vimentin deletion reduces the nuclear segmentation of eosinophils. Nuclear lobes were counted in the eosinophils of Vim^−/− or control Vim^+/+ mice (200 cells per condition) by blood smears. The percentage of eosinophils with 1, 2, 3, or 4 nuclear lobes (L1, L2, L3, or L4) was calculated. Vim^+/+: 1.97 ± 0.05, Vim^−/−: 1.78 ± 0.05; mean ± SEM; P = 0.010. [Scale bars, 2 µm (B), 200 nm (C and D), and 50 nm (C, Inset).]

Discussion

Mutations of the human LBR reduce the number of nuclear lobes in neutrophils (13), but how such an integral protein of the inner-layer membrane generates contractile forces to establish the concave nuclear morphology remains unknown. Notably, LBR is broadly expressed in a variety of mammalian cells without segmented nuclei (19, 20), suggesting the involvement of additional component(s) in the distinct morphology of granulocyte nuclei. Our findings illustrated that the exogenous expression of vimentin in nonimmune HeLa cells could be sufficient to form cytoplasmic filaments wrapping on the concave surface of otherwise unsegmented nuclei, supporting that vimentin is involved in this unique subcellular feature. However, it is worth noting that the genetic deletion of vimentin would not completely abolish nuclear segmentation in granulocytes, suggesting that this process is regulated by multiple factors.

Studies utilizing cytochalasin-B have demonstrated a minor role of actin filaments in lobulated nuclear morphology, while the involvement of microtubules is still controversial (21, 22). Among 49 tomograms that we collected at the nuclear concave regions of neutrophils, microtubules were only found within 4 tomograms. In addition, there appeared no direct interactions between microtubules and vimentin bundles, implying that microtubules are not directly involved in the process of nuclear segmentation.

During the spreading of in vitro cultured retinal pigmented epithelial cells, ring structures formed by vimentin filaments might correlate with nuclear deformation (23). A canonical view has been that the intermediate-filament network alone does not act as a contractile system (24). Instead, recent research has indicated that IF-dependent forces can be accomplished through integration with actin filaments (25–27), e.g., in the process of cell migration (28). However, we did not observe any colocalization of vimentin and actin filaments in the cryo-ET analyses of neutrophils and HeLa cells. On the other hand, the spatial correlation between vimentin filaments and the concave surface has raised the possibility that they may serve as “anchors” for a currently undefined motor molecule(s) to squeeze the NE. In addition, though there were no discernible components tethering vimentin filaments to the neutrophil NE in the current study, their existence may not be excluded due to the limitations of subtomogram averaging. The detailed mechanism of vimentin-mediated establishment of segmented nuclei warrants more future research.

In sum, this study has reported for the first time the involvement of the cytoplasmic intermediate-filament protein vimentin in the nuclear segmentation of granulocytes, which could have broad implications for their basic biology and diverse functions in pathophysiological contexts.

Materials and Methods

Mouse Information.

All the experimental procedures were performed in compliance with the protocol approved by the Institutional Animal Care and Use Committee of Peking University. Mice were maintained on the 12 h/12 h light/dark cycle (light period 7:00 am ~ 7:00 pm) at ambient temperature (22 °C ~ 25 °C) with water and standard chow diet available ad libitum.

C57BL/6 wild-type mice were purchased from Charles River International. Vim^+/− male and female mice (deletion of exon 3 ~ 7 of the gene, #KOCMP-22352-Vim-B6J-VA) on the C57BL/6J background were purchased from Cyagen Biosciences and in-house bred to obtain Vim^−/− and Vim^+/+ mice. All the experiments involved Vim^−/− mice included Vim^+/+ littermates as the control mice. Mice utilized in the experiments were 8 ~ 12 wk old.

For the LPS treatment, C57BL/6 wild-type mice were intraperitoneally injected with LPS (Sigma-Aldrich, 1 mg/ml dissolved in sterile saline) at 10 mg/kg of body weight, and tissues were harvested at 4 h post-injection.

FACS Sorting and Analyses.

Blood samples were collected from the mice of indicated conditions via tail bleeding into phosphate-buffered saline (PBS)/2% heat-inactivated fetal bovine serum (HI-FBS, Sigma-Aldrich)/10 mM ethylenediaminetetraacetic acid (EDTA) (pH 8.0). The samples were centrifuged at 500 g for 5 min and resuspended in the ammonium-chloride-potassium (ACK) buffer (Thermo Fisher Scientific) to lyse red blood cells. The cells were then centrifuged at 500 g for 5 min, resuspended in PBS/2% HI-FBS/10 mM EDTA, and stained with the intended FACS antibodies.

Spleen tissues were freshly dissected from the mice of indicated conditions and cut into small pieces on ice. The tissues were mashed through a 70-μm cell strainer in PBS/2% HI-FBS/10 mM EDTA, and cell suspensions were centrifuged at 500 g for 5 min. The cells were resuspended in the ACK buffer to lyse red blood cells and centrifuged at 500 g for 5 min. The cells were then processed through the EasySep Mouse Neutrophil Enrichment Kit (Stemcell Technologies, #19762), resuspended in PBS/2% HI-FBS/10 mM EDTA, and stained with the intended FACS antibodies.

FACS-stained neutrophils were sorted on the Beckman MoFlo Astrios EQ. Alternatively, FACS-stained immune cells were analyzed on the BD LSRFortessa, and the results were processed by FlowJo (https://www.flowjo.com).

Immune cell types were identified as follows: neutrophils (CD45⁺ CD11b⁺ Ly6G⁺ Ly6C⁻), eosinophils (CD45⁺ CD11b⁺ Siglec-F⁺ Ly6G⁻), basophils (CD45⁺ CD69⁺ CD200R3⁺ CD11c⁻), monocytes (CD45⁺ CD11b⁺ Ly6C⁺ Ly6G⁻), NK cells (CD45⁺ NK1.1⁺ CD3⁻), B cells (CD45⁺ CD19⁺ CD3⁻), CD4⁺ T cells (CD45⁺ CD3⁺ CD4⁺ CD8⁻ NK1.1⁻), and CD8⁺ T cells (CD45⁺ CD3⁺ CD8⁺ CD4⁻ NK1.1⁻).

Cell Culture, Plasmids, and Stable Line Generation.

The Flp-In™ T-REx™ HeLa cells were cultured in Dulbecco’s Modified Eagle’s Medium supplemented with 2 mM GlutaMAX-I, 10% FBS, and 100 U/mL HyClone penicillin-streptomycin at 37 °C and 5% CO₂.

The pcDNA5-GFP-Vimetin construct was generated by inserting vimentin cDNA (NCBI Reference Sequence: NM_003380.5) into the pcDNA5-GFP backbone. Plasmid mCherry-LaminA-C-18 (#55068) was ordered at ADDGENE.

The cells were seeded at 1 to 2 × 10⁵ per 6-well, and plasmid transfections were performed using Lipofectamine™ 3,000 transfection reagent. The fluorescent-positive cells were sorted with FACS into a 6-well plate. The positive cells were grown in a selection medium with G418 (200 μg/mL), hygromycin (100 μg/mL), and blasticidin (15 μg/mL). The expression of the vimentin protein was induced with tetracycline at 1 μg/mL for 24 h.

Cryo-EM Grid Preparation.

Neutrophils were FACS-sorted from C57BL/6 wild-type mice without or with the LPS treatment. The cells were centrifuged at 500 g for 5 min and resuspended in RPMI 1640 (Thermo Fisher Scientific) containing 10% HI-FBS, 100 U/mL penicillin, 100 mg/mL streptomycin, and 10% glycerol at a density of 3.5 × 10⁶ cells per mL.

The R 2/1 EM grids (Quantifoil) were glow-discharged for 60 s using the Model 950 Advanced Plasma System (GATAN) before sample preparations. Neutrophils were loaded onto the glow-discharged EM grids and vitrified using the Vitrobot Mark IV (Thermo Fisher Scientific). The grids were stored in liquid nitrogen before cryo-FIB milling.

For HeLa cells, the cells were seeded on the grids, and 4 µL blotting buffer (PBS, 10% glycerol, 0.5 mg/mL Dynabeads™ MyOne™ Carboxylic Acid) was added to the grids before blotting and vitrification with Vitrobot Mark IV (Thermo Fisher Scientific).

Correlative Cryo-Light and Electron Microscopy and Lamellae Preparation.

The vitrified neutrophil samples were milled by the Aquilos 2 Cryo-FIB System (Thermo Fisher Scientific) to achieve the thickness requirement for high-quality tomographic data collection as described (29). Organometallic platinum deposition via GIS was first performed for 30 s, followed by sputtering with 30 mA current for 15 s to protect the samples during milling process. After applying the Thermo Scientific AutoTEM software for automatic milling with the current from 1 nA to 100 pA, lamellae were then manually polished to make their thickness less than 200 nm with the current from 50 pA to 10 pA. Finally, sputtering was performed with 7 mA current for 7 s to reduce charging during tilt series collection. The resulting lamellae were stored in liquid nitrogen before cryo-ET imaging.

The cryo-CLEM and cryo-FIB milling for HeLa cells were conducted as described. Briefly, the grids were imaged with a Leica SP8 cryo-confocal microscope for eGFP-Vimentin (488 nm excitation) and mCherry-LaminA-C (552 nm excitation). The deconvolved images with HUYGENS (Huygens Professional version 21.10, Scientific Volume Imaging) were employed for correlation with the SEM image from Aquilos 2 cryo-FIB system with 3D-Correlation Toolbox (30). FIB milling was performed at the indicated positions with a series current beam.

Cryo-ET Data Collection.

Tilt series of targeted regions in lamellae were collected with a 300 kV Titan Krios transmission electron microscope equipped with a K3 camera and a Gatan energy filter. Images were collected using dose-symmetric tilt-scheme (31) from −60° to 60° (relative to the pre-tilt angle) with 2° or 3° increments with SerialEM software (32). The defocus was set between −4.0 and −8.0 µm. The total electron dose for each tilt series was limited to ~110 e⁻/ Ų. The super-resolution frames (10 frames per tilt) of each tilt were motion-corrected and 2× binned using MotionCor2 software (33) to obtain tilt series with a pixel size of 3.33 Å for neutrophil and 2.24 Å for HeLa cells. Tilt series were aligned using patch-tracking mode, and tomograms were reconstructed by back projection by IMOD (34). For further segmentation and particle picking, tomograms were downscaled by a binning factor of 4, and deconvolution was performed in WARP (35).

Segmentation and Subtomogram Averaging.

The binned tomograms were first processed using IsoNet 0.2 software (36) to improve the following performance of segmentation. The NE was automatically segmented using a tensor voting-based method (37) and then manually refined in Amira 2020.3 software.

IFs were traced by Amira 2020.3 software using Cylinder Correlation tools (38). A hollow cylinder model with a length of 60 nm, an outer radius of 5 nm, and an inner radius of 2.6 nm was generated as a reference. A total of 34,248 subtomograms (2× binned) with a box size of 60 cubic pixels were cropped along the traced filaments with 4 nm space using helical tools in RELION 2.1 (39) and WARP. The prior information of Euler angle was kept for further alignment.

For the analyses of filament diameters, filaments were first rotated to be parallel to XY plane according to the prior information using MATLAB based TOM toolbox (40). The resulting subtomograms were projected along Z axis with a projection thickness of 30 pixels. The generated 2D images were aligned and classified by the 2D classification in RELION 2.1 (39). Six classes with clear filament features were selected for the diameter analyses by plotting mean gray values as line charts, while particles from poor averaged classes were discarded.

For the 3D reconstruction of IFs, 4,519 particles from the most distributed class were selected for 3D auto-refine in RELION 2.1 (39).

Distance Analysis and Statistics.

The distance between IF-NE or IF-IF was calculated based on segmentation results obtained above. For IFs, points were picked every 4 nm for the calculation. The distance between points and nearest segmentation of the outer-layer membrane of NE was defined as the IF–NE distance. The distance between the nearest points of different filaments was defined as the IF–IF distance.

For NE, each pixel of the inner-layer membrane was selected and its closest distance to the segmentation of outer-layer membrane segmentation was defined as the distance between two-layer membranes of the NE.

A total of 20 slices of each inner-layer nuclear membrane were selected for curvature analysis and statistics. For each slice, three points were drawn on the inner nuclear membrane layer, spaced evenly apart by 50 pixels (65 nm). These three points defined a circle, with radius r. The curvature of the membrane at the middle point was estimated by taking the inverse of the circle’s radius, 1/r.

Immuno-Electron Microscopy.

Immuno-electron microscopy was performed according to the procedures described previously (41). Briefly, FACS-sorted cells were fixed with 4% paraformaldehyde and 0.2% glutaraldehyde in PBS, then washed with 150 mM glycine solution in PBS two times for 5 min each for quenching the free aldehyde. Cells were then resuspended and embedded in 12% gelatin (PanReac AppliChem, 147116.1210), solidified in ice, and cut into 1 mm³ blocks on a cold plate. The blocks were transferred to a 2.3 M sucrose solution and rotated at 4 °C overnight to prepare for cryo-sectioning. Afterwards the blocks were glued to aluminum specimen holder and frozen in liquid nitrogen. Then, 70-nm cryosections were cut at −117 °C using an ultramicrotome (Leica Microsystem FC7), and mounted on 100-mesh formvar/carbon-coated nickel grids. For single labelling, the grids were incubated with Vimentin antibody (Cell Signaling, #5741T, 1:10). After washing with PBS, secondary antibody [10-nm protein A-gold (Cell Microscopy Center, University Medical Center Utrecht, Utrecht, The Netherlands), 1:50] was applied for 30 min. After labelling, the sections were counterstained with 2% methylcellulose and 2% uranyl acetate and then examined under an electron microscope (FEI, Tecnai G2 Spirit) operating at 120 kV with a Gatan 832 CCD camera (Gatan).

Confocal Fluorescent Microscopy.

HeLa cells were seeded in μ-dishes (Ibidi company, 35 mm). pcDNA5-GFP and pcDNA5-GFP-Vimentin constructs were transiently transfected using Lipofectamine™ 3000 transfection reagent. Confocal fluorescent images were acquired with an AIRSCAN 2 microscope (Zeiss, LSM980).

RNA-seq Analyses.

Neutrophils were FACS-sorted from the spleens of C57BL/6 wild-type mice without or with the LPS treatment. The cells were centrifuged at 500 g for 5 min and then homogenized in the TRIzol Reagent (Thermo Fisher Scientific). Total RNAs were extracted by the RNeasy Mini Kit (Qiagen) and processed for single-end RNA-seq analyses by the Beijing Genomics Institute. Gene expression levels were normalized as fragments per kilobase per million.

Blood Smears.

Blood samples were collected from four Vim^−/− male mice and four Vim^+/+ male littermates of 8 to 12 wk old via tail bleeding. Blood smears were immediately prepared on glass slides and stained with the Giemsa Stain Kit (Solarbio, #G4640). The stained samples were scanned under a bright field by Zeiss Axio Scan Z1 equipped with a 20× objective. Fifty cells per mouse (a total of 200 cells per genotype) were evaluated, and the number of nuclear lobes in each cell was manually counted.

Statistical Methods.

Student’s t test (unpaired, two-tailed) was performed using GraphPad Prism (http://www.graphpad.com/scientific-software/prism). Statistical details of the experiments are included in figure legends.

Supplementary Material

Appendix 01 (PDF)

Movie S1.

Molecular architecture of the concave nuclear region of neutrophil. Related to Figure 1.

Data, Materials, and Software Availability

The code used for double-membrane spacing analysis has been deposited to GitHub: github.com/GuoQLabPKU/Double-membrane_spacing (42). The RNA-seq data have been deposited to the BioSample Database (https://www.ncbi.nlm.nih.gov/sra) with the accession numbers SRR23974780 (43), SRR23974781 (44), SRR23974782 (45), and SRR23974783 (46). Representative cryo-ET tomogram has been deposited to EMDB with accession number EMD-35979 (47), the corresponding raw data and segmentation have been deposited to EMPIAR with accession number EMPIAR-11709 (48).