Bicaudal D2, dynein, and kinesin-1 associate with nuclear pore complexes and regulate centrosome and nuclear positioning during mitotic entry

Abstract

BICD2 is one of the two mammalian homologues of the Drosophila Bicaudal D, an evolutionarily conserved adaptor between microtubule motors and their cargo that was previously shown to link vesicles and mRNP complexes to the dynein motor. Here, we identified a G2-specific role for BICD2 in the relative positioning of the nucleus and centrosomes in dividing cells. By combining mass spectrometry, biochemical and cell biological approaches, we show that the nuclear pore complex (NPC) component RanBP2 directly binds to BICD2 and recruits it to NPCs specifically in G2 phase of the cell cycle. BICD2, in turn, recruits dynein-dynactin to NPCs and as such is needed to keep centrosomes closely tethered to the nucleus prior to mitotic entry. When dynein function is suppressed by RNA interference-mediated depletion or antibody microinjection, centrosomes and nuclei are actively pushed apart in late G2 and we show that this is due to the action of kinesin-1. Surprisingly, depletion of BICD2 inhibits both dynein and kinesin-1-dependent movements of the nucleus and cytoplasmic NPCs, demonstrating that BICD2 is needed not only for the dynein function at the nuclear pores but also for the antagonistic activity of kinesin-1. Our study demonstrates that the nucleus is subject to opposing activities of dynein and kinesin-1 motors and that BICD2 contributes to nuclear and centrosomal positioning prior to mitotic entry through regulation of both dynein and kinesin-1.

Figure 1. BICD2 interacts directly with RanBP2.

(A) Schematic representation of the domains of RanBP2 and BICD2 and their fragments used for binding studies. (B) Co-IP of endogenous BICD2 and RanBP2 from HeLa cells. IPs were performed with antibodies against BICD2 or the control IgG and analyzed by Western blotting with the indicated antibodies. 2.5% of the input was loaded on gel. RanBP2, but not the negative control, GM130, is co-precipitated with BICD2. (C) Co-IP of endogenous RanBP2 and BICD2 from nocodazole-arrested HeLa cells. IPs were performed with antibodies against RanBP2 or the control IgG and analyzed by Western blotting with the indicated antibodies. 3% of the input was loaded on gel. BICD2 is co-precipitated with RanBP2, while its close homologue BICD1 and another dynactin-binding protein, CLIP-170, are not. (D) Co-IPs from HEK293 cells co-transfected with CFP-tagged RanBP2 fragments or GFP-RanGAP1 and mCherry-fused BICD2-CT. IPs were performed using mouse anti-GFP antibodies and analyzed by Western blotting using rabbit anti-GFP or anti-BICD2 antibodies. BICD2-CT and the IgG bands are indicated by an arrowhead and an arrow. BICD2-CT is coprecipitated with RanBP2 fragment 3, but not with other RanBP2 fragments or RanGAP1. (E) HIS-tagged BICD2-CT, GST, and GST fusions of RanBP2 fragments 3 and 4 were purified from E. coli and analyzed by SDS-polyacrilamide gel electrophoresis and Coomassie staining. (F) GST pull-down assays with the indicated RanBP2 fusions and purified HIS-tagged BICD2-CT, which was detected by Western blotting with anti-HIS tag antibodies. 10% of the input and the bound fractions were loaded on gel. Using 0.8 µM BICD2-CT and 1.4 µM GST-RanBP2 fragment 3, an almost quantitative binding was observed. This experiment confirms a direct interaction between RanBP2 fragment 3 and BICD2-CT. (G) Yeast two-hybrid analysis. BICD2 fragments were linked to GAL4 activation domain and tested in a pairwise fashion for interaction with RanBP2, GTP-bound Rab6A (Q72R), kinesin-1 (KIF5A), and kinesin-4 (KIF21B) tail regions cloned into LexA fusion vector. Interaction strength was scored according to the time needed for β-galactosidase reporter to generate visible blue-colored yeast colonies on X-Gal containing filters: +++0–30 min, ++30–60 min, +60–180 min and - no β-galactosidase activity. Similar to the previously characterized BICD2 partner Rab6, RanBP2 fragment 3 (2147–2287) specifically interacts with BICD2-CT, but not with other parts of BICD2. (H) HeLa cells were co-transfected with the indicated plasma membrane targeted CFP-RanBP2 fusions and HA-tagged BICD2-CT, fixed with paraformaldehyde, and stained with anti-HA antibodies. CFP fluorescence was visualized directly. CFP-tagged palmitoylated RanBP2 fragments are present on the plasma membrane (upper panels), while BICD2-CT (upper panels) localizes to the Golgi and cytoplasmic vesicles (middle panels). RanBP2 fragment 3, but not the other RanBP2 fragments, recruits BICD2-CT to the plasma membrane (see overlay in the bottom panels).

Figure 2. BICD2 associates with the NE and AL in G2 phase in a RanBP2-dependent manner.

(A) HeLa cells were transfected with a control siRNA, fixed with cold methanol followed by paraformaldehyde 3 d later, and stained for endogenous BICD2, RanBP2, and cyclin B1. Note that in the cyclin B1-positive cell, BICD2 and RanBP2 co-localize at the NE and at a perinuclear spot that is an accumulation of AL. (B) HeLa cells were treated with 10 µM nocodazole for 1 h and stained as described for (A). Note that BICD2 strongly stains the NE and cytoplasmic NPC-positive dots (AL) in the cyclin B1-positive but not the cyclin B1 negative cells. (C) HeLa cells that were either treated with 10 µM nocodazole or transfected with control siRNAs or a mixture of RanBP2 siRNAs #1 and #2 were stained as described for (A), and the percentage of cyclin B1-positive cells showing BICD2 accumulation at the NE was counted. In case of RanBP2 knockdown, only the cells in which RanBP2-specific nuclear staining was reduced to background levels were included in the quantification. Error bars represent SD; the number of experiments for each condition is shown in parentheses below the graph; ∼40–100 cells were counted per experiment. (D) Fluorescence intensity ratio between NE and the adjacent cytoplasm in cyclin B1 positive and negative cells (∼50 cells per measurement); cells were either untreated or treated with 10 µM nocodazole for 1 h. Measurements were performed in a region of 0.3 by 30 µm at the border between cytoplasm and the nucleus (NE) and in a directly adjacent region (cytoplasm). Error bars represent SEM. (E) HeLa cells were transfected with a mixture of RanBP2 siRNAs #1 and #2, fixed, and stained as described for (A). Note that the two cyclin B1-positive cells that have strongly reduced RanBP2 levels show no accumulation of BICD2 at the NE.

Figure 3. BICD2 to localizes to the nuclear pores at the NE.

(A) HeLa cells were fixed with cold methanol and stained with antibodies against BICD2 and nucleoporins (MAB414). These panels show focal planes in the middle part of the nuclei. (B) Enlargements of the flattened NE surface from the same cells as in (A) (the area from which the enlarged image is taken is indicated in (A) by a white rectangle). The insets show further enlargements, in which individual NPC can be distinguished. Colors used for the overlays are indicated above the corresponding images. Note that most NPC dots display signal in both channels, indicating that BICD2 and MAB414 antigens (NPCs) co-localize. (C) Representative fluorescence intensity profile of BICD2 (green) and MAB414 staining (red) at the NE surface. The profile was obtained using Linescan function of MetaMorph at the position indicated by a blue line in (B). Vertical axes are in arbitrary units. Note that most peaks are present in both BICD2 and MAB414 channels, although their intensities often differ, in agreement with the fact that BICD2 is not a nucleoporin. (D) Correlation between BICD2 and MAB414 signals in Figure 3B. The intensity of each pixel in the green and red channel is represented by a dot. (E,F) The same images and analysis as in (B) and (D), but with the MAB414 panel rotated by 180°. Note that there is no co-localization between the green and red signals and no correlation is visible in the plot. (G) Average coefficient of linear correlation between BICD2 and MAB414 signals, determined from plots such as in (B) and (D) obtained for 19 cells. Note that the coefficient is much higher for properly aligned images compared to rotated ones, indicating that co-localization is not spurious. Error bars represent SD.

Figure 4. BICD2 is required for targeting dynein/dynactin to the NE and AL.

(A–D) Control HeLa cells or HeLa cells treated with 10 µM nocodazole for 1 h were fixed with paraformaldehyde (A–C) or cold methanol (D) and stained for the indicated endogenous proteins. Dynactin is visualized with an antibody to p150^Glued and dynein with an antibody to dynein IC. Colors used for the overlays are indicated above the corresponding images. Note that BICD2 shows two types of localization: it colocalizes either with Rab6 or with RanBP2. In cells where BICD2 associates with RanBP2-positive NE and AL, dynein and dynactin are also targeted to these structures. (E) HeLa cells were transfected with a control siRNA (upper panels) or BICD2#1 siRNA (bottom panels). Three days later, cells were treated with 10 µM nocodazole for 5 h, fixed with cold methanol, and stained for endogenous RanBP2, phosphorylated histone H3, and dynein IC. NE staining by dynein antibodies is indicated by an arrow. Colors used for the overlays are indicated above the corresponding images. Note that dynein is enriched at the NE and annulate lamellae in control phospho-histone H3 positive cell, but not in BICD2-knockdown cell. (F) Percentage of HeLa cells positive for phosphorylated histone H3 that show strong accumulation of dynein IC at the RanBP2-positive NE and AL in control or BICD2-depleted cells 3 d after siRNA transfection. Only the cells with clearly visible AL were included in the quantification. Error bars represent SD; ∼25–30 cells were counted in two experiments.

Figure 5. Dynein inactivation causes rapid separation of nuclei and centrosomes in G2.

(A) U2OS cells were transfected with the control or DHC#3 siRNAs, fixed with cold methanol, and stained for α-tubulin (green in overlay) and γ-tubulin (red in overlay), as well as DNA (DAPI, blue in overlay). Note that centrosomes are located close to the nucleus in control cells but are removed far away from the nucleus in dynein-depleted cells. (B) mCherry-α-tubulin stable U2OS cell line was imaged with a 3 min time interval 2.5 d after transfection with the indicated siRNA mixtures, and the distance between centrosomes and the nucleus at the time of NEB was measured. Error bars represent SD. The number of experiments is indicated in parentheses below the graph. ∼20 cells per experiment were analyzed. (C) mCherry-α-tubulin stable U2OS cells were imaged with a 3 min interval 2.5 d after transfection with DHC#1 siRNA. Nuclei and centrosomes undergo G2-specific displacements (indicated by asterisks and arrows, respectively), resulting in their separation. (D,E) Quantification of displacement of nuclei and centrosomes starting from the moment when G2-specific enhanced nucleation of MTs at the centrosome became visible (D), and velocity of nuclear movement during 1 h before NEB (E). Nuclear movement in dynein-depleted cells was initiated 54±4 min (mean±SD) before NEB. Measurements were performed in 10 cells in 3 independent experiments. (F–H) U2OS cells stably expressing mCherry-α-tubulin were microinjected with antibodies against dynein IC (70.1) or the Myc tag. Late G2 cells were chosen based on the presence of separated centrosomes. (F) An example of nucleus-centrosome separation after microinjection with dynein-inhibiting antibodies. (G,H) Distance between the nucleus and the centrosomes (G) and percentage of cells showing strong centrosome detachment (H) at 30 min after microinjection.

Figure 6. Depletion of BICD2 and RanBP2 causes centrosome detachment in prophase U2OS cells.

(A) Western blots with the indicated antibodies were performed with equal amounts of extracts of U2OS cells 3 d after transfection with the indicated siRNAs. Note that BICD1 antibody cross-reacts with BICD2 (arrows). Both BICD proteins can be independently depleted, although BICD1#1 siRNA causes some co-depletion of BICD2. Levels of dynactin, dynein, KIF5B, and RanBP2 are not significantly affected BICD knockdown. (B) U2OS cells were transfected with BICD2#1 and BICD1#1 siRNAs (upper panel) or with GFP-BICD2-CT, fixed with cold methanol, and stained for α-tubulin and γ-tubulin, as well as DNA (DAPI). Colors used for the overlays are indicated above the corresponding images. Note centrosome detachment caused by the loss of BICD function through siRNA-mediated depletion or overexpression of the dominant negative mutant BICD2-CT. (C) The distance between the nucleus and the centrosomes in fixed prophase U2OS cells under different conditions. (a,b) Cells were transfected with the indicated siRNAs against dynein HC or KIF5B. (c) Cells were transfected with the indicated siRNAs against BICD1 and/or BICD2. (d) Cells were transfected with GFP-BICD2-CT. (e) Cells were transfected with the indicated siRNAs against RanBP2 or NUP214. (f) The effect of RanBP2 knockdown (induced with the RanBP2#3 siRNA) was rescued by expression of the GFP-BICD2-NT-nesprin-3 fusion but not by GFP alone. (g) Cells were transfected with the indicated siRNAs against BICD2, and either untreated or incubated with 4 µM STLC, an Eg5 inhibitor. Error bars represent SD; the number of experiments for each condition is shown in parentheses below the graph; 20–30 cells were counted per experiment. Strong separation of the nucleus and the centrosomes is induced by dynein depletion. Mild nuclear-centrosome separation is caused by overexpression of BICD2-CT or depletion of BICD2 or RanBP2, but not BICD1 or NUP214, which serve as negative controls. A fusion of BICD2 N-terminus to the NE-targeting domain can compensate for the loss of RanBP2. The effect of BICD2 knockdown on nuclear-centrosome positioning can be suppressed by inhibiting Eg5, suggesting that the MT sliding activity of this motor is responsible for centrosome detachment in BICD2-depleted cells. (D) HeLa cells were transfected with GFP-BICD2-NT-nesprin-3 fusion, fixed with paraformaldehyde, and stained for p150^Glued. Note the recruitment of dynactin to the NE in the transfected cell.

Figure 7. Dynein, kinesin-1, and BICD2 control the localization of AL in G2 phase.

(A) HeLa cells were transfected with control, DHC#1, p150^Glued, or KIF5B#1 siRNAs, fixed with paraformaldehyde 3 d later, and stained for endogenous BICD2, RanBP2, and cyclin B1. In the overlays, BICD2 is shown in green and RanBP2 in red. The outline of the cyclin B1-positive cell is indicated. Note AL displacement to the cell periphery in cyclin B1-positive cells depleted of dynein or dynactin, and the centripetal relocalization of AL in kinesin-1 (KIF5B) depleted cyclin B1-stained cell. (B) GFP-RanGAP1 stable HeLa cell line was imaged with a 2 or 3 min time interval 2 d after transfection with the control, p150^Glued, or KIF5B#1 siRNAs. 0 min indicates the first frame after NEB (defined as the time when GFP-RanGAP1 enters the nucleus). Contrast is inverted. Arrows show the accumulation of AL at the cell periphery or the cell center. Peripheral displacement of AL after dynein or dynactin knockdown occurred at 1 h±30 min before mitotic onset (mean ± SD, measured in 6 and 10 cells, respectively); strong accumulation of AL near the nucleus in KIF5B depleted cells started at 3 h±30 min before NEB (mean ± SD, measured in 18 cells). (C) AL localization after different siRNA co-transfections. Black bars represent the percentage of HeLa cells showing a strong AL accumulation near the nucleus (like in the bottom panel in Figure 7A) and gray bars illustrate the percentage of HeLa cells showing displacement of AL to the outmost cell periphery (like in the middle panels of Figure 7A). In each case, cells were transfected with a combination of two siRNAs: control, KIF5B or dynein HC siRNAs (as indicated on top of the panel) in combination with the control siRNA or the siRNAs against BICD1 and BICD2 (the cotransfected siRNA is indicated at the bottom), or the combination of KIF5B siRNAs and dynein HC siRNAs. Cells with AL accumulation in the cell center or the cell periphery were scored in KIF5B or dynein knockdown cells, respectively; in a double KIF5B/dynein HC knockdown, both types of cells were counted. Error bars represent SD. ∼30–100 cells were counted in three experiments. Cells with fully aggregated AL represent a minority of the population (∼20%) in control G2 cells but become predominant (∼75%) after depletion of kinesin-1. This effect can be suppressed by co-depletion of BICD2 or dynein HC, but not BICD1. Similarly, control cells never display peripheral AL localization, while ∼40% of dynein-depleted cyclin B1-positive cells show this phenotype, which can be suppressed by co-depletion of BICD2 or KIF5B, but not BICD1.

Figure 8. A model of the concerted action of BICD2, dynein, and kinesin-1 in different phases of the cell cycle.

In G1 and S phase, BICD2, dynein, and kinesin-1 associate with Rab6 vesicles; kinesin-1 activity predominates in this complex. In G2, BICD2 with the associated motors accumulates at the NPCs, where the dynein-mediated movement predominates, resulting in tight association of the centrosome and nucleus.