|Home | About | Journals | Submit | Contact Us | Français|
Cytokinesis partitions the cytoplasm of a parent cell into two daughter cells and is essential for the completion of cell division. The final step of cytokinesis in animal cells is abscission, which is a process leading to the physical separation of two daughter cells. Abscission requires membrane traffic and microtubule disassembly at a specific midbody region called the secondary ingression. Here, we report that WD repeat-containing protein 5 (WDR5), a core subunit of COMPASS/MLL family histone H3 lysine 4 methyltransferase (H3K4MT) complexes, resides at the midbody and associates with a subset of midbody regulatory proteins, including PRC1 and CYK4/MKLP1. Knockdown of WDR5 impairs abscission and increases the incidence of multinucleated cells. Further investigation revealed that the abscission delay is primarily due to slower formation of secondary ingressions in WDR5 knockdown cells. Consistent with these defects, midbody microtubules in WDR5 knockdown cells also display enhanced resistance to depolymerization by nocodazole. Recruitment of WDR5 to the midbody dark zone appears to require integrity of the WDR5 central arginine-binding cavity, as mutations that disrupt histone H3 and MLL1 binding to this pocket also abolish the midbody localization of WDR5. Taken together, these data suggest that WDR5 is specifically targeted to the midbody in the absence of chromatin and that it promotes abscission, perhaps by facilitating midbody microtubule disassembly.
Post-translational modifications of histones, including methylation of specific lysine residues of histones H3 and H4, regulate chromatin structure and function by promoting the assembly of protein complexes at specific genomic loci (1, 2). Methylation of histone H3 lysine 4 (H3K4) is enriched within actively transcribed genes and is viewed as a hallmark of transcriptional activity (3,–6). In mammals, the COMPASS/MLL family of H3K4 methyltransferases (H3K4MTs) function as a complex consisting of one catalytic subunit (SET1A, SET1B, MLL1, MLL2, MLL3, or MLL4) and four core regulatory subunits (WDR5, RbBP5, ASH2L, and mDPY-30) (1, 7,–9). These core regulatory subunits form a subcomplex that dramatically enhances the enzymatic activity of the catalytic subunit, and depletion of any one of the core regulatory subunits leads to a defect in H3K4 methylation (10, 11). The physiological significance of H3K4MTs is highlighted by the fact that translocation of MLL1 and knock-out of MLL3 cause cancer in humans and mice, respectively (12). Recently, mutations in MLL2 were found to be the most common cause of Kabuki syndrome (13). In addition, sequencing data from the human cancer genome indicate that several H3K4MT subunits are frequently mutated in a variety of cancers (12).
Although the precise structure of assembled catalytic and core regulatory subunits remains elusive (14,–16), WDR5 is essential for the association of RbBP5, ASH2L, and mDPY-30 with MLL1 (11, 15). WDR5 is a highly conserved ~36-kDa protein with a short unstructured N terminus followed by seven WD40 repeats that adopt a seven-bladed β-propeller fold (17). Hierarchical assembly of MLL1 with RbBP5, ASH2L, and mDPY-30 occurs via two distinct binding sites located on opposite faces of the WDR5 β-propeller (15, 18, 19). One site, referred to as the arginine binding cavity, is occupied by the arginine-containing WIN (WDR5-interacting) motif of the MLL/SET catalytic subunit and the other by a motif within the RbBP5 C-terminal tail. Almost all studies of WDR5 have been conducted regarding its nuclear function, and whether this protein has a cytoplasmic role remains unclear. In 2010, Wang et al. (20) reported that nuclear WDR5 translocates to the mitochondrial outer membrane where it mediates host response after viral infection. Subsequently, a quantitative proteomics study of SET1/MLL complex stoichiometry identified a large number of novel WDR5-associated proteins, several of which have known cytoplasmic functions (21). These findings suggest that WDR5, like other β-propellers, may function as a scaffolding hub for cytoplasmic signaling modules yet to be identified.
While investigating the role of Golgi-localized mDPY-30 in vesicular transport (22), we observed phenotypes in WDR5-depleted RPE1 (an immortalized nontumor human cell line) and HeLa cells characteristic of cytokinetic defects. Cytokinesis, the final step of cell division that results in two separated daughter cells, is critical for preserving genomic integrity (23,–26). Failure of cytokinesis can cause tetra- and polyploidization, a state of chromosomal instability that is thought to precede cancer formation (27, 28). Mechanistically, cytokinesis in animal cells can be divided into two stages, cleavage furrow ingression and abscission (24,–26, 29). Upon completion of cleavage furrow ingression, the actomyosin ring is converted to the midbody ring, and the midbody matures to a thickness of ~1–2 μm. Three groups of proteins are essential for the formation of midbody microtubules as follows: (a) PRC1-KIF4 (an antiparallel microtubule cross-linker and a PRC1-interacting kinesin, respectively); (b) the central spindlin complex MKLP1-CYK4 (an antiparallel microtubule-bundling complex); and (c) the chromosomal passenger complex consisting of Aurora B kinase and three additional regulatory subunits (a modulator of MKLP1-CYK4) (30, 31). Although these proteins exhibit dynamic localization during midbody maturation, each group localizes to specific regions of the mature midbody as follows: the dark zone (PRC1-KIF4); the ring (MKLP1-CYK4); and the flanking region (chromosomal passenger complex) (32).
Following maturation of the midbody, cytokinesis enters its second stage, abscission, resulting in the final separation of the two daughter cells (24, 25). During abscission, MKLP1-CYK4 recruits CEP55 to the midbody ring. CEP55 then binds ESCRTIII targeting factors (ALIX and TSG101), which in turn localize ESCRTIII to the ring. Shortly before the final cut, ESCRTIII migrates to an adjacent site known as the secondary ingression where it will mediate membrane fission. The secondary ingression is characterized by the further narrowing of the intercellular bridge from ~1–2 μm to ~ 100 nm (33, 34), and its formation requires the temporal and spatial orchestration of vesicular transport/membrane fusion and clearance of F-actin and microtubules. Although spastin, an ESCRTIII-associated microtubule-severing enzyme (35, 36), has long been postulated to mediate microtubule disassembly at the secondary ingression, a recent study shows that the formation of secondary ingressions does not depend on ESCRTIII (37). Furthermore, additional microtubule disassembly mechanisms have been observed within the midbody (37, 38), and whether spastin is involved in these processes is unknown.
Here, we provide evidence suggesting that the H3K4MT core regulatory subunit WDR5 localizes to the center of midbody in tumor (HeLa) and nontumor (RPE1) cell lines. In particular, the central arginine binding cavity of WDR5 appears to be required for this targeting. Depletion of WDR5 increases the incidence of multinucleated cells, a hallmark of cytokinesis failure. Live cell imaging revealed that midbody abscission is delayed in WDR5-depleted cells and that this phenotype could be largely attributed to slower formation of secondary ingressions. Because the midbody microtubules in WDR5-depleted cells display enhanced resistance to depolymerization by nocodazole compared with control cells, knockdown of WDR5 may cause the abscission defects by enhancing midbody microtubule stability. In its entirety, our data reveal a novel subcellular pool of WDR5 localized to the midbody and support a model in which WDR5 promotes abscission, possibly by facilitating midbody microtubule disassembly.
Unless otherwise specified, all reagents/materials were purchased from Fisher Scientific, VWR, Sigma, Life Technologies, or Clontech.
All antibodies used for Western blot analysis (WB),3 immunocytochemistry (ICC), and immunoprecipitation (IP) are as follows: rabbit anti-ALIX (Santa Cruz Biotechnology sc-271975, WB 1:500); rabbit anti-Aurora B (Abiocode R0117-1, WB 1:2000); mouse anti-centriolin (Santa Cruz Biotechnology sc-365521, WB 1:500); rabbit anti-CEP55 (Santa Cruz Biotechnology sc-134622, WB 1:1000); rabbit anti-CHMP1B (gift from W. I. Sundquist UT592 (39), WB 1:500); mouse anti-CYK4 (Santa Cruz Biotechnology sc-271110, WB 1:1250); rabbit anti-ECT2 (Santa Cruz Biotechnology sc-1005, WB 1:600); mouse anti-GAPDH (Biochain Institute Y3322GAPDH, WB 1:6000); mouse anti-GFP (Roche Applied Science, catalog no. 11814460001, WB 1:1600); rabbit anti-histone H3 (Cell Signaling Technology catalog no. 2650, WB 1:4000); rabbit anti-KIF4A (Bethyl A301-074A, WB 1:2500); rabbit anti-mDPY-30 (Ma Laboratory previously characterized (22), WB 0.25 μg/ml); rabbit anti-MKLP1 (Santa Cruz Biotechnology sc-867, WB 1:1000); goat anti-mouse IgG (Thermo Scientific, DyLight 680 catalog no. 35518, WB 1:6000); goat anti-mouse IgG (Jackson ImmunoResearch, Alexa Fluor 488 catalog no. 115-545-062, Rhodamine Red-X catalog no. 115-295-062, and cyanine Cy5 catalog no. 115-175-146, ICC 1:200); goat anti-rabbit IgG (Jackson ImmunoResearch, Alexa Fluor 488 catalog no. 111-545-144 and Rhodamine Red-X catalog no. 111-295-144, ICC 1:200); rabbit IgG (Invitrogen catalog no. 02-6102, IP 4 μg/ml); goat anti-rabbit IgG (Thermo Scientific, DyLight 680 catalog no. 35568, WB 1:6000); rabbit anti-PRC1 (Santa Cruz Biotechnology sc-8356, WB 1:1000); rabbit anti-RbBP5 (Bethyl A300-107A, WB 1:2000); rabbit anti-TSG101 (Epitomics catalog no. 5347-1, WB 1:2500); mouse anti-α-tubulin (Covance MMS-407R, ICC 1:6000); mouse anti-acetylated-α-tubulin (Santa Cruz Biotechnology sc-23950, ICC 1:1000); rabbit anti-β-tubulin (Santa Cruz Biotechnology sc-9104, WB 1:200); rabbit anti-WDR5 (Bethyl A302-429A, IP 4 μg/ml); and rabbit anti-WDR5 (Bethyl A302-430A, WB 1:3500).
All shRNA constructs were purchased from Sigma as glycerol stocks (pLKO.1-puro vector). Lentiviral particles were generated from shRNA constructs as described below. Target sequences for shRNAs are as follows: control shRNA (SHC002, nontargeting in human or mouse, 5′-CAACAAGATGAAGAGCACCAA-3′); WDR5 shRNA 1 (TRCN0000118047, 5′-GCCTCCTCTCTGAAGATGATT-3′); and WDR5 shRNA 2 (TRCN0000118050, 5′-CCAACCTTATTGTCTCAGGAT-3′).
For pLVXpuro and pLNCX constructs, lentiviral particles were generated as described below. Human and mouse WDR5 cDNAs code for identical amino acid sequences. Subcloning details are as follows: pLVXpuro-EGFP (Clontech, 5′ SmaI, 3′ XbaI); pLVXpuro-WDR5-EGFP (Clontech, 5′XhoI/Klenow, 3′ Bsp1407I, from mouse cDNA, C-terminal EGFP); pLVXpuro-mWDR5-HAC (Clontech, 5′ BamHI, 3′ XbaI/Klenow, from mouse cDNA, C-terminal HA); pEGFP-C3-WDR5_71-173 (Clontech, 5′ XhoI, 3′ SmaI, from mouse cDNA, N-terminal EGFP); pEGFP-C2-EGFP-WDR5 (Clontech, 5′ EcoRI, 3′ BamHI, from mouse cDNA, N-terminal EGFP); pEGFP-C3-EGFP-WDR5-S91K (Clontech, 5′BglII, 3′ EcoRI, subcloned from pHisParallel, gift of M. Cosgrove, N-terminal EGFP); pEGFP-C3-EGFP-WDR5-F133A (Clontech, 5′BglII, 3′ EcoRI, subcloned from pHisParallel, gift of M. Cosgrove, N-terminal EGFP); pLNCX-mCherry (Clontech, 5′ BamHI/Klenow, 3′ HindIII); pLNCX-mCherry-α-Tubulin (Clontech, 5′ NruI, 3′ HindIII, from Addgene plasmid catalog no. 21043, N-terminal mCherry).
CD8-CIMPR-expressing HeLa (HeLa) (22), hTERT RPE-1 (RPE1, gift from L. Johnson, NRI, University of California at Santa Barbara), and HEK293T cells were cultured in Advanced DMEM (Hyclone DMEM-RS) supplemented with 5% fetal bovine serum, 2 mm glutamine (Life Technologies), and 1× penicillin/streptomycin (Life Technologies). Stably transduced cell lines were grown in the same media with additional supplementation as follows: pLVXpuro and pLKO.1puro (3 μg/ml puromycin); pLNCX (400 μg/ml G418); hTERT (200 μg/ml hygromycin). Cells were dissociated for passage using TrypLE dissociation reagent (Life Technologies). For immunocytochemistry, glass coverslips were coated with 1:500 Matrigel (BD Biosciences) in serum-free medium overnight in a 37 °C incubator and rinsed with 1× PBS before use.
Production of lentiviral particles and infection of HeLa and RPE1 cell lines were performed as described previously (40). HEK293T cells were grown to 70% confluence in 100-mm plates, and the spent medium was replaced with fresh medium 2 h before transfection. In a sterile transformation tube pMD2.G (2.9 μg), psPAX2 (5.4 μg), and either pLVXpuro, pLKO.1puro, or pLNCX vector containing the appropriate insert (8.3 μg) were mixed. The following components were then added in the following order: 242 μl of 0.1× TE (1 mm Tris, 0.1 mm EDTA, pH 8.8), 128 μl of sterile water, and 41.5 μl of 2.5 m calcium chloride. After brief mixing, 481 μl of 2× HBS (280 mm NaCl, 100 mm HEPES, 1.5 mm Na2HPO4, 7.11 ≤ pH ≤ 7.13) was added dropwise under gentle vortexing and was incubated for 5 min. After incubation, 830 μl of this transfection mixture was added dropwise to the cells, and medium was replaced 14–16 h post-transfection. Supernatant-containing viral particles was harvested three times, once every 12 h, and pooled supernatants were kept at 4 °C over the collection period. After the last collection, the supernatant was centrifuged for 5 min at 1500 rpm, filtered with a 0.22-μm filter, and stored in aliquots at −80 °C.
HeLa or RPE1 cells were trypsinized and re-plated at 60–80% confluence in media containing 10 μg/ml Polybrene (Calbiochem). After cell re-attachment (1–3 h), the medium was removed to a separate well and replaced with viral supernatant. Cells were incubated for 30–60 min before Polybrene-containing medium was added back, and the cells were grown overnight. Approximately 24 h post-transfection, the medium was replaced, and antibiotic selection began at 48 h post-transfection. Transduced cell lines were grown under antibiotic selection for at least 7 days before starting experiments or making frozen stocks.
HeLa cells were infected with lentiviral particles, selected using the appropriate antibiotic, and isolated by limiting dilution into 48-well plates at both low (0.75 cells/well) and high (3 cells/well) cell concentrations. One week after seeding, each well was examined for the presence of single cell colonies, and wells containing more than one colony were excluded. Colonies received half-media changes every 2–3 days and were expanded sequentially from 48-well to 12- and 6-well and T75 culture vessels. Effective knockdown of the target protein was assessed by SDS-PAGE and Western blotting.
HeLa or RPE1 cells stably expressing WDR5-EGFP were lysed in ice-cold Nonidet P-40 buffer (50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 1 mm EDTA, and 1% Nonidet P-40), with 1× Halt Protease and Phosphatase Inhibitor Mixture (Thermo Scientific) and 1 mm PMSF added just before lysis. Lysates were rocked at 4 °C and cleared by 4 °C centrifugation at maximum speed, and the total protein concentration of the supernatant was normalized across samples using the CB-X protein assay kit (G-Biosciences). Adjusted lysates were bound to a mouse monoclonal anti-GFP antibody covalently conjugated to magnetic microbeads while rocking at 4 °C for 3 h (μMACS GFP-tagged protein isolation kit, Miltenyi Biotec). The beads were washed on μMACS columns with Nonidet P-40 lysis buffer and eluted according to the manufacturer's instructions. HeLa or RPE1 cells stably expressing EGFP alone were included as a negative control.
Midbodies were isolated from thymidine/nocodazole-synchronized HeLa cells as described previously by Skop et al. (41). Briefly, cells undergoing synchronous cytokinesis were pelleted, resuspended in spindle isolation buffer (2 mm PIPES, pH 6.9, 0.25% Triton X-100, and 20 μg/ml Taxol, 1× Halt mixture protease inhibitor, 1 mm PMSF) and divided into 2 aliquots of equal volume. Pellets, which contain both mitotic spindles and midbodies, were obtained by centrifugation of each aliquot. The total nonspindle/midbody supernatant was collected and mixed with SDS-PAGE sample buffer. To obtain the spindle fraction, one spindle/midbody pellet was resuspended in SDS-PAGE sample buffer (at half of the total supernatant volume). The other spindle/midbody pellet was chilled on ice, washed, and resuspended in 50 mm MES, pH 6.3, and subjected to centrifugation through a cushion of 40% glycerol, yielding isolated midbodies. The isolated midbody pellet was combined with SDS-PAGE sample buffer (again at half of the total supernatant volume). To generate a “non-midbody” fraction for immunoblotting, equal volumes of nonspindle/midbody supernatant and spindle fractions were combined. This was electrophoresed alongside the isolated midbody fraction at a ratio of 2:1 to maintain equal cell numbers in each lane.
For immunoprecipitation from the spindle fraction, the spindle/midbody pellet was resuspended in ice-cold RIPA buffer and incubated with primary antibody (or an equivalent amount of rabbit IgG as a negative control) for 3 h followed by protein A-Sepharose (Invitrogen) for 2 h at 4 °C. The Sepharose beads were washed four times in ice-cold lysis buffer, and bound proteins were eluted with SDS-PAGE sample buffer at 95 °C for 15 min.
Cells were lysed in ice-cold RIPA buffer (50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 1 mm EDTA, 1% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% SDS), with 1× Halt Protease and Phosphatase Inhibitor Mixture (Thermo Scientific) and 1 mm PMSF added as above. After rocking and clearing lysates at 4 °C, the total protein concentration was quantified using the CB-X protein assay (G-Biosciences) and denatured in SDS loading buffer. An equal amount of total protein per lane was loaded into a polyacrylamide gel (6–15%, depending on the proteins to be resolved), and samples were separated by SDS-PAGE and transferred to Immobilon 0.45-μm PVDF membranes (Millipore) using an Owl HEP-1 semi-dry electroblotting system (Thermo Scientific). Membranes were incubated overnight at 4 °C with the appropriate primary antibody diluted in a 1:1 mixture of Odyssey blocking buffer (LI-COR Biosciences) and PBST (1× PBS supplemented with 0.1% Tween 20). Membranes were washed with PBST (three times for 10 min each), incubated in the dark for 45 min with the appropriate secondary antibody (goat anti-mouse or anti-rabbit IgG (H+L) DyLight 680 conjugated, Thermo Scientific), washed with PBST (three times, 5 min each), followed by PBS for 5 min, and dried in the dark. Fluorescence imaging and quantification was performed on an Odyssey Infrared Imaging System (LI-COR Biosciences).
Cells were fixed in 3% UltraPure EM-grade formaldehyde (Polysciences Inc.) for 10 min, quenched with PBS containing 100 mm glycine (two times, 5 min each), and permeabilized with PBS containing either 0.1% saponin or 0.1% Triton X-100 for 15 min. Fixed cells were blocked in buffer (BLOK casein in PBS (G-Biosciences) containing 0.1% saponin and 5% goat serum) for 30 min, incubated with primary antibody diluted in blocking buffer for 2 h, washed in PBS (three times, 5 min each), incubated in fluorophore-conjugated secondary antibody diluted in blocking buffer for 1 h, washed in PBS (three times, 5 min each), incubated in PBS containing 1 μg/ml DAPI for 5 min, and washed in PBS (three times, 5 min each). Coverslips were then allowed to air dry and mounted in PPD antifade (90% glycerol, 10% PBS, 1 mg/ml p-phenylenediamine, pH 8.0).
Slides were imaged on an Olympus IX-81 fluorescent microscope using a Plan-Apochromat 60× oil total internal reflection fluorescence objective (Olympus, NA = 1.45), a 1.6× tube factor, Semrock excitation and emission filters, and an X-Cite Exacte illumination source (Lumen Dynamics). Images were captured using a monochrome digital camera (Evolution QEi, Media Cybernetics) controlled with In Vivo acquisition software version 3.2.0 (Media Cybernetics). Image analysis was performed using ImageJ (National Institutes of Health), Photoshop, and Illustrator (Adobe).
Clonal HeLa cell lines expressing WDR5 shRNAs were transduced with mCherry-tubulin, grown on Matrigel-coated (1:500) glass-bottomed Delta T dishes (Bioptechs), and imaged using an Olympus IX-81 as described above. Because the timing of cytokinesis can vary significantly depending on surface coating conditions and cell density (42), these two variables were controlled as tightly as possible between individual experiments. All live cell imaging experiments were performed at 37 °C using a Delta T Open Dish System equipped with a heated lid and 5% CO2 perfusion (Bioptechs). For measurements of abscission duration, a 5-image z-series (1.5 μm steps) was acquired at 15-min intervals between anaphase initiation and midbody formation, and at 30-min intervals between midbody formation and abscission. To minimize the potential for phototoxicity, lamp power was set at 10% of maximum, and each exposure was kept <200 ms. For nocodazole treatment, cells were imaged as above, except that treatment with nocodazole in a small volume of prewarmed media (final dish concentration, 10 μm) was initiated upon midbody formation. The nocodazole concentration was chosen based on a similar study of midbody microtubule depolymerization in fixed HeLa cells (43). Midbody formation was defined as the first appearance of an intercellular bridge containing a bulge and mostly condensed flanking microtubules, and abscission was defined as the first appearance of a gap lacking a detectable mCherry-tubulin signal in the midbody connecting the daughter cells. For live cell snapshots of EGFP-WDR5, EGFP-WDR5-S91K, and EGFP-WDR5-F133A, and EGFP-WDR5_71–173, HeLa cells were imaged 24 h after transient co-transfection with mCherry-tubulin and one of the above EGFP-containing plasmids using TurboFect (Thermo Scientific) at a 3:1 ratio of TurboFect to plasmid according to the manufacturer's instructions. To quantify the cytokinesis phenotype of cells overexpressing EGFP or EGFP-WDR5_71–173, live cell exposures were captured once every 30 min for a maximum of 3 h following mature midbody formation (defined by the presence of a midbody bulge and mostly condensed flanking midbody microtubules). Cell death was defined by extreme membrane blebbing events in one or both connected daughter cells that led to the collapse of overall cell shape.
Cells co-expressing WDR5-EGFP (pLVXpuro) and mCherry (pLNCX) or EGFP (pLVXpuro) and mCherry (pLNCX) were immunostained for acetylated tubulin as described above and imaged on an Olympus FLV1000S spectral laser scanning confocal. A PlanApo N 60XOSC objective (N.A. = 1.4) was used throughout, with excitation laser lines at 405, 488, 559, and 635 nm for DAPI, EGFP, mCherry, and Cy5-conjugated secondary antibody, respectively. Images were captured as z-series of 8–15 planes each (0.5-μm steps). Quantification of WDR5 or EGFP midbody enrichment was performed on the middle z-slice in each series using ImageJ (National Institutes of Health). A profile plot (intensity versus distance) was constructed for EGFP, mCherry, and acetylated tubulin channels along a 6-μm linear selection overlaid on the midbody-acetylated tubulin signal. The center of the midbody dark zone was defined as the location on this line with the lowest acetylated tubulin signal, and peak mCherry or EGFP values were taken as the maximum intensity value within 1.5 μm of this dark zone center. The ratio of peak mCherry or EGFP signal to the average signal of a 1-μm length of midbody flanking region immediately outside the boundary of the dark zone center (1.5 μm to 2.5 μm from the minimum acetylated tubulin point) was calculated as the fold enrichment of mCherry or EGFP in the dark zone compared with the flanking region.
The following phenotypes were quantified in fixed cells using conventional fluorescence microscopy as described above. Multinucleated cells were counted using DAPI (1 μg/ml) as a nuclear marker and the lipophilic dye DiOC6 (1 μg/ml) to visualize cell membranes. Cells containing two or more similarly sized nuclei within one cytoplasm were classified as multinucleated. At least 1000 cells per condition were counted, and data are presented as the mean ± S.E. of at least three independent experiments. Midbody width was measured from images acquired under nonsaturating exposure conditions. Using acetylated tubulin to visualize midbody microtubules, midbody width was measured at the thickest point for 100 cells per condition with ImageJ (National Institutes of Health). Data are presented as the mean ± S.E. of four independent experiments. Secondary ingressions, defined by at least one dark gap lacking tubulin or acetylated tubulin signal accompanied by narrower microtubules, were quantified in 100 cells per condition. Data represent the mean ± S.E. of five independent experiments. For quantification of nocodazole-induced microtubule disassembly in live cells, images acquired under nonsaturating exposure conditions were background-subtracted using ImageJ (National Institutes of Health); the midbody was bound with a freehand selection, and the average pixel intensity within this region was measured. Values for 15- and 30-min time points were normalized to the average midbody pixel intensity at 0 min in the same cell. For each group, n = 11–22 cells.
All statistical tests were performed using IBM SPSS Statistics 20 (IBM). Unless otherwise specified, statistical significance was assessed by univariate ANOVA with Tukey's HSD post hoc, and data are presented as mean ± S.E. error bars. Homogeneity of variance was assessed using Levene's test. For all analyses, α = 0.05.
Although WDR5 is generally thought to function in the nucleus, the EGFP fusion of WDR5 produced a stronger fluorescent signal in the midbody central dark zone than the neighboring areas in both HeLa (Fig. 1a) and RPE1 (Fig. 1b) cells. Because the fixation and permeabilization of cells during immunostaining can lead to imaging artifacts, particularly the differential extraction of some soluble proteins from the cytoplasm (44), we also performed fluorescence imaging of live HeLa cells stably expressing mCherry-tubulin (to label the midbody) and either EGFP or WDR5-EGFP. Although midbody-localized WDR5-EGFP was apparent in nearly all telophase cells, EGFP alone was not concentrated at this location in live cells (Fig. 1c). To confirm that WDR5 is indeed enriched in the dark zone, we performed a quantitative confocal microscopic analysis of HeLa cells co-expressing WDR5-EGFP and mCherry (a soluble protein expected to distribute uniformly throughout the midbody). As shown in Fig. 1, d–f, WDR5-EGFP was concentrated ~2.5-fold in the dark zone relative to mCherry. Because extreme overexpression of WDR5-EGFP might lead to a spurious localization at the midbody, we compared the degree of dark zone WDR5-EGFP enrichment in cells from the same stable line that expressed either high or low levels of WDR5-EGFP. Midbody WDR5-EGFP enrichment was found to be independent of the WDR5-EGFP expression level (Fig. 1f). Finally, we compared the degree of nonspecific midbody enrichment exhibited by EGFP or mCherry alone. Although the signal of EGFP alone is slightly enhanced within the midbody in fixed cells, this difference was not statistically significant relative to mCherry (Fig. 1f). Central dark zone proteins are typically inaccessible to antibodies (32, 45), which may explain why the midbody localization of WDR5 has not been reported previously.
The following observations suggest that the midbody localization of WDR5-EGFP is not a consequence of its association with lagging chromosomes, which are occasionally observed at the midbodies of dividing cells. First, lagging chromosomes were found in fewer than 10% of telophase HeLa cells (a number consistent with a previous report (46)), whereas midbody-localized WDR5-EGFP was present in virtually all telophase cells. Second, the midbody EGFP signal of WDR5 was readily detected when DAPI-labeled lagging chromosomes were absent (Fig. 1, a, b, and d). These results indicate that WDR5 can be found at the midbody independent of lagging chromosomes, and we therefore examined several known midbody-localized proteins for interaction with WDR5.
We generated RPE1 and HeLa cells stably expressing WDR5-EGFP and tested the interaction between WDR5-EGFP and known midbody proteins using co-immunoprecipitation. In addition to ASH2L, RbBP5, mDPY-30, and histone H3 (Fig. 2a), WDR5-EGFP efficiently co-immunoprecipitated with several dark zone and midbody ring proteins (PRC1, MKLP1, CYK4, and CEP55) in both RPE1 (Fig. 2b) and HeLa cells (Fig. 2c). The WDR5 fusion also co-immunoprecipitated, although to a lesser extent, with proteins known to be targeted to the midbody ring by their interaction with MKLP1, CYK4, and CEP55, including ECT2, ALIX, TSG101, and CHMP1B. Consistent with the dark zone localization of WDR5-EGFP revealed by our imaging analysis, little or undetectable association between WDR5-EGFP and Aurora B kinase (Fig. 2, b and c), a protein flanking the dark zone, was observed. Interestingly, some dark zone (KIF4) or midbody ring (Centriolin) proteins also exhibited little or undetectable co-immunoprecipitation with WDR5-EGFP (Fig. 2, b and c), suggesting that WDR5 preferentially associates with a subset of dark zone/ring proteins. Compared with EGFP alone, we noticed that stable expression of WDR5-EGFP led to an increase in the levels of PRC1, MKLP1, CYK4, and CEP55 in RPE1 cells (Fig. 2b, lysate lane). Subsequent analysis using nontransfected RPE1 cells showed that this observation was due to a down-regulation of these proteins in EGFP-expressing RPE1 cells. However, co-immunoprecipitation of WDR5-EGFP with these proteins cannot be simply explained by the increased protein levels, because HeLa cells stably expressing EGFP or WDR5-EGFP have similar levels of PRC1, MKLP1, CYK4, and CEP55 (Fig. 2c, lysate lane). Because many of the midbody proteins found to associate with WDR5-EGFP interact directly or indirectly with microtubules, we asked whether tubulin itself co-immunoprecipitates with WDR5-EGFP. Although nearly all β-tubulin remains soluble under our lysis and immunoprecipitation conditions, it does not appreciably associate with WDR5-EGFP (Fig. 2d). This result suggests that WDR5-EGFP associates with midbody proteins independent of a direct interaction with microtubules.
To address the issue of whether co-immunoprecipitation of midbody proteins with WDR5-EGFP reflects the behavior of endogenous WDR5 at the midbody, we first examined the degree of WDR5-EGFP overexpression in this HeLa stable line. As shown in Fig. 3a, WDR5-EGFP is moderately overexpressed (~5-fold greater abundance) compared with endogenous WDR5. Furthermore, we confirmed that endogenous WDR5 co-immunoprecipitates with MKLP1 and CYK4 from HeLa spindle/midbody isolate under the stringent conditions required to solubilize the isolated spindle/midbody pellet (1% Nonidet P-40, 0.5% sodium deoxycholate, and 0.1% SDS) (Fig. 3b), suggesting a specific interaction that is independent of WDR5 epitope tagging or exogenous expression. Moreover, a larger fraction of endogenous WDR5 is present in isolated midbodies compared with lamin A (a nuclear marker), Hsp70, or GAPDH (two proteins predominantly localized to the cytoplasm) (Fig. 3c). These results suggest that endogenous WDR5 is a bona fide component of the midbody.
Previously, large scale proteomic studies using multidimensional protein identification technology (MudPIT) mass spectrometry identified WDR5 as a component of the mitotic spindle (47) but not of the midbody (41) in CHO cells. Although MudPIT is especially useful for bulk proteomic analysis, its detection sensitivity for specific proteins can be limited by factors, including low abundance (48), small size (49), presence of modifications (50), and amino acid composition (50). Some combination of these limitations may explain why MudPIT did not identify WDR5 in purified midbodies despite the presence of nuclear contamination (1 nucleus per 60 to 70 midbodies), and other nuclear proteins (41).
The midbody localization of WDR5 raises the possibility that it might play a role in cytokinesis. Because multinucleation is a hallmark of cytokinesis failure (23, 27), we determined the fraction of cells containing two or more nuclei in HeLa cells stably expressing a nontargeting control shRNA or either of two different shRNAs targeting WDR5. Both WDR5 shRNAs increased the number of multinucleated cells relative to the control shRNA (Fig. 4, a and b). To obtain a more uniform knockdown efficiency, we generated clonal HeLa cell lines stably expressing WDR5 shRNA 1. Although WDR5 could not be depleted by more than 80% in any of the clones tested (Fig. 4e, ~55 to 72% knockdown), all clones exhibited significantly increased numbers of multinucleated cells compared with two clones derived from the control shRNA (Fig. 4, c and d). Because clones derived from four additional WDR5 shRNAs expressed even higher levels of residual WDR5 (data not shown), further experiments were performed using clones 6 and 8 derived from WDR5 shRNA 1, both of which exhibit an intermediate level of WDR5 knockdown and multinucleation (Fig. 4, c and d). Moreover, the multinucleation phenotype exhibited by clones 6 and 8 was mostly rescued by stable expression of HA-tagged mouse WDR5 (mWDR5-HAC), which encodes an identical amino acid sequence to human WDR5 but is resistant to shRNA 1 (Fig. 4f). It remains possible that epitope tagging or nonphysiological expression levels of mWDR5-HAC may hinder the full rescue of the multinucleation phenotype to control levels.
To further examine the role of WDR5 in cytokinesis, we performed fluorescent live cell imaging of cytokinesis in clonal WDR5 knockdown HeLa stable lines. Although knockdown of WDR5 had little effect on the early events of cytokinesis, including cleavage furrow ingression and midbody maturation (see below), it significantly delayed abscission (Fig. 5a; only those cells that completed abscission were included). Pairwise comparisons revealed significantly longer mean abscission times in both WDR5 knockdown clonal lines compared with either control clone (ANOVA with Tukey's HSD post hoc, p < 0.001). Midbody abscission was defined as the sudden appearance of a gap in the mCherry-tubulin signal connecting the daughter cells (arrows in Fig. 5b). Appearance of this gap was not due to photobleaching of mCherry-tubulin based on the short duration of the exposures (<200 ms per exposure) as compared with the rate of mCherry-tubulin photobleaching in our system (Fig. 5f; ~1% signal loss/s of continuous exposure). Although ~8% multinucleation is observed in WDR5 knockdown clones 6 and 8, our live cell imaging data did not reveal any instances of furrow regression in control or WDR5 knockdown stable lines (n = 13–21 cells per group). It is worth noting that a recent study indicates that even if depletion of a protein yields 5–10% multinucleation, the rate of regression could be as low as ~1%, and it may be restricted to a minor subset of telophase cells, for example those containing lagging chromosomes (51).
To gain insight into the mechanism by which WDR5 facilitates abscission, we asked whether WDR5 knockdown affects the formation of secondary ingressions, a process preceding the final scission and characterized by gaps or narrowing in the midbody microtubules adjacent to the dark zone. We investigated secondary ingression formation by measuring the time between the formation of mature midbodies and the appearance of secondary ingressions using the live cell imaging data set previously described in Fig. 5a. Our analysis shows that the formation of secondary ingressions is significantly delayed in WDR5 knockdown cells (Fig. 5c), yet the time between anaphase onset and midbody maturation is not substantially affected (Fig. 5d). Although completion of final abscission after secondary ingression is also delayed in WDR5 knockdown cells (Fig. 5g, ANOVA with Tukey's HSD post hoc, p < 0.05), this duration represents only ~20% of the total delay observed in Fig. 5a. Consistent with this phenotype, WDR5 knockdown cells examined after fixation display a reduced fraction of telophase cells containing a secondary ingression (Fig. 5e, 15% for control clones versus 8%, and 7% for WDR5 knockdown clones 6 and 8, respectively). Taken together, our data suggest that WDR5 promotes abscission primarily by facilitating the formation of secondary ingressions.
Secondary ingression formation depends on midbody microtubule disassembly and FIP3-positive recycling endosomes (37, 38). Compared with the telophase midbody microtubules of control cells, those of WDR5 knockdown HeLa cells occupy a larger area (Fig. 6a, ~30–35% greater width), raising the possibility that the disassembly of midbody microtubules may be impaired in these knockdown cells. To directly test whether this is the case, we added 10 μm nocodazole to control or WDR5 knockdown cells after midbody maturation and utilized live cell imaging to monitor microtubule disassembly within the midbody. Indeed, we found that the midbody microtubules of WDR5 knockdown cells are more resistant to nocodazole, compared with those of control cells (Fig. 6, b and c). These observations suggest that a defect in midbody microtubule disassembly may account for the delayed formation of secondary ingressions in WDR5 knockdown cells.
Elucidation of the specific factor(s) that confers the midbody targeting of WDR5 may provide further insight into the function of midbody-localized WDR5. Although several studies have examined the sequence and structural features required for specific WDR5 protein-protein interactions (15, 19, 52), the mechanisms that regulate its subcellular localization remain largely unexplored. In principle, the top, bottom, and sides of the WDR5 β-propeller are all potential surfaces for protein-protein interactions (53), although interactions that involve a site located at the top (acidic end) of the WDR5 central cavity are best characterized (18, 19, 54). This cavity binds to an ARX containing “WIN” (WDR5-interacting) motif in the target protein. To determine whether interactions at this site influence the midbody localization of WDR5, we investigated the localization of EGFP fusions of WDR5 containing either the S91K or F133A point mutation, each of which blocks the binding of WIN motif-containing proteins to WDR5 (11, 18, 19), in HeLa cells. Interestingly, although wild type EGFP-WDR5 localizes to the nucleus and midbody (Fig. 7a), mutation of either S91K or F133A substantially reduced the midbody localization of WDR5 without affecting its nuclear localization in all cells examined (Fig. 7, b and c). This result suggests that distinct mechanisms control the midbody and nuclear targeting of WDR5 and indicates that a WIN motif-containing protein likely recruits WDR5 to the midbody dark zone.
Based on the observation that mutations within the arginine binding cavity of WDR5 greatly reduce its midbody (but not nuclear) localization, it is possible that uncharacterized structural features of WDR5 in the vicinity of this pocket facilitate its selective targeting to the midbody. Indeed, an EGFP fusion of a 103-amino acid truncation of WDR5 (positions 71–173), which contains many of the residues required for efficient binding to WIN motifs in histone H3 and MLL1 (18, 19, 55,–57), exhibited a highly enriched EGFP signal at the midbody with relatively little EGFP accumulation in the nucleus (Fig. 7d). The selective enrichment of EGFP-WDR5(71–173) at the midbody prompted us to examine whether cells transiently expressing this construct displayed a disrupted cytokinesis phenotype. Although most cells overexpressing EGFP alone completed cytokinesis normally, overexpression of EGFP-WDR5(71–173) led to either cell death during cytokinesis or abscission delays greater than 3 h in 90% of the cells imaged (Fig. 7e), presumably due to the dominant-negative effect of this truncation. These observations indicate that WDR5 contains a domain sufficient for its midbody (but not nuclear) targeting and suggest that midbody-localized WDR5 functions in the completion of cytokinesis.
Because of the intimate link between H3K4 methylation and transcription (58), the H3K4MT literature almost exclusively focuses on the roles of these complexes in the nucleus, and much less is known regarding the cytoplasmic localization and function of H3K4MT subunits. This study identifies a novel function of WDR5 in abscission. We show that the EGFP fusion of WDR5 localizes within the mature midbody during cytokinesis (Figs. 11–3) in the absence of DNA. Knockdown of WDR5 increases the prevalence of multinucleated cells (Fig. 4), a hallmark of cytokinesis failure. Importantly, depletion of WDR5 prolongs the duration between midbody maturation and final scission without significantly affecting the time between anaphase and midbody maturation (Fig. 5). Collectively, these data show that WDR5 promotes abscission, possibly by performing a local function at the midbody. In this regard, it is interesting that that the S91K and F133A mutations impair the localization of WDR5 to the midbody (but not to the nucleus), and that WDR5(71–173) is selectively targeted to the midbody and induces cytokinesis failure (Fig. 7, d and e). These observations suggest that identification of the putative WIN motif-containing midbody protein may allow future dissection of nuclear and midbody functions of WDR5. In addition to multinucleation and abscission delay, cell death during cytokinesis is also associated with the perturbation of several different cytokinesis regulators, either via knockdown (37, 59,–62) or by ectopic expression of the midbody-targeting region of the protein (59). However, the mechanism of cytokinesis-associated cell death has not been extensively investigated.
To the best of our knowledge, WDR5 is the first H3K4MT subunit to be found at the midbody, which raises several intriguing questions regarding the mechanism and functional significance of its recruitment to this location Recently, transient knockdown of WDR5 and other H3K4MT complex members, including RbBP5, ASH2L, mDPY-30, and the catalytic subunit. MLL1, was reported to increase the degree of multinucleation in U2OS cells, although the underlying mechanism was not explored (63). Given this finding, one possibility is that some combination of additional H3K4MT subunits localizes to the midbody and functions together with WDR5 in the context of a methyltransferase complex or subcomplex. In this regard, histone H3 phosphorylated at either Ser-10 or Thr-11 was also previously reported to reside at the midbody in MCF-7 and CHO cells (64, 65). Given the difficulty of obtaining highly specific immunostaining of midbody proteins, this result should be verified with EGFP-histone H3 in live cells. It is also possible that an H3K4MT methyltransferase complex or subcomplex promotes abscission by methylating novel non-histone midbody substrate(s). The recent observation that a midbody-localized demethylase ALKBH4 facilitates cytokinesis by demethylating actin at the cleavage furrow suggests that regulated methylation/demethylation of non-histone proteins may be important for proper cytokinesis (66).
Alternatively, WDR5 may assemble with a set of non-H3K4MT midbody effectors to perform an activity besides protein methylation. Although our immunoprecipitation data suggest that EGFP-WDR5 robustly associates with at least five different midbody proteins (Fig. 2, b and c; PRC1, CYK4, MKLP1, CEP55, and ECT2), it seems unlikely that all are direct interacting partners. If WDR5 performs a scaffolding role for non-H3K4MT proteins, as is common for other WD40 β-propellers (53), it will be important to identify the direct midbody binding partner(s). In this aspect, our study suggests that a WIN motif-containing protein directly binds to the central Arg-binding cavity of WDR5 and targets it to the midbody. The WIN domain has been defined and structurally characterized using 12–19 amino acid peptides designed around six central residues that are highly conserved among human MLL-SET family catalytic subunits in the pattern G(S/C/A)AR(A/S)E (54, 67, 68). However, a 3-mer peptide containing the central invariant alanine and arginine residues (acetyl-ARA-NH2) was recently shown to have a nearly identical Ki (0.12 μm) as the full 12-residue WIN peptide (0.16 μm) (69, 70), suggesting that the ARA peptide captures most of the high affinity interactions between the 12-mer WIN domain and WDR5. Similarly, an alanine-arginine motif is required for the high affinity interaction of WDR5 with the N terminus of histone H3 (ARTKQ) (71), suggesting that a minimal AR(AST) sequence within a putative midbody targeting protein may be sufficient to recruit WRD5.
Only two non-H3K4MT-binding interactions with WDR5 at this site have been characterized so far. WDR5 is required for efficient assembly of the nonspecific lethal complex in the nucleus, and the KANSL1 subunit of this complex binds to the central cavity of WDR5 via a WIN motif (72). Additionally, the influenza virus protein NS1, which primarily localizes to the nucleus of infected cells (73) and is thought to function as a histone mimic, contains two adjacent WIN motifs near its C terminus that bind WDR5 in a similar manner (74). Notably, none of these proteins have been reported to localize to the midbody or function in cytokinesis, suggesting that if WDR5 is recruited to the midbody by a protein other than SET/MLL or histone H3, its identity remains to be discovered. Of the midbody proteins that robustly co-immunoprecipitate with WDR5-EGFP (PRC1, MKLP1, CYK4, ECT2, and CEP55), only MKLP1 and ECT2 contain AR(AST) motifs (MKSARAK and QMDARAG, respectively), and whether either directly recruits WDR5 to the midbody remains to be determined.
Several observations indicate that WDR5 modulates the stability of midbody microtubules during abscission. First, the midbody microtubules in WDR5 knockdown cells are more resistant to nocodazole-mediated depolymerization (Fig. 6, b and c). Second, depletion of WDR5 results in a larger area of midbody microtubules compared with the control (Fig. 6a), and secondary ingression formation is delayed in these cells (Fig. 5, c and e). Both findings are consistent with a defect in midbody microtubule disassembly. Interestingly, PRC1 and MKLP1-CYK4, three known midbody-localized microtubule regulators, represent the most prominent proteins co-immunoprecipitated with WDR5 (Fig. 2, b and c). Identification of the direct binding partner(s) of WDR5 at the midbody should help elucidate the molecular mechanism by which WDR5 knockdown modulates microtubule disassembly. Another question that remains to be answered is the site where WDR5 exerts its effect on microtubules. Although WDR5-EGFP is highly enriched at the center of the midbody, it can also be detected at the midbody ring and along the flanking regions.4
In summary, we have uncovered an unexpected role of the H3K4MT regulatory subunit WDR5 in cytokinesis at the midbody. Considering the intimate relationship between cytokinesis and tumorigenesis, it will be important to elucidate the molecular mechanism by which WDR5 promotes abscission. Moreover, in addition to its tight association with human cancers (75), mutations or misregulation of H3K4MT subunits are directly linked to life span/aging (76), immunity (77), diabetes (78), mental retardation (79), and stem cell differentiation (80,–82). Discovering nontranscriptional functions of H3K4MT subunits, as described here, is essential for understanding the association between these proteins and the above pathophysiological events.
We are grateful to M. Cosgrove for the WDR5 S91K and F133A constructs, W. I. Sundquist for the CHMP1B antibody, L. Johnson for the hTERT-RPE-1 cell line, and to the many undergraduates whose technical assistance contributed to this work. We also thank M. Glotzer for helpful suggestions, C. A. Vandenberg and K. R. Foltz for critically reading this manuscript, T. E. Kippin and K. K. Szumlinski for statistical advice, and Jennifer Knappe for editing assistance. We acknowledge the use of the NRI-MCDB Microscopy Facility and the Spectral Laser Scanning Confocal supported by the Office of The Director, National Institutes of Health Award S10OD010610.
*This work was supported, in whole or in part, by National Institutes of Health Grants 5R21MH086853-02 and 1R21MH086853-01A1 from National Institute of Mental Health (to D. M.).
4D. Ma and J. Bailey, unpublished data.
3The abbreviations used are: