Live Imaging of Drosophila Brain Neuroblasts Reveals a Role for Lis1/Dynactin in Spindle Assembly and Mitotic Checkpoint Control^V⃞

Generation and Characterization of Glued Alleles

We generated a new Glued allele, Gl^Δ22, by imprecise excision of the P-element l(3)KG07739 inserted into the 5′-UTR encoding region of the Glued gene (Bellen et al., 2004). PCR and sequence analysis revealed that Gl^Δ22 contains a ∼3.2-kb deletion removing the presumptive transcriptional start site of the Glued gene (Figure 1B). In addition, the molecular nature of the Gl^1–3 allele (Harte and Kankel, 1982) was determined by sequence comparison of genomic DNA amplified from homozygous Gl^1–3 mutant and Oregon R larvae using standard PCR techniques, revealing a point mutation converting codon 932 into a premature stop codon (Figure 1A). Homozygous Gl^1–3, homozygous Gl^Δ22, and hemizygous Gl^1–3/Df(3L)fz-GF3b animals die as 2nd instar larvae.

Open in a separate window

Figure 1.

Generation and characterization of Glued alleles. (A) Schematic representation of structural domains within the Glued (Gl) protein as predicted by the SMART algorithm: black box: cytoskeleton-associated protein-Glycine (CAP-Gly) MT-binding motif; gray boxes: coiled-coil domains. Also shown is the point mutation in Gl^1–3 converting codon 932 (CAG) into a premature stop codon. (B) Schematic representation of the Glued gene and neighboring annotated genes CG8833 and CG32137. Black boxes: Glued coding region; gray boxes: untranslated regions; black arrow: translational start site; dashed line: insertion site of the P-element l(3)KG07739. Open brackets indicate the extent of the deletion in the Gl^Δ22 allele generated by imprecise excision of the l(3)KG07739 P-element. We also obtained viable lines with precise P-element excision, indicating that lethality in Gl^Δ22 is due to the indicated deletion. Homozygous Gl^1–3, homozygous Gl^Δ22, and transheterozygous Gl^1–3/Gl^Δ22 mutant neuroblasts show indistinguishable phenotypes. The function of the gene CG8833 is unknown.

Generation of Transgenic Fly Lines Expressing GFP-tagged Lis1 Protein

The complete Lis1 coding sequence was amplified from Drosophila EST RE28987 and subcloned into the pUAST vector downstream of, and in-frame with, three repeats encoding EmeraldGFP (Tsien, 1998). Transgenic flies were generated by standard methods. GFP-Lis1 was expressed in larval neuroblasts by crossing pUAST-3xEmeraldGFP-Lis1 transgenic flies to a worniu-Gal4 driver line (Albertson et al., 2004; see Figure 10 and Supplementary Movie 10). In addition, one of us (R.S.) subcloned the Lis1 coding sequence into the pUASP vector downstream of a single GFP coding sequence; this GFP-Lis1 was ubiquitously expressed in embryos using the maternal nanos-Gal4 driver for the immunoprecipitation experiments (see Figure 8B).

Open in a separate window

Figure 8.

Lis1 coimmunoprecipitates with dynactin and cytoplasmic dynein subunits. (A) Immunoprecipitation of Dhc (top row) and Gl (bottom row) proteins from wild-type embryonic lysates with anti-Dhc and anti-Lis1 antibodies (lanes 3 and 4, respectively). The reduced amount of the upper Gl polypeptide in lane 4 accompanies some degradation of the sample and is not the result of selective precipitation of a single Glued polypeptide (see also B, lane 2). (B) Immunoprecipitation of Dhc (top row), Gl (middle row), and GFP-Lis1 (bottom row) protein from lysates of transgenic embryos expressing a GFP-Lis1 fusion protein using anti-GFP and anti-Dhc antibodies (lanes 2 and 3, respectively).

Open in a separate window

Figure 10.

Dynamics of GFP-Lis1 protein localization in live neuroblasts. (A) GFP-Lis1 protein was expressed in wild-type 3rd instar larval brains (Supplementary Movie 10). Shortly after NEB, GFP-Lis1 protein was recruited into dotlike accumulations, likely reflecting kinetochore association (1:15, arrowhead). During metaphase GFP-Lis1 redistributed along spindle microtubules and became enriched on spindle poles (4:15–7:30, arrows). During anaphase and telophase (14: 00–18:00) elevated levels of GFP-Lis1 were only detectable on spindle poles (arrow); only the apical spindle pole is in focus at 18:00. Mitosis was slightly delayed in these neuroblasts, presumably because of higher laser intensities used to visualize the GFP-Lis1 fusion protein. Bar, 5 μm.

Time-lapse Analysis of Neuroblast Cell Division in Larval Brain Explants

Wild-type, Lis1^–, Lis1^– rod^H4.8 (all 48 h ALH), or Gl^1–3 (96 h ALH) late 2nd instar larvae (expressing either G147-GFP, His2AvD-GFP, or GFP-Rod under control of their native promoters) or worniu-Gal4 UAS-GFP-Lis1 wandering 3rd instar larvae were dissected in D-22 (pH 6.75) insect medium (US Biological, Swampscott, MA) supplemented with 7.5% fetal bovine serum (FBS). Four to five larval brains were immediately transferred into 200 μl D-22 medium supplemented with 7.5% FBS, 0.5 mM ascorbic acid, and fatbodies obtained from 10 wild-type wandering 3rd instar larvae. Brains were mounted with fatbody tissue on a standard membrane (Yellow Springs Instruments, Yellow Springs, OH) and placed on a stainless steel slide as previously described (Kiehart et al., 1994). Brains were imaged using a Bio-Rad Radiance 2000 laser scanning confocal microscope equipped with a 60×1.4 NA oil immersion objective. For the analysis of spindle assembly (neuroblasts expressing 2 copies of G147-GFP), stacks containing four focal planes spaced by 1.5 μm were acquired at intervals of 10 s. For the analysis of chromosome movements and cell cycle timing (neuroblasts expressing one copy of His2AvD-GFP) or GFP-Rod localization, stacks containing six focal planes spaced by 1.5 μm were acquired at intervals of 15 s. For the analysis of GFP-Lis1 localization, stacks containing four focal planes spaced by 1.5 μm were acquired at intervals of 15 s. Time-lapse image series were converted into movies using MetaMorph (Universal Imaging, West Chester, PA) and ImageJ. All movie frames are maximum intensity projections.

Antibodies and Immunofluorescent Staining

Drosophila EST RE28987 was used to amplify a Lis1 cDNA encoding amino acids 1–90, which was subcloned into bacterial expression vectors to express 6xHis-tagged Lis1 protein. Purified 6xHis-Lis1 was injected into rats to generate polyclonal antibodies.

Larvae were dissected in Schneider's medium (Sigma, St. Louis, MO), fixed in 100 mM Pipes (pH 6.9), 1 mM EGTA, and 1 mM MgCl₂ for 25 min and blocked for 1 h in 1× phosphate-buffered saline (PBS) containing 1% bovine serum albumin and 0.1% Triton X-100 (PBS-BT). For labeling of DNA with propidium iodide, RNase A was added to a final concentration of 1 μg/ml. After blocking, specimen were extensively washed in PBS-BT for 1 h and incubated with primary antibodies in PBS-BT overnight at 4°C. Primary antibodies were: rat anti-Lis1 (1:2500; this study); rabbit anti-Gl (raised against Gl C-terminus, 1:150; Waterman-Storer and Holzbaur, 1996); rabbit anti-Cnn (1:1000; Heuer et al., 1995); rabbit anti-nPKCζ (Santa Cruz Biotechnology, Santa Cruz, CA; 1:500); rat anti-Miranda (1:1000; Irion et al., 2004); mouse anti-α-tubulin (DM1A, Sigma, 1:2000); rat anti-α-tubulin (MCA78S, Serotec, Raleigh, NC; 1:100); rabbit anti-Cid (1:500; Henikoff et al., 2000); mouse anti-γ-tubulin (GTU-88, Sigma, 1:2000); rabbit anti-phospho-histone H3 (Upstate Biotechnology, Lake Placid, NY; 1:1000); rabbit anti-Rod (1:200; Scaerou et al., 1999). Primary antibodies were extensively rinsed off with PBS-BT for 1 h at room temperature, and specimens were incubated with fluorescently conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA, and Molecular Probes, Eugene, OR) diluted in PBS-BT, followed by extensive rinsing with PBS-BT. For DNA labeling, specimens were mounted in FluoroGuard Antifade Reagent (Bio-Rad, Richmond, CA) containing 2.5 μg/ml propidium iodide. Brains were imaged using a Bio-Rad Radiance 2000 or Leica TCS SP2 laser scanning confocal microscope (Deerfield, IL) equipped with a 60× 1.4 NA or 63× 1.4 NA oil immersion objective, respectively. Figures were assembled in Adobe Photoshop (San Jose, CA).

Cell Cycle Analysis in Fixed Specimens

Larval brains were labeled for phospho-histone H3 (mitotic DNA), α-tubulin, and Miranda. Neuroblasts were identified by size and expression of Miranda protein, and scored for cell cycle stage using phospho-histone H3 and α-tubulin.

Immunoprecipitation Experiments

Immunoprecipitation experiments were conducted in high-speed supernatants at 4°C, and no microtubules were present. In detail, 0–24-h embryos were homogenized in 2.5 volumes IP buffer (50 mM HEPES, pH 7.2, 150 mM KCl, 0.9 M glycerol, 0.5 mM dithiothreitol, with protease inhibitors 10 μg/ml aprotinin, 1 μg/ml each leupeptin and pepstatin, 0.1 μg/ml each of soybean trypsin inhibitor, n-tosyl-l-arginine methylester, and benzamidine), and then supplemented with Triton X-100, 0.1%. After ultracentrifugation at 25,000 rpm for 20 min in a 50ti rotor, the supernatant was collected and precleared against washed protein A-Sepharose beads (Sigma). Cleared extract, 700 μl, was incubated 2 h at 4 °C with beads that had previously been bound anti-Dhc monoclonal P1H4 (dynein heavy chain; McGrail and Hays, 1997), anti-GFP monoclonal 3E6 (Molecular Probes), or anti-Lis1 antibodies. Beads were washed in IP buffer three times, the last two washes without detergent. Pellets were eluted into 20 μl 2× sample buffer, and samples were analyzed by SDS-PAGE followed by Western analysis using anti-Dhc monoclonal P1H4 (McGrail and Hays, 1997); rabbit anti-Gl C-terminal (Waterman-Storer and Holzbaur, 1996) or anti-GFP monoclonal JL-8 (Clontech, Palo Alto, CA) antibodies.

Lis1 and Dynactin Are Required for Centrosome Separation and Spindle Assembly

Time-lapse imaging of wild-type second instar larval neuroblasts expressing G147-GFP showed that duplicated centrosomes stay in close proximity to each other until they separate during prophase. By the onset of prometaphase, the pair of centrosomes was always completely separated and positioned on opposite sides within the neuroblast (average separation 171 ± 7°, n = 15; Figure 2, A (0:00) and C, Supplementary Movie 1). As prometaphase progressed, the centrosomes nucleated MTs that formed a straight bipolar spindle (Figure 2A (0:00–6:10), Supplementary Movie 1).

Open in a separate window

Figure 2.

Lis1 mutant neuroblasts exhibit defects in centrosome separation and spindle formation. (A and B) Dynamics of centrosome separation and spindle formation in larval neuroblasts expressing the G147-GFP MT-associated protein. In this and all subsequent figures, time is shown in min:sec relative to the onset of prometaphase (NEB, 0:00). Onset of prometaphase was determined as the moment the spindle microtubules penetrated into the cell center. Red lines indicate degree of centrosome separation calculated as the angle encompassed by two lines that were connected in the cell center and whose outer points were defined by the spindle poles at the time of prometaphase onset (0:00). Bars indicate cell cycle stages (see legend below B). (A) In wild-type neuroblasts, centrosomes were well separated and positioned on opposite sites of the nucleus at late prophase (–5:00). Centrosomes could rotate after this stage, but both remained positioned on opposite sides of the nucleus. After NEB (0:00), MTs formed a straight bipolar spindle (0:00–6:10; Supplementary Movie 1). (B) In Lis1^– mutant neuroblasts, centrosome separation was frequently incomplete during prophase (–8:50–0:00). After NEB, MTs often had a wavy morphology and frequently assembled into a Y-shaped spindle (2:40). A central bipolar spindle generally formed after several minutes, although occasionally both MTOCs remained associated with the same spindle half (1:30–25:40, arrowheads). A few Lis1^– neuroblasts contained more than two MTOCs, which usually reassociated with the spindle apparatus before cytokinesis (9:40–17:50, arrowheads; Supplementary Movie 2). (C) Quantification of centrosome separation defects in Lis1^– mutant neuroblasts. (D–M) Wild-type, Lis1^–, and Gl^Δ22 mutant metaphase neuroblasts of 2nd instar larvae were double-labeled for α-tubulin and the centrosomal markers Centrosomin (Cnn; D–H) or γ-tubulin (I–M). Arrowheads indicate centrosomes. (D and I) Wild-type metaphase neuroblasts formed straight bipolar spindles with two centrosomes focusing MTs at the spindle poles. (E–H and J–M) Lis1^– and Gl^Δ22 metaphase spindles exhibited various morphological defects, including curved spindles (E and H), one or two centrosomes unattached to spindle (F), two centrosomes on the same half spindle (J), or occasionally extra centrosomes (G, L, and M). Similar defects were observed in Gl^1–3/Gl^Δ22 mutant neuroblasts; see Table 1. Lis1 and Gl mutant larval neuroblasts were on average smaller than wild-type counterparts. Bars, 5 μm.

To investigate the role of Lis1 in centrosome separation and spindle formation, we performed a similar time-lapse analysis on 2nd instar larval neuroblasts transheterozygous for two previously described Lis1 alleles, Lis1^G10.14 and Lis1^k13209 (Liu et al., 1999). In Lis1^G10.14/Lis1^k13209 (hereafter referred to as Lis1^–) larval brains, Lis1 protein was reduced below detectable levels (see below and Figure 9). In contrast to wild-type, time-lapse imaging of Lis1^– mutant neuroblasts expressing G147-GFP revealed that centrosome separation was frequently incomplete at the onset of prometaphase (average separation 124 ± 38°, n = 12; Figure 2, B (0:00) and C, Supplementary Movie 2), and MTs emanating from both spindle poles often developed a Y-shaped morphology during prometaphase, presumably because of the close apposition of the centrosomes (Figure 2B (0:00–2:40), Supplementary Movie 2). Strikingly, these Y-shaped spindles generally “recovered” to form seemingly bipolar arrays of kMTs by late metaphase (Figure 2B (17:50–24:10), Supplementary Movie 2). However, despite the apparent bipolar spindle organization, microtubule organizing centers (MTOCs) occasionally detached from spindles or two MTOCs appeared to maintain attachment to the same side of a half-spindle. In few cases neuroblasts contained more than the normal two MTOCs (Figure 2B (17:50–24:10), Supplementary Movie 2). We conclude that Lis1 is required for centrosome separation, proper spindle assembly, and regulation of MTOC number.

Open in a separate window

Figure 9.

Lis1 and Gl protein localize to spindle poles/centrosomes, kinetochores, and spindle microtubules. (A–K) Neuroblasts of wild-type (A, B, F, and G), dhc^6–10 (C and H), Lis1^– (D and I), Gl^1–3 (E and J), and Gl^Δ22 (K) mutant larvae were double-labeled for α-tubulin and Lis1 (A–E) or α-tubulin and Gl (antibody raised against Gl carboxy-terminus; Waterman-Storer and Holzbaur, 1996; F–K). Spindle poles/centrosomes (arrows) and kinetochores/kMTs (arrowheads) are indicated. Lis1 and Gl protein levels are reduced to background levels in Lis1^– and Gl mutant metaphase neuroblasts, respectively, demonstrating the specificity of the Lis1 and Gl antibodies (D, J, and K). Bar, 5 μm. (L) Lis1 protein (∼45 kDa) is almost undetectable in lysates of 2nd instar Lis1^– mutant larvae compared with those of similarly aged wild-type (wt) larvae (arrow). Bottom panel: α-tubulin as loading control. The Lis1 antibody unspecifically detects a band of ∼150 kDa (asterisk); this cross reactivity is unlikely to interfere with the Lis1 localization in neuroblasts as signals obtained with this antibody are reduced to background levels in Lis1^– mutant neuroblasts (D). (M) Gl protein is undetectable in Gl^1–3 and Gl^Δ22 2nd instar larval lysates using an antibody raised against the Gl amino-terminus (Fan and Ready, 1997). Bottom panel: α-tubulin as loading control.

To analyze spindle formation defects at higher spatial resolution and to determine the nature of the observed supernumerary MTOCs, we assayed centrosome (Centrosomin and γ-tubulin) and spindle markers (α-tubulin) in fixed neuroblasts. We found that wild-type metaphase neuroblasts had a straight bipolar spindle with each spindle pole tightly focused on a single centrosome (Figure 2, D and I, Table 1). In contrast, Lis1^– mutant neuroblasts showed curved spindles, unfocused spindle poles, spindles with one or two unattached centrosomes, or even spindles in which both centrosomes were attached to the same half-spindle (Figure 2, E, F, J, and K; Table 1). Strikingly, some of these metaphase neuroblasts appeared to contain 3–4 centrosomes with multipolar spindles, bipolar spindles with more than one centrosome on each half spindle, or bipolar spindles with 1–2 unattached centrosomes (Figure 2, G and L, Table 1). Because of the lack of suitable Drosophila centriole markers, we were unable to determine whether all ectopic centrosomelike (Centrosomin/γ-tubulin positive) structures contained centrioles. We conclude that Lis1 has a critical role in correct spindle assembly by regulating centrosome separation, focusing of spindle poles, centrosome attachment, and centrosome number and/or centrosome integrity in larval neuroblasts. The presence of a subset of Lis1^– mutant neuroblasts with normal bipolar spindles, however, allowed us to examine these neuroblasts for defects in later steps of the cell cycle (see next section).

Lis1 functions together with dynactin in many cell types, so we wanted to determine if dynactin was also required for spindle formation. Well-characterized loss-of-function dynactin mutants do not exist in Drosophila, so we generated a null mutation in Glued (Gl^Δ22), which encodes the largest dynactin subunit, as well as molecularly characterized an extant allele, Gl^1–3 (Harte and Kankel, 1982; Figure 1; see Materials and Methods). Both Gl^Δ22 and Gl^1–3 appeared to comprise protein-null alleles (see below, Figure 9). We found that Gl^Δ22 homozygous or Gl^1–3 hemizygous mutant neuroblasts phenocopied centrosome and spindle formation defects observed in Lis1^– mutants. Defects included metaphase neuroblasts with curved spindles, unfocused spindle poles, spindles with one or two unattached centrosomes, spindles in which both centrosomes were attached to the same half-spindle, or occasionally neuroblasts with three or four centrosomes forming bipolar or multipolar spindles (Figure 2, H and M, Table 1, and unpublished data). Thus, both dynactin and Lis1 promote centrosome separation and proper spindle formation in larval Drosophila neuroblasts.

Lis1 and Dynactin Are Required for Timely Anaphase Onset

In wild-type neuroblasts (expressing G147-GFP), the duration of prometaphase and metaphase (from NEB to anaphase onset) was quite rapid and highly reproducible, lasting 6 min 12 s ± 0 min 49 s (n = 18; Figure 3, A and D, Supplementary Movie 1). In contrast, Lis1^– mutant neuroblasts showed dramatic lengthening of the prometaphase/metaphase interval to an average of 46 min 54 s ± 18 min 33s(n = 11, Figure 3, B and D, Supplementary Movie 3). We reasoned that this delay in anaphase onset might be due to extended mitotic checkpoint activity. To test this, we reduced Rod function to bypass the checkpoint (Basto et al., 2000). In Lis1^– rod^H4.8 double mutant neuroblasts, progression through prometaphase/metaphase only took 11 min 31 s ± 4 min 23 s (n = 11, Figure 3, C and D, Supplementary Movie 4), confirming that the delay in anaphase onset is checkpoint-dependent. Note that centrosome separation and spindle assembly defects were still present in Lis1^– rod^H4.8 double mutant neuroblasts, indicating that these defects were independent of altered mitotic checkpoint signaling (Figure 3C, Supplementary Movie 4, and unpublished data). In addition, we also calculated mitotic index and metaphase:anaphase ratio in fixed specimens of wild type, Gl single, Lis1^– single, and Lis1^– rod^H4.8 double mutants. Both mitotic index and metaphase:anaphase ratio of neuroblasts were increased in Lis1^– and Gl single mutants compared with wild-type and Lis1^– rod^H4.8 double mutant neuroblasts (Table 2). Thus, the observed delays in anaphase onset in Lis1 and Gl mutant neuroblasts are consistent with extended mitotic checkpoint activity, indicating a role for Lis1/dynactin in satisfying or inactivating the mitotic checkpoint.

Open in a separate window

Figure 3.

Lis1 mutant neuroblasts exhibit mitotic checkpoint-dependent delays in anaphase onset. Cell cycle timing was determined in larval neuroblasts expressing the G147-GFP MT-associated protein. Bars indicate cell cycle stages (see legend below C). (A) In wild-type neuroblasts, prometaphase and metaphase took 5–8 min (similar to Supplementary Movie 1). (B) In Lis1^– mutant neuroblasts, prometaphase/metaphase took 27–71 min (Supplementary Movie 3). Arrows indicate apparent detachment of kMTs from MTOCs. (C) In Lis1^– rod^H4.8 double mutant neuroblasts, prometaphase/metaphase took 6–22 min (Supplementary Movie 4). Note that spindle assembly was still defective and MTOCs could detach from the spindle apparatus (arrowheads). The GFP signal appears dimmer (compared with signal of neuroblasts depicted in A and B) because the neuroblast was positioned deeper in the brain. All neuroblasts were imaged using identical laser settings. Bar, 5 μm. (D) Quantification of prometaphase/metaphase duration in wild-type, Lis1^– single mutant, and Lis1^– rod^H4.8 double mutant neuroblasts. The left histogram depicts duration of prometaphase/metaphase for 10 neuroblasts of each genotype (including neuroblasts with shortest and longest prometaphase/metaphase as well as 8 randomly chosen neuroblasts for each genotype). The right histogram depicts the calculated mean duration of prometaphase/metaphase + SD for each genotype (wild type (n = 18), Lis1^– (n = 11), Lis1^– rod^H4.8 (n = 11)).

To measure the length of prometaphase and metaphase individually, we expressed a single copy of a GFP-tagged histone variant His2AvD under control of its native promoter (His2AvD-GFP; Clarkson and Saint, 1999) in larval neuroblasts. We defined prometaphase as the interval from NEB (evident by a slight increase of cytoplasmic His2AvD-GFP fluorescence intensity not associated with chromosomes) to the moment of completed chromosome congression on the metaphase plate, and we defined metaphase as the interval between completion of chromosome congression and initiation of poleward chromosome movement. In wild-type neuroblasts, prometaphase and metaphase took 4 min 14 s ± 1 min 31 s and 2 min 24 s ± 1 min 13 s, respectively (n = 12, Figure 4, A and D, Supplementary Movie 5). Throughout metaphase, all chromosomes remained aligned in a tight metaphase plate. In contrast, in Lis1^– mutant neuroblasts both prometaphase (mean duration 32 min 18 s ± 15 min 55 s) and metaphase (mean duration 19 min 21 s ± 16 min 35 s) were significantly prolonged (n = 10, Figure 4, B–D, Supplementary Movies 6 and 7). These data indicate that Lis1/Gl affects cell cycle timing in at least two ways. First, Lis1/Gl is required for timely prometaphase progression, possibly by promoting spindle assembly and thereby facilitating efficient MT-kinetochore capturing/chromosome congression. Second, Lis1/Gl is also required for the timely initiation of the metaphase-to-anaphase transition, possibly by contributing to mitotic checkpoint inactivation.

Open in a separate window

Figure 4.

Lis1 mutant neuroblasts show delayed chromosome congression to the metaphase plate and prolonged metaphase. Chromosome movement and cell cycle timing was assayed in larval neuroblasts expressing GFP-tagged His2AvD protein. Bars indicate cell cycle stages (see legend below C). (A) In wild-type neuroblasts, chromosomes completed congression into a tight metaphase plate within 2–7 min after NEB (Supplementary Movie 5). These neuroblasts stayed in metaphase for 2–4 min. (B and C) In Lis1^– mutant neuroblasts, chromosome congression was severely delayed and prometaphase took 17–63 min (Supplementary Movies 6 and 7). Metaphase was also prolonged and took 3–49 min. In two neuroblasts we observed chromosomes that were lost from and recongressed to the metaphase plate (C, 25:30–36:15, arrowheads, Supplementary Movie 7). Bar, 5 μm. (D) Quantification of prometaphase and metaphase duration in wild-type and Lis1^– mutant neuroblasts. Depicted are durations of prometaphase and metaphase for 10 neuroblasts of each genotype (for wild type including neuroblasts with shortest and longest prometaphase/metaphase interval as well as 8 randomly chosen neuroblasts).

Lis1 and Gl Are Required to Transport Checkpoint Proteins from Kinetochores and Generate Interkinetochore Tension

Loss of Lis1/dynactin function could cause prolonged mitotic checkpoint activity by several mechanisms, including impairment of MT-kinetochore attachment (Rieder et al., 1995), a reduction of interkinetochore tension (Li and Nicklas, 1995), or a defect in the transport of checkpoint proteins off kinetochores (Howell et al., 2001; Wojcik et al., 2001). Here we tested which, if any, of these mechanisms were affected in Lis1 and Gl mutant neuroblasts.

We first tested whether defective MT-kinetochore attachment was the primary cause of delayed metaphase-to-anaphase transition in Lis1 and Gl mutant neuroblasts. Analysis of fixed preparations stained for spindle, DNA, and/or kinetochore markers revealed that in Lis1 and Gl single mutant metaphase neuroblasts typically all chromosomes had congressed into a tight metaphase plate with kinetochore fibers abutting kinetochores (Lis1^–: 72.5%, n = 102; Gl^1–3: 75.0%, n = 40; Gl^1–3/Df(3L)fz-GF3b: 69.2%, n = 107; Figure 5, C and D). In addition, we followed chromosome movement (visualized with His2AvD-GFP) in Lis1^– mutant neuroblasts using time-lapse analysis. We found that during prometaphase chromosomes showed delayed congression to the equatorial plate but eventually aligned into a tight metaphase plate (Figure 4, B and C, Supplementary Movies 6 and 7, n = 10). In two Lis1^– mutant neuroblasts, we observed formation of a tight metaphase plate, subsequent chromosome loss, and recongression of the lost chromosome to the metaphase plate (Figure 4C (17:15–40:45), Supplementary Movie 7). Importantly, even after congression of all chromosomes into a tight metaphase plate, Lis1^– mutant metaphase neuroblasts showed delayed transition into anaphase (Figure 4B (25:15–33:15) and 4C (40:45–45:30), Supplementary Movies 6 and 7). Although chromosome separation in Lis1^– mutant anaphase neuroblasts was slightly slower than in wild-type counterparts, chromosomes were always completely partitioned in both daughter cells in these mutants (n = 10). We conclude that defective MT-kinetochore attachment is unlikely to be the only cause of extended checkpoint activity in Lis1 (and Gl) mutant neuroblasts.

Open in a separate window

Figure 5.

Tension between sister centromeres is reduced in Lis1 and Gl mutant metaphase neuroblasts. Wild-type (A, B), Lis1^– (C), and Gl^1–3 (D) mutant neuroblasts were triple labeled for α-tubulin, Cid (centromeres), and DNA. Top and middle rows show maximum intensity projections of three-dimensional image stacks containing 10–20 0.2-μm sections. Squares indicate regions shown at higher magnification in bottom row. Bottom row shows individual sister centromere pairs in single focal planes at high magnification. Brackets indicate sister centromere pairs. (A) In wild-type prophase neuroblasts, centromere pairs were closely apposed due to the absence of MT attachment and tension. (B) In wild-type metaphase neuroblasts, centromere pairs were aligned at the equatorial plate. Because of MT attachment and generation of tension on kinetochores, the distance between sister centromeres was increased. (C and D) In Lis1^– (C) and Gl^1–3 (D) mutant metaphase neuroblasts, distance between sister centromeres was slightly smaller than its wild-type metaphase counterparts, but still significantly larger than between wild-type prophase sister centromeres. (E) Quantification of measured distances between sister centromeres in prophase and metaphase neuroblasts of the indicated genotypes. Depicted are means + SDs. Student's t test analysis revealed no significant difference in sister centromere distance in wild-type, Lis1^–, and Gl^1–3 mutant prophase neuroblasts (p > 0.2 in all pairwise combinations). Sister centromere distance was slightly but significantly reduced in Lis1^– and Gl^1–3 mutant metaphase neuroblasts compared with wild-type counterparts (p < 0.001, indicated by brackets and asterisks). No statistically significant difference in sister centromere distance was observed in Lis1^– and Gl^1–3 mutant neuroblasts (p > 0.1). Numbers of analyzed centromere pairs are depicted at the bottom of each histogram bar. Bars, 5 μm in top and middle rows, 1 μm in bottom row.

We next analyzed whether Lis1 or Gl mutants showed loss of interkinetochore tension. Kinetochores assemble on specific DNA regions called centromeres. Centromeric nucleosomes contain a histone H3-like protein, named CENP-A or Cid (Centromere Identifier) in flies (Henikoff et al., 2000). Increased tension between sister kinetochores is evident by increased distance between sister centromeres (intercentromere distance). During prophase, in the absence of MT-kinetochore attachment, the measured intercentromere distance reflects the “resting length” corresponding to 0% tension. We found that wild-type, Lis1 mutant, and Gl mutant prophase neuroblasts showed statistically indistinguishable intercentromere resting lengths (wild type: 0.51 ± 0.08 μm, n = 36; Lis1^–: 0.51 ± 0.07 μm, n = 20; Gl^1–3: 0.55 ± 0.07 μm, n = 5; Student's t test analysis revealed p values > 0.2 in all pair wise combinations, Figure 5, A and E). On entry into metaphase, the wild-type intercentromere length increased to 1.10 ± 0.13 μm (n = 36, Figure 5, B and E), which presumably reflects 100% tension. In Lis1 and Gl mutant metaphase neuroblasts the intercentromere distance was slightly but significantly reduced compared with wild-type metaphase counterparts (Lis1^–: 0.97 ± 0.16 μm, n = 58; Gl^1–3: 0.93 ± 0.13 μm, n = 25; p < 0.001; Figure 5, C–E). Assuming a linear correlation between measured intercentromere distance and generated tension, we calculated that tension was reduced to ∼77 and ∼70% in Lis1 and Gl mutant metaphase neuroblasts, respectively. Thus, it is possible that the observed reduction in interkinetochore tension activates the mitotic checkpoint and delays anaphase onset in Lis1 and Gl mutant neuroblasts (see Discussion).

To test whether checkpoint protein localization was normal in Lis1 and Gl mutant neuroblasts, we assayed the dynamic localization of GFP-Rod. GFP-Rod was expressed under the control of the native Rod promoter in a rod mutant background (Scaerou et al., 1999; Basto et al., 2004). In wild-type neuroblasts, GFP-Rod was excluded from the nucleus during interphase and prophase, but accumulated on unattached kinetochores at the onset of prometaphase (Figure 6A (0:00), Supplementary Movie 8). At metaphase, GFP-Rod moved from kinetochores onto kMTs (Figure 6A (2:15–5:30), Supplementary Movie 8), presumably as a result of poleward streaming along kMTs (Basto et al., 2004). In contrast, Lis1^– mutant neuroblasts lacked poleward GFP-Rod streaming and thus showed persistent GFP-Rod kinetochore localization during metaphase and anaphase (Figure 6B (0:00–49:15), Supplementary Movie 9). In addition, we confirmed that GFP-Rod resembled endogenous Rod localization by analyzing fixed preparations. In wild-type neuroblasts, Rod localized to prometaphase kinetochores and redistributed along kMTs during metaphase (Figure 7, B and C). In Lis1, Gl, and dhc (dynein heavy chain) mutant neuroblasts we observed persistent high level of Rod localization at metaphase kinetochores that were seemingly attached to MTs (Figure 7, D–F), consistent with the notion that a Lis1/dynactin/dynein complex is required for timely removal of the Rod checkpoint protein from kinetochores at metaphase.

Open in a separate window

Figure 6.

Dynamics of Rod localization in wild-type and Lis1 mutant neuroblasts. GFP-Rod protein was expressed in wild-type (A) and Lis1^– mutant (B) larval neuroblasts. Dotted circles indicate neuroblast outlines. Bars indicate cell cycle stages (see legend below B). (A) In wild-type interphase neuroblasts, GFP-Rod was weakly detectable in the cytoplasm, but not in the nucleus (–3:00). During prometaphase, GFP-Rod accumulated on unattached kinetochores shortly after NEB (0:00–1:15, arrowheads), and was detectable along spindle fibers (arrows) during prometaphase (2:15) and metaphase (3:45–5:30). During anaphase GFP-Rod gradually disappeared from spindle kMTs, but was still detectable on kinetochores (6:45–7:45; Supplementary Movie 8). (B) In Lis1^– mutant neuroblasts, GFP-Rod was detectable on kinetochores during prometaphase (0:00–17:45), but failed to redistribute along spindle MTs during metaphase (23:15–40:15). GFP-Rod protein remained localized to kinetochores during anaphase/telophase (42:30–49:15; Supplementary Movie 9). Bar, 5 μm.

Open in a separate window

Figure 7.

Rod protein transport off kinetochores is impaired in Lis1, Gl, and dhc mutant neuroblasts. Wild type (A–C), dhc^6–10 (D), Gl^1–3 (E), and Lis1^– (F) mutant neuroblasts were triple-labeled for Lis1 (top row), Rod (middle row), and α-tubulin. Merged images are shown in bottom row. Arrows and arrowheads indicate spindle poles and kinetochores, respectively. (A) Wild-type prophase neuroblast with Lis1 but not Rod on centrosomes. (B) Wild-type prometaphase neuroblast showing colocalization of Lis1 and Rod on kinetochores. (C) Wild-type metaphase neuroblast showing Lis1 and Rod codistributed along spindle MTs. (D) dhc^6–10 mutant late prometaphase/early metaphase neuroblast showing colocalization of Lis1 and Rod protein on kinetochores. (E) Gl^1–3 mutant metaphase neuroblast with Rod localization on kinetochores. Lis1 protein is not detectable on kinetochores above cytoplasmic levels. (F) Lis1^– mutant metaphase neuroblasts with Rod localization on kinetochores. Bar, 5 μm.

Lis1 and Dynactin Coimmunoprecipitate with Dynein and Colocalize with the Checkpoint Protein Rod on Kinetochores

We have shown that Lis1 and dynactin have similar functions throughout the cell cycle, ranging from centrosome separation to generation of interkinetochore tension and checkpoint protein transport. We next determined whether Lis1/dynactin are physically associated in vivo, and we analyzed subcellular localization of Lis1/dynactin throughout the neuroblast cell cycle. Using anti-Lis1 antibodies, we could immunoprecipitate both dynein and dynactin subunits, and in a reciprocal experiment we used anti-Dhc antibodies to immunoprecipitate Gl protein (Figure 8A). We also expressed full-length Lis1 protein fused to GFP (GFP-Lis1) in wild-type embryos and showed it could immunoprecipitate Dhc and Gl proteins; similarly, both anti-Dhc and anti-Gl antibodies could immunoprecipitate GFP-Lis1 (Figure 8B and unpublished data). From these results we conclude that some Lis1, dynactin, and dynein proteins are associated in a complex in Drosophila embryos. However, our immunoprecipitations were not quantitative, and we did not determine the percentage of each protein that exists together in a complex. Our findings support previous observations reported for the interaction of dynein, Lis1, and dynactin in mammalian brain cytosolic extracts (Faulkner et al., 2000; Smith et al., 2000).

We next assayed the subcellular localization of Lis1 and Gl proteins in mitotic neuroblasts. Lis1 and Gl showed spindle pole/centrosome association from late prophase through telophase (Figure 9, A-B and F-G, and unpublished data). Both proteins were colocalized with Rod on prometaphase kinetochores (Figure 7B and unpublished data), and distributed along kMTs during metaphase (Figures ​(Figures7C7C and 9, B and G). During anaphase/telophase, Lis1/dynactin staining intensity was diminished on kMTs (unpublished data). A similar localization at centrosomes and kinetochores has been reported for dynein (Pfarr et al., 1990; Steuer et al., 1990; Starr et al., 1998; Gonczy et al., 1999; Wojcik et al., 2001). In Lis1 and Gl mutant neuroblasts, Lis1 and Gl proteins were undetectable at all of these locations, respectively, demonstrating that labeling of these structures with the anti-Lis1 and anti-Gl antibodies was specific (Figure 9, D, J, and K). Furthermore, we observed the same localization of a GFP-tagged full-length Lis1 (GFP-Lis1) protein in live neuroblasts. In prophase neuroblasts, GFP-Lis1 was excluded from the nucleus. With the beginning of prometaphase GFP-Lis1 was strongly associated with kinetochores. During late prometaphase/metaphase, GFP-Lis1 made a transition from kinetochore to centrosomal/spindle localization, whereas during anaphase and telophase there was progressively less GFP-Lis1 associated with the mitotic spindle (Figure 10A, Supplementary Movie 10).

In summary, Lis1/dynactin/dynein coimmunoprecipitate, localize to centrosome/spindle poles, and are transiently colocalized with the checkpoint protein Rod on prometaphase kinetochores before enrichment on kMTs. These results are consistent with a function of Lis1/dynactin/dynein in centrosome separation and spindle assembly as well as kinetochore-based checkpoint function (see Discussion).

The Role of Lis1/Dynactin in Regulating Cell Cycle Timing and Mitotic Checkpoint Signaling

Our time-lapse imaging experiments showed that loss of Lis1/Gl in neuroblasts results in extension of both prometaphase and metaphase. Prometaphase in Lis1 mutant neuroblasts was characterized by delayed congression of chromosomes to the equatorial plate, which is likely to be largely due to inefficient kinetochore capturing as an indirect result of spindle assembly defects. Importantly, in Lis1 mutant neuroblasts congression of all chromosomes into a tight metaphase plate eventually occurred, suggesting that Lis1/Gl are not absolutely critical for MT/kinetochore attachment per se.

In addition we observed severe delays in metaphase-to-anaphase transition. A few of these neuroblasts showed individual chromosomes that were transiently lost from and recongressed to the metaphase plate. Thus, consistent with findings in mammalian cells (Faulkner et al., 2000), Lis1 appears to play some role in maintaining stable chromosome alignment in metaphase neuroblasts. However, in contrast to Faulkner et al. (2000), we found that loss of Lis1 function caused delays in metaphase-to-anaphase transition even when all chromosomes stayed aligned in a tight metaphase plate. Thus, mitotic checkpoint activity remained high even after apparent bipolar kinetochore attachment. Two defects appear to contribute to prolonged checkpoint activity in Lis1 mutant metaphase neuroblasts: reduced interkinetochore tension and failure to transport checkpoint proteins (e.g., Rod) off kinetochores. Reduced interkinetochore tension may be due to lack of Lis1/dynactin on kinetochores or on spindle pole/MTs (which may affect forces acting on kinetochore pairs as a consequence of altered spindle morphology or MT dynamics). Defects in Rod checkpoint protein transport off kinetochores can be explained as a direct consequence of depletion of kinetochore-associated Lis1/dynactin/dynein motor complex, which in wild-type cells is loaded with Rod at kinetochores. However, previous studies indicated that Rod and Zw10 are removed from kinetochores in response to interkinetochore tension not MT attachment (Williams et al., 1996; Scaerou et al., 2001; Basto et al., 2004). Therefore, in addition to its direct role as a “carrier,” Lis1/dynactin/dynein may also play an indirect role in modulating Rod transport by generating the interkinetochore tension required to trigger initiation of Rod streaming.

In summary, our data are consistent with and extends a model recently proposed for dynein function in checkpoint protein transport in Drosophila and mammalian cells (Howell et al., 2001; Wojcik et al., 2001; Basto et al., 2004). According to this model a Lis1/dynactin/dynein-Rod/Zw10 complex, preassembled on unattached kinetochores, is critical for timely anaphase onset by promoting poleward streaming of checkpoint proteins away from kinetochores after correct kinetochore-MT attachment has occurred (Figure 11B). Our data demonstrate that in Drosophila, the Lis1 protein is an obligate component in this process. Although the Lis1-binding proteins NudE/Nudel have been implicated in facilitating dynein-dependent checkpoint protein transport (Yan et al., 2003), it remains to be directly tested whether Lis1 has a similar function in mammalian cells.

What is the link between Rod/Zw10 and Mad2 in mitotic checkpoint function? Two recent studies demonstrated that the Rod/Zw10 complex is required for efficient recruitment of Mad2 to unattached kinetochores in mammalian cells and Drosophila neuroblasts (Buffin et al., 2005; Kops et al., 2005) and that Mad2 and Rod colocalize during poleward transport along kMTs in Drosophila neuroblasts (Buffin et al., 2005). Although a physical link between the Rod/Zw10 complex and Mad2 has not been discovered, an attractive model is that Rod/Zw10 links Mad2 to the Lis1/dynactin/dynein complex during poleward checkpoint protein transport (Buffin et al., 2005; Kops et al., 2005).

We are grateful to Drs. Douglas Kankel, Don Ready, Roger Karess, Erika Holzbaur, Thomas Kaufman, Liqun Luo, Xavier Morin, Gary Karpen, Steve Henikoff, Robert Saint, and the Bloomington Stock Center for sharing reagents and fly lines. We especially thank Dr. Roger Karess for communicating results before publication, Sarah Siegrist for her help in developing the time-lapse imaging technique in larval brain explants, Amy Sheehan and Dr. Melissa Rolls for generating the pUAST-3xEmeraldGFP construct, Susie Shrimpton and Laurina Manning for technical assistance, and Sarah Siegrist as well as Drs. Stephan Schneider, Julie Canman, Bruce Bowerman, and Roger Karess for stimulating discussions. This work was supported by the American Heart Association (predoctoral fellowship to K.H.S.), the National Institutes of Health (Grant GM 053695 to T.S.H.), and the Howard Hughes Medical Institute (where C.Q.D. is an Investigator).