Serine/threonine phosphorylation marks generated by ~ 400 kinases are reversed by a handful of phosphatases that are targeted to substrates via dozens of regulatory subunits10
. For most signaling networks, including those required for the establishment of chromosome-microtubule attachments during prometaphase, it remains unclear which, or even if, a specific phosphatase regulatory subunit is important. In mammals, PP1 and PP2A, the most abundant eukaryotic serine/threonine phosphatases, target to kinetochores (PP1)11
or centromeres (PP2A)12–14
during mitosis. PP1, however, localizes to kinetochores only after chromosomes have bi-oriented and preventing PP1 targeting to the kinetochore does not impair chromosome alignment15, 16
, suggesting it is not essential for establishment of kinetochore-microtubule interactions in prometaphase. Therefore, we decided to examine if PP2A plays a role in this process.
We reasoned that a high-resolution live-cell localization analysis might inform on PP2A functions at centromeres. PP2A holoenzymes are composed of a common catalytic and scaffold subunit, and a variable regulatory subunit17
(). To analyze the dynamics of PP2A localization during mitosis we fused the core subunits to green fluorescent protein (GFP). As expected, this approach was unsuccessful for the catalytic subunit, an abundant protein that cannot be over-expressed18
. Therefore, we examined the localization dynamics of GFP-scaffold stably expressed human RPE1 cells. Near-simultaneous differential interference contrast (DIC) and real-time confocal microscopy of mitotic cells revealed that the scaffold was enriched at centrosomes (, arrows) and to discrete regions on chromosomes (, arrowheads), which corresponded to centromere/kinetochore targeting (). Unexpectedly, we found that scaffold centromere/kinetochore targeting decreased between prometaphase and metaphase in RPE1 () and HeLa cells (Fig. S1a
). GFP-scaffold signal was enriched on centromeres of chromosomes that had not congressed to the metaphase plate (, compare centromere 1 and 2), suggesting that its targeting may be sensitive to chromosome-microtubule attachment status. To test this, we arrested cells at metaphase and tracked distribution of GFP-scaffold before and after depolymerization of microtubules by nocodazole. Strikingly, within minutes of nocodazole addition, GFP-scaffold became enriched on centromeres (), confirming attachment-sensitive localization.
Figure 1 Microtubule-sensitive targeting PP2A to centromeres/kinetochores during cell division. (a) Schematic showing PP2A’s scaffold, catalytic, and regulatory subunits. (b) Maximum intensity confocal projections show distributions of GFP-scaffold expressed (more ...)
To identify regulatory subunits that also reveal a microtubule attachment-dependent localization, we performed a comprehensive localization analysis for all regulatory subunit genes, using stable cell lines expressing GFP fusions of individual subunits. In humans, there are at least fifteen regulatory subunits, distributed over four evolutionarily conserved families17
but the localization dynamics of only two subunits have been reported thus far19
. By live-imaging, only the five members of the B56 (B’) family of regulatory subunits were observed at centromeres (, and Fig. S1b–e
). Similar to the scaffold, B56 targeting was highest in prometaphase, and reduced (B56α, ε) or undetectable (B56β, γ, δ) by metaphase (). Furthermore, when we repeated the nocodazole wash-in assay, GFP-B56α, β, δ, and ε targeting to centromeres increased within minutes (), confirming microtubule-attachment sensitive targeting. We did not consistently detect targeting of B56γ to centromeres in this assay (data not shown). Microtubule-sensitive localization of endogenous B56α to centromeres/kinetochores was confirmed by immunofluorescence in unperturbed cells (Fig. S2
), and in a nocodazole wash-in assay (). Together, these data indicate that B56-PP2A targeting to centromeres/kinetochores is sensitive to kinetochore-microtubule attachment status.
Figure 2 Microtubule-sensitive targeting of B56 regulatory subunits to centromeres/kinetochores. (a) Maximum intensity confocal projections show distributions of GFP in different cell lines stably expressing GFP-B56α-ε proteins. (b) RPE1 cells (more ...)
Microtubule-attachment sensitive kinetochore targeting is a hallmark of proteins that regulate microtubule binding (e.g. dynein, Cenp-E20
) and/or mitotic checkpoint signaling (e.g. Mad221
). However, we considered a requirement for B56-PP2A in the spindle checkpoint unlikely because depletion of the scaffold results in a mitotic arrest in human cells12, 14
In mammals, one challenge in assigning PP2A functions during the cell cycle is the potential for redundancy in regulatory subunit function. Thus, while PP2A is essential for mitosis12, 14
, knockdown of individual regulatory subunits in human cells has not been observed to perturb chromosome segregation22
. Redundancy may be particularly relevant to the B56 family, which share a pseudo HEAT repeat structure with ~ 80% sequence identity23, 24
. Furthermore, depletion of the scaffold or chemical inhibition of the catalytic subunit cannot inform on specific regulatory subunits. Therefore, we chose to deplete B56α-ε proteins using RNAi. We used two non-overlapping pools, composed of one siRNA targeting each B56 subunit, and analyzed the extent of depletion by two methods. First, by western blot analysis, each pool partially reduced protein levels of endogenous B56α, β, and δ (Fig. S3a
). We confirmed that GFP-fusions of B56γ and ε were depleted, as we were unable to detect the endogenous proteins using available antibodies (Fig. S3b
). Second, we confirmed that B56-PP2A siRNA cells had reduced levels of GFP-scaffold at centromeres/kinetochores (). As expected12, 14
, nocodazole-treated B56-PP2A siRNA cells accumulated in mitosis (Fig. S4a
), indicating an intact spindle checkpoint.
Figure 3 B56-PP2A is required for stable kinetochore-microtubule attachments and chromosome alignment. (a) Analysis of GFP-scaffold levels at centromeres/kinetochores after B56-PP2A siRNA. An RPE1 cell line expressing GFP-scaffold was transfected with control (more ...)
To examine whether B56-PP2A siRNA treatment impairs chromosome-microtubule attachment, we made use of the fact that K-fibers are preferentially stable during a brief incubation at 4 °C, while other spindle microtubules depolymerize25
. First, we scored the presence of K-fibers in any mitotic cell that had not yet entered anaphase. B56-PP2A siRNA increased the fraction of mitotic cells that contained few or no K-fibers (), indicating that B56-PP2A is required for proper chromosome-spindle interactions. Second, we used transient nocodazole arrest to accumulate mitotic cells lacking microtubule attachments and then released cells into media containing proteasome inhibitor for 40 minutes, sufficient time for control cells to generate K-fibers (). In contrast, B56-PP2A siRNA cells typically had numerous kinetochores lacking K-fibers () and most contained few or no K-fibers (). To confirm the specificity of our phenotype, we generated stable cell lines over-expressing siRNA-resistant B56α or B56β (Fig. S3c
). When these cell lines were transfected with B56-PP2A siRNA, K-fibers persisted ().
To examine whether B56-PP2A siRNA impairs alignment of chromosomes at the spindle equator, we analyzed chromosome alignment in metaphase-arrested cells. Consistent with defects in chromosome-spindle attachments, individual B56-PP2A siRNA cells had numerous misaligned chromosomes (). Overall, B56-PP2A siRNA resulted in an eight-fold increase in the fraction of cells with misaligned chromosomes, and over-expression of an siRNA-resistant B56 regulatory subunit rescued this defect (). We conclude that B56-PP2A is required to establish stable kinetochore-microtubule attachments and align chromosomes at the spindle equator.
Recent reports indicate that PP2A plays an important role in regulating sister-chromatid cohesion12–14
. B56 regulatory subunits have been linked to this function via their association with Sgo112–14
, a centromeric protein that maintains cohesion and regulates K-fiber stability26–28
. To examine if cohesion is lost in our experiments, we isolated chromosomes from B56-PP2A siRNA cells and found that at least 97% of chromatids had paired kinetochores (), indicating that centromeric cohesion is preserved. Additionally, we measured inter-kinetochore distances in metaphase-arrested cells, which are expected to increase if cohesion at the centromere is compromised. We found, however, that inter-kinetochore distances in B56-PP2A siRNA cells were equal to or less than in control cells, consistent with intact cohesion (Fig. S5
). Finally, we confirmed that centromere targeting of Sgo1 is preserved in B56-PP2A siRNA cells ( and S4b–c
). Together, these data indicate that perturbations in centromeric cohesion and/or Sgo1 targeting cannot account for the defects in kinetochore-microtubule attachment observed in B56-PP2A siRNA cells.
Defects in K-fiber stability in B56-PP2A siRNA cells could be due to an imbalance in substrate phosphorylation and/or failure to recruit proteins that bind kinetochores to microtubules. To test the latter possibility, we examined the kinetochore targeting of three proteins in the KMN network (Dsn1, Knl1, and Hec1), the core microtubule binding complex at the kinetochore29
. It has been shown that Dsn1 and Hec1 levels at kinetochores are not sensitive to microtubule binding, while Knl1 recruitment increases 27% in nocodazole-arrested cells compared to metaphase cells6
. Therefore, to exclude effects of microtubule sensitivity, we included nocodazole in our analyses. Under these conditions, Dsn1 and Hec1 levels were unchanged by B56-PP2A siRNA treatment, while Knl1 levels were modestly increased (1.4-fold), indicating that overall KMN network targeting is preserved in B56-PP2A siRNA cells ().
Figure 4 B56-PP2A depletion increases the phosphorylation of Aurora B substrates and Aurora inhibition suppresses the B56-PP2A siRNA phenotype. (a–c) RPE1 cells were transfected with control or B56-PP2A siRNA (pool 2) and (a, c) incubated with nocodazole (more ...)
To test whether phosphorylation at kinetochores is increased in B56-PP2A siRNA cells, we analyzed substrates of Aurora B, a key regulator of microtubule attachment stability30
. We chose two KMN network substrates (Ser100 on Dsn1 and Ser24 on Knl1), whose phosphorylation decreases microtubule binding affinity6
. In prometaphase cells, phosphorylation of both substrates was increased in B56-PP2A siRNA cells compared to controls (). However, because modification of these substrates is sensitive to microtubule attachment status, as is Knl1 recruitment6
(and because B56-PP2A siRNA destabilizes K-fibers), it was necessary to compare phosphorylation levels on kinetochores with similar inter-kinetochore spacing, a read-out for microtubule attachments. This analysis revealed that on kinetochores under comparable microtubule-dependent pulling forces (defined as a 1.2 to 1.5 µm inter-kinetochore stretch), B56-PP2A siRNA increased the mean phosphorylation of Dsn1 and Knl1 by 1.8- and 2.3-fold, respectively (). After accounting for changes in Knl1 targeting (), the net increase in phospho-Knl1 is 1.6-fold. Together, these analyses suggest that B56-PP2A limits the phosphorylation of these Aurora B substrates. In nocodazole, B56-PP2A siRNA did not increase Dsn1 phosphorylation and the slight increase in Knl1 phosphorylation could be attributed to higher levels of Knl1 at the kinetochore (). It is noteworthy that phosphorylation of Dsn1 and Knl1 at these sites increases 7- and 2-fold respectively in nocodazole-treated compared to prometaphase cells6
. Thus, phosphorylation at these sites may approach 100% in nocodazole, in which case loss of B56-PP2A would not further increase phosphorylation.
To test whether suppressing Aurora B can rescue the B56-PP2A siRNA phenotype, cells were arrested in mitosis, followed by wash-in of chemical inhibitors of Aurora31, 32
(Hesperadin or ZM447439) or control solvent. As expected, control cells had cold-stable K-fibers, whereas many B56-PP2A siRNA cells did not (). Strikingly, Aurora inhibition was sufficient to restore K-fibers in B56-PP2A siRNA treated cells () and both inhibitors restored K-fibers to nearly all kinetochores (). Together, these data suggest that the defects in kinetochore-microtubule interactions in B56-PP2A siRNA cells are due, at least in part, to increased phosphorylation of Aurora B substrates.
To examine whether B56-PP2A attenuates the signaling of kinetochore kinases other than Aurora B, we examined Plk1. Remarkably, wash-in of the Plk1 inhibitor BI253633
also restored K-fibers in B56-PP2A siRNA cells (). Consistent with this, phosphorylation of a Plk1 kinetochore substrate, Ser 676 on BubR134
was increased in prometaphase cells following B56-PP2A siRNA (). Furthermore, even in the absence of attachments, B56-PP2A siRNA treatment resulted in a three-fold increase in phospho-BubR1 staining, without affecting BubR1 protein levels at kinetochores (), indicating that B56-PP2A modulates the phosphorylation level of a Plk1 substrate in addition to Aurora B substrates. BubR1 phosphorylation was undetectable in a cell line stably over-expressing siRNA-resistant B56β (), consistent with this site as a potential B56-PP2A substrate. Taken together, these analyses reveal that reduction of PP2A at the centromere increases the phosphorylation of multiple kinetochore proteins.
Figure 5 B56-PP2A regulates Plk1 substrate phosphorylation and Plk1 targeting to the kinetochore, and Plk1 inhibition suppresses the B56-PP2A siRNA phenotype. (a–b) RPE1 cells transfected with control or either of two B56-PP2A siRNA pools (1, 2) were incubated (more ...)
Considering that Plk1 targeting depends on docking to phospho-epitopes35
, we reasoned that B56-PP2A siRNA treatment might increase Plk1 recruitment to kinetochores. Consistent with this, B56-PP2A siRNA increased Plk1 kinetochore targeting in prometaphase cells (). However, because Plk1 recruitment to kinetochores is sensitive to microtubule attachment status2
, we examined Plk1 targeting in nocodazole-treated cells. Strikingly, B56-PP2A siRNA resulted in a 2.5-fold increase in Plk1 kinetochore targeting (). This result was confirmed with a second pool of B56-PP2A siRNA (data not shown). Critically, over-expression of siRNA-resistant B56β rescued this defect (). We conclude that, in addition to regulating the phosphorylation of kinase substrates, B56-PP2A also controls targeting of Plk1.
At the start of mitosis, all chromosomes are unattached to the spindle. Cytological studies indicate that initial kinetochore-microtubule interactions during prometaphase consist of one or a few microtubules, with microtubule occupancy increasing to ~ 25 microtubules per kinetochore on aligned chromosomes by metaphase36, 37
. Several kinases, including Aurora B and Plk12, 8
, are enriched at unattached kinetochores, suggesting high substrate phosphorylation. For Aurora B, a vast body of data involving phospho-mimetic substrates, kinase mis-targeting, and kinase inhibition suggests that high substrate phosphorylation at the kinetochore would lead to unstable attachments, both by disrupting kinetochore-microtubule contacts and by increasing the turnover of kinetochore-bound microtubules30
. This creates a paradox for how K-fibers form during prometaphase. Additionally, it has been shown that phosphorylation can ‘fine-tune’ the strength of kinetochore-microtubule interactions6
, but contributions from PP2A, or any other phosphatase, during the capture of microtubules by kinetochores remained unclear.
Our data reveal that B56-PP2A is an essential regulator of chromosome-spindle attachments. At the start of mitosis, B56-PP2A is enriched on unattached chromosomes (, top) where it counteracts kinases, reducing phosphorylation to levels that stabilize kinetochore-microtubule binding (, middle). Notably, even a modest reduction in B56-PP2A alters the phosphorylation landscape at kinetochores, preventing the stabilization of K-fibers. As microtubules contact the kinetochore, B56-PP2A is removed, as are a subset of kinases (e.g. Plk12
). Kinetochore accessibility of the remaining kinase, Aurora B, is reduced when chromosomes are bi-oriented and proper inter- and intra-kinetochore tension is established8
(, bottom). Interestingly, the re-distribution of B56-PP2A from the centromere towards the kinetochore on chromatids that have come under tension (, middle) may ensure timely dephosphorylation of Aurora B and Plk1 substrates on kinetochores that have bi-oriented. While this function has been attributed to PP115, 16
, recent work in yeast reveals the essential function of kinetochore PP1 is to silence the spindle checkpoint 38
Misregulation of PP2A is considered a pre-requisite for malignancy in human cells, but less is understood about which phospho-signaling networks are associated with tumorigenesis9
. Interestingly, point mutations in the scaffold that disrupt binding to the B56 regulatory subunits have been identified in lung and breast carcinomas39
. Our data suggest these mutations will increase the frequency of whole chromosome gain or loss, the most common form of chromosomal instability in human tumors40
, through disruption of kinetochore-microtubule interactions.