Centrosomes: Please keep your social distance!

Accurate control of centrosome number is essential for proper chromosome segregation, and it is well established that centrosome abnormalities can trigger a p53‐dependent cell cycle arrest. Two new studies published in The EMBO Journal demonstrate how PIDD1 is recruited to centrosomes and that the localization of PIDD1 to distal appendages of centrosomes is required for PIDDosome activation at clustered supernumerary centrosomes.

Two recent studies identify centriolar distal appendages to be key for PIDD1 recruitment and subsequent p53‐dependent cell cycle arrest in the precent of supernumerary centrosomes.

Centrosomes act as the microtubule‐organizing centers of the cell and are involved in diverse cellular processes such as polarity, motility, regulation of cell shape and cell division. Accurate control over centrosome number is vital, as both increased and decreased centrosome numbers can cause severe errors in cell division, leading to aneuploidy and possibly tumor development. In addition, supernumerary centrosomes may contribute to tumor cell invasiveness and could deregulate centrosome‐based signaling pathways (Godinho & Pellman, 2014; Nigg & Holland, 2018; Ben‐David & Amon, 2020).

Cells have several mechanisms in place to avoid the adverse effects caused by deviating centrosome numbers. First, centrosome duplication is tightly regulated to ensure that each centrosome only duplicates once during a single cell cycle (Nigg & Holland, 2018). The centrosome consists of two centrioles: a mother centriole and a daughter centriole. Maturation of centrioles is regulated such that a cell only contains two mature mother centrioles during mitosis. The mature mother centrioles are decorated with distal and subdistal appendages. Second, the mitotic spindle is organized in a way that ensures that each daughter cell inherits only one centrosome. Thus, a newborn cell inherits one centrosome consisting of one mature mother centriole and one daughter centriole (Nigg & Holland, 2018).

However, centrosome abnormalities can arise when expression of specific proteins involved in centrosome duplication is deregulated, when centrosomes fail to separate, or when cytokinesis fails, leading to loss or gain of centrosomes (Nigg & Holland, 2018). In fact, such abnormalities are quite common in cancer (Godinho & Pellman, 2014). To study the cellular effects of centrosome abnormalities, several ways to generate cells carrying extra centrosomes have been developed. For instance, knock‐in alleles resulting in the stabilization and hyperactivation of different centriole duplication factors (such as PLK4 and SAS6) cause centriole overduplication and supernumerary centrosomes, and were shown to trigger a p53‐dependent proliferation arrest that is independent of DNA damage (Holland et al, 2012). Similarly, tetraploid cells generated by the inhibition of cell division during mitosis, e.g., using cytokinesis inhibitors, produce cells with supernumerary centrosomes that arrest in a p53‐dependent manner (Fig 1). These cells were long thought to arrest due to the tetraploid DNA content, but more recent data demonstrated that such cells activate the PIDDosome protein complex (Fava et al, 2017), which contains caspase‐2 that can cleave and inactivate Mdm2, resulting in stabilization of p53 (Sladky et al, 2017). Depletion of PIDDosome components PIDD1, RIADD, or caspase‐2 prevented Mdm2 cleavage, p53 activation, and the proliferation arrest that normally occurs following cytokinesis failure (Fava et al, 2017). Interestingly, tetraploid cells harboring only one centrosome failed to activate the PIDDosome. Also, depletion of distal appendage proteins prevented PIDDosome activation. Taken together, these data indicated that the presence of supernumerary centrosomes and, more specifically, of excess mature centrioles, is the cause of PIDDosome and p53 activation and arrested proliferation in response to supernumerary centrosomes. However, the molecular mechanism leading to PIDDosome activation by supernumerary centrosomes remained elusive. Now, the groups of Fava (Burigotto et al, 2021) and Holland (Evans et al, 2021) have begun to elucidate the mechanistical basis for the PIDDosome‐dependent activation of p53 in response to supernumerary centrosomes.

Figure 1. Supernumerary centrosomes may arise from erroneous centrosome duplication, failed centrosome separation during prophase, or cytokinesis failure.

In interphase, clustering of supernumerary centrosomes causes PIDDosome activation and cell cycle arrest. Conversely, unclustered supernumerary centrosomes fail to activate the PIDDosome, thus allowing continued proliferation (see text for details).

The two new studies published in this issue of The EMBO Journal address the question how supernumerary centrosomes inhibit proliferation from different angles. On one hand, Burigotto et al (2021) performed yeast‐two‐hybrid screening using PIDD1 as a bait, based on their previous finding that PIDD1 is required for p53 activation by supernumerary centrosomes, and the observation that PIDD1 itself localizes to the mother centriole (Fava et al, 2017). This resulted in the identification of the distal appendage protein ANKRD26 as a PIDD1 interactor. Indeed, the authors confirm that ANKRD26 itself, as well as other known distal appendage proteins, are essential to arrest cells in response to supernumerary centrosomes. On the other hand, Evans et al (2021) devised a clever CRISPR‐Cas9‐based knockout screen to identify genes required for a proliferation arrest in response to supernumerary centrosomes. In addition to known factors such as p53 and PIDD1, this screen identified several putative genes required for growth inhibition in response to supernumerary centrosomes, including ANKRD26 and several other distal appendage proteins.

Following their identification of ANKRD26 as a PIDD1‐interacting protein, both groups continued to investigate the molecular basis of ANKRD26 localization to the centrosome, as well as the ANKRD26‐dependent recruitment of PIDD1 to the centrosome. Collectively, these studies were able to map the domains required for the centrosomal localization of ANKRD26, as well as the domains required for the interaction between PIDD1 and ANKRD26. Importantly, the mapping of these interaction domains allowed elegant perturbation experiments, specifically abolishing PIDD1 centrosome recruitment. This demonstrated that ANKRD26‐dependent recruitment of PIDD1 to the centrosome is a requirement for PIDDosome activation upon supernumerary centrosomes.

While these data clearly demonstrate that PIDD1 needs to localize to the centrosome, they do not resolve how the PIDDosome can sense the presence of supernumerary centrosomes. The simplest model would be that PIDD1 centrosome localization and/or autoproteolysis (required for PIDDosome activity (Tinel et al, 2007)) is limited to cells with supernumerary centrosomes. However, Burigotto et al (2021) show that PIDD1 also localizes to centrosomes and autoproteolysis occurs in cells with normal centrosome numbers. Furthermore, PIDD1 autoproteolysis occurs even in the absence of its localization to centrosomes. Therefore, other cues must be required to allow PIDDosome activation upon supernumerary centrosomes. One hint toward such a cue came from the observation that supernumerary centrosomes tend to cluster during interphase (Godinho & Pellman, 2014). Burigotto et al investigated the possibility that interphase clustering of centrosomes could be important for PIDDosome activation by supernumerary centrosomes. They showed that supernumerary centrosomes normally cluster within a few hours after a failed cytokinesis, but that the destabilization of microtubules during telophase in cells that undergo cytokinesis failure prevents clustering of centrosomes in G1 phase. Importantly, this also prevented cleavage of Mdm2, indicating that PIDDosome activation indeed requires the efficient clustering of supernumerary centrosomes during G1 (Fig 1). Thus, it could simply be that the increased local concentration of PIDD1 at clustered centrosomes leads to PIDDosome activation and a concomitant increase in p53.

The PIDDosome can also be activated in response to DNA damage (Sladky et al, 2017), and both groups also investigated if centrosome localization of PIDD1 is required for DNA damage‐induced PIDDosome activation. The conclusions appear to diverge at this point, however, with Burigotto et al (2021) showing that PIDDosome activation in response to replication stress depends on distal appendage proteins and PIDD1. In addition, they show that PIDDosome activation by drug‐induced stabilization of p53 depends on distal appendage proteins and PIDD1. Interestingly, p53 stabilization leads to increased PIDD1 expression. In contrast, Evans et al (2021) present data that imply that PIDDosome/caspase‐2 activation in response to DNA double‐strand breaks (DSB) occurs independently of ANKRD26, PIDD1, and p53. In line with this, others have previously shown that caspase‐2 activation occurs within a few hours following DSB induction in thymocytes derived from PIDD1‐knockout mice (Manzl et al, 2009). These data indicate that PIDD1‐independent caspase‐2 activation may take part as the immediate response to DSBs. However, there are indications that DNA lesions can result in PIDDosome activation when cells fail to accurately arrest in response to DSBs (Tsabar et al, 2020). This will lead to an illegitimate cell division and subsequent PIDDosome activation, resulting in hyperactivation of p53. This delayed PIDDosome activation likely acts to prevent the continued proliferation of damaged cells that have escaped the checkpoint.

Important to note is that replication stress and DSBs result in activation of different signaling cascades (Ciccia & Elledge, 2010), which may differentially affect proliferation. DSBs effectively block cell cycle progression to prevent cell division in the presence of broken DNA. In contrast, replication stress‐induced under‐replicated DNA does not necessarily block cell division. Thus, the data presented here are compatible with a model where DSB‐signaling initially induces PIDDosome‐independent activation of caspase‐2. However, in cases where the DNA damage load or the extent of replication stress is insufficient to cause direct caspase‐2 activation and to prevent cell division, cells will progress through mitosis, causing PIDDosome‐dependent activation of caspase‐2 and hyperactivation of p53 in the respective daughter cells.

Collectively, the data presented by Burigotto et al and Evans et al fit a model in which the clustering of supernumerary centrosomes is required to locally increase PIDD1 levels beyond a certain threshold level that is required for PIDDosome activation. In light of this, it is interesting to note that PIDDosome activation in response to p53 activation requires PIDD1 centrosome localization, but can occur in the absence of supernumerary centrosomes. How can this be reconciled with a model in which the increase in local concentration of PIDD1 at clustered centrosomes is key to PIDDosome activation? Possibly, the answer to this question lies in the fact that p53 itself can also activate expression of PIDD1. In this setting, the increase in local PIDD1 concentration at centrosomes would be achieved by enhancing PIDD1 levels, rather than through clustering of multiple centrosomes. Alternatively, as a consequence of the globally increased PIDD1 levels, a single centrosome may already be sufficient to locally elevate PIDD1 levels beyond the threshold level required for PIDDosome activation. This implies that p53‐dependent induction of PIDD1 will act in a positive feedback loop to further enhance local PIDDosome activation in cells with supernumerary centrosomes.

To directly test this model, it will be interesting to see whether artificial tethering of PIDD1 to other locations than the centrosome could induce PIDDosome activation in acentrosomal cells or cells depleted of distal appendage proteins. Also, one could test whether overexpression of PIDD1 alone can result in PIDDosome activation and whether this no longer occurs in cells lacking centrosomes. Additionally, drugs that reverse the clustering of supernumerary centrosomes during interphase (Pannu et al, 2014) can be used to test this model and address whether continued clustering is required for sustained PIDDosome activation, or whether transient clustering is sufficient for continuous PIDDosome activation and p53 activation. Collectively, these experiments would elucidate whether it is sufficient for PIDDosome activation if PIDD1 levels, either locally or globally, surpass a certain threshold level. In addition, they would address whether additional centrosome‐based ques are required for PIDDosome activation and if PIDDosome activation is irreversible.

The EMBO Journal (2021) 40: e107525.