Gelling Bodies: Understanding the Mechanisms Underlying Arsenic Trioxide Action

Compartmentalization of the nucleus into subnuclear domains has been implicated in the regulation of gene expression, nuclear structure, and response to stress (1). One of the most studied interchromatin domains is the promyelocytic leukemia nuclear body (PML-NB). From the start, the PML-NB attracted the attention of the scientific community because of the identification of a chromosomal translocation fusing the PML gene to the retinoic acid receptor α (RARα) gene in the majority of patients with acute promyelocytic leukemia (APL; refs. 2–4). The resulting fusion gene encodes the PML–RARα oncogene, which is the main driver of APL and disrupts the PML-NBs into multiple microspeckles. Combination of all-trans retinoic acid (ATRA) and the anticancer drug arsenic trioxide (ATO) causes PML–RARα degradation, reformation of PML-NBs, and reactivation of a senescence program in leukemic blasts, leading to eradication of the disease (2). ATO and ATRA–mediated targeting of PML–RARα represents a paradigm of targeted therapy along with tyrosine kinase inhibitors in chronic myelogenous leukemia.

Significant efforts have been made in trying to decipher the mechanisms governing formation and degradation of PML-NBs in the context of normal hematopoietic cells and leukemic blasts. Hugues de Thé and Valérie Lallemand-Breitenbach conducted seminal work in deciphering how oxidation and SUMOylation coordinate formation and demise of this nuclear subdomain (5). However, the response of PML-NBs to ATO in both physiologic and pathologic settings remains incompletely understood. Their latest study, published in this issue of Cancer Discovery (6), substantially advances our understanding of fundamental mechanisms regulating the genesis and dynamics of this subnuclear structure upon therapy. Furthermore, it has more general implications on our grasp of how membraneless organelles (MLO) form in the cell.

Among the main questions at the core of this work is whether biogenesis of PML-NBs occurs via canonical liquid–liquid phase separation (LLPS), known to govern formation of several MLOs. Indeed, Bercier and colleagues (6) showed that PML separates from diffuse nucleoplasmic fractions into PML-NBs, and its diffusion coefficient and viscosity are similar to the ones of other MLOs, like nucleoli or nuclear speckles. Furthermore, PML dynamically exchanges between PML-NBs and the nucleoplasm, suggesting that PML-NBs indeed display the hallmarks of LLPS. However, the residency time of PML at PML-NBs is longer than for other MLOs nucleators, such as Cajal bodies. Another difference is the fact that nearly one third of PML-NBs fail to recover after photobleaching, suggesting that there are two kinds of PML-NBs, of which only one efficiently exchanges with the nucleoplasm. This challenges the idea that PML-NB biogenesis and dynamics are governed by a canonical LLPS mechanism. ATO appears to immobilize PML as well as PML-RARα within PML-NBs, suggesting a transition from liquid- to gel-like structures. This prompted the authors to investigate the mechanisms underlying these properties.

ATO-driven gel-like features of PML-NBs appear to rely on the B-Box domain of PML and not on disulfide bridge formation. This is elegantly demonstrated using N-ethylmaleimide, which alkylates cysteine, as well as via mutations in key cysteines in the B-Box or RING finger domains. Interestingly, mutation of a single cysteine in B-Box 2 (B2 C213A) is sufficient to make PML mostly diffuse and insensitive to ATO, while formation of disulfide bonds mediated by C213 or C389 in the RING domain is not required for the response to ATO. Overall, these data suggest that C213 may contribute to arsenic binding.

Intriguingly, a short sequence across C212/C213 residues is conserved in PML proteins across evolution, and this is where mutations found in patients with ATO-resistant APL fall. To gain more insights into the property of this sequence, the authors successfully generated a crystal structure of the B2 domain, a feat previously not achieved by others. This structure highlights the presence of an α-helix that is situated adjacent to the C213 residue critical for PML-NB formation and response to ATO. The authors then generated a number of mutants affecting zinc coordination within B2 or residues within the α-helix, such as C213. All these mutants make PML more soluble, indicating that they are essential for PML-NB formation. Interestingly, mutations in the α-helix do not affect B2 folding and make PML reside fully in a liquid compartment.

The next question the authors decided to address was whether the PML B2 domain formed trimers like other TRIM proteins and if such trimers were important for PML-NB formation. Mutagenesis of specific residues within the α-helix including the A216V patient-associated mutation predicted to affect trimerization dramatically affected PML-NB formation. The C213 residue is organized as a triad at the center of the PML B2 helix trimer. Then, the authors hypothesized that such conformation represents an ATO docking site. Indeed, ATO is unable to cause PML-NB transition to a more immobile, gel-like state when key residues affecting trimerization were mutated, including C213. Another elegant experiment using ReAsH, an ATO derivative that fluoresces when bound to proteins, showed that trimerization and the C213 residue are critical for ATO binding to PML and that trimerization is required for subsequent irreversible binding of ATO to C213.

An important question arising from the study is whether C213 and the α-helix are also important for ATO-induced SUMOylation, another key consequence of ATO treatment. Indeed, all mutations affecting the α-helix and C213 impair PML basal as well as ATO-induced SUMOylation. Furthermore, knocking in a mutation impairing α-helix formation in embryonic stem cells results in abolished partner SUMOylation as well. Finally, mutation of C213 to valine does not affect basal SUMOylation but still impairs ATO-induced SUMOylation. It would be informative to determine basal SUMOylation targets in wild-type (WT) cells and cells carrying C213V or C213S, as this could reveal important differences in partner SUMOylation related to structural changes of the triad domain.

Would what is described above for WT PML apply also to PML–RARα? The authors conducted in-depth investigation of the impact of mutations affecting α-helix trimerization and C213 in the context of PML–RARα and found similar effects as for what concerned PML. Furthermore, these mutations also impair PML–RARα degradation upon ATO. The C213S mutation markedly affects the distribution of PML–RARα, which becomes almost completely nuclear diffused. It would be very important to assess whether these changes in subnuclear localization correlate with alterations of PML–RARα genomic distribution using chromatin immunoprecipitation or similar approaches. One could hypothesize that impaired trimerization would modify PML–RARα's ability to bind and repress its canonical target genes. Furthermore, gain effects could not be excluded. Another possibility is that chromatin structure and/or marks could be affected at the sites where WT or mutant PML–RARα binds. Cancer-associated mutations impairing response to ATO could also behave differently with respect to genome distribution and effects on chromatin.

The last part of the study deals with effects of impaired α-helix trimerization on gene expression. The authors decided to use an in vivo model, in which the mutation corresponding to the APL-associated A216V, A220V, is knocked in in mice. They focused on hepatocytes given their particular sensitivity to oxidative stress and damage. One interesting observation is that A220V leads to increased γH2AX foci in the liver, suggesting induction of DNA damage, due to increased oxidative damage and/or impaired DNA repair. Although several studies have linked PML-NB to genome maintenance mechanisms (7, 8), the relevance of these interactions in in vivo settings was limited. Therefore, this represents an opportunity to investigate the mechanisms underlying the induction of DNA damage due to impaired PML-NB formation in vivo. A220V induces marked gene expression changes at steady state, with several gene ontology categories coming up, including metabolism-related, MYC, DNA repair, and the unfolded protein response. As with many studies on the effects of PML loss on transcription, it remains to be clarified what the underlying mechanisms would be. For instance, some of the observed changes would be compatible with impaired p53 function, but not all. The authors decided to explore the effect of the A220V mutation on the response to oxidative stress. Also in these settings, A220V has marked effects on transcription, in great part overlapping with changes caused by Pml loss. Interestingly, most of the differentially affected genes were downregulated. Gene ontology of downregulated genes reveals pathways like p53, apoptosis, and TNFα signaling. Among the upregulated categories, MYC and IFN response are interesting. In particular, the IFN response would be worth investigating further because of its potential association with nucleic acid sensing, which could be caused by intermediates of DNA damage and its repair.

Overall, these findings addressed a long-lasting question in the field of ATO-mediated anticancer therapy, in particular how PML-NBs are targeted by ATO. This could have implications on how ATO targets other proteins in the cell. Furthermore, this work further highlights the importance of these nuclear subdomains in the response to stress, such as oxidative stress, rather than at steady state, in agreement with the mild phenotype of germline Pml inactivation. An increased understanding of how PML-NBs are targeted during the response to stress will pave the way to new efforts aimed at reducing the negative consequences of supraphysiologic oxidative stress or inflammation.