Hras helps hippo heterodimerize to evade tumor suppression

ABSTRACT

Ras oncoproteins can promote or suppress cellular apoptosis, but the mechanisms underlying these varied responses remain incompletely understood. Ras is linked to the Hippo tumor suppressor pathway, a highly conserved signaling cassette that regulates organ size in animals ranging from flies to humans. The proximal members of this pathway, Mammalian Ste20-like kinases (Msts) −1 and −2, self-associate in homodimers and also form heterodimers with other proteins. Formation of such complexes is known to regulate Mst kinase activity and thus, the Hippo pathway. In a manuscript that recently appeared in Current Biology, we showed that activated Hras promotes the formation of Mst1/Mst2 heterodimers, that activation of Erk was required for this event, and that these heterodimers were much less active than Mst1/Mst1 or Mst2/Mst2 homodimers. Interestingly, the formation of such heterodimers was required to deactivate the Hippo pathway and to enable transformation by Hras. In this Commentary, we discuss the background for this study and surprising implications thereof.

When stressed, cells can be thought of as being delicately poised on a molecular weighing device, teetering between survival and death. The Ras GTPases – Kras, Nras, and Hras - encoded by the most commonly mutated oncogenes in human cancer, are known to put their thumbs firmly on the apoptotic scale, but, depending on isoform, cell type, and circumstances, not always on the same side.^1-3 For example, in primary human cells, expression of activated Kras causes cell cycle arrest and apoptosis, whereas the same oncogene induces transformation in immortalized cells in which key tumor suppressors have been disabled or sequestered. The three Ras isomers differ in their propensity to elicit apoptosis, with Kras in general having the strongest effect, and Hras the weakest. Why this is so is not fully understood, but one idea is that Hras engages pro-survival enzymes such as PI3K more efficiently than does Kras.⁴

Recently, it has been shown that the 3 Ras isoforms also differ in their effects on the pro-apoptotic mammalian sterile-20 like (Mst) kinases, Mst1 and Mst2. These enzymes are highly conserved serine/threonine protein kinases that are best known for their role in activating the Hippo signaling pathway. In Drosophila, the main role of this pathway is to control organ size via effects on apoptosis and cell proliferation.^5,6 In mammals, Mst function appears to be more complex. While loss of Mst1 or Mst2 alone in mice is not associated with tumor formation, simultaneous loss of both these genes in the liver has oncogenic effects, consistent with the idea that these genes encode potential tumor suppressors with at least partly redundant functions.^7-9 Matallanas et al. showed that, in Hke3 colonic epithelial cells, Hras and Nras have strong inhibitory effects on Mst2 activity, whereas Kras stimulates Mst2.¹⁰ Kras has also been shown to activate Mst1 at mitochondria in cardiomyocytes, leading to the phosphorylation of Bcl-xL and subsequent apoptotic cell death.¹¹ These results raised interesting mechanistic questions regarding the relationship between Ras and Mst that we attempted to address in a recent paper in Current Biology.¹²

The activity of Mst1 and Mst2 is regulated at least in part by dimerization, either with themselves as homodimers or with other proteins as heterodimers (Fig. 1). Dimerization formation is mediated by a C-terminal, coiled-coil section of Mst1 and Mst2 segment termed the SARAH (Salvador (in mammals, WW45), Rassf, Hippo (in mammals, Mst)) domain.¹³ In homodimers, Mst1 or Mst2 transphosphorylates a key threonine residue in the activation loop of its partner's kinase domain (T183 in Mst1, T180 in Mst2). However, other, non-kinase SARAH domain proteins can also complex with Mst1 or Mst2 with high affinity.^14,15 When Mst1 or Mst2 are complexed with non-kinase SARAH partners, such as WW45 or Rassf proteins, transphosphorylation is not possible.¹⁶ This inability to transphosphorylate may underlie the low activity state observed in some Mst heterodimers.^16,17 In other cases, such as that of the c-Raf/Mst2 complex, low activity appears to result via recruitment of a protein phosphatase to the complex.^18,19 It should be noted, however, that not all heterodimers are inactive, as binding of non-kinase partners such as Rassf5 does not block Mst2 activation if it occurs after phosphorylation of the activation loop.¹⁶

Figure 1.

Mst homo- and heterodimers. In their homodimeric form, Mst1 and Mst2 are highly active and promote cell cycle delay, apoptosis, and tumor suppression. Mst1 and Mst2 can also heterodimerize with additional proteins, some of which containing SARAH domains (WW45, Rassf; SARAH domain indicated by extension from rectangle), and some which do not (c-Raf, Abl, Pdx1, and mTORC2 (via Rictor). In many cases, these heterodimers are inactive, and the Hippo pathway is suppressed. Mst heterodimers are inactivated due to a variety of mechanisms, including inhibitory phosphorylation of Mst kinase domain (Akt), phosphorylation-induced dissociation of Mst homodimers (mTORC2), importation of complexed phosphatases (c-Raf), absence of a kinase partner for transphosphorylation (Rassf, if bound to Mst prior to activation), and unknown mechanisms (Mst1/Mst2 heterodimers).

Given the complexities of Mst regulation, we sought to identify additional interactors for these kinases. Using specialized “Flip-In” HEK293 cells that had been engineered to express epitope-tagged Mst1 at endogenous levels, we unexpectedly found abundant Mst2 peptides in Mst1 immunoprecipitates.^12,20 This finding was not due to cross-reactive antibodies, as epitope-tagged Mst1 had been immunoprecipitated with anti-Flag-tag antibodies, and the Mst1 and Mst2 antibodies used for immunoblots were shown, using recombinant proteins, to recognize only their intended targets. Further experiments using endogenous proteins and a different technique (proximity ligation assays (PLA)) confirmed that Mst1 and Mst2 interact in cells. Interestingly, we found that the affinity of isolated SARAH domains was higher for Mst1/Mst2 heterodimers than for Mst1/Mst1 or Mst2/Mst2 homodimers, suggesting that Mst1/Mst2 heterodimers may be favored, at least in some circumstances. This supposition was supported by measuring the levels of Mst1/Mst2 heterodimers in cells, which showed that a substantial fraction of total Mst1 and Mst2 is found in such heterodimers. It should be noted that Hwang et al. reported that the Mst1/Rassf5 SARAH domain heterodimer also showed higher affinity than Mst1/Mst1 SARAH domain homodimers,¹⁴ suggesting that homodimers may not, in general, represent the favored state for Mst1 or Mst2 in cells.

A second, bigger surprise was that the Mst1/Mst2 heterodimers were much less active than either Mst1/Mst1 or Mst2/Mst2 homodimers (Fig. 1). Unlike, say, WW45/Mst1 or Rassf/Mst1 heterodimers, we expected that Mst1/Mst2 heterodimers should be capable of transphosphorylation, and thus, full activation. Given that such was not the case, we assumed that the geometry of the heterodimer is somehow not conducive to activation. Such “bad” geometries might result in kinase domains that are not properly aligned for transphosphorylating the partner's activation loop threonine residue, or might shield the PPxF motifs in the kinase domains of Mst1 and Mst2 that are needed for docking to WW45, and hence for efficient coupling to LATs.²¹ If so, that would add an interesting new layer to Hippo pathway regulation; namely, the modulation of adaptor protein association by steric hindrance. To determine the structural basis for these observations, we attempted to crystallize a heterodimer comprising Mst1 and Mst2 SARAH domains, but were unsuccessful in obtaining a suitable crystal; thus the mechanism underlying the low activity of Mst1/Mst2 heterodimers remains a puzzle for now.

What is clear, however, is that the low activity of such heterodimers could have important biological consequences, provided that a switching mechanism exists to regulate their formation in cells. Given the known role of Ras proteins in regulating apoptotic pathways, we tested Hras as well as the Ras effectors c-Raf, Akt, and Nore1 (a.k.a. Rassf5) for their ability to regulate Mst1/Mst2 heterodimerization. Using PLA to monitor the level of Mst1/Mst2 heterodimers, we found that activated Hras, and to a much lesser extent, c-Raf, had a profound positive effect on Mst heterodimerization. Interestingly, blockade of Erk signaling by a Mek inhibitor abrogated this effect. These results demonstrate that Hras, acting through the Raf/Mek/Erk cascade, somehow promotes Mst1/Mst2 heterodimerization.

As Mst1/Mst2 heterodimers are much less active than Mst1/Mst1 or Mst2/Mst2 homodimers, we then tested a series of predictions that at first blush seemed counterintuitive. For example, if we assume that the switch from active Mst1/Mst1 and Mst2/Mst2 homodimers to inactive Mst1/Mst2 heterodimers is required for Hras-mediated transformation, then immortalized cells lacking either Mst1 or Mst2 (e.g., Mst1^+/+;Mst2^−/− or, alternatively, Mst1^−/−;Mst2^+/+ mouse embryo fibroblasts) should resist transformation by Hras, since heterodimerization is not possible when one partner is absent. That means that such cells, though lacking a putative tumor suppressor, should be harder to transform. And that, indeed, is just what we found. Cells lacking Mst2 were highly resistant to Hras-induced focus formation and soft agar colony growth. Moreover, such cells showed elevated Hippo pathway signaling, characterized by high levels of phospho-Mst, phospho-LATS, and phospho-Yap. In contrast, lost of both Mst isoforms together had the expected effect: such cells were easy to transform and showed low Hippo pathway activity.

Among the unanswered questions are: 1) how does ERK lead to Mst1/2 heterodimerization and 2) Is this phenomenon unique to Hras, and if so, why? While we do not, at present, have answers to either question, there are hints that could lead to answers soon. For example, Sciarretta et al. recently reported that, in cardiomyocytes, active mTORC2 can phosphorylate Mst1 at a serine residue in its SARAH domain, leading to disassociation of the Mst1/Mst1 homodimer and inactivation of the Hippo pathway.²² These authors did not report if Mst1/Mst2 heterodimers were promoted under these conditions, but this is a reasonable possibility, especially given that the S438 phosphorylation site in the Mst1 SARAH domain is not present in the equivalent position of Mst2 (Fig. 2A, arrow). Neither Mst1 or Mst2 have consensus ERK phosphorylation sites in their SARAH domains, so it seems unlikely that the effect of Ras, via ERK, on impeding Mst1/Mst1 homodimerization is mediated by direct, ERK-catalyzed alterations within the SARAH domain, but it is possible that ERK phosphorylates other residues in Mst1 or Mst2, or acts indirectly through mTORC2 or another kinase. It is also possible that innate structural differences in the SARAH domains of Mst1 and Mst2 cause or contribute to the lack of activity of heterodimers. While similar in overall sequence and structure, the Mst1 SARAH domain differs from the Mst2 SARAH domain at 16 of 48 positions (Fig. 2); these differences may result in different relationships between the 2 kinase domains. Regarding Ras isoforms, since Nras, like Hras, is relatively weak in its ability to induce apoptosis, we predict that Nras, like Hras, will promote Mst1/Mst2 inactivation via heterodimezation, but that Kras, which has strong proapoptotic activity, will lack this effect. However, even if this proves to be the case, it will not explain why, as all 3 Ras isoforms are potent activators of ERK. The explaination may well lie in differences in location and/or engagement of scaffold proteins by the 3 Ras isoforms. For example, the SARAH-domain-containing Rassf1a adaptor protein is found in preferential association with Kras.²³ In this regard, it would be interesting to test the effects of Ras-independent ERK activation of Mst1/2 heterodimer formation. This could be achieved by expressing mutationally activated Raf or Mek in cells, thereby bypassing the need for Ras activation.

Figure 2.

Mst SARAH domains. (A) Alignment of SARAH domains of Mst1 and Mst2. Asterisks indicate identical residues; colons indicate strongly similar residues; periods indicate weakly similar residues. (B) Yellow ribbon locations mark the 16 of 48 positions that differ between Mst1 and Mst2.

Among the surprising predictions: Mice (or men?) lacking one but not both of the Mst genes should not be cancer-prone, but might in fact be cancer-resistant in the face of certain oncogenic stimuli. For example, we predict that, compared to wild-type mice, single Mst knock-out mice (e.g., Mst1^−/−mice) will resist tumorigenesis when bred to an mutant Hras expressing mouse. Thus we propose that, in the case of the Msts, losing a putative tumor suppressor might, in some respects, be a good thing. Of course, cancer proneness aside, loss of Mst1 is not without issues: Mst1^−/− mice, and humans bearing disabling mutations in MST1, suffer immune dysfunction due to defects in T-cell migration.^24-26 Be that as it may, Msts appear to be most unusual tumor suppressors, and heterodimerization an unusual mechanism for inactivation.

Funding Statement

This work was supported by grants from the Department of Defense to J.C. (DOD 14-1-0141) and the NIH to J.C. (R01 CA148805) and Heinrich Roder (R01 GM056250), and the Fox Chase Cancer Center (P30 CA006927), as well as by an appropriation from the Commonwealth of Pennsylvania.

Acknowledgments

We thank Mark Andrake and Roland Dunbrack (Molecular Modeling Facility, Fox Chase Cancer Center) for their structural analysis of Mst1/Mst2 heterodimers.