Nucleophosmin 1 (NPM1, or B23, numatrin, NO38) is an extremely abundant human protein (top 0.25%, according to the PaxDB database1 with up to 30 known sites of phosphorylation that are thought to regulate its activity and interaction with other proteins. NPM1 forms stable pentamers and supramolecular assemblies inducing liquid-liquid phase separation (LLPS), and is the main component of the membraneless organelle nucleolus[2], [3]. NPM1 regulates ribosome biogenesis, histone assembly, centrosome duplication, DNA integrity and cell stress responses[2], [4], [5], [6], [7]. The regulation of NPM1 functions is accompanied with its phosphorylation, oligomer disassembly and nucleocytoplasmic shuttling[7], [8], [9], [10], [11], [12], [13], [14]. Many mutations in NPM1 cause its permanent egress to the cytoplasm and induce various cancer types including acute myeloid leukemia (AML)15. However, the exact mechanisms underlying NPM1 regulation, including multi-site phosphorylation and recognition by the major phosphoserine/phosphothreonine (pS/pT)-binding protein known as 14-3-3, are not well understood.

14-3-3 proteins were the first phospho-Ser/Thr binding protein modules discovered16 and belong to the top 1% most abundant human proteins1. Composed primarily of α-helices, the ∼30 kDa 14-3-3 subunits assemble into homo- or heterodimers containing one groove per subunit that bind specific phospho-motifs located internally or at the C-terminus of their partner proteins17 (Fig. 1A). As dimers, 14-3-3 proteins can bind phosphorylated clients at a single phospho-site or multi-valently through multiple phospho-sites. The complexity of 14-3-3/client interactions is evident when considering that clients commonly have more than two 14-3-3-binding motifs and can thus bind 14-3-3 in a variety of multi-valent modes depending on which combination of sites are phosphorylated, i.e. their “phosphorylation code”. In addition, humans express seven 14-3-3 subtypes encoded by separate genes, which are historically termed ‘isoforms' and designated by the Greek letters (β (beta), γ (gamma), ε (epsilon), ζ (zeta), η (eta), σ (sigma), τ (tau)); these isoforms are highly conserved but differ by the affinity of interaction they have with their phospho-targets18. While the 14-3-3 interactome comprises hundreds of different proteins, only a minority have been well-studied when in complex with 14-3-3.

Recent work has implicated 14-3-3 in regulating NPM1 function. For example, centrosome amplification depended on cytosolic NPM1 cooperating with 14-3-3[14], [19] in an NPM1 phosphorylation-dependent manner20. Specifically, 14-3-3ε was found among the NPM1 interactors in the nucleolus5, though another study could not reproduce the direct 14-3-3/phospho-NPM1 interaction21. Reasons for this discrepancy are not clear but it could be because NPM1/14-3-3 interactions depend on the specific pattern of NPM1 phosphorylation across multiple sites. Intriguingly, recent crystal structures have provided snapshots of 14-3-3 interaction with two NPM1 phospho-fragments, around pSer4822 and pSer29323. NPM1 phosphorylation at Ser48 by prosurvival kinase Akt led to NPM1 pentamer disassembly and exit from the nucleolus7, and so not surprisingly NPM1 phosphorylated at Ser48 is prevalent in human tumors7. Yet, the details of the 14-3-3/NPM1 interaction were not analyzed further and the hierarchy of these two (and potentially other) NPM1 phospho-sites in controlling complexation, remain unknown.

In this work, we addressed these questions by dissecting the interaction of 14-3-3 proteins with unphosphorylated and phosphorylated recombinant NPM1 variants. Current approaches to decipher the roles of single and multi-site phosphorylations in context of the whole protein are limited as phosphosites are not always solvent accessible for kinases, and the complexity of kinase specificity and activation mechanisms can lead to incomplete and/or off-target phosphorylation. Further, their inadvertent exposure to phosphatases leads to heterogeneity in sample preparation. Although still widely used because of the simplicity of their incorporation and stability, phospho-mimicking mutations (S/T → D/E) are rather poor mimics of pS/pT. Here, we use a combination of in-cell kinase-mediated phosphorylation strategies[24], [25], [26], [27] as well as site-directed translational installation of a non-hydrolyzable phosphonate analog of phosphoserine[28], [29] to generate an array of NPM1 phospho-variants and peptides to study their interactions with 14-3-3 proteins. We show that native NPM1 forms a stable pentamer generally resistant to phosphorylation, but once phosphorylated it disassembles into monomers that expose the phospho-sites for complexation with 14-3-3. Measuring the affinity of four putative 14-3-3 phospho-sites to all seven 14-3-3 isoforms reveals a hierarchical binding mechanism that supports the relevance of several structurally distinct, multi-valent NPM1/14-3-3 binding complexes. Additionally, we identified a selective, differential modulation of these interactions by a small molecule, fusicoccin (FSC), depending on a given NPM1 phospho-motif - from 1.4-fold inhibition to 115-fold stabilization. These results provide important clues regarding NPM1 regulation and how therapeutics targeting 14-3-3/NPM1 complexes could be used to treat disease, while also highlighting the utility of recently developed genetic code expansion (GCE) tools for generating and studying phosphorylated proteins and peptides.