Analysis of Interactions Stabilized by Fusicoccin A Reveals an Expanded Suite of Potential 14–3–3 Binding Partners

Abstract

Fusicoccin A (FC) is a diterpene glycoside that stabilizes protein–protein interactions (PPIs) between 14–3–3 adapter proteins and their phosphoprotein interaction partners. Recently, FC has gained attention for its pro-apoptotic and neuroprotective properties in cell culture. Although the exact molecular mechanism(s) is (are) unresolved, 14–3–3 PPIs are central to this activity. With the goal of refining the pharmacology of this chemotype, we conducted a systematic analysis of the structural features that govern FC-induced stabilization of 14–3–3 PPIs utilizing a C-terminal phosphorylation recognition motif. This study confirmed that a C-terminal amino acid with a small alkyl group is required for the interaction of FC at canonical C-terminal 14–3–3 PPI interfaces. Using bioinformatics, this structural insight was leveraged to assemble a database of 119 candidate 14–3–3 PPIs that can serve as targets for FC. This group includes a subset of proteins with experimentally determined C-terminal phosphosites that have not been explored as potential targets of FC.

Graphical Abstract

Protein–protein interactions (PPIs) encompass an important contemporary class of drug targets.^1,2 Existing small-molecule-based strategies for PPI modulation focus on inhibition. This approach has been effective in instances where a well-defined conformational binding motif (e.g., an α-helix) at the PPI interface supports the rational design of a peptidomimetic.^3,4 In contrast, the strategy of using a stabilizer molecule to promote PPIs is comparatively underexplored.⁵ In principle, this mode of action provides an avenue to interrogate PPI networks where inhibition is intractable; however, the development of PPI stabilizers is intrinsically challenging because there is no clear starting point for initial ligand identification.

Natural products offer a rich source of molecular scaffolds that can be used to develop PPI stabilizers.⁶ With this idea in mind, we became interested in fusicoccin A (1, Figure 1), a complex diterpene glycoside that stabilizes PPIs involving 14–3–3 proteins.^7,8 These adapter proteins are tightly integrated into phosphorylation-dependent cell signaling pathways, and their aberrant function is linked to the pathobiology of cancer and certain neurological disorders.^9,10 Nevertheless, 14–3–3 is widely perceived as an intractable target because it forms PPIs with over 500 client proteins.^11,12 In most cases, these interactions are mediated through a canonical “internal” (RXX_pZXP, X = any residue, _pZ = phosphorylated S or T) phosphorylation recognition motif presented on the surface of client proteins that locates to the amphipathic binding groove of 14–3–3.¹³ Intriguingly, 1 is not compatible with this group of 14–3–3 PPIs. Instead, the selectivity profile of 1 is largely constrained to a distinct part of the 14–3–3 interactome where binding is mediated though an atypical C-terminal phosphorylation recognition motif (X_pZX_COOH).^14–17

Figure 1.

A key contact between 1 (khaki) and the C-terminal valine of 2 (purple) in the 14–3–3 binding groove (gray) revealed by analysis of the 14–3–3σ·ERα-ctp·1 ternary complex (PDB 4JDD).

Recently, 1 has gained attention for its pro-apoptotic and neuroregenerative activity in human cell culture.^18,19 Although the precise molecular mechanism(s) remain(s) undefined, 14–3–3 PPIs contribute to the pharmacology.²⁰ This view is supported by characterized interactions between 1 and various systems that recapitulate C-terminal 14–3–3 PPIs in vitro.¹¹ As a result, further development of this chemotype requires a systematic investigation of the structural features that govern stabilization of C-terminal 14–3–3 PPIs and a comprehensive profile of 14–3–3 client proteins that are compatible with 1. To this end, Aducci and Camoni reported that the stabilizing effect of 1 is primarily dependent on the identity of the C-terminal residue of the client phospholigand.²¹ Using in silico site-directed mutagenesis, this study predicted that C-terminal recognition motifs with hydrophobic, polar, or charged C-terminal residues could be targets for 1. Ohkanda subsequently showed that only 14–3–3 clients possessing the aliphatic residues I, V, and L at the C-terminus were strongly stabilized by 1.²² However, this study utilized a semisynthetic variant of 1 harboring a modified diterpene core attached to a non-natural furanose ring system. As a result, we set out to resolve this discrepancy and, in the process, develop a tool to predict C-terminal 14–3–3 PPIs that are compatible with biosynthetic 1.

Our investigation began by selecting a suitable C-terminal 14–3–3 PPI to benchmark the activity of 1. As shown in Figure 1, the complex formed between 14–3–3σ and estrogen receptor alpha (ERα) possessed several attractive features: (i) The 14–3–3 interaction site of ERα is located at the C-terminus of the F-domain, and binding to 14–3–3 is known to be dependent on phosphorylation of the penultimate threonine (T594) in vivo. (ii) The dynamics between 14–3–3σ and ERα have been assayed in vitro using a synthetic peptide (15 amino acids) harboring the 14–3–3 recognition sequence (⁵⁹⁰GFPA_pTV_COOH), referred to herein as the ERα C-terminal phospholigand (ERα-ctp, 2), and (iii) an X-ray structure of the 14–3–3σ·ERα-ctp·1 ternary complex (Protein Data Bank [PDB] ID: 4JDD, 2.1 A resolution) was available to guide our analysis.²³ Notably, among the seven isoforms of 14–3–3 present in humans, 14–3–3σ is arguably the most important in cancer biology.^9,24 Therefore, we focused on analyzing interactions between full-length, recombinant 14–3–3σ (see the SI for details) and the synthetic ERα-ctp mimetic 2 in the presence and absence of 1.

As shown in Figure 2, we started by establishing a uniform method to evaluate the stabilization induced by biosynthetic 1. Using isothermal titration calorimetry (ITC), we found that peptide 2 binds to 14–3–3σ with a K_d of 0.75 ± 0.14 μM. In contrast, the K_d of 2 for 14–3–3σ decreased to 0.02 ± 0.05 μM when 200 μM 1 was present. Thus, 1 stabilizes the 14–3–3σ·ERα-ctp complex by approximately 40-fold. This result was confirmed by fluorescence polarization (FP) experiments. Using this technique, we determined that the apparent affinity of fluorescein-labeled 2 (N-FAM-2) and 14–3–3σ was enhanced by approximately 70-fold, from an apparent K_d of 3.61 ± 0.41 μM to 0.05 ± 0.02 μM, when titrated in the presence of 80 μM 1. We also found that 1 stabilized the 14–3–3·N-FAM-2 complex with an EC₅₀ of 3.16 ± 1.03 μM. Given that the apparent affinity determined by FP using N-FAM-2 was comparable to the K_d measured by ITC using 2, we concluded that the more convenient fluorescence polarization-based assay would be used to analyze 14–3–3/phospholigand interactions moving forward.

Figure 2.

Stabilization of the 14–3–3σ·ERα-ctp complex by fusicoccin A (1): (a) Isothermal titration calorimetry (ITC) of ERα-ctp (2) binding to 14–3–3σ in the absence (innate) or presence (stabilized) of 200 μM 1. (b) Fluorescence polarization (FP) measurements of fluorescein-labeled N-FAM-ERα-ctp (N-FAM-2) in the absence (innate) or presence (stabilized) of 80 μM 1 titrated with 14–3–3σ to obtain the apparent K_d of the 14–3–3σ·ERα-ctp complex. (c) Fluorescence polarization (FP) measurements of N-FAM-2 and 14–3–3σ titrated with 1 to obtain an EC₅₀ value for its stabilizing activity toward the 14–3–3σ·ERα-ctp complex. (d) A comparison of ITC and FP methods for measuring the fold stabilization (S_F) induced by 1. Error bars indicate the mean ± SD of at least two experiments.

Next, we chose to revisit a series of structurally distinct C-terminal 14–3–3 PPIs from the literature. The five targets we selected have been analyzed in the presence and absence of 1 previously;²¹ however, in most cases, existing studies relied on alternative analytical methods, different 14–3–3 isoforms, and/or longer synthetic phosphopeptides. As shown in Table 1, we focused on the interactions of 14–3–3σ and a series of fluorescein-labeled hexaphosphopeptides replicating the C-terminal recognition sequences of the potassium ion channel Task-3,²⁵ the platelet adhesion receptor GpIBα,²⁶ the G-protein coupled receptor GPR15,²⁷ the interleukin 9 receptor α chain (IRL9),²⁸ and the cyclin-dependent kinase inhibitor p27Kip1.²⁹ Importantly, our analysis of the 14–3–3σ·Task3-ctp complex in the absence (K_d = 3.15 ± 1.12 μM) and presence of 80 μM 1 (K_d = 0.10 ± 0.05 μM) were in agreement with data reported by Ottmann using a similar FP assay and identical interaction partners.²⁵ We also observed a stabilizing effect of 1 in combination with 14–3–3σ and either GpIBα-ctp (S_F = 14 ± 2) or GPR15-ctp (S_F = 7 ± 2). Notably, both sequences possess a C-terminal leucine, and the magnitude of increased affinity induced by 1 is consistent with literature reports.²¹ In contrast, the IRL9-ctp, which terminates in phenylalanine, was not strongly stabilized by 1 (S_F = 3 ± 1). Similarly, 1 did not markedly enhance the affinity of 14–3–3σ for the p27Kip1-ctp, which possesses a C-terminal phosphothreonine (S_F = 3 ± 2), and intriguingly, the innate affinity of the 14–3–3σ·p27Kip1-ctp complex itself was diminished (apparent K_d = 82.2 ± 31.0 μM). Taken together, these results reinforce the idea that the identity of the final residue within the C-terminal recognition motif is a key determinant of the activity of 1 at 14–3–3 PPI interfaces. These data also suggest that the C-terminal residue plays a direct role in innate binding to 14–3–3.

Table 1.

Apparent K_d and Fold Stabilization (S_F) of Known C-Terminal 14–3–3 PPIs in the Absence and Presence of 80 μM 1^a

client protein	phospholigand sequenceb	innate Kd (μM)	Kd (μM) in presence of 1c	SF
Task3	RKKRpSVCOOH	3.15 ± 1.12	0.10 ± 0.05	14 ± 2
GpIBα	YSGHpSLCOOH	22.5 ± 9.71	1.69 ± 0.30	14 ± 2
GPR15	KRSVpSLCOOH	7.36 ± 0.36	1.10 ± 0.09	7 ± 1
IRL9	ARSWpTFCOOH	3.16 ± 0.32	1.54 ± 0.26	3 ± 1
p27Kip1	ALRRQpTCOOH	82.2 ± 31.0	39.9 ± 8.09	3 ± 2

^aReported apparent K_d is the mean ± standard deviation of at least two experiments.

^bSynthetic hexapeptides were labeled at the N-termini with the fluorophore fluorescein.

^cTitrations carried out in the presence of 80 μM 1. See the Supporting Information for details.

Analysis of the 14–3–3σ·ERα-ctp·1 ternary structure shown in Figure 1 shows that interactions between 1 and client phospholigand 2 in the 14–3–3σ binding groove are restricted to a hydrophobic contact between the 5–8–5 tricyclic core of 1 and the isopropyl side chain of the C-terminal valine within 2. On the basis of this observation, we hypothesized that amino acids with small alkyl side chains could make similar contacts with 1 and therefore support formation of a ternary complex. However, it was more challenging to predict which of these C-terminal residues might support binding to 14–3–3σ in the absence of 1, particularly given the intrinsic plasticity of the 14–3–3 binding groove that can accommodate hundreds of client phosphoproteins.¹² Therefore, to interrogate these features systematically, we analyzed a focused library of N-FAM-2 variants with single amino acid changes at the C-terminus.

As summarized in Table 2, we observed a significant difference in the affinity of ERα-ctp variants for 14–3–3σ depending on the identity of the C-terminal residue. The results were divided into three groups: The first group contains hydrophobic residues I and V, which had micromolar affinities for 14–3–3σ (apparent K_d ~ 3.0–3.5 μM). A second group includes residues F, L, and A, which exhibited moderate affinities for 14–3–3σ (apparent K_d ~ 10.5–20.0 μM). The last group contained T, N, S, G, and D, which showed comparably weak affinity for 14–3–3σ (apparent K_d 50.0–200 μM). These titrations highlight a clear preference for docking of a hydrophobic C-terminal residue in the 14–3–3 binding groove.

Table 2.

Impact of the Modulation of the C-Terminal Residue within ERα-ctp on the Apparent Affinity of the 14–3–3σ·ERα-ctp Complex in Both the Absence and Presence of 1^a

C-terminal residue (X)	innate Kd (μM)	Kd (μM) in the presence of 1b	ΔG stabilization (kcal·mol–1)c
I	3.03 ± 0.31	0.06 ± 0.01	2.3
V	3.61 ± 0.41	0.05 ± 0.02	2.5
F	10.5 ± 3.36	4.03 ± 0.26	0.6
L	11.0 ± 2.72	0.17 ± 0.02	2.5
A	19.9 ± 6.99	0.70 ± 0.05	2.0
T	50.0 ± 13.3	1.32 ± 0.14	2.2
S	79.8 ± 9.62	3.54 ± 0.44	1.8
N	59.5 ± 13.5	32.4 ± 4.75	0.3
G	144 ± 60.0	36.5 ± 17.3	0.8
D	173 ± 82.3	48.2 ± 0.26	0.7

^aApparent K_d reported as the mean ± standard deviation of two experiments.

^bTitrations carried out in the presence of 80 μM 1. See the SI for details.

^cΔG stabilization was calculated using mean K_d values in the absence and presence of 1.

Next, we compared the affinities of our phospholigand library with 14–3–3σ in the presence of 80 μM 1. Consistent with our hypothesis, variants of N-FAM-2 bearing a C-terminal residue with an alkyl side chain were stabilized by 1. We found that residues I, V, L, A, T, and S supported a substantial decrease in apparent K_d in the presence of 1 (>20-fold, Table 2). The effect of 1 on the alanine variant was particularly striking. In this case, the apparent affinity to 14–3–3σ was enhanced by approximately 30-fold, from a K_d of 19.9 ± 6.99 μM to 700 ± 50 nM, in the presence of 80 μM 1. When interpreted in terms of ΔG of stabilization (kcal·mol^–1), the relative ranking of C-terminal residues was found to be V or L (2.5) > I (2.3) > T (2.2) > A (2.0) > S (1.8). In contrast, variants of N-FAM-2 harboring F, N, G, and D at the C-terminus were not stabilized to a large extent by 1 (ΔG < 0.9 kcal/mol). These results point to an upper (N) and lower (G) constraint on the size of amino acid side chains that support interactions with 1 in the 14–3–3 binding groove. In this regard, hydrophobic residues L, V, and I with branched side chains were optimal, and the stabilizing effect of 1 decreased as the size of the C-terminal residue was decreased (i.e., V → A → G). A direct comparison of the V and T variants of N-FAM-2 indicated that polar functional groups were tolerated but diminished the stabilizing effect of 1. Consistent with this model, we found that the S variant of N-FAM-2 was stabilized by 22-fold in the presence of 1, whereas the corresponding A variant was stabilized by 29-fold under identical conditions. In contrast, slightly larger polar C-terminal residues (i.e., N) and C-terminal residues with small, charged side chains (i.e., D) disrupt hydrophobic contacts with 1.

With an improved picture of the structural criteria required for stabilization by 1, we employed bioinformatics to establish the scope of potential 14–3–3 client phosphoproteins that might serve as biological targets for 1 in vivo. Thus, using the C-terminal recognition motif as a search criterion, we performed a BLAST search to identify human proteins terminating in the motif -XZB_COOH (Z = T or S and B = V, L, I, T, A, or S).³⁰ This inquiry returned over 800 unique FASTA sequences, which were further subjected to filtering algorithms. Using NetPhos 3.1,³¹ we compiled a list of proteins with predicted phosphorylation at the penultimate S or T residue. These putative client phosphoproteins were then filtered using the 14–3–3 Prediction Tool.³² This analysis returned a list of 119 C-terminal 14–3–3 PPIs that are candidate targets for 1 (the full list is reported in the SI). Next, utilizing the PhosphoSitePlus database, we evaluated each candidate phosphoprotein for experimentally validated phosphorylation at the predicted phosphosite.³³ This search uncovered a subset of proteins within our initial list that have a reported C-terminal 14–3–3 phosphorylation recognition motif (Figure 3). These proteins are the most likely to be new targets for 1 in vivo. Importantly, our bioinformatics analysis also captured the C-terminal 14–3–3 PPIs previously reported with 1, including the sequences shown in Table 1.

Figure 3.

Select proteins that are putative targets for fusicoccin A (1): (a) FP measurements of fluorescein-labeled Nup160-ctp (3) in the absence (innate) or presence (stabilized) of 80 μM 1 titrated with 14–3–3σ to obtain the apparent K_d of the 14–3–3σ·Nup160-ctp complex. (b) The same FP experiment repeated with fluorescein-labeled TNFRSF19-ctp (4). (c) A comparison of the fold-stabilization (S_F) induced by 1 using phospholigands 3 and 4. Error bars indicate the mean ± SD of at least two experiments.

As shown in Figure 3, this study identified several disease relevant proteins that have not been explored as interaction partners for 14–3–3 or as biological targets of 1. Thus, in an initial effort to validate the results of our search, we probed interactions between 14–3–3σ and phosphopeptides replicating the putative C-terminal 14–3–3 recognition domains of two structurally distinct candidate proteins from our list. The first was nucleoporin 160 (Nup160-ctp, 3),³⁴ which binds to 14–3–3σ with an apparent K_d of 5.34 ± 0.31 μM in the absence of 1 and with an apparent K_d of 0.25 ± 0.02 μM when titrated in the presence of 80 μM 1 (Figure 3A). The affinity of the second candidate protein, TNF receptor 19 (TNFRSF19-ctp, 4),³⁵ was also improved from an apparent K_d of 538 ± 141 μM to 38.0 ± 2.91 μM, when titrated in the presence of 80 μM 1 (Figure 3B). Importantly, the 80 μM concentration of 1 used in these experiments exceeds the concentrations of 1 reported to illicit phenotypic changes in cell culture (~2–20 μM) and is far beyond the EC₅₀ recorded for the 14–3–3σ·ERα-cpt PPI in vitro (~3 μM).^18,23 It is also noteworthy that the sequences of these phospholigands are different at every position except for the C-termini, and although 3 was a superior ligand for 14–3–3σ, the magnitude of stabilization afforded by 1 was comparable for 3 and 4 (20-fold vs 14-fold, Figure 3C). As such, these results demonstrate that the presence of a small alkyl residue at the C-terminus is the main requirement for stabilization of canonical C-terminal 14–3–3 PPIs by 1.

In conclusion, we have shown that for C-terminal 14–3–3 PPIs, 1 is best accommodated in the 14–3–3 biding groove when the client phosphoprotein terminates with V, L, I, T, A, or S. Using bioinformatics, we have leveraged this observation to assemble a database of 119 candidate 14–3–3 PPIs that can be targets for 1 in vivo. This group is notably enriched in proteins terminating with L, V, or I (68 out of 119), the residues found to be most compatible with 1 in Table 2. To further validate this analysis, we have demonstrated that interactions between 14–3–3σ and potential phospholigand targets identified from our search are enhanced by 1 in vitro. More broadly, this study points to an expanded family of potential molecular targets for 1. As a result, we contend that chemical synthesis will play a critical role unlocking the utility of 1 as a chemical probe of 14–3–3 functions.

Supplementary Material

Supplementary Material

Figure S1. Representative 12% SDS-PAGE gel of purified His-14–3-3σ (27 kD band). Lanes from right to left indicate: (1) ladder, (2) post-induction fraction, (3) whole cell lysate, (4) supernatant, (5) column flow-through, (6) wash, (7) 4 μg, (8) 8 μg, (9) 10 μg, (10) 12 μg.