Cd(II)- and Pb(II)-Induced Self-Assembly of Peripheral Membrane Domains from Protein Kinase C

Abstract

Cd²⁺ and Pb²⁺ are xenobiotic heavy metal ions that use ionic mimicry to interfere with the cellular function of biomacromolecules. Using a combination of SAXS, electron microscopy, FRET, and solution NMR spectroscopy, we demonstrate that treatment with Cd²⁺ and Pb²⁺ causes self-assembly of protein kinase C regulatory domains that peripherally associate with membranes. The self-assembly process successfully competes with ionic mimicry and is mediated by conserved protein regions that are distinct from the canonical Ca²⁺-binding motifs of protein kinase C. The ability of protein oligomers to interact with anionic membranes is enhanced compared to the monomeric species. Our findings suggest that metal-ion-dependent peripheral membrane domains can be utilized for generating protein–metal-ion nanoclusters and serve as biotemplates for the design of sequestration agents.

Cadmium (Cd) and lead (Pb) are potent environmental toxins whose harmful effects on human health are well-documented.^1–4 Ionic mimicry is considered to be the primary mechanism through which these metal ions interfere with cellular processes. It can involve (i) the cellular uptake of Cd²⁺ and Pb²⁺, through hijacking the transporters responsible for the import of essential metal ions^5–7 and (ii) intracellular competition for metal-ion-binding sites of proteins involved in cellular signaling,^8–13 metabolism,^14,15 and metal-ion homeostasis.^16,17 Here, we report another mode of Cd²⁺ and Pb²⁺ action that coexists and successfully competes with ionic mimicry. The phenomenon in question is the M²⁺ (M = Cd, Pb)-mediated self-assembly of peripheral membrane regions that belong to the lipid- and Ca²⁺-activated α isoform of protein kinase C (PKC). Both Pb²⁺ and Cd²⁺ activate PKC.^8,18–20

The N-term region (Figure 1A) controls the enzymatic activity of PKC: its membrane insertion in response to binding second messengers Ca²⁺ and diacylglycerol relieves the autoinhibition of the enzyme. The region comprises two types of peripheral membrane modules: the tandem conserved homology 1 (C1A and C1B) and conserved homology 2 (C2) domains. C1 domains are the treble-clef Zn²⁺ fingers. They bind membrane-embedded diacylglycerol as part of the PKC activation process (Figure 1B). The C2 domain in its metal-ion free state does not interact with membranes. Binding of Ca²⁺ to the negatively charged apical loop region drives the C2 domain to anionic membranes (Figure 1B). The divalent metal ions therefore play two distinct roles: Zn²⁺ ensures the structural integrity of the diacylglycerol-sensing C1 domains, while Ca²⁺ facilitates the C2 membrane insertion as part of the PKC activation process. We previously demonstrated that Pb²⁺ and Cd²⁺ bind to the loop regions of isolated C2 domains^9,10,21,22 with a high affinity and thereby can act as competitive inhibitors of Ca²⁺ binding.

Figure 1.

Modular structure (A) and peripheral domains of conventional PKC isoenzymes (B). Shown is the ribbon representation of C1B (NMR structure, 2ELI) with highlighted structural Zn²⁺ sites, and C2 (crystal structure, 1DSY) in complex with Ca²⁺ and phosphatidylserine analogue, PSF. DAG stands for diacylglycerol.

To determine the effect of divalent metal ions on the N-term regulatory region of PKC, we conducted SAXS experiments (sections S1–3 and Table S2) on the 22.3 kDa C1B–C2 region from PKCα and the 30.6 kDa chimeric full regulatory region C1C2c that additionally contains the C1A domain from the γ isoform.²³ Three metal ions were used: Cd²⁺, Pb²⁺, and native activator Ca²⁺. Unexpectedly, we found that Cd²⁺ and Pb²⁺ caused significant changes in protein morphology compared to Ca²⁺ (Figure 2A). Moreover, based on the similarity of the Cd²⁺ and Pb²⁺ scattering curves, we concluded that the effect of these metal ions on the N-term regulatory regions is similar.

Figure 2.

Cd²⁺- and Pb²⁺-induced C1B–C2 self-association results in loss of conformational flexibility and formation of globular species.(A) SAXS scattering curves of the 2- and 3-domain regulatory regions showing the differences between Ca²⁺- and Cd²⁺/Pb²⁺-bound proteins. Changes in size and flexibility/shape caused by Cd²⁺ and Pb²⁺ are evident from the comparison of pair-distance distribution functions P(r) (B) and dimensionless Kratky plots (C), respectively.

Analysis of the pair-distance distribution functions, P(r), revealed that interactions with Cd²⁺ and Pb²⁺ cause the self-assembly of the N-term regions. This is manifested in (i) a ~2-fold increase in the radius of gyration, R_g, and (ii) 2- to 3-fold increase in the maximum distance present in the scattering particle, D_max, both relative to the Ca²⁺-bound species (Figure 2B). Estimation of the molecular mass produced the following values in kDa: 20 (Ca²⁺/C1B–C2, monomer mass 22.3 kDa), 32 (Ca²⁺/C1C2c, monomer mass 30.6 kDa), 310 (Cd²⁺/ C1B–C2), 310 (Cd²⁺/C1C2c), and 260 (Pb²⁺/C1C2c). These data indicate that Ca²⁺-complexed proteins remain mostly monomeric, whereas Cd²⁺ and Pb²⁺-complexed proteins self-assemble with n ~ 8–14. The Pb²⁺/C1B–C2 scattering curves were not suitable for quantitative analysis (section S3).

To evaluate the flexibility and shape of the protein complexes, we constructed dimensionless Kratky plots.²⁴ The plots of globular proteins converge to zero at high scattering angles and have a maximum of 1.1 at qR_g = √3. Ca²⁺-complexed C1B–C2 and C1C2c lack these features and have a plateau-like region at high a qR_g that is typical of flexible multidomain proteins (Figure 2C). In contrast to Ca²⁺, all Cd²⁺ and Pb²⁺ plots have maxima at (√3, 1.1) and show a significant decrease in intensity at high qR_g values. We conclude that the Ca²⁺-complexed regulatory regions behave as flexible monomers in solution, consistent with their modular architecture and unstructured linkers. Interactions with Cd²⁺ and Pb²⁺ result in the formation of oligomeric species that are more globular in shape, with attenuated mobility of the linker regions. For the rest of the work reported here, we used a 2-domain C1B–C2 membrane-binding region due to its favorable solubility properties.

We conducted negative-stain electron microscopy experiments to visualize the oligomers (Figure 3A,B; Ca²⁺ control, Figure S1). The overall globular (Cd²⁺) and ellipsoidal (Pb²⁺) shapes of C1B–C2 oligomers are evident from the inspection of characteristic class averages (bottom panel of Figure 3A,B). The Cd²⁺-containing particles appear to be composed of 4 spherical densities, each with an estimated molecular mass of 80 kDa. Given that the molecular mass of the monomer is 22.3 kDa, this would correspond to a 320 kDa oligomer consisting of ~16 monomers. This is in good agreement with SAXS results. The footprint of Pb²⁺-containing particles appears as two parallel ellipsoids, with each ellipsoid having an estimated molecular mass of ~40 kDa. This arrangement would produce an 80 kDa tetramer. The distributions of particle sizes are centered at ~12.5 and ~15.0 nm for the Pb²⁺- and Cd²⁺- containing particles, respectively (Figure 3C).

Figure 3.

Electron micrographs and class averages of (A) Cd²⁺- and (B) Pb²⁺-complexed C1B–C2. (C) Distribution of Feret diameters in Cd²⁺- and Pb²⁺-complexed C1B–C2. (D) Schematic representation of FRET experiments. C2 tryptophan residues and the dansyl-labeled lipids in the LUVs serve as a donor and acceptor, respectively. (E, F) Fluorescence emission spectra of C1B–C2 in the presence of LUVs as a membrane mimic. The appearance of dansyl emission peak at 495 nm due to membrane binding is shown with an arrow. Inset in part F: spectra of the isolated C2 domain in the presence of LUVs and Cd²⁺.

Protein-to-membrane FRET experiments (Figure 3D) demonstrated that both Pb²⁺ and Cd²⁺-complexed C1B–C2 oligomers associate with anionic lipid membranes. This is manifested in the appearance of the dansyl emission peak upon addition of Pb²⁺ (Figure 3E) and Cd²⁺ (Figure 3F) to C1B–C2. Ca²⁺ control is given in Figure S2. EGTA treatment eliminated the dansyl peak, indicating that C1B–C2 dissociates from membranes upon sequestration of metal ions. In Pb²⁺ experiments, the intensity of the Trp peak does not fully recover upon EGTA treatment because C2 has one high-affinity Pb²⁺ site⁹ that remains populated. We were surprised to find that Cd²⁺ supported the membrane association of C1B–C2 oligomers (Figure 3F), because we knew from previous work that, unlike Pb²⁺, Cd²⁺ binding to the isolated C2 domain does not drive the membrane association (inset of Figure 3F and ref 10). The most plausible explanation for the enhancement of membrane-binding affinity of the C1B–C2 oligomers is the increase in the effective concentration of C2 membrane-binding functionalities in the oligomers. We recently observed a similar behavior for the two C2 domains of Synaptotagmin I,²¹ where a high local concentration was created through protein tethering to the membranes via the N-terminal segment.

There exists experimental evidence that Ca²⁺ binding to the C2 domain can facilitate self-association of PKCα.^25,26 We previously demonstrated, using NMR relaxation techniques, that Ca²⁺-dependent dimerization of the isolated C2 domain in solution is extremely weak, with ~2–3% population of dimeric species.²⁷ Similarly, our Ca²⁺/C1B–C2 SAXS (Figure 2) and EM data (Figure S1) indicate that Ca²⁺ does not cause significant self-association of C1B–C2. The sharp contrast between Ca²⁺- and Cd²⁺/Pb²⁺-dependent behaviors of C1B–C2 suggested that xenobiotic metal ions act through a mechanism that is distinct from Ca²⁺. To gain insight into the mechanism of Cd²⁺- and Pb²⁺-driven self-assembly, we used 2D¹⁵N–¹H NMR spectroscopy.

Treatment of C1B–C2 with Cd²⁺ and Pb²⁺ resulted in (i) ¹H and ¹⁵N chemical shift changes due to Cd²⁺ and Pb²⁺ binding to the C2 loops, consistent with our previous findings;^9,10 (ii) progressive decrease in cross-peak intensities due to self-assembly; and (iii) extreme attenuation of the backbone N–H cross-peak intensities in three protein regions (Figure 4A). The latter indicates that the affected residues participate in a chemical exchange process that is intermediate on the NMR chemical shift time scale. We speculated that the broadened regions: loop 3 of the C2 domain, the C-terminal α-helix of the C2 domain, and the linker region connecting C1B and C2 may be directly involved in metal-ion-mediated self-assembly.

Figure 4.

Self-assembly is mediated by noncanonical metal-ion-binding motifs. (A) Normalized intensities of the N–H backbone cross-peaks in Pb²⁺-complexed C1B–C2. Arrows indicate three regions that show extreme broadening. (B) Ribbon representation of C1B–C2 with a modeled linker region. Cd²⁺ and Zn²⁺ are depicted as purple and tan spheres, respectively. Also shown is the sequence alignment of the linker and helical regions in conventional PKCs, with putative Cd²⁺- and Pb²⁺-coordinating residues highlighted in red. (C, D) ¹⁵N-filtered ¹H NMR spectra of the amide region in wild-type C1B–C2 and its LH variant. (E, F) Size-exclusion chromatography of C1B–C2 and its LH variant in the presence of Cd²⁺ (E, Superdex 200 Increase 10/300) and Pb²⁺ (F, Superdex 75 10/300 GL). The population of oligomeric species is dominant in the wild-type species, but is undetectable in the LH variant.

To test this, we first eliminated the ability of the C2 domain to coordinate metal ions, by mutating Asp246 of loop 3 to Asn (D246N), or mutating Asp187 and Asp193 of loop 1 to Asn (D187N/D193N) (see Figure 4B for the location of the residues). Size-exclusion chromatography (Figure S3A) revealed that the effect of these mutations on self-assembly is either modest (D246N) or nonexistent (D187N/D193N) (Figure S3B). We concluded that ionic mimicry, i.e. binding of Cd²⁺ and Pb²⁺ to the canonical Ca²⁺ binding sites, is not the major driving force behind the self-assembly of the N-term region.

The C-terminal α-helix has three glutamic acids, Glu281, Glu282, and Glu284, and a glutamine, Gln280. The linker region contains Asp154, His155, and Glu157 (Figure 4B). All of these residues can potentially coordinate divalent metal ions. We reasoned that if indeed these regions contribute to the intermolecular coordination of Cd²⁺ and Pb²⁺, the mutation of charged residues would reduce the extent of self-assembly. Charge reversal of three acidic residues: Asp154, Glu157, and Glu281 and replacement of His155 with Gly resulted in a significant increase in monomeric C1B–C2 species (Figure S3C).

We then constructed a “Linker-Helix” (LH) C1B–C2 variant, in which we replaced the entire linker region with glycines and reversed the charge of the α-helix by replacing the three glutamic acids with lysines. The ¹⁵N-filtered ¹H NMR spectra in the presence of Cd²⁺ (Figure 4C) and Pb²⁺ (Figure 4D) clearly show that while in the wild-type C1B–C2 the integrated intensity of amide region decreases due to self-assembly; no such decrease is observed in the LH variant. Consistent with the NMR results, the LH elutes from the size-exclusion columns as monomeric species, in contrast to the mostly oligomeric wild-type C1B–C2 (Figure 4E,F).

Replacing only the linker region with glycines was not sufficient to completely abolish the self-assembly (Figure S4). In aggregate, our data indicate that both the linker region connecting C1B and C2 and the α-helix of the C2 domain are the major drivers of the C1B–C2 self-assembly. These regions are highly conserved in all Ca²⁺-dependent PKC isoenzymes across different species (Figure S5). The C1B domain does not appear to have a significant role in the self-assembly of the N-term region. According to our NMR data, structural integrity of C1B is not perturbed by treatment with Cd²⁺ and Pb²⁺.

In summary, Cd²⁺- and Pb²⁺-driven self-assembly of the N-term regulatory region is the most unexpected and surprising finding of our study. The implication of our results is that, in addition to ionic mimicry, one of the mechanisms of Pb²⁺ and Cd²⁺ toxicity could be the clustering of proteins at the membrane surface. Using sedimentation velocity experiments, we estimated that the apparent affinity of Cd²⁺-driven self-association is < 7 μM at a physiological pH of 7.4 (Figure S6). This is comparable to the 5–40 μM range reported for the peptoid ligands that were specifically designed for Cd²⁺ sequestration from biological fluids.²⁸ The C1B–C2 system, specifically the amino acid composition of regions mediating self-assembly, can therefore provide a useful biological template for designing highly efficient sequestration agents that are free of limitations associated with chelation therapy. Finally, our data and recent reports^29–31 on the supra-molecular ring structures of C2 domains from Synaptotagmin 1 suggest that Ca²⁺-dependent C2 domains in combination with xenobiotic metal ions could be useful building blocks for creating protein–metal-ion nanostructures.