Interactions between peptides/proteins and lipid membranes are involved in every aspect of cellular function. They are also central to numerous pathological processes [1], cellular infections by viruses and bacterial toxins 2, 3, the action of a new generation of peptide-based antibiotics [4], and cancer therapeutics [5]. Thus, understanding the basic physicochemical principles governing protein-lipid interactions and our ability to predict these interactions under various conditions are of utmost importance.

The bilayer interactions of membrane-active peptides can be divided into the three steps illustrated in Fig. 1a: membrane partitioning, interfacial (IF) folding, and transmembrane (TM) insertion [6]. Here, we use three different peptide systems to study the influence of Ca2+ and Mg2+ on these events: (1) Bid-BH3 is a 25 a.a. peptide derived from the BH3 domain of the apoptotic regulator Bid [7]; (2) prph2-CTER, a 68 a.a. peptide corresponding to the cytoplasmic C-terminus of the integral membrane protein peripherin-2 8, 9; (3) the cancer-targeting peptide pHLIP, initially derived from a 36 a.a. sequence of the helix-C of bacteriorhodopsin [10] (all sequences are presented in the Supplement). These three systems correspond to the three interaction modes in Fig. 1a (Bid-BH3, IF partitioning without folding; prph2-CTER, IF partitioning with partial folding; pHLIP, TM insertion), and for each of them, the addition of Ca2+ or Mg2+ leads to significantly more favorable membrane interaction.

The induction of reversible membrane binding by the addition of divalent cations is illustrated by the shift in the position of Trp fluorescence of Bid-BH3 in the presence of cardiolipin-containing large unilamellar vesicles (LUV) (Fig. 1b). This shift is only observed when both membranes and divalent cations are present, and it is reversed by chelating divalent cations with EDTA. To quantitively characterize the membrane partitioning of Bid-BH3 and determine the free energy of transfer from water to the bilayer interface (ΔGIF), we have utilized a well-established methodology based on measuring changes in Trp fluorescence intensity upon lipid titration 6, 11. The results presented in Fig. 1c demonstrate that fluorescence increases associated with the membrane partitioning of Bid-BH3 are only evident when either Ca2+ or Mg2+ is present. Estimated free energies of partitioning are favorable (ΔGIF = -6.6 kcal/ mol at 2 mM Ca2+ and -6.3 kcal/mol at 2 mM Mg2+), which is unexpected given that both peptides and membranes have a negative charge. Note that the effects of Ca2+ are different from that of the monovalent cation Na+, which inhibits insertion (Fig. 1e). (The effect of Na+ are consistent with the expectations from the reduction in the surface potential of the anionic membranes and have been reported in other systems 12, 13) A similar impact on Ca2+- and Mg2+-dependent partitioning is also observed with a very different system; the prph2-CTER peptide (Fig. 1d), which is largely disordered [14] but includes an interfacial helix 8, 9. The estimated free energy values are even more favorable for prph2-CTER (ΔGIF = -9 kcal/ mol), possibly due to the additional gain via partitioning folding coupling [15], which has been observed for this peptide previously [8]. Noticeably, no substantial folding is observed upon partitioning of Bid-BH3 (Fig. 1f), and prph2-CTER exhibits partial folding, with an estimated 20% of its sequence in a helical conformation (Fig. 1g).

The cancer-targeting peptide pHLIP can insert into lipid bilayers in a transmembrane conformation [16], which makes it a valuable model for studying the thermodynamics of the insertion transition. Originally, the insertion was shown to be triggered by acidification; however, our recent studies demonstrate that it also can be induced by divalent cations at neutral and even basic pH [17]. We have previously compared the thermodynamics of protonation-induced and divalent cation-induced insertion and demonstrated that their effects depend on lipid composition 17, 18, 19. (Note that thermodynamic analysis of the transition from interfacial to the inserted state of any peptide, unlike that from soluble to interfacial states illustrated in Fig. 1c-d, does not rely on titration experiments with increasing amounts of lipid. Details on the determination of ΔGTM for pHLIP insertion are described in Supplementary Methods and our previous publication [17]). Here, we present a summary of the free energies of Ca2+- and Mg2+-induced insertion of pHLIP at pH 7.5 (Fig. 1h). In all cases, the presence of divalent cations leads to a favorable insertion. The magnitude, however, clearly depends on lipid composition. For zwitterionic POPC (squares), the presence of 2 mM of either Ca2+ or Mg2+ led to a similar free energy change of 1.5-2 kcal/mol, sufficient to convert a predominantly non-inserted population of pHLIP into a predominantly inserted population. The difference is much more pronounced for pHLIP insertion into LUV containing 25% of the anionic lipid POPS (circles): 2.5 kcal/mol gain in favorable ΔGTM for 2 mM Mg2+ and 4.5 kcal/mol gain for 2 mM Ca2+.

The data presented here demonstrate that divalent cations can strongly influence distinct aspects of polypeptide-lipid interaction. These interactions should not be confused with the well-characterized cellular signaling with Ca2+ as a second messenger, which requires a highly specific action of Ca2+ over that of Mg2+. It is clearly not the case here, as the ability of Mg2+ to trigger interactions is similar to that of Ca2+ (Fig. 1). The implications of this lack of selectivity are rather important for the mediation of protein-bilayer interactions, especially given that cytoplasmic concentrations of Mg2+ are in millimolar range. We can reasonably assume that the two-fold effect of Ca2+ over Mg2+ demonstrated in Fig. 1b will be entirely negated by the several orders higher concentration of Mg2+, making it a predominant actor inside the cell. In contrast, Ca2+ will be expected to play a leading role outside the cell and inside Ca2+-rich compartments, such as the endoplasmic reticulum or mitochondria, or transiently during Ca2+ waves.

The fact that concentrations of Mg2+ (inside the cell) and Ca2+ (outside the cell) are relatively constant suggests that regulation of such interactions relies on factors other than changes in divalent cations concentrations. Instead, several recent publications point to changes in lipid composition as such triggers. For example, membrane interactions of the Bcl-2 family of apoptotic factors have been suggested to be regulated at a relatively constant intracellular level of Mg2+ by redistributing cardiolipin in mitochondria [13]. Similarly, the selective interactions of pHLIP with cells possessing elevated levels of exposed PS, which require the presence of Ca2+, can be one of the mechanisms by which pHLIP targets cancer cells (which, unlike non-transformed cells, expose PS at their surface) 20, 21.

How well can we predict membrane interactions in the presence of physiologically relevant concentrations of Ca2+ and Mg2+? The current prediction tools are insufficient to identify and describe such interactions properly. For example, the application of the MPEx webtool [22], recently updated to account for interfacial binding to anionic membranes [23], suggests a very unfavorable free energy of +8.4 kcal/mol for the interfacial partitioning of unfolded Bid-BH3 to LUV composed of 1TOCL:2POPC (same composition used in Fig. 1c). However, as shown in Fig. 1c, the presence of divalent cations leads to a very favorable interaction for this system. We suggest that the likely mechanism of the action of divalent cations involves the coupling of lipid headgroups and anionic sidechains, similar to that recently demonstrated for pHLIP with the help molecular dynamics simulations [24]. This difference of nearly 15 kcal/mol illustrates that predictive methods should be extended to include/consider the action of Ca2+ and Mg2+ and that systematic biophysical experiments to develop and test predictive models are imperatively needed.