Cytosolic Ca²⁺ Buffers

Abstract

“Ca²⁺ buffers,” a class of cytosolic Ca²⁺-binding proteins, act as modulators of short-lived intracellular Ca²⁺ signals; they affect both the temporal and spatial aspects of these transient increases in [Ca²⁺]_i. Examples of Ca²⁺ buffers include parvalbumins (α and β isoforms), calbindin-D9k, calbindin-D28k, and calretinin. Besides their proven Ca²⁺ buffer function, some might additionally have Ca²⁺ sensor functions. Ca²⁺ buffers have to be viewed as one of the components implicated in the precise regulation of Ca²⁺ signaling and Ca²⁺ homeostasis. Each cell is equipped with proteins, including Ca²⁺ channels, transporters, and pumps that, together with the Ca²⁺ buffers, shape the intracellular Ca²⁺ signals. All of these molecules are not only functionally coupled, but their expression is likely to be regulated in a Ca²⁺-dependent manner to maintain normal Ca²⁺ signaling, even in the absence or malfunctioning of one of the components.

Calcium-binding proteins, such as parvalbumin, modulate spatiotemporal aspects of calcium signals. They are integral components of a system of mutually dependent pumps, channels, and other proteins that preserves calcium homeostasis.

DEFINITION OF A CYTOSOLIC Ca²⁺ BUFFER

Molecules serving as chelators for Ca²⁺ ions must contain negatively charged groups arranged in such a way as to fulfill the necessary geometrical constraints for chemical coordination. Proteins with appropriately spaced acidic side-chain residues (e.g., glutamate, aspartate) and/or backbone carbonyl groups provide the “cage” in which a Ca²⁺ ion may fit in. Evolutionarily well-conserved protein families differing in the way the Ca²⁺ ions are bound include the annexins, the C2-domain proteins, the EF-hand proteins, the pentraxins, the vitamin-K-dependent proteins, and the intraorganellar low-affinity, high-capacity Ca²⁺-binding proteins; for more details on the structural diversity of EF-hand motifs, see Gifford et al. (2007), and for other Ca²⁺-binding sites, see Bindreither and Lackner (2009). However, the term Ca²⁺ buffer is applied only to a small subset of cytosolic proteins of the EF-hand family, including parvalbumins (PV; alpha and beta isoforms), calbindin-D9k (CB-D9k), calbindin-D28k (CB-D28k), and calretinin (CR). The majority of EF-hand proteins belong to the group of “Ca²⁺ sensors”; that is, binding of Ca²⁺ ions induces a conformational change, which permits them to interact with specific targets in a Ca²⁺-regulated manner. The prototypical examples of Ca²⁺ sensors are calmodulin (Chin and Means 2000) and proteins of the S100 family. However, if present at sufficiently high concentrations, Ca²⁺ sensors may also function as Ca²⁺ buffers. Importantly, cytosolic Ca²⁺ buffers do not act as buffers in analogy to chemical buffers such as pH buffers. The latter serve to clamp the pH to a predetermined value in such a way that the addition of an acid or a base elicits only a minor change in the pH of the solution. Therefore, the buffering capacity is highest when the pH is close to the pK value of the corresponding acid/base pair. The situation is fundamentally different for the cytosolic Ca²⁺ buffers. Under basal conditions, [Ca²⁺]_i is in the order of 20–100 nM. Yet, the dissociation constants for Ca²⁺ (K_D,Ca) of most Ca²⁺ buffers are almost one order of magnitude larger, ≈200 nM–1.5 µM (Table 1). Thus, in a resting cell, Ca²⁺ buffers are at large in their Ca²⁺-free state, ready to bind Ca²⁺ ions whenever [Ca²⁺]_i increases. As a result of their specific properties, Ca²⁺ buffers act in different ways to modulate the spatiotemporal aspects of cytosolic Ca²⁺ signals. How a given Ca²⁺ buffer affects intracellular Ca²⁺ signals depends on several parameters, including the intracellular concentration (Intracellular Concentration), the affinity for Ca²⁺ and other metal ions (Metal-binding Affinities), the kinetics of Ca²⁺ binding and release (Metal-binding Kinetics), and the intracellular mobility (Mobility and Interaction with Ligands). When measuring the total Ca²⁺-buffering capacity (κ_S) of a cell, often the distinction is made between mobile and immobile buffers (Zhou and Neher 1993). The latter ones are defined as molecules capable of binding cytosolic Ca²⁺ that are not washed out when, for example, the plasma membrane is patched with a pipette. Very little is known about the molecular identity of immobile Ca²⁺ buffers, except their relatively low Ca²⁺ affinity; presumably, they are made up of cytosol-exposed stretches of membrane proteins and/or membrane-associated proteins with a rather low affinity for Ca²⁺ ions. Additionally, negatively charged phospholipid headgroups of the inner leaflet of the plasma membrane serve as “weak” Ca²⁺ chelators (McLaughlin et al. 1981). Thus, the slowly mobile or immobile buffers, together with mobile Ca²⁺ buffers, are responsible for the rather slow diffusion of Ca²⁺ ions inside a cell (Mobility and Interaction with Ligands).

Table 1.

Properties of selected Ca²⁺-binding proteins (adapted from Schwaller 2009).

	α PV	β PV (OM)	CB-D9k	CB-D28k	CR
Ca2+-binding sites (functional)	3 (2)	3 (2)	2 (2)	6 (4)	6 (5)
Ca2+-specific/ mixed Ca2+/Mg2+ sites	0/2	1/1	2/0	4/0a	5/0
KD,Ca (nM)	4–9b	Mixed: 42–45d	KD1 ≈ 200–500e KD2 ≈ 60–300	high aff. (h)f KD1 ≈ 180–240	KD(T) 28 µMg KD(R) 68 KD(app) 1.4 µM
		Ca2+-specific: 590–780d		medium aff. (m) KD2 ≈ 410–510	EF5: 36 µM
KD,Mg	≈ 30 µMb	160–250 µMd		714 µMa	4.5 mMh
KD,Ca(app) (nM) at [Mg2+] of 0.5–1 mM	150–250c	230–310c
kon,Ca (µM−1s−1)	6b		up to 1000i	h sites ≈ 12f	T sites: 1.8g
				m sites ≈ 82	R sites: 310
kon,Mg (µM−1s−1)	0.1–1c				site EF5: 7.3
cooperativity	nok	? yesl	yesm	yes	yes
				nH ≈ 1.1–1.2a	nH ≈ 1.3–1.9g
DCabuffer (µm2s−1)	37–43n			>100o	≈ 25p
	∼12			≈ 25

^aAlthough considered as Ca²⁺-specific, at physiological [Mg²⁺]_i, the apparent Ca²⁺ affinity is approximately two-fold lower (Berggard et al. 2002a).

^b(Eberhard and Erne 1994).

^cK_D,Ca and k_on for PV are [Mg²⁺] dependent; calculated values are estimates at [Mg²⁺]_i 0.6–0.9 mM.

^d(Cox et al. 1990).

^e(Linse et al. 1991).

^fCB-D28k has high-affinity (h) and medium-affinity (m) sites, the stochiometry h/m is either 2/2 or 3/1 (Nagerl et al. 2000).

^gFor details, see text and Faas et al. (2007).

^h(Stevens and Rogers 1997).

ⁱThe value represents the diffusion limit, assuming a maximal Ca²⁺ diffusion rate of ≈ 200 µm²s⁻¹ (Martin et al. 1990).

^kPV has 2 essentially identical Ca²⁺-binding sites with n_H close to 1.

^l(Yin et al. 2006).

^m(Akke et al. 1991).

ⁿThe diffusion coefficients D_Cabuffer for PV in muscle myoplasm (37 µm²s⁻¹) (Maughan and Godt 1999) and Purkinje cell dendrites of 43 µm²s⁻¹; (Schmidt et al. 2003a); smaller values are measured in PC soma and axons (≈12 µm²s⁻¹); (Schmidt et al. 2007a).

^oCB-D28k’s mobility in PC dendrites is 26 µm²s⁻¹ (Schmidt et al. 2005), clearly slower than in water (Gabso et al. 1997), and also slower than PV in PC dendrites.

^pEstimation based on the similar size of CB-D28k and CR.

IMPORTANT PARAMETERS TO CHARACTERIZE Ca²⁺ BUFFERS

Intracellular Concentration

The difficulty in obtaining reliable values for Ca²⁺-buffer concentrations is linked to the fact that those cells that strongly express Ca²⁺ buffers are frequently only a subset of cells (often with complex morphologies) within composite tissues; for example, in a subset of neurons in the brain, in specific segments of the kidney nephron, etc. The concentration of PV is as high as 1.5 mM in the superfast swimbladder muscle of toadfish (Tikunov and Rome 2009) and approximately 1 mM in mouse fast-twitch muscle, while it is lower in other muscles and highly correlated with the speed of muscle relaxation (Heizmann et al. 1982). Within different neuron subpopulations, PV is on average one order of magnitude lower (50–150 µM): 80 µM in mouse (Schmidt et al. 2003b) and 120 µM in rat (Hackney et al. 2005) Purkinje cells; 150 µM in mouse cerebellar interneurons (Collin et al. 2005), and 100–300 µM in inner and outer hair cells from inner ear (Hackney et al. 2005). The concentration of mammalian β PV called oncomodulin (OM) is particularly high in rat cochlear outer hair cells (2–3 mM) (Hackney et al. 2005). High concentrations of CR (1.2 mM) are also present in tall saccular hair cells of the frog (Edmonds et al. 2000). In other cells, calretinin concentration is much lower: approximately 20 µM and 35 µM in rat inner and outer hair cells, respectively (Hackney et al. 2005), and 30–40 µM in mouse cerebellar granule cells (Gall et al. 2003). The concentration of CB-D28k is 150–360 µM in Purkinje cells (for a review, see Schwaller et al. 2002), and 40–50 µM in mature hippocampal dentate gyrus granule cells, CA3 interneurons, and in CA1 pyramidal cells (Muller et al. 2005). In cells expressing various mobile and immobile Ca²⁺ buffers of unknown identities and Ca²⁺-binding properties, the concept of the Ca²⁺-binding ratio of endogenous buffers has proven to be useful. The ratio of buffer-bound Ca²⁺ changes over free Ca²⁺ changes (κ_S ∼ [Ca²⁺ buffer]/K_D,Ca) serves as a measure to compare the Ca²⁺-buffering capacity of different cell types. According to the single compartment and linear approximation model (Neher 1998), motor neurons with low cytosolic Ca²⁺ buffer expression also have a very low Ca²⁺-buffering capacity (κ_S < 50) (Lips and Keller 1998). Hippocampal principal neurons and PV-expressing interneurons have κ_S values in the order of 60 and 150, respectively (Lee et al. 2000b). The highest Ca²⁺-buffering capacity (κ_S of 900–2000) is seen in cerebellar Purkinje cells expressing high levels of CB-D28k and PV (Fierro and Llano 1996).

Metal-binding Affinities

Structurally, different types of EF-hand domains exist: the canonical EF-hand domain, comprising a Ca²⁺-binding loop of 12 amino acids, in which Ca²⁺ ions are mostly coordinated via oxygen atoms of carboxyl side chain groups, and several non-canonical ones (Gifford et al. 2007). Typical for proteins of the S100 family, including CB-D9k (Marenholz et al. 2004), is the presence of a noncanonical loop, termed pseudo (Ψ) EF-hand, consisting of a loop of 14 amino acids with an inside-out conformation compared to the canonical loop; that is, Ca²⁺-coordination is preferentially provided by backbone carbonyl groups (Nelson et al. 2002). Functionally, two types of EF-hand Ca²⁺-binding sites are discernable due to their different selectivity and affinity for Ca²⁺ and Mg²⁺ ions (Celio et al. 1996). The Ca²⁺-specific sites display affinities for Ca²⁺ (K_Ca) in the order of 10⁻³–10⁻⁷ M and significantly lower ones for Mg²⁺ (K_Mg = 10⁻¹–10⁻² M). The proteins CB-D28k and CR have 4 and 5 functional Ca²⁺-binding sites of this type, respectively. The mixed Ca²⁺/Mg²⁺ sites bind Ca²⁺ with high and Mg²⁺ with moderate affinity in a competitive manner (dissociation constants: K_Ca = 10⁻⁷–10⁻⁹ M; K_Mg = 10⁻³–10⁻⁵ M) (Table 1). PV is the prototypical example of a protein with two mixed sites. Based on PV’s affinities for Ca²⁺ and Mg²⁺ and with [Ca²⁺]_i (of approximately 50 nM) and [Mg²⁺]_i (0.5–1 mM) inside a cell under basal conditions, PV’s two Ca²⁺/Mg²⁺ sites are, to a large degree, occupied by Mg²⁺. In most proteins, EF-hand domains are paired; two helix-Ca²⁺-binding loop-helix regions are connected by a short stretch of 5–10 amino acid residues and form a functional unit. Aside from providing structural stability of EF-hand domains, due to the close apposition also of the Ca²⁺-binding loops in such a tandem domain (Fig. 1), binding of a Ca²⁺ ion to one loop may allosterically affect the other, both with respect to affinity and binding kinetics (Faas et al. 2007; Nelson et al. 2002). As a result of this building principle, the majority of EF-hand proteins have an even number of Ca²⁺-binding domains (2 in CB-D9k, 6 in CB-D28k and CR); the uneven number (3) for PV (alpha and beta) is, together with the group of penta-EF-hand Ca²⁺-binding proteins (Maki et al. 2002), rather the exception.

Figure 1.

3D-structures of selected EF-hand Ca²⁺ buffers. (A) Consensus sequence of the canonical EF-hand Ca²⁺-binding loop of 12 amino acids. Amino acids X, Y, Z, and –Z provide side-chain oxygen ligands, * provides the backbone carbonyl oxygen, and at –X, a water molecule is hydrogen-bonded to a loop residue. Amino acids most often present at a given position are shown below, and shaded residues are the most conserved ones (Marsden et al. 1990). At positions X and –Z, Asp (D) and Glu (E) are generally present, respectively. The seven oxygen ligands coordinating the Ca²⁺ ion are located at the seven corners of a pentagonal bipyramid, and the Ca²⁺ ion (not shown) is in the center (right). (B) Solution structure of Ca²⁺-bound human α PV; PDB: 1RJV. Both the CD domain (green) and EF domain (yellow/red) bind one Ca²⁺ ion each (green spheres) in canonical Ca²⁺-binding loops of 12 amino acids. The orthogonally oriented helices E and F (gray-shaded) are connected by the Ca²⁺-binding loop. Both Ca²⁺-binding sites in PV are of the Ca²⁺/Mg²⁺ mixed type. The N-terminal AB domain (blue) is necessary for protein stability. (C) NMR solution structure of bovine CB-D9k; PDB: 1B1G. The shown structure takes into account the Ca²⁺ ions and explicit solvent molecules. The N-terminal domain EF1 is a pseudo (Ψ) EF-hand with a larger loop of 14 amino acids, while the second domain (EF2) has a canonical Ca²⁺-binding loop of 12 amino acids. In both loops, the Glu residue at the position –Z with the 2 carboxyl oxygen atoms (red) serves as a bidentate ligand representing two corners of the pentagonal bipyramid. This residue, most often Glu (rarely Asp), is a critical determinant for the Ca²⁺ affinity of the entire loop; Ca²⁺ ions are shown as green spheres. The two Ca²⁺-binding loops are in close proximity and stabilized via short β-type interactions (gray-shaded area). (D) 3D NMR structure of CB-D28k; PDB: 2G9B. CB-D28k has a relatively compact structure comprising three Ca²⁺-binding units, each unit consisting of a pair of EF-hands. Ca²⁺-binding is restricted to the Ca²⁺-binding loop 1 in the N-terminal unit (blue), to both loops 3 and 4 in the middle unit (green), and to loop 5 in the C-terminal pair (yellow/red). EF-hands 2 and 6 are nonfunctional, with respect to Ca²⁺-binding. The Ca²⁺-binding loops flanked by two almost perpendicular alpha-helical regions are numbered from 1 to 6. Images B–D were generated with PDB ProteinWorkshop 1.50 (Moreland et al. 2005).

Metal-binding Kinetics

The majority of Ca²⁺ buffers have dissociation constants (K_D,Ca) in the low micromolar range (Table 1). Thus, in a resting cell (e.g., muscle fibers, neurons), Ca²⁺ buffers are mostly in the Ca²⁺-free form. It is the kinetics of various Ca²⁺ buffers that strongly affect the spatiotemporal aspects of Ca²⁺ signals, in particular, in excitable cells (Schmidt et al. 2007b; Schmidt and Eilers 2009; Schwaller 2009). [Ca²⁺]_i transients last for tens of milliseconds to several hundred milliseconds and, thus, are differently modulated by kinetically distinct Ca²⁺ buffers. Their on-rates for Ca²⁺ binding (k_on) vary by more than two orders of magnitude. Typical fast buffers with Ca²⁺-specific sites, including CB-D9k and troponin C, have k_on rates >10⁸ M⁻¹s⁻¹ comparable to the on-rate of the synthetic buffer BAPTA. At the other end of the kinetic scale, on-rates for the slow-onset buffer PV are ≈3 × 10⁶ M⁻¹s⁻¹ under physiological conditions with [Mg²⁺]_i of ≈0.5–1.0 mM. PV’s slow on-rate is the result of PV’s Mg²⁺/Ca²⁺ binding sites, where the rate for Ca²⁺-binding is determined by the slow Mg²⁺ off-rate (Lee et al. 2000c) (Table 1). Of importance, in a nonphysiological setting, that is, in the absence of Mg²⁺, the on-rate of Ca²⁺ binding to PV is very rapid (1.08 × 10⁸ M⁻¹s⁻¹), almost as fast as the “fast” buffers (Lee et al. 2000c).

Since all endogenous Ca²⁺ buffers have more than one Ca²⁺-binding site, a given protein may have sites with different affinities and kinetics; results for CB-D28k and CR are summarized here. CB-D28k contains two types of binding sites differing in K_D,Ca and k_on: One type is a high-affinity site (K_D,Ca ≈ 200 nM) that binds Ca²⁺ with a k_on comparable to that of EGTA (≈1 × 10⁷ M⁻¹ s⁻¹), and the second type binds Ca²⁺ with intermediate affinity (K_D,Ca ≈ 400–500 nM), but with an approximately 8-fold faster k_on (Table 1). The ratio between high:intermediate affinity sites is either 3:1 or 2:2. The experimental data could be equally well modeled with either stochiometry for the two types of binding sites (Nagerl et al. 2000). For determining CR’s kinetic Ca²⁺-binding properties in vitro, cooperativity was included in the mathematical model (Table 1). CR contains four 4 high-affinity, cooperative binding sites (Schwaller et al. 1997; Stevens and Rogers 1997) organized into two indistinguishable pairs, probably 1 & 2 and 3 & 4 (Faas et al. 2007) and one independent low-affinity Ca²⁺-binding site (EF5) with an intrinsic dissociation constant (K’_D) of ≈0.5 mM (Schwaller et al. 1997). Within cooperative pairs, the two binding sites influence each other in an allosteric manner. In a pair, initially both binding sites are in a T- (tense) state, with a low affinity and slow binding rate. When the first Ca²⁺ ion is bound to a pair, the unoccupied site changes to an R- (relaxed) state, characterized by a high affinity and a fast binding rate. This leads to a gradual increase in the Ca²⁺ association rate (k_on) as [Ca²⁺]_i increases from a concentration in a resting neuron (≈50 nM) to approximately 1–10 µM after opening of Ca²⁺ channels (Schwaller 2009). Therefore, the kinetics of Ca²⁺ buffering of CR powerfully depends on the prevailing Ca²⁺ concentration prior to a perturbation, resulting in non-linear Ca²⁺ buffering by CR (for more details, see Fig. 3 in Schwaller 2009 and Faas et al. 2007).

Mobility and Interaction with Ligands

In an aqueous solution, the mobility defined as the diffusion coefficient (D) of a molecule is approximately proportional to the hydrodynamic radius (i.e., approximately proportional to the relative molecular mass; M_r). For relatively large molecules, such as dextrans and globular proteins, D should be proportional to the inverse cubic root of M_r. In Purkinje cells, PV is freely mobile, but D can vary considerably in different compartments: ≈12 µm²/s in axons, somata, and nuclei (Schmidt et al. 2007a) versus ≈43 µm²/s in dendrites (Schmidt et al. 2003a), most likely as the result of different cytoplasmic properties (e.g., tortuosity; i.e., diffusion in a porous medium). The latter value is very similar to PV’s mobility in frog myoplasm, 43 and 32 µm²/s for transverse and longitudinal diffusion, respectively (Maughan and Godt 1999), but clearly smaller than in an aqueous solution: 140 µm²/s (Feher 1984). As expected from the larger M_r of CB-D28k (≈29 kDa) compared to PV (≈12 kDa), D in Purkinje cell dendrites is smaller: 26 µm²/s (Schmidt et al. 2005). How the presence of a Ca²⁺ buffer affects the spatiotemporal aspects of a cytosolic Ca²⁺ transient not only depends on the mobility of the Ca²⁺ buffer, but also on the mobility of the free Ca²⁺ ions. In the cytosol isolated from Xenopus laevis oocytes, the diffusion coefficient D of inositol 1,4,5-trisphophate (InsP₃) is much bigger than that of Ca²⁺ ions in a solution with [Ca²⁺]_i of 90 nM: 283 µm²/s versus 13 µm²/s, respectively (Allbritton et al. 1992). Even when [Ca²⁺]_i is increased to 1 µM, presumably saturating slowly mobile and immobile buffers, D_Ca only reaches a value of 65 µm²/s. This indicates that the slow diffusion of Ca²⁺ ions in a resting cell ([Ca²⁺]_i ≈ 50 nM) is caused by slowly mobile or immobile buffers “acting like velcro” for Ca²⁺ ions, limiting the effective range of an unbuffered free Ca²⁺ ion to ≈0.1 µm. Thus, the range of Ca²⁺ can be increased by buffered diffusion, that is, mobile Ca²⁺ buffers acting as shuttles transporting Ca²⁺ through the “mesh of immobile buffers” (Schmidt et al. 2007b).

Parvalbumins

Structural Aspects of Parvalbumins

The first prototypical structure of an EF-hand domain was determined in PV (Kretsinger and Nockolds 1973), an atypical EF-hand protein; the protein (M_r ≈ 12 kDa; human gene symbol: PVALB) has an uneven number (3) of EF-hand domains, and the Ca²⁺ -binding sites are Ca²⁺/Mg²⁺ mixed sites (Table 1). While the two C-terminal domains CD and EF are functional metal-binding sites, the N-terminal AB site is necessary for the PV’s stability (Fig. 1) (Cox et al. (1999). The CD and EF domains form a pair consisting of two helix-loop-helix regions linked by a short stretch of 5–10 amino acid residues. In both sites, the Ca²⁺ ion in the center of the loop is coordinated by seven ligands sitting in the corners of a pentagonal bipyramid (Swain et al. 1989). Results based on solution structures of α PV and β PV (Babini et al. 2004) in the Ca²⁺-loaded and the apo (metal-free) form are summarized: (I) PV’s Ca²⁺-loaded EF-hand domains and the linker region connecting the CD and EF domains are rather rigid structures; also the N- and C-termini of PV have a low intrinsic mobility (Baldellon et al. 1998); (II) Differences in the structure of apo- and Ca²⁺-loaded forms of rat PV are small, mostly confined to the loop region. Thus, Ca²⁺ binding does not require major structural rearrangements (Henzl and Tanner 2008); (III) The first two points also hold true for the Ca²⁺/Mg²⁺ (EF) site in rat β PV, while the noncanonical CD site undergoes significant structural alterations, when Ca²⁺ is removed from β PV (Henzl and Tanner 2007). Thus, the global rigidity of α PV favors this molecule to serve as a “simple” Ca²⁺ buffer, while the Ca²⁺-dependent conformational changes in β PV may provide β PV also with a Ca²⁺ sensor function.

Functional Aspects of Parvalbumin and Oncomodulin

Cells with high PV expression levels include a subset of mostly GABA-ergic neurons, fast-twitch muscle fibers, and epithelial cells in the early distal convoluted tubule (DCT1) in nephrons of the kidney. PV’s slow-onset Ca²⁺-binding properties affect Ca²⁺ transients in a particular way: The rate in rise in [Ca²⁺]_i is hardly affected, but the initial rate of [Ca²⁺]_i decay is increased. In the later phase of the decay, the unbinding of Ca²⁺ ions from PV prolongs the late phase of the [Ca²⁺]_i decay. Thus, PV’s hallmark is the conversion of a monoexponential [Ca²⁺]_i decay into a biexponential one (Collin et al. 2005; Lee et al. 2000c). In fast-twitch muscles, this increases muscle relaxation of an electrically induced muscle twitch, while barely affecting the contraction phase. In PV knockout mice (PV^−/−), twitch half-relaxation rates in fast muscles are slower than in the PV-expressing wild-type (WT) muscles (Schwaller et al. 1999). Conversely, the overexpression of PV by injection of Pvalb cDNA into the rat slow-twitch muscle, soleus, significantly increases the speed of relaxation, without affecting the contraction (Muntener et al. 1995). Pvalb gene delivery in rat heart in vivo increases the rate of heart relaxation in normal hearts and in an animal model of slowed cardiac muscle relaxation (Szatkowski et al. 2001). Thus, PV or genetically “tuned” PV variants are discussed as potential tools to enhance cardiac diastolic function (Rodenbaugh et al. 2007; Wang and Metzger 2008). An often-neglected aspect is the role of Ca²⁺ buffers in acting as transient Ca²⁺ sources, prolonging the [Ca²⁺]_i decay. In fast-twitch muscles subjected to long tetanic contractions, PV saturates with Ca²⁺ and consequently slows down relaxation (Raymackers et al. 2000). The slow decay component mediated by Ca²⁺-bound PV also leads to a robust, PV-dependent, delayed transmitter release at cerebellar interneuron–interneuron synapses subsequent to presynaptic bursts of action potentials (Collin et al. 2005).

The PV-mediated, biexponential [Ca²⁺]_i decay is observed in: (I) PV-injected chromaffin cells (Lee et al. 2000c); (II) PV-containing hippocampal interneurons (Lee et al. 2000b); (III) PV-expressing molecular layer interneurons (MLI) in the cerebellum (Collin et al. 2005); (IV) presynaptic terminals of the calyx of Held (Muller et al. 2007); and (V) in Purkinje cell dendrites (Schmidt et al. 2007a). PV’s acceleration of the early phase of [Ca²⁺]_i decay limits or slows down the buildup of residual [Ca²⁺]_i in presynaptic terminals, thus affecting short-term plasticity. The effect is most pronounced at timepoints, when [Ca²⁺]_i decay curves in the presence or absence of PV show the largest differences. PV’s effect on decreasing/preventing paired-pulse facilitation at synapses between MLI and Purkinje cells is most pronounced at ≈33 Hz (Caillard et al. 2000). Also in the presynaptic terminals of the calyx of Held, PV accelerates the decay of spatially averaged [Ca²⁺]_i and paired-pulse facilitation (Muller et al. 2007). In hippocampal PV-interneurons, differences in paired-pulse modulation between WT and PV^−/− mice are apparent only when trains at 33, 50, and 100 Hz are delivered (Vreugdenhil et al. 2003), likely due to differences in components of the Ca²⁺ signaling toolkit and/or more efficient presynaptic Ca²⁺ extrusion mechanisms. The largest relative effect of PV in preventing facilitation is seen at approximately 33 Hz, within the range of gamma frequency (30–80 Hz) oscillations. As a result, the power of kainite-induced gamma oscillations in area CA3 in vitro is approximately 3-fold higher in PV^−/− versus WT tissue. This can be explained by an increased facilitation of GABA release at persistent high frequencies. In accordance with the hypothesis that changes in the inhibitory activity of PV neurons in the neocortex—often critically involved in strong perisomatic inhibition—may be a major mechanism underlying epileptic seizures (Mihaly et al. 1997). PV^−/− mice have a lower threshold for pentylenetetrazole (PTZ)-induced seizures (Schwaller et al. 2004). The subpopulation of PV-immunoreactive (PV-ir) neurons is critically involved in controlling the output of principal neurons (Freund 2003); moreover, PV is not only a marker for these GABA-ergic interneurons, but contributes to controlling the network activity. The absence of PV in the cerebellar circuitry leads to the emergence of 160-Hz oscillations in vivo sustained by synchronous, rhythmic-firing Purkinje cells aligned along the parallel fiber axis (Servais et al. 2005). Also, PV^−/− Purkinje cell-firing properties are different from WT ones: The complex spike duration and the spike pause are decreased, and the simple spike-firing rate is increased. These differences in firing properties, together with the oscillations, are the likely cause for the mild locomotor phenotype, that is, a slight impairment of motor coordination/motor learning (Farre-Castany et al. 2007).

Much less is known about OM’s specific function. OM is present in the organ of Corti (Thalmann et al. 1995), more precisely in cochlear outer hair cells (Sakaguchi et al. 1998) in gerbil, rat, and mouse. An extracellular role for OM in retinal ganglion cell regeneration (Yin et al. 2006) was reported, but see also Hauk et al. (2008) and Schwaller (2009). An up-regulation of OM occurs in PV^−/− mice in a sparse subpopulation of neurons in the thalamus and in the dentate gyrus, as well as in partly varicose axons in the diencephalon (Csillik et al. 2010). The functional significance of ectopic OM expression and the exact identity of neurons expressing OM in PV^−/− mice remain to be shown.

Structural and Functional Aspects of Calbindin-D9k

CB-D9k (human gene symbol: S100G) is the smallest protein with four alpha-helical regions forming an EF-hand pair consisting of a canonical (EF2) and a noncanonical/pseudo (EF1) EF-hand domain, joined by a linker region of 10 amino acids (Fig. 1). The tandem domain is stabilized by a short beta-type interaction between the two Ca²⁺-binding loops (Kordel et al. 1993). The Ca²⁺-binding affinities of individual subdomains are several orders of magnitude lower than for the corresponding sites within the intact protein. Thus, EF-hands organized in tandem domains are the physiological relevant structures (Finn et al. 1992; Nelson et al. 2002). K_D,Ca values are almost identical for both sites (Table 1), and the two Ca²⁺ ions bind with positive cooperativity (Linse et al. 1991). CB-D9k undergoes Ca²⁺-induced conformational changes; they are, however, less pronounced than in the prototypical sensor calmodulin. This, together with no identified binding partners, indicates that CB-D9k most likely functions as a Ca²⁺ buffer, rather than a Ca²⁺ sensor (Skelton et al. 1994).

In a rat kidney, CB-D9k is expressed in the loop of Henle, the distal convoluted tubule, and in intercalated cells of the collecting duct (Bindels et al. 1991a). In a mouse, CB-D9k expression is present in late distal convoluted tubules (Lee et al. 2006) and the connecting tubules. Strong expression of CB-D9k is restricted to the first 2 cm of the duodenum (Huybers et al. 2007). CB-D9k is assumed to be a freely mobile molecule in the cytoplasm of specific epithelial cells of the kidney and duodenum. Its Ca²⁺-binding properties, the regulation by 1,25(OH)₂ vitamin D3 in the intestine, and CB-D9k’s relative electrophoretic mobility led to the name calbindin-D9k (Kallfelz et al. 1967). In other tissues, CB-D9k expression is regulated also in other ways (Choi and Jeung 2008), for example, by estrogen in the uterus (Darwish et al. 1991) or by PTH in mouse primary renal tubular cells (Cao et al. 2002). Suggested functions of CB-D9k include a role in the regulation of Ca²⁺ transport processes across epithelial cells (Bindels et al. 1991b), but also as a Ca²⁺ buffer/shuttle optimally tuned for transcellular Ca²⁺ transport (Choi and Jeung 2008).

Calbindin-D28k

Structural and General Aspects of CB-D28k

CB-D28k (M_r ≈ 29 kDa; human gene symbol: CALB1) has six EF-hand domains, four of which bind Ca²⁺ with medium/high affinity (Cheung et al. 1993). EF-hand 2 is nonfunctional and under physiological conditions EF6 most likely is as well. The four medium/high affinity sites (Nagerl et al. 2000) are considered Ca²⁺-specific, albeit low affinity Mg²⁺ binding (K_D,Mg ≈ 700 µM) to the same sites at physiological [Mg²⁺]_i decreases the apparent Ca²⁺ affinity approximately two-fold; additionally, Mg²⁺ binding increases the cooperativity of Ca²⁺ binding (Berggard et al. 2002a). The NMR solution structure of Ca²⁺-bound rat CB-D28k reveals that it consists of a single, almost globular (ellipsoid) fold with six distinguishable EF-hand domains (Kojetin et al. 2006) (Fig. 1). The on-rates of CB-D28k’s fast binding sites (k_on ≈ 8. x 10⁷ M⁻¹s⁻¹) are fast enough to affect the early rising phase of Ca²⁺ transients; the peak amplitude is significantly decreased in WT Purkinje cells, when compared to Purkinje cells from CB-D28k^−/− mice (Airaksinen et al. 1997). The fact that the time to peak is not significantly different in CB-D28k^−/− and WT Purkinje cells indicates that the initial rise in [Ca²⁺]_i is principally governed by the properties (density, kinetics) of the Ca²⁺ channels. In Purkinje cell dendrites, CB-D28k acts as a fast Ca²⁺ buffer for the first approximately 100 ms, reducing the peak [Ca²⁺]_i amplitude to about one half, while later on prolonging the decay by acting as a Ca²⁺ source (Schmidt et al. 2003b).

Although principally considered as a freely mobile protein, CB-D28k binds to several identified target proteins including Ran-binding protein (RanBP) M (Lutz et al. 2003); caspase-3 (Bellido et al. 2000); 3′,5′-cyclic nucleotide phosphodiesterase (Reisner et al. 1992); plasma membrane ATPase (Morgan et al. 1986); L-type Ca²⁺ channel α subunit (CANAC1C) (Christakos et al. 2007); myo-inositol monophosphatase (IMPase) (Berggard et al. 2002b); and in the kidney to TRPV5 (Lambers et al. 2006). In most cases, binding studies were performed in vitro; the physiological implications of these interactions are not clear yet. In dendrites and spines of Purkinje cells, approximately 20% of CB-D28k molecules are temporarily, that is, for several seconds, immobile by their binding to IMPase, a key enzyme of the InsP₃-signaling cascade, and the fraction of immobilized CB-D28k increases by climbing fiber stimulation (Schmidt et al. 2005). In summary, the above findings, together with CB-D28k’s Ca²⁺-dependent conformational changes, indicate additional Ca²⁺ sensor functions (Berggard et al. 2002a).

Functional Aspects of CB-D28k

Reported functions for CB-D28k include a role in Ca²⁺ resorption in the kidney (Boros et al. 2009) and modulation of insulin production and secretion in pancreatic beta cells (Reddy et al. 1997; Sooy et al. 1999). Data on a putative neuroprotective role against excitotoxicity have not yet resulted in a consistent picture (for a review, see Schwaller et al. 2002 and Schwaller 2009). Results obtained in CB-D28k^−/− mice are summarized. At first glance, these mice show no phenotype related to development, the general morphology of the nervous system, the visual (Wassle et al. 1998) and auditory (Airaksinen et al. 2000) systems, or behavior under standard housing conditions (Airaksinen et al. 1997). CB-D28k^−/− mice show a mild—however more severe than PV^−/− mice—impairment in motor coordination/motor learning (Airaksinen et al. 1997; Farre-Castany et al. 2007), likely resulting from the 160-Hz oscillations in the cerebellum (Cheron et al. 2004; Servais et al. 2005). The motor coordination phenotype is due to CB-D28k’s absence in Purkinje cells, since this phenotype and alterations in Purkinje cell physiology also occur in mice with Purkinje cell-specific Calb1 ablation (Barski et al. 2003). At the cellular level, short-term plasticity between cortical multipolar bursting cells and pyramidal cells, or at the mossy fiber-CA3 pyramidal cell synapse in the hippocampus, is affected by CB-D28k (Blatow et al. 2003). The rapid saturation of presynaptic CB-D28k transiently decreases the Ca²⁺ buffering capacity, leading to enhanced facilitation (Blatow et al. 2003) by a mechanism called “facilitation by buffer saturation” (Maeda et al. 1999; Neher 1998). Ca²⁺ buffers, such as CB-D28k, also affect Ca²⁺-dependent inactivation (CDI) of voltage-dependent Ca²⁺ currents (I_Ca). In dentate gyrus granule cells with low or absent CB-D28k expression resulting from Ammon’s horn sclerosis in humans (AHS) (Nagerl and Mody 1998) or in mice with Calb1 gene ablation (Klapstein et al. 1998), CDI is increased, compared to CB-D28k-expressing granule cells, thereby decreasing the total Ca²⁺ load. Increased CDI in the absence of CB-D28k may be viewed as a protective/homeostatic mechanism to limit Ca²⁺ influx in order to augment the resistance against excitotoxicity and to protect the surviving neurons.

Structural and Functional Aspects of Calretinin

Human calretinin (M_r ≈ 31 kDa; gene symbol: CALB2) consists of 271 amino acids and has 6 EF-hand domains, five of which are able to bind Ca²⁺ ions (Schwaller et al. 1997; Stevens and Rogers 1997). Structural data (NMR) is available only for the N-terminal 100 amino acids of rat CR comprising EF-hand domains 1 and 2 (Palczewska et al. 2001). As in CB-D28k, the two domains form a relatively tight structure. CR’s Ca²⁺-dependent conformational changes, together with results from other in vitro studies, suggest that CR also may have Ca²⁺-sensor functions (Billing-Marczak and Kuznicki 1999).

A role for CR in neuroprotection against glutamate toxicity was postulated, but evidence is most often indirect or obtained in model systems in vitro (D’Orlando et al. 2001; Lukas and Jones 1994; Pike and Cotman 1995). For more details, see Schwaller 2009. CR^−/− mice show impaired long-term potentiation (LTP) in the hippocampus (Gurden et al. 1998; Schurmans et al. 1997). While the effect on hippocampal LTP is indirect, the uniform expression of CR in cerebellar granule cells, together with the stereotypic cerebellar organization, has allowed for a detailed investigation of CR in vivo. In CR^−/− granule cells, the excitability is increased, they show faster action potentials, and, under conditions generating repetitive spike discharges, show enhanced increases in frequency with injected currents (Gall et al. 2003). This leads to altered Ca²⁺ homeostasis in Purkinje cells. The firing properties of Purkinje cells are altered in alert CR^−/− mice: The simple spike-firing rate increased the complex spike duration and the spike pause is shorter (Schiffmann et al. 1999). As in PV^−/− and CB-D28k^−/− mice, alert CR^−/− mice show 160 Hz oscillations (Cheron et al. 2004) that appear phenotypically as an impairment of motor coordination. In alert “rescue” mice, where CR in CR^−/− mice is selectively re-expressed in granule cells, granule cell excitability, as well as Purkinje cell firing, resembles that in WT mice. As a consequence, neither 160-Hz oscillations, nor motor coordination impairment, are detected in the rescue mice (Bearzatto et al. 2006).

The similarity of the oscillations and motor coordination deficits in mice deficient for either one of the three CaBPs points toward an effect at the cerebellar network level. In all three knockout strains, the oscillations are temporarily reduced by blocking of: (I) gap junctions between interneurons; (II) N-methyl-D-aspartate receptors; or (III) GABA_A receptors. This indicates that oscillations emerge via a mechanism that synchronizes assemblies of Purkinje cells (mediated by parallel fiber excitation) and the network of chemically-coupled MLI. In addition, recurrent Purkinje cell collaterals (Orduz and Llano 2007) may be implicated in these oscillations (de Solages et al. 2008).

COMPARISON OF THE PHYSIOLOGICAL EFFECTS BROUGHT ABOUT BY CYTOSOLIC Ca²⁺ BUFFERS AND IMMOBILE “Ca²⁺ BUFFERS/STORES”, IN PARTICULAR BY MITOCHONDRIA

Cytosolic Ca²⁺ buffers may be viewed as transitory Ca²⁺ sinks/stores, and together with the plasmalemmal extrusion systems, Ca²⁺ uptake into ER compartments and mitochondria serve as a cell’s Ca²⁺ “off mechanisms”; Ca²⁺-loaded buffers, together with release from Ca²⁺-filled organelles (ER, mitochondria), subserve as the intracellular “on mechanisms” (Berridge et al. 2003). Here, I briefly put side-by-side the role of mitochondria in Ca²⁺ buffering/sequestration with the role of mobile cytosolic Ca²⁺ buffers. The comparison is primarily focused on excitable cells, mostly neurons (for a more detailed role on mitochondria and Ca²⁺ signaling, see Rimessi et al. 2008 and Szabadkai and Duchen 2008). In the large glutamatergic presynaptic terminals of the calyx of Held, mitochondria contribute to increase the rate in [Ca²⁺]_i decay, when peak [Ca²⁺]_i is >2.5 µM (Kim et al. 2005). This mitochondria-mediated increase in [Ca²⁺]_i decay closely resembles the action of the “slow buffer” PV in the same terminals (Muller et al. 2007). At the physiological level, this delayed buffering by PV and mitochondria affects plasticity of synaptic transmission. More importantly, it has an effect on both short-term facilitation and short-term depression. Blocking mitochondrial Ca²⁺ uptake in the calyx of Held slows down the recovery from synaptic depression (Billups and Forsythe 2002), an effect that can be reverted by the addition of 1 mM EGTA. Slow release of mitochondrial Ca²⁺, but not from ER stores, leads to the post-tetanic potentiation (PTP) in motor axons contacting the opener muscle of the crayfish Procambarus clarkii leg (Tang and Zucker 1997). In analogy, at synapses between molecular layer interneurons (MLI), release of Ca²⁺, likely from Ca²⁺-bound PV, increases delayed transmitter release after an AP train (Collin et al. 2005). Both PV and mitochondria hardly affect basal synaptic transmission and show similar effects with respect to short-term modulation. The increased removal of intracellular Ca²⁺ by PV prevents the buildup of residual [Ca²⁺] and thus reduces paired-pulse facilitation at the calyx of Held and in MLI axon terminals; these findings were deduced by comparing PV^−/− and WT mice (Collin et al. 2005; Muller et al. 2007). The reduced density of mitochondria in presynaptic axon terminals of synaptophilin knockout mice (snph^−/−) affects short-term facilitation in cultured snph^−/− hippocampal neurons. While basal synaptic transmission evidenced by single EPSCs and miniature AMPA currents is unaltered, facilitation is increased (Kang et al. 2008). This effect closely resembles the one observed at hippocampal interneuron/CA1 pyramidal neuron synapses in PV^−/− mice, where facilitation of IPSCs is augmented at stimulation frequencies >33 Hz (Vreugdenhil et al. 2003).

Both mitochondria and cytosolic Ca²⁺ buffers have an effect on the spreading of intracellular Ca²⁺ waves. While in Xenopus laevis oocytes, energized mitochondria promote the propagation of Ca²⁺ waves (Boitier et al. 1999), mitochondria in astrocytes limit the rate and extent of Ca²⁺ wave propagation (Jouaville et al. 1995). Also, Ca²⁺ buffers affect Ca²⁺ waves. The fast buffer CR promotes the spreading of InsP₃-evoked Ca²⁺ signals in oocytes, while the slow buffer PV shortens the duration of these Ca²⁺ signals and restricts the global responses to discrete localized events (puffs) (Dargan et al. 2004). In summary: (I) Both mitochondria and cytosolic Ca²⁺ buffers participate in the shaping of Ca²⁺ signals in presynaptic terminals and consequently have an effect on short-term modulation of synaptic plasticity, that is, facilitation, potentiation, and depression; (II) They also affect the spreading of Ca²⁺ waves; the effect depends on the cell type, on the kinetic properties of the cytosolic Ca²⁺ buffers, and also on ER luminal regulatory mechanisms involving the luminal Ca²⁺-binding protein calreticulin (Camacho and Lechleiter 1995); (III) In the systems investigated so far, PV and mitochondria mostly behave as slow-onset buffers, rarely affecting the maximal amplitude of Ca²⁺ signals, but increasing the rate of [Ca²⁺]_i decay. In presynaptic terminals, the time window most strongly affected by the Ca²⁺ buffering action of PV and mitochondria is ≈10–200 ms after peak [Ca²⁺]_I; (IV) The physiological effect of Ca²⁺ buffering/sequestering by mitochondria and Ca²⁺ buffers is dependent on the cell type, morphology of involved compartments (e.g., presynaptic terminal, soma) and, importantly, the contribution of all other components from the Ca²⁺ signaling toolkit; (V) Evidently, PV and mitochondria cannot completely replace one another with respect to Ca²⁺ buffering. They are still different with respect to several parameters: mobility, Ca²⁺ storing capacity, effects of Ca²⁺ binding/uptake on metabolism, Mg²⁺ effects, kinetics of Ca²⁺ binding/release, kinetics of synthesis/degradation, etc. The finding that mitochondria volume and PV expression levels are inversely correlated in several systems is discussed later in this article.

THE Ca²⁺ HOMEOSTASOME

How can a simple change in [Ca²⁺]_i observed in, for example, muscle contraction, neurotransmission, or cell cycle regulation be used by cells to elicit the correct downstream events, as diverse as membrane fusion of neurosecretory vesicles with the plasma membrane or activation/repression of genes? The obvious parameters are the amplitude of the Ca²⁺ signal and the duration or the frequency at which these signals are generated. Subtler regulations comprise the cell morphology, where Ca²⁺ signals are restricted to certain regions: dendrites, soma, or axon terminals of nerve cells. Finally, molecules implicated in Ca²⁺ signaling may be spatially restricted; for example, Ca²⁺ channel subunits in active zones (Bucurenciu et al. 2008). To achieve the necessary precision of Ca²⁺ signals, cells require an accurately tunable system for regulating [Ca²⁺]_i. Opening of plasma membrane Ca²⁺ channels or Ca²⁺ release from internal stores results in an initial increase in [Ca²⁺]_i. The shape of the Ca²⁺ signal in the cytosol, both with respect to space and time, is then modulated by immobile and, if present, by mobile Ca²⁺ buffers. Finally, extrusion systems such as plasma membrane Ca²⁺ pumps (PMCA), the Na⁺/Ca²⁺ exchanger (NCX), and organellar uptake by the ER and/or mitochondria restore the initial situation with respect to [Ca²⁺]_i. All of the above components are part of a cell’s “Ca²⁺-signaling toolkit” that is able to regulate the expression of its own components necessary for accurate and cell-specific Ca²⁺ signaling (Berridge et al. 2003; Schwaller 2009). One of the gene regulators is Ca²⁺ itself, and effects are mediated by Ca²⁺/calmodulin-dependent kinases (CaMK) and Ca²⁺-regulated phosphatases (e.g., calcineurin). As an example, long-term survival of cultured cerebellar granule cells is dependent on accurate Ca²⁺ signals necessitating temporal changes in the transcription of Ca²⁺-signaling toolkit components: IP₃R and PMCAs 2 and 3 are up-regulated, while a PMCA4 splice variant and plasma membrane NCX2 are down-regulated in a calcineurin-dependent manner (Carafoli et al. 1999). Such adaptative/homeostatic mechanisms are also induced if Ca²⁺-signaling toolkit components are functionally compromised (e.g., in genetic diseases) or purposely eliminated in knockout mice. The network of molecules implicated in Ca²⁺ signaling, homeostasis, and its own regulation is termed the Ca²⁺ homeostasome (Schwaller 2007, 2009).

Results on the modulation of the Ca²⁺ homeostasome brought about by altered expression of CaBPs (e.g., in transgenic mice) are summarized. The most surprising finding is that in essentially all CaBP knockout mice and in a given cell type, the deleted CaBP is not compensated by up-regulating expression of one of the more than 240 other EF-hand family members. That is, in the subset of identified “PV-ergic” neurons (e.g., Purkinje cells, stellate, and basket cells in the cerebellum), none of the other Ca²⁺ buffers (CB-D28k, CB-D9k, or CR) are expressed in the above-mentioned neuron subtypes of PV^−/− mice (Schwaller et al. 2004). The same holds true for CR-immunoreactive or CB-D28k-immunoreactive neurons in the respective knockout strains, CB-D28k^−/− and CR^−/− (Airaksinen et al. 1997; Schiffmann et al. 1999). The most notable exception is the up-regulation of CB-D9k in epithelial kidney cells in CB-D28k^−/− mice (Zheng et al. 2004). Two plausible explanations for the absence of compensation/homeostatic mechanisms at the level of other Ca²⁺ buffers are presented: (I) Neurons once committed to express a certain Ca²⁺ buffer have permanently inactivated/repressed the promoter for other Ca²⁺ buffers; (II) The specific properties (affinities, kinetics, cooperativity, mobility) of any other Ca²⁺ buffer would not be adequate to restore “normal” Ca²⁺ signaling (Schwaller 2009). Thus, if not at the level of other Ca²⁺ buffers, how do cells cope with the absence of a particular Ca²⁺ buffer? Also, does the absence or impairment of other components of the Ca²⁺-signaling toolkit affect the expression of Ca²⁺ buffers?

Purkinje cells are characterized by extensive Ca²⁺ signaling and high expression levels of PV and CB-D28k and, thus, are well-suited to address these questions. The two Ca²⁺ buffers are present in the soma, axon, dendrites, and spines, indicating that they are principally mobile proteins. While PV is freely mobile in all compartments (Schmidt et al. 2007a), a fraction of CB-D28k molecules is immobilized in dendrites and spines by its binding to IMPase (Schmidt et al. 2005). The most striking alterations in the absence of PV are the morphological changes observed in the soma. The volume of mitochondria, Ca²⁺ sequestrating organelles that also serve as transient Ca²⁺ stores (Billups and Forsythe 2002; Murchison and Griffith 2000), is increased by about 40% selectively in a narrow compartment underneath the plasma membrane (Chen et al. 2006). Concomitantly, the subplasmalemmal smooth ER compartment is decreased (Fig. 2). These changes in the soma don’t occur in CB-D28k^−/− Purkinje cells. In the latter, subtle changes in the spine morphology are evident: Spines are longer and spine head volume is increased (Vecellio et al. 2000). In spiny pyramidal cells, spine heads are considered as separate biochemical compartments with negligible Ca²⁺ diffusion via the spine neck (Sabatini et al. 2002). However, modeling studies in Purkinje cell spines have revealed that Ca²⁺ buffers are not only involved in modulating the shape of Ca²⁺ transients within the spines, but together with the spine neck geometry also define the amount of Ca²⁺ ions that may reach the parental dendrite and lead to activation of Ca²⁺-/CaM-dependent signaling cascades (Schmidt and Eilers 2009). Of the two principal Purkinje cell Ca²⁺ buffers, mostly CB-D28k is involved in spino-dendritic coupling by buffered Ca²⁺ diffusion, while the contribution of PV is minute. This is a likely explanation for the unaltered PV^−/− Purkinje cell spine morphology. The absence of CB-D28k and PV not only affects Purkinje cell morphology, but also components directly involved in Ca²⁺ signaling. Ca_v2.1 (P/Q type) channels are the major voltage-operated Ca²⁺ channels of mature Purkinje cells (>90% of the whole-cell voltage-gated Ca²⁺ current) and regulate Ca²⁺ signaling and excitability of these cells. These channels are regulated by Ca²⁺-dependent feedback mechanisms consisting of both Ca²⁺-dependent facilitation (CDF) and inactivation (CDI). While the former process is essentially mediated by the Ca²⁺ sensor calmodulin (CaM) (Lee et al. 2000a), CDI is modulated by synthetic Ca²⁺ buffers (EGTA, BAPTA) and by the Ca²⁺ buffers PV and CB-D28k in vitro (Kreiner and Lee 2005). Of note, PV and CB-D28k affect Ca_v2.1 channel function differently than the synthetic buffers EGTA and BAPTA, often presumed to serve as close substitutes for endogenous Ca²⁺ buffers. CDI of Ca_v2.1 channel is assumed to depend on intracellular Ca²⁺ microdomains around Ca²⁺ channels. These microdomains are expected to be differently affected by various Ca²⁺ buffers, which in turn specifically influence the inactivation properties of the Ca²⁺ channels. Contrary to the expectation based on in vitro experiments, CDI in Purkinje cells of double knockout (CB-D28k^−/−PV^−/−) mice is not increased. However, P-type currents recorded in these cells exhibit increased voltage-dependent inactivation as the result of a decreased expression of the auxiliary Ca_vβ2a subunit compared to WT neurons (Kreiner et al. 2010) (Fig. 2). This, together with the observation that spontaneous action potentials are not different in CB-D28k^−/−PV^−/− and WT Purkinje cells, indicates that increased inactivation due to molecular switching of Ca_v2.1 beta subunits may preserve normal activity-dependent Ca²⁺ signals in the absence of PV and CB-D28k.

Figure 2.

Homeostatic/adaptive changes in the soma of Purkinje cells (PC) caused by malfunctioning or elimination of Ca²⁺-signaling toolkit component(s); regulation by the Ca²⁺ homeostasome. (A) A detailed situation is depicted for wild-type mice. Increases in [Ca²⁺]_i (red arrows) result from influx via Ca_V2.1 (P/Q) channels or release from internal stores (light blue) via the IP₃ receptor. IP₃ is generated by the activation of metabotropic glutamate receptors (mGluR). Ca²⁺ removal systems (blue arrows) include PMCA and NCX in the plasma membrane, SERCA pumps, and mitochondria (green). Identified Ca²⁺-signaling toolkit components including organelles, which are up- or down-regulated, are marked in yellow and magenta, respectively. (B) In PV^−/−, PC subplasmalemmal mitochondria are increased, while ER volume directly underneath the plasma membrane is decreased. (C) In addition to the changes observed in PV^−/−, in PC lacking both, PV and CB-D28k the auxiliary Ca_vβ2a subunit of Ca_V2.1 (P/Q), is decreased, leading to increased voltage-dependent inactivation of P-type currents. Model studies indicate an up-regulation of Ca²⁺ extrusion systems, possibly PMCA. (D) In PMCA2^−/− PC, expression of mGluR1 and of IP₃ receptor type 1 (IP3R1), responsible for the Ca²⁺ release from ER stores, is decreased. Also, the cytosolic Ca²⁺ buffering capacity mediated by CB-D28k is decreased. (E) In leaner mice PC that are characterized by strongly attenuated Ca_v2.1 Ca²⁺ channel function, the rapid Ca²⁺ buffering/sequestering capacity is reduced: PV and CB-D28k are down-regulated and (subplasmalemmal) ER is decreased/impaired, leading to reduced Ca²⁺ uptake.

A cross talk between the regulation of a Ca²⁺ channel and Ca²⁺ buffer expression in Purkinje cells is also observed in leaner mice that have a mutation in the pore-forming alpha subunit of Ca_v2.1. The strongly-reduced Ca_v2.1 Ca²⁺ channel function leads to adaptive changes consisting of a diminished rapid Ca²⁺ buffering/sequestering capacity of Purkinje cells (Dove et al. 2000). The Ca²⁺ buffering capacity is less than 50% compared to Purkinje cells from WT mice, due to reduced PV and CB-D28k expression levels (Fig. 2). In addition, reduced Ca²⁺ uptake by the (likely subplasmalemmal) ER further contributes to the reduced buffering ability of leaner mice Purkinje cells (Murchison et al. 2002). Also, impairment of proteins responsible for Ca²⁺ extrusion in Purkinje cells activates the Ca²⁺ homeostasome. In Purkinje cells of mice with a mutation in PMCA2 characterized by a reduced Ca²⁺ extrusion, the rise in [Ca²⁺]_i during high K⁺-induced depolarization is decreased. This is indicative of a Ca_v channel down-regulation likely to regulate [Ca²⁺]_i toward normal homeostasis (Ueno et al. 2002). PMCA2^−/− mice have decreased expression levels of CB-D28k (Hu et al. 2006), metabotropic glutamate receptor 1 (mGluR1), and of InsP₃ receptor type 1 (IP3R1), responsible for the Ca²⁺ release from ER stores (Kurnellas et al. 2007) (Fig. 2). Again, this reduction in mGluR1-mediated [Ca²⁺]_i elevation may be viewed as an adaptive mechanism to cope with the reduced Ca²⁺ extrusion. The authors suggest that “the decrease in the expression of mGluR1 and its downstream effectors and perturbations in the mGluR1-signaling complex in the absence of PMCA2 may cumulatively result in aberrant mGluR signaling in Purkinje neurons, leading to cerebellar deficits in the PMCA2-null mouse.” However, the severely distorted PMCA2^−/− Purkinje cell morphology (smaller cell body, distorted dendritic tree) may also be a likely cause for the ataxic phenotype (Empson et al. 2007). In addition to the identified changes occurring in Purkinje cells of CB-D28k^−/−PV^−/− mice, an up-regulation of Ca²⁺ extrusion/uptake mechanisms was hypothesized, since the decay of dendritic Ca²⁺ signals could be accurately fitted only when applying a two-fold higher Ca²⁺ extrusion rate, compared to the rates sufficient to model the Ca²⁺ transients in PV^−/− and WT Purkinje cells (Schmidt et al. 2003b). Currently, no experimental data is available to account for the increased dendritic [Ca²⁺]_i decay in CB-D28k^−/−PV^−/− Purkinje cells; putative candidates are PMCA isoforms, NCX isoforms, or increased uptake into stores.

What are the evidences that the changes discussed above are not “simple” compensation mechanisms, but may be considered as “truly homeostatic” mechanisms? One argument is the generality of the mechanism and a second one the reciprocality as exemplified for the relationship between PV content and mitochondrial volume. In the absence of PV in PV^−/− mice, an up-regulation of mitochondria occurs not only in PV-ergic Purkinje cells (Chen et al. 2006) or cerebellar stellate and basket cells (B Schwaller, unpubl.), but is also seen in PV^−/− fast-twitch muscle fibers (Chen et al. 2001). The latter are characterized by high PV expression levels in WT mice. Vice versa, in transgenic mice ectopically expressing PV in striatal neurons (Van Den Bosch et al. 2002), a neuron subpopulation normally not expressing PV, the mitochondrial volume is decreased by almost 50% (Maetzler et al. 2004). This reduction accounts for the heightened excitotoxic injury provoked by a local injection of ibotenic acid. A last example of reciprocality: elimination of PV and CB-D28k from Purkinje cells alters Ca_v2.1 channel function (Kreiner et al. 2010), while a reduced Ca²⁺ influx due to a mutation in the Ca_v2.1 channel down-regulates PV and CB-D28k (Dove et al. 2000). The elucidation of the pathways and molecular mechanisms responsible for the regulation of the Ca²⁺ homeostasome remains an exciting topic for future research.