A CELLULAR ATLAS OF CALCINEURIN SIGNALING

Abstract

Intracellular Ca²⁺ signals are temporally controlled and spatially restricted. Signaling occurs adjacent to sites of Ca²⁺ entry and/or release, where Ca²⁺-dependent effectors and their substrates co-localize to form signaling microdomains. Here we review signaling by calcineurin, the Ca²⁺/calmodulin regulated protein phosphatase and target of immunosuppressant drugs, Cyclosporin A and FK506. Although well known for its activation of the adaptive immune response via NFAT dephosphorylation, systematic mapping of human calcineurin substrates and regulators reveals unexpected roles for this versatile phosphatase throughout the cell. We discuss calcineurin function, with an emphasis on where signaling occurs and mechanisms that target calcineurin and its substrates to signaling microdomains, especially binding of cognate short linear peptide motifs (SLiMs). Calcineurin is ubiquitously expressed and regulates events at the plasma membrane, other intracellular membranes, mitochondria, the nuclear pore complex and centrosomes/cilia. Based on our expanding knowledge of localized CN actions, we describe a cellular atlas of Ca²⁺/calcineurin signaling.

1. Calcium signaling is temporally and spatially regulated

Calcium, Ca²⁺, is a universal messenger that regulates cell functions throughout the life of an organism, from fertilization to differentiation and from survival to aging and death. Intracellular Ca²⁺ concentrations are highly regulated and range from ~100 nM at basal, resting conditions to 0.5–1 μM in localized microdomains during signaling [1]. Dynamic changes in Ca²⁺ concentration control essential cell processes such as proliferation, secretion, migration metabolism, gene expression, and even apoptosis. But how do cells translate one message into such diverse responses? The answer lies in the complexity of the message, whose duration, amplitude, oscillatory frequency, and source all encode specific instructions [2]. In brief, cytosolic Ca²⁺ concentrations can increase via 1) entry of external Ca²⁺ through ion channels or exchangers at the plasma membrane (PM), 2) exit of Ca²⁺ from internal sources, i.e. mitochondria, the endo-lysosomal system, Golgi, nuclear envelope and especially the endoplasmic reticulum (ER), through Ca²⁺ release channels [3], or 3) a combination of both mechanisms, which influence each other through both positive and negative feedback loops that finely tune the temporal and spatial properties of the signal [2]. Similarly, an extensive network of Ca²⁺ pumps, exchangers and buffering proteins ensure that cytosolic Ca²⁺ increases are transient and highly localized by rapidly removing Ca²⁺ ions from the cytoplasm and either expelling it or sequestering it within organelles [4, 5]. These elaborate Ca²⁺ dynamics can be locally monitored in real time thanks to a large tool-box of genetically encoded Ca²⁺ indicators that can be targeted to different cellular compartments and tuned to detect a wide range of Ca²⁺ concentrations [6]. Similarly, Ca²⁺-sequestering proteins such as calbindin can be targeted to locally perturb these signals (as shown in [7, 8]).

Because Ca²⁺ ions are spatially restricted, Ca²⁺ effectors as well as their regulators and substrates cluster into signaling microdomains through interactions with Ca²⁺-conducting proteins. Ca²⁺ ions bind directly to a wide variety of proteins that participate in signaling. Some of these modulate the Ca²⁺ signal itself, i.e. pumps, channels and buffering/sequestering proteins, while other Ca²⁺-binding proteins control downstream cellular processes, i.e. kinases and other enzymes, cytoskeletal proteins and transcription factors [1]. Calmodulin (CaM), a highly conserved protein that contains four Ca²⁺-binding motifs called EF hands, is a key intracellular Ca²⁺ receptor that, once bound to Ca²⁺ regulates a host of proteins including Ca²⁺ channels and pumps as well as structural proteins and enzymes [9]. Ca²⁺-bound CaM regulates phosphorylation by activating a variety of protein kinases [10] and a single conserved protein phosphatase, called calcineurin (CN), also known as PP2B or PPP3, which is the focus of this review.

2. Calcineurin, the Ca²⁺/calmodulin-regulated protein phosphatase

Calcineurin was first isolated as the most abundant Ca²⁺/CaM binding protein in the brain [11] and was soon determined to be ubiquitously expressed and to function as a protein phosphatase [12]. Notably, as the only phosphatase regulated by Ca²⁺ and Ca²⁺-bound CaM, CN effects Ca²⁺ signaling in eukaryotes ranging from fungi to mammals but excluding plants. In yeast, CN regulates environmental stress responses and is required for pathogenesis [13]. In mammals, CN has well-established roles in the cardiovascular, nervous, and immune systems. Humans that are heterozygous for loss of function mutations in CN display epilepsy, developmental delay, and autism spectrum disorder, while mutations that lead to increased CN activity cause movement and craniofacial disorders as well as seizures [14, 15]. CN’s most highly studied function is to dephosphorylate the Nuclear Factor of Activated T-cells (NFAT) family of transcription factors, causing them to translocate to the nucleus where they activate gene expression [16]. In T-cells, NFAT activation initiates the adaptive immune response. Thus, the CN inhibitors, FK506 and cyclosporin A (CsA) have been used clinically as immunosuppressants for 40 years, and are routinely prescribed to solid organ transplant recipients to prevent organ rejection and for autoimmune disorders such as atopic dermatitis and psoriasis [17]. However, by reducing CN function in non-immune tissues these drugs also cause wide-ranging adverse effects, including diabetes, hypertension, hypomagnesemia, seizures, pain and increased hair growth [18, 19]. Most of these complications occur via unknown etiologies, highlighting the importance of elucidating tissue and disease-specific roles for CN and discovering fundamental knowledge of the signaling pathways that are regulated by this phosphatase.

2.1. CN structure, isoforms and activation

Calcineurin is an obligate heterodimer that is always composed of one catalytic subunit, CNA, complexed with a dedicated regulatory subunit, CNB. In this regard, CN differs from related PPP phosphatases such as PP1 and PP2A whose arsenal of regulator subunits give rise to a large number of isozymes with distinct properties [20]. In humans, the CNA subunit is encoded by three genes: PPP3CA (CNAα), PPP3CB (gives rise to CNAβ1 and CNAβ2) and PPP3CC (CNAγ), which show tissue-specific expression. Of these α, β2 and γ are canonical isoforms that diverge somewhat at their amino and carboxyl termini, but overall display a highly homologous domain architecture (Figure 1A): CNA contains a catalytic domain with an active site that coordinates two dinuclear (Zn²⁺ and Fe³⁺) metal ions, followed by a CNB-binding helix, an autoinhibitory sequence (AIS) which blocks an essential substrate recognition site for LxVP motifs (see section 2.3) [21], a CaM-binding domain which tightly binds one molecule of Ca²⁺-bound CaM (in low picomolar range), and a C-terminal autoinhibitory domain (AID) which blocks the active site under resting conditions ([Ca²⁺] ≤ 100 nM) [22, 23]. The small regulatory CNB subunit, encoded by PPP3R1 and PPP3R2, contains two pairs of Ca²⁺ binding EF-hands; one with high affinity (K_D in nanomolar range) that constitutively binds Ca²⁺, and the other with a significantly lower Ca²⁺ affinity (10–50 μM) that serves as a calcium sensor [24–26] (Figure 1A). CN activation occurs when intracellular [Ca²⁺] rises, and maximal Ca²⁺ occupancy of CNB induces conformational changes in CNA that promote Ca²⁺/CaM binding, which causes autoinhibition by the AIS and the AID to be disrupted (Figure 1B) [21, 27]. Thus Ca²⁺ binding to CNB and Ca²⁺/CaM binding to CNA both contribute to Ca²⁺-dependent regulation of CN. CN activation is generally reversible, although under some pathophysiological conditions including Alzheimer’s Disease, cleavage of the C-terminal region removes the autoinhibitory domain and produces a constitutively active phosphatase (Figure 1B) [28, 29]. This highly conserved CN activation mechanism was also leveraged to develop genetically encoded, fluorescence resonance energy transfer-based reporters, CaNAR, CaNARi [30, 31] and DuoCaN /UniCaN [32] which measure spatio-temporal dynamics of CN activity in living cells. Such studies revealed that the level of Ca²⁺/CN activity is determined locally by the availability of free Ca²⁺/CaM. In vivo, CN activity is also modulated by endogenous protein regulators including Regulator of Calcineurin (RCAN1–3), Calcineurin Binding Protein-1 (CABIN/CAIN), and Calcineurin and Ras Binding protein (CARABIN), which will not be discussed here but were recently reviewed [33].

Fig. 1. Calcineurin domain structure and activation.

(A) Schematic of canonical (α, β2 and γ) and β1 catalytic subunit (CNA) and regulatory (CNB) subunits of calcineurin. CNB binding helix (BBH); calmodulin (CaM)-binding domains (CBD); autoinhibitory sequence (AIS); autoinhibitory domain (AID); EF-hand Ca²⁺ -binding sites (EF1 and EF2 are low-affinity, EF3 and EF4 are high affinity sites) are represented in the cartoons. (B) Cartoons of activation mechanisms for canonical (α, β2 and γ) and non-canonical calcineurin-β1 (CN- β1) isozymes. CN is inactive under resting conditions. For canonical isozymes, Ca²⁺ binding to CNB and CaM binding to CNA relieves the autoinhibition by the AIS (purple) and AID (blue). Under pathophysiological conditions, cleavage of the regulatory domain can produce a constitutively active enzyme (boxed). CN-β1 is only partially activated by Ca²⁺ and CaM binding due persistent occlusion of a substrate-binding grove by the LAVP sequence (green) found within the non-canonical carboxy terminal. Full activation may involve tethering of the inhibitory tail to membranes via palmitoylation of two cysteines (red).

2.2. Calcineurin-β1 Isozyme

The Calcineurin-β1 (CN-β1) isozyme (i.e. CNAβ1 complexed with CNB) exhibits an alternative mechanism of CN autoinhibition and activation (Figure 1B). This isozyme is conserved in vertebrates and ubiquitously expressed, albeit at low levels [34–36]. Alternative 3’ end processing of the PPP3CB transcript gives rise to CNAβ1, which is identical to canonical CNAβ2 through the CaM-binding domain but contains a unique C-tail (Figure 1A) [35]. This divergent C-terminal domain lacks the AID but contains a sequence motif, LAVP, which autoinhibits CN-β1 by blocking substrate-binding [36]. In vitro, this novel autoinhibitory mechanism confers higher basal and Ca²⁺-dependent activity to CN-β1 compared to canonical CN-β2 and decreases its dependence on CaM [36]. Furthermore, while CN-β2 is fully activated by Ca²⁺ and Ca²⁺/CaM, CN-β1 is still partially inhibited due to LAVP binding to the substrate-binding pocket (Figure 1B) [36]. Recent studies show that the unique CNAβ1 C-tail also contains a pair of conserved cysteine residues that are palmitoylated in vivo [37], and mediate CN-β1 localization to the Golgi, intracellular vesicles, and PM where it encounters a distinct set of interactors and substrates (see section 3.1.3). Palmitoylation is reversible, and palmitates on CNAβ1 turn over rapidly in vivo suggesting that CN-β1-modifying enzymes may be key regulators of this enzyme [37]. Which of the 23 human palmitoyl transferases target CN-β1 is not yet known, but the PM-localized thioesterase, α/β-Hydrolase Domain-Containing protein 17A (ABHD17A), depalmitoylates CN-β1 as well as other PM proteins [37, 38]. CN-β1 does not activate NFAT [39] and mice lacking CNAβ1 (but still expressing CNAβ2) are viable but display metabolic abnormalities and develop cardiac hypertrophy [40]. The Phosphoinositide-4-kinase type IIIα (PI4KIIIα) protein complex was identified as a substrate for CN-β1 (see section 3.1.3), but its full range of functions and substrates remain to be discovered [37].

2.3. Short Linear Motifs (SLiMs) are key for CN signaling

Despite its ubiquitous expression and essential roles beyond the immune system, only a ~100 proteins have been identified as CN substrates since its discovery more than 40 years ago, in part due to the transient nature of CN-substrate interactions [41]. However, it is now clear that CN and other protein phosphatases bind to their substrates via cognate short linear peptide motifs (SLiMs) [20]. This concept explains phosphatase substrate specificity and can be leveraged to map phosphatase signaling pathways systematically [20, 41, 42]. SLiMs are degenerate 3–10 aa-long sequences that lie within intrinsically disordered regions [43]. By forming transient interactions with globular domains that can be tuned and regulated, SLiMs form the backbone of signaling networks [43]. CN specifically recognizes two SLiMs, termed PxIxIT and LxVP, whose positioning with respect to each other, and to sites of dephosphorylation varies between substrates [19, 44]. This mechanism of substrate selection is highly conserved, and the two motifs serve different functions during dephosphorylation: PxIxIT binds to the CNA subunit of active or inactive CN, thus tethering CN to substrates and regulators and determining its distribution within cells [41, 45]. High affinity PxIxIT motifs inhibit CN in vivo by preventing substrate engagement [44, 46, 47] and mutations in CNA that impair PxIxT binding compromise its function [45, 48]. Furthermore, the primary sequence of each PxIxIT motif, as well as post translational modifications determine PxIxIT-CN affinity, which in turn specifies the Ca²⁺-concentration dependence of dephosphorylation in vivo [48–50]. Finally, acquisition and loss of PxIxIT motifs drive evolutionary rewiring of the CN signaling network [51]. In contrast, LxVP docks in a hydrophobic groove at the CNA/B interface that is accessible only after enzyme activation and thus engages only during dephosphorylation [21, 44]. Small molecules and peptides that block LxVP binding, including the immunosuppressants, FK506 and cyclosporin A and autoinhibitory sequences in CNA regulatory domains, inhibit dephosphorylation of protein substrates by CN without engaging the catalytic center of the enzyme [21, 36, 44].

Combining computational (http://slim.icr.ac.uk/motifs/calcineurin/) and experimental identification of PxIxIT and LxVP motifs in humans has identified hundreds of proteins that contain putative CN-binding SLiMs, several of which have been verified as new CN substrates, while others (as indicated in this review) are ripe for investigation [37, 41, 42, 52]. Furthermore, identifying SLiMs in a candidate substrate increases confidence that it interacts directly with CN, and mutating confirmed SLiMs allows the functional consequences of CN-binding to be assessed [37, 41, 42, 52]. These systematic approaches have revealed hitherto unknown functions for CN in Notch signaling, nucleocytoplasmic transport, phosphoinositide signaling and at centrosomes and cilia [37, 41, 42, 52].

Although there is no single phosphorylation motif that is uniquely dephosphorylated by CN, it does show a preference for sequences that contain basic residues upstream of the phosphosite at the −3 position, slightly prefers phosphothreonine over phosphoserine and, unlike other phosphatases, is not negatively impacted by the presence of a proline just after the phosphorylated residue [53–55]. Finally, within substrates, CN seems to preferentially dephosphorylate sites that contain the T/S(p)xxP motif, which is also found in the CN regulator, RCAN1 [54, 56].

2.4. Proximity labelling reveals SLiM-dependent CN interactions throughout the cell

Experimental approaches to identify CN substrates have historically relied on phosphoproteomics and/or physical capture of CN-substrate complexes [37, 42, 51, 57]. However, recently developed proximity-dependent labeling methods allow low affinity protein-protein interactions within a living cell to be detected [57]. Chief among these are proximity-dependent biotinylation methods in which a promiscuous biotin ligase is coupled to a bait protein of choice to identify a variety of labelled prey proteins [58–60]. For CN, proximal proteins that rely on SLiM binding were identified via proximity labelling coupled to mass spectrometry using fusions of both wild-type and SLiM-binding defective CNA to biotin ligases [41]. Of 397 CN-proximal proteins identified in HEK293 cells, approximately half showed SLiM-dependent proximity to CN and were enriched for the presence of CN-binding SLiMs [41]. These studies revealed CN proximity to multiple cellular locations. While some of these are known sites of Ca²⁺ signaling, i.e. the PM, ER, mitochondria, and endo-lysosomal compartments, others, especially the centrosome and nuclear pore complex, were unexpected and suggest new locations for Ca²⁺ signaling [41, 61].

3. CN containing signalosomes occur throughout mammalian cells

Because dephosphorylation occurs only when the phosphatase and its substrate co-localize with a source of Ca²⁺, CN signaling pathways must be mapped both spatially and temporally. The subcellular distribution of CN is determined by SLiM-mediated binding to its substrates, regulators and signaling scaffolds. In particular, these scaffolds include members of the A Kinase Anchoring Protein (AKAP) family, which target protein kinase A (PKA), cAMP-related enzymes and other signaling proteins to discrete subcellular compartments [52]. Several AKAPs coordinate CN and cAMP signaling including AKAP5 (also termed AKAP79/AKAP150 in human and rodent cells, respectively), AKAP6 and AKAP1, which will be discussed [52]. Below we review the evidence for localized CN functions in different intracellular compartments, with the goal of establishing a cellular atlas of CN signaling.

3.1. CN signaling at the plasma membrane

With multiple sites of Ca²⁺ entry, the PM is a major site for Ca²⁺ signaling. Canonical CN isoforms, CN-β2 and CNα, localize to the PM indirectly, via SLiM-mediated binding to PM substrates or to signaling scaffolds, such as AKAP5. In contrast, the CN-β1 isozyme associates directly with membranes via palmitoylation of its C-tail. Some of CN’s many functions and targets at the PM are highlighted below and shown in Figure 2A.

Fig. 2. CN signaling occurs in localized microdomains throughout the cell.

CN substrates and scaffolds in different subcellular locations are shown as discussed in the text. (A) Plasma Membrane. (B) Endocytic compartments. Endosomes, Lysosome and the Golgi Apparatus are shown. (C) Mitochondria (D) Endoplasmic Reticulum and Nucleus. In the nucleus the nuclear pore complex (NPC) is shown. (E) Centrosome and Cilia. All figures were created with BioRender.com.

3.1.1. CN regulates PM proteins that contain CN-binding SLiMS

NHE1:

CN dephosphorylates and inhibits sodium-hydrogen exchanger isofom-1 (NHE1), which couples cellular entry of Na⁺ with proton extrusion to regulate intracellular pH and cell volume [54]. This 12-transmembrane domain-containing protein contains a disordered, phosphorylated cytosolic tail that serves as a signaling scaffold by binding extracellular signal regulated kinase 2 (ERK2) and CN [54, 62]. The NHE1 tail binds to CN via a PxIxIT and an LxVP motif, separated by a dynamic linker that sweeps the CN surface. These interactions target CN to a particular phosphorylated residue (T779), which lies within a favored TxxP dephosphorylation motif. Thus, CN antagonizes ERK2 to inhibit NHE1, but pH_i increases caused by NHE1 also activate CN-NFAT signaling [63] suggesting that these proteins may dynamically regulate each other’s function in vivo.

TRESK:

TWIK-related spinal cord potassium channel (TRESK) is a member of the two-pore K+ channel family that is encoded by the KCNK18 gene and harbors mutations associated with migraines [64]. TRESK is abundantly expressed in the dorsal root and trigeminal ganglia, where it contributes to resting potential and thus negatively regulates neuronal activity associated with neuropathic pain [64, 65]. TRESK also contributes to the development of neuropathic pain by regulating inflammation and apoptosis mediated by MAPK signaling [66]. Ca²⁺/CN activates TRESK, which contains a PxIxIT and an LxVP motif that mediate dephosphorylation of the channel [67, 68]. TRESK activation may decrease activity of the nociceptive transient receptor potential cation channel subfamily V member 1 (TRPV1) Ca²⁺ channel [69], which is also desensitized directly by CN [70].

TRPV1:

This non-selective cation channel responds to capsaicin, heat and other noxious stimuli to transduce signals that produce pain [71]. TRPV1 associates with AKAP5, which promotes sensitization of the channel via phosphorylation by PKA and/or protein kinase C (PKC) [72]. CN dephosphorylates and desensitizes TRPV1, which does not require AKAP5 [70] but may be mediated via two computationally predicted PxIxIT motifs in its C-tail [41]. Modulation of both TRESK and TRPV1 by CN may explain why some patients taking CN inhibitors experience pain [73]. In addition, CN associates with and reduces the current density of the T-type Ca²⁺ channel, CaV3.2 (encoded by CACNA1H), which is implicated in chronic pain, and also regulates pace-making in the heart [74, 75]. Both CACNA1H and closely related CACNA1G (CaV3.1) contain CN-binding motifs that were identified in silico and have not yet been verified experimentally [41].

Other ion channels that contain CN-binding SLiMs:

Systematic discovery of CN-binding SLiMs in the human proteome revealed that the resulting CN network is statistically enriched in ion channels, and especially K⁺ channels [41]. The CN-regulated ATP-sensitive inward rectifier l, Kir6.1 (KCNJ8) [76], voltage gated, Kv7.1 (KCNQ1), and Ca²⁺-regulated small conductance or SK channels (KCNN2, KCNN3) all contain predicted CN-binding motifs. These K⁺ channels regulate heart contractions and give rise to mutations that are associated with arrhythmogenic disorders [77]. Interestingly, mice with cardiac-specific deletions of CNB (and thus lacking CN activity) develop arrythmia [78]. Thus, possible regulation of these channels by CN should be further examined, and could underly the arrythmia observed in CN mutant mice. In sum, in silico prediction of CN-binding SLiMs suggests that additional roles for CN at the PM, and particularly in regulation of ion channel activity, await experimental validation.

3.1.2. AKAP5 organizes CN-regulated signalosomes at the PM

AKAP5 is a member of the AKAP family which, in addition to binding PKA, also anchors calmodulin, PKC, adenylate cyclases, phosphodiesterases and CN [79, 80]. CN interacts with AKAP5 via, a high affinity atypical PxIxIT motif, IAIIIT [81], an LxVP motif [82], and additional, low affinity sequences [83]. The role of CN in many AKAP5-mediated functions has been elucidated through studies of mice that express AKAP5 lacking the PxIxIT site [84]. AKAP5 is broadly expressed, and associates with a wide range of ion channels and receptors to create localized signalosomes that coordinate the antagonistic regulation of substrates by cAMP/PKA and Ca²⁺/CN [79]. Below we highlight only functions for AKAP5-PKA-CN signaling in synaptic plasticity and NFAT activation. Note that CN is highly abundant in neural tissue [85], and its functions in neuronal signaling and neurodegenerative disease have been reviewed elsewhere [86, 87].

Regulation of synaptic plasticity by AKAP5-PKA-CN:

Learning and memory occurs in the hippocampus, largely through phosphorylation dependent modification of synaptic strength, i.e. synaptic plasticity [88]. AKAP5 plays an important role in this process by targeting PKA and CN to neuronal membranes, where they phosphorylate/dephosphorylate the same residue on a critical Ca²⁺ channel, i.e. Ca²⁺ permeable AMPA receptors (CP-AMPAR) [89]. CP-AMPAR conductance as well as its insertion into vs removal from synaptic membranes (via endocytosis) is dictated by its phosphorylation status and AKAP5 targets CN and PKA to either synaptic or endosomal membranes depending on its palmitoylation status [90]. This activity-dependent regulation of CP-AMPAR directly impacts synaptic strength [90], and is disrupted by the β-amyloid (Aβ) oligomers that are produced during Alzheimer’s disease [91].

Regulation of CN-NFAT signaling by AKAP5:

Long lasting forms of memory require coupling of neuronal activity to changes in gene expression, and this is also regulated by AKAP5-CN-PKA signaling [92, 93]. AKAP5 associates with L-type voltage gated Ca²⁺ channels (LTCCs) and mediates their rapid phosphorylation by PKA, which activates Ca²⁺ influx and allows CN to dephosphorylate and inactivate the channel [92, 93]. Thus, AKAP5-targeted actions of PKA and CN on the LTCC determines the shape and magnitude of the Ca²⁺ signal generated by this channel. Furthermore, Ca²⁺ signals from LTCCs induce gene expression by NFAT, which requires dephosphorylation by CN to translocate to the nucleus [16]. AKAP organizes CN-NFAT signaling by directly recruiting both NFAT [94], and CN, which binds to a PxIxIT SLiM in AKAP5 [81]. CN affinity for this PxIxIT must be accurately tuned to allow both concentration of CN near the mouth of the LTCC, and sufficient dissociation so CN can engage with NFAT [81]. This dynamic AKAP-CN-LTCC-NFAT complex explains how Ca²⁺ entry through LTCCs selectively activates NFAT-mediated gene expression. Furthermore, in a neuron, Ca²⁺ signals generated at the dendrites, far from the nucleus, must be propagated through the cell to the neuronal cell body via a series of LTCC-Ca²⁺ spikes to achieve efficient activation of NFAT [95]. Thus, NFAT molecules that enter the nucleus are those localized to LTCCs in the soma, and the level of NFAT activity is determined by the total somatic Ca²⁺ signal [95].

Similarly, in T-cells AKAP5 associates with Orai1, the pore forming subunit of the Ca²⁺ release activated (CRAC) channel, which mediates store-operated calcium entry (SOCE) [96], to couple Ca²⁺ entry through this channel with NFAT activation [89]. This signaling is critical for adaptive immunity, as mutations in Orai1 lead to severe combined immunodeficiency [97]. Orai is activated by the ER Ca²⁺-sensing stromal interaction molecule (STIM) proteins [96]. In response to Ca²⁺ depletion from the ER lumen, caused by prolonged signaling, STIM clusters at ER-PM junctions, where it binds to and activates Orai, allowing both refilling of the Ca²⁺ store and CN-mediated activation of NFAT [96, 98]. The binding site for AKAP5 in the N terminus of Orai1 (specifically the Ora1alpha or long isoform) is absent in other Orai family members (Orai2 and 3) [99]. Loss-of-function mutations in Orai1, disruption of its AKAP5 binding site or knockdown of AKAP5 all block the activation and nuclear translocation of NFAT1, whereas buffering of cytoplasmic Ca²⁺ with the slow-acting Ca²⁺ chelator EGTA has no effect [99]. Overall, the requirement for AKAP5 in CN-NFAT signaling in both excitable and non-excitable cells illustrates how precise localization of these signaling complexes to sites of Ca²⁺ influx allows focal signaling events at the PM to directly impact nuclear processes.

3.1.3. The CN-β1 isozyme regulates the PI4KIIIα complex at the PM

Palmitoylation of the unique CNAβ1 C-tail allows this isozyme to associate directly with the PM, Golgi and intracellular vesicles where it is targeted to membrane-associated substrates [37]. Compared to cytosolic CN-β2, CN-β1, preferentially interacts with membrane-associated proteins, including the PM-localized phosphatidylinositol 4-kinase type IIIα (PI4KIIIα) complex, which phosphorylates phosphatidylinositol at the fourth position of the inositol ring to produce phosphatidylinositol-4 phosphate (PI4P), a rate limiting precursor for the signaling lipid, phosphatidylinositol 4,5-bisphosphate (PI(4,5)P₂). The PI4KIIIα complex is composed of four proteins: PI4KA, TTC7B, FAM126A and EFR3B. PI4KA, the catalytic subunit, combines with regulatory proteins TTC7A/B and FAM126A in the cytosol, and this trimer is recruited to the PM via binding to the anchoring protein, EFR3A/B [100] (Figure 2A). The disordered C-tail of FAM126A regulates PI4KIIIα complex activity [101, 102] and contains a high affinity CN-binding PxIxIT motif that is required for dephosphorylation of FAM126A by CN, while PI4KA contains a second non-canonical PxIxIT motif [37]. CN-β1 is thought to regulate PI4P synthesis by the PI4KIIIα complex during signaling by G-protein coupled receptors (GPCRs). GPCR receptors that couple to Gq-type members of the heterotrimeric family of GTP-binding proteins initiate Ca²⁺ release from the ER by activating phospholipase C-mediated production of inositol 1,4,5,-trisphosphate (IP3). Under these conditions, PI(4,5)P₂ is rapidly depleted from the PM, and activation of PI4P synthesis by the PI4KIIIα complex is required to sustain signaling [103]. Inhibiting either PKC or CN reduce this activation of PI4P synthesis, presumably by modifying PI4KIIIα complex phosphorylation, although further elucidation of this mechanism is required. These studies show that palmitoylation allows CN-β1 to carry out unique regulatory roles at membranes. Furthermore, association of the palmitoylated CNAβ1 C-terminus with the PM may also displace the autoinhibitory LxVP sequence, causing activation of the phosphatase (Figure 1B). Additional substrates and functions for this enzyme await identification as studies of mice lacking this isoform are viable but exhibit cardiac hypertrophy and metabolic changes [40].

3.2. CN signaling at endocytic compartments:

The Golgi apparatus, endosomes, and lysosomes all store and release Ca²⁺ [104, 105] and are emerging as sites for CN signaling. Little is known about roles for CN at the Golgi, although CN-β1 was reported to bind to COG8 and activate signaling by mechanistic target of rapamycin (mTORC2) and AK strain transforming (Akt) kinases [106] (Figure 2B). CN regulates the biogenesis of lysosomes via activation of transcription factor EB (TFEB) [107, 108] and also regulates the formation of endosomes via the clathrin-independent mechanism termed activity-dependent bulk endocytosis in neurons, where it also participates in synaptic vesicle recycling [109–112]. Here, we focus on newly identified roles for CN signaling at endo-lysosomal compartments.

3.2.1. CN signaling at endo-lysosomal Ca²⁺ nanodomains regulates phagocytosis

Macrophages kill pathogens through a process termed phagocytosis, which is initiated when immunoglobulin G-coated pathogens bind to Fc-γ receptors on the surface of macrophages, causing an increase in intracellular [Ca²⁺] [113]. The pathogen is subsequently internalized into a phagosome, which then fuses with lysosomes to create an acidic compartment that digests the pathogen [113]. Surprisingly, phagocytosis is initiated by a Ca²⁺ signal that originates from lysosomes and locally activates CN (Figure 2B) [7] Upon activation of the Fc-γ receptor the Ca²⁺ mobilizing messenger, nicotinic acid adenine dinucleotide phosphate, triggers highly localized Ca²⁺ release from lysosomes via two-pore channels (TPCs), which is efficiently blocked by targeting the Ca²⁺-binding protein, calbindin, to the lysosome but not by disrupting Ca²⁺ signaling more globally with the slow Ca²⁺ chelator, EGTA. This TPC- released Ca²⁺ rapidly activates CN, which in turn dephosphorylates and stimulates the GTPase dynamin-2, allowing its trafficking to the PM where it is required for phagosome scission and endocytosis. Thus, this study elegantly reveals that peri-lysosomal Ca²⁺ signaling nanodomains drive phagocytosis. However, mechanisms that recruit CN and phosphorylated dynamin-2 to this Ca²⁺ signaling nanodomain remain to be identified.

3.2.2. CN regulates directional transport of neurotrophic signaling endosomes

CN on endosomes was recently shown to regulate neurotrophic signaling, which promotes neural connectivity (Figure 2B) [114]. At the pre-synapse of neurons, Brain-derived Neurotrophic Factor binds to and activates its receptor, tyrosine receptor kinase B (TrkB), and ligand-receptor complexes are then packaged into ‘signaling endosomes’ which must travel to the cell soma to deliver survival signals. CN dephosphorylates the huntingtin protein (HTT) on the surface of signaling endosomes to determine their transport direction. When phosphorylated, HTT promotes anterograde endosome transport (i.e. away from the soma) via interaction with the kinesin microtubule motor. However, in response to the Ca²⁺ signal generated downstream of TrkB activation, CN dephosphorylates HTT, causing dynein-mediated retrograde transport of endosomes toward the soma. Thus, CN acts locally, allowing endosomal vesicles to “choose” their transport direction, a mechanism that may be generalizable to other signaling endosomes. How CN associates with the endosomal surface and whether endosomal Ca²⁺ stores contribute to signaling are yet to be investigated. However, this new pathway promises to shed new light on CN signaling in pathologies such as Huntington`s Disease and Rett syndrome, in which retrograde trafficking of signaling endosomes is disrupted [115].

3.3. CN signaling at the mitochondria

In addition to being sites of cellular ATP production, mitochondria also regulate lipid metabolism, apoptosis, and participate in Ca²⁺ signaling [116]. Mitochondria sequester Ca²⁺ via the Mitochondrial Calcium Uniporter and expel Ca²⁺ via specific efflux pathways [117]. These activities alter the amplitude and dynamics of cytosolic Ca²⁺ signals and regulate mitochondrial metabolism [2]. CN signaling at mitochondria promotes apoptosis through substrates dynamin-related protein 1 (Drp1) and Bcl-2 associated death promoter protein (BAD), and sperm motility via Spermatogenesis-Associated 33 protein (SPATA33) (Figure 2C).

3.3.1. Regulation of mitochondrial dynamics

Mitochondria undergo dynamic fusion and fission which regulate their shape and function, with mitochondrial fragmentation being an early step in apoptosis [118]. Drp1, a large GTPase that causes fission by physically constricting and severing the outer mitochondrial membrane, is regulated by phosphorylation [119]. PKA inhibits Drp1, and phospho-mimetic mutations decrease fission and cell sensitivity to apoptosis [120]. CN dephosphorylates Drp1 through an LxVP motif to promote fission which sensitizes cells for apoptosis [121]. Modulating the affinity of this SLiM for CN directly impacts dephosphorylation efficiency, mitochondrial shape and the amount of neuronal death induced by oxygen-glucose deprivation [121]. PKA is targeted to mitochondria by binding the AKAP1 scaffold, which has also been shown to associate with CN [122] and with NCX3, a mitochondrial Na⁺/Ca²⁺ exchanger, to facilitate Ca²⁺ efflux from the mitochondria, and potentially activate CN [123]. Thus, like AKAP5, AKAP1 may organize a mitochondrially localized signalosome that fine-tunes Drp1 function and allows PKA and CN to antagonistically regulate mitochondrial function. This interplay is particularly critical in cardiomyocytes and hippocampal neurons, where preventing Drp1 mediated mitochondrial fission is linked to defects in cardiac metabolism, hypertrophy, neuronal development/morphogenesis and Parkinson’s disease [124].

3.3.2. Activation of apoptosis

CN also regulates the pro-apoptotic protein BAD [104]. Under nonapoptotic conditions, phosphorylated BAD is sequestered in the cytosol via binding to 14-3-3 proteins [125]. Upon apoptotic stimuli, such as starvation, CN dephosphorylates BAD, causing its translocation to the mitochondrial outer membrane where it binds to the antiapoptotic protein B-cell lymphoma-2 (Bcl-2) and B-cell lymphoma-extra large (Bcl-XL) proteins [104]. This displaces Bcl-2-associated X protein (Bax), which initiates mitochondrial membrane permeability and the caspase-3 apoptotic signaling cascade [126]. PKA is a key kinase that inactivates BAD, and as for Drp1, AKAP1 targeting of PKA to the outer mitochondrial membrane modulates BAD phosphorylation and sensitivity to apoptotic signals [124]. Thus, AKAP1 seems to organize a mitochondrial signaling hub that coordinates signaling by PKA and CN to critically regulate mitochondrial function in multiple tissues [127].

3.3.3. Regulation of sperm motility

CN is essential for male fertility and sperm motility [128], where a specialized form of CN composed of PPP3CC (catalytic subunit) and PPP3R2 (regulatory subunit) is expressed [129]. Recently the sperm protein SPATA33 was shown to recruit CN to mitochondria in these cells [130]. SPATA33 was identified first as a testis-enriched PxIxIT-containing protein whose deletion compromises sperm motility [41, 130]. SPATA33 localizes to mitochondria in the sperm midpiece by binding to the mitochondrial Voltage-Dependent Anion Channel 2 (VDAC2), and may function in sperm motility by targeting CN to mitochondria, where CN potentially regulates VDAC2 and/or other mitochondrial proteins [130].

Together, these studies show that CN targeting to mitochondria enables rapid regulation of these organelles. Future studies are required to identify all CN substrates at mitochondria as well as the source and regulation of Ca²⁺ signals that activate this phosphatase.

3.4. CN signaling at the Endoplasmic Reticulum

The ER is a major intracellular Ca²⁺ store that contains approximately 2mM Ca²⁺ [2]. In response to a variety of cellular stimuli, Ca²⁺ is released from the ER through either inositol 1,4,5-trisphosphate receptors (IP3Rs) and/or ryanodine receptors (RyRs), leading to global increases in intracellular Ca²⁺ [ 2, 119]. Thus, Ca²⁺ homeostasis within the ER is not only necessary for ER functions such as protein folding, quality control and lipid synthesis, but also to generate intracellular Ca²⁺ signals, and to activate SOCE, which refills this Ca²⁺ store when it is depleted [131]. Roles for CN at this major source of Ca²⁺ are summarized below and shown in Figure 2D.

3.4.1. CN negatively regulates Ca²⁺ release from IP3R

Under many signaling conditions, IP3 binds to ER-localized IP3Rs to release Ca²⁺ from this organelle [2]. Thus, IP3Rs are signaling hubs that bind to and are regulated by many proteins, including kinases and phosphatases, in part because excessive Ca²⁺ release leads to cell death [132]. PKA phosphorylates IP3R which stimulates Ca²⁺ release, and Protein Phosphatase 1 (PP1) reverses this phosphorylation [133]. Bcl-2 also binds directly to the IP3R and inhibits Ca²⁺ release, which contributes to Bcl-2 antiapoptotic activity, especially in cancers such as Chronic Lymphocytic Leukemia (CLL) in which Bcl-2 is highly expressed [134, 135]. This occurs through a rapid negative feedback pathway that requires CN and Dopamine- and cAMP-regulated phosphoprotein-32 (DARPP-32), which inhibits PP1 when phosphorylated by PKA, but is dephosphorylated and inactivated by CN [122]. Bcl-2 binds directly to DARPP-32 and CN and targets both signaling molecules to the IP3R [136, 137]. Thus, in Bcl-2 overexpressing cells, when PKA phosphorylates the IP3R and stimulates Ca²⁺ release, activation of CN then dephosphorylates DARPP-32 allowing activation of PP1 and IP3R dephosphorylation, which inhibits Ca²⁺ release. Treating primary human CLL cells with a peptide that releases Bcl-2 from the IP3R causes elevated Ca²⁺ and apoptosis, likely because this Bcl2-CN-DARPP-32-dependent negative feedback is disrupted [137, 138].

3.4.2. CN promotes refilling of ER calcium

The efficient refilling of ER Ca²⁺ stores from the cytosol by the sarcoendoplasmic reticulum Ca²⁺ ATPase, SERCA, is regulated by a CN-dependent feedback loop: Calnexin, a Ca²⁺-transmembrane ER protein, interacts with and inhibits the activity of SERCA when the ER lumen is optimally loaded with Ca²⁺ [124]. However, following IP₃-mediated Ca²⁺ release from ER, CN dephosphorylates calnexin on its cytosolic domain, causing its release from SERCA, which activates the pump to refill the Ca²⁺ store [125].

3.4.3. CN regulates the unfolded protein response pathway

In addition to activating SERCA, CN modulates one arm of the Unfolded Protein Response (UPR), a signaling pathway that restores ER function when an imbalance between protein folding and degradation causes the accumulation of misfolded proteins [139]. Under normal conditions, the luminal domain of Protein kinase R-like Endoplasmic Reticulum kinase (PERK), binds to the major ER chaperone, Binding Immunoglobulin Protein (BIP) (Figure 2D) [139]. However, during ER stress, increased binding of unfolded proteins to BIP cause its dissociation from PERK, and activation of the cytoplasmic PERK kinase domain via autophosphorylation [139]. Active PERK then phosphorylates eukaryotic translation initiation factor 2 (eIF2a), causing attenuation of protein translation which allows the ER to recover [139]. In Xenopus oocytes, Ca²⁺ release during ER stress was shown to induce CN binding to and activation of PERK through an undetermined mechanism [140]. Furthermore, knockdown or inhibition of CN attenuated PERK-activated UPR and potentiated apoptotic responses. PERK was also shown to phosphorylate and inhibit CN suggesting mutual regulation of these proteins [140]. UPR activation is especially critical in the nervous system, where accumulation of misfolded proteins and ER stress is the hallmark of many neurodegenerative diseases including Alzheimer’s, Parkinson’s and Huntington’s diseases [141]. Recent studies have also suggested a role for the CN-PERK interaction in regulating Ca²⁺ dynamics during insulin secretion in pancreatic β cells [142], induction of ventricular arrhythmias in diabetic cardiomyopathies [143], as well as the coordination of ER/PM contact site formation through F-actin remodeling [144].

Thus, CN may regulate ER function through complementary mechanisms that promote refilling of the Ca²⁺ store and induce the ER response to stress. How CN associates with the ER is not yet clear, although CN-proximal and SLiM-containing proteins include the ER structuring Reticulon family proteins, Rtn1 and Rtn4, and BIP, which also has roles in the cytoplasm [41, 145].

3.5. Nuclear functions for CN

CN has roles in the nucleus (Figure 2D), and Ca²⁺ dynamics in this compartment are distinct from those in the cytosol [146, 147]. Ca²⁺ can enter the nucleus from the cytosol through nuclear pores, or be released from the nuclear envelope, which is contiguous with the ER and contains both IP3Rs and RyRs [148, 149]. CN contains both nuclear localization (NLS) and nuclear export sequences (NES) that allow it to access nuclear proteins [150] and can also ‘hitchhike’ with partner proteins, such as NFAT, which has been reported to shuttle CN in and out of the nucleus [151]. In addition to NFAT, CN dephosphorylates and activates other transcription factors, including CREB-Regulated Transcription Coactivator 1, TFEB, and Myocyte Enhancer Factor-2 (MEF2), to mediate Ca²⁺ regulated gene expression [108, 152, 153]. How these events are spatially organized is unknown. By analogy with NFAT-CN-AKAP signaling, as yet undiscovered mechanisms may couple activation of these transcription factors to specific Ca²⁺ microdomains.

3.5.1. CN signaling at the Nuclear Envelope

CN localizes both to the nuclear envelope via binding to AKAP6, and to nuclear pore complexes (Figure 2D). AKAP6, also known as mAKAPβ, is expressed primarily in striated muscle including cardiomyocytes, and associates with the nuclear envelope via a cluster of spectrin-like repeats [154]. AKAP6 directly scaffolds multiple proteins including RyR, calcineurin (the CNAβ isoform), PKA and MEF2, and is required for CN-dependent regulation of NFAT and MEF2 that promotes cardiac hypertrophy in response to β–adrenergic signaling [154]. Thus, AKAP6 is proposed to organize a ‘signalosome’ at the nuclear envelope that locally organizes and integrates CN activity with other signaling pathways during cardiac hypertrophy.

3.5.2. CN signaling at the Nuclear Pore Complex

Recently, CN was shown to interact in a SLiM-dependent manner with several proteins in the nuclear pore complex (NPC), including components of the nuclear basket which projects into the nucleoplasm (Nucleoporin-153, Translocated promoter region protein and Nucleoporin-50) (Figure 2D) [41]. In vitro, CN directly dephosphorylates these nuclear basket proteins after phosphorylation by ERK, a kinase that phosphorylates NPC proteins in vivo to inhibit nucleocytoplasmic transport [41, 155, 156]. Incubating cells with CN inhibitors decreased the nuclear accumulation of an NLS-containing transport reporter, and increased phosphorylation of several nuclear pore proteins, also called nucleoporins [41]. These surprising findings suggest that a pool of NPC-localized CN dephosphorylates nucleoporins to increase transport through the pore, and that the NPCs may function as a local Ca²⁺ signaling microdomain. Indeed, Proximity-Ligation assays showed that endogenous CN and NPC proteins interact in a PxIxIT-dependent manner in vivo [41]. While previous work showed that changes in cytosolic [Ca²⁺] alter the structure of key transport nucleoporins [148, 157–159], there are conflicting reports in the literature as to whether Ca²⁺ regulates nucleocytoplasmic transport [160, 161]. Recent evidence that CN binds to and regulates NPC proteins indicates that Ca²⁺ signaling at NPCs should be re-examined using modern tools.

3.6. CN signaling at centrosomes and cilia

Centrosomes are microtubule organizing centers, composed of two centrioles associated with pericentriolar material, that nucleate microtubules in interphase and form the poles of the mitotic spindle [162]. Centrosomes contain Ca²⁺-binding proteins, calmodulin and centrin, and Ca²⁺ signals have been observed at these organelles during mitosis [163]. A pool of CN localizes to centrosomes, where CN binds directly to POC5, one component of a protein scaffold in the centriole lumen that stabilizes this structure [61, 164] (Figure 2E). CN dephosphorylates POC5 in vitro and CN inhibitors alter POC distribution within the centriole lumen [51]. Whether CN regulates POC5 phosphorylation and function in vivo and how it is targeted to centrosomes remains to be determined.

In most non-proliferating cells, centrioles transform into basal bodies which direct formation of primary cilia, i.e. non-motile appendages that are specialized sites for Hedgehog and Ca²⁺ signaling whose disruption causes a group of disorders known as ciliopathies (Figure 2E) [165, 166]. POC5 is also required for ciliogenesis [167]. However, inhibiting CN does not affect cilia number but rather increases cilia length [51], consistent with previous findings that deletion of RCAN2, a negative regulator of CN that also localizes to cilia, shortens cilia [168] (Figure 2E). How CN alters cilia length is not yet known, but may occur via regulation of POC5 or other centrosomal proteins that contain predicted CN-binding SLiMs such as HAUS6, a component of the augmin complex [41, 169]. Interestingly, AKAP5 also localizes to primary cilia [170], and could mediate CN-dependent modulation of ciliary cAMP/PKA signaling, which also impacts cilia length [171]. Thus, unraveling CN actions in cilia and at centrosomes promises to reveal new aspects of CN signaling and possible connections to ciliopathies.

4. Conclusion

Our current state of knowledge shows that CN plays critical roles throughout the body, and that NFAT is just one of an arsenal of CN substrates. New SLiM-directed approaches are revealing CN functions systematically, and techniques like proximity-dependent biotinylation allow these SLiM-dependent interactions, which influence the subcellular targeting of CN, to be identified. Elucidating where and when CN acts is essential for mapping CN and its substrates to specific Ca²⁺ signaling microdomains throughout the cell, and disrupting these local interactions provides exquisite insights into the discrete functions of each localized signaling event. In total, these advances are providing a much more comprehensive view of CN signaling than ever before, which promises to explain adverse effects of CN inhibitors (CysA, FK506) and the clinical presentations associated with mutations in human CN genes. Far from being a one-note song, it is clear that CN directs a symphony of Ca²⁺ dependent signaling events, many of which are yet to be revealed.

Highlights.

• Calcineurin, the Ca²⁺/calmodulin-activated phosphatase acts throughout the cell

• Calcineurin regulates substrates in localized Ca²⁺ signaling microdomains

• Calcineurin acts at PM, endosomes, ER, mitochondria, nuclear pores and centrosomes.

• Calcineurin binds cognate Short Linear Motifs in substrates, scaffolds, regulators

• Discovery of SLiMs systematically maps calcineurin signaling pathways