The unique C terminus of the calcineurin isoform CNAβ1 confers non-canonical regulation of enzyme activity by Ca2+ and calmodulin

Calcineurin, the conserved Ca2+/calmodulin-regulated phosphatase and target of immunosuppressants, plays important roles in the circulatory, nervous, and immune systems. Calcineurin activity strictly depends on Ca2+ and Ca2+-bound calmodulin (Ca2+/CaM) to relieve autoinhibition of the catalytic subunit (CNA) by its C terminus. The C terminus contains two regulatory domains, the autoinhibitory domain (AID) and calmodulin-binding domain (CBD), which block the catalytic center and a conserved substrate-binding groove, respectively. However, this mechanism cannot apply to CNAβ1, an atypical CNA isoform generated by alternative 3′-end processing, whose divergent C terminus shares the CBD common to all isoforms, but lacks the AID. We present the first biochemical characterization of CNAβ1, which is ubiquitously expressed and conserved in vertebrates. We identify a distinct C-terminal autoinhibitory four-residue sequence in CNAβ1, 462LAVP465, which competitively inhibits substrate dephosphorylation. In vitro and cell-based assays revealed that the CNAβ1-containing holoenzyme, CNβ1, is autoinhibited at a single site by either of two inhibitory regions, CBD and LAVP, which block substrate access to the substrate-binding groove. We found that the autoinhibitory segment (AIS), located within the CBD, is progressively removed by Ca2+ and Ca2+/CaM, whereas LAVP remains engaged. This regulatory strategy conferred higher basal and Ca2+-dependent activity to CNβ1, decreasing its dependence on CaM, but also limited maximal enzyme activity through persistence of LAVP-mediated autoinhibiton during Ca2+/CaM stimulation. These regulatory properties may underlie observed differences between the biological activities of CNβ1 and canonical CNβ2. Our insights lay the groundwork for further studies of CNβ1, whose physiological substrates are currently unknown.

Calcineurin, the Ca 2ϩ /calmodulin (CaM) 4 regulated serine/ threonine-protein phosphatase, transduces Ca 2ϩ signals to regulate diverse physiological functions. Calcineurin is a key signaling enzyme in eukaryotic cells, from yeast, where it regulates survival during environmental stress (1), to mammals, where it plays well-established roles in synaptic plasticity (2), heart development (3), and adaptive immunity (4). Signaling by calcineurin through its best characterized substrates, the nuclear factor of activated T-cells (NFAT) family of transcription factors, is critical for T-cell activation, and for this reason the calcineurin inhibitors FK506 and cyclosporin A are in wide clinical use as immunosuppressants (5). Calcineurin, a heterodimer of catalytic (CNA) and regulatory (CNB) subunits, is ubiquitously expressed in humans, but encoded by several genes whose expression is tissue-dependent. In mammals, three genes encode isoforms of CNA (␣, ␤, and ␥), which are highly homologous, differing mostly at their N and C termini, and display a domain architecture and activation mechanism that are conserved throughout eukaryotes (4). An exception is CNA␤1, a transcript variant of the CNA␤ gene that, due to alternative 3Ј processing, lacks the C-terminal autoinhibitory domain (AID) present in all other forms of calcineurin ( Fig. 1A) (6,7). Understanding the distinct mechanisms by which this isoform is regulated may provide novel insights into the full range of calcineurin functions and activities.
Calcineurin utilizes evolutionarily conserved surfaces to recognize substrates containing short linear motifs (SLiMs), conserved peptides found in disordered regions (8 -10). Calcineurin relies on SLiM interaction at docking surfaces distinct from the catalytic center to achieve substrate specificity, rather than recognition of a phosphosite motif (11). One such surface along ␤ sheet ␤14 of CNA binds SLiMs with the consensus sequence PXIXIT (8,12). This docking site anchors substrates as well as scaffolds and regulators. A second surface, formed at the interface between CNA and its obligate regulatory subunit CNB, is available only in the Ca 2ϩ /CaM-activated enzyme and binds SLiMs with the consensus sequence LXVP (13,14). This interaction is required for dephosphorylation, and is blocked by inhibitors, including the immunosuppressantimmunophilin complexes, FK506-FKBP and cyclosporin A-cyclophilin, as well as the viral protein, A238L (15). Furthermore, a sequence within the regulatory domain of CNA termed the autoinhibitory segment (AIS) contributes to autoinhibition by occupying the LXVP-docking groove in the unstimulated enzyme (16). Thus, recognition of LXVP SLiMs is an essential step in substrate interaction with and dephosphorylation by calcineurin.
Calcineurin is activated by a well-documented mechanism that is shared between fungal and animal kingdoms, because of the high degree of sequence and structural conservation among homologs. As shown in Fig. 1A, CNA contains an N-terminal catalytic domain followed by a regulatory domain consisting of a CNB-binding helix (BBH), CaM-binding domain (CBD), which includes the AIS mentioned above, and the C-terminal AID. When cellular Ca 2ϩ levels are low, the inactive enzyme adopts a closed conformation that shields substrate access to the active site. Upon cellular stimulation that gives rise to elevated Ca 2ϩ levels, Ca 2ϩ binding to CNB and to CaM results in a series of conformational changes that release the regulatory domain and reveal both the LXVP substrate-binding pocket and the catalytic center, allowing substrate dephosphorylation (16 -20).
In contrast to the well-studied cytosolic CNA␤2 isoform, the non-canonical CNA␤1 variant lacks the C-terminal AID and instead contains unique sequences that share no homology with AID (Fig. 1, A and C). This alternative C terminus confers distinct protein-protein interactions and contains a membrane localization sequence that targets CNA␤1 to the Golgi apparatus (21,22). CNA␤1 has been reported to perform functions distinct from canonical calcineurin isoforms. In mouse, CNA␤1 is highly expressed in regenerating muscle tissue and stem cells, is cardioprotective rather than pro-hypertrophic, and regulates signaling pathways that are unaffected by CNA␤2 (7,(21)(22)(23). Surprisingly, the biochemical properties of this interesting isozyme, referred to here as CN␤1, have not been studied.
Because it lacks the AID, CN␤1 was originally thought to be constitutively active. However, the failure of CNA␤1 overexpression to yield cardiac hypertrophy, an outcome observed for expression of constitutively active calcineurin, challenged the idea that CN␤1 activity was unregulated (22,24). Additionally, unlike CN␤2, CN␤1 does not dephosphorylate NFAT or impact NFAT-responsive gene expression (22). In fact, although dozens of substrates of calcineurin have been identified, which, if any, are dephosphorylated by CN␤1 is unknown.
Here, we investigate the mechanism of CN␤1 activation. Using in vitro and cell-based assays, we identify a conserved substrate-like LXVP SLiM within the CNA␤1 C terminus, and demonstrate that it behaves as a pseudosubstrate to block substrate recognition. Thus, the atypical domain structure of CNA␤1 gives rise to Ca 2ϩ /CaM-dependent regulation of CN␤1 phosphatase activity through a non-canonical mechanism that confers unique functional characteristics to this isoform.

The CNA␤1 transcript is ubiquitously distributed in human tissues
CNA␤1 and CNA␤2 were the first calcineurin catalytic subunits to be cloned, and were identified as variant isoforms present in human brain mRNA (6). Differential 3Ј processing, carried out in part by the splicing and polyadenylation factor muscle-blind-like 1, produces the shorter CNA␤1 transcript, which lacks exons 13 and 14 found in CNA␤2 (21). In addition to the brain, CNA␤1 transcripts are present in mouse heart, testis, and spleen (7), but neither the tissue distribution of this transcript nor its abundance relative to canonical CNA␣ and CNA␤2 isoforms has been systematically studied. Therefore, we analyzed RNAseq datasets from a panel of adult human tissues (25) and determined the relative levels of CNA␣, CNA␤2, and CNA␤1 mRNA. These analyses revealed that the CNA␤1 transcript was present in all tissues examined at a consistently low level, in contrast to CNA␣ and CNA␤2, whose expression levels were more variable between tissues (Fig. 1B). Thus CNA␤1 protein is likely found throughout the body, and may not predominate, but could contribute distinctly to calcineurin-dependent regulation of cellular processes.

CN␤1 and CN␤2 display different activities in vivo
We sought to compare the biochemical properties of CNA␤1or CNA␤2-containing isozymes in vivo, but were challenged by the lack of known substrates for CN␤1 in human cells. Therefore, we took advantage of the yeast Saccharomyces cerevisiae, where calcineurin promotes survival during exposure to metal ions and other types of environmental stress, in part by dephosphorylating and activating the Crz1 transcription factor (1, 26 -28). Because mechanisms of calcineurin substrate recognition are highly conserved (8), we reasoned that human calcineurin would likely dephosphorylate yeast substrates, and thus provide a method to compare, directly, the activity of CN␤2 and CN␤1 in a simple in vivo system. In the absence of stress, calcineurin-deficient yeast (cna1⌬ cna2⌬ cnb1⌬) transformed with human CNB (hCNB) and either CNA␤2 or an empty vector grew identically ( Fig. 2A). Under stress conditions neither of these strains survived, suggesting failure of CNA␤2 to function in yeast. However, expression of a constitutively active form of human CNA␤, which lacks the CBD and C-terminal regulatory sequences, did complement the growth defects of yeast calcineurin mutants when co-expressed with hCNB (data not shown). Therefore, we considered whether endogenous yeast CaM, which is only 59% identical to human calmodulin (29), might be unable to activate human calcineurin. Indeed, co-expression of CNA␤2 and hCNB with sea urchin CaM (sCaM), which is 93% identical to human CaM, supported yeast growth under stress conditions ( Fig. 2A). Surprisingly, when CNA␤1 was co-expressed with hCNB, cells displayed modest growth under stress in the absence of sCaM ( Fig.  2A). Growth of CNA␤1-expressing cells improved upon co-expression of sCaM, but was less robust than that of CNA␤2-expressing cells, despite the identical CBD sequences and similar expression levels for both isoforms (Fig. 2B). Both isozymes were sensitive to FK506, as yeast expressing either CNA␤1 or CNA␤2

Mechanism of CNA␤1 autoinhibition
failed to grow under stress conditions in the presence of this inhibitor ( Fig. 2A). These results suggest that CNA␤1-and CNA␤2containing isozymes both dephosphorylate yeast substrates, but may differ in their maximal activity and dependence on CaM.
In light of these findings, we examined the distinct C termini of CNA␤1 and CNA␤2 for possible regulatory sequences. The CNA␤1 C terminus shares no homology with the AID, but does contain a 4-residue sequence, 462 LAVP 465 , which is well-conserved across vertebrate CNA␤1 homologs, and contains the core residues found in LXVP motifs, which mediate substrate binding to calcineurin (Fig. 1C) (8,13,14). We hypothesized that this LAVP sequence might block substrate access to the conserved hydrophobic substrate-docking groove of CN␤1, and thus inhibit its activity.

The LAVP sequence from CNA␤1 blocks substrate engagement at the substrate-recognition groove
To compare the regulatory properties of the different CNA␤ C termini directly, we synthesized peptides derived from these sequences or from the calcineurin substrate NFATc1 (Table 1) and added them in trans to CN␤ trunc , a heterodimer composed of hCNB with a truncated catalytic subunit (residues 1-400), which lacks all sequences C-terminal to the BBH, and is thus constitutively active. Notably, both the catalytic center and substrate-binding pocket are exposed in this form. The activity of this recombi-nant, purified human CN␤ trunc was assessed in vitro using two different substrates, para-nitrophenyl phosphate (pNPP) and RII phosphopeptide, which probe the availability of the catalytic center and substrate-docking site, respectively (Fig. 3, A and C).

Mechanism of CNA␤1 autoinhibition
which the Leu, Val, and Pro residues were substituted by alanine, had no effect on activity. Together, these findings suggest that the LAVP sequence from CNA␤1 engages the substrate-docking cleft of calcineurin, and does not block the catalytic site.
We predicted that peptides containing an LAVP sequence would directly compete with RII phosphopeptide at the substrate-docking site. To test this, we measured CN␤ trunc activity toward RII while varying the substrate and inhibitor peptide concentrations. Indeed, we observed that LAVP-containing peptides NFATc1-LAVP wt , ␤1-LAVP wt , and ␤1-DWGT mut were all competitive inhibitors (Fig. 3E). In contrast, the ␤2-AID peptide inhibited RII dephosphorylation noncompetitively, in agreement with previous reports (16,18,19), and suggesting that RII and AID may bind CN simultaneously. CN␤ trunc activity was not inhibited in the presence of up to 100 M NFATc1-LAVP mut or ␤1-LAVP mut peptide. Taken together, results of these analyses suggest that the LAVP sequence in CNA␤1 is a bona fide LXVP motif that binds calcineu- , hCNB, and either sCaM or vector (strains RBY01-06) were spotted on selective SC media in the absence or presence of 10 mM MnCl 2 (stress) and 1 g/ml of FK506 or ethanol/Tween vehicle (veh). Growth was visualized after incubation at 25°C for 6 days and is representative of three independent experiments. B, immunoblot shows protein expression levels from yeast strains used in A, representative of three independent experiments. Graph shows quantification of expression level of CNA, normalized to PGK, from at least three biological replicates. Error bars are Ϯ S.D. (n Ն 3). Means were not significantly different (ns) as assessed by one-way ANOVA.

Table 1 Sequences of peptides used in trans-inhibition assays
Mutated residues are underlined.

Peptide
Sequence

Mechanism of CNA␤1 autoinhibition
rin with similar affinity to the LXVP motif of NFATc1, and can occlude substrate binding. We also note that a peptide encoding the AIS ( 419 FSVL 422 ), an LXVP-like motif within the CBD of CNA (16), failed to affect dephosphorylation in these analyses (supplemental Fig. S1), suggesting that this sequence has very low affinity for calcineurin when isolated from the rest of the CBD.

CN␤1 is autoinhibited in vitro by the LXVP site within its C terminus
Next, we investigated potential autoinhibition of intact CN␤1 and its regulation by Ca 2ϩ and CaM in vitro. To this end we expressed and purified, from yeast, recombinant human CN␤2, CN␤1 wt , and mutants CN␤1 LAVPwt,AISmut , CN␤1 LAVPmut,AISwt , where K D is the apparent dissociation constant and B max is the maximum specific binding. When K D was not determinable, data points were connected by straight lines. E, kinetic analysis of peptide inhibitory mechanism. The concentration of RII phosphopeptide substrate was varied between 100 and 1000 M, and inhibitory peptide concentration was varied from 0 to 100 M as indicated in each plot. Best-fit curves were chosen by mechanistic analysis as described under "Experimental procedures." For ␤2-AID peptide, data were fit to the model for noncompetitive inhibition. For NFATc1-LAVP wt , ␤1-LAVP wt , and ␤1-DWGT mut peptides, data were fit to the competitive inhibition model. For NFATc1-LAVP mut and ␤1-LAVP mut peptides, the Michaelis-Menten model was fit globally to the data. Error bars are Ϯ S.D. (n Ն 3) from at least three independent experiments, each with three technical replicates. All assays were performed in the presence of 0.4 mM CaCl 2 .

Mechanism of CNA␤1 autoinhibition
and CN␤1 LAVPmut,AISmut , in which 462 LAVP 465 and/or 419 FSVL 422 was mutated to AAAA or ASAA, respectively. Initial attempts to purify the CN␤1 heterodimers were unsuccessful due to aggregation, which was alleviated by deleting the highly hydrophobic C-terminal 23 aa of CNA␤1, but leaving the LAVP sequence intact. This small truncation did not perturb enzyme expression and had little effect on function, as assessed under calcineurin-activating conditions in yeast (supplemental Fig. S2). Therefore, these forms of CN␤1 were purified to homogeneity and analyzed below (supplemental Fig. S3). Phosphatase activities of purified CN␤ isozymes were first measured in vitro using pNPP under three conditions: 1) in the presence of excess Ca 2ϩ chelator (EGTA), 2) with Ca 2ϩ , or 3) in the presence of saturating levels of Ca 2ϩ -bound CaM. As shown in Fig. 4, in the absence of Ca 2ϩ /CaM, CN␤2 displayed low basal activity toward pNPP. Addition of Ca 2ϩ increased activity 5-fold, reflecting the elevated rate of hydrolysis that has been previously observed when all four Ca 2ϩ -binding sites of CNB are occupied (37). Finally, the binding of Ca 2ϩ /CaM to CN␤2 increased activity an additional 4.5-fold, because of structural changes that release the AID from the active site upon CaM binding (38,39). CN␤1 wt differed from CN␤2 in exhibiting an increased rate of pNPP hydrolysis in EGTA, and maximal stimulation by Ca 2ϩ in the absence of calmodulin (Fig. 4). This is consistent with the catalytic center of CN␤1 wt being open, i.e. not blocked by an AID that requires Ca 2ϩ /calmodulin for removal. The mutant enzymes, CN␤1 LAVPwt,AISmut , CN␤1 LAVPmut,AISwt , and CN␤1 LAVPmut,AISmut , displayed similar regulatory properties as CN␤1 wt , but had slightly higher maximal activities.
To investigate the regulation of calcineurin isozymes in the context of a peptide substrate, we performed in vitro phosphatase assays with RII phosphopeptide in the absence or presence of Ca 2ϩ and Ca 2ϩ /CaM (Fig. 5). As expected, CN␤2 was inactive in the absence of Ca 2ϩ or CaM and exhibited a steep dependence on Ca 2ϩ /CaM, which stimulated activity 20-fold relative to Ca 2ϩ alone. Dephosphorylation of RII phosphopeptide by CN␤1 wt was similarly very low in EGTA. Activity was stimulated 13-fold by Ca 2ϩ , but in contrast to CN␤2, Ca 2ϩ /CaM only modestly raised the activity of CN␤1 wt , and maximal activity was significantly lower than for CN␤2 under the same conditions. Mutating the AIS (CN␤1 LAVPwt,AISmut ) resulted in no significant differences in enzyme activity or regulation. However, mutation of the LAVP sequence (CN␤1 LAVPmut,AISwt ) caused a significant increase in activity both in the presence of Ca 2ϩ and Ca 2ϩ /calmodulin. Thus, LAVP is responsible for the low phosphatase activity observed for CN␤1 under stimulated conditions, and this sequence, which is unique to the C terminus of CNA␤1, acts as an autoinhibitory domain by blocking substrate engagement, not the active site. Analysis of the double mutant, CN␤1 LAVPmut,AISmut , revealed a significant (18-fold) increase in phosphatase activity in EGTA compared with CN␤1 wt . This is consistent with the proposed inhibitory role of the AIS (16)

Mechanism of CNA␤1 autoinhibition
AIS, other elements within the regulatory domain contribute to autoinhibition.

CN␤1 is autoinhibited in vivo by both the AIS and the LXVP motif
Having analyzed calcineurin phosphatase activity in vitro, we sought to investigate how CN␤1 activity was regulated in cells, especially in the context of protein substrates that include both PXIXIT and LXVP calcineurin interaction motifs. Thus, we reexamined the activity of human calcineurin in yeast, and measured the ability of human calcineurin to specifically dephosphorylate and activate the Crz1 transcription factor. To assay calcineurin-dependent Crz1 activity quantitatively, we incorporated a reporter gene consisting of four tandem repeats of the DNA sequence that Crz1 binds (CDRE) placed upstream of the ␤-galactosidase gene (4X-CDRE-lacZ) (26). The amount of ␤-galactosidase activity in extracts of yeast carrying this reporter gene reflects the activity of calcineurin in the cell.
To measure human calcineurin activity, hCNB was expressed together with CNA␤2, CNA␤1 wt , CNA␤1 LAVPwt,AISmut , CNA␤1 LAVPmut,AISwt , or CNA␤1 LAVPmut,AISmut in calcineurin-deficient yeast. Furthermore, to examine CaM dependence, each isozyme was expressed alone or together with sCaM. To mirror the conditions of in vitro assays, ␤-galactosidase activity was measured under three conditions: 1) unstimulated: growth in standard medium, which contains low amounts of Ca 2ϩ , 2) Ca 2ϩ -stimulated: following exposure to metal ion stress which induces Ca 2ϩ signaling, 3) Ca 2ϩ /CaM-stimulated: following stress treatment of cells co-expressing sCaM. In the absence of stress, expression of sCaM did not significantly affect observed CDRE-lacZ activity (data not shown).
Expression of CN␤2 failed to generate detectable CDRE-lacZ activity under unstimulated or Ca 2ϩ -stimulated (stress) conditions (Fig. 6). However, statistically significant amounts of ␤-galactosidase were produced under stress when sCaM was co-expressed. These results mirror the initial growth assays ( Fig. 2A) and confirm that canonical calcineurin absolutely requires both Ca 2ϩ and CaM for activation in vivo (Fig. 6). In cells expressing CN␤1 wt , CDRE-lacZ activity was low but detectable under unstimulated conditions and progressively increased upon stress treatment and co-expression of sCaM, suggesting that Ca 2ϩ and Ca 2ϩ /CaM each contributed to enzyme activation. By contrast, in cells expressing CN␤1 LAVPwt,AISmut , CDRE-lacZ activity was statistically equivalent across all three conditions, and rose to the level of activity of stress/CaM-stimulated CN␤1 wt , revealing a contribution of the AIS to autoinhibition. However, mutation of the LAVP (CN␤1 LAVPmut,AISwt and CN␤1 LAVPmut,AISmut ) resulted in high CDRE-lacZ activity under all conditions tested. This suggests that these enzymes are active even under standard growth conditions, causing ␤-galactosidase to continually accumulate regardless of stimulation conditions (supplemental Fig. S4). Expression levels of hCNB and sCaM were similar between strains and reproducible in experimental replicates (data not shown). Each CN␤1 isoform was also comparably expressed, with the exception of substrate: RII phosphopeptide CN

Mechanism of CNA␤1 autoinhibition
CN␤1 LAVPmut,AISmut , whose levels were reduced by ϳ50% (data not shown). Together, these findings suggest that for activation of a protein substrate (the yeast Crz1 transcription factor), both the AIS and the LXVP motif in CNA␤1 contribute to autoinhibition, with loss of the LXVP motif resulting in high levels of activity.

Discussion
These studies establish the biochemical mechanisms that regulate phosphatase activity of CNA␤1, a conserved isoform that is generated through alternative 3Ј end mRNA processing and consequently contains unique C-terminal regulatory sequences. Both in vitro and when expressed in yeast, CN␤1 differs from CN␤2 by exhibiting significant CaM independence, a more restricted activity range, and lower maximal activity in the presence of Ca 2ϩ /CaM. Our data show that the conserved LAVP sequence in the C-terminal domain of CNA␤1 is an LXVP SLiM that competitively inhibits substrate dephosphorylation in trans, and serves as a cis-autoinhibitory sequence, with mutation of key motif residues abolishing inhibition. These distinct regulatory properties likely contribute to the difference in biological functions observed for CN␤1 and CN␤2.
Elegant studies by Klee and others (4,(17)(18)(19)(20) have elucidated the complex mechanism of canonical calcineurin activation, which is consistent with our results for CN␤2 and is described below. In Fig. 7A, we present a model, based on this body of work, in which a snapshot of possible conformations for each signaling condition is shown, with thicker lines surrounding forms proposed to predominate in the ensemble. Under basal Ca 2ϩ conditions (EGTA), the enzyme is in an inactive, closed conformation (form I). Only two of the four Ca 2ϩ -binding EF hands in CNB are occupied (17), preventing the N-terminal lobe of CNB from interacting with the BBH (20). This positioning of CNB facilitates BBH-CBD binding, including engagement of the AIS at the LXVP docking surface (16). Additionally, the AID blocks access to the catalytic center (31). Extremely low activity observed with both substrates (Figs. 4 and 5) suggests that form I is dominant under these conditions, with other possible forms (II-IV) existing transiently, if at all. Upon Ca 2ϩ addition, the two low-affinity Ca 2ϩ sites in CNB become occupied, promoting a conformational change that disrupts the BBH-CBD interaction (20). Release of the CBD (including the AIS) increases accessibility of the substrate-binding pocket. Meanwhile, Ca 2ϩ also alters the structure of the LXVP-binding surface and the catalytic center to enhance dephosphorylation (17,37). The inactive conformation, I-Ca 2ϩ , still predominates, but the modest increase in activity observed upon Ca 2ϩ addition (Fig. 5) suggests some availability of the catalytically active

Mechanism of CNA␤1 autoinhibition
forms (II-IV-Ca 2ϩ ). Ca 2ϩ /CaM-bound calcineurin is fully active and cannot be further activated except through proteolysis that removes the regulatory domain (37). Ca 2ϩ binding to CNB is essential for CaM stimulation of calcineurin and Ca 2ϩbound CaM binds calcineurin at very high affinity (K D Յ 0.1 nM) (17,40). Upon CaM binding, the CBD adopts an ␣-helical structure that results in the full removal of the domain, including the AIS, from the LXVP pocket, and triggers a conformational change that displaces, but may not fully remove, the AID from the catalytic center (16,39). In summary, most calcineurin isoforms, including CNA␤2, form a heterodimer with CNB whose activity is strictly controlled by Ca 2ϩ /CaM. Both Ca 2ϩ and CaM are required in concert to relieve autoinhibition at two separate locations that form the extended active site: the LXVP SLiM docking surface and the catalytic center. In contrast to CNA␤2, CNA␤1 lacks the AID and instead possesses a unique C-terminal regulatory domain. Based on our results, we propose a model for the multistep regulation of this isoform (Fig. 7B). In the presence of EGTA, the catalytic center of CN␤1 is open, as demonstrated by the elevated level of pNPP hydrolysis by CN␤1 compared with CN␤2 (Fig. 4), and the enzyme exists as an ensemble of inactive (forms I and II) and active (form III) conformations. In forms I and II, the substratebinding pocket is engaged by the LXVP motif or AIS, respectively. Mutation of either (Fig. 6) or both sequences (Figs. 5 and 6) significantly increased basal enzyme activity suggesting that both forms are present, but the overall low amount of RII phosphopeptide dephosphorylation suggests that a minor fraction of the enzyme adopts the fully active form III (Fig. 5). By analogy with CN␤2, Ca 2ϩ weakens the BBH/CBD interaction (20), thus shifting the equilibrium away from the CBD-bound form (II-Ca 2ϩ ) and toward forms I-Ca 2ϩ and III-Ca 2ϩ . This aspect of the model is supported by the modest activation observed for CN␤1 LAVPwt,AISmut in vivo in the presence of stress (Fig. 6), compared with the dramatic activity increase observed for CN␤1 LAVPmut,AISwt both in vivo and in vitro (Figs. 5 and 6). The effects of the AIS mutation in CN␤1 LAVPwt,AISmut may be more apparent in vivo due to the high effective concentration of an endogenous substrate containing both calcineurin interaction motifs (PXIXIT and LXVP). The presence of the active form III-Ca 2ϩ in the ensemble explains the ability of CN␤1-expressing yeast to grow under stress in the absence of sCaM compared with yeast expressing CN␤2 ( Fig. 2A). Ca 2ϩ /CaM fully dissociates the CBD, including the AIS, from the substrate-binding pocket. In contrast to CNA␤2, however, the LXVP motif in CNA␤1 limits activity even in the presence of Ca 2ϩ /CaM. This is apparent in vitro, where Ca 2ϩ /CaM-stimulated CN␤1 activity never reaches the level of CN␤2, versus CN␤1 LAVPmut,AISwt and CN␤1 LAVPmut,AISmut , which are highly active (Fig. 5). Indeed, this evidence suggests that of the two CaM-bound states (forms I-CaM and II-CaM), the inactive, LXVP-bound form I-CaM may predominate.
In conclusion, we propose that CN␤1 exhibits a unique mechanism of regulation by Ca 2ϩ /CaM. In contrast to CN␤2, CN␤1 is autoinhibited at a single site, but by two inhibitory regions that compete with each other and with substrates. The first, the AIS-containing CBD, is progressively removed by the addition of Ca 2ϩ and CaM, but the second, LAVP, remains bound. This confers higher basal and Ca 2ϩ -dependent activity, but less dependence on CaM, and limits the maximum activity achievable by this isozyme.

Mechanism of CNA␤1 autoinhibition
These unique biochemical properties may allow CN␤1 to play distinct regulatory roles in vivo. In vitro, autoinhibition of CN␤1 by its C-terminal LXVP motif limits its activity toward the RII phosphopeptide. However, in vivo, this LAVP sequence may rather serve as a filter that restricts CN␤1 substrate specificity. Our model suggests that CN␤1 will selectively dephosphorylate substrates whose high-affinity LXVP SLiMs can outcompete the C-terminal LAVP sequence for binding to the substrate docking groove on calcineurin. SLiM properties do vary, as NFAT family members contain LXVP motifs with differing affinities for calcineurin (41). Indeed, altering PXIXIT affinity in either Crz1 or NFAT modulates signaling strength (42,43), and the same is likely to apply to LXVP motifs. An unexpected finding is that CN␤1 activity in vitro and in vivo is lower compared with CN␤2 in the presence of maximal Ca 2ϩ / CaM stimulation. This limited activity range might be biologically advantageous in vivo under extreme Ca 2ϩ conditions, such as in microdomains near Ca 2ϩ entry sites, or during sustained signaling (44). We also show that CN␤1 has significant CaM-independent activity, suggesting that it can signal under CaM-limiting conditions (45). Finally, in vivo, the C-terminal 23 aa of CNA␤1, deleted in our studies, directs the enzyme to the Golgi, in contrast to the cytosolic localization of CNA␤2 (21). This membrane association potentially targets CNA␤1 to a unique pool of substrates, including those regulating mechanistic target of rapamycin (mTOR) and AKT signaling (21). The CNA␤1 C terminus also mediates unique protein-protein interactions, including binding to Rictor, a subunit of mTORC2 (22). Thus, in vivo, interaction of this C-terminal tail with membranes and/or proteins might relieve autoinhibition by the LAVP pseudosubstrate sequence to activate the enzyme in a spatially regulated manner. Overexpression and depletion studies have clearly indicated that CNA␤1 has unique physiological functions that are distinct from CNA␤2 (7,(21)(22)(23). The biological mechanisms underlying these effects remain to be elucidated as CN␤1 substrates are identified and their unique regulation by CN␤1 can be studied in vivo.

Growth media and general methods
Yeast media and culturing methods were followed as described in synthetic complete (SC) medium (46) except twice the amount of amino acids and nucleotides were used. Yeast transformations were performed by the lithium acetate method (47). Plasmids and yeast strains used in this study are listed in supplemental Tables S1 and S2, respectively.

RNAseq analysis
FASTQ files from the ArrayExpress database at EMBL-EBI under accession number E-MTAB-1733 were downloaded and paired reads were aligned to unique sequence regions of CNA␣, CNA␤2, CNA␤1, and TATA box-binding protein using Bow-tie2 version 2.2.3. Reads were filtered for quality with SAMtools version 1.1. Raw counts for each CNA variant were divided by the length of the reference sequence and then normalized to TATA box-binding protein, whose expression variation is low across most tissues (48).

Protein expression and purification
For MBP-CN␤ trunc -His (MBP, maltose-binding protein), CNA and CNB subunits were expressed in tandem in Escherichia coli BL21(DE3) RIL cells (Invitrogen) and cultured in LB containing chloramphenicol (34 g/ml) and ampicillin (50 g/ml) at 37°C to mid-log phase. Expression was induced for 2 h with 1 mM isopropyl 1-thio-␤-D-galactopyranoside. Cells were harvested by centrifugation (4,000 ϫ g, 20 min) and frozen at Ϫ80°C. For purification, all steps were performed at 4°C. Cells were resuspended in lysis buffer (50 mM Tris, pH 7.4, 100 mM NaCl, 1 mM DTT, protease inhibitors) and supplemented with 2 g/ml of DNase. Cells were lysed by two passes through a French press at 35,000 psi. Crude extract was clarified by three rounds of centrifugation (15,000 ϫ g, 30 min). MBP-CN␤ trunc -His heterodimers were bound in batch to amylose resin (New England Biolabs) for 2 h, washed with lysis buffer, and eluted with lysis buffer containing 20 mM maltose. MBP-CN␤ trunc -His was then affinity purified with Ni-NTA-agarose (Invitrogen) in the presence of 15 mM imidazole for 1 h, washed with lysis buffer containing 15 mM imidazole, and eluted with lysis buffer containing 300 mM imidazole. Pure MBP-CN␤ trunc -His was brought to 20% glycerol, aliquoted, and stored at Ϫ80°C.

Enzyme activity assays
The rate of RII dephosphorylation was determined by measuring continuous PO 4 release detected with the phosphatebinding fluorophore MDCC-PBP (Invitrogen). Reactions were performed in assay buffer (50 mM Tris, pH 8.0, 100 mM NaCl, 6 mM MgCl 2 , 100 g/ml of BSA, 1 mM DTT, 0.4 mM CaCl 2 , 0.25 M MDCC-PBP). For peptide trans-inhibition assays, reactions contained 5 nM MBP-CN␤ trunc , 50 M RII phosphopeptide substrate, and 0 -100 M peptide. For assays to determine the mode of inhibition, reactions contained 5 nM MBP-CN␤ trunc , 100 -1000 M RII phosphopeptide substrate, and 0 -100 M peptide. The mode of inhibition was determined by comparing alternative models with Akaike Information Criterion using Prism software (GraphPad) and selecting the more statistically probable model, which was always favored by Ͼ90%.
For assays with calcineurin isozymes (CN␤2, CN␤1 wt , CN␤1 LAVPmut,AISwt , CN␤1 LAVPwt,AISmut , and CN␤1 LAVPmut,AISmut ) reactions contained 2.5 nM CN, 50 M RII phosphopeptide substrate, and 2 mM EGTA, 0.4 mM CaCl 2 , or 0.4 mM CaCl 2 ϩ 0.25 M bovine CaM (Calbiochem). Experiments were performed in black low-binding half-area 96-well plates in a Biotek Neo plate reader (BioTEK, Winooski, VT) with continuous shaking at room temperature. Fluorescence was measured at 425 nm excitation/460 nm emission. Reaction rates were linear and constituted less than 10% of product formation. Phosphate standards were used to convert fluorescence signal to PO 4 concentration. Enzyme rate is reported as v/ [E], where v is initial velocity and [E] is total enzyme concentration.
The rate of pNPP hydrolysis was measured in a continuous assay by monitoring the production of pNP at 405 nm. Reactions were performed in assay buffer (100 mM Tris, pH 8.0, 100 mM NaCl, 0.4 mM CaCl 2 , 100 g/ml of BSA, 6 mM MgCl 2 , 1 mM DTT). For peptide inhibition assays, reactions contained 10 nM MBP-CN␤ trunc , 20 mM pNPP (New England Biolabs), and 0 -100 M peptide. For assays with calcineurin isozymes (CN␤2, CN␤1 wt , CN␤1 LAVPmut,AISwt , CN␤1 LAVPwt,AISmut , and CN␤1 LAVPmut,AISmut ), reactions contained 2.5 nM CN, 20 mM pNPP, and either 2 mM EGTA, 0.4 mM CaCl 2 , or 0.4 mM CaCl 2 ϩ 0.25 M bovine CaM (Calbiochem). Experiments were performed in clear 96-well plates in a Biotek Neo plate reader with continuous shaking at room temperature. Reaction rates were linear and constituted less than 1% of product formation. Standards of known pNP (Sigma) concentration were used to convert absorbance units to pNP concentration, which was used as a proxy for PO 4 concentration. Enzyme rate is reported as v/[E], where v is initial velocity and [E] is total enzyme concentration.

Peptides
Peptides were synthesized and purified by the Tufts University Core Facility (Boston, MA) or by Peptide 2.0 (Chantilly, VA). The amino acid sequences of peptides used in the transinhibition assay are described in Table 1. Sequences of peptides used in supplemental Fig. S1 are as follows: AIS-FSVL wt peptide, ARVFSVLREESESVL; AIS-FSVL mut peptide, ARVASAA-REESESVL and RII phosphopeptide, DLDVPIPGRFDRRVSp VAAE. Peptides were dissolved in 10 mM Tris, pH 8.0.

␤-Galactosidase assay
Yeast strains (RBY05-14) were grown in SC lacking leucine, uracil, and tryptophan, and were supplemented with 80 g/ml of adenine and 2% dextrose to mid-log phase. Calcineurin expression was induced with 100 M CuSO 4 for 30 min prior to the 1.5-h treatment in 5 mM MnCl 2 . Cultures were transferred to 96-well plate for lacZ analysis, whereas 1 ml was reserved to confirm even protein expression (see below). Calcineurin-dependent activity was determined using substrate fluorescein di-␤-D-galactopyranoside (F-1179, Invitrogen) as previously described (49). Fluorescence emission at 530 nm was detected using a Biotek Neo plate reader and normalized to A 600 . Activity of each strain was measured in triplicate, and at least 3 independent transformants were evaluated per strain. Under standard growth conditions without CuSO 4 induction or stress treatment, calcineurin subunits were expressed at a low level (supplemental Fig. S4A) yet CN␤1 LAVPmut,AISmut remained highly active (supplemental Fig. S4B).

Yeast growth assay
Yeast strains (RBY01-06) were grown in selective media to mid-log phase, then 0.2 A 600 units of cells from each culture were 5-fold serially diluted in water and spotted onto plates. Plates contained SC lacking leucine, uracil, and tryptophan, and were supplemented with 80 g/ml of adenine, 2% dextrose, and 0 or 10 mM MnCl 2 . Where noted, plates were supplemented with 1 g/ml of FK506 (LC Laboratories) dissolved in 90% eth-