Dissecting physical structure of calreticulin, an intrinsically disordered Ca²⁺-buffering chaperone from endoplasmic reticulum

Abstract

Calreticulin (CALR) is a Ca²⁺ binding multifunctional protein that mostly resides in the endoplasmic reticulum (ER) and plays a number of important roles in various physiological and pathological processes. Although the major functions ascribed to CALR are controlling the Ca²⁺ homeostasis in ER and acting as a lectin-like ER chaperon for many glycoproteins, this moonlighting protein can be found in various cellular compartments where it has many non-ER functions. To shed more light on the mechanisms underlying polyfunctionality of this moonlighting protein that can be found in different cellular compartments and that possesses a wide spectrum of unrelated biological activities, being able to interact with Ca²⁺ (and potentially other metal ions), RNA, oligosaccharides, and numerous proteins, we used a set of experimental and computational tools to evaluate the intrinsic disorder status of CALR and the role of calcium binding on structural properties and conformational stability of the full-length CALR and its isolated P- and C-domains.

1. INTRODUCTION

Calreticulin (CALR) is a multifunctional protein that participates in various physiological and pathological processes. Being preferentially resided in the endoplasmic reticulum (ER), CALR is mostly known for its roles in Ca²⁺ homeostasis (in ER lumen, more than 50% of Ca²⁺ associates with CALR) (Nakamura et al., 2001) and acting as a lectin-like ER chaperon for many glycoproteins (Labriola, Cazzulo, & Parodi, 1999; Spiro, Zhu, Bhoyroo, & Soling, 1996; Vassilakos, Michalak, Lehrman, & Williams, 1998; Zapun et al., 1998). CALR binds monoglucosylated oligosaccharides (Helenius & Aebi, 2004; Helenius, Trombetta, Hebert, & Simons, 1997), as well as interacts with misfolded proteins that can be either glycosylated or non-glycosylated (Lum et al., 2016; Saito, Ihara, Leach, Cohen-Doyle, & Williams, 1999). As a matter of fact, together with another lectin chaperone (or carbohydrate-binding chaperone), calnexin (CNX), CALR is known to interact with practically all glycoproteins in the ER (Ellgaard & Frickel, 2003). In response to Ca²⁺, CALR might escort numerous client proteins to their active sites, thereby regulating their functions (Fadel et al., 1999; Jiang, Dey, & Matsunami, 2014). Besides being highly enriched within the ER lumen, CALR is present in other cellular compartments, where it has multiple non-ER functions (Gold et al., 2010). For example, in the cytoplasm, CALR shuttles other proteins between Golgi and endoplasmic reticulum (Fadel et al., 1999; D. B. Williams, 2006). It can also be present in the nucleus, where CALR mediates the export of all nuclear receptors, including the glucocorticoid receptor (GRα) (Dedhar et al., 1994; Holaska et al., 2001). CALR can translocate to the surface of cells undergoing immunogenic cell death (ICD), or immunogenic apoptosis, acting as a part of a damage-associated molecular pattern (DAMP) involved in the phagocytic response (Chao et al., 2010; Gardai et al., 2005). In fact, CALR expression is abnormally high on the cell-surface of a wide variety of cancer cells, including hematological malignancy (Wemeau et al., 2010; Zamanian et al., 2016; Zamanian, Veerakumarasivam, Abdullah, & Rosli, 2013). It has been suggested that, by antagonizing the “eat-me” signal provided by CD47, this abnormal cell-surface expression of CALR may inhibit clearance of cancer cells by the immune system, favoring disease progression (Chao et al., 2010; Obeid et al., 2007). Since there is no transmembrane domain in CALR, this protein is anchored to the membrane via binding of its C-terminal domain to specific adaptors, such as CD59 in neutrophils (Ghiran, Klickstein, & Nicholson-Weller, 2003). On the plasma membrane, CALR acts as receptor for orphan ligands, such as thrombospondin (Goicoechea, Pallero, Eggleton, Michalak, & Murphy-Ullrich, 2002). Furthermore, CALR is involved in cell adhesion by playing a role in regulation of expression of several genes related to focal contacts (Fadel et al., 1999; Villagomez et al., 2009) or via direct interaction with the cytoplasmic KXGFFKR motif of the integrin α-subunit (Coppolino & Dedhar, 1999; Dedhar et al., 1994; Rojiani, Finlay, Gray, & Dedhar, 1991). Finally, CALR was shown to be an RNA-binding protein that regulates mRNA stability (Lu, Weng, & Lee, 2015; Mueller et al., 2008; Nickenig et al., 2002; Totary-Jain et al., 2005) and can be found in stress granules (SGs) in the cytoplasm (Carpio, Lopez Sambrooks, Durand, & Hallak, 2010).

Therefore, these and many other studies clearly show that CALR is a multifunctional moonlighting protein. As with many other moonlighting proteins, the ability of CALR to have a highly diversified functional repertoire is related to the tightly controlled distribution of this protein within the various cellular compartments that allows CALR to act as the right man that appears in the right place at a right time. Furthermore, it is recognized also that the polyfunctionality of CALR can be further controlled and regulated by various posttranslational modifications (PTMs). For example, CALR phosphorylation is essential for mRNA binding (Mueller et al., 2008), whereas CALR association with SGs, being promoted by the decrease in intracellular Ca²⁺, is arginylation-dependent (Carpio et al., 2010).

What makes CALR a moonlighting protein? Human CALR is synthesized in a form of a precursor containing cleavable N-terminal signal peptide (residues 1-17) needed for targeting this protein to the ER lumen, where it is accumulated due to the presence of a C-terminal KDEL ER-retention motif (Afshar, Black, & Paschal, 2005). Structurally and functionally, mature human CALR (residues 18-417; UniProt ID: P27797) can be divided into three domains, a globular N-terminal domain (N-domain, N-CARL, residues 18-197), an extended proline-rich P-domain (residues 198-308), and an acidic C-terminal domain (C-domain, C-CARL, residues 309-417) (Chouquet et al., 2011; Michalak, Groenendyk, Szabo, Gold, & Opas, 2009). The chaperone function of this protein requires N-CALR and P-CALR that bind other proteins. Furthermore, this chaperone module is helped by C-CALR that in addition to binding other proteins contains the KDEL motif required for translocation of the protein and its complexes across membranes. Furthermore, since C-CALR contains a large number of negatively charged amino acids, this domain is involved in high-capacity Ca²⁺ storage (see Figure 1 that shows linear representation of the CALR and indicates domains with their boundaries used in this study).

Figure 1.

Schematic representation of human CALR where positions of the domains with their boundaries used in this study are shown and some basic functions ascribed to different domains are shown.

It became clear at the turn of the 20^th century that many biologically active proteins lack unique 3D structures under physiological conditions and instead exist as highly dynamic conformational ensembles consisting of a multitude of interconverting structures (Dunker et al., 2001; Habchi, Tompa, Longhi, & Uversky, 2014; Oldfield & Dunker, 2014; Tompa, 2002, 2012; Uversky, 2013a, 2013c, 2016a; Uversky & Dunker, 2010; Uversky, Gillespie, & Fink, 2000; van der Lee et al., 2014; Wright & Dyson, 1999). This phenomenon is known as protein intrinsic disorder, and it defines high structural heterogeneity of proteins, where either entire protein or any part of it can be disordered to different degree. The inability of intrinsically disordered proteins (IDPs) and intrinsically disordered protein regions (IDPRs) to (completely) fold and form globular well-defined structures with hydrophobic cores, as seen in “normal” ordered proteins, is determined by the presence of specific differences in the amino acid sequences of ordered proteins and IDPs. Since the ability of a polypeptide chain to collapse into compact globule in aqueous media is determined by an interplay between its hydrophobicity and the overall net charge, combination of low overall hydrophobicity and high net charge defines the existence of proteins with extended disorder (Uversky et al., 2000). Statistically, amino acid sequence of IDPs and IDPRs are characterized by the enrichment in the so-called disorder-promoting residues (Arg, Gln, Glu, Gly, Lys, Ser, and Pro) and the depletion in the order-promoting residues (Asn, Cys, Ile, Leu, Phe, Trp, Tyr, and Val) (Dunker et al., 2001; Radivojac et al., 2007; P. Romero et al., 2001; Vacic, Uversky, Dunker, & Lonardi, 2007; R. M. Williams et al., 2001).

According to several bioinformatics studies, IDPs and hybrid proteins possessing ordered domains and IDPRs constitute significant fraction of all proteomes, with the per-proteome amounts of differently disordered proteins increasing with the increase in the organism complexity (Dunker, Obradovic, Romero, Garner, & Brown, 2000; Z. Peng et al., 2015; Walsh et al., 2015; Ward, Sodhi, McGuffin, Buxton, & Jones, 2004; Xue, Dunker, & Uversky, 2012). According to the Protein Data Bank (PDB) analyses, many proteins with known 3-D structures commonly contain IDPRs of different length (DeForte & Uversky, 2016b; Le Gall, Romero, Cortese, Uversky, & Dunker, 2007). Furthermore, even enzymes, which have long been considered an exception to the rule of the prevalence of protein intrinsic disorder due to the structural requirements for catalysis, were recently shown to abundantly contain functional IDPRs (DeForte & Uversky, 2016a, 2017). High natural abundance of IDPs/IDPRs is determined by their functional multifariousness (Dunker, Brown, Lawson, Iakoucheva, & Obradovic, 2002; Dunker, Brown, & Obradovic, 2002; Dunker et al., 2001; Habchi et al., 2014; Uversky et al., 2000; van der Lee et al., 2014), especially in biological processes related to the control and regulation of various signaling pathways (Dunker, Cortese, Romero, Iakoucheva, & Uversky, 2005; Iakoucheva, Brown, Lawson, Obradovic, & Dunker, 2002; Leonova & Galzitskaya, 2015; Oldfield et al., 2005). IDPs/IDPRs are tightly controlled themselves by different means, such as alternative splicing (Buljan et al., 2013; Buljan et al., 2012; P. R. Romero et al., 2006) and numerous enzymatically-catalyzed posttranslational modifications (Iakoucheva et al., 2004; Lenton et al., 2017; Pejaver et al., 2014; Reddy et al., 2017). Intrinsic disorder is crucial for functions of ‘moonlighting’ proteins (Tompa, Szasz, & Buday, 2005), control of generation and disintegration of various cellular membrane-less organelles (Uversky, 2016b, 2017; Uversky, Kuznetsova, Turoverov, & Zaslavsky, 2015), programmed cell death (Z. Peng, Xue, Kurgan, & Uversky, 2013), innate immunity (Xue & Uversky, 2014), antimicrobial peptides of the NK-lysin family (Yacoub, Al-Maghrabi, Ahmed, & Uversky, 2017), βγ-crystallin Hahella chejuensis (Gao, Yang, Zhang, Su, & Huang, 2017), transcription regulation (Fuxreiter et al., 2008; Liu et al., 2006; Toth-Petroczy et al., 2008), regulation of kinase activity (Kathiriya et al., 2014), induction of pluripotent stem cells (Xue, Oldfield, Van, Dunker, & Uversky, 2012), assembly and functionality of various protein complexes (Fuxreiter et al., 2014), and conditional or transient disorder (Jakob, Kriwacki, & Uversky, 2014; Uversky, 2015). This list is far from being complete, and it is believed now that the biological activities of IDPs/IDPRs are complementary to functions of ordered proteins and domains (Dunker et al., 2001; Radivojac et al., 2007; Uversky, 2002). Finally, pathogenesis of different human diseases, such as cancer, cardiovascular disease, amyloidoses and neurodegeneration, is commonly associated with dysfunctions of IDPs/IDPRs (Uversky et al., 2014; Uversky, Oldfield, & Dunker, 2008).

Curiously, several specific structural features and important functions were assigned to the P- (residues 198–308 of the unprocessed human protein, UniProt ID: P27797) and C-terminal domains (residues 309–417 of the unprocessed human protein, UniProt ID: P27797) of CALR. For example, SAXS analysis of the monomeric CALR in solution showed that the P-domain (or P-arm) is characterized by high flexibility, protrudes from the extended globular structure formed by N- and C-domains, and might adopt a spiral-like conformation (Norgaard Toft et al., 2008), whereas molecular dynamics simulations revealed that the CALR P-domain is characterized by high conformational flexibility and that interactions between CALR and its binding partners promotes transition of the P-domain into “open” conformation (Yan, Murphy-Ullrich, & Song, 2010). Subsequent studies indicated that CALR differently interacts with glycosylated and non-glycosylated target proteins, binding of which display distinct kinetic profiles because they induce “open” and “closed” conformations of P-domain, respectively (Wijeyesakere, Rizvi, & Raghavan, 2013). Recently, the presence of such a transition from “open” to “closed” conformation that regulates the accessibility of the dual specificity substrate-binding site, which is able to interact with both carbohydrates and/or proteins, was demonstrated during the structural characterization of CALR isolated from two distinct parasites, Trypanosoma cruzi and Entamoeba histolytica (Moreau et al., 2016). It was shown that the deletion of the acidic tail (residues 359-417) stimulates the ability of CALR to interact with polypeptide substrates and enhance chaperone activity of this protein (Rizvi, Mancino, Thammavongsa, Cantley, & Raghavan, 2004). Application of chemical cross-linking, mass spectrometry, bioinformatics analysis, and computational modelling revealed that Ca²⁺ binding is accompanied by the structural rearrangement of CALR, where the position of flexible P-domain is dependent on the concentration of Ca²⁺, part of the disordered acidic C-terminal tail is stabilized by Ca²⁺, and where Ca²⁺ induces interactions between the P-loop and the acidic C-terminal tail of CALR (Boelt et al., 2016).

In this article, we used a set of spectroscopic techniques to analyze the effect of calcium on structural properties and conformational stability of the recombinant human CALR and its P- and C-domains. We also utilized a wide spectrum of computational tools to look at the peculiarities of distribution of intrinsic disorder predisposition within the amino acid sequence of this protein and to find if intrinsic disorder may be related to its functional multifariousness.

2. MATERIALS AND METHODS

2.1. Materials

The full-length recombinant CALR and its isolated P- and C-domains (residues 198–308 and 309–417 of the unprocessed human protein, UniProt ID: P27797, respectively) were a kind gift of Prof. Marlene Bouvier (School of Pharmacy, University of Connecticut, Storrs, CT 06269 and Boston Biomedical Research Institute, Watertown, MA 02472). The E. coli strain BNN103 based on a glutathione S-transferase (GST) fusion protein system using the pGEX-3X plasmid (Pharmacia) was utilized for the CALR expression (Bouvier & Stafford, 2000). The expression plasmids for human CALR and its domains have been described previously (Bouvier & Stafford, 2000; Tan, Chen, Li, Mabuchi, & Bouvier, 2006). A glutathione-Sepharose 4B affinity column was used for the purification of the crude GST-CRT fusion protein at 4°C. The eluted protein was digested with Factor Xa. After the proteolytic digestion of the GST tag, the N-terminus of the resulting CALR protein contained additional GIPG residues. Digested GST tag was removed by direct application of the proteolytic mixture onto the glutathione-Sepharose 4B affinity column. Final stage of the protein purification was achieved by applying the CALR containing flow-through of the glutathione-Sepharose 4B affinity column to the Uno Q-6 (Bio-Rad) FPLC ion-exchange column. Purified protein was analyzed by the N-terminal amino acid sequencing and MALDI mass spectrometry to validate the Factor Xa cleavage and to evaluated the protein molecular mass (Bouvier & Stafford, 2000; Tan et al., 2006).

2.2. Computational evaluation of the intrinsic disorder predisposition of human CALR

2.2.1. Evaluation of the per-residue disorder propensity

The intrinsic disorder predisposition of human CALR (UniProt ID: P27797) was evaluated by four algorithms from the PONDR family, PONDR-FIT, PONDR^® VLXT, PONDR^® VSL2, and PONDR^® VL3 (X. Li, Romero, Rani, Dunker, & Obradovic, 1999; Obradovic, Peng, Vucetic, Radivojac, & Dunker, 2005; K. Peng, Radivojac, Vucetic, Dunker, & Obradovic, 2006; K. Peng et al., 2005; P. Romero et al., 2001; Xue, Dunbrack, Williams, Dunker, & Uversky, 2010), as well as by the IUPred web server with its two versions for predicting long and short disordered regions (Dosztanyi, Csizmok, Tompa, & Simon, 2005). PONDR^® VSL2 is one of the most accurate stand-alone disorder predictors (Obradovic et al., 2005), PONDR^® VL3 possesses a high accuracy for finding long disordered regions (Obradovic et al., 2003), PONDR^® VLXT is not the most accurate predictor, but has a high sensitivity to local sequence peculiarities which are often associated with disorder-based interaction sites (Dunker et al., 2001), PONDR-FIT is a metapredictor that is moderately more accurate than each of the component predictors and that is one of the most accurate disorder predictors (Xue et al., 2010), whereas IUPred utilizes the pair-wise energy estimation approach for finding intrinsically disordered residues and regions. The consensus disorder propensity of human CALR was evaluated by averaging disorder profiles of individual predictors. Use of consensus for evaluation of intrinsic disorder is motivated by empirical observations that this approach usually increases the predictive performance compared to the use of a single predictor (Fan & Kurgan, 2014; Z. Peng & Kurgan, 2012; Walsh et al., 2015).

2.2.2. MobiDB-based intrinsic disorder analysis

We further characterized the overall disorder status of human CALR using the MobiDB database (http://mobidb.bio.unipd.it/) (Di Domenico, Walsh, Martin, & Tosatto, 2012; Potenza, Domenico, Walsh, & Tosatto, 2015), that generates consensus disorder scores by combining the outputs of ten predictors, such as two versions of DisEMBL (Linding, Jensen, et al., 2003), two versions of ESpritz (Walsh, Martin, Di Domenico, & Tosatto, 2012), GlobPlot (Linding, Russell, Neduva, & Gibson, 2003), JRONN (Yang, Thomson, McNeil, & Esnouf, 2005), PONDR^® VSL2B (Obradovic et al., 2005; K. Peng et al., 2006), and two versions of IUPred (Dosztanyi et al., 2005). MobiDB also has manually curated annotations related to protein function and structure derived from UniProt (Apweiler et al., 2004) and DisProt (Sickmeier et al., 2007), as well as from Pfam (Finn et al., 2014) and PDB (Berman et al., 2000).

2.2.3. Analysis of the CALR functional disorder by D²P²

The D²P² database (http://d2p2.pro/) (Oates et al., 2013) was used for the complementary evaluation of intrinsic disorder propensity of human CALR. Some important disorder-related functional information was retrieved from this database too. D²P² is a database of predicted disorder for a large library of proteins from completely sequenced genomes (Oates et al., 2013) that uses outputs of IUPred (Dosztanyi et al., 2005), PONDR^® VLXT (P. Romero et al., 2001), PrDOS (Ishida & Kinoshita, 2007), PONDR^® VSL2B (Obradovic et al., 2005; K. Peng et al., 2006), PV2 (Oates et al., 2013), and ESpritz (Walsh et al., 2012). The database is further supplemented by data concerning location of various functional domains, sites of curated posttranslational modifications, and predicted disorder-based protein binding sites.

2.2.4. Prediction of potential disorder-based binding sites by ANCHOR and MoRFpred

The presence of the potential disorder-based protein binding sites in human CALR was evaluated by the ANCHOR algorithm [58,59]. This algorithm utilizes the pair-wise energy estimation approach originally used by IUPred [41,60]. This approach is based on the hypothesis that long disordered regions might include localized potential binding sites, which are not capable of folding on their own due to not being able to form enough favorable intrachain interactions, but can obtain the energy to stabilize via interaction with a globular protein partner [58,59].

In addition to ANCHOR, we utilized a MoRFpred algorithm that was designed to find all types of MoRFs, α-helix, β-strand, coil, and complex using a comprehensive dataset of annotated MoRFs (Disfani et al., 2012). This tool utilizes several sequence-derived hallmarks of potential MoRFs, such as their appearance as dips in the disorder predictions and high hydrophobicity and stability. High accuracy of MoRFpred in comparison with several existing tool for finding disorder-based binding sites in proteins is achieved by combining information about evolutionary profiles with several sequence-derived features, such as B-factors, predicted disorder, selected physiochemical properties of amino acids, and solvent accessibility (Disfani et al., 2012).

2.2.5. Analysis of the CALR interactivity by STRING

Additional functional information for these proteins was retrieved using STRING (Search Tool for the Retrieval of Interacting Genes, http://string-db.org/). STRING generates a network of predicted associations based on predicted and experimentally-validated information on the interaction partners of a protein of interest (Szklarczyk et al., 2011). In the corresponding network, the nodes correspond to proteins, whereas the edges show predicted or known functional associations. Seven types of evidence are used to build the corresponding network, where they are indicated by the differently colored lines: a green line represents neighborhood evidence; a red line - the presence of fusion evidence; a purple line - experimental evidence; a blue line – co-occurrence evidence; a light blue line - database evidence; a yellow line – text mining evidence; and a black line – co-expression evidence (Szklarczyk et al., 2011). In our analysis, the most stringent criteria were used for selection of interacting proteins by choosing the highest cut-off of 0.9 as the minimal required confidence level.

2.3. Spectroscopic analysis of the full-length CALR and its isolated P- and C-domains

2.3.1. Circular dichroism measurements

CD spectra were collected on an AVIV-60DS spectropolarimeter (AVIV, Lakewood, N.J.), equipped with a temperature-controlled cell-holder. In the far-UV region CD spectra were recorded in 0.01 cm cell from 250–190 nm with a step size of 0.5 nm, a bandwidth of 1.5 nm, and an averaging time of 5 sec. For all spectra, averages of 5 scans were obtained. CD spectra of the appropriate buffer were recorded and subtracted from the protein spectra. Protein concentrations were 1 mg/mL. The following buffers were used in the CD studies: 2 mM HEPES, 200 mM NaCl, pH 7.0 supplemented with either 2 mM EDTA, 50 μM CaCl₂, or 1000 μM CaCl₂. The molar ellipticity, θ was calculated as the CD signal (millidegrees) × MW (Da)/[number of residues × CALR concentration (mg/mL) × cell pathlength (mm)].

2.3.2. Fluorescence measurements

Since fluorescence intensity dramatically depends on temperature, the temperature should be kept constant to ensure that the observed effects are associated with protein and not changes in temperature. Therefore, all fluorescence measurements were performed using a FluoroMax-2 fluorescence spectrophotometer (Jobin Yvon-Spex, NJ, USA) equipped with a temperature-controlled cell-holder. All fluorescence measurements were performed at room temperature with the final protein concentrations of 0.01 mg/ml.

Intrinsic fluorescence measurements were conducted using an excitation wavelength of 280 nm and recording the emission spectra from 300 to 420 nm. The following buffers were used in the fluorescence experiments: 2 mM HEPES, 200 mM NaCl, pH 7.0 supplemented with either 2 mM EDTA, 50 μM CaCl₂, or 1000 μM CaCl₂.

In ANS binding studies, a fresh 10 mM ANS stock solution was prepared in double distilled water. The concentration of ANS was determined by UV absorption at 350 nm using a molar extinction coefficient of 5000. An aliquot of 1 μL of the ANS solution was added to 1 mL of buffer to a final concentration of 10 μM. A blank spectrum without protein was recorded before 10 μL of protein solution was added to the solution to a final protein concentration of 0.01 mg/ml. Fluorescence measurements were performed immediately after the addition of protein to the ANS solution in semimicro quartz cuvettes (Hellma, Germany). Spectra were recorded at ambient temperatures with a 1-cm excitation light path using a FluoroMax-2 spectrofluorometer (Instruments S.A., Inc. Jobin Yvon-Spex, USA). The light source was a 150-W xenon lamp. Emission spectra were recorded from 460 nm to 600 nm with excitation at 350 nm, an increment of 1 nm, an integration time of 1 s, and slits of 5 nm for both excitation and emission. For all spectra, averages of 5 scans were obtained. Data were processed using DataMax software.

2.3.3. Acrylamide Quenching

Acrylamide quenching studies of intrinsic fluorescence were performed by adding aliquots from a stock solution of the quencher into a cuvette containing the protein solution. Fluorescence intensities were corrected for dilution effects. Fluorescence quenching data were analyzed using the general form of the Stern-Volmer equation, taking into account not only the dynamic, but also static quenching (Eftink & Ghiron, 1981):

$$\frac{I_0}{I}=(1/K_{SV}[Q])e^{V[Q]}$$ (1)

where I_o and I are the fluorescence intensities in the absence and presence of quencher, K_SV is the dynamic quenching constant, V is a static quenching constant, and [Q] is the quencher concentration. All related experiments were conducted in triplicates.

2.4. Urea-induced unfolding of the full-length CALR and its P- and C-domains

Conformational stability of various CALR-related constructs was evaluated by measuring the intrinsic fluorescence after incubation of proteins in increasing amounts of urea (from 0 M to 8 M). All samples were incubated for four hours at room temperature to insure complete unfolding equilibrium. The final protein concentration was 0.01 mg/ml. The following buffers were used in the urea-induced unfolding studies: 2 mM HEPES, 200 mM NaCl, pH 7.0 supplemented with either 2 mM EDTA, 50 μM CaCl₂, or 1000 μM CaCl₂. The fluorescence measurements were performed using a FluoroMax-2 fluorescence spectrophotometer (Jobin Yvon-Spex, NJ, USA) with an excitation wavelength of 280 nm (5 nm band pass) and recording the emission spectra from 300 to 420 nm (5 nm band pass). Assuming a two-state folding mechanism, the fraction of the unfolded conformation in the unfolding of the full-length CALR was obtained using the following equation:

$${\mathrm{f}}_{\mathrm{u}}=\left({\mathrm{y}}_{\mathrm{f}}-\mathrm{y}\right)/\left({\mathrm{y}}_{\mathrm{f}}-{\mathrm{y}}_{\mathrm{u}}\right)$$ (2)

where y_f and y_u represent the intrinsic fluorescence value of y characteristic of the folded and unfolded conformations, respectively, under the conditions where y is being measured. The values of y_f and y_u were obtained by linear regression on the data points before and after the unfolding transition, respectively. A sigmoidal curve-fit was subsequently used to determine the midpoints of the unfolding transition (C_m). All related experiments were conducted in triplicates.

3. RESULTS AND DISCUSSION

3.1. Effect of Ca²⁺ binding on structural properties and conformational stability of CALR and its isolated P- and C-domains

CALR is an endoplasmic reticulum (ER) luminal Ca²⁺-buffering chaperone that serves as one of the ER Ca²⁺-binding/storage proteins (Nash, Opas, & Michalak, 1994), plays a role in regulation of the intracellular Ca²⁺ homoeostasis, and is involved in the control of the ER Ca²⁺ capacity (Fadel et al., 1999; Gelebart, Opas, & Michalak, 2005; Michalak et al., 2009). Therefore, it is not surprising that structure and function of CALR are known to be controlled by Ca²⁺ binding (Andrin et al., 1998; Carpio et al., 2010; Conte, Keith, Gutierrez-Gonzalez, Parodi, & Caramelo, 2007; Corbett et al., 1999; Holaska, Black, Rastinejad, & Paschal, 2002; Villamil Giraldo et al., 2010; Wijeyesakere, Gafni, & Raghavan, 2011). It was also emphasized that different domains of CALR may play different roles in the calcium sensitivity of this protein. For example, the CALR globular domain is involved in high affinity Ca²⁺ binding, whereas its acidic C-tail, with high content of acidic residues (22 glutamates and 12 aspartates are spread over the 50 C-terminal residues of this domain defining its net negative charge of −25), is responsible for the high capacity (~20 mol of Ca²⁺ per 1 mol of CALR) and low affinity (K_d ~ 2 mM) Ca²⁺-binding activity of CALR and is needed for the Ca²⁺-buffering properties of this protein (Baksh & Michalak, 1991). Dynamic light scattering and size exclusion chromatography-based analyses revealed that the full-length CALR undergoes noticeable compaction in the presence of Ca²⁺ (Villamil Giraldo et al., 2010), and the thermal stability of the full-length protein was increased in the Ca²⁺-dependent manner (Z. Li, Stafford, & Bouvier, 2001; Wijeyesakere et al., 2011). Here, the thermal denaturation midpoint temperatures (T_m) progressively increased from 40.2 °C to 44.3 °C and 46.4 °C when CaCl₂ concentration was increased from 0 to 50 μM and 1000 μM, respectively (Z. Li et al., 2001). Earlier studies also showed that the far-UV CD spectrum of the full-length CALR was characterized by a minimum at 208 nm and a prominent shoulder at 229 nm, which was attributed to the contributions from aromatic side chains (Bouvier & Stafford, 2000). In fact, ~11% of all residues in CALR are aromatic (this protein has 18 phenylalanines, 14 tyrosines, and 11 tryptophans) (Bouvier & Stafford, 2000). The shape of the far-UV CD spectrum of the full-length CALR was characterized by an apparent lack of features typical for the proteins having a high content of well-defined secondary structure such as α-helices and β-sheets (Bouvier & Stafford, 2000) and was not affected by the addition of CaCl₂ (Z. Li et al., 2001). Although these experiments uncovered interesting effects of calcium on the CALR structure one can argued that the low NaCl concentrations (<10mM) investigated by Wijeyesakere et al. (Wijeyesakere et al., 2011) may not be predictive of the conformational changes induced by Ca²⁺ at more “physiological” NaCl concentrations (i.e., 150–200 mM) in the human full-length CALR protein.

To gain a better understanding of how Ca²⁺ affects the structure and conformational stability of human CALR, a comprehensive biophysical analysis of the human full-length protein and of its isolated P- and C-domain at 0, 50 and 1000 μM Ca²⁺ was conducted in solutions containing 200 mM NaCl. In agreement with the results of previous studies, Figure 2A and 2B show that the addition of calcium decreases the overall hydrophobicity of the full-length CALR (measured as changes in the fluorescence parameters of the hydrophobic fluorescent probe ANS, and suggesting the Ca²⁺-induced increase in the compaction of this protein) and noticeably increases its conformational stability toward the urea-induced unfolding. The presence of some calcium-induced compaction of the full-length CALR is also evidenced by the slight blue shift of its intrinsic fluorescence from 335 nm to 334 nm, and to 333 nm when the CaCl₂ concentration was increased from 0 to 50 μM, and to 1000 μM, respectively (see Figure 2C). The presence of the calcium-induced partial compaction is further supported by the reduction in the efficiency of the dynamic fluorescent quenching (as evaluated by the Stern-Volmer constant, K_SV), from 4.19±0.08 to 3.91±0.06 and to 3.68±0.08 M⁻¹ for CALR in the presence of 0, 50 and 1000 μM Ca²⁺ in solutions containing 200 mM NaCl. For comparison, the urea-unfolded CALR in the presence of 10.5 M urea is characterized by the K_SV of 19.29 ±0.08 M⁻¹. Furthermore, Figures 2D, 2E, and 2F show that the urea-induced unfolding of the full-length CALR in its apo-form or in the presence of 50 and 1000 μM Ca²⁺ was accompanied not only by the noticeable red shift of the maximal fluorescence from 333–335 nm in their native states to 353–354 nm in the urea-unfolded states (that reflects a conformational transition from the compact form whose tryptophan residues are mostly located in the hydrophobic environment and protected from interaction with the solvent to the highly expanded state with solvent-accessible tryptophan residues) but also by a dramatic increase in the intrinsic fluorescence intensity, reflecting unfolding-promoted release of some quenching groups from the close vicinity of the tryptophan residues of this protein. We would like to emphasize here that although we came to the conclusions that are similar to those reported in previous studies (Wijeyesakere et al., 2011), the obtained results on the effect of Ca²⁺ on CALR structure and conformational stability in the presence of 200 mM NaCl are not trivial, since one might expect that protein interactions with Ca²⁺ (which are obviously electrostatically driven) may be affected by salt.

Figure 2. Analysis of the effect of calcium on structural properties and conformational stability of the full-length human CALR. The following buffers were used in these spectroscopic studies: 2mM HEPES, 200mM NaCl, pH=7.0 supplemented with either 2mM EDTA (Apo), 50μM CaCl₂, or 1000μM CaCl₂.

A. ANS fluorescence spectra measured for the full-length human CALR in the absence calcium (black line) or in the presence of 50μM CaCl₂ (red line), or 1000μM CaCl₂ (green line). Spectra of free ANS (blue line) and CALR completely unfolded by 8M urea (pink line) are shown for comparison. In these experiments, protein concentration was 0.01mg/mL, and ANS concentration was kept at the level of 10μM.

B. Urea-induced unfolding of the apo-form of the full-length human CALR (black circles) and the CALR in the presence of 50μM CaCl₂ (red triangles), or 1000μM CaCl₂ (green squares). Urea-induced unfolding was detected by the characteristic red shift of the intrinsic fluorescence spectrum monitored by the urea-induced changes in the F₃₆₀/F₃₂₀ values, where F₃₆₀ and F₃₂₀ correspond to the fluorescence intensity at 360 and 320nm, respectively. Experiments were performed in triplicate. Plots represents averaged F₃₆₀/F₃₂₀ values. Conformational stability of the full-length protein is progressively increased with the increase in calcium concentration.

C. Normalized intrinsic fluorescence spectra of the full-length CALR in the absence calcium (black line) or in the presence of 50μM CaCl₂ (red line), or 1000μM CaCl₂ (green line), and CALR completely unfolded by 8M urea (pink line). Spectra were normalized, since there is a dramatic increase in their maximal intensity which makes the visual analysis of their position difficult.

D. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the apo-CALR. Spectra were measured in the presence of 0.0, 0.82, 1.20, 1.65, 2.15, 2.61, 2.94, 3.44, 4.07, 4.59, 5.25, 5.73, 6.59, 7.21, and 7.97 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

E. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the full-length human CALR in the presence of 50μM CaCl₂. Spectra were measured in the presence of 0.00, 0.83, 1.68, 2.25, 2.59, 2.94, 3.41, 3.92, 4.21, 4.91, 5.45, 5.98, 6.21, and 6.49 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

F. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the full-length human CALR in the presence of 1000μM CaCl₂. Spectra were measured in the presence of 0.00, 0.97, 1.68, 2.10, 2.52, 2.98, 3.44, 3.87, 4.28, 4.78, 5.27, 5.69, 6.10, 6.53, 6.96, 7.39, 7.83, 8.73, and 9.07 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

In the existing structural models, CALR is typically depicted as a rigid body (comprised of the N-CALC and N-terminal part of C-domain) with two flexible arms (P-domain and C-tail) (Chouquet et al., 2011; Michalak et al., 2009; Wijeyesakere et al., 2011). Therefore, technically, analysis of the isolated domains of this protein is justifiable. Furthermore, there is an important question on whether the individual isolated domains of CALR can sense calcium in the presence of 200 mM NaCl. Therefore, to understand if structure and conformational stability of the isolated P-domain and C-domain are impacted by calcium, we analyzed how the addition of different CaCl₂ concentrations affects their far-UV CD spectra (Figures 3A and 4A), intrinsic fluorescence (Figures 3B and 4B), and conformational transition curves describing their urea-induced unfolding (Figures 3C and 4C). Figure 3A shows that the far-UV CD spectrum of the isolated P-domain is characterized by a rather unusual shape and is insensitive to the presence or absence of calcium. This shape of the P-domain far-UV CD spectrum, with a strong negative band at ~200 nm, a weak positive band at ~222 nm, and another negative band at ~230 nm, can be attributed to a polyproline II-like conformation that can be present in highly charged polypeptides with high proline content. In fact, this 101 residue-long domain has a net charge of −20 containing 21 aspartates, 13 glutamates, 12 lysines and 2 arginines and 19 prolines account for 17.4% of all its residues.

Figure 3. Analysis of the effect of calcium on structural properties and conformational stability of the isolated P-domain of human CALR. The following buffers were used in these spectroscopic studies: 2mM HEPES, 200mM NaCl, pH=7.0 supplemented with either 2mM EDTA (Apo), 50μM CaCl₂, or 1000μM CaCl₂.

A. Far-UV CD spectra measured for the isolated P-domain of human CALR in the absence calcium (black line) or in the presence of 50μM CaCl₂ (red line), or 1000μM CaCl₂ (green line). Spectrum of the P-domain of human CALR completely unfolded by 6.0M urea (pink line) is shown for comparison.

B. Normalized intrinsic fluorescence spectra of human P-CALR in the absence calcium (black line) or in the presence of 50μM CaCl₂ (red line), or 1000μM CaCl₂ (green line), and P-CALR completely unfolded by 6.0M urea (pink line). Spectra were normalized, since there is a dramatic increase in their maximal intensity which makes the visual analysis of their position difficult.

C. Urea-induced unfolding of the apo-form of the P-domain of human CALR (black circles) and the P-domain of human CALR in the presence of 50μM CaCl₂ (red triangles), or 1000μM CaCl₂ (green squares). Urea-induced unfolding was detected by the characteristic red shift of the intrinsic fluorescence spectrum monitored by the urea-induced changes in the F₃₆₀/F₃₂₀ values, where F₃₆₀ and F₃₂₀ correspond to the fluorescence intensity at 360 and 320nm, respectively. Experiments were performed in triplicate, and plot represents averaged F₃₆₀/F₃₂₀ values. Calcium does not affect the conformational stability of the P-domain.

D. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the apo-P-CALR. Spectra were measured in the presence of 0.00, 0.47, 0.83, 1.42, 2.01, 2.37, 2.72, 3.32, 4.04, 5.01, and 5.74 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

E. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the P-CALR in the presence of 50μM CaCl₂. Spectra were measured in the presence of 0.00, 0.41, 0.62, 0.91, 1.23, 1.58, 2.00, 2.32, 2.53, 2.79, 2.99, 3.32, 4.05, 4.76, 5.31, and 6.01 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

F. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the P-CALR in the presence of 1000μM CaCl₂. Spectra were measured in the presence of 0.00, 0.59, 1.18, 1.65, 2.01, 2.37, 2.84, 3.32, 3.92, 4.52, 5.25, and 5.98 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

Figure 4. Analysis of the effect of calcium on structural properties and conformational stability of the isolated C-domain of human CALR. The following buffers were used in these spectroscopic studies: 2mM HEPES, 200mM NaCl, pH=7.0 supplemented with either 2mM EDTA (Apo), 50μM CaCl₂, or 1000μM CaCl₂.

A. Far-UV CD spectra measured for the isolated C-domain of human CALR in the absence calcium (black line) or in the presence of 50μM CaCl₂ (red line), or 1000μM CaCl₂ (green line). Spectrum of the C-domain of human CALR completely unfolded by 6 M urea (pink line) is shown for comparison.

B. Intrinsic fluorescence spectra of the isolated human C-CALR in the absence calcium (black line) or in the presence of 50μM CaCl₂ (red line), or 1000μM CaCl₂ (green line), and P-CALR completely unfolded by 6.0M urea (pink line).

C. Urea-induced unfolding of the apo-form of the C-domain of human CALR (black circles) and the C-domain of human CALR in the presence of 50μM CaCl₂ (red triangles), or 1000μM CaCl₂ (green squares). Since the position of the intrinsic fluorescence spectrum of the isolated C-domain is red-shifted (352 nm) and is affected by the addition of CaCl₂ or high urea concentrations, urea-induced unfolding of this domain was detected by the urea-induced increase in the fluorescence intensity. Experiments were performed in triplicate, and plots represents averaged F_max values. Calcium does not affect the conformational stability of the C-domain.

D. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the isolated apo-C-CALR. Spectra were measured in the presence of 0.00, 0.59, 0.94, 1.35, 2.01, 2.72, 3.72, 4.88, and 5.96 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

E. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the C-CALR in the presence of 50μM CaCl₂. Spectra were measured in the presence of 0.00, 0.43. 0.88, 1.46, 1.98, 2.67, 3.62, 4.12, 5.08, and 5.92 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

F. Effect of the increasing urea concentration on the intrinsic fluorescence spectrum of the C-CALR in the presence of 1000μM CaCl₂. Spectra were measured in the presence of 0.00, 0.59, 1.54, 2.37, 3.32, 4.36, and 5.37 M urea. Increase in urea concentration causes red shift and increase in fluorescence intensity.

Figure 3B shows that the intrinsic fluorescence of the isolated P-domain both in the absence or presence of CaCl₂ was slightly blue-shifted in comparison with that of the urea-unfolded polypeptide (346 nm vs. 352 nm, respectively). This suggested the presence of some protection of tryptophan residues of the isolated P-domain from solvent, which, however, was not affected by the addition of calcium to any significant degree. This conclusion is further supported by the results of the acrylamide-induced fluorescence quenching experiments, which showed that the isolated apo-P-CALR and P-CALR in the presence of 50 and 1000 μM Ca²⁺ in solutions containing 200 mM NaCl where characterized by the K_SV of 5.68±0.08, 5.41±0.07, and 5.29±0.06 M⁻¹, respectively, whereas the urea-unfolded P-CALR had the K_SV of 19.93±0.06 M⁻¹.

In comparison with the unfolding curves recorded for the full-length protein (see Figure 3A), unfolding of the isolated P-domain detected by the urea-induced changes in its intrinsic fluorescence is characterized by very low cooperativity, which is typical for the unfolding patterns of mostly disordered proteins (Uversky, 2009). Despite its high negative charge, neither P-domain structure (Figures 3A and 3B) nor conformational stability (Figure 3C) are affected by the addition of CaCl₂, indicating the lack of the structure-inducing and/or stabilizing effects of Ca²⁺ binding for the P-domain of CALR. Similar to the full-length CALR, urea-induced unfolding of the isolated P-CALR was accompanied by both a noticeable red shift of the fluorescence spectrum and a large increase in the intrinsic fluorescence intensity.

Figure 4A shows that in agreement with predictions (see below), the far-UV CD spectrum of the isolated C-domain is characterized by a strong negative band near 203 nm typical for the mostly disordered polypeptides but has a pronounced negative band in the vicinity of 217–227 nm suggesting the presence of some ordered secondary structure. Although the CD spectral properties of this domain are minimally affected by Ca²⁺, Figure 4A shows the intensity of the negative band near 203 nm decreases and the intensity of the 217–227 shoulder increases in the calcium-dependent manner, suggesting that the structure of the isolated C-domain is sensitive to calcium, shifting from more disordered structure in the absence of Ca²⁺ to a more ordered form when calcium concentration is raised. This observation is in the line with the previously published study, where a systematic analysis of the various fragments of the C-domain revealed that the addition of calcium mostly affects structure of the C-terminal half of this domain, whereas its N-terminal half was not affected by calcium (Villamil Giraldo et al., 2010). Figure 4B shows that the position of the intrinsic fluorescence spectrum of the isolated C-domain (352 nm) was not noticeably affected by the addition of CaCl₂ or high urea concentrations, suggesting high solvent accessibility of its tryptophan residues in all states analyzed in this study. This conclusion is in line with the outputs of the acrylamide-induced fluorescence quenching experiments, where the lack of significant protection of the tryptophan residues in the isolated apo-C-CALR alone or C-CALR in the presence of 50 and 1000 μM Ca²⁺ or 6 M urea is evidenced by the K_SV of 10.5±0.3, 13.2±0.4, 12.6±0.2, and 15.1±0.3 M⁻¹, respectively. However, Figure 4B shows that the addition of urea caused a detectable increase in the intrinsic fluorescence intensity of this domain, and this phenomenon was used to follow urea-induced changes in its structure. Figure 4C represents the corresponding unfolding curve and shows that the urea-induced unfolding of the isolated C-domain is completely non-cooperative and is not affected by calcium. Spectra used for the generation of these unfolding curves are presented in Figures 4D, 4E, and 4F.

Therefore, our analysis revealed that neither structure nor conformational stability of the P-domain were sensitive to calcium. Although highly disordered C-CALR gained a bit more ordered structure at calcium binding, it still remained mostly disordered, and its conformational stability was not affected by calcium. Taken together, our observations indicate that calcium has a noticeable effect on the structure and conformational stability of the full-length protein, making it somewhat more compact and significantly more stable. Since no detectable changes in the structure and conformational stability of the isolated P-domain was induced by calcium, and since the addition of calcium had minimal effects of the intrinsically disordered nature of the isolated C-domain and did not affect its conformational stability, the calcium-induced changes in the behavior of the full-length CALR can be attributed to the calcium-dependent stabilization of an C-embedded structure, where the portion of the C-domain is incorporated into a jelly-roll-like structure formed by its N-domain.

3.2. Peculiarities of structure and intrinsic disorder of human CALR

3.2.1. Peculiarities of the CALR structure

There are three domains in the mature CALR (residues 18-417; UniProt ID: P27797), a globular N-terminal domain (N-domain) followed by an extended proline-rich P-domain and acidic C-terminal domain (C-domain), spanning through residues 18-197, 198-308, and 309-417, respectively (Chouquet et al., 2011; Michalak et al., 2009). Despite serious efforts no 3D structure is available for the full-length protein, suggesting the presence of significant conformational dynamics in some of its parts. In fact, it was emphasized that the flexible nature of the P-domain and the C-terminal tail preclude crystallization of the full-length CALR (Chouquet et al., 2011).

In agreement with this hypothesis, the use of programs for secondary structure prediction revealed that the P-domain is expected to be devoid of regular secondary structure elements mostly due to the high content of proline residues (Ellgaard et al., 2001). The P-domain is characterized by the presence of multiple copies of types 1 and 2 repeat sequences arranged in a specific ‘111222’ pattern (Ellgaard et al., 2001). Figure 5A represents solution NMR structure of the P-domain of the rat CALR (residues 189–288 in mature protein or residues 206-305 in UniProt ID: P18418) and shows that this domain is characterized by an unusual extended hairpin fold with high conformational flexibility. The most stable part of this flexible domain is its central loop (residues 227–247) closed by an anti-parallel β-sheet containing two very short β-strands, residues 224–226 and 248–250 (PDB ID: 1HHN; (Ellgaard et al., 2001)). There are two additional two-stranded antiparallel β-sheets containing pairs of short β-strands, residues 207–209 and 262–264, and 190–192 and 276–278 (Ellgaard et al., 2001). Since remaining parts of this 101 residue-long domain are mostly unstructured, the P-domain is expected to have 1% of α-helical structure (1 helix; 4 residues) and 11% of β-sheet structure (6 strands; 12 residues) (Ellgaard et al., 2001).

Figure 5. Structural characterization and intrinsic disorder propensity of human CALR.

A. Solution NMR structure of the P-domain of the rat CALR (residues 189–288 in mature protein or residues 206-305 in UniProt ID: P18418) (PDB ID: 1HHN; (Ellgaard et al., 2001)).

B. X-ray crystal structure of the N-domain human of CALR (residues 18-204) and the N-terminal half of the C-domain (residues 302-368) connected by a GSG tripeptide (PDB ID: 3POW; (Chouquet et al., 2011)).

C. Evaluating intrinsic disorder propensity of human CALR (UniProt ID: P27797) by series of per-residue disorder predictors. Disorder profiles generated by PONDR^® VLXT, PONDR FIT, PONDR^® VL3, PONDR^® VSL2, IUPred_short and IUPred_long are shown by black, red, green, yellow, blue, and pink lines, respectively. Light cyan line shows the mean disorder propensity calculated by averaging disorder profiles of individual predictors. Light pink shadow around the PONDR^® FIT shows error distribution. In these analyses, the predicted intrinsic disorder scores above 0.5 are considered to correspond to the disordered residues/regions. Positions of the N-domain (residues 1-180 in mature form that correspond to the residues 18-197 in UniProt ID: P27797) and C-domain (residues 293-400 in mature form that correspond to the residues 309-417 of the in UniProt entry P27797) are shown as gray shaded areas.

Also, the X-ray crystal structure was solved for the construct of the human CALR containing the N-domain (residues 18-204) and the N-terminal half of the C-domain (residues 302-368) connected by a GSG tripeptide instead of the extended P-domain (PDB ID: 3POW; (Chouquet et al., 2011)). Figure 5B represents structure of this construct and shows that it is characterized by a jelly-roll fold formed by one convex and one concave anti-parallel β-sheets (Chouquet et al., 2011). The part of the C-domain being integrated into the globular domain, provides two central strands to the β-sandwich structure and adds a long kinked C-terminal α-helix (Chouquet et al., 2011). Based on these observations, a structural model of CALR was proposed, according to which the N-domain and a part of C-domain jointly form a globular structure, from which a hook-like arm of a β-stranded hairpin of P-domain protrudes, with the tip of this arm possessing the binding site for the protein partners (e.g., oxidoreductase ERp57; (Frickel et al., 2002)), whereas the acidic C-terminal domain, also protruding from the globular body, contains a large number of negatively charged amino acids and acts as a high-capacity Ca²⁺ storage, within the ER lumen (Michalak et al., 2009; Wijeyesakere et al., 2011).

3.2.2. Peculiarities of the intrinsic disorder distribution in CALR

Therefore, currently available structural data suggest that the overall CALR structure can be viewed as a globular body with two flexible and functionally different arms corresponding to the P-domain and C-tail. In agreement with these structural data and model, Figure 5C represents the results of the multiparametric analysis of the intrinsic disorder predisposition of human CALR evaluated by four algorithms from the PONDR family, PONDR-FIT, PONDR^® VLXT, PONDR^® VSL2, and PONDR^® VL3 (X. Li et al., 1999; Obradovic et al., 2005; K. Peng et al., 2006; K. Peng et al., 2005; P. Romero et al., 2001; Xue et al., 2010), as well as by the IUPred web server that predicts probability of a query protein to have short and long IDPRs (Dosztanyi et al., 2005). We also estimated the mean disorder propensity of human CALR which was calculated by averaging disorder profiles of individual predictors. The use of the consensus of multiple predictors for evaluation of intrinsic disorder is motivated by empirical observations that this approach usually increases the predictive performance compared to the use of a single predictor (Fan & Kurgan, 2014; Z. Peng & Kurgan, 2012; Walsh et al., 2015). In fact, 38.25% and 43.65% of the human CALR residues are predicted to be disordered (have disorder scores above the 0.5 threshold) based on the averaging the outputs of the PONDR^® VLXT, PONDR^® VL3, PONDR^® VSL2, PONDR^® FIT, IUPred_short, and IUPred_long predictors and by the MobiDB platform that aggregates the outputs from ten disorder predictors (http://mobidb.bio.unipd.it/) (Di Domenico et al., 2012; Potenza et al., 2015), respectively. This clearly places CALR into the category of highly disordered proteins, if the classification of proteins based on their content of predicted disordered residue (CPDR) values is used, where proteins are considered as highly ordered, moderately disordered, or highly disordered if their CPDR<10%, 10% ≤ CPDR<30%, or CPDR≥30%, respectively (Rajagopalan, Mooney, Parekh, Getzenberg, & Kulkarni, 2011)

Figure 5C shows that the N-domain is expected to be mostly ordered, whereas P-domain and the C-terminal tail of C-domain are predicted to be highly disordered. Importantly, the part of the C-domain which was co-crystallized with the N-domain (residues 302-368, see Figure 5B) is predicted to be mostly ordered. Therefore, our computational data where the mostly disordered nature of the C-terminal part of CALR that includes P- and C-domains is shown based on the peculiarities of the amino acid sequence of CALR provides support and explanation for the “two armed body” structural model of this protein.

3.2.3. Potential functional roles of disorder in CALR

To further illustrate abundance and functionality of intrinsic disorder in human CALR, we used the output of the D²P² platform (http://d2p2.pro/) (Oates et al., 2013), which shows results of the intrinsic disorder predisposition analysis of a target protein using several common disorder predictors, such as IUPred (Dosztanyi et al., 2005), PONDR^® VLXT (P. Romero et al., 2001), PrDOS (Ishida & Kinoshita, 2007), PONDR^® VSL2B (Obradovic et al., 2005; K. Peng et al., 2006), PV2 (Oates et al., 2013), and ESpritz (Walsh et al., 2012). In addition, the D²P² profile of a query protein contains information on the location of various posttranslational modifications and predicted disorder-based protein binding sites. Figure 6A represents the D²P² profile of human CALR that provides further support to the notion that this protein is expected to be highly disordered. Furthermore, Figure 6A shows that CALR has multiple sites of various posttranslational modifications, such as phosphorylation, acetylation, and glycosylation.

Figure 6. Computational analysis of functional disorder in human CALR.

A. Intrinsic disorder propensity and some important disorder-related functional information generated for human CALR (UniProt ID: P27797) by the D²P² database (http://d2p2.pro/) (Oates et al., 2013). Outputs of nine disorder predictors are shown by differently colored bars. The green-and-white bar in the middle of the plot shows the predicted disorder agreement between nine predictors, with green parts corresponding to disordered regions by consensus. Yellow bars show the locations of the predicted disorder-based binding sites (molecular recognition features, MoRFs), whereas colored circles at the bottom of the plot show location of various PTMs, such as phosphorylation (red), acetylation (yellow), and ubiquitination (purple).

B. Analysis of the interactivity of human CALR by STRING computational platform that produces the network of predicted associations for a particular group of proteins (Szklarczyk et al., 2011). In the corresponding network, the nodes correspond to proteins, whereas the edges show predicted or known functional associations. Seven types of evidence are used to build the corresponding network, where they are indicated by the differently colored lines: a green line represents neighborhood evidence; a red line - the presence of fusion evidence; a purple line - experimental evidence; a blue line – co-occurrence evidence; a light blue line - database evidence; a yellow line – text mining evidence; and a black line – co-expression evidence (Szklarczyk et al., 2011). In our analysis, the most stringent criteria were used for selection of interacting proteins by choosing the highest cut-off of 0.9 as the minimal required confidence level.

Multiple studies revealed that some IDPs/IDPRs are able to undergo at least partial disorder-to-order transitions upon binding, and such binding-induced folding is crucial for recognition, regulation, and signaling functions of these proteins (Dunker et al., 2001; Dyson & Wright, 2002, 2005; Mohan et al., 2006; Oldfield et al., 2005; Z. Peng et al., 2013; Uversky, 2013b; Uversky et al., 2000; Vacic, Oldfield, et al., 2007; Wright & Dyson, 1999). Among these disorder-based functional sites are short order-prone motifs typically located within the long IDPRs. Such motifs are able to undergo disorder-to-order transition as a result of binding to a specific partner. These motifs are known as molecular recognition features (MoRFs) that can be identified computationally (Cheng et al., 2007; Oldfield et al., 2005). In our study, we used a specialized algorithm, ANCHOR, to find potential disorder-based protein binding sites that can fold as a result of interaction with specific partners (Dosztanyi, Meszaros, & Simon, 2009; Meszaros, Simon, & Dosztanyi, 2009). Figure 6A represents results of this analysis and shows that human CALR contains five MoRFs identified by ANCHOR, residues 170-176; 198-204, 259-269, 304-352, 409-417, distributed over the C-terminal part of N-domain, P-domain, and C-domain, suggesting that intrinsic disorder in CALR can be used for protein-protein interactions. This hypothesis is further supported by the results of the pre-CALR analysis by MoRFpred, which is an alternative computational tool for accurate evaluation of disorder-based binding sites undergoing disorder-to-order transition at interaction with specific partners (Disfani et al., 2012). Results of this analysis shows that the pre-CALR contains a set of relatively short regions that have a potential to fold at interaction with binding partners. These regions are residues 1-10, 19-25, 50-55, 76-80, 199-206, 221-225, 236-242, 247-252, 312-314, 332-334, 378-389, and 409-417. Regions in this list shown by bold font either coincide or are contained within the ANCHOR-predicted MoRFs (residues 198-204, 304-352, and 409-417). Curiously, although ANCHOR-predicted MoRFs in regions 170-176 and 259-269 are not identified by MoRFpred, these two regions have noticeably higher average MoRFpred scores when compared to adjacent (in the chain) regions, suggesting that do have increased binding potential. Therefore, ANCHOR and MoRFpred data indicate that CALR can be a highly promiscuous binder that utilizes intrinsic disorder to interact with a broad range of binding partners.

In agreement with this hypothesis, Figure 6B represents the STRING-based interactome of the human CALR and indicates that this protein indeed has several putative binding partners. Note that data shown in this plot were obtained using the most stringent confidence level of 0.9. Furthermore, MobiDB lists 2,791 proteins that interact with human CALR and that are characterized by the levels of intrinsic disorder ranging from 0% to 100% (http://mobidb.bio.unipd.it/entries/P27797). Overall, the vast majority of proteins interacting with CALR are expected to be ordered. In fact, 24, 75, and 2,692 of CALR partners are predicted to be highly or moderately disordered or highly ordered, possessing CPDR≥30%, 10% ≤ CPDR<30%, and CPDR<10%, respectively. Although indicated number of CALR binding partners seems to be very high, one should keep in mind that together with another lectin chaperone, calnexin (CNX), CALR is known to interact with practically all glycoproteins investigated to date (Ellgaard & Frickel, 2003) that makes it indeed a highly promiscuous binder.

4. CONCLUSIONS

Data reported in this study indicate that human CALR contains significant amount of intrinsic disorder mostly concentrated in its P- and C-domains. These observations are in line with the previous notion that the flexible nature of the P-domain and of the C-terminal tail has precluded crystallization of the full-length human CALR so far (Chouquet et al., 2011). Our study represents results of careful structural and conformational analyses of the full-length and the isolated P- and C-domains of human CALR that provide some mechanistic insights on how Ca²⁺ may affect exposure of both N-terminal and C-terminal halves of CALR (N-CALR and C-CALR, respectively) in vitro and in vivo. In fact, in the existing structural models, CALR is depicted as a rigid body (comprised of the N-CALR and N-terminal part of the C-domain) with two flexible arms (P-domain and C-tail). However, whether the individual isolated domains may sense Ca²⁺ has not been established as yet. Our analysis revealed that neither structure nor conformational stability of the P-domain were sensitive to Ca²⁺. Similarly, although the highly disordered C-CALR gained some level of ordered structure after Ca²⁺-binding, it still remained mostly disordered, and its conformational stability was not affected by Ca²⁺. These observations indicate that Ca²⁺ has a noticeable effect on the structure and conformational stability of the full-length protein, making it somewhat more compact and significantly more stable. Since no detectable changes in the structure and conformational stability of the isolated P-domain was induced by Ca²⁺ and its addition had minimal effects of the intrinsically disordered nature of the isolated C-domain without affecting its conformational stability, the Ca²⁺-induced changes in the behavior of the full-length CALR can be attributed to Ca²⁺-dependent stabilization of an C-embedded structure, where the portion of the C-domain is incorporated into a jelly-roll-like structure formed by its N-domain. This close spatial relationship between the two domains make their conformations sensitive to Ca²⁺ and determines which of the five MoRF motifs of the protein would be exposed and may exert their binding functions.