Self-Oligomerization Is Essential for Enhanced Immunological Activities of Soluble Recombinant Calreticulin

We have recently reported that calreticulin (CRT), a luminal resident protein, can be found in the sera of patients with rheumatoid arthritis and also that recombinant CRT (rCRT) exhibits extraordinarily strong immunological activities. We herein further demonstrate that rCRT fragments 18–412 (rCRT/18-412), rCRT/39-272, rCRT/120-308 and rCRT/120-250 can self-oligomerize in solution and are 50–100 fold more potent than native CRT (nCRT, isolated from mouse livers) in activating macrophages in vitro. We narrowed down the active site of CRT to residues 150–230, the activity of which also depends on dimerization. By contrast, rCRT/18-197 is almost completely inactive. When rCRT/18-412 is fractionated into oligomers and monomers by gel filtration, the oligomers maintain most of their immunological activities in terms of activating macrophages in vitro and inducing specific antibodies in vivo, while the monomers were much less active by comparison. Additionally, rCRT/18-412 oligomers are much better than monomers in binding to, and uptake by, macrophages. Inhibition of macrophage endocytosis partially blocks the stimulatory effect of rCRT/18-412. We conclude that the immunologically active site of CRT maps between residues 198–230 and that soluble CRT could acquire potent immuno-pathological activities in microenvironments favoring its oligomerization.


Introduction
Calreticulin (CRT) is a 46 kDa Ca 2+ -binding glycoprotein in the endoplasmic reticulum of eukaryotic cells [1][2]. It folds into 3 domains including a lectin-like globular N domain (amino acid residues , a proline-rich P domain (residues 198-308) and a Ca 2+ -binding C domain (residues 309-412) [3][4][5]. CRT can also appear at the surface of various types of cells exhibiting multiple biological functions [6][7][8][9][10][11][12]. Recently it has been shown that soluble CRT is present in the sera of patients with rheumatoid arthritis and with SLE [13][14] and that natural CRT (nCRT), isolated from human or mouse tissues, can directly activate macrophages in vitro [15]. Additionally, rCRT/39-272, a prokaryotically-expressed murine CRT fragment covering amino acid residues 39-272 fused with an N-terminal His-tag, was extraordinarily potent in activating B cells and macrophages in vitro and also in eliciting specific Ab production in mice [16]. This recombinant polypeptide also exhibited potent adjuvanticity, effectively assisting IgG production against conjugated target Ags with or without T cell help [16][17]. However, molecular mechanisms underlying this phenomenon are far from clear. Recent X-ray crystographical studies by Kozlov [18,19]. The sequence of rCRT/39-272 encompasses most of the globular N domain (aa residues , and we have previously shown that it possess lectin-like activity (selective binding with polysaccharides including carrageenan, alginic acids, and hyaluronic acids in ELISAs) [16], implying that the prokaryotically expressed recombinant polypeptide retained the lectin activity of CRT. It is of interest to determine if destroying the carbohydrate binding and/or peptide-binding sites (by deleting first half of the N domain sequence) would also abolish the potent immunological activities observed in rCRT/ 39-272. Additionally, the fact that rCRT/39-272 is substantially more potent than nCRT in activating macrophages(see below) raised concerns regarding the possibility of LPS contamination in the E. coli.-expressed recombinant product. The ''LPS contamination'' hypothesis suggests tight binding between bacterial LPS and rCRT and also that, due to a synergistic effect, the LPS-CRT complex is a more potent immune activator than free LPS and rCRT alone. Based upon the observation that interaction between the N and C domains of CRT influences its structural stability as well as functional activity [20], a ''C-domain deletion'' hypothesis has also been postulated, suggesting that deletion of the C-domain (as in rCRT/39-272) leads to exposure of an immunologically active site (IAS) in the N and/or P domains of CRT. The present study compares the biochemical characteristics of nCRT and a series of truncated rCRT polypeptides and investigates the molecular mechanisms underlying the potent immunological activities of soluble rCRT. The results arising from this study have important implications for our understanding of the potential role of soluble CRT in immunopathological conditions.

Biochemical Properties of rCRT and nCRT
Native CRT was extracted from mouse livers by (NH 4 ) 2 SO 4 precipitation followed by ion exchange chromatography on a DEAE-A50 column using a linear gradient of 280-500 mM NaCl for elution. Samples of the eluted fractions were assayed by SDS-PAGE (Fig. 1A), which showed that the eluent between 360-380 mM NaCl contained protein bands of the expected molecular weight for nCRT (55 kDa), which showed approximately 90% homogeneity judging by density of the major band in Coomassie blue (CBB)-stained gel (Fig. 1C).
Contaminating RPL2 does not Contribute to rCRT Activity When preparing rCRT/18-412, Cp32 was a relatively persistent contaminant protein (Figs. 1 & 3). To characterize this protein, the Cp32 band was sliced out of SDS-PAGE gels and then (i) used to immunize C57/BL6 mice for preparation of specific antisera; and (ii) subjected to Q-TOF mass spectrometry analysis for sequence identification. The resultant antisera (Cp32-Abs) were able to recognize the immunizing protein band, but not the rCRT-60 kDa and rCRT-46 kDa bands, in WB (Fig. 3B). The MS result identified Cp32 as bacterial 50S ribosomal protein L2 (RPL2), which was confirmed by positive recognition of recombinant RPL2 (rRPL2, commercially available) by Cp32-Abs (Fig. 3B). It is noteworthy that rRPL2 was unable to activate macrophages, as evidenced in NO 2 2 production assays (Fig. 3C), and also that rRPL2 was not recognized by CRT-Abs (Figs. 4C & 4D). These results exclude the possibility that the contaminant Cp32 could make a substantial contribution to the immunological activities of rCRT. 2 in the culture supernatant was then determined using Griess Reagent and the results are expressed as mean NO 2 2 concentration (mM) 6 SD. LPS-based ELISAs were performed for detection of LPS binding with CRT (B) or lactoferrin (C). Lf or nCRT (2 mg/ml) were added to wells in polyvinyl plates pre-coated with LPS (10 mg/ml), with BSA as a negative control. Combination of polyclonal rabbit Abs against CRT, or lactoferrin, and HRP-labeled goat-anti-rabbit IgG was used for detection with OPD as substrate. The results are expressed as absorbance at OD492 nm6SD. For sinergy analysis (D), freshly isolated mouse peritoneal macrophages were stimulated with nCRT (0.3-30 mg/ml) in the presence, or absence, of LPS (0.1 ng/ml) for 24 h. Cells in medium alone (Medium) or stimulated with LPS (0.1 ng/ml) alone (LPS) were included as controls. TNF-a in the culture supernatant was then quantitated using an ELISA kit and the results are expressed as mean concentration (pg/ml)6SD. These are representatives of 3 independent experiments. doi:10.1371/journal.pone.0064951.g002

OrCRTs Activate Macrophages in an Endocytosisdependent Manner
In order to investigate the mechanisms for the relationship between rCRT oligomerization and its potent immunological activities, fractionated OrCRTs and MrCRT-60 kDa (monomeric rCRT/18-412) were conjugated with FITC and then compared for ability to stain macrophages. After 30 min incubation at 4uC, FITC-OrCRTs showed stronger binding to murine macrophages than FITC-MrCRTs, as evidenced by flow cytometric analysis and confocal laser scanning microscopy (Fig. 7A&B). Substantially more FITC-OrCRTs than FITC-MrCRTs were endocytosed by macrophages after 30 min incubation at 37uC (Fig. 7B). Moreover, monodansylcadaverine (MDC), an endocytosis inhibitor [21], partially suppressed NO 2 2 production by macrophages under stimulation with OrCRTs, while LPS-triggered macrophage activation was unaffected by the same treatment (Fig. 7C).

Discussion
In this study, we have examined different hypotheses (i.e. Cdomain deletion hypothesis, LPS contamination hypothesis, RPL2 hypothesis and oligomerization hypothesis) for explanation of the much stronger immunogenicity and immunostimulatory activities of rCRT than nCRT. Our data strongly suggests that selfoligomerization of the rCRT polypeptides is a key factor for their strong immunological activities. As summarized in Table 1, all four rCRT fragments containing residues 120-250 (including rCRT/18-412, rCRT/39-272, rCRT/120-250 and rCRT/120-308) self-oligomerized and exhibited potent macrophage activating ability in vitro and strong immunogenicity in vivo, while nCRT, which existed mainly in monomeric form, showed only modest stimulatory activities towards macrophages and was non-immunogenic in mice (Figs. 1 & 3). Moreover, fractionated rCRT/18-412 oligomers were 50-100 folds more active than MrCRTs in activating macrophages in vitro (Fig. 5). The N domain polypep-tide (rCRT/18-197) also formed higher-molecular-weight oligomers (Fig. 6A), but it is almost completely inactive in terms of stimulating macrophages in vitro (Fig. 6B) and inducing Ab responses in vivo (Table 1). Clearly, oligomerization is necessary but not sufficient to arm the rCRT polypeptides with potent immunological activities.
In general, oligomerized (aggregated) proteins are more immunogenic than their monomeric counterparts. However, the immunogenicity of OrCRTs is by far the most impressive and not comparable by other protein aggregates. Even in the absence of any adjuvant, minute amount (1 ng/mouse) rCRT/18-412 or rCRT/39-272 can elicit strong IgG responses in mice ( Fig. 2; Ref. 16). Most, if not all, other protein antigens are unable to induce IgG production in T-cell-deficient nude mice, yet rCRT/39-272 and rCRT/18-412 could do so relatively efficiently [16,22]. Moreover, the potent adjuvanticity of the rCRT polypeptides is also quite phenomenal. For instance, rCRT/39-272 (mostly in oligomeric forms) is able to assist the production of IgG Abs against fused target proteins or conjugated polysaccharides in healthy mice or T-cell-deficient nude mice [17,22]. Dimerization/oligomerization of soluble CRT was also observed by previous investigators. For example, Jorgensen et al documented that shielding of the free Cys163 in the N domain is the main reason that nCRT exists mainly in monomeric form under physiological conditions. Under partial unfolding conditions such as high temperature or low PH, however, the free Cys could be exposed and subsequently help CRT oligomerization [23]. nCRT (isolated from human placenta) formed homodimers and higher-molecular-weight species through disulfide bonding as well as non-covalent association, and that oligomerized nCRT showed higher binding affinity to peptides and denatured proteins [23]. In the case of prokaryotically expressed rCRT polypeptides (the folding of which may differ from nCRT), all Cys residues in sequence could contribute to its olimerization, although it might not be absolutely necessary that all 3 Cys residues have to be available for inter-molecular cross-linking at the same time. Mancino and colleagues illustrated that self-oligomerized rCRT could better assist HLA folding in vitro [24]. There are 3 conserved cysteine residues in the amino acid sequence of CRT. Cys105 and Cys137 form intramolecular disulfide bonds, while Cys163 is free [25][26]. It is likely that rCRT polypeptides are unable to form appropriate intra-molecular disulfide bonds like in nCRT, thereby allowing all 3 cysteine residues to participate in self-oligomerization, although formation of higher-molecularweight oligomers could also occur through non-covalent association of CRT [23,27]. CRT is considered one of the heat shock proteins (HSPs) that share many immunological and biochemical activities [28][29]. Interestingly, self-oligomerization also occurs to other HSPs such as GRP94 and HSP90, which is likely associated with their chaperone function [30][31][32].
Koslov et al and Chouquet et al have recently solved the crystal structure of the lectin site as well as a peptide-binding site in the CRT N domain [18,19], which apparently play important roles in the physiological function of CRT. However, our data maps the IAS of CRT to a region of 80 residues between aa150-230 (Fig. 6). As rCRT-N was almost completely inactive in functional assays (Fig. 6, Table 1), the IAS may be narrowed down further to aa198-230 in the P domain, although a series of truncated synthetic peptides covering this region would be needed for a concrete conclusion. Interestingly, this sequence of 30 amino acid residues contains the RA shared epitope (SE)-binding site (residues 217-224) of CRT recently mapped by Ling et al. [33]. Such coinciding results from 2 independent groups further emphasize the importance of the aa198-230 region of the P domain to the immunological activities of CRT. It has been documented that the CRT P domain adopts a hairpin-shaped structure, its sequence is consisted of 3 copies of a repeat motif (type 1: IXDPXXXK-PEDWD) followed by 3 copies of another repeat motif (type 2) [34][35][36]. Interestingly, the SE-binding site almost completely overlaps the type 1 motif [33]. It is reasonable to suggest that the type 1 motif might also represent the core of IAS of CRT, responsible for direct biding with activation receptors on the surface of immune cells. Functional comparison data showed that rCRT/120-308 and rCRT/120-250 (possessing 3 type 1 repeats, the latter without type 2 repeats) are 10 times more active than rCRT/120-230 (with 2 type 1 repeats, no type 2 repeats) in activating macrophages (Table 1, Fig. 6B), implying that (i) type 2 repeats do not contribute to the immunological activity of the molecule; and (ii) the presence of 3 copies of the type 1 motif is of crucial importance for the potent immunological function of OrCRTs. It can be envisaged that oligomerization of CRT multiplies its binding avidity to immune cells with receptors for the type 1 motif, thereby enabling it to deliver stimulatory signals to the cells in a highly efficient manner. We further predict that the high-avidity binding of OrCRTs to macrophages may easily trigger their uptake process. Indeed, OrCRTs can activate macrophages in an endocytosis-dependent pathway (Fig. 7). Perhaps OrCRTs could use certain intracellular sensors to deliver immunostimulatory activity to the responding immune cells. An example of an intracellular sensor for endocytosed polymeric proteins is the NOD-like receptor (NLR) protein NLRP3, a key component of the NLRP3 inflammasome and an important intracellular sensor for microbial ligands and endogenous danger signals [37]. Masters and colleagues demonstrated that soluble oligomers of islet amyloid polypeptide (IAPP), a protein that forms amyloid deposits in the pancreas during type 2 diabetes, could be endocytosed and trigger the NLRP3 inflammasome and generate mature IL-1b [38].
Finally, the region of aa150-230 of CRT has a 98% homology between mouse and human. It would be of interest to examine if there is a functional relationship between the SE binding site and the IAS of CRT. Irrespective of such a relationship, oligomerization might occur to extracellular CRT released by tissue cells thereby converting CRT into a highly active form, which may play important roles in the development and pathogenesis of autoimmune disorders in humans.

Materials and Methods
Purification of nCRT nCRT was purified from mouse livers using a modification of a previously described methods [39,40]. Briefly, fresh mouse liver cells (erythrocytes depleted) were collected and centrifuged at 1200 rpm for 5 min. The cell pellet was lysed in 3 volumes of lysis buffer (1% Triton-X 100, 0.2 mM PMSF in PBS) for 30 min on ice, followed by centrifugation at 35,000 g for 60 minutes. The supernatant was then precipitated using (NH 4 ) 2 SO 4 and the final precipitate dissolved in binding buffer (150 mM NaCl, 20 mM Tris, PH7.4) followed by dialysis against this buffer. The sample was applied to a DEAE Sephadex A50 column (1062 cm, GE Healthcare, US) which was then sequentially washed with binding buffer and washing buffer (280 mM NaCl, 20 mM Tris, PH7.4) at 1 ml/min to remove contaminating proteins. The fractions were eluted with a linear salt (280-500 mM NaCl) gradient.

Western Blotting
The separated protein bands in SDS-PAGE gels were electroblotted onto PVDF membranes, at a constant current of 250 mA in transbuffer (50 mM Tris, pH 8.0, containing 0.192 M glycine and 20% methanol), using a Bio-Rad Trans-Blot Cell. The strips were incubated for 1 hr at room temperature in blocking buffer (TBS containing 5% nonfat milk), followed by a overnight incubation at 4uC with constant agitation in indicated antibody diluted in blocking buffer. After 3 washes with TBS containing 0.05% Tween 20, strips were incubated for 1 hr with HRPconjugated secondary antibody (Southern Biotechnology Associates Inc., USA) and visualized using the ECL detection system as recommended by the manufacturer (Applygen Technologies Inc., Beijing, China).

ELISAs
LPS-based and CRT-based ELISAs were as previously described [41]. Briefly, ELISA plates were coated at 4uC overnight with rCRT or LPS and subsequently incubated with blocking solution (1% BSA in PBS) for 2 hrs at 37uC. The wells were washed five times with PBS containing 0.05% Tween 20 (PBS-T) prior to incubation at 37uC with 100 ml of diluted mouse sera or with indicated protein (nCRT and lactoferrin) followed by corresponding antibody in triplicate. After 5 washes with PBS-T, the plates were further incubated with HRP-labeled goat-antimouse or goat-anti-rabbit IgG Abs (Southern Biotechnology Associates Inc., AL., USA) for 1 hr at 37uC. The reaction was developed with 100 ml of o-phenylenediamine (OPD, Sigma) for 5 min and stopped with 100 ml 2 M H 2 SO 4 . Optical density (OD) was measured at 492 nm in an ELISA spectrophotometer (Titertek Multiscan Plus MK II; ICN Flow Laboratories, Irvine, UK).

Cells and Tissue Culture
All cells were cultured in complete R10 medium: RPMI-1640 supplemented with 10% (v/v) fetal bovine serum (Hyclone, USA), penicillin/streptomycin (100 U/ml), L-glutamine (2 mM), and b-ME (5610 25 M). For preparation of mouse peritoneal macrophages, mice were injected i.p. with 3% thioglycollate (1 ml/ mouse) and the macrophages retrieved from the peritoneum 3 days later using a syringe. The resultant cells were.90% positive for F4/80 marker, as determined by FACS analysis. Relative macrophage stimulation activity (RMSA) of CRT is estimated using nCRT as reference, which is able to induce TNF-a production by macrophages in vitro at a concentration of 10 mg/ml or above (see Fig. 5A). The listed results are based on several batches of independent experiments including Fig. 5A. b) Ability to elicit specific IgG responses in healthy BALB/c mice after s.c. immunization, in the absence of adjuvant, with 100 mg protein and a booster immunization with 50 mg protein a fortnight later. The mice were monitored for up to 28 days after the second immunization. ND: not detected; -: not immunogenic; +: strong humoral response; 2/+: weak response.

Macrophage Activation Assays
Freshly prepared C57/BL6 mouse peritoneal macrophages (1.5610 5 cells/well) were stimulated with rCRT fragments, or nCRT, or LPS, in R10 medium in 96-well tissue culture plates for 24 hrs in a 5% CO 2 incubator at 37uC. The concentration of TNF-a in the culture supernatant was determined using ELISA kits (Biolegend, San Diego, USA) following the manufacturer's instructions. The concentration of NO 2 2 in the supernatant was determined by Griess Reagent. Standard curves were established using NaNO 2.

Mice and their Immunization
Female C57BL/6 mice between the age of 6-8 weeks were purchased from the Model Animal Research Center, Nanjing, China. All animals were maintained under specified-pathogen-free (SPF) conditions and animal usage was conducted according to protocols approved by the Soochow University Institutional Animal Care and Use Committee. For immunization with rCRT, mice were immunized s.c. at the base of the tail with 100 mg protein in total 100 ml PBS. When booster immunization was needed, 50 mg of protein in 200 ml PBS was injected intraperitoneally (i.p.). For immunization with PAGE gel slices containing Cp32, the band was cut out with a razorblade and then frozen in liquid nitrogen and emulsioned with CFA, which was then injected s.c. into female C57/BL6 mice. Serum samples were collected by tail bleeding, aliquoted and kept at 220uC until use.

Separation of OrCRTs and MrCRTs
A Sephadex G-75 (GE Healthcare, US) column of 8062 cm was employed. 5 ml of rCRT/18-412 at 10 mg/ml was loaded into the column, followed by elution with 0.9% NaCl at 20 ml/h and collected every 2 ml.

FITC Labeling of Proteins
Protein was labeled by FITC using FluoroTag TM FITC Conjugation Kit (Sigma, US) according to the manufacturer's instructions. In brief, 1 mg of protein was dialyzed against 0.1 M Na 2 CO 3 , pH 9.5 at 5 mg/ml. Fluoresceine isothiocyanate (FITC, Sigma) was dissolved in the same buffer with DMSO at 1 mg/ml and 50 ml of this solution was added to the protein. The sample was gently mixed for 2 h at RT. Separation of labeled protein from unbound FITC was performed on a Sephadex G-25 column.
Flow Cytometric Analysis 10 6 freshly isolated peritoneal macrophages were stained with APC-anti-F4/80 and then incubated with 15 mg/ml of FITC-OrCRT, FITC-MrCRT or FITC-OVA for 30 min at 4uC. The binding of CRT by macrophages was determined by flow cytometry (BD FACS Canto II, US).

Confocal Laser Scanning Microscopy
Macrophages (1.5610 5 /well) were plated onto poly-l-lysine coated glass slides and allowed to adhere. The cells were then incubated in a total volume of 200 ml 0.5% BSA in PBS with 15 mg/ml FITC-OrCRT, FITC-MrCRT or FITC-OVA for 30 min at 4uC or 37uC. After washing to remove unbound protein, macrophages were fixed in 1% paraformaldehyde and stored at 4uC until microscopic analysis. Finally, 1% of glycerol was added and slides were counterstained with 1 mg/ml DAPI. The cells were imaged with a Nikon confocal microscope system A1.

Statistical Analysis
All experiments were repeated at least 3 times and the results are expressed as mean6standard deviation of the mean (SD). Statistical analysis was performed using the Independent-Samples t test or two-side paired t test between groups using the SPSS 14.0 program (SPSS, Chicago, IL). Differences were considered statistically significant at p,0.05.