A total of 63 randomly selected dog-allergic subjects (35 women [20 Finnish and 15 Spanish], mean age 38.8±11.2 years(data were presented as mean±standard deviation, otherwise indicated), and 28 men [17 Finnish and 11 Spanish] mean age 39.3±9.8 years) presenting with allergic symptoms upon exposure to dog, together with 12 healthy nonatopic individuals as control subjects (9 women [Finnish], mean age 35.9±11.9 years, and 3 men [Finnish], mean age 28.7±1.5 years) were recruited to the study. A patient was classified as allergic to dog if the result of the dog dander UniCAP specific IgE measurement (Pharmacia, Uppsala, Sweden) was >0.35 kU/L (mean IgE response of the dog allergic cohort to dog epithelium extract was 12.5 kU/L [range; 0.38-100 kU/L]) or the result of the skin prick test (SPT) using commercial dog epithelial extract (ALK Abelló, Hørsholm, Denmark) was ≥3 mm (mean; 5.8 mm, range; 3-18 mm). SPTs against nCan f 4 and rCan f 4 were performed to 16 dog-allergic patients together with 6 control subjects (both groups were Finnish) according to European recommendations, as previously described.15,16 The study was approved by the Ethics Committee of the Kuopio University Hospital with the permit number 182/1999, valid until 2015.

Dog dander raw material (mixed breed RME64P, Greer laboratories, Lenoir, NC, USA) was extracted in PBS overnight at 4℃, essentially, as previously described,12 to prepare in-house dog allergen extract. A cocktail of 3 Can f 4 allergen-specific monoclonal antibodies (mAbs) (for the generation of mAbs 26D, 41G, and 48F, see Supplementary Material 1) was coupled to the HiTrap NHS column (GE Healthcare, Amersham Biosciences, Uppsala, Sweden), and the allergen was purified from the extract according to the manufacturer's instructions. The protein concentration of the allergen was determined by the Bio-Rad Protein Assay Kit (Bio-Rad Laboratories, Inc., Hercules CA, USA) using bovine serum albumin as a standard. The purity of the allergen preparation was verified by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE)7 with silver staining.17 For details, see Supplementary Material 1.

For mass spectrometric analysis and N-terminal sequencing, the purified natural Can f 4 allergen was, after desalting by reversed phase chromatography, subjected to MALDI-TOF mass spectrometry and Edman degradation. For further amino acid sequence determination, the protein was reduced, alkylated, and digested with trypsin, and the peptides were separated by reversed phase chromatography. Selected individual peptides were then subjected to MALDI-TOF mass spectrometry and sequencing by Edman degradation (Supplementary Material 1).

Carbohydrate moieties covalently linked to the nCan f 4 allergen backbone were examined with SDS-PAGE using the Glycoprotein Detection Kit (Sigma-Aldrich, St. Louis, MO, USA), according to the manufacturer's instructions.

Molecular cloning of the DNA encoding the Can f 4 sequence into the Pichia pastoris yeast, the production of the recombinant allergen,18 and its purification in both monomeric and multimeric forms are described in Supplementary Material 1. Molecular mass and identity of the purified rCan f 4 was determined with ESI-Quad-TOF by direct infusion. The identity of rCan f 4 was confirmed by LC-MS/MS and subsequent Mascot database search, as previously described for details (Supplementary Materials 1 and 2).

The production and purification of other recombinant proteins of the lipocalin family, dog Can f 1/Can f 2, horse Equ c 1, mouse Mus m 1, cow Bos d 2, human tear lipocalin (TL; lipocalin-1/von Ebner's gland protein), and the recombinant control protein psoriasin have been previously described in our publications.7,14,19,20

Immunization of the BALB/c mice and hybridoma production are described in detail in Supplementary Material 1. In brief, supernatants of the generated hybridomas were tested for reactivity to the 18 kDa component of dog dander (that was later designated as Can f 4) with ELISA and immunoblotting. The antibodies 26D, 41G, and 48F were purified by affinity chromatography on a protein-G column (GE Healthcare) and the isotypes of the antibodies were determined by the Clonacell InstantCHEK One-minute Isotype Kit (Stem Cell, Vancouver, Canada). Protein concentrations were determined by the Bio-Rad Protein Assay Kit using bovine gamma globulin as a standard.

For assessing whether the 3 mAbs specific to Can f 4 recognize distinct epitopes on the allergen, they were first labeled with biotin. Biotinylation was carried out using aminohexanoyl-biotin-N-hydroxysuccinimide ester (AH-BNHS, Zymed Laboratories, Inc., San Francisco, CA, USA) at a 1:10 w/w ratio of AH-BNHS and mAb, according to the manufacturer's instructions. Then, one of the mAbs that was biotinylated was mixed in a test tube with several dilutions of 1 of the 2 other mAbs which were not biotinylated. Next, the solution was added to the ELISA plate coated with nCan f 4 (1 µg/mL). After an incubation of 1 hour at 37℃, the bound biotinylated mAb was detected with Streptavidin-HRP (1:10,000, GE Healthcare). All 6 combinations were tested. The color reaction was developed by the TMB Single Solution reagent (Zymed Laboratories, Inc. San Francisco, CA, USA) and measured at 450 nm. The results were expressed as inhibition percentages.

Experimental dust samples were collected from homes with and without dogs (5 and 4 samples, respectively) by vacuum cleaning the carpet in the living room for 5 minutes. They were extracted overnight in PBS (1:10 w/v, pH 7.4) at 4℃. After centrifugation and sterile filtration, the protein concentration of the extracted samples was determined by the Bio-Rad Protein Assay Kit using bovine serum albumin as a standard.

For measuring the Can f 4 content in the samples, microtiter plates were coated with mAb 26D (5 µg/mL) overnight at 4℃. Then test samples (diluted in PBS containing 1% BSA/0.05% Tween 20), and rCan f 4 standards (0.01 to 2,500 ng/mL) were incubated at 37℃ for 2 hours. After an incubation (1 hour, 37℃) with biotinylated mAb 48F mAb (1 µg/mL), followed by an incubation with Streptavidin-HRP (1:10,000, GE Healthcare, Buckinghamshire, UK, 0.5 hour, room temperature), the color reaction was developed and measured, as described above. To determine the amount of Can f 4 in test samples, a standard curve of absorbance against the log concentration of the rCan f 4 standard was plotted. The results were calculated from the straight part of this curve.

Levels of IgE antibodies to the Can f 4 preparations were measured with indirect ELISAs, as previously described,7 with a few modifications. The Nunc® MaxiSorp microtiter plates (Thermo Fisher Scientific, Roskilde, Denmark) were first coated with nCan f 4, rCan f 4 or multimeric form (mu) of rCan f 4 (5 µg/mL) and then incubated with sera diluted 1:10 for 1 hour at 37℃. The diluent contained 2% of mouse serum (Sigma-Aldrich). The bound IgE was detected by biotinylated mouse anti-human IgE (1:1,000; Southern Biotechnology Associates, Inc., Birmingham, AL, USA) combined with Streptavidin-HRP (1:10,000). The color reaction was developed by the TMB Single Solution reagent. The results were expressed as optical density (OD) values at 450 nm. The cutoff for a positive reaction was defined as the mean OD of the healthy control subjects plus 3 SD.

Patients' IgE reactivity to Can f 4 in the commercial SPT dog dander extract (ALK Abelló) and to the purified allergens nCan f 4 and rCan f 4 was detected with SDS-PAGE immunoblotting, basically as previously reported.21,22 The same technique was used in detecting the reactivity of the mAbs 26D, 41G, and 48F against the commercial extract and the in-house dog dander extract. In brief, the protocol used for all Western blotting experiments was as follows: After SDS-PAGE and protein transfer, the membrane strips with the dog dander extract or the allergens were incubated overnight with sera (diluted 1:10) or each of the mAbs (0.5 µg/mL), followed by incubation for 2 hours with monoclonal horseradish peroxidase-labeled mouse anti-human IgE at 1:2,000 or ZyMax™ rabbit-anti-mouse IgG-HRP (Zymed Laboratories) at 1:5,000, respectively. The reactive bands were visualized with the ECL Western Blotting Detection Reagents (GE Healthcare).

The reactivity of antibodies to the denatured or nondenatured nCan f 4 was analyzed by the dot blot method. The quantification of dot intensities was performed by using the public domain ImageJ software (http://imagej.nih.gov/ij/)23 and the MicroArray Profile.jar algorithm (http://www.optinav.com/download/MicroArray_Profile.jar). For details, see Supplementary Material 1.

Individual serum samples or pooled sera from 6 dog-allergic patients with diverse levels of specific IgE to Can f 4 or the Can f 4-specific mAbs produced were used in ELISA inhibitions, basically as previously described.14 In brief, the tests were performed essentially as the indirect ELISAs but with an additional preincubation (1 hour at 37℃) of the sera or the mAbs with the inhibitor proteins (nCan f 4, rCanf 4 , rCan f 1, rCan f 2, rEqu c 1, rBos d 2, rMus m 1, rTL, or rPsoriasin). In the preincubation phase, the final serum dilution ranged from 1:30 to 1:75 depending on the IgE level. In ELISA inhibitions with the biotinylated mAbs, appropriate dilutions were used. The final concentrations of the inhibitors ranged from 0.3 to 200 µg/mL. The results were expressed as inhibition percentages.

All statistical analyses were performed using GraphPad Prism (GraphPad Software, San Diego, CA, USA) and SPSS (IBM SPSS Statistics 21 I; IBM Corp, Armonk, NY, USA) The two-tailed t test and the Mann-Whitney U test were used for determining statistically significant differences between the groups. For assessing the inter-rater agreement (categorical items) and for analyzing correlations, the Kappa (κ) statistic and the Spearman Rank correlation test were used, respectively.

To generate mAbs specific to Can f 4, a protein fraction with a molecular mass of 12-18 kDa7 was isolated by size exclusion chromatography (SEC) from in-house dog dander extract and used for immunizing mice. Exploiting the 3 mAbs (41G, 26D, and 48F) with reactivity to the 18kDa component in dog dander (Fig. 1), nCan f 4 was isolated from the extract by affinity chromatography. Its high level of purity was verified by silver-stained SDS-PAGE (Fig. E1A in Supplementary Material 2). The glycoprotein staining of the nCan f 4 protein suggested that the protein is nonglycosylated (data not shown).

The cDNA encoding the Can f 4 (GenBank accession number KF192077) was amplified from a mixture of cDNAs derived from dog parotid gland, mandibular gland, and tongue tissues. It was found that the cDNA contains an open reading frame encoding a 174 amino-acid protein. At the beginning, there is a putative 16 amino-acid leader, followed by a 158 amino-acid mature protein. The predicted molecular mass is 17.6 kDa, and the predicted pI is 6.19. Comparison of the amino acid sequence with the previously published one8 revealed 4 differences: Ile36Val, Asp46Glu, Met54Leu, and Ser66Leu (Fig. E2 in Supplementary Material 2). In order to investigate the importance of Can f 4 as an allergen, it was produced as a recombinant protein using the P. pastoris expression system.

We have previously reported that 60% of dog dust-sensitized subjects exhibited IgE reactivity to a band of dog dander extract corresponding to Can f 4 in immunoblotting.7 In contrast, Mattsson et al.8 found that only 35% of dog-allergic subjects were sensitized to rCan f 4 as examined by ImmunoCAP (Phadia, Uppsala, Sweden). Here, we first examined the IgE reactivity of 63 dog-allergic patients and 12 control subjects to the isolated nCan f 4 by immunoblotting. Twenty-nine (46%) out of the 63 allergic subjects showed IgE binding to the natural allergen (Fig. E4 in Supplementary Material 2), thus pointing to a possibility that Can f 4 is recognized less frequently by dog-allergic subjects than previously assumed. In the immunoblotting analysis with rCan f 4, 31 (49%) out of the 63 subjects showed IgE reactivity to the allergen. The agreement between the 2 analyses was very good, as 29 samples were positive with both allergen preparations (κ=0.94, P<0.001). None of the control subjects was positive in the analyses (Fig. E4 in Supplementary Material 2).

IgE reactivity to the Can f 4 preparations was further evaluated by indirect ELISAs. In comparison to the immunoblot results, a greatly increased prevalence of IgE reactivity to the immunoaffinity-purified nCan f 4 was found: 51 out of the 63 serum samples (81%) were positive (Fig. 2A). However, when the sera were analyzed for specific IgE to rCan f 4, the prevalence of IgE reactivity decreased to 54%, as only 34 out of the 63 sera were positive (Fig. 2A). Therefore, the agreement between the analyses was only fair (κ=0.37, P<0.001). Interestingly, when multimeric (mu) rCan f 4 was used as antigen at the same protein concentration, 39 samples (62%) were found to be positive (Fig. 2A), and the agreement with the nCan f 4-based ELISA was better (κ=0.40, P<0.001). The specific IgE levels of dog-allergic subjects to nCan f 4, rCan f 4, or murCan f 4 were comparable (Fig. 2A; P>0.05, two-tailed t test). The prevalences of Can f 4-specific IgE reactivities among Finnish and Spanish dog-allergic patients according to the assays and allergen preparations used are shown in Fig. 2B.

The results above hint that the discrepancy in IgE seropositivity to Can f 4 with different methods and that allergen preparations could be due to the conformation dependency of IgE binding. To examine this, a dot blot analysis with the 3 mAbs and 3 dog-allergic patients' sera was conducted using natural undenatured or denatured Can f 4. Fig. E5 shows (see Supplementary Material 2) that denaturation greatly attenuate antibody binding to the natural allergen. On average, the intensity of dot blot reactions reduced by 44.8% with mAbs and by 54.4% with sera analyzed (analysis by ImageJ Microarray).

Sixteen dog-allergic patients and 6 nonatopic volunteers were examined with SPTs for reactivity to the natural and recombinant Can f 4 preparations. Fourteen (88%) and 7 (44%) out of the 16 allergic subjects showed a positive reaction to nCan f 4 or rCan f 4, respectively. Therefore, the agreement between the tests was only fair (κ=0.20, P=0.18). All the control subjects were negative in the tests. The mean wheal diameters induced by nCan f 4 (4.4±0.51 [mean±SEM] mm) and rCan f 4 (3.8±0.92 [mean±SEM] mm) did not differ statistically (P>0.05, Mann Whitney U test), and there was a good correlation between the wheal sizes (Spearman rank correlation; r=0.89, P<0.001).

To quantitatively compare the antibody-binding capacities of recombinant and natural Can f 4, the reactivity of each mAb to rCan f 4 was inhibited with nCan f 4 in ELISA, and then vice versa. As shown in Fig. 3A, the binding of mAbs to rCan f 4 was completely inhibited with nCan f 4. The reverse experimental setting (inhibition with rCan f 4) produced the same result (Fig. 3A). However, the inhibition curve with mAb 41G did not overlap entirely between the preparations (as was the case with 26D and 48F), as nCan f 4 was needed about 10 times more than the recombinant preparation to obtain the 50% inhibition level (Fig. 3A).

IgE-binding capacities of the Can f 4 preparations were further examined by IgE ELISA inhibitions. As shown in Fig. 3B, the IgE reactivity of 3 individual sera to rCan f 4 was completely inhibited by prior incubation with nCan f 4, whereas the reverse was not; rCan f 4 inhibited 50%-70% of the reactivity to nCan f 4.

The specificity of IgE binding to Can f 4 was examined by ELISA inhibition using nCan f 4 in the solid phase. As shown in Fig. 4, none of the 6 lipocalin proteins or the control proteins psoriasin or tear lipocalin showed IgE cross-reactivity with Can f 4.

For measuring the amount of Can f 4 in environmental samples or allergen preparations, a sandwich ELISA was developed. To that end, the 3 Can f 4-specific mAbs (all IgG1, data not shown) were analyzed by competitive ELISAs to provide an optimal pair for capturing Can f 4. Fig. 5A demonstrates that 2 of the antibodies (41G and 48F) recognize the identical or overlapping epitope. Therefore, after further experiments, the mAbs 26D and 48F were adopted for the ELISA. Fig. 5B shows that the linear part of the ELISA curve lays between the Can f 4 concentrations of 10 and 1,000 ng/mL. Next, we measured the concentrations of Can f 4 in the experimental dust samples collected with a vacuum cleaner from 5 homes with and 4 homes without dogs. Fig. 5C shows that the variation of the measured Can f 4 concentration is large. The mean Can f 4 concentration in the samples from homes with dogs was 67.3±61.0 µg/g of dust whereas it was 4.0±0.1 ng/g in the samples from homes without dogs.