A Novel Sandwich ELISA for Tartrate-Resistant Acid Phosphatase 5a and 5b Protein Reveals that Both Isoforms are Secreted by Differentiating Osteoclasts and Correlate to the Type I Collagen Degradation Marker CTX-I In Vivo and In Vitro

Healthy human adult male serum samples (Novakemi AB, Handen, Sweden) were used as quality control samples. They were aliquoted and stored at − 20 °C. Median age was 47.8 years (range 29 –65 years). Human peripheral blood monocytes (hPBMC) were isolated from buffy coats of donated blood of healthy men at Karolinska University Hospital Laboratory, Huddinge, Sweden. The surplus blood products were coded for research use. The study was performed in accordance of guidelines from Swedish Research Council and Karolinska Institutet. All donors involved in the study gave written informed consent in accordance with the Helsinki Declaration.

In order to separate TRAP 5a from TRAP 5b, the method was designed as follows (Fig. 1): In the first step, a sample was added to TRAP 5a-specific antibody (mAb 46)-coated wells. TRAP 5a was captured on mAb 46 antibodies and, after overnight incubation at + 4 °C, the sample, depleted of TRAP 5a but having the original amount of TRAP 5b, was transferred to wells coated with mAb 25.44, recognizing both isoforms. Subsequently, the bound TRAP 5a in the first well was detected using biotinylated mAb 12.56 biotin/streptavidin-conjugated HRP and the remaining TRAP 5b in the second well was measured using biotinylated mAb 12.56 and biotin/streptavidin-conjugated HRP.

General protocol for double-sandwich ELISA quantifying TRAP 5a and 5b protein independently in one and the same sample: 96-well ½ volumes ELISA plates (Costar, New York, NY) were coated with 5 μg/ml (100 μl/well) mAb 46 or (50 μl/well) mAb 25.44 diluted in PBS (0.01 M pH 7.4) and left at 4 °C overnight. Next day, mAb 46-coated plates were washed 3 times in TBST (200 µl/well; 25 mM Tris 7.6, 150 mM NaCl, 0.1% Tween 20). After washing, pre-treated standards and samples, for procedure see “Sample preparation: pre-treatment” (65 μl/well) were added to mAb 46-coated plate and incubated overnight at 4 °C. Supernatant (samples and TRAP 5a standard curve) was then transferred (50 μl/well) from mAb 46-coated plate to mAb 25.44-coated plate (Fig. 1). Also, fresh standard curve for TRAP 5a was added to the mAb 25.44-coated plates and incubated for 2 h at room temperature. Plates were washed 3 times in TBST (200 μl/well) and incubated with 0.25 μg/ml biotinylated detection mAb 12.56 (50 μl/well) diluted in 0.1% bovine serum albumin (BSA) + TBST for 1 h at room temperature. Next, plates were washed 3 times in TBST (200 μl/well) and incubated with streptavidin-HRP (Mabtech) diluted 1:1000 in 0.1% BSA + TBST (50 μl/well) for 1 h at room temperature. Finally, plates were washed 3 times in TBST (200 μl/well) and developed using K-Blue^® Substrate (TMB) (Neogen, Lansing, MI) (50 μl/well) and the reaction was stopped with 1 M H₂SO₄ (50 μl/well). Absorbance was read at 450 nm using BioTek’s PowerWave HT microplate spectrophotometer (BioTek, Winooski, VT). Affinities of mAb 46 and mAb 25.44 were compared in a sandwich ELISA. TRAP 5a and TRAP 5b standards were added to mAb 46- or mAb 25.44-coated plates followed by incubation overnight. Next day, ELISA assay was performed as described above.

Total TRAP sandwich ELISA was performed as the previously described TRAP 5a ELISA [28] and as described above under section “General protocol for double sandwich ELISA quantifying TRAP 5a and 5b protein independently in one and the same sample” with the exception that plates were coated with 5 µg/ml mAb 25.44 and no transfer is needed.

Limit of blank (LoB), limit of detection (LoD), and limit of quantification (LoQ) were calculated as previously described [29].

Inter-assay variation was estimated by calculating CV % of four separate runs. It was calculated using human serum samples with TRAP concentrations above or equal to LoD (see “Human Serum Samples”) (n = 18).

Intra-assay variation was determined using 16 technical replicates of 11 human serum samples of TRAP 5a and 5b concentrations at different points of the detection range (see “Human Serum Samples”).

For pre-treatment, serum samples (30 μl) were diluted 1:1 with 1.5 M glycine pH 2.5 (30 μl) and incubated for 1 h at 37 °C. After incubation, samples were neutralized by addition of 1 volume (30 μl) 1 M Tris–HCl pH 8.5. Samples were further diluted by an additional 3 volumes (90 μl) of Diluent ELISA (Mabtech) leading to final dilution of serum samples 1:6. Standard curves were directly diluted in 1.5 M glycine pH 2.5, followed by the addition of 1 volume 1 M Tris–HCl pH 8.5, and 3 volumes of Diluent ELISA giving a final dilution of 1:6.

First, the amount of TRAP 5a on mAb 46 plate was calculated from the TRAP 5a standard curve. Then the TRAP 5a standard curve was transferred to mAb 25.44 plate and the values from samples were recorded to produce a carry-over standard curve to obtain the expected absorbance left over at different concentrations. The carry-over of TRAP 5a value was then subtracted from the TRAP 5b sample values from mAb 25.44 before the actual concentrations were interpolated from the TRAP 5b standard curve.

Standard curves for TRAP 5a and TRAP 5b were compared using curve fit with linear regression. Serum total TRAP and TRAP 5b + 5a without and with correction were analyzed using Wilcoxon matched-pairs signed-rank test. Correlations between CTX-I, TRAP 5a, and TRAP 5b were made using Spearman correlation. Differences in TRAP 5a and 5b protein levels in lysate and media were analyzed using Mann–Whitney U test. p values < 0.05 were considered significant.

The sandwich ELISA (Fig. 1) using mAb 46 as capture antibody, raised against the loop region specific for TRAP 5a, detected in a dose-dependent manner specifically TRAP 5a, but not TRAP 5b (p = 0.0003; Fig. 2a), in line with previous studies [28]. mAb 25.44 capture antibody detected both TRAP 5a and 5b in a dose-dependent manner with the same affinity (p = 0.9591; Fig. 2b). In cross screening of antibodies raised against recombinant TRAP 5a, detector antibody mAb12.56 was superior against both capture antibodies.

Examination of TRAP 5a and 5b carry-over showed that TRAP 5b protein concentration was not affected by previous incubation on mAb 46 (Fig. 2c). However, a fraction of TRAP 5a was not immobilized on mAb 46 but was carried over to mAb 25.44 during sample transfer (Fig. 2c). This carry-over of TRAP 5a would lead to falsely high TRAP 5b protein levels, since the mAb 25.44 does not discriminate between the isoforms. The error caused by carry-over TRAP 5a could be compensated by transferring also the TRAP 5a standard curve to mAb 25.44 and subtracting the amount of TRAP 5a carry-over to the wells detecting TRAP 5b (Fig. 2d, f). In serum, correcting for TRAP 5a carry-over resulted in TRAP 5a + 5b protein concentrations that were equal to total TRAP protein concentration while the sum of TRAP 5a + 5b protein concentrations without correction gave significantly higher TRAP concentrations than total TRAP measurement (Fig. 2e).

In serum, TRAP can be found in a complex formed with a2M, which may impair TRAP 5a and 5b detection, therefore the effect of pre-treatment of serum samples using acidification to promote dissociation of a2M-TRAP complexes was investigated. TRAP 5a concentrations measured in pre-treated serum were significantly higher (p < 0.0450) compared to non-treated samples (Fig. 3a). On the other hand, pre-treatment of serum had no effect on amount of TRAP 5b or total TRAP detected on mAb 25.44 (Fig. 3b, c). This effect of pre-treatment on TRAP 5a on mAb 46 was not observed with the recombinant proteins or samples from in vitro human OC cultures (data not shown).

Both inter- and intra-assay variations for TRAP 5a and TRAP 5b in the double-TRAP 5a/5b sandwich ELISA were within an acceptable range (below 20%) (Table 1). The LoD values were 0.3 ng/ml and 0.05 ng/ml for TRAP 5a and 5b, respectively. No degradation of TRAP 5b recombinant protein or decrease in TRAP 5b protein concentrations in serum samples was seen over time (Supplementary figure 1).

In healthy adult men, TRAP 5a concentration was 4.4 ± 0.6 ng/ml and TRAP 5b was 1.3 ± 0.2 ng/ml, respectively (Table 1). Thus, isoform 5a constituted 75% of serum TRAP protein while isoform 5b accounted for 25%, i.e. a ratio of 3:1 (Table 1). In serum, TRAP 5a and 5b protein concentrations correlated moderately positively (for definition see [30]) to each other (Fig. 4a). Also, TRAP 5b protein levels correlated moderately positively with the collagen degradation marker CTX-I (Fig. 4b) which is in line with previous results [31]. Interestingly, TRAP 5a also showed a positive moderate correlation to CTX-I (Fig. 4c).

To validate TRAP isoforms as biomarkers, we measured their production on active (bone) and inactive (plastic) substrates by human CD14+ PBMCs. The PBMCs were isolated from 6 healthy male donors and cultured on cell culture plastic in the presence of 30 ng/ml macrophage colony stimulating factor (M-CSF) and 2 ng/ml RANKL or 30 ng/ml M-CSF alone as a control. In the presence of RANKL + M-CSF, multinucleated TRAP-positive osteoclast-like cells started to appear at day 4, whereas in cultures with M-CSF alone, aggregated cell clusters and a few TRAP-positive mononuclear cells but no multinucleated cells were observed at this time-point (Supplementary figure 3A). On day 6 with RANKL and M-CSF, there were abundant osteoclast-like multinuclear cells with strong TRAP signal and only a few TRAP-positive mononuclear cells. With M-CSF alone, there was an abundance of mononuclear TRAP-positive cells but no multinucleated cells.

On bone, the difference between RANKL + M-CSF and M-CSF alone on CD14+ PBMCs was even more pronounced, however, the time course of OC differentiation was somewhat delayed compared to differentiation on plastic (Supplementary figure 3B). At day 4, with both M-CSF alone and RANKL + M-CSF, only TRAP-positive mononuclear cells were attached to bone, whereas at day 8 and onwards, TRAP-positive multinucleated cells appeared only in M-CSF + RANKL-stimulated cultures. With M-CSF alone, there were only few TRAP-positive mononuclear cells. Furthermore, deep resorption pits were observed on bone slices on which RANKL-stimulated cells were cultured, confirming their identity as OCs (Supplementary figure 3C). These pits were stained TRAP-positive, especially the edges of the pit, indicating secretion of TRAP during resorption.

To confirm that the mAb 46 and mAb 25.44 indeed identify different targets and different timepoints of OC differentiation and function, we conducted immunocytochemical (ICC) staining of the TRAP isoforms. The distribution and localization of intracellular TRAP isoforms 5a and 5b in CD14+ PBMCs differentiated with M-CSF-and RANKL were analyzed by ICC staining with TRAP 5a-specific mAb 46 and total (5a + 5b) polyclonal TRAP-antibody on glass (Fig. 5a) or on bone (Fig. 5b) using confocal microscopy. M-CSF + RANKL-treated PBMCs cultured on glass demonstrated TRAP 5a signal (green) at 4 days of culture in a vast majority of both mono- and multinuclear cells (Fig. 5a). In addition, total TRAP (magenta) was present in most cells, although a small population of mononuclear as well as a few multinuclear cells displayed very high total TRAP staining (Fig. 5a). The levels of total TRAP staining varied considerably between cells, and in the cells with high total TRAP signal, the isoforms appeared to be predominantly co-localized (white areas) in a perinuclear area of the cells. At day 7, both mononuclear and multinucleated cells stained positive for TRAP 5a, but the multinucleated cells displayed only weak intracellular total TRAP staining, whereas high total TRAP signals were detected in a subset of mononuclear cells. Furthermore, the co-localization of TRAP 5a and total TRAP in multinucleated cells was lower on day 7 compared to day 4, suggesting that the antibodies stain different compartments, and the compartments stained exclusively magenta in the merged figure could represent stored TRAP 5b.

On bone-coated coverslips [32], TRAP 5a was more widely spread throughout the cell compared to total TRAP and abundant in both mono- and multinuclear cells stimulated with M-CSF + RANKL (Fig. 5b). At day 4, most mononuclear cells were positive for TRAP 5a, whereas a subset of these mononuclear cells were also positive for total TRAP. The localization of the two stains showed very little overlap on day 4 and several compartments stained exclusively magenta in the merged image, suggesting the presence of intracellular compartments with TRAP 5b (Fig. 5b). The most striking difference regarding TRAP 5a was observed in multinucleated cells exhibiting actin structures resembling sealing zones at day 8. These cells exhibited perinuclear structures that were strongly stained for TRAP 5a and this staining overlapped to a high degree with staining for total TRAP (Fig. 5b), indicating the presence of TRAP 5a also in resorbing multinuclear cells. In addition to the strong perinuclear staining, a few vesicles in the same cells were positive for only total TRAP. Similar vesicles positive for only total TRAP, indicating the presence of TRAP 5b, were found also in other multinuclear and mononuclear cells (Fig. 5b).

To assess kinetics of generation and secretion of TRAP isoforms from CD14+ derived OCs, intracellular (cell lysate) and extracellular (culture media) TRAP 5a and 5b concentrations were measured (Fig. 6). In cultures on plastic, secretion (medium) of TRAP 5b and TRAP 5a from cultures differentiated with RANKL + M-CSF were significantly higher from day 4 onwards compared to day 3 of culture (Fig. 6a). Concentrations of TRAP 5b showed a constant increase over the culture period and it was the predominant secreted isoform throughout the culture. Cells stimulated with M-CSF alone secreted low basal amounts of both isoforms without a significant increase over the culture period.

On plastic, intracellular (lysate) TRAP 5b protein concentrations increased rapidly between days 3 and 4 in M-CSF + RANKL-stimulated cells and were significantly higher from day 4 of culture. The increase of TRAP 5a concentrations occurred later and they were significantly higher from day 6 of culture. In the cell culture lysates stimulated with M-CSF alone, concentrations of both TRAP isoforms were low, consistent with the TRAP activity staining (Supplementary figure 3). During OC differentiation, TRAP 5b concentrations increased over 20-fold, both in lysate and medium, compared to day 3 where no multinuclear cells were present.

In cultures grown on bone chips, a significant increase in the secretion of TRAP 5b was apparent from day 6 of culture in M-CSF + RANKL-stimulated cells compared to day 4 (Fig. 6c).

Thus, secretion occurred later on bone than in cultures grown on plastic, and secreted concentrations were much higher than on plastic, both matching observations from the TRAP activity staining (Supplementary figure 3). Also, secretion of TRAP 5a increased significantly on day 8 of culture in M-CSF + RANKL-stimulated cells (Fig. 6c), but remained lower than secreted TRAP 5b. Overall, the pattern of TRAP isoform secretion seems to be similar on bone and plastic. In lysates, there was a corresponding increase in both isoforms (high for TRAP 5b and moderate for 5a), from day 8 coinciding with the appearance of OCs (cf. Supplementary figure 3B). Put together, cultures grown on bone produced more TRAP isoforms prompting us to investigate its connection with bone resorption, especially as cells with sealing zones display high levels of intracellular TRAP isoforms (Fig. 5b).

The bone resorption marker CTX-I was detectable in medium from day 6 and increased progressively to day 12 (Fig. 6e). Intracellular TRAP 5b was detectable prior to production of CTX-I, indicating that TRAP production emerges before onset of resorption and intracellular processing of TRAP 5a to TRAP 5b (Fig. 6c, e). The ratio of CTX-I/TRAP 5b in the medium showed a strong increase on day 8 of the culture period suggesting an active role for TRAP 5b in facilitating bone resorption activity (Fig. 6f). The relative ratios of TRAP isoforms 5b/5a and CTX-I/5a increased up to day 8 and decreased thereafter. This decrease was due to a continued secretion of TRAP 5a at later culture time points and coincided with gradual saturation of CTX-I-output.

Concentrations of TRAP 5a and 5b protein from in vitro OC differentiation on bone correlated moderately high, and positively to each other (Fig. 6g). Also, both TRAP 5a and TRAP 5b protein levels correlated positively with CTX-I (Figs. 4i, 6h) similar to the observation in human serum (cf. Fig. 3), suggesting that a considerable fraction of serum TRAP 5a could derive from differentiating OCs.