Results

Separation of CTX species by preparative chromatography and off-line 2D LC/LC

Concentrated pools of serum and plasma, separated by preparative chromatography, exhibited similar elution profiles regardless of the kits used for quantitating the fractions, as illustrated in Figure 3. However, there were significant variations in the concentrations calculated by the different kits for highly concentrated fractions. This discrepancy can potentially be attributed to the matrix effect, which often has a substantial impact on immunoassays. Since our fractions differ significantly from a biological matrix, it’s likely that the various immunoassays are affected differently. In light of these findings, it was determined that reliance on the IDS-iSYS CTX-I (CrossLaps®) kit for further antibody-based quantitation would be more appropriate. Regarding retention time, CTX molecules appear to be highly hydrophilic, given their elution at a very low percentage of ACN. The differences observed in the elution profiles between urine and plasma/serum could be attributed to modifications that CTX molecules undergo, such as conjugation, in order to be excreted in urine. Consequently, different CTX metabolites may be present in higher concentrations in urine. Concerning the off-line 2D LC/LC separation, out of all the fractions obtained during the initial step of preparative chromatography (figure 4A), nine (fractions 3, 5, and 17-23) displayed concentrations above 1 ng/mL, the arbitrarily chosen threshold, and were retained for the second round of preparative chromatography (figure 4B). Two types of elution profiles were observed for fractions 3, 5, 17-23. Fractions 5 and 17 displayed very similar profiles, both with two peaks at 1 minute and 3.75 minutes, while the remaining fractions yielded elution profiles with only one peak at different retention time. Given the similarity in the elution profiles of fractions 5 and 17, it’s conceivable that the species present in these fractions may be isomers. In the initial chromatography, both isomers are separated into distinct fractions. It is plausible that following post-column flow splitting, the isolated isomer undergoes isomerization to transform into the other isomer, reaching equilibrium in solution. The fact that fractions 19 to 23 exhibited concentrations exceeding 1 ng/mL could be attributed to either significant tailing of the peak at 3.75 minutes or the presence of different molecules than those responsible for the peak at 3.75 minutes. In light of these results, the hypothesis that different CTX molecules coexist and are recognized by the immunoassays gains credence, as the elution profiles obtained from fractions 19 to 23 do not exhibit a peak at 3.75 minutes but rather at 4, 4.25, 4.5, 4.75, and 5 minutes.

Peptide identification

From the samples obtained after protein precipitation coupled with preparative chromatography, a total of 502 linear peptides derived from type I collagen were successfully identified. Among these, 22 peptides containing the C-terminal pyridinoline crosslink site were identified. Pyridinoline crosslinks involve the fusion of two telopeptide hydroxyallysine and one helix lysine. While only hydroxylysine residues, and by extension, hydroxyallysine, are directly implicated in pyridinoline crosslinks at the telopeptide level, our focus also encompasses peptides containing a lysine residue at the crosslink site of the telopeptide. This broader consideration aims to gather information about the cleavage sites around the crosslink site, rather than identifying precursors of the crosslink. Peptides featuring hydroxylation of proline residues (n=7) in the telopeptidic regions were excluded, as this PTM is reported to be absent in this region of collagen molecules [24,38–40]. Additional investigations are required to validate the presence of hydroxyproline in the telopeptidic peptides. For the samples purified by affinity chromatography, 23 peptides containing the epitope EKAHDGGR or EhylHDGGR were identified. These peptides varied in length, ranging from 7 to 60 residues. Hydroxyprolines were identified at various sites in 9 of these peptides. However, no sugar-mediated PTMs were detected. It is worth noting that the absence of glycosylation detection is not unexpected, as glycosylation can be challenging to analyze using electrospray ionization, the method employed in this study. Additionally, it is crucial to acknowledge that certain PTMs, such as the oxidation of hydroxylysine residues by lysyl oxidase—an essential step preceding crosslinking—were not considered in this analysis. Consequently, more linear peptides may have been present but not identified. Significantly, no divalently crosslinked peptides were found in the analysis. Divalent crosslinks are typically associated with newly-synthetized bone tissue, serving as precursors to trivalent crosslinks found in older bone tissue. The newly-synthetized bone tissue rarely undergoes bone resorption compared to older bone tissue except in some metabolic bone disorders, such as Paget’s disease. Therefore, founding no divalently crosslinked peptides in the plasma and serum is in line with the expectations. All peptides containing a lysine or hydroxylysine residue involved in pyridinoline crosslinks are detailed in Table 1.

Analysis of theorical trivalently crosslinked models

Given the multitude of peaks observed in the chromatogram resulting from affinity chromatography coupled to nano-LC-HR-MS, we postulated that the number of crosslinked species was substantial. To achieve a comprehensive identification of these species, particularly acknowledging that trivalently crosslinked peptides cannot be directly identified using standard peptide/protein identification software, we developed model molecules. These model molecules consisted of three peptides selected from the previously identified list of linear peptides, crosslinked together by a pyridinoline or deoxypyridinoline. Subsequently, these model molecules were employed in the chromatogram search using specialized software. A total of 3,230 model molecules, outlined in Supplemental Info 1, were generated by combining the various peptides. The chemical formula for each model molecule was calculated as part of this process. Skyline was employed to identify traces corresponding to our model molecules based on their chemical formulas. All model molecules were successfully identified in samples obtained through preparative chromatography (fractions 3, 5, 17-23) and samples obtained via affinity chromatography in both plasma and serum. For samples purified by preparative chromatography, the majority of the molecules coeluted within the first 3 minutes of liquid chromatography. Unfortunately, the peak shapes were poorly defined, characterized by low intensity. This phenomenon could be attributed to the inherent complexity of the sample, even after separation via preparative chromatography. The important competition for ionization in the ionization source becomes apparent when numerous molecules coelute. Additionally, molecules like phospholipids and albumin, which significantly compete for the signal, were not eliminated, further contributing to this challenge. In contrast, samples purified by affinity chromatography exhibited chromatograms (figure 5) with fewer interferences and better peak shapes, which can be attributed to the higher level of sample purity. The peak intensities in these chromatograms were more than 50 times higher. Following Skyline analysis, similar chromatograms were obtained for both plasma and serum using the same antibodies. However, commonalities were identified among the four chromatograms, including the coelution of numerous CTX species at 1.5 minutes and a prominent peak corresponding to the ”3+3+4” species. While C-terminal pyridinoline and deoxypyridinoline predominantly involve the α1 chain helix of type I collagen and less the α2 chain helix, the ”3+3+4” species, composed of two peptides, PQEKAHDGGR (α1 chain of type I collagen) crosslinked to the peptide FKGIRGHNG (α2 chain of type I collagen) by a deoxypyridinoline, was designated as the primary CTX species due to its consistent presence in all chromatograms. It is important to note, however, that this designation does not necessarily imply it is the most concentrated species in the blood. The prominence of the “3+3+4” species may stem from its superior ionization, possibly due to better isolation from other molecules or its specific structure and size[41,42]. Nonetheless, it is evident that different species were present in plasma and serum, and distinct species were also captured by the antibodies. In terms of differences between plasma and serum, it is apparent that plasma contains larger species than serum. This distinction can be explained by the fact that proteolysis occurs to a lesser extent in plasma after collection, as plasma proteases are inhibited by the chelation of divalent ions by EDTA. Moreover, it is evident that degradation products were simultaneously captured with the initial species they originated from, as very similar species were identified within the same sample. Regarding the differences between the species captured by the antibodies, the most prevalent species differed for both antibodies. There were no significant differences in terms of properties or structures between the group of species recognized by antibody 1M0161 and the group of species recognized by antibody 1M0122, except that the species captured by 1M0161 tended to have a lower retention time and were thus more polar than the species captured by 1M0122. However, a big number of the identified species were captured by the same antibody in both matrices. Therefore, it can be hypothesized that the capture of different species is reproducible.

Discussion

Osteoporosis, characterized by excessive bone resorption over bone formation, diminishes bone mass, heightening fracture vulnerability. Currently endorsed by the International Osteoporosis Foundation and the International Federation of Clinical Chemistry and Laboratory Medicine for the assessment of bone resorption in osteoporotic patients, CTX faces challenges due to variable immunoassay outcomes, eroding clinician confidence and limiting clinical use. Addressing this, the development of a reliable reference method is imperative, but the incomplete understanding of CTX structure hinders progress. In this work, we employed a multi-step approach in order to fully characterize CTX. We extracted CTX directly from human biological matrices using preparative liquid chromatography (LC) and affinity chromatography techniques. Preparative LC served as a straightforward preparation step for the subsequent identification of uncrosslinked linear peptides derived from type I collagen via high-resolution mass spectrometry (HR-MS). The objective of the preparative chromatography-based separation was to identify a comprehensive set of peptides originating from type I collagen present in the blood, independent of antibodies, in order to use them to build CTX models. Affinity chromatography was then used for the analysis of CTX species, as it provided a significantly more concentrated and purified sample thanks to its high selectivity for CTX species. We then applied our previously described proteomics workflow[36] to characterize all the distinct CTX species present in both plasma and serum samples. This marks a significant step towards enhancing our understanding of these critical markers in the context of osteoporosis management. The substantial number of CTX species identified in our study can be attributed to the complex nature of CTX proteolysis, occurring at multiple stages. Firstly, our prior research on cathepsin K cleavage sites has demonstrated that the digestion of type I collagen by cathepsin K remains non-reproducible[36]. Consequently, during bone resorption, multiple CTX species are produced. Moreover, once released into the bloodstream, these species are susceptible to cleavage by circulating metalloproteases and undergo further proteolysis in the liver by Kupffer cells, leading to an increased number of coexisting species in the bloodstream. Depending on the type of blood collection tube used, proteolysis by circulating proteases may still occur after blood collection. In addition to cleavages, the presence of PTMs that are not uniform further increases the variety of species observed in the chromatograms. These PTMs primarily involve the hydroxylation of lysine and proline residues, which are highly prevalent in type I collagen. Given that an aspartic acid residue is located near the crosslink site, many of the species may undergo isomerization resulting in the manifestation of multiple species. However, β-isomerization of type I collagen is associated with older bone tissue, suggesting that the majority of resorbed bone tissue should exhibit β-isomerization. Consequently, CTX should also predominantly exist in the β-isomeric form, as it is released during bone resorption. Nevertheless, in specific bone diseases such as Paget’s disease, α-CTX may be released during bone resorption. It is worth noting that smaller species were more prominent in the chromatograms, possibly due to their higher abundance in the blood as a result of in situ proteolysis. Additionally, larger molecules may ionize less efficiently in the mass spectrometer source, and some of them may have partially precipitated during the sample preparation process. The observation that more large species were found in plasma aligns with this theory. A significant limitation in our study is the inability to doubly confirm the presence of our model molecules through MS/MS spectra, given the unknown and thus, unpredictable fragmentation pattern of such molecules. Nevertheless, the superposition of traces from multiple states of charge at the same retention time serves as an intriguing indicator of our target, making it less likely to be an interference. The diversity of CTX species identified and the differences in the species recognized by the antibodies in the three commercially available kits may contribute to the variations in results obtained by these assays[43–45]. Each kit may use a different combination of antibodies, leading to differential recognition of CTX species. However, it’s worth noting that despite these differences, the overall results from the kits do not show significant discrepancies. This suggests that a significant portion of CTX species is still effectively captured by the antibodies used in these assays, leading to relatively consistent measurements across the kits. In conclusion, this study successfully identified a substantial number of CTX species extracted from human plasma and serum. Our findings highlight the complexity of CTX proteolysis, and the diverse array of species present in these biological matrices. In light of our findings, it becomes evident that, prior to the development of a LC-MS/MS method for the quantitation of all CTX species, the development of a digestion step aimed at yielding a single, standardized CTX species is crucial. Once this digestion step is optimized within the biological matrix, we will be well-positioned to progress towards the development and subsequent validation of an LC-MS/MS method for the quantitation of the total CTX. This method is anticipated to play a crucial role in advancing the precision and standardization of CTX immunoassays, thereby improving osteoporosis monitoring. Following standardization, these immunoassays are expected to exhibit reduced variability, becoming fully interchangeable. This, in turn, is projected to instill greater confidence among clinicians, fostering enhanced utilization of the marker in patient follow-up. With improved assessment of medication compliance, the risk of fractures is anticipated to decrease in osteoporotic patients.