Sensitive asprosin detection in clinical samples reveals serum/saliva correlation and indicates cartilage as source for serum asprosin

The C-terminal pro-fibrillin-1 propeptide asprosin is described as white adipose tissue derived hormone that stimulates rapid hepatic glucose release and activates hunger-promoting hypothalamic neurons. Numerous studies proposed correlations of asprosin levels with clinical parameters. However, the enormous variability of reported serum and plasma asprosin levels illustrates the need for sensitive and reliable detection methods in clinical samples. Here we report on newly developed biochemical methods for asprosin concentration and detection in several body fluids including serum, plasma, saliva, breast milk, and urine. Since we found that glycosylation impacts human asprosin detection we analyzed its glycosylation profile. Employing a new sandwich ELISA revealed that serum and saliva asprosin correlate strongly, depend on biological sex, and feeding status. To investigate the contribution of connective tissue-derived asprosin to serum levels we screened two cohorts with described cartilage turnover. Serum asprosin correlated with COMP, a marker for cartilage degradation upon running exercise and after total hip replacement surgery. This together with our finding that asprosin is produced by primary human chondrocytes and expressed in human cartilage suggests a contribution of cartilage to serum asprosin. Furthermore, we determined asprosin levels in breast milk, and urine, for the first time, and propose saliva asprosin as an accessible clinical marker for future studies.


Scientific Reports
| (2022) 12:1340 | https://doi.org/10.1038/s41598-022-05060-x www.nature.com/scientificreports/ monoclonal antibody therapy was recently shown to be promising in the treatment of metabolic syndrome as it reduced appetite and body weight in mouse models 3 . Due to its proposed crucial metabolic function, several clinical studies have been conducted to establish a correlation of asprosin levels with obesity, diabetes mellitus type 1 and 2, insulin resistance, metabolic syndrome, and physical exercise [4][5][6] . However, reported measurements of asprosin in clinical samples showed considerable variations across studies raising serious concerns about the reliability of the applied ELISA detection approaches 5 . For instance, asprosin concentrations determined in blood samples from children showed huge variations of four orders of magnitude from < 1 ng/ml up to > 100 ng/ml [7][8][9][10] (Supplementary Table 1). In adults reported asprosin concentrations in plasma and serum even range from < 0.5 ng to > 350 ng/ml (Supplementary Table 1). Additionally, the sensitivity of the employed ELISA kits appears to vary by more than one order of magnitude regarding standard detection range (0.156-10 ng/ml to 7.80-500 ng/ml), and sensitivity (0.1-2.1 ng/ml) (Supplementary Table 1).
This documented variability illustrates the need for new approaches for the sensitive and reliable measurement of asprosin in human body fluids. We therefore aimed to develop new biochemical tools and methods for asprosin detection in several body fluids. Since fibrillin-1 is known to be a structural component of the extracellular matrix architecture of connective tissue microenvironments 11 , it is plausible that connective tissue derived asprosin contributes to blood asprosin levels. To test a potential contribution of cartilage derived asprosin to total serum asprosin levels, we measured asprosin levels in two cohorts for which release of cartilage derived components into the blood were already documented by monitoring serum levels of cartilage oligomeric matrix protein (COMP), a marker for cartilage degradation and osteoarthritis 12,13 . We measured serum samples from athletes before and after mechanical loading by acute physical activity as well as and from osteoarthritis patients before and after hip replacement surgery. Our results not only provide new biochemical insight into the molecular requirements for sensitive asprosin detection in so far not investigated body fluids but also revealed a new correlation between asprosin levels and cartilage metabolism.

Establishment of sensitive methods for human asprosin detection. Generation of recombinant
human asprosin and analysis of its glycosylation profile. For the generation of a specific polyclonal antiserum against asprosin, we recombinantly overexpressed the human asprosin protein sequence (Fig. 1C) in HEK293 EBNA cells with a C-terminally placed double Strep-tag II for efficient affinity purification. Quality control SDS-PAGE analysis showed that asprosin was obtained with high purity after resin-coupled streptactin affinity chromatography as assessed by Coomassie staining (Fig. 1B). The molecular weight of recombinant asprosin corresponded to approximately 37 kDa, suggesting that approximately 20 kDa were added by glycosylation in HEK293 cells when compared to its calculated mass based on the amino acid sequence (Fig. 1B,C). Deglycosylation by using an enzyme mixture to remove all N-linked as well as many common O-linked glycans resulted in a band at 20 kDa which corresponds to the calculated mass of the asprosin core protein (Fig. 1D,G).
It was previously predicted that human asprosin possesses three N-glycosylation sites 1 , and recently its N-glycosylation in yeast was investigated 14 . However, until now no data is available describing the actual sites and profiles of asprosin N-glycosylation used by human cells. To investigate these structural profiles of N-linked chains as well as to identify which N-glycosylation sites are used in human cells, we performed mass spectrometry of asprosin overexpressed in HEK293 cells after enzymatic glycan liberation and proteolytic digestion Figure 1. Generation of a specific polyclonal anti-asprosin antibody. (A) Domain structure of fibrillin-1 and its N-and C-terminal halves. The C-terminal propeptide sequence of fibrillin-1 after furin cleavage (furin cleavage site marked by arrow and dashed line, furin cleavage sequence underlined) yields asprosin (marked in grey). The 140 amino acids representing human asprosin S 2732 -H 2871 were recombinantly produced in HEK293 cells. (B) Coomassie stained quality control gel of recombinantly produced and purified (> 95% purity) human asprosin. Asprosin was overexpressed with a C-terminally placed 2 × -Strep-tag II, and eluted fractions (F2-F6) after affinity chromatography were subjected to reducing SDS-PAGE using a 12% gel. (C) Mass spectrometry analysis revealed that all of the predicted three N-linked glycosylation sites (marked in red, position regarding the fibrillin-1 sequence in parentheses) are used as indicated by the + 1 mass shift expected for de-N-glycosylated peptides (see also Supplementary Table S2). Underlined residues mark predicted cleavage sites of V8 and trypsin used to generate peptides of de-N-glycosylated asprosin. No unequivocal assignments of O-glycosylation sites were possible on the basis of peptide masses in MS1 spectra. Residues representing linker regions are indicated in green, thrombin cleavage (LVPRGS) site is underlined, and Strep-tag II sequences are marked in blue. (D) Removal of N-linked, as well as many common O-linked glycans, after incubation in a deglycosylation mix results in significant loss of molecular weight of human asprosin. (E) (left) Polyclonal anti-asprosin serum (pc-asp anti-asprosin antibody) raised in rabbit showed a high titer in ELISA assay detecting coated asprosin (100 ng/well). Half maximum signal was already reached at a 1:10,000 dilution. Rabbit pre-serum (before rabbit immunization) served as a control. Recombinant asprosin was coated in duplicates (100 ng/well). (right) ELISA showing high specificity of the pc-asp antibody. No cross-reactivity to the N-terminal region of fibrillin-1 (rF90), C-terminal region of fibrillin-1 (rF6), human albumin, or human IgG was detected. Data points represent mean ± SD of duplicates. (F) (left) Pc-asp antibody specifically recognizes human asprosin in western blot analysis, with no observable cross-reactivity to the N-terminal region (rF90) and C-terminal region (rF6) of fibrillin-1. (middle) Detection of N-terminal fibrillin-1 (rF90) with labmade fibrillin-1 rF90 antibody. (right) Specific detection of the C-terminal fibrillin-1 (rF6) with CPTC-FBN1-3 antibody (DSHB, Iowa, USA) which was raised against a synthetic peptide in the rF6 region. (G) Reduced signal of deglycosylated asprosin in western blot by pc-asp anti-asprosin antibody. Data were analyzed using Graphpad Prism version 8.0.2. www.nature.com/scientificreports/ of de-N-glycosylated peptides. Profiles of PNGaseF liberated N-glycans were characterized by one dominant high-mannose-type chain (M5) and by a prominent series of core-fucosylated complex-type chains with bi-to tetra-antennarity and sialylation with up to four residues (Supplementary Fig. S1 and Table S2). Besides these major glycans with fully processed complex-type structures, a considerable proportion of incompletely processed species of the truncated -Gal series was detected. N-Glycosylation sites were identified in the peptide mass fingerprints after de-N-glycosylation and conversion of Asn to Asp by the + 1 mass shift (see Supplementary Fig. S3  and Table S3). According to these peptide mass fingerprints, all three predicted sites (N3 (N2734), N19 (N2750), and N36 (N2767)) were found to be glycosylated (Fig. 1C). Asprosin could be also O-glycosylated, as three sites in the amino-terminal region (Ser1, Thr5, Thr16) were predicted (NetOGlyc version 4.0.0.13). Chemical cleavage of O-linked chains by reductive beta-elimination revealed one signal in the MALDI survey spectrum at m/z 1706 that could be structurally assigned by PSD-MALDI-MS to a hexasaccharide alditol based on a core 2 tetrasaccharide substituted with two sialic acid residues ( Supplementary Fig. S2). Based on peptide mass fingerprints of de-N-glycosylated asprosin peptides, no assignments of the O-glycosylation sites were possible.

Scientific
Establishment of ELISA and SPR assays for sensitive asprosin detection. Purified asprosin was injected into a rabbit for polyclonal antibody production. The obtained serum after 60 days of immunization showed a half-maximum titer at 1:10 4 dilution. The affinity-purified antibody showed a high specificity against human asprosin and showed no cross-reactivity to mouse asprosin or the N-terminal and C-terminal halves of fibrillin-1 (Fig. 1A) when used in western blot and direct ELISA (Fig. 1E,F and Supplementary Fig. S5D). To assess the sensitivity of the newly generated polyclonal anti-asprosin antibody (pc-asp) we employed surface plasmon resonance (SPR). Anti-asprosin antibody was immobilized onto a sensor chip and a concentration series of recombinant asprosin (0-80 nM) was flown over in solution ( Fig. 2A). Obtained signals showed a linear response still in the range of 0-1.25 nM (0-50 ng/ml) ( Fig. 2A, (right)). An even higher sensitivity was observed by direct ELISA when asprosin was directly coated to the microtiter well, showing a linear response in the range between 0 and 1.0 ng/ml (0-28 pM) (Fig. 2B). Next, we established a sandwich ELISA by using the newly generated pc-asp as capture and a commercially available monoclonal anti-asprosin antibody (mab-asp) (clone Birdy-1, AdipoGen Life Sciences Inc.) as detector. Titration experiments showed a linear detection signal in the range of 0-0.15 ng/ ml (0-4 pM) (Fig. 2C). The linear detection range of the asprosin sandwich ELISA was 0-50 ng/ml (Supplementary Fig. S4A). Coating concentration of pc-asp as capture antibody was optimized at 3 µg/ml, as higher coating concentrations lead to a marked decrease of the detection signal ( Supplementary Fig. S4B). SPR measurements of immobilized asprosin showed that pc-asp has a molecular affinity in the < 1 nM range (K D = 0.29 ± 0.30 nM), while the K D of mab-asp is 75-fold higher (22 ± 2 nM) ( Supplementary Fig. S5A). SPR experiments also revealed that the newly generated pc-asp has a more than 100-fold higher sensitivity for immobilized asprosin compared to mab-asp ( Supplementary Fig. S5B) which was also indicated by direct ELISA measurements ( Supplementary  Fig. S5C). The established sandwich ELISA assay can detect human asprosin with a sensitivity of < 65 pg/ml with an inter-assay coefficient of variability (CV) < 6.6% (n = 32), and an intra-assay CV < 5.5% (n = 32). We also tested any potential cross-reactivity of pc-asp against mouse asprosin and human placensin, the C-terminal propeptide of fibrillin-2 which was recently proposed to also have a metabolic function 15 . In direct ELISA and western blot no signals of mouse asprosin were detectable ( Supplementary Fig. S5D). Also our asprosin sandwich ELISA did not detect placensin which we also overexpressed und purified with a double C-terminal Strep-tag II (Supplementary Fig. S5E-G). This finding even excluded a potential cross-reactivity of pc-asp against the Strep-tag II sequence. Deglycosylation resulted in a better detection sensitivity of asprosin by mab in ELISA (Supplementary Fig. S5H (right)) and western blot ( Supplementary Fig. S5I), while direct ELISA detection by pc-asp was not altered (Supplementary Fig. S5H (left)).
Affinity column chromatography allows specific asprosin concentration and determination in serum and urine. Depending on the sensitivity of the employed assay asprosin levels in body fluids may not be detectable without pre-concentration. To selectively concentrate asprosin from body fluids we generated an affinity column by immobilizing pc-asp via CNBr to sepharose beads (Fig. 3A). After sample application, the column was washed and the captured amounts of asprosin were eluted at pH 2.5 and neutralized (see "Materials and Methods" section). Using this approach, we were able to selectively pull-down asprosin from the cell culture supernatant of RPE cells (Fig. 3B). ELISA and SPR analysis revealed an eight-fold increase of asprosin concentration by using this affinity column approach (Fig. 3B). With this approach, we were able to obtain sufficient serum asprosin amounts for western blot detection. Thereby, we could demonstrate that human asprosin circulating in serum has a molecular weight of around 40 kDa (Fig. 3C), similar to the recombinantly expressed asprosin (Fig. 1B). Utilizing the pc-asp affinity column approach allowed us also to concentrate asprosin from urine, for the first time. However, after concentration of 200 ml urine, we were able to detect about 3 ng/ml of asprosin which corresponds to 15 pg/ml present in non-concentrated urine (Fig. 3D). www.nature.com/scientificreports/ with the secreted amounts of fibrillin-1 which were significantly higher in cells derived from extracellular matrix (ECM) protein-producing connective tissues (Fig. 4A, bottom).

Measurement of asprosin in human cell lines and clinical samples.
Asprosin levels in serum and saliva correlate strongly and depend on biological sex and feeding status. To test the accuracy of our sandwich ELISA in clinical samples we performed spike-and-recovery and linearity-ofdilution experiments in serum and plasma ( Supplementary Fig. S6). The signal recovery was not affected when our recombinant asprosin standard was serially diluted in human plasma or serum ( Supplementary Fig. S6). www.nature.com/scientificreports/ Measurement of serum samples from age, biological sex, and BMI matched individuals showed that asprosin levels were significantly higher in women than men (Fig. 4B). It was previously reported that asprosin serum levels increase upon fasting 1 and that asprosin can also be detected in saliva 16 . We were therefore interested whether a fasting-induced increase of asprosin can be also detected in saliva, and whether serum and saliva levels show a correlation. Assessment of saliva samples from the same individuals 3 h after lunch (non-fasted) or after overnight fasting (fasted) showed that fasting induced an increase of asprosin levels in saliva (Fig. 4C). We then quantified asprosin amounts in simultaneously taken serum and saliva samples from several individuals after overnight fasting and performed a Spearman correlation analysis. Our results revealed a strong correlation (Spearman correlation coefficient: ρ = 0.9429, *P = 0.0167) between asprosin levels in serum and saliva (Fig. 4D). Asprosin levels in serum and plasma from the same individuals were about ten-fold higher than in saliva. Similar  www.nature.com/scientificreports/ to the low asprosin concentrations detected in saliva, we detected asprosin in human breast milk in the range of (1.1-3.3 ng/ml). An overview of the ranges of detected asprosin in human body fluids is given in Table 1.

Measurement of asprosin in patient cohorts with documented cartilage turnover.
Asprosin serum levels increase after 30 min of running exercise. Recent studies reported that asprosin serum levels are regulated by physical activity 17,18 . For instance, it was found that plasma asprosin levels were elevated only in women upon 20 s of anaerobic exercise 19 . To further investigate whether physical exercise affects asprosin blood levels in men, and whether cartilage may be a potential source of serum asprosin we measured serum levels of asprosin before and after running exercise in a cohort for which release of the cartilage degradation marker cartilage oligomeric matrix protein (COMP) was already reported 20     www.nature.com/scientificreports/ ately before (after 30 min rest in supine position) and after the 30 min running exercise. Further blood samples were taken 30, 60, and 120 min after the end of the running exercise ( Fig. 5A). Serum asprosin levels increased by about 25% on average (123.7 ± 42.3%) immediately (t1) after 30 min running exercise and remained at an elevated but not significant level (12.5% increase) for at least 120 min after (111.5 ± 31.06%) (Fig. 5B). An overview of the determined asprosin serum levels determined in this cohort is given in Supplementary Table S4.
Asprosin levels in serum correlate with the cartilage marker COMP upon hip replacement surgery. Since we found asprosin to be secreted by primary human chondrocytes (Fig. 4A), we were interested whether serum asprosin levels correlate with cartilage related pathologies. Therefore, we measured serum asprosin levels of patients (n = 14) diagnosed with symptomatic hip osteoarthritis (OA) prior and after total hip replacement (THR) surgery (Fig. 6). We previously reported that COMP, a biomarker of cartilage degeneration and osteoarthritis showed a reduction, one week (t1) after hip joint replacement surgery, and returned to pre-operative serum levels three months (t2), and one year (t3) post-surgery 21 . Our measurements revealed that serum asprosin levels strongly correlated with COMP levels, showing a 50% reduction at one week post-surgery before reaching 90% of pre-operative levels after three months (t2) and up to one year (t3) post-surgery (Fig. 6A

Analysis of asprosin localization suggests its storage in chondrocytes and cartilage tissue. As-
prosin localization in primary human chondrocyte culture. To further investigate the possibility of local asprosin storage in cartilage, primary human chondrocyte cultures were analyzed by immunofluorescence using pc-asp antibody (Fig. 7A). In primary human chondrocytes asprosin showed predominantly two staining patterns: a slight diffuse signal distributed in cytoplasm and in cytoplasmic processes, and intensive vesicle like structures, distributed predominately in the cytoplasm (Fig. 7A). In most cells, both types of signals were observed simultaneously. However, an increase in signal intensity and the formation of fiber-like structures were observed in the periphery of some cytoplasmic processes (Fig. 7A), whereas less closely spaced vesicle-like structures were also detected around or in the vicinity of the cells. www.nature.com/scientificreports/ Asprosin localization in human cartilage. Furthermore, we investigated asprosin localization in human cartilage. The immunoreactivity of pc-asp antibody was observed, independent of the used fixatives, as an intensive signal localized to chondrocytes and their pericellular microenvironment 22 (Fig. 7B, Supplementary Fig. S7).
The immunoreactivity was detected in chondrocytes distributed throughout all cartilage zones. Close inspection of chondrocytes in acetone treated cryosections revealed, similar to the analysis of cultured primary chondrocytes, diffuse distributed signals surrounding the nuclei and vesicle like structures. Accumulation of vesicle like structures was observed perinuclear or in the area comparable to the lacunar space. It is noteworthy that scattered vesicles were detectable, albeit to a much lower degree and signal intensity, in the interterritorial regions. Compared to only acetone treated cryosections, the signal intensity in acetone-metanol fixed specimens appeared weaker. In 4% PFA fixed cryosections a predominantly diffuse staining pattern was seen (Supplementary Fig. S7). To increase the availability of pc-asp epitopes, we performed enzymatic antigen retrieval with empirically selected proteases. For this purpose, digestion solutions containing pepsin, hyaluronidase, proteinase K and their combination or collagenase type 1 were applied to acetone-treated sections. As expected, treatment of cartilage sections with the different enzymes separately or in combination resulted in a broader signal distribution of asprosin within chondrocyte associated areas. The strongest intensity and the largest area of chondrocyte associated asprosin signal distribution were observed when sections were treated with pepsin or enzyme mixtures ( Supplementary Fig. S8). However, enzyme treatment resulted in a low preservation of morphological structures. When the sections were treated with collagenase type 1, a better preservation of cell nuclei and a significant increase of the asprosin signal were observed. Analysis of collagenase type 1 treated sections at higher magnification and in Z-stacks showed clear perinuclear signal accumulation while accumulations or clusters of vesicles-like structures distributed in the periphery of nuclei revealed a staining pattern comparable to chondrons 22 (Fig. 7C).
*** serum asprosin in OA patients before/ after total hip replacement surgery relative serum asprosin in OA patients before/ after total hip replacement surgery serum COMP in OA patients before/ after total hip replacement surgery relative serum COMP in OA patients before/ after total hip replacement surgery A B asprosin (ng/ml)

Discussion
Since its discovery as a new metabolic hormone, several clinical studies have been conducted to establish a correlation of asprosin levels with obesity, diabetes mellitus, insulin resistance, metabolic syndrome, or diabetic nephropathy 4,6 (Supplementary Table S1). Pathologically elevated asprosin levels were reported in patients with obesity 7,16 , insulin resistance 23,24 , and diabetes mellitus type 1 25 and type 2 [26][27][28] . Therefore, current research attempts focus on the possibility of using anti-asprosin antibody therapy to regain metabolic health 3 . However, the considerable variability of reported asprosin levels in human serum, plasma, and saliva, ranging from < 0.5 to > 350 ng/ml raises serious concerns regarding the reliability of sensitive asprosin measurements (Supplementary Table S1). Without the availability of reliable methods for asprosin detection in clinical samples any informative correlation studies or future administration of effective anti-asprosin therapy remain futile. This prompted us to establish new biochemical approaches for the sensitive and reliable measurement of asprosin in clinical samples. Analysis of human asprosin produced by human embryonic kidney cells revealed a significant amount of N-terminal glycosylation similar to physiological asprosin circulating in human blood (Fig. 3C) which contributes to about 50% of its molecular mass. As shown in yeast 14 , our analysis demonstrated that all three predicted N-glycosylation sites (N 3 , N 19 , and N 36 ) were utilized ( Supplementary Fig. S3). Our analysis also showed that additional O-glycosylation of asprosin is likely (Supplementary Fig. S2). Currently, the function of asprosin glycosylation is not clear, however, similar to other proteins it is possible that it keeps asprosin more soluble and may therefore prevent its aggregation in blood 29 . By employing our newly raised polyclonal asprosin antibody we established sensitive direct and sandwich ELISAs with a reliable detection below 0.15 ng/ml which is comparable to the stated sensitivity of commercially available ELISA kits used in previous studies 5 (Supplementary Table S1). The high affinity of the generated pc-asp (K D ~ 300 pM) also enabled us to selectively concentrate asprosin from complex clinical samples via antibody affinity chromatography. Using this approach we were able to determine asprosin in urine for the first time (Fig. 3D) and retrieved sufficient amounts from serum for subsequent detection by western blotting (Fig. 3C).
Our measurements suggest the following range of asprosin levels in clinical samples: serum (9-20 ng/ml), plasma (9-14 ng/ml), saliva (0.5-2 ng/ml), and breast milk (1-3 ng/ml) ( Table 1). Asprosin serum and plasma levels determined by us are in agreement with ranges previously measured in large patient cohorts with over 100 individuals [30][31][32] (Supplementary Table S1), suggesting that these ranges represent physiological relevant asprosin concentrations. Our investigations also provided a possible explanation for the strong deviation of reported serum and plasma asprosin levels by other studies (Supplementary Table S1). The first report about blood asprosin levels was by Romere et al. 1 who employed similar to our sandwich ELISA, a combination of polyclonal and mouse mab (capture: mouse mab against human profibrillin amino acids 2838-2865; detector: polyclonal goat against human profibrillin amino acids 2737-2750 by Abnova). However, two additional studies using the same detector but different capture antibodies reported serum asprosin levels ranging from 10-220 to 307-7454 ng/ml 7,33 . This indicates that the choice of the capture antibody is critical for obtaining reliable asprosin measurements. Also using non-glycosylated recombinant asprosin produced in E. Coli for standard curve 1 generation may result in artificially skewed concentration ranges. Our biochemical investigations showed that glycosylation of asprosin may affect detection sensitivity in ELISA and western blot analysis (Fig. 1G, Supplementary Fig. S5H,I). For instance, ELISA detection sensitivity of deglycosylated asprosin by mab (clone Birdy-1) appeared to be increased ( Supplementary Fig. S5H, right). Therefore, it is plausible that differences in detection sensitivity between E.Coli produced asprosin standards and glycosylated physiological asprosin are the underlying cause for the enormous variance of reported asprosin values. Despite 92% sequence identity between murine and human asprosin, our polyclonal antibody raised against folded and glycosylated full length human asprosin did not cross-react with murine asprosin (Supplementary Fig. S5D). Also, we so far failed to reliably detect asprosin amounts in mouse serum or plasma samples by using commercially available www.nature.com/scientificreports/ sandwich ELISA kits (unpublished results). This suggests that the availability of surface epitopes for human and murine asprosin differs. Overall, our data indicate that reported absolute asprosin levels in humans and mice should be considered with caution. For instance, it was reported that plasma asprosin levels display a circadian oscillation in mice and vary from 8 to 14 nM 1 . However, similar human and murine plasma asprosin levels in the range of 5-14 nM as reported by Romere et al. 1 would correspond to 185-440 ng/ml which are at least one order of magnitude higher compared to our determined values (Table 1) and those recently measured in larger cohorts (Supplementary Table S1). Our developed sandwich ELISA showed sufficient sensitivity to reliably detect asprosin at lower amounts in the range of 1-2 ng/ml. This allowed us to measure asprosin in breast milk, for the first time. Asprosin levels in breast milk might be an indicator of the metabolic health status of the mother, however, it may also contribute to the mother-infant communication. In the context of perinatal programming, assessing asprosin levels in breast milk might be informative, since it could be shown that other metabolic hormones like leptin, ghrelin, insulin and GLP-1 are present in breast milk and frequently correlate with the maternal BMI 34 . Therefore, asprosin concentrations in breast milk may also depend on the maternal BMI and elicit metabolic effects in the neonate.
Our serum measurements showed that asprosin was significantly elevated in females (Fig. 4B), which is similar to previous findings 35 . Currently, there is only limited information available regarding sex-dependent expression rates of fibrillin-1 from which asprosin is derived after furin-mediated cleavage within the C-terminal region. However, analysis of fibrillin-1 mutant mice revealed sex-dependent differences in thoracic aorta contractility, as well as aortic media injury 36,37 . Interestingly, addition of 17β-estradiol was found to promote fibrillin-1 production by human aortic smooth muscle cells, but not by fibroblasts 38 , suggesting a sex-and tissue-dependent regulation of fibrillin-1 expression which may explain the observed increase of asprosin in serum samples from females.
Saliva asprosin concentration is also increased upon fasting which strongly correlated with serum asprosin levels (Fig. 4C,D). Due to the high sensitivity and low intra-and inter-assay variability compared to commercially available asprosin detection kits (Supplementary Table S1), our established sandwich ELISA may be routinely employed for the analysis of saliva as useful alternative to serum samples as its collection is simple, non-invasive and painless 39,40 . Recently, a correlation of saliva with serum asprosin was already demonstrated and seems to be BMI-dependent 16 . However, the reported asprosin levels in saliva from donors with normal BMI similar to our tested cohort was reported to be around 30 ng/ml, which is 30-times higher than what we observed on average under fasting conditions (Fig. 4C, Table 1). Also recently, asprosin saliva amounts were reported to be about 25 ng/ml in a normal control group using the same commercially available ELISA kit 41 . Unfortunately, the kit used by both studies (Sunred Bioscience) is no longer commercially available, so that detailed information on its components and characteristics could not be retrieved.
There has been conflicting evidence reported regarding the effect of exercise on asprosin blood levels. The first report stated a sex-dependent increase of asprosin upon acute anaerobic exercise 19 . Serum levels of asprosin were found to be elevated only in women 15 min, 30 min, 60 min, and 24 h after a 20 s bicycle exercise. However, aerobic exercise in diabetic male rats led to a decrease of hepatic asprosin levels 42 . Recently, a significant decrease in serum asprosin was also observed in adult male individuals subjected to an aerobic exercise protocol in moderate intensity for 30 min at two different time periods of the day (morning: 08:00-10:00 h, evening: 20.00-22.00 h) at least 3 d apart. A significant decrease in asprosin serum levels was observed in normal and overweight, or obese participants of the study 18 . However, a comparison of absolute blood asprosin amounts determined in both studies raises concern about the informative value of the data. While Wiecek et al. 19 report plasma asprosin levels in the range of 106-362 ng/ml, Ceylan et al. 18 report serum asprosin levels within 0.53-1.17 ng/ml (Supplementary Table S1). Recent studies conducted with large cohorts including > 100 BMI, age, and biological sex matched controls showed that values for serum and plasma asprosin are in the same range as determined by our measurements (13-32 ng/ml) ( Table 1 and Supplementary Table S1), but out of the reported ranges by both previously published exercise studies 18,19 .
Our findings show a significant increase of asprosin after 30 min of running exercise (Fig. 5). Before the experiment, which took place in the morning, subjects were asked not to exercise for 24 h. Participants remained seated for 30 min before the first blood sample (baseline) was drawn. Our results clearly showed that upon acute running exercise asprosin release into the blood is induced and remains stable for at least 2 h afterwards. So far it is not understood by which mechanism running exercise leads to increased asprosin levels. Previously, plasma asprosin levels were reported to be increased upon a 20 s bicycle sprint in women 19 which represents an exercise regime with less mechanical loading compared to treadmill running. Since these increased asprosin levels upon a low impact bicycle sprint also correlated with irisin, an exercise-induced myocyte produced cytokine 43 , and fibrillin-1 is known to be ubiquitously expressed in connective tissues including skeletal muscle 1,44 , muscle tissue was suggested as source of asprosin secretion 19 . However, since irisin and asprosin are also produced by adipocytes 1,45 it remains difficult to assess the tissue-specific contribution to circulating asprosin levels in blood upon exercise. Measurement of asprosin in combination with tissue-specific markers may allow dissecting which musculoskeletal compartment might be challenged by a certain exercise regimen and activate asprosin synthesis or release. In this context, it is also not clear how the applied mechanical impact due to the chosen exercise regimen may be implicated in altered asprosin synthesis or release. Mechanical loading could result in cellular adaptation leading to anabolic reactions and asprosin synthesis. Alternatively, mechanical loading may lead to connective tissue degradation and thereby cause release of tissue resident asprosin into the circulation. Mechanical loading of joints due to exercise leads to degradation of cartilage and release of cartilage extracellular matrix components such as COMP into the serum 20,46 . We previously reported that COMP levels increased upon running exercise in the same cohort 20 . If asprosin is a component of cartilage and produced by chondrocytes (Fig. 4A) www.nature.com/scientificreports/ patients before and after THR (Fig. 6). Our analysis showed a clear correlation of asprosin and COMP serum levels in both the exercise cohort and the THR cohort suggesting asprosin as component of cartilage, for the first time. Fibrillin-1 is known to be a component of cartilage and where it assembles into supramolecular microfibrils that structurally contribute to the cellular microenevironment of chondrocytes 47 . It is therefore possible that also aprosin is locally stored within cartilage and released upon mechanical loading.
To further investigate the idea of local asprosin storage in cartilage, we analyzed asprosin localization in primary human chondrocyte culture and cartilage tissue (Fig. 7). Close inspection of the signal distribution in cultured primary chondrocytes and in chondrocytes embedded in cartilage tissue revealed two similar staining patterns that appeared as diffuse signals in the cytoplasm and in cytoplasmic processes, as well as vesicle-like structures (Fig. 7). Interestingly, in the periphery of some cytoplasmic processes detected signals were accumulated to fiber-like structures (Fig. 7A). Cytoplasmic processes in chondrocytes were described to be formed as active chondrocyte responses due to changes in the ECM that can result in release of cytokines or degradative enzymes 48 . It appears therefore plausible that asprosin may be released into the extracellular space upon mechanical loading. Within each zone of articular cartilage the composition of the ECM is region-specific 49 . This property of the matrix can significantly influence the results of immunohistochemistry, for example by inhibiting the penetration of antibodies 50 . We therefore performed additional pretreatment of cartilage sections with different enzymes partially removing collagen fibers or proteoglycans. Enzyme treatment indicated that collagens, hyaluronan and other ECM components contribute to masking of asprosin detection (Fig. 7C). However, this finding still supports the hypothesis that asprosin is secreted by chondrocytes and deposited within their pericellular matrix.
Our results support the idea of local asprosin storage within tissue-specific microenvironments. Potential extracellular storage platforms could be proteoglycans or other ECM supramolecular scaffolds such as the fibronectin/elastic fiber network which have been known to act as sinks or targeting scaffolds for connective tissue derived growth factors 11,51 . Asprosin stored in tissues may function as sensor for local energy demand. Experiments with murine myoblasts (C2C12) showed that asprosin is not only able to interfere with muscle cell insulin sensitivity 52 , but also up-regulates glucose transporter 4 (GLUT4) expression in C2C12 myotubes and thereby enhance local glucose uptake by tissue muscle cells 14 .
More extensive studies are required to investigate how asprosin is released into different body fluids and what exact function it serves there. Also, the mechanisms of asprosin release upon exercise require further investigation, especially considering different exercise regimen from low to high mechanical loading. Another important question is whether asprosin release is influenced by sex hormone levels which may explain the sex-specific differences we observed. Overall, our results suggest that reported absolute asprosin levels in clinical samples have to be considered with caution, since used commercially available ELISAs show a considerable variation in detection sensitivity. Our biochemical approaches and generated tools will now allow in-depth analysis of asprosin biological functions on a molecular level.

Materials and methods
Ethics statement. All research on human tissues was performed in accordance with local regulations and was approved by the Human Ethics Committee of the University Hospital of Cologne (No. 18/124, No. 15/368, No. 14-422), the local ethics committee of the Goethe University Frankfurt in Germany (No. 497/15) and the German Sport University Cologne Germany 20 . Written Informed consent was obtained from all participating probands and patients who agreed to have materials examined for research purposes. The study was conducted in accordance with the Declaration of Helsinki.
Expression and purification of proteins. Human asprosin (S 2732 -H 2871 ) and human placensin (S 2780 -Y 2912 ) DNA sequences were generated by GeneArt Strings DNA Fragments service (Thermo Fisher Scientific, Massachusetts, USA) including the restriction enzymes sequences for NheI at the 5′ and XhoI at the 3′ ends. Double strained synthetic cDNAs encoding for the respective protein sequences were cloned via the Nhe1/ Xho1 sites of the modified pCEP-Pu vector containing an N-terminal BM-40 signal peptide and a C-terminal double Strep-tag II downstream of the Xho1 restriction site 53 . Ligated vectors were transformed into One Shot TOP10 Chemically Competent E. coli (Thermo Fisher Scientific, Massachusetts, USA), and clones were verified by Sanger sequencing (Microsynth Seqlab GmbH, Göttingen, Germany). HEK-293 EBNA cells cultured in Dulbecco's modified eagle medium (DMEM) (Thermo Fisher Scientific, Massachusetts, USA) were transfected with asprosin or placensin expression vectors by using FuGENE HD Transfection Reagent (Promega GmbH, Germany) according to the manufacturer's instructions. One day after transfection, cells were selected by supplementing the medium with puromycin (1 µg/ml). The cells were expanded for large scale production in TripleFlask Cell Culture Flasks (Thermo Fisher Scientific, Massachusetts, USA). At 90% confluency, cells were incubated with serum-free DMEM containing puromycin (0.5 µg/ml) and collected after 48 h. Collected medium was filtered with a suitable membrane filter, then subjected to Strep-Tactin XT gravity flow columns (2 ml beads; IBA GmbH, Germany) at 4 °C overnight. Asprosin or placensin were eluted with elution buffer (100 mM Tris/HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA, 2.5 mM desthiobiotin). The collected fractions were analyzed by Coomassie Blue Staining, concentrated and the buffer exchanged to PBS by using Amicon Ultra Centrifugal Filters (cut-off: 3 kDa) (Merck Millipore, Burlington, MA, USA). The N-terminal half of fibrillin-1 (rF90, amino acids positions: M 1 -V 1527 ) with a C-terminal His 6 -tag was overexpressed in HEK-293 EBNA cells followed by purification as described previously 54  www.nature.com/scientificreports/ Antibodies. The pc-asp anti-asprosin antibody was generated against full-length human asprosin. Recombinantly expressed human asprosin was used to immunize a rabbit for antibody production (Pineda Antibody Service, Berlin, Germany). Preimmune serum (0.5 ml) was obtained before the immunization of the rabbit and did not show cross reactivity to asprosin tested by ELISA. The rabbit was immunized with a total amount of 500 µg asprosin in five consecutive immunizations at days 1, 14, 28, 42, and 56. The antiserum was obtained 60 days after immunization. The antiserum was affinity purified on a column of CNBr-activated Sepharose 4B (Cytiva, Uppsala, Sweden) conjugated with the recombinantly expressed asprosin (300 µg) according to the manufacturer's instructions. The antibody was eluted with 0.1 M glycine (pH 2.5) and neutralized with 3 M Tris/ HCl, pH 8. The eluted antibody was concentrated by using Amicon Ultra Centrifugal Filters (cut-off: 10 kDa), (starting volume: 6 ml, end volume: 1 ml). The antibody was used as a capture antibody for sandwich ELISA at 3 µg/ml, at a 1:5000 dilution (0.2 µg/ml) for western blot, and at a 1:70 dilution (14 µg/ml) for cells and tissues immunofluorescence. Mab anti-asprosin antibody (clone Birdy-1¸ AG-20B-0073) (AdipoGen Life Sciences Inc., San Diego, USA) was used at a 1:1000 dilution (1 µg/ml) for western blot and at a 1:2000 dilution (0.5 µg/ml) as detection antibody in sandwich ELISA. Fibrillin-1 polyclonal antibody, raised against C-terminally His 6 -tagged rF90 representing the N-terminal half of fibrillin1 51 , was used at a 1:5000 dilution (0.5 µg/ml) for western blot. Fibrillin-1 rabbit monoclonal antibody (CPTC-FBN1-3) (DSHB, Iowa, USA), raised against the synthetic peptide: TCVDINECLLEPR, which is contained within the C-terminal half of fibrillin-1, was used at a 1:2500 dilution (1 µg/ml).
Protein quantification. Concentrations of proteins, which were used in this study including pc-asp antibody were colorimetrically measured by BCA Protein Assay Kit (Thermo Fisher Scientific, Massachusetts, USA). In addition to the BCA assay, the concentrations of purified asprosin and placensin were spectrophotometrically measured by NanoDrop One (Thermo Fisher Scientific, Massachusetts, USA) after buffer exchange into PBS using Amicon Ultra Centrifugal Filters (cut off: 3 kDa). The protein quantification was performed using the program "Protein A280" which measures the protein absorbance at 280 nm and calculates the concentration depending on the protein extinction coefficient. The predicted extinction coefficient for human asprosin-2 × Strep-tag II:

Characterization of pc-asp anti-asprosin antibody.
To test the specificity and sensitivity of pc-asp anti-asprosin antibody, recombinant asprosin, N-terminal fibrillin-1 (rF90), C-terminal fibrillin-1 (rF6), human albumin, and human IgG (Sigma-Aldrich, Darmstadt, Germany) were diluted in PBS and coated to the wells of a 96-well ELISA plate in 100 µl/well followed by overnight incubation at 4 °C. After washing with PBS containing 0.05% Tween 20 (PBST), the wells were incubated with 5% milk in PBS for 2 h for blocking. Purified pc-asp antiasprosin antibody was diluted 1:10,000 in 2.5% milk in PBS and applied to the wells for 1 h at RT. The wells were washed three times with PBST, and mouse anti-rabbit IgG-HRP conjugate (#211-032-171, Jackson ImmunoResearch, Pennsylvania, USA) was then added at dilution 1:2000 in 2.5% milk in PBS and incubated for 30 min at RT. After washing the wells with PBS, the signal was developed by adding 1-Step Ultra TMB-ELISA Substrate Solution (Thermo Fisher Scientific, Massachusetts, United States) followed by an incubation for 1-3 min. The reaction was stopped with an equal volume of 30% H 2 SO 4 (stopping solution) and the optical density (OD) was measured 450 nm. www.nature.com/scientificreports/ plateau and further processed. A piece of cartilage was dissected using a scalpel and embedded in OCT (Thermo Fisher Scientific, Massachusetts, USA), snap frozen and kept in -80 °C freezer until cryosectioning.
Immunofluorescence and toluidine blue staining. To identify the optimal tissue fixation and treatment for asprosin detection by the pc-asp anti-asprosin antibody, air-dried cartilage cryosections were subjected to one of the following preparations: (1)  For immunofluorescence chondrocyte cultures, chondrocytes were seeded on uncoated glass coverslips at a density of 1 × 10 5 cells per well in 12-well plates. Cells were kept in culture for 12 days, then washed with PBS for 5 min, fixed with ice cold acetone:methanol (1:1) for 10 min at − 20 °C, blocked with 1% bovine serum albumin (#11,930.03, SERVA Electrophoresis GmbH, Heidelberg, Germany) in PBS, and then incubated with pc-asp antiasprosin antibody (1:70) in 0.5% BSA in PBS for 1 h at room temperature. The coverslips were washed two times with PBST for 5 min and incubated with secondary antibody goat anti-rabbit Alexa 555 (# A27039, Invitrogen, Thermo Fisher Scientific, Massachusetts, USA) for 30 min in dark at room temperature, then washed two times with PBST for 1 min and one time with PBS for 1 min. Subsequently, coverslips were air-dried for 10 min and mounted to the slide using ProLong Gold Antifade Mountant with DAPI. Confocal images were obtained using a Leica SP8 confocal microscope (Leica Camera AG, Germany) and Leica LAS AF Lite 4.0 software. Images were further processed using Fiji/ImageJ software to obtain average Intensity Z-projection.
Cartilage cryosections with 10 µm thickness were stained with Toluidine Blue O (#1B481, Chroma-Gesellschaft Schmid GmbH & Co., Köngen, Germany) according to the following procedures: sections were dried for 1 h at 37 °C, washed in distilled water for three times, stained in Toluidine Blue O working solution for 2 min, washed in distilled water for 3 times, dehydrated quickly in 95% ethanol and then in 100% ethanol for two times, washed in xylol two times for 3 min, and coverslipped with EUKITT UV O. Kindler EUKITT UV mounting media (ORSAtec GmbH, Germany). Toluidine-stained cartilage sections were imaged with Zeiss AX10 light microscope equipped with Zeiss plan-Neofluar 10 × /0.3 objective, Hitachi HV-F202 camera and Diskus software (version number 5.0.6353 #7599).
Surface plasmon resonance. SPR experiments were performed as described previously 57 using a BIAcore 2000 system (BIAcore AB, Uppsala, Sweden). Pc-asp anti-asprosin antibody was immobilized at 8000 RUs to a CM5 sensor chip using the amine coupling kit following the manufacturer's instructions (Cytiva, Uppsala, Sweden). Interaction studies were performed by injecting 0-80 nM recombinant asprosin and RPE supernatant pulldown fractions in HBS-EP buffer (0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% (v/v) surfactant P20) (Cytiva, Uppsala, Sweden). Human asprosin was immobilized at 500 RUs to a CM5 sensor chip using the amine coupling kit. Interaction studies were performed by injecting 0-40 nM pc-asp, mab and Strep-ab in HBS-EP buffer. Kinetic constants were calculated by nonlinear fitting (1:1 interaction model with mass transfer) to the association and dissociation curves according to the manufacturer's instructions (BIAevaluation version 3.0 software). Apparent equilibrium dissociation constants (K D values) were then calculated as the ratio of k d /k a .
Exercise study. Fifteen male healthy participants, age = 27.5 ± 3.1 years, weight = 78 ± 7.7 kg, height = 1.81 ± 0.05 m, BMI = 23.9 ± 1.9 kg/m 2 were recruited by German Sport University Cologne, Cologne, Germany to study the effect of exercise on the serum asprosin level (Supplementary Table S7). Inclusion criteria were age from 20 to 35 years and physical fitness. Exclusion criteria were musculoskeletal disorder, acute and chronic injuries, trauma or surgery of lower extremities' joints. Subjects were asked not to exercise 24 h prior to the study and to come to the institute either by car or public transportation. The subjects remained seated for 30 min before the first blood sampling (t0). Further blood samples were drawn immediately afterwards (t1), and then 30 min (t2), 60 min (t3), and 120 min (t4) after running. Blood (5.0-7.5 ml) was collected by venipuncture www.nature.com/scientificreports/ using serum-gel monovettes1 (Sarstedt, Nümbrecht, Germany). After blood coagulation (30 min, RT), serum was isolated by centrifugation (10 min, 3000 rpm) and stored at − 80 °C.

Osteoarthritis patients study.
Fourteen patients diagnosed with unilateral and bilateral symptomatic hip OA, validated by conventional anterior/posterior and lateral pelvic radiographs (7 females, 7 males, age = 61.4 ± 10.5 years, height = 1.74 ± 0.07 m, weight = 79.7 ± 16.4 kg, BMI = 26.2 ± 4.7 kg/m 2 ) at the Department of Orthopaedics at the University Hospital of the Goethe University Frankfurt in Germany voluntarily participated in the study. All patients reported pain for the ipsilateral hip although, the contralateral hip was free from any symptoms before the surgery. At the time of all follow-up measurements all patients were pain-free. The period of prospective examination began before surgery and ended nearly 1 year after initial THR. The patients were recruited and examined as previously reported 21 . Anthropometric measures of the two analyzed cohorts in this study are shown in Supplementary Table S7.
Blood and saliva samples. Fasted blood (3-5 ml) and saliva samples (1-2 ml) were collected 8-9 h after overnight fasting and for non-fasted samples, they were collected 3 h after lunch meal. Serum separator tube (SST) was used to separate serum from the whole blood specimen. Samples were allowed to clot for 30 min before centrifugation for at least 15 min at 1000 × g. serum was removed, aliquoted and kept frozen at − 80 °C. For saliva samples, the required amount of saliva samples was collected, and subjects were asked to refrain from eating, drinking, smoking, chewing gum, brushing their teeth, or using mouthwash for at least 30 min before providing their samples. Samples were centrifuged for 5 min at 1000 × g and aliquoted and kept frozen at − 80 °C.
Statistical analysis. Data are expressed as mean ± SD. Statistical analyses were performed using GraphPad Prism software version 8.0.2 and the significance of differences between groups was determined by applying an unpaired two-tailed Student's test. Values of P ≤ 0.05 were considered significant. For the exercise study, one-way analysis of variance (ANOVA) with repeated measures and Dunnett's multiple comparisons test for post hoc analysis were performed to detect significant differences in asprosin levels between time points. The significance level was set at P ≤ 0.05.