Analytical Performance of a Standardized Kit for Mass Spectrometry-based Measurements of Human Glycosaminoglycans

Glycosaminoglycans (GAGs) are long linear sulfated polysaccharides implicated in processes linked to disease development such as mucopolysaccharidosis, respiratory failure, cancer, and viral infections, thereby serving as potential biomarkers. A successful clinical translation of GAGs as biomarkers depends on the availability of standardized GAG measurements. However, owing to the analytical complexity associated with the quantification of GAG concentration and structural composition, a standardized method to simultaneously measure multiple GAGs is missing. In this study, we sought to characterize the analytical performance of a ultra-high-performance liquid chromatography coupled with triple-quadrupole tandem mass spectrometry (UHPLC-MS/MS)-based kit for the quantification of 17 GAG disaccharides. The kit showed acceptable linearity, selectivity and specificity, accuracy and precision, and analyte stability in the absolute quantification of 15 GAG disaccharides. In native human samples, here using urine as a reference matrix, the analytical performance of the kit was acceptable for the quantification of CS disaccharides. Intra- and inter-laboratory tests performed in an external laboratory demonstrated robust reproducibility of GAG measurements showing that the kit was acceptably standardized. In conclusion, these results indicated that the UHPLC-MS/MS kit was standardized for the simultaneous measurement of GAG disaccharides allowing for comparability of measurements and enabling translational research. Summary Analytical performance of a kit for standardized GAG measurements, based on an established UHPLC-MS/MS method


7
the provided highest calibration level (the nominal concentration of the disaccharide at each 1 calibration level was provided in Table S1); a mixture of 17 purified disaccharides (one per analyte) 2 to aid the operator in the adjustment of eventual drifts in retention times, and four quality control 3 (QC) samples to be used in every MS run to monitor inter-sequence variability. In Figure 1B, 4 representative chromatograms are illustrated for the afore-mentioned mixture of 17 purified 5 disaccharides quantified using the kit (see Figure S1 for chromatograms derived from the standard 6 GAG solution). We characterized the analytical performance of the kit in terms of calibration capability and 7 performance in native human samples. 8

Characterization of the kit analytical performance: calibration 9
First, we characterized the calibration curve parameters: linearity, detection capability, selectivity 10 and specificity, accuracy and precision, carry-over, and disaccharide stability in the auto-sampler. 11 This process established the performance of the calibration curve over a specific range for each 12 disaccharide. 13 In the second part, we performed the characterization of the kit in terms of GAGs extraction, 14 detection, and quantification in native (human) samples by measuring the following parameters: 15 selectivity and specificity, recovery, matrix effect, linearity response, accuracy and precision, and 16 disaccharide stability. 17

Linearity and detection capability of the calibrators 18
We tested the linearity of the calibration curve for each disaccharide. We prepared nine levels of 19 calibration of each disaccharide in triplicates, injecting each replicate two times. We pre-specified 20 acceptance criteria for a level to be included in the final calibration curve in terms of acceptable 21 coefficient of variation (CV), which was required to be lower than 25% (30% for the lowest level), 22 and deviation with the respect to the nominal concentration after back-calculation, which was 1 required to be lower than 25%. We defined the upper limit of quantification (ULoQ) and low limit 2 of quantification (LLoQ) for each disaccharide as the highest and lowest calibration levels meeting 3 the acceptance criteria, respectively. We defined the range the linearity as the concentration 4 between LLoQ and ULoQ. 5 In Table 1, we reported for each disaccharide the number of calibration levels within the range of 6 linearity between LLoQ and ULoQ, including the coefficient of determination (R 2 ) of the 7 calibration curve and the coefficient of variation (CV) and deviation of back-calculated 8 concentration with respect to the nominal concentration at the LLoQ and ULoQ. 9 Table 1. Linearity of the calibration curve for each disaccharide including the number of calibrator 10 levels, coefficient of determination (R 2 ), and nominal concentration, coefficient of variation (CV), 11 and deviation from back-calculated concentration for the lower limit of quantification (LLoQ) and 12 upper limit of quantification (ULoQ For 15 of 17 disaccharides (all except Tris CS and 2s6s HS), we obtained a calibration curve with 1 acceptable linearity and detection capability. In cases where the calibration curve could not be 2 constructed (for Tris CS and 2s6s HS), GAG quantification relied on the ratio between the observed 3 peak area and the corresponding peak area at the highest level of the provided calibrator. 4

Selectivity and specificity of the calibrators 5
We tested the selectivity and specificity to each disaccharide by inspecting the presence of peaks 6 of an area greater than 20 % LLoQ for that disaccharide in blank samples. For each disaccharide, 7 no peak could be detected at the expected retention time for that disaccharide in any of the blank 8 11 samples. Selectivity and specificity for the kit were therefore deemed acceptable for all 1 disaccharides. 2

Accuracy and precision of the calibrators 3
We tested the accuracy and precision of the calibration curves by creating a set of three standard 4 GAG solutions at three different concentrations (low, medium, and high). In Table 2 and 3, we  5 reported the accuracy (percentage difference between nominal concentration for a disaccharide in 6 the standard GAG solution and measured concentration) over a 2-day experiment and the precision 7 (as CV) for each disaccharide at each standard GAG solution concentration (low, medium, and 8 high). 9     The accuracy was acceptable for virtually all detected CS and HA disaccharides at all 3 1 concentrations on both Day 1 and Day 2, as well as for most detected HS disaccharides at medium 2 and high concentrations on Day 1 (but not on Day 2). Similarly, both intraday and inter-day 3 precision were acceptable for all detected CS and HA disaccharides at all 3 concentrations as well 4 14 as for intraday precision for most detected HS disaccharides at all 3 concentrations. We attributed 1 the results reported in Tables 2 and 3 that did not meet the acceptance criteria to poor signal 2 acquisition. 3

Carry-over in the calibrators 4
We tested the impact of the carry-over by inspecting the presence of peaks of an area greater than 5 20 % LLoQ for that disaccharide in blank samples immediately after the acquisition of the highest 6 calibration curve level. For each disaccharide, no peak could be detected at the expected retention 7 time for that disaccharide in any of the blank samples. Carry-over was therefore considered 8 negligible for all disaccharides. 9

Disaccharide stability in the autosampler 10
We tested the stability of disaccharides stored in the autosampler at 10 °C by monitoring the peak 11 area of each analyte over 14 days at the third-highest calibration curve level for that disaccharide. 12 In Table S2, we reported the change in peak areas for a given disaccharide at a given time point 13 relative to the corresponding peak area at the initial time point. 14 The stability in the auto-sampler of all disaccharides except Ns2s HS was acceptable over 6 days 15 (at least 5 of 6-time points with <30% deviation from Day 0). We found that all disaccharides 16 except 4s CS had poor stability after 14 days in the autosampler. 17

Characterization of the kit analytical performance: concentration estimation 18
Having established the performance of the calibration curve for each disaccharide, we could 19 estimate disaccharide concentrations using the so-characterized calibration curves. In the second 20 part, we, therefore, performed the characterization of the kit in terms of estimation of the 21 disaccharide concentration in native samples by measuring the following performance parameters: 22 1 disaccharide stability. For the sake of consistency, we chose human urine as the reference matrix 2 for the native samples. 3

Recovery 4
We tested recovery by first generating a GAG-depleted sample (referred to as proxy urine, SI 5 Methods); and next by spiking the proxy urine with a set of three standard GAG solutions at three 6 different concentrations (low, medium, and high) either at the beginning of the sample preparation 7 or immediately after filtration during sample preparation. In Table S3, we reported the recovery of 8 each disaccharide after filtration across five replicates. 9 The results indicated acceptable recovery (<25% deviation) for all detectable CS disaccharides at 10 all three concentration levels (except 2s CS, acceptable only at the highest level); for all detectable 11 HS disaccharides at the "low" and "medium" concentration levels (except for 2s HS); and for HA 12 at the "medium" concentration level. 13

Matrix effect 14
We tested matrix effects to evaluate how endogenous compounds in native urine interfered with 15 the measurement of the analytes. We spiked proxy urine from 6 healthy donors and milli-Q water 16 samples in triplicates with the above-described set of standard GAG solutions at two concentration 17 levels, "high" and low". In Table S4, we reported the matrix effect as the ratio (in %) between the 18 disaccharide concentration in 6 proxy urine samples versus the milli-Q water sample at the two 19 concentration levels, as well as the average matrix effect in proxy urine. 20 The matrix effect was moderate (between 42% and 89%) in CS disaccharides at the "high" level 21 and low (78% to 103%) at the "low" level, indicative of signal suppression due to matrix effects. 22 We found moderate-to-high matrix effects in HA and HS disaccharides at the "high" level and 1 moderate-to-low at the "low" level. 2

Accuracy and precision in native samples 3
We tested the accuracy of disaccharide concentrations in native samples by spiking the proxy with 4 the same set of standard GAG solutions as described above at three different concentration levels 5 (low, medium, and high). In Table 4, we reported the accuracy in the concentration of each CS 6 disaccharide in a three-day experiment by two operators on a given analysis day. We computed the 7 intraday precision in terms of coefficient of variation from the same data and reported in Table 5. 8 We then computed the intraday precision as the coefficient of variation from the same data and 9 reported in Table 5. 10 We were unable to reliably determine the accuracy and precision for HA and HS disaccharides 11 because their estimated concentrations were close to the respective LLoQ even for the "high" level 12 standard GAG solutions. We attributed this partly to matrix effects as later discussed. 13    The accuracy in native samples was acceptable (<30% deviation from the nominal concentration) 1 for all detectable CS disaccharides except for 0s CS, where the accuracy was below 70% in 2 "medium" and "high" level samples -albeit never below 60%. The intraday precision in native 3 samples was acceptable (CV < 25%) for all detectable CS disaccharides. 4 5 We tested the stability of each disaccharide in the native matrix over 14 days. Specifically, we 6 prepared two proxy urine samples and spiked them with the above-described set of standard GAG 7 solutions at two concentration levels ("high" and low"). We stored the spiked samples at -20 °C 8 and measured disaccharide concentrations on day 1 and day 14. In Table S5, we reported the 9 percentage difference in the disaccharide concentration at a given level ("high" or "low" at Day 1 10 and Day 14 compared to the corresponding nominal concentration. 11

Disaccharide stability in native samples
The stability in native samples was acceptable (<30% deviation from the nominal concentration) 12 for all detectable CS disaccharides on Day 1 and for 4 of 6 detectable CS disaccharides on Day 14, 13 where the remaining two CS disaccharides (0s CS and 4s6s CS) deviated from the nominal 14 concentrations between 33% and 37%, respectively. Note that the stability for HA and HS 15 disaccharides could not be reliably estimated in this experiment and it was omitted. 16

Selectivity and specificity in native samples 17
We tested selectivity and specificity to each disaccharide by inspecting the presence of peaks of an 18 area greater than 20 % LLoQ in proxy urine -without any spiked GAG solution. For each 19 disaccharide, no peak could be detected in proxy urine at the expected retention time for that 20 disaccharide. Selectivity and specificity in native samples were therefore acceptable for all 1 disaccharides. 2

Linearity in native samples 3
We tested the linearity of disaccharide concentrations in native samples by spiking proxy with the 4 set of standard GAG solutions as described above at nine different concentration levels. In Table  5 6, we reported the linearity for each disaccharide in terms of coefficient of determination (R 2 ) for 6 the linear regression between peak areas and concentration across the nine levels. 7 Table 6. Linearity of disaccharides in proxy urine samples (in terms of coefficient of determination 8 R 2 between peak areas and concentration levels) of a standard GAG solution serially diluted to 9 generate nine concentration levels. Note that stability for HA and HS disaccharides could not be 10 reliably estimated in this experiment and it was omitted. Acceptable values marked in bold. The linearity was acceptable for all detectable CS disaccharides (R 2 > 0.95). Note that we were 1 unable to reliably estimate the linearity for HA and HS disaccharides. 2

External validation of kit analytical performance 3
We sought to validate the hereby presented analytical performance specifications by performing 4 tests of intra-laboratory and interlaboratory precision in a GLP-compliant external laboratory 5 (Lablytica Life Science AB, Uppsala, Sweden). 6

Intra-laboratory precision 7
We tested intra-laboratory precision by monitoring the disaccharide concentration in four QC 8 samples included in the kit throughout 14 independent experiments (runs). We focused on the two 9 properties of the GAG profile, non-sulfated CS (0s CS) and total CS, namely the sum of all 10 measured CS disaccharides in a sample. Two QC samples were synthetic samples spiked with 11 standard GAG solutions at high or low concentration (~14.5 µg mL -1 and ~1.7 µg mL -1 for total 12 CS, respectively). The other two QC samples were native samples with known high or low total 13 GAG concentration (~12.5 µg mL -1 and ~4.2 µg mL -1 for total CS, respectively). In Figure 2, the 14 estimated concentration of the two key GAG properties (total CS and 0s CS) were plotted across 15 runs. The concentrations for all measured CS GAGs were shown in Figure S2. In Table S6, we  16 reported the CV for each disaccharide in the 4 QC samples. The intra-laboratory precision was acceptable for all major CS disaccharides. The precision for HS 5 disaccharides was acceptable only in "low" concentration samples, while for HA it was acceptable 6 only in synthetic QC samples. In general, the concentration for di-sulfated and tri-sulfated 7 disaccharides as well as 2s disaccharides was below LLoQ in all QC samples and therefore 8 precision estimates should be interpreted with caution. 9

Inter-laboratory precision 10
We tested inter-laboratory precision by comparing the disaccharide concentration of two key GAG 11 properties (total CS and 0s CS) in a panel of nine native urine samples from healthy donors 12 independently analyzed in the reference laboratory versus the external laboratory. We found that 1 the total CS and 0s CS concentration estimates from the reference laboratory versus the external 2 laboratory, were strongly correlated (Pearson correlation coefficient R > 0.95) (Figure 3) GAGs non-invasively, for example in blood and urine (5)(6)(7)(8)29). However, moving beyond 1 biomarker discovery to clinical translation necessitates the standardization of GAG measurements. 2 UHPLC-MS/MS has emerged in the last decade as the gold standard for rapid quantification of 3 GAG concentration and disaccharide composition. Many methods were published that illustrated 4 efficient separation of up to 19 disaccharides in biological samples (11-23). However, none of 5 these reported extensive analytical performance testing. Therefore, here we comprehensively 6 characterized the analytical performance of a kit for GAG measurements using an UHPLC-MS/MS 7 method which has been previously described (12). 8 Overall, the kit efficiently separated 17 disaccharides and exhibited excellent selectivity and 9 specificity to all disaccharides with negligible carryover and sufficient stability for typical 10 laboratory work shifts. The calibrators were accurate and precise for 15 of 17 disaccharides over a 11 range of concentrations covering approximately one-order of magnitude for each disaccharide. Compared to published methods, the here-characterized kit for absolute quantification of GAGs 18 adequately was found to be capable to calibrate as many as 15 disaccharides simultaneously. 19 In native samples, here using urine as the reference matrix, we demonstrated the robust and accurate 20 analytical performance characteristics of the kit. Critically for analyses of biological and clinical 21 samples, the kit enabled the quantification of CS disaccharides within a calibration range that 22 captured physiological values spanning one order of magnitude. We were unable to validate the 1 results on HA and HS disaccharides because the concentrations recovered in urine were below the 2 LLoQ for virtually all of these disaccharides. This could simply reflect a low abundance of urinary 3 HA and HS in physiological conditions compared to CS. The hypothesis is in line with previous 4 reports in which the total HA and HS concentration in urine were measured as ~20% of the total 5 CS concentration, almost an order of magnitude less abundant (8, 18). Nevertheless, we could not 6 rule out the hypothesis that the kit was underperforming in the quantification of HA and HS 7 disaccharides in the urine. We attributed one possibility to the here-observed matrix effects, which 8 showed moderate to strong peak area suppression in HA and HS in the urine. Another explanation 9 could be a less efficient extraction yield during enzymatic digestion of HA and HS disaccharides 10 as compared to CS disaccharides. Overall, we showed that the kit had an acceptable analytical 11 performance for the quantification of CS disaccharides in native samples, while its performance in 12 HA and HS warranted further investigation. 13 We deemed the kit standardized for CS measurements given that the intra-laboratory and inter-14 laboratory precision tests produced acceptable results across two independent laboratories. 15 Specifically, the here-described kit proved capable to simultaneously calibrate 15 of 17 16 disaccharides with high linearity (R 2 > 0.99 for all disaccharides except 6s CS wherein R 2 = 0.98) 17 and with high intra-laboratory precision using 14 replicates of four control urine samples (CV 18 ranging 8 to 22% at "high" concentration and 8 to 42% at "low" concentration for the major CS 19 disaccharides, here defined as >5% of total CS). In comparison, Tomatsu et al. (2014) reported a 20 method that could quantify 7 disaccharides (2 keratan sulfate disaccharide, 3 HS, and 2 CS) with 21 linearity R 2 ranging 0.982 to 0.993 across two orders of magnitude and with an intra-assay precision 22