A substantial amount of optimization was performed (outlined in Supplementary Methods) to adapt the existing LC-MS assays to the AEMS platform for multiplexed protein quantification. Carrier solvent composition, flow rate, and sample ejection volume were optimized using a commercial standard peptide mixture (PepCalMix, SCIEX, Concord) in a simple protein digest (Beta-Galactosidase) to mimic anticipated sample complexities post-SISCAPA enrichment. Fine-tuning of the ejection volume was conducted, optimizing for peak shape and maintaining linearity between ejection volume and MRM peak area (Supplementary Fig. 2). Finally, the tolerance towards reagents used in the original (e.g. PBS, CHAPS) and modified SISCAPA protocol (ABC) was evaluated by investigating a range of concentrations on the ejection peaks (Supplementary Fig. 3). The narrow ejection peaks of AEMS necessitated careful optimization of the MS method. The chosen strategy involved 4 MRMs per method with a 10 ms dwell time, ensuring adequate data points across the ejection peak for precise quantification (Supplementary Fig. 4). Finally, these optimized parameters were used to determine the LLOQs for the PepCalMix peptides on two different AEMS systems, and good linearity was observed for 19/20 peptides (average R2 of 0.997) with average LLOQs of 260 and 520 amol/µL, respectively (Supplementary Fig. 5 and Supplementary Table 6). A similar approach was applied for the ten peptides of the APR panel (Supplementary Table 7).
Assessing platform reproducibility for peptide quantification
The reproducibility of an analytical system is key to its ability to characterize human biological variability. A mixture of the light and heavy peptides of the SARS-CoV-2 NCAP peptide AYNVTQAFGR (5 fmol/µL) in elution buffer was prepared and aliquoted across a 384-well plate (with 1 Marker Well). The peptide peak areas were recorded for the 383 sample wells with one ejection per well, then the plate was run 30 consecutive times acquiring 11490 sample ejections in total. This dataset comprises both the plate reproducibility and the long-term stability of the Echo MS system (Fig. 2A). The peak area reproducibility was found to be 4.6 and 6.4% CV for the light and heavy peptide, respectively, and the L/H peak area ratio for the peptide was maintained between 6.2–7.2% CV across the 11490 sample measurements (Fig. 2B). Data acquisition time for one plate was 10.5 min, so 30 plate runs required 5.25 h. This represents a sample acquisition rate of over 2000 samples per hour at very high reproducibility.
A The measurement of a peptide across 384 sample wells at the acquisition rate of 1.5 secs per well requires 10.5 min to acquire. B This measurement of the complete plate was repeated 30 times (acquiring 11490 sample ejections in total) and the peak area and peak area ratio percent coefficient of variation (%CV) were determined for each plate run. The %CV for the peak areas for the light and heavy peptides (Light AYNVTQAFGR (blue), heavy AYNVTQAFGR (orange)) were between 4.6 and 6.4% and the peak area ratios (summed L/H peak area ratio (green)) were between 6.2 and 7.2%.
Optimization of automated immunoenrichment sample preparation
The SISCAPA protocol has previously been optimized on various automation stations for LC-MS analysis23,25,26. Here, the protocol was adapted for the Biomek i7 automation station to make it compatible with AEMS analysis (Supplementary Fig. 1). To further improve the efficiency of bead washing and reduce the remaining salt concentrations, the buffer used during bead washing was switched from phosphate buffered saline (PBS) to ammonium bicarbonate (ABC), the plate type was changed from deep well U-bottom plates to deep well V-bottom plates to assist in the liquid removal, and new tips were used for each buffer removal step. Total time for the full sample preparation was ~6 h (including a 3-h digestion step at 37 °C), with final transfer of the 96 wells into a ready-to-analyze Echo qualified 384-well plate. The reproducibility of the total workflow, comprising the i7 sample preparation step, and the AEMS measurement was assessed by processing 16 wells of pooled healthy human plasma and using a pool and split strategy. The total workflow %CV for the light peptides from the 10 APR proteins ranged from 4.9–11.9% (Supplementary Table 5). The imprecision of upstream SISCAPA sample preparation ranged from 2.0–7.2%, encompassing both the imprecision arising from digestion variability and the liquid handling.
Acute phase response proteins enriched from plasma samples—data quality
After optimization of the automated sample preparation protocol for immunoenrichment of the 10-plex APR peptides, a 14-point standard addition curve was created for the 10-plex to establish the endogenous levels of the ten analytes within pooled human plasma. The endogenous levels of the analytes, observed by the plateau in Fig. 3A, encompass a range spanning over six orders of magnitude. Next, a cohort of 225 plasma samples from confirmed COVID-19 positive and negative subjects, as well as 23 healthy plasma samples and one standard healthy plasma pooled sample (19 technical replicates) were processed. The samples were randomized across three 96-well plates for sample preparation, with one plate being processed per day on the Biomek i7 workstation. The %CV for the L/H-ratio sum for the 10 APR proteins ranged from 6.68–40.87% (Supplementary Fig. 6).
A The dynamic range of the 10-plex APR was explored in pooled human plasma using a standard addition curve, spanning a range of almost > 1000000, highlighting their varying abundance levels and potential implications for physiological processes. B Reproducibility of endogenous peptide areas for the acute phase response peptide areas from plasma captures (n = 268). As expected, the reproducibility observed for the endogenous peptides from triplicate technical measurements correlates with the observed peptide area or abundance, with the higher abundance peptides showing very good reproducibility (<10%) and the less abundant, lower area peptides having more variance. Data was subjected to the outlier rejection strategy and rejected data points were not plotted. Inset shows the reproducibility of acute phase response SIL peptide areas from plasma captures, in a violin plot. Data are represented by the median, first and third quartiles, and range. The reproducibility of the SIL peptide for each enriched sample measured in triplicate was found to be very good, with average %CV values across the 267 measured samples between 4.2% and 10.5%. C Correlation of measured AEMS L/H peptide ratios with LC-MS data (n = 71) for the most biologically relevant acute phase response proteins, namely A1AG, C3, LBP, CRP and SAA. The ratios measured by LC-MS were very similar to the ratios determined using the Echo MS system. After outlier rejection, the slopes for all proteins were very close to 1 and the R² values were 0.96 and higher.
Datapoints for downstream biological measurements were deemed irreproducible and removed if the L/H peak area ratio between the two fragments monitored per peptide was greater than the average fragment ratio difference plus 2-sigma. A dilution series was additionally used to define the area observed at the LLOQ of each peptide, and values below this area threshold were also removed (Supplementary Fig. 7A). After application of this outlier rejection process, the reproducibility of the heavy peptide peak area (sum of both fragments monitored) across triplicate measurements (Fig. 3B inset) was between 4.2% and 10.5% CV.
In Fig. 3B, the peak areas for the endogenous light peptides were plotted vs the triplicate %CVs. As expected, a strong correlation between the peak area and the reproducibility was observed. Six targeted peptides were easily detected by AEMS across all conditions, SAA and CRP were easily measured in unhealthy samples, which contrasted with healthy samples in which they were mostly undetectable27,28. The peptides from MPO and MBL protein were near or below the LLOQ in all samples. Supplementary Fig. 7B shows the proportion of datapoints per protein that were rejected across the entire dataset.
The utility of applying the SISCAPA workflow for protein quantification has been demonstrated using LC-MS analysis, MALDI analysis, and rapid trap-elute strategies29,30,31. Therefore, to benchmark AEMS, a subset of the samples were also run by LC-MRM on a SCIEX QTRAP 6500+ system using microflow chromatography. For eight peptides with quantifiable ratios, of which 5 are shown in Fig. 3C, a near-perfect linear correlation between LC-MS and AEMS was observed with average slopes of 1.022 and average R2 values of 0.979. The linear correlation and regression between LC-MS and AEMS for the three other proteins (Alb, Hx and IgM) are shown in Supplementary Fig. 8.
The ejection volume and time between ejections was optimized based on the observed peptide signal and by the final plate, optimal ejection intervals were found to be 1.5 or 2 s, while the ejection volumes used were either 100 nL (Hx, Alb), 200 nL (C3, IgM, A1AG, LBP, MBL and MPO) or 300 nL (CRP and SAA). The time required to process all the samples in triplicate from the 96-well based sample preparation, using an ejection time of 1.5 s, was 1.6 h for a total of 4.8 h to analyze all three 96-well plates. Nevertheless, the robust reproducibility observed in the SIL peptide area (Fig. 3B, inset) serves as evidence that triplicate measurements are dispensable. This revelation presents a noteworthy prospect for a substantial reduction in analysis time, allowing for the completion of a single 96-well plate in just 32 min. In contrast, the run time per sample for the LC-MS assay was 5.5 mins to acquire all 10 peptides. For triplicate analysis, the total LC-MS run time for the cohort would be 73.5 h, meaning the Echo MS system workflow for analysing SISCAPA enriched samples is ~15x faster than microflow LC-MS, with identical peptide biomarker results and adequate signal.
Acute phase response proteins and COVID-19 disease severity
The cohort of plasma samples was collected from ill individuals admitted to either the medical floor or the ICU of the Cedars-Sinai Medical Center between March and May 2020. Samples were classified as positive if subjects had tested positive for SARS-CoV-2 within the first few days after admission, while the negative group consisted of symptomatic individuals whose RT-qPCR results were negative. Commercially purchased pooled healthy plasma and a cohort of 23 plasma samples collected from healthy individuals 5 years prior to the outbreak of the COVID-19 pandemic (2015) were both included. Samples were classified into nine distinct sub-groups by the Immunophenotyping Assessment in a COVID-19 Cohort (IMPACC) study and the WHO R&D Blueprint and COVID-19 classifications32,33 (Supplementary Table 8).
The L/H peptide ratios were plotted according to their disease classification (Fig. 4). As expected, CRP was substantially higher in unhealthy samples (COVID-19 positive and negative) relative to healthy individuals. Interestingly, there was minimal difference in CRP levels between the COVID-19 negative and COVID-19 positive samples. LBP and SAA were also found to have a smaller but statistically significant increase in the unhealthy samples. A1AG gradually increased from healthy to severely ill SARS-CoV-2 samples. Other proteins (C3, Hx, and IgM) did not show a significant change in this cohort, although the C3 protein in the SARS-CoV-2 infected, non-admitted class of samples seemed to show a spike which requires confirmation in future studies on larger populations. Finally, Alb showed small decreases in abundance among the unhealthy samples in the cohort, also consistent with its role as a negative acute phase reactant.
Samples were classified according to SARS-CoV-2 status (Positive/Negative) and disease severity (Supplementary table 8), then the light endogenous peptide ratio to the stable isotope labelled peptide was plotted (L/H ratio). Data are presented as following: minima, the smallest data point within 1.5 times the interquartile range (IQR) below the first quartile; maxima, the largest data point within 1.5 times the IQR above the third quartile; centre, the median of the dataset, representing the midpoint of the data; bounds of box, the lower and upper bounds of the box represent the first quartile and third quartile respectively, defining the IQR; whiskers, extending from the bounds of the box to the minimum and maximum values within 1.5 times the IQR from the first and third quartile respectively; and percentile, where Q1 represents the 25th percentile and Q3 represents the 75th percentile of the dataset. Four proteins were seen to increase in the disease samples (CRP, LBP, SAA, A1AG), and one protein (ALB) was found to decrease slightly in all disease samples. Figure 4, SARS-CoV-2 icon, created with http://BioRender.com, released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License.
Our results were consistent with Messner et al. (2020)4, who found that CRP, LBP and SAA1:SAA2 were increased in COVID-19 positive samples, with some separation between the COVID positive and severe disease samples. This group similarly reported Alb downregulated with COVID-19 disease. Other studies also found SAA1:SAA2 and CRP upregulated in COVID-19 disease3,34. Razavi et al. (2016) followed 22 proteins in 14 individuals over time from self-collected dried blood spots35, and observed large elevations in SAA, CRP, and LBP levels upon symptomatic undefined (presumably viral) infection, along with a small persistent drop in albumin levels consistent with an acute phase response. A similar pattern of expression was observed in our cohort, with elevated levels of SAA (p < 0.0001*), CRP (p < 0.0001*), and LBP (p < 0.0001*) alongside a decrease in ALB (p < 0.0001*) in all unhealthy individuals presenting ill at the hospital, regardless of COVID-19 status (Fig. 4). A similar correlation between CRP and SAA was observed, as well as between CRP and LBP and between SAA and LBP proteins. However, since the ratios among CRP, SAA, and LBP change dramatically during the time course of an infection (Supplementary Fig. 9), the significance of this finding remains to be confirmed35. As this is only a discovery cohort of 225 samples, other comparisons did not directly result in statistically significant differences. Moreover, it’s worth highlighting that with the increased speed of data acquisition facilitated by AEMS, such comparative analyses comprising thousands of samples can now be performed more readily, potentially enhancing the precision of our findings in future studies.
Quantification of SARS-CoV-2 NCAP peptide enriched from nasopharyngeal swabs
Recently, the capability of the SISCAPA assay in combination with LC-MS to detect the NCAP protein from up to 500 nasopharyngeal swabs per day up to the limit of infectiousness was shown, corresponding to an estimated RT-qPCR Ct-value of 32-338,9. Considering the overwhelming number of SARS-CoV-2 tests performed worldwide during the pandemic, the prospect of applying AEMS for measuring tens of thousands of samples a day on a single platform could greatly increase society’s pandemic preparedness. A small cohort of 145 nasopharyngeal swab samples was processed in two 96-well plates using the semi-automated sample preparation protocol described above, including an additional SPE step for further removal of background matrix (see Supplementary Fig. 10 for Ct distribution).
The sensitivity of the AEMS method was tested on the three SARS-CoV-2 NCAP peptides (AYNVTQAFGR, ADETQALPQR, and KQQTVTLLPAADLDDFSK) from a dilution series of recombinant SARS2_NCAP protein (0–150 fmol/µL) in the commonly used UTM medium. With the Echo MS system, an LLOQ in UTM of 0.195 fmol/µL was achieved for both the AYNVTQAFGR and ADETQALPQR peptides (Fig. 5A), while KQQTVTLLPAADLDDFSK demonstrated a higher LLOQ of 3.12 fmol/µL. The respective peak area values were used for outlier removal (Supplementary Fig. 11). In a previous study using LC-MS, excellent linearity down to 4 amol/µL was observed, corresponding to 144 mol on column9. This LC-MS method applies 10 µL on column, whereas the ejection volume with AEMS was 300 nL, resulting in a 33.3-fold higher loading in LC-MS. It is this sample loading difference that explains the bulk of difference in assay sensitivity between LC-MS and AEMS. Very good reproducibility of the triplicate measurements of the SIL peptides was observed (Supplementary Fig. 12), with average %CV values of 5% to 6.5%.
A Linearity of the average summed Light to Heavy (L/H) ratio with the spiked concentration of NCAP protein in Universal Transport Medium (UTM). B Overlay of the XIC of the three target peptides across samples with a prior dilution series in UTM and ending with a dilution series in ammonium bicarbonate (ABC). Each peak represents one patient sample. C Secondary axis plots of the raw measurements of E-gene Cycle threshold (Ct) (red dots) and AYNVTQAFGR logarithmically transformed MS Peak Area (Log2Area) (green bars) for results sorted from low to high Log2Area. A strong linear correlation illustrates the level of agreement between RT-qPCR and AEMS, with Log2Area flattening at 7.5 (green line), i.e. beyond Ct > 26 (red line). A high percent positive (PPA = TP/(TP + FN)) and negative agreement (PNA = TN/(TN + FP)) between RT-qPCR (Ct) and MS (AYN Log2Area) is achieved, especially below Ct 26.
A selection of samples was run by LC-MS, and again very good correlation was observed between the peptide ratios measured by AEMS and LC-MS, (Supplementary Fig. 13) with average slopes of 0.983 and an R2 of 0.943. Notably, while AEMS is less sensitive compared to LC-MS and therefore starts losing accuracy at the lowest peptide concentration, the variation in the correlation plot for the KQQTVTLLPAADLDDFSK peptide derives from the LC-MS runs, which suffer from column carry-over for this peptide, as described earlier (9). Interestingly, as there is no chromatography with AEMS, this peptide behaved very well in the Echo MS system. Next, a total of 142 nasopharyngeal swab samples were screened using the AEMS assay for SARS-CoV-2 NCAP peptides. An overlay of the Extracted Ion Chromatograms (XIC) for each peptide run (with dilution series before and after the sample batch) show the intuitive nature of the data (Fig. 5B), with each visible peak indicating an infection measured in 1.5 s.
A binary comparison between RT-qPCR for the E-gene (red dots, right axis) and Log2Area of the AYNVTQAFGR peptide (green bars, left axis) is shown in Fig. 5C. The cohort comprised 113 qPCR positive and 29 RT-qPCR negative samples, with a Ct value >40 being considered negative. In accordance to the Clinical and Laboratory Standards Institute (CLSI) user protocol for evaluation of qualitative test performance (EP 12-A2), percent positive and negative agreement (PPA and PNA) were calculated. The PNA/PPA matrix in Fig. 5C depicts these numbers when a summed intensity of AYNVTQAFGR of 7.5 is used. However, there are several concerns with this representation as previously described by Van Puyvelde et al., therefore results were sorted from low to high virus measurement, i.e., from low to high Log2Area, which inversely correlates to the Ct value9. A clear correlation between both tests was found, as described earlier for LC-MS measurement. When a Log2Area of 7.5 for AYNVTQAFGR was applied as a cutoff (see dilution series in Fig. 5A), eight samples with Ct < 26 are wrongly classified as negative (PPA = 91.30%). Whether these outliers have an analytical cause (in the RT-qPCR or the AEMS assay) or whether this is a case of residual RNA36, cannot be addressed at this point. Inversely, at this threshold, 3 samples with Ct 40 were classified as positive by AEMS but as negative based on the RT-qPCR Ct value (PNA = 93.88%). Overall, within the current small sample cohort, Ct 26 seems a good first estimate for the NCAP detection limit by AEMS in nasopharyngeal swabs. This aligns well to the published LC-MS method, where a threshold of Ct 30-32 in UTM could be attained, but now with 33-fold lower sample loading on AEMS (25 = 32 and thus 5 Ct values lower). However, AEMS provides a potential rate of 2400 samples per hour, as opposed to 30 per hour using LC-MS9.
- SEO Powered Content & PR Distribution. Get Amplified Today.
- PlatoData.Network Vertical Generative Ai. Empower Yourself. Access Here.
- PlatoAiStream. Web3 Intelligence. Knowledge Amplified. Access Here.
- PlatoESG. Carbon, CleanTech, Energy, Environment, Solar, Waste Management. Access Here.
- PlatoHealth. Biotech and Clinical Trials Intelligence. Access Here.
- Source: https://www.nature.com/articles/s41467-024-48563-z