Absolute Quantification of Aβ_1-42 in CSF Using a Mass Spectrometric Reference Measurement Procedure

Alzheimer's disease (AD) is the most common form of dementia and affects about 35 million people worldwide¹. The neuropathological hallmarks of the disease widely believed to lie at the core of AD pathogenesis are intracellular neurofibrillary tangles of hyperphosphorylated tau protein² and extracellular plaques consisting of aggregated amyloid-beta (Aβ) peptides³. In line with this, the assessment of plaque pathology in vivo by biomarkers has recently been included in the research diagnostic criteria for AD⁴. For CSF measurements of Aβ_1-42, several immunoassays are available and are used in many clinical laboratories⁵. The concentration of Aβ_1-42 in CSF is approximately 50% lower in AD patients than in the cognitively normal elderly, reflecting the deposition of the peptide in plaques in the brain^6,7.

These biomarkers are mainly analyzed using immunoassays (i.e., antibody-based techniques), but these assays may be influenced by matrix effects⁸. The use of immunoassays on different technology platforms and the lack of assay standardization^9,10 makes the introduction of global cut-off concentrations difficult^11,12. An analytically validated RMP would permit the uniform calibration of different assay platforms, ideally resulting in better comparability across analytical platforms and in better control of factors contributing to the overall measurement variability.

The absolute quantification of Aβ_1-42 using the developed LC-MS/MS method overcomes many of the issues associated with antibody-based techniques. The method, listed as an RMP by the Joint Committee for Traceability in Laboratory Medicine (JCTLM database identification number C11RMP9), will be used to determine the absolute concentration of Aβ_1-42 in a Certified Reference Material (CRM) to harmonize CSF Aβ_1-42 measurements across techniques and analytical platforms. The described workflow should be of relevance for the development of candidate reference methods for peptides and proteins within other areas of medicine.

NOTE: This protocol requires aliquots of at least 50 µL, with a concentration of 50 µg/mL for each Aβ peptide, as starting material. The Aβ peptides should be dissolved in 20% acetonitrile (ACN) and 4% concentrated ammonia solution in deionized water (v/v) and stored at 80 °C.

Caution: See Table 1 for safety information.

1. Preparation of the Solutions

1. Prepare 100 mL of 20% ACN and 4% concentrated ammonia solution in deionized water (v/v) by diluting 20 mL of ACN and 4 mL of concentrated ammonia (~25%) in deionized water. Adjust the final volume to 100 mL with deionized water. Make fresh daily.
2. Prepare 50 mL of 5 M guanidine-hydrochloride by dissolving 26.08 g of guanidine-hydrochloride in deionized water to a final volume of 50 mL. Store at 20 °C and make fresh monthly.
3. Prepare 200 mL of 4% phosphoric acid in deionized water (v/v) by diluting 9.4 mL of concentrated phosphoric acid (~85%) in deionized water. Adjust the final volume to 200 mL with deionized water. Store in the refrigerator and make fresh weekly.
4. Prepare 50 mL of 75% ACN and 10% concentrated ammonia solution (v/v) in deionized water by diluting 37.5 mL of ACN and 5 mL of concentrated ammonia (~25%) in deionized water. Adjust the final volume to 50 mL with deionized water. Make fresh daily.
5. Thaw at least 2.5 mL of human CSF for the calibrators, obtained from de-identified leftover samples from routine clinical analysis.
6. Prepare artificial CSF containing 150 mM Na, 3.0 mM K, 1.4 mM Ca, 0.8 mM Mg, 1.0 mM P, and 155 mM Cl in deionized water and add bovine serum albumin to a final concentration of 4 mg/mL. Only 1 mL is needed per analysis, but prepare a large volume, aliquot it, and store for future use.

2. Preparation of the Calibrators

1. Prepare 0.5 mL of 4 µg/mL ¹⁵N-Aβ_1-42 peptide by adding 40 µL of 50 µg/mL ¹⁵NAβ₁₄₂ to 0.46 mL of 20% ACN and 4% concentrated ammonia in a 0.5-mL microcentrifuge tube. Mix on a vortex mixer for 1 min.
2. Prepare 2 mL of 100 ng/mL ¹⁵N-Aβ_1-42 peptide by adding 50 µL of the 4 µg/mL ¹⁵NAβ₁₄₂ to 1.95 mL of 20% ACN and 4% concentrated ammonia in a 2-mL microcentrifuge tube. Mix on a vortex mixer for 1 min.
3. Prepare six calibrator solutions (A-F) by mixing the volumes of each solution indicated in Table 2. Use 0.5, 1.5, and 2 mL microcentrifuge tubes. Mix on a vortex mixer for 1 min.
4. Prepare the final calibrators (in duplicate) in 0.5 mL microcentrifuge tubes by adding the corresponding calibration solutions and human CSF according to Table 3. Mix on a vortex mixer for 1 min.

3. Preparation of the Internal Standard

1. Prepare 2 mL of 0.8 µg/mL ¹³C-Aβ_1-42 peptide by adding 32 µL of 50 µg/mL ¹³CAβ₁₄₂ to 1.968 mL of 20% ACN and 4% concentrated ammonia in a 2-mL microcentrifuge tube. Mix on a vortex mixer for 1 min.
2. Prepare 5 mL of 16 ng/mL ¹³C-Aβ_1-42 peptide by adding 0.1 mL of 0.8 µg/mL to 4.9 mL of 20% ACN and 4% concentrated ammonia in a 5 mL microcentrifuge tube. Mix on a vortex mixer for 1 min.

4. Preparation of the Response Factor Sample

NOTE: The response factor (RF) determination is performed to determine the concentration of the labeled peptide used for calibration (¹⁵N-Aβ_1-42). This requires that the concentration of the native Aβ_1-42 peptide has been determined using amino acid analysis (AAA). Thus, the volume and concentration of native Aβ_1-42 peptide aliquots need to fulfill the requirements of the AAA.

1. Prepare 0.5 mL of 4 µg/mL native (unlabeled)^Aβ_1-42 by adding 40 µL of 50 µg/mL native Aβ_1-42 to 0.46 mL of 20% ACN and 4% concentrated ammonia in a 0.5-mL microcentrifuge tube. Mix on a vortex mixer for 1 min.
2. Prepare a 2-mL 40 ng/mL mix of native and ¹⁵N-Aβ_1-42 by adding 20 µL of 4 µg/mL native Aβ_1-42 and 20 µL of 4µg/mL ¹⁵N-Aβ_1-42 to 1.96 mL of 20% ACN and 4% concentrated ammonia in a 2 mL microcentrifuge tube. Mix on a vortex mixer for 1 min.
3. Add 20 µL of the 40 ng/mL mix to 0.38 mL of artificial CSF in a 0.5 mL microcentrifuge tube. Prepare duplicates and mix on a vortex mixer for 1 min.

5. Sample Preparation

NOTE: Thaw the samples to be measured at room temperature on a roller.

1. Add 0.18 mL of each calibrator, response factor, and unknown sample (including quality control [QC] samples, if used) to a 1 mL protein 96-deep-well plate, according to Figure 1 (assuming a full plate is used). Make sure to add the samples in, or close to, the bottom of the wells.
2. Add 20 µL of internal standard to each well (i.e., calibrators, response factors, QCs, and unknowns); it is crucial to release the drop on the side of the well close to the surface of the sample without submerging the pipette tip.
3. Add 0.2 mL of 5 M guanidine-hydrochloride to each well.
4. Place the sample plate on a microplate shaker and mix the samples for 45 min at 1,100 rpm. The optimal frequency might differ depending on instrumentation. Set the frequency and amplitude of the mixer so that the solutions are thoroughly mixed and no drops of internal standard or CSF are left unmixed on the side of the wells.
5. Add 0.2 mL of 4% phosphoric acid to each well. Vortex mix briefly.

6. Solid Phase Extraction

NOTE: In all washing, loading, and elution steps, apply the lowest possible vacuum after adding the solution and gradually increase as needed to load or elute the solution. Disable the vacuum between each loading and elution step.

1. Put a reservoir tray for waste under a mixed-mode, cation-exchange, 96-well solid phase extraction (SPE) plate in the extraction plate manifold chamber.
2. Condition the SPE sorbent by adding 0.2 mL of methanol to each well.
3. Equilibrate the sorbent by adding 0.2 mL of 4% phosphoric acid to each well.
4. Transfer all samples (about 0.62 mL in each well) from the deep-well plate to the SPE plate. Use an eight-channel pipette when transferring the samples from the deep-well plate to the SPE plate; it is not crucial to transfer the entire or equal volumes of all samples, since the samples contain an internal standard that will compensate for variations.
5. Wash the sorbent after the samples have passed through by adding 0.2 mL of 4% phosphoric acid to each well.
6. After the washing solvent has eluted from the sorbent, replace the reservoir tray with a collection plate or tubes.
7. Elute the sample from the sorbent by twice adding 50 µL of 75% ACN/10% concentrated ammonia, and note that this solution requires a very low vacuum to pass through the sorbent. Remember to disable the vacuum between each addition.
   1. OPTIONAL: Seal the collection plate or tubes and freeze them at -80 °C. Remove the seal from the collection plate or tubes before proceeding to step 6.8.2.
   2. Dry the eluates by using vacuum centrifugation (without applying heat); this can take from one to several hours, depending on the vacuum centrifuge.
   3. Seal the containers and freeze them at -80 °C.

7. Liquid Chromatography

1. Prepare mobile phase A (5% ACN and 0.3% concentrated ammonia in deionized water [v/v]), B (4% deionized water and 0.1% concentrated ammonia in ACN [v/v]), and needle wash (50% ACN and 4% concentrated ammonia in deionized water [v/v]).
   1. For 500 mL of mobile phase A, add 25 mL of ACN and 1.5 mL of concentrated ammonia to deionized water. Adjust the final volume to 500 mL with deionized water.
   2. For 500 mL of mobile phase B, add 500 µL of concentrated ammonia and 25 mL of deionized water to ACN. Adjust the final volume to 500 mL with ACN.
   3. Prepare 250 mL of needle wash by adding 120 mL of ACN and 10 mL of concentrated ammonia to deionized water. Adjust the final volume to 250 mL with deionized water.
   4. Put the mobile phases A and B and needle wash bottles open in a sonication bath for 20 min before use with the LC system
2. Dissolve each sample with 25 µL of 20% ACN and 4% concentrated ammonia solution and place them on a shaker for 20 min. Centrifuge down the samples and place them in the autosampler (keep at 7 °C).
3. Inject 20 µL of sample on a 1 x 250 mm² polystyrene-divinylbenzene (reversed-phase) monolithic column maintained at 50 °C.
   1. Use the LC gradient shown in Table 4 with a flow rate of 0.3 mL/min. Divert the first two and the last five min to waste (post-column) using a divert valve to reduce the contamination of the mass spectrometer.

8. Mass Spectrometric Analysis

NOTE: These parameters were used for a quadrupole-orbitrap hybrid mass spectrometer equipped with aheated electrospray ionization source.

1. Set the parameters for the ion source according to Table 5.
2. Set the MS instrument to isolate the 4+ charge states of native Aβ_1-42 (1,129.48 mass-to-charge ratio [m/z]), ¹⁵N-Aβ_1-42 (1,143.00 m/z), and ¹³C-Aβ_1-42 (1,179.50 m/z) in the quadrupole mass analyzer with an isolation width of 2.5 m/z.
3. Fragment the isolated peptides in the collision cell with a normalized collision energy (NCE) of 17.0. This might need to be tuned for each instrument, even of the same type (and especially if using other types of instruments, such as a triple quadrupole MS).
4. Record the fragment spectra with a resolution of 17.500, with an automatic gain control target of 2 x 10⁵ charges and a maximum injection time of 250 ms.

9. Data Processing

1. Use the sum of the product ions (with a mass tolerance of ± 250 milli mass units [mmu]) in Table 6 to calculate the chromatographic areas for each peptide. Note that the ion types and numbers are only shown for native Aβ_1-42 product ions, since they are the same for both ¹⁵N-Aβ_1-42 and ¹³C-Aβ_1-42.
2. Determine the average response factor of the two response factor samples by dividing the area under the curve (chromatographic peak) of ¹⁵N-Aβ_1-42 with the area under the curve of native Aβ_1-42.
3. Adjust the concentration of the ¹⁵N-Aβ_1-42 used for calibration by multiplying it with the response factor calculated in step 9.2.
4. Construct a calibration curve by plotting the area ratios of ¹⁵N-Aβ_1-42 to ¹³C-Aβ_1-42 from the two sets of calibrators against the concentration.
5. Calculate the slope and intercept of the calibration curve using linear regression.
6. Calculate the area ratio of native Aβ_1-42 to the internal standard (¹³C-Aβ_1-42) for unknown samples.
7. Extrapolate the concentration of unknown samples from the calibration curve using the slope and intercept obtained in step 9.5.

The plate setup in Figure 1 is used for a full plate of samples. If fewer unknown samples are to be analyzed, the second calibrator, RF, and QC sets should be placed after the first half of the unknown samples.

As seen in Figure 2, the calibrators are close to the regression line, with low standard deviations. This method has a lower level of quantification of 150 pg/mL and an upper level of quantification of 4,000 pg/mL. The residual standard deviation of the calibration should of course be as low as possible. If calibration is non-linear, the run should be discarded (depending on severity), and the deviation is most likely due to incorrect pipetting technique and/or errors in the dilution of calibrators. The coefficient of variation (CV) of replicates should be below 20%, but preferably below 10%.

The native ¹⁵N- and ¹³C-Aβ_1-42 elute from the LC column simultaneously (since they only differ on the isotopic level), with close to symmetrical peaks and without significant tailing (Figure 3). At least ten measurements should be performed for each chromatographic peak and can be adjusted with the maximum injection time in the instrument method. All three peptides can be measured simultaneously for all measurements (calibration, RF, and unknowns) during the MS analysis. However, if the sensitivity of the method is suboptimal, only peptides of interest should be measured for each injection (i.e., only measure ¹⁵N- and ¹³C-Aβ_1-42 for calibrators, native and ¹⁵N-Aβ_1-42 for RF samples, and native and ¹³C-Aβ_1-42 for unknown samples).

Figure 1: Layouts of SPE and deep-well plates. Typical layout of calibrators (A-F), response factor sample (RF), quality control samples (QC), and unknowns. Please click here to view a larger version of this figure.

Figure 2: Calibration curve. Calibration curve constructed using ¹⁵N-Aβ_1-42 at 172, 572, 1,144, 2,287, 3,431, and 4,574 pg/mL (adjusted using the response factor) and ¹³C-Aβ_1-42 as the internal standard in human CSF (n = 2). The area ratio of ¹⁵N-Aβ_1-42/¹³C-Aβ_1-42 is plotted (Y-axis) against the concentration (X-axis). Please click here to view a larger version of this figure.

Figure 3: Chromatogram. Chromatogram of 0.500 ng/mL native (endogenous) Aβ_1-42 (top panel) and 1.6 ng/mL ¹³C-Aβ_1-42 (bottom panel) in human CSF. Please click here to view a larger version of this figure.

Figure 4: Quantification of unknown Aβ_1-42 in unknown samples. The peak area ratio is calculated by dividing the native Aβ_1-42 chromatographic peak area with the internal standard (¹³C-Aβ_1-42) chromatographic peak area. The concentration of native Aβ_1-42 in the sample is extrapolated from the calibration curve. Please click here to view a larger version of this figure.

Table 1: Safety information. Safety information for chemicals used for this protocol.

Table 2: Calibrator solutions. Calibrator solutions prepared in 20% ACN and 4% concentrated ammonia used for spiking CSF calibrators.

Table 3: Calibrators. Calibrators prepared in human CSF.

Table 4: LC gradient. The LC gradient used with a constant flow rate of 300 µL/min.

Table 5: Ion source settings. Parameters for the ion source to be set in the instrument tune software.

Table 6: Ions used for quantification. The 4+ charge states of the precursor ions are isolated in the quadrupole mass analyzer, with an isolation width of 2.5 m/z. The product ions (with a mass tolerance of ± 250 mmu) are used to calculate the chromatographic areas for each peptide. Ion types and numbers are only shown for native Aβ_1-42 product ions, since they are the same for both ¹⁵N-Aβ_1-42 and ¹³C-Aβ_1-42.

For the described method, instead of using a surrogate matrix, we used the surrogate analyte approach^13,14,15,16, which enables calibration in human CSF. The surrogate analyte approach involves two different isotopically labeled standards. One (¹⁵NAβ₁₄₂) is used to generate the calibration curve in human CSF, while another (¹³CAβ₁₄₂) is used as the internal standard. Unknown endogenous Aβ₁₄₂ concentrations are then extrapolated from the calibration curve constructed using the ¹⁵NAβ₁₄₂/¹³CAβ₁₄₂ ratio by the calculated endogenous Aβ₁₄₂/¹³CAβ₁₄₂ ratio. The surrogate analyte approach was used since there is no analyte-free CSF available, and low Aβ₁₄₂ recovery was observed when using native Aβ₁₄₂ in a surrogate matrix during method development.

Since ¹⁵NAβ₁₄₂ and native Aβ₁₄₂ may give different responses in the MS, the concentration of ¹⁵NAβ₁₄₂ is adjusted by measuring an RF sample-an artificial CSF sample containing equal concentrations of ¹⁵NAβ₁₄₂ and native Aβ₁₄₂, with a known concentration determined by AAA.The response factor might differ between different mass spectrometers due to possible variations in the isotopic purity of the ¹⁵N-labeled peptide between batches. Thus, the response factor should be determined for each measurement day.

The most critical steps in this protocol are the preparation of the calibrators and the RF samples. Aβ peptides, especially Aβ_1-42, are very hydrophobic and easily stick to pipette tips and tubes surfaces^8,17,18. To minimize the loss of Aβ peptides during pipetting, it is extremely important to saturate the pipette tips prior to delivery. Preferably, three volumes of peptide solution should be discarded prior to delivery to a new tube containing solution. Depending on the volume and concentration of the stock solution, this is not always possible. The second-best approach is of course to pipette the peptide solution up and down three times prior to delivery. For the same reason, it is important to use appropriate sizes for tubes, avoiding large void volumes.

Previously published data for the method show that recovery was within 100% (15%)¹⁵. The relative errors for the back-calculated calibrators were below 15% of the whole range defined by the calibrator curve¹⁹.

One obvious limitation of this technique is that it is low-throughput compared to automated immunoassays. However, the purpose of the described method is high accuracy and not throughput. This method can also be expanded to include Aβ_1-38 and Aβ_1-40, which are shorter¹⁹. Another limitation of this method is that the operator will need extensive mass spectrometry training before running the analysis on the instrument.

Quantification using immunoassays is dependent upon the interaction between the antibody and the antigen. This interaction could be affected by the presence of sample components that may interfere or compete with the interaction. In addition, the interaction may also be affected by the conformation of the antigen. These effects are difficult to control and are believed to be the main reason why it has been difficult to harmonize results between immunoassay platforms and between laboratories. Because quantification with MS is based on directly counting the target molecules relative to a stable, isotope-labeled standard, quantification is absolute and generally unaffected by such matrix effects. In addition, diagnostic protein measurements by immunoassays should be supported by an unbroken chain of higher-order measurement procedures and materials, from validated LC-MS/MS and stable, isotope-labeled internal standards to RMPs and a CRM, thus improving the comparability and reliability of results^20,21.

In conclusion, the described RMP for the measurement of Aβ₁₄₂ in CSF is an important step in developing a CRM that will help to establish general cut-off concentration for Aβ₁₄₂ in CSF. Exact cut-offs are highly important for accurately diagnosing AD early and are of the utmost importance when new types of disease-modifying drugs reach the clinic.