Study materials

Recombinant HbA2 (α2δ2) in natural and U-15N-labeled form was kindly provided by the European Commission, Joint Research Centre, Geel, Belgium. The materials were produced by Trenzyme GmbH as previously described [17] using P69905 and P02042 (UniProtKB) as templates for the co-expression in E. coli of the α and δ subunits, respectively. Both materials were obtained as solutions of about 0.42 mg/g (natural form) and 0.47 mg/g (labeled form) in 50 mmol/L 2-Amino-2-(hydroxymethyl)propane-1,3-diol (Tris), pH 7.5, and 100 mmol/L NaCl. The sample volumes were 28 mL (HbA2) and 26 mL (U-15N-HbA2). The recombinant HbA2 (α2δ2) of natural isotopic composition was ultra-purified by semi-preparative strong anion exchange chromatography using a MONO-Q 4.6/100 PE column, yielding 8 mL of ultrapurified HbA2 at a concentration of 0.43 mg/g. This was then used as the third study material. All study materials were stored in a fridge at 4–8 °C until used.

Calibrators HbA2 used for preparation of the calibrators

The material was obtained from SIGMA-Aldrich, cat. No.: H0266; lot: SLBK8749V, as a neat substance. The “protein-purity” was 99.0% using the LFQ method described herein. For value-assignment, a stock solution was prepared from this solid material by dissolving ∼2 mg in 1 g of Tris (10 mmol/L, pH 7.8). The mass fraction of HbA2 in the solid material was 482.7 mg/g as determined by AAA. A stock solution of the HbA2 material in Tris (10 mmol/L, pH 7.8) was prepared. An aliquot of this was used as a constant component present in each of the calibrators (red in Fig. 1A).

Fig. 1

LFQ-based measurement of HCPs (E. coli) mass fraction in recombinant HbA2. A Calibrators were obtained by spiking aliquots of HbA2 stock solution (red) with increasing amounts of E. coli lysate (green). Amounts per mass (mg) of both components (HbA2 and E. coli) and mass fractions of E. coli were hence known for each calibrator through amino acid analysis. B Quantitative information was acquired by shotgun proteomics. C MS1 intensities were integrated over all features associated with peptides identified from either E. coli or HbA2. D The sum of all MS1 peak intensities from peptides associated with E. coli per sum of all peak intensities associated with HbA2 peptides for the calibrators were plotted vs. mass fractions. The fitting linear function was then used to calculate the E. coli fraction in the investigated materials from sample measurements (red line and arrows)

E. coli proteome sample

Lyophilized E. coli protein material was obtained from BIO-RAD (ReadyPrep, Catalog 163 2110, L9703999, Control 310,004,134). For a stock solution, the material was reconstituted in water (30% acetonitrile, 0.1% formic acid); the mass fraction of E.coli proteins as determined by AAA in that stock solution was 0.345 ± 0.012 mg/g. A series of calibrator samples was prepared by mixing an aliquot of HbA2 stock solution with the appropriate amount of E. coli stock solution (red and green, respectively, in Fig. 1A). The mass fractions of HbA2 in these sample solutions were 0.390, 0.383, 0.372, 0.361, and 0.350 mg/g (solution). The corresponding mass fractions of E. coli protein, relative to HbA2 in the calibrator samples, was 10.3, 25.0, 52.7, 80.1, and 111.6 mg (E. coli)/g(HbA2).

Yeast proteome sample

Yeast-based calibrators were prepared in the same way as described for E. coli. A whole-cell protein extract of Saccharomyces cerevisiae (Promega, V7341, lot 434,786) was deployed. The material came as a solution in 50 mmol/L Tris and 6.5 mol/L urea. Solutions containing mass fractions of HbA2 (0.390, 0.387, 0.383, 0.378, and 0.372 mg/g) were prepared. Mass fractions of yeast proteins in these solutions were 12.7, 26.5, 52.4, 76.6, and 113.6 mg/g (relative to HbA2).

Human K562 proteome sample

A whole-cell protein extract from human K562 cells (Promega, V6941, lot 444,583) was dissolved in 50 mmol/L Tris and 6.5 mol/L urea, as above. Solutions containing mass fractions of HbA2 (0.390, 0.387, 0.383, 0.378, and 0.372 mg/g) were prepared, comprising also K562-proteins at 15.0, 32.7, 63.9, 94.3, and 136.2 mg/g as mass fractions relative to HbA2.

Protein mix

Human C-reactive protein (CRM GBW09228, National Institute of Metrology, China), human insulin analog (insulin aspart, NovoLog) and human β2-microglobulin (kindly provided by the European Commission, Joint Research Centre, Geel, Belgium) were obtained as solutions. Bovine serum albumin (Sigma-Aldrich, cat. No. 05470, lot No. 1099572), myoglobin from horse skeletal muscle (Sigma-Aldrich, cat. No. 70025, lot No. 381848/1), cytochrome-c from bovine heart (Sigma-Aldrich, cat. No. C3131, lot SLBZ0555), somatotropin (NIBSC, WHO International Standard 98/574), human ceruloplasmin (Athens Research & Technology, cat. No. 16–16-030518), and serotransferrin (Sigma-Aldrich, cat. No. T3309, lot BCBR1763V) were obtained as solids and had to be dissolved to known concentrations in water prior to use. The mass fractions of somatotropin, ceruloplasmin, serotransferrin, β2-microglobulin, and insulin were determined by mass spectrometry based AAA, while certified values were used, as provided by the supplier, for C-reactive protein, albumin, cytochrome-c, and myoglobin. Aliquots of these solutions were mixed to yield a stock solution containing somatotropin, ceruloplasmin, serotransferrin, β2-microglobulin, insulin, C-reactive protein, albumin, cytochrome-c, and myoglobin in the mass-ratio of 0.1055:0.1176:0.124:0.1251:0.1247:0.0317:0.1215:0.1260:0.1233. Aliquots of this mixed solution were spiked with aliquots of the HbA2 stock solution, resulting in HbA2 mass fractions of 0.393, 0.388, 0.383, 0.378, 0.372, and 0.367 mg/g, and protein mass fractions, relative to HbA2, of 22.7, 47.4, 69.9, 92.8, 117.1, and 140.8 mg/g. To additionally cover the low HCPs fraction range as needed for the ultra-purified HbA2 material, a second series of calibrator samples was prepared. These calibrators were of 0.427 mg/g HbA2 mass fraction and 0.2, 0.3, 0.4, 0.7, 1.1, and 1.4 mg/g protein mass fractions relative to HbA2.

Determination of protein mass fractions in the calibrators

For the stock solutions used to prepare the calibrators, the mass fractions of amino acids were determined by mass spectrometry based AAA, as detailed in Arsene et al. [17]. These mass fractions were then combined with the known mass fractions (or relative amounts) of these amino acids in the protein or proteome to yield the protein mass fraction in that stock solution. In the cases of E. coli, yeast and K562, relative amounts (by mass) of amino acids were used, as was previously published [18,19,20]. The contribution to the overall estimated measurement uncertainty from AAA (in our laboratory) and uncertainties published with literature data were combined to yield expanded (95%) uncertainties of 3.5%, 2.7%, 3.0%, and 3.2%, respectively, for the E. coli, yeast, K562, and protein-mix calibrator stock solutions.

Proteolysis

To a 30 µL aliquot of sample (recombinant HbA2) or calibrator, 70 µL of Tris solution (35 mg Tris base, 46 mg Tris HCl, dissolved in 1 mL water) were added. Proteolysis (37 °C) was started by the addition of 10 µL of trypsin solution (1 mg/mL in 50 mM acetic acid). Trypsin from porcine pancreas was obtained from Sigma-Aldrich, St Louis, USA; cat. No.: T0303. After 10, 70, 130, 190, and 250 min, further 10 µL aliquots of trypsin solution were added. In parallel, 40 µL aliquots of acetonitrile were added after 10, 30, 60, 90, 120, and 150 min, respectively. The sample or calibrator was further incubated at 37 °C overnight. For reduction, 0.8 mg of dithiothreitol (DTT) was added. After incubation (37 °C) for 1 h, 3 mg of 2-iodoacetamide were added for alkylation (30 min at room temperature). The excess of 2-iodoacetamide was quenched with 3 mg of DTT. The reaction was stopped by the addition of 10 µL of formic acid (10 vol.-%). The sample or calibrator was desalted using solid-phase extraction (SPE) C18 ec cartridges (Chromabond, 100 mg, Macherey Nagel, Düren, Germany). After lyophilization, residues were redissolved in 40 µL of water (0.1% formic acid) and subjected to nLC-MS/MS analysis.

Liquid chromatography-mass spectrometry

An UltiMate 3000 RSLCnano HPLC system (Thermo Fisher Scientific) coupled to a timsTOF Pro mass spectrometer (Bruker Daltonics) was used for the analysis of the proteolysed samples and calibrators. Peptides were trapped on a pre-column (Acclaim PepMap C18, 5 µm, 0.3 × 5 mm) and then separated on a Bruker Fifteen nanoFlow column (15 cm × 75 µm, C18, 1.9 µm, 120 Å) using a linear water-acetonitrile gradient from 1 to 60% B in 210 min and then from 60 to 80% B in 20 min (with solvent A: water, 0.1 vol.-% formic acid and B: acetonitrile, 0.1 vol.-% formic acid) at 40 °C. The flow rate was 300 nl/min. The timsTOF Pro mass spectrometer was equipped with a CaptiveSpray ion source. The mass spectrometer was run using the DDA-PASEF-standard-1.1 s-cycle time method, as provided by Bruker. Briefly, the settings were 10 PASEF MS/MS scans per acquisition cycle with a trapped ion mobility accumulation and elution time of 100 ms. Spectra were acquired in a m/z range of 100 to 1700 and in an (inverse) ion mobility range (1/K0) of 0.60 to 1.60 Vs/cm2. The collision energy was set up as a linear function of ion mobility starting from 20 eV for 1/K0 of 0.6 to 59 eV for 1/K0 of 1.6.

Protein database search

PEAKS Studio Xpro (Bioinformatics Solutions Inc.) was used for feature detection/database searching and precursor ion (MS1) quantification. Databases for E. coli, Saccharomyces cerevisiae, human proteome, and the mixture of nine proteins were obtained as FASTA files (uniprot.org, accessed: 27. Aug. 2021). FASTA files of human hemoglobin subunit alpha and delta (Uniprot: P69905 and P02042) were added to the databases of non-human proteomes. The following settings were applied for data analysis: carbamidomethylation of cysteine as fixed modification, methionine oxidation, and glutamine or asparagine deamidation as variable modifications. A maximum of two modifications per peptide were allowed. With the human proteome, glycosylation was set as an additional variable modification using the built-in glycosylation list. Trypsin/P was set as the enzyme, and no more than two missed cleavages per peptide were allowed. The mass tolerance for the monoisotopic mass of precursor ions and fragment ions was 15 ppm and 0.05 Da, respectively. For the retention time and ion-mobility of an identified peptide, the shift tolerance between different runs was 3 min and 5%, respectively. Mass correction was enabled for precursor ions. The false discovery rate (FDR) was 1% at the peptide and protein level. The minimum length of identified peptides was seven amino acids. Results of quantification were obtained as peak areas at the protein-level.

Mass fraction of impurity in the labeled HbA2 material

The mass fraction of impurity in the labeled HbA2 material was 78.1 ± 8.6 mg/g. The individual results from n = 6 repetitions of label-free quantification were 70.8, 72.4, 84.2, 92.8, 73.2, and 75.0 mg/g.

Fractions of co-purifying proteins

The fraction of E. coli proteins known to frequently be co-purified [13] was calculated as the ratio of the (MS1) intensity of these proteins to the intensity of all E. coli proteins identified in the HbA2 material or in the E. coli proteome sample. For Fig. 3, fractions were calculated for the two HbA2 materials and compared to the fraction for an E. coli proteome sample containing a similar amount of E. coli proteins (80.1 mg/g).

Downstream data analysis

Further analysis of the data exported from PEAKS was based on Python 3.8 with the modules pandas, numpy, numpy.linalg, and Matplotlib imported as needed. For datafit and cross-validation, Scikit-learn [21] 1.0.2 was used.

Data availability

The mass spectrometry data and the tables of identified and quantified proteins have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD041736.