Purpose: Multi-attribute method (MAM) is an LC-MS based technique that uses peptide mapping to identify and quantitate product quality attributes (PQAs) in protein therapeutics. MAM has been proposed as a replacement for traditional quality release tests used for protein therapeutics that is increasingly being tested for implementation in Quality by Design (QbD) and quality control (QC) settings within the biopharmaceutical industry. Here, the ability of MAM to demonstrate lot-to-lot variations was tested by relative quantitation of PQAs in five different rituximab drug product lots.
Methods: Tryptic digestion of five rituximab drug product lots was performed using 50 µg rituximab at a 1:10 enzyme:substrate ratio. Separation of digested samples was performed on a Thermo Accela LC system at 50 °C using an Agilent Zorbax C18 300-SB column (300 Å, 2.1 mm x 150 mm, 1.8 µm). Mobile phase A consisted of 0.1% formic acid in water while mobile phase B was 0.1% formic acid in acetonitrile. 3 µg of sample was loaded per injection and eluted with a 95-minute gradient of 1 – 90 % buffer B at a flow rate of 250 µL/min. MS analysis of the separated samples was performed on a Thermo Q Exactive hybrid quadrupole-orbitrap mass spectrometer with a HESI source. Mass spectra were acquired in the positive ion mode ranging from 300-1800 m/z with a resolution of 140,000, AGC target of 3e6 and max IT of 200 ms. Raw LC-MS data were then imported into Chromeleon for identification and quantitation of rituximab PQAs. Components were defined when at least two of the following MS criteria were met: isotopic dot product ≥ 0.9, mass accuracy ≤ 0.5 and peak apex alignment ≤ 0.5 min. AUC integration was performed using the Genesis algorithm and a mass tolerance of 5.0 ppm was used for XIC extraction. The sum of all observable mass-to-charge ratios of the top four isotopes at each charge was used to calculate the areas.
Results: The average % relative abundance levels of all five lots for each of the 21 rituximab PQAs detected by MAM were as follows: H1-pyroglutamination (99.8%); L1-pyroglutamination (98.1%); Lys451 clipping (98.4%); Asn388 deamidation (0.61%); methionine oxidation (Met20, 1.6%; Met21, 1.3%; Met34, 3.2%; Met41, 3.6%; Met 256, 3.5%; Met432, 1.1%); Asn301 glycosylation (none, 0.76%; G0F-N, 1.0%; G1F-N, 0.68%; G0, 1.1%; G0F, 38.9%; G1F, 43.2%; G1FSa, 0.45%; G2F, 9.9%; G2FSa, 1.2%; G2FSa2, 1.4%; M5, 1.3%). Significant lot-to-lot differences were demonstrated using one-way analysis of variance (ANOVA) tests (p<0.05), and if the % relative abundance level for each PQA deviated by >20% from the average of all five lots (n = 15). Three of the 21 PQAs showed lot-to-lot variation due to one of the five rituximab lots having Met20, Met21, and Met34 oxidation levels that were higher than 120% of the average of all lots. For the other 18 PQAs, relative abundance levels were within 20% of the average for all lots, showing high lot-to-lot similarity with regards to those 18 PQAs.
Conclusion: MAM successfully detected and quantified 21 rituximab PQAs for five different lots of rituximab drug product. Lot variability was established for three rituximab PQAs, thus demonstrating the sensitivity of MAM in distinguishing product differences.