Mass Spectrometry in Late Development and QC: Practical Considerations for Multi-Attribute Monitoring and Beyond

Photo Credit: CNStock/Shutterstock.com

Scott Berger and Joe Fredette, Waters Corporation, Milford, Massachusetts, USA

As a result of the pharmaceutical cGMP for the 21st century and quality by design (QbD) initiatives championed by regulators, the biopharmaceutical industry has been looking for ways to introduce more automated and higher information content analyses into manufacturing, late-development, and quality control (QC). Mass spectrometry (MS-) based attribute monitoring assays have been proposed as key tools to provide the sensitivity, throughput, selectivity, and flexibility required for monitoring critical product and process attributes for biopharmaceutical production and release. Two analytical workflows, subunit multi-attribute monitoring (MAM) and peptide MAM, have emerged to dominate this discussion, and this article is intended to reflect on the active debates over the needs, challenges, and practical limitations for adopting MS-based attribute monitoring for late-development and QC.

A recent review by the FDA states that the use of mass spectrometry (MS) technologies detailing the characterization of biotherapeutic proteins in regulatory submissions has continued to grow and diversify over the last decade (1). The improvements in both capabilities and robustness of liquid chromatography–mass spectrometry (LC–MS) technology for biopharmaceutical analysis has led to its expanding role beyond product characterization in early development, to the monitoring of product and process attributes in late development, manufacturing, and quality control (QC) (2). However, this expansion has not been without controversy, and discussions over the analytical benefits and challenges of using LC–MS technologies for multi-attribute monitoring (MAM) in downstream laboratory environments are actively taking shape in the literature, and at industry events. This article will focus on the developments in the use of MS for biopharmaceutical development, manufacturing, and QC, and discuss some of the key pieces that must fall into place before MS-based methodologies can be more widely adopted within regulated (GxP) laboratories.

Defining MAM and Its Scope

In 2009, the FDA outlined the quality by design (QbD) initiative, which mandated that quality in manufacturing must be built in at the design stage (3). This implied a better understanding early-on of both the product and process, and deployment of tools to monitor and maintain high quality throughout the production process. Following that mandate, many biopharmaceutical companies began to invest in MS technologies for monitoring both product and process quality attributes in upstream and downstream development, and even for QC lot release applications. Drug development and QC organizations have also been driven by the desire for increased process understanding, enhanced product quality, and a desire to achieve gains in organizational productivity. Achieving these goals is a high priority for biopharmaceutical manufacturers, who are facing increased competitive pressure from both innovator and biosimilar producers, where speed to market, high product quality, and lower manufacturing costs are seen as essential to securing and retaining the market position of their drugs. 

Many companies have established “factory-of-the-future” programmes that include MAM-related initiatives that strive to achieve this competitive advantage. In recent years, the acronym MAM has gained considerable visibility within the biopharmaceutical industry, although the term has yet to be clearly defined. In its broadest form, multi-attribute monitoring refers to analytical approaches that can quantify multiple product and process attributes within a single analysis. The latest generation of these approaches has involved the use of mass spectrometry to provide increased selectivity, sensitivity, and flexibility to attribute analysis. 

Effective attribute monitoring typically requires existing product and process characterization data from which a panel of product or process attributes are selected for targeted monitoring. For mAbs, which represent half of all biotherapeutics being developed today, two distinct MAM-based LC–MS workflows have emerged, one focused on mAb subunit-based analysis, and the other on peptide mapping analysis. In the following sections, we discuss the progress and challenges with each of these workflows.

Subunit-Based Attribute Monitoring and MAM Methodologies

The value of MS has been widely acknowledged using top-down and middle-down approaches that can detect and measure product attributes, and are capable of being deployed within regulated laboratories. Early on, researchers at Roche deployed several MS‑based product release assays, which they discussed in a recent review article (4). In particular, they described a reduced subunit quadrupole time-of-flight electrospray ionization mass spectrometry (QTOF-ESI-MS) test for assessing Herceptin glycosylation variation in use since 2004, where data from over 500 drug batches has been obtained with only three system suitability deviations during this period. 

More recently, scientists at Eli Lilly used rapid peptic subunit LC–MS-based approaches to independently characterize the N-linked glycosylation patterns in the fragment antigen binding (Fab) domain, independent from the canonical glycosylation found within the fragment crystallizable (Fc) mAb domain (5). This approach provided quick and precise glycosylation profiling data during cell line and clone selection studies. In a subsequent presentation at a 2015 USP Workshop on Glycosylation Analysis for Biopharmaceuticals, scientists from Eli Lilly described the development of a glycosylation assay based on this analysis and detailed their ICH validation of this approach as a candidate release assay for their biotherapeutic. 

Scientists from Janssen have extended the utility of subunit-based analysis to include additional attributes not targeted by the researchers at Lilly. They developed a method that combined high-throughput protein purification along with rapid LC–MS analysis to simultaneously quantify multiple product attributes on aglycosylation, glycosylation, and glycation using very low volumes of cell culture fermentation broth (6). This high-throughput at-line analysis method enabled near real-time process monitoring that was utilized to identify correlations between process parameters and product quality in this study. The authors envisioned further development of this automated workflow into an efficient real‑time feedback control platform that could automatically adjust key process inputs based on pre-established detection limits to achieve target product profiles and improve overall product quality. 

 In a second study published by the scientists at Janssen, the authors detailed a high-throughput subunit MAM platform assay for monitoring global oxidation on mAb drugs (7). Reproducibility and intermediate precision of an automated subunit mass-based monitoring assay for oxidation monitoring was developed and qualified for all three IdeS digest subunits of an IgG1 (LC, Fc, and Fd’) using GMP‑compliant MS hardware and software. In this study, N-glycosylation of Fc domain was removed by enzymatic treatment, and oxidation was monitored at multiple methionine residues located throughout the molecule. The study highlighted the advantage of applying the subunit mass method to assess oxidation profiles, and the potential for monitoring additional attributes, such as glycosylation, glycation, clipping, and terminal modifications, with this high‑throughput platform assay. 

In recent years, the subunit attribute monitoring or MAM methodology has become the method of choice for some leading biopharmaceutical GMP laboratories, providing the benefit of simple sample preparation, robust relative quantification, and a streamlined workflow conducive to at-line high throughput monitoring or QC release testing. This methodology, however, is limited in its ability to monitor the small mass changes accompanying modifications, such as deamidation or isomerization, or generate site-specific information when multiple modification sites are located within the same subunit fragment. For these situations, peptide‑based analyses are viewed as more attractive workflows.

Peptide-Based Attribute Monitoring and MAM Methodologies

Despite increased challenges with sample preparation and data analysis, peptide‑based attribute monitoring and MAM approaches have received significant interest in the past few years. LC–MS-based peptide mapping assays have historically been used in regulated development and QC laboratories for targeted assessment of individual sites of modification in situations where optical‑based methods were deemed impractical. Typically, these analyses involved running a focused peptide map with either extracted ion chromatogram (XIC) or single ion recording (SIR)-based relative quantification of targeted peptides in their modified and unmodified forms for post-translational modification (PTM) analysis, and for generation of more selective peptide profiles that can be used to confirm product identity.

Recently, researchers in China established a validated identity assay for biotherapeutic mAb QC capable of distinguishing between 11 molecules produced at that manufacturing site (8). In this study, eight complementarity‑determining region (CDR) peptides were monitored and used to establish a signature CDR profile that could uniquely identify each product. This work was accomplished using a single quadrupole MS detector, which provided sufficient selectivity and specificity for this qualitative analysis.

Roche researchers also described their use of a peptide mapping LC–ESI‑MS product identity release assay for pegylated Interferon alpha-2a that has been in use since 2002, having been applied to over 350 batches of drug substance (4).

The applicability of MS detection technology for both quantitative and qualitative analysis within peptide maps has also been demonstrated in recent recorded presentations on attributes, such as glycosylation, oxidation, and deamidation (9).

In 2015, researchers at Amgen described their efforts to monitor multiple product attributes for characterization and process development using high resolution MS1 analysis, and described their vision of a multi-attribute methodology workflow that can greatly reduce the number of assays needed for process development, product disposition, and in-process control (10).

In subsequent publications and presentations, they have demonstrated the practical implementation of this approach, albeit with a much smaller subset of these potential attributes and discussed its use in support of a regulatory filing for clinical material (11). Similarly, Roche Diagnostics recently published on a peptide-based MAM approach for high-throughput sample preparation along with LC–MS analysis for quantification of deamidation, isomerization, oxidation, and glycosylation (12).

As mentioned previously, peptide-based MAM workflows are challenged by a more complicated sample preparation procedure that requires effort to optimize for robust digestion and recovery, while minimizing artifacts that can alter attribute levels, such as deamidation, isomerization, and oxidation. In addition, data interpretation for multiple peaks, each with multiple charge states, is fundamentally more complex when compared to subunit‑based MAM workflows based on spectral summation and spectral deconvolution. Certain targeted attributes are distributed throughout the molecule, such as oxidation and glycation, hindering the ability to directly correlate attribute profile changes with the potential root causes of bioactivity changes, as discussed in the Janssen study (7). Such changes may be diluted or missed at the peptide level, and are more easily monitored by aggregate changes at the global or subunit level. 

Some scientists have expressed the desire to detect new peaks beyond the targeted monitoring assay to satisfy regulators’ potential objections to replacing traditional assays. Given the complexity and variability of background ions within LC–MS peptide maps, the discussion of properly setting the parameters for meaningful detection of new peaks in a peptide-based MAM assay remains an unresolved element of peptide-based MAM discussions. Peptide-based monitoring workflows that enable rapid detection and identification of the “new peak” using the original monitoring data would require data independent global acquisition of peptide fragmentation data. An integrated workflow for monitoring and new peak characterization would maintain data integrity of the monitoring assay, while avoiding product release delays from reanalysis, as part of a batch failure investigation. Overall, this “new peak” detection requirement will be important for support of formulations and stability functions, but it remains a key discussion point on its appropriateness for the QC and lot-release function.

Deployment Challenges for Moving LC–MS Platforms Into GxP Laboratories

The practicality of deploying high‑resolution (HR)MS-based MAM approaches, especially when they apply to product quality control and release, has been questioned at many recent MS and biopharmaceutical industry events. As it stands today, HRMS systems require a degree of care and training exceeding that of the technicians and scientists typically deployed in the QC environment, and thus more experienced scientists will be required to operate these systems. This “grey-matter” challenge, combined with the “green-matter” challenge of deploying and maintaining LC–MS systems that cost 3–10 times more than typical optical detection LC installations in QC laboratories, can make it difficult for quality control laboratories to justify going down this road. The cost is further increased by the need for redundancy because operator absence and instrument downtime would not be tolerated as sources of delay for batch release. Additionally, the issue of instrument or informatics lifecycle management has been raised as a concern because of the assay re-validation requirements when LC–MS platforms evolve and new models are released. Companies will need to take this factor into consideration as they evaluate MS platforms to deploy within their organizations. The economics of this proposal are better borne by larger companies, leaving smaller organizations struggling to understand a path for implementation. In this sense, the move to simpler and less expensive mass detection‑based peptide MAM analysis could spur broader adoption.

Moving Forward: Where Do We Stand With MS in Regulated Laboratories and MAM?

Great strides have been made in adapting the MS technologies historically used for the characterization of biopharmaceuticals and applying them towards standard, routine assays that can be deployed in regulated development and QC laboratories. The dominant approaches today, subunit-based MAM and peptide‑based MAM, are demonstrating their usefulness, but issues with robustness, ease-of-use, method transferability, and cost of deployment still need to be overcome.

The “standard” instrument platform for biotherapeutic MAM analysis has almost certainly not yet been developed commercially, and many fundamental questions associated with moving these methodologies past product development remain, such as what does the regulatory path look like going forward?, and how soon after the acceptance of initial MAM‑based assays for approved products will requirements for the myriad of traditional assays they are intended to displace start to dissipate?

 Comments from regulators speaking at industry events such as the CASSS WCBP meetings indicate concern, and the need for longer term trending before these orthogonal assays can be dropped from filings and analytical release assay packages. Differential timing for this acceptance across regulatory bodies is of equal concern to those who are considering the adoption of these methods. 

Ultimately, the MAM paradigm envisions a range of fit-for-purpose LC–MS platforms that add significant analytical and business value to every stage of biotherapeutic development and commercialization. Building effective partnerships between tool providers, regulators, and biopharmaceutical producers will be key to realizing this vision. Continued honest debate and the sharing of knowledge and experience is required to catalyze the critical conversations that will shape the evolution of biopharmaceutical analytics for many years to come.

References

1. S. Rogstad et al., J. Am. Soc. Mass Spectrom.28, 786–794 (2017).
2. J. Fredette and D. Diehl, The Column12(12), 10–13 (2016).
3. A.S. Rathore et al., Biotechnol.27, 26−34 (2009).
4. M. Haberger et al., Am Pharm Review. 19(5), (2016).
5. A. Lim et al., Anal. Biochem. 375, 163–172 (2008).
6. J. Dong et al., Anal. Chem. 88, 8673−8679 (2016).
7. I. Sokolowska et al., mAbs9(3), 498–505 (2017).
8. J. Zhang et al., Chromatographia79(7), 395–403 (2016).
9. Webcast “Monitoring Product and Process Attributes in Biopharmaceutical Development and QC”, presented by Waters Corporation
10. R.S. Rogers et al., MAbs7(5), 881–90 (2015).
11. T. Wang et al., Anal. Chem.89, 3562−3567 (2017).
12. K. Bomans et al., American Pharm. Review16–21 (2016).

Scott Berger is Sr. Manager, Marketing, Biopharmaceutical Business Operations, at Waters Corporation, Milford, Massachusetts, USA.

Joe Fredette is Sr. Business Development Manager, Biopharmaceuticals, at Waters Corporation, Milford, Massachusetts, USA.

E-mail:scott_berger@waters.comWebsite: www.waters.com/biopharm