Validating MCSGP: Process Characterization and PPQ for Peptide Purification

Process Characterization and Performance Qualification of MCSGP

Ralf Eisenhuth and Thomas Müller-Späth

Processes 2025, 13(12), 3950 · Published 6 December 2025
Special Issue: New Frontiers in Chromatographic Separation Technology
DOI: 10.3390/pr13123950 · Open access, CC BY

Read the full article Download PDF

Competing interests, as disclosed by the authors: Ralf Eisenhuth was employed by Bachem AG and Thomas Müller-Späth by YMC ChromaCon. The research received no external funding.

Citation (BibTeX)

@article{eisenhuth2025process,
  title   = {Process Characterization and Performance Qualification of MCSGP},
  author  = {Eisenhuth, Ralf and M{\"u}ller-Sp{\"a}th, Thomas},
  journal = {Processes},
  volume  = {13},
  number  = {12},
  pages   = {3950},
  year    = {2025},
  doi     = {10.3390/pr13123950}
}

MCSGP purifies continuously, recycling impure side fractions between two columns so that high yield and high purity are reached together. It does that by running to predefined recycling and collection windows, cycle after cycle, for hours or days — without the mid-run adjustment to pooling boundaries that an operator can make in batch chromatography. The chromatographic separation therefore has a greater influence on final product quality, and correspondingly more weight falls on the control strategy around it.

That makes validation a more demanding exercise. How do you show, in a regulatory filing, that a process running to fixed windows for days will deliver the right material every cycle?

The study is a collaboration between Bachem AG and YMC ChromaCon, with Bivalirudin as the test system. It sits alongside an earlier characterization and validation study for CaptureSMB, the simpler twin-column capture process, and extends that thinking to gradient purification.

1. Why This Study Matters: A Gap Between Practice and Paperwork

MCSGP is in operation at several companies and research institutes, purifying synthetic peptides and oligonucleotides at production scale. Its appeal is well documented: internal recycling of impure side fractions lets it reach high yield and high purity at the same time, with a high degree of automation, fewer in-process controls and a lower process mass intensity than batch chromatography.

Regulatory validation is a separate matter from process performance, and it has its own structure. The FDA Process Validation guideline sets out three elements:

• Process Design Builds scientific understanding of the process. Process Characterization is the central piece: identifying which parameters could affect critical quality attributes, testing the high-ranked ones experimentally, and defining the ranges within which quality is met.
• Process Performance Qualification Confirms that the equipment and process perform as intended and can manufacture reproducibly at commercial scale. At least three consecutive representative batches are run under normal operating conditions, in the actual facility, with trained operators and qualified equipment.
• Continued Process Verification Ongoing monitoring once routine manufacturing is under way. This study covers the first two elements; the third is outside its scope.

None of that framework is specific to continuous chromatography, and the paper’s finding is that it largely does not need to be. What was missing was a worked example showing how it applies to a process whose defining feature — recycling material between two columns — has no counterpart in batch operation.

Where MCSGP differs, and where it does not

The paper separates the purification into two parts: the chromatographic part, which is identical for batch and MCSGP, and the product collection and recycling part, which is specific to MCSGP. The list of process parameters is therefore largely the same as for a batch process, with a few additions covering the recycling and collection step.

The practical difference is how many elutions it takes to generate meaningful data. For a batch purification, one elution per tested parameter is considered sufficient. An MCSGP experiment needs at least three cycles with two elutions each — enough to confirm that the process has reached cyclic steady state, that impurities are not accumulating through the weak or strong recycling streams, and that changes in column load caused by internal recycling do not affect the outcome.

2. The Study at a Glance

Bivalirudin was used as the test system: a 20-amino-acid direct thrombin inhibitor with a molecular weight of 2,180 g/mol, produced by solid-phase peptide synthesis and supplied as crude material of 81% purity. The target for the product pool was more than 99.0% purity, with the added constraint that the impurity profile had to be comparable to the legacy batch process.

What was run, and on what

3. How Far the Parameters Were Pushed

Most of the 34 identified parameters were never tested experimentally, and the reasoning behind that is worth following. Large-scale MCSGP equipment holds most parameters within narrow limits by construction. A Contichrom TWIN system has four three-head pumps — two for gradient generation, one for feeding, one for in-line dilution, cleaning-in-place and equilibration — and each is controlled by a dedicated flow meter with a calibration tolerance of ±2%.

Whether that ±2% matters depends entirely on what the pump is doing. For equilibration, in-line dilution, cleaning-in-place and chase washes, a 2% variation is negligible. For column load it is not: a 2% change alters the peak shape, which can shift what ends up inside the recycling zones. So the same tolerance makes one parameter well-controlled and another potentially critical, and only the second needs an experiment.

That assessment left four parameters to test. Each was then varied well beyond the range it covers in production.

4. What Process Characterization Found

Every set point produced product above the 99.0% purity specification. Across all the characterization runs the average pool purity was 99.30%, and every individual value fell inside a band of ±3 standard deviations, that is 99.30% ± 0.32%. Since ±3σ captures roughly 99.7% of expected values for a normal distribution, the effect of the parameter variations was statistically indistinguishable from ordinary process noise. On purity alone, all four parameters would have been classified as non-critical, and the proven acceptable range would have been the full characterized range.

Each eluate pool was also examined for individual impurities. Those profiles were consistent across every run but one: at −25% displacer slope, a single impurity rose from below the limit of quantitation to 0.13%. Small in absolute terms, but the specification required the impurity profile to be comparable to the legacy process, and it no longer was.

On that basis gradient slope was designated a critical process parameter, and the variance permitted during routine manufacturing was reduced through organizational measures. Column load, elution flow rate and detection wavelength were designated non-critical. The gross-to-gross yields were also evaluated; excluding the shallow-slope run, they fell between 60% and 65%.

Why one factor at a time, rather than a designed experiment

The choice of OFAT over DoE is argued explicitly, and the reasoning is specific to peptides. Related substances in a peptide tend to behave as regular samples, with similar retention behavior under gradient conditions, so the primary observable effect of parameter variation is the width of the overall peak rather than a change in resolution. Purity changes in MCSGP then tend to come from misalignment between the peak and the fixed recycling and collection zones.

That creates a practical hazard for a designed experiment. Combinations that cumulatively broaden or narrow the peak — a lower column load with a steeper slope and a wavelength shift that lowers the signal, for instance — can push the peak far enough out of alignment with fixed zones to fail the run, forcing repeated iterations of the DoE setup. OFAT does not reveal interactions between parameters, which is a real limitation, but it allows much wider ranges to be tested. For molecules with stringent purity specifications the paper recommends including the combinations that produce the narrowest and broadest peaks as fixed experiments, and considering smaller ranges.

5. Supervising a Process That Runs for Days

An MCSGP process producing eluate continuously for several days presents a sampling problem that batch chromatography does not. Collecting and analyzing every main cut separately is not cost-effective at production scale, but sampling too little leaves the process unsupervised. The answer here is a two-phase strategy.

• Start-up: intensive monitoring Every main cut is analyzed by an in-process control method until three criteria for cyclic steady state are met — comparable impurity profiles for the main cuts, good overlay of the UV traces on each column, and comparable main cut volume on a column basis.
• Steady state: strategic sampling Once those criteria are satisfied, HPLC sampling drops to two samples per day. Supervision then rests on three elements together: the strategic samples, the online UV trace overlay, and the recorded main cut volume per column.

The approach is aligned with the risk management expectations for product availability set out in ICH Q9 (R1).

AutoPeak as Process Analytical Technology

Underpinning the strategy is AutoPeak, which monitors UV at the column outlets and starts and stops the recycling and collection phases from absolute and relative UV triggers rather than from the clock. The main cut borders are set as UV thresholds entered into the control. Because the cut points track the peak rather than elapsed time, retention time shifts that accumulate over a long run do not move the pool composition.

The qualification runs illustrate this. Over PPQ batch 1, a slight peak shift appeared from cycle to cycle; when the traces were realigned to the AutoPeak trigger point marking the start of product collection, they overlaid exactly. On that basis the paper positions AutoPeak as Process Analytical Technology in the sense of the FDA PAT guidance — a framing supported by the FDA Process Validation and PAT guidance documents and by ICH Q8, Q11 and Q13, which explicitly support real-time monitoring as part of a control strategy.

6. The PPQ Campaign: Full Results

A pre-PPQ GMP run came first: eleven elutions, 1.5 kg of crude purified to 0.8 kg of intermediate at 99.6% purity, with a PMI of 1,402 kg/kg for the purification alone. The qualification campaign then ran four multiday MCSGP runs, the last batch split into two parts, to cover a range of batch sizes.

All four produced consistent elution profiles, main cut sizes and product purities. Comparing the third elution — the first at cyclic steady state — with elution 106 of the longest run showed the impurity profile holding across the full duration, and purity for every full elution stayed within the ±3σ band anticipated from characterization.

• Crude in 54.5 kg at 81% purity
• MCSGP 397 elutions over 235 hours, producing 1,576 L of main cut
• SPE 42 elutions over 65 hours on the same system and columns, alternately loading and eluting with a sharp gradient, producing 631 L of main cut
• API out 33.6 kg at 99.2% purity — a gross-to-gross yield of 62%, or 76% effective yield taking product and crude purity into account
• Total 300 hours of chromatography operating time

Against the legacy process

Swipe the table sideways to see the legacy and MCSGP columns.

Source: Eisenhuth and Müller-Späth, Processes 2025, 13, 3950 (CC BY). The legacy figure is the existing Bivalirudin batch process, which includes extensive side-cut re-chromatography.

Experimental conditions

7. What This Means for Peptide Manufacturing

On the evidence here, the regulatory framework does not need reinventing for continuous chromatography. The parameter list is largely the batch list, the risk assessment is the standard one, and the qualification requirement is the standard three-batch expectation. What changes is the number of elutions needed to generate data, and the shift from correcting a process to supervising one.

That is a useful result for anyone weighing continuous purification against the validation work it would imply. Of 34 identified parameters, four warranted experiments and one turned out to be critical — and the paper notes that those same four would also be investigated during characterization of the corresponding batch process. The characterization burden is comparable; the difference is material consumption, since cyclic steady state has to be confirmed from successive identical elution profiles. In this study that extra experimental time was minimized by daisy-chaining the required MCSGP runs and executing them automatically.

The operational advantages appear in the campaign figures rather than in the validation argument. Running uninterrupted over several days, with the control strategy supported by online measurement, produced a substantial reduction in analytical sampling and process mass intensity alongside a higher gross yield. The authors also note that data integrity and electronic batch record handling requirements are identical for batch and MCSGP.

8. Key Terms in This Study (Glossary)

9. Frequently Asked Questions