Continuous Anion Exchange Purification of a Conjugated siRNA

Continuous Purification of a Conjugated Short Interfering RNA Therapeutic Using Anion Exchange Twin-Column Chromatography (MCSGP)

Richard Weldon, Tobias Vandermeersch and Thomas Müller-Späth

Organic Process Research & Development 2025, 29(6), 1400–1410 · Published 2 June 2025
DOI: 10.1021/acs.oprd.4c00513 · Open access, CC BY 4.0

Read the full article Supporting information

The authors declare no competing financial interest. Richard Weldon and Thomas Müller-Späth are at YMC ChromaCon; Tobias Vandermeersch is at Johnson & Johnson, which provided funding and the feed material for the study.

Citation (BibTeX)

@article{weldon2025continuous,
  title   = {Continuous Purification of a Conjugated Short Interfering RNA
             Therapeutic Using Anion Exchange Twin-Column Chromatography (MCSGP)},
  author  = {Weldon, Richard and Vandermeersch, Tobias
             and M{\"u}ller-Sp{\"a}th, Thomas},
  journal = {Organic Process Research \& Development},
  volume  = {29},
  number  = {6},
  pages   = {1400--1410},
  year    = {2025},
  doi     = {10.1021/acs.oprd.4c00513}
}

In oligonucleotide manufacturing, the expensive step comes before the chromatography. Solid-phase synthesis assembles the sequence one coupling at a time, and every incomplete coupling leaves a chemically similar impurity behind — shortmers and longmers that the purification then has to separate from a target they closely resemble. Product lost at the polishing step has to be replaced by synthesizing more, which is where the cost sits.

That makes purification yield unusually valuable here. A polishing step that recovers more of what was loaded lets the synthesis batch shrink proportionally, taking reagents, solvents, waste and column volume down with it.

It also sets out how the conversion was done — a software-aided procedure for taking a batch chromatogram and turning it into a running continuous method — and projects what the same process would deliver on production-scale equipment.

1. Why This Study Matters: Yield Is the Expensive Variable

Oligonucleotide drugs have moved from promise to product, with 15 approvals to date and hundreds of candidates in clinical trials. Making them at commercial scale remains challenging and expensive. At the time of writing, the maximum commercial GMP scale is limited to a 2,000 mmol synthesis scale using solid-phase oligonucleotide synthesis, and losses during downstream purification limit the maximum output per batch further still.

What batch polishing costs

The standard polishing step is single-column ion exchange or reversed-phase chromatography, with the pure center cut of the chromatogram collected and the rest set aside. Because the target does not fully separate from chemically similar impurities, the lower-purity side cuts have to be reprocessed or discarded.

• Fractionation and mock pooling Yields in batch mode are maximized by fractionating and analyzing the elution peak to find the optimal center-cut boundaries, then generating multiple mock product pools for analytical testing and QC. The paper describes these tasks as essential for batch release and a significant proportion of the overall manufacturing burden in a GMP environment.
• Rechromatography of side cuts The common alternative is to pool out-of-specification side fractions from several batches and re-run them. Side cuts are transferred into custom storage bags, held in a dedicated walk-in freezer, pooled and diafiltered to adjust buffer composition before reloading. It improves overall yield, but reduces the productivity of the process.
• Synthesis has to make up the difference Whatever the polishing step loses must be replaced upstream. Oligonucleotide synthesis is scaled up to compensate for product lost during purification, which the paper notes is very costly.

MCSGP addresses this by recycling the low-purity side cuts internally to a second column, with in-line dilution to aid readsorption, so that rechromatography becomes unnecessary. The technique has been demonstrated across a range of biomolecules; what this study adds is a case in anion exchange mode, on a conjugated siRNA, taken from an existing batch process.

2. The Study at a Glance

The molecule is the sense strand of a therapeutic siRNA, synthesized by standard phosphoramidite chemistry on the solid phase with the sugar conjugation performed during synthesis. The crude entered purification at 85 area% product purity, with the critical N-1 shortmer impurity at 0.93 area%, against a target purity specification of more than 95%.

What was run, and on what

Both eluents were the same in kind: 20 mM sodium phosphate with 10 vol% acetonitrile at pH 8.0, with sodium chloride providing the gradient. The pilot method used 0.9 M NaCl in mobile phase B and the downscale runs 1.1 M, which produced a narrower peak in the downscaled runs — a difference the authors expected the internal recycling to compensate for.

3. Three Ways to Run the Same Separation

The study reports three processes on the same chemistry: the pilot batch method, the same method downscaled, and MCSGP built from that downscaled method. Comparing all three is what separates the effect of the operating mode from the effect of column geometry.

4. From Batch Method to Continuous Process

MCSGP is more complex to set up than single-column chromatography, and the paper notes that it requires specialized process development tools to optimize and implement. Its practical contribution is a procedure that starts from something a process chemist already has — a fractionated batch chromatogram with purity data — and turns it into a running twin-column method.

• Run a batch design method The downscaled batch run was repeated at an elevated elution flow rate of 200 cm/h to maximize MCSGP performance, the eluate fractionated, and each fraction analyzed by HPLC. The chromatogram was then overlaid with the measured fraction purities.
• Set the phase boundaries on screen That chromatogram and its purity data were uploaded to MCSGP Wizard v8.1. The boundaries between the recycling and collection phases were set by drag-and-drop against the purity profile, to optimize the purity of the product collection window.
• Let the software size the rest Gradient parameters, the load per switch and the in-line dilution factors were calculated automatically from the imported chromatogram and the relative position of the phase boundaries. Washing, cleaning and regeneration steps were entered into a template table.
• Run the generated methods The wizard produced a set of startup, main and shutdown methods ready to run on the Contichrom CUBE hardware.

What the four phases do

Each MCSGP switch has four phases, and the two that define the technique are the recycling steps. In phase 1 a gradient starts on the first column while the second is cleaned and regenerated in parallel. In phase 2 the columns are interconnected and the early-eluting, low-purity side cut is transferred from the upstream column to the downstream one, with in-line dilution lowering the modifier content so the product readsorbs. In phase 3 the columns run in parallel again: pure product is collected from the first while the second receives fresh crude feed. In phase 4 the late-eluting, low-purity side cut is recycled the same way as in phase 2. The columns then swap positions and the sequence repeats.

5. What the 13-Cycle Run Showed

A 13-cycle MCSGP run was carried out, giving 27 product elutions including startup and shutdown, over a total run time of 17 hours. The cycle number was limited only by how much crude was available for the study.

Convergence

A continuous process has to settle before it can be judged, and the study tracks that settling three ways. The area under the UV curve for the product collection phase reached a constant value by cycle 5, indicating that the mass of product entering and leaving the system was in steady state. Product purity followed the same pattern: it starts at 95.6 area%, declines and stabilizes at around 95.1 area%. The critical N-1 impurity starts at 0.5 area%, rises and stabilizes at around 0.7 area% — inside the specification of less than 1.0%.

Per-cycle yield was steady at roughly 94–95%, with product recovery holding between 127 and 130 mg per cycle. The shutdown step returned a much lower 62%, which is a normal feature of MCSGP: it is calculated against the larger load applied during startup, and its effect on the overall figure shrinks as the number of cycles rises. Including startup and shutdown, the overall yield was 93%.

Impurity accumulation, watched rather than assumed

Some accumulation was detectable in the region undergoing recycling of strongly adsorbing impurities. The paper treats this as expected rather than alarming: impurity accumulation in the recycling regions is typical in MCSGP, and only becomes a problem if it extends into the product pool and causes the specification to be missed. It did not here, and the stated remedy if it had would be to narrow the recycling windows during process design and send more impurities to waste.

The conductivity traces overlaid reproducibly across the whole run, indicating stable system performance, and the authors note that monitoring area under the curve cycle to cycle gives an operator a live signal: deviation from predefined limits can trigger a run to be adjusted, paused or aborted before time and material are wasted.

6. Full Results and Scale-Up

Swipe the table sideways to see all three processes.

Source: Table 4, Weldon, Vandermeersch and Müller-Späth, Org. Process Res. Dev. 2025, 29, 1400–1410 (CC BY). * The downscaled batch was evaluated by HPLC-UV only, not by mass spectrometry, so those two values are not strictly comparable to the others.

Reading the productivity figure carefully

Productivity per liter of resin was 9.88 g/L_resin/h for MCSGP against 10.09 for the downscaled batch — effectively the same. The gain over the pilot batch process, at 5.28, comes from bed height: the downscaled column is half as tall, so residence time during elution halved and productivity roughly doubled. In practice, half the bed height with twice the productivity gives similar overall throughput, since throughput is productivity multiplied by column volume.

The 87% throughput advantage the paper reports is a different comparison: an MCSGP system runs two columns rather than one, at the same column diameter. Put another way, one elution in MCSGP completed in 38 minutes against 90 minutes for one batch elution from the same 10 cm column. Three things contribute: the higher yield, the higher elution flow rate used in MCSGP (200 against 150 cm/h, which carries less risk because recycling safeguards yield), and the fact that regeneration, cleaning and feeding run in parallel with elution on the other column.

Scale-up scenarios

MCSGP processes developed on the Contichrom CUBE are directly scalable to the Contichrom TWIN, a production system capable of operating under GMP. The study projects four configurations.

Swipe the table sideways to see all four configurations.

Source: Table 5 of the study. These are calculated scale-up scenarios rather than demonstrated runs. The paper states that the largest configuration, assuming 24 hours of continuous operation, enables multiton-scale production per year.

7. What This Means for Oligonucleotide Manufacturing

The headline number is a 13 percentage point gain in yield, but the argument the paper makes is about what that gain does upstream. Higher chromatographic yield allows the chemical synthesis batch to be scaled down proportionally for the same output of product.

That matters because synthesis is where the raw material cost, the custom intermediates and the solvent burden sit. The authors report that for equivalent product output, MCSGP carries 14% lower cost of goods, and they attribute the saving to fewer raw materials and consumables, fewer custom oligonucleotide intermediates, smaller synthesis batches and smaller volumes of chromatography resin. On resin specifically, they calculate that a production-scale single-column batch system requires 86% more resin than MCSGP for equivalent output — and note that this figure may understate the difference, because batch chromatography is not typically run continuously.

Alongside the measured process improvements, the paper sets out a series of operational benefits it expects from automation: fewer in-process control samples and so a reduced GMP analytical burden, the elimination of rechromatography with its storage bags, freezer space and diafiltration, less operator interaction and less scope for user error, and smaller columns that are easier to pack. These are argued rather than measured in this study.

One methodological note worth carrying forward: the anion exchange method here reached above 95% purity in both batch and MCSGP modes, where the authors say NaOH-based AIEX methods in their experience typically give 88–92%. They suggest the phosphate-buffered mobile phase with 10% acetonitrile may have improved selectivity by reducing unwanted hydrophobic interactions, but also note the crude entered at a high starting purity of 85%.

8. Key Terms in This Study (Glossary)

9. Frequently Asked Questions