HPLC 2025 Preview: Fundamentally Speaking (Part 2)

Michael Lämmerhofer: The Swedish separation science community (with Göran Schill and coworkers) has a long tradition of ion-pair extraction and thereafter ion-pair chromatography. Ion-pair chromatography’s reputation was in the past not always the best. A particular problem of ion-pair LC is system peaks. For the average chromatographer they are hard to understand or grasp, and probably for this reason they are often also termed ghost peaks. You found a way to simulate ghost peaks that means you understand their origin. Why should we not be afraid of ghost peaks? How can they be explained?

Michael Lämmerhofer from the Institute of Pharmaceutical Sciences, University of Tübingen, Germany, spoke to JFK Huber Lecture Award winner of 2024 Torgny Fornstedt, professor in analytical chemistry and leader of the Fundamental Separation Science Group, Karlstad University, Sweden, about his pioneering work in high performance liquid chromatography (HPLC) with a focus on fundamentals and industrial applications.

Torgny Fornstedt: A specific subset of ghost peaks—known as system peaks—arises from disturbances in the chromatographic system’s equilibrium, particularly when additives such as ion-pairing agents are involved. These peaks are not contaminants but are actually generated by the injection itself. System peaks are typically not visible under normal conditions; they travel through the column as undetected zones and only appear as visible peaks at the column outlet if the additive is detectable by the chosen detection method.

When a sample is injected, differences between the sample and the mobile phase can create transient zones of additive concentration. These zones travel through the column, interacting with the stationary phase and leading to detectable signals—even in the absence of analytes. Such system peaks are not contaminants but are predictable outcomes of the system’s response to changes in additive concentration. As discussed earlier, understanding these dynamics is crucial for accurate interpretation.

Through our research, we’ve developed simulations that accurately predict the occurrence and behavior of these system peaks. By understanding their origin, chromatographers can anticipate and manage these signals, turning potential challenges into opportunities for deeper insight into the chromatographic process.

ML: Ion-pair LC has experienced a renaissance, driven mostly by its great utility for oligonucleotide chromatography. Is it possible to simulate adsorption processes in ion-pair separation mode, considering the challenges posed by multivalency and the significant, yet difficult-to-predict, contributions of secondary structure effects to the separation? Have you studied the details of ion-pair LC of oligos with hexafluoro-propan-2-ol (HFIP) and its peculiarities? Do we properly understand the effects of this fluorinated agent?

TF: Yes, simulating ion-pair LC for oligonucleotides is feasible, despite complexities such as multivalency and secondary structure. We’ve developed mechanistic models that account for ion-pair formation and analyte adsorption, accurately capturing retention and peak behavior. Secondary structures do add challenges, but under denaturing conditions, oligonucleotides behave predictably enough for modeling.

Interestingly, despite their overall complexity, ion-pair systems for oligonucleotides are less prone to ghost or system peak issues. Unlike traditional ion-pair LC where strongly adsorbing additives can distort peaks, oligonucleotide separations often use small, volatile, and polar ion-pairing reagents that don’t adsorb strongly to the stationary phase. This makes these systems paradoxically safer from the strange band shapes described earlier, even if they are mechanistically more intricate.

ML: You are collaborating with the pharmaceutical industry in the field of oligonucleotide therapeutics. I understand that chromatography is the workhorse for preparative isolation of oligonucleotide therapeutics. How many chromatographic process steps are required for producing pharmaceutical-grade oligos that can be administered to humans?

TF: Yes, chromatography is the foundation of oligonucleotide purification and is essential for producing pharmaceutical-grade material. In our collaborations with industry, purification is never a one-step process—it typically involves three to six chromatographic steps, and up to eight for complex molecules like siRNA, which is a double-stranded duplex requiring additional, separate purification.

These steps generally fall into five key categories:

• Initial Capture and Desalting: Crude oligonucleotides contain truncated sequences, salts, and byproducts. Capture is typically done using ion-exchange chromatography (IEX), ion-pair reversed-phase liquid chromatography (IP-RPLC), or anion exchange chromatography (AIEX); size-exclusion chromatography (SEC) or ultrafiltration is used for desalting.

• Main Purification: This step removes closely related impurities, often via IP-RPLC in early development or IEX with sodium in later stages. These are typically overloaded, nonlinear processes that benefit from modeling.

• Polishing: Trace-level impurities are removed using a second, orthogonal method—commonly another IEX, reversed-phase, or HILIC-like step. IP-RPLC followed by AIEX is a well-established approach.

• Counterion Exchange and Desalting: To replace volatile counterions (tetraethylammonium [TEA]) with pharmaceutically acceptable ones (sodium), techniques such as SEC or IEX are used—typically followed by lyophilization.

• Intermediate Purification (if needed): For modified or conjugated oligos, particularly siRNA, additional purification steps may be required for individual strands or intermediates.

ML: Why so many steps?

TF: Oligonucleotides are structurally complex—charged, flexible, and prone to secondary structures. Even minor impurities can have biological effects, so regulatory expectations are extremely high. Each step serves a specific purpose: removing impurities, exchanging counterions, or ensuring formulation compatibility.

To support this, we apply mechanistic modeling and digital chromatography to simulate and optimize each step—from early development through GMP-scale manufacturing.

In summary, producing therapeutic oligonucleotides requires a customized, multi-step process, designed not only for purity but also for scalability, robustness, and regulatory compliance.

ML: A keyword that appears as a red line in your recent research is digital technology. Can you tell us what it means in the context of your fundamental separation science research work? How can you benefit from this tool?

TF: In our research, digital technology means using mechanistic models, simulations, and algorithmic tools to understand and predict chromatographic behavior. We build digital twins of separation systems to simulate experiments, extract key parameters, and design methods virtually. This shifts chromatography from trial-and-error to a predictive science, saving time, improving robustness, and providing deeper molecular insight.

Recently, we launched a new project after a long-time industry colleague highlighted persistent issues with chromatographic integration in quality control (QC)—particularly when reporting results to regulatory authorities. At first, we were skeptical, but the issue turned out to be real and surprisingly widespread (1–3).

Current QC software often relies on basic integration methods such as droplines, peak height, and skimming, and typically requires time-consuming manual tuning. These approaches frequently struggle with overlapping or skewed peaks, baseline drift, and accurate peak identification, may result in significant errors and inconsistent outcomes. From a regulatory perspective, the outcome of manual tuning might be problematic because it varies based on individual skill and experience.

To address this, we’ve developed a new deterministic software that performs fully automated, model-based peak integration using, if possible, model-based fitting. It requires no manual tuning and is currently being validated with pharmaceutical partners to improve QC accuracy, reproducibility, and regulatory confidence.

While artificial intelligence (AI) and machine learning offer promise in many fields, we believe they may not be ideal for this application. In QC, consistency is critical—results must be repeatable and transparent, not dependent on how much the system has “learned.” Our deterministic approach ensures the same outcome every time, making it better suited for regulated environments.

References

(1) Dolan, J. Peak Integration Errors: Common Issues and How to Fix Them. Sep. Science 2021, https://www.sepscience.com/hplc-solutions-129-peak-integration-part-3common-integration-errors-6962 (accessed 2025-05-06).

(2) Bicking, M. K. L. Integration Errors in Chromatographic Analysis, Part I: Peaks of Approximately Equal Size. LCGC North Am. 2006, 24, 402–414.

(3) Bicking, M. K. L. Integration Errors in Chromatographic Analysis, Part II: Large Peak Size Ratios. LCGC North Am. 2006, 24, 604–616.