Automated Liquid-Liquid Extraction

Some analytical sample preparation techniques have been used successfully for decades, even centuries, but can still pose a challenge depending on the sample matrix. Liquid-Liquid Extraction (LLE) is such a method. The separation principle used is easily explained along with its practical implementation: Analytes are distributed between two immiscible phases, typi-cally an aqueous matrix and an organic solvent, according to their relative solubilities. The steps that follow, however, are far less simple, considerably more labor-intensive, and each car-ries the potential to introduce errors into the process. And the incorrect analysis results subsequently obtained are not even easily identified as such. In other words: A significant portion of the total measurement uncertainty in chemical analysis does not originate within the analytical instrument and the associated method, but rather in the sample preparation and other steps preceding it. Automation does not solve this problem by simpli-fying the chemistry; rather, automation establishes superior process control. To illustrate this, this article presents four ex-amples from the fields of veterinary medicine, forensic toxicolo-gy, and food safety from the GERSTEL Application Laborato-ries. Although the applications rely on some different tech-niques, they lead to the same conclusion.

Case 1: Ketamine in Horse Serum – Turning Nine Manual Steps into one Automated Workflow [1]

The manual LLE protocol for the determination of ketamine in biological matrices consists of nine steps: alkalization of the sample with a potassium hydroxide solution, addition of an MTBE/dichloromethane mixture (7:3 v/v), vortex mixing at 2000 rpm for 5 minutes, centrifugation, transfer of the organic phase into a clean glass vial, evaporation at 40 °C, reconstitution in a mixture of formic acid, water, and methanol, filtration, and finally injection into the LC-MS/MS system. In a manual operation, each of these steps constitutes a potential source of error.

The manual protocol was transferred to an automated Prep Sequence on a GERSTEL MPS Dual Head WorkStation equipped with an mVORX mixing module, an mVAP evapora-tion station, and a filtration option. This resulted in a coefficient of variation of 0.467% across four replicates, with a recovery rate of 88.7%. Three replicates of the pure standard solution showed a variation of 0.182% CV. The automated extraction from this complex biological matrix yielded almost the same precision as the direct measurement of the standard. This is a result that would be hard to achieve using the manual proce-dure, especially in routine operation, even under strictly con-trolled laboratory conditions.

The MPS WorkStation also enables automated method devel-opment: For the determination of buprenorphine and norbu-prenorphine in bovine plasma, four extraction solvents of vary-ing polarity—hexane/IPA (99:1), methyl *tert*-butyl ether, meth-ylene chloride, and ethyl acetate—were systematically com-bined with three pH conditions (acidic, neutral, and basic) and evaluated. The result: Hexane/IPA under acidic conditions yielded the best recovery rates for both analytes. The method development procedure was performed overnight without labor-atory staff intervention; The Analysis Instrument was set up to make the results were available the following morning. What would take weeks to perform manually was finalized in a matter of hours.

Case 2: Bisphenol A in Beverages: The required Limit of Quantitation cannot be reached using the official Method [2]

In April 2023, the European Food Safety Authority (EFSA) low-ered the tolerable daily intake (TDI) for Bisphenol A by a factor of 20,000 to 0.2 ng/kg of body weight per day. For a 60-kilogram adult, this corresponds to a tolerable daily intake of 12 ng. Based on the simplified assumption of a daily beverage consumption of one liter, this limit would be reached at a BPA concentration of 12 ng/L. The established AOAC reference method (Official Method 2017.15) operates with a validated LOQ of 300 ng/L, more than 25 times higher than the regulato-ry-relevant concentration. This is of course nowhere near suffi-cient sensitivity for exposure assessment in accordance with the new EFSA TDI.

Automated Salting-out Assisted Liquid-Liquid Extraction (SALLE) - performed on a GERSTEL MPS roboticPRO equipped with a quickMIX mixing module and CF-200 centrifuge is able to bridge the gap and upgrade the AOAC method. The underlying principle is the salting-out effect: The addition of so-dium chloride to the aqueous sample displaces the polar sol-vent, acetonitrile, from the aqueous phase, thereby inducing phase separation. Since acetonitrile is directly compatible with LC-MS/MS, the evaporation step is entirely eliminated. This not only removes a potential source of error but also yields a signif-icant time saving. Four milliliters of sample, one gram of sodium chloride, five pulse-mixing cycles of 15 seconds each at 1500 rpm, five minutes of continuous mixing, ten minutes of centrifu-gation—done.

The results are compelling: an LOQ of 0.1 ng/mL, an R² value of 0.997, and an average accuracy of 99.9% with a relative standard deviation (RSD) of 2.32%, spanning seven distinct beverage matrices ranging from soft drinks, orange juice, and iced tea to protein drinks. Matrix-dependent recovery variations ranging from 89.4% (iced tea) to 130% (pulp-free orange juice) reflect actual ionization effects within the LC-MS/MS system; this is not a methodological error, but rather a matrix effect that was reliably compensated for through the use of an isotopically labeled internal standard (d₁₆-BPA).

Case 3: Fatty Acids in Infant Formula—When a few Sec-onds Determine Data Quality [3]

Infant formula ranks among the most strictly regulated food products worldwide. Its fatty acid profile, specifically concerning DHA and arachidonic acid, is essential for neurological devel-opment during the first months of a life and requires reliable analytical determination. The reference method used is AOAC 2012.13, a transesterification technique for the determination of fatty acid methyl esters (FAMEs) via GC-FID. Its most critical parameter is not the chemistry, but the timing: The hexane addi-tion must occur exactly 180 (±10) seconds after the sodium methoxide is added, followed by the neutralization solution 30 seconds later. These specific time windows govern the reaction kinetics of the transesterification process. When processing multiple samples manually in parallel, consistent adherence to the timing a real challenge.

Automated implementation on an MPS robotic/MPS

roboticPRO system, equipped with a Valco M50 pump module and a Motion 40 agitator, ensures consistent perfect timing. The MAESTRO software accurately controls every addition point to the second, while the pump module dispenses the neutralization solution at a defined flow rate of 35 mL/min. Accuracy is em-bedded in the process, independent of the number of samples and operator performance on any given day.

The most significant finding of this process optimization was not a gain in efficiency, but rather an improvement in quality. For example, a controlled agitation of 500 rpm proved optimal, as opposed to maximum rpm. Higher rotational speeds actually promoted emulsion formation and impaired phase separation. Consequently, automation in this context serves not just to ramp up the throughput, but rather to reproducibly establish the best possible operating conditions. Two powdered infant formu-la samples, with fat contents of 25.5% and 27.1% respectively, yielded mean precision values of 2.86% and 10.0% CV. The higher value observed in the second sample is attributable to matrix-specific effects—effects that are revealed through auto-mation rather than being obscured by operator-induced variabil-ity.

Case 4: Xylazine and Medetomidine in Plasma and Urine - When Automation Goes All The Way [4]

In the USA and Canada, xylazine has emerged as a forensic challenge due to its use as an adulterant in illicit fentanyl-containing street drugs: The U.S. Drug Enforcement Admin-istration (DEA) has detected its presence in seized samples across 48 of the 50 states. Medetomidine is a selective α₂-adrenoceptor agonist used in veterinary anesthesia, which in-creasingly is gaining relevance in forensic contexts involving cases of poisoning and substance abuse. Both compounds exist in biological matrices (plasma and urine) in a conjugated form, necessitating enzymatic pretreatment prior to the actual extraction step. This significantly increases the complexity of the workflow.

The automated method, executed on the MPS roboticPRO sys-tem equipped with quickMIX, a centrifuge, and a MultiSample Evaporation Station (mVAP), comprises the entire workflow: The enzymatic hydrolysis of conjugated analytes using β-glucuronidase (15-minute incubation at room temperature) and dual chloroform extraction with subsequent pooling of the or-ganic phases followed by evaporation of the extract and recon-stitution of the residue in a 1:1 ethanol-water mixture and, final-ly, injection into the LC-MS/MS system. The process is fully automated and requires no manual intervention.

The performance data are remarkable for a multi-step workflow involving biological matrices: An LOQ of 0.1 ng/mL for both ana-lytes in both matrices, R² values ≥ 0.997, recoveries ranging from 88.7% (Medetomidine in plasma) to 115% (Xylazine in plasma), and precision (RSD) of less than 3.6%. For plasma, the mean accuracy was 101% for Xylazine (range: 87.3–104%) and 101% for Medetomidine (94.7–107%); for urine, the corre-sponding values were 101% and 102%, respectively. Matrix effects were compensated for through the use of isotopically labeled internal standards (Xylazine-d6 and Medetomidine-¹³C-d3). The added value of automation is reflected not only in pre-cision and sample throughput, but also in terms of safety: The exposure of laboratory personnel to organic solvents and po-tentially infectious sample material is consistently minimized.

Four Applications: Much in Common, a few Differences

Veterinary pharmacology, food safety, infant nutrition, forensic toxicology, these are fields of application that could hardly be more diverse. What unites them is a structurally identical prob-lem: They rely on a proven extraction technique that is prone to error, when performed manually. The critical method parame-ters, including volume precision, mixing intensity, timing, tem-perature, and pressure, must be tightly controlled, precisely what automation achieves.

Three caveats, however, deserve explicit mention. First, auto-mation cannot transform a chemically unsuitable method into a good one: Matrix effects in the detection system do not vanish with increased dispensing precision. Those must be addressed separately through appropriate strategies such as isotope dilu-tion. Second, automation is not economically viable for every application: For very low sample amounts or for methods that are used infrequently, manual operation may remain the more advantageous option. Third, automation does not replace the expert’s assessment: Solvent selection, pH optimization, and extraction strategy, all of these require expertise in analytical chemistry. Automation translates these decisions into a repro-ducible and efficient process.

Was bleibt, ist ein Befund mit einer klaren Richtung. Verschärfte Grenzwerte wie der neue EFSA-TDI für BPA, wachsende Do-kumentationspflichten nach ISO/IEC 17025, GMP und GLP sowie zunehmend komplexere Matrices machen präzise, be-dienerunabhängige und validierbare Probenvorbereitungssys-teme nicht zu einer Option, sondern zu einer methodischen Notwendigkeit. Die automatisierte LLE, vor allem in der SALLE-Variante für polare Analyten, ist dafür methodisch gut positio-niert. Aus einem handwerklich geprägten Schritt wird eine defi-nierte, steuerbare, dokumentierbare Prozesskette. Das ist kein Selbstzweck, sondern schlicht moderne Analytik.

What remains is a conclusion that points in a clear direction. Stricter regulatory limits, such as the new EFSA TDI for BPA, and ever growing requirements for documentation under ISO/IEC 17025, GMP, and GLP, as well as increasingly com-plex matrices, mean that precise, operator-independent sample preparation systems that can be fully validated are no longer merely an option, but a necessity in the modern analytical la-boratory. Automated liquid-liquid extraction (LLE) systems are well-positioned to meet this challenge.

References

[1] Foster F et al. Automating Liquid-Liquid Extractions using a Bench-top Workstation. GERSTEL AppNote 177.

[2] Foster FD, Stuff JR. Automated Salting-out Assisted Liquid-Liquid Extraction and Determination of Bisphenol A in Beverage Samples. GERSTEL AppNote 230.

[3] Foster F et al. Automated Liquid-Liquid Extraction and Determi-nation of Fatty Acids Composition in Infant Formula according to AOAC® 2012.13. GERSTEL AppNote 267.

[4] Foster FD, Harper-Kerr M. Automated Liquid-Liquid Extraction and Determination of Xylazine and Medetomidine in Plasma and Urine Samples. GERSTEL AppNote 289.