Matrix effect evaluation using multi-component post-column infusion in untargeted hydrophilic interaction liquid chromatography-mass spectrometry plasma metabolomics

The goal of this study is to evaluate and address matrix effects (ME) in untargeted hydrophilic interaction liquid chromatography–mass spectrometry (HILIC-MS)-based plasma metabolomics. Matrix effects, caused by co-eluting compounds that affect ionization efficiency, can significantly impact the accuracy of metabolite detection. The study compares two ME evaluation methods—stable isotope labeled internal standards (SIL-IS) and post-column infusion (PCI)—and demonstrates PCI’s advantages in untargeted workflows.

By testing 18 SIL standards across different HILIC columns and mobile phase pH conditions, the researchers found the BEH-Z-HILIC column at pH 4 with 10 mM ammonium formate minimized matrix effects while maintaining strong analytical performance. The PCI method was further validated across 40 plasma samples, confirming its robustness and consistency with SIL-IS in measuring both absolute and relative matrix effects. Overall, the study supports PCI as a powerful tool for monitoring and minimizing matrix effects during method development and biological sample analysis.

The original article

Matrix effect evaluation using multi-component post-column infusion in untargeted hydrophilic interaction liquid chromatography-mass spectrometry plasma metabolomics

Mengle Zhu, Lieke Lamont, Pascal Maas, Amy C. Harms, Marian Beekman, P. Eline Slagboom, Anne-Charlotte Dubbelman, Thomas Hankemeier 

Journal of Chromatography A, Volume 1740, 11 January 2025, 465580

https://doi.org/10.1016/j.chroma.2024.465580

licensed under CC-BY 4.0

Selected sections from the article follow. Formats and hyperlinks were adapted from the original.

Polar metabolites play a crucial role in the immune response and inflammatory process and are key biomarkers in clinical research [13,14]. These metabolites, such as amino acids, organic acids and acylcarnitines, are for instance linked to rheumatoid arthritis [15] and oxidative stress in patients with COVID-19 [16]. Polar metabolites do not only have a diagnostic role but also a predictive capability for mortality [17]. Despite their biochemical importance, the analysis of polar metabolites by LC-MS is often challenged by poor retention on reversed-phase LC columns. Hydrophilic interaction liquid chromatography (HILIC) has emerged as a valuable chromatographic technique well-suited for polar metabolites, addressing this issue effectively. However, despite its potential advantages such as improved column retention and higher MS sensitivity because of increased ionization efficiency facilitated by a high proportion of organic mobile phase [18], HILIC has not been widely utilized by the scientific community. Common challenges associated with HILIC include complicated retention mechanisms, poor reproducibility, and low peak capacity [19]. The HILIC separation mechanism depends on the chemical properties of the stationary phase and mobile phase condition, including the type and concentration of the salt buffer, the pH, and the analyte structure [20]. Different HILIC columns and mobile phase conditions were systematically investigated in previous research by Contrepois et al. [11] and Hosseinkhani and coworkers [21]. Even though there is no consensus regarding the preferred HILIC stationary phase in the metabolomics field, the zwitterionic stationary phase seems to have better performance and wide coverage of polar metabolites in different biological matrices [11,[21], [22], [23], [24]].

Despite LC-MS technology offering advanced features, developing a robust method that can be validated for routine application is challenging. A major challenge is the occurrence of matrix effect (ME) during biological analysis [[25], [26], [27], [28]]. IUPAC defines ME as ‘the combined effects of all components of the sample other than the analyte on the measurement of the quantity. If a specific component can be identified as causing an effect, then this is referred to as interference’ [29]. A prevalent interpretation of ME hypothesizes that the presence of co-eluting matrix interferences can impact the ionization efficiency of target analytes and influence the signal intensity due to the competition for available charges and access to the droplet surface during the electrospray process [27,30]. ME can be observed either as a decrease in signal (ion suppression) or as an increase in signal (ion enhancement). Apart from ionization, matrix interferences may also affect the extraction efficiency during sample preparation [31]. The occurrence of ME can significantly impact the precision, accuracy, linearity and limits of quantification and detection of the analytical method, leading to distorted or false results. The traditional way of minimizing ME involves the removal of all interfering co-eluting compounds during sample preparation prior to chromatography and ionization. This approach is only viable if the analytical method focuses on a narrow range of analytes with similar physicochemical properties. However, in metabolomics, analytical methods often target a plethora of compounds with varying physicochemical properties, particularly in untargeted analyses. Another possibility to reduce ME is to separate interferences from target analytes during chromatography by parameter optimization, including the chosen stationary phase, the particle size of the column and mobile phase conditions [32]. For instance, Chambers et al. [33] observed an obvious reduction in ME under different mobile phase pH conditions. Therefore, it is essential to assess and minimize ME during method development. The stable isotope labelled-internal standard (SIL-IS) method is a common and compelling approach to assess the ME. In this method, SIL standards are added to the samples prior to sample preparation and LC-MS analysis to evaluate the ME influence from sample preparation and ionization. Although effective, this becomes a costly approach if the analytical method, in the field of metabolomics, screens hundreds of metabolites and is hampered by the lack of commercial standards for some (unknown) metabolites. Alternative techniques for ME assessment include the post-extraction spike method [26], slope ratio analysis [34], and post-column infusion (PCI) of standards method [35]. The post-extraction spike method investigates the ME by comparing the response of the analyte in a standard solution to its response in a matrix which does not contain this analyte and spiked with the analyte at the same concentration after sample preparation. The Slope Ratio Analysis uses a calibration line added to the blank and matrix and compares the slopes of both calibration lines. Those two approaches are more focused on the quantitative evaluation of ME during validation, are retention time-dependent and require a blank matrix to be included. The PCI approach uses (a mixture of) SIL standard(s) or other physicochemically related standard(s) which is (are) continuously infused after the chromatographic separation prior to ionization, which only can evaluate the ME from ionization. Although it requires an additional pump, PCI is the most suitable approach for qualitative ME assessment to facilitate the optimal performance of untargeted method development, due to its retention time-independent character.

We hypothesize that ME evaluation by PCI method can facilitate untargeted HILIC method optimization and provide valuable information during biological analysis. To investigate this hypothesis, we first compared the ME evaluated by SIL-IS and PCI methods to explore the capacity of the PCI approach for ME assessment and the contribution of sample preparation to ME. Next, the PCI method was used for a HILIC column comparison between HILIC-Z, ZIC-cHILIC and BEH-Z-HILIC columns. In addition, the chromatographic performance of three columns was compared by using 18 SIL standards from different polar metabolite classes as our targets. Based on our results, the optimal column was selected and the salt concentration and pH of the mobile phases were optimized. Subsequently, linearity, repeatability, recovery, ME, and precision were characterized for this method. Finally, our HILIC-MS method with PCI was employed to investigate ME differences of polar metabolites in plasma samples from healthy and metabolically compromised older adults.

2. Materials and methods

2.5. LC-MS analysis

LC-MS analysis was performed using a Waters Acquity UPLC H-Class system coupled with a SCIEX TripleTOF 5600 mass spectrometer with an electrospray ionization source (ESI) that operated at negative mode. The ESI source parameters were set as follows: spray voltage - 4.5 kV, capillary temperature 400 °C, ion source gas 1: 20 psi, ion source gas 2: 50 psi, curtain gas 25 psi. Data were acquired in full scan mode at the m/z range of 50–900 Dalton (Da). Mobile phase A consisted of 90 % ACN and 10 % Milli-Q water with 10 mM ammonium formate at pH 4. Mobile phase B consisted of 10 % ACN and 90 % 10 mM ammonium formate in Milli-Q water at pH 4. The gradient started at 100 % A and was kept for 1 min, then B linearly increased to 15 % over 2 min and to 21 % from 2 to 5 min, then to 26 % from 2 to 7.5 min, to 40 % from 7.5 to 10 min whereat it was held for 1 min, then returned to 100 % A in 0.5 min and equilibrated the column for 6.5 min, resulting in an 18 min run time per analysis. A 90 % ACN solution was used as the weak needle wash, while 10 % ACN served as the strong needle wash. The wash program was set to 12 s prior to each injection. The autosampler temperature was set at 10 °C and the oven temperature was maintained at 30 °C. The injection volume was 3 µL.

During the column comparison, the buffer condition was 5 mM ammonium formate at pH 7 and the flow rate was 0.5 mL/min. Three HILIC columns were investigated in this study; the InfinityLab Poroshell 120 HILIC-Z column (2.7 μm, 2.1 mm × 100 mm, Agilent), the Atlantis Premier BEH Z-HILIC column (1.7 μm, 2.1 mm × 100 mm, Waters) and the SeQuant® ZIC®-cHILIC column (3 μm, 2.1 mm × 100 mm, Merck). Detailed information of the three columns is included in Table S3. The following buffer conditions: 5 mM, 10 mM and 20 mM ammonium formate and pH 4, pH 7 and pH 8 were investigated after selecting the optimal column at a 0.5 mL/min flow rate.

For method characterization and LLS cohort measurement, the buffer condition was 10 mM ammonium formate at pH 4 and the flow rate was 0.3 mL/min. In this study, the acidic and alkaline pH of the aqueous ammonium formate were adjusted using formic acid and ammonium hydroxide, respectively.

3. Results and discussion

In this study, ME evaluation was investigated during the HILIC method optimization. We first evaluated the ME by comparing the peak area of SILs spiked in plasma and neat solution. This was compared with ME evaluation using four standards that were post-column infused during injection of plasma and neat solution. The PCI method was applied to further method optimization, characterization and biological application.

3.5. AME and RME evaluation in LLS cohort by PCI

To assess the applicability of this untargeted HILIC-MS method with PCI, it was applied to analyze 40 plasma samples for polar metabolites. These plasma samples were collected from patients with high MetaboHealth score (n = 20) or low MetaboHealth score (n = 20). One sample was excluded because of the low signal, leaving 39 samples for further analysis. Based on commercial authentic standards, 50 endogenous polar metabolites were detected and identified in these samples, with the RT and m/z information of measured compounds summarized in Table S10. The objective was to evaluate the AME and RME of those 50 compounds across 39 plasma samples using the PCI approach. The AME of endogenous compounds cannot be directly calculated due to absent signal in the neat solution samples. Hence, the median AME assessed by four PCI standards was used to represent the AME of each endogenous compound, and the average AME across all samples was calculated for each compound (Table S10 and Fig. 5). Twelve compounds, presented as green circles, exhibited AME between 80 % to 120 %, indicating acceptable ME across 39 plasma samples in Fig. 5A. Twenty compounds displayed slightly higher ion suppression, represented by yellow circles. The remaining compounds, shown as red circles, experienced significant ion suppression. The RME of 50 compounds was illustrated in Fig. 5B, plotting the m/z and RT of compounds. Four compounds (4-hydroxyproline, alanine, taurine and threonine) showed a slightly increased variation (RME > 10 %) over 39 plasma samples, also experiencing high ion suppression. The majority of compounds demonstrate RME below 10 %. Overall, those 39 plasma samples showed a comparable ME between samples, but they still exhibited relatively high ME.

Journal of Chromatography A, Volume 1740, 11 January 2025, 465580: Fig. 5. (A): AME of 50 endogenous compounds. Green circles mean those compounds have an acceptable AME; yellow circles mean those compounds experienced between 20 % and 30 % ion suppression in those plasma samples; Red circles mean those compounds exhibited above 30 % ion suppression. The two lines represent AME values of 80 % and 120 % .(B): RME of 50 endogenous compounds. Green rhombi mean their RME is equal to or below 10 %, and orange rhombi mean their RME is between 10 % and 15 %.

4. Conclusions

In this study, an untargeted HILIC-MS method was developed for polar metabolite analysis by applying a multi-component PCI approach for ME evaluation. This strategy facilitated column selection and optimization of mobile phase conditions and demonstrated that ME can be minimized by selecting the most suitable column and buffer condition, emphasizing the significance of ME evaluation in method development. The strong correlation observed between ME assessment using PCI and SIL-IS approaches provides compelling evidence for the efficacy of PCI in ME evaluation for untargeted analysis. Furthermore, a link to clinical metabolomics analysis was made through the LLS cohort which was used to evaluate the AME and RME of endogenous compounds and demonstrates an impressive capability to monitor ME during untargeted metabolomics analysis.

In conclusion, ME assessment provides valuable information during method development and application. Optimization of chromatographic parameters has shown significant improvements when guided by ME evaluation. The integration of the PCI approach with HILIC-MS provides an auspicious ability to evaluate ME during untargeted metabolomics analysis, paving the way for more accurate and reliable metabolite quantification in complex biological systems. Future analysis may benefit from the use of more chemically representative PCI standards to cover a broader range of targets. Moreover, this PCI approach could be used as a powerful tool to correct ME for accurate quantification of polar metabolites in clinical studies. Moving forward, its use could further advance clinical metabolomics, contributing to more precise biomarker discovery and a deeper understanding of metabolic pathways in health and disease.