Automated Rapid Organic-Solvent-Free Metabolomics via Arrow–GC–MS: An Optimized Green Workflow for Human Plasma Profiling

Metabolomics is an emerging field that facilitates the comprehensive characterization of low molecular weight metabolites, which reflect the dynamic interplay between environmental exposures and endogenous biological processes [1]. As the downstream output of complex molecular networks, including the genome, transcriptome, proteome, and epigenome, the metabolome is regarded as the closest biochemical representation of phenotype [2]. Consequently, metabolomic signatures offer a sensitive and functionally relevant readout of physiological processes, pathophysiological alterations, and responses to therapeutic interventions [3]. Despite its demonstrated potential for biomarker discovery, precision diagnostics, and disease monitoring, the integration of metabolomics into routine clinical workflows remains limited [4]. Significant barriers include reliance on technically demanding analytical platforms, high operational costs, and complex sample preparation protocols that necessitate highly skilled personnel [5].

Among various biofluids, human plasma has attracted considerable attention in the field of metabolomics due to its molecular diversity, clinical accessibility, and capacity to reflect systemic metabolic processes [2]. Plasma contains a diverse array of biomolecules, including amino acids, organic acids, sugars, lipids, xenobiotics, and volatile organic compounds (VOCs) [[6], [7], [8]] , among others, which serve as indicators of metabolic alterations associated with aging [9], cancer progression [10,11], cardiometabolic disorders [12], neurological diseases [13,14], immune activation [15], and related systemic processes. Moreover, plasma is extensively employed in clinical diagnostics and biobanking [16], making it a suitable matrix for translational metabolomics [2]. However, traditional plasma metabolomics methodologies often rely on labour-intensive procedures such as solvent-based extraction, chemical derivatization, and the use of large injection volumes [4]. These approaches increase analysis time, generate chemical waste, reduce reproducibility, and limit implementation in routine diagnostics [5].

Gas chromatography coupled with mass spectrometry (GC–MS) represents a robust and high-throughput analytical platform that has historically been pivotal in metabolomics, particularly for the analysis of volatile and semi-volatile compounds [17]. GC–MS provides excellent chromatographic resolution, quantitative accuracy, spectral reproducibility, and compatibility with extensive public spectral libraries such as the National Institute of Standards and Technology Mass Spectral Library (NIST) and the Human Metabolome Database (HMDB) [18]. Nevertheless, its application in plasma metabolomics has traditionally relied on sample preparation steps that increase protocol complexity and environmental impact due to the use of hazardous solvents [4]. Furthermore, most GC–MS workflows utilize liquid injection [10,19,20], which necessitates large solvent volumes, contributes to system contamination, and diminishes instrument longevity.

To address these limitations, the principles of green analytical chemistry (GAC) advocate for the reduction of hazardous reagents, the simplification of workflows, and the development of solvent-free extraction strategies [21]. Headspace solid-phase microextraction (SPME) represents a powerful alternative that facilitates solvent-minimized extraction of volatile and semi-volatile compounds directly from biological samples. Recent advancements in SPME technology have led to the development of the PAL SPME Arrow system, which overcomes the limitations of traditional SPME fibers by increasing sorbent volume, enhancing extraction capacity, improving mechanical robustness, and enabling full automation and replicability [22]. Although the SPME Arrow technique has been successfully employed in environmental testing [23], food chemistry [[24], [25], [26], [27]], and forensic toxicology [23], its utilization in biological research and clinical metabolomics remains limited despite its potential to deliver rapid and highly informative metabolic profiles.

In this study, we introduce AROMA-GC-MS, an expedited, solvent-minimized, and automatable GC–MS workflow utilizing headspace PAL SPME Arrow for the metabolomic profiling of human plasma. This method necessitates minimal sample volume, circumvents derivatization, reduces chemical waste, and is compatible with routine laboratory automation. We conducted a systematic evaluation of key parameters affecting extraction efficiency and sensitivity, including sample dilution, extraction temperature, incubation time, and desorption conditions. Our findings demonstrate that headspace SPME Arrow, when coupled with GC–MS, facilitates the reliable detection of a wide array of clinically relevant metabolites, offering a practical and scalable solution for translational metabolomics. By aligning analytical performance with the principles of GAC and clinical feasibility, this workflow supports the integration of metabolomics into biomedical research and future diagnostic and prognostic applications.

2. Materials and methods

2.6. GC–MS-Based Metabolome Profiling

Metabolomic profiling of the samples was conducted using an Agilent Technologies 8890 gas chromatograph coupled with a 5977C mass selective detector (Santa Clara, USA), equipped with a PAL RSI 120 Series 2 autosampler. The chromatographic separation was performed on an Agilent J&W HP-5ms Ultra Inert capillary column (30 m × 0.25 mm i.d., 0.25 µm film; Agilent Technologies, Santa Clara, USA). The oven temperature program was set as follows: initial temperature of 50 °C held for 3 minutes, followed by a ramp of 5 °C per minute to 210 °C, maintained for 2 minutes, and then increased at 20 °C per minute to 325 °C. The injection was conducted in splitless mode, with helium serving as the carrier gas at a constant flow rate of 1.0 mL per minute. Post-run conditions were set at 325 °C for 3 minutes with a flow rate of 3.0 mL per minute. The transfer line, ion source, and quadrupole temperatures were maintained at 320, 230, and 150 °C, respectively. Electron impact ionization at 70 eV was employed. Full-scan data were acquired over a mass-to-charge ratio (m/z) range of 45–450, with normal scan speed and a threshold of 150 counts; the solvent delay was set to 0 minutes, and the total acquisition time was 42 minutes 45 seconds. All samples were analyzed in triplicate.

3. Results

3.2. AROMA-GC-MS optimization: incubation and desorption parameters

To define the optimal thermal and temporal conditions for HS-SPME extraction in the AROMA-GC-MS workflow, incubation temperature and incubation/desorption times were systematically evaluated. We first assessed the effect of incubation temperature on metabolome coverage and signal response. The number of detected compounds increased significantly with incubation temperature, with both 60 °C and 70 °C yielding higher compound counts than 50 °C (Figure 2A). In parallel, normalized signal intensity also increased with temperature, with statistically significant differences observed between 50 °C and the higher-temperature conditions (Figure 2B).

Green Analytical Chemistry, Volume 18, 2026, 100376: Figure 2. Optimization of incubation temperature, incubation time, and desorption time for the AROMA-GC-MS workflow. A. Average number of detected compounds as a function of incubation temperature (50, 60, and 70 °C). Bars represent mean compound counts across technical replicates, and error bars indicate the standard error of the mean (SEM). B. Normalized integrated signal intensity obtained at each incubation temperature. Bars represent mean values across technical replicates, with error bars indicating SEM. C. Correlation analysis between incubation temperature and the average number of detected compounds. Pearson correlation coefficient (r) and corresponding P-value are reported; shaded areas represent the 95% confidence interval. D. Venn diagram illustrating the overlap and uniqueness of compounds detected under different desorption–incubation time combinations. Numbers indicate absolute compound counts, with percentages referring to the proportion relative to the total number of detected compounds. E. Average number of detected compounds across combinations of desorption time (3 and 5 min) and incubation time (10 and 20 min). Bars represent mean compound counts across technical replicates, with error bars indicating SEM. Statistical significance was assessed using one-way ANOVA followed by post hoc testing; significance levels are indicated as *P ≤ 0.05 and **P ≤ 0.01.

The increase in signal response and compound coverage with temperature is consistent with improved release of volatile and semi-volatile metabolites from the plasma matrix into the headspace. Higher temperatures increase analyte vapor pressure and accelerate equilibration between the liquid phase, headspace, and the Arrow coating, thereby enhancing extraction efficiency.

Despite these increases, correlation analysis across the tested temperature range revealed a positive but non-significant association between incubation temperature and the number of detected compounds (Figure 2C), indicating that gains in coverage tend to plateau at higher temperatures. Consistent with this observation, no statistically significant differences were observed between the 60 °C and 70 °C conditions for either compound count or signal intensity. Based on these results, 60 °C was selected as the optimal incubation temperature, as it provides enhanced analytical sensitivity and metabolome coverage relative to 50 °C while avoiding diminishing returns and potentially increased thermal stress at 70 °C under PAL autosampler conditions.

Temporal parameters were subsequently optimized by evaluating four desorption-incubation combinations (3–10, 3–20, 5–10, and 5–20 min). The number of detected compounds remained comparable across all conditions (Figure 2D-E), with no significant pairwise differences, indicating that feature-level recovery is robust to variations in incubation and desorption time.

At the identification level, extending the incubation time to 20 min resulted in a substantial increase in annotated compounds, as reflected by the large fraction of species uniquely detected under the longer incubation condition (Figure 2D-E). The improvement observed with longer incubation time likely reflects more complete analyte equilibration during HS-SPME Arrow extraction. Extending the incubation period favors the transfer of metabolites with slower diffusion kinetics or lower volatility from the salted plasma matrix into the headspace and subsequently onto the Arrow coating. This suggests that prolonged incubation enhances analyte equilibration and headspace enrichment without compromising overall compound detection. Accordingly, incubation time was standardized at 20 min to maximize compound coverage while maintaining acceptable analytical throughput.

Comparisons between desorption times of 3 and 5 min revealed no significant differences in compound counts across incubation settings (Figure 2D–E). Increasing desorption time produced only limited additional gains, suggesting that most analytes were efficiently released under the tested inlet conditions. A desorption time of 5 min was therefore selected to provide a conservative margin for complete analyte release and to minimize potential carryover, while remaining compatible with routine autosampler operation.

3.4. Analytical-space expansion through Arrow-compatible derivatization in AROMA-GC-MS

The AROMA-GC-MS workflow was further evaluated to assess whether the inclusion of an Arrow-compatible MeOx–TMS derivatization step could provide complementary analytical coverage relative to the non-derivatized protocol. As shown in Figure 4A, the overlap between derivatized and underivatized analyses was minimal, with the majority of detected compounds being specific to each mode (66 compounds uniquely detected following derivatization and 115 compounds uniquely detected in the non-derivatized run). The list of identified compounds in both derivatized and non-derivatized experiments is provided in Supplementary Table 1. To further increase confidence in compound annotation, experimentally determined linear retention index (LRI) values were compared with literature- or spectral library-reported values for representative identified compounds (Supplementary Table 2). The curated metabolite list therefore represents putatively annotated compounds supported by complementary evidence, including spectral library matching, LRI agreement, and biological plausibility. This limited intersection indicates that derivatization enables access to a largely orthogonal chemical space rather than redundantly expanding the native AROMA-GC-MS coverage.

Green Analytical Chemistry, Volume 18, 2026, 100376: Figure 4. Analytical-space expansion through Arrow-compatible derivatization in the AROMA-GC-MS workflow. A. Venn diagram showing the overlap and uniqueness of compounds detected using the optimized non-derivatized AROMA-GC-MS protocol and the Arrow-compatible derivatization protocol. Numbers indicate compounds detected in at least 3 out of 5 technical replicates per condition. B. Normalized integrated signal intensity obtained under non-derivatized and derivatized conditions. Bars represent mean values across technical replicates, and error bars indicate the standard error of the mean (SEM). C. Average number of detected compounds for non-derivatized and derivatized analyses. Bars represent mean compound counts across technical replicates, with error bars indicating SEM. Statistical significance was assessed using one-way ANOVA followed by post hoc testing; significance levels are indicated as ***P ≤ 0.001. Circular plot depicting the hierarchical chemical class composition of compounds detected under D. derivatized and E. non-derivatized AROMA-GC-MS conditions. Inner rings represent primary molecular classes (color-coded), while outer segments correspond to subclass-level annotations across all technical replicates.

At the global analytical level, integrated signal intensity was comparable between derivatized and non-derivatized analyses (Figure 4B), indicating that the inclusion of derivatization does not compromise overall signal response. In contrast, the total number of detected compounds was significantly higher in the underivatized condition for this dataset (Figure 4C). Nevertheless, the derivatized workflow contributed a substantial set of unique compound identifications, highlighting its added value despite the lower overall compound count.

To further characterize the chemical nature of this complementarity, the hierarchical molecular class composition of compounds detected under derivatized and non-derivatized conditions was examined (Figure 4D–E). The derivatized workflow preferentially expanded coverage toward highly polar and low-volatility compound classes, including carbohydrates, polyols, organic acids, amino alcohols, and nitrogen- and sulfur-containing species, compound families that are classically challenging to access under solvent-free HS-SPME conditions (Figure 4D). In contrast, the non-derivatized AROMA-GC-MS analysis was enriched in more lipophilic and semi-volatile classes, such as hydrocarbons, fatty acids and esters, organosilicon compounds, and aromatic species (Figure 4E). These distinct class-level profiles provide a mechanistic basis for the limited overlap observed between the two analytical modes and further support the interpretation of derivatization as a source of orthogonal chemical information.

These findings demonstrate that Arrow-compatible derivatization functions as a complementary extension of the AROMA-GC-MS platform by expanding the accessible chemical space without altering the core analytical workflow. Nevertheless, within the validated pipeline, the non-derivatized AROMA-GC-MS protocol remains the default environmentally friendly and high-throughput approach.

3.7. Biological interpretation of the plasma metabolome captured by AROMA-GC-MS

To explore the biological relevance of the plasma metabolome captured by AROMA-GC-MS, the interpretation was focused on the curated set of putatively identified metabolites supported by spectral matching, LRI agreement, and biological or chemical plausibility. These LRI-supported annotations were interpreted in the context of metabolite families and biochemical processes previously reported in an evidence-based comparative study of GC–MS-detectable metabolites related to oxidative stress and neurological conditions [14].

5. Conclusions

AROMA-GC-MS establishes a solvent-free, fully automated, and environmentally sustainable workflow for plasma metabolomics that aligns with the principles of green analytical chemistry. The method achieves high analytical depth and reproducibility from minimal sample input while maintaining compatibility with clinical automation. By integrating HS-SPME Arrow extraction with optional MeOx–TMS derivatization, AROMA-GC-MS captures both volatile and semi-volatile metabolite fractions, broadening biochemical coverage beyond conventional liquid-injection approaches. The resulting plasma metabolome reflects physiologically relevant pathways linked to lipid metabolism, oxidative stress, and antioxidant defense, supporting the translational potential of this approach in biomedical and clinical research. Importantly, HS-SPME Arrow complements liquid-chromatography MS by emphasizing volatile and semi-volatile fractions, enabling multi-platform metabolomics strategies without expanding solvent or resource footprints.