Evaluation of GC-APCI/MS and GC-FID As a Complementary Platform

Abstract

With a development of the metabolomics field, complementary cross-platform approaches started to attract attention, as none of the contemporary analytical methods had the capacity to cover the entire space of the human metabolome. In the current manuscript, we have evaluated an online coupling of gas chromatography (GC)-mass spectrometry (MS) and flame ionization detector (FID) as ways of cross-detector analysis. The possible value of this combination was recognized from the very first days of GC-MS history but was never explored in detail. We have compared the basic analytical parameters of both detectors, such as limit of detection (LOD) and limit of quantification, with intra- and interday reproducibility. We show that for the majority of the tested compounds, MS detector demonstrates lower LOD. At the same time, FID appeared to be more robust, showing lower relative standard deviations (RSDs) for intra- and interday reproducibility. We conclude that the gain of this dual detector acquisition appears to be most evident for complex biological samples, where wide dynamic range and predictable response of FID are useful for an initial quantitative overview of sample composition and estimation of molar proportions of different metabolites. MS provides reliable, structural information and superior, at least in the case of atmospheric pressure chemical ionization, sensitivity. Taken together, both detectors represent a flexible tool for explorative studies and if supported by a powerful data-processing algorithm, would appear to be useful in any metabolic profiling study.

INTRODUCTION

The concept of metabolomics as a global overview of an organism's metabolites emerged in the early 1970s,^1,2 and from the beginning, it was evident that its technical realization demands complementary methodologies, as none of the contemporary analytical methods had the capacity to cover the entire space of the human metabolome. Despite the booming development of this field, 40 years later, we are still facing the same problem.

A classical way to expand the set of analyzable compounds [coming from different chemical families and with different molecular weights (MWs)] and their linear dynamic range is the use of multidimensional chromatography. In routine analysis, multidimensional chromatography is usually divided in heart-cut and comprehensive³ approaches. The latter one is of primary importance for metabolic profiling, as it implies the separation of the complete sample in both chromatographic dimensions. A combination of two or more stationary phases leads to the exhaustive fractionation of the sample and eventually helps to use the dynamic range of the detector, usually a mass spectrometer, more efficiently. Multidimensional comprehensive chromatography is a relatively time-consuming method, and the depth of the coverage is proportional to the analysis time. When expensive, high-end mass spectrometry (MS) instruments are used as detectors, this can be a disadvantage. In addition, multidimensional comprehensive chromatography is, in practice, a single detector method, and regardless of the fractionation quality, the physical properties of the detector are the factors that “constrain” the analysis. In the case of MS, those factors are the ionization technique and the type of mass analyzer.

An alternative way to extend the coverage of the metabolome is a cross-platform analysis. The essence of cross-platform analysis is the acquisition of data using instruments with entirely different physics of detectors, followed by postacquisition data fusion. A classical example is a combination of NMR spectroscopy and MS. NMR is, without reservation, the most powerful method of modern metabolomics.^4–6 With regard to metabolic profiling of body fluids, ¹H-NMR, which provides a fingerprint of proton-containing species, has become a routine tool. Moreover, NMR is probably the only analytical method where quantitative and structural information is “embedded” in the physics of the detector. Still, a coanalysis of NMR with MS data expands the analytical domains of both methods and opens new ways of data interpertation.⁷ For example, it has been shown that application of correlation-based algorithms, such as statistical heterospectroscopy, can be a useful way to dissect and structurally resolve xenobiotic metabolites in human body fluids.⁸

The combination of NMR and MS is undoubtedly a powerful platform, but it requires costly instrumentation (NMR) and specialized expertise. In the current manuscript, we propose a way of cross-detector analysis, which can be practiced in almost any analytical laboratory: a combination of gas chromatography-mass spectrometry (GC-MS) and flame ionization detector (FID). GC is, without doubt, one of the most important and widely applied techniques in modern analytical chemistry, and FID has often been the first choice for GC routine analysis of complex biological samples. Indeed, since its introduction in 1958,⁷ FID has quickly become the most popular detector for GC and not without a reason: FID responds practically to all organic compounds, it is resistant to small fluctuations of the gas flow, it is insensitive to gas impurities, and most importantly, the FID response is very predictable, obeying the rule of equal carbon response.⁹ The additional value of a combination of MS and FID has been recognized already in the early days of GC-MS history.^10,11 Recently, the usefulness of FID as an auxiliary detector in NMR and/or MS metabolic profiling studies has been demonstrated once more.¹² However, in practice, the combination of GC-MS and GC-FID has never developed into a routine analytical method, and there are a number of explanations for this. One of these is that for routine analytical applications, the benefits were not clearly obvious, and at the same time, technical issues with the synchronization of FID and vacuum-stage MS were considerable.^11,12 With the reintroduction of atmospheric pressure (AP) sources for GC-MS,^13,14 the technical problems preventing optimal use of a combination of MS and FID have been largely overcome. In the current study, we demonstrate not only the simplicity of the coupling but also the possible complementarities of both detectors, in particular, for explorative studies about complex biological samples.

MATERIALS AND METHODS

A standard solution of 17 aa at 1 mM each in 1 M HCl was purchased from Sigma-Aldrich (Zwijndrecht, Nederland). Dopamine hydrochloride and phenyl-glycine (Phe-Gly) hydrate were from Fluka (UK). Sarcosine, theophylline, caffeine, nortriptyline hydrochloride, hippuric acid, 4-O-methyldopamine hydrochloride, benzoic acid, uric acid, and 5-hydroxyindole-3-acetic acid were acquired from Sigma-Aldrich. Stock standard solutions of the 28 compounds were prepared in methanol at a concentration of 200 μM. N-methyl-N-trimethylsilyltrifluoroacetamide plus 1% trimethylchlorosilane (MSTFA+1% TMS) were used as derivatization reagents (Pierce, Rockford, IL, USA). These reagents were used from freshly opened 1 ml bottles. Methoxyamine hydrochloride was purchased from Supelco (Park Bellefonte, PA, USA). Methanol (HPLC grade) and acetonitrile (GC grade) were acquired from Sigma-Aldrich, and pyridine (>99%, ultra-pure GC grade) was from Fluka. 4-Chlorophenol, decylamine, methyl caprate, prometryn, aldrin, and endosulfan α were purchased from Sigma-Aldrich. Stock solutions of these analytes were prepared in acetonitrile at a concentration of 10 mM. Reserpine was purchased from Sigma-Aldrich.

Sample Preparation

Aliquots of the standard mix were evaporated and derivatized by adding 100 μl methoxyamine (60 min, 40°C) and subsequently, 50 μl MSTFA (30 min, 40°C). Cerebrospinal fluid (CSF) samples were taken by lumbar puncture. The study was approved by the Ethical Committee of the Leiden University Medical Center (The Netherlands). Samples were processed within 1 h, centrifuged at 300 g to remove cells, aliquoted, and stored at –80°C until use. The CSF samples were prepared as described previously¹⁵: the proteins were precipitated with cold methanol, and then the supernatant was evaporated under a gentle stream of nitrogen. The derivatization was carried out with methoxyamine and MSTFA + 1% TMCS.

GC-MS-FID Analysis

The samples (1 μl) were applied by splitless injection with a programmable Agilent CTC PAL multipurpose sampler (Agilent, Atlanta, GA, USA) into an Agilent 7890A GC (Agilent, Palo Alto, CA, USA). The chromatograph was equipped with a single HP-5-MS column (30 m, 0.25 mm ID, 0.25 μm film thickness) and a column flow splitter with deactivated capillary (0.6 m, 0.25 mm ID, 0.25 μm film thickness), which allow a simultaneous acquisition with MS and FID detectors. For sample injection, a septumless programmed temperature vaporizing (PTV) injector was used. For each analysis, the purge time of the PTV injector was set to 60 s at a purge flow rate of 20 ml/min and an equilibration time of 1 min. Helium was used as carrier gas at a constant flow rate of 1 ml/min through the column.

For the underivatized compounds, the PTV worked with the following temperature program: 60°C held for 0.1 min, and then the temperature was increased to 250°C at a rate of 500°C/min. The column temperature was first kept at 40°C for 0.1 min and then raised to 280°C at 30°C/min and held for 10 min. For the biological samples, the following temperature program was used: 90°C held for 0.1 min, followed by the temperature increase to 250°C at 500°C/min. The column temperature was first kept at 70°C for 5 min and then raised to 280°C at 5°C/min and held for 10 min.

The FID was operated at 300°C; hydrogen (H) and air flows were set at 40 ml/min and 370 ml/min, respectively.

The transfer line to the mass spectrometer was kept at 300°C. In the AP chemical ionization (APCI) source operating in positive mode, temperature and flow rate of the dry gas (nitrogen) were 300°C and 2.00 l/min, respectively. The APCI vaporizer temperature was 300°C; the pressure of the nebulizer gas (nitrogen) was set to 3.5 Bar, and the voltage of the corona discharge needle was +4000 nA. Capillary voltage was set to –2000 V and the end-plate offset to –500 V.

The maXis (Bruker Daltonics, Bremen, Germany), an orthogonal-accelerated TOF-MS, was used as a MS detector. The polarity of the APCI interface and all of the parameters of the TOF-MS detector were optimized with infusion of reserpine. Spectra were acquired in scan mode from mass-to-charge ratio (m/z) 100 to m/z 1000 with 1 Hertz (Hz) frequency.

External Calibration Was Performed Using the Tune Mix

The linear dynamic range for the mix of six underivatized compounds was estimated for a concentration range from 0.05 to 10,000 μM. For the standard mix, a concentration range, 0.5–50 μM was used.

The two calibration lines (APCI-TOF and FID) were used for calculation of limit of detection (LOD) and limit of quantification (LOQ; Tables 1 and 2 in Supplementary Data); these were calculated on the basis of a signal-to-noise ratio of 3 for LOD and 10 for LOQ. The background noise was estimated from the peak baseline close to the analyte peak. Each calibration point is the result of three independent injections.

TABLE 1.

Analytical parameters for underivatized compounds

	Calibration curve	r2	Linear dynamic range (μM)	LOD (μM)	LOQ (μM)	Retention time (min)
4-Chlorophenol	y = 583,54× – 54192	0,9951	50–5000	40,0	133,3	5,27
	y = 0,1639× – 14,011	0,9934	10–5000	8,33	27,78	5,27
Decylamine	y = 5317,7× + 96980	0,9933	10–2500	4,29	14,29	5,51
	y = 0,2578× – 27,218	0,9972	25–5000	12,00	40,0	5,50
Methyl caprate	y = 11382× + 1E + 06	0,9932	0,5–5000	0,38	1,25	5,84
	y = 0,1682× + 12,262	0,996	0,5–10000	0,59	1,95	5,83
Prometryn	y = 55477× + 36030	0,9941	0,05–500	0,06	0,20	8,58
	y = 0,2207× – 1,2977	0,9972	1–1000	0,85	2,82	8,56
Aldrin	y = 279,97× – 3776,1	0,9974	5–1000	3,75	12,50	9,07
	y = 0,4799× – 7,9728	0,9960	1–1000	1,58	5,26	9,05
Endosulfan α	y = 42543× – 97395	0,9907	0,05–100	0,03	0,09	10,19
	y = 0,2981× – 4,2042	0,9982	5–1000	1,64	5,48	10,16

The gray lines are referred to as GC-APCI-TOF MS detector; the white lines are referred to FID detector. r², Linear regression.

TABLE 2.

Standard mix analytical parameters

Compounds	Retention time (min)	Calibration curve	r2	LOD (μM)	LOQ (μM)	Repeatability intraday	Repeatability interday (24 h)	Repeatability interday (48 h)
Alanine + 2 TMS + H	12,39	y = 13490× – 45721	0,9686	0,19	0,63	7%	18%	23%
	12,40	y = 0,2716× + 2,6521	0,9756	0,24	0,79	5%	14%	14%
Sarcosine + 2 TMS + H	13,41	y = 6145,7× – 32450	0,9665	3,00	6,47	17%	18%	26%
	13,41	y = 0,1097× + 0,0803	0,9537	8,00	26,88	6%	7%	11%
Proline + 1 TMS + H	14,53	y = 5473,2× – 59848	0,9943	21,58	71,94	16%	16%	21%
	14,55	y = 0,2041× – 0,5898	0,9969	18,29	60,98	3%	6%	8%
Valine +2 TMS + H	15,83	y = 21344× – 73007	0,9199	0,10	0,33	9%	20%	35%
	15,83	y = 0,2403× + 0,0935	0,945	0,78	2,60	1%	3%	5%
Benzoic acid + 1 TMS + H	16,51	y = 895,29× – 5812,6	0,9998	10,00	33,33	18%	18%	18%
	16,50	y = 0,5802× – 1,3997	0,9887	6,66	22,20	2%	6%	5%
Ile + 2 TMS + H	17,42	y = 23962× – 80202	0,9911	0,34	1,11	19%	24%	25%
	17,44	y = 0,3432× – 1,0109	0,9928	5,00	16,60	2%	3%	6%
Leu + 2 TMS + H	18,02	y = 30845× – 120980	0,8989	0,13	0,42	26%	17%	16%
	18,03	y = 0,6134× – 2,2667	0,9736	1,71	5,71	2%	4%	10%
Glycine + 3 TMS + H	18,34	y = 3430,7× + 2601,9	0,9747	0,06	0,20	15%	39%	37%
	18,34	y = 0,3445× + 1,1661	0,9997	2,70	9,90	14%	30%	27%
Serine + 3 TMS + H	19,92	y = 34064× – 87840	0,9373	0,07	0,22	12%	13%	19%
	19,92	y = 0,4867× – 1,18	0,9279	0,67	2,22	3%	3%	9%
Threonine + 3 TMS + H	20,62	y = 33259× – 58986	0,9442	0,05	0,16	13%	11%	20%
	20,63	y = 0,5272× – 1,0441	0,9241	1,55	5,15	2%	4%	7%
Methionine + 2 TMS + H	23,75	y = 44225× – 113892	0,9622	0,12	0,40	6%	17%	30%
	23,76	y = 0,4499× + 1,2307	0,9839	3,33	11,11	5%	6%	12%
Aspartic acid + 3 TMS + H	23,92	y = 36461× – 15137	0,9972	0,07	0,24	13%	24%	28%
	23,91	y = 0,5083× – 0,7629	0,9784	3,00	10,00	4%	6%	19%
Glutamic acid + 3 TMS + H	26,20	y = 29834× – 49222	0,9957	0,12	0,38	12%	21%	32%
	26,19	y = 0,4277× – 0,5318	0,9933	3,50	11,68	9%	11%	12%
Phenylalanine + 2 TMS + H	26,26	y = 49218× – 25330	0,9974	0,13	0,44	16%	18%	20%
	26,25	y = 0,7029× – 1,1104	0,9924	3,50	11,68	4%	5%	14%
Phenyl-Glycine + H	28,06	y = 119260× – 165927	0,9972	0,06	0,18	13%	22%	34%
	28,04	y = 0,8171× – 0,7237	0,9956	0,35	1,18	5%	5%	8%
Hyppuric acid + 1 TMS + H	30,66	y = 11673× – 26449	0,9918	0,26	0,86	16%	15%	16%
	30,64	y = 0,3526× – 0,5933	0,9835	4,77	15,92	6%	13%	11%
Caffeine + H	30,78	y = 114348× – 170144	0,9962	0,23	0,77	5%	10%	9%
	30,76	y = 0,3152× + 0,0298	0,9934	1,66	5,55	2%	6%	7%
Theophylline + 1 TMS + H	32,12	y = 45565× – 62580	0,9952	0,33	1,11	6%	16%	26%
	32,13	y = 0,2126× – 0,3059	0,9875	2,43	8,09	2%	37%	45%
Lysine + 4 TMS + H	32,53	y = 56140× – 14749	0,9948	0,11	0,36	20%	27%	35%
	32,53	y = 0,2126× – 0,3059	0,9875	0,75	2,50	7%	11%	11%
Tyrosine + 3 TMS + H	32,86	y = 75274× – 46887	0,9963	0,09	0,83	12%	21%	23%
	32,86	y = 0,6914× – 0,0326	0,9974	0,59	1,95	6%	8%	6%
4-Methyldopamine + 3 TMS	34,40	y = 91126× + 94865	0,9816	0,03	0,10	10%	16%	18%
	34,38	y = 0,8979× – 0,4778	0,9992	0,70	2,34	2%	6%	5%
Dopamine + 4 TMS + H	35,53	y = 87513× + 30795	0,9786	0,05	0,18	13%	14%	21%
	35,50	y = 1,0494× – 1,0955	0,9945	1,20	4,00	1%	4%	4%
Uric acid + 4 TMS + H	36,15	y = 29816× – 35130	0,9984	0,14	0,45	12%	28%	38%
	36,15	y = 0,2464× + 0,1531	0,9999	0,94	3,13	6%	10%	34%
5-Hydroxyindole-3-acetic + 3 TMS + H	37,90	y = 67994× – 66186	0,995	0,08	0,25	11%	22%	18%
	37,91	y = 0,4126× – 0,1815	0,9981	0,58	1,92	4%	5%	13%
Nortriptyline + H	38,32	y = 8649,4× + 5316,9	0,995	0,57	1,89	10%	16%	25%
	38,26	y = 0,0882× + 0,2164	0,9907	3,33	11,11	8%	27%	39%

The gray lines are referred to GC-APCI-TOF MS; the white lines are referred to FID. Ile, Isoleucine; Leu, leucine; Si, silicon.

The intra- and interday repeatability test was performed using standard mix as a reference. Results of the test were expressed as relative sd (RSD). Four independent injections were carried out at 25 μM concentration level. The interday repeatability was calculated after 24 and 48 h.

The accurate mass data of the molecular ions were processed with DataAnalysis 4.0 software (Bruker Daltonics). A list of possible elemental formulas was generated by using the SmartFormula module of the DataAnalysis 4.0. It uses a CHNO algorithm, which provides standard functionalities, such as minimum/maximum elemental range, electron configuration, and ring-plus double bond equivalents, as well as a sophisticated comparison of the theoretical with the measured isotope pattern (σ value).

RESULTS

To illustrate the complementarity of APCI-MS and FID, we used a mix of six organic compounds: 4-chlorophenol, decylamine, methyl caprate, prometryn, aldrin, and endosulfan α. The selection of these compounds is not random; it includes aliphatic and cyclic structures, with and without chlorine or sulfur. A comparison of the FID chromatogram and the base peak chromatogram (BPC) shows (Fig. 1) that the response of the MS detector is more compound-dependent than FID. For example, the peaks at 5.3 and 9.2 min (namely, 4-chlorophenol and aldrin) are quite intense in FID but are nearly absent in the BPC generated by the MS detector. Table 1 summarizes the analytical parameters for both detectors. Calibration lines, LODs, and LOQs were calculated, plotting peak areas as a function of concentration [extracted ion chromatograms (EICs) were used for calculations of peak areas in MS]. According to the correlation coefficients, each detector demonstrates a high degree of linearity (r²>0.991 in all cases). LOD and LOQ are significantly lower for GC-APCI/TOF-MS than GC-FID, except for 4-chlorophenol and aldrin. On the other hand, the linear range of detection is wider in the case of FID. It is important to mention that the retention times of a given compound are almost identical for both detectors (FID and APCI-MS). Thus, using this relatively simple example, we have shown that there is substantial differences in response factors and basic analytical parameters (LOD and LOQ) between FID and APCI-MS detectors.

FIGURE 1.

An example of comparison of the chromatograms recorded with GC-APCI-TOF MS (upper) and FID (lower) for six underivatized compounds (100 μM each): 1, 4-chlorophenol; 2, decylamine; 3, methyl caprate; 4, prometryn; 5, aldrin; 6, endosulfan α.

A combination of six organic compounds, even when selected specifically with regard to chemical diversity, does not reflect the complexity of biological samples. Besides, a large proportion of body fluid metabolites is nonvolatile polar compounds and needs a derivatization before GC analysis. Therefore, as a next step, we analyzed a standard mix of compounds typically reported as components of body fluids. This mix included compounds belonging to different chemical families: amines, amino acids, organic acids, alcohols, and xanthines. All chemical species were selected with the specific aim to cover a wide range of polarities and MWs, mimicking as closely as possible a real-life situation. Fig. 2 shows typical chromatograms of FID and MS detectors. Table 2 summarizes analytical data for compounds of the mix (only the most stable silylation form of a compound is shown). In agreement with previous data (Table 1), retention times for both detectors were found to be identical, and the MS detector showed a lower LOD (<3 μM for MS and <8 μM for FID), except for proline and benzoic acid (Table 2). The maximum intraday deviations were 9% for FID and 26% for MS; the maximum interday deviations (after 48 h) were 39% for FID and 38% for MS. The RSD after 48 h for both detectors indicates degradation in the samples once they had been derivatized. The linearity of response was found to be equally good for FID and MS (r²>0.95). Even when the correlation factor was not ideal (r²<0.95), as in the case of valine, serine, and threonine, this deviation affects both detectors, which indicates a problem of derivatization, a thorny step in the analytical procedure, rather than instability of the detector.

FIGURE 2.

Representative chromatograms of the standard mix (25 μM): APCI-TOF-MS (BPC; upper) and FID (lower). Zoom-in: Ile (17.6 min) and Leu (18.2 min).

The detailed analysis of the chromatograms presented in Fig. 2 shows how differently both detectors respond to certain compounds despite their superficial similarity. For example, Ile and Leu are “classical” positional isomers, which have the same atomic composition, number of single and double bonds, and consequently, the same masses. The peak areas observed for these compounds give a good approximation of the relative detector response. Calculating the peak area ratios for FID and MS, we obtain the following numbers: Leu FID/Ile FID = 0.52; Leu MS/Ile MS = 0.69. Thus, there is a significant difference of the ratios in the Leu/Ile peak areas between FID and MS detectors. The effect is also reflected in the angular coefficients of the calibration lines: Ile has a higher angular coefficient than Leu (Table 2). This increase was found to be larger for FID (+44%) than for APCI-TOF (+22%). Analysis of the compound mass spectra (Fig. 3) provides a possible explanation of this effect. Leu and Ile are represented by doubly silylated forms. In GC-APCI MS, the parent ion is the dominant feature of the spectra, unlike in conventional GC-electronic impact (EI) spectra, where the parent is a minor peak or even absent.¹⁵ Main fragments, for Leu as well as for Ile, are the result of a loss of neutral groups such as Si(CH₃)₃OH (m/z 90) or Si(CH₃)₃OH + CO (m/z 118), leading to formation of m/z 186.1 and m/z 158.1 fragments, respectively. However, Ile has an additional intense fragment at m/z 260.1. This fragment is the result of a loss of the methyl group, most probably at position 5; the remaining tertiary carbon at position 3 stabilizes the positive charge with formation of the m/z 260.1 fragment. An additional loss of Si(CH₃)₃OH (m/z 90) leads to the formation of a fragment with m/z 170.1 (Fig. 3).

FIGURE 3.

APCI-TOF spectra for Leu (upper) and Ile (lower).

Next, we used human CSF to test the applicability of our approach for biological material. Fig. 4A shows typical BPC and FID chromatograms. With both detectors, very complex chromatograms were recorded. A number of molecular features extracted from MS data and overview of compounds detected in human CSF with GC-APCI-MS were reported elsewhere.¹⁵ At first glance, there is no serious dissimilarity between the chromatograms. However, as the representative examples (Fig. 4B and C) show, there is a number of differences between FID and APCI-MS. Fig. 4B shows a few abundant compounds that saturate the MS detector. It is a common situation for biological fluids with their enormous concentration range. It is evident that realistic quantitative estimation based on MS data is impossible. FID, on the contrary, is still far from its saturation point within its linear range and therefore, can provide accurate quantitative information. Fig. 4C illustrates a different situation: in the case of coeluting or poorly resolved peaks, MS plays an essential role in the interpretation of the results.

FIGURE 4.

(A) Representative chromatograms of CSF sample: APCI-TOF-MS (BPC; upper); FID chromatogram (lower). (B) Zoom-in showing examples of peaks saturating the MS detector but remaining within linear dynamic range of FID. (C) Zoom-in showing an example of EIC used to resolve coeluting compounds.

DISCUSSION

GC-MC at AP has a peculiar history. The first experiments with corona discharges have shown the potential of ionization processes at AP.¹⁶ The first instrumental designs of GC-APCI/MS were described in the early 1970s, but for a number of reasons, vacuum-stage instruments took over the market. For decades, GC-APCI/MS remained an exotic application.

The combination of GC-MS and GC-FID also is an old idea, with roots going back to the 1960s.^10,11 Horning² demonstrated the feasibility of the combination GC-MS/GC-FID in one of his first reports about APCI-GC. His source design is regularly quoted in reviews about APCI MS, but his idea to use MS and FID as parallel detectors remained unnoticed.^17,18 One of the common arguments against the routine use of the combination of FID and MS is the lack of complementarity.¹⁹ It has been postulated, for example, that MS and FID yield a similar chromatogram, and response factors for the majority of organic compounds are comparable.¹⁹ This statement, however, is based on the comparison of FID signals, which were obtained with classical GC-MS ionization techniques: EI and CI. As we have shown here, the difference between APCI-MS and FID detectors becomes evident even with a simple mix of six organic compounds. Thus, the statement of equal response factors for FID and MS would appear not to be applicable to GC-APCI-TOF/MS. However, the difference in response factors and basic analytical parameters (LOD and LOQ) is not entirely surprising, considering the different nature of the physics of both detectors. Besides, the important question is not whether APCI-MS and FID produce different responses but whether those differences are complementary and whether the combination of both detectors might thus be beneficial for analysis of complex biological samples.

Our next example, a mix representing the most common body fluid metabolites (standard mix), is an attempt to address this question. Our “standard mix” includes compounds from several chemical families: amines, amino acids, organic acids, alcohols, and xanthines. It is certainly a much more complex sample than the previous one and at a first glance, much more diverse. However, the derivatization procedure introduces a common bias for all metabolites. In addition, there are not so many heteroatoms, such as, for example, chloride. A simple visual inspection of the trace of the chromatogram is not sufficient for the comparison of detectors, but the analysis of the basic analytical parameters reveals a few interesting features. For example, the MS detector consistently demonstrates lower LOD and LOQ values, and the intraday repeatability is significantly better for FID (RSD between 1% and 9%, as compared with 6–26% for MS). Moreover, a comparison of the data acquired 24 h and 48 h after derivatization indicates that there is a common trend for both detectors toward higher RSDs, which is, most probably, a side-effect of derivatization instability. Considering that the effects of derivatization on the analytical performance of a given method can be circumvented by, for example, an in-line derivatization approach and that the MS detector shows better LOD/LOQ, one of the possible preliminary conclusions could be that the combination of FID and APCI-MS as parallel detectors is indeed of little practical use. However, our analysis of the response of two detectors to such classical positional isomers as Leu and Ile proves the opposite, or at least, makes the last statement questionable. The difference in Leu/Ile ratios observed with the two detectors is the result of ”sensitivities” of the detector to different properties of compounds: although FID responds mainly to the total carbon number, the response of MS is affected by the structural differences between the analytes. The tertiary carbon of Ile affects the fragment formation during the in-source/in-funnel fragmentation process, which leads to the differences in compound spectra and eventually changes the response of the detector. On the other hand, FID is also not an ideal “carbon counter”, and its response can be affected by the presence of such heteroatoms as chlorine or sulfur. Nevertheless, there are not that many physiologically relevant metabolites containing a large number of chlorine atoms, and if there are any, their presence is easily detected by unique isotopic patterns.

Our next example was human CSF, which plays a key role in the mechanical and immunological protection of the brain, the maintenance of its homeostasis, and metabolism and as such, is an important source of analytical information. We have used it to outline areas where a parallel data collection by two detectors (FID and APCI-MS) could be useful for the interpretation of GC results. It would thus appear that the combination of detectors has at least two possible advantages. Firstly, the wide dynamic range of FID can be useful for the exact quantitative estimation of compounds present in biological samples at concentrations close to saturation of the MS detector, and those quantitative results are free of any ionization and instrumental bias. Secondly, the TOF mass analyzer provides high-quality structural information, which is essential for the data interpretation in metabolomics studies. Moreover, the advantages of APCI mass spectra for structural identification of compounds become more evident if we compare them with spectra of traditional GC-MS sources, such as EI and CI. The two latter ionization techniques produce spectra, where the quasi-molecular ion is low-abundant or even absent. This can be a serious disadvantage for the analysis of biological material, where a significant part of compounds is usually unknown. With an APCI source, the quasi-molecular ion ([M+H]⁺) is a dominant one, which opens the possibility for structural characterization of unknown compounds using the combination of accurate mass and isotopic pattern. In addition, an ionization process resulting in the formation of a stable, quasi-molecular ion and the technical characteristics of modern TOF instruments makes it possible to use such acquisition modes as MS2 or multiple reaction monitoring, which are common in LC-MS but seldom used in GC-MS.

Thus, the power of MS as a tool for the structural elucidation of metabolites generally needs no additional confirmation, but quantification remains “the weakest link”. The quantitative response of a mass spectrometer cannot be attributed to a single physical phenomenon: it is the product of hyphenation, ionization technique, efficiency of ion optics, and the type of mass analyzer. In other words, MS is a selective detector. In routine work, dealing with the measurements of known compounds, this selectivity is not a problem: the response of the MS detector can be perfectly linear for every single compound. However, in explorative studies, where the relative abundance of compounds within the complete metabolite profile is as important as the absolute quantity, the mass spectrometer may introduce an undesired bias. On the contrary, the quantitative response of the FID detector is free from ionization bias and those biases introduced by the type of mass analyzer or the instrumental design of a mass spectrometer. Consequently, the FID gives a better overall quantitative representation of complex biological samples. As a result, a parallel detection for GC-FID, GC-APCI, might offer a quick way to obtain simultaneously the quantitative and structural information in metabolic profiling studies.

There are, of course, a few technical issues needed to be resolved to turn parallel data acquisition with GC-FID/GC-MS into an efficient method. Those are: the different acquisition rates (few Hz for MS and routine 50 Hz for FID), the lack of a proper algorithm for data fusion, the development of an optimal strategy for treatment of overlapping peaks, and so on. However, the main purpose of this manuscript was to draw attention to the elegant but forgotten idea of the simultaneous use of FID and MS, rather than resolve all of the technical problems of this combination. With the reintroduction of APCI sources for GC in their modern, user-friendly form, the parallel acquisition of FID and MS chromatograms has become a simple instrumental option that requires no additional technical modification of the chromatograph. The gain of this dual detector acquisition would appear to be most evident for the analysis of complex biological samples. FID is the only detector capable of handling concentration ranges typical for body fluids. Its wide dynamic range and predictable response are useful for an initial quantitative overview of sample composition and the estimation of molar proportions of the different metabolites. MS provides reliable, structural information and superior, at least in the case of APCI ionization, sensitivity. Taken together, the use of both detectors represents a flexible tool for explorative studies and if supported by a powerful data-processing algorithm, would appear to be useful in any metabolic profiling study.