3. 1 Define the optimal quenching and extraction procedure for quantitative metabolomics
Accurate determination of intracellular metabolite levels requires well-established procedures for sampling and sample pretreatment. The ideal procedure should immediately quench cellular metabolism and quantitatively extract all metabolites, while a significant challenge associated with the rapid turnover, varied abundance and physicochemical diversity of intracellular metabolites still requires systematic and quantitative evaluation. Otherwise, the obtained biased metabolite level will lead to misinterpretation of functional information about the biochemical and physiological states of cells [40]. In recent years, different quenching and extraction methods have been proposed and used for quantitative metabolomics of mammalian cells including tumor cells (Table 1 ). Early quenching studies on microorganisms established the use of cold methanol as a standard quenching protocol for its high efficiency of freezing enzyme activities [43-45], and cold methanol has also been reported frequently as the quenching agent in mammalian cell metabolomics [20, 24, 46, 47]. In addition to cold methanol, the use of cold normal saline has also become common and cold 0.9% (w/v) NaCl has been preferably used for quenching mammalian cell metabolism with minimal metabolite leakage [48]. Meanwhile, liquid nitrogen was also a commonly used quenching method in recent years which could rapidly and efficiently stop cell metabolism [18, 49-52]. After cell harvesting and quenching, metabolite extraction was the next critical step. Several methods existed for metabolite extraction, but the literatures were sometimes contradictory regarding the adequacy and performance of each technique [41]. To the best of our knowledge, the polarity of metabolites ranged from very hydrophilic to hydrophobic compounds, and polar metabolites were supposed to be extracted with ethanol, methanol, acetonitrile, or other mixtures of these solvents, etc [18, 24, 26, 46, 48, 51, 52], while more lipophilic metabolites could be extracted by chloroform/methanol/water mixture, etc [20, 47, 49]. At the same time, in order to ensure that the extraction of intracellular metabolites was as complete as possible, the extraction was usually combined with several freeze-thaw cycles [18].
However, no previous study has yet simultaneously optimized these critical procedures, i.e., quenching and extraction, to explore cell metabolomics using a strictly quantitative approach. Therefore, in this study, we systematically evaluated 12 combinations of three commonly used quenchers and four extractants using an IDMS approach and the specific experimental workflow were shown in Figure 1 . The prepared samples were analyzed using gas/liquid chromatography-IDMS method, and a total of 43 metabolites were quantitatively determined in Hela cells, including amino acids, sugar phosphates, organic acids, adenosine nucleotides and coenzymes (Figure 2 ).
As shown in Figure 1 , the medium was first removed and the cells attached to the flask surface were washed twice with PBS. The conventional method generally used trypsin to detach cells from their growth surface, which has been proven to inevitably change the profile of cellular metabolites since the detachment of cells from the extracellular matrix altered their physiology [18, 53-55]. In order to avoid changing the cellular environment before quenching metabolic activities within the cell, in this study, tumor cells cultivated in 2D mode were directly quenched and then scraped with the extraction solvent in all experiments. Meanwhile, 3D MTSs were first digested into single cell suspension, and the cell pellet was obtained after centrifugation. The subsequent operations were the same as those of the 2D cells. We estimated the cell number by determining the protein content of cell pellet. The standard curve of the protein was shown in Figure S1 . With respect to quenching efficiency, ideal quenching solvent should immediately inhibit cellular metabolic activity and inactivate enzymes within the cells without leaking metabolites from the cells. Quenching of metabolism was usually achieved by rapid changes in temperature or pH. The rate of quenching was critical because many metabolic reactions in glycolysis, as well as those related to ATP, featured high turnover rates in the timescale of seconds [45]. Numerous studies have reported the adverse effects of prolonged exposure to quenching solutions which demonstrated that the longer the cells were exposed to the quenching solution, the exacerbated leakage of intracellular metabolites [22, 45, 56]. This seemed reasonable that prolonged exposure to organic solvents and cold shock could cause cell membrane leakage, which also implied that sample processing should proceed as quickly as possible to the extraction step. In addition, quantitative methods should be used to determine the extent of cellular leakage so that true intracellular metabolite concentrations could be determined [45]. To achieve this, we quantitatively assessed the number and amount of leaked metabolites in the quenching agent, and investigated the factors that influenced the extent of leaked metabolites.
In order to determine the most effective extraction method, four different extraction methods were evaluated after Hela cells were quenched with liquid nitrogen, 50% methanol, and normal saline respectively. The evaluated extraction methods included 80% methanol, 50% acetonitrile, methanol: chloroform: water (1:1:1, v/v/v) and 70% ethanol (75℃). Sellick et al. found that the addition of a water-based extraction step combined the properties of solvent-based and water-based methods, and extracts produced in this method showed the greatest amount of recovered metabolites in CHO cell [26]. Therefore, we also adopted an additional water extraction to increase metabolite coverage for all 12 combination methods. The metabolites in the obtained cell extracts were absolutely quantified using GC-IDMS and LC-MS/MS-IDMS method. Subsequently, effectiveness of the quenching and extraction procedures should be quantitatively evaluated for their suitability for Hela carcinoma cells grown as 2D monolayer cultures and 3D MTS. For this purpose, previous studies have reported some evaluation criteria with respect to relative metabolite abundance and metabolite coverage. For example, LORENZ et al. established a method for relative quantification of metabolites in adherent mammalian cells using the clonal β-cell line INS-1 as a model sample, without considering absolute quantification of metabolites [49]. Dietmair et al. evaluated several quenching and extraction procedures for mammalian cells grown in suspension, but did not take into account the reproducibility of each method, with large errors in the metabolite level [48].
To allow an accurate view of intracellular metabolome, the quenching step should stop metabolism as quickly as possible while the extraction should cover a wide spectrum of metabolite. Furthermore, target metabolites should not undergo any physical or chemical modification and degradation should be minimized during quenching and extraction process [45]. In this study, we evaluated the compatibility of three different quenchers such as liquid nitrogen, -40°C 50% methanol, 0.5°C normal saline and four extractants (80% methanol, methanol: chloroform: water (1:1:1, v/v/v), 50% acetonitrile, 75°C 70% ethanol) for global metabolite profiling of adherent Hela carcinoma cells. To comprehensively evaluate 12 combinations of quenching and extraction procedures in this study, we analyzed the acquired metabolite data based on the following four criteria: 1) Metabolite leakage in the quenching solution; 2) Metabolite coverage and abundance identified in cell extracts; 3) Absolute quantitation of intracellular metabolite concentrations; 4) Reproducibility of metabolite quantification [57].
3.1.1 Metabolite leakage in the quenching solution
To study the effect of quenching solutions on the integrity of cell membrane, the potential metabolite leakage during quenching was evaluated by quantifying the metabolites in the quenching solution. The metabolites analyzed include compounds with different properties, such as amino acids, organic acids, sugar phosphates, adenosine nucleotides and coenzymes. Figure 3 showed the effect of these two quenchers (50% methanol and normal saline) on the recovery of different representative metabolites. Liquid nitrogen did not cause intracellular metabolites leak into the quencher because liquid nitrogen would volatilize directly. In both cases, all determined amino acids were found in the quenching solution. Compared with normal saline, cold methanol gives rise to more serious leakage of intracellular amino acids (Figure 3a ). Also, an obvious phenomenon showed that the more leakage comes with the higher amounts of intracellular amino acids, such as alanine (Ala), valine (Val), proline (Pro), lysine (Lys), aspartate (Asp), glutamate (Glu). In consistency with amino acids, organic acids, sugar phosphates and energy metabolism intermediates also exhibited a significant higher leakage when tumor cells were quenched with the cold methanol. Meanwhile, the results also indicated that intracellular sugar phosphates were more inclined to leak in response to cold methanol treatment (Figure 3b ). According to previous reports, the higher degree of leakage caused by cold methanol may be partly attributed to the following aspects. The first was the increase in cell membrane fluidity or the decrease in thickness during the quenching process [57]; The sudden increase in cell membrane permeability caused by ”cold shock” was also reported [58]; In addition, methanol was a small molecule which was easier to penetrate cells which may also be the reason for the leakage of intracellular metabolites [22]. Moreover, Canelas et al. found that the extent of metabolite leakage inS. cerevisiae was affected by the methanol content of the quencher, and reducing the methanol concentration resulted in the increase of intracellular metabolite leakage [44]. Therefore, the necessity of fine-tuning quenching strategy should be emphasized. In general, it could be concluded that methanol quenching was not suitable for accurate quantification of the metabolites in Hela cells. Normal saline was the better of the two quenchers, although it could not completely prevent leakage.
Furthermore, we investigated the factors such as the molecular weight which has been reported to affect the degree of metabolite leakage [56]. Figure 4 showed the relationship between the degree of metabolite leakage and molecular size which was represented by molecular weight. We found that metabolites with a smaller molecular weight were more seriously leaked than metabolites with a larger molecular weight when normal saline was used as the quencher. This was consistent with the findings by Canelas et al. observed in S. cerevisiae that the leakage of the metabolites into the quenching solution was mostly driven by diffusion [56] and the diffusion rate of smaller molecules was higher [44]. In contrast with this, when cold methanol was used as the quencher, the degree of metabolite leakage showed no clear relation with the molecular weight, and severe leakage was observed to be widespread in the detected metabolites.
3.1.2 Metabolite coverage and abundance identified in cell extracts
The comparison of quantitative whole-cell metabolites showed that the number and total amount of metabolites extracted under each condition were significantly different (Figures 5 and 6) . For example, Oxaloacetic acid (OAA) was below the detection limit with methanol/chloroform as the extractant, but detectable in all the other extractants. Most sample preparation procedures frequently used cold organic solvents to stop cellular metabolic activities. In this study, we also quenched cellular metabolism directly by adding liquid N2 on the flask plate, followed by the addition of extraction solvent. Our results showed relatively less metabolites were obtained with methanol as the quencher compared to cells quenched with normal saline or directly quenched with liquid N2. Kapoore et al. also found that methanol quenching treatments resulted in severe leakage of nearly all metabolite classes in metastatic breast cancer cell line MDA-MB-231 [29]. In addition to this, the result also showed that direct quenching with liquid N2produced the highest recovery of all metabolites compared with methanol and normal saline (Figure 6 ). Additionally, the overall effect of normal saline as the quenching agent was between liquid nitrogen and methanol. Consistently, one previous study also reported that quenching with 60% methanol (buffered or unbuffered) resulted in leakage of intracellular metabolites from the cells. Whereas, quenching with cold normal saline (0.9% [w/v] NaCl, 0.5℃) did not damage the cells and effectively prevented the conversion of ATP to adenosine diphosphate (ADP) and adenosine monophosphate (AMP), indicating metabolic arrest in CHO cells [48].
Furthermore, based on the number of metabolites extracted, acetonitrile shows the best extraction efficiency (Figure 5 ). In addition to the total number of extracted metabolites, the total amount of extracted metabolites was also an important indicator for evaluating the extraction methods. The results showed that using liquid nitrogen as the quencher and 50% acetonitrile as the extractant, the maximum total amount of intracellular metabolites could be harvested, about 295 nmol/million cells, which was 1.67 times to 13.73 times higher than other combination methods. In addition, it could be observed that this optimal method allowed a high extraction capacity for adenosine nucleotides and coenzymes (Figure 6 ). For the four different extractants, it can be inferred that the extraction efficiency of acetonitrile and ethanol was generally higher, while the extraction efficiency of methanol: chloroform: water (1:1:1, v/v/v) (M/C) was the worst regardless of the type of quenchers (Figure 6 ).