高通量分析表型组的方法

生物谷按：在基因组，蛋白质组后面，还有许多“组学”，如代谢组，转录组，表型组等。尤其是表型组研究一直缺乏有效的段，传统的手段研究十分缓慢，这篇文章作者建立的新方法有效地用于研究代谢组学和表型组学，值得关注。

Introduction

The genomics revolution has produced massive datasets, and large-scale experiments for generating gene disruptions and analyzing phenotypes are underway to ascertain gene function. Such functional analyses are also referred to as phenomics, meaning any form of phenotypic analysis of genomic information or entire mutant collections with the goal of understanding the relationship between genes and higher levels of organization in the cell. Beyond qualitative or semi-quantitative phenotype profiling, the development and implementation of genome-wide analytical techniques has spawned a generation of `omics' enterprises with much of the present emphasis on the genome-wide mRNA, protein or organic metabolite complements of cells. The operational unit of function in complex biological systems, however, is more properly seen as fully assembled biochemical networks [1., 2. and 3.], as connectivity, interactions and dynamic properties such as kinetics and regulation are not defined by genome sequences or expression array information and cannot be inferred directly from measurement of the components ( Figure 1).

(8K)

Figure 1. Schematic overview of the connection between compositional and functional units in metabolic systems.

The functional determinants of cellular physiology are in vivo molecular fluxes through metabolic pathways, because they reflect the integration of genetic and metabolic regulation (Figure 1). Measurement of metabolism-wide fluxes, the fluxome [4.], thus allows us to observe the functional output of the compositional transcriptome, proteome and metabolome changes and addresses the missing link in contemporary functional analyses to the cellular phenotype. The broad connective nature has made it very difficult to study metabolism comprehensively, but isotopic tracer experiments in combination with material balances has enabled metabolism-wide quantification of in vivo molecular fluxes [3., 5. and 6.]. Until recently, fluxome analyses were labor- and instrumentation-intensive, and this review focuses on advances in tailoring ¹³C-based flux analysis for high-throughput phenotypic characterization in microbes. In contrast to the analysis of dynamic flux responses in metabolic subsystems [2.], the emphasis is on metabolism-wide characterization of microbes in (quasi) steady state. This appears to be most pertinent for large-scale systems response analyses of the integrated metabolic regulation in functional genomics, metabolic engineering, and the nascent field of systems biology.

Established experimental approaches

From its early days when material fluxes were balanced within assumed reaction networks [7.], metabolic network analysis [6. and 8.] has matured to actually identify the topology of active reactions and pathways and to quantify the molecular flux through them on a variety of substrates [9., 10., 11., 12., 13. and 14.]. The pivotal elements of this advance were elaborate ¹³C labeling experiments, new measurement techniques, and mathematical data evaluation methods. Recent discoveries include the unexpected activity of so-called futile cycles in several microbes [15., 16. and 17.] and the identification of a novel pathway for glucose catabolism in Escherichia coli [18.].

Three general approaches for the interpretation of ¹³C labeling patterns from isotopic tracer experiments can be distinguished: comparative, analytical and integrated (Figure 2). Comparative multivariate statistics may be used to discriminate strains or conditions (e.g. the presence of toxic chemicals) in massive `raw' datasets in a bioinformatics approach. To gain biochemical insight, however, further data interpretation is necessary. Direct analytical interpretation of a selected ¹³C labeling pattern in pathway intermediates or products has been used extensively for biochemical pathway elucidation [2. and 6.]. Generally, algebraic or probabilistic equations relate the ¹³C pattern to flux partitioning ratios [17. and 19.]. A particularly informative methodology is metabolic flux ratio analysis, which quantifies the relative contribution of multiple converging pathways and reactions to the formation of target metabolic intermediates from the ¹³C pattern of proteinogenic amino acids. Originally developed for nuclear magnetic resonance (NMR)-based analysis by Szyperski [20. and 21.] and extended by others [15.], this approach was recently modified to quantify 14 independent flux ratios from [1-¹³C] and [U-¹³C] glucose experiments by gas chromatography mass spectrometry (GC-MS) analysis of amino acids [22.]. In the integrated approach, all available ¹³C labeling data, extracellular material fluxes, and biomass composition are simultaneously interpreted with metabolic models of varying complexity [5.]. Briefly, the labeling state of metabolic intermediates is balanced within a model to map metabolic fluxes in an iterative fitting procedure on the isotopomer (isotope isomers) [5., 16. and 23.], summed fractional labels [24.] or bondomer (entities that vary only in numbers and position of intact C–C bonds) [25.] pattern of network metabolites. The in vivo flux distribution in different microorganisms has been quantified by this approach with ¹³C data from NMR analysis [15., 16., 17. and 26.], MS analysis [9., 27., 28., 29. and 30.] or both [31. and 32.].

(28K)

Figure 2. Schematic flow chart for the interpretation of mass isotopomer and physiological data by bioinformatics, analytical metabolic flux ratio analysis, iterative isotopomer balancing, and ¹³C-constrained flux analysis. The latter is a mixture of analytical mass flux ratio analysis and integrated flux balancing of physiological data within the flux ratio constraints. Example results that may be obtained by the different approaches are depicted in the lower panels. The size of the arrows in the absolute flux panel is proportional to the molecular flux.

High-throughput ¹³C pattern analysis

Although NMR and MS analyses were both used successfully, only the latter has the potential for high-throughput analysis at the microscale because of its extraordinary sensitivity, speed and comparatively low cost [33.]. Thus, it has been shown that on the order of 0.5–1 mg of dry biomass –– which can be generated easily in microcultivation systems –– suffice to detect mass isotopomer distributions of proteinogenic amino acids by GC-MS [22. and 34.]. After sample pretreatment, GC-MS analysis of a culture sample takes approximately 5–20 min; setting an upper bound of about 300 analyses per day and per instrument. Speed and sensitivity were improved further with matrix-assisted laser desorption ionization - time of flight (MALDI-TOF) MS for quantification of the ¹³C labeling pattern in three accumulated metabolites from 1 l of culture supernatant [35.]. Only a few microbes will secrete those particular compounds, but many extracellular metabolites are accessible to MS for metabolome footprinting [36.] and would potentially also be available for ¹³C labeling pattern analysis from tracer experiments [37.]. The concentration and ¹³C pattern of intracellular metabolites are valuable to resolve dynamic flux responses [2., 38. and 39.], but their inherent high turnover times and low concentrations are a technical challenge. To achieve high-throughput, steady-state flux analysis based on proteinogenic amino acids or extracellular metabolites appears to be more promising.

Generally, cell separation is necessary, but protein-based protocols also require high temperature hydrolysis to liberate the amino acids. Independent of the analytes, pretreatment for GC-MS analysis involves drying and derivatization, whereas samples only require dilution for MALDI-TOF MS [35.]. As pretreatment is amenable to parallel processing, however, it is unlikely to constitute a serious limitation to throughput. The choice of analytes for high-throughput MS analysis will therefore depend on factors such as flux observability and methodological robustness, including evaporation, chemical stability and formation/consumption during the cultivation. At least for protein-based protocols, the harvesting phase is not critical [4. and 22.], but extracellular metabolites are typically not produced in a constant manner and might even be consumed upon depletion of the primary substrate. It should be noted that the ability to observe a particular flux is also a function of the administered ¹³C-labeled tracers that are not restricted to the frequently used glucose [9., 10., 11., 12., 13. and 14.].

High-throughput fluxome mapping

Before fluxome mapping, the measured mass isotopomer data must be corrected for natural stable isotopes, and an elegant method and software for automated correction, statistical data treatment and error recognition of MS data was recently described [40.]. With the availability of MS data at an appropriate throughput, the next issue is the flux resolution to be achieved, unless simple mutant or condition discrimination by multivariate statistics is sufficient ( Figure 2). Metabolic flux partitioning ratios are almost instantaneously available from the processed MS data without the need for physiological data and with computation times below a second [22.]. Thus, these ratios are readily available for high-throughput fluxome mapping, but do not reveal absolute flux values. As an important difference to integrated flux fitting, analytically determined flux ratios bear high biochemical veracity because the evidence for a particular flux change is direct [17.]. High-resolution mapping of absolute fluxes by iterative isotopomer balancing dominates rigorous analysis of metabolically engineered strains [28., 29. and 30.], and computational expenses for iterative fitting procedures may be greatly reduced through explicit solution of non-linear isotopomer balance equations [41.], isotopomer path tracing [42.] or efficient statistical analysis [43.]. Although the computation time per single simulation run was reduced to about a second, many runs are still necessary to obtain flux solution and sensitivity analysis for one dataset (strain).

To determine absolute fluxes at high-throughput, the necessity for high-quality physiological data is probably more critical than computation time, as such data are difficult and expensive to gather at high-throughput. Notably, absolute fluxes were also quantified from a subset of physiological data and isotopomer balancing of only selected MS data in shake flask [28. and 35.] and microtiter plate cultures [44.] (see also Update). A mathematically much simpler approach for the quantification of absolute fluxes with negligible computational effort implements analytically determined flux ratios as constraints for flux balancing [45.]. Such comparatively simple and rapid ¹³C-constrained flux balancing (Figure 2) allowed the quantification of fluxes from selected physiological data and GC-MS-derived flux ratios in shake flask cultures [27.] (see also Update). To acquire physiological data from parallel microscale cultures, extracellular metabolite concentration changes can be assessed either with standard enzymatic or chromatographic methods or with sensitive MALDI-TOF MS [46.]. Importantly, MALDI-TOF methods can simultaneously quantify several low molecular mass compounds in a short time and their applicability may be further extended by the derivatization of analytes [37.].

High-throughput cultivation systems

Effectively, the above analytical and computational advances have shifted the limitation of high-throughput flux analysis to suitable microscale cultivation systems. For large-scale approaches, fluxes are best studied in physiological (quasi) steady state using cultivation systems that are amenable for high-throughput. Most accurate, but not applicable, are continuous bioreactor cultures that were used extensively for rigorous analysis of tracer experiments. Technically much simpler, a (quasi) steady state can also be achieved during maximum exponential growth in batch culture. At a first level of parallel analysis, shake flask batch cultures were used for flux analysis [27., 28. and 35.], with the presently highest number of 68 flux profiles obtained in a single study [22.].

Yet another order of magnitude is necessary for high-throughput applications. Although the sensitivity of MS-based analyses is sufficient to monitor the ¹³C pattern in either secreted metabolites [35. and 44.] or proteinogenic amino acids [22. and 34.] (see also Update) of microtiter cultures, the physiological characteristics in such cultivation devices must be quantitatively reproducible, robust and as comparable as possible to those at larger scale. The latter point is critical, because otherwise mass transfer problems, in particular oxygen supply, may limit growth and lead to uncharacteristic behavior. Exceptional oxygen mass transfer and cross-contamination free operation can be achieved in silicone-covered deep-well microtiter plates with 1 ml cultures [47.] (see also Update). For routine application, however, robotic instrumentation is better achieved in standard microtiter plates with 200–400 l culture volumina, which also appear to be suitable [44., 48. and 49.] provided cross-contamination-free operation is ensured.

Conclusions and outlook

The methods and tools for high-throughput flux analysis of entire mutant collections or sets of environmental conditions (e.g. the presence of toxic chemicals) are seemingly all in place. For functional genomics, it may suffice to first discriminate mutants from massive mass isotopomer and physiological datasets by multivariate statistics. The available data for interesting outliers or discriminated groups are then biochemically interpreted by simple, rapid and robust methods such as metabolic flux ratio analysis [22.] or ¹³C-constrained analysis of absolute fluxes [27. and 35.]. Direct analytical flux analysis of all perturbed cases would be more appropriate for quantitative characterization of metabolic network responses or screening for interesting flux phenotypes in systems biology or metabolic engineering.

The next years will witness large-scale profiling of fluxes through central carbon metabolism in microbes for functional genomics, metabolic engineering and systems biology. The quantitative functional data on systems behavior that emerges from the structural and regulatory interactions of network components in thousands of genetically or environmentally perturbed cases will create an enormous opportunity for systems-oriented research by supplanting the traditional approach of investigating only the system components (e.g. genes or proteins). Scientifically, the anticipated results will undoubtedly provide new insights into metabolic network behavior and global regulation patterns to foster understanding of the biochemical complexity of entire cells. Upon extension to suitable mammalian models, such as cultured cell lines [50., 51. and 52.], high-throughput flux profiling also has potential for pharmaceutical research through the functional characterization of diseased metabolic states that will allow the screening of drugs and toxins.

Update

Recent work demonstrated analytical robustness of ¹³C-constrained flux balancing at different cultivation scales [53.]. In particular, it was shown that, upon proper batch culture handling, metabolic fluxes are directly comparable between aerobic bioreactors and 1 ml deep-well microtiter plates.

Original article:

Related articles:

M.K. Hellerstein, In vivo measurement of fluxes through metabolic pathways: the missing link in functional genomics and pharmaceutical research. Annu Rev Nutr 23 (2003), pp. 379–402.  Abstract-MEDLINE   

E. Fischer and U. Sauer, A novel metabolic cycle catalyzes glucose oxidation and anaplerosis in hungry Escherichia coli. J Biol Chem 278 (2003), pp. 46446–46451. Full Text 

E. Fischer and U. Sauer, Metabolic flux profiling of Escherichia coli mutants in central carbon metabolism by GC-MS. Eur J Biochem 270 (2003), pp. 880–891.  Abstract-MEDLINE