Sample Preparation in Global Metabolomics of Biological Fluids and Tissues - Proteomic and Metabolomic Approaches to Biomarker Discovery - page 57

were also found to be ef cient for protein removal

(98%) but resulted in much lower metabolite
coverage versus organic solvent extraction. 54
Although the precipitation with acetonitrile or
acetone may perform better in terms of protein
removal, 52,53 precipitation with methanol,
ethanol, or methanol/ethanol mixture is consis-
tently reported to result in better metabolite
coverage and method precision. 53,54 Conse-
quently, precipitation with methanol and meth-
anol/ethanol dominates the

of overlap between different precipitation
methods (white boxes, pairwise comparison).
Acetone/methanol mixtures and methanol
resulted in the highest metabolite coverage. In
an early study, Bruce et al. also compared the
performance of protein precipitation using
various aqueous solutions of methanol or aceto-
nitrile for the analysis of human plasma and
found the best performance with methanol. 56
However, the overall sample procedure was
very long and demanding and incorporated
a two-hour standing period and an ultra

field, with greater
than 50% of studies reported to date carried out
using this solvent, while precipitation with aceto-
nitrile was used in about 25% of studies to date. 9
For ef

ltration
step as well as an evaporation/reconstitution
step. Such a long, multiple-step sample prepara-
tion procedure can be expected to result in signif-
icant loss of metabolites, and in fact only 516 to
735 features were observed, which is quite low
in comparison to other reported methods. The
same group further re

cient protein removal, minimum plasma
to precipitant ratio of 1 to 2.5 (v/v) was recom-
mended, 52 although most current studies tend to
employ plasma to precipitation ratios of 1 to 4. 53
Clearly, none of the precipitation methods
provides complete protein removal, with esti-
mates of 2 to 10% of proteins remaining in

ned their methodology
by testing four different solvents (methanol,
ethanol, acetonitrile, and acetone) and combina-
tions of solvents in a more comprehensive
study. 53 The best method performance for
precipitation of human plasma samples was
obtained using methanol/ethanol (1/1, v/v)
and methanol/acetonitrile/acetone (1/1/1, v/
v/v) using a 1-to-4 plasma to precipitant ratio.
In terms of analytical variability, median rela-
tive standard deviation (RSD) values of 15 to
20% in combination with LC-MS seem to be
a reasonable target and can be achieved with
the majority of methods tested. Considering the
profound effects of ionization suppression on
LC-MS results, there is surprisingly very limited
evaluation of matrix effects in the context of
LC-MS metabolomics, and no in-depth studies
have been reported to date to compare the
performance of various sample preparation
methods with respect to this important param-
eter. In general, organic solvent precipitation is
favored over acid or metal salt precipitation
due to less ionization suppression, but post e
column infusion experiments indicate large
regions of suppression for the commonly used
solvents, 52

nal
extract depending on the solvent and precipitant
ratios selected. 53 This result contributes to short-
ened column lifetimes for metabolomic studies,
and protein buildup may affect the accuracy and
quality of metabolomic results. 41,42

Selection of Extraction Solvent: Metabolite
Coverage and Method Reproducibility
The previous discussion shows that there is an
increased interest in recent years to research the
effect of different sample preparation methods
on global metabolomic data, but there is still
no clear consensus regarding the optimum
method(s) to employ for a given application
although recently published detailed protocols
represent an invaluable starting point for the
best current practices. 44
In a comprehensive comparison of 14 extrac-
tion methods tested, the precipitation with meth-
anol was found to perform the best in terms of
metabolite coverage and method reproduc-
ibility. 54 The results of this in-depth study are
summarized in Figure 4 as a function of total
metabolite coverage (gray boxes) and the degree

indicating that this parameter is far