Biomedical Engineering Reference
In-Depth Information
often be distinguished from healthy tissue on the basis of molecular markers,
even in the absence of obvious histopathological abnormalities. The most com-
prehensive of such markers is the entire spectrum of expressed proteins.
In 2D protein electrophoresis, two principal approaches are pursued to
compare expression states between different experimental conditions. The first
is to carry out two electrophoretic protein separations, one for each of the de-
sired conditions. The resulting spot patterns then need to be matched across gels,
a formidable task given the limited reproducibility of individual gels (26). In a
second approach, proteins are chemically labeled before separation to distin-
guish proteins that are expressed under different experimental conditions (27).
(The very same approach is also used in direct mass-spectrometric identification
of proteins in complex mixtures.) One commonly used labeling technique is that
of in vivo labeling, where one population of cells is grown in a standard me-
dium, and the other population in a medium enriched in a stable isotope such as
heavy nitrogen (
15
N). The two cell populations are mixed before protein extrac-
tion. Such differential labeling results in slightly different masses of the same
gene product in two cell samples, which can be used to quantify to what extent
its expression has changed. One of the disadvantages of this technique is that
tissue samples, e.g., as obtained through biopsies, cannot be differentially la-
beled. In situations like these, the technique of post-extraction labeling is useful,
as exemplified by isotope-coded affinity tagging (ITAG). Here, proteins ex-
tracted from two cell types are chemically modified through the addition of par-
ticular chemical moieties such as alkyl groups to specific amino acids such as
cysteine. The modifying agent has a different isotope composition for the two
samples. After pooling the modified proteins, they can again be analyzed jointly,
because their shifted masses distinguish them (1,27).
The data resulting from these approaches is complex and represents relative
abundances of the thousands of proteins in a cell. Fortunately, the problem of
analyzing data of similar complexity has arisen earlier in large-scale measure-
ments of mRNA gene expression through microarrays. Thus, a plethora of com-
putational tools are available to analyze such data (3,5,8,29).
I will briefly comment on the identification of protein networks through the
analysis of large-scale mRNA expression data. Most large-scale analysis of
mRNA gene expression has goals similar to that of quantifying protein expres-
sion, namely, to provide hints about interacting gene products that form part of a
network controlling a biological process. The key difference is that mRNA is
only an intermediate in the process of gene expression. From this shortcoming
arise a number of disadvantages to reconstructing protein networks from mRNA
expression (15). For example, the expression of many proteins is translationally
(and not transcriptionally) regulated. For such proteins, mRNA gene expression
is a poor indicator of protein abundance. Second, and similarly, differential pro-
tein degradation is an important factor in determining protein concentration. It
cannot be captured by mRNA gene expression analysis. Finally, the activity of
