Westermeier R., Naven T., H?pker H.-R. Proteomics in Practice: A Guide to Successful Experimental Design

Подождите немного. Документ загружается.

2 Critical Points 9

Fig. 6: Graph showing the dependence of overall recovery versus number of steps.

Note: The way of sample acquisition and

sample preparation prior to the separation

needs to be optimized for each sample type.

This optimization work can take up to several

months.

The concentrations of proteins are not evenly distributed within a

sample. It is very difficult for all separation methods and for mass

spectrometry to detect and analyze very low concentrated proteins in

the presence of highly abundant proteins. While regulatory proteins

and transcription factors are expressed in very low copy numbers

(a few proteins per cell), house keeping proteins exist in some million

copies per cell: e.g. b-actin in 15 million molecules per cell (Kis-

lauskis et al. 1997). The protein concentrations in human plasma

even span over ten orders of magnitude (Anderson and Anderson

2002).

Often the complex protein samples have limited stability. Because

many proteins exhibit biological activities, some of them start to mod-

ify other proteins in the sample mixture immediately during and

after taking the sample. Some enzymes, like proteases are even active

under “denaturing” conditions. The measures for sample stabiliza-

tion need to be optimized for different sample types. Only when the

sample acquisition and sample preparation is under control, it is

worth to invest time and money in a major experiment.

In tissue samples protein

concentrations span a dynamic

range of six orders of magni-

tudes.

Kislauskis EH, Zhu X, Singer

RH. Cell Biol 136 (1997)

1263–1270.

Anderson NL, Anderson NG.

Mol Cell Proteomics 1 (2002)

845–867.

Enzyme inhibitors can only

partly stabilize the sample.

Introduction10

Many parameters influence the composition of a proteome. In a

proteomics experiment it is critical to distinguish between induced

biological changes and inherent biological variations. Those are, for

instance, genetic differences, gender and age of patients, slightly dif-

ferent cell or plant growth conditions. Therefore it is important to

always analyze several biological replicates. The required number of

replicates can be different for different sample types. Some guide-

lines can be found in the paper by Hunt et al. 2005.

The expression level differences of target proteins and biomarkers

associated with induced biological changes are mostly subtle, as

shown for instance in the paper by Moumen et al. (2005). On and off

effects are very rare. Therefore it is very important to be able to detect

and prove the significance of the experimental effect over the back-

ground of bio-variability of the sample source and the “noise” of the

analytical tool with high statistical confidence. Also this fact greatly

influences the required sample size (number of replicates).

Membrane proteins are very difficult to solubilize, highly hydro-

phobic, and they can get easily lost during sample preparation and

separation by sticking to a surface or by aggregation.

Post-translational modifications like phosphorylation and glycosyla-

tion require sophisticated analysis tools like MS

(MS/MS/MS),

where the peptide ions get several times fragmented. At present

about 350 different ways of post-translational modifications are pub-

lished (see Delta mass database: http://www.abrf.org/index.cfm/

dm.home). It is very difficult to keep the overview and have all tools

available for their identification and characterization.

The presently available detection methods are not sufficiently sensi-

tive. It is desirable to reach down to a LOD of one protein expressed

in a cell. Should the protein detection method be sensitive enough,

then there is still another limitation of detection in mass spectrome-

try for protein identification.

When the proteome analysis starts on the peptide level, as it is

nowadays still the case for the MudPIT type of workflow, any type of

correlation to the parent proteins such as information on post-transla-

tional modifications and different protein isoforms can easily get lost

or is difficult to trace.

Automation as much as possible or available is desirable, in order

to keep the results independent of differences of human skills and to

increase the sample throughput.

Statisticians would demand at

least five replicates, in many

cases three replicates can

already deliver highly confident

results. In clinical proteomics

the numbers of required

patients are much higher.

Hunt SMN, Thomas MR,

Sebastian LT, Pedersen SK,

Harcourt RL, Sloane AJ,

Wilkins MR. J Proteome Res. 4

(2005) 809–819.

Moumen A, Masterson P,

O’Connor MJ, Jackson SP. Cell

123 (2005) 1065–1078.

It is often necessary to apply

special methods, which are

suited for membrane protein

analysis, like for instance blue

native electrophoresis.

. . .not to mention the many

different possibilities of glycosy-

lation types.

Alternative identification strate-

gies like, for instance, western

blotting need to be further

developed.

2 Critical Points 11

2.2

Challenges of the Analysis Systems

Depending on their methodological focus proteomics researchers

have different points of view on the analytical steps:

Mass spectrometry experts call every technology

used prior to mass spectrometry “Sample pre-

paration”; this includes also electrophoresis,

chromatography, and protein arrays.

Separation technologists sometimes call mass

spectrometry “Detection”.

There are inherent variabilities in each analysis system, from pre-frac-

tionation and separation to detection and mass spectrometry. The

major sources of variations are: gel-to-gel variations in 2-D electro-

phoresis, protein losses inherent with each analysis step, retention

time drifts and shifts in liquid chromatography, and ionization differ-

ences in mass spectrometry. Therefore of each biological replicate

several analytical replicates are required.

Most of the techniques employed in proteomics are quite demand-

ing for hands-on skills, or they are highly sophisticated to operate.

Good operator skills and knowledge are at least as important as the

performance of the equipment. Education programs and tutorials

have become an indispensable part of the various proteomics socie-

ties, which have been established almost everywhere around the

world.

Because of the points described above, there are standards needed

for quality control of the results. Development of standards for pro-

teomics is on the way. There is the International Proteome Standardi-

zation Initiative (PSI), and the relevant scientific journals regularly

review their instructions to the authors.

For the top-down approach of analysis of intact proteins with high

resolution mass spectrometry, the resolving power of the sample pre-

paration procedure should be as high as possible. Polyacrylamide gel

electrophoresis would be the preferred method of choice. However, it

is very difficult to elute intact proteins from the polyacrylamide gel

matrix.

Quantitative proteomics is based on relative quantifications. Abso-

lute quantification would be desirable, but is still not possible.

Univariate statistics like Student’s t-test and ANOVA (Analysis of

variance between groups) need to be applied to analyze whether the

detected biological event is significant or not.

The application of multivariate statistical tools like principal com-

ponent analysis (PCA), hierarchical clustering, and K-means cluster-

This should be kept in mind,

when methodical issues are

discussed.

Statisticians would demand at

least five replicates, in many

cases three replicates can

already deliver highly confident

results.

Also this book is intended as a

contribution.

Even with very sophisticated

equipment and highly devel-

oped analysis concepts, errors

like mixing up samples or

wrong annotation cannot be

completely avoided.

Introduction12

ing on large experimental data sets help to discover errors and deliver

very important information:

They can reveal experimental errors like wrong

sample annotation, for instance by detecting out-

liers.

They find proteins with similar expression pro-

files, prognostic or diagnostic markers, and clas-

sifiers without bias of the operator.

Proteomics Strategies

In short the following proteomics strategy is pursued: At first target

proteins are identified, characterized and correlated with biochemical

pathways. This part is often called “Classical Proteomics” or “Expres-

sion Proteomics”. Once some structural information is known, smal-

ler subsets of proteins are analyzed with milder separation and mea-

suring techniques: for instance, some proteins are fished out of a cell

lysate with affinity chromatography and then proteins with intact ter-

tiary structure, or protein–protein complexes are studied. This is

called “Functional Proteomics”.

3.1

Proteome Mapping

As already mentioned above, non-hypothesis-driven proteomics

experiments do not deliver meaningful results. In contrast to a gen-

ome a proteome is never static. Therefore a snapshot of a protein

expression status without any biological context cannot contain valu-

able information. However, proteome maps are necessary for estab-

lishing databases for protein identification and characterization.

3.2

Differential Analysis

Differential analysis between wild type and mutant, diseased and

healthy material, treated and control, is a powerful approach to detect

significant changes in protein expression levels. In practice it is very

rare that a protein is either present or completely absent. In most

cases only partly up- and down-regulations of certain proteins are

observed. These quantitative changes should be determined with pre-

cision and confidence.

When samples are pre-labeled with spectrally distinct fluorescent

dyes, mixed, and separated together under completely identical condi-

See for instance Swiss-prot

(now Uni-prot) and TrEMBL.

And it is very helpful to have

software tools available which

are dedicated to differential

display analysis.

In such a case the protein

detection is carried out with a

multifluorescence imager at

different wavelengths.