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

Подождите немного. Документ загружается.

1 Electrophoretic Techniques130

The development of software for 2-D electrophoresis gel image analy-

sis is a continuously ongoing process. The functions become more

reliable, reproducible and automated from year to year. With the lat-

est developed program it is already possible to compare gels of differ-

ent sizes, shapes, and even damaged gels. However, irreproducible

results cannot be fixed, even not by the most sophisticated software.

On the contrary: the highly developed programs recognize inconsis-

tencies between different patterns. Exact and tidy laboratory work is

still requested. The software cannot turn bad separation results into

good results.

There are two types of 2-D imaging software on the market:

Cheap to free of charge tools for basic image

analysis;

Professional programs, which evaluate with high

reproducibility and offer many valuable func-

tions for image analysis, statistical evaluation,

and reporting.

& The most critical parts in image analysis are spot

detection and background subtraction. Manual

spot editing and background definitions are

always user-biased.

Furthermore it is very important that the image analysis program

does not modify the raw data.

There are two different concepts of 2-D electrophoresis, which

require different approaches for image analysis and evaluation:

One sample per gel separations;

Multiplex separations like achieved with DIGE.

1.6.2.1 Traditional One Sample Per Gel Analysis

The following brief description is based on the concept of one of the

most advanced software packages, which is available under the name

ImageMaster Platinum, also known under the previous name “Mela-

nie”. The goal is to apply automatic procedures wherever possible, to

exclude manual editing and human interference.

Spot detection The most important step is spot detection, where dif-

ferent parameters have to be optimized:

With a “smooth” factor splitting of spots can be

influenced.

The selection of a certain “minimal area” elimi-

nates spots below the set threshold.

This can vary with different

sample types, pH gradients, gel

size and composition, detection

techniques.

1.6 Image Analysis 131

The definition of “saliency” discriminates

between real spots and artifacts (real spots have

higher saliency values).

These parameters are defined in a selected enlarged restricted area

and then applied on all gels of an experiment. Properly optimized

spot detection reduces spot editing close to zero.

Background correction A background subtraction function can easily

lead to errors. After many years of experience with spot volume quan-

tification it has been found that by defining the spot boundary at 75%

of the peak maximum and calculating the spot volume for only for

the values above the boundary, the relative spot volume quantification

comes very close to the real situation and ensures a very high repro-

ducibility.

Normalization Normalization takes care of the variation in protein

loading and staining. The spot volumes are normalized against the

total spot volume of the gel fully automatically without any user inter-

action.

pI and M

calibration There are two ways to calibrate the image for

pI and M

annotation (see also pages 98 and 118):

1-D calibration. Calibration curves are derived

from IPG pH gradient graphs for pI determina-

tion for one dimension and from the positions

of co-run molecular weight markers for the sec-

ond dimension by importing a ladder.

2-D calibration. pI and M

values are calibrated at

the same time by interpolating between known

values. These values usually come from mass

spectrometry analysis. A number of spots with

known protein properties are selected, the pI

and M

information are imported from the

respective protein lists belonging to these spots.

Gel viewing For visual inspection the spots they can be viewed in

two- and three-dimensional view.

Matching For differential analysis it is important to be able to super-

impose the spot patterns. This cannot be done directly, because typi-

cally there are local distortions caused by imperfections in gel matrix,

variations in gel running conditions, temperature effects, uneven

focusing, and polymerization issues. Formerly spot matching tried to

find corresponding spots in pairs of gels. However for automatic

For controlling the matching

efficiency vectors between the

spots of the gel pair are

displayed. The directions and

lengths of the vectors between

the corresponding spots make it

easy to detect mismatches.

1 Electrophoretic Techniques132

matching it is more efficient to search for pairs of features, using

spot clusters, spot shapes, sizes, and positions together. Even this fea-

ture-based matching can sometimes produce mismatches. Therefore

it is necessary to control this process by visual inspection, and per-

form corrections with setting landmarks.

Data analysis 2-D gel patterns are compared to detect qualitative

and/or quantitative protein expression changes between individual

samples or different populations. As already mentioned in the intro-

duction, the challenge is to find biologically significant changes

between different samples against the background of inherent biolo-

gical variations, induced variations during sample preparation, and

gel-to-gel variations.

Populations of replicate gels are compared to each other with so

called “match sets”. Arbitrary reference gels are no longer employed.

The match sets can be organized in various ways; this allows creating

different groups of gels.

Several statistical tools and clustering techniques are available to

extract results and to discover similarities or significant differences.

Preparing the spot picking list The x/y positions of the selected spots,

including the reference markers, are exported into a table. The num-

bers of the reference markers are automatically named “IR1” and

“IR2” (internal reference).

Reporting and exporting of data The results can be printed out in

images with or without annotations and tables. The files can also be

exported into Excel and other office software.

1.6.2.2 The Analysis of DIGE Gels

DIGE gels are obtained by labeling the proteins in different samples

with different, spectrally distinct fluorescent dyes, the CyDye fluors,

and running the mixed samples together in the same gel. Because of

the multiplex feature and the co-migration of the proteins to the iden-

tical spot coordinates, DIGE gels can be analyzed in a completely dif-

ferent way than traditional gels. This analysis runs almost fully auto-

matically.

The main philosophy in the differential approach of the DeCyder

software is the assumption that the protein compositions of different

samples within an experiment are very similar; that there are only a

few changes in protein expression levels, which are significant and

should be detected.

This reduces the user bias to a

minimum.

1.6 Image Analysis 133

The key points for the evaluation of DIGE gels are:

The co-detection of spots;

The co-separation of the pooled internal standard

in each gel;

Utilizing the internal standard for in-gel normal-

ization.

Co-detection As already mentioned before, because of the co-migra-

tion of the same proteins the spots do not need to be matched within

a gel. Thus the sample spot volumes can be directly related to the

standard spot volumes, which are originating from protein mixtures

of all samples. DeCyder does this automatically by employing a co-

detection algorithm. Co-detection means that for the definition and

assessment of a spot the information of all three channels is com-

bined. Each particular spot receives the identical spot boundary

across all three channels. In this way the relative spot volume calcula-

tion becomes very accurate, and splitting of double or multiple spots

is more reliable than manual splitting. If co-detection would not be

available, the operator of the software, as the supervisor of evaluation,

would detect the spot shoulders and perform manual splitting. Figure

1.49 shows a typical situation. With triple co-detection the correct

spot boundaries are applied. Single detection would lead to erroneous

spot matching.

Fig. 1.49: Comparison of spot boundary definition with triple

co-detection and single detection. On the left hand side the

three-dimensional view of all three channels; on the right hand

side the two-dimensional view of the Cy3 channel.

From Josef Blles with kind permission.

With the help of the internal standard the software can automati-

cally normalize the spot volumes.

1 Electrophoretic Techniques134

Linking the images of different gels via the internal standard A 2-D

electrophoresis experiment always consists of more than one gel.

Therefore the spot volume ratios have also to be compared across dif-

ferent gels. In this case a spot matching procedure has to be carried

out, because there are always some gel-to-gel variations, which do not

allow direct superimposing of the images. However, in each gel one

channel contains the image of the pooled internal standard. Because

these standard images originate from the identical sample mix, spot

matching is greatly facilitated. Figure 1.50 shows how the spot

volume ratios are compared across different gels with the help of the

internal standard.

Fig. 1.50: Schematic diagram showing the linking of data from

several gels via the internal standard images. Note also the reversed

labeling and the planned randomization of sample application.

With a batch processor a high number of images are evaluated fully

automatically, which saves hands-on time and keeps user bias to a

minimum. The significance of changes and quantitative differences

between protein expression levels can easily be checked with several

inbuilt statistical tools: Student’s t-test, one- and two-way ANOVA,

and false discovery rate adjustment. Differences in protein expression

levels of less than 10% can be detected with over 95% confidence.

The unique difference of this

approach to other programs is

the very high evaluation speed

from importing the images to a

reliable result.

1.6 Image Analysis 135

Two examples By using the internal standard concept gel-to-gel var-

iations and technical variations of software are considerably

decreased. Alban et al. (2003) have demonstrated the power of incor-

porating the pooled internal standard using an experimental model.

They generated a model for a time / dose experiment by spiking four

different commercial proteins into E. coli lysate: eight samples have

been prepared with different levels of the spikes. These samples were

analyzed with conventional one color 2-D electrophoresis and with 2-

D DIGE employing the pooled internal standards. The ability of both

methods to measure abundance changes and monitor trends in pro-

tein expression were compared. Besides the fact that the conventional

method is much slower, is more work intensive and needs more gels,

the results were inaccurate and the trends were incorrect when com-

pared to the defined spiking amounts and the results achieved with

the DIGE concept using the internal standard.

In the paper by Friedman et al. (2004) a very good real life example

for the usefulness of the internal standard is presented: In this study

healthy and tumor tissue of six patients were analyzed with 2-D

DIGE. Image analysis and data evaluation of the obtained results

were performed two times, with and without taking the internal stan-

dard in respect. Without the internal standard ten significant differ-

ences between the healthy and the tumor samples were found; but

using the internal standard revealed in total 52 significant differ-

ences.

Statistical evaluation Some valuable statistical aspects on difference

gel analysis can be found in a book chapter on DIGE applications in

clinical research (Sitek et al. 2006).

Extended data analysis Although 2-D DIGE in conjunction with the

powerful image analysis program can be applied for rather complex

experiments like multifactorial time/dose studies, there is a limita-

tion of the sample number contained in one experiment. Also, when

too many samples are analyzed in one experiment, one-time events

can be diluted out of the internal standard. For this reason and in

order to extract as much information as possible from a series of

experiments some additional multivariate statistical tools have been

made available:

Principal component analysis (PCA) reduces the

dimensionality of the data, in order to find the

underlying sources of variation. It is a useful

tool to find initial groupings of the data and

detect outliers.

Different clustering techniques reveal proteins

with similar expression profiles to find complex

Alban A, David S, Bjçrkesten L,

Andersson C, Sloge E, Lewis S,

Currie I. Proteomics 3 (2003)

36–44.

Friedman DB, Hill S, Keller JW,

Merchant NB, Levy SE, Coffey

RJ, Caprioli RM. Proteomics 4

(2004) 793–811.

Sitek B, Scheibe B, Jung K,

Schramm A, Sthler K. In:

Proteomics in Drug Research

(M Hamacher et al. Eds.)

Wiley-VCH, Weinheim (2006)

pp 33–55.

1 Electrophoretic Techniques136

partners and proteins which function together in

certain biological pathways.

With discriminant analysis proteins are found,

which could serve as classifiers as well as diag-

nostic or prognostic markers. With unsupervised

clustering unknown samples can be classified

without user bias.

Fig. 1.51: Application of DeCyder

EDA: (A) False-colored representative

gel containing three differentially

labeled samples. Cy2-labeled internal

standard (blue), Cy3-labeled control

No. 1 (green), and Cy5-labeled iron-

starved No. 1 (red) are overlaid. (B)

Unsupervised PCA properly groups the

16 individual DIGE expression maps

differentiated by two principle compo-

nents (PC1 and PC2) and demonstrates

high reproducibility between the repli-

cate samples within each group.

patterns representing the five proteins

that increase abundance in the presence

of hemin, the 29 Class II protein

features negatively affected by Fur and

iron, and the 30 Class III protein

features positively affected by Fur and

iron. (D) Unsupervised hierarchical

clustering of 16 individual DIGE expres-

sion maps (groups, shown along top)

and of individual proteins (shown on

the left), with relative expression values

for each protein displayed as a heat

map using a relative scale ranging from

–0.5 (green) to +0.5 (red). From: DOI:

10.1371/journal.ppat.0020087.g001

1.7 Spot Handling 137

Friedman et al. (2006) have applied these multivariate statistical tools

on their study on the influences of Staphylococcus aureus on the cen-

tral metabolism of the host. Figure 1.51 shows combined images of

the 2-D electrophoresis result in false color representation, a PCA

grouping, K-means cluster, and a heat map of unsupervised hierarch-

ical clustering.

It is very useful, when the images of the results and the 2-D DIGE

experiment can be directly linked. The add-on software package

DeCyder EDA is programmed in such a way that the original experi-

mental data are directly available via a hyperlink by clicking on a PCA

spot or a band on the heat map.

Spot picking list Because the fluorescent dyes are not visible with

the eye, it is particularly useful to pick the spots from 2-D DIGE gels

according the x/y data of the image file.

1.6.3

Use of 2-D Electrophoresis Data

Most of the image analysis software includes a web browser to check

results of other laboratories in the World Wide Web. Because there is

no standardization for sample preparation and for running 2-D elec-

trophoresis, the spot patterns of different laboratories are not easy to

compare in practice.

A protein cannot be identified on the basis of its position in the gel

alone. Identification is only possible by further analysis of the pro-

tein, for instance with mass spectrometry. It is furthermore incorrect

to conclude that a protein is up or down regulated, just using the

information from growing or shrinking spot intensity. It is also possi-

ble that another protein had changed its pI due to phosphorylation or

another post-translational modification, and is now sitting on top of

the protein observed.

Therefore proteins of interest have to be analyzed further. This is

done mainly with mass spectrometry: The gel plugs containing the

protein are cut out of the gel slab subsequently to image analysis.

1.7

Spot Handling

Between image analysis and further analysis the gels can be kept in a

refrigerator for several weeks. It is also possible to dry the gels for sto-

rage and soak them in water before spot cutting.

In high-throughput proteomics it is very important to keep track of

each sample and of each protein to be analyzed. Before the proteins

Friedman DB, Stauff DL, Pish-

chany G, Whitwell CW, Torres

VJ, Skaar EP.

PLoS Pathog 2 (2006):

e87. DOI:

10.1371/journal.ppat.0020087

In this way the results can be

traced back very easily.

1 Electrophoretic Techniques138

contained in the spots can be analyzed with mass spectrometry a ser-

ies of steps have to be performed:

Spot picking;

Destaining and washing of the gel plug;

In-gel digestion;

Peptide extraction;

Spotting of peptides on MALDI target slides.

During this procedure mix-ups and contaminations have to be pre-

vented to avoid wrong conclusions.

In order to keep control over all these steps and parallel processes,

as well as the results, two things are necessary:

Automation;

Workflow database.

Therefore a workstation has been developed, which handles all steps

automatically. Figure 1.52 shows a schematic drawing of such an inte-

grated system. In the fully automated system a robotic arm transports

gels, microtiter plates and target slides between the stations. The

workstation is controlled by a laboratory workflow system software.

This software can control the entire proteome analysis workflow

including sample tracing.

image

analysis

2-D Gel

fixed on a film

or glass plate

digester

spot picker

gel hotel

spotter

drying oven

robotic

arm

incubator

refrigerator

trypsin

MALDI ToF MS

Fig. 1.52: Schematic drawing of an automated workstation for

spot picking, in-gel digestion, and spotting peptides on a

MALDI target slide.

The gels are identified with the help of barcodes on the support

films and the trays. After spot picking the gel is transported back to

the gel hotel. Many moves are needed for the digestion step. Spot

1.7 Spot Handling 139

picking and spotting are combined to be carried out by the same

instrument.

The laboratory workflow system can be extended to include all

instruments and software involved in the entire proteomics analysis,

from sample preparation to mass spectrometry.

In most of the laboratories the different steps are performed by

stand-alone instruments.

1.7.1

Spot Picking

Picking spots from a gel manually is a very cumbersome job. The

spots, marked in the printout of the image analysis for further analy-

sis, have to be found in the gel again and transferred to the correct

reaction tube or well of a microtiter plate. Sample tracing is not easy

and errors can easily occur. Additionally, contaminations with keratin

and other stuff must be avoided. This is a typical work for a laboratory

robot.

Robotic spot pickers are much more reliable for excising selected

spots from the gel slab and transferring them to defined wells of

microtiter plates. This automated picking can be controlled with a

CCD camera or by transferring the pixel coordinates from image ana-

lysis into the machine coordinates of the picker instrument.

For the second technique, the gel has to be immobilized on a rigid

support – a glass plate or a plastic film – prior to scanning. Two self-

adhesive reference markers, glued onto the support plate or film, are

necessary to enable the machine to recalculate the x/y positions of

the proteins from the imported picking list. The two markers are

automatically recognized by the camera and used as calculation

points for the spot positions. This procedure provides a very high

picking efficiency and accuracy, because it utilizes the high resolution

and sensitivity of the scanning device. Furthermore, using the image

analysis data, also fluorescent, non-stained and radioactive labeled

spots can be excised.

& Note: For this approach it is necessary and very

important that the image analysis software does

not modify the raw data.

Figure 1.53 shows the assembly of a robotic spot picker working

according this concept. The gel is placed into a liquid-containing tray

to prevent drying during the picking process. The glass plate or film

support is fixed at the bottom of the tray.

A picking file has to be created to assign the spot position to the

respective microtiter plate well.

The camera in this system has

only the function to find the

positions of the two reference

markers.