Antibody Microarrays

1
Antibody Microarrays
Merrill Birkner ph296December 1, 2003
2
Antibody (Ab) Microarray

• A complete microarray-based system for profiling
  protein expression in biological samples used to
  compare two biological samples to measure the
  relative differences in protein expression.
• The microarray consists of hundreds of monoclonal
  antibodies covalently bound in an ordered layout
  to a glass slide.
• A protein which can be synthesized in pure form
  by a single clone (population) of cells. These
  antibodies can be made in large quantities and
  have a specific affinity for certain target
  molecules called antigens which can be found on
  the surface of cells and those that are
  malignant.
• The array can be used as a means to correlate
  specific proteins with physiological or
  pathological process of interest, by comparing
  hundreds of proteins at a time.
• It is used for toxicity testing, disease
  investigation, and drug discovery.

3
Antigens Antibodies

• Antigens
• Molecules that stimulate the production of
  specific antibodies and combine specifically with
  the antibodies produced. Most antigens are
  foreign to the blood and other bodily fluids.
• Antibodies
• Antibody proteins (immunoglobulins) are found in
  the gamma globulin class of plasma proteins.
  There are five main subclasses IgG, IgA, IgM,
  IgD, and IgE. (ex. Most antibodies in serum are
  from the class IgG).

4
Antibody Structure

• Consists of four interconnected polypeptide
  chains. Two heavy chains (H-chains) and joined
  to two shorter chains (L-chains).
• These four chains are arranged in the form of a
  Y with the stalk of the Y is called the
  crystallizable fragment and the top of the Y is
  known as the antigen-binding fragment.

5
Antigen/Antibody Interaction
6
Ab Array Procedure

• Extraction of total cellular protein from
  biological samples of interest (eg. Serum
  samples).
• Labeling of extracted protein with fluorescent
  dyes Cy5 and Cy3 (direct labeling, direct
  labeling with hapten tag, paired Ab sandwich
  assay).
• Removal of unbound dye.
• Incubation of labeled protein with the array.
• Scanning of the array and the analysis of the
  results.

7
(No Transcript)
8

• This procedure is a fluorescence-based analysis
  covalently immobilized antibodies are used to
  capture fluorescently labeled antigens.
• They do not measure absolute concentrations-
  instead they provide a relative measure of
  protein abundance i.e. the abundance of protein
  in one sample as compared to another sample.
• As part of array development, all antibodies are
  printed and tested against their specific
  purified antigen (when available) and against
  cell lines and tissues samples (for quality
  control).
• A reference pool is also used, and similar to the
  gene expression microarrays, a pool of equal
  aliquots from each sample to be measured is used,
  thus ensuring that all proteins from the samples
  are represented in the reference.

9
(No Transcript)
10

• Direct Labeling
• Direct Labeling with a hapten tag.
• Paired Ab sandwich assays.

11
Direct Labeling (w/ hapten tag)

• A convenient method to measure multiple proteins
  in a complex mixture. All proteins are labeled
  with either a fluorophore or a hapten tag such as
  biotin.
• Advantages
• Only one captured antibody per target is
  required, as compared to the next method- easier
  to expand detection to new targets for which
  matched antibody pairs may not be available.
• Can label different samples with different tags
  and to co-incubate the samples on the same
  arrays.
• Disadvantages
• Potential for a high background all proteins are
  labeled from the sample, including high
  concentration proteins such as albumin in serum
  nonspecific binding or adsorption of these
  proteins to Ab could cause interference ? reduce
  detection sensitivity or data accuracy.
• Potential for disruption of antibody-antigen
  interactions if the labeling reaction severely
  alters an antigens binding site.

12
Dual Antibody Sandwich

• Antibodies spotted onto microarray substrates
  capture specific antigens, and a cocktail of
  detection antibodies, each antibody matched to
  one of the spotted antibodies, is incubated on
  the arrays.
• Advantages
• Quantification of the bound detection antibodies
  provides a measure of each antigens abundance.
• Sandwich assays are more sensitive than the
  direct labeling method because background is
  reduced through the specific detection of two
  antibodies instead of one.
• Disadvantages
• The development and validation of assays
  measuring many targets in parallel is difficult
  because of the cross reactivity and precipitation
  when using many detection antibodies.

13
ELISA as a validation method

• The Enzyme-Linked Immunoabsorbent Assay is
  serologic test used as a general screening tool
  for the detection of antibodies or antigens in a
  sample. ELISA technology links a measurable
  enzyme to either an antigen or antibody.
• These tests are often used to validate the
  microarray results

14
Ab/Ag interaction in ELISA wells
15
Gene Expression vs. Ab Microarray

• Gene expression, in most cases, does not
  necessarily correlate with changes in protein
  expression.
• In cases when there is a correlation between mRNA
  and protein abundance, the correlation is often
  time shifted.
• This time shift is likely to be different for
  each mRNA-protein pair.
• With these arrays it is now possible to compare
  changes in gene expression with changes in
  protein expression using similar technologies.
• There are also many reasons for merely studying
  protein abundance.

16
Ab Microarrays Cancer Research

• Information from protein profiling experiments
  may reveal associations between proteins or
  groups of proteins and disease states or
  experimental conditions.
• Biomarkers in cancer are potentially valuable for
  early detection, staging of patients,
  classification of patients, or as surrogate
  markers for drug response.
• These microarrays increase the number of proteins
  that can be conveniently measured, therefore
  taking advantage of the benefit of using combined
  markers in diagnostics.

17

• Important in this field because there is a low
  volume requirement and the multiplex detection
  capability of microarrays make optimal use of
  precious clinical samples.
• Work continues on the optimization of various
  aspects of the protocols, such as substrates for
  Ab attachment, the methods of Ab attachment, Ab
  buffers and concentrations, wash conditions, etc.
  

18
Antibody microarray profiling of human prostate
cancer sera Antibody screening and
identification of potential biomarkers.
Proteomics 2003, 3, 56-63.

• Miller, J., Zhou, H., Kwekel, J., Cavallo, R.,
  Burke, J., Butler, E.B., Teh, B., Haab, B.

19
Background

• Protein Biomarkers in the serum hold great
  promise for noninvasive disease detection and
  classification.
• Ab protein microarrays can have many
  applications including protein profiling of
  cancer tissue, autoimmune diagnostics, protein
  interaction screening, and Ab-based detection of
  multiple antigens.
• Certain parts of the Ab microarray technology
  have not been perfected
• An optimized protein immobilization method is
  needed that retains native structure and
  reactivity and decreases nonspecific protein
  adsorption.
• Ab can be immobilized by adsorption to
  poly-L-lysin membranes, by chemical crosslinking
  to derivatized glass surfaces. Hydrogels
  recently have also been introduced as a protein
  microarray substrate.

20

• Another important issue is to create an efficient
  method of validating antibody performance in the
  microarray assay.
• Previous work in the development of the antibody
  microarray methods made use of solutions of known
  target antigen concentrations to characterize
  antibody performance.
• This is often very expensive and the antigens are
  often unavailable.

21
Goals

• 1. Compare two surfaces and antibody
  immobilization schemes poly-L-lysine coated
  glass with a second photoreactive cross linking
  layer, polyacrylamide-based hydrogels on glass.
  
• 2. Establish an efficient method to screen
  antibodies for those that are functional in the
  microarray assay.
• Hypothesis a statistical filter could identify
  antibody measurements that are consistent with
  specific and quantitatively accurate antigen
  binding. This hypothesis is tested by comparing
  microarray measurements to ELISA tests.
• 3. Demonstrate the use of this technology to
  screen serum samples for potential biomarkers, by
  analyzing the relative protein abundances in
  serum samples from prostate cancer patients and
  controls.

22
Serum samples

• 33 males with prostate cancer ages 39-85 at the
  Methodist Hospital Houston, TX, USA, prior to
  commencement of radiotherapy.
• PSA (prostate-specific antigen) concentration
  2.5-335 ng/mL (from ELISA) median 6.4 ng/mL
• Histological grades of cancer tissue samples
  ranged from a Gleason combined scores of 6-9
• 20 serum samples taken from healthy males aged
  30-69
• Normal PSA levels 0.2-3.2 ng/mL median 0.85
  ng/mL

23
Microarray preparation

• The microarrays were deposited on two different
  types of substrates the poly-L-lysine (HSBA) and
  the hydrogel details found in paper.
• The serum samples and reference pool were diluted
  and mixed with the Cy5 or Cy3.
• A reference pool is a pool of equal aliquots from
  each sample, thus ensuring that all proteins from
  the samples are represented in the reference.

24
Data analysis and statistics.

• The local background in each color channel was
  subtracted from the signal at each antibody spot
  (spots with defects or no detectable signal
  removed).
• The ratio of the net signal from the
  sample-specific channel to the net signal from
  the reference specific channel was calculated for
  each antibody spot ratios from replicate
  antibody measurements in the same array were
  averaged.
• The resulting ratios were multiplied by a
  normalization factor for each array (next slide)
• Hierarchical clustering and visualization were
  performed using Cluster and Treeview. Ratios
  were log transformed median centered.
• Antibodies that did not have good measurements in
  at least 75 of the samples were removed from
  subsequent analysis.
• The permutation t-test was calculated using the
  program Cluster Identification Tool.

25
Normalization Method

• The resulting ratios were multiplied by a
  normalization factor for each array N, calculated
  by
• N (SIgG / µIgG)/RIgG
• SIgG the ELISA-measured IgG concentration of
  the serum sample on that array.
• µIgG the mean ELISA-measured IgG concentration
  of all of the samples.
• RIgG the average ratio of the replicate
  anti-IgG antibody spots on a particular array.

26
Internal Normalization

• 4 slides per sample were created (label
  sample A with Cy3 and Cy5 and sample B with Cy3
  and Cy5). Combine one of each of these samples
  and scan determine the signal ratios of the 2
  slides (Cy5/Cy3). With this method potential
  variability is eliminated because each protein
  sample labeled with each dye.

27
Results

• Serum control samples were analyzed using a
  two-color comparative fluorescence assay on
  microarrays containing 184 different antibodies
  spotted in quadruplicate.
• 40 of the antibodies targeted 32 unique proteins
  that are typically found in the serum of healthy
  serum samples.
• Another 13 antibodies targeted 9 proteins that
  have been detected in the serum of cancer
  patients, and the rest of the antibodies targeted
  normally intracellular proteins.

28
Goal 1

• The samples were repeated twice on the 2 types of
  slides (internal normalization) there were 4
  microarray experiments per sample.
• The hydrogel substrate generally produced a
  lower, more consistent background than the other
  surface.
• The fluorescent signal from the Ab on the
  hydrogels had an average six-fold higher S/N
  ratio than the corresponding antibodies on the
  other surface.
• The Ab showed measurable signal above background
  using the hydrogels (78 Ab) as compared to the
  other surface (23 Ab).
• The hydrogel allowed weak detection from the
  greater number of Ab, reflecting decreased
  detection limits as a result of higher S/N ratio
  measurements.

29
Goal 2

• Define a statistical test that could filter the
  Ab measurement for those that are consistent with
  specific and accurate antigen binding.
• They examined the overall variation in the
  reproducibility of Ab measurements from the 2
  different surfaces after reverse labeling
  (mentioned before).
• Developed an effective an efficient method to
  screen antibodies for those that function well in
  the microarray assay.
• Using ELISA measurements as standards, they
  examined the ability of the statistical filter
  based on the correlation of the data from
  reversed labeling experiments to distinguish
  between reliable and unreliable microarray
  measurements.

30

• First, in order to view patterns of similarity
  between sets of microarray measurements, average
  linkage hierarchical clustering is used.
• There are 4 slides hydrogel sample in red,
  hydrogel sample in green, HSBA sampled labels in
  red, and HSBA labeled in green. These were
  combined and clustered.
• Each colored square represents one Ab measurement
  from one array. The color and intensity of each
  square represents the relative protein binding of
  the sample versus the reference.
• Ab measurements that reproduce well between the
  different experiments are clustered together.

31
(No Transcript)
32

• The correlation of measurements from replicate
  data sets as an initial screen to identify
  reliable antibodies.
• The Pearson correlation of measurements between
  the reverse-labeled experiment set was
  calculated, both for hydrogel and HSBA. For both
  surfaces progressively fewer Ab exceeded the
  threshold as the threshold was increased.
• In order to assess the degree to which the
  correlation parameter predicted specific and
  accurate antigen detection, microarray
  measurements from 7 of the Ab were compared to
  ELISA measurements for the corresponding
  antigens.
• For these 7 antibodies a high inter-experiment
  set of correlations predicted a good agreement
  between the microarray and ELISA measurements.
• They found no examples of Ab measurements that
  have high inter-experiment set correlation but
  poor agreement with ELISA measurements!

33
(No Transcript)
34
(No Transcript)
35
Goal 3 Detection of biomarkers

• As a result of the previous analysis, only Ab
  that passed the stringent correlation threshold
  for inclusion were used in the following
  analysis.
• They used a correlation threshold of 0.7, because
  the microarray measurements exceeding this
  threshold agreed well with the ELISA
  measurements.
• In order to estimate the significance of the
  association between expression patterns and
  sample groups (cancer and normal) permutation
  t-tests were used.
• Determines the statistical significance of each
  genes discrimination using a user defined
  segregation of samples.

36
Permutation t-test

• Estimate the distribution of the t-test
  statistics under the null hypothesis by
  permutation of the sample labels.
• The p-value pg is given as the fraction of
  permutations producing a test statistic that is
  at least as extreme as the observed one. It is
  the probability under the null hypothesis that
  the test statistic is at least as extreme as Tg.
  
• Standard t-tests assume normally distributed data
  in each class and equal variance within classes.
  This test will be more accurate than the normal
  t-test for non-normal distributions and small
  samples.

37

• When applied to the discrimination of cancer
  patients from the controls, CIT identified
• vWF, IgM, alpha-anti-chymotrypsin, Villin, and
  IgG with p-values below 0.01.

Marker P-value Correlation w/ PSA
vWF lt0.000007 0.18
IgM lt0.00006 -0.19
ACT lt0.001 -0.15
Villin lt0.001 -0.36
IgG lt0.01 -0.16
38

• Hemoglobin was also discriminated but was found
  to be an artifact of hemolysis of controls.
• None of these markers significantly correlated
  with PSA, and all varied independently of PSA.
• IgM and IgG were lower and vWF higher in cancer
  patients therefore similar to previous studies.
• Since none of the proteins correlated with PSA,
  they could potentially bolster diagnostic
  accuracy if used in conjunction with PSA.

39
Remarks/Future work.

• Larger studies are needed to further examine the
  relationship between serum proteins and prostate
  cancer.
• Further development in this technology will have
  significant utility in medical diagnostics as
  well as broader clinical and research
  application.
• Using the D/S/A algorithm to analyze the data.
  (Data Van Andel Inst. Michigan Dr. Brian Haab)

40
(No Transcript)