EXPERIMENTAL DESIGN FOR

Yeast GENOMIC EXPRESSION ANALYSES

 

As discussed at the 2000 Yeast Genetics Meeting in Seattle, Washington

 

 

I.  Experimental setup

A.  Choose the experimental parameters

 

1.  Choosing the strain background:  Many labs routinely work with their favorite

strains, but it is worthwhile to consider which strain is best for your genomic

studies, because different strains can have very different genotypes and

phenotypes

 

a.  Be aware of the differences between common lab strains (eg. S288C vs

W303, etc.) and how those differences might the experimental results

b.  The mating type can affect phenotype, including gene expression

c.  Best to use strains with minimal auxotrophic markers to simplify analysis of

the results

 

 2.  Choosing the experimental conditions

a.  Determine the optimal media, culture type, temperature for your needs

 

b.  Investigate the optimal “dosage” of the stimulus:  Many genomic

studies observe the response of cells to various stimuli.  It is worth

investigating the dosage of the applied stimulus to pick conditions that

will give meaningful results

 

1)  Lethal conditions may result in data that are difficult to interpret,

while mild conditions may not provoke a detectible genomic   

expression program

 

c.  Investigate the appropriate time points for each timecourse experiment

 

1)  Timecourse experiments yield a higher level of detail than single-

timepoint experiments, including temporal information

 

2)  Determine the appropriate time points that will reveal the genomic

expression response … it is easy to miss rapid responses that occur and subside within a short period (eg 15 minutes)

 

 

B.  Plan the experiment:  ONE VARIABLE ONLY!

As in any experiment, it is very important to insure only One variable in genomic expression experiments.  Often, experimental variables that overlooked can provoke substantial changes in genomic expression and confound analysis of the results.

 

1.  Hypothetical and real examples of multiple variable experiments:

 

a.  **  Diauxic shift during experiment:  likely the most common oversight

is the progression of the cells through diauxic shift, when the cells  become limited for glucose and alter their metabolism accordingly.  The expression of thousands of genes is altered during this phase of growth.  The timing of diauxic shift is dependent on the culture conditions (strain, media, growth temperature, aeration, environmental stress) so it is very important to know when diauxic shift occurs under your conditions and avoid it (see more below).

 

b.  Pleiotropic drugs will result in pleiotropic cellular effects and thus genomic

     expression

 

            1)  The “DNA damaging agent” Methyl-methane sulfonate (MMS)

methylates many cellular targets in addition to DNA

2)  High sodium – alters ionic strength, osmotic strength, as well as Na+ concentration

 

c.  Experiments with extensive cell handling:  account for cell handing in a 

control experiment

     1)  Changes in culture aeration can lead to hypoxia

 

d.  Drugs suspended in a carrier solution:  add carrier alone in mock control

 

C.  Choose the reference for microarrays:  The main goal in choosing a reference is to ensure significant hybridization signal in every spot on the arrays so that the Ratio of R/G signal in each spot can be quantitated … therefore the identity of the reference is somewhat arbitrary; the data can be subsequently mathematically transformed to reveal the biologically-relevant ratios.

 

1.  Example reference samples

a.  Genomic DNA

b.  An arbitrary RNA reference pool

c.  Time zero RNA, taken just before beginning the experiment

d.  A pool of all of the RNA samples recovered from an experiment

RNA taken from the control sample

 

2.  Regardless of which reference is used, be sure to use the identical reference

on all arrays in a given timecourse so as to compare the timepoints to eachother

 

3.  Mathematical transformation example:

 

a.  for a timecourse in which genomic DNA is used as the reference, there will

be one array for each time point INCLUDING the time = 0 sample

 

b.  Each ratio for each spot = Red/Green signal = signal from time point    

RNA sample/signal from genomic DNA

 

       c.  To transform the data, divide the R/G ratio measured for each gene on the t>0 arrays  by the corresponding R/G ratio measured on the t = 0  array to cancel the “genomic DNA” denomenator: 

 

(R/G t > 0 array) / (R/G t = 0 array)  = (RNA t > 0 /Genomic DNA) / (RNA t=0 /Genomic DNA)  =

RNA t > 0/ RNA t = 0

 

 

II.  Execution of the experiment

 

A.  Before beginning sample collection, allow the subculture to recover from stationary

phase

1.  At least 2 doublings (not absolute time)

 

B.  Begin the experiment at a cell density that will avoid diauxic shift at end of the

experiment

1.  Aware that timing of diauxic shift is condition-specific (media, temperature,

aeration, environmental stress)

 

2.  For long experiments use a chemostat to maintain culture conditions

 

C.  Record ALL possible details – you’ll be glad you did

 

1.  Examples:

a.  OD600, cell number, cell volume over time

b.  Cell viability through experiment

c.  Cell morphology:  take periodic photographs of the cells to

characterize cell shape, cell cycle arrest

d.  Nutrient concentrations (glucose, NH4, ethanol, etc.):  Always freeze a

small aliquot of the culture media to measure such things later

e.  If possible, measure drug concentrations in the culture during

experiment

2.  Record any anomalies

3.  Record anything you can think of … you many not know what details

will be valuable until After you see the results

 

III.  Sample collection

A.  Be as controlled as possible!  Collect all samples as identically as possible

 

B.  Collect by centrifugation (3-5 min at ~3g) or filtration (<1 min by filtering culture

over a sterile 0.45 um filter and collecting entire filter)

 

C.  Again, few variables

 

1.  Collect cells ~experimental temperature to avoid

temperature shock     

2.  Collect cells as quickly as possible

3.  Do not wash or handle cells unnecessarily

 

 

D.  Example problems:

1.  washing cells induces many variables and can induce the stress response

( induces changes in nutrients, osmolarity, ionic, pH, etc.)

2.  collecting cells on ice can induce cold shock

3.  lengthy collection can induce hypoxia

4.  variable collection time is Very Bad!

 

V.  Microarray hybridization

A.  Again, as controlled as possible

1.   For optimal comparison within a timecourse, perform sample collection, RNA

preparation and labeling, and especially hybridizations together for consistency

 

       B.  Always be consistent with labeling and hybridizations (time, temperature) to improve

reproducibility

 

 

VI.  Duplicate experiments

A.  Duplicate experiments as identically as possible

1.  Maintain as many experimental details (such as starting OD600, culture shaker

speed, timing of experiment, etc.) as possible

 

2.  The most common differences between experiments seem to be differences in

the  expression of metabolic genes

 

 

V.  The practical art of data Analysis

A.  Remember all of the experimental details when analyzing the data

 

1.  Be aware of:

 

a.  Pleiotropic conditions during the experiment

b.  Potential diauxic shift problems

c.  Strain background (auxotrophic markers, mating type)

d.  Cell cycle progression or arrest

e.  Secondary effects of the primary stimulus

 

     B.  Keep an open mind when interpreting the results!

 

     C.  To identify responses that are specific to your conditions, compare the data to other

datasets of unrelated conditions … remember that many of the observed responses

may not be specific to your conditions.

 

      D.  Remember that whatever analytical method used to analyze the data (hierarchical

clustering, Self Organizing Maps, K-means clustering, Singular Value

Decomposition), these are methods of Organizing the data

 

1.  Which analysis method is the best:  NONE! 

 

a.  Different methods have different strengths and weaknesses which make

each method suited to different analytical problems

 

b.  Often the most thorough analysis involves multiple permutations of multiple

computational methods

 

c.  A given cluster is not necessarily “THE” answer