Transcript rnai_presentation

Kyle Breakey, Kylie Broadbent,
Christine Mineart, Matt Perez
http://www.life.uiuc.edu/shapiro/RNAipathway.jpg
Limitations
Full-genome RNAi profiling of early
embryogenesis in Caenorhabditis elegans
B. Sönnichsen1, L. B. Koski1,6, A. Walsh1, P. Marschall1,6, B. Neumann1,6, M. Brehm1, A.M. Alleaume1,6, J. Artelt1,6, P. Bettencourt1,6, E. Cassin2,6, M. Hewitson1, C. Holz1, M.
Khan1, S. Lazik1, C. Martin1, B. Nitzsche1,6, M. Ruer2, J. Stamford2, M. Winzi1, R.
Heinkel1,6, M. Röder1,6, J. Finell1,6, H. Häntsch1, S. J. M. Jones3, M. Jones4,6, F. Piano5, K.
C. Gunsalus5, K. Oegema2,6, P. Gönczy2,6, A. Coulson4,6, A. A. Hyman2 and C. J.
Echeverri1
2005
2000
http://www.newsdesk.umd.edu/scitech/images/worm-briggsae_WT_male.gif
Limitations
• Endpoint Analysis: (older)
– Phenotypes scored when researcher made observations
– Phenotypes were missed-such as delays in cell division
– Can’t differentiate between secondary effects and the direct knockdown of
genes
• Movies: (less old)
– DIC relies on manual phenotyping
• Biased and not quantitative
• Expensive and time consuming
• Live-Cell Assay: (2006)
– Image captured every 30 min. and made into a video clip
– Computerized analysis of phenotypes using fluorescence imaging
HeLa?
•
Henrietta Lacks
Died of cervical cancer on the 4th October, 1951 at
Johns Hopkins University Hospital.
•
Biopsy
Sample sent for analysis and deemed malignant – Then
revealed to be the 1st immortal human cell line.
•
Immortal
Capable of dividing outside of the body an unlimited
number of times as long as fundamental survival
conditions are sustained.
•
Ethical?
Went to the supreme court who ruled that discarded
tissue and cells are no longer the property of the
individual and are free to be commercialized…
http://www.hno.harvard.edu/gazette/2001/07.19/photos/4-Gilbert-250.jpg
HeLa?
• Instrumental in creating
the Polio vaccine
HeLa Karyotype
• May be the key to
curing cancer?
• Karyotype shows 84
chromosomes
– Thought to have fused with
HPV genome through
horizontal gene transfer
http://cancerres.aacrjournals.org/cgi/content/full/59/1/141#SEC6
siRNA Cell Array
(See references #8-9)
http://www.nature.com/ng/journal/v37/n6s/full/ng1560.html
• siRNA-gelatin transfection
solution was prepared in 384
plates.
– Transfection=the introduction of
foreign material into eukaryotic
cells
• HeLa cells seeded on top of the
array
• Spotted siRNA are taken up
from the solid phase by the cells
morphology
Reporter
expression and
immunostaining
Viability/cell
count
-49 target genes known to be related to chromosome segregation or nuclear
structure
-synthesized one siRNA for each gene and assayed on HeLa cells
dsRNAi sequences used
-49 target genes known to be related to chromosome structure
-synthesized one siRNA for each and assayed on HeLa cells
-printed automatically prepared ‘transfection cocktails’ for 50 siRNAs into
imaging chambers
-49 target genes known to be related to chromosome structure
-synthesized one siRNA for each and assayed on HeLa cells
- printed automatically prepared ‘transfection cocktails’ for 50 siRNAs into
imaging chambers
-siRNA taken up by sample cells
-49 target genes known to be related to chromosome structure
-synthesized one siRNA for each and assayed on HeLa cells
-printed automatically prepared ‘transfection cocktails’ for 50 siRNAs into
imaging chambers
-siRNA taken up by sample cells
-43/49 gene knockdown efficiencies measured by qRT-PCR
-all siRNAs showed partial knockdown
dsRNA primers used to
quantify expression levels
of targeted gene products.
Supplementary Figure 1, Neumann et al.
relative remaining mRNA (%)
siRNA knock-down efficiency. mRNA knock-down efficiency of 43 siRNAs targeting 42 of the 49 endogenous
genes with predicted cell cycle function as well as the peripheral Golgi coatamer protein COPB as a known
suppressible gene. For the remaining six siRNAs qRT-PCR did not yield products due to difficulties in primer
design. mRNA expression levels were determined 48 hours after transfection and are shown as average and
standard deviations of three independent experiments.
-49 target genes known to be related to chromosome structure
-synthesized one siRNA for each and assayed on HeLa cells
-printed automatically prepared ‘transfection cocktails’ for 50 siRNAs into
imaging chambers
-siRNA taken up by sample cells
-43/49 gene knockdown efficiencies measured by qRT-PCR
-all siRNAs showed partial knockdown
-scored phenotypes by fluorescence microscopy
Figure 1. Workflow of highthroughput RNAi screening by
time-lapse imaging.
Flowchart of the steps, including
time lines of each process based
on one live-cell microarray
containing 384 siRNAs.
-49 target genes known to be related to chromosome structure
-synthesized one siRNA for each and assayed on HeLa cells
-printed automatically prepared ‘transfection cocktails’ for 50 siRNAs into
imaging chambers
-siRNA taken up by sample cells
-43/49 gene knockdown efficiencies measured by qRT-PCR
-all siRNAs showed partial knockdown
-scored phenotypes by fluorescence microscopy
-TPX2 and COPB RNAi knockouts
-49 target genes known to be related to chromosome structure
-synthesized one siRNA for each and assayed on HeLa cells
-printed automatically prepared ‘transfection cocktails’ for 50 siRNAs into
imaging chambers
-siRNA taken up by sample cells
-43/49 gene knockdown efficiencies measured by qRT-PCR
-all siRNAs showed partial knockdown
-scored phenotypes by fluorescence microscopy
-TPX2 and COPB RNAi knockouts
-can store microarrays for 7 months w/o loss of transfection
-49 target genes known to be related to chromosome structure
-synthesized one siRNA for each and assayed on HeLa cells
-printed automatically prepared ‘transfection cocktails’ for 50 siRNAs into
imaging chambers
-siRNA taken up by sample cells
-43/49 gene knockdown efficiencies measured by qRT-PCR
-all siRNAs showed partial knockdown
-scored phenotypes by fluorescence microscopy
-TPX2 and COPB RNAi knockouts
-can store microarrays for 7 months w/o loss of transfection
- Scrambled siRNA samples did not affect cell growth
-http://www.nature.com/nmeth/journal/v3/n5/suppinfo/nmeth876_S1.html
Automated Phenotype Analysis
• To begin this process, the boundaries of the
fluorescent chromosomes were defined by
segmentation using “optimized local adaptive
thresholding”
– Histones were tagged with GFP
– An algorithm is developed to segment the image to be
studied
– Segmentation is based on the gray levels in the pixels
of the image
• This identified interphase nuclei and mitotic
chromosome sets with over 99% accuracy
(compared to manual analysis)
• Second, the texture and shape properties of
each chromosome set were computed.
• Extracted numerical features from each
segmented chromosome and defined values:
– Shape
– Texture
• From these properties of texture and shape,
biological classes were defined
• The classes defined:
– Interphase
– Mitosis
– Apoptosis
• Several different cell-death phenotypes
– Shape
• Nuclei of abnormal shape
• Automatic classification into these classes was
done by machine-learning – multiclass SVMs
• SVM - support vector machine
– Vectors are used to “teach” the classifier to identify and
classify images into classes
– In this example, vectors that defined shape and texture
features from the chromosome sets were used to teach
the SVM to identify different biological classes
• The SVM classifier was trained on the 45 bestranked features of 1,000 manually labeled cells
– Classifier achieved 97% accuracy on all samples
• There were 89 frames per movie for each siRNA,
for every single frame of the time-lapse movie, of
each spotting position, a class was assigned to the
cells
• This gave plots of the change in phenotype
over time
• They used automated phenotyping for the
analysis of these images as it allowed for a
“quantitative, unbiased, and reproducible
analysis of phenotypes”
– A project like this generates lots of data, to
efficiently analyze it all a method like this had
to be developed
• Segmentation and classification analyses were
done for 2 normal cell divisions on a scrambled
siRNA spot and the software correctly assigned
the interphase and mitotic chromosomes.
• Apoptotic cells were visualized by
suppressing nuclear envelope proteins
– SYNE2 and PLK1
• Knock-down of an inner centromere protein
INCENP resulted in segregation defects that
were correctly assigned to the shape class
• An entire movie was analyzed for classification
– For siSYNE2, siPLK1, siINCENP
– Plots of “mitotic”, “shape”, and “apoptotic” analyses were generated
• Each analysis was done by taking the number of the cells in the
area of the movie that corresponded to a class and that was
compared to the total number of cells in that same area.
• The percentage of cells in a class were compared to
the data generated for the scrambled siRNA cells.
• If any of the three classes were significantly
different from the scrambled siRNA cells (control),
the phenotype appearance and phenotype
penetrance were defined.
• By observing the effects of a specific gene
knockdown over time, more detailed information
regarding the phenotype of the knock down could
be gathered
– Phenotype penetrance
• The maximum difference from the control
– Phenotype appearance
• the earliest time at which the difference from the control cells
Automated Phenotype Analysis
• The software developed to observe the phenotype
of the cells performed:
• 1.
Found location of chromosomes in each cell for each
frame
• 2.
Classified the chromosomes by morphology
• 3.
Detected cell-cycle phenotypes based on classification
results for an entire movie
• This method developed a way to analyze a
genome-wide RNAi study efficiently and
accurately
Quality Analysis
• Compared automatic phenotyping
(Computer) to manual phenotyping
(Human)
• For mitotic phenotypes: 86% agreement
(43/49)
– 14% false postive (7/49)
• Apoptotic phenotypes: 88% agreement
(44/49)
– 10% false postive (5/49)
Phenotypes Defined
• Mitotic – Abnormalities in cell
division
• Shape – Unusual shape of nucleus
• Apoptosis - DEATH
Classification of Genes
• Phenotype displayed
-Mitotic
- Shape
- Apoptotic
• Penetrance
- Used the mean of the
maximum values
Classification of Genes (cont.)
• Order of appearance of
phenotypes
– Not possible with
endpoint assays
– Increases ability to
categorize
• Indicates stability of
targeted protein and
efficiency of siRNA
Multiparameter Clustering
• Grouped genes by
phenoprint
– Phenotype observed
over time
This is cool but…
• Machine learning has limitations
– Only recognize phenotypes it was
programmed to recognize
– Explains 14% false postive rate
• Lots of programming to recognize
phenotype of all the different cell types
Important to Remember
• The point of this paper was to design a high
throughput method of detecting phenotypes
using RNAi
– Not to discover the specific phenotypes of the
genes used in the experiment
 49 RNAi constructs were used to assay mitotic assembly
 Time-lapse imaging resolved phenotypes
 developed a method for high-throughput RNAi data analysis
 clear advantages for time-lapse imaging over end-point analysis
 detects mitotic delays, rather than just mitotic arrest
The End
(Apoptosis)
Questions?