Measuring gene expression in individual live cells

1
Measuring gene expression in individual live
cells

• David A. Ball
• Synthetic Biology Group
• July 14, 2009

2
Overview

• Background
• What is gene expression?
• Why do we need to measure proteins in single
  cells?
• Application to Yeast cell cycle
• Current Measurement System
• Hardware
• Software
• Future Measurements
• New instrument Cyto?IQ

2
3
Gene expression

• Transcription
• Genes in DNA copied to messenger RNA
• Translation
• Ribosome converts mRNA codons (3 bases) into
  Amino Acids
• Amino Acids form protein

3
4
Gene expression

• Necessary to characterize both steps
• mRNA (1-10/cell)
• Protein (10-10,000/cell)

5
Saccharomyces cerevisiae (budding yeast)

• Well-studied organism
• Genome completely sequenced
• Asymmetrical division
• Population doubles in 100 minutes
• Cell cycle is similar to higher Eukaryotes

5
6
Budding yeast cell cycle
7
Classic protein measurements

• Synchronize cells
• chemical blocks cells in specific cell cycle
  phase
• release block
• take measurements of population at regular
  intervals
• chemical may interfere with process of interest
• Measurements Western Blot
• Kill lots of cells
• Extract all proteins, separate by electrophoresis
• Transfer to membrane, stain with
  fluorescent-labeled antibody
• Results in mean protein concentration

7
8
Cell cycle model

• gt30 equations
• 100 parameters
• Predicts behavior of 100 mutants

Chen et al., Mol Biol Cell 2004
8
9
Life is noisy...

• Many sources of variation in cells
• Fluctuation of molecular interactions
• Fluctuations of the cell size
• Distribution of molecules among cells

Need to observe gene expression in individual
cells
Elowitz, et al, Science 2002
9
10
Green Fluorescent Protein

• Responsible for bioluminescence of Aequorea
  victoria
• 1962 Purified GFP
• 1992 DNA Sequence
• 1996 Crystal structure

11
The many colors of GFP

• Mutations in DNA sequence cause shifts in color
• Blue, Cyan, Green, Yellow
• Reef Coral produces red fluorescent protein
• Red, Orange, Yellow
• Currently 100 flavors

11
12
GFP in other organisms
13
Current workflow Hardware

• Computer-controlled microscope
• Point-visiting
• Auto-focus
• Motorized filters shutters
• CCD camera
• Collect 2 images/time-point
• Bright-field
• Fluorescence
• 5 GB of images

13
14
Movies
GAL1
CLN2
15
Current workflow Software

• Identify individual cells (segmentation)
• Track fluorescence characteristics over time
• Average
• Localized concentration
• Analysis of fluorescence time-series

15
16
Software Image processing

• Cell Identification
• Flood-fill
• Locate plateaus
• Erode
• Label Cells
• Dilate
• Cells mapped between frames by calculation of
  intersectionunion
• Cell-body pixels transferred to fluorescence
  images for quantifying FP expression over time

17
Example GAL1-YFP

• GAL1 part of the galactose metabolism pathway
• Cells grown overnight in glucose
• Transferred to galactose
• Observed for 8 hours

Track expression in gt400 cells
18
Cell cycle regulators

• Obtained GFP-fusions of 17 regulators each in
  separate cell-line (Invitrogen)
• For each
• Observed 20 regions (500 cells)
• Collected Phase-Contrast Fluorescence Images
• Temporal Resolution 4 min

19
Budding yeast cell cycle
20
Regulating proteins Conventional wisdom
21
Cell cycle regulator CLN2
600 cells
22
Cell cycle regulator CLB2
23
Cell cycle regulator NET1
24
Cell cycle regulator CDC14
25
Cells without GFP
26
Further analysis Noise reduction

• Signal measurements generally have noise
• Instrument
• System that is measured (cells)
• Need to extract signal from noise
• Regression (curve-fitting)
• Filtering

26
27
Sine fitting

• Fit data to a sine function with variable
  amplitude, period, and phase
• Also include a cubic polynomial to account for
  trend

28
Data filtering

• Use Fourier analysis to extract frequency
  components of an oscillating signal
• Find noise floor (N), remove components below (2N)

28
29
Cell cycle regulator CLN2
30
Cell cycle regulator CLB2
31
Cell cycle regulator NET1
32
Cell cycle regulator CDC14
33
Cells without GFP
Sine fit
34
Quantifying localized concentration

• Find pixel with peak intensity
• Check if neighboring pixels are above threshold
• Iterate until no more pixels found above
  threshold or cell boundary is hit

35
Localization
CDC14
CLN2
CLB2
NET1
36
Bud-Bud fluorescence
37
Regression Bud-Bud
Phase
CLB2
CLN2
NET1
38
Limitations of current workflow

• 2 days between start of experiment output of
  image processing
• Limited to the use of 3 fluorescent proteins
• Limited number of cells
• Measure fluorescence rather than of proteins

38
39
Cyto?IQ
40
Features of Cyto?IQ

• Process images in real-time
• Returns statistical model
• Spectral imaging to allow for study of many
  proteins simultaneously
• Count Protein numbers
• Count mRNA

40
41
Features of Cyto?IQ

• Process images in real-time
• Returns statistical model
• Spectral imaging to allow for study of many
  proteins simultaneously
• Count Protein numbers
• Count mRNA

41
42
Spectral Imaging

• Traditional fluorescence imaging use filters to
  selectively excite/collect light from FPs
• Measure entire spectrum of fluorescence
• Unmix the contributions of each FP

43
Spectral Imaging
44
Features of Cyto?IQ

• Process images in real-time
• Returns statistical model
• Spectral imaging to allow for study of many
  proteins simultaneously
• Count Protein numbers
• Count mRNA

44
45
Stroboscopic Imaging

• Generally, photos captured with constant light,
  and opening/closing shutter
• Moving objects produce blur due to finite shutter
  speed
• To reduce blur
• decrease shutter time
• keep shutter open, flash light

46
Single-molecule counting

• Proteins randomly move through cell (D 10
  m2/s)
• Movement causes blur
• Excite fluorescence with intense, short laser
  pulse

47
Features of Cyto?IQ

• Process images in real-time
• Returns statistical model
• Spectral imaging to allow for study of many
  proteins simultaneously
• Count Protein numbers
• Count mRNA

47
48
Quantifying mRNA

• Probes designed to bind mRNA
• Series of 4 probes, each with 5 fluorescent tags
• Downsides
• Requires dead cells
• Each set of probes 1000

Zenklusen, et al, Nat Struct Mol Biol, 2008
49
Quantifying mRNA

• MS2 mRNA binding protein
• Express MS2-GFP fusion
• Target mRNA must be modified to include binding
  motif

Bertrand, et al, Mol Cell, 1998
49
50
Cyto?IQ Goals

• Count and track gt10 proteins and mRNAs over time
• Observe gt1000 cells
• Return statistical model to the user

51
Summary

• To fully understand gene expression, we must
  measure transcription and translation in
  individual cells
• Able to extract amplitude, period, and phase of
  the oscillations of cell-cycle proteins
• Current techniques have limitations such as
  number of proteins we can observe
• Future work will aim at counting expression of up
  to 10 genes simultaneously

51
52
Acknowledgments

• Synthetic Biology Group
• Jean Peccoud
• Julie Marchand
• Michael Czar
• Patrick Cai
• Matthew Lux
• Sarah Zheng
• Fred Cross
• (Rockefeller University)

John Tyson Kathy Chen William Baumann
NIH Grant 5R01GM078989-02
53
Composite movies

• CCD records in grayscale
• Sequentially collect bright-field image and
  fluorescence image
• Combine 2 images into the R, G, and B channels

53
54
Spectral Detection with NanoDrop 3300

• Collected spectra from 4 highly overlapping
  fluorescent proteins
• EGFP, acGFP, vYFP, and Citrine
• Made mixtures with known combinations, and
  unmixed the spectra

Fluorescent Protein Spectra in Living Cells
55
Spectral Unmixing
Input Spectrum
acGFP Spectrum
Citrine Spectrum
D
B
C
A

vYFP Spectrum
EGFP Spectrum

• Have m equations (spectral data points),
  and n unknowns (coefficients)
• If the reference spectra are known, then we can
  solve for the coefficients (A,B,C) that produce
  the input spectra
• Currently using a least-squares minimization

56
Preliminary Results

• Mixtures of cells with different fluorescent
  proteins

Mix 1
Mix 2
Mix 3
Mix 4
Mix 5