1

Next-Generation Cytomics: Spectral Fingerprinting

J. Paul Robinson

SVM Professor of CytomicsProfessor of Immunopharmacology

Professor of Biomedical Engineering

Director, Purdue University Cytometry LaboratoriesDeputy Director, Bindley Biosciences Center

Cell measurement technologies in ‘cytomics’

“…the systematic study of biological organization and behavior at the cellular level…

….has developed out of computational imaging and flow cytometry and promises to provide essential data for systems biology.”

2

The National Center for Applied CytomicsAn NIH Research Resource

Core 3: Biomolecular Design and Delivery

Core 1: Detection & Analysis

Core 2: Separation & Sorting

Increasing informationIncreasing homogeneity

Cellular Axis

Increased manipulation of cell/particle function

Molecular Axis

Cytomics Axis = IntegrationIncreased diagnostic utility Increased molecular knowledge

Increased opportunity for application & translation

Sabbatical visit program funded by the Center as well as postdoctoral opportunities

Purdue University

West Lafayette, IN

3

Biophotonic Projects• High speed spectral analysis of particles• Angle-resolved light scattering of particles• Whole animal spectral imaging• Bacterial classification by angular scattering

fingerprints• Spectral endoscopy using multimodal imaging• Calibrated fluorescence opthalmoscopy• Microdroplet delivery/manipulation systems

Microdroplet Generation & Manipulation for Drug Delivery

Water

WaterPLGAPLGA

Water

Controlledpressureregulationchanges

Additionof smallbeads

hydrogels

4

Focus the advanced capabilities offered by angle-resolved light scattering on

cytometry,

Sheath in

Piezoelectriccrystal oscillator

Sample in

Sheath

Flow cell

Particle

channel

Laser beam

32 ChSpectralDetector

• Collect 32 channels of spectral data

• Collect 20-32 channels of calibrated scatter data

• Use scattering “fingerprint” analysis to extract information on size, shape, and “refractive index” distribution

• 1000-3000 particles/sec for 64 channel data

8

Scatterometer >1500 scatter patterns from cultures of 108 Listeriastrains were measured and analyzed69 – L. monocytogenes16 - L. innocua12 - L. ivanovii5 - L. seeligeri3 - L. welshimeri3 - L. grayi

A

B

C

D

E

Robinson, Bhunia, Hirleman

Bacterial classification by angular scattering fingerprints

Purdue University Patent Pending

5

L. grayi LM37L. seeligeri LA 15 L. welshimeri ATCC35897

L. innocua F4248L. monocytogenes ATCC19113 L. ivanovii ATCC19119

Purdue University Patent Pending

Every organism has a very specific scatter pattern

Sal. enteritidis ATCC 13096

Sal. typhimurium Sal. typhimurium(Copenhagen)

Sal. typhimurium(Kentucky)

Enterobacter aerogenesEscherichia coli K12

Purdue University Patent Pending

Questions Currently being approached:•Can we determine antibiotic resistant strains from normals•Can we identify minor biochemical changes in a strain•How does age of colony alter classification accuracy•What is the earliest age colony one can measure•Can we do the classification in real time

6

MultimodeEndoscope

Purdue University Patent Pending

IF

λ

Multispectral imaging of tissue withsimultaneous white light imaging

Calibrated Fluorescence Opthalmoscopy

7

Whole Animal Spectral Imaging

• Use AOTF based imaging system with a variety of cameras & light sources

• Multispectral imaging:– Use of multiple fluorescent probes.– Discrimination of autofluorescence.

Cell Analysis Tools• Measurement of the optical properties of single

cells – Example: blood cells – separation of white blood cell

populations– Example: mixed bacterial suspensions –classification

by species– Example: multiplexed beads – identification of

analytes in fluids such as serum, CSF, cell culture medium, etc

– Example: functional analysis of cell populations –oxidative metabolism, enzyme, cell cycle analysis

8

Integration of Technologies• Redefine the fundamental basis for flow cytometry

design by integration of principles of chemical analysis and image analysis– Use a spectral analysis technology as opposed to a

fluorescence intensity profile analysis system– Implement capabilities for automated classification

IMPACT: • Next generation clinical diagnostics in real time• Fast classification for any complex system IF

λ

IF

λλ1 λ2 λ3

Hydrodynamically focused fluidics

•Increase Pressure:•Wider Core•Increased Turbulence

Signal

Cells pass though the chamber and are excited by the laser light. Scatter from this excitation is directed to detectors. A signal is collected for each cell and its characteristics are documented as shown on left.

LaserExcitationvolume

1-10 microsecondsper measurement

9

Optical Design of a basic flow Optical Design of a basic flow cytometercytometer

PMT 1

PMT 2

PMT 5

PMT 4

DichroicFilters

BandpassFilters

Laser

Flow chamber

PMT 3

Scatter

Sensor

Sample

Review of the principles that govern current flow cytometry

Signal (fluorescence) Intensity

Signal

Puls

e he

ight

Cou

nts

Dim redBright red

Single band measurement

10

Basic gating and analysispopulation identification

90 LS

FS

Side Scatter

Forw

ard

Sca

tter

Log-

PE

Variables 1 & 2 Variables 3 & 4

Multicolor (polychromatic) vs. multispectral cytometry (I)

• What is the difference between polychromatic and multispectral cytometry.

• Is it the number of colors?

excitation

emission

scatter

IF

λ

IS

α

IF

λλ1 λ2 λ3

IS

αα1 α2

excitation

emission

scatter

IF

λ

IS

α

IF

λλ1 λ2 λ3

IS

αα1 α2

11

12

Spectral Signature

Match Signature

Resolution of Data Point

One full spectra per particle

Single intensity as a parameter (usually) Spectrum as a

parameter

IF

λ

IF

λ

IF

λ

IF

λ

IF

λ

IS

α

IS

α

IS

αIS

α

IS

α

IF

λIF

λIF

λ

IS

α

IS

αIS

α

Multicolor (polychromatic) vs. multispectral cytometry (I)

P1 P2 P3…..

13

History of Multispectral ImagingA Purdue solution – Prof David A. Landgrebe

The Laboratory for Applications of Remote Sensing (LARS) • Launching of the first weather satellite, TIROS-1, on April 1, 1960 • In 1964, the U.S. Department of Agriculture's Economic Research Service

agreed to fund a small grant at Purdue to have data from the optical portion of the electromagnetic spectrum collected over Purdue agricultural sites by the Univ. Michigan equipment and analyzed at Purdue.

• LARS was formed in 1966 to take an interdisciplinary team approach to the creation and use of space-based technology for observing and managing of agricultural resources world wide.

– Goal to analyze land masses for agricultural uses from spacecraft altitudes– Major problem was spatial resolution too costly at the time

• It was thus decided to rely primarily upon the spectral response of the subject matter, thus creating a new type of spectroscopy.

• It became the basis for the LANDSAT series of Earth satellites and is the basis for a new generation of instruments to be carried on Space Station platforms in the 1990's.

Material Summarized from http://www.lars.purdue.edu/home/LARSHistory.html

Basic imaging…

Color imageMultispectralimage

Greyscaleimage

14

Color composition is a mixture of spectral bands

Proc. SPIE Vol. 4056, p. 50-64, Wavelet Applications VII, Harold H. Szu, Martin Vetterli; William J. Campbell, James R. Buss, Ed.

Images of soybean plants collected with

Landsat system

Images from LARS Laboratory at PURDUE

Long Island, NY

15

Multispectral detection toolsMultiple sets of optical filters Gratings and multianode PMTs

Multianode PMT withvariety of gratings

AOTF

High-resolution cytology segmentation

ConventionalRGB Image

Spectrallysegmented Image

Wavelength (nm)

CharacteristicSpectra

High spectral resolution increases utility of spectrally responsive indicator dyesSlide from Dr. Richard Levenson, CRi, Inc.,35B Cabot Rd.,Woburn, MA 01801, www.cri-inc.com

16

Spectral “unmixing”

505 515 525 535 545 555 565

FITC filter blockUnmixedDiOC6(3)And bis-oxonol

Dye 1: bis-(1,3-dibutylbarbituric acid)trimethine oxonol, DiBAC4(3)Dye 2: 3,3'-dihexyloxacarbocyanine iodide, DiOC6(3)

Green

Green

Before After

Why multispectral cytometry?

Traditional methodology of cytometry is univariate or bivariate - the intensity of fluorescence at one or two, three, etc., given excitation/emission wavelength pairs are used to identify a phenotype.

This type of analysis focuses on the variance of a single wavelength and does not take advantage of the information content of a complete spectrum.

Current cell analysis technologies

17

Advanced polychromatic

cytometry

8 PMTs and 16 filters

Figure from Roederer et al

Data processing in flow and imaging applications

Compensation: we can model the spreading of light among the color channels as a linear transformationLet the matrix, C, specify how the colors are spread among the three channels in a multi-color fluorescence application. Each element cij is the proportion of the brightness from fluorophore j that appears in color channel i of the digitized image.

• Spectral classification– PCA– LDA– …

• Spectral deconvolutionb][yCx

bCxy1 −=

+=−

• x is the n by 1 vector of actual fluorophore brightness values

• vector b accounts for the black level offset of the digitizer

Polychromatic Cytometry Multispectral Cytometry

18

Cell/system complexity can be reduced using tools such as flow cytometry

Figures from Roederer, et al

Multispectral Cytometry. Why?

• Identification of multiple spectrally overlapping stains (multiplexing)

• Spectral barcoding• Potential capability of spectral un-mixing

(multiple stains in a single particle)• Identification of intrinsic fluorescence

(autofluorescence classification)• Opportunity for intelligent systems

approach to classification

19

Multispectral cytometry – early designs• As early as in 1979 Wade, et al. reported recording the fluorescence spectrum

of particles in a flow system. The design of the instrumentation allowed only collection of integrated spectra from many particles (not from an individual particle) and assumed homogeneity of the sample.

• Steen and Stokke, 1986 measured fluorescence spectra of rat thymocytes, using a custom-built cytometer and grating monochromator

• In 1990 Buican used a Fourier transform interferometer to obtain single-cell spectra.

• Yet another design based on a flint-glass prism and intensified photodiode array was proposed by Gauci et al. in 1996. Again, the data rate of the instrument was too slow for a practical use. The same year Asbury et al. measured spectra of cells and chromosomes using a monochromator changing the wavelength during the course of a run

• SoftRay Inc. and a group of researchers from Wyoming and Utah Universities recently pursued another prism-based concept

Bottom line: None of these interesting technologies and design ideas have had any impact on development of widely used clinical or research-grade FC instruments available today

Multiplexing & spectral coding • FCM multiplexing, is based on various coding

procedures. In 1990 Robinson et al. developed a simple classification method using a combinatorial system. It allowed us to separate a number of clinical syndromes on a basis of the classification results of the flow cytometry data. The specific ratios of fluorescence stains gave rise to unique patterns, which we originally termed “phenograms” Robinson, J. Paul, Durack, Gary & Kelley, Steve: "An innovation in flow cytometry data collection & analysis producing a correlated multiple sample analysis in a single file". Cytometry 12:82-90,1991

A related approach was used by Luminex Corporation to build the FlowMetrix system. The most significant and original idea was a proposition to use microbeads as a solid support for optical tags.

20

Nanocrystals/Micro-Dots multiplexed systems

• New probes• Potentially 1000’s of

combinations• Sensitive, long lived,

less bleaching• Difficult to make• Will require some

advanced classificationCode:121 2112

Add

Wash

Code:121 1110

Code:111 1020

Code:121 2110

Nano-crystals – multicolor sets

Nano-crystals were generously supplied by Crystalplex

1 2 3

4 5 6

7 8 9

21

E. faecalisE. coliS. aureus

Example: identification of bacterial species by

fluorescence spectroscopy and PCA

“Rapid Identification of Bacterial Species by Fluorescence Spectroscopy and Classification Through Principle Components Analysis” by Héctor Enrique Giana, Landulfo Silveira Jr., RenatoAmaro Zângaro, Marcos Tadeu T. Pacheco, Journal of Fluorescence, Vol. 13, No. 6, November 2003.

PUCL multispectral cytometer

22

Multianode PMT – sensitivity and uniformity

Hamamatsu 32 Ch PMT

LatestPMT

Particle Spectrometer

Power supply and high-speed electronics for 32 channel collection(64 channel in new instrument)

Purdue University Patent Pending

Grating

Fiber to photon source

23

A 32-channel multispectral flow cytometer

Initial Spectral data

Multianode PMT – gain and spectral filtering

Now asimple4 colorcytometer

24

Basic flow Cytometry Systems

Spectral Overlap makes for very complex analysis

25

Overlapping Spectra

Calibration is accurate and against an easily obtainable calibration lamp($300 lamp is from Lightform, Inc www.lightform.com)

26

Principle of Operation

US & foreign patents pending

Analysis using “Multispec” software (Initially Principal Component Analysis)

Flow system

Imaging system

1 12 2

3

27

Spectra of gated zones

phycoerythrin Rh123

chlorophyll

Bang’s Labs, IncDyedBeads

PERh-123Chlorophyll

Bang’sLabsDyedBeads*

These CANNOT be resolved

Accept using advancedClassification tools like PCA

2 color analysis

28

Spectral analysis allows classification

Fluorescence intensity (QD 525) [A.U.]500 600 700 800 900

Fluo

resc

ence

inte

nsity

(QD

575

) [A.

U.]

600

700

800

900

1000

Fluorescence intensity (QD 525) [A.U.]500 600 700 800

Fluo

resc

ence

inte

nsity

(QD

575

) [A

.U.]

600

700

800

900

1000

0 50 100 150 200 250 300 350 400 450 500 550 600 650 700

Principal component 1-0.15 -0.10 -0.05 0.00 0.05 0.10

Prin

cipa

l com

pone

nt 2

-0.05

0.00

0.05

0.10

0.15

600

700

800

900

1000

Cha

nnel

18

500 600 700 800 900 1000Channel 27

2 color system 32 channel system

Develop signal processing tools suitable as a foundation for clinical instruments

Analysis of complex samples (mixed nanocrystals)

29

Problems?• Problems with dispersion elements• Low sensitivity - narrow bandwidth = high

spectral resolution AND low numbers of photons• Data collection – relatively expensive electronics

is required at present• Data analysis – traditional methods fail. • Quantitation is no doubt more complex• Standards development will be required

How do we bring clinical problems to the table?

20

60

100

Enrico Lugli et al, Università di Modena e Reggio EmiliaOral Presentation Immunology(Group of Andrea Cossarizza, University of Modena)

Classification ?

30

Conclusions• Spectral cytometry may fundamentally change the current

concepts of cytometry• Spectral cytometry may be a far better alternative for

some specific applications requiring classification• Spectral cytometry can separate probes of “apparently”

the same emission (eg FITC, DIOC5, DIOC6, Oxonol, etc)• There is less dependence on differences in intensity as

there are in the spectral changes• Traditional spectral compensation (hard to do) is not used• There are complex problems to solve mathematically• Much larger number of fluorochromes can be separated

using the spectral cytometer (???)• May be the choice for clinical diagnostic instrumentation

Acknowledgements:

BiophysicistComputer Scientist Coffee

maker

Not shown: Peng Xi, China; Dominik Lenz, Germany)Tytus Bernas, Poland; Sang Youp Lee, S. Korea)

31

AcknowledgementsContributors – Collaborators

Dan Hirleman (ME)Yinlong Sun (CS)Kinam Park (IP)Eli Asem (BMS)Sophie LeLevrie (BMS)Arun Bhunia (FS)Gerald J. Gregori (CNRS)

Senior ScientistBartek Rajwa (Biophysics)

PostdocsValery Patsekin (Photonics)Tytus Bernas (Proteomics)Sang Youp Lee (Fluidics)Dominik Lenz (Cytometry)Krishna Mishra (Molecular Biology)

Graduate StudentsWamiq Ahmed (Computational)Murugesan Venkatapathi (M.Eng)Bulent Bayraktar (Computational)Silas Leavesley (BioEng)Jia Liu (Cell Biology)Connie Paul (Pharmaceuticals)

StaffJennie Sturgis (Imaging)Kathy Ragheb (Flow)Cheryl Holdman (Flow)Gretchen Lawler (CPC)Steve Kelley (Network)

Departments & CentersPurdue University Cytometry LabsBasic Medical ScienceBiomedical Engineering Electrical & Computing EngineeringMechanical EngineeringBindley Bioscience CenterDiscovery ParkPurdue Cancer Center

Funding:NIH, NSF, USDA, Purdue UniversityCorporate: Beckman-Coulter, Becton Dickinson, Point-Source, Parker-Hannifin, Bio-Rad,Polysciences, Bangs labs, MediaCybernetics,