proliferation tutorial

16
FlowJo Data Analysis Software for Flow Cytometry PC User Documentation Proliferation Tutorial

Upload: arghyashree-roychowdhury

Post on 02-Mar-2015

455 views

Category:

Documents


8 download

TRANSCRIPT

Page 1: Proliferation Tutorial

FlowJo Data Analysis Software for Flow Cytometry

PC User Documentation

Proliferation Tutorial

Page 2: Proliferation Tutorial

2

FlowJo was written by Adam Treister and Mario Roederer beginning in 1996,

based on concepts developed at the Herzenberg laboratory at Stanford. We are

indebted to our active and enthusiastic users worldwide for their ideas, discussions

and tireless testing of new versions.

FlowJo, its tutorials, documentation and web site are Copyright © Tree Star, Inc.

1997-2011. All Rights Reserved.

• FlowJo Advanced Tutorial •

• © MMX•

Revision Date: 1 January 2011

Version 7.6.2

Page 3: Proliferation Tutorial

3

FlowJo Proliferation Tutorial

FlowJo is a software application designed to be a comprehensive tool for analyzing

flow cytometric data. Because proliferation is common task in cytometric analysis,

FlowJo includes a platform for applying mathematical models to cell cytometric

data fully integrated into the software.

This tutorial is designed to instruct you on how to use the FlowJo proliferation

platform to analyze proliferation data. A brief overview on the biology of cell

proliferation is provided mainly for the purpose of defining the terms, but this

manual does not include a comprehensive discussion. This tutorial is also written

with the assumption that the user is familiar with the most basic operations in

FlowJo, such as loading data, making simple gates, dragging analysis nodes to

other samples groups, or the layout and table editors. If you are not familiar with

these operations, a general tutorial of FlowJo is available at www.Flowjo.com as

well.

This tutorial is designed to introduce you to the proliferation platform. Reading

through it, you will learn how to operate FlowJo. Run the program as you perform

the steps in the tutorial so that you can get the best feel for how the program

works! Included with the demo data are a series of completed workspaces that will

enable you to jump into the tutorial at any point and follow along.

As a note, we are pleased to be able to frequently update FlowJo to provide new

features & analysis capabilities. Therefore, it is possible that the graphics shown in

this tutorial may not exactly match the windows that you see when you run the

most recent version of FlowJo. You can always download the most recent version

of FlowJo from http://www.flowjo.com/home/windows.html.

Page 4: Proliferation Tutorial

4

Table of Contents

Introduction ..................................................................................5

Proliferation Basics......................................................................5

The Experimental Data ................................................................5

Lesson 1: The Proliferation Platform .......................................6

Creating a New Model .................................................................6

Proliferation Statistics in FlowJo.................................................7

An Example Summary Statistics Calculation ..............................8

Lesson 2: Applying Constraints................................................9

Fixing the Undivided Mean .........................................................9

Fixing the Ratio............................................................................10

Fixing the CV...............................................................................10

Fixing the Background.................................................................10

Creating Gates..............................................................................11

Lesson 3: Creating Proliferation Outputs ................................12

Applying the Model to other Samples .........................................12

Creating a Table of Proliferation Statistics..................................12

Creating a Layout of Proliferation Graphics................................13

Appendix ..................................................................................14

Resources ..................................................................................15

Page 5: Proliferation Tutorial

5

Introduction

This tutorial will guide you through the process of applying and modifying

mathematical models to proliferation data in FlowJo, from adding data to

producing tables of statistics and publication quality graphics.

The tutorial is divided up into 5 sections so you can perform it piecewise if desired.

The tutorial is written for the user to perform all operations, but the completed

workspaces represent the outcome of the operations performed in each chapter so

that you can jump to any stage of the tutorial. If you would like to perform the

tutorial starting with a lesson other than 1, just open the workspace from the

preceding lesson and you will have all of the work completed through that lesson.

Proliferation Basics

The Proliferation assays serve as powerful tools to understand the functionality of

different cell types. While phenotypic characteristics provide the first level of

analysis, it is only through the functional assays that one can fully understand the

capacity of these different phenotypes of cells to respond or activate to different

conditions.

There are a number of dyes used in cell proliferation assays, with CFSE being the

most common. Several companies have introduced new dyes in recent years.

However, all of the dyes bind to lipids or proteins on the cell membrane, and so the

mechanism is the same. When a cell divides, each daughter cell gets

approximately half of the initial bound dye, and when measured with flow

cytometry, the daughter progeny have half the fluorescence intensity compared to

the parent.

The Experimental Data

The data for this tutorial is a set of seven files of peripheral blood mononuclear

cells; one unstimulated control and six files at subsequent time points labeled

Sample 1 through 6. Each file is stained with a panel of Foxp3 GFP, CD25 PE, 7-

AAD, CD4 PECy7, and a protein binding dye from eBiosciences called eFluor®

670. This dye is excited at 647nm and emits at around 660nm.

Page 6: Proliferation Tutorial

6

Lesson 1: The Proliferation Platform

This lesson will guide you through the basic use of the proliferation platform.

Creating a New Model

1) Begin the lesson by opening the initial workspace, Prolif_Lesson_1.wsp. The

primary gating has already been done.

2) In sample 1, click on the ‘new live’ gate and right click.

3) Select Proliferation from the pop up menu.

A proliferation node will be

created under the new live

gate and the proliferation

interface will open. FlowJo

will attempt to determine

which parameter is the

proliferation marker. If the

correct parameter is not

displayed by default, click

the drop down parameter list

below the x-axis and select

the correct parameter.

Set the parameter to Comp-

Red_600_20_APC. Once

you have done so, your

platform will look like the

graphic to the right.

4) The default number of peaks is set to eight. This is arbitrary, so the next step is

to set the number of peaks to the appropriate number based on the data. To do so:

• Open the Options disclosure triangle at the bottom of the proliferation

platform interface.

• Count the number of peaks to get a starting estimate. By “eyeball metrics”

there are five peaks in this data file.

Page 7: Proliferation Tutorial

7

Note: Inputting the “correct” number of peaks is not critical. If you overestimate,

FlowJo will still attempt to fit peaks in at roughly half the intensity of the

preceding peak, and will include sparsely populated additional populations. If you

underestimate, FlowJo will not have enough peaks to fit the data which will

produce a poor model. As a rule of thumb is to count the number of peaks you can

see and then add one. So begin this tutorial with 6 peaks.

Proliferation Statistics in FlowJo

FlowJo keeps track of descriptive statistics regarding the model and a series of

summary statistics that can be used to summarize the

proliferation of the cells from the time of staining to the time of

collection. The statistics that FlowJo uses are shown in the

figure to the right.

The descriptive statistics are the number of peaks, the CV, or

width, of the peaks, and the location of the undivided peak mean.

By identifying the number of generations and the count in each

generation, FlowJo uses the model as a basis to calculate a series

of statistics that allow for the end result to be summarized in a

number or two. The summary statistics are the percent of cells

that divided at least once, the division index, the proliferation

index, and the mean root mean square error.

• Percent Divided is how many cells divided at least once.

• Proliferation Index is the average number of divisions of just the responding

cells (cells that underwent at least one division).

• The Division Index is the average number of divisions for all of the cells in the

original starting population.

• mRMS is an acronym for mean root mean square error. The distance of the

composite model line from the histogram of the data is calculated, squared, and

then the square root is taken. The mean is taken over all of the peaks. We use

this process so that portions of the model above the actual data (producing a

positive distance) do not cancel out the error on portions of the model below

the actual data (producing a negative distance), resulting in a low RMS and a

terrible model. Since RMS is a measure of the distance from the model to the

Page 8: Proliferation Tutorial

8

data, a smaller RMS indicates a better fit. RMS is not appropriate to compare

between experiments, as the experimental condition will play an important role

in determining what a good RMS is, making between experiment comparisons

a case of comparing apples to oranges. RMS can be used to determine which

model fit the data tighter, and whether a constraint improved the fit or not.

An Example Summary Statistics Calculation

Consider a culture of 1 million cells where upon stimulation, 50% of the cells each

divide twice. You will now have a culture with 2.5 million cells (0.5 million never

divided, 0.5 million divided twice resulting in 2 million cells). CFSE analysis

would show 2 peaks: an undivided, no cells in the first generation, and cells in the

second generation. The CFSE platform would return the following statistics:

Percent Divided: 50% 50% of the original cells divided

Proliferation Index: 2 Responding cells divided twice on average

Division Index: 1 Half the cells divided twice, half never divided for

an average of 1.

Note: the following is always true: (DI) * (%Div) = PI

Thus, of these three values, there are only two independent measurements.

The ends lesson 1. The work to this point is saved as Prolif_Lesson_2.wsp.

Page 9: Proliferation Tutorial

9

Lesson 2: Applying Constraints

The models within FlowJo will not fit all data files properly without additional

input by an expert user. There are many reasons for this, but a common cause is

that the models are initialized to fit the “standard” shape that we’ve seen so far,

and many experiments perturb the cell cycle and cause the data to deviate from the

standard form. This lesson will teach you to use constraints to limit how the

model can fit the data to an appropriate solution. By imparting some of your

expert knowledge of the biological system onto the model, you can improve the fit.

To begin the tutorial from this point, open Prolif_Lesson_2.wsp.

Fixing the Undivided Mean

You can determine from the undivided sample that Generation 0 is at 37,600

fluorescence units. In the workspace right click on the ‘new live’ gate under the

undivided sample, select the Proliferation model from the Tools menu and change

the peak count to 1. You will get a model like the one pictured below to the left.

Return to Sample 1 and check the box for ‘Fix Peak 0’. The data entry box will

enable. Enter 37,600. The model will now look like the figure above to the right.

Page 10: Proliferation Tutorial

10

The RMS can often be lowered by constraining the model. In this case the RMS

dropped to 1.8, indicating that this was a good constraint to apply.

Fixing the Ratio

Checking the Fixed Ratio box and entering a value sets the fluorescence ratio

between peaks to the number you entered. The standard starting point is 0.5,

indicating perfect conservation of the proliferation dye with 50% of the dye going

to each daughter cell. For example, if the MFI of Generation 0 is 200, than the

MFI of Generation 1 will be 100 fluorescence intensity units. The ratio will

usually be less than 0.5, as cells typically loose some dye during the division

process. In the example we are using, the ratio was calculated at about 0.56,

indicating the background fluorescence wasn’t properly modeled (which can

happen sometimes). A ratio of greater than 0.5 is biologically impossible. It

implies that both daughter cells received more than half of the dye.

Click the Fixed Ratio box and enter 0.5 in the tutorial. Notice that the RMS

decreases to roughly 2.9, indicating that this is an improvement to the model, and

biologically possible compared to the 0.56 that the model initially generated. You

can try varying the input number to see if another ratio produces a better fit.

Fixing the CV

Fixing the CV sets the coefficient of variation in the distribution, which is the

width of each population. Typically, the CV of the undivided peak is a good place

to start. The unstimulated sample had a CV of 2.7. Fix the CV to match the

unstimulated as well. You will notice that the RMS now drops to 1.95, again an

improvement.

Fixing the Background

Fixing the Background sets the amount of fluorescence to be subtracted as

background from every cell. The model assumes that fluorescence of each

generation is equal to the fluorescence of the previous generation times the ratio,

adjusted for background noise. Expressed mathematically the fluorescence is:

F(n) = [ F(n-1) - B] * r + B.

F(i) = fluorescence of the ith

generation; B = background, r = ratio.

Page 11: Proliferation Tutorial

11

The background was set

properly by the model so

for this experiment, we

do not need to adjust it.

Try constraining it and

observe that a constraint

negatively impacts the

model, resulting in a

higher RMS.

The final model will look

like the figure to the

right.

Creating Gates

Press the Create Gates

button at the top of the

workspace, and accept

the “Generation” prefix.

It is important to note

that when you press the

Create Gates button

FlowJo places hard line

gates on the model at the

points where it becomes less probable that a cell belongs to one population and

more probable that it belongs to another. Cells that are in the gate are treated as

having a probability of membership of 100% for the “winning” phase and 0% for

all other. This is a different manner of calculating the frequencies and thus the

statistics that display in the workspace next to created gates will usually be slightly

different then the statistics displayed in the model. If you have the opportunity to

use the model statistics, do so. The stochastic approach models the data better.

However, if you take the proliferation model created subpopulations, continue the

analysis, and in the end need to come up with fractions that add up to 100%, you

will need to use the gate frequencies.

Note: These data were chosen because the samples need modification to fit

properly. Many data sets do not need any modification (other than the peak

setting) and will fit perfectly.

The ends lesson 2. The work to this point is saved as Prolif_Lesson_3.wsp.

Page 12: Proliferation Tutorial

12

Creating Proliferation Outputs

When satisfied with the proliferation model you can easily take advantage of all

the other basic tools within FlowJo for expediting analyses. In this chapter we will

cover applying a proliferation model to other samples, creating tables of

proliferation model statistics, and creating layouts of proliferation graphs.

To begin the tutorial from this point, open workspace Prolif_Lesson_3.wsp.

Applying the Model to other Samples

The cell cycle model or the gates created using the model, like any other analysis

node in FlowJo, can be dragged and dropped to other levels of the hierarchy, to

other samples, or to the group. Drag the proliferation node up to the group under

the ‘new live’ gate. This will apply the proliferation model to every sample.

Of course, you will have to go back and adjust the model for the unstimulated

sample to have 1 peak, but scroll through the other samples and notice that they

look pretty good!

Creating a Table of Proliferation Statistics

Drag the proliferation node into the table editor to place all the statistics in a table.

The resulting table editor will look like the figure below.

Page 13: Proliferation Tutorial

13

Feel free to delete any of the statistics that are not necessary. If you batch, you’ll

get those statistics for all samples.

Creating a Layout of Proliferation Graphics

Drag the proliferation node into the layout editor to place all the model graphics

into a layout. The resulting layout will look like the figure below.

The workspaces completed to this point are saved as Prolif_Lesson_4.wsp.

Page 14: Proliferation Tutorial

14

This ends the tutorial. There is more documentation available in the reference web

pages, which you’ll reach from any of the help menu items within the program, or

by looking at: http://www.flowjo.com/

If you have any questions, or ideas for improvements, please contact us at:

[email protected]

FlowJo Proliferation Tutorial and Web Site are Copyright © Tree Star, Inc. 1997-

2011.

Revision Date: January 1, 2011 Version 7.6.2

Page 15: Proliferation Tutorial

15

Appendix

Equation used:

Sum (i * Ni / 2^i) / Sum (Ni / 2^i); where Ni = number of events in peak i.

For example:

The Division Index is the number of number of divisions that took place during

culture divided by the number of cells at start of culture.

The Proliferation Index is the number of number of divisions that took place

divided by the number of cells of the original population that went into division.

The number of cells that you had at start of culture is:

G0 + G1/2 + G2/4 + G3/8 + ... + Gn/(2^n)

The number of cells that went into division is the number of cells that you had at

start of culture minus G0.

The total number of divisions is:

G1/2 * 1 + G2/4*2 + G3/8*3 + G4/16*4 + ... + Gn/(2^n)*n

G0 = 15888

G1 = 32922

G2 = 13647

G3 = 897

The number of cells at start of culture:

15888 + (32922/2) + (13647/4) + (897/8) = 35872.87

The total number of divisions:

(32922/2)*1 + (13647/4)*2 + (897/8)*3 = 23620.87

The number of cells that went into division:

35872.875 - 15888 = 19984.875

Division Index: 23620.875 / 35872.875 = 0.66

Proliferation Index: 23620.875 / 19984.875 = 1.18

Page 16: Proliferation Tutorial

16

Resources

http://www.flowjo.com/v9/html/proliferation.html

http://www.flowjo.com/v76/en/proliferation.html

http://www.flowjo.com/v76/en/prolifmodeladjust.html

http://flowjo.typepad.com/the_daily_dongle/2007/05/dongleoids_inde.html