Embed Size (px)
Citation preview
Adapta&on and Design of Panels for the CyTOF
December 5, 2014 CyTOF User Mee&ng – University of Virginia
Mike Leipold
Fluorophores
Two Different Approaches to Simultaneous Detec&on
Elemental Tagging
Fluorometer Induc&vely Coupled Plasma Mass Spectrometry
(ICP-‐MS)
0
10
20
30
40
50
60
70
80
90
100
380
403
426
449
472
495
518
541
564
587
610
633
656
679
702
725
748
771
794
wavelength
intensity
8 Alexa Fluorophores
Two Different Approaches to Simultaneous Detec&on
Elemental Tagging
LSR II CyTOF
Overview
1. CyTOF – the machine, pros and cons 2. Panel development – background, desired signals 3. Experimental procedures – an&bodies, reagents 4. Running samples – example data
-‐ LSRII vs CyTOF -‐ Flu 2010, 2011
Part 1 – The CyTOF
Single Cell Analysis -‐ CyTOFTM Machine
Flow
Figure courtesy O. Ornatsky
Flow Interface to ICP-‐MS -‐ Nebulizer – vaporizes -‐ Argon Plasma – atomizes
and ionizes
-‐ Mass analyzer
Published in: Dmitry R. Bandura; Vladimir I. Baranov; Olga I. Ornatsky; Alexei Antonov; Robert Kinach; Xudong Lou; Serguei Pavlov; Sergey Vorobiev; John E. Dick; Scob D. Tanner; Anal. Chem. 2009, 81, 6813-‐6822. DOI: 10.1021/ac901049w Copyright © 2009 American Chemical Society
Single Cell Analysis -‐ CyTOFTM Machine
Monocyte?
B cell?
CD8+ T cell?
-‐”Push” analogous to &me: 76,800 pushes/sec: 220 displayed here (1/400 of data)
Display of Cell Events
-‐ Advantages: -‐ Minimal "spectral" overlap – higher dimensionality (more probes at once)
-‐ Quan&ta&ve – broad dynamic range
-‐ Not light-‐ or &me-‐sensi&ve
-‐ Minimal background -‐ no analogy to autofluorescence: low/nil biological background for lanthanides
Pros of CyTOF
-‐ Disadvantages: -‐ Destruc&ve: (currently) no way to recover interes&ng cells
-‐ Slower: limit of 1000 cells/sec; prac&cal limit for best resolu&on oken in ~400 cells/sec range
-‐ Cell transmission efficiency: only ~20-‐30% of cells that enter machine get counted
-‐ Ion transmission efficiency: only ~1 in 10,000 ions that enter machine get counted
Cons of CyTOF
-‐ Disadvantages: -‐ Postprocessing: cannot set “on fly” gates to get only “live
intact singlets” -‐ Events registered by CyTOF contains debris, doublets, etc -‐ Might only get ~50% of total events as “live intact singlets”
-‐ No analogy to FSC or SSC: MUST have marker (Mn+) for any ga&ng
Cons of CyTOF
Single Cell Analysis -‐ CyTOFTM Machine
-‐ CyTOFv1 – ~2010-‐May 2013 -‐ Mass window ~93 units: eg, AW 103-‐195
-‐ CyTOFv2 – May 2013-‐current
-‐ Mass window ~135 units: eg, AW 78-‐212 -‐ More automated tuning, including recordkeeping -‐ Improvements in ion op&cs -‐ &ghter peaks, less "spillover"
Part 2 – Panel Design
Relevant Issues to Panel Design
1. Background – any non-‐desired signal -‐ contamina&ng signals -‐ an&body &ter (not discussed)
2. Desired Signal
-‐ an&body clone -‐ metal label – analogous to fluorophore choice
Relevant Issues to Panel Design – Background Signals
* Desired signal only at Mass "M" 1. M+16 – oxide – cannot eliminate, can only limit 2. Metal salt impuri&es – M+1, M-‐1, etc; Ln 3. Environmental contamina&on – sample 4. Instrument * All poten&al sources of background, even spillover!
* Desired signal only at Mass "M" 1. M+16 – oxide – cannot eliminate, can only limit
-‐ Proper daily machine tuning -‐ Lower AW Ln = worse
Relevant Issues to Panel Design – Background Signals
Daily Tuning – Liquid Tuning Solu&on
Current profile Make-‐up Gas profile
DVS Elemental Beads – "Cells" -‐ Polystyrene beads: contain nat. abund. Eu; or La, Pr, Tb, Tm, Lu -‐ Act as "cells" run in acquisi&on mode (vs. tuning/solu&on mode)
0 101 102 103 104
(Pr141)Dd: (Pr141)Dd
0
101
102
103
104
(La1
39)D
d: (L
a139
)Dd
0 101 102 103 104
(Tm169)Dd: (Tm169)Dd
0
101
102
103
104
(Tb1
59)D
d: (T
b159
)Dd
Makeup Gas Flow Rates - Median Dual counts vs AW
140 150 160 170 1800
500
1000
1500
2000
Atomic MassM
edia
n D
ual C
ount
MG = 0.74
Effect of Tuning – Beads -‐ Oxida&on
Median Dual counts
0.5 0.6 0.7 0.8 0.9 1.00
500
1000
1500
2000
Makeup Gas Flow Rate (L/min)
Med
ian
Dua
l Cou
nt
La139 Pr141Tb159Tm169 Lu175 "Gd155" "Gd157" "Re185"
Median Dual counts
0.5 0.6 0.7 0.8 0.9 1.00.01
0.1
1
10
100
1000
10000
Makeup Gas Flow Rate (L/min)M
edia
n D
ual C
ount
La139 Pr141Tb159Tm169 Lu175 "Gd155" "Gd157" "Re185"
Linear Scale Log Scale
-‐ At high MG flow rate, oxida&on can be significant (decreases M, increases M+16 spillover)
-‐ Even a few percent can be significant for high-‐abundance markers (eg, CD57, histone proteins, CD45, etc)
* *
* Desired signal only at Mass "M" 2. Metal salt impuri&es
a) Few lanthanides are naturally 100% monoisotopic -‐ e.g., Nd144 signal in Nd145 salt -‐ M+1, M-‐1, etc
b) Other lanthanide contamina&ons – all Ln chemistry is similar -‐ La139 most common contaminant
Relevant Issues to Panel Design – Background Signals
Purity of Isotopes – Liquid Stock Solu&ons
CyTOF Channel
Single-‐Isotop
e Sample
-‐ M+1 impuri&es – some isotopes worse than others -‐ M+16 oxida&on impuri&es – affects all isotopes; lower AW worst
BD Ig Κ Capture Beads and MAXPAR An&bodies
0 101 102 103 104
IgD(Nd146)Dd: IgD
0
1000
2000
3000
4000
# C
ells
0 101 102 103 104
CD85j(Sm147)Dd: CD85j
0
500
1000
1500
# C
ells
0 101 102 103 104
Nd148(Nd148)Dd: Nd148
0
100
200
300
400
500
# C
ells
Sm147 148 146
No "Sm146" -‐ 146 is Nd
M M+1 M-‐1
* Spillover due to elemental and/or isotopic impuri&es, not strictly mass proximity
Major Contaminants/Spillovers: Percentage of M
-‐ M+1 and M+16 are usually the major spillovers -‐ Only Fluidigm metals shown; non-‐Fluidigm metals can be worse
* Ir191 contam?
Major Contaminants/Spillovers: Cell Data
* Formerly, Fluidigm sold Dy162 and Dy164, but not Dy163 – Dy purity issues
Regular Phenotyping panel: CD45RA-‐Dy162 CD20-‐Dy164 * no Dy163 an&body in experiment, so not usually monitored
Major Contaminants/Spillovers: Cell Data
CD45RA-‐Dy162 CD20-‐Dy164 * no Dy163 an&body in experiment!
0 101 102 103 104
(Dy163)Dd
0
20
40
60
80
100
% o
f Max
CD3+
0 101 102 103 104
(Dy163)Dd
0
20
40
60
80
100
% o
f Max
B cells
Magenta: both Dy162, Dy164 Blue: Dy162 only Red: Dy164 only
-‐ no CD20 -‐ high CD20, high CD45RA
3. Environmental contamina&on
-‐ Reagents – PFA lots with micropar&cles, MeOH with free metal, Ar gas with Sn
-‐ Biological sourcing – Iodine (free I vs IdU), Pt in cispla&n-‐treated donor samples
-‐ Lab dust
-‐ Barium, lead, etc – dish soap, syringes, reagent sourcing, striker flint, etc
Relevant Issues to Panel Design – Background Signals
Relevant Issues to Panel Design – Background Signals
4. Instrument Abundance Sensi&vity – lek (M-‐1) and right (M+1) leg of ion peak -‐ CyTOFv1: 0-‐1% -‐ CyTOFv2: <0.3%
0 101 102 103 104
(Sm147)Dd: CD85j
0
100
200
300
400
# C
ells
-‐ Worse with high AW masses: peak less Gaussian, M+1 leg wider
Major Contaminants/Spillovers: Percentage of M
* CyTOF spillover -‐ s&ll much less than Fluorescence spillover! * All propor&onal to signal at Mass "M"..... 1. Match marker abundance with metal brightness
-‐ 1% of 300 is 3 (background) -‐ 1% of 3000 is 30 (probably not background)
2. Mutually exclusive lineages
-‐ T cell marker followed by B cell marker 3. Dump channels for most contaminated isotopes
-‐ analyze the "nega&ve" cells
Desired Signal Intensity-‐ Choice of Metal
Reminder: you cannot detect anything that doesn't have bound metal -‐ No scaber to gate monocytes from lymphocytes, etc
-‐ Only about 3-‐fold difference in signal across lanthanides -‐ Switching lanthanides might help only in borderline cases
Higher sensitivity (165-170)
Lower sensitivity (139-150)
0 101 102 103 104
CD3(Nd150)Dd: CD3
0
500
1000
1500
# C
ells
0 101 102 103 104
CD3(Tm169)Dd
0
500
1000
1500
2000
# C
ells
0 101 102 103 104CD16(Ho165)Dd
0
101
102
103
104
CD56(Yb174)Dd
0 101 102 103 104
CD16(Sm149)Dd: CD16
0
101
102
103
104
CD
56(Y
b174
)Dd:
CD
56
Desired Signal Intensity-‐ Choice of Metal
1. "Dim" metals – In, La, Pr, Nd, Sm -‐ Bright/abundant markers (eg, CD57) -‐ True bimodal (eg, CD27)
0 101 102 103 104
CD27(Sm152)Dd: CD27
0
200
400
600
800
# C
ells
0 101 102 103 104
(In113)Dd: CD57
0
1000
2000
3000
# C
ells
CD8+ CD4+
Desired Signal Intensity-‐ Choice of Metal
2. "Bright" metals – Tb, Dy, Tm, Yb -‐ Dim/rare markers -‐ "Smear" distribu&on (eg, CD45RA, CCR7) -‐ Small fold-‐change
0 101 102 103 104
CD45RA(Dy162)Dd: CD45RA
0
50
100
150
200
# C
ells
CD8+
0 101 102 103 104
CD3(Nd150)Dd: CD3
0
101
102
103
104
TCR
gd(Y
b171
)Dd:
TC
Rgd
TCRgd3.29
Desired Signal Intensity-‐ Choice of Metal
3. Recommend bivariates -‐ one bright, one dim -‐ two mediums – "smear", or subpopula&ons
0 101 102 103 104
CD19(Nd142)Dd: CD19
0
101
102
103
104
CD
20(D
y164
)Dd:
CD
20
B cells48.2
0 101 102 103 104
(Sm154)Dd: CD14
0
101
102
103
104
(Er1
66)D
d: C
D33
CD14+ CD33+14.9
CD14- CD33-84.8
Desired Signal Intensity-‐ Choice of Metal
0 101 102 103 104
(Sm147)Dd: CD85j
0
1000
2000
3000
# C
ells
CD85j+3.07
CD85j-96.9
0 101 102 103 104
(Sm147)Dd: CD85j
0
500
1000
1500
2000
# C
ells
CD85j+17.9
CD85j-82.1
0 101 102 103 104
(Sm147)Dd: CD85j
0
50
100
150
200
# C
ells
CD85j+96.2
CD85j-3.85
CD4+ CD8+ Monocytes
"nega&ve" "posi&ve" subpopula&on
Desired Signals – CD85j as a Borderline Case
Confirm New Markers With More Than One Donor
TCRgd
CD33
0 101 102 103 104
(Nd150)Dd: CD3
0
101
102
103
104
(Yb1
71)D
d: T
CR
gd
TCRgd3.08
0 101 102 103 104
(Nd150)Dd: CD3
0
101
102
103
104
(Yb1
71)D
d: T
CR
gd
TCRgd0.377
0 101 102 103 104
(Sm154)Dd: CD14
0
101
102
103
104
(Er1
66)D
d: C
D33
CD14+ CD33+16.9
CD14- CD33-82.7
0 101 102 103 104
(Sm154)Dd: CD14
0
101
102
103
104
(Er1
66)D
d: C
D33
CD14+ CD33+8.66
CD14- CD33-91.1
Part 3 – Experimental Procedure
CyTOF Sample Info
1. Reagent purity is paramount -‐ MilliQ water only -‐ trace metal pure salts and buffers, if possible
2. All samples -‐ fixed and permeabilized
-‐ Thorough fixa&on -‐ Fresh PFA! 3. MilliQ water washes at the end
-‐ Proper fixa&on ensure intact cells -‐ Minimizes buffer salt introduc&on into machine
4. Titrate all reagents for your assay condi&ons
-‐ Commercial -‐ Every batch of in-‐house reagents – re-‐verify before use
CyTOF Sample Info – Star&ng Cell Count
* Remember: only 20-‐30% cell transmission efficiency -‐ also, processing losses
Titration of cell numbers - Plated vs Live Intact Singlets
0 500,000 1,000,000 1,500,0000
50,000
100,000
150,000
Cell #, plated
Live
inta
ct s
ingl
ets
SlopeY-intercept when X=0.0X-intercept when Y=0.0
Live intact Singlets0.1829 ± 0.01088-14297 ± 690078149
R squareLive intact Singlets0.9895
Cells Plated
Live intact singlets
200K 28K
500K 76K
1M 175K
Yao et al: down to 10K (major only) HIMC: recommend at least 500K
Metal Staining of Cells -‐ DNA Intercalator
Schaferet al J.Organomet.Chem., 2007, 692, 1300−1309 Ornatsky et al, Anal. Chem., 2008, 80, 2539–2547
-‐ Labels any cell containing DNA regardless of whether any labelled an&bodies bind
Chela&ng Polymer Structures
DN3 branched dendrimer, ~16 chela&on sites
X8
linear polymer with ~22 chela&on sites
An&bodies – X8 vs DN3 Comparison
Live-‐Dead
maleimide-‐DOTA cispla&n
Normaliza&on – EQ Four-‐Element Beads
-‐ Polystyrene beads: contain elemental Ce, Eu, Ho, Lu -‐ burn like cells, but are defined composi&on
Normaliza&on Beads -‐ Long Run&me Decrease in Signal Intensity
10:33 AM 6:55 PM
Normaliza&on Affects All Signals Noise amplifica&on!
* Global Mean as baseline, rather than Minimum or Maximum -‐ limits noise amplifica&on
Normaliza&on CANNOT correct for improper machine cleaning/tuning (undetectable analyte signals).
Finck et al, Cytometry A, 2013
Signal Decrease Over Very Long Run-‐Times
73% of start
98% of start
-‐ Single file: >2 million cells = >3 hr
* Highlights need for moving window normaliza&on Finck et al, Cytometry A, 2013
Effects of Normaliza&on
* Freq Parent fairly robust; only small gains from normaliza&on
Freq Parent Median Intensity
* Median Intensity more affected by normaliza&on (phosphoflow...)
-‐ Control sample split in half, run 8.5hr apart; ra&o aker manual ga&ng
Part 4 – Sample Running
Running Samples
1. Warm up machine – ~15-‐20 min to create and maintain fully hot and stable plasma
2. Tune machine –maximize M signal while minimizing M+16
oxide forma&on 3. Finish washing cells – final wash in MilliQ water, then
resuspension and dilu&on in MilliQ water. -‐ residual buffer salts can affect Current value
4. Resuspend and filter immediately before injec&on
-‐ Reduces aggregates (clogs) 5. Appropriate dilu&on – minimize doublets
-‐ Around 1 cell event/screen refresh; usually ~750K/mL on cell counter
Standard CyTOF Output Appearance
-‐ Spike at ~0 is legi&mate (“true zero”) -‐ Flow data is usually distributed above and below zero due to comps and auto-‐fluorescence. * FlowJo se~ngs must be changed for CyTOF data
0 101 102 103 104
CD3(Nd150)Dd: CD3
0
500
1000
1500
# C
ells
CD3+72.2
CD3-27.8
0 101 102 103 104
DNA2(Ir193)Dd: DNA2
0
101
102
103
104
DN
A1(
Ir191
)Dd:
DN
A1
Intact cells76.1
0 30 60 90 120Cell_length: Cell_length
0
101
102
103
104
DN
A1(
Ir191
)Dd:
DN
A1
Intact singlets71.6
0 101 102 103 104
CD14(Sm154)Dd: CD14
0
101
102
103
104
CD
33(E
r166
)Dd:
CD
33
Lymphocytes69.1
Monocytes30.7
0 101 102 103 104
Dead(In115)Dd: Dead
0
101
102
103
104
DN
A1(
Ir191
)Dd:
DN
A1
Live cells96.1
B Cells CD3+ Cells CD3-‐ Cells Fluo
rescen
ce
Mass
0 101 102 103 104
IgD(Nd146)Dd: IgD
0
101
102
103
104
CD
27(S
m15
2)D
d: C
D27
14.3 7.64
726.03
0 101 102 103 104
CD4(Nd143)Dd: CD4
0
101
102
103
104
CD
8(N
d144
)Dd:
CD
8
CD4+61
CD8+30.3
0 101 102 103 104
CD16(Sm149)Dd: CD16
0
101
102
103
104
CD
56(Y
b174
)Dd:
CD
56
NK cells20.1
0 102 103 104 105
<Pacific Blue-A>: CD16
0102
103
104
105
<FIT
C-A
>: C
D56
NK cells19.5
0 102 103 104 105
<Am Cyan-A>: IgD
0
103
104
105<P
E-C
y7-A
>: C
D27
13.6 7.03
69.89.51
0 102 103 104 105
<PerCP-Cy5-5-A>: CD4
0
103
104
105
<AP
C-C
y7-A
>: C
D8
CD4+59.9
CD8+32.1
CyTOF vs LSRII – Representa&ve Examples
Flu 2010-‐2011 -‐ CyTOF Phenotyping Panel
-‐ Collapse 6 tube LSRII panel into 1 tube CyTOF panel – 23 markers
CD14 CD33 CD3 CD4 CD8 TCRgd CD19 CD20 CD16 CD56
CD45RA CCR7 CD25 CD127 CD27 CD28 CD38 CD85j IgD HLADR CD24 CD94 CD161
Flu 2010, Flu 2011 Studies
-‐ On-‐going Flu vaccine studies – U19 -‐ CyTOF Phenotyping – 2010, 2011 years
-‐ Day 0 (pre-‐vaccina&on) only – Study 15, 17, 18 -‐ Day 0, 1, 4, 7 – Study 23-‐2011
-‐ 2010-‐2011: >300 unique donors – between-‐donor variability
Age Band (y/o) <35 35-65 >65 Total
Male 62 22 46 130 Female 84 47 65 196
Total 146 69 111 326
2010 Flu Studies – Consistency – 284 Control -‐ Same Draw, Same Year Replicates
Flu 2011, 2012 Studies – Consistency – 885 Control
Naive CD8+
Year
Per
cent
age
of P
aren
t
2011
2012
Men 35
-650
20
40
60
80
IgD+ CD27- B cells
Year
Per
cent
age
of P
aren
t
2011
2012
Men 35
-650
20
40
60
80
100
-‐ Generally good year-‐to-‐year agreement from same 885 donor draw
CD4+ Treg
Year
Per
cent
age
of P
aren
t
2011
2012
Men 35
-650
2
4
6
8
10
Flu 2011, 2012 Studies – Consistency – 885 Control
-‐ Some&mes greater variability in less-‐frequent popula&ons
B cells total
Men <3
5
Men 35
-65
Men >6
5
Wom
en <3
5
Wom
en 35
-65
Wom
en >6
50
20
40
60
80
Gender and Age Band
Per
cent
age
of P
aren
t
Flu 2010, Flu 2011 Studies – Donor Variability
**** ***
**** ****
Mann-‐Whitney two-‐tailed * <0.05 ** <0.01 *** <0.001 **** <0.0001
CD4+ Effector Memory
Men <3
5
Men 35
-65
Men >6
5
Wom
en <3
5
Wom
en 35
-65
Wom
en >6
50
20
40
60
80
100
Gender and Age Band
Per
cent
age
of P
aren
t
CD4+ Naive
Men <3
5
Men 35
-65
Men >6
5
Wom
en <3
5
Wom
en 35
-65
Wom
en >6
50
20
40
60
80
100
Gender and Age Band
Per
cent
age
of P
aren
t
CD4+ total
Men <3
5
Men 35
-65
Men >6
5
Wom
en <3
5
Wom
en 35
-65
Wom
en >6
50
20
40
60
80
100
Gender and Age Band
Per
cent
age
of P
aren
t
Flu 2010, Flu 2011 Studies – Donor Variability
**** ***
****
**** *** **
**** **** *
*** * *
Mann-‐Whitney two-‐tailed * <0.05 ** <0.01 *** <0.001 **** <0.0001
CD8+ Effector Memory
Men <3
5
Men 35
-65
Men >6
5
Wom
en <3
5
Wom
en 35
-65
Wom
en >6
50
20
40
60
80
100
Gender and Age Band
Per
cent
age
of P
aren
t
CD8+ Naive
Men <3
5
Men 35
-65
Men >6
5
Wom
en <3
5
Wom
en 35
-65
Wom
en >6
5-20
0
20
40
60
80
100
Gender and Age Band
Per
cent
age
of P
aren
t
CD8+ total
Men <3
5
Men 35
-65
Men >6
5
Wom
en <3
5
Wom
en 35
-65
Wom
en >6
50
20
40
60
80
Gender and Age Band
Per
cent
age
of P
aren
t
Flu 2010, Flu 2011 Studies – Donor Variability
* *
**** *** *
**** **** *
**** *** ****
**** **** ****
**** **** ****
Mann-‐Whitney two-‐tailed * <0.05 ** <0.01 *** <0.001 **** <0.0001
Study 23-‐2011
0 101 102 103 104
(Eu151)Dd: CD38
0
101
102
103
104
(Sm
152)
Dd:
CD
27CD27+ CD38+
0.465
0 101 102 103 104
(Eu151)Dd: CD38
0
101
102
103
104
(Sm
152)
Dd:
CD
27
CD27+ CD38+0.338
0 101 102 103 104
(Eu151)Dd: CD38
0
101
102
103
104
(Sm
152)
Dd:
CD
27
CD27+ CD38+0.566
0 101 102 103 104
(Eu151)Dd: CD38
0
101
102
103
104
(Sm
152)
Dd:
CD
27
CD27+ CD38+2.26
Day 0 Day 1
Day 4 Day 7
Conclusions
1. Proper tuning of the CyTOF ensures minimal sample oxida&on. 2. Metal purity (both elemental and isotopic) is highly important. 3. The choice of metal label is dependent upon marker abundance,
modality, and resolu&on needed (fold-‐change). 4. All samples must be stained with metals to be detected by the
CyTOF. Proper fixa&on is important, as are MilliQ water washes. 5. Head-‐to-‐head comparisons with LSRII assays are consistent. 6. The Flu studies are an example of how a CyTOF panel can be
used to monitor the surface phenotype for hundreds of donors to detect varia&ons between gender, age-‐band, and &me-‐point.
Acknowledgements
-‐ HIMC staff -‐ Henrik Mei, Meena Malipotlalla, Holden Maecker
-‐ External CyTOF:
-‐ Nolan lab: Sean Bendall, Erin Simonds, Eli Zunder, Bernd Bodenmiller -‐ Davis lab: Evan Newell, Bill O'Gorman, Peber Brodin -‐ Fluidigm/DVS Sciences
-‐ U19 NIH Grants and HIMC Service Center fees
hbp://himc.stanford.edu
