+1(978)310-4246 credencewriters@gmail.com
  

The overall goal of the article is to develop a phosphospecific flow cytometry-based assay to measure the effects of the compound library on signaling pathways. Identified hits were then confirmed in their pathway selectivity assays and finally lead compound selectivity was confirmed in vivo in mice. With this background, answer the following questions pertaining to each of the figures in the journal article (20 points):

Fig 1:

What cells were used in the primary screening, what simulants were used to induce phosphorylation of p38 and pSTAT1 and what was the outcome. How did they validate the assay – 2 point

Fig 2:

Describe briefly (step-wise description) how the authors performed the screening of the NCI natural product library (what stimulants were used and what signaling pathways were they targeting/measuring (2 points). To identify the hits what threshold did they use and how did multiparameter kinase pathway screening help identify pathway selective inhibitors (2 point)

Fig 3:

Describe how validation of the hits identified from initial screen was performed and what was the outcome (2 points)

Fig 4:

Screening was performed in heterogeneous primary cell populations with a goal to identify: i) Pathway specific inhibitors; ii) cell-type specific inhibitors and iii) patterns of pathway and cell type druggability. As related to the figure 4, answer the following questions

What was the origin (organ/tissue) of the primary cells and what markers were used to identify the different cell population;

How did the authors demonstrate quantitative nature of the phosphoflow platform;

What is Fluorescence cell barcoding and what is the advantage of its use;

What is selectivity factor (SI) and how did they apply it to interpret the data in Fig. 4c,d. (4 points).

Fig. 5:

Briefly describe the experiment and outcome of the results that lead to the identification of pathway selective compound (2 points)

Fig. 6 & 8:

What experiment and results lead the authors to believe that the inhibitor Streptonigrin is a cell-type selective compound. What was the outcome of evaluating this compound in vivo in mice (4 points)

Fig. 7:

What is selectivity factor clustering and how did application of this statistical tool help them identify pathways and cell types that are druggable (2 points)

Please if you cant read articles and understand whats going on do not pick question. Each question must be 4-6 value sentences long minimum.

© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
High-content single-cell drug screening with
phosphospecific flow cytometry
Peter O Krutzik, Janelle M Crane, Matthew R Clutter & Garry P Nolan
Drug screening is often limited to cell-free assays involving purified enzymes, but it is arguably best applied against systems that
represent disease states or complex physiological cellular networks. Here, we describe a high-content, cell-based drug discovery
platform based on phosphospecific flow cytometry, or phosphoflow, that enabled screening for inhibitors against multiple
endogenous kinase signaling pathways in heterogeneous primary cell populations at the single-cell level. From a library of
small-molecule natural products, we identified pathway-selective inhibitors of Jak-Stat and MAP kinase signaling. Dose-response
experiments in primary cells confirmed pathway selectivity, but importantly also revealed differential inhibition of cell types and
new druggability trends across multiple compounds. Lead compound selectivity was confirmed in vivo in mice. Phosphoflow
therefore provides a unique platform that can be applied throughout the drug discovery process, from early compound screening
to in vivo testing and clinical monitoring of drug efficacy.
Screening for small-molecule inhibitors of kinases and other enzymes
is typically performed in high-throughput formats using purified
enzymes as the target. Recently, more emphasis has been placed on
cell-based screens using reporter systems such as green fluorescent
protein or enzymatic reporters such as luciferase. Further refinement
of cell-based assays has lead to the development of high-content
screens that produce multiple readouts for each cell or measurement,
typically using microscopy platforms1–3.
However, in even the most advanced cell-based screens, it can be
difficult to assay multiple targets across the many cell subsets present
in heterogeneous primary cell populations (such as human peripheral
blood or mouse splenocytes4). In human peripheral blood mononuclear cells for instance, T cells typically comprise 70% of the total
population, B cells 10% and monocytes another 10%. Any assay or
screening platform that analyzes such a diverse population as a whole,
without sorting individual cell types, will necessarily be biased toward
more prevalent populations (T cells) and will miss important changes
that might occur in rarer cell types (monocytes and B cells). Moreover,
cell-based screens that focus on single cell types and single biochemical
targets ignore the potential of compounds to exhibit differential
behavior across cell types and to affect multiple pathway components
in a signaling network.
Therefore, a platform that can distinguish cell types from one
another and simultaneously measure the effects of a drug on several
kinase pathways in each cell type, providing crucial insights into onand off-target effects, would be beneficial. Phosphospecific flow
cytometry, or phosphoflow, is a single-cell, multiparameter assay
platform capable of overcoming these limitations in traditional
approaches. Phosphoflow is a cell-based assay that enables quantitative
measurement of the phosphorylation levels of intracellular signaling
proteins5–9. The platform has been applied to profile disease and
normal signaling cascades, to elucidate roles of aberrant signaling in
cancer10–12 and to study the signaling biology of immune cell subsets8,13–16. Its application to drug screening has been somewhat limited
and focused on single drug effects in cell lines or primary cells5,17.
Importantly, using phosphoflow, measurements are made on endogenous proteins after phosphorylation is induced by extracellular
ligands such as cytokines, which leads to greater confidence in the
derived mechanistic conclusions18. In general, the technique has been
applied by measuring effects on physiologic targets within cells, thereby
circumventing the need for overexpression of reporter substrates and
eliminating possible disruption of normal signaling mechanisms that
can result from perturbation of protein content in cells.
Furthermore, phosphoflow is highly quantitative, enabling doseresponse titration curves to be generated for multiple pathways and
cell types simultaneously6,17. Throughput of flow cytometry has
recently been improved by the development of high-throughput
autosamplers and a technique called fluorescent cell barcoding developed in our laboratory17,19–21. The ability of the flow cytometer to
measure 15 or more fluorescent parameters for every single cell being
analyzed enables rapid interrogation of two fundamental properties of
drug-like compounds: pathway selectivity and cell type selectivity.
Pathway selectivity of compounds by the phosphoflow approach is
assessed by simultaneously analyzing multiple signaling cascades using
antibody reagents specific for different phosphoepitopes. For instance,
it is possible to analyze signaling via several constituents of the
mitogen-activated protein kinase (MAPK) pathway while simultaneously assaying multiple Jak-Stat signaling pathways, thus enabling
network-based screening in complex populations4,8. Determination of
the cell type selectivity of a compound is also enabled because cell
Department of Microbiology and Immunology, Baxter Laboratory in Genetic Pharmacology, Stanford University, 269 Campus Drive, Stanford, California 94305, USA.
Correspondence should be addressed to G.P.N. (gnolan@stanford.edu).
Received 15 June; accepted 30 October; published online 23 December 2007; doi:10.1038/nchembio.2007.59
132
VOLUME 4
NUMBER 2
FEBRUARY 2008
NATURE CHEMICAL BIOLOGY
a
Unstimulated
Stimulated
780
160
5
10
104
103
102
3,220
296
0
pStat1
Figure 1 Cell-based screening with phosphoflow. (a) The U937 cell line was
stimulated with IFN-g and anisomycin simultaneously to induce Stat1 and
p38 phosphorylation. Cells were stained with pStat1 Pacific Blue and p-p38
PE antibodies. Note the shift of the cell population to the upper right
quadrant of the flow cytometric plot upon stimulation (top plots). Prior
treatment with Jak inhibitor I (JakI) or SB203580 blocked signaling via
Stat1 or p38, respectively (bottom plots). Values in the corners of the plots
represent median fluorescence intensities for the pStat1 and p-p38 axes.
Numbers in parentheses represent percent inhibition of the pathway. Arrows
indicate direction of phosphorylation induction or inhibition. (b) Doseresponse experiment with JakI. U937 cells were treated with ten-fold
dilutions of JakI (from 10–5 to 10–11 M), stimulated with IFN-g and
analyzed for pStat1 levels by phosphoflow. The IC50 curve generated from
the phosphoflow median fluorescence intensity data is shown to the right.
Mean and s.d. from triplicate experiments are shown. Data were fit by
logistic regression with R2 ¼ 0.99.
0
102
103
104
105
0 102
103
104
105
Stimulated
Inhibitor:
5 166
(99%)
SB203580
Jakl
Jakl+SB203580
182
(96%)
660
10
104
103
RESULTS
A quantitative cell-based flow cytometric screen
Our initial screen of the NCI natural product compound library was
carried out in U937 cells, a cell line derived from a monocytic
leukemia. This cell line is often used in our laboratory due to its
robust and consistent induction of Jak-Stat and MAPK signaling
pathways in response to stimulation. The goal of this initial screening
in a myeloid cell line was to ‘trim’ the compound library to a smaller
number of molecules that were able to modulate cytokine signaling in
intact cells of the immune system. These lead molecules were then
carried on to more complex experiments in primary cell populations
(Supplementary Fig. 1 online).
Treatment of U937 cells with interferon-g (IFN-g) and anisomycin
(2) lead to the induction of Stat1 and p38 phosphorylation, respectively (Fig. 1a). In staining during phosphoflow, fluorescently labeled
phosphospecific antibodies are added to cells after the cells have been
fixed and permeabilized. The median fluorescence intensity (MFI) of
antibody staining correlates to the level of phosphorylation of each
signaling protein6. Stimulation of signaling increases the MFI value,
while inhibition of signaling leads to a decrease in MFI relative to
positive controls. In U937 cells, simultaneous addition of IFN-g and
anisomycin (and later granulocyte-macrophage colony-stimulating
factor, GM-CSF) was equivalent to adding each stimulus individually,
NATURE CHEMICAL BIOLOGY
VOLUME 4
NUMBER 2
FEBRUARY 2008
102
(90%)
521
1,660
0
0
102
103
104
5
0 102
10
b
103 104
p-p38
IF
N
Ja γ
kl
–
–
+
–
100
+
+
+
+
+
+
0
0 102 103 104 105
pStat1
105
(97%)
323
0 102
103
104
105
100
Percent inhibition
surface markers can be used to delineate cell subsets while concurrent
measures of intracellular phosphoepitopes are used to determine
pathway specificity7. This allows for compound screening in primary
cell populations such as human peripheral blood and mouse blood
cells or tissues8,22.
Here, we describe phosphoflow-based screening of the natural
product library from the US National Cancer Institute (NCI) for
pathway and cell type–selective inhibitors of cytokine-induced Jak-Stat
signaling23,24. Initial screening in the U937 monocytic cell line revealed
several lead compounds with pathway selectivity against the Jak-Stat
family of proteins. These compounds were evaluated in primary
mouse splenocytes and blood, and we identified one compound,
streptonigrin (1), that displayed greater than 100-fold selectivity for
inhibition of Stat activation in B cells relative to T cells. We verified
these cell type–specific effects in vivo in mice, both in the spleen and in
blood. Therefore, phosphoflow is a platform capable of screening
compound libraries for inhibitors of specific kinase signaling pathways
or particular cell types in physiologically and disease-relevant primary
cell populations. The platform can be used throughout drug discovery,
from initial screening in cell lines and primary cells to in vivo studies
in mice to clinical trials monitoring drug efficacy in people.
Percent inhibition
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
80
60
40
20
IC50 = 10 nM
0
10–11
10–9
10–7
[Jakl] (M)
10–5
which indicates no obvious synergistic interactions of the stimuli on
the pathways being measured. Adding kinase inhibitors such as Jak
inhibitor I (JakI, 3) or SB203580 (4) before stimulation prevented
induction of the expected pathways, IFN-g:phospho-Stat1 (pStat1)
and anisomycin:phospho-p38 (p-p38), respectively. Percent inhibition
was calculated using the MFI values from treated samples as well as the
untreated positive (stimulated) and negative (unstimulated) control
samples. Performing a dose-response experiment with Jak inhibitor I,
we were able to plot percent inhibition against compound concentration and determine the half-maximal inhibitory concentration (IC50)
as 10 nM (Fig. 1b).
Multiparameter screening for pathway-selective compounds
With a quantitative cell-based assay for signaling modulators in hand,
we screened the NCI natural product library of 235 compounds at a
single dose (20 mM) to determine their effects on three signaling
pathways: IFN-g:pStat1, GM-CSF:pStat5 and anisomycin:p-p38
(Fig. 2a). Ideally, the goal of the screen was to find compounds that
selectively modulate Jak-Stat cytokine signaling without affecting
MAPK signaling. Compounds were added for 30 min followed by
stimulation with simultaneous addition of IFN-g, GM-CSF and
anisomycin. The cells were fixed with formaldehyde, permeabilized
with methanol and simultaneously stained with phosphospecific
antibodies against Stat1, Stat5 and p38 (ref. 6). The percent inhibition
of each compound on each of the three signaling pathways was then
calculated (Fig. 2a).
Overall, assay performance was robust with Z ¢ factors of 0.72, 0.65
and 0.76 for pStat1, pStat5 and p-p38 measurements, respectively
(Supplementary Table 1 online). The Z ¢ factor is a measurement
of assay robustness, combining the signal-to-noise ratio and s.d.
of control samples. Values of Z ¢ greater than 0.5 are considered
133
Pathway
IFN-γ :pStat1
GM-CSF:pStat5
Anisomycin:p-p38
100
80
Percent inhibition
Pathway selectivity confirmed in cell line
To confirm the hits obtained in the initial screen, we further tested 18
compounds in a dose-response experiment: 13 potent inhibitors
(which showed 465% inhibition in the initial screen), 3 weak
inhibitors (which showed 25–65% inhibition) and 2 compounds
that did not display any effect (Fig. 3). Of the 13 potent inhibitors,
12 showed nearly identical effects on each of the 3 pathways as were
observed in the initial screen. The 3 weak inhibitors again displayed
b
+3σ
60
Hit threshold
40
20
0
Potency
70
–20
60
50
Selectivity
a
For example, we were particularly interested in finding compounds
that selectively inhibit Stat5 signaling. If our screen had been onedimensional, we would have found eight compounds that potently
inhibit the pathway of interest (Fig. 2c, red points). When data from
the simultaneous p38 analysis is included, the number of selective hits
is narrowed to four compounds. Adding a third simultaneously
measured pathway, in this case IFN-g:pStat1, revealed that three out
of the four compounds also inhibit Stat1 phosphorylation. A single
compound from the library, cryptosporiopsin (NSC137442, 5), selectively inhibited the GM-CSF:pStat5 pathway.
Applying similar analysis to the anisomycin:p-p38 pathway revealed
seven selective inhibitors (Fig. 2d). No inhibitors were found to block
IFN-g:pStat1 signaling specifically. Thus, increasing the number of
simultaneous pathways analyzed improved the ability to find compounds with specific inhibition profiles and increased confidence that
unwanted side effects would be minimized.
Max – avg inhibition (%)
excellent25. We set the threshold for selection of hits at 3 s.d. from the
means of the compound-treated samples (that is, at approximately
65% inhibition) (Fig. 2a). Of the 235 compounds in the NCI natural
product library, 15 compounds inhibited one or more pathways above
this hit threshold (for a 6% hit rate). This relatively high hit rate
is not unexpected from a natural product library of biologically
active molecules in an assay that measures effects on three pathways
at once.
The advantages of screening multiple pathways simultaneously were
readily apparent. By evaluating the effect of each compound on two
Jak-Stat pathways and one MAPK pathway, we obtained from a single
analysis data that would traditionally require three independent
screens. This allowed us to generate a plot of the maximum inhibition,
or potency, of each compound against any of the three signaling
pathways versus the maximum value minus the average inhibition
value, or selectivity, across all three measured pathways (Fig. 2b).
Compounds that display potent inhibition of any pathway appear in
the right side of the plot. Those that inhibit all three pathways potently
have low selectivity values, whereas pathway-selective compounds
have high selectivity values. This type of multiparameter screening
greatly reduces the need for counter-screens to validate hits from the
initial screen and saves time spent pursuing compounds that show
nonselective or undesirable inhibition profiles.
From the initial screen, it was clear that as the number of pathways
analyzed increased, the hit rate for selective compounds decreased.
±1σ
40
30
20
10
0
0
Natural product library compounds
2
3
100
80
60
pStat1 >65%
inhibition
40
20
0
–20
Compounds
20 40 60 80 100 120
Maximum inhibition (%)
d
Number of pathways analyzed
1
Number of selective hits
c
Percent inhibition
GM-CSF:pStat5
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
IFN-γ :pStat1
GM-CSF:pStat5
Anisomycin:p-p38
12
10
8
6
4
2
0
1
3
2
Number of pathways analyzed
–20 0 20 40 60 80 100
–20 0 20 40 60 80 100
Percent inhibition anisomysin:p-p38
Figure 2 Initial screen reveals Jak-Stat and MAP kinase pathway-selective compounds. (a) The inhibitory activity of the 235 compounds in the NCI natural
product library on the three pathways measured in U937 cells (each pathway represented by different colored squares). The hit threshold was drawn at
approximately three s.d. away from the mean inhibition value of zero (65% inhibition). (b) Plot of maximum minus average inhibition of compounds
(selectivity) versus maximum inhibition (potency). Compounds that specifically target one pathway have high selectivity and high potency (top right of plot).
Compounds that nonspecifically inhibit more than one pathway have high potency, but low selectivity (bottom right of plot). Background colors in plot
represent no inhibition (blue), weak inhibition (light yellow) and strong inhibition (dark yellow). (c) Advantage of multiparameter kinase pathway screening
when searching for pathway-selective inhibitors. Eight compounds inhibit GM-CSF:pStat5 signaling; however, four of these compounds also inhibit the
anisomycin–p-p38 pathway. Three of the remaining compounds also block IFN-g:pStat1 signaling, leaving a single compound that is selective for the
GM-CSF:pStat5 pathway. Red boxes designate compounds selective for GM-CSF:pStat5 given the indicated number of pathways analyzed. Yellow triangles
indicate compounds that inhibited IFN-g:pStat1 by greater than 65%. (d) Increasing the number of pathways analyzed reduced the number of pathwayselective hits identified. The analysis shown in c was performed for all three pathways, revealing seven compounds selective for the anisomycin:p-p38
pathway, and no compounds selective for the IFN-g:pStat1 pathway.
134
VOLUME 4
NUMBER 2
FEBRUARY 2008
NATURE CHEMICAL BIOLOGY
but lacked potency at 20 mM, and therefore was not examined further.
The final compound that blocked two pathways and was carried
forward to primary cell screening was streptonigrin (NSC45383, 1)31.
This compound was particularly noteworthy because it blocked GMCSF:pStat5 and anisomycin:p-p38, but not IFN-g:pStat1, thus inhibiting both tyrosine and serine-threonine kinases while still maintaining
some selectivity.
Finally, one molecule, striatin E (NSC312033, 10), inhibited all
three signaling pathways nearly completely at 20 mM. Two other
compounds from the initial screen showed this nonspecific inhibition
profile but were excluded from the secondary dose-response screen
because the desired profile was for pathway-selective compounds, not
general kinase inhibitors. However, as noted, the platform enables
screening for any desired inhibition phenotypes, including nonselectivity of compound effects.
intermediate levels of inhibition at 20 mM, whereas the previously
ineffective compounds showed no inhibition. Therefore inhibition
patterns from 17 of 18 (94%) compounds were confirmed.
The 12 potent inhibitors were classified into three categories
based on the number of pathways they inhibited (Fig. 3). Six
compounds selectively inhibited one pathway. Of these six, five
blocked the anisomycin:p-p38 pathway. One of these molecules,
homoharringtonine (HHT, NSC141633, or cephalotaxine, 6), was
extremely potent, with an IC50 value of approximately 200 nM
(Fig. 3). Homoharringtonine is currently being used in treatment
of the cancers acute myelogenous leukemia (AML) and chronic
myelogenous leukemia (CML) and is a known protein synthesis
inhibitor26,27. The other compound that selectively inhibited one
pathway was cryptosporiopsin, which blocked the GM-CSF:
pStat5 pathway at 20 mM. Cryptosporiopsin (NSC137442) is an
antibiotic that has not received much use due to potential detrimental effects on RNA synthesis28. However, because of its high
selectivity for Jak-Stat signaling we carried this compound on to
screening in primary cells.
Four compounds selectively inhibited two out of the three pathways.
Of these, three inhibited IFN-g:pStat1 and GM-CSF:pStat5, a desired
cytokine signaling pathway inhibition profile. MHK (methyl-13hydroxy-15-oxokaurenoate, NSC620358, 7) and crassin (NSC210236,
8)29 completely blocked Stat1 and Stat5 phosphorylation at 20 mM,
with little effect on p38 activity. These compounds were carried on to
primary cell screening. Nanaomycin (NSC267461, 9)30 was selective
a
IFN-γ:
pStat1
GM-CSF:
pStat5
Multiplexed, quantitative primary cell drug screening
We next used the multiparameter capabilities of phosphoflow to
evaluate the four Jak-Stat selective inhibitors from the NCI natural
product library (cryptosporiopsin, streptonigrin, MHK and crassin)
in primary mouse splenocytes (Fig. 4). In addition to the library
compounds, we tested three commercially available Jak-Stat inhibitors:
Jak inhibitor I (JakI; also known as pyridone 6 or P6)32, WHI-P154
(WHI, Jak3 inhibitor II, 11)33 and ZM449829 (ZM, Jak3 inhibitor
V, 12)23,24. WHI-P154 has been used extensively as an immunosuppressive agent in transplantation23. Indirubin-3-monoxime (I3M, 13),
b
Anisomycin:
p-p38
[Compound]
Percent inhibition
Nagilactone C (14)
Coumermycin (15)
Eupatal (16)
Helenalin (17)
Streptovitacin A (18)
100
Unstimulated
Stimulated
HHT
0
Natural products
Solanine (19)
1 pathway
inhibited
B-1 (20)
Tubulosine (21)
Cryptosporiopsin
HHT (6)
Rottlerin (22)
Fastigillin B (23)
Cryptosporiopsin (5)
Streptonigrin
Nanaomycin (9)
Streptonigrin (1)
2 pathways
inhibited
MHK (7)
MHK
Crassin (8)
Patulin (24)
Controls
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
Striatin E (10)
Jakl (3)
SB203580 (4)
Striatin E
3 pathways
inhibited
0 102 103 104 105 0 102 103 104 105
pStat1
pStat5
0 102 103 104 105
p-p38
Figure 3 Validation of hits from initial cell line screen. (a) 18 compounds from the initial screen were subjected to dose-response titration in U937 cells
to measure inhibition of the IFN-g:pStat1, GM-CSF:pStat5 and anisomycin:p-p38 pathways. Percent inhibition of each pathway at increasing doses of each
compound is represented with a heatmap. Compounds with a range of inhibitory effects in the initial screen were chosen to test the robustness of the
phosphoflow platform in identifying selective compounds. Text background box represents percent inhibition each compound displayed in the initial screen:
blue, no inhibition; light yellow, 25–65%; dark yellow, 465%. (b) Histogram plots of the effects of five compounds on the three signaling pathways
measured. Homoharringtonine and cryptosporiopsin inhibit one pathway, streptonigrin and MHK inhibit two pathways, and striatin E nonselectively inhibits
all three pathways. Note that all three pathways were analyzed simultaneously at the single-cell level with pStat1 Pacific Blue, p-p38 PE and pStat5 Ax647
phosphospecific antibodies. Histograms are colored according to percent inhibition. Wedges represent increasing compound doses of 0.2, 2 and 20 mM. B-1,
baccharis principle B-1.
NATURE CHEMICAL BIOLOGY
VOLUME 4
NUMBER 2
FEBRUARY 2008
135
previously identified in our laboratory and others as having Jak-Stat
selectivity, was also tested17,34.
In the primary cell screen, we greatly expanded on the number of
measurements in our initial cell line screen. We examined 24 signaling
pathways: six cytokines (IFN-g, interleukin 4 (IL-4), IL-6, IL-7, IL-10
and IL-15) across four Stat proteins (Stat1, Stat3, Stat5 and Stat6). The
effect of each compound on each of these pathways was measured in
five cell types for a possible 120-point profile of each compound. As
would be expected from normal physiology, none of the cytokines
induced all of the possible signaling molecules in all of the cell types,
so we limited our analysis to the 27 cell type–signaling pathway
combinations that showed greater than approximately two-fold induction of phosphorylation after stimulation (Supplementary Fig. 2
2,000
1,000
104
10
0
0
4
10
5
0
B cells
105
CD4 PE-Cy7
104
3
0
CD4+ T cells
104
CD4– T cells
103
0
100
80
60
40
20
0
80
60
CD4 T cells
20
0
103 104 105
TCRβ PE
d
IL-6:pStat1
IL-15:pStat5
IL-10:pStat3
IL-7:pStat5
IFN-γ:pStat1
IL-4:pStat6
IL-6:pStat3
IL-4:pStat5
100
IC50 (µM): 0.7
2.5
18
Selectivity factor: 1.7
0
–2.9
105
µM
100
IC50
SF = –log2
median IC50
6
1,000
4
100
2
Selectivity factor
10
0
1
–2
0.1
–4
ZM
I
H
W
K
H
M
R
PN
si
n
ra
s
ST
C
R
I3
M
ZM
I
K
H
W
H
M
PN
R
n
ST
N
ra
ss
i
C
C
R
PS
Ja
kl
M
I3
N
–6
0.01
C
IC50 value (µM)
10,000
103 104
SSC
0
1
10
[Crassin] (µM)
10
1
0.1
0.1
0
40
–
l
10
103 104 105
TCRβ PE
Median IC50 curve
100
PS
105
B220 QD605
3
10
10
SSC
All 27 pathway–cell type
combinations for crassin
100
80
60
40
20
0
Ja
k
0
c
IL-4:pStat6 pathway
CD11b-hi
CD11b-int
3
Percent inhibition
FSC
3,000
b
CD11b-hi
105
CD11b PerCP-Cy5.5
a 4,000
online). Stimulation conditions were chosen to provide maximum
coverage of cell types and signaling pathways.
The primary cell screen was different from our initial screen in
three critical aspects. First, the mouse spleen is a complex, heterogeneous population of immune cell types. In order to examine the
effects of each molecule on multiple cell types simultaneously to gain
information on cell type selectivity, we stained with antibodies against
surface markers in addition to the phosphospecific antibodies used in
the initial screen so that we could define cell types of interest (Fig. 4a).
The cell types included CD11b-hi cells (neutrophils), CD11b-intermediate cells (monocytes, macrophages, dendritic cells and natural
killer cells), B cells (B220+TCRb–), CD4+ T cells (TCRb+CD4+
B220–) and CD8+ T cells (TCRb+CD4-B220–) (Fig. 4a).
Relative strength of inhibition
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
Figure 4 Screening in heterogeneous primary cell populations. (a) Cell type identification. Using four surface markers, five cell types were identified:
B cells, CD4+ T cells, CD4– (that is, CD8+) T cells, CD11b-hi cells (neutrophils) and CD11b-int cells (macrophages, dendritic cells and other myeloid
populations). Each cell type is color coded. CD11b-hi cells, cyan; CD11b-int cells, green; B cells, red; CD4+ T cells, black; CD4– T cells, yellow. FSC,
forward scatter; SSC, side scatter. (b) Dose-response titration curves of crassin on the IL-4:pStat6 pathway in CD11b-hi cells and CD4– T cells. Data were fit
by logistic regression with R2 ¼ 0.99. (c) Determination of SF. Using the IC50 values from all 27 cell type–pathway combinations for crassin, the median
IC50 was calculated, and all other IC50 values normalized to this value using the equation shown. The SF calculation for the two cell types in b are
displayed. (d) The entire set of 216 data points (8 compounds 27 cell type–pathway combinations) showing both absolute IC50 values and SF values for
the eight compounds analyzed in primary cells. Cell types are color coded as in a. Most compounds display median IC50 values (shown in black dashed
lines) between 1 and 20 mM. Note that the SF is a dimensionless number that represents the relative selectivity of a compound for a particular pathway or
cell type. Positive values represent relatively strong inhibition, whereas negative values represent relatively weak inhibition. Cyan and yellow arrows indicate
the two cell type–pathway combinations highlighted in b and c. CRPSN, cryptosporiopsin; STRPN, streptonigrin.
136
VOLUME 4
NUMBER 2
FEBRUARY 2008
NATURE CHEMICAL BIOLOGY
Crassin (8)
+2
–2
Selectivity factor
OH
100
80
60
40
20
0
IL-10:pStat3
IL-4:pStat5
IL-4:pStat6
B cells (B220+TCRβ-)
0.1
1 1.5
10 16 100
[Crassin] (µM)
IL-4
IL-10
100
IL-4
C
yt
C okin
ra e
ss
in
IL-4:pStat6
IFN-γ :pStat1
IL-4:pStat5
IL-10:pStat3
IL-15:pStat5
IL-6:pStat3
IL-6:pStat1
IL-7:pStat5
Percent inhibition
O
d
B cell pathways
Percent inhibition
D
C
O
c
D
b
HO
C
a
4+
T
c
4 – ell
B Tc
ce e
ll
l
C l
D
11
C b-i
D
11 nt
bhi
ARTICLES
– –
+ –
+
+
+
+
+
0
0
3
4
5
10 10 10
pStat3
3
0 10
4
5
10 10
pStat5
0
3
4
5
10 10 10
pStat6
Cytokine
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
B cells
e Cytokine Purified
– + + + +
Figure 5 Identification of pathway-selective compound. (a) Chemical structure of crassin. (b) Heatmap representation
Crassin (µM) – – 2 10 50
of the 27-point inhibition profile of crassin. Note the horizontal pattern of SF values, indicating pathway-selective
Stat3(pY705)
IL-10
inhibition of IFN-g:pStat1, IL-4:pStat5 and IL-10:pStat3. Also note potent inhibition of CD11b-hi cells as illustrated in
IL-4
Stat5(pY694)
Figure 4. Arrows indicate pathways examined in c. (c) Titration curves derived from phosphoflow analysis of crassin
IL-4
Stat6(pY641)
showing IL-10:pStat3, IL-4:pStat5 and IL-4:pStat6 pathways in B cells. (d) Histogram analysis of original phosphoflow
data of the three pathways in c. Wedge indicates increasing doses of 0.08, 0.4, 2, 10 and 50 mM of crassin. Plots are
colored according to percent inhibition, yellow representing 100% inhibition. pStat3 Ax488, pStat5 Ax647 and pStat6 Ax488 monoclonal antibodies (in two
staining sets) were used for staining. (e) B cells were purified from total splenocytes by magnetic isolation, treated with three concentrations of crassin, then
stimulated with IL-10 and IL-4. Cells were lysed and subjected to western blot analysis with the same phosphospecific antibodies used for phosphoflow.
Nonspecific bands on the western blots were of insignificant intensity compared with specific bands.
Second, each compound was tested with a five-point dose-response
titration over a 600-fold range of concentrations. This enabled us to
plot percent inhibition versus drug concentration and determine IC50
values for each cell type and pathway of interest (Fig. 4b). Data was
curve fit with logistic regression equations in Spotfire DecisionSite.
Over 77% of the curves had R2 values of 4 0.99, and 93% had R2
values 4 0.95, thus demonstrating the quantitative nature of the
phosphoflow platform.
Finally, to enable much higher throughput of samples, we applied
fluorescent cell barcoding (FCB), a technique developed in our
laboratory that labels individual samples with unique signatures of
fluorescent molecules, allowing the samples to be combined into one
tube for staining and analysis17. Application of the FCB method
substantially reduced antibody consumption and acquisition time
on the cytometer and improved data robustness by combining control
wells together with compound-treated wells.
Inhibition profiles described by selectivity factor
The goals of our primary cell screen were to find (i) pathway-specific
inhibitors, (ii) cell type–specific inhibitors and (iii) patterns of pathway and cell type druggability.
To more easily identify compounds with specific inhibition profiles
across cell types and signaling pathways, we represented the compound selectivity as a dimensionless value that described selectivity
without regard to its IC50 value. We designated this term the
‘selectivity factor’ (SF) (Fig. 4c,d). Although absolute potency is
important in drug screening, we were focused on highlighting pathway or cell type–selective compounds, regardless of their actual IC50
value (though all compounds had average IC50 values below 20 mM,
Fig. 4d). The SF is a straightforward normalization equation whereby
the IC50 value on a particular pathway in a particular cell type is
divided by the median IC50 of the compound across all the measured
pathways and cell types. Log2 conversion of this IC50 ratio results in
pathways having the median IC50 to have SF values of zero. Inverting
the values ensures that pathways inhibited potently are represented by
positive values, and those inhibited relatively poorly are represented by
negative values. Figure 4c shows the effect of crassin on CD11b-hi and
CD4– T cell signaling through the IL-4:pStat6 pathway. Shown are the
IC50 values and the corresponding SF values. Compounds with a large
range of SF values have a broad range of inhibitory action and are
selective. Those with small ranges of SF values tend to inhibit all
NATURE CHEMICAL BIOLOGY
VOLUME 4
NUMBER 2
FEBRUARY 2008
pathways with equal potency and are nonselective (though they may
be extremely potent). Note that IC50 values and SF values are
interchangeable, requiring only the median IC50 value for conversion
back and forth (Fig. 4d).
Discovery of pathway-selective inhibitors
To identify pathway- and cell type–selective compounds, we plotted
the 27 SF values for each compound as a heatmap with cell types on
the x axis and pathways on the y axis (Supplementary Fig. 3 online).
Because not all pathways are induced in every cell type, white gaps are
present in the heatmap display. By representing the data in this
manner, compounds that were pathway selective showed horizontal
patterns of high and low SF values. Compounds that were cell type
selective showed vertical patterns of high and low SF values.
In searching for pathway-selective compounds, crassin
(NSC210236, Fig. 5a) showed the clear and characteristic horizontal
‘bands’ of SF values (Fig. 5b) for which the data representation was
developed. Looking at B cell pathways alone, crassin was approximately ten-fold more selective for IL-10:pStat3 and IL-4:pStat5 (IC50
values of 1–2 mM) relative to IL-4:pStat6 (IC50 value of 16 mM,
Fig. 5c). Examining the phosphoflow histogram data (Fig. 5d), we
observed a clear dose-response titration curve, with the IL-10:pStat3
and IL-4:pStat5 pathways decreasing in antibody fluorescence at lower
doses of crassin than the IL-4:pStat6 pathway. To confirm the
selectivity profile of crassin, we purified B cells via magnetic isolation,
treated with the compound, then stimulated with IL-4 and IL-10. We
then performed western blots with pStat3, pStat5 and pStat6 antibodies (Fig. 5e). Blotting confirmed not only the concentration of
inhibition (about 1 mM for IL-10:pStat3 and IL-4:pStat5), but also the
selectivity (about 10 mM for IL-4:pStat6).
Discovery of cell type–selective inhibitors
A major advantage of a single-cell drug discovery platform based on
phosphoprotein analysis is the ability to discern signaling effects of
drugs in complex, heterogeneous populations of cells. Depending on
the disease in question, it may be desirable to develop a drug against
the cell type that is known to cause or drive the disease, and perhaps
more importantly, to avoid inhibition in surrounding normal cell
types. We therefore analyzed the dataset to find compounds that act
only against specific cell types. In this case, we looked for compounds
that display vertical patterns in their 27-point inhibition profile
137
O
O
HN
HO
O
O
Streptonigrin (1)
+2
–2
Selectivity factor
N
H2N
N
100
IF
B cells
+
CD4 T cells
100
80
60
40
20
0
– –
+ –
+
+
+
+
+
0
3
1 3.5 10
[STRPN] (µM)
0.1
+
CD4 T cells
B cells
0 10
100
4
5
10 10
0 10
3
4
5
10 10
pStat1
e
IFN-γ
Purified
– + +
Figure 6 Identification of cell type–selective compound. (a) Chemical structure of cell type–selective compound
STRPN (20 µM) – – +
streptonigrin (STRPN). (b) Heatmap representation of the 27-point inhibition profile of streptonigrin. Note the
Stat1(pY701)
B cells
vertical pattern of SF values, indicating potent inhibition of B cells and CD11b-int cells. Arrows indicate pathways
+
examined in c. (c) Dose-response titration curves derived from phosphoflow analysis of the effect of streptonigrin
Stat1(pY701)
CD4 T cells
on IFN-g:pStat1 signaling in B cells and CD4+ T cells. Note that even at the highest dose tested of 100 mM, no
+
pathways in CD4 T cells were inhibited greater than 50%. (d) Histogram analysis of original phosphoflow data for the IFN-g:pStat1 pathway in B cells and
CD4+ T cells. Wedge indicates increasing doses of 0.16, 0.8, 4, 20 and 100 mM of streptonigrin. Plots are colored according to percent inhibition, yellow
representing 100% inhibition. pStat1 Ax647 monoclonal antibody was used for staining. (e) B cells and CD4+ T cells were isolated from total splenocytes
via magnetic separation, treated with streptonigrin at 20 mM and stimulated with IFN-g. Cells were lysed and subjected to western blot analysis with the
same phosphospecific Stat1 antibody used for phosphoflow analysis. Nonspecific bands were not present on the western blot.
Druggability of signaling pathways and cell types
Having analyzed 27 unique SF values for each compound, we
examined whether there were any observable biological trends
among their inhibition profiles. In particular, it would be important
to know whether particular pathways or cell types were inherently
more or less capable of being inhibited than others across multiple
compounds—that is, whether they were more ‘druggable’ targets.
For this analysis, we combined the cell type and pathway variables
into 27 cell type–pathway combinations and performed hierarchical
clustering of the cell type–pathway combinations and the compounds using Ward’s method, which is commonly used in microarray analysis (Fig. 7a). By this approach, cell type–pathway
combinations that are inhibited more easily have positive SF values
and appear at the bottom of the clustering analysis, whereas those
that are inhibited weakly or not at all appear at the top of the
clustering analysis.
a
b
CD4+/IL-4:pStat6
–
B220/IL-4:pStat6
CD11bint/IL-10:pStat3
138
B220/IL-10:pStat3
Selectivity factor
CD4+/IL-6:pStat3
CD4 /IL-6:pStat3
CD11bint/IL-4:pStat6
CD11bhi/IL-4:pStat6
2
0
–2
–4
+
CD4 /IL-6:pStat1
CD4+/IL-7:pStat5
IL-4:pStat5 IL-4:pStat6
CD11bhi/IFN-γ :pStat1
CD4+/IFN-γ:pStat1
c
Druggable cell type
–
CD4 /IL-15:pStat5
–
CD4 /IL-6:pStat1
IL-6:pStat3 pathway
–
CD4 /IL-7:pStat5
+
CD4 /IL-15:pStat5
2
CD4+/IL-4:pStat5
CD4+/IL-10:pStat3
CD4–/IL-10:pStat3
CD11bint/IL-15:pStat5
CD11bint/IL-4:pStat5
B220/IFN-γ:pStat1
Selectivity factor
0
–2
–
CD4 /IFN-γ:pStat1
–4
–
CD4 /IL-4:pStat5
B220/IL-4:pStat5
B220/IL-6:pStat3
VOLUME 4
NUMBER 2
–3
+3
Selectivity factor
FEBRUARY 2008
ce
C
lls
D
4+
T
c
el
C
ls
D
4–
T
ce
lls
Figure 7 Druggable signaling pathways and cell types revealed by selectivity
factor clustering. (a) The SF values of each compound on the 27 unique cell
type–pathway combinations were clustered via hierarchical clustering using
Ward’s method. Note the unique profile of the cell type–selective compound
streptonigrin (STRPN). Red circles indicate pathways analyzed in b, whereas
green circles indicate cell types analyzed in c. (b) Druggable pathways
susceptible to inhibition. The two signaling proteins induced by IL-4, pStat5
and pStat6 were separated in the clustering analysis. Each circle on the
plot represents a drug–cell type combination. On average across all the
compounds and cell types examined, IL-4:pStat5 was inhibited at
approximately ten-fold lower doses than IL-4:pStat6. This indicates that
the Stat5 pathway is a more druggable target of IL-4 signaling. Black bars
represent mean SF values. (c) Druggable cell types susceptible to inhibition.
IL-6 induced Stat3 phosphorylation potently in B cells and T cells. However,
these cell types were at opposite ends of the clustering analysis. Each circle
represents a different compound. On average across all of the compounds
tested, B cells were inhibited at ten-fold lower doses than CD4+ and
CD4– T cells. Therefore, B cells are more druggable for this pathway, and
T cells may be difficult to inhibit selectively without affecting B cells.
B220, B cells.
Druggable pathway
–
CD4 /IL-4:pStat6
B
heatmaps, which indicate inhibition of cell types rather than pathways
(Supplementary Fig. 3).
One compound, streptonigrin (NSC45383, Fig. 6a), was unique in
its clear vertical SF pattern, which indicated potent inhibition of
B cells and CD11b-int cells and almost no inhibition of CD4+ T cells
and CD11b-hi cells (Fig. 6b). Examining the IFN-g:pStat1 pathway in
detail (Fig. 6c) demonstrated that B cells were inhibited by streptonigrin at 3.5 mM, but CD4+ T cells were not inhibited by more than 50%
at even the highest dose of compound used (100 mM). Notably,
CD4– T cells were inhibited at intermediate drug levels (20 mM).
The phosphoflow histogram data representation comparing the
IFN-g:pStat1 pathway in B cells and CD4+ T cells clearly shows
the inhibition of IFN-g:pStat1 in B cells but not T cells (Fig. 6d).
Note the potent stimulation of Stat1 (inducing 15- to 30-fold changes
in phosphorylation) by IFN-g in both cell types. To confirm these
results, we isolated B cells and CD4+ T cells by magnetic purification,
then treated the two purified cell populations with the compound,
stimulated with IFN-g, and measured Stat1 phosphorylation by
western blot (Fig. 6e). Streptonigrin completely inhibited Stat1
signaling in B cells at 20 mM, but not in CD4+ T cells. Owing to
this unique cell type selectivity, streptonigrin was evaluated in vivo as
described below.
ST
R
PN
W
H
I
l3
M
Ja
C k1
ra
C ssi
R n
PS
N
M
H
K
ZM
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ST γ
R
PN
IL-4:pStat6
IFN-γ:pStat1
IL-4:pStat5
IL-10:pStat3
IL-15:pStat5
IL-6:pStat3
IL-6:pStat1
IL-7:pStat5
Percent inhibition
O
NH2
d
IFN-γ:pStat1 pathway
Percent inhibition
D
C
O
c
D
b
O
C
a
4+
T
c
4 – ell
B Tc
ce e
ll
l
C l
D
11
C b-in
D
11 t
bhi
ARTICLES
NATURE CHEMICAL BIOLOGY
pathways induced by a single cytokine are clearly regulated differently
in these three cell types. Note that the IL-10:pStat3 pathway is
separated in an opposite fashion, with T cells being inhibited strongly
and B cells weakly.
By examining the effects of all eight compounds together, we were
able to (i) derive mechanistic insights into cytokine signaling in
immune cells and (ii) offer hypotheses for further testing that
would have been difficult to obtain via traditional methods.
When examining the clustered dendrogram for potentially druggable pathways, it is notable that IL-4 induction of Stat5 and Stat6
phosphorylation appeared at opposite ends of the dendrogram. The
IL-4:pStat5 pathway in several cell types appeared at the bottom of
the dendrogram, which indicates that it was inhibited at low concentrations, whereas the IL-4:pStat6 pathway was clustered strongly at
the top of the dendrogram, which indicates that it was inhibited
at high concentrations (Fig. 7b). In these instances, one cytokine,
IL-4, induced two phosphoproteins simultaneously, but the two
proteins were inhibited differentially by these eight compounds.
This suggests a clear hypothesis that the two pathways are induced
and regulated through different intracellular mechanisms. It should be
noted that we did not find a strong correlation between the fold
change in phosphorylation of a pathway and SF values (Supplementary Fig. 4 online).
We next examined the dataset for indications of cell subsets that
appeared more susceptible to drug action than others (Fig. 7c).
Examining the cluster analysis, it was not clear that any cell subsets
were inherently more sensitive to drug action with the compound set
tested than any other cell subset. However, within the IL-6:pStat3
signaling pathway, we did observe a considerable separation of B cell
susceptibility relative to CD4+ and CD4– T cells (Fig. 7c). The
IL-6:pStat3 pathway was inhibited potently in B cells by most
compounds, but weakly or not at all in the T cell populations.
As emphasized by their separation in the row dendrogram, the
b
4,000
73.5
Spleen 2,000
1,000
FSC
15 min
Phosphoflow
0
4,000
76.1
3,000
Blood 2,000
CD11b PerCP-Cy5.5
3,000
1,000
0
Harvest spleen
and blood
IL-6:pStat3
IL-4:pStat6
[STRPN]
d
1.05
105
104
3.12
104
103
0
105
8.58
104
6.22
105
53.3
103
0
105
9.74
104
24.5
103
103
0
66.5
104
30.3
32.3
103
0
105
69.6
104
103
0
26.8
0
0 103 104 105
0 103 104 105
0 103 104 105
0 103 104 105
SSC
TCRβ PE
TCRβ PE
SSC
Total cells
B cells
CD4+ T cells
CD4– T cells
CD11b-int
IL
c
105
Spleen
Total cells
B cells
+
CD4 T cells
–
CD4 T cells
-4
+
ST IL
R -6
PN
30 min
CD4 Ax700
Cytokine
Compound
Preclinical in vivo monitoring of compound selectivity
An optimal drug development assay platform should allow monitoring of the effects of a compound in early screening and also throughout in vivo testing in preclinical animal models and in human clinical
trials. Therefore, we evaluated the utility of phosphoflow for
analysis of drug efficacy in mice. We chose to test streptonigrin, the
cell type–selective molecule, due to its unique inhibition profile of
B cell signaling (Fig. 8).
We treated mice intravenously with 1.0, 3.0 or 10 mg kg–1 of the
compound for 30 min, then stimulated Stat6 and Stat3 phosphorylation by simultaneous intravenous injection of IL-4 and IL-6, respectively (Fig. 8a). As both drug and stimulus were administered in vivo,
the natural environment of these signaling pathways was mimicked8.
Soon after stimulation, we drew peripheral blood and removed the
spleen from the mice to determine whether the molecule inhibits the
targeted pathways in specific cell types.
B220 PacBlu
a
0
103 104 105
0
103 104 105
0
CD11b-int
103 104 105
0
103 104 105
0
–
+
+
+
+
–
–
–
+
+
+
+
–
–
103 104 105
pStat3
Percent inhibition
Total cells
B cells
Blood
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
+
CD4 T cells
–
CD4 T cells
CD11b-int
100
0
103 104 105
0
103 104 105
0
103 104 105
0
103 104 105
0
103 104 105
pStat6
0
Figure 8 Preclinical monitoring of drug efficacy in vivo. (a) Experimental setup of in vivo drug testing. Compounds were administered intravenously, followed
by intravenous injection of cytokines to induce signaling. Signaling was analyzed in whole blood samples and splenocytes. (b) Surface marker analysis used
to reveal cell types for in vivo drug testing experiments. Note that B cells comprise 50–60% of the spleen, but only 5–15% of the blood. (c) Streptonigrin
was administered at three different doses (1, 3 and 10 mg kg–1), and inhibition of in vivo signaling was measured in four cell types and in the total cell
population. B cell signaling was inhibited potently in both peripheral blood and the spleen. Note however that CD11b-int cells were not inhibited as they
were ex vivo. Analysis of the total cell population showed no effect in blood, but a broad effect in spleen due to the proportion of B cells in each
compartment. (d) Underlying phosphoflow data used for the heatmap analysis. pStat3 Ax488 and pStat6 Ax647 monoclonal antibodies were used for
staining. Wedge indicates increasing doses of compound.
NATURE CHEMICAL BIOLOGY
VOLUME 4
NUMBER 2
FEBRUARY 2008
139
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
Streptonigrin was indeed selective for B cells relative to CD4+ and
CD4– T cells, both in the spleen and blood (Fig. 8c). Both IL-4:pStat6
and IL-6:pStat3 were inhibited effectively, though IL-6–pStat3 was
blocked at approximately three-fold lower dose in both tissues. The
phosphoflow platform therefore allows extremely accurate determination of drug effects in vivo, providing detailed pharmacodynamic and
dosing assessments.
The importance of single-cell analysis by phosphoflow is emphasized when the effects of streptonigrin in specific cell subsets are
compared to the effects on the total population. Without cell subsetting, it would have appeared that streptonigrin had no effect in the
blood and a moderate inhibitory effect on signaling in the spleen
(Fig. 8c). This is explained by the proportion of B cells in each tissue;
in the spleen B cells are approximately 50–60% of the total population,
whereas in the blood they comprise only 5–15% (Fig. 8b). Therefore,
with a molecule like streptonigrin that is cell type–specific, measurements made on total populations of cells will be biased toward the
most prevalent cell type (in this case B cells in the spleen and T cells in
the blood), leading to aberrant conclusions about drug activity.
Phosphoflow screening avoids these problems by measuring compound effects in each cell type individually.
DISCUSSION
Using phosphospecific flow cytometry to analyze intracellular protein
phosphorylation, compounds from the natural product library from
the NCI were evaluated for cell type– and pathway-specific effects on
cytokine-induced signaling. We found several potent Jak-Stat inhibitors, three of which were selective relative to p38 MAP kinase
inhibition. We carried these hits forward into primary cells and
found both pathway- and cell type–selective compounds. Importantly,
we were able to analyze the in vivo efficacy of streptonigrin at
inhibiting B cell signaling both in whole blood and in splenic cell
populations. Phosphoflow was therefore applied throughout the drug
discovery process, from initial screening all the way through preclinical testing.
A key prior limitation in applying flow cytometry in drug
discovery has been throughput. However, using our recently developed FCB technique increased the throughput 10- to 20-fold and
allowed us to screen several 96-well plates per day without difficulty17. With the method, it is readily possible to screen 1,000–2,000
compounds per day with manual staining and sample acquisition.
Therefore, focused libraries of approximately 10,000 compounds or
libraries of lead compounds from prior screening efforts can be
screened within one week.
Perhaps more important than throughput, however, is the quality of
the results from a quantitative platform that can measure signaling at
the single-cell level, such as phosphoflow. In the screens we performed
we obtained extremely high Z ¢ factors (40.7) and were able to
generate accurate IC50 curves (R2 4 0.95)25. These values allowed
us to quickly move from initial screening in cell lines to finding cell
type– and pathway-specific inhibitors in primary cells. It is possible
that with such a two-stage screening process—that is, moving from
a human cell line to primary mouse cells—one might follow leads
that work only in one species and not the other. However, we
believe that performing screening in two similar yet disparate
systems will generally lead to a higher number of clinically effective
lead compounds.
In this study, we identified pathway-selective inhibitors such as
crassin (NSC210236) that demonstrated the ability to target a particular signaling pathway in multiple cell types. What we did not find
was a compound that was able to target one pathway in only one cell
140
type selectively. Such molecules are perhaps the ‘holy grail’ of drug
screening, and may be identified more rapidly via phosphoflow.
Importantly though, our ability to simultaneously analyze multiple
signaling pathways substantially reduced the need for secondary
counter-screens and improved hit choice by selecting only for compounds with desired inhibition profiles from the initial screen35.
Unlike traditional in vitro kinase assays or immortalized cell line
assays (such as applied in our initial U937 screen), screening using
primary cells enabled us to assay, for each lead molecule, the cellsubtype-dependent pathway selectivity in a manner not possible with
any prior approach. This type of analysis is typically inaccessible
because classical drug screening platforms do not assay drug activity in
heterogeneous populations of cells. Indeed, using phosphoflow we
were able to identify streptonigrin (NSC45383), which displayed
4100-fold selectivity for B cells relative to T cells, both ex vivo and
in vivo. In designing drugs against diseases restricted to one or a few
cell types, this sort of cell-specific analysis will likely become critical
to success4.
By comparing the inhibition profiles of the eight compounds
simultaneously, we gained mechanistic insight into the susceptibility,
or ‘‘druggability,’’ of multiple pathways and cell subtypes36–38. Generally, pathways that were induced less strongly (that is, that had lower
fold induction of phosphorylation) were inhibited more selectively by
all the compounds39. However, there was not a clear correlation
between fold change and SF (Supplementary Fig. 4). Our results
suggest that although several pathways can be induced by one
stimulus, not all the pathways are equal as drug targets; pathways
are affected by protein abundance, and some pathways might have
different modes of regulation and negative feedback. The trends we
observed indicate that more robust phosphorylation events may be
very difficult to target selectively without affecting similar but
‘‘weaker’’ pathways at the same time. For instance, it may be very
difficult to target IL-6:pStat3 in T cells without modulating the
same pathway in B cells, given that most compounds targeted
B cells more potently. We expect that profiling many additional
Jak-Stat inhibitors will allow for the definition of an ‘‘inhibition
landscape’’ of cell types and pathways to determine those that are
more resistant to drug inhibition. Notably, compounds that are able to
act in unique ways might actually target unknown regulatory mechanisms or proteins that are not known to be part of a particular signaling
pathway. In this way, phosphoflow-based drug screening may reveal
new signaling networks and feedback loops and identify new therapeutic targets.
We envision that single-cell, multiparameter measures of proteins,
such as kinase and phosphatase signaling states, via techniques like
phosphoflow, can be applied as a useful tool in the drug development
process. Indeed, phosphoflow has applications at each stage of drug
discovery40: disease profiling and target identification10–12, highthroughput and high-content drug screening, preclinical drug testing,
and clinical monitoring of drug efficacy in patient blood samples5.
Thus, the platform is ideally suited to drug development for immune
system diseases such as leukemias, systemic lupus erythematosus,
rheumatoid arthritis, acquired immunodeficiency syndrome, transplant rejection and immune suppression, and asthma—and it potentially can be expanded into other tissues such as solid tumors with
appropriate modifications of the technique (data not shown) or into
multiparametric analysis of cell cycle effects of chemotherapeutics
(ref. 41 and data not shown). In all these diseases, researchers and
drug developers will benefit from an ability to decipher protein
functions within individual cell types in complex and otherwise
difficult-to-analyze cell populations.
VOLUME 4
NUMBER 2
FEBRUARY 2008
NATURE CHEMICAL BIOLOGY
ARTICLES
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
METHODS
Antibodies. Phosphospecific monoclonal antibodies against Stat1 (pY701,
clone 4a), Stat3 (pY705, clone 49), Stat5 (pY694, clone 47), Stat6 (pY641,
clone J71-773.58.11) and p38 (pT180/Y182, clone 36) conjugated to Ax488,
Ax647 or PE (BD Phosflow reagents) and surface marker antibodies to mouse
TCR-b (H57-597), CD11b (M1/70), B220 (RA3-6B2) and CD4 (RM4-5)
conjugated to PE, PerCP-Cy5.5 or PE-Cy7 were kindly provided by Becton
Dickinson Pharmingen. B220 was conjugated to Quantum Dot 605 (Molecular
Probes) using suggested protocols. Stat1 phosphospecific antibody was conjugated to Pacific Blue using Pacific Blue succinimidyl ester (Molecular Probes).
Human recombinant (r) IFN-g and GM-CSF and mouse rIFN-g, IL-6,
IL-10, IL-4 and IL-7 were kindly provided by BD Pharmingen. Mouse rIL-15
was from Peprotech. Bovine serum albumin (BSA) and methanol were from
Sigma. Sealed ampules of 16% formaldehyde in water were from Electron
Microscopy Sciences.
Small molecules. The NCI natural product library consisting of 235 purified
natural products at 10 mM concentration in DMSO was kindly provided by the
NCI Developmental Therapeutics Program (NCI DTP; http://dtp.nci.nih.gov/
index.html). SB203580 (SB), JakI, WHI-P154, I3M and ZM-449829 were from
Calbiochem and were greater than 95% pure. Anisomycin and larger quantities
of streptonigrin were purchased from Sigma. Commercial streptonigrin was
found to be identical to the library streptonigrin, both in selectivity and
potency, confirming identity and purity of the library molecules.
Initial cell line screening of NCI natural product library. Initial screening
was performed in U937 cells, a monocytic lymphoma cell line, cultured in
RPMI-1640 containing 10% fetal bovine serum (FBS), 100 U ml–1 penicillin,
100 mg ml–1 streptomycin and 1 mM L-glutamine (RPMI-10). Cells at a density
of B106 cells ml–1 were distributed to the wells of a 96-well deep block plate
(2-ml capacity). Compounds from the natural product library were added at
20 mM, along with the control inhibitors JakI and SB203580, for 30 min.
Stimuli (IFN-g 10 ng ml–1, GM-CSF 10 ng ml–1 and anisomycin 2 mg ml–1)
were then added simultaneously for 15 min to induce phosphorylation.
Samples were then subjected to phosphoflow and simultaneously stained with
pStat1 Pacific Blue, pStat5 Ax647 and p-p38 PE phosphospecific antibodies.
Phosphoflow cytometry. Phosphoflow was performed as previously
described6,8,42. Briefly, cells were fixed with formaldehyde (1.6% final concentration), permeabilized with cold methanol (100%), washed twice with staining
medium (phosphate-buffered saline (PBS), pH 7.3 containing 0.5% BSA and
0.02% sodium azide) and finally stained with phosphospecific antibodies in
staining medium.
Flow cytometry was performed on a BD LSRII flow cytometer, equipped
with 405 nm, 488 nm and 633 nm lasers. After acquisition, data were analyzed
using Flowjo software to generate MFI values for each phosphoprotein in each
cell type of interest. The medians were then exported to Microsoft Excel
to calculate percent inhibition (PI) as follows: PI ¼ (MFIstim – MFItreated)/
(MFIstim – MFI unstim) 100, where MFItreated represents compound-treated
and stimulated samples, MFIunstim is the value for unstimulated negative control
and MFIstim represents the stimulated positive control. Data were visualized in
Spotfire DecisionSite to generate heatmaps and perform clustering analysis.
Primary cell screening. Splenocytes were obtained from 6–10 week old BALB/c
mice by homogenization and resuspended in RPMI-10 at 5 106 cells ml–1.
Splenocytes were rested for 2–3 h at 37 1C before stimulation7,8. All animal
work was performed under approved Stanford Administrative Panel on
Laboratory Animal Care/Institutional Animal Care and Use Committee
(APLAC/IACUC) protocols. Eight different compounds, four natural products
and four commercially available Jak inhibitors, were added at five different
concentrations (five-fold dilutions). Samples were incubated for 30 min, then
stimulated separately with rIFN-g, IL-6, IL-10, IL-4, IL-7 or IL-15 at 10 ng ml–1.
Cells were then fixed and permeabilized as described above. During the
permeabilization step, samples were subjected to the fluorescent cell barcoding
method described previously17. Briefly, samples were labeled with 2.0, 0.66, 0.22
or 0 mg ml–1 of Ax700-NHS (encoding four different compounds) and 3.0,
1.0, 0.3, 0.1, 0.03 or 0 mg ml–1 Pacific Blue-NHS (encoding the six-point
dose-response curve for each compound, including the positive control). The
NATURE CHEMICAL BIOLOGY
VOLUME 4
NUMBER 2
FEBRUARY 2008
barcoded samples were then washed twice, combined and stained with B220
QD605, TCRb PE, CD11b PerCP-Cy5.5 and CD4 PE-Cy7 (previously validated
for phosphoflow7) and the combination of pStat1 Ax647 and pStat3 Ax488 or
pStat5 Ax647 and pStat6 Ax488. SF was calculated as follows: SF ¼ –log2(IC50/
median IC50), where IC50 is the IC50 value of a compound on a particular cell
type–pathway combination and median IC50 is the median IC50 value of the
same compound across all 27 cell type–pathway combinations analyzed.
Western blotting. Splenocytes were isolated from BALB/c mice as described
above, and resuspended in MEM containing 5% FBS, 100 U ml–1 penicillin,
100 mg ml–1 streptomycin and 1 mM L-glutamine (MEM-5). B cells and CD4+
T cells were purified using the BD IMag separation system according to
manufacturer protocols. Briefly, CD4+ T cells were isolated first using antiCD4 magnetic beads. B cells were then purified using anti-B220 magnetic
beads. Isolated cells were 90–95% pure. After isolation, cells were rested for 2 h
in RPMI-10 at 37 1C, and were then treated with inhibitors for 30 min before
stimulation with cytokines for 15 min. Cells were put on ice, then pelleted at
4 1C for 3 min and resuspended in lysis buffer (PBS containing 1% Triton
X-100, protease inhibitor tablet (Roche), 5 mM sodium fluoride and 1 mM
sodium orthovanadate). After 30 min, cellular debris were pelleted, and equal
protein amounts were loaded per lane on reducing SDS-PAGE gels. Gels were
transferred to polyvinylidene fluoride (PVDF) membranes and blotted with the
same antibody clones used for phosphoflow analysis. Luminescent images were
acquired after horseradish peroxidase development on a Lumi-Imager CCDbased digital system (Roche).
In vivo compound treatment. In vivo compound treatment and stimulation
were performed as previously described8. Briefly, streptonigrin was dissolved
in 200 ml sterile PBS and injected into the tail vein (intravenously) of BALB/c
mice (at 1, 3 or 10 mg kg–1). After 30 min, IL-4 and IL-6 (500 ng each) were
injected intravenously. After 15 additional min, mice were killed, the spleen
was removed and homogenized directly in BD Lyse/Fix buffer, and blood
was removed via cardiac puncture and diluted 20-fold in BD Lyse/Fix buffer.
Both tubes were incubated for 15 min. After pelleting, cells were resuspended in 4 1C methanol and standard phosphoflow methods were used as
described above.
Note: Supplementary information and chemical compound information is available on
the Nature Chemical Biology website.
ACKNOWLEDGMENTS
We thank E. Danna, K. Schulz and K. Gibbs for critical reading of this
manuscript. We thank BD Pharmingen, in particular G. Gao, R. Campos and
B. Balderas, for kindly providing antibody reagents and technical expertise. We
also thank the NCI DTP and J. Johnson for providing the natural product library
used in this study. P.O.K. was supported in part by a Howard Hughes Medical
Institute Predoctoral Fellowship. G.P.N. was supported by US National Heart,
Lung and Blood Institute contract N01-HV-28183 and US National Institutes of
Health grant AI35304.
AUTHOR CONTRIBUTIONS
P.O.K. designed the study, performed experiments, analyzed data and wrote the
manuscript. J.M.C. performed experiments and analyzed data. M.R.C. performed
experiments and helped write the manuscript. G.P.N. helped to design the study
and write the manuscript.
COMPETING INTERESTS STATEMENT
The authors declare competing financial interests: details accompany the full-text
HTML version of the paper at http://www.nature.com/naturechemicalbiology/.
Published online at http://www.nature.com/naturechemicalbiology
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions
1. Nicholson, R.L., Welch, M., Ladlow, M. & Spring, D.R. Small-molecule screening:
advances in microarraying and cell-imaging technologies. ACS Chem. Biol. 2, 24–30
(2007).
2. Valet, G. Cytomics as a new potential for drug discovery. Drug Discov. Today 11,
785–791 (2006).
3. Lang, P., Yeow, K., Nichols, A. & Scheer, A. Cellular imaging in drug discovery. Nat.
Rev. Drug Discov. 5, 343–356 (2006).
141
© 2008 Nature Publishing Group http://www.nature.com/naturechemicalbiology
ARTICLES
4. Nolan, G.P. What’s wrong with drug screening today. Nat. Chem. Biol. 3, 187–191
(2007).
5. Chow, S., Patel, H. & Hedley, D.W. Measurement of MAP kinase activation by flow
cytometry using phospho-specific antibodies to MEK and ERK: potential for pharmacodynamic monitoring of signal transduction inhibitors. Cytometry 46, 72–78 (2001).
6. Krutzik, P.O. & Nolan, G.P. Intracellular phospho-protein staining techniques for flow
cytometry: monitoring single cell signaling events. Cytometry A 55, 61–70 (2003).
7. Krutzik, P.O., Clutter, M.R. & Nolan, G.P. Coordinate analysis of murine immune cell
surface markers and intracellular phosphoproteins by flow cytometry. J. Immunol. 175,
2357–2365 (2005).
8. Krutzik, P.O., Hale, M.B. & Nolan, G.P. Characterization of the murine immunological
signaling network with phosphospecific flow cytometry. J. Immunol. 175, 2366–2373
(2005).
9. Krutzik, P.O., Irish, J.M., Nolan, G.P. & Perez, O.D. Analysis of protein phosphorylation
and cellular signaling events by flow cytometry: techniques and clinical applications.
Clin. Immunol. 110, 206–221 (2004).
10. Irish, J.M. et al. Single cell profiling of potentiated phospho-protein networks in cancer
cells. Cell 118, 217–228 (2004).
11. Irish, J.M., Kotecha, N. & Nolan, G.P. Mapping normal and cancer cell signalling
networks: towards single-cell proteomics. Nat. Rev. Cancer 6, 146–155 (2006).
12. Jacobberger, J.W. et al. Immunoreactivity of Stat5 phosphorylated on tyrosine as a cellbased measure of Bcr/Abl kinase activity. Cytometry A 54, 75–88 (2003).
13. Irish, J.M., Czerwinski, D.K., Nolan, G.P. & Levy, R. Kinetics of B cell receptor
signaling in human B cell subsets mapped by phosphospecific flow cytometry.
J. Immunol. 177, 1581–1589 (2006).
14. Zell, T. et al. Single-cell analysis of signal transduction in CD4 T cells stimulated by
antigen in vivo. Proc. Natl. Acad. Sci. USA 98, 10805–10810 (2001).
15. Ilangumaran, S., Finan, D. & Rottapel, R. Flow cytometric analysis of cytokine receptor
signal transduction. J. Immunol. Methods 278, 221–234 (2003).
16. Kersh, E.N. et al. TCR signal transduction in antigen-specific memory CD8 T cells.
J. Immunol. 170, 5455–5463 (2003).
17. Krutzik, P.O. & Nolan, G.P. Fluorescent cell barcoding in flow cytometry allows highthroughput drug screening and signaling profiling. Nat. Methods 3, 361–368 (2006).
18. Sachs, K., Perez, O., Pe’er, D., Lauffenburger, D.A. & Nolan, G.P. Causal proteinsignaling networks derived from multiparameter single-cell data. Science 308,
523–529 (2005).
19. Edwards, B.S., Oprea, T., Prossnitz, E.R. & Sklar, L.A. Flow cytometry for highthroughput, high-content screening. Curr. Opin. Chem. Biol. 8, 392–398 (2004).
20. Young, S.M. et al. High-throughput screening with HyperCyt flow cytometry to detect
small molecule formylpeptide receptor ligands. J. Biomol. Screen. 10, 374–382
(2005).
21. Jacobberger, J.W. Increasing the power of cytometry. Nat. Methods 3, 343–344
(2006).
22. Van Meter, M.E. et al. K-RasG12D expression induces hyperproliferation and aberrant
signaling in primary hematopoietic stem/progenitor cells. Blood 109, 3945–3952
(2007).
142
23. O’Shea, J.J., Pesu, M., Borie, D.C. & Changelian, P.S. A new modality for immunosuppression: targeting the JAK/STAT pathway. Nat. Rev. Drug Discov. 3, 555–564
(2004).
24. Luo, C. & Laaja, P. Inhibitors of JAKs/STATs and the kinases: a possible new cluster of
drugs. Drug Discov. Today 9, 268–275 (2004).
25. Zhang, J.H., Chung, T.D. & Oldenburg, K.R. A simple statistical parameter for use in
evaluation and validation of high throughput screening assays. J. Biomol. Screen. 4,
67–73 (1999).
26. Quintas-Cardama, A. et al. Phase I/II study of subcutaneous homoharringtonine in
patients with chronic myeloid leukemia who have failed prior therapy. Cancer 109,
248–255 (2007).
27. Chen, R., Gandhi, V. & Plunkett, W. A sequential blockade strategy for the design of
combination therapies to overcome oncogene addiction in chronic myelogenous
leukemia. Cancer Res. 66, 10959–10966 (2006).
28. Dessureault, J. & Krause, M.O. The effect of a new antibiotic, cryptosporiopsin, on
RNA synthesis in L-cells. Can. J. Biochem. 53, 149–154 (1975).
29. Burres, N.S., Hunter, J.E. & Wright, A.E. A mammalian cell agar-diffusion assay for the
detection of toxic compounds. J. Nat. Prod. 52, 522–527 (1989).
30. Marumo, H., Kitaura, K., Morimoto, M., Tanaka, H. & Omura, S. The mode of action
of nanaomycin A in Gram-positive bacteria. J. Antibiot. (Tokyo) 33, 885–890
(1980).
31. Suzuki, H. et al. Immunosuppressive activity of streptonigrin in vitro and in vivo.
Biosci. Biotechnol. Biochem. 60, 789–793 (1996).
32. Thompson, J.E. et al. Photochemical preparation of a pyridone containing tetracycle: a
Jak protein kinase inhibitor. Bioorg. Med. Chem. Lett. 12, 1219–1223 (2002).
33. Narla, R.K., Liu, X.P., Myers, D.E. & Uckun, F.M. 4-(3¢-Bromo-4¢hydroxylphenyl)amino-6,7-dimethoxyquinazoline: a novel quinazoline derivative with potent cytotoxic
activity against human glioblastoma cells. Clin. Cancer Res. 4, 1405–1414 (1998).
34. Nam, S. et al. Indirubin derivatives inhibit Stat3 signaling and induce apoptosis in
human cancer cells. Proc. Natl. Acad. Sci. USA 102, 5998–6003 (2005).
35. Feng, B.Y., Shelat, A., Doman, T.N., Guy, R.K. & Shoichet, B.K. High-throughput
assays for promiscuous inhibitors. Nat. Chem. Biol. 1, 146–148 (2005).
36. Hopkins, A.L. & Groom, C.R. The druggable genome. Nat. Rev. Drug Discov. 1,
727–730 (2002).
37. Overington, J.P., Al-Lazikani, B. & Hopkins, A.L. How many drug targets are there?
Nat. Rev. Drug Discov. 5, 993–996 (2006).
38. Cheng, A.C. et al. Structure-based maximal affinity model predicts small-molecule
druggability. Nat. Biotechnol. 25, 71–75 (2007).
39. Kitano, H. A robustness-based approach to systems-oriented drug design. Nat. Rev.
Drug Discov. 6, 202–210 (2007).
40. Clutter, M.R., Krutzik, P.O. & Nolan, G.P. Phospho-specific flow cytometry in drug
discovery. Drug Discov. Today Technol. 2, 295–302 (2005).
41. Darzynkiewicz, Z., Crissman, H. & Jacobberger, J.W. Cytometry of the cell cycle:
cycling through history. Cytometry A 58, 21–32 (2004).
42. Schulz, K.R., Danna, E.A., Krutzik, P.O. & Nolan, G.P. Single-cell phospho-protein
analysis by flow cytometry. Curr. Protoc. Immunol. 78, 8.17 (2007).
VOLUME 4
NUMBER 2
FEBRUARY 2008
NATURE CHEMICAL BIOLOGY

Purchase answer to see full
attachment

  
error: Content is protected !!