 870 Data Supplement - Supplemental Data_files/shim.gif) Science, Vol 294,
Issue 5543, 870-875 , 26 October 2001
 870 Data Supplement - Supplemental Data_files/abstract.gif)
|
The Plasticity of Dendritic Cell
Responses to Pathogens and Their Components Q.
Huang, D. Liu, P. Majewski, L. C. Schulte, J. M. Korn, R. A.
Young, E. S. Lander, N. Hacohen
|
Supplementary MaterialPreparation
of dendritic cells
Elutriated human monocytes (Advanced Biotechnology Inc.)
were grown in RPMI with 10% fetal bovine serum (Life Technologies)
supplemented with 1000 U/ml GM-CSF (R&D Systems) and 1000 U/ml
IL-4 (R&D Systems) for 7 days in 24 well plates (106
cells in 1 ml medium/well), and fed every 2 days after plating
(1).
Stimulation of dendritic cells
On day 7, DC were harvested and aliquoted into 100mm
plates at 107 cells/plate and incubated at 37° for 60
min. Pathogens or their components were then added to DC cultures at
amounts shown below. Stimulated cells showed mature DC dendrite
formation, were less adherent at 24 hours, expressed typical DC
markers (cd83, cd86) (Web fig. 4), and stimulated allogeneic T cell
proliferation at DC:T ratios of 1:1000-1:10. The following
microorganisms were used at the titers shown: E. coli SD54
(ATCC) (5:1 MOI) (kindly provided by J. Nau), Influenza
A/PR/8/34 (750 HAU/ml, Hemagglutinin units; the amount of virus
needed to infect 50% of the cells was ~ 5-10 HAU/ml, gift of J.
Hermann lab), Candida albicans HLC54 (5:1 MOI) (gift of G.
Fink lab); and the following components at the concentrations shown:
LPS, lipopolysaccharide from E. coli 055:B5 (1 g/ml, Sigma L-2880), polyI:C (25 g/ml, Pharmacia; endotoxin levels were <0.2 EU/ml, which
are not sufficient to induce TNF expression in DC), mannan from S. cerevisiae (1
mg/ml, Sigma, M 7504). One concern is that the observed pathogen
specific responses are due to different titers of each pathogen,
rather than to intrinsic differences in pathogen recognition. This
is unlikely because many genes within the common response are
induced to equal levels by all stimuli, suggesting that some
pathways can be engaged equally by all pathogens; the titers used
were shown to saturate expression of known maturation markers; all
the pathogens were phagocytosed by DC; and in some cases, we
repeated microarray measurements with lower titers and
concentrations and found similar gene expression profiles (data not
shown).
Supplemental Table 1. Summary of donors with pathogens
and components
The following table is a summary of the stimuli added to DC from
each donor.
DONOR |
STIMULI |
Donor D |
polyI:C |
Donor F |
E. coli, influenza, LPS, polyI:C |
Donor H |
E. coli, influenza, LPS |
Donor I |
C. albicans |
Donor M |
C. albicans, mannan |
Donor O |
mannan |
Donor T |
E. coli, influenza, C.
albicans |
RNA preparation and microarray hybridization
At each timepoint (0, 1 hr, 2 hr, 4 hr, 8 hr, 12 hr, 18
hr, 24 hr, 36 hr), 107 cells were harvested and lysed
using Trizol (Molecular Research Center). Total RNA was isolated,
labeled and prepared for hybridization to HuGeneFL oligonucleotide
arrays (Affymetrix) using standard methods (2).
Hybridization was carried out overnight with 15 g of labeled RNA product, and arrays were scanned on
Affymetrix scanners. Measurements were made at multiple timepoints
after DC stimulation in order to allow comparison of response levels
between six unrelated human donors with variable kinetics and to
define different phases in DC activation. Influenza and E.
coli responses were always measured in the same donor.
Statistical analysis
Data collection and validation. 1.7 ×
106 individual gene measurements were stored, analyzed
and visualized using a set of database and analysis tools developed
in the lab. Arrays measurements were normalized with a reference
array hybridized with sample from the same donor, using the median
of the hybridization signals of all genes with P-calls (P-calls
according to the Affymetrix software) as a scaling factor. Data for
several missing timepoints were interpolated using flanking
timepoints. Gene expression profiles were validated by measuring the
protein levels of four secreted factors (TNF , IL12p40, IL10, MCP-1) using standard ELISA measurements
which confirmed both the kinetics and the relative induction levels
in response to each stimulus (Web fig. 1).
Scoring system. We collected a time series of mRNA
fluorescence levels, R={R1, R2,
R3,..., Rn}, in DC exposed to each pathogen or
compound, and a control series of mRNA levels, C={ C1,
C2, C3,..., Cn }, in untreated DC
from the same donor. Ri and Ci are
steady-state mRNA hybridization measurements ("average difference"
in Affymetrix terminology) at the ith timepoint; n is the
total number of timepoints. We devised a score,
Si=(Ri- C)/ C, to measure significant deviation of the
stimulated expression level at one timepoint, Ri from the
mean Cof the control timecourse. By using C, which is the standard deviation of the
control timecourse, the score penalizes genes with high noise in the
media control, thus allowing us to extract the most robust data.
Regulated genes. Upregulated genes were selected
according to the following criteria applied at the same time to all
three donors: (i) Si>1.2 for 2 consecutive timepoints or (ii) Si>4 for
1 timepoint. Downregulated genes were selected if
Si <-1.4 for 4 points; however, since the signal-to-noise was lower than
for upregulated genes, we relaxed the filter further and selected
the most robust down-regulated genes that passed our filter 2 out of
3 donors and had the same trend in the third donor.
Common regulated genes. Common upregulated genes were
selected based on an intersection of the pathogen upregulated genes
defined above (i & ii). Common downregulated genes were selected
based on Si <-1.2 for 3 points across all pathogens.
Stimulus-specific genes. Fold expression level in
response to a stimulus, A, and relative to control is defined as
FA=max{L1,L2,...,Ln-1}
where Li=geomean{Ri
Ri+1}/geomean{Ci Ci+1}. The ratio
of fold expression levels for a gene between stimulus A and B is
defined as RFAB=FA/FB. Genes that
were regulated more strongly by one stimulus (A) than another
stimulus (B) were identified in one of two ways: (i) average
RFAB 2.5 in three donors; or (ii) in three donors,
Si>1.4 for 2 points for stimulus A (selects for a gene profile that is
significantly different from control timecourse) and
Si=1.2 for all points for stimulus B (gene profile
closely matches the control timecourse). The final set of
stimulus-specific genes (Figs. 1 and 3, stippled circles) were used
to describe how the DC response differs biologically for each
stimulus. The remaining genes that were regulated in stimulus A but
not in stimulus B and did not pass the stimulus-specific filters
above were not used in a comparative analysis since they are on the
border between the groups of common and stimulus-specific genes.
All filters were used as described above except for:
(1) Two-way comparisons (between DC
responses to stimulus A vs. stimulus B) when the responses to A and
B were not measured in the same donor for all replicates: (i)
Component-pathogen comparisons. All component responses were
measured in two donors while all pathogen responses were measured in
three donors. To increase the stringency of comparison between
component and pathogen, we required that the average
RFAB 2.5 (as for all other comparisons) and that
FA/FB 2 for one experiment in which the component and pathogen
are measured in the same batch of DC. (ii) C.
albicans-pathogen comparisons. Since the same batch of DC were
exposed only in one experiment (donor T) to all the pathogens
(C. albicans, E. coli and influenza), we increased the
stringency of any pathogen comparison to C. albicans by requiring
that the average RFAB 2.5 and that FA/FB 2 for the experiment in which all the pathogen responses
were measured (donor T). (2) Regulated
genes for component responses, which were measured in two donors,
require a more stringent filter: Upregulated genes were
selected according to the following criteria for one experiment: (i)
Si>2 for 2 consecutive timepoints or (ii) Si>1.2 for
3 consecutive timepoints or (iii) Si>5 for
1 timepoint; the same criteria were applied to an
independent experiment from a second donor, except (i) was replaced
with a less stringent filter, Si>1.4 for 2 points, to take into account variations in level of
expression between donors. Downregulated genes were selected if
Si <-1.4 for 2 points.
Temporal clustering
A self-organizing map algorithm (3) was used to
cluster genes together based on the similarity of their temporal
expression profiles. We used this procedure to aid us in classifying
genes into six basic groups: three major groups which consist of
genes that are expressed in an early, middle or late phase of the
timecourse; and for each of these groups, the genes are further
divided up into those that are expressed transiently or in a
sustained fashion (Fig. 2A).
Functional categories
We also assigned each of the regulated genes into
functional categories (e.g. chemokines, glycolysis, etc.) according
to the public databases. We verified and refined our assignment of
functional groups by comparing to the Proteome annotated database of
genes, Human PSD(tm) (kindly provided through a collaboration with
Proteome, Inc).
Neutrophil migration
Neutrophil isolation and migration assays were done with
standard techniques (4). Directed neutrophil migration was
measured using 96-well chambers with filters (Neuroprobe). 20,000
neutrophils were fluorescently labelled with PKH26 (Sigma) and
placed on top of the filter; DC supernatants were placed below (1:10
dilution). After incubating for 30 min at 37°C and 10 min at 4°C,
neutrophils were counted on the bottom of each well.
References 1. M. Cella et al.,
J. Exp. Med. 189, 821 (1999). 2. T. R.
Golub et al., Science 286, 531
(1999). 3. P. Tamayo et al., Proc. Natl. Acad. Sci.
U.S.A. 96, 2907 (1999). 4. J. E. Coligan
et al., Eds., Current Protocols in Immunology
(Wiley, 1999).
Supplemental Figure 1. mRNA and protein
expression levels for several cytokines and chemokines.
(A) TNFa. Left panels are hybridization
fluorescence levels from Affymetrix microarrays (average
difference). Right panels are standard ELISA measurements (R&D
Systems) of protein levels from the supernatant of the same DC
incubated with different pathogens or components (see legend on
right). (B) IL-12/p40. (C) mcp-1.
(D) IL-10.
 870 Data Supplement - Supplemental Data_files/1064314S1_thumb.gif)
Medium
version | Full
size version
Supplemental Figure 2. Comparison of E.
coli-regulated and component-regulated gene expression in human
monocyte-derived dendritic cells. (A) Right panel,
Overlapping sets of E. coli- and LPS regulated genes(see
methods). The number of the pathogen-specific genes are shown inside
the stippled circles (see methods). Some stimulus-specific genes are
also found in the common group but are much more strongly regulated
in one stimulus that in others and are used to define
stimulus-specific properties of DC. Left panel, representation of
mRNA expression levels at 0, 1, 2, 4, 8, 12, 24 hours in response to
E.coli and LPS. Each gene is represented by a single row of
colored bars, and each time point is represented by a single column.
Color bars represent the ratio of hybridization measurements between
corresponding timepoints in the pathogen and control medium
profiles, according to the scale shown. Genes are placed in groups
corresponding to pair-wise overlaps shown in accompanying Venn
diagrams. From top to bottom: E. coli- but not
LPS-regulated genes; common regulated genes; LPS but not E.
coli regulated genes. The vertical grey bar on the right marks
stringent stimulus-specific genes corresponding to the stippled
circles in the Venn diagram. (B) Gene profiles and
overlap of E.coli and mannan. (C) Gene
profiles and overlaps for E.coli and dsRNA.
 870 Data Supplement - Supplemental Data_files/1064314S2_thumb.gif)
Medium
version | Full
size version
Supplemental Figure 3. Comparison of pathogen
and component-regulated gene expression in human monocyte-derived
dendritic cells. (A) Gene profiles and overlap of
C. albicans and mannan. (B) Influenza and
dsRNA. All conventions are as in Supplementary Figure 2.
 870 Data Supplement - Supplemental Data_files/1064314S3_thumb.gif)
Medium
version | Full
size version
Supplemental Figure 4. FACS analysis of DC
preparations. (A) cd86 (B7-2) staining of untreated
DC or DC treated with LPS or polyI:C. (B) cd83
staining of untreated DC or DC treated with LPS.
(C) cd14 staining of monocytes, untreated DC or DC
treated with LPS. Dendritic cells have no cd14 expression and are
off-scale to the left of the x-axis.
 870 Data Supplement - Supplemental Data_files/1064314S4_thumb.gif)
Medium
version | Full
size version
Supplemental Figure 5. Differential effects of
pathogens on neutrophil chemotaxis. Migration of neutrophils (in a
chemotaxis chamber) toward conditioned cell culture medium collected
from DC exposed to control medium, influenza or E. coli for
different times. Results were reproduced in two donors in five
independent experiments.
 870 Data Supplement - Supplemental Data_files/1064314S5_thumb.gif)
Medium
version | Full
size version
View the Excel
tables
Volume 294, Number 5543, Issue of 26 Oct 2001,
p. 870. Copyright © 2004 by The American Association for the
Advancement of Science. All rights reserved.
|